October 2003 327ndash341Lead Review Article

Ketones Metabolismrsquos Ugly DucklingTheodore B VanItallie MD and Thomas H Nufert BA

Ketones were rst discovered in the urine of dia-betic patients in the mid-19th century for almost 50years thereafter they were thought to be abnormaland undesirable by-products of incomplete fat ox-idation In the early 20th century however theywere recognized as normal circulating metabolitesproduced by liver and readily utilized by extrahe-patic tissues In the 1920s a drastic ldquohyperketo-genicrdquo diet was found remarkably effective for treat-ment of drug-resistant epilepsy in children In 1967circulating ketones were discovered to replace glu-cose as the brainrsquos major fuel during the markedhyperketonemia of prolonged fasting Until thenthe adult human brain was thought to be entirelydependent upon glucose During the 1990s diet-induced hyperketonemia was found therapeuticallyeffective for treatment of several rare genetic disor-ders involving impaired neuronal utilization of glu-cose or its metabolic products Finally growingevidence suggests that mitochondrial dysfunctionand reduced bioenergetic efciency occur in brainsof patients with Parkinsonrsquos disease (PD) and Alz-heimerrsquos disease (AD) Because ketones are ef-ciently used by mitochondria for ATP generationand may also help protect vulnerable neurons fromfree radical damage hyperketogenic diets shouldbe evaluated for ability to benet patients with PDAD and certain other neurodegenerative disorders

Key words ketogenic diet hyperketonemia Par-kinsonrsquos disease Alzheimerrsquos disease mitochon-driopathies seizure disorders ketone metabolismcopy 2003 International Life Sciences Institute

doi 10131nr2003oct327ndash341

An Inauspicious Debut

From the very beginning ketone bodiesmdashb-hydroxybu-tyric acid (BHB) acetoacetic acid (AcAc) and ace-

tonemdashsuffered the stigma so often in icted on thosediscovered in bad company When ketones rst came tothe attention of physicians in the latter part of the 19thcentury it was because they were found in abundance inthe urine of patients in diabetic coma12 It soon becameevident that a vast overproduction of ketones in the bodywas largely responsible for the devastating clinical man-ifestations of what is now called diabetic ketoacidosisNegative views about ketones prevailed and the nature ofthe relationship between impaired glucose utilization andketone body metabolism continued to be misunderstoodfor nearly a half century In the words of Peters and VanSlyke3 ldquoThe formation of b-hydroxybutyric and aceto-acetic acids instead of CO2 H2O was believed todenote incomplete combustion of fat It was thereforededuced that complete combustion of fat required simul-taneous oxidation of carbohydrate an opinion vividlyexpressed by Naunyn in the aphorism lsquoFats burn in the ame of carbohydratersquo As analytical techniques havegained in precision and sensitivity ketone bodies haveproved to be normal components of blood not productsthat appear only when the metabolism of carbohydrate isdisorderedrdquo

Physiology and Metabolism

Starting in the 1930s studies of the physiology of keto-genesis and ketolysis established the hepatic origin ofketone bodies and their rapid utilization by extrahepatictissues4ndash6 Diabetic animals were found to use ketones asrapidly as normal animals do78 Early studies involvinginjection of BHB into laboratory animals showed that itsutilization increased as its concentration in the bloodrose As the blood level of BHB continued to risehowever its utilization slowed and then reached a pointat which further BHB administration did not increaseperipheral uptake910 Such observations underlie thecurrently held view that hyperketonemia results whenhepatic production of ketone bodies exceeds the abilityof the extrahepatic tissues to remove them

In 1958 Johnson et al11 reported on the normalvariations in total ketone bodies measured in serum andurine of 208 healthy young men eating normal diets andengaging in moderate daily physical activity They found

Dr VanItallie and Mr Nufert are with the Divisionof Endocrinology Diabetes and Nutrition Departmentof Medicine St Lukersquos-Roosevelt Hospital Center inNew York City 10025 USA Dr VanItallie is ProfessorEmeritus of Medicine Columbia University College ofPhysicians amp Surgeons

327Nutrition Reviews Vol 61 No 10

that there are always measurable concentrations of ke-tones in serum and urine Regardless of season the meanpostabsorptive blood serum level for this group was 07mML (It is now evident that the method used byJohnson et al to measure serum ketones systematicallyoverestimated ketone concentrations When currentlypreferred enzymatic methods are used blood ketonelevels after an overnight fast rarely exceed 05 mML andare usually much lower)

The postabsorptive serum levels of ketones thatJohnson and associates thought normal should be con-trasted with the dramatically higher concentrations (inmML) of BHB (ca 44ndash55) and AcAc (ca 10) thatoccur after a week or more of fasting in non-diabeticobese individuals12 Among hospitalized patients in se-vere diabetic ketoacidosis initial BHB levels of 230mML are not unusual13

By 1949 investigators knew that two molecules ofacetyl-CoA (the primary metabolic intermediate arisingfrom b-oxidation of fatty acids) could condense to formAcAc14 The principal pathway by which AcAc isformed in the liver from acetoacetyl-CoA remained un-clear until 1958 however when Lynen et al15 reportedelegant studies indicating that deacylation of acetoacetyl-CoA did not take place directly but involved two steps(1) condensation of acetoacetyl-CoA with acetyl-CoA toform b-hydroxy-b-methylglutaryl-CoA (HMG-CoA)and (2) cleavage of HMG-CoA to form AcAc and acetyl-CoA As shown in Figure 1 the pathway proposed byLynen et al is the one currently accepted Formation ofHMG-CoA from acetoacetyl CoA is catalyzed by mito-chondrial HMG-CoA synthase a step that is stimulatedby starvation low levels of insulin and consumption ofa low-carbohydratelow-protein high-fat diet AcAc isproduced from HMG-CoA by HMG-CoA lyase16 Mech-anisms by which ketones are formed in the liver and usedin extrahepatic tissues are shown in Figure 1

Ketone Body Kinetics in Humans

As measured by Wastney and associates17 the ketonebodies have a turnover time in the blood circulation ofapproximately 2 minutes From the circulation theyrapidly enter a compartment (probably the body cellmass) that turns over approximately once a minute In1984 Hall et al18 reported measurements of postabsorp-tive plasma concentrations and kinetics of AcAc andBHB release and removal in 11 normal subjects 6patients with newly diagnosed and untreated insulin-dependent (type 1) diabetes and 10 severely obese indi-viduals who were rst studied in the postabsorptive stateand then again after 1 to 2 weeks of starvation (Table 1)

As shown in Table 1 the mean postabsorptiveplasma ketone levels (012 mML) for normal subjects

studied by Hall et al were well below the mean of 07mML reported by Johnson et al11 as ldquonormalrdquo Hall etal measured AcAc and BHB enzymatically adapting theprocedure described by Williamson et al in 196219 Inthe normal subjects mean rates of ketone release andremoval were of the order of 100 mmol min 1 m 2 Inthe patients with newly diagnosed untreated type 1diabetes mellitus the mean postabsorptive ketone levelwas 246 mML approximately 20 times higher than thatin normal subjects Mean ketone release rate in thediabetic patients was approximately triple that of thenormal subjects Mean ketone removal rate in the dia-betic patients was approximately double that of thenormal subjects however the mean rate coef cient forketone removal (the fraction of the total uptake lost byoxidation and excretion) was reduced to 13 of normal

In the 10 severely obese subjects the mean postab-sorptive plasma ketone level (042 mML) was higherthan that in the normal individuals The obese subjectsalso exhibited substantially higher than normal meanrates of ketone release and removal (171 vs 81 and 210vs 111 mmol min 1 m 2 respectively) Similar to thediabetic patients the obese subjects showed a lower thannormal mean rate coef cient for ketone removal

In the obese subjects who had been deprived of foodfor 1 to 2 weeks the mean postabsorptive ketone levelwas 701 mML almost 60 times normal At the sametime mean ketone release and removal rates in thefood-deprived obese individuals were approximately vetimes normal whereas the rate coef cient for ketoneremoval was reduced to approximately 16 of normalWhenever ketone body levels were elevated in the dia-betic and obese subjects the increase was almost alwaysin the form of BHB

These observations indicate that ketone bodies arevery rapidly cleared from the circulation and metabo-lized by extrahepatic tissues In untreated patients withtype 1 diabetes ketone production and removal rates aregreatly augmented In the presence of such high plasmaketone levels however the mean rate coef cient forketone removal is signi cantly reduced This reductionin the rate coef cient for ketone removal (also observedin the postabsorptive and food-deprived obese subjects)was found by Hall et al18 to be coupled with an increasein the interconversion between the ketones with a largerfraction of AcAc being converted to BHB than thereverse (we now know that this re ects reduction of thefree [NAD ][NADH] ratio with which ketones are innear equilibrium)20

The biochemical mechanism underlying the in-creased ketonemia that occurs when food deprivation isprolonged is incompletely understood Because the lipol-ysis rate after 5 weeks remains essentially unchangedfrom that seen during the rst few days of food depriva-

328 Nutrition Reviews Vol 61 No 10

tion and the rate of ketogenesis also remains un-changed21 it is evident that the rate of peripheral clear-ance of ketones must diminish in order to account for theelevated blood ketone levels measured during the food-deprivation period

Laffel22 suggested that the inverse relationship be-tween level of ketonemia and rate coef cient for ketoneremoval could arise from down-regulation of an enzymeinvolved in ketone utilization in response to a rise inintracellular levels of AcAc during starvation in associ-ation with dietary carbohydrateprotein restriction or intype 1 diabetes This explanation seems reasonable how-ever the evidence thus far is inconclusive that any suchenzyme is suf ciently inhibited or decreased to accountfor the observed diminution in ketone utilization It isnoteworthy however that the brain adapts to starvationwith an increased rate of ketone utilization and that anincrease in 3-oxoacid-CoA transferase (OCT) activityhas been demonstrated in the brain of the starved rat Bycontrast decreased OCT activity has been described inthe skeletal muscle of rats with diabetes of more than 10daysrsquo duration16 To complicate matters further exercisetraining for a number of weeks appears to increase thecapacity of muscle to use ketones Because exercise alsopromotes ketogenesis the muscles presumably accom-modate to such increased production by inducing oractivating relevant ketone-utilizing enzymes1623

The physiologic explanation for the increased ke-tonemia during the early phase of food deprivation ap-pears to lie in data reported by Owen and Reichard24

who showed that the rise in blood ketones during the rst2 weeks of starvation results from a decreasing rate ofketone utilization in muscle As food deprivation contin-ues a dramatic change occurs in glucose and ketoneutilization by the nervous system and muscle Glucoseutilization in the brain drops to 13 that in the fed statewith ketones lling the energy gap A nearly reciprocalrelationship exists for muscle which seems to adapt sofully to prolonged fasting that it virtually ignores ketonesas an energy source drawing instead on fatty acids21 AsCahill25 stated ldquo early in starvation ketoacids serveas muscle fuel but with more prolonged starvation theyare less well used In fact as acetoacetate is taken up bythe muscle it is returned back to the blood as b-hydroxy-butyrate signifying a more reduced state of the musclemitochondriamdashsecondary to free fatty acid utilizationThus fatty acids appear to take preference for oxidationand ketoacid oxidation ceases sparing the ketoacids forthe brain a superb survival processrdquo

According to Laffel22 healthy adult liver can pro-duce as much as 185 g of ketone bodies per day Ketonessupply 2 to 6 of the bodyrsquos energy needs after anovernight fast and 30 to 40 after a 3-day fast

Effect of Total Starvation and CarbohydrateRestriction on Blood and Urine Ketones

In the late 19th century physiologists focused on thestriking association between ketonuria and worseningdiabetes mellitus Later as methods for detecting andquantifying ketones in biologic uids improved atten-tion shifted to the effect of food deprivation changes inenergy expenditure and modi cations of dietary compo-sition on the urinary excretion and blood levels of ketonebodies It was soon evident that the ketonuria associatedwith total starvation does not become substantial until 3to 5 days have elapsed This is the time it takes for liverglycogen to be thoroughly depleted In 1931 Behre26

reported that ketonuria begins to increase within a fewhours after a meal is skipped In three subjects studied byKartin et al27 during a 6-day period of total starvationblood ketones after 2 days of food deprivation variedfrom 11 to 25 mML After 6 days blood ketones had risento 23 to 40 mML Peters and Van Slyke3 point out thatsuch rises ldquo are not enough to tax severely the mecha-nisms for the preservation of acid-base equilibrium theserum bicarbonate does not fall below 35 to 40 volumes percent Benedictrsquos subject averaged about 6 grams ofb-hydroxybutyricacid in the urine daily for the last 2 weeksof his 31-day fast In diabetic acidosis ketonuria has beenknown to reach 10 times as much as thisrdquo

By accelerating the rate of carbohydrate utilizationhuman physical exercise2328 administration of dinitro-phenol or thyroid hormone to rabbits10 and increasedbovine lactation29 can give rise to appreciable ketosisPregnant women maintain higher than normal ketonelevels and show an exaggerated ketogenic response tofasting30 In human subjects undergoing experimentalstarvation consumption of relatively small amounts ofglucose per day (eg 50ndash75 g) greatly reduces ketonebody production31

A low-carbohydrate diet is not necessarily a keto-genic diet This is particularly true of diets with unre-stricted content of meat and other protein-rich foodsHeinbecker32 reported in 1928 that Baf n Island Eski-mos subsisting on their usual diet of meat (virtually theonly source of carbohydrate in their food was the glyco-gen in seal muscle) showed minimal ketonuria (As areference point muscle glycogen concentration in restinghumans on a mixed diet is approximately 144 gkg33) Itis unlikely that these very small amounts of glycogencould have accounted for the absence of appreciableketonuria A much more likely explanation is that theglucose derived from catabolism of ingested meat pro-tein was suf cient to prevent ketosis McClellan andDuBois34 fed two human volunteers ldquocarbohydrate-freerdquodiets high in meat content (an Eskimo-type diet) formany months in a metabolic ward setting Their ndingsled them to conclude that in persons subsisting on diets

329Nutrition Reviews Vol 61 No 10

330 Nutrition Reviews Vol 61 No 10

very low in carbohydrate ketosis varies inversely withthe quantity of protein eaten This occurs because ap-proximately 48 to 58 of the amino acids in most dietaryproteins are glucogenic3536 For every 2 grams of proteinconsumed in a carbohydrate-free diet somewhere be-tween 10 and 12 grams are potentially convertible to

glucose Therefore to obtain a degree of hyperketonemia(approximately 2ndash7 mML) believed to be therapeuti-cally effective in certain important medical conditionssuch as epilepsy patients must rigorously restrict proteinas well as carbohydrate intake37 and when possibleincrease their level of physical activity23

Table 1 Postabsorptive Plasma Concentrations of Acetoacetate (AcAc) and b-Hydroxybutyrate (BHB)and Rates of Ketone Release and Removal Together with Rate Coefcients for Ketone Removal (meanSD) in Normal Diabetic and Normally Fed and Subsequently Food-deprived Severely Obese Subjects

DiagnosisMetabolic

StateAcAc

(mML)BHB

(mML)

KetoneRelease

Rate( mol

min 1 m 2)

KetoneRemoval

Rate( mol

min 1 m 2)

RateCoef cientfor KetoneRemoval(min 1)

Normal(n 5 11) Postabsorptive 005 002 007 004 81 66 1107 1059 0168 0169

Diabeticdagger

(n 5 6) Postabsorptive 067 075 179 263 208 118 2188 874 0055 0040ObeseDagger

(n 5 10) Postabsorptive 015 006 027 012 171 70 2101 932 0066 0040Food-deprived

1ndash2 weekssect 104 026 597 113 569 286 5673 2562 0027 0019

Fraction lost by oxidation and excretiondaggerRecently diagnosed untreated patients with insulin-dependent diabetes mellitusDaggerSubjects severely obesesectSubjects received 50 kcal and 1 multivitamin tabletdayData from Reference 18

Figure 1 Hepatic formation and peripheral utilization of ketone bodies Mobilization of non-esteri ed fatty acids (NEFA) fromadipose tissue fat stores by the activation of hormone-sensitive lipase is an initial step that provides ample substrate for the hepaticproduction of ketone bodies (b-hydroxybutyrate and acetoacetate) Once fatty acids enter the hepatocyte from the blood circulationthey are activated to CoA esters by the family of membrane-bound fatty acid acyl-CoA synthetases The action of one of thesesynthetases (long-chain acyl CoA synthetase) is identi ed as (3) After CoA esteri cation the metabolic paths diverge based on thefatty acid (acyl) chain length Short- and medium-chain acyl-CoAs readily cross the outer and inner mitochondrial membranes andundergo b-oxidation into acetyl-CoA and acetoacetate (the terminal 4-carbon fragment) Long-chain acyl-CoAs readily cross theouter mitochondrial membrane after which they must be converted to long-chain acylcarnitine by membrane-bound carnitinepalmitoyltransferaseI or CPT-1 (1) for subsequent transport across the inner mitochondrial membrane The inner membranendash boundenzyme carnitine acylcarnitine translocase (4) exchanges one long-chain acyl-CoA from the intermembrane space with one free(unacylated)carnitine from the matrix thereby permitting carnitinersquos return to the intermembrane space The inner membranendash boundenzyme carnitine palmitoyltransferase II or CPT-II (2) converts the fatty acid back into its CoA ester for subsequent b-oxidationfreeing carnitine for recycling Very longndash chain acyl-CoAs and unsaturated fatty acids of various lengths are transported intoperoxisomes along with long- and medium-chain acyl-CoAs for chain-length reduction via b-oxidation Unlike mitochondrialb-oxidation peroxisomal b-oxidation does not proceed to completion This results in peroxisomal export of medium-chain acyl andacylcarnitinecompounds to the mitochondria for further oxidationMitochondrial b-oxidation produces acetyl-CoA which enters theTCA cycle for further oxidation to produce ATP Ketogenesis occurs when hepatic energy requirements are met allowing excessacetyl-CoA to be acted upon by thiolase which condenses two acetyl-CoAs to form acetoacetyl-CoA Hydroxymethylglutyaryl-CoAsynthase (HMG-CoA synthase) condenses acetoacetyl-CoA with acetyl-CoA to form hydroxymethylglutaryl-CoA which is thendissociated into acetoacetate and acetyl-CoA by hydroxymethylglutaryl-CoA lyase (HMG-CoA lyase) The action of theenzyme b-hydroxybutyrate dehydrogenase (5) reduces much of the acetoacetate to b-hydroxybutyrate using NADH as areductant Both acetoacetate and b-hydroxybutyrate are transported back through the mitochondrial membranes for export intothe blood circulation The pathway of lipogenesis illustrated by dashed lines becomes activated in the fed state and results inthe synthesis of fatty acids and their esteri cation into triacylglycerol (triglycerides) The rst committed intermediary in fattyacid synthesis is malonyl-CoA which inhibits CPT-1 (enzyme 1) This inhibition is illustrated as a thin dashed line pointingto CPT-1 Inhibition of CPT-1 prevents mitochondrial uptake of fatty acyl-CoA (and its subsequent b-oxidation) and therebyblocks ketogenesis In the fed (non-ketotic) state triglyceride is exported from the liver as very lowndash density lipoprotein(VLDL)

331Nutrition Reviews Vol 61 No 10

Ketones Substrates in Search of a Mission

By the early 1960s a great deal of information hadaccumulated about the biochemistry and physiology ofthe ketone bodies yet their role in human metabolismremained obscure38 Nevertheless ve facts were knownto medical scientists that had they been viewed in properperspective would have suggested strongly that ketoneshave to be the brainrsquos back-up fuel during starvationFirst scientists recognized that when people are deprivedof food for 3 days or longer they become hyperketone-mic and hyperketonuric Second catheterization studiesshowed that the adult human brain uses approximately100 to 150 g glucose per day39 Third scientists under-stood that non-esteri ed fatty acids do not cross theblood-brain barrier (BBB) and are not directly availableto the brain as an energy source40 Fourth Benedict41

showed that during 31 days of total caloric starvation hisldquofasting manrdquo excreted an average of 872 gday ofnitrogen (N) which corresponds to a rate of body proteincatabolism of approximately 55 gday Approximately32 g of this amount (058 550) could have beenavailable as fuel for the brainmdashapproximately 27 ofthe brainrsquos daily needs In another subject deprived offood for 15 days mean total N excretion was 482 gdaymaking only 174 g of glucoseday potentially availableto the brain42

Finally because glucose was widely assumed to bethe exclusive fuel of the adult human brain38 one mightinfer that during continuing starvation (after liver glyco-gen stores were exhausted) the brain would have to besustained by glucose derived from intracellular proteinMaking 100 to 150 g of glucoseday available to thebrain would require the gluconeogenic breakdown ofsome 172 to 259 gday of body protein At this unsus-tainable rate of protein attrition death might be expectedto occur within approximately 2 weeksmdash certainly notthe 57 to 73 days (mean 61 25 [SEM] days) that itreportedly takes for totally starved young men of averagebody composition to die43 If the nutrition scientists ofthe time had given more thought to the meaning of thisunexplained ldquoprolongationrdquo of life during total starva-tion they might have reasoned that during extended fooddeprivation the brain is able to use the ketones generatedin such abundance as an alternative energy source Sucha scenario would have explained the greater than ex-pected longevity and the far lower than expected dailynitrogen de cit exhibited by starved individuals

In 1966 Cahill et al39 ldquoconnected the dotsrdquo andconcluded that ketones might provide an alternate fuelfor the brain during starvation To test this hypothesisthe authors performed brain catheterization studies inobese patients undergoing 5 to 6 weeks of therapeuticfasting44 This landmark study showed that under condi-tions of sustained starvation (during which the subjects

achieved blood ketone levels close to 7 mML and bloodglucose concentrations of ca 38 mML) BHB and AcAcreplaced glucose as the predominant fuel(s) for brainmetabolism The ldquostarvation paradoxrdquo was thus resolvedIt also became clear that during starvation the liverrsquosability to generate ketones for use by the brain representsa biologic strategy for maintaining strength and prolong-ing life during conditions of severe nutritional privationIn this way the burden of sustaining life during starva-tion is shifted from the bodyrsquos protein stores to itsreserves of fatmdasha far more effective and ef cient adap-tation Thanks to the pioneering studies of Cahill et almetabolismrsquos ugly duckling nally began to emerge asan incipient swan

Ketones Uniquely Efcient Metabolic Fuel

In the mid 1940s Lardy and associates4546 demonstratedthat when tested on sperm BHB and AcAc comparedwith other metabolic substrates have the ability to de-crease oxygen consumption while increasing motilityFor almost half a century this apparent increase in met-abolic ef ciency remained unexplained In 1994 and1995 however RL Veechrsquos group47ndash49 at the NationalInstitutes of Health showed that in the working perfusedrat heart when 5 mM ketone bodiesL were added to theglucose-containing perfusate there was a 25 increasein hydraulic work with a signi cant decrease in oxygenconsumption Addition of ketones also brought about areduction of the mitochrondrial NAD couple with oxi-dation of the mitochondrial co-enzyme Q couple Theconsequence is an increase in the energy of the redoxspan between site I and site II of the electron transportsystem

As Veech et al50 described ldquoAn increase in theredox span between two sites will result in an increasedenergy release by the electron traveling across that span With an increase in redox energy of the respiratorychain there is a corresponding increase in the energy ofthe protons ejected from the mitochondria at the energy-conserving sites which is then re ected in an increase inthe energy of ATP hydrolysisrdquo

As shown in Table 2 Veech and associates47ndash49 alsoobserved that the introduction of ketone bodies into therat heart perfusate produced a 16-fold rise in acetyl-CoAcontent and increases in TCA cycle intermediates Thesestimulatory effects on TCA cycle activity were similar tothose obtained with the administration of saturatingdoses of insulin Insulin increases glucose transport inthe myocardium by promoting translocation of insulin-sensitive GLUT4 to the plasma membrane Insulin alsostimulates the activity of pyruvate dehydrogenase mul-tienzyme complex (PDH)mdasha rate-limiting step in themitochondrial transformation of pyruvate to acetyl-CoAAs Veech et al50 put it ldquoThe implication was that

332 Nutrition Reviews Vol 61 No 10

ketosis which is the physiological response to insulindeprivation during starvation was equivalent in meta-bolic effects to the actions of insulin (Moreover)ketones by-passed the block in glucose transport causedby lack of insulin (and) also by-passed the block inPDH induced by insulin de ciency by providing analternative source of mitochondrial acetyl CoArdquo Thepathway by which ketones are able to achieve this effectis shown in Figure 2

Veech et al50 pointed out that the gradients of Kand Na (and Ca ) between the extra- and intracellularphases of the cell are normally in virtual equilibriumwith the resting electrical potential between these twophases and with the energy ( G) of ATP hydrolysiswhich sustains the action of the Na pump By increas-ing metabolic ef ciency and thereby the energy pro-duced by ATP hydrolysis BHB augments the potentialbetween the extra- and intracellular phases giving rise toan increase in the energy of the K gradient (ratio of[K ]out[K ]in) in muscle and brain A larger mitochon-drial membrane potential would lead to larger energyproduction because it is the mitrochondrial membranepotential difference that drives the production of ATPMaintenance of a suitable K gradient in the cell is ofcourse indispensable to its proper functioning As willbe discussed below impairment of ion homeostasiswithin neuronal cells in certain brain areas might resultin a degree of electrical instability suf cient to trigger anepileptic seizure (or in the heart a cardiac arrhythmia)

Studies of blood ow and oxygen consumption inthe brains of food-deprived obese human subjects re-vealed values (45 mL min 1 100 g 1 and 27 mL O2

min 1 100 mL 1)44 that were well below the normallevels for adult human brains of 57 and 36 respective-ly51 Although these data need con rmation they suggestan increase in the metabolic ef ciency in human brainsusing ketoacids as their principal energy source in placeof glucose

Examples of mechanisms by which diet-inducedhyperketonemia might overcome or circumvent meta-bolic impediments to utilization of such substrates asglucose and pyruvate are shown schematically in Fig-ure 3

Use of a ldquoHyperketogenicrdquo Diet in Treatmentof Seizure Disorders

Fasting has long been acknowledged as a method toprevent or reduce seizures in epileptic individuals37

Because of the obvious limitations of this treatment useof a ketogenic diet to mimic the metabolic effects of fooddeprivation was proposed in 1921 by both Woodyatt52

and Wilder53 at the Mayo Clinic In 1924 Peterman54

(also from Mayo) reported the clinical application andeffectiveness of the regimen suggested by Woodyatt andWilder Petermanrsquos diet provided (per day) 1 g of pro-teinkg body weight (considerably less in adults) 10 to15 g of carbohydrate and the remainder of the calories asfat The Peterman ketogenic diet is sometimes called the41 diet because it provided a ratio of approximately 4parts (by weight) fat to 1 part (by weight) of a mixture ofprotein and carbohydrate This distribution of the majormacronutrients is almost identical to that of most anti-convulsant diets used today37 The term ldquohyperketogenicdietrdquo (HKD) will be used in this review to distinguish the41 diet (and certain variations thereof) from the far lessdrastic low-carbohydrate mildly ketogenic diets thathave been advocated for weight control in a successionof popular diet books

In the 1970s several investigators5556 modi ed thePeterman HKD by substituting medium-chain triglycer-ides (MCT) in varying amounts for long-chain triglycer-ides (MCTrsquos ketogenic effect is not dependent on con-current restriction of dietary carbohydrate57) Althoughall such variations of the HKD were approximately equalin reducing seizures many of the children given MCT-

Table 2 Insulin and Ketone Effects on Cardiac Efciency Acetyl-CoA Production [NADH][NAD ] Ratioand Free Energy of ATP Hydrolysis ( GA TP ) in the Perfused Rat Heart

Substrates Addedto Perfusate

AcetylCoAdagger

[NADH][NAD ]Daggersect

(mitochondrial)

G of ATPHydrolysis(cytosolic)

kJmol

CardiacEf ciency

()

CardiacEf ciency( relativeto glucose)

Glucose 1 005 562 105 100Glucose Insulin 9 022para 589para 134para 128Glucose Ketones 15 062para 576para 130para 124Glucose Insulin

Ketones 18 043para 589para 143para 136

Data of Sato K et al4 8 and DaggerKashiwaya et al4 7 4 9

daggerNumbers represent multiples of that measured with perfusate containing only glucosesectExpressed by Kashiwaya et al4 7 4 9 as [NAD ][NADH]paraIndicates a signi cant difference (P 005) from perfusate containing 10 mmolL glucose ketones added at 4 mMLD-b-hydroxybutyrate 1 mML acetoacetate

333Nutrition Reviews Vol 61 No 10

augmented diets complained of nausea diarrhea andoccasional vomiting58

Among those patients who adhere to the 41 HKDfor 12 months 40 experience greater than 90 reduc-tion in seizures whereas an additional 40 experience50 to 90 reduction Although the precise mechanism ofaction responsible for the clinical ef cacy of the diet hasnever been clari ed it is likely that the hyperketonemiainduced by the 41 HKD is the key factor responsible forthe regimenrsquos anticonvulsant effect rather than the ac-companying metabolic acidosis59 or dehydration37 In

1972 Uhleman and Neims60 demonstrated that micepups fed a high-fat ketogenic diet exhibited signi cantlyincreased thresholds against investigator-applied electro-shock challenges This resistance to electroshock disap-peared promptly after the animals were shifted to ahigh-carbohydrate diet Subsequently a number of stud-ies have shown an increase in the electroconvulsivethreshold in rats maintained on a high-fat diet or de-prived of food for 48 hours De Vivo et al61 found thatthe ratios of ATP to adenosine diphosphate (ADP) weresigni cantly increased in the brains of rats fed the high-

Figure 2 Production of ATP (adenosine triphosphate) in extrahepatic tissues arising from use of ketones and glucose Glucose isregulated by a family of GLUT transporter proteins GLUT4 transporters found in most body tissues are mediated by insulinGLUT1 transporters common in tissues such as brain and erythrocytes allow glucose uptake without insulin mediation By contrastuptake of ketone bodies (shown as BHB and AcAc) occurs via the family of monocarboxylate transporters (MCTs) which are notinsulin mediated MCT proteins enable ketones to pass readily through the blood-brain barrier Many types of peripheral cellsincluding brain cells not only use glucose but also use ketones to produce acetyl-CoA Glucose utilization depends on glycolyticenzymes (not shown) which produce pyruvate Pyruvate in turn enters the mitochondrion and is converted to acetyl-CoA by thepyruvate dehydrogenase multienzyme complex (PDH) Ketone substrates enter the mitochondrion in the form of (1) acetoacetate(AcAc) and (2) b-hydroxybutyrate (BHB) which is readily converted to acetoacetate by b-hydroxybutyrate dehydrogenase (HBD)Two enzymes principally control ketolysis First 3-oxoacid-CoA transferase (OCT) adds coenzyme-A to AcAc which is then splitinto two molecules of acetyl-CoA by acetoacetyl-CoA thiolase (ACT) Acetyl-CoA undergoes oxidative degradation in the TCAcycle to reduce the electron carrriers NAD (nicotinamide adenine dinucleotide) and FAD ( avin adenine dinucleotide) to NADHand FADH2 respectively Acetyl-CoA is also used for lipid or sterol synthesis (not shown) NADH and FADH2 donate electrons tothe protein complexes I and II of the mitochondrial respiratory chain The ow of these electrons down the electron-transport chainto oxygen (O2) allows complexes I III and IV to pump protons (shown as H ) out of the matrix and into the intermembrane spacewhich generates a proton motive force (pmf) between the intermembrane space and the matrix The pmf provides the energy to driveprotons back through complex V (ATP synthase) to produce ATP from ADP (adenosine diphosphate) and inorganic phosphate (Pi)

334 Nutrition Reviews Vol 61 No 10

fat diet They concluded ldquo the anticonvulsant propertyof the ketogenic diet derives from the observed increasein cerebral energy reserve which in turn re ects thehigher ATPADP ratio [this] might provide a reason-able explanation for the increased neuronal stability thatdevelops during chronic ketosisrdquo In 1999 Pan et al62

carried out 31P magnetic resonance spectroscopic imag-ing studies in seven patients with intractable epilepsybefore and after institution of a ketogenic diet Theymeasured signi cant increases (P 005) in high-energy phosphates in association with adherence to theketogenic diet in these subjects They concluded thatenhancement of brain energy metabolism occurs withuse of the ketogenic diet in humans as in lab animals

Thavendiranathan et al63 have recently reportedstudies of the effect of the ldquoclassicrdquo ketogenic diet onseizures triggered by two threshold testsmdashthe pentyle-netetrazol infusion test and the threshold electroconvul-sive shock test These authors observed small but signif-icant threshold elevations (15ndash20) against bothchemically and electrically triggered seizures Similarlyprotective effects were not seen when suprathresholdstimuli were employed

Janigro64 summarized studies indicating that there isa reduction in both glucose uptake and metabolic activityin seizure foci in epileptic patients In his words ldquoDi-minished ion homeostasis together with increased meta-bolic demand of hyperactive neurons may further aggra-vate neuropathological consequences of BBB loss ofglucose uptake mechanismsrdquo Therefore one explanationfor the effectiveness of the 41 HKD in preventingseizures could be the increased transport of BHB at theBBB and its increased availability as an energy source tothe affected neurons BHB is not only an alternate sourceof energy for seizure foci in the brainmdashbypassing animpairment of carbohydrate utilization such as thatwhich can occur from a block in mitochondrial PDHactivitymdashbut it may actually promote ketone and glu-cose uptake in such foci In rats diet-induced ketosisproduces an eightfold increase in the activity of themonocarboxylate transporter and increases the level ofthe glucose transporter GLUT1 in brain endothelialcells and neuropil65 The hypothesis put forth by Janigroto explain how ketosis might eliminate or reduce seizuresis supported by reports that a suitable ketogenic diet isthe treatment of choice for two rare conditions that may

Figure 3 Conditions that can impede acetyl-CoA production from pyruvate and its precursors Examples include (1) insuf cientinsulin (2) insulin receptor abnormalities (3) diminished transport of glucose owing to impaired release of GLUT4 transportersassociated with insulin resistance [see below] (4) pyruvate dehydrogenase complex (PDH) de ciency or dysfunction (5)unavailabilityof glucose at the cell membrane and (6) GLUT1 mutations Provision of ketones to the cell can effectively circumventor overcome such impediments and compensate for the resulting de cits of glycolytically derived acetyl-CoA (GLUT1 designatesthe insulin-independent membrane protein responsible for glucose transport in red blood cells and throughout the blood-brain barrier[BBB] GLUT4 designates the insulin-dependentglucose transporterprotein expressed in most cells Many forms of insulin resistanceentail putative post-receptor defects of insulin-signal transduction resulting in the failure of insulin to release GLUT4 transportersfrom the GLUT4 pool at a rate suf cient to bind and transport enough extracellular glucose to maintain cellular pyruvate levels)

335Nutrition Reviews Vol 61 No 10

be complicated by seizuresmdashthe facilitated glucosetransporter protein type 1 (GLUT1) de ciency syn-drome6667 and the PDH complex de ciency syndrome68

Treatment with a ketogenic diet reportedly brought aboutdramatic clinical improvement in a one-year-old boywith Leigh syndrome which is associated with a de -ciency of PDH69 In this child treatment with a keto-genic diet was associated with striking improvement ofthe cerebral lesions as measured by neuroimaging

In 1973 Buckley and Williamson70 proposed thatneural lipid synthesis preferentially uses AcAc overglucose Patel and Owen71 obtained con rmatory datafor this hypothesis in 1976 In 1984 Koper et al72

compared the rate of incorporation of AcAc and glucoseinto fatty acids and cholesterol using labeled AcAc andglucose At saturating levels of glucose the authorsobserved that the rate at which AcAc label incorporatedinto fatty acids and cholesterol exceeded that of glucoseby a factor of 5 to 10 in both oligodendrocytes andastrocytes These observations provide a basis for argu-ing that some of the bene ts associated with ketogenicdiets may be related to their enhancing effects on lipo-genesis andor sterol synthesis (ie damaged neuronsmight repair more readily in brains given an adequateketone supply)

Therapeutic Potential of Diet-inducedHyperketonemia in Neurodegenerative andOther Disorders

Veech et al50 suggest that maintenance of a suitabledegree of diet-induced hyperketonemia (in the range of2ndash7 mML) has the potential to aid management ofseveral neurologic diseases including common neurode-generative diseases like Alzheimerrsquos disease (AD) andParkinsonrsquos disease (PD) In their ldquohypothesis paperrdquothe authors single out AD PD Friedreichrsquos AtaxiaLeprechaunisms and lesser forms of insulin disorders asexamples of illnesses that might bene t from therapeuticketosis They also describe experience with one case ofLaFora body disease (until now a uniformly fatal geneticdisease associated with progressive dementia) in whichuse of the 41 HKD resulted in unprecedented improve-ment in cognition gait and awareness after one month73

In this disease mutations in a gene encoding a novelprotein tyrosine phosphatase cause progressive myo-clonic epilepsy

Calabrese et al74 and DiMauro and Schon75 haverecently called attention to evidence suggesting the mi-tochondrial genome may play a key role in neurodegen-erative diseases Decreases in respiratory chain complexactivities have been identi ed in AD PD amyotrophiclateral sclerosis and Huntingtonrsquos disease Mitochondri-opathies (MCPs) possibly associated with oxidantanti-oxidant balance perturbation may underlie defects in

energy metabolism and contribute to cellular damage anddestruction

Parkinsonrsquos DiseaseAlthough the etiology of PD is complex involving bothgenetic and environmental factors there is growing in-terest in the role of mitochondrial dysfunction in thepathogenesis of this disorder Schapira76 pointed out thata proportion of PD patients exhibit a de ciency in mito-chondrial respiratory chain function most notably adefect of complex I within neurons of the substantianigra In his words ldquoBoth mitochondrial mutations andtoxic agents (endogenous and exogenous) have beendemonstrated to cause a de ciency in complex I func-tion This is achieved by increasing the degree of orsusceptibility to oxidative stress which may contributeto apoptopic cell deathrdquo

Finsterer77 reported metabolic studies indicating thatPD can be a manifestation of MCP In his series 63 ofPD patients had features of generalized MCP howeverhe used lactate stress testing to detect patients withMCPmdashan approach that could not be expected to pick upmitochondrial lesions limited to brain sites

According to Veech et al ldquoPD appears to resultfrom an acquired defect in the mitochondrial NADHmultienzyme complex of the dopaminergic cells of themesencephalonrdquo78 They point out although the diseaseis treatable for a period by administration of levodopa(the metabolic precursor of dopamine) continuing freeradical damage eventually reduces the effectiveness ofthis therapy Kashiwaya et al79 reported that primarycultures of mesencephalic dopaminergic neurons ex-posed to the meperidine analogue MPP (a molecule thatinhibits NADH dehydrogenase [complex 1] and causesoxygen free radical formation) can be protected fromdeath by addition of 4 mM D-b-hydroxybutyrate Veechet al50 suggest that BHB probably acts by decreasing thesource of mitochondrial oxygen radical formationldquo by oxidizing the coenzyme Q couple while at thesame time reducing the redox potential of the NADPcouple which through glutathione is the nal detoxi -cation step of H2O2rdquo In a recent review Schon andManfredi80 wrote ldquoBecause complexes I and III are theprincipal sources of free radicals in the cell alteredcomplex I function in the substantia nigra pars compactacould be responsible for the increased DNA damage andlipid peroxidation found in PD brainsrdquo

To the extent that ketones crossing the BBB couldcircumvent or compensate for mitochondrial defects as-sociated with PD or could protect dopaminergic neuronsfrom free radical damage a suitable ketogenic diet mightwell have bene cial effects in certain patients sufferingfrom Parkinsonism

336 Nutrition Reviews Vol 61 No 10

Alzheimerrsquos DiseaseIn 1996 Ogawa et al81 determined regional cerebralblood ow and studied energy utilization in the brains ofseven patients with AD by measuring arteriovenousdifferences of an array of substrates Single-photon emis-sion computed tomography showed that the patients hadsigni cantly decreased regional cerebral blood ow lim-ited to the parietotemporal region The glucose extractionfraction and global cerebral metabolic rate of glucosewere signi cantly decreased whereas the global cerebralmetabolic rate of oxygen was only slightly decreasedBecause the cerebral metabolic rates of the various sub-strates (ketones lactate etc) did not change the authorsconcluded that the markedly elevated oxygenglucoseutilization ratio indicated an altered energy metabolismin AD

Conceivably the impairment of glucose metabolismin AD patients described by Ogawa et al is attributableat least in part to local insulin resistance Watson andCraft82 called attention to emerging evidence that insulinabnormalities and insulin resistance in AD may contrib-ute to the pathophysiology and clinical symptoms of thisdisorder In their words ldquoIt has been recently demon-strated that insulin-sensitive glucose transporters are lo-calized in the same regions supporting memory and thatinsulin plays a role in memory functions Collectivelythese ndings suggest that insulin may contribute tonormal cognitive functioning and that insulin abnormal-ities may exacerbate cognitive impairments such asthose associated with ADrdquo The authors suggest thatimproving insulin effectiveness may have therapeuticbene t for patients with AD The suggestion of Watsonand Craft that insulin abnormalities may play a role inthe pathogenesis of AD has to be reconciled with the nding by Molina et al83 that cerebrospinal uid insulinlevels in 27 AD patients did not differ from those in 16matched controls

In 1996 Hoshi et al84 reported that a fragment ofthe beta chain of amyloid inhibits PDH activity in pri-mary cultures of hippocampal neurons A year later thesame authors found that the 1ndash42 fragment of the betachain of amyloid Ab1ndash42 can inhibit acetyl cholinesynthesis of septal neurons85 Veech et al50 note thatcultured hippocampal neurons die when exposed toAb1ndash42 However addition of 4 mM D-b-hydroxybu-tyrate protects these neurons from amyloid fragmentndashinduced death Casley et al86 showed that beta-amyloidcauses a signi cant reduction in state 3 and state 4mitochondrial respiration which is diminished further bynitric oxide These authors also found that cytochromeoxidase -ketoglutarate dehydrogenase and PDH activ-ities are inhibited by b-amyloid They concluded thatb-amyloid can directly disrupt mitochondrial function

inhibit key enzymes and contribute to the impairment ofenergy metabolism observed in AD

Veech et al50 contend that the foregoing studiesprovide a rationale for trial of a ketogenic diet in thetreatment of AD Ketones can overcome localized insulinresistance can bypass a block in PDH and have beenshown to protect dopaminergic neurons from free radicaldamage and from apoptosis induced by the 1ndash42 amy-loid fragment in tissue-culture experiments

What Next

In considering the potential usefulness of therapeutichyperketonemia in the treatment of AD PD and certainother neurodegenerative conditions it is essential to beclear about what is meant by terms like ldquoketogenic dietrdquoand ldquohyperketonemiardquo

There is widespread confusion among both physi-cians and lay individuals about what constitutes a keto-genic diet Similar to glucose ketones are present in theblood at all times As shown in Table 3 the range ofketone levels that can be achieved by prolonged fooddeprivation or by adherence to various diets is quitewide Hence it makes no sense to speak of a ldquoketogenicdietrdquo without also specifying the degree of serum ketoneelevation that the diet is intended to achieve

The serum concentrations sought for treatment ofepilepsy range from 2 to 7 mML Such levels can beachieved only by strict adherence to a regimen like the41 HKDmdashone most patients nd burdensome On theother hand the kind of low-carbohydrate ldquoketogenicdietrdquo (eg 30 protein 8 carbohydrate 61 fat [asproportions of total energy]) that has recently becomepopularmdashalthough much easier to followmdashachievesBHB levels of only 028ndash040 mML87 (This ldquopopularrdquoketogenic diet at a 2000 kcalday intake provides 127 gof presumptive carbohydrate [40 preformed] plus 87 gderivable from 150 g protein not including the smallquantity of glycerol released in the course of adipocytetriglyceride hydrolysis At the same energy level the 41HKD would yield 33 g of presumptive carbohydrate[10 g preformed] plus 23 g derivable from 40 g proteinThus the effective quantity of carbohydrate provided bythe popular ketogenic diet is almost four times thatsupplied by the 41 HKD)

Future research must address the problem of match-ing diet-induced elevation of serum ketone levels withtherapeutic bene ts Because the drastic type of hyper-ketogenic dietmdashthe one used to treat epilepsy and relatedneurologic disordersmdashis so onerous and dif cult to fol-low it is especially important to determine the extent towhich such a diet can be liberalized without losingdesirable therapeutic effects Given the daunting natureof the 41 HKD it is encouraging that there exists apotential nutritional stratagem which eventually may

337Nutrition Reviews Vol 61 No 10

make it possible to induce therapeutic hyperketonemia inpatients without requiring them to adhere to a Draconianregimen The stratagem entails administration by mouthof substantial quantities of ketones in a formmdashnamelyas esters or small polymersmdashsuitable for ingestion Intheir article about the potential therapeutic uses of ketonebodies Veech et al50 suggest that the problems associ-ated with ldquothe very unappetizing high-fat ketogenicdietrdquo might be obviated by feeding patients up to 100 to150 g per day of a nutritionally acceptable form ofketone bodies Unfortunately neither ketone esters noroligomers of BHB are presently available in suf cientquantities for the conduct of clinical trials

The traditional 41 HKD is dif cult but far fromunmanageable Numerous epileptic children and adultswith seizures resistant to anticonvulsant medication havemanaged to remain on this regimen for years withoutexperiencing serious adverse effects There are justi ableconcerns about its hyperlipidemic effect however andthe fact that the diet has been reported to increase risk ofside effects including nephrolithiasis88 depression ofplatelet function (manifested by easy bruising)89 andoccasionally prolongation of the electrocardiographicQTc interval in children90 It should be possible to mitigateor perhaps prevent the adverse effects of this very highndashfatdiet on plasma lipids by modifying its fatty acid contentWhat is most important now however is to get on with thejob of determining whether diet-induced hyperketonemia(at any level) will prove clinically bene cial in the man-agement of a number of neurodegenerative disorders par-ticularly Parkinsonrsquos disease and Alzheimerrsquos disease

Acknowledgment

We thank Richard L Veech MD DPhil for hisencouragement and very helpful suggestions We alsothank Steven W Fowkes for creative and elegant ren-dering of Figures 1ndash3 and editorial advice

1 Gerhardt J Diabetes mellitus und aceton WienerMedizinische Presse 18656672

2 Hirschfeld F Beobachtungen uber die Acentonurieund das Coma diabeticum Z klin Med 189528176 ndash209

3 Peters JP Van Slyke DD Quantitative ClinicalChemistry Interpretations Volume 1 2nd editionBaltimore Williams amp Wilkins 1946439

4 Cohen PP Studies in ketogenesis J Biol Chem1937119333ndash346

5 Cohen PP Stark IE Hepatic ketogenesis and ketol-ysis in different species J Biol Chem 193812697ndash107

6 Jowett M Quastel JH Fat metabolism III The for-mation and breakdown of acetoacetic acid in animaltissues Biochem J 1935292181ndash2191

7 Chaikoff IL Soskin S The utilization of acetoaceticacid by normal and diabetic dogs before and afterevisceration Am J Physiol 19288758 ndash72

8 Barnes RH Drury DR Greeley PO Wick AN Utili-zation of the ketone bodies in normal animals and inthose with ketosis Am J Physiol 1940130144 ndash150

9 Mirsky IA Broh-Kahn RH The inuence of in-creased metabolism on b-hydroxybutyric acid utili-zation Am J Physiol 1937120446 ndash 450

10 Waters ET Fletcher JP Mirsky IA The relationbetween carbohydrate and b-hydroxybutyric acid

Table 3 Relationship of Metabolic State and Diet Composition to Plasma Levels of Acetoacetate (AcAc)and b-Hydroxybutyrate (BHB) in Normal-weight Overweight and Obese Adults

Metabolic StateDiet

Composition CommentsAcAc

(mML)BHB

(mML) References

Postabsorptive Malesn 5 11 Normal weight 005 007 Hall et al18

n 5 12 Normal weight 008 Sharman et al87

n 5 10 Obese 015 027 Hall et al18

n 5 6 Overweight 001 002 Cahill et al39

Ketogenic diets Males (normal weight)(eucaloric) 3 weeks on diet (n 5 12) 040 Sharman et al87

C8 P30 F61 6 weeks on diet 028 Sharman et al87

Female (normal weight)C2 P8 F90 2 weeks on diet (n 5 1) 070 602 VanItallie Nonas amp Heyms elddagger

Fasting Malesn 5 6 (overweight)

1-week fast 095 358 Cahill et al39

n 5 6 (obese)1ndash2-week fast 104 597 Hall et al18

Distribution (percent total calories) of carbohydrate (C) protein (P) and fat (F)daggerUnpublished study (St Lukersquos-Roosevelt Hospital 2003)Where n 1 all values are means

338 Nutrition Reviews Vol 61 No 10

that there are always measurable concentrations of ke-tones in serum and urine Regardless of season the meanpostabsorptive blood serum level for this group was 07mML (It is now evident that the method used byJohnson et al to measure serum ketones systematicallyoverestimated ketone concentrations When currentlypreferred enzymatic methods are used blood ketonelevels after an overnight fast rarely exceed 05 mML andare usually much lower)

The postabsorptive serum levels of ketones thatJohnson and associates thought normal should be con-trasted with the dramatically higher concentrations (inmML) of BHB (ca 44ndash55) and AcAc (ca 10) thatoccur after a week or more of fasting in non-diabeticobese individuals12 Among hospitalized patients in se-vere diabetic ketoacidosis initial BHB levels of 230mML are not unusual13

By 1949 investigators knew that two molecules ofacetyl-CoA (the primary metabolic intermediate arisingfrom b-oxidation of fatty acids) could condense to formAcAc14 The principal pathway by which AcAc isformed in the liver from acetoacetyl-CoA remained un-clear until 1958 however when Lynen et al15 reportedelegant studies indicating that deacylation of acetoacetyl-CoA did not take place directly but involved two steps(1) condensation of acetoacetyl-CoA with acetyl-CoA toform b-hydroxy-b-methylglutaryl-CoA (HMG-CoA)and (2) cleavage of HMG-CoA to form AcAc and acetyl-CoA As shown in Figure 1 the pathway proposed byLynen et al is the one currently accepted Formation ofHMG-CoA from acetoacetyl CoA is catalyzed by mito-chondrial HMG-CoA synthase a step that is stimulatedby starvation low levels of insulin and consumption ofa low-carbohydratelow-protein high-fat diet AcAc isproduced from HMG-CoA by HMG-CoA lyase16 Mech-anisms by which ketones are formed in the liver and usedin extrahepatic tissues are shown in Figure 1

Ketone Body Kinetics in Humans

As measured by Wastney and associates17 the ketonebodies have a turnover time in the blood circulation ofapproximately 2 minutes From the circulation theyrapidly enter a compartment (probably the body cellmass) that turns over approximately once a minute In1984 Hall et al18 reported measurements of postabsorp-tive plasma concentrations and kinetics of AcAc andBHB release and removal in 11 normal subjects 6patients with newly diagnosed and untreated insulin-dependent (type 1) diabetes and 10 severely obese indi-viduals who were rst studied in the postabsorptive stateand then again after 1 to 2 weeks of starvation (Table 1)

As shown in Table 1 the mean postabsorptiveplasma ketone levels (012 mML) for normal subjects

studied by Hall et al were well below the mean of 07mML reported by Johnson et al11 as ldquonormalrdquo Hall etal measured AcAc and BHB enzymatically adapting theprocedure described by Williamson et al in 196219 Inthe normal subjects mean rates of ketone release andremoval were of the order of 100 mmol min 1 m 2 Inthe patients with newly diagnosed untreated type 1diabetes mellitus the mean postabsorptive ketone levelwas 246 mML approximately 20 times higher than thatin normal subjects Mean ketone release rate in thediabetic patients was approximately triple that of thenormal subjects Mean ketone removal rate in the dia-betic patients was approximately double that of thenormal subjects however the mean rate coef cient forketone removal (the fraction of the total uptake lost byoxidation and excretion) was reduced to 13 of normal

In the 10 severely obese subjects the mean postab-sorptive plasma ketone level (042 mML) was higherthan that in the normal individuals The obese subjectsalso exhibited substantially higher than normal meanrates of ketone release and removal (171 vs 81 and 210vs 111 mmol min 1 m 2 respectively) Similar to thediabetic patients the obese subjects showed a lower thannormal mean rate coef cient for ketone removal

In the obese subjects who had been deprived of foodfor 1 to 2 weeks the mean postabsorptive ketone levelwas 701 mML almost 60 times normal At the sametime mean ketone release and removal rates in thefood-deprived obese individuals were approximately vetimes normal whereas the rate coef cient for ketoneremoval was reduced to approximately 16 of normalWhenever ketone body levels were elevated in the dia-betic and obese subjects the increase was almost alwaysin the form of BHB

These observations indicate that ketone bodies arevery rapidly cleared from the circulation and metabo-lized by extrahepatic tissues In untreated patients withtype 1 diabetes ketone production and removal rates aregreatly augmented In the presence of such high plasmaketone levels however the mean rate coef cient forketone removal is signi cantly reduced This reductionin the rate coef cient for ketone removal (also observedin the postabsorptive and food-deprived obese subjects)was found by Hall et al18 to be coupled with an increasein the interconversion between the ketones with a largerfraction of AcAc being converted to BHB than thereverse (we now know that this re ects reduction of thefree [NAD ][NADH] ratio with which ketones are innear equilibrium)20

The biochemical mechanism underlying the in-creased ketonemia that occurs when food deprivation isprolonged is incompletely understood Because the lipol-ysis rate after 5 weeks remains essentially unchangedfrom that seen during the rst few days of food depriva-

328 Nutrition Reviews Vol 61 No 10

tion and the rate of ketogenesis also remains un-changed21 it is evident that the rate of peripheral clear-ance of ketones must diminish in order to account for theelevated blood ketone levels measured during the food-deprivation period

Laffel22 suggested that the inverse relationship be-tween level of ketonemia and rate coef cient for ketoneremoval could arise from down-regulation of an enzymeinvolved in ketone utilization in response to a rise inintracellular levels of AcAc during starvation in associ-ation with dietary carbohydrateprotein restriction or intype 1 diabetes This explanation seems reasonable how-ever the evidence thus far is inconclusive that any suchenzyme is suf ciently inhibited or decreased to accountfor the observed diminution in ketone utilization It isnoteworthy however that the brain adapts to starvationwith an increased rate of ketone utilization and that anincrease in 3-oxoacid-CoA transferase (OCT) activityhas been demonstrated in the brain of the starved rat Bycontrast decreased OCT activity has been described inthe skeletal muscle of rats with diabetes of more than 10daysrsquo duration16 To complicate matters further exercisetraining for a number of weeks appears to increase thecapacity of muscle to use ketones Because exercise alsopromotes ketogenesis the muscles presumably accom-modate to such increased production by inducing oractivating relevant ketone-utilizing enzymes1623

The physiologic explanation for the increased ke-tonemia during the early phase of food deprivation ap-pears to lie in data reported by Owen and Reichard24

who showed that the rise in blood ketones during the rst2 weeks of starvation results from a decreasing rate ofketone utilization in muscle As food deprivation contin-ues a dramatic change occurs in glucose and ketoneutilization by the nervous system and muscle Glucoseutilization in the brain drops to 13 that in the fed statewith ketones lling the energy gap A nearly reciprocalrelationship exists for muscle which seems to adapt sofully to prolonged fasting that it virtually ignores ketonesas an energy source drawing instead on fatty acids21 AsCahill25 stated ldquo early in starvation ketoacids serveas muscle fuel but with more prolonged starvation theyare less well used In fact as acetoacetate is taken up bythe muscle it is returned back to the blood as b-hydroxy-butyrate signifying a more reduced state of the musclemitochondriamdashsecondary to free fatty acid utilizationThus fatty acids appear to take preference for oxidationand ketoacid oxidation ceases sparing the ketoacids forthe brain a superb survival processrdquo

According to Laffel22 healthy adult liver can pro-duce as much as 185 g of ketone bodies per day Ketonessupply 2 to 6 of the bodyrsquos energy needs after anovernight fast and 30 to 40 after a 3-day fast

Effect of Total Starvation and CarbohydrateRestriction on Blood and Urine Ketones

In the late 19th century physiologists focused on thestriking association between ketonuria and worseningdiabetes mellitus Later as methods for detecting andquantifying ketones in biologic uids improved atten-tion shifted to the effect of food deprivation changes inenergy expenditure and modi cations of dietary compo-sition on the urinary excretion and blood levels of ketonebodies It was soon evident that the ketonuria associatedwith total starvation does not become substantial until 3to 5 days have elapsed This is the time it takes for liverglycogen to be thoroughly depleted In 1931 Behre26

reported that ketonuria begins to increase within a fewhours after a meal is skipped In three subjects studied byKartin et al27 during a 6-day period of total starvationblood ketones after 2 days of food deprivation variedfrom 11 to 25 mML After 6 days blood ketones had risento 23 to 40 mML Peters and Van Slyke3 point out thatsuch rises ldquo are not enough to tax severely the mecha-nisms for the preservation of acid-base equilibrium theserum bicarbonate does not fall below 35 to 40 volumes percent Benedictrsquos subject averaged about 6 grams ofb-hydroxybutyricacid in the urine daily for the last 2 weeksof his 31-day fast In diabetic acidosis ketonuria has beenknown to reach 10 times as much as thisrdquo

By accelerating the rate of carbohydrate utilizationhuman physical exercise2328 administration of dinitro-phenol or thyroid hormone to rabbits10 and increasedbovine lactation29 can give rise to appreciable ketosisPregnant women maintain higher than normal ketonelevels and show an exaggerated ketogenic response tofasting30 In human subjects undergoing experimentalstarvation consumption of relatively small amounts ofglucose per day (eg 50ndash75 g) greatly reduces ketonebody production31

A low-carbohydrate diet is not necessarily a keto-genic diet This is particularly true of diets with unre-stricted content of meat and other protein-rich foodsHeinbecker32 reported in 1928 that Baf n Island Eski-mos subsisting on their usual diet of meat (virtually theonly source of carbohydrate in their food was the glyco-gen in seal muscle) showed minimal ketonuria (As areference point muscle glycogen concentration in restinghumans on a mixed diet is approximately 144 gkg33) Itis unlikely that these very small amounts of glycogencould have accounted for the absence of appreciableketonuria A much more likely explanation is that theglucose derived from catabolism of ingested meat pro-tein was suf cient to prevent ketosis McClellan andDuBois34 fed two human volunteers ldquocarbohydrate-freerdquodiets high in meat content (an Eskimo-type diet) formany months in a metabolic ward setting Their ndingsled them to conclude that in persons subsisting on diets

329Nutrition Reviews Vol 61 No 10

330 Nutrition Reviews Vol 61 No 10

very low in carbohydrate ketosis varies inversely withthe quantity of protein eaten This occurs because ap-proximately 48 to 58 of the amino acids in most dietaryproteins are glucogenic3536 For every 2 grams of proteinconsumed in a carbohydrate-free diet somewhere be-tween 10 and 12 grams are potentially convertible to

glucose Therefore to obtain a degree of hyperketonemia(approximately 2ndash7 mML) believed to be therapeuti-cally effective in certain important medical conditionssuch as epilepsy patients must rigorously restrict proteinas well as carbohydrate intake37 and when possibleincrease their level of physical activity23

Table 1 Postabsorptive Plasma Concentrations of Acetoacetate (AcAc) and b-Hydroxybutyrate (BHB)and Rates of Ketone Release and Removal Together with Rate Coefcients for Ketone Removal (meanSD) in Normal Diabetic and Normally Fed and Subsequently Food-deprived Severely Obese Subjects

DiagnosisMetabolic

StateAcAc

(mML)BHB

(mML)

KetoneRelease

Rate( mol

min 1 m 2)

KetoneRemoval

Rate( mol

min 1 m 2)

RateCoef cientfor KetoneRemoval(min 1)

Normal(n 5 11) Postabsorptive 005 002 007 004 81 66 1107 1059 0168 0169

Diabeticdagger

(n 5 6) Postabsorptive 067 075 179 263 208 118 2188 874 0055 0040ObeseDagger

(n 5 10) Postabsorptive 015 006 027 012 171 70 2101 932 0066 0040Food-deprived

1ndash2 weekssect 104 026 597 113 569 286 5673 2562 0027 0019

Fraction lost by oxidation and excretiondaggerRecently diagnosed untreated patients with insulin-dependent diabetes mellitusDaggerSubjects severely obesesectSubjects received 50 kcal and 1 multivitamin tabletdayData from Reference 18

Figure 1 Hepatic formation and peripheral utilization of ketone bodies Mobilization of non-esteri ed fatty acids (NEFA) fromadipose tissue fat stores by the activation of hormone-sensitive lipase is an initial step that provides ample substrate for the hepaticproduction of ketone bodies (b-hydroxybutyrate and acetoacetate) Once fatty acids enter the hepatocyte from the blood circulationthey are activated to CoA esters by the family of membrane-bound fatty acid acyl-CoA synthetases The action of one of thesesynthetases (long-chain acyl CoA synthetase) is identi ed as (3) After CoA esteri cation the metabolic paths diverge based on thefatty acid (acyl) chain length Short- and medium-chain acyl-CoAs readily cross the outer and inner mitochondrial membranes andundergo b-oxidation into acetyl-CoA and acetoacetate (the terminal 4-carbon fragment) Long-chain acyl-CoAs readily cross theouter mitochondrial membrane after which they must be converted to long-chain acylcarnitine by membrane-bound carnitinepalmitoyltransferaseI or CPT-1 (1) for subsequent transport across the inner mitochondrial membrane The inner membranendash boundenzyme carnitine acylcarnitine translocase (4) exchanges one long-chain acyl-CoA from the intermembrane space with one free(unacylated)carnitine from the matrix thereby permitting carnitinersquos return to the intermembrane space The inner membranendash boundenzyme carnitine palmitoyltransferase II or CPT-II (2) converts the fatty acid back into its CoA ester for subsequent b-oxidationfreeing carnitine for recycling Very longndash chain acyl-CoAs and unsaturated fatty acids of various lengths are transported intoperoxisomes along with long- and medium-chain acyl-CoAs for chain-length reduction via b-oxidation Unlike mitochondrialb-oxidation peroxisomal b-oxidation does not proceed to completion This results in peroxisomal export of medium-chain acyl andacylcarnitinecompounds to the mitochondria for further oxidationMitochondrial b-oxidation produces acetyl-CoA which enters theTCA cycle for further oxidation to produce ATP Ketogenesis occurs when hepatic energy requirements are met allowing excessacetyl-CoA to be acted upon by thiolase which condenses two acetyl-CoAs to form acetoacetyl-CoA Hydroxymethylglutyaryl-CoAsynthase (HMG-CoA synthase) condenses acetoacetyl-CoA with acetyl-CoA to form hydroxymethylglutaryl-CoA which is thendissociated into acetoacetate and acetyl-CoA by hydroxymethylglutaryl-CoA lyase (HMG-CoA lyase) The action of theenzyme b-hydroxybutyrate dehydrogenase (5) reduces much of the acetoacetate to b-hydroxybutyrate using NADH as areductant Both acetoacetate and b-hydroxybutyrate are transported back through the mitochondrial membranes for export intothe blood circulation The pathway of lipogenesis illustrated by dashed lines becomes activated in the fed state and results inthe synthesis of fatty acids and their esteri cation into triacylglycerol (triglycerides) The rst committed intermediary in fattyacid synthesis is malonyl-CoA which inhibits CPT-1 (enzyme 1) This inhibition is illustrated as a thin dashed line pointingto CPT-1 Inhibition of CPT-1 prevents mitochondrial uptake of fatty acyl-CoA (and its subsequent b-oxidation) and therebyblocks ketogenesis In the fed (non-ketotic) state triglyceride is exported from the liver as very lowndash density lipoprotein(VLDL)

331Nutrition Reviews Vol 61 No 10

Ketones Substrates in Search of a Mission

By the early 1960s a great deal of information hadaccumulated about the biochemistry and physiology ofthe ketone bodies yet their role in human metabolismremained obscure38 Nevertheless ve facts were knownto medical scientists that had they been viewed in properperspective would have suggested strongly that ketoneshave to be the brainrsquos back-up fuel during starvationFirst scientists recognized that when people are deprivedof food for 3 days or longer they become hyperketone-mic and hyperketonuric Second catheterization studiesshowed that the adult human brain uses approximately100 to 150 g glucose per day39 Third scientists under-stood that non-esteri ed fatty acids do not cross theblood-brain barrier (BBB) and are not directly availableto the brain as an energy source40 Fourth Benedict41

showed that during 31 days of total caloric starvation hisldquofasting manrdquo excreted an average of 872 gday ofnitrogen (N) which corresponds to a rate of body proteincatabolism of approximately 55 gday Approximately32 g of this amount (058 550) could have beenavailable as fuel for the brainmdashapproximately 27 ofthe brainrsquos daily needs In another subject deprived offood for 15 days mean total N excretion was 482 gdaymaking only 174 g of glucoseday potentially availableto the brain42

Finally because glucose was widely assumed to bethe exclusive fuel of the adult human brain38 one mightinfer that during continuing starvation (after liver glyco-gen stores were exhausted) the brain would have to besustained by glucose derived from intracellular proteinMaking 100 to 150 g of glucoseday available to thebrain would require the gluconeogenic breakdown ofsome 172 to 259 gday of body protein At this unsus-tainable rate of protein attrition death might be expectedto occur within approximately 2 weeksmdash certainly notthe 57 to 73 days (mean 61 25 [SEM] days) that itreportedly takes for totally starved young men of averagebody composition to die43 If the nutrition scientists ofthe time had given more thought to the meaning of thisunexplained ldquoprolongationrdquo of life during total starva-tion they might have reasoned that during extended fooddeprivation the brain is able to use the ketones generatedin such abundance as an alternative energy source Sucha scenario would have explained the greater than ex-pected longevity and the far lower than expected dailynitrogen de cit exhibited by starved individuals

In 1966 Cahill et al39 ldquoconnected the dotsrdquo andconcluded that ketones might provide an alternate fuelfor the brain during starvation To test this hypothesisthe authors performed brain catheterization studies inobese patients undergoing 5 to 6 weeks of therapeuticfasting44 This landmark study showed that under condi-tions of sustained starvation (during which the subjects

achieved blood ketone levels close to 7 mML and bloodglucose concentrations of ca 38 mML) BHB and AcAcreplaced glucose as the predominant fuel(s) for brainmetabolism The ldquostarvation paradoxrdquo was thus resolvedIt also became clear that during starvation the liverrsquosability to generate ketones for use by the brain representsa biologic strategy for maintaining strength and prolong-ing life during conditions of severe nutritional privationIn this way the burden of sustaining life during starva-tion is shifted from the bodyrsquos protein stores to itsreserves of fatmdasha far more effective and ef cient adap-tation Thanks to the pioneering studies of Cahill et almetabolismrsquos ugly duckling nally began to emerge asan incipient swan

Ketones Uniquely Efcient Metabolic Fuel

In the mid 1940s Lardy and associates4546 demonstratedthat when tested on sperm BHB and AcAc comparedwith other metabolic substrates have the ability to de-crease oxygen consumption while increasing motilityFor almost half a century this apparent increase in met-abolic ef ciency remained unexplained In 1994 and1995 however RL Veechrsquos group47ndash49 at the NationalInstitutes of Health showed that in the working perfusedrat heart when 5 mM ketone bodiesL were added to theglucose-containing perfusate there was a 25 increasein hydraulic work with a signi cant decrease in oxygenconsumption Addition of ketones also brought about areduction of the mitochrondrial NAD couple with oxi-dation of the mitochondrial co-enzyme Q couple Theconsequence is an increase in the energy of the redoxspan between site I and site II of the electron transportsystem

As Veech et al50 described ldquoAn increase in theredox span between two sites will result in an increasedenergy release by the electron traveling across that span With an increase in redox energy of the respiratorychain there is a corresponding increase in the energy ofthe protons ejected from the mitochondria at the energy-conserving sites which is then re ected in an increase inthe energy of ATP hydrolysisrdquo

As shown in Table 2 Veech and associates47ndash49 alsoobserved that the introduction of ketone bodies into therat heart perfusate produced a 16-fold rise in acetyl-CoAcontent and increases in TCA cycle intermediates Thesestimulatory effects on TCA cycle activity were similar tothose obtained with the administration of saturatingdoses of insulin Insulin increases glucose transport inthe myocardium by promoting translocation of insulin-sensitive GLUT4 to the plasma membrane Insulin alsostimulates the activity of pyruvate dehydrogenase mul-tienzyme complex (PDH)mdasha rate-limiting step in themitochondrial transformation of pyruvate to acetyl-CoAAs Veech et al50 put it ldquoThe implication was that

332 Nutrition Reviews Vol 61 No 10

ketosis which is the physiological response to insulindeprivation during starvation was equivalent in meta-bolic effects to the actions of insulin (Moreover)ketones by-passed the block in glucose transport causedby lack of insulin (and) also by-passed the block inPDH induced by insulin de ciency by providing analternative source of mitochondrial acetyl CoArdquo Thepathway by which ketones are able to achieve this effectis shown in Figure 2

Veech et al50 pointed out that the gradients of Kand Na (and Ca ) between the extra- and intracellularphases of the cell are normally in virtual equilibriumwith the resting electrical potential between these twophases and with the energy ( G) of ATP hydrolysiswhich sustains the action of the Na pump By increas-ing metabolic ef ciency and thereby the energy pro-duced by ATP hydrolysis BHB augments the potentialbetween the extra- and intracellular phases giving rise toan increase in the energy of the K gradient (ratio of[K ]out[K ]in) in muscle and brain A larger mitochon-drial membrane potential would lead to larger energyproduction because it is the mitrochondrial membranepotential difference that drives the production of ATPMaintenance of a suitable K gradient in the cell is ofcourse indispensable to its proper functioning As willbe discussed below impairment of ion homeostasiswithin neuronal cells in certain brain areas might resultin a degree of electrical instability suf cient to trigger anepileptic seizure (or in the heart a cardiac arrhythmia)

Studies of blood ow and oxygen consumption inthe brains of food-deprived obese human subjects re-vealed values (45 mL min 1 100 g 1 and 27 mL O2

min 1 100 mL 1)44 that were well below the normallevels for adult human brains of 57 and 36 respective-ly51 Although these data need con rmation they suggestan increase in the metabolic ef ciency in human brainsusing ketoacids as their principal energy source in placeof glucose

Examples of mechanisms by which diet-inducedhyperketonemia might overcome or circumvent meta-bolic impediments to utilization of such substrates asglucose and pyruvate are shown schematically in Fig-ure 3

Use of a ldquoHyperketogenicrdquo Diet in Treatmentof Seizure Disorders

Fasting has long been acknowledged as a method toprevent or reduce seizures in epileptic individuals37

Because of the obvious limitations of this treatment useof a ketogenic diet to mimic the metabolic effects of fooddeprivation was proposed in 1921 by both Woodyatt52

and Wilder53 at the Mayo Clinic In 1924 Peterman54

(also from Mayo) reported the clinical application andeffectiveness of the regimen suggested by Woodyatt andWilder Petermanrsquos diet provided (per day) 1 g of pro-teinkg body weight (considerably less in adults) 10 to15 g of carbohydrate and the remainder of the calories asfat The Peterman ketogenic diet is sometimes called the41 diet because it provided a ratio of approximately 4parts (by weight) fat to 1 part (by weight) of a mixture ofprotein and carbohydrate This distribution of the majormacronutrients is almost identical to that of most anti-convulsant diets used today37 The term ldquohyperketogenicdietrdquo (HKD) will be used in this review to distinguish the41 diet (and certain variations thereof) from the far lessdrastic low-carbohydrate mildly ketogenic diets thathave been advocated for weight control in a successionof popular diet books

In the 1970s several investigators5556 modi ed thePeterman HKD by substituting medium-chain triglycer-ides (MCT) in varying amounts for long-chain triglycer-ides (MCTrsquos ketogenic effect is not dependent on con-current restriction of dietary carbohydrate57) Althoughall such variations of the HKD were approximately equalin reducing seizures many of the children given MCT-

Table 2 Insulin and Ketone Effects on Cardiac Efciency Acetyl-CoA Production [NADH][NAD ] Ratioand Free Energy of ATP Hydrolysis ( GA TP ) in the Perfused Rat Heart

Substrates Addedto Perfusate

AcetylCoAdagger

[NADH][NAD ]Daggersect

(mitochondrial)

G of ATPHydrolysis(cytosolic)

kJmol

CardiacEf ciency

()

CardiacEf ciency( relativeto glucose)

Glucose 1 005 562 105 100Glucose Insulin 9 022para 589para 134para 128Glucose Ketones 15 062para 576para 130para 124Glucose Insulin

Ketones 18 043para 589para 143para 136

Data of Sato K et al4 8 and DaggerKashiwaya et al4 7 4 9

daggerNumbers represent multiples of that measured with perfusate containing only glucosesectExpressed by Kashiwaya et al4 7 4 9 as [NAD ][NADH]paraIndicates a signi cant difference (P 005) from perfusate containing 10 mmolL glucose ketones added at 4 mMLD-b-hydroxybutyrate 1 mML acetoacetate

333Nutrition Reviews Vol 61 No 10

augmented diets complained of nausea diarrhea andoccasional vomiting58

Among those patients who adhere to the 41 HKDfor 12 months 40 experience greater than 90 reduc-tion in seizures whereas an additional 40 experience50 to 90 reduction Although the precise mechanism ofaction responsible for the clinical ef cacy of the diet hasnever been clari ed it is likely that the hyperketonemiainduced by the 41 HKD is the key factor responsible forthe regimenrsquos anticonvulsant effect rather than the ac-companying metabolic acidosis59 or dehydration37 In

1972 Uhleman and Neims60 demonstrated that micepups fed a high-fat ketogenic diet exhibited signi cantlyincreased thresholds against investigator-applied electro-shock challenges This resistance to electroshock disap-peared promptly after the animals were shifted to ahigh-carbohydrate diet Subsequently a number of stud-ies have shown an increase in the electroconvulsivethreshold in rats maintained on a high-fat diet or de-prived of food for 48 hours De Vivo et al61 found thatthe ratios of ATP to adenosine diphosphate (ADP) weresigni cantly increased in the brains of rats fed the high-

Figure 2 Production of ATP (adenosine triphosphate) in extrahepatic tissues arising from use of ketones and glucose Glucose isregulated by a family of GLUT transporter proteins GLUT4 transporters found in most body tissues are mediated by insulinGLUT1 transporters common in tissues such as brain and erythrocytes allow glucose uptake without insulin mediation By contrastuptake of ketone bodies (shown as BHB and AcAc) occurs via the family of monocarboxylate transporters (MCTs) which are notinsulin mediated MCT proteins enable ketones to pass readily through the blood-brain barrier Many types of peripheral cellsincluding brain cells not only use glucose but also use ketones to produce acetyl-CoA Glucose utilization depends on glycolyticenzymes (not shown) which produce pyruvate Pyruvate in turn enters the mitochondrion and is converted to acetyl-CoA by thepyruvate dehydrogenase multienzyme complex (PDH) Ketone substrates enter the mitochondrion in the form of (1) acetoacetate(AcAc) and (2) b-hydroxybutyrate (BHB) which is readily converted to acetoacetate by b-hydroxybutyrate dehydrogenase (HBD)Two enzymes principally control ketolysis First 3-oxoacid-CoA transferase (OCT) adds coenzyme-A to AcAc which is then splitinto two molecules of acetyl-CoA by acetoacetyl-CoA thiolase (ACT) Acetyl-CoA undergoes oxidative degradation in the TCAcycle to reduce the electron carrriers NAD (nicotinamide adenine dinucleotide) and FAD ( avin adenine dinucleotide) to NADHand FADH2 respectively Acetyl-CoA is also used for lipid or sterol synthesis (not shown) NADH and FADH2 donate electrons tothe protein complexes I and II of the mitochondrial respiratory chain The ow of these electrons down the electron-transport chainto oxygen (O2) allows complexes I III and IV to pump protons (shown as H ) out of the matrix and into the intermembrane spacewhich generates a proton motive force (pmf) between the intermembrane space and the matrix The pmf provides the energy to driveprotons back through complex V (ATP synthase) to produce ATP from ADP (adenosine diphosphate) and inorganic phosphate (Pi)

334 Nutrition Reviews Vol 61 No 10

fat diet They concluded ldquo the anticonvulsant propertyof the ketogenic diet derives from the observed increasein cerebral energy reserve which in turn re ects thehigher ATPADP ratio [this] might provide a reason-able explanation for the increased neuronal stability thatdevelops during chronic ketosisrdquo In 1999 Pan et al62

carried out 31P magnetic resonance spectroscopic imag-ing studies in seven patients with intractable epilepsybefore and after institution of a ketogenic diet Theymeasured signi cant increases (P 005) in high-energy phosphates in association with adherence to theketogenic diet in these subjects They concluded thatenhancement of brain energy metabolism occurs withuse of the ketogenic diet in humans as in lab animals

Thavendiranathan et al63 have recently reportedstudies of the effect of the ldquoclassicrdquo ketogenic diet onseizures triggered by two threshold testsmdashthe pentyle-netetrazol infusion test and the threshold electroconvul-sive shock test These authors observed small but signif-icant threshold elevations (15ndash20) against bothchemically and electrically triggered seizures Similarlyprotective effects were not seen when suprathresholdstimuli were employed

Janigro64 summarized studies indicating that there isa reduction in both glucose uptake and metabolic activityin seizure foci in epileptic patients In his words ldquoDi-minished ion homeostasis together with increased meta-bolic demand of hyperactive neurons may further aggra-vate neuropathological consequences of BBB loss ofglucose uptake mechanismsrdquo Therefore one explanationfor the effectiveness of the 41 HKD in preventingseizures could be the increased transport of BHB at theBBB and its increased availability as an energy source tothe affected neurons BHB is not only an alternate sourceof energy for seizure foci in the brainmdashbypassing animpairment of carbohydrate utilization such as thatwhich can occur from a block in mitochondrial PDHactivitymdashbut it may actually promote ketone and glu-cose uptake in such foci In rats diet-induced ketosisproduces an eightfold increase in the activity of themonocarboxylate transporter and increases the level ofthe glucose transporter GLUT1 in brain endothelialcells and neuropil65 The hypothesis put forth by Janigroto explain how ketosis might eliminate or reduce seizuresis supported by reports that a suitable ketogenic diet isthe treatment of choice for two rare conditions that may

Figure 3 Conditions that can impede acetyl-CoA production from pyruvate and its precursors Examples include (1) insuf cientinsulin (2) insulin receptor abnormalities (3) diminished transport of glucose owing to impaired release of GLUT4 transportersassociated with insulin resistance [see below] (4) pyruvate dehydrogenase complex (PDH) de ciency or dysfunction (5)unavailabilityof glucose at the cell membrane and (6) GLUT1 mutations Provision of ketones to the cell can effectively circumventor overcome such impediments and compensate for the resulting de cits of glycolytically derived acetyl-CoA (GLUT1 designatesthe insulin-independent membrane protein responsible for glucose transport in red blood cells and throughout the blood-brain barrier[BBB] GLUT4 designates the insulin-dependentglucose transporterprotein expressed in most cells Many forms of insulin resistanceentail putative post-receptor defects of insulin-signal transduction resulting in the failure of insulin to release GLUT4 transportersfrom the GLUT4 pool at a rate suf cient to bind and transport enough extracellular glucose to maintain cellular pyruvate levels)

335Nutrition Reviews Vol 61 No 10

be complicated by seizuresmdashthe facilitated glucosetransporter protein type 1 (GLUT1) de ciency syn-drome6667 and the PDH complex de ciency syndrome68

Treatment with a ketogenic diet reportedly brought aboutdramatic clinical improvement in a one-year-old boywith Leigh syndrome which is associated with a de -ciency of PDH69 In this child treatment with a keto-genic diet was associated with striking improvement ofthe cerebral lesions as measured by neuroimaging

In 1973 Buckley and Williamson70 proposed thatneural lipid synthesis preferentially uses AcAc overglucose Patel and Owen71 obtained con rmatory datafor this hypothesis in 1976 In 1984 Koper et al72

compared the rate of incorporation of AcAc and glucoseinto fatty acids and cholesterol using labeled AcAc andglucose At saturating levels of glucose the authorsobserved that the rate at which AcAc label incorporatedinto fatty acids and cholesterol exceeded that of glucoseby a factor of 5 to 10 in both oligodendrocytes andastrocytes These observations provide a basis for argu-ing that some of the bene ts associated with ketogenicdiets may be related to their enhancing effects on lipo-genesis andor sterol synthesis (ie damaged neuronsmight repair more readily in brains given an adequateketone supply)

Therapeutic Potential of Diet-inducedHyperketonemia in Neurodegenerative andOther Disorders

Veech et al50 suggest that maintenance of a suitabledegree of diet-induced hyperketonemia (in the range of2ndash7 mML) has the potential to aid management ofseveral neurologic diseases including common neurode-generative diseases like Alzheimerrsquos disease (AD) andParkinsonrsquos disease (PD) In their ldquohypothesis paperrdquothe authors single out AD PD Friedreichrsquos AtaxiaLeprechaunisms and lesser forms of insulin disorders asexamples of illnesses that might bene t from therapeuticketosis They also describe experience with one case ofLaFora body disease (until now a uniformly fatal geneticdisease associated with progressive dementia) in whichuse of the 41 HKD resulted in unprecedented improve-ment in cognition gait and awareness after one month73

In this disease mutations in a gene encoding a novelprotein tyrosine phosphatase cause progressive myo-clonic epilepsy

Calabrese et al74 and DiMauro and Schon75 haverecently called attention to evidence suggesting the mi-tochondrial genome may play a key role in neurodegen-erative diseases Decreases in respiratory chain complexactivities have been identi ed in AD PD amyotrophiclateral sclerosis and Huntingtonrsquos disease Mitochondri-opathies (MCPs) possibly associated with oxidantanti-oxidant balance perturbation may underlie defects in

energy metabolism and contribute to cellular damage anddestruction

Parkinsonrsquos DiseaseAlthough the etiology of PD is complex involving bothgenetic and environmental factors there is growing in-terest in the role of mitochondrial dysfunction in thepathogenesis of this disorder Schapira76 pointed out thata proportion of PD patients exhibit a de ciency in mito-chondrial respiratory chain function most notably adefect of complex I within neurons of the substantianigra In his words ldquoBoth mitochondrial mutations andtoxic agents (endogenous and exogenous) have beendemonstrated to cause a de ciency in complex I func-tion This is achieved by increasing the degree of orsusceptibility to oxidative stress which may contributeto apoptopic cell deathrdquo

Finsterer77 reported metabolic studies indicating thatPD can be a manifestation of MCP In his series 63 ofPD patients had features of generalized MCP howeverhe used lactate stress testing to detect patients withMCPmdashan approach that could not be expected to pick upmitochondrial lesions limited to brain sites

According to Veech et al ldquoPD appears to resultfrom an acquired defect in the mitochondrial NADHmultienzyme complex of the dopaminergic cells of themesencephalonrdquo78 They point out although the diseaseis treatable for a period by administration of levodopa(the metabolic precursor of dopamine) continuing freeradical damage eventually reduces the effectiveness ofthis therapy Kashiwaya et al79 reported that primarycultures of mesencephalic dopaminergic neurons ex-posed to the meperidine analogue MPP (a molecule thatinhibits NADH dehydrogenase [complex 1] and causesoxygen free radical formation) can be protected fromdeath by addition of 4 mM D-b-hydroxybutyrate Veechet al50 suggest that BHB probably acts by decreasing thesource of mitochondrial oxygen radical formationldquo by oxidizing the coenzyme Q couple while at thesame time reducing the redox potential of the NADPcouple which through glutathione is the nal detoxi -cation step of H2O2rdquo In a recent review Schon andManfredi80 wrote ldquoBecause complexes I and III are theprincipal sources of free radicals in the cell alteredcomplex I function in the substantia nigra pars compactacould be responsible for the increased DNA damage andlipid peroxidation found in PD brainsrdquo

To the extent that ketones crossing the BBB couldcircumvent or compensate for mitochondrial defects as-sociated with PD or could protect dopaminergic neuronsfrom free radical damage a suitable ketogenic diet mightwell have bene cial effects in certain patients sufferingfrom Parkinsonism

336 Nutrition Reviews Vol 61 No 10

Alzheimerrsquos DiseaseIn 1996 Ogawa et al81 determined regional cerebralblood ow and studied energy utilization in the brains ofseven patients with AD by measuring arteriovenousdifferences of an array of substrates Single-photon emis-sion computed tomography showed that the patients hadsigni cantly decreased regional cerebral blood ow lim-ited to the parietotemporal region The glucose extractionfraction and global cerebral metabolic rate of glucosewere signi cantly decreased whereas the global cerebralmetabolic rate of oxygen was only slightly decreasedBecause the cerebral metabolic rates of the various sub-strates (ketones lactate etc) did not change the authorsconcluded that the markedly elevated oxygenglucoseutilization ratio indicated an altered energy metabolismin AD

Conceivably the impairment of glucose metabolismin AD patients described by Ogawa et al is attributableat least in part to local insulin resistance Watson andCraft82 called attention to emerging evidence that insulinabnormalities and insulin resistance in AD may contrib-ute to the pathophysiology and clinical symptoms of thisdisorder In their words ldquoIt has been recently demon-strated that insulin-sensitive glucose transporters are lo-calized in the same regions supporting memory and thatinsulin plays a role in memory functions Collectivelythese ndings suggest that insulin may contribute tonormal cognitive functioning and that insulin abnormal-ities may exacerbate cognitive impairments such asthose associated with ADrdquo The authors suggest thatimproving insulin effectiveness may have therapeuticbene t for patients with AD The suggestion of Watsonand Craft that insulin abnormalities may play a role inthe pathogenesis of AD has to be reconciled with the nding by Molina et al83 that cerebrospinal uid insulinlevels in 27 AD patients did not differ from those in 16matched controls

In 1996 Hoshi et al84 reported that a fragment ofthe beta chain of amyloid inhibits PDH activity in pri-mary cultures of hippocampal neurons A year later thesame authors found that the 1ndash42 fragment of the betachain of amyloid Ab1ndash42 can inhibit acetyl cholinesynthesis of septal neurons85 Veech et al50 note thatcultured hippocampal neurons die when exposed toAb1ndash42 However addition of 4 mM D-b-hydroxybu-tyrate protects these neurons from amyloid fragmentndashinduced death Casley et al86 showed that beta-amyloidcauses a signi cant reduction in state 3 and state 4mitochondrial respiration which is diminished further bynitric oxide These authors also found that cytochromeoxidase -ketoglutarate dehydrogenase and PDH activ-ities are inhibited by b-amyloid They concluded thatb-amyloid can directly disrupt mitochondrial function

inhibit key enzymes and contribute to the impairment ofenergy metabolism observed in AD

Veech et al50 contend that the foregoing studiesprovide a rationale for trial of a ketogenic diet in thetreatment of AD Ketones can overcome localized insulinresistance can bypass a block in PDH and have beenshown to protect dopaminergic neurons from free radicaldamage and from apoptosis induced by the 1ndash42 amy-loid fragment in tissue-culture experiments

What Next

In considering the potential usefulness of therapeutichyperketonemia in the treatment of AD PD and certainother neurodegenerative conditions it is essential to beclear about what is meant by terms like ldquoketogenic dietrdquoand ldquohyperketonemiardquo

There is widespread confusion among both physi-cians and lay individuals about what constitutes a keto-genic diet Similar to glucose ketones are present in theblood at all times As shown in Table 3 the range ofketone levels that can be achieved by prolonged fooddeprivation or by adherence to various diets is quitewide Hence it makes no sense to speak of a ldquoketogenicdietrdquo without also specifying the degree of serum ketoneelevation that the diet is intended to achieve

The serum concentrations sought for treatment ofepilepsy range from 2 to 7 mML Such levels can beachieved only by strict adherence to a regimen like the41 HKDmdashone most patients nd burdensome On theother hand the kind of low-carbohydrate ldquoketogenicdietrdquo (eg 30 protein 8 carbohydrate 61 fat [asproportions of total energy]) that has recently becomepopularmdashalthough much easier to followmdashachievesBHB levels of only 028ndash040 mML87 (This ldquopopularrdquoketogenic diet at a 2000 kcalday intake provides 127 gof presumptive carbohydrate [40 preformed] plus 87 gderivable from 150 g protein not including the smallquantity of glycerol released in the course of adipocytetriglyceride hydrolysis At the same energy level the 41HKD would yield 33 g of presumptive carbohydrate[10 g preformed] plus 23 g derivable from 40 g proteinThus the effective quantity of carbohydrate provided bythe popular ketogenic diet is almost four times thatsupplied by the 41 HKD)

Future research must address the problem of match-ing diet-induced elevation of serum ketone levels withtherapeutic bene ts Because the drastic type of hyper-ketogenic dietmdashthe one used to treat epilepsy and relatedneurologic disordersmdashis so onerous and dif cult to fol-low it is especially important to determine the extent towhich such a diet can be liberalized without losingdesirable therapeutic effects Given the daunting natureof the 41 HKD it is encouraging that there exists apotential nutritional stratagem which eventually may

337Nutrition Reviews Vol 61 No 10

make it possible to induce therapeutic hyperketonemia inpatients without requiring them to adhere to a Draconianregimen The stratagem entails administration by mouthof substantial quantities of ketones in a formmdashnamelyas esters or small polymersmdashsuitable for ingestion Intheir article about the potential therapeutic uses of ketonebodies Veech et al50 suggest that the problems associ-ated with ldquothe very unappetizing high-fat ketogenicdietrdquo might be obviated by feeding patients up to 100 to150 g per day of a nutritionally acceptable form ofketone bodies Unfortunately neither ketone esters noroligomers of BHB are presently available in suf cientquantities for the conduct of clinical trials

The traditional 41 HKD is dif cult but far fromunmanageable Numerous epileptic children and adultswith seizures resistant to anticonvulsant medication havemanaged to remain on this regimen for years withoutexperiencing serious adverse effects There are justi ableconcerns about its hyperlipidemic effect however andthe fact that the diet has been reported to increase risk ofside effects including nephrolithiasis88 depression ofplatelet function (manifested by easy bruising)89 andoccasionally prolongation of the electrocardiographicQTc interval in children90 It should be possible to mitigateor perhaps prevent the adverse effects of this very highndashfatdiet on plasma lipids by modifying its fatty acid contentWhat is most important now however is to get on with thejob of determining whether diet-induced hyperketonemia(at any level) will prove clinically bene cial in the man-agement of a number of neurodegenerative disorders par-ticularly Parkinsonrsquos disease and Alzheimerrsquos disease

Acknowledgment

We thank Richard L Veech MD DPhil for hisencouragement and very helpful suggestions We alsothank Steven W Fowkes for creative and elegant ren-dering of Figures 1ndash3 and editorial advice

1 Gerhardt J Diabetes mellitus und aceton WienerMedizinische Presse 18656672

2 Hirschfeld F Beobachtungen uber die Acentonurieund das Coma diabeticum Z klin Med 189528176 ndash209

3 Peters JP Van Slyke DD Quantitative ClinicalChemistry Interpretations Volume 1 2nd editionBaltimore Williams amp Wilkins 1946439

4 Cohen PP Studies in ketogenesis J Biol Chem1937119333ndash346

5 Cohen PP Stark IE Hepatic ketogenesis and ketol-ysis in different species J Biol Chem 193812697ndash107

6 Jowett M Quastel JH Fat metabolism III The for-mation and breakdown of acetoacetic acid in animaltissues Biochem J 1935292181ndash2191

7 Chaikoff IL Soskin S The utilization of acetoaceticacid by normal and diabetic dogs before and afterevisceration Am J Physiol 19288758 ndash72

8 Barnes RH Drury DR Greeley PO Wick AN Utili-zation of the ketone bodies in normal animals and inthose with ketosis Am J Physiol 1940130144 ndash150

9 Mirsky IA Broh-Kahn RH The inuence of in-creased metabolism on b-hydroxybutyric acid utili-zation Am J Physiol 1937120446 ndash 450

10 Waters ET Fletcher JP Mirsky IA The relationbetween carbohydrate and b-hydroxybutyric acid

Table 3 Relationship of Metabolic State and Diet Composition to Plasma Levels of Acetoacetate (AcAc)and b-Hydroxybutyrate (BHB) in Normal-weight Overweight and Obese Adults

Metabolic StateDiet

Composition CommentsAcAc

(mML)BHB

(mML) References

Postabsorptive Malesn 5 11 Normal weight 005 007 Hall et al18

n 5 12 Normal weight 008 Sharman et al87

n 5 10 Obese 015 027 Hall et al18

n 5 6 Overweight 001 002 Cahill et al39

Ketogenic diets Males (normal weight)(eucaloric) 3 weeks on diet (n 5 12) 040 Sharman et al87

C8 P30 F61 6 weeks on diet 028 Sharman et al87

Female (normal weight)C2 P8 F90 2 weeks on diet (n 5 1) 070 602 VanItallie Nonas amp Heyms elddagger

Fasting Malesn 5 6 (overweight)

1-week fast 095 358 Cahill et al39

n 5 6 (obese)1ndash2-week fast 104 597 Hall et al18

Distribution (percent total calories) of carbohydrate (C) protein (P) and fat (F)daggerUnpublished study (St Lukersquos-Roosevelt Hospital 2003)Where n 1 all values are means

338 Nutrition Reviews Vol 61 No 10

tion and the rate of ketogenesis also remains un-changed21 it is evident that the rate of peripheral clear-ance of ketones must diminish in order to account for theelevated blood ketone levels measured during the food-deprivation period

Laffel22 suggested that the inverse relationship be-tween level of ketonemia and rate coef cient for ketoneremoval could arise from down-regulation of an enzymeinvolved in ketone utilization in response to a rise inintracellular levels of AcAc during starvation in associ-ation with dietary carbohydrateprotein restriction or intype 1 diabetes This explanation seems reasonable how-ever the evidence thus far is inconclusive that any suchenzyme is suf ciently inhibited or decreased to accountfor the observed diminution in ketone utilization It isnoteworthy however that the brain adapts to starvationwith an increased rate of ketone utilization and that anincrease in 3-oxoacid-CoA transferase (OCT) activityhas been demonstrated in the brain of the starved rat Bycontrast decreased OCT activity has been described inthe skeletal muscle of rats with diabetes of more than 10daysrsquo duration16 To complicate matters further exercisetraining for a number of weeks appears to increase thecapacity of muscle to use ketones Because exercise alsopromotes ketogenesis the muscles presumably accom-modate to such increased production by inducing oractivating relevant ketone-utilizing enzymes1623

The physiologic explanation for the increased ke-tonemia during the early phase of food deprivation ap-pears to lie in data reported by Owen and Reichard24

who showed that the rise in blood ketones during the rst2 weeks of starvation results from a decreasing rate ofketone utilization in muscle As food deprivation contin-ues a dramatic change occurs in glucose and ketoneutilization by the nervous system and muscle Glucoseutilization in the brain drops to 13 that in the fed statewith ketones lling the energy gap A nearly reciprocalrelationship exists for muscle which seems to adapt sofully to prolonged fasting that it virtually ignores ketonesas an energy source drawing instead on fatty acids21 AsCahill25 stated ldquo early in starvation ketoacids serveas muscle fuel but with more prolonged starvation theyare less well used In fact as acetoacetate is taken up bythe muscle it is returned back to the blood as b-hydroxy-butyrate signifying a more reduced state of the musclemitochondriamdashsecondary to free fatty acid utilizationThus fatty acids appear to take preference for oxidationand ketoacid oxidation ceases sparing the ketoacids forthe brain a superb survival processrdquo

According to Laffel22 healthy adult liver can pro-duce as much as 185 g of ketone bodies per day Ketonessupply 2 to 6 of the bodyrsquos energy needs after anovernight fast and 30 to 40 after a 3-day fast

Effect of Total Starvation and CarbohydrateRestriction on Blood and Urine Ketones

In the late 19th century physiologists focused on thestriking association between ketonuria and worseningdiabetes mellitus Later as methods for detecting andquantifying ketones in biologic uids improved atten-tion shifted to the effect of food deprivation changes inenergy expenditure and modi cations of dietary compo-sition on the urinary excretion and blood levels of ketonebodies It was soon evident that the ketonuria associatedwith total starvation does not become substantial until 3to 5 days have elapsed This is the time it takes for liverglycogen to be thoroughly depleted In 1931 Behre26

reported that ketonuria begins to increase within a fewhours after a meal is skipped In three subjects studied byKartin et al27 during a 6-day period of total starvationblood ketones after 2 days of food deprivation variedfrom 11 to 25 mML After 6 days blood ketones had risento 23 to 40 mML Peters and Van Slyke3 point out thatsuch rises ldquo are not enough to tax severely the mecha-nisms for the preservation of acid-base equilibrium theserum bicarbonate does not fall below 35 to 40 volumes percent Benedictrsquos subject averaged about 6 grams ofb-hydroxybutyricacid in the urine daily for the last 2 weeksof his 31-day fast In diabetic acidosis ketonuria has beenknown to reach 10 times as much as thisrdquo

By accelerating the rate of carbohydrate utilizationhuman physical exercise2328 administration of dinitro-phenol or thyroid hormone to rabbits10 and increasedbovine lactation29 can give rise to appreciable ketosisPregnant women maintain higher than normal ketonelevels and show an exaggerated ketogenic response tofasting30 In human subjects undergoing experimentalstarvation consumption of relatively small amounts ofglucose per day (eg 50ndash75 g) greatly reduces ketonebody production31

A low-carbohydrate diet is not necessarily a keto-genic diet This is particularly true of diets with unre-stricted content of meat and other protein-rich foodsHeinbecker32 reported in 1928 that Baf n Island Eski-mos subsisting on their usual diet of meat (virtually theonly source of carbohydrate in their food was the glyco-gen in seal muscle) showed minimal ketonuria (As areference point muscle glycogen concentration in restinghumans on a mixed diet is approximately 144 gkg33) Itis unlikely that these very small amounts of glycogencould have accounted for the absence of appreciableketonuria A much more likely explanation is that theglucose derived from catabolism of ingested meat pro-tein was suf cient to prevent ketosis McClellan andDuBois34 fed two human volunteers ldquocarbohydrate-freerdquodiets high in meat content (an Eskimo-type diet) formany months in a metabolic ward setting Their ndingsled them to conclude that in persons subsisting on diets

329Nutrition Reviews Vol 61 No 10

330 Nutrition Reviews Vol 61 No 10

very low in carbohydrate ketosis varies inversely withthe quantity of protein eaten This occurs because ap-proximately 48 to 58 of the amino acids in most dietaryproteins are glucogenic3536 For every 2 grams of proteinconsumed in a carbohydrate-free diet somewhere be-tween 10 and 12 grams are potentially convertible to

glucose Therefore to obtain a degree of hyperketonemia(approximately 2ndash7 mML) believed to be therapeuti-cally effective in certain important medical conditionssuch as epilepsy patients must rigorously restrict proteinas well as carbohydrate intake37 and when possibleincrease their level of physical activity23

Table 1 Postabsorptive Plasma Concentrations of Acetoacetate (AcAc) and b-Hydroxybutyrate (BHB)and Rates of Ketone Release and Removal Together with Rate Coefcients for Ketone Removal (meanSD) in Normal Diabetic and Normally Fed and Subsequently Food-deprived Severely Obese Subjects

DiagnosisMetabolic

StateAcAc

(mML)BHB

(mML)

KetoneRelease

Rate( mol

min 1 m 2)

KetoneRemoval

Rate( mol

min 1 m 2)

RateCoef cientfor KetoneRemoval(min 1)

Normal(n 5 11) Postabsorptive 005 002 007 004 81 66 1107 1059 0168 0169

Diabeticdagger

(n 5 6) Postabsorptive 067 075 179 263 208 118 2188 874 0055 0040ObeseDagger

(n 5 10) Postabsorptive 015 006 027 012 171 70 2101 932 0066 0040Food-deprived

1ndash2 weekssect 104 026 597 113 569 286 5673 2562 0027 0019

Fraction lost by oxidation and excretiondaggerRecently diagnosed untreated patients with insulin-dependent diabetes mellitusDaggerSubjects severely obesesectSubjects received 50 kcal and 1 multivitamin tabletdayData from Reference 18

Figure 1 Hepatic formation and peripheral utilization of ketone bodies Mobilization of non-esteri ed fatty acids (NEFA) fromadipose tissue fat stores by the activation of hormone-sensitive lipase is an initial step that provides ample substrate for the hepaticproduction of ketone bodies (b-hydroxybutyrate and acetoacetate) Once fatty acids enter the hepatocyte from the blood circulationthey are activated to CoA esters by the family of membrane-bound fatty acid acyl-CoA synthetases The action of one of thesesynthetases (long-chain acyl CoA synthetase) is identi ed as (3) After CoA esteri cation the metabolic paths diverge based on thefatty acid (acyl) chain length Short- and medium-chain acyl-CoAs readily cross the outer and inner mitochondrial membranes andundergo b-oxidation into acetyl-CoA and acetoacetate (the terminal 4-carbon fragment) Long-chain acyl-CoAs readily cross theouter mitochondrial membrane after which they must be converted to long-chain acylcarnitine by membrane-bound carnitinepalmitoyltransferaseI or CPT-1 (1) for subsequent transport across the inner mitochondrial membrane The inner membranendash boundenzyme carnitine acylcarnitine translocase (4) exchanges one long-chain acyl-CoA from the intermembrane space with one free(unacylated)carnitine from the matrix thereby permitting carnitinersquos return to the intermembrane space The inner membranendash boundenzyme carnitine palmitoyltransferase II or CPT-II (2) converts the fatty acid back into its CoA ester for subsequent b-oxidationfreeing carnitine for recycling Very longndash chain acyl-CoAs and unsaturated fatty acids of various lengths are transported intoperoxisomes along with long- and medium-chain acyl-CoAs for chain-length reduction via b-oxidation Unlike mitochondrialb-oxidation peroxisomal b-oxidation does not proceed to completion This results in peroxisomal export of medium-chain acyl andacylcarnitinecompounds to the mitochondria for further oxidationMitochondrial b-oxidation produces acetyl-CoA which enters theTCA cycle for further oxidation to produce ATP Ketogenesis occurs when hepatic energy requirements are met allowing excessacetyl-CoA to be acted upon by thiolase which condenses two acetyl-CoAs to form acetoacetyl-CoA Hydroxymethylglutyaryl-CoAsynthase (HMG-CoA synthase) condenses acetoacetyl-CoA with acetyl-CoA to form hydroxymethylglutaryl-CoA which is thendissociated into acetoacetate and acetyl-CoA by hydroxymethylglutaryl-CoA lyase (HMG-CoA lyase) The action of theenzyme b-hydroxybutyrate dehydrogenase (5) reduces much of the acetoacetate to b-hydroxybutyrate using NADH as areductant Both acetoacetate and b-hydroxybutyrate are transported back through the mitochondrial membranes for export intothe blood circulation The pathway of lipogenesis illustrated by dashed lines becomes activated in the fed state and results inthe synthesis of fatty acids and their esteri cation into triacylglycerol (triglycerides) The rst committed intermediary in fattyacid synthesis is malonyl-CoA which inhibits CPT-1 (enzyme 1) This inhibition is illustrated as a thin dashed line pointingto CPT-1 Inhibition of CPT-1 prevents mitochondrial uptake of fatty acyl-CoA (and its subsequent b-oxidation) and therebyblocks ketogenesis In the fed (non-ketotic) state triglyceride is exported from the liver as very lowndash density lipoprotein(VLDL)

331Nutrition Reviews Vol 61 No 10

Ketones Substrates in Search of a Mission

By the early 1960s a great deal of information hadaccumulated about the biochemistry and physiology ofthe ketone bodies yet their role in human metabolismremained obscure38 Nevertheless ve facts were knownto medical scientists that had they been viewed in properperspective would have suggested strongly that ketoneshave to be the brainrsquos back-up fuel during starvationFirst scientists recognized that when people are deprivedof food for 3 days or longer they become hyperketone-mic and hyperketonuric Second catheterization studiesshowed that the adult human brain uses approximately100 to 150 g glucose per day39 Third scientists under-stood that non-esteri ed fatty acids do not cross theblood-brain barrier (BBB) and are not directly availableto the brain as an energy source40 Fourth Benedict41

showed that during 31 days of total caloric starvation hisldquofasting manrdquo excreted an average of 872 gday ofnitrogen (N) which corresponds to a rate of body proteincatabolism of approximately 55 gday Approximately32 g of this amount (058 550) could have beenavailable as fuel for the brainmdashapproximately 27 ofthe brainrsquos daily needs In another subject deprived offood for 15 days mean total N excretion was 482 gdaymaking only 174 g of glucoseday potentially availableto the brain42

Finally because glucose was widely assumed to bethe exclusive fuel of the adult human brain38 one mightinfer that during continuing starvation (after liver glyco-gen stores were exhausted) the brain would have to besustained by glucose derived from intracellular proteinMaking 100 to 150 g of glucoseday available to thebrain would require the gluconeogenic breakdown ofsome 172 to 259 gday of body protein At this unsus-tainable rate of protein attrition death might be expectedto occur within approximately 2 weeksmdash certainly notthe 57 to 73 days (mean 61 25 [SEM] days) that itreportedly takes for totally starved young men of averagebody composition to die43 If the nutrition scientists ofthe time had given more thought to the meaning of thisunexplained ldquoprolongationrdquo of life during total starva-tion they might have reasoned that during extended fooddeprivation the brain is able to use the ketones generatedin such abundance as an alternative energy source Sucha scenario would have explained the greater than ex-pected longevity and the far lower than expected dailynitrogen de cit exhibited by starved individuals

In 1966 Cahill et al39 ldquoconnected the dotsrdquo andconcluded that ketones might provide an alternate fuelfor the brain during starvation To test this hypothesisthe authors performed brain catheterization studies inobese patients undergoing 5 to 6 weeks of therapeuticfasting44 This landmark study showed that under condi-tions of sustained starvation (during which the subjects

achieved blood ketone levels close to 7 mML and bloodglucose concentrations of ca 38 mML) BHB and AcAcreplaced glucose as the predominant fuel(s) for brainmetabolism The ldquostarvation paradoxrdquo was thus resolvedIt also became clear that during starvation the liverrsquosability to generate ketones for use by the brain representsa biologic strategy for maintaining strength and prolong-ing life during conditions of severe nutritional privationIn this way the burden of sustaining life during starva-tion is shifted from the bodyrsquos protein stores to itsreserves of fatmdasha far more effective and ef cient adap-tation Thanks to the pioneering studies of Cahill et almetabolismrsquos ugly duckling nally began to emerge asan incipient swan

Ketones Uniquely Efcient Metabolic Fuel

In the mid 1940s Lardy and associates4546 demonstratedthat when tested on sperm BHB and AcAc comparedwith other metabolic substrates have the ability to de-crease oxygen consumption while increasing motilityFor almost half a century this apparent increase in met-abolic ef ciency remained unexplained In 1994 and1995 however RL Veechrsquos group47ndash49 at the NationalInstitutes of Health showed that in the working perfusedrat heart when 5 mM ketone bodiesL were added to theglucose-containing perfusate there was a 25 increasein hydraulic work with a signi cant decrease in oxygenconsumption Addition of ketones also brought about areduction of the mitochrondrial NAD couple with oxi-dation of the mitochondrial co-enzyme Q couple Theconsequence is an increase in the energy of the redoxspan between site I and site II of the electron transportsystem

As Veech et al50 described ldquoAn increase in theredox span between two sites will result in an increasedenergy release by the electron traveling across that span With an increase in redox energy of the respiratorychain there is a corresponding increase in the energy ofthe protons ejected from the mitochondria at the energy-conserving sites which is then re ected in an increase inthe energy of ATP hydrolysisrdquo

As shown in Table 2 Veech and associates47ndash49 alsoobserved that the introduction of ketone bodies into therat heart perfusate produced a 16-fold rise in acetyl-CoAcontent and increases in TCA cycle intermediates Thesestimulatory effects on TCA cycle activity were similar tothose obtained with the administration of saturatingdoses of insulin Insulin increases glucose transport inthe myocardium by promoting translocation of insulin-sensitive GLUT4 to the plasma membrane Insulin alsostimulates the activity of pyruvate dehydrogenase mul-tienzyme complex (PDH)mdasha rate-limiting step in themitochondrial transformation of pyruvate to acetyl-CoAAs Veech et al50 put it ldquoThe implication was that

332 Nutrition Reviews Vol 61 No 10

ketosis which is the physiological response to insulindeprivation during starvation was equivalent in meta-bolic effects to the actions of insulin (Moreover)ketones by-passed the block in glucose transport causedby lack of insulin (and) also by-passed the block inPDH induced by insulin de ciency by providing analternative source of mitochondrial acetyl CoArdquo Thepathway by which ketones are able to achieve this effectis shown in Figure 2

Veech et al50 pointed out that the gradients of Kand Na (and Ca ) between the extra- and intracellularphases of the cell are normally in virtual equilibriumwith the resting electrical potential between these twophases and with the energy ( G) of ATP hydrolysiswhich sustains the action of the Na pump By increas-ing metabolic ef ciency and thereby the energy pro-duced by ATP hydrolysis BHB augments the potentialbetween the extra- and intracellular phases giving rise toan increase in the energy of the K gradient (ratio of[K ]out[K ]in) in muscle and brain A larger mitochon-drial membrane potential would lead to larger energyproduction because it is the mitrochondrial membranepotential difference that drives the production of ATPMaintenance of a suitable K gradient in the cell is ofcourse indispensable to its proper functioning As willbe discussed below impairment of ion homeostasiswithin neuronal cells in certain brain areas might resultin a degree of electrical instability suf cient to trigger anepileptic seizure (or in the heart a cardiac arrhythmia)

Studies of blood ow and oxygen consumption inthe brains of food-deprived obese human subjects re-vealed values (45 mL min 1 100 g 1 and 27 mL O2

min 1 100 mL 1)44 that were well below the normallevels for adult human brains of 57 and 36 respective-ly51 Although these data need con rmation they suggestan increase in the metabolic ef ciency in human brainsusing ketoacids as their principal energy source in placeof glucose

Examples of mechanisms by which diet-inducedhyperketonemia might overcome or circumvent meta-bolic impediments to utilization of such substrates asglucose and pyruvate are shown schematically in Fig-ure 3

Use of a ldquoHyperketogenicrdquo Diet in Treatmentof Seizure Disorders

Fasting has long been acknowledged as a method toprevent or reduce seizures in epileptic individuals37

Because of the obvious limitations of this treatment useof a ketogenic diet to mimic the metabolic effects of fooddeprivation was proposed in 1921 by both Woodyatt52

and Wilder53 at the Mayo Clinic In 1924 Peterman54

(also from Mayo) reported the clinical application andeffectiveness of the regimen suggested by Woodyatt andWilder Petermanrsquos diet provided (per day) 1 g of pro-teinkg body weight (considerably less in adults) 10 to15 g of carbohydrate and the remainder of the calories asfat The Peterman ketogenic diet is sometimes called the41 diet because it provided a ratio of approximately 4parts (by weight) fat to 1 part (by weight) of a mixture ofprotein and carbohydrate This distribution of the majormacronutrients is almost identical to that of most anti-convulsant diets used today37 The term ldquohyperketogenicdietrdquo (HKD) will be used in this review to distinguish the41 diet (and certain variations thereof) from the far lessdrastic low-carbohydrate mildly ketogenic diets thathave been advocated for weight control in a successionof popular diet books

In the 1970s several investigators5556 modi ed thePeterman HKD by substituting medium-chain triglycer-ides (MCT) in varying amounts for long-chain triglycer-ides (MCTrsquos ketogenic effect is not dependent on con-current restriction of dietary carbohydrate57) Althoughall such variations of the HKD were approximately equalin reducing seizures many of the children given MCT-

Table 2 Insulin and Ketone Effects on Cardiac Efciency Acetyl-CoA Production [NADH][NAD ] Ratioand Free Energy of ATP Hydrolysis ( GA TP ) in the Perfused Rat Heart

Substrates Addedto Perfusate

AcetylCoAdagger

[NADH][NAD ]Daggersect

(mitochondrial)

G of ATPHydrolysis(cytosolic)

kJmol

CardiacEf ciency

()

CardiacEf ciency( relativeto glucose)

Glucose 1 005 562 105 100Glucose Insulin 9 022para 589para 134para 128Glucose Ketones 15 062para 576para 130para 124Glucose Insulin

Ketones 18 043para 589para 143para 136

Data of Sato K et al4 8 and DaggerKashiwaya et al4 7 4 9

daggerNumbers represent multiples of that measured with perfusate containing only glucosesectExpressed by Kashiwaya et al4 7 4 9 as [NAD ][NADH]paraIndicates a signi cant difference (P 005) from perfusate containing 10 mmolL glucose ketones added at 4 mMLD-b-hydroxybutyrate 1 mML acetoacetate

333Nutrition Reviews Vol 61 No 10

augmented diets complained of nausea diarrhea andoccasional vomiting58

Among those patients who adhere to the 41 HKDfor 12 months 40 experience greater than 90 reduc-tion in seizures whereas an additional 40 experience50 to 90 reduction Although the precise mechanism ofaction responsible for the clinical ef cacy of the diet hasnever been clari ed it is likely that the hyperketonemiainduced by the 41 HKD is the key factor responsible forthe regimenrsquos anticonvulsant effect rather than the ac-companying metabolic acidosis59 or dehydration37 In

1972 Uhleman and Neims60 demonstrated that micepups fed a high-fat ketogenic diet exhibited signi cantlyincreased thresholds against investigator-applied electro-shock challenges This resistance to electroshock disap-peared promptly after the animals were shifted to ahigh-carbohydrate diet Subsequently a number of stud-ies have shown an increase in the electroconvulsivethreshold in rats maintained on a high-fat diet or de-prived of food for 48 hours De Vivo et al61 found thatthe ratios of ATP to adenosine diphosphate (ADP) weresigni cantly increased in the brains of rats fed the high-

Figure 2 Production of ATP (adenosine triphosphate) in extrahepatic tissues arising from use of ketones and glucose Glucose isregulated by a family of GLUT transporter proteins GLUT4 transporters found in most body tissues are mediated by insulinGLUT1 transporters common in tissues such as brain and erythrocytes allow glucose uptake without insulin mediation By contrastuptake of ketone bodies (shown as BHB and AcAc) occurs via the family of monocarboxylate transporters (MCTs) which are notinsulin mediated MCT proteins enable ketones to pass readily through the blood-brain barrier Many types of peripheral cellsincluding brain cells not only use glucose but also use ketones to produce acetyl-CoA Glucose utilization depends on glycolyticenzymes (not shown) which produce pyruvate Pyruvate in turn enters the mitochondrion and is converted to acetyl-CoA by thepyruvate dehydrogenase multienzyme complex (PDH) Ketone substrates enter the mitochondrion in the form of (1) acetoacetate(AcAc) and (2) b-hydroxybutyrate (BHB) which is readily converted to acetoacetate by b-hydroxybutyrate dehydrogenase (HBD)Two enzymes principally control ketolysis First 3-oxoacid-CoA transferase (OCT) adds coenzyme-A to AcAc which is then splitinto two molecules of acetyl-CoA by acetoacetyl-CoA thiolase (ACT) Acetyl-CoA undergoes oxidative degradation in the TCAcycle to reduce the electron carrriers NAD (nicotinamide adenine dinucleotide) and FAD ( avin adenine dinucleotide) to NADHand FADH2 respectively Acetyl-CoA is also used for lipid or sterol synthesis (not shown) NADH and FADH2 donate electrons tothe protein complexes I and II of the mitochondrial respiratory chain The ow of these electrons down the electron-transport chainto oxygen (O2) allows complexes I III and IV to pump protons (shown as H ) out of the matrix and into the intermembrane spacewhich generates a proton motive force (pmf) between the intermembrane space and the matrix The pmf provides the energy to driveprotons back through complex V (ATP synthase) to produce ATP from ADP (adenosine diphosphate) and inorganic phosphate (Pi)

334 Nutrition Reviews Vol 61 No 10

fat diet They concluded ldquo the anticonvulsant propertyof the ketogenic diet derives from the observed increasein cerebral energy reserve which in turn re ects thehigher ATPADP ratio [this] might provide a reason-able explanation for the increased neuronal stability thatdevelops during chronic ketosisrdquo In 1999 Pan et al62

carried out 31P magnetic resonance spectroscopic imag-ing studies in seven patients with intractable epilepsybefore and after institution of a ketogenic diet Theymeasured signi cant increases (P 005) in high-energy phosphates in association with adherence to theketogenic diet in these subjects They concluded thatenhancement of brain energy metabolism occurs withuse of the ketogenic diet in humans as in lab animals

Thavendiranathan et al63 have recently reportedstudies of the effect of the ldquoclassicrdquo ketogenic diet onseizures triggered by two threshold testsmdashthe pentyle-netetrazol infusion test and the threshold electroconvul-sive shock test These authors observed small but signif-icant threshold elevations (15ndash20) against bothchemically and electrically triggered seizures Similarlyprotective effects were not seen when suprathresholdstimuli were employed

Janigro64 summarized studies indicating that there isa reduction in both glucose uptake and metabolic activityin seizure foci in epileptic patients In his words ldquoDi-minished ion homeostasis together with increased meta-bolic demand of hyperactive neurons may further aggra-vate neuropathological consequences of BBB loss ofglucose uptake mechanismsrdquo Therefore one explanationfor the effectiveness of the 41 HKD in preventingseizures could be the increased transport of BHB at theBBB and its increased availability as an energy source tothe affected neurons BHB is not only an alternate sourceof energy for seizure foci in the brainmdashbypassing animpairment of carbohydrate utilization such as thatwhich can occur from a block in mitochondrial PDHactivitymdashbut it may actually promote ketone and glu-cose uptake in such foci In rats diet-induced ketosisproduces an eightfold increase in the activity of themonocarboxylate transporter and increases the level ofthe glucose transporter GLUT1 in brain endothelialcells and neuropil65 The hypothesis put forth by Janigroto explain how ketosis might eliminate or reduce seizuresis supported by reports that a suitable ketogenic diet isthe treatment of choice for two rare conditions that may

Figure 3 Conditions that can impede acetyl-CoA production from pyruvate and its precursors Examples include (1) insuf cientinsulin (2) insulin receptor abnormalities (3) diminished transport of glucose owing to impaired release of GLUT4 transportersassociated with insulin resistance [see below] (4) pyruvate dehydrogenase complex (PDH) de ciency or dysfunction (5)unavailabilityof glucose at the cell membrane and (6) GLUT1 mutations Provision of ketones to the cell can effectively circumventor overcome such impediments and compensate for the resulting de cits of glycolytically derived acetyl-CoA (GLUT1 designatesthe insulin-independent membrane protein responsible for glucose transport in red blood cells and throughout the blood-brain barrier[BBB] GLUT4 designates the insulin-dependentglucose transporterprotein expressed in most cells Many forms of insulin resistanceentail putative post-receptor defects of insulin-signal transduction resulting in the failure of insulin to release GLUT4 transportersfrom the GLUT4 pool at a rate suf cient to bind and transport enough extracellular glucose to maintain cellular pyruvate levels)

335Nutrition Reviews Vol 61 No 10

be complicated by seizuresmdashthe facilitated glucosetransporter protein type 1 (GLUT1) de ciency syn-drome6667 and the PDH complex de ciency syndrome68

Treatment with a ketogenic diet reportedly brought aboutdramatic clinical improvement in a one-year-old boywith Leigh syndrome which is associated with a de -ciency of PDH69 In this child treatment with a keto-genic diet was associated with striking improvement ofthe cerebral lesions as measured by neuroimaging

In 1973 Buckley and Williamson70 proposed thatneural lipid synthesis preferentially uses AcAc overglucose Patel and Owen71 obtained con rmatory datafor this hypothesis in 1976 In 1984 Koper et al72

compared the rate of incorporation of AcAc and glucoseinto fatty acids and cholesterol using labeled AcAc andglucose At saturating levels of glucose the authorsobserved that the rate at which AcAc label incorporatedinto fatty acids and cholesterol exceeded that of glucoseby a factor of 5 to 10 in both oligodendrocytes andastrocytes These observations provide a basis for argu-ing that some of the bene ts associated with ketogenicdiets may be related to their enhancing effects on lipo-genesis andor sterol synthesis (ie damaged neuronsmight repair more readily in brains given an adequateketone supply)

Therapeutic Potential of Diet-inducedHyperketonemia in Neurodegenerative andOther Disorders

Veech et al50 suggest that maintenance of a suitabledegree of diet-induced hyperketonemia (in the range of2ndash7 mML) has the potential to aid management ofseveral neurologic diseases including common neurode-generative diseases like Alzheimerrsquos disease (AD) andParkinsonrsquos disease (PD) In their ldquohypothesis paperrdquothe authors single out AD PD Friedreichrsquos AtaxiaLeprechaunisms and lesser forms of insulin disorders asexamples of illnesses that might bene t from therapeuticketosis They also describe experience with one case ofLaFora body disease (until now a uniformly fatal geneticdisease associated with progressive dementia) in whichuse of the 41 HKD resulted in unprecedented improve-ment in cognition gait and awareness after one month73

In this disease mutations in a gene encoding a novelprotein tyrosine phosphatase cause progressive myo-clonic epilepsy

Calabrese et al74 and DiMauro and Schon75 haverecently called attention to evidence suggesting the mi-tochondrial genome may play a key role in neurodegen-erative diseases Decreases in respiratory chain complexactivities have been identi ed in AD PD amyotrophiclateral sclerosis and Huntingtonrsquos disease Mitochondri-opathies (MCPs) possibly associated with oxidantanti-oxidant balance perturbation may underlie defects in

energy metabolism and contribute to cellular damage anddestruction

Parkinsonrsquos DiseaseAlthough the etiology of PD is complex involving bothgenetic and environmental factors there is growing in-terest in the role of mitochondrial dysfunction in thepathogenesis of this disorder Schapira76 pointed out thata proportion of PD patients exhibit a de ciency in mito-chondrial respiratory chain function most notably adefect of complex I within neurons of the substantianigra In his words ldquoBoth mitochondrial mutations andtoxic agents (endogenous and exogenous) have beendemonstrated to cause a de ciency in complex I func-tion This is achieved by increasing the degree of orsusceptibility to oxidative stress which may contributeto apoptopic cell deathrdquo

Finsterer77 reported metabolic studies indicating thatPD can be a manifestation of MCP In his series 63 ofPD patients had features of generalized MCP howeverhe used lactate stress testing to detect patients withMCPmdashan approach that could not be expected to pick upmitochondrial lesions limited to brain sites

According to Veech et al ldquoPD appears to resultfrom an acquired defect in the mitochondrial NADHmultienzyme complex of the dopaminergic cells of themesencephalonrdquo78 They point out although the diseaseis treatable for a period by administration of levodopa(the metabolic precursor of dopamine) continuing freeradical damage eventually reduces the effectiveness ofthis therapy Kashiwaya et al79 reported that primarycultures of mesencephalic dopaminergic neurons ex-posed to the meperidine analogue MPP (a molecule thatinhibits NADH dehydrogenase [complex 1] and causesoxygen free radical formation) can be protected fromdeath by addition of 4 mM D-b-hydroxybutyrate Veechet al50 suggest that BHB probably acts by decreasing thesource of mitochondrial oxygen radical formationldquo by oxidizing the coenzyme Q couple while at thesame time reducing the redox potential of the NADPcouple which through glutathione is the nal detoxi -cation step of H2O2rdquo In a recent review Schon andManfredi80 wrote ldquoBecause complexes I and III are theprincipal sources of free radicals in the cell alteredcomplex I function in the substantia nigra pars compactacould be responsible for the increased DNA damage andlipid peroxidation found in PD brainsrdquo

To the extent that ketones crossing the BBB couldcircumvent or compensate for mitochondrial defects as-sociated with PD or could protect dopaminergic neuronsfrom free radical damage a suitable ketogenic diet mightwell have bene cial effects in certain patients sufferingfrom Parkinsonism

336 Nutrition Reviews Vol 61 No 10

Alzheimerrsquos DiseaseIn 1996 Ogawa et al81 determined regional cerebralblood ow and studied energy utilization in the brains ofseven patients with AD by measuring arteriovenousdifferences of an array of substrates Single-photon emis-sion computed tomography showed that the patients hadsigni cantly decreased regional cerebral blood ow lim-ited to the parietotemporal region The glucose extractionfraction and global cerebral metabolic rate of glucosewere signi cantly decreased whereas the global cerebralmetabolic rate of oxygen was only slightly decreasedBecause the cerebral metabolic rates of the various sub-strates (ketones lactate etc) did not change the authorsconcluded that the markedly elevated oxygenglucoseutilization ratio indicated an altered energy metabolismin AD

Conceivably the impairment of glucose metabolismin AD patients described by Ogawa et al is attributableat least in part to local insulin resistance Watson andCraft82 called attention to emerging evidence that insulinabnormalities and insulin resistance in AD may contrib-ute to the pathophysiology and clinical symptoms of thisdisorder In their words ldquoIt has been recently demon-strated that insulin-sensitive glucose transporters are lo-calized in the same regions supporting memory and thatinsulin plays a role in memory functions Collectivelythese ndings suggest that insulin may contribute tonormal cognitive functioning and that insulin abnormal-ities may exacerbate cognitive impairments such asthose associated with ADrdquo The authors suggest thatimproving insulin effectiveness may have therapeuticbene t for patients with AD The suggestion of Watsonand Craft that insulin abnormalities may play a role inthe pathogenesis of AD has to be reconciled with the nding by Molina et al83 that cerebrospinal uid insulinlevels in 27 AD patients did not differ from those in 16matched controls

In 1996 Hoshi et al84 reported that a fragment ofthe beta chain of amyloid inhibits PDH activity in pri-mary cultures of hippocampal neurons A year later thesame authors found that the 1ndash42 fragment of the betachain of amyloid Ab1ndash42 can inhibit acetyl cholinesynthesis of septal neurons85 Veech et al50 note thatcultured hippocampal neurons die when exposed toAb1ndash42 However addition of 4 mM D-b-hydroxybu-tyrate protects these neurons from amyloid fragmentndashinduced death Casley et al86 showed that beta-amyloidcauses a signi cant reduction in state 3 and state 4mitochondrial respiration which is diminished further bynitric oxide These authors also found that cytochromeoxidase -ketoglutarate dehydrogenase and PDH activ-ities are inhibited by b-amyloid They concluded thatb-amyloid can directly disrupt mitochondrial function

inhibit key enzymes and contribute to the impairment ofenergy metabolism observed in AD

Veech et al50 contend that the foregoing studiesprovide a rationale for trial of a ketogenic diet in thetreatment of AD Ketones can overcome localized insulinresistance can bypass a block in PDH and have beenshown to protect dopaminergic neurons from free radicaldamage and from apoptosis induced by the 1ndash42 amy-loid fragment in tissue-culture experiments

What Next

In considering the potential usefulness of therapeutichyperketonemia in the treatment of AD PD and certainother neurodegenerative conditions it is essential to beclear about what is meant by terms like ldquoketogenic dietrdquoand ldquohyperketonemiardquo

There is widespread confusion among both physi-cians and lay individuals about what constitutes a keto-genic diet Similar to glucose ketones are present in theblood at all times As shown in Table 3 the range ofketone levels that can be achieved by prolonged fooddeprivation or by adherence to various diets is quitewide Hence it makes no sense to speak of a ldquoketogenicdietrdquo without also specifying the degree of serum ketoneelevation that the diet is intended to achieve

The serum concentrations sought for treatment ofepilepsy range from 2 to 7 mML Such levels can beachieved only by strict adherence to a regimen like the41 HKDmdashone most patients nd burdensome On theother hand the kind of low-carbohydrate ldquoketogenicdietrdquo (eg 30 protein 8 carbohydrate 61 fat [asproportions of total energy]) that has recently becomepopularmdashalthough much easier to followmdashachievesBHB levels of only 028ndash040 mML87 (This ldquopopularrdquoketogenic diet at a 2000 kcalday intake provides 127 gof presumptive carbohydrate [40 preformed] plus 87 gderivable from 150 g protein not including the smallquantity of glycerol released in the course of adipocytetriglyceride hydrolysis At the same energy level the 41HKD would yield 33 g of presumptive carbohydrate[10 g preformed] plus 23 g derivable from 40 g proteinThus the effective quantity of carbohydrate provided bythe popular ketogenic diet is almost four times thatsupplied by the 41 HKD)

Future research must address the problem of match-ing diet-induced elevation of serum ketone levels withtherapeutic bene ts Because the drastic type of hyper-ketogenic dietmdashthe one used to treat epilepsy and relatedneurologic disordersmdashis so onerous and dif cult to fol-low it is especially important to determine the extent towhich such a diet can be liberalized without losingdesirable therapeutic effects Given the daunting natureof the 41 HKD it is encouraging that there exists apotential nutritional stratagem which eventually may

337Nutrition Reviews Vol 61 No 10

make it possible to induce therapeutic hyperketonemia inpatients without requiring them to adhere to a Draconianregimen The stratagem entails administration by mouthof substantial quantities of ketones in a formmdashnamelyas esters or small polymersmdashsuitable for ingestion Intheir article about the potential therapeutic uses of ketonebodies Veech et al50 suggest that the problems associ-ated with ldquothe very unappetizing high-fat ketogenicdietrdquo might be obviated by feeding patients up to 100 to150 g per day of a nutritionally acceptable form ofketone bodies Unfortunately neither ketone esters noroligomers of BHB are presently available in suf cientquantities for the conduct of clinical trials

The traditional 41 HKD is dif cult but far fromunmanageable Numerous epileptic children and adultswith seizures resistant to anticonvulsant medication havemanaged to remain on this regimen for years withoutexperiencing serious adverse effects There are justi ableconcerns about its hyperlipidemic effect however andthe fact that the diet has been reported to increase risk ofside effects including nephrolithiasis88 depression ofplatelet function (manifested by easy bruising)89 andoccasionally prolongation of the electrocardiographicQTc interval in children90 It should be possible to mitigateor perhaps prevent the adverse effects of this very highndashfatdiet on plasma lipids by modifying its fatty acid contentWhat is most important now however is to get on with thejob of determining whether diet-induced hyperketonemia(at any level) will prove clinically bene cial in the man-agement of a number of neurodegenerative disorders par-ticularly Parkinsonrsquos disease and Alzheimerrsquos disease

Acknowledgment

We thank Richard L Veech MD DPhil for hisencouragement and very helpful suggestions We alsothank Steven W Fowkes for creative and elegant ren-dering of Figures 1ndash3 and editorial advice

1 Gerhardt J Diabetes mellitus und aceton WienerMedizinische Presse 18656672

2 Hirschfeld F Beobachtungen uber die Acentonurieund das Coma diabeticum Z klin Med 189528176 ndash209

3 Peters JP Van Slyke DD Quantitative ClinicalChemistry Interpretations Volume 1 2nd editionBaltimore Williams amp Wilkins 1946439

4 Cohen PP Studies in ketogenesis J Biol Chem1937119333ndash346

5 Cohen PP Stark IE Hepatic ketogenesis and ketol-ysis in different species J Biol Chem 193812697ndash107

6 Jowett M Quastel JH Fat metabolism III The for-mation and breakdown of acetoacetic acid in animaltissues Biochem J 1935292181ndash2191

7 Chaikoff IL Soskin S The utilization of acetoaceticacid by normal and diabetic dogs before and afterevisceration Am J Physiol 19288758 ndash72

8 Barnes RH Drury DR Greeley PO Wick AN Utili-zation of the ketone bodies in normal animals and inthose with ketosis Am J Physiol 1940130144 ndash150

9 Mirsky IA Broh-Kahn RH The inuence of in-creased metabolism on b-hydroxybutyric acid utili-zation Am J Physiol 1937120446 ndash 450

10 Waters ET Fletcher JP Mirsky IA The relationbetween carbohydrate and b-hydroxybutyric acid

Table 3 Relationship of Metabolic State and Diet Composition to Plasma Levels of Acetoacetate (AcAc)and b-Hydroxybutyrate (BHB) in Normal-weight Overweight and Obese Adults

Metabolic StateDiet

Composition CommentsAcAc

(mML)BHB

(mML) References

Postabsorptive Malesn 5 11 Normal weight 005 007 Hall et al18

n 5 12 Normal weight 008 Sharman et al87

n 5 10 Obese 015 027 Hall et al18

n 5 6 Overweight 001 002 Cahill et al39

Ketogenic diets Males (normal weight)(eucaloric) 3 weeks on diet (n 5 12) 040 Sharman et al87

C8 P30 F61 6 weeks on diet 028 Sharman et al87

Female (normal weight)C2 P8 F90 2 weeks on diet (n 5 1) 070 602 VanItallie Nonas amp Heyms elddagger

Fasting Malesn 5 6 (overweight)

1-week fast 095 358 Cahill et al39

n 5 6 (obese)1ndash2-week fast 104 597 Hall et al18

Distribution (percent total calories) of carbohydrate (C) protein (P) and fat (F)daggerUnpublished study (St Lukersquos-Roosevelt Hospital 2003)Where n 1 all values are means

338 Nutrition Reviews Vol 61 No 10

330 Nutrition Reviews Vol 61 No 10

very low in carbohydrate ketosis varies inversely withthe quantity of protein eaten This occurs because ap-proximately 48 to 58 of the amino acids in most dietaryproteins are glucogenic3536 For every 2 grams of proteinconsumed in a carbohydrate-free diet somewhere be-tween 10 and 12 grams are potentially convertible to

glucose Therefore to obtain a degree of hyperketonemia(approximately 2ndash7 mML) believed to be therapeuti-cally effective in certain important medical conditionssuch as epilepsy patients must rigorously restrict proteinas well as carbohydrate intake37 and when possibleincrease their level of physical activity23

Table 1 Postabsorptive Plasma Concentrations of Acetoacetate (AcAc) and b-Hydroxybutyrate (BHB)and Rates of Ketone Release and Removal Together with Rate Coefcients for Ketone Removal (meanSD) in Normal Diabetic and Normally Fed and Subsequently Food-deprived Severely Obese Subjects

DiagnosisMetabolic

StateAcAc

(mML)BHB

(mML)

KetoneRelease

Rate( mol

min 1 m 2)

KetoneRemoval

Rate( mol

min 1 m 2)

RateCoef cientfor KetoneRemoval(min 1)

Normal(n 5 11) Postabsorptive 005 002 007 004 81 66 1107 1059 0168 0169

Diabeticdagger

(n 5 6) Postabsorptive 067 075 179 263 208 118 2188 874 0055 0040ObeseDagger

(n 5 10) Postabsorptive 015 006 027 012 171 70 2101 932 0066 0040Food-deprived

1ndash2 weekssect 104 026 597 113 569 286 5673 2562 0027 0019

Fraction lost by oxidation and excretiondaggerRecently diagnosed untreated patients with insulin-dependent diabetes mellitusDaggerSubjects severely obesesectSubjects received 50 kcal and 1 multivitamin tabletdayData from Reference 18

Figure 1 Hepatic formation and peripheral utilization of ketone bodies Mobilization of non-esteri ed fatty acids (NEFA) fromadipose tissue fat stores by the activation of hormone-sensitive lipase is an initial step that provides ample substrate for the hepaticproduction of ketone bodies (b-hydroxybutyrate and acetoacetate) Once fatty acids enter the hepatocyte from the blood circulationthey are activated to CoA esters by the family of membrane-bound fatty acid acyl-CoA synthetases The action of one of thesesynthetases (long-chain acyl CoA synthetase) is identi ed as (3) After CoA esteri cation the metabolic paths diverge based on thefatty acid (acyl) chain length Short- and medium-chain acyl-CoAs readily cross the outer and inner mitochondrial membranes andundergo b-oxidation into acetyl-CoA and acetoacetate (the terminal 4-carbon fragment) Long-chain acyl-CoAs readily cross theouter mitochondrial membrane after which they must be converted to long-chain acylcarnitine by membrane-bound carnitinepalmitoyltransferaseI or CPT-1 (1) for subsequent transport across the inner mitochondrial membrane The inner membranendash boundenzyme carnitine acylcarnitine translocase (4) exchanges one long-chain acyl-CoA from the intermembrane space with one free(unacylated)carnitine from the matrix thereby permitting carnitinersquos return to the intermembrane space The inner membranendash boundenzyme carnitine palmitoyltransferase II or CPT-II (2) converts the fatty acid back into its CoA ester for subsequent b-oxidationfreeing carnitine for recycling Very longndash chain acyl-CoAs and unsaturated fatty acids of various lengths are transported intoperoxisomes along with long- and medium-chain acyl-CoAs for chain-length reduction via b-oxidation Unlike mitochondrialb-oxidation peroxisomal b-oxidation does not proceed to completion This results in peroxisomal export of medium-chain acyl andacylcarnitinecompounds to the mitochondria for further oxidationMitochondrial b-oxidation produces acetyl-CoA which enters theTCA cycle for further oxidation to produce ATP Ketogenesis occurs when hepatic energy requirements are met allowing excessacetyl-CoA to be acted upon by thiolase which condenses two acetyl-CoAs to form acetoacetyl-CoA Hydroxymethylglutyaryl-CoAsynthase (HMG-CoA synthase) condenses acetoacetyl-CoA with acetyl-CoA to form hydroxymethylglutaryl-CoA which is thendissociated into acetoacetate and acetyl-CoA by hydroxymethylglutaryl-CoA lyase (HMG-CoA lyase) The action of theenzyme b-hydroxybutyrate dehydrogenase (5) reduces much of the acetoacetate to b-hydroxybutyrate using NADH as areductant Both acetoacetate and b-hydroxybutyrate are transported back through the mitochondrial membranes for export intothe blood circulation The pathway of lipogenesis illustrated by dashed lines becomes activated in the fed state and results inthe synthesis of fatty acids and their esteri cation into triacylglycerol (triglycerides) The rst committed intermediary in fattyacid synthesis is malonyl-CoA which inhibits CPT-1 (enzyme 1) This inhibition is illustrated as a thin dashed line pointingto CPT-1 Inhibition of CPT-1 prevents mitochondrial uptake of fatty acyl-CoA (and its subsequent b-oxidation) and therebyblocks ketogenesis In the fed (non-ketotic) state triglyceride is exported from the liver as very lowndash density lipoprotein(VLDL)

331Nutrition Reviews Vol 61 No 10

Ketones Substrates in Search of a Mission

By the early 1960s a great deal of information hadaccumulated about the biochemistry and physiology ofthe ketone bodies yet their role in human metabolismremained obscure38 Nevertheless ve facts were knownto medical scientists that had they been viewed in properperspective would have suggested strongly that ketoneshave to be the brainrsquos back-up fuel during starvationFirst scientists recognized that when people are deprivedof food for 3 days or longer they become hyperketone-mic and hyperketonuric Second catheterization studiesshowed that the adult human brain uses approximately100 to 150 g glucose per day39 Third scientists under-stood that non-esteri ed fatty acids do not cross theblood-brain barrier (BBB) and are not directly availableto the brain as an energy source40 Fourth Benedict41

showed that during 31 days of total caloric starvation hisldquofasting manrdquo excreted an average of 872 gday ofnitrogen (N) which corresponds to a rate of body proteincatabolism of approximately 55 gday Approximately32 g of this amount (058 550) could have beenavailable as fuel for the brainmdashapproximately 27 ofthe brainrsquos daily needs In another subject deprived offood for 15 days mean total N excretion was 482 gdaymaking only 174 g of glucoseday potentially availableto the brain42

Finally because glucose was widely assumed to bethe exclusive fuel of the adult human brain38 one mightinfer that during continuing starvation (after liver glyco-gen stores were exhausted) the brain would have to besustained by glucose derived from intracellular proteinMaking 100 to 150 g of glucoseday available to thebrain would require the gluconeogenic breakdown ofsome 172 to 259 gday of body protein At this unsus-tainable rate of protein attrition death might be expectedto occur within approximately 2 weeksmdash certainly notthe 57 to 73 days (mean 61 25 [SEM] days) that itreportedly takes for totally starved young men of averagebody composition to die43 If the nutrition scientists ofthe time had given more thought to the meaning of thisunexplained ldquoprolongationrdquo of life during total starva-tion they might have reasoned that during extended fooddeprivation the brain is able to use the ketones generatedin such abundance as an alternative energy source Sucha scenario would have explained the greater than ex-pected longevity and the far lower than expected dailynitrogen de cit exhibited by starved individuals

In 1966 Cahill et al39 ldquoconnected the dotsrdquo andconcluded that ketones might provide an alternate fuelfor the brain during starvation To test this hypothesisthe authors performed brain catheterization studies inobese patients undergoing 5 to 6 weeks of therapeuticfasting44 This landmark study showed that under condi-tions of sustained starvation (during which the subjects

achieved blood ketone levels close to 7 mML and bloodglucose concentrations of ca 38 mML) BHB and AcAcreplaced glucose as the predominant fuel(s) for brainmetabolism The ldquostarvation paradoxrdquo was thus resolvedIt also became clear that during starvation the liverrsquosability to generate ketones for use by the brain representsa biologic strategy for maintaining strength and prolong-ing life during conditions of severe nutritional privationIn this way the burden of sustaining life during starva-tion is shifted from the bodyrsquos protein stores to itsreserves of fatmdasha far more effective and ef cient adap-tation Thanks to the pioneering studies of Cahill et almetabolismrsquos ugly duckling nally began to emerge asan incipient swan

Ketones Uniquely Efcient Metabolic Fuel

In the mid 1940s Lardy and associates4546 demonstratedthat when tested on sperm BHB and AcAc comparedwith other metabolic substrates have the ability to de-crease oxygen consumption while increasing motilityFor almost half a century this apparent increase in met-abolic ef ciency remained unexplained In 1994 and1995 however RL Veechrsquos group47ndash49 at the NationalInstitutes of Health showed that in the working perfusedrat heart when 5 mM ketone bodiesL were added to theglucose-containing perfusate there was a 25 increasein hydraulic work with a signi cant decrease in oxygenconsumption Addition of ketones also brought about areduction of the mitochrondrial NAD couple with oxi-dation of the mitochondrial co-enzyme Q couple Theconsequence is an increase in the energy of the redoxspan between site I and site II of the electron transportsystem

As Veech et al50 described ldquoAn increase in theredox span between two sites will result in an increasedenergy release by the electron traveling across that span With an increase in redox energy of the respiratorychain there is a corresponding increase in the energy ofthe protons ejected from the mitochondria at the energy-conserving sites which is then re ected in an increase inthe energy of ATP hydrolysisrdquo

As shown in Table 2 Veech and associates47ndash49 alsoobserved that the introduction of ketone bodies into therat heart perfusate produced a 16-fold rise in acetyl-CoAcontent and increases in TCA cycle intermediates Thesestimulatory effects on TCA cycle activity were similar tothose obtained with the administration of saturatingdoses of insulin Insulin increases glucose transport inthe myocardium by promoting translocation of insulin-sensitive GLUT4 to the plasma membrane Insulin alsostimulates the activity of pyruvate dehydrogenase mul-tienzyme complex (PDH)mdasha rate-limiting step in themitochondrial transformation of pyruvate to acetyl-CoAAs Veech et al50 put it ldquoThe implication was that

332 Nutrition Reviews Vol 61 No 10

ketosis which is the physiological response to insulindeprivation during starvation was equivalent in meta-bolic effects to the actions of insulin (Moreover)ketones by-passed the block in glucose transport causedby lack of insulin (and) also by-passed the block inPDH induced by insulin de ciency by providing analternative source of mitochondrial acetyl CoArdquo Thepathway by which ketones are able to achieve this effectis shown in Figure 2

Veech et al50 pointed out that the gradients of Kand Na (and Ca ) between the extra- and intracellularphases of the cell are normally in virtual equilibriumwith the resting electrical potential between these twophases and with the energy ( G) of ATP hydrolysiswhich sustains the action of the Na pump By increas-ing metabolic ef ciency and thereby the energy pro-duced by ATP hydrolysis BHB augments the potentialbetween the extra- and intracellular phases giving rise toan increase in the energy of the K gradient (ratio of[K ]out[K ]in) in muscle and brain A larger mitochon-drial membrane potential would lead to larger energyproduction because it is the mitrochondrial membranepotential difference that drives the production of ATPMaintenance of a suitable K gradient in the cell is ofcourse indispensable to its proper functioning As willbe discussed below impairment of ion homeostasiswithin neuronal cells in certain brain areas might resultin a degree of electrical instability suf cient to trigger anepileptic seizure (or in the heart a cardiac arrhythmia)

Studies of blood ow and oxygen consumption inthe brains of food-deprived obese human subjects re-vealed values (45 mL min 1 100 g 1 and 27 mL O2

min 1 100 mL 1)44 that were well below the normallevels for adult human brains of 57 and 36 respective-ly51 Although these data need con rmation they suggestan increase in the metabolic ef ciency in human brainsusing ketoacids as their principal energy source in placeof glucose

Examples of mechanisms by which diet-inducedhyperketonemia might overcome or circumvent meta-bolic impediments to utilization of such substrates asglucose and pyruvate are shown schematically in Fig-ure 3

Use of a ldquoHyperketogenicrdquo Diet in Treatmentof Seizure Disorders

Fasting has long been acknowledged as a method toprevent or reduce seizures in epileptic individuals37

Because of the obvious limitations of this treatment useof a ketogenic diet to mimic the metabolic effects of fooddeprivation was proposed in 1921 by both Woodyatt52

and Wilder53 at the Mayo Clinic In 1924 Peterman54

(also from Mayo) reported the clinical application andeffectiveness of the regimen suggested by Woodyatt andWilder Petermanrsquos diet provided (per day) 1 g of pro-teinkg body weight (considerably less in adults) 10 to15 g of carbohydrate and the remainder of the calories asfat The Peterman ketogenic diet is sometimes called the41 diet because it provided a ratio of approximately 4parts (by weight) fat to 1 part (by weight) of a mixture ofprotein and carbohydrate This distribution of the majormacronutrients is almost identical to that of most anti-convulsant diets used today37 The term ldquohyperketogenicdietrdquo (HKD) will be used in this review to distinguish the41 diet (and certain variations thereof) from the far lessdrastic low-carbohydrate mildly ketogenic diets thathave been advocated for weight control in a successionof popular diet books

In the 1970s several investigators5556 modi ed thePeterman HKD by substituting medium-chain triglycer-ides (MCT) in varying amounts for long-chain triglycer-ides (MCTrsquos ketogenic effect is not dependent on con-current restriction of dietary carbohydrate57) Althoughall such variations of the HKD were approximately equalin reducing seizures many of the children given MCT-

Table 2 Insulin and Ketone Effects on Cardiac Efciency Acetyl-CoA Production [NADH][NAD ] Ratioand Free Energy of ATP Hydrolysis ( GA TP ) in the Perfused Rat Heart

Substrates Addedto Perfusate

AcetylCoAdagger

[NADH][NAD ]Daggersect

(mitochondrial)

G of ATPHydrolysis(cytosolic)

kJmol

CardiacEf ciency

()

CardiacEf ciency( relativeto glucose)

Glucose 1 005 562 105 100Glucose Insulin 9 022para 589para 134para 128Glucose Ketones 15 062para 576para 130para 124Glucose Insulin

Ketones 18 043para 589para 143para 136

Data of Sato K et al4 8 and DaggerKashiwaya et al4 7 4 9

daggerNumbers represent multiples of that measured with perfusate containing only glucosesectExpressed by Kashiwaya et al4 7 4 9 as [NAD ][NADH]paraIndicates a signi cant difference (P 005) from perfusate containing 10 mmolL glucose ketones added at 4 mMLD-b-hydroxybutyrate 1 mML acetoacetate

333Nutrition Reviews Vol 61 No 10

augmented diets complained of nausea diarrhea andoccasional vomiting58

Among those patients who adhere to the 41 HKDfor 12 months 40 experience greater than 90 reduc-tion in seizures whereas an additional 40 experience50 to 90 reduction Although the precise mechanism ofaction responsible for the clinical ef cacy of the diet hasnever been clari ed it is likely that the hyperketonemiainduced by the 41 HKD is the key factor responsible forthe regimenrsquos anticonvulsant effect rather than the ac-companying metabolic acidosis59 or dehydration37 In

1972 Uhleman and Neims60 demonstrated that micepups fed a high-fat ketogenic diet exhibited signi cantlyincreased thresholds against investigator-applied electro-shock challenges This resistance to electroshock disap-peared promptly after the animals were shifted to ahigh-carbohydrate diet Subsequently a number of stud-ies have shown an increase in the electroconvulsivethreshold in rats maintained on a high-fat diet or de-prived of food for 48 hours De Vivo et al61 found thatthe ratios of ATP to adenosine diphosphate (ADP) weresigni cantly increased in the brains of rats fed the high-

Figure 2 Production of ATP (adenosine triphosphate) in extrahepatic tissues arising from use of ketones and glucose Glucose isregulated by a family of GLUT transporter proteins GLUT4 transporters found in most body tissues are mediated by insulinGLUT1 transporters common in tissues such as brain and erythrocytes allow glucose uptake without insulin mediation By contrastuptake of ketone bodies (shown as BHB and AcAc) occurs via the family of monocarboxylate transporters (MCTs) which are notinsulin mediated MCT proteins enable ketones to pass readily through the blood-brain barrier Many types of peripheral cellsincluding brain cells not only use glucose but also use ketones to produce acetyl-CoA Glucose utilization depends on glycolyticenzymes (not shown) which produce pyruvate Pyruvate in turn enters the mitochondrion and is converted to acetyl-CoA by thepyruvate dehydrogenase multienzyme complex (PDH) Ketone substrates enter the mitochondrion in the form of (1) acetoacetate(AcAc) and (2) b-hydroxybutyrate (BHB) which is readily converted to acetoacetate by b-hydroxybutyrate dehydrogenase (HBD)Two enzymes principally control ketolysis First 3-oxoacid-CoA transferase (OCT) adds coenzyme-A to AcAc which is then splitinto two molecules of acetyl-CoA by acetoacetyl-CoA thiolase (ACT) Acetyl-CoA undergoes oxidative degradation in the TCAcycle to reduce the electron carrriers NAD (nicotinamide adenine dinucleotide) and FAD ( avin adenine dinucleotide) to NADHand FADH2 respectively Acetyl-CoA is also used for lipid or sterol synthesis (not shown) NADH and FADH2 donate electrons tothe protein complexes I and II of the mitochondrial respiratory chain The ow of these electrons down the electron-transport chainto oxygen (O2) allows complexes I III and IV to pump protons (shown as H ) out of the matrix and into the intermembrane spacewhich generates a proton motive force (pmf) between the intermembrane space and the matrix The pmf provides the energy to driveprotons back through complex V (ATP synthase) to produce ATP from ADP (adenosine diphosphate) and inorganic phosphate (Pi)

334 Nutrition Reviews Vol 61 No 10

fat diet They concluded ldquo the anticonvulsant propertyof the ketogenic diet derives from the observed increasein cerebral energy reserve which in turn re ects thehigher ATPADP ratio [this] might provide a reason-able explanation for the increased neuronal stability thatdevelops during chronic ketosisrdquo In 1999 Pan et al62

carried out 31P magnetic resonance spectroscopic imag-ing studies in seven patients with intractable epilepsybefore and after institution of a ketogenic diet Theymeasured signi cant increases (P 005) in high-energy phosphates in association with adherence to theketogenic diet in these subjects They concluded thatenhancement of brain energy metabolism occurs withuse of the ketogenic diet in humans as in lab animals

Thavendiranathan et al63 have recently reportedstudies of the effect of the ldquoclassicrdquo ketogenic diet onseizures triggered by two threshold testsmdashthe pentyle-netetrazol infusion test and the threshold electroconvul-sive shock test These authors observed small but signif-icant threshold elevations (15ndash20) against bothchemically and electrically triggered seizures Similarlyprotective effects were not seen when suprathresholdstimuli were employed

Janigro64 summarized studies indicating that there isa reduction in both glucose uptake and metabolic activityin seizure foci in epileptic patients In his words ldquoDi-minished ion homeostasis together with increased meta-bolic demand of hyperactive neurons may further aggra-vate neuropathological consequences of BBB loss ofglucose uptake mechanismsrdquo Therefore one explanationfor the effectiveness of the 41 HKD in preventingseizures could be the increased transport of BHB at theBBB and its increased availability as an energy source tothe affected neurons BHB is not only an alternate sourceof energy for seizure foci in the brainmdashbypassing animpairment of carbohydrate utilization such as thatwhich can occur from a block in mitochondrial PDHactivitymdashbut it may actually promote ketone and glu-cose uptake in such foci In rats diet-induced ketosisproduces an eightfold increase in the activity of themonocarboxylate transporter and increases the level ofthe glucose transporter GLUT1 in brain endothelialcells and neuropil65 The hypothesis put forth by Janigroto explain how ketosis might eliminate or reduce seizuresis supported by reports that a suitable ketogenic diet isthe treatment of choice for two rare conditions that may

Figure 3 Conditions that can impede acetyl-CoA production from pyruvate and its precursors Examples include (1) insuf cientinsulin (2) insulin receptor abnormalities (3) diminished transport of glucose owing to impaired release of GLUT4 transportersassociated with insulin resistance [see below] (4) pyruvate dehydrogenase complex (PDH) de ciency or dysfunction (5)unavailabilityof glucose at the cell membrane and (6) GLUT1 mutations Provision of ketones to the cell can effectively circumventor overcome such impediments and compensate for the resulting de cits of glycolytically derived acetyl-CoA (GLUT1 designatesthe insulin-independent membrane protein responsible for glucose transport in red blood cells and throughout the blood-brain barrier[BBB] GLUT4 designates the insulin-dependentglucose transporterprotein expressed in most cells Many forms of insulin resistanceentail putative post-receptor defects of insulin-signal transduction resulting in the failure of insulin to release GLUT4 transportersfrom the GLUT4 pool at a rate suf cient to bind and transport enough extracellular glucose to maintain cellular pyruvate levels)

335Nutrition Reviews Vol 61 No 10

be complicated by seizuresmdashthe facilitated glucosetransporter protein type 1 (GLUT1) de ciency syn-drome6667 and the PDH complex de ciency syndrome68

Treatment with a ketogenic diet reportedly brought aboutdramatic clinical improvement in a one-year-old boywith Leigh syndrome which is associated with a de -ciency of PDH69 In this child treatment with a keto-genic diet was associated with striking improvement ofthe cerebral lesions as measured by neuroimaging

In 1973 Buckley and Williamson70 proposed thatneural lipid synthesis preferentially uses AcAc overglucose Patel and Owen71 obtained con rmatory datafor this hypothesis in 1976 In 1984 Koper et al72

compared the rate of incorporation of AcAc and glucoseinto fatty acids and cholesterol using labeled AcAc andglucose At saturating levels of glucose the authorsobserved that the rate at which AcAc label incorporatedinto fatty acids and cholesterol exceeded that of glucoseby a factor of 5 to 10 in both oligodendrocytes andastrocytes These observations provide a basis for argu-ing that some of the bene ts associated with ketogenicdiets may be related to their enhancing effects on lipo-genesis andor sterol synthesis (ie damaged neuronsmight repair more readily in brains given an adequateketone supply)

Therapeutic Potential of Diet-inducedHyperketonemia in Neurodegenerative andOther Disorders

Veech et al50 suggest that maintenance of a suitabledegree of diet-induced hyperketonemia (in the range of2ndash7 mML) has the potential to aid management ofseveral neurologic diseases including common neurode-generative diseases like Alzheimerrsquos disease (AD) andParkinsonrsquos disease (PD) In their ldquohypothesis paperrdquothe authors single out AD PD Friedreichrsquos AtaxiaLeprechaunisms and lesser forms of insulin disorders asexamples of illnesses that might bene t from therapeuticketosis They also describe experience with one case ofLaFora body disease (until now a uniformly fatal geneticdisease associated with progressive dementia) in whichuse of the 41 HKD resulted in unprecedented improve-ment in cognition gait and awareness after one month73

In this disease mutations in a gene encoding a novelprotein tyrosine phosphatase cause progressive myo-clonic epilepsy

Calabrese et al74 and DiMauro and Schon75 haverecently called attention to evidence suggesting the mi-tochondrial genome may play a key role in neurodegen-erative diseases Decreases in respiratory chain complexactivities have been identi ed in AD PD amyotrophiclateral sclerosis and Huntingtonrsquos disease Mitochondri-opathies (MCPs) possibly associated with oxidantanti-oxidant balance perturbation may underlie defects in

energy metabolism and contribute to cellular damage anddestruction

Parkinsonrsquos DiseaseAlthough the etiology of PD is complex involving bothgenetic and environmental factors there is growing in-terest in the role of mitochondrial dysfunction in thepathogenesis of this disorder Schapira76 pointed out thata proportion of PD patients exhibit a de ciency in mito-chondrial respiratory chain function most notably adefect of complex I within neurons of the substantianigra In his words ldquoBoth mitochondrial mutations andtoxic agents (endogenous and exogenous) have beendemonstrated to cause a de ciency in complex I func-tion This is achieved by increasing the degree of orsusceptibility to oxidative stress which may contributeto apoptopic cell deathrdquo

Finsterer77 reported metabolic studies indicating thatPD can be a manifestation of MCP In his series 63 ofPD patients had features of generalized MCP howeverhe used lactate stress testing to detect patients withMCPmdashan approach that could not be expected to pick upmitochondrial lesions limited to brain sites

According to Veech et al ldquoPD appears to resultfrom an acquired defect in the mitochondrial NADHmultienzyme complex of the dopaminergic cells of themesencephalonrdquo78 They point out although the diseaseis treatable for a period by administration of levodopa(the metabolic precursor of dopamine) continuing freeradical damage eventually reduces the effectiveness ofthis therapy Kashiwaya et al79 reported that primarycultures of mesencephalic dopaminergic neurons ex-posed to the meperidine analogue MPP (a molecule thatinhibits NADH dehydrogenase [complex 1] and causesoxygen free radical formation) can be protected fromdeath by addition of 4 mM D-b-hydroxybutyrate Veechet al50 suggest that BHB probably acts by decreasing thesource of mitochondrial oxygen radical formationldquo by oxidizing the coenzyme Q couple while at thesame time reducing the redox potential of the NADPcouple which through glutathione is the nal detoxi -cation step of H2O2rdquo In a recent review Schon andManfredi80 wrote ldquoBecause complexes I and III are theprincipal sources of free radicals in the cell alteredcomplex I function in the substantia nigra pars compactacould be responsible for the increased DNA damage andlipid peroxidation found in PD brainsrdquo

To the extent that ketones crossing the BBB couldcircumvent or compensate for mitochondrial defects as-sociated with PD or could protect dopaminergic neuronsfrom free radical damage a suitable ketogenic diet mightwell have bene cial effects in certain patients sufferingfrom Parkinsonism

336 Nutrition Reviews Vol 61 No 10

Alzheimerrsquos DiseaseIn 1996 Ogawa et al81 determined regional cerebralblood ow and studied energy utilization in the brains ofseven patients with AD by measuring arteriovenousdifferences of an array of substrates Single-photon emis-sion computed tomography showed that the patients hadsigni cantly decreased regional cerebral blood ow lim-ited to the parietotemporal region The glucose extractionfraction and global cerebral metabolic rate of glucosewere signi cantly decreased whereas the global cerebralmetabolic rate of oxygen was only slightly decreasedBecause the cerebral metabolic rates of the various sub-strates (ketones lactate etc) did not change the authorsconcluded that the markedly elevated oxygenglucoseutilization ratio indicated an altered energy metabolismin AD

Conceivably the impairment of glucose metabolismin AD patients described by Ogawa et al is attributableat least in part to local insulin resistance Watson andCraft82 called attention to emerging evidence that insulinabnormalities and insulin resistance in AD may contrib-ute to the pathophysiology and clinical symptoms of thisdisorder In their words ldquoIt has been recently demon-strated that insulin-sensitive glucose transporters are lo-calized in the same regions supporting memory and thatinsulin plays a role in memory functions Collectivelythese ndings suggest that insulin may contribute tonormal cognitive functioning and that insulin abnormal-ities may exacerbate cognitive impairments such asthose associated with ADrdquo The authors suggest thatimproving insulin effectiveness may have therapeuticbene t for patients with AD The suggestion of Watsonand Craft that insulin abnormalities may play a role inthe pathogenesis of AD has to be reconciled with the nding by Molina et al83 that cerebrospinal uid insulinlevels in 27 AD patients did not differ from those in 16matched controls

In 1996 Hoshi et al84 reported that a fragment ofthe beta chain of amyloid inhibits PDH activity in pri-mary cultures of hippocampal neurons A year later thesame authors found that the 1ndash42 fragment of the betachain of amyloid Ab1ndash42 can inhibit acetyl cholinesynthesis of septal neurons85 Veech et al50 note thatcultured hippocampal neurons die when exposed toAb1ndash42 However addition of 4 mM D-b-hydroxybu-tyrate protects these neurons from amyloid fragmentndashinduced death Casley et al86 showed that beta-amyloidcauses a signi cant reduction in state 3 and state 4mitochondrial respiration which is diminished further bynitric oxide These authors also found that cytochromeoxidase -ketoglutarate dehydrogenase and PDH activ-ities are inhibited by b-amyloid They concluded thatb-amyloid can directly disrupt mitochondrial function

inhibit key enzymes and contribute to the impairment ofenergy metabolism observed in AD

Veech et al50 contend that the foregoing studiesprovide a rationale for trial of a ketogenic diet in thetreatment of AD Ketones can overcome localized insulinresistance can bypass a block in PDH and have beenshown to protect dopaminergic neurons from free radicaldamage and from apoptosis induced by the 1ndash42 amy-loid fragment in tissue-culture experiments

What Next

In considering the potential usefulness of therapeutichyperketonemia in the treatment of AD PD and certainother neurodegenerative conditions it is essential to beclear about what is meant by terms like ldquoketogenic dietrdquoand ldquohyperketonemiardquo

There is widespread confusion among both physi-cians and lay individuals about what constitutes a keto-genic diet Similar to glucose ketones are present in theblood at all times As shown in Table 3 the range ofketone levels that can be achieved by prolonged fooddeprivation or by adherence to various diets is quitewide Hence it makes no sense to speak of a ldquoketogenicdietrdquo without also specifying the degree of serum ketoneelevation that the diet is intended to achieve

The serum concentrations sought for treatment ofepilepsy range from 2 to 7 mML Such levels can beachieved only by strict adherence to a regimen like the41 HKDmdashone most patients nd burdensome On theother hand the kind of low-carbohydrate ldquoketogenicdietrdquo (eg 30 protein 8 carbohydrate 61 fat [asproportions of total energy]) that has recently becomepopularmdashalthough much easier to followmdashachievesBHB levels of only 028ndash040 mML87 (This ldquopopularrdquoketogenic diet at a 2000 kcalday intake provides 127 gof presumptive carbohydrate [40 preformed] plus 87 gderivable from 150 g protein not including the smallquantity of glycerol released in the course of adipocytetriglyceride hydrolysis At the same energy level the 41HKD would yield 33 g of presumptive carbohydrate[10 g preformed] plus 23 g derivable from 40 g proteinThus the effective quantity of carbohydrate provided bythe popular ketogenic diet is almost four times thatsupplied by the 41 HKD)

Future research must address the problem of match-ing diet-induced elevation of serum ketone levels withtherapeutic bene ts Because the drastic type of hyper-ketogenic dietmdashthe one used to treat epilepsy and relatedneurologic disordersmdashis so onerous and dif cult to fol-low it is especially important to determine the extent towhich such a diet can be liberalized without losingdesirable therapeutic effects Given the daunting natureof the 41 HKD it is encouraging that there exists apotential nutritional stratagem which eventually may

337Nutrition Reviews Vol 61 No 10

make it possible to induce therapeutic hyperketonemia inpatients without requiring them to adhere to a Draconianregimen The stratagem entails administration by mouthof substantial quantities of ketones in a formmdashnamelyas esters or small polymersmdashsuitable for ingestion Intheir article about the potential therapeutic uses of ketonebodies Veech et al50 suggest that the problems associ-ated with ldquothe very unappetizing high-fat ketogenicdietrdquo might be obviated by feeding patients up to 100 to150 g per day of a nutritionally acceptable form ofketone bodies Unfortunately neither ketone esters noroligomers of BHB are presently available in suf cientquantities for the conduct of clinical trials

The traditional 41 HKD is dif cult but far fromunmanageable Numerous epileptic children and adultswith seizures resistant to anticonvulsant medication havemanaged to remain on this regimen for years withoutexperiencing serious adverse effects There are justi ableconcerns about its hyperlipidemic effect however andthe fact that the diet has been reported to increase risk ofside effects including nephrolithiasis88 depression ofplatelet function (manifested by easy bruising)89 andoccasionally prolongation of the electrocardiographicQTc interval in children90 It should be possible to mitigateor perhaps prevent the adverse effects of this very highndashfatdiet on plasma lipids by modifying its fatty acid contentWhat is most important now however is to get on with thejob of determining whether diet-induced hyperketonemia(at any level) will prove clinically bene cial in the man-agement of a number of neurodegenerative disorders par-ticularly Parkinsonrsquos disease and Alzheimerrsquos disease

Acknowledgment

We thank Richard L Veech MD DPhil for hisencouragement and very helpful suggestions We alsothank Steven W Fowkes for creative and elegant ren-dering of Figures 1ndash3 and editorial advice

1 Gerhardt J Diabetes mellitus und aceton WienerMedizinische Presse 18656672

2 Hirschfeld F Beobachtungen uber die Acentonurieund das Coma diabeticum Z klin Med 189528176 ndash209

3 Peters JP Van Slyke DD Quantitative ClinicalChemistry Interpretations Volume 1 2nd editionBaltimore Williams amp Wilkins 1946439

4 Cohen PP Studies in ketogenesis J Biol Chem1937119333ndash346

5 Cohen PP Stark IE Hepatic ketogenesis and ketol-ysis in different species J Biol Chem 193812697ndash107

6 Jowett M Quastel JH Fat metabolism III The for-mation and breakdown of acetoacetic acid in animaltissues Biochem J 1935292181ndash2191

7 Chaikoff IL Soskin S The utilization of acetoaceticacid by normal and diabetic dogs before and afterevisceration Am J Physiol 19288758 ndash72

8 Barnes RH Drury DR Greeley PO Wick AN Utili-zation of the ketone bodies in normal animals and inthose with ketosis Am J Physiol 1940130144 ndash150

9 Mirsky IA Broh-Kahn RH The inuence of in-creased metabolism on b-hydroxybutyric acid utili-zation Am J Physiol 1937120446 ndash 450

10 Waters ET Fletcher JP Mirsky IA The relationbetween carbohydrate and b-hydroxybutyric acid

Table 3 Relationship of Metabolic State and Diet Composition to Plasma Levels of Acetoacetate (AcAc)and b-Hydroxybutyrate (BHB) in Normal-weight Overweight and Obese Adults

Metabolic StateDiet

Composition CommentsAcAc

(mML)BHB

(mML) References

Postabsorptive Malesn 5 11 Normal weight 005 007 Hall et al18

n 5 12 Normal weight 008 Sharman et al87

n 5 10 Obese 015 027 Hall et al18

n 5 6 Overweight 001 002 Cahill et al39

Ketogenic diets Males (normal weight)(eucaloric) 3 weeks on diet (n 5 12) 040 Sharman et al87

C8 P30 F61 6 weeks on diet 028 Sharman et al87

Female (normal weight)C2 P8 F90 2 weeks on diet (n 5 1) 070 602 VanItallie Nonas amp Heyms elddagger

Fasting Malesn 5 6 (overweight)

1-week fast 095 358 Cahill et al39

n 5 6 (obese)1ndash2-week fast 104 597 Hall et al18

Distribution (percent total calories) of carbohydrate (C) protein (P) and fat (F)daggerUnpublished study (St Lukersquos-Roosevelt Hospital 2003)Where n 1 all values are means

338 Nutrition Reviews Vol 61 No 10

very low in carbohydrate ketosis varies inversely withthe quantity of protein eaten This occurs because ap-proximately 48 to 58 of the amino acids in most dietaryproteins are glucogenic3536 For every 2 grams of proteinconsumed in a carbohydrate-free diet somewhere be-tween 10 and 12 grams are potentially convertible to

glucose Therefore to obtain a degree of hyperketonemia(approximately 2ndash7 mML) believed to be therapeuti-cally effective in certain important medical conditionssuch as epilepsy patients must rigorously restrict proteinas well as carbohydrate intake37 and when possibleincrease their level of physical activity23

Table 1 Postabsorptive Plasma Concentrations of Acetoacetate (AcAc) and b-Hydroxybutyrate (BHB)and Rates of Ketone Release and Removal Together with Rate Coefcients for Ketone Removal (meanSD) in Normal Diabetic and Normally Fed and Subsequently Food-deprived Severely Obese Subjects

DiagnosisMetabolic

StateAcAc

(mML)BHB

(mML)

KetoneRelease

Rate( mol

min 1 m 2)

KetoneRemoval

Rate( mol

min 1 m 2)

RateCoef cientfor KetoneRemoval(min 1)

Normal(n 5 11) Postabsorptive 005 002 007 004 81 66 1107 1059 0168 0169

Diabeticdagger

(n 5 6) Postabsorptive 067 075 179 263 208 118 2188 874 0055 0040ObeseDagger

(n 5 10) Postabsorptive 015 006 027 012 171 70 2101 932 0066 0040Food-deprived

1ndash2 weekssect 104 026 597 113 569 286 5673 2562 0027 0019

Fraction lost by oxidation and excretiondaggerRecently diagnosed untreated patients with insulin-dependent diabetes mellitusDaggerSubjects severely obesesectSubjects received 50 kcal and 1 multivitamin tabletdayData from Reference 18

Figure 1 Hepatic formation and peripheral utilization of ketone bodies Mobilization of non-esteri ed fatty acids (NEFA) fromadipose tissue fat stores by the activation of hormone-sensitive lipase is an initial step that provides ample substrate for the hepaticproduction of ketone bodies (b-hydroxybutyrate and acetoacetate) Once fatty acids enter the hepatocyte from the blood circulationthey are activated to CoA esters by the family of membrane-bound fatty acid acyl-CoA synthetases The action of one of thesesynthetases (long-chain acyl CoA synthetase) is identi ed as (3) After CoA esteri cation the metabolic paths diverge based on thefatty acid (acyl) chain length Short- and medium-chain acyl-CoAs readily cross the outer and inner mitochondrial membranes andundergo b-oxidation into acetyl-CoA and acetoacetate (the terminal 4-carbon fragment) Long-chain acyl-CoAs readily cross theouter mitochondrial membrane after which they must be converted to long-chain acylcarnitine by membrane-bound carnitinepalmitoyltransferaseI or CPT-1 (1) for subsequent transport across the inner mitochondrial membrane The inner membranendash boundenzyme carnitine acylcarnitine translocase (4) exchanges one long-chain acyl-CoA from the intermembrane space with one free(unacylated)carnitine from the matrix thereby permitting carnitinersquos return to the intermembrane space The inner membranendash boundenzyme carnitine palmitoyltransferase II or CPT-II (2) converts the fatty acid back into its CoA ester for subsequent b-oxidationfreeing carnitine for recycling Very longndash chain acyl-CoAs and unsaturated fatty acids of various lengths are transported intoperoxisomes along with long- and medium-chain acyl-CoAs for chain-length reduction via b-oxidation Unlike mitochondrialb-oxidation peroxisomal b-oxidation does not proceed to completion This results in peroxisomal export of medium-chain acyl andacylcarnitinecompounds to the mitochondria for further oxidationMitochondrial b-oxidation produces acetyl-CoA which enters theTCA cycle for further oxidation to produce ATP Ketogenesis occurs when hepatic energy requirements are met allowing excessacetyl-CoA to be acted upon by thiolase which condenses two acetyl-CoAs to form acetoacetyl-CoA Hydroxymethylglutyaryl-CoAsynthase (HMG-CoA synthase) condenses acetoacetyl-CoA with acetyl-CoA to form hydroxymethylglutaryl-CoA which is thendissociated into acetoacetate and acetyl-CoA by hydroxymethylglutaryl-CoA lyase (HMG-CoA lyase) The action of theenzyme b-hydroxybutyrate dehydrogenase (5) reduces much of the acetoacetate to b-hydroxybutyrate using NADH as areductant Both acetoacetate and b-hydroxybutyrate are transported back through the mitochondrial membranes for export intothe blood circulation The pathway of lipogenesis illustrated by dashed lines becomes activated in the fed state and results inthe synthesis of fatty acids and their esteri cation into triacylglycerol (triglycerides) The rst committed intermediary in fattyacid synthesis is malonyl-CoA which inhibits CPT-1 (enzyme 1) This inhibition is illustrated as a thin dashed line pointingto CPT-1 Inhibition of CPT-1 prevents mitochondrial uptake of fatty acyl-CoA (and its subsequent b-oxidation) and therebyblocks ketogenesis In the fed (non-ketotic) state triglyceride is exported from the liver as very lowndash density lipoprotein(VLDL)

331Nutrition Reviews Vol 61 No 10

Ketones Substrates in Search of a Mission

By the early 1960s a great deal of information hadaccumulated about the biochemistry and physiology ofthe ketone bodies yet their role in human metabolismremained obscure38 Nevertheless ve facts were knownto medical scientists that had they been viewed in properperspective would have suggested strongly that ketoneshave to be the brainrsquos back-up fuel during starvationFirst scientists recognized that when people are deprivedof food for 3 days or longer they become hyperketone-mic and hyperketonuric Second catheterization studiesshowed that the adult human brain uses approximately100 to 150 g glucose per day39 Third scientists under-stood that non-esteri ed fatty acids do not cross theblood-brain barrier (BBB) and are not directly availableto the brain as an energy source40 Fourth Benedict41

showed that during 31 days of total caloric starvation hisldquofasting manrdquo excreted an average of 872 gday ofnitrogen (N) which corresponds to a rate of body proteincatabolism of approximately 55 gday Approximately32 g of this amount (058 550) could have beenavailable as fuel for the brainmdashapproximately 27 ofthe brainrsquos daily needs In another subject deprived offood for 15 days mean total N excretion was 482 gdaymaking only 174 g of glucoseday potentially availableto the brain42

Finally because glucose was widely assumed to bethe exclusive fuel of the adult human brain38 one mightinfer that during continuing starvation (after liver glyco-gen stores were exhausted) the brain would have to besustained by glucose derived from intracellular proteinMaking 100 to 150 g of glucoseday available to thebrain would require the gluconeogenic breakdown ofsome 172 to 259 gday of body protein At this unsus-tainable rate of protein attrition death might be expectedto occur within approximately 2 weeksmdash certainly notthe 57 to 73 days (mean 61 25 [SEM] days) that itreportedly takes for totally starved young men of averagebody composition to die43 If the nutrition scientists ofthe time had given more thought to the meaning of thisunexplained ldquoprolongationrdquo of life during total starva-tion they might have reasoned that during extended fooddeprivation the brain is able to use the ketones generatedin such abundance as an alternative energy source Sucha scenario would have explained the greater than ex-pected longevity and the far lower than expected dailynitrogen de cit exhibited by starved individuals

In 1966 Cahill et al39 ldquoconnected the dotsrdquo andconcluded that ketones might provide an alternate fuelfor the brain during starvation To test this hypothesisthe authors performed brain catheterization studies inobese patients undergoing 5 to 6 weeks of therapeuticfasting44 This landmark study showed that under condi-tions of sustained starvation (during which the subjects

achieved blood ketone levels close to 7 mML and bloodglucose concentrations of ca 38 mML) BHB and AcAcreplaced glucose as the predominant fuel(s) for brainmetabolism The ldquostarvation paradoxrdquo was thus resolvedIt also became clear that during starvation the liverrsquosability to generate ketones for use by the brain representsa biologic strategy for maintaining strength and prolong-ing life during conditions of severe nutritional privationIn this way the burden of sustaining life during starva-tion is shifted from the bodyrsquos protein stores to itsreserves of fatmdasha far more effective and ef cient adap-tation Thanks to the pioneering studies of Cahill et almetabolismrsquos ugly duckling nally began to emerge asan incipient swan

Ketones Uniquely Efcient Metabolic Fuel

In the mid 1940s Lardy and associates4546 demonstratedthat when tested on sperm BHB and AcAc comparedwith other metabolic substrates have the ability to de-crease oxygen consumption while increasing motilityFor almost half a century this apparent increase in met-abolic ef ciency remained unexplained In 1994 and1995 however RL Veechrsquos group47ndash49 at the NationalInstitutes of Health showed that in the working perfusedrat heart when 5 mM ketone bodiesL were added to theglucose-containing perfusate there was a 25 increasein hydraulic work with a signi cant decrease in oxygenconsumption Addition of ketones also brought about areduction of the mitochrondrial NAD couple with oxi-dation of the mitochondrial co-enzyme Q couple Theconsequence is an increase in the energy of the redoxspan between site I and site II of the electron transportsystem

As Veech et al50 described ldquoAn increase in theredox span between two sites will result in an increasedenergy release by the electron traveling across that span With an increase in redox energy of the respiratorychain there is a corresponding increase in the energy ofthe protons ejected from the mitochondria at the energy-conserving sites which is then re ected in an increase inthe energy of ATP hydrolysisrdquo

As shown in Table 2 Veech and associates47ndash49 alsoobserved that the introduction of ketone bodies into therat heart perfusate produced a 16-fold rise in acetyl-CoAcontent and increases in TCA cycle intermediates Thesestimulatory effects on TCA cycle activity were similar tothose obtained with the administration of saturatingdoses of insulin Insulin increases glucose transport inthe myocardium by promoting translocation of insulin-sensitive GLUT4 to the plasma membrane Insulin alsostimulates the activity of pyruvate dehydrogenase mul-tienzyme complex (PDH)mdasha rate-limiting step in themitochondrial transformation of pyruvate to acetyl-CoAAs Veech et al50 put it ldquoThe implication was that

332 Nutrition Reviews Vol 61 No 10

ketosis which is the physiological response to insulindeprivation during starvation was equivalent in meta-bolic effects to the actions of insulin (Moreover)ketones by-passed the block in glucose transport causedby lack of insulin (and) also by-passed the block inPDH induced by insulin de ciency by providing analternative source of mitochondrial acetyl CoArdquo Thepathway by which ketones are able to achieve this effectis shown in Figure 2

Veech et al50 pointed out that the gradients of Kand Na (and Ca ) between the extra- and intracellularphases of the cell are normally in virtual equilibriumwith the resting electrical potential between these twophases and with the energy ( G) of ATP hydrolysiswhich sustains the action of the Na pump By increas-ing metabolic ef ciency and thereby the energy pro-duced by ATP hydrolysis BHB augments the potentialbetween the extra- and intracellular phases giving rise toan increase in the energy of the K gradient (ratio of[K ]out[K ]in) in muscle and brain A larger mitochon-drial membrane potential would lead to larger energyproduction because it is the mitrochondrial membranepotential difference that drives the production of ATPMaintenance of a suitable K gradient in the cell is ofcourse indispensable to its proper functioning As willbe discussed below impairment of ion homeostasiswithin neuronal cells in certain brain areas might resultin a degree of electrical instability suf cient to trigger anepileptic seizure (or in the heart a cardiac arrhythmia)

Studies of blood ow and oxygen consumption inthe brains of food-deprived obese human subjects re-vealed values (45 mL min 1 100 g 1 and 27 mL O2

min 1 100 mL 1)44 that were well below the normallevels for adult human brains of 57 and 36 respective-ly51 Although these data need con rmation they suggestan increase in the metabolic ef ciency in human brainsusing ketoacids as their principal energy source in placeof glucose

Examples of mechanisms by which diet-inducedhyperketonemia might overcome or circumvent meta-bolic impediments to utilization of such substrates asglucose and pyruvate are shown schematically in Fig-ure 3

Use of a ldquoHyperketogenicrdquo Diet in Treatmentof Seizure Disorders

Fasting has long been acknowledged as a method toprevent or reduce seizures in epileptic individuals37

Because of the obvious limitations of this treatment useof a ketogenic diet to mimic the metabolic effects of fooddeprivation was proposed in 1921 by both Woodyatt52

and Wilder53 at the Mayo Clinic In 1924 Peterman54

(also from Mayo) reported the clinical application andeffectiveness of the regimen suggested by Woodyatt andWilder Petermanrsquos diet provided (per day) 1 g of pro-teinkg body weight (considerably less in adults) 10 to15 g of carbohydrate and the remainder of the calories asfat The Peterman ketogenic diet is sometimes called the41 diet because it provided a ratio of approximately 4parts (by weight) fat to 1 part (by weight) of a mixture ofprotein and carbohydrate This distribution of the majormacronutrients is almost identical to that of most anti-convulsant diets used today37 The term ldquohyperketogenicdietrdquo (HKD) will be used in this review to distinguish the41 diet (and certain variations thereof) from the far lessdrastic low-carbohydrate mildly ketogenic diets thathave been advocated for weight control in a successionof popular diet books

In the 1970s several investigators5556 modi ed thePeterman HKD by substituting medium-chain triglycer-ides (MCT) in varying amounts for long-chain triglycer-ides (MCTrsquos ketogenic effect is not dependent on con-current restriction of dietary carbohydrate57) Althoughall such variations of the HKD were approximately equalin reducing seizures many of the children given MCT-

Table 2 Insulin and Ketone Effects on Cardiac Efciency Acetyl-CoA Production [NADH][NAD ] Ratioand Free Energy of ATP Hydrolysis ( GA TP ) in the Perfused Rat Heart

Substrates Addedto Perfusate

AcetylCoAdagger

[NADH][NAD ]Daggersect

(mitochondrial)

G of ATPHydrolysis(cytosolic)

kJmol

CardiacEf ciency

()

CardiacEf ciency( relativeto glucose)

Glucose 1 005 562 105 100Glucose Insulin 9 022para 589para 134para 128Glucose Ketones 15 062para 576para 130para 124Glucose Insulin

Ketones 18 043para 589para 143para 136

Data of Sato K et al4 8 and DaggerKashiwaya et al4 7 4 9

daggerNumbers represent multiples of that measured with perfusate containing only glucosesectExpressed by Kashiwaya et al4 7 4 9 as [NAD ][NADH]paraIndicates a signi cant difference (P 005) from perfusate containing 10 mmolL glucose ketones added at 4 mMLD-b-hydroxybutyrate 1 mML acetoacetate

333Nutrition Reviews Vol 61 No 10

augmented diets complained of nausea diarrhea andoccasional vomiting58

Among those patients who adhere to the 41 HKDfor 12 months 40 experience greater than 90 reduc-tion in seizures whereas an additional 40 experience50 to 90 reduction Although the precise mechanism ofaction responsible for the clinical ef cacy of the diet hasnever been clari ed it is likely that the hyperketonemiainduced by the 41 HKD is the key factor responsible forthe regimenrsquos anticonvulsant effect rather than the ac-companying metabolic acidosis59 or dehydration37 In

1972 Uhleman and Neims60 demonstrated that micepups fed a high-fat ketogenic diet exhibited signi cantlyincreased thresholds against investigator-applied electro-shock challenges This resistance to electroshock disap-peared promptly after the animals were shifted to ahigh-carbohydrate diet Subsequently a number of stud-ies have shown an increase in the electroconvulsivethreshold in rats maintained on a high-fat diet or de-prived of food for 48 hours De Vivo et al61 found thatthe ratios of ATP to adenosine diphosphate (ADP) weresigni cantly increased in the brains of rats fed the high-

Figure 2 Production of ATP (adenosine triphosphate) in extrahepatic tissues arising from use of ketones and glucose Glucose isregulated by a family of GLUT transporter proteins GLUT4 transporters found in most body tissues are mediated by insulinGLUT1 transporters common in tissues such as brain and erythrocytes allow glucose uptake without insulin mediation By contrastuptake of ketone bodies (shown as BHB and AcAc) occurs via the family of monocarboxylate transporters (MCTs) which are notinsulin mediated MCT proteins enable ketones to pass readily through the blood-brain barrier Many types of peripheral cellsincluding brain cells not only use glucose but also use ketones to produce acetyl-CoA Glucose utilization depends on glycolyticenzymes (not shown) which produce pyruvate Pyruvate in turn enters the mitochondrion and is converted to acetyl-CoA by thepyruvate dehydrogenase multienzyme complex (PDH) Ketone substrates enter the mitochondrion in the form of (1) acetoacetate(AcAc) and (2) b-hydroxybutyrate (BHB) which is readily converted to acetoacetate by b-hydroxybutyrate dehydrogenase (HBD)Two enzymes principally control ketolysis First 3-oxoacid-CoA transferase (OCT) adds coenzyme-A to AcAc which is then splitinto two molecules of acetyl-CoA by acetoacetyl-CoA thiolase (ACT) Acetyl-CoA undergoes oxidative degradation in the TCAcycle to reduce the electron carrriers NAD (nicotinamide adenine dinucleotide) and FAD ( avin adenine dinucleotide) to NADHand FADH2 respectively Acetyl-CoA is also used for lipid or sterol synthesis (not shown) NADH and FADH2 donate electrons tothe protein complexes I and II of the mitochondrial respiratory chain The ow of these electrons down the electron-transport chainto oxygen (O2) allows complexes I III and IV to pump protons (shown as H ) out of the matrix and into the intermembrane spacewhich generates a proton motive force (pmf) between the intermembrane space and the matrix The pmf provides the energy to driveprotons back through complex V (ATP synthase) to produce ATP from ADP (adenosine diphosphate) and inorganic phosphate (Pi)

334 Nutrition Reviews Vol 61 No 10

fat diet They concluded ldquo the anticonvulsant propertyof the ketogenic diet derives from the observed increasein cerebral energy reserve which in turn re ects thehigher ATPADP ratio [this] might provide a reason-able explanation for the increased neuronal stability thatdevelops during chronic ketosisrdquo In 1999 Pan et al62

carried out 31P magnetic resonance spectroscopic imag-ing studies in seven patients with intractable epilepsybefore and after institution of a ketogenic diet Theymeasured signi cant increases (P 005) in high-energy phosphates in association with adherence to theketogenic diet in these subjects They concluded thatenhancement of brain energy metabolism occurs withuse of the ketogenic diet in humans as in lab animals

Thavendiranathan et al63 have recently reportedstudies of the effect of the ldquoclassicrdquo ketogenic diet onseizures triggered by two threshold testsmdashthe pentyle-netetrazol infusion test and the threshold electroconvul-sive shock test These authors observed small but signif-icant threshold elevations (15ndash20) against bothchemically and electrically triggered seizures Similarlyprotective effects were not seen when suprathresholdstimuli were employed

Janigro64 summarized studies indicating that there isa reduction in both glucose uptake and metabolic activityin seizure foci in epileptic patients In his words ldquoDi-minished ion homeostasis together with increased meta-bolic demand of hyperactive neurons may further aggra-vate neuropathological consequences of BBB loss ofglucose uptake mechanismsrdquo Therefore one explanationfor the effectiveness of the 41 HKD in preventingseizures could be the increased transport of BHB at theBBB and its increased availability as an energy source tothe affected neurons BHB is not only an alternate sourceof energy for seizure foci in the brainmdashbypassing animpairment of carbohydrate utilization such as thatwhich can occur from a block in mitochondrial PDHactivitymdashbut it may actually promote ketone and glu-cose uptake in such foci In rats diet-induced ketosisproduces an eightfold increase in the activity of themonocarboxylate transporter and increases the level ofthe glucose transporter GLUT1 in brain endothelialcells and neuropil65 The hypothesis put forth by Janigroto explain how ketosis might eliminate or reduce seizuresis supported by reports that a suitable ketogenic diet isthe treatment of choice for two rare conditions that may

Figure 3 Conditions that can impede acetyl-CoA production from pyruvate and its precursors Examples include (1) insuf cientinsulin (2) insulin receptor abnormalities (3) diminished transport of glucose owing to impaired release of GLUT4 transportersassociated with insulin resistance [see below] (4) pyruvate dehydrogenase complex (PDH) de ciency or dysfunction (5)unavailabilityof glucose at the cell membrane and (6) GLUT1 mutations Provision of ketones to the cell can effectively circumventor overcome such impediments and compensate for the resulting de cits of glycolytically derived acetyl-CoA (GLUT1 designatesthe insulin-independent membrane protein responsible for glucose transport in red blood cells and throughout the blood-brain barrier[BBB] GLUT4 designates the insulin-dependentglucose transporterprotein expressed in most cells Many forms of insulin resistanceentail putative post-receptor defects of insulin-signal transduction resulting in the failure of insulin to release GLUT4 transportersfrom the GLUT4 pool at a rate suf cient to bind and transport enough extracellular glucose to maintain cellular pyruvate levels)

335Nutrition Reviews Vol 61 No 10

be complicated by seizuresmdashthe facilitated glucosetransporter protein type 1 (GLUT1) de ciency syn-drome6667 and the PDH complex de ciency syndrome68

Treatment with a ketogenic diet reportedly brought aboutdramatic clinical improvement in a one-year-old boywith Leigh syndrome which is associated with a de -ciency of PDH69 In this child treatment with a keto-genic diet was associated with striking improvement ofthe cerebral lesions as measured by neuroimaging

In 1973 Buckley and Williamson70 proposed thatneural lipid synthesis preferentially uses AcAc overglucose Patel and Owen71 obtained con rmatory datafor this hypothesis in 1976 In 1984 Koper et al72

compared the rate of incorporation of AcAc and glucoseinto fatty acids and cholesterol using labeled AcAc andglucose At saturating levels of glucose the authorsobserved that the rate at which AcAc label incorporatedinto fatty acids and cholesterol exceeded that of glucoseby a factor of 5 to 10 in both oligodendrocytes andastrocytes These observations provide a basis for argu-ing that some of the bene ts associated with ketogenicdiets may be related to their enhancing effects on lipo-genesis andor sterol synthesis (ie damaged neuronsmight repair more readily in brains given an adequateketone supply)

Therapeutic Potential of Diet-inducedHyperketonemia in Neurodegenerative andOther Disorders

Veech et al50 suggest that maintenance of a suitabledegree of diet-induced hyperketonemia (in the range of2ndash7 mML) has the potential to aid management ofseveral neurologic diseases including common neurode-generative diseases like Alzheimerrsquos disease (AD) andParkinsonrsquos disease (PD) In their ldquohypothesis paperrdquothe authors single out AD PD Friedreichrsquos AtaxiaLeprechaunisms and lesser forms of insulin disorders asexamples of illnesses that might bene t from therapeuticketosis They also describe experience with one case ofLaFora body disease (until now a uniformly fatal geneticdisease associated with progressive dementia) in whichuse of the 41 HKD resulted in unprecedented improve-ment in cognition gait and awareness after one month73

In this disease mutations in a gene encoding a novelprotein tyrosine phosphatase cause progressive myo-clonic epilepsy

Calabrese et al74 and DiMauro and Schon75 haverecently called attention to evidence suggesting the mi-tochondrial genome may play a key role in neurodegen-erative diseases Decreases in respiratory chain complexactivities have been identi ed in AD PD amyotrophiclateral sclerosis and Huntingtonrsquos disease Mitochondri-opathies (MCPs) possibly associated with oxidantanti-oxidant balance perturbation may underlie defects in

energy metabolism and contribute to cellular damage anddestruction

Parkinsonrsquos DiseaseAlthough the etiology of PD is complex involving bothgenetic and environmental factors there is growing in-terest in the role of mitochondrial dysfunction in thepathogenesis of this disorder Schapira76 pointed out thata proportion of PD patients exhibit a de ciency in mito-chondrial respiratory chain function most notably adefect of complex I within neurons of the substantianigra In his words ldquoBoth mitochondrial mutations andtoxic agents (endogenous and exogenous) have beendemonstrated to cause a de ciency in complex I func-tion This is achieved by increasing the degree of orsusceptibility to oxidative stress which may contributeto apoptopic cell deathrdquo

Finsterer77 reported metabolic studies indicating thatPD can be a manifestation of MCP In his series 63 ofPD patients had features of generalized MCP howeverhe used lactate stress testing to detect patients withMCPmdashan approach that could not be expected to pick upmitochondrial lesions limited to brain sites

According to Veech et al ldquoPD appears to resultfrom an acquired defect in the mitochondrial NADHmultienzyme complex of the dopaminergic cells of themesencephalonrdquo78 They point out although the diseaseis treatable for a period by administration of levodopa(the metabolic precursor of dopamine) continuing freeradical damage eventually reduces the effectiveness ofthis therapy Kashiwaya et al79 reported that primarycultures of mesencephalic dopaminergic neurons ex-posed to the meperidine analogue MPP (a molecule thatinhibits NADH dehydrogenase [complex 1] and causesoxygen free radical formation) can be protected fromdeath by addition of 4 mM D-b-hydroxybutyrate Veechet al50 suggest that BHB probably acts by decreasing thesource of mitochondrial oxygen radical formationldquo by oxidizing the coenzyme Q couple while at thesame time reducing the redox potential of the NADPcouple which through glutathione is the nal detoxi -cation step of H2O2rdquo In a recent review Schon andManfredi80 wrote ldquoBecause complexes I and III are theprincipal sources of free radicals in the cell alteredcomplex I function in the substantia nigra pars compactacould be responsible for the increased DNA damage andlipid peroxidation found in PD brainsrdquo

To the extent that ketones crossing the BBB couldcircumvent or compensate for mitochondrial defects as-sociated with PD or could protect dopaminergic neuronsfrom free radical damage a suitable ketogenic diet mightwell have bene cial effects in certain patients sufferingfrom Parkinsonism

336 Nutrition Reviews Vol 61 No 10

Alzheimerrsquos DiseaseIn 1996 Ogawa et al81 determined regional cerebralblood ow and studied energy utilization in the brains ofseven patients with AD by measuring arteriovenousdifferences of an array of substrates Single-photon emis-sion computed tomography showed that the patients hadsigni cantly decreased regional cerebral blood ow lim-ited to the parietotemporal region The glucose extractionfraction and global cerebral metabolic rate of glucosewere signi cantly decreased whereas the global cerebralmetabolic rate of oxygen was only slightly decreasedBecause the cerebral metabolic rates of the various sub-strates (ketones lactate etc) did not change the authorsconcluded that the markedly elevated oxygenglucoseutilization ratio indicated an altered energy metabolismin AD

Conceivably the impairment of glucose metabolismin AD patients described by Ogawa et al is attributableat least in part to local insulin resistance Watson andCraft82 called attention to emerging evidence that insulinabnormalities and insulin resistance in AD may contrib-ute to the pathophysiology and clinical symptoms of thisdisorder In their words ldquoIt has been recently demon-strated that insulin-sensitive glucose transporters are lo-calized in the same regions supporting memory and thatinsulin plays a role in memory functions Collectivelythese ndings suggest that insulin may contribute tonormal cognitive functioning and that insulin abnormal-ities may exacerbate cognitive impairments such asthose associated with ADrdquo The authors suggest thatimproving insulin effectiveness may have therapeuticbene t for patients with AD The suggestion of Watsonand Craft that insulin abnormalities may play a role inthe pathogenesis of AD has to be reconciled with the nding by Molina et al83 that cerebrospinal uid insulinlevels in 27 AD patients did not differ from those in 16matched controls

In 1996 Hoshi et al84 reported that a fragment ofthe beta chain of amyloid inhibits PDH activity in pri-mary cultures of hippocampal neurons A year later thesame authors found that the 1ndash42 fragment of the betachain of amyloid Ab1ndash42 can inhibit acetyl cholinesynthesis of septal neurons85 Veech et al50 note thatcultured hippocampal neurons die when exposed toAb1ndash42 However addition of 4 mM D-b-hydroxybu-tyrate protects these neurons from amyloid fragmentndashinduced death Casley et al86 showed that beta-amyloidcauses a signi cant reduction in state 3 and state 4mitochondrial respiration which is diminished further bynitric oxide These authors also found that cytochromeoxidase -ketoglutarate dehydrogenase and PDH activ-ities are inhibited by b-amyloid They concluded thatb-amyloid can directly disrupt mitochondrial function

inhibit key enzymes and contribute to the impairment ofenergy metabolism observed in AD

Veech et al50 contend that the foregoing studiesprovide a rationale for trial of a ketogenic diet in thetreatment of AD Ketones can overcome localized insulinresistance can bypass a block in PDH and have beenshown to protect dopaminergic neurons from free radicaldamage and from apoptosis induced by the 1ndash42 amy-loid fragment in tissue-culture experiments

What Next

In considering the potential usefulness of therapeutichyperketonemia in the treatment of AD PD and certainother neurodegenerative conditions it is essential to beclear about what is meant by terms like ldquoketogenic dietrdquoand ldquohyperketonemiardquo

There is widespread confusion among both physi-cians and lay individuals about what constitutes a keto-genic diet Similar to glucose ketones are present in theblood at all times As shown in Table 3 the range ofketone levels that can be achieved by prolonged fooddeprivation or by adherence to various diets is quitewide Hence it makes no sense to speak of a ldquoketogenicdietrdquo without also specifying the degree of serum ketoneelevation that the diet is intended to achieve

The serum concentrations sought for treatment ofepilepsy range from 2 to 7 mML Such levels can beachieved only by strict adherence to a regimen like the41 HKDmdashone most patients nd burdensome On theother hand the kind of low-carbohydrate ldquoketogenicdietrdquo (eg 30 protein 8 carbohydrate 61 fat [asproportions of total energy]) that has recently becomepopularmdashalthough much easier to followmdashachievesBHB levels of only 028ndash040 mML87 (This ldquopopularrdquoketogenic diet at a 2000 kcalday intake provides 127 gof presumptive carbohydrate [40 preformed] plus 87 gderivable from 150 g protein not including the smallquantity of glycerol released in the course of adipocytetriglyceride hydrolysis At the same energy level the 41HKD would yield 33 g of presumptive carbohydrate[10 g preformed] plus 23 g derivable from 40 g proteinThus the effective quantity of carbohydrate provided bythe popular ketogenic diet is almost four times thatsupplied by the 41 HKD)

Future research must address the problem of match-ing diet-induced elevation of serum ketone levels withtherapeutic bene ts Because the drastic type of hyper-ketogenic dietmdashthe one used to treat epilepsy and relatedneurologic disordersmdashis so onerous and dif cult to fol-low it is especially important to determine the extent towhich such a diet can be liberalized without losingdesirable therapeutic effects Given the daunting natureof the 41 HKD it is encouraging that there exists apotential nutritional stratagem which eventually may

337Nutrition Reviews Vol 61 No 10

make it possible to induce therapeutic hyperketonemia inpatients without requiring them to adhere to a Draconianregimen The stratagem entails administration by mouthof substantial quantities of ketones in a formmdashnamelyas esters or small polymersmdashsuitable for ingestion Intheir article about the potential therapeutic uses of ketonebodies Veech et al50 suggest that the problems associ-ated with ldquothe very unappetizing high-fat ketogenicdietrdquo might be obviated by feeding patients up to 100 to150 g per day of a nutritionally acceptable form ofketone bodies Unfortunately neither ketone esters noroligomers of BHB are presently available in suf cientquantities for the conduct of clinical trials

The traditional 41 HKD is dif cult but far fromunmanageable Numerous epileptic children and adultswith seizures resistant to anticonvulsant medication havemanaged to remain on this regimen for years withoutexperiencing serious adverse effects There are justi ableconcerns about its hyperlipidemic effect however andthe fact that the diet has been reported to increase risk ofside effects including nephrolithiasis88 depression ofplatelet function (manifested by easy bruising)89 andoccasionally prolongation of the electrocardiographicQTc interval in children90 It should be possible to mitigateor perhaps prevent the adverse effects of this very highndashfatdiet on plasma lipids by modifying its fatty acid contentWhat is most important now however is to get on with thejob of determining whether diet-induced hyperketonemia(at any level) will prove clinically bene cial in the man-agement of a number of neurodegenerative disorders par-ticularly Parkinsonrsquos disease and Alzheimerrsquos disease

Acknowledgment

We thank Richard L Veech MD DPhil for hisencouragement and very helpful suggestions We alsothank Steven W Fowkes for creative and elegant ren-dering of Figures 1ndash3 and editorial advice

1 Gerhardt J Diabetes mellitus und aceton WienerMedizinische Presse 18656672

2 Hirschfeld F Beobachtungen uber die Acentonurieund das Coma diabeticum Z klin Med 189528176 ndash209

3 Peters JP Van Slyke DD Quantitative ClinicalChemistry Interpretations Volume 1 2nd editionBaltimore Williams amp Wilkins 1946439

4 Cohen PP Studies in ketogenesis J Biol Chem1937119333ndash346

5 Cohen PP Stark IE Hepatic ketogenesis and ketol-ysis in different species J Biol Chem 193812697ndash107

6 Jowett M Quastel JH Fat metabolism III The for-mation and breakdown of acetoacetic acid in animaltissues Biochem J 1935292181ndash2191

7 Chaikoff IL Soskin S The utilization of acetoaceticacid by normal and diabetic dogs before and afterevisceration Am J Physiol 19288758 ndash72

8 Barnes RH Drury DR Greeley PO Wick AN Utili-zation of the ketone bodies in normal animals and inthose with ketosis Am J Physiol 1940130144 ndash150

9 Mirsky IA Broh-Kahn RH The inuence of in-creased metabolism on b-hydroxybutyric acid utili-zation Am J Physiol 1937120446 ndash 450

10 Waters ET Fletcher JP Mirsky IA The relationbetween carbohydrate and b-hydroxybutyric acid

Table 3 Relationship of Metabolic State and Diet Composition to Plasma Levels of Acetoacetate (AcAc)and b-Hydroxybutyrate (BHB) in Normal-weight Overweight and Obese Adults

Metabolic StateDiet

Composition CommentsAcAc

(mML)BHB

(mML) References

Postabsorptive Malesn 5 11 Normal weight 005 007 Hall et al18

n 5 12 Normal weight 008 Sharman et al87

n 5 10 Obese 015 027 Hall et al18

n 5 6 Overweight 001 002 Cahill et al39

Ketogenic diets Males (normal weight)(eucaloric) 3 weeks on diet (n 5 12) 040 Sharman et al87

C8 P30 F61 6 weeks on diet 028 Sharman et al87

Female (normal weight)C2 P8 F90 2 weeks on diet (n 5 1) 070 602 VanItallie Nonas amp Heyms elddagger

Fasting Malesn 5 6 (overweight)

1-week fast 095 358 Cahill et al39

n 5 6 (obese)1ndash2-week fast 104 597 Hall et al18

Distribution (percent total calories) of carbohydrate (C) protein (P) and fat (F)daggerUnpublished study (St Lukersquos-Roosevelt Hospital 2003)Where n 1 all values are means

338 Nutrition Reviews Vol 61 No 10

Ketones Substrates in Search of a Mission

By the early 1960s a great deal of information hadaccumulated about the biochemistry and physiology ofthe ketone bodies yet their role in human metabolismremained obscure38 Nevertheless ve facts were knownto medical scientists that had they been viewed in properperspective would have suggested strongly that ketoneshave to be the brainrsquos back-up fuel during starvationFirst scientists recognized that when people are deprivedof food for 3 days or longer they become hyperketone-mic and hyperketonuric Second catheterization studiesshowed that the adult human brain uses approximately100 to 150 g glucose per day39 Third scientists under-stood that non-esteri ed fatty acids do not cross theblood-brain barrier (BBB) and are not directly availableto the brain as an energy source40 Fourth Benedict41

showed that during 31 days of total caloric starvation hisldquofasting manrdquo excreted an average of 872 gday ofnitrogen (N) which corresponds to a rate of body proteincatabolism of approximately 55 gday Approximately32 g of this amount (058 550) could have beenavailable as fuel for the brainmdashapproximately 27 ofthe brainrsquos daily needs In another subject deprived offood for 15 days mean total N excretion was 482 gdaymaking only 174 g of glucoseday potentially availableto the brain42

Finally because glucose was widely assumed to bethe exclusive fuel of the adult human brain38 one mightinfer that during continuing starvation (after liver glyco-gen stores were exhausted) the brain would have to besustained by glucose derived from intracellular proteinMaking 100 to 150 g of glucoseday available to thebrain would require the gluconeogenic breakdown ofsome 172 to 259 gday of body protein At this unsus-tainable rate of protein attrition death might be expectedto occur within approximately 2 weeksmdash certainly notthe 57 to 73 days (mean 61 25 [SEM] days) that itreportedly takes for totally starved young men of averagebody composition to die43 If the nutrition scientists ofthe time had given more thought to the meaning of thisunexplained ldquoprolongationrdquo of life during total starva-tion they might have reasoned that during extended fooddeprivation the brain is able to use the ketones generatedin such abundance as an alternative energy source Sucha scenario would have explained the greater than ex-pected longevity and the far lower than expected dailynitrogen de cit exhibited by starved individuals

In 1966 Cahill et al39 ldquoconnected the dotsrdquo andconcluded that ketones might provide an alternate fuelfor the brain during starvation To test this hypothesisthe authors performed brain catheterization studies inobese patients undergoing 5 to 6 weeks of therapeuticfasting44 This landmark study showed that under condi-tions of sustained starvation (during which the subjects

achieved blood ketone levels close to 7 mML and bloodglucose concentrations of ca 38 mML) BHB and AcAcreplaced glucose as the predominant fuel(s) for brainmetabolism The ldquostarvation paradoxrdquo was thus resolvedIt also became clear that during starvation the liverrsquosability to generate ketones for use by the brain representsa biologic strategy for maintaining strength and prolong-ing life during conditions of severe nutritional privationIn this way the burden of sustaining life during starva-tion is shifted from the bodyrsquos protein stores to itsreserves of fatmdasha far more effective and ef cient adap-tation Thanks to the pioneering studies of Cahill et almetabolismrsquos ugly duckling nally began to emerge asan incipient swan

Ketones Uniquely Efcient Metabolic Fuel

In the mid 1940s Lardy and associates4546 demonstratedthat when tested on sperm BHB and AcAc comparedwith other metabolic substrates have the ability to de-crease oxygen consumption while increasing motilityFor almost half a century this apparent increase in met-abolic ef ciency remained unexplained In 1994 and1995 however RL Veechrsquos group47ndash49 at the NationalInstitutes of Health showed that in the working perfusedrat heart when 5 mM ketone bodiesL were added to theglucose-containing perfusate there was a 25 increasein hydraulic work with a signi cant decrease in oxygenconsumption Addition of ketones also brought about areduction of the mitochrondrial NAD couple with oxi-dation of the mitochondrial co-enzyme Q couple Theconsequence is an increase in the energy of the redoxspan between site I and site II of the electron transportsystem

As Veech et al50 described ldquoAn increase in theredox span between two sites will result in an increasedenergy release by the electron traveling across that span With an increase in redox energy of the respiratorychain there is a corresponding increase in the energy ofthe protons ejected from the mitochondria at the energy-conserving sites which is then re ected in an increase inthe energy of ATP hydrolysisrdquo

As shown in Table 2 Veech and associates47ndash49 alsoobserved that the introduction of ketone bodies into therat heart perfusate produced a 16-fold rise in acetyl-CoAcontent and increases in TCA cycle intermediates Thesestimulatory effects on TCA cycle activity were similar tothose obtained with the administration of saturatingdoses of insulin Insulin increases glucose transport inthe myocardium by promoting translocation of insulin-sensitive GLUT4 to the plasma membrane Insulin alsostimulates the activity of pyruvate dehydrogenase mul-tienzyme complex (PDH)mdasha rate-limiting step in themitochondrial transformation of pyruvate to acetyl-CoAAs Veech et al50 put it ldquoThe implication was that

332 Nutrition Reviews Vol 61 No 10

ketosis which is the physiological response to insulindeprivation during starvation was equivalent in meta-bolic effects to the actions of insulin (Moreover)ketones by-passed the block in glucose transport causedby lack of insulin (and) also by-passed the block inPDH induced by insulin de ciency by providing analternative source of mitochondrial acetyl CoArdquo Thepathway by which ketones are able to achieve this effectis shown in Figure 2

Veech et al50 pointed out that the gradients of Kand Na (and Ca ) between the extra- and intracellularphases of the cell are normally in virtual equilibriumwith the resting electrical potential between these twophases and with the energy ( G) of ATP hydrolysiswhich sustains the action of the Na pump By increas-ing metabolic ef ciency and thereby the energy pro-duced by ATP hydrolysis BHB augments the potentialbetween the extra- and intracellular phases giving rise toan increase in the energy of the K gradient (ratio of[K ]out[K ]in) in muscle and brain A larger mitochon-drial membrane potential would lead to larger energyproduction because it is the mitrochondrial membranepotential difference that drives the production of ATPMaintenance of a suitable K gradient in the cell is ofcourse indispensable to its proper functioning As willbe discussed below impairment of ion homeostasiswithin neuronal cells in certain brain areas might resultin a degree of electrical instability suf cient to trigger anepileptic seizure (or in the heart a cardiac arrhythmia)

Studies of blood ow and oxygen consumption inthe brains of food-deprived obese human subjects re-vealed values (45 mL min 1 100 g 1 and 27 mL O2

min 1 100 mL 1)44 that were well below the normallevels for adult human brains of 57 and 36 respective-ly51 Although these data need con rmation they suggestan increase in the metabolic ef ciency in human brainsusing ketoacids as their principal energy source in placeof glucose

Examples of mechanisms by which diet-inducedhyperketonemia might overcome or circumvent meta-bolic impediments to utilization of such substrates asglucose and pyruvate are shown schematically in Fig-ure 3

Use of a ldquoHyperketogenicrdquo Diet in Treatmentof Seizure Disorders

Fasting has long been acknowledged as a method toprevent or reduce seizures in epileptic individuals37

Because of the obvious limitations of this treatment useof a ketogenic diet to mimic the metabolic effects of fooddeprivation was proposed in 1921 by both Woodyatt52

and Wilder53 at the Mayo Clinic In 1924 Peterman54

(also from Mayo) reported the clinical application andeffectiveness of the regimen suggested by Woodyatt andWilder Petermanrsquos diet provided (per day) 1 g of pro-teinkg body weight (considerably less in adults) 10 to15 g of carbohydrate and the remainder of the calories asfat The Peterman ketogenic diet is sometimes called the41 diet because it provided a ratio of approximately 4parts (by weight) fat to 1 part (by weight) of a mixture ofprotein and carbohydrate This distribution of the majormacronutrients is almost identical to that of most anti-convulsant diets used today37 The term ldquohyperketogenicdietrdquo (HKD) will be used in this review to distinguish the41 diet (and certain variations thereof) from the far lessdrastic low-carbohydrate mildly ketogenic diets thathave been advocated for weight control in a successionof popular diet books

In the 1970s several investigators5556 modi ed thePeterman HKD by substituting medium-chain triglycer-ides (MCT) in varying amounts for long-chain triglycer-ides (MCTrsquos ketogenic effect is not dependent on con-current restriction of dietary carbohydrate57) Althoughall such variations of the HKD were approximately equalin reducing seizures many of the children given MCT-

Table 2 Insulin and Ketone Effects on Cardiac Efciency Acetyl-CoA Production [NADH][NAD ] Ratioand Free Energy of ATP Hydrolysis ( GA TP ) in the Perfused Rat Heart

Substrates Addedto Perfusate

AcetylCoAdagger

[NADH][NAD ]Daggersect

(mitochondrial)

G of ATPHydrolysis(cytosolic)

kJmol

CardiacEf ciency

()

CardiacEf ciency( relativeto glucose)

Glucose 1 005 562 105 100Glucose Insulin 9 022para 589para 134para 128Glucose Ketones 15 062para 576para 130para 124Glucose Insulin

Ketones 18 043para 589para 143para 136

Data of Sato K et al4 8 and DaggerKashiwaya et al4 7 4 9

daggerNumbers represent multiples of that measured with perfusate containing only glucosesectExpressed by Kashiwaya et al4 7 4 9 as [NAD ][NADH]paraIndicates a signi cant difference (P 005) from perfusate containing 10 mmolL glucose ketones added at 4 mMLD-b-hydroxybutyrate 1 mML acetoacetate

333Nutrition Reviews Vol 61 No 10

augmented diets complained of nausea diarrhea andoccasional vomiting58

Among those patients who adhere to the 41 HKDfor 12 months 40 experience greater than 90 reduc-tion in seizures whereas an additional 40 experience50 to 90 reduction Although the precise mechanism ofaction responsible for the clinical ef cacy of the diet hasnever been clari ed it is likely that the hyperketonemiainduced by the 41 HKD is the key factor responsible forthe regimenrsquos anticonvulsant effect rather than the ac-companying metabolic acidosis59 or dehydration37 In

1972 Uhleman and Neims60 demonstrated that micepups fed a high-fat ketogenic diet exhibited signi cantlyincreased thresholds against investigator-applied electro-shock challenges This resistance to electroshock disap-peared promptly after the animals were shifted to ahigh-carbohydrate diet Subsequently a number of stud-ies have shown an increase in the electroconvulsivethreshold in rats maintained on a high-fat diet or de-prived of food for 48 hours De Vivo et al61 found thatthe ratios of ATP to adenosine diphosphate (ADP) weresigni cantly increased in the brains of rats fed the high-

Figure 2 Production of ATP (adenosine triphosphate) in extrahepatic tissues arising from use of ketones and glucose Glucose isregulated by a family of GLUT transporter proteins GLUT4 transporters found in most body tissues are mediated by insulinGLUT1 transporters common in tissues such as brain and erythrocytes allow glucose uptake without insulin mediation By contrastuptake of ketone bodies (shown as BHB and AcAc) occurs via the family of monocarboxylate transporters (MCTs) which are notinsulin mediated MCT proteins enable ketones to pass readily through the blood-brain barrier Many types of peripheral cellsincluding brain cells not only use glucose but also use ketones to produce acetyl-CoA Glucose utilization depends on glycolyticenzymes (not shown) which produce pyruvate Pyruvate in turn enters the mitochondrion and is converted to acetyl-CoA by thepyruvate dehydrogenase multienzyme complex (PDH) Ketone substrates enter the mitochondrion in the form of (1) acetoacetate(AcAc) and (2) b-hydroxybutyrate (BHB) which is readily converted to acetoacetate by b-hydroxybutyrate dehydrogenase (HBD)Two enzymes principally control ketolysis First 3-oxoacid-CoA transferase (OCT) adds coenzyme-A to AcAc which is then splitinto two molecules of acetyl-CoA by acetoacetyl-CoA thiolase (ACT) Acetyl-CoA undergoes oxidative degradation in the TCAcycle to reduce the electron carrriers NAD (nicotinamide adenine dinucleotide) and FAD ( avin adenine dinucleotide) to NADHand FADH2 respectively Acetyl-CoA is also used for lipid or sterol synthesis (not shown) NADH and FADH2 donate electrons tothe protein complexes I and II of the mitochondrial respiratory chain The ow of these electrons down the electron-transport chainto oxygen (O2) allows complexes I III and IV to pump protons (shown as H ) out of the matrix and into the intermembrane spacewhich generates a proton motive force (pmf) between the intermembrane space and the matrix The pmf provides the energy to driveprotons back through complex V (ATP synthase) to produce ATP from ADP (adenosine diphosphate) and inorganic phosphate (Pi)

334 Nutrition Reviews Vol 61 No 10

fat diet They concluded ldquo the anticonvulsant propertyof the ketogenic diet derives from the observed increasein cerebral energy reserve which in turn re ects thehigher ATPADP ratio [this] might provide a reason-able explanation for the increased neuronal stability thatdevelops during chronic ketosisrdquo In 1999 Pan et al62

carried out 31P magnetic resonance spectroscopic imag-ing studies in seven patients with intractable epilepsybefore and after institution of a ketogenic diet Theymeasured signi cant increases (P 005) in high-energy phosphates in association with adherence to theketogenic diet in these subjects They concluded thatenhancement of brain energy metabolism occurs withuse of the ketogenic diet in humans as in lab animals

Thavendiranathan et al63 have recently reportedstudies of the effect of the ldquoclassicrdquo ketogenic diet onseizures triggered by two threshold testsmdashthe pentyle-netetrazol infusion test and the threshold electroconvul-sive shock test These authors observed small but signif-icant threshold elevations (15ndash20) against bothchemically and electrically triggered seizures Similarlyprotective effects were not seen when suprathresholdstimuli were employed

Janigro64 summarized studies indicating that there isa reduction in both glucose uptake and metabolic activityin seizure foci in epileptic patients In his words ldquoDi-minished ion homeostasis together with increased meta-bolic demand of hyperactive neurons may further aggra-vate neuropathological consequences of BBB loss ofglucose uptake mechanismsrdquo Therefore one explanationfor the effectiveness of the 41 HKD in preventingseizures could be the increased transport of BHB at theBBB and its increased availability as an energy source tothe affected neurons BHB is not only an alternate sourceof energy for seizure foci in the brainmdashbypassing animpairment of carbohydrate utilization such as thatwhich can occur from a block in mitochondrial PDHactivitymdashbut it may actually promote ketone and glu-cose uptake in such foci In rats diet-induced ketosisproduces an eightfold increase in the activity of themonocarboxylate transporter and increases the level ofthe glucose transporter GLUT1 in brain endothelialcells and neuropil65 The hypothesis put forth by Janigroto explain how ketosis might eliminate or reduce seizuresis supported by reports that a suitable ketogenic diet isthe treatment of choice for two rare conditions that may

Figure 3 Conditions that can impede acetyl-CoA production from pyruvate and its precursors Examples include (1) insuf cientinsulin (2) insulin receptor abnormalities (3) diminished transport of glucose owing to impaired release of GLUT4 transportersassociated with insulin resistance [see below] (4) pyruvate dehydrogenase complex (PDH) de ciency or dysfunction (5)unavailabilityof glucose at the cell membrane and (6) GLUT1 mutations Provision of ketones to the cell can effectively circumventor overcome such impediments and compensate for the resulting de cits of glycolytically derived acetyl-CoA (GLUT1 designatesthe insulin-independent membrane protein responsible for glucose transport in red blood cells and throughout the blood-brain barrier[BBB] GLUT4 designates the insulin-dependentglucose transporterprotein expressed in most cells Many forms of insulin resistanceentail putative post-receptor defects of insulin-signal transduction resulting in the failure of insulin to release GLUT4 transportersfrom the GLUT4 pool at a rate suf cient to bind and transport enough extracellular glucose to maintain cellular pyruvate levels)

335Nutrition Reviews Vol 61 No 10

be complicated by seizuresmdashthe facilitated glucosetransporter protein type 1 (GLUT1) de ciency syn-drome6667 and the PDH complex de ciency syndrome68

Treatment with a ketogenic diet reportedly brought aboutdramatic clinical improvement in a one-year-old boywith Leigh syndrome which is associated with a de -ciency of PDH69 In this child treatment with a keto-genic diet was associated with striking improvement ofthe cerebral lesions as measured by neuroimaging

In 1973 Buckley and Williamson70 proposed thatneural lipid synthesis preferentially uses AcAc overglucose Patel and Owen71 obtained con rmatory datafor this hypothesis in 1976 In 1984 Koper et al72

compared the rate of incorporation of AcAc and glucoseinto fatty acids and cholesterol using labeled AcAc andglucose At saturating levels of glucose the authorsobserved that the rate at which AcAc label incorporatedinto fatty acids and cholesterol exceeded that of glucoseby a factor of 5 to 10 in both oligodendrocytes andastrocytes These observations provide a basis for argu-ing that some of the bene ts associated with ketogenicdiets may be related to their enhancing effects on lipo-genesis andor sterol synthesis (ie damaged neuronsmight repair more readily in brains given an adequateketone supply)

Therapeutic Potential of Diet-inducedHyperketonemia in Neurodegenerative andOther Disorders

Veech et al50 suggest that maintenance of a suitabledegree of diet-induced hyperketonemia (in the range of2ndash7 mML) has the potential to aid management ofseveral neurologic diseases including common neurode-generative diseases like Alzheimerrsquos disease (AD) andParkinsonrsquos disease (PD) In their ldquohypothesis paperrdquothe authors single out AD PD Friedreichrsquos AtaxiaLeprechaunisms and lesser forms of insulin disorders asexamples of illnesses that might bene t from therapeuticketosis They also describe experience with one case ofLaFora body disease (until now a uniformly fatal geneticdisease associated with progressive dementia) in whichuse of the 41 HKD resulted in unprecedented improve-ment in cognition gait and awareness after one month73

In this disease mutations in a gene encoding a novelprotein tyrosine phosphatase cause progressive myo-clonic epilepsy

Calabrese et al74 and DiMauro and Schon75 haverecently called attention to evidence suggesting the mi-tochondrial genome may play a key role in neurodegen-erative diseases Decreases in respiratory chain complexactivities have been identi ed in AD PD amyotrophiclateral sclerosis and Huntingtonrsquos disease Mitochondri-opathies (MCPs) possibly associated with oxidantanti-oxidant balance perturbation may underlie defects in

energy metabolism and contribute to cellular damage anddestruction

Parkinsonrsquos DiseaseAlthough the etiology of PD is complex involving bothgenetic and environmental factors there is growing in-terest in the role of mitochondrial dysfunction in thepathogenesis of this disorder Schapira76 pointed out thata proportion of PD patients exhibit a de ciency in mito-chondrial respiratory chain function most notably adefect of complex I within neurons of the substantianigra In his words ldquoBoth mitochondrial mutations andtoxic agents (endogenous and exogenous) have beendemonstrated to cause a de ciency in complex I func-tion This is achieved by increasing the degree of orsusceptibility to oxidative stress which may contributeto apoptopic cell deathrdquo

Finsterer77 reported metabolic studies indicating thatPD can be a manifestation of MCP In his series 63 ofPD patients had features of generalized MCP howeverhe used lactate stress testing to detect patients withMCPmdashan approach that could not be expected to pick upmitochondrial lesions limited to brain sites

According to Veech et al ldquoPD appears to resultfrom an acquired defect in the mitochondrial NADHmultienzyme complex of the dopaminergic cells of themesencephalonrdquo78 They point out although the diseaseis treatable for a period by administration of levodopa(the metabolic precursor of dopamine) continuing freeradical damage eventually reduces the effectiveness ofthis therapy Kashiwaya et al79 reported that primarycultures of mesencephalic dopaminergic neurons ex-posed to the meperidine analogue MPP (a molecule thatinhibits NADH dehydrogenase [complex 1] and causesoxygen free radical formation) can be protected fromdeath by addition of 4 mM D-b-hydroxybutyrate Veechet al50 suggest that BHB probably acts by decreasing thesource of mitochondrial oxygen radical formationldquo by oxidizing the coenzyme Q couple while at thesame time reducing the redox potential of the NADPcouple which through glutathione is the nal detoxi -cation step of H2O2rdquo In a recent review Schon andManfredi80 wrote ldquoBecause complexes I and III are theprincipal sources of free radicals in the cell alteredcomplex I function in the substantia nigra pars compactacould be responsible for the increased DNA damage andlipid peroxidation found in PD brainsrdquo

To the extent that ketones crossing the BBB couldcircumvent or compensate for mitochondrial defects as-sociated with PD or could protect dopaminergic neuronsfrom free radical damage a suitable ketogenic diet mightwell have bene cial effects in certain patients sufferingfrom Parkinsonism

336 Nutrition Reviews Vol 61 No 10

Alzheimerrsquos DiseaseIn 1996 Ogawa et al81 determined regional cerebralblood ow and studied energy utilization in the brains ofseven patients with AD by measuring arteriovenousdifferences of an array of substrates Single-photon emis-sion computed tomography showed that the patients hadsigni cantly decreased regional cerebral blood ow lim-ited to the parietotemporal region The glucose extractionfraction and global cerebral metabolic rate of glucosewere signi cantly decreased whereas the global cerebralmetabolic rate of oxygen was only slightly decreasedBecause the cerebral metabolic rates of the various sub-strates (ketones lactate etc) did not change the authorsconcluded that the markedly elevated oxygenglucoseutilization ratio indicated an altered energy metabolismin AD

Conceivably the impairment of glucose metabolismin AD patients described by Ogawa et al is attributableat least in part to local insulin resistance Watson andCraft82 called attention to emerging evidence that insulinabnormalities and insulin resistance in AD may contrib-ute to the pathophysiology and clinical symptoms of thisdisorder In their words ldquoIt has been recently demon-strated that insulin-sensitive glucose transporters are lo-calized in the same regions supporting memory and thatinsulin plays a role in memory functions Collectivelythese ndings suggest that insulin may contribute tonormal cognitive functioning and that insulin abnormal-ities may exacerbate cognitive impairments such asthose associated with ADrdquo The authors suggest thatimproving insulin effectiveness may have therapeuticbene t for patients with AD The suggestion of Watsonand Craft that insulin abnormalities may play a role inthe pathogenesis of AD has to be reconciled with the nding by Molina et al83 that cerebrospinal uid insulinlevels in 27 AD patients did not differ from those in 16matched controls

In 1996 Hoshi et al84 reported that a fragment ofthe beta chain of amyloid inhibits PDH activity in pri-mary cultures of hippocampal neurons A year later thesame authors found that the 1ndash42 fragment of the betachain of amyloid Ab1ndash42 can inhibit acetyl cholinesynthesis of septal neurons85 Veech et al50 note thatcultured hippocampal neurons die when exposed toAb1ndash42 However addition of 4 mM D-b-hydroxybu-tyrate protects these neurons from amyloid fragmentndashinduced death Casley et al86 showed that beta-amyloidcauses a signi cant reduction in state 3 and state 4mitochondrial respiration which is diminished further bynitric oxide These authors also found that cytochromeoxidase -ketoglutarate dehydrogenase and PDH activ-ities are inhibited by b-amyloid They concluded thatb-amyloid can directly disrupt mitochondrial function

inhibit key enzymes and contribute to the impairment ofenergy metabolism observed in AD

Veech et al50 contend that the foregoing studiesprovide a rationale for trial of a ketogenic diet in thetreatment of AD Ketones can overcome localized insulinresistance can bypass a block in PDH and have beenshown to protect dopaminergic neurons from free radicaldamage and from apoptosis induced by the 1ndash42 amy-loid fragment in tissue-culture experiments

What Next

In considering the potential usefulness of therapeutichyperketonemia in the treatment of AD PD and certainother neurodegenerative conditions it is essential to beclear about what is meant by terms like ldquoketogenic dietrdquoand ldquohyperketonemiardquo

There is widespread confusion among both physi-cians and lay individuals about what constitutes a keto-genic diet Similar to glucose ketones are present in theblood at all times As shown in Table 3 the range ofketone levels that can be achieved by prolonged fooddeprivation or by adherence to various diets is quitewide Hence it makes no sense to speak of a ldquoketogenicdietrdquo without also specifying the degree of serum ketoneelevation that the diet is intended to achieve

The serum concentrations sought for treatment ofepilepsy range from 2 to 7 mML Such levels can beachieved only by strict adherence to a regimen like the41 HKDmdashone most patients nd burdensome On theother hand the kind of low-carbohydrate ldquoketogenicdietrdquo (eg 30 protein 8 carbohydrate 61 fat [asproportions of total energy]) that has recently becomepopularmdashalthough much easier to followmdashachievesBHB levels of only 028ndash040 mML87 (This ldquopopularrdquoketogenic diet at a 2000 kcalday intake provides 127 gof presumptive carbohydrate [40 preformed] plus 87 gderivable from 150 g protein not including the smallquantity of glycerol released in the course of adipocytetriglyceride hydrolysis At the same energy level the 41HKD would yield 33 g of presumptive carbohydrate[10 g preformed] plus 23 g derivable from 40 g proteinThus the effective quantity of carbohydrate provided bythe popular ketogenic diet is almost four times thatsupplied by the 41 HKD)

Future research must address the problem of match-ing diet-induced elevation of serum ketone levels withtherapeutic bene ts Because the drastic type of hyper-ketogenic dietmdashthe one used to treat epilepsy and relatedneurologic disordersmdashis so onerous and dif cult to fol-low it is especially important to determine the extent towhich such a diet can be liberalized without losingdesirable therapeutic effects Given the daunting natureof the 41 HKD it is encouraging that there exists apotential nutritional stratagem which eventually may

337Nutrition Reviews Vol 61 No 10

make it possible to induce therapeutic hyperketonemia inpatients without requiring them to adhere to a Draconianregimen The stratagem entails administration by mouthof substantial quantities of ketones in a formmdashnamelyas esters or small polymersmdashsuitable for ingestion Intheir article about the potential therapeutic uses of ketonebodies Veech et al50 suggest that the problems associ-ated with ldquothe very unappetizing high-fat ketogenicdietrdquo might be obviated by feeding patients up to 100 to150 g per day of a nutritionally acceptable form ofketone bodies Unfortunately neither ketone esters noroligomers of BHB are presently available in suf cientquantities for the conduct of clinical trials

The traditional 41 HKD is dif cult but far fromunmanageable Numerous epileptic children and adultswith seizures resistant to anticonvulsant medication havemanaged to remain on this regimen for years withoutexperiencing serious adverse effects There are justi ableconcerns about its hyperlipidemic effect however andthe fact that the diet has been reported to increase risk ofside effects including nephrolithiasis88 depression ofplatelet function (manifested by easy bruising)89 andoccasionally prolongation of the electrocardiographicQTc interval in children90 It should be possible to mitigateor perhaps prevent the adverse effects of this very highndashfatdiet on plasma lipids by modifying its fatty acid contentWhat is most important now however is to get on with thejob of determining whether diet-induced hyperketonemia(at any level) will prove clinically bene cial in the man-agement of a number of neurodegenerative disorders par-ticularly Parkinsonrsquos disease and Alzheimerrsquos disease

Acknowledgment

We thank Richard L Veech MD DPhil for hisencouragement and very helpful suggestions We alsothank Steven W Fowkes for creative and elegant ren-dering of Figures 1ndash3 and editorial advice

1 Gerhardt J Diabetes mellitus und aceton WienerMedizinische Presse 18656672

2 Hirschfeld F Beobachtungen uber die Acentonurieund das Coma diabeticum Z klin Med 189528176 ndash209

3 Peters JP Van Slyke DD Quantitative ClinicalChemistry Interpretations Volume 1 2nd editionBaltimore Williams amp Wilkins 1946439

4 Cohen PP Studies in ketogenesis J Biol Chem1937119333ndash346

5 Cohen PP Stark IE Hepatic ketogenesis and ketol-ysis in different species J Biol Chem 193812697ndash107

6 Jowett M Quastel JH Fat metabolism III The for-mation and breakdown of acetoacetic acid in animaltissues Biochem J 1935292181ndash2191

7 Chaikoff IL Soskin S The utilization of acetoaceticacid by normal and diabetic dogs before and afterevisceration Am J Physiol 19288758 ndash72

8 Barnes RH Drury DR Greeley PO Wick AN Utili-zation of the ketone bodies in normal animals and inthose with ketosis Am J Physiol 1940130144 ndash150

9 Mirsky IA Broh-Kahn RH The inuence of in-creased metabolism on b-hydroxybutyric acid utili-zation Am J Physiol 1937120446 ndash 450

10 Waters ET Fletcher JP Mirsky IA The relationbetween carbohydrate and b-hydroxybutyric acid

Table 3 Relationship of Metabolic State and Diet Composition to Plasma Levels of Acetoacetate (AcAc)and b-Hydroxybutyrate (BHB) in Normal-weight Overweight and Obese Adults

Metabolic StateDiet

Composition CommentsAcAc

(mML)BHB

(mML) References

Postabsorptive Malesn 5 11 Normal weight 005 007 Hall et al18

n 5 12 Normal weight 008 Sharman et al87

n 5 10 Obese 015 027 Hall et al18

n 5 6 Overweight 001 002 Cahill et al39

Ketogenic diets Males (normal weight)(eucaloric) 3 weeks on diet (n 5 12) 040 Sharman et al87

C8 P30 F61 6 weeks on diet 028 Sharman et al87

Female (normal weight)C2 P8 F90 2 weeks on diet (n 5 1) 070 602 VanItallie Nonas amp Heyms elddagger

Fasting Malesn 5 6 (overweight)

1-week fast 095 358 Cahill et al39

n 5 6 (obese)1ndash2-week fast 104 597 Hall et al18

Distribution (percent total calories) of carbohydrate (C) protein (P) and fat (F)daggerUnpublished study (St Lukersquos-Roosevelt Hospital 2003)Where n 1 all values are means

338 Nutrition Reviews Vol 61 No 10

ketosis which is the physiological response to insulindeprivation during starvation was equivalent in meta-bolic effects to the actions of insulin (Moreover)ketones by-passed the block in glucose transport causedby lack of insulin (and) also by-passed the block inPDH induced by insulin de ciency by providing analternative source of mitochondrial acetyl CoArdquo Thepathway by which ketones are able to achieve this effectis shown in Figure 2

Veech et al50 pointed out that the gradients of Kand Na (and Ca ) between the extra- and intracellularphases of the cell are normally in virtual equilibriumwith the resting electrical potential between these twophases and with the energy ( G) of ATP hydrolysiswhich sustains the action of the Na pump By increas-ing metabolic ef ciency and thereby the energy pro-duced by ATP hydrolysis BHB augments the potentialbetween the extra- and intracellular phases giving rise toan increase in the energy of the K gradient (ratio of[K ]out[K ]in) in muscle and brain A larger mitochon-drial membrane potential would lead to larger energyproduction because it is the mitrochondrial membranepotential difference that drives the production of ATPMaintenance of a suitable K gradient in the cell is ofcourse indispensable to its proper functioning As willbe discussed below impairment of ion homeostasiswithin neuronal cells in certain brain areas might resultin a degree of electrical instability suf cient to trigger anepileptic seizure (or in the heart a cardiac arrhythmia)

Studies of blood ow and oxygen consumption inthe brains of food-deprived obese human subjects re-vealed values (45 mL min 1 100 g 1 and 27 mL O2

min 1 100 mL 1)44 that were well below the normallevels for adult human brains of 57 and 36 respective-ly51 Although these data need con rmation they suggestan increase in the metabolic ef ciency in human brainsusing ketoacids as their principal energy source in placeof glucose

Examples of mechanisms by which diet-inducedhyperketonemia might overcome or circumvent meta-bolic impediments to utilization of such substrates asglucose and pyruvate are shown schematically in Fig-ure 3

Use of a ldquoHyperketogenicrdquo Diet in Treatmentof Seizure Disorders

Fasting has long been acknowledged as a method toprevent or reduce seizures in epileptic individuals37

Because of the obvious limitations of this treatment useof a ketogenic diet to mimic the metabolic effects of fooddeprivation was proposed in 1921 by both Woodyatt52

and Wilder53 at the Mayo Clinic In 1924 Peterman54

(also from Mayo) reported the clinical application andeffectiveness of the regimen suggested by Woodyatt andWilder Petermanrsquos diet provided (per day) 1 g of pro-teinkg body weight (considerably less in adults) 10 to15 g of carbohydrate and the remainder of the calories asfat The Peterman ketogenic diet is sometimes called the41 diet because it provided a ratio of approximately 4parts (by weight) fat to 1 part (by weight) of a mixture ofprotein and carbohydrate This distribution of the majormacronutrients is almost identical to that of most anti-convulsant diets used today37 The term ldquohyperketogenicdietrdquo (HKD) will be used in this review to distinguish the41 diet (and certain variations thereof) from the far lessdrastic low-carbohydrate mildly ketogenic diets thathave been advocated for weight control in a successionof popular diet books

In the 1970s several investigators5556 modi ed thePeterman HKD by substituting medium-chain triglycer-ides (MCT) in varying amounts for long-chain triglycer-ides (MCTrsquos ketogenic effect is not dependent on con-current restriction of dietary carbohydrate57) Althoughall such variations of the HKD were approximately equalin reducing seizures many of the children given MCT-

Table 2 Insulin and Ketone Effects on Cardiac Efciency Acetyl-CoA Production [NADH][NAD ] Ratioand Free Energy of ATP Hydrolysis ( GA TP ) in the Perfused Rat Heart

Substrates Addedto Perfusate

AcetylCoAdagger

[NADH][NAD ]Daggersect

(mitochondrial)

G of ATPHydrolysis(cytosolic)

kJmol

CardiacEf ciency

()

CardiacEf ciency( relativeto glucose)

Glucose 1 005 562 105 100Glucose Insulin 9 022para 589para 134para 128Glucose Ketones 15 062para 576para 130para 124Glucose Insulin

Ketones 18 043para 589para 143para 136

Data of Sato K et al4 8 and DaggerKashiwaya et al4 7 4 9

daggerNumbers represent multiples of that measured with perfusate containing only glucosesectExpressed by Kashiwaya et al4 7 4 9 as [NAD ][NADH]paraIndicates a signi cant difference (P 005) from perfusate containing 10 mmolL glucose ketones added at 4 mMLD-b-hydroxybutyrate 1 mML acetoacetate

333Nutrition Reviews Vol 61 No 10

augmented diets complained of nausea diarrhea andoccasional vomiting58

Among those patients who adhere to the 41 HKDfor 12 months 40 experience greater than 90 reduc-tion in seizures whereas an additional 40 experience50 to 90 reduction Although the precise mechanism ofaction responsible for the clinical ef cacy of the diet hasnever been clari ed it is likely that the hyperketonemiainduced by the 41 HKD is the key factor responsible forthe regimenrsquos anticonvulsant effect rather than the ac-companying metabolic acidosis59 or dehydration37 In

1972 Uhleman and Neims60 demonstrated that micepups fed a high-fat ketogenic diet exhibited signi cantlyincreased thresholds against investigator-applied electro-shock challenges This resistance to electroshock disap-peared promptly after the animals were shifted to ahigh-carbohydrate diet Subsequently a number of stud-ies have shown an increase in the electroconvulsivethreshold in rats maintained on a high-fat diet or de-prived of food for 48 hours De Vivo et al61 found thatthe ratios of ATP to adenosine diphosphate (ADP) weresigni cantly increased in the brains of rats fed the high-

Figure 2 Production of ATP (adenosine triphosphate) in extrahepatic tissues arising from use of ketones and glucose Glucose isregulated by a family of GLUT transporter proteins GLUT4 transporters found in most body tissues are mediated by insulinGLUT1 transporters common in tissues such as brain and erythrocytes allow glucose uptake without insulin mediation By contrastuptake of ketone bodies (shown as BHB and AcAc) occurs via the family of monocarboxylate transporters (MCTs) which are notinsulin mediated MCT proteins enable ketones to pass readily through the blood-brain barrier Many types of peripheral cellsincluding brain cells not only use glucose but also use ketones to produce acetyl-CoA Glucose utilization depends on glycolyticenzymes (not shown) which produce pyruvate Pyruvate in turn enters the mitochondrion and is converted to acetyl-CoA by thepyruvate dehydrogenase multienzyme complex (PDH) Ketone substrates enter the mitochondrion in the form of (1) acetoacetate(AcAc) and (2) b-hydroxybutyrate (BHB) which is readily converted to acetoacetate by b-hydroxybutyrate dehydrogenase (HBD)Two enzymes principally control ketolysis First 3-oxoacid-CoA transferase (OCT) adds coenzyme-A to AcAc which is then splitinto two molecules of acetyl-CoA by acetoacetyl-CoA thiolase (ACT) Acetyl-CoA undergoes oxidative degradation in the TCAcycle to reduce the electron carrriers NAD (nicotinamide adenine dinucleotide) and FAD ( avin adenine dinucleotide) to NADHand FADH2 respectively Acetyl-CoA is also used for lipid or sterol synthesis (not shown) NADH and FADH2 donate electrons tothe protein complexes I and II of the mitochondrial respiratory chain The ow of these electrons down the electron-transport chainto oxygen (O2) allows complexes I III and IV to pump protons (shown as H ) out of the matrix and into the intermembrane spacewhich generates a proton motive force (pmf) between the intermembrane space and the matrix The pmf provides the energy to driveprotons back through complex V (ATP synthase) to produce ATP from ADP (adenosine diphosphate) and inorganic phosphate (Pi)

334 Nutrition Reviews Vol 61 No 10

fat diet They concluded ldquo the anticonvulsant propertyof the ketogenic diet derives from the observed increasein cerebral energy reserve which in turn re ects thehigher ATPADP ratio [this] might provide a reason-able explanation for the increased neuronal stability thatdevelops during chronic ketosisrdquo In 1999 Pan et al62

carried out 31P magnetic resonance spectroscopic imag-ing studies in seven patients with intractable epilepsybefore and after institution of a ketogenic diet Theymeasured signi cant increases (P 005) in high-energy phosphates in association with adherence to theketogenic diet in these subjects They concluded thatenhancement of brain energy metabolism occurs withuse of the ketogenic diet in humans as in lab animals

Thavendiranathan et al63 have recently reportedstudies of the effect of the ldquoclassicrdquo ketogenic diet onseizures triggered by two threshold testsmdashthe pentyle-netetrazol infusion test and the threshold electroconvul-sive shock test These authors observed small but signif-icant threshold elevations (15ndash20) against bothchemically and electrically triggered seizures Similarlyprotective effects were not seen when suprathresholdstimuli were employed

Janigro64 summarized studies indicating that there isa reduction in both glucose uptake and metabolic activityin seizure foci in epileptic patients In his words ldquoDi-minished ion homeostasis together with increased meta-bolic demand of hyperactive neurons may further aggra-vate neuropathological consequences of BBB loss ofglucose uptake mechanismsrdquo Therefore one explanationfor the effectiveness of the 41 HKD in preventingseizures could be the increased transport of BHB at theBBB and its increased availability as an energy source tothe affected neurons BHB is not only an alternate sourceof energy for seizure foci in the brainmdashbypassing animpairment of carbohydrate utilization such as thatwhich can occur from a block in mitochondrial PDHactivitymdashbut it may actually promote ketone and glu-cose uptake in such foci In rats diet-induced ketosisproduces an eightfold increase in the activity of themonocarboxylate transporter and increases the level ofthe glucose transporter GLUT1 in brain endothelialcells and neuropil65 The hypothesis put forth by Janigroto explain how ketosis might eliminate or reduce seizuresis supported by reports that a suitable ketogenic diet isthe treatment of choice for two rare conditions that may

Figure 3 Conditions that can impede acetyl-CoA production from pyruvate and its precursors Examples include (1) insuf cientinsulin (2) insulin receptor abnormalities (3) diminished transport of glucose owing to impaired release of GLUT4 transportersassociated with insulin resistance [see below] (4) pyruvate dehydrogenase complex (PDH) de ciency or dysfunction (5)unavailabilityof glucose at the cell membrane and (6) GLUT1 mutations Provision of ketones to the cell can effectively circumventor overcome such impediments and compensate for the resulting de cits of glycolytically derived acetyl-CoA (GLUT1 designatesthe insulin-independent membrane protein responsible for glucose transport in red blood cells and throughout the blood-brain barrier[BBB] GLUT4 designates the insulin-dependentglucose transporterprotein expressed in most cells Many forms of insulin resistanceentail putative post-receptor defects of insulin-signal transduction resulting in the failure of insulin to release GLUT4 transportersfrom the GLUT4 pool at a rate suf cient to bind and transport enough extracellular glucose to maintain cellular pyruvate levels)

335Nutrition Reviews Vol 61 No 10

be complicated by seizuresmdashthe facilitated glucosetransporter protein type 1 (GLUT1) de ciency syn-drome6667 and the PDH complex de ciency syndrome68

Treatment with a ketogenic diet reportedly brought aboutdramatic clinical improvement in a one-year-old boywith Leigh syndrome which is associated with a de -ciency of PDH69 In this child treatment with a keto-genic diet was associated with striking improvement ofthe cerebral lesions as measured by neuroimaging

In 1973 Buckley and Williamson70 proposed thatneural lipid synthesis preferentially uses AcAc overglucose Patel and Owen71 obtained con rmatory datafor this hypothesis in 1976 In 1984 Koper et al72

compared the rate of incorporation of AcAc and glucoseinto fatty acids and cholesterol using labeled AcAc andglucose At saturating levels of glucose the authorsobserved that the rate at which AcAc label incorporatedinto fatty acids and cholesterol exceeded that of glucoseby a factor of 5 to 10 in both oligodendrocytes andastrocytes These observations provide a basis for argu-ing that some of the bene ts associated with ketogenicdiets may be related to their enhancing effects on lipo-genesis andor sterol synthesis (ie damaged neuronsmight repair more readily in brains given an adequateketone supply)

Therapeutic Potential of Diet-inducedHyperketonemia in Neurodegenerative andOther Disorders

Veech et al50 suggest that maintenance of a suitabledegree of diet-induced hyperketonemia (in the range of2ndash7 mML) has the potential to aid management ofseveral neurologic diseases including common neurode-generative diseases like Alzheimerrsquos disease (AD) andParkinsonrsquos disease (PD) In their ldquohypothesis paperrdquothe authors single out AD PD Friedreichrsquos AtaxiaLeprechaunisms and lesser forms of insulin disorders asexamples of illnesses that might bene t from therapeuticketosis They also describe experience with one case ofLaFora body disease (until now a uniformly fatal geneticdisease associated with progressive dementia) in whichuse of the 41 HKD resulted in unprecedented improve-ment in cognition gait and awareness after one month73

In this disease mutations in a gene encoding a novelprotein tyrosine phosphatase cause progressive myo-clonic epilepsy

Calabrese et al74 and DiMauro and Schon75 haverecently called attention to evidence suggesting the mi-tochondrial genome may play a key role in neurodegen-erative diseases Decreases in respiratory chain complexactivities have been identi ed in AD PD amyotrophiclateral sclerosis and Huntingtonrsquos disease Mitochondri-opathies (MCPs) possibly associated with oxidantanti-oxidant balance perturbation may underlie defects in

energy metabolism and contribute to cellular damage anddestruction

Parkinsonrsquos DiseaseAlthough the etiology of PD is complex involving bothgenetic and environmental factors there is growing in-terest in the role of mitochondrial dysfunction in thepathogenesis of this disorder Schapira76 pointed out thata proportion of PD patients exhibit a de ciency in mito-chondrial respiratory chain function most notably adefect of complex I within neurons of the substantianigra In his words ldquoBoth mitochondrial mutations andtoxic agents (endogenous and exogenous) have beendemonstrated to cause a de ciency in complex I func-tion This is achieved by increasing the degree of orsusceptibility to oxidative stress which may contributeto apoptopic cell deathrdquo

Finsterer77 reported metabolic studies indicating thatPD can be a manifestation of MCP In his series 63 ofPD patients had features of generalized MCP howeverhe used lactate stress testing to detect patients withMCPmdashan approach that could not be expected to pick upmitochondrial lesions limited to brain sites

According to Veech et al ldquoPD appears to resultfrom an acquired defect in the mitochondrial NADHmultienzyme complex of the dopaminergic cells of themesencephalonrdquo78 They point out although the diseaseis treatable for a period by administration of levodopa(the metabolic precursor of dopamine) continuing freeradical damage eventually reduces the effectiveness ofthis therapy Kashiwaya et al79 reported that primarycultures of mesencephalic dopaminergic neurons ex-posed to the meperidine analogue MPP (a molecule thatinhibits NADH dehydrogenase [complex 1] and causesoxygen free radical formation) can be protected fromdeath by addition of 4 mM D-b-hydroxybutyrate Veechet al50 suggest that BHB probably acts by decreasing thesource of mitochondrial oxygen radical formationldquo by oxidizing the coenzyme Q couple while at thesame time reducing the redox potential of the NADPcouple which through glutathione is the nal detoxi -cation step of H2O2rdquo In a recent review Schon andManfredi80 wrote ldquoBecause complexes I and III are theprincipal sources of free radicals in the cell alteredcomplex I function in the substantia nigra pars compactacould be responsible for the increased DNA damage andlipid peroxidation found in PD brainsrdquo

To the extent that ketones crossing the BBB couldcircumvent or compensate for mitochondrial defects as-sociated with PD or could protect dopaminergic neuronsfrom free radical damage a suitable ketogenic diet mightwell have bene cial effects in certain patients sufferingfrom Parkinsonism

336 Nutrition Reviews Vol 61 No 10

Alzheimerrsquos DiseaseIn 1996 Ogawa et al81 determined regional cerebralblood ow and studied energy utilization in the brains ofseven patients with AD by measuring arteriovenousdifferences of an array of substrates Single-photon emis-sion computed tomography showed that the patients hadsigni cantly decreased regional cerebral blood ow lim-ited to the parietotemporal region The glucose extractionfraction and global cerebral metabolic rate of glucosewere signi cantly decreased whereas the global cerebralmetabolic rate of oxygen was only slightly decreasedBecause the cerebral metabolic rates of the various sub-strates (ketones lactate etc) did not change the authorsconcluded that the markedly elevated oxygenglucoseutilization ratio indicated an altered energy metabolismin AD

Conceivably the impairment of glucose metabolismin AD patients described by Ogawa et al is attributableat least in part to local insulin resistance Watson andCraft82 called attention to emerging evidence that insulinabnormalities and insulin resistance in AD may contrib-ute to the pathophysiology and clinical symptoms of thisdisorder In their words ldquoIt has been recently demon-strated that insulin-sensitive glucose transporters are lo-calized in the same regions supporting memory and thatinsulin plays a role in memory functions Collectivelythese ndings suggest that insulin may contribute tonormal cognitive functioning and that insulin abnormal-ities may exacerbate cognitive impairments such asthose associated with ADrdquo The authors suggest thatimproving insulin effectiveness may have therapeuticbene t for patients with AD The suggestion of Watsonand Craft that insulin abnormalities may play a role inthe pathogenesis of AD has to be reconciled with the nding by Molina et al83 that cerebrospinal uid insulinlevels in 27 AD patients did not differ from those in 16matched controls

In 1996 Hoshi et al84 reported that a fragment ofthe beta chain of amyloid inhibits PDH activity in pri-mary cultures of hippocampal neurons A year later thesame authors found that the 1ndash42 fragment of the betachain of amyloid Ab1ndash42 can inhibit acetyl cholinesynthesis of septal neurons85 Veech et al50 note thatcultured hippocampal neurons die when exposed toAb1ndash42 However addition of 4 mM D-b-hydroxybu-tyrate protects these neurons from amyloid fragmentndashinduced death Casley et al86 showed that beta-amyloidcauses a signi cant reduction in state 3 and state 4mitochondrial respiration which is diminished further bynitric oxide These authors also found that cytochromeoxidase -ketoglutarate dehydrogenase and PDH activ-ities are inhibited by b-amyloid They concluded thatb-amyloid can directly disrupt mitochondrial function

inhibit key enzymes and contribute to the impairment ofenergy metabolism observed in AD

Veech et al50 contend that the foregoing studiesprovide a rationale for trial of a ketogenic diet in thetreatment of AD Ketones can overcome localized insulinresistance can bypass a block in PDH and have beenshown to protect dopaminergic neurons from free radicaldamage and from apoptosis induced by the 1ndash42 amy-loid fragment in tissue-culture experiments

What Next

In considering the potential usefulness of therapeutichyperketonemia in the treatment of AD PD and certainother neurodegenerative conditions it is essential to beclear about what is meant by terms like ldquoketogenic dietrdquoand ldquohyperketonemiardquo

There is widespread confusion among both physi-cians and lay individuals about what constitutes a keto-genic diet Similar to glucose ketones are present in theblood at all times As shown in Table 3 the range ofketone levels that can be achieved by prolonged fooddeprivation or by adherence to various diets is quitewide Hence it makes no sense to speak of a ldquoketogenicdietrdquo without also specifying the degree of serum ketoneelevation that the diet is intended to achieve

The serum concentrations sought for treatment ofepilepsy range from 2 to 7 mML Such levels can beachieved only by strict adherence to a regimen like the41 HKDmdashone most patients nd burdensome On theother hand the kind of low-carbohydrate ldquoketogenicdietrdquo (eg 30 protein 8 carbohydrate 61 fat [asproportions of total energy]) that has recently becomepopularmdashalthough much easier to followmdashachievesBHB levels of only 028ndash040 mML87 (This ldquopopularrdquoketogenic diet at a 2000 kcalday intake provides 127 gof presumptive carbohydrate [40 preformed] plus 87 gderivable from 150 g protein not including the smallquantity of glycerol released in the course of adipocytetriglyceride hydrolysis At the same energy level the 41HKD would yield 33 g of presumptive carbohydrate[10 g preformed] plus 23 g derivable from 40 g proteinThus the effective quantity of carbohydrate provided bythe popular ketogenic diet is almost four times thatsupplied by the 41 HKD)

Future research must address the problem of match-ing diet-induced elevation of serum ketone levels withtherapeutic bene ts Because the drastic type of hyper-ketogenic dietmdashthe one used to treat epilepsy and relatedneurologic disordersmdashis so onerous and dif cult to fol-low it is especially important to determine the extent towhich such a diet can be liberalized without losingdesirable therapeutic effects Given the daunting natureof the 41 HKD it is encouraging that there exists apotential nutritional stratagem which eventually may

337Nutrition Reviews Vol 61 No 10

make it possible to induce therapeutic hyperketonemia inpatients without requiring them to adhere to a Draconianregimen The stratagem entails administration by mouthof substantial quantities of ketones in a formmdashnamelyas esters or small polymersmdashsuitable for ingestion Intheir article about the potential therapeutic uses of ketonebodies Veech et al50 suggest that the problems associ-ated with ldquothe very unappetizing high-fat ketogenicdietrdquo might be obviated by feeding patients up to 100 to150 g per day of a nutritionally acceptable form ofketone bodies Unfortunately neither ketone esters noroligomers of BHB are presently available in suf cientquantities for the conduct of clinical trials

The traditional 41 HKD is dif cult but far fromunmanageable Numerous epileptic children and adultswith seizures resistant to anticonvulsant medication havemanaged to remain on this regimen for years withoutexperiencing serious adverse effects There are justi ableconcerns about its hyperlipidemic effect however andthe fact that the diet has been reported to increase risk ofside effects including nephrolithiasis88 depression ofplatelet function (manifested by easy bruising)89 andoccasionally prolongation of the electrocardiographicQTc interval in children90 It should be possible to mitigateor perhaps prevent the adverse effects of this very highndashfatdiet on plasma lipids by modifying its fatty acid contentWhat is most important now however is to get on with thejob of determining whether diet-induced hyperketonemia(at any level) will prove clinically bene cial in the man-agement of a number of neurodegenerative disorders par-ticularly Parkinsonrsquos disease and Alzheimerrsquos disease

Acknowledgment

We thank Richard L Veech MD DPhil for hisencouragement and very helpful suggestions We alsothank Steven W Fowkes for creative and elegant ren-dering of Figures 1ndash3 and editorial advice

1 Gerhardt J Diabetes mellitus und aceton WienerMedizinische Presse 18656672

2 Hirschfeld F Beobachtungen uber die Acentonurieund das Coma diabeticum Z klin Med 189528176 ndash209

3 Peters JP Van Slyke DD Quantitative ClinicalChemistry Interpretations Volume 1 2nd editionBaltimore Williams amp Wilkins 1946439

4 Cohen PP Studies in ketogenesis J Biol Chem1937119333ndash346

5 Cohen PP Stark IE Hepatic ketogenesis and ketol-ysis in different species J Biol Chem 193812697ndash107

6 Jowett M Quastel JH Fat metabolism III The for-mation and breakdown of acetoacetic acid in animaltissues Biochem J 1935292181ndash2191

7 Chaikoff IL Soskin S The utilization of acetoaceticacid by normal and diabetic dogs before and afterevisceration Am J Physiol 19288758 ndash72

8 Barnes RH Drury DR Greeley PO Wick AN Utili-zation of the ketone bodies in normal animals and inthose with ketosis Am J Physiol 1940130144 ndash150

9 Mirsky IA Broh-Kahn RH The inuence of in-creased metabolism on b-hydroxybutyric acid utili-zation Am J Physiol 1937120446 ndash 450

10 Waters ET Fletcher JP Mirsky IA The relationbetween carbohydrate and b-hydroxybutyric acid

Table 3 Relationship of Metabolic State and Diet Composition to Plasma Levels of Acetoacetate (AcAc)and b-Hydroxybutyrate (BHB) in Normal-weight Overweight and Obese Adults

Metabolic StateDiet

Composition CommentsAcAc

(mML)BHB

(mML) References

Postabsorptive Malesn 5 11 Normal weight 005 007 Hall et al18

n 5 12 Normal weight 008 Sharman et al87

n 5 10 Obese 015 027 Hall et al18

n 5 6 Overweight 001 002 Cahill et al39

Ketogenic diets Males (normal weight)(eucaloric) 3 weeks on diet (n 5 12) 040 Sharman et al87

C8 P30 F61 6 weeks on diet 028 Sharman et al87

Female (normal weight)C2 P8 F90 2 weeks on diet (n 5 1) 070 602 VanItallie Nonas amp Heyms elddagger

Fasting Malesn 5 6 (overweight)

1-week fast 095 358 Cahill et al39

n 5 6 (obese)1ndash2-week fast 104 597 Hall et al18

Distribution (percent total calories) of carbohydrate (C) protein (P) and fat (F)daggerUnpublished study (St Lukersquos-Roosevelt Hospital 2003)Where n 1 all values are means

338 Nutrition Reviews Vol 61 No 10

augmented diets complained of nausea diarrhea andoccasional vomiting58

Among those patients who adhere to the 41 HKDfor 12 months 40 experience greater than 90 reduc-tion in seizures whereas an additional 40 experience50 to 90 reduction Although the precise mechanism ofaction responsible for the clinical ef cacy of the diet hasnever been clari ed it is likely that the hyperketonemiainduced by the 41 HKD is the key factor responsible forthe regimenrsquos anticonvulsant effect rather than the ac-companying metabolic acidosis59 or dehydration37 In

1972 Uhleman and Neims60 demonstrated that micepups fed a high-fat ketogenic diet exhibited signi cantlyincreased thresholds against investigator-applied electro-shock challenges This resistance to electroshock disap-peared promptly after the animals were shifted to ahigh-carbohydrate diet Subsequently a number of stud-ies have shown an increase in the electroconvulsivethreshold in rats maintained on a high-fat diet or de-prived of food for 48 hours De Vivo et al61 found thatthe ratios of ATP to adenosine diphosphate (ADP) weresigni cantly increased in the brains of rats fed the high-

Figure 2 Production of ATP (adenosine triphosphate) in extrahepatic tissues arising from use of ketones and glucose Glucose isregulated by a family of GLUT transporter proteins GLUT4 transporters found in most body tissues are mediated by insulinGLUT1 transporters common in tissues such as brain and erythrocytes allow glucose uptake without insulin mediation By contrastuptake of ketone bodies (shown as BHB and AcAc) occurs via the family of monocarboxylate transporters (MCTs) which are notinsulin mediated MCT proteins enable ketones to pass readily through the blood-brain barrier Many types of peripheral cellsincluding brain cells not only use glucose but also use ketones to produce acetyl-CoA Glucose utilization depends on glycolyticenzymes (not shown) which produce pyruvate Pyruvate in turn enters the mitochondrion and is converted to acetyl-CoA by thepyruvate dehydrogenase multienzyme complex (PDH) Ketone substrates enter the mitochondrion in the form of (1) acetoacetate(AcAc) and (2) b-hydroxybutyrate (BHB) which is readily converted to acetoacetate by b-hydroxybutyrate dehydrogenase (HBD)Two enzymes principally control ketolysis First 3-oxoacid-CoA transferase (OCT) adds coenzyme-A to AcAc which is then splitinto two molecules of acetyl-CoA by acetoacetyl-CoA thiolase (ACT) Acetyl-CoA undergoes oxidative degradation in the TCAcycle to reduce the electron carrriers NAD (nicotinamide adenine dinucleotide) and FAD ( avin adenine dinucleotide) to NADHand FADH2 respectively Acetyl-CoA is also used for lipid or sterol synthesis (not shown) NADH and FADH2 donate electrons tothe protein complexes I and II of the mitochondrial respiratory chain The ow of these electrons down the electron-transport chainto oxygen (O2) allows complexes I III and IV to pump protons (shown as H ) out of the matrix and into the intermembrane spacewhich generates a proton motive force (pmf) between the intermembrane space and the matrix The pmf provides the energy to driveprotons back through complex V (ATP synthase) to produce ATP from ADP (adenosine diphosphate) and inorganic phosphate (Pi)

334 Nutrition Reviews Vol 61 No 10

fat diet They concluded ldquo the anticonvulsant propertyof the ketogenic diet derives from the observed increasein cerebral energy reserve which in turn re ects thehigher ATPADP ratio [this] might provide a reason-able explanation for the increased neuronal stability thatdevelops during chronic ketosisrdquo In 1999 Pan et al62

carried out 31P magnetic resonance spectroscopic imag-ing studies in seven patients with intractable epilepsybefore and after institution of a ketogenic diet Theymeasured signi cant increases (P 005) in high-energy phosphates in association with adherence to theketogenic diet in these subjects They concluded thatenhancement of brain energy metabolism occurs withuse of the ketogenic diet in humans as in lab animals

Thavendiranathan et al63 have recently reportedstudies of the effect of the ldquoclassicrdquo ketogenic diet onseizures triggered by two threshold testsmdashthe pentyle-netetrazol infusion test and the threshold electroconvul-sive shock test These authors observed small but signif-icant threshold elevations (15ndash20) against bothchemically and electrically triggered seizures Similarlyprotective effects were not seen when suprathresholdstimuli were employed

Janigro64 summarized studies indicating that there isa reduction in both glucose uptake and metabolic activityin seizure foci in epileptic patients In his words ldquoDi-minished ion homeostasis together with increased meta-bolic demand of hyperactive neurons may further aggra-vate neuropathological consequences of BBB loss ofglucose uptake mechanismsrdquo Therefore one explanationfor the effectiveness of the 41 HKD in preventingseizures could be the increased transport of BHB at theBBB and its increased availability as an energy source tothe affected neurons BHB is not only an alternate sourceof energy for seizure foci in the brainmdashbypassing animpairment of carbohydrate utilization such as thatwhich can occur from a block in mitochondrial PDHactivitymdashbut it may actually promote ketone and glu-cose uptake in such foci In rats diet-induced ketosisproduces an eightfold increase in the activity of themonocarboxylate transporter and increases the level ofthe glucose transporter GLUT1 in brain endothelialcells and neuropil65 The hypothesis put forth by Janigroto explain how ketosis might eliminate or reduce seizuresis supported by reports that a suitable ketogenic diet isthe treatment of choice for two rare conditions that may

Figure 3 Conditions that can impede acetyl-CoA production from pyruvate and its precursors Examples include (1) insuf cientinsulin (2) insulin receptor abnormalities (3) diminished transport of glucose owing to impaired release of GLUT4 transportersassociated with insulin resistance [see below] (4) pyruvate dehydrogenase complex (PDH) de ciency or dysfunction (5)unavailabilityof glucose at the cell membrane and (6) GLUT1 mutations Provision of ketones to the cell can effectively circumventor overcome such impediments and compensate for the resulting de cits of glycolytically derived acetyl-CoA (GLUT1 designatesthe insulin-independent membrane protein responsible for glucose transport in red blood cells and throughout the blood-brain barrier[BBB] GLUT4 designates the insulin-dependentglucose transporterprotein expressed in most cells Many forms of insulin resistanceentail putative post-receptor defects of insulin-signal transduction resulting in the failure of insulin to release GLUT4 transportersfrom the GLUT4 pool at a rate suf cient to bind and transport enough extracellular glucose to maintain cellular pyruvate levels)

335Nutrition Reviews Vol 61 No 10

be complicated by seizuresmdashthe facilitated glucosetransporter protein type 1 (GLUT1) de ciency syn-drome6667 and the PDH complex de ciency syndrome68

Treatment with a ketogenic diet reportedly brought aboutdramatic clinical improvement in a one-year-old boywith Leigh syndrome which is associated with a de -ciency of PDH69 In this child treatment with a keto-genic diet was associated with striking improvement ofthe cerebral lesions as measured by neuroimaging

In 1973 Buckley and Williamson70 proposed thatneural lipid synthesis preferentially uses AcAc overglucose Patel and Owen71 obtained con rmatory datafor this hypothesis in 1976 In 1984 Koper et al72

compared the rate of incorporation of AcAc and glucoseinto fatty acids and cholesterol using labeled AcAc andglucose At saturating levels of glucose the authorsobserved that the rate at which AcAc label incorporatedinto fatty acids and cholesterol exceeded that of glucoseby a factor of 5 to 10 in both oligodendrocytes andastrocytes These observations provide a basis for argu-ing that some of the bene ts associated with ketogenicdiets may be related to their enhancing effects on lipo-genesis andor sterol synthesis (ie damaged neuronsmight repair more readily in brains given an adequateketone supply)

Therapeutic Potential of Diet-inducedHyperketonemia in Neurodegenerative andOther Disorders

Veech et al50 suggest that maintenance of a suitabledegree of diet-induced hyperketonemia (in the range of2ndash7 mML) has the potential to aid management ofseveral neurologic diseases including common neurode-generative diseases like Alzheimerrsquos disease (AD) andParkinsonrsquos disease (PD) In their ldquohypothesis paperrdquothe authors single out AD PD Friedreichrsquos AtaxiaLeprechaunisms and lesser forms of insulin disorders asexamples of illnesses that might bene t from therapeuticketosis They also describe experience with one case ofLaFora body disease (until now a uniformly fatal geneticdisease associated with progressive dementia) in whichuse of the 41 HKD resulted in unprecedented improve-ment in cognition gait and awareness after one month73

In this disease mutations in a gene encoding a novelprotein tyrosine phosphatase cause progressive myo-clonic epilepsy

Calabrese et al74 and DiMauro and Schon75 haverecently called attention to evidence suggesting the mi-tochondrial genome may play a key role in neurodegen-erative diseases Decreases in respiratory chain complexactivities have been identi ed in AD PD amyotrophiclateral sclerosis and Huntingtonrsquos disease Mitochondri-opathies (MCPs) possibly associated with oxidantanti-oxidant balance perturbation may underlie defects in

energy metabolism and contribute to cellular damage anddestruction

Parkinsonrsquos DiseaseAlthough the etiology of PD is complex involving bothgenetic and environmental factors there is growing in-terest in the role of mitochondrial dysfunction in thepathogenesis of this disorder Schapira76 pointed out thata proportion of PD patients exhibit a de ciency in mito-chondrial respiratory chain function most notably adefect of complex I within neurons of the substantianigra In his words ldquoBoth mitochondrial mutations andtoxic agents (endogenous and exogenous) have beendemonstrated to cause a de ciency in complex I func-tion This is achieved by increasing the degree of orsusceptibility to oxidative stress which may contributeto apoptopic cell deathrdquo

Finsterer77 reported metabolic studies indicating thatPD can be a manifestation of MCP In his series 63 ofPD patients had features of generalized MCP howeverhe used lactate stress testing to detect patients withMCPmdashan approach that could not be expected to pick upmitochondrial lesions limited to brain sites

According to Veech et al ldquoPD appears to resultfrom an acquired defect in the mitochondrial NADHmultienzyme complex of the dopaminergic cells of themesencephalonrdquo78 They point out although the diseaseis treatable for a period by administration of levodopa(the metabolic precursor of dopamine) continuing freeradical damage eventually reduces the effectiveness ofthis therapy Kashiwaya et al79 reported that primarycultures of mesencephalic dopaminergic neurons ex-posed to the meperidine analogue MPP (a molecule thatinhibits NADH dehydrogenase [complex 1] and causesoxygen free radical formation) can be protected fromdeath by addition of 4 mM D-b-hydroxybutyrate Veechet al50 suggest that BHB probably acts by decreasing thesource of mitochondrial oxygen radical formationldquo by oxidizing the coenzyme Q couple while at thesame time reducing the redox potential of the NADPcouple which through glutathione is the nal detoxi -cation step of H2O2rdquo In a recent review Schon andManfredi80 wrote ldquoBecause complexes I and III are theprincipal sources of free radicals in the cell alteredcomplex I function in the substantia nigra pars compactacould be responsible for the increased DNA damage andlipid peroxidation found in PD brainsrdquo

To the extent that ketones crossing the BBB couldcircumvent or compensate for mitochondrial defects as-sociated with PD or could protect dopaminergic neuronsfrom free radical damage a suitable ketogenic diet mightwell have bene cial effects in certain patients sufferingfrom Parkinsonism

336 Nutrition Reviews Vol 61 No 10

Alzheimerrsquos DiseaseIn 1996 Ogawa et al81 determined regional cerebralblood ow and studied energy utilization in the brains ofseven patients with AD by measuring arteriovenousdifferences of an array of substrates Single-photon emis-sion computed tomography showed that the patients hadsigni cantly decreased regional cerebral blood ow lim-ited to the parietotemporal region The glucose extractionfraction and global cerebral metabolic rate of glucosewere signi cantly decreased whereas the global cerebralmetabolic rate of oxygen was only slightly decreasedBecause the cerebral metabolic rates of the various sub-strates (ketones lactate etc) did not change the authorsconcluded that the markedly elevated oxygenglucoseutilization ratio indicated an altered energy metabolismin AD

Conceivably the impairment of glucose metabolismin AD patients described by Ogawa et al is attributableat least in part to local insulin resistance Watson andCraft82 called attention to emerging evidence that insulinabnormalities and insulin resistance in AD may contrib-ute to the pathophysiology and clinical symptoms of thisdisorder In their words ldquoIt has been recently demon-strated that insulin-sensitive glucose transporters are lo-calized in the same regions supporting memory and thatinsulin plays a role in memory functions Collectivelythese ndings suggest that insulin may contribute tonormal cognitive functioning and that insulin abnormal-ities may exacerbate cognitive impairments such asthose associated with ADrdquo The authors suggest thatimproving insulin effectiveness may have therapeuticbene t for patients with AD The suggestion of Watsonand Craft that insulin abnormalities may play a role inthe pathogenesis of AD has to be reconciled with the nding by Molina et al83 that cerebrospinal uid insulinlevels in 27 AD patients did not differ from those in 16matched controls

In 1996 Hoshi et al84 reported that a fragment ofthe beta chain of amyloid inhibits PDH activity in pri-mary cultures of hippocampal neurons A year later thesame authors found that the 1ndash42 fragment of the betachain of amyloid Ab1ndash42 can inhibit acetyl cholinesynthesis of septal neurons85 Veech et al50 note thatcultured hippocampal neurons die when exposed toAb1ndash42 However addition of 4 mM D-b-hydroxybu-tyrate protects these neurons from amyloid fragmentndashinduced death Casley et al86 showed that beta-amyloidcauses a signi cant reduction in state 3 and state 4mitochondrial respiration which is diminished further bynitric oxide These authors also found that cytochromeoxidase -ketoglutarate dehydrogenase and PDH activ-ities are inhibited by b-amyloid They concluded thatb-amyloid can directly disrupt mitochondrial function

inhibit key enzymes and contribute to the impairment ofenergy metabolism observed in AD

Veech et al50 contend that the foregoing studiesprovide a rationale for trial of a ketogenic diet in thetreatment of AD Ketones can overcome localized insulinresistance can bypass a block in PDH and have beenshown to protect dopaminergic neurons from free radicaldamage and from apoptosis induced by the 1ndash42 amy-loid fragment in tissue-culture experiments

What Next

In considering the potential usefulness of therapeutichyperketonemia in the treatment of AD PD and certainother neurodegenerative conditions it is essential to beclear about what is meant by terms like ldquoketogenic dietrdquoand ldquohyperketonemiardquo

There is widespread confusion among both physi-cians and lay individuals about what constitutes a keto-genic diet Similar to glucose ketones are present in theblood at all times As shown in Table 3 the range ofketone levels that can be achieved by prolonged fooddeprivation or by adherence to various diets is quitewide Hence it makes no sense to speak of a ldquoketogenicdietrdquo without also specifying the degree of serum ketoneelevation that the diet is intended to achieve

The serum concentrations sought for treatment ofepilepsy range from 2 to 7 mML Such levels can beachieved only by strict adherence to a regimen like the41 HKDmdashone most patients nd burdensome On theother hand the kind of low-carbohydrate ldquoketogenicdietrdquo (eg 30 protein 8 carbohydrate 61 fat [asproportions of total energy]) that has recently becomepopularmdashalthough much easier to followmdashachievesBHB levels of only 028ndash040 mML87 (This ldquopopularrdquoketogenic diet at a 2000 kcalday intake provides 127 gof presumptive carbohydrate [40 preformed] plus 87 gderivable from 150 g protein not including the smallquantity of glycerol released in the course of adipocytetriglyceride hydrolysis At the same energy level the 41HKD would yield 33 g of presumptive carbohydrate[10 g preformed] plus 23 g derivable from 40 g proteinThus the effective quantity of carbohydrate provided bythe popular ketogenic diet is almost four times thatsupplied by the 41 HKD)

Future research must address the problem of match-ing diet-induced elevation of serum ketone levels withtherapeutic bene ts Because the drastic type of hyper-ketogenic dietmdashthe one used to treat epilepsy and relatedneurologic disordersmdashis so onerous and dif cult to fol-low it is especially important to determine the extent towhich such a diet can be liberalized without losingdesirable therapeutic effects Given the daunting natureof the 41 HKD it is encouraging that there exists apotential nutritional stratagem which eventually may

337Nutrition Reviews Vol 61 No 10

make it possible to induce therapeutic hyperketonemia inpatients without requiring them to adhere to a Draconianregimen The stratagem entails administration by mouthof substantial quantities of ketones in a formmdashnamelyas esters or small polymersmdashsuitable for ingestion Intheir article about the potential therapeutic uses of ketonebodies Veech et al50 suggest that the problems associ-ated with ldquothe very unappetizing high-fat ketogenicdietrdquo might be obviated by feeding patients up to 100 to150 g per day of a nutritionally acceptable form ofketone bodies Unfortunately neither ketone esters noroligomers of BHB are presently available in suf cientquantities for the conduct of clinical trials

The traditional 41 HKD is dif cult but far fromunmanageable Numerous epileptic children and adultswith seizures resistant to anticonvulsant medication havemanaged to remain on this regimen for years withoutexperiencing serious adverse effects There are justi ableconcerns about its hyperlipidemic effect however andthe fact that the diet has been reported to increase risk ofside effects including nephrolithiasis88 depression ofplatelet function (manifested by easy bruising)89 andoccasionally prolongation of the electrocardiographicQTc interval in children90 It should be possible to mitigateor perhaps prevent the adverse effects of this very highndashfatdiet on plasma lipids by modifying its fatty acid contentWhat is most important now however is to get on with thejob of determining whether diet-induced hyperketonemia(at any level) will prove clinically bene cial in the man-agement of a number of neurodegenerative disorders par-ticularly Parkinsonrsquos disease and Alzheimerrsquos disease

Acknowledgment

We thank Richard L Veech MD DPhil for hisencouragement and very helpful suggestions We alsothank Steven W Fowkes for creative and elegant ren-dering of Figures 1ndash3 and editorial advice

1 Gerhardt J Diabetes mellitus und aceton WienerMedizinische Presse 18656672

2 Hirschfeld F Beobachtungen uber die Acentonurieund das Coma diabeticum Z klin Med 189528176 ndash209

3 Peters JP Van Slyke DD Quantitative ClinicalChemistry Interpretations Volume 1 2nd editionBaltimore Williams amp Wilkins 1946439

4 Cohen PP Studies in ketogenesis J Biol Chem1937119333ndash346

5 Cohen PP Stark IE Hepatic ketogenesis and ketol-ysis in different species J Biol Chem 193812697ndash107

6 Jowett M Quastel JH Fat metabolism III The for-mation and breakdown of acetoacetic acid in animaltissues Biochem J 1935292181ndash2191

7 Chaikoff IL Soskin S The utilization of acetoaceticacid by normal and diabetic dogs before and afterevisceration Am J Physiol 19288758 ndash72

8 Barnes RH Drury DR Greeley PO Wick AN Utili-zation of the ketone bodies in normal animals and inthose with ketosis Am J Physiol 1940130144 ndash150

9 Mirsky IA Broh-Kahn RH The inuence of in-creased metabolism on b-hydroxybutyric acid utili-zation Am J Physiol 1937120446 ndash 450

10 Waters ET Fletcher JP Mirsky IA The relationbetween carbohydrate and b-hydroxybutyric acid

Table 3 Relationship of Metabolic State and Diet Composition to Plasma Levels of Acetoacetate (AcAc)and b-Hydroxybutyrate (BHB) in Normal-weight Overweight and Obese Adults

Metabolic StateDiet

Composition CommentsAcAc

(mML)BHB

(mML) References

Postabsorptive Malesn 5 11 Normal weight 005 007 Hall et al18

n 5 12 Normal weight 008 Sharman et al87

n 5 10 Obese 015 027 Hall et al18

n 5 6 Overweight 001 002 Cahill et al39

Ketogenic diets Males (normal weight)(eucaloric) 3 weeks on diet (n 5 12) 040 Sharman et al87

C8 P30 F61 6 weeks on diet 028 Sharman et al87

Female (normal weight)C2 P8 F90 2 weeks on diet (n 5 1) 070 602 VanItallie Nonas amp Heyms elddagger

Fasting Malesn 5 6 (overweight)

1-week fast 095 358 Cahill et al39

n 5 6 (obese)1ndash2-week fast 104 597 Hall et al18

Distribution (percent total calories) of carbohydrate (C) protein (P) and fat (F)daggerUnpublished study (St Lukersquos-Roosevelt Hospital 2003)Where n 1 all values are means

338 Nutrition Reviews Vol 61 No 10

fat diet They concluded ldquo the anticonvulsant propertyof the ketogenic diet derives from the observed increasein cerebral energy reserve which in turn re ects thehigher ATPADP ratio [this] might provide a reason-able explanation for the increased neuronal stability thatdevelops during chronic ketosisrdquo In 1999 Pan et al62

carried out 31P magnetic resonance spectroscopic imag-ing studies in seven patients with intractable epilepsybefore and after institution of a ketogenic diet Theymeasured signi cant increases (P 005) in high-energy phosphates in association with adherence to theketogenic diet in these subjects They concluded thatenhancement of brain energy metabolism occurs withuse of the ketogenic diet in humans as in lab animals

Thavendiranathan et al63 have recently reportedstudies of the effect of the ldquoclassicrdquo ketogenic diet onseizures triggered by two threshold testsmdashthe pentyle-netetrazol infusion test and the threshold electroconvul-sive shock test These authors observed small but signif-icant threshold elevations (15ndash20) against bothchemically and electrically triggered seizures Similarlyprotective effects were not seen when suprathresholdstimuli were employed

Janigro64 summarized studies indicating that there isa reduction in both glucose uptake and metabolic activityin seizure foci in epileptic patients In his words ldquoDi-minished ion homeostasis together with increased meta-bolic demand of hyperactive neurons may further aggra-vate neuropathological consequences of BBB loss ofglucose uptake mechanismsrdquo Therefore one explanationfor the effectiveness of the 41 HKD in preventingseizures could be the increased transport of BHB at theBBB and its increased availability as an energy source tothe affected neurons BHB is not only an alternate sourceof energy for seizure foci in the brainmdashbypassing animpairment of carbohydrate utilization such as thatwhich can occur from a block in mitochondrial PDHactivitymdashbut it may actually promote ketone and glu-cose uptake in such foci In rats diet-induced ketosisproduces an eightfold increase in the activity of themonocarboxylate transporter and increases the level ofthe glucose transporter GLUT1 in brain endothelialcells and neuropil65 The hypothesis put forth by Janigroto explain how ketosis might eliminate or reduce seizuresis supported by reports that a suitable ketogenic diet isthe treatment of choice for two rare conditions that may

Figure 3 Conditions that can impede acetyl-CoA production from pyruvate and its precursors Examples include (1) insuf cientinsulin (2) insulin receptor abnormalities (3) diminished transport of glucose owing to impaired release of GLUT4 transportersassociated with insulin resistance [see below] (4) pyruvate dehydrogenase complex (PDH) de ciency or dysfunction (5)unavailabilityof glucose at the cell membrane and (6) GLUT1 mutations Provision of ketones to the cell can effectively circumventor overcome such impediments and compensate for the resulting de cits of glycolytically derived acetyl-CoA (GLUT1 designatesthe insulin-independent membrane protein responsible for glucose transport in red blood cells and throughout the blood-brain barrier[BBB] GLUT4 designates the insulin-dependentglucose transporterprotein expressed in most cells Many forms of insulin resistanceentail putative post-receptor defects of insulin-signal transduction resulting in the failure of insulin to release GLUT4 transportersfrom the GLUT4 pool at a rate suf cient to bind and transport enough extracellular glucose to maintain cellular pyruvate levels)

335Nutrition Reviews Vol 61 No 10

be complicated by seizuresmdashthe facilitated glucosetransporter protein type 1 (GLUT1) de ciency syn-drome6667 and the PDH complex de ciency syndrome68

Treatment with a ketogenic diet reportedly brought aboutdramatic clinical improvement in a one-year-old boywith Leigh syndrome which is associated with a de -ciency of PDH69 In this child treatment with a keto-genic diet was associated with striking improvement ofthe cerebral lesions as measured by neuroimaging

In 1973 Buckley and Williamson70 proposed thatneural lipid synthesis preferentially uses AcAc overglucose Patel and Owen71 obtained con rmatory datafor this hypothesis in 1976 In 1984 Koper et al72

compared the rate of incorporation of AcAc and glucoseinto fatty acids and cholesterol using labeled AcAc andglucose At saturating levels of glucose the authorsobserved that the rate at which AcAc label incorporatedinto fatty acids and cholesterol exceeded that of glucoseby a factor of 5 to 10 in both oligodendrocytes andastrocytes These observations provide a basis for argu-ing that some of the bene ts associated with ketogenicdiets may be related to their enhancing effects on lipo-genesis andor sterol synthesis (ie damaged neuronsmight repair more readily in brains given an adequateketone supply)

Therapeutic Potential of Diet-inducedHyperketonemia in Neurodegenerative andOther Disorders

Veech et al50 suggest that maintenance of a suitabledegree of diet-induced hyperketonemia (in the range of2ndash7 mML) has the potential to aid management ofseveral neurologic diseases including common neurode-generative diseases like Alzheimerrsquos disease (AD) andParkinsonrsquos disease (PD) In their ldquohypothesis paperrdquothe authors single out AD PD Friedreichrsquos AtaxiaLeprechaunisms and lesser forms of insulin disorders asexamples of illnesses that might bene t from therapeuticketosis They also describe experience with one case ofLaFora body disease (until now a uniformly fatal geneticdisease associated with progressive dementia) in whichuse of the 41 HKD resulted in unprecedented improve-ment in cognition gait and awareness after one month73

In this disease mutations in a gene encoding a novelprotein tyrosine phosphatase cause progressive myo-clonic epilepsy

Calabrese et al74 and DiMauro and Schon75 haverecently called attention to evidence suggesting the mi-tochondrial genome may play a key role in neurodegen-erative diseases Decreases in respiratory chain complexactivities have been identi ed in AD PD amyotrophiclateral sclerosis and Huntingtonrsquos disease Mitochondri-opathies (MCPs) possibly associated with oxidantanti-oxidant balance perturbation may underlie defects in

energy metabolism and contribute to cellular damage anddestruction

Parkinsonrsquos DiseaseAlthough the etiology of PD is complex involving bothgenetic and environmental factors there is growing in-terest in the role of mitochondrial dysfunction in thepathogenesis of this disorder Schapira76 pointed out thata proportion of PD patients exhibit a de ciency in mito-chondrial respiratory chain function most notably adefect of complex I within neurons of the substantianigra In his words ldquoBoth mitochondrial mutations andtoxic agents (endogenous and exogenous) have beendemonstrated to cause a de ciency in complex I func-tion This is achieved by increasing the degree of orsusceptibility to oxidative stress which may contributeto apoptopic cell deathrdquo

Finsterer77 reported metabolic studies indicating thatPD can be a manifestation of MCP In his series 63 ofPD patients had features of generalized MCP howeverhe used lactate stress testing to detect patients withMCPmdashan approach that could not be expected to pick upmitochondrial lesions limited to brain sites

According to Veech et al ldquoPD appears to resultfrom an acquired defect in the mitochondrial NADHmultienzyme complex of the dopaminergic cells of themesencephalonrdquo78 They point out although the diseaseis treatable for a period by administration of levodopa(the metabolic precursor of dopamine) continuing freeradical damage eventually reduces the effectiveness ofthis therapy Kashiwaya et al79 reported that primarycultures of mesencephalic dopaminergic neurons ex-posed to the meperidine analogue MPP (a molecule thatinhibits NADH dehydrogenase [complex 1] and causesoxygen free radical formation) can be protected fromdeath by addition of 4 mM D-b-hydroxybutyrate Veechet al50 suggest that BHB probably acts by decreasing thesource of mitochondrial oxygen radical formationldquo by oxidizing the coenzyme Q couple while at thesame time reducing the redox potential of the NADPcouple which through glutathione is the nal detoxi -cation step of H2O2rdquo In a recent review Schon andManfredi80 wrote ldquoBecause complexes I and III are theprincipal sources of free radicals in the cell alteredcomplex I function in the substantia nigra pars compactacould be responsible for the increased DNA damage andlipid peroxidation found in PD brainsrdquo

To the extent that ketones crossing the BBB couldcircumvent or compensate for mitochondrial defects as-sociated with PD or could protect dopaminergic neuronsfrom free radical damage a suitable ketogenic diet mightwell have bene cial effects in certain patients sufferingfrom Parkinsonism

336 Nutrition Reviews Vol 61 No 10

Alzheimerrsquos DiseaseIn 1996 Ogawa et al81 determined regional cerebralblood ow and studied energy utilization in the brains ofseven patients with AD by measuring arteriovenousdifferences of an array of substrates Single-photon emis-sion computed tomography showed that the patients hadsigni cantly decreased regional cerebral blood ow lim-ited to the parietotemporal region The glucose extractionfraction and global cerebral metabolic rate of glucosewere signi cantly decreased whereas the global cerebralmetabolic rate of oxygen was only slightly decreasedBecause the cerebral metabolic rates of the various sub-strates (ketones lactate etc) did not change the authorsconcluded that the markedly elevated oxygenglucoseutilization ratio indicated an altered energy metabolismin AD

Conceivably the impairment of glucose metabolismin AD patients described by Ogawa et al is attributableat least in part to local insulin resistance Watson andCraft82 called attention to emerging evidence that insulinabnormalities and insulin resistance in AD may contrib-ute to the pathophysiology and clinical symptoms of thisdisorder In their words ldquoIt has been recently demon-strated that insulin-sensitive glucose transporters are lo-calized in the same regions supporting memory and thatinsulin plays a role in memory functions Collectivelythese ndings suggest that insulin may contribute tonormal cognitive functioning and that insulin abnormal-ities may exacerbate cognitive impairments such asthose associated with ADrdquo The authors suggest thatimproving insulin effectiveness may have therapeuticbene t for patients with AD The suggestion of Watsonand Craft that insulin abnormalities may play a role inthe pathogenesis of AD has to be reconciled with the nding by Molina et al83 that cerebrospinal uid insulinlevels in 27 AD patients did not differ from those in 16matched controls

In 1996 Hoshi et al84 reported that a fragment ofthe beta chain of amyloid inhibits PDH activity in pri-mary cultures of hippocampal neurons A year later thesame authors found that the 1ndash42 fragment of the betachain of amyloid Ab1ndash42 can inhibit acetyl cholinesynthesis of septal neurons85 Veech et al50 note thatcultured hippocampal neurons die when exposed toAb1ndash42 However addition of 4 mM D-b-hydroxybu-tyrate protects these neurons from amyloid fragmentndashinduced death Casley et al86 showed that beta-amyloidcauses a signi cant reduction in state 3 and state 4mitochondrial respiration which is diminished further bynitric oxide These authors also found that cytochromeoxidase -ketoglutarate dehydrogenase and PDH activ-ities are inhibited by b-amyloid They concluded thatb-amyloid can directly disrupt mitochondrial function

inhibit key enzymes and contribute to the impairment ofenergy metabolism observed in AD

Veech et al50 contend that the foregoing studiesprovide a rationale for trial of a ketogenic diet in thetreatment of AD Ketones can overcome localized insulinresistance can bypass a block in PDH and have beenshown to protect dopaminergic neurons from free radicaldamage and from apoptosis induced by the 1ndash42 amy-loid fragment in tissue-culture experiments

What Next

In considering the potential usefulness of therapeutichyperketonemia in the treatment of AD PD and certainother neurodegenerative conditions it is essential to beclear about what is meant by terms like ldquoketogenic dietrdquoand ldquohyperketonemiardquo

There is widespread confusion among both physi-cians and lay individuals about what constitutes a keto-genic diet Similar to glucose ketones are present in theblood at all times As shown in Table 3 the range ofketone levels that can be achieved by prolonged fooddeprivation or by adherence to various diets is quitewide Hence it makes no sense to speak of a ldquoketogenicdietrdquo without also specifying the degree of serum ketoneelevation that the diet is intended to achieve

The serum concentrations sought for treatment ofepilepsy range from 2 to 7 mML Such levels can beachieved only by strict adherence to a regimen like the41 HKDmdashone most patients nd burdensome On theother hand the kind of low-carbohydrate ldquoketogenicdietrdquo (eg 30 protein 8 carbohydrate 61 fat [asproportions of total energy]) that has recently becomepopularmdashalthough much easier to followmdashachievesBHB levels of only 028ndash040 mML87 (This ldquopopularrdquoketogenic diet at a 2000 kcalday intake provides 127 gof presumptive carbohydrate [40 preformed] plus 87 gderivable from 150 g protein not including the smallquantity of glycerol released in the course of adipocytetriglyceride hydrolysis At the same energy level the 41HKD would yield 33 g of presumptive carbohydrate[10 g preformed] plus 23 g derivable from 40 g proteinThus the effective quantity of carbohydrate provided bythe popular ketogenic diet is almost four times thatsupplied by the 41 HKD)

Future research must address the problem of match-ing diet-induced elevation of serum ketone levels withtherapeutic bene ts Because the drastic type of hyper-ketogenic dietmdashthe one used to treat epilepsy and relatedneurologic disordersmdashis so onerous and dif cult to fol-low it is especially important to determine the extent towhich such a diet can be liberalized without losingdesirable therapeutic effects Given the daunting natureof the 41 HKD it is encouraging that there exists apotential nutritional stratagem which eventually may

337Nutrition Reviews Vol 61 No 10

make it possible to induce therapeutic hyperketonemia inpatients without requiring them to adhere to a Draconianregimen The stratagem entails administration by mouthof substantial quantities of ketones in a formmdashnamelyas esters or small polymersmdashsuitable for ingestion Intheir article about the potential therapeutic uses of ketonebodies Veech et al50 suggest that the problems associ-ated with ldquothe very unappetizing high-fat ketogenicdietrdquo might be obviated by feeding patients up to 100 to150 g per day of a nutritionally acceptable form ofketone bodies Unfortunately neither ketone esters noroligomers of BHB are presently available in suf cientquantities for the conduct of clinical trials

The traditional 41 HKD is dif cult but far fromunmanageable Numerous epileptic children and adultswith seizures resistant to anticonvulsant medication havemanaged to remain on this regimen for years withoutexperiencing serious adverse effects There are justi ableconcerns about its hyperlipidemic effect however andthe fact that the diet has been reported to increase risk ofside effects including nephrolithiasis88 depression ofplatelet function (manifested by easy bruising)89 andoccasionally prolongation of the electrocardiographicQTc interval in children90 It should be possible to mitigateor perhaps prevent the adverse effects of this very highndashfatdiet on plasma lipids by modifying its fatty acid contentWhat is most important now however is to get on with thejob of determining whether diet-induced hyperketonemia(at any level) will prove clinically bene cial in the man-agement of a number of neurodegenerative disorders par-ticularly Parkinsonrsquos disease and Alzheimerrsquos disease

Acknowledgment

We thank Richard L Veech MD DPhil for hisencouragement and very helpful suggestions We alsothank Steven W Fowkes for creative and elegant ren-dering of Figures 1ndash3 and editorial advice

1 Gerhardt J Diabetes mellitus und aceton WienerMedizinische Presse 18656672

2 Hirschfeld F Beobachtungen uber die Acentonurieund das Coma diabeticum Z klin Med 189528176 ndash209

3 Peters JP Van Slyke DD Quantitative ClinicalChemistry Interpretations Volume 1 2nd editionBaltimore Williams amp Wilkins 1946439

4 Cohen PP Studies in ketogenesis J Biol Chem1937119333ndash346

5 Cohen PP Stark IE Hepatic ketogenesis and ketol-ysis in different species J Biol Chem 193812697ndash107

6 Jowett M Quastel JH Fat metabolism III The for-mation and breakdown of acetoacetic acid in animaltissues Biochem J 1935292181ndash2191

7 Chaikoff IL Soskin S The utilization of acetoaceticacid by normal and diabetic dogs before and afterevisceration Am J Physiol 19288758 ndash72

8 Barnes RH Drury DR Greeley PO Wick AN Utili-zation of the ketone bodies in normal animals and inthose with ketosis Am J Physiol 1940130144 ndash150

9 Mirsky IA Broh-Kahn RH The inuence of in-creased metabolism on b-hydroxybutyric acid utili-zation Am J Physiol 1937120446 ndash 450

10 Waters ET Fletcher JP Mirsky IA The relationbetween carbohydrate and b-hydroxybutyric acid

Table 3 Relationship of Metabolic State and Diet Composition to Plasma Levels of Acetoacetate (AcAc)and b-Hydroxybutyrate (BHB) in Normal-weight Overweight and Obese Adults

Metabolic StateDiet

Composition CommentsAcAc

(mML)BHB

(mML) References

Postabsorptive Malesn 5 11 Normal weight 005 007 Hall et al18

n 5 12 Normal weight 008 Sharman et al87

n 5 10 Obese 015 027 Hall et al18

n 5 6 Overweight 001 002 Cahill et al39

Ketogenic diets Males (normal weight)(eucaloric) 3 weeks on diet (n 5 12) 040 Sharman et al87

C8 P30 F61 6 weeks on diet 028 Sharman et al87

Female (normal weight)C2 P8 F90 2 weeks on diet (n 5 1) 070 602 VanItallie Nonas amp Heyms elddagger

Fasting Malesn 5 6 (overweight)

1-week fast 095 358 Cahill et al39

n 5 6 (obese)1ndash2-week fast 104 597 Hall et al18

Distribution (percent total calories) of carbohydrate (C) protein (P) and fat (F)daggerUnpublished study (St Lukersquos-Roosevelt Hospital 2003)Where n 1 all values are means

338 Nutrition Reviews Vol 61 No 10

be complicated by seizuresmdashthe facilitated glucosetransporter protein type 1 (GLUT1) de ciency syn-drome6667 and the PDH complex de ciency syndrome68

Treatment with a ketogenic diet reportedly brought aboutdramatic clinical improvement in a one-year-old boywith Leigh syndrome which is associated with a de -ciency of PDH69 In this child treatment with a keto-genic diet was associated with striking improvement ofthe cerebral lesions as measured by neuroimaging

In 1973 Buckley and Williamson70 proposed thatneural lipid synthesis preferentially uses AcAc overglucose Patel and Owen71 obtained con rmatory datafor this hypothesis in 1976 In 1984 Koper et al72

compared the rate of incorporation of AcAc and glucoseinto fatty acids and cholesterol using labeled AcAc andglucose At saturating levels of glucose the authorsobserved that the rate at which AcAc label incorporatedinto fatty acids and cholesterol exceeded that of glucoseby a factor of 5 to 10 in both oligodendrocytes andastrocytes These observations provide a basis for argu-ing that some of the bene ts associated with ketogenicdiets may be related to their enhancing effects on lipo-genesis andor sterol synthesis (ie damaged neuronsmight repair more readily in brains given an adequateketone supply)

Therapeutic Potential of Diet-inducedHyperketonemia in Neurodegenerative andOther Disorders

Veech et al50 suggest that maintenance of a suitabledegree of diet-induced hyperketonemia (in the range of2ndash7 mML) has the potential to aid management ofseveral neurologic diseases including common neurode-generative diseases like Alzheimerrsquos disease (AD) andParkinsonrsquos disease (PD) In their ldquohypothesis paperrdquothe authors single out AD PD Friedreichrsquos AtaxiaLeprechaunisms and lesser forms of insulin disorders asexamples of illnesses that might bene t from therapeuticketosis They also describe experience with one case ofLaFora body disease (until now a uniformly fatal geneticdisease associated with progressive dementia) in whichuse of the 41 HKD resulted in unprecedented improve-ment in cognition gait and awareness after one month73

In this disease mutations in a gene encoding a novelprotein tyrosine phosphatase cause progressive myo-clonic epilepsy

Calabrese et al74 and DiMauro and Schon75 haverecently called attention to evidence suggesting the mi-tochondrial genome may play a key role in neurodegen-erative diseases Decreases in respiratory chain complexactivities have been identi ed in AD PD amyotrophiclateral sclerosis and Huntingtonrsquos disease Mitochondri-opathies (MCPs) possibly associated with oxidantanti-oxidant balance perturbation may underlie defects in

energy metabolism and contribute to cellular damage anddestruction

Parkinsonrsquos DiseaseAlthough the etiology of PD is complex involving bothgenetic and environmental factors there is growing in-terest in the role of mitochondrial dysfunction in thepathogenesis of this disorder Schapira76 pointed out thata proportion of PD patients exhibit a de ciency in mito-chondrial respiratory chain function most notably adefect of complex I within neurons of the substantianigra In his words ldquoBoth mitochondrial mutations andtoxic agents (endogenous and exogenous) have beendemonstrated to cause a de ciency in complex I func-tion This is achieved by increasing the degree of orsusceptibility to oxidative stress which may contributeto apoptopic cell deathrdquo

Finsterer77 reported metabolic studies indicating thatPD can be a manifestation of MCP In his series 63 ofPD patients had features of generalized MCP howeverhe used lactate stress testing to detect patients withMCPmdashan approach that could not be expected to pick upmitochondrial lesions limited to brain sites

According to Veech et al ldquoPD appears to resultfrom an acquired defect in the mitochondrial NADHmultienzyme complex of the dopaminergic cells of themesencephalonrdquo78 They point out although the diseaseis treatable for a period by administration of levodopa(the metabolic precursor of dopamine) continuing freeradical damage eventually reduces the effectiveness ofthis therapy Kashiwaya et al79 reported that primarycultures of mesencephalic dopaminergic neurons ex-posed to the meperidine analogue MPP (a molecule thatinhibits NADH dehydrogenase [complex 1] and causesoxygen free radical formation) can be protected fromdeath by addition of 4 mM D-b-hydroxybutyrate Veechet al50 suggest that BHB probably acts by decreasing thesource of mitochondrial oxygen radical formationldquo by oxidizing the coenzyme Q couple while at thesame time reducing the redox potential of the NADPcouple which through glutathione is the nal detoxi -cation step of H2O2rdquo In a recent review Schon andManfredi80 wrote ldquoBecause complexes I and III are theprincipal sources of free radicals in the cell alteredcomplex I function in the substantia nigra pars compactacould be responsible for the increased DNA damage andlipid peroxidation found in PD brainsrdquo

To the extent that ketones crossing the BBB couldcircumvent or compensate for mitochondrial defects as-sociated with PD or could protect dopaminergic neuronsfrom free radical damage a suitable ketogenic diet mightwell have bene cial effects in certain patients sufferingfrom Parkinsonism

336 Nutrition Reviews Vol 61 No 10

Alzheimerrsquos DiseaseIn 1996 Ogawa et al81 determined regional cerebralblood ow and studied energy utilization in the brains ofseven patients with AD by measuring arteriovenousdifferences of an array of substrates Single-photon emis-sion computed tomography showed that the patients hadsigni cantly decreased regional cerebral blood ow lim-ited to the parietotemporal region The glucose extractionfraction and global cerebral metabolic rate of glucosewere signi cantly decreased whereas the global cerebralmetabolic rate of oxygen was only slightly decreasedBecause the cerebral metabolic rates of the various sub-strates (ketones lactate etc) did not change the authorsconcluded that the markedly elevated oxygenglucoseutilization ratio indicated an altered energy metabolismin AD

Conceivably the impairment of glucose metabolismin AD patients described by Ogawa et al is attributableat least in part to local insulin resistance Watson andCraft82 called attention to emerging evidence that insulinabnormalities and insulin resistance in AD may contrib-ute to the pathophysiology and clinical symptoms of thisdisorder In their words ldquoIt has been recently demon-strated that insulin-sensitive glucose transporters are lo-calized in the same regions supporting memory and thatinsulin plays a role in memory functions Collectivelythese ndings suggest that insulin may contribute tonormal cognitive functioning and that insulin abnormal-ities may exacerbate cognitive impairments such asthose associated with ADrdquo The authors suggest thatimproving insulin effectiveness may have therapeuticbene t for patients with AD The suggestion of Watsonand Craft that insulin abnormalities may play a role inthe pathogenesis of AD has to be reconciled with the nding by Molina et al83 that cerebrospinal uid insulinlevels in 27 AD patients did not differ from those in 16matched controls

In 1996 Hoshi et al84 reported that a fragment ofthe beta chain of amyloid inhibits PDH activity in pri-mary cultures of hippocampal neurons A year later thesame authors found that the 1ndash42 fragment of the betachain of amyloid Ab1ndash42 can inhibit acetyl cholinesynthesis of septal neurons85 Veech et al50 note thatcultured hippocampal neurons die when exposed toAb1ndash42 However addition of 4 mM D-b-hydroxybu-tyrate protects these neurons from amyloid fragmentndashinduced death Casley et al86 showed that beta-amyloidcauses a signi cant reduction in state 3 and state 4mitochondrial respiration which is diminished further bynitric oxide These authors also found that cytochromeoxidase -ketoglutarate dehydrogenase and PDH activ-ities are inhibited by b-amyloid They concluded thatb-amyloid can directly disrupt mitochondrial function

inhibit key enzymes and contribute to the impairment ofenergy metabolism observed in AD

Veech et al50 contend that the foregoing studiesprovide a rationale for trial of a ketogenic diet in thetreatment of AD Ketones can overcome localized insulinresistance can bypass a block in PDH and have beenshown to protect dopaminergic neurons from free radicaldamage and from apoptosis induced by the 1ndash42 amy-loid fragment in tissue-culture experiments

What Next

In considering the potential usefulness of therapeutichyperketonemia in the treatment of AD PD and certainother neurodegenerative conditions it is essential to beclear about what is meant by terms like ldquoketogenic dietrdquoand ldquohyperketonemiardquo

There is widespread confusion among both physi-cians and lay individuals about what constitutes a keto-genic diet Similar to glucose ketones are present in theblood at all times As shown in Table 3 the range ofketone levels that can be achieved by prolonged fooddeprivation or by adherence to various diets is quitewide Hence it makes no sense to speak of a ldquoketogenicdietrdquo without also specifying the degree of serum ketoneelevation that the diet is intended to achieve

The serum concentrations sought for treatment ofepilepsy range from 2 to 7 mML Such levels can beachieved only by strict adherence to a regimen like the41 HKDmdashone most patients nd burdensome On theother hand the kind of low-carbohydrate ldquoketogenicdietrdquo (eg 30 protein 8 carbohydrate 61 fat [asproportions of total energy]) that has recently becomepopularmdashalthough much easier to followmdashachievesBHB levels of only 028ndash040 mML87 (This ldquopopularrdquoketogenic diet at a 2000 kcalday intake provides 127 gof presumptive carbohydrate [40 preformed] plus 87 gderivable from 150 g protein not including the smallquantity of glycerol released in the course of adipocytetriglyceride hydrolysis At the same energy level the 41HKD would yield 33 g of presumptive carbohydrate[10 g preformed] plus 23 g derivable from 40 g proteinThus the effective quantity of carbohydrate provided bythe popular ketogenic diet is almost four times thatsupplied by the 41 HKD)

Future research must address the problem of match-ing diet-induced elevation of serum ketone levels withtherapeutic bene ts Because the drastic type of hyper-ketogenic dietmdashthe one used to treat epilepsy and relatedneurologic disordersmdashis so onerous and dif cult to fol-low it is especially important to determine the extent towhich such a diet can be liberalized without losingdesirable therapeutic effects Given the daunting natureof the 41 HKD it is encouraging that there exists apotential nutritional stratagem which eventually may

337Nutrition Reviews Vol 61 No 10

make it possible to induce therapeutic hyperketonemia inpatients without requiring them to adhere to a Draconianregimen The stratagem entails administration by mouthof substantial quantities of ketones in a formmdashnamelyas esters or small polymersmdashsuitable for ingestion Intheir article about the potential therapeutic uses of ketonebodies Veech et al50 suggest that the problems associ-ated with ldquothe very unappetizing high-fat ketogenicdietrdquo might be obviated by feeding patients up to 100 to150 g per day of a nutritionally acceptable form ofketone bodies Unfortunately neither ketone esters noroligomers of BHB are presently available in suf cientquantities for the conduct of clinical trials

The traditional 41 HKD is dif cult but far fromunmanageable Numerous epileptic children and adultswith seizures resistant to anticonvulsant medication havemanaged to remain on this regimen for years withoutexperiencing serious adverse effects There are justi ableconcerns about its hyperlipidemic effect however andthe fact that the diet has been reported to increase risk ofside effects including nephrolithiasis88 depression ofplatelet function (manifested by easy bruising)89 andoccasionally prolongation of the electrocardiographicQTc interval in children90 It should be possible to mitigateor perhaps prevent the adverse effects of this very highndashfatdiet on plasma lipids by modifying its fatty acid contentWhat is most important now however is to get on with thejob of determining whether diet-induced hyperketonemia(at any level) will prove clinically bene cial in the man-agement of a number of neurodegenerative disorders par-ticularly Parkinsonrsquos disease and Alzheimerrsquos disease

Acknowledgment

We thank Richard L Veech MD DPhil for hisencouragement and very helpful suggestions We alsothank Steven W Fowkes for creative and elegant ren-dering of Figures 1ndash3 and editorial advice

1 Gerhardt J Diabetes mellitus und aceton WienerMedizinische Presse 18656672

2 Hirschfeld F Beobachtungen uber die Acentonurieund das Coma diabeticum Z klin Med 189528176 ndash209

3 Peters JP Van Slyke DD Quantitative ClinicalChemistry Interpretations Volume 1 2nd editionBaltimore Williams amp Wilkins 1946439

4 Cohen PP Studies in ketogenesis J Biol Chem1937119333ndash346

5 Cohen PP Stark IE Hepatic ketogenesis and ketol-ysis in different species J Biol Chem 193812697ndash107

6 Jowett M Quastel JH Fat metabolism III The for-mation and breakdown of acetoacetic acid in animaltissues Biochem J 1935292181ndash2191

7 Chaikoff IL Soskin S The utilization of acetoaceticacid by normal and diabetic dogs before and afterevisceration Am J Physiol 19288758 ndash72

8 Barnes RH Drury DR Greeley PO Wick AN Utili-zation of the ketone bodies in normal animals and inthose with ketosis Am J Physiol 1940130144 ndash150

9 Mirsky IA Broh-Kahn RH The inuence of in-creased metabolism on b-hydroxybutyric acid utili-zation Am J Physiol 1937120446 ndash 450

10 Waters ET Fletcher JP Mirsky IA The relationbetween carbohydrate and b-hydroxybutyric acid

Table 3 Relationship of Metabolic State and Diet Composition to Plasma Levels of Acetoacetate (AcAc)and b-Hydroxybutyrate (BHB) in Normal-weight Overweight and Obese Adults

Metabolic StateDiet

Composition CommentsAcAc

(mML)BHB

(mML) References

Postabsorptive Malesn 5 11 Normal weight 005 007 Hall et al18

n 5 12 Normal weight 008 Sharman et al87

n 5 10 Obese 015 027 Hall et al18

n 5 6 Overweight 001 002 Cahill et al39

Ketogenic diets Males (normal weight)(eucaloric) 3 weeks on diet (n 5 12) 040 Sharman et al87

C8 P30 F61 6 weeks on diet 028 Sharman et al87

Female (normal weight)C2 P8 F90 2 weeks on diet (n 5 1) 070 602 VanItallie Nonas amp Heyms elddagger

Fasting Malesn 5 6 (overweight)

1-week fast 095 358 Cahill et al39

n 5 6 (obese)1ndash2-week fast 104 597 Hall et al18

Distribution (percent total calories) of carbohydrate (C) protein (P) and fat (F)daggerUnpublished study (St Lukersquos-Roosevelt Hospital 2003)Where n 1 all values are means

338 Nutrition Reviews Vol 61 No 10

Alzheimerrsquos DiseaseIn 1996 Ogawa et al81 determined regional cerebralblood ow and studied energy utilization in the brains ofseven patients with AD by measuring arteriovenousdifferences of an array of substrates Single-photon emis-sion computed tomography showed that the patients hadsigni cantly decreased regional cerebral blood ow lim-ited to the parietotemporal region The glucose extractionfraction and global cerebral metabolic rate of glucosewere signi cantly decreased whereas the global cerebralmetabolic rate of oxygen was only slightly decreasedBecause the cerebral metabolic rates of the various sub-strates (ketones lactate etc) did not change the authorsconcluded that the markedly elevated oxygenglucoseutilization ratio indicated an altered energy metabolismin AD

Conceivably the impairment of glucose metabolismin AD patients described by Ogawa et al is attributableat least in part to local insulin resistance Watson andCraft82 called attention to emerging evidence that insulinabnormalities and insulin resistance in AD may contrib-ute to the pathophysiology and clinical symptoms of thisdisorder In their words ldquoIt has been recently demon-strated that insulin-sensitive glucose transporters are lo-calized in the same regions supporting memory and thatinsulin plays a role in memory functions Collectivelythese ndings suggest that insulin may contribute tonormal cognitive functioning and that insulin abnormal-ities may exacerbate cognitive impairments such asthose associated with ADrdquo The authors suggest thatimproving insulin effectiveness may have therapeuticbene t for patients with AD The suggestion of Watsonand Craft that insulin abnormalities may play a role inthe pathogenesis of AD has to be reconciled with the nding by Molina et al83 that cerebrospinal uid insulinlevels in 27 AD patients did not differ from those in 16matched controls

In 1996 Hoshi et al84 reported that a fragment ofthe beta chain of amyloid inhibits PDH activity in pri-mary cultures of hippocampal neurons A year later thesame authors found that the 1ndash42 fragment of the betachain of amyloid Ab1ndash42 can inhibit acetyl cholinesynthesis of septal neurons85 Veech et al50 note thatcultured hippocampal neurons die when exposed toAb1ndash42 However addition of 4 mM D-b-hydroxybu-tyrate protects these neurons from amyloid fragmentndashinduced death Casley et al86 showed that beta-amyloidcauses a signi cant reduction in state 3 and state 4mitochondrial respiration which is diminished further bynitric oxide These authors also found that cytochromeoxidase -ketoglutarate dehydrogenase and PDH activ-ities are inhibited by b-amyloid They concluded thatb-amyloid can directly disrupt mitochondrial function

inhibit key enzymes and contribute to the impairment ofenergy metabolism observed in AD

Veech et al50 contend that the foregoing studiesprovide a rationale for trial of a ketogenic diet in thetreatment of AD Ketones can overcome localized insulinresistance can bypass a block in PDH and have beenshown to protect dopaminergic neurons from free radicaldamage and from apoptosis induced by the 1ndash42 amy-loid fragment in tissue-culture experiments

What Next

In considering the potential usefulness of therapeutichyperketonemia in the treatment of AD PD and certainother neurodegenerative conditions it is essential to beclear about what is meant by terms like ldquoketogenic dietrdquoand ldquohyperketonemiardquo

There is widespread confusion among both physi-cians and lay individuals about what constitutes a keto-genic diet Similar to glucose ketones are present in theblood at all times As shown in Table 3 the range ofketone levels that can be achieved by prolonged fooddeprivation or by adherence to various diets is quitewide Hence it makes no sense to speak of a ldquoketogenicdietrdquo without also specifying the degree of serum ketoneelevation that the diet is intended to achieve

The serum concentrations sought for treatment ofepilepsy range from 2 to 7 mML Such levels can beachieved only by strict adherence to a regimen like the41 HKDmdashone most patients nd burdensome On theother hand the kind of low-carbohydrate ldquoketogenicdietrdquo (eg 30 protein 8 carbohydrate 61 fat [asproportions of total energy]) that has recently becomepopularmdashalthough much easier to followmdashachievesBHB levels of only 028ndash040 mML87 (This ldquopopularrdquoketogenic diet at a 2000 kcalday intake provides 127 gof presumptive carbohydrate [40 preformed] plus 87 gderivable from 150 g protein not including the smallquantity of glycerol released in the course of adipocytetriglyceride hydrolysis At the same energy level the 41HKD would yield 33 g of presumptive carbohydrate[10 g preformed] plus 23 g derivable from 40 g proteinThus the effective quantity of carbohydrate provided bythe popular ketogenic diet is almost four times thatsupplied by the 41 HKD)

Future research must address the problem of match-ing diet-induced elevation of serum ketone levels withtherapeutic bene ts Because the drastic type of hyper-ketogenic dietmdashthe one used to treat epilepsy and relatedneurologic disordersmdashis so onerous and dif cult to fol-low it is especially important to determine the extent towhich such a diet can be liberalized without losingdesirable therapeutic effects Given the daunting natureof the 41 HKD it is encouraging that there exists apotential nutritional stratagem which eventually may

337Nutrition Reviews Vol 61 No 10

make it possible to induce therapeutic hyperketonemia inpatients without requiring them to adhere to a Draconianregimen The stratagem entails administration by mouthof substantial quantities of ketones in a formmdashnamelyas esters or small polymersmdashsuitable for ingestion Intheir article about the potential therapeutic uses of ketonebodies Veech et al50 suggest that the problems associ-ated with ldquothe very unappetizing high-fat ketogenicdietrdquo might be obviated by feeding patients up to 100 to150 g per day of a nutritionally acceptable form ofketone bodies Unfortunately neither ketone esters noroligomers of BHB are presently available in suf cientquantities for the conduct of clinical trials

The traditional 41 HKD is dif cult but far fromunmanageable Numerous epileptic children and adultswith seizures resistant to anticonvulsant medication havemanaged to remain on this regimen for years withoutexperiencing serious adverse effects There are justi ableconcerns about its hyperlipidemic effect however andthe fact that the diet has been reported to increase risk ofside effects including nephrolithiasis88 depression ofplatelet function (manifested by easy bruising)89 andoccasionally prolongation of the electrocardiographicQTc interval in children90 It should be possible to mitigateor perhaps prevent the adverse effects of this very highndashfatdiet on plasma lipids by modifying its fatty acid contentWhat is most important now however is to get on with thejob of determining whether diet-induced hyperketonemia(at any level) will prove clinically bene cial in the man-agement of a number of neurodegenerative disorders par-ticularly Parkinsonrsquos disease and Alzheimerrsquos disease

Acknowledgment

We thank Richard L Veech MD DPhil for hisencouragement and very helpful suggestions We alsothank Steven W Fowkes for creative and elegant ren-dering of Figures 1ndash3 and editorial advice

1 Gerhardt J Diabetes mellitus und aceton WienerMedizinische Presse 18656672

2 Hirschfeld F Beobachtungen uber die Acentonurieund das Coma diabeticum Z klin Med 189528176 ndash209

3 Peters JP Van Slyke DD Quantitative ClinicalChemistry Interpretations Volume 1 2nd editionBaltimore Williams amp Wilkins 1946439

4 Cohen PP Studies in ketogenesis J Biol Chem1937119333ndash346

5 Cohen PP Stark IE Hepatic ketogenesis and ketol-ysis in different species J Biol Chem 193812697ndash107

6 Jowett M Quastel JH Fat metabolism III The for-mation and breakdown of acetoacetic acid in animaltissues Biochem J 1935292181ndash2191

7 Chaikoff IL Soskin S The utilization of acetoaceticacid by normal and diabetic dogs before and afterevisceration Am J Physiol 19288758 ndash72

8 Barnes RH Drury DR Greeley PO Wick AN Utili-zation of the ketone bodies in normal animals and inthose with ketosis Am J Physiol 1940130144 ndash150

9 Mirsky IA Broh-Kahn RH The inuence of in-creased metabolism on b-hydroxybutyric acid utili-zation Am J Physiol 1937120446 ndash 450

10 Waters ET Fletcher JP Mirsky IA The relationbetween carbohydrate and b-hydroxybutyric acid

Table 3 Relationship of Metabolic State and Diet Composition to Plasma Levels of Acetoacetate (AcAc)and b-Hydroxybutyrate (BHB) in Normal-weight Overweight and Obese Adults

Metabolic StateDiet

Composition CommentsAcAc

(mML)BHB

(mML) References

Postabsorptive Malesn 5 11 Normal weight 005 007 Hall et al18

n 5 12 Normal weight 008 Sharman et al87

n 5 10 Obese 015 027 Hall et al18

n 5 6 Overweight 001 002 Cahill et al39

Ketogenic diets Males (normal weight)(eucaloric) 3 weeks on diet (n 5 12) 040 Sharman et al87

C8 P30 F61 6 weeks on diet 028 Sharman et al87

Female (normal weight)C2 P8 F90 2 weeks on diet (n 5 1) 070 602 VanItallie Nonas amp Heyms elddagger

Fasting Malesn 5 6 (overweight)

1-week fast 095 358 Cahill et al39

n 5 6 (obese)1ndash2-week fast 104 597 Hall et al18

Distribution (percent total calories) of carbohydrate (C) protein (P) and fat (F)daggerUnpublished study (St Lukersquos-Roosevelt Hospital 2003)Where n 1 all values are means

338 Nutrition Reviews Vol 61 No 10

make it possible to induce therapeutic hyperketonemia inpatients without requiring them to adhere to a Draconianregimen The stratagem entails administration by mouthof substantial quantities of ketones in a formmdashnamelyas esters or small polymersmdashsuitable for ingestion Intheir article about the potential therapeutic uses of ketonebodies Veech et al50 suggest that the problems associ-ated with ldquothe very unappetizing high-fat ketogenicdietrdquo might be obviated by feeding patients up to 100 to150 g per day of a nutritionally acceptable form ofketone bodies Unfortunately neither ketone esters noroligomers of BHB are presently available in suf cientquantities for the conduct of clinical trials

The traditional 41 HKD is dif cult but far fromunmanageable Numerous epileptic children and adultswith seizures resistant to anticonvulsant medication havemanaged to remain on this regimen for years withoutexperiencing serious adverse effects There are justi ableconcerns about its hyperlipidemic effect however andthe fact that the diet has been reported to increase risk ofside effects including nephrolithiasis88 depression ofplatelet function (manifested by easy bruising)89 andoccasionally prolongation of the electrocardiographicQTc interval in children90 It should be possible to mitigateor perhaps prevent the adverse effects of this very highndashfatdiet on plasma lipids by modifying its fatty acid contentWhat is most important now however is to get on with thejob of determining whether diet-induced hyperketonemia(at any level) will prove clinically bene cial in the man-agement of a number of neurodegenerative disorders par-ticularly Parkinsonrsquos disease and Alzheimerrsquos disease

Acknowledgment

We thank Richard L Veech MD DPhil for hisencouragement and very helpful suggestions We alsothank Steven W Fowkes for creative and elegant ren-dering of Figures 1ndash3 and editorial advice

1 Gerhardt J Diabetes mellitus und aceton WienerMedizinische Presse 18656672

2 Hirschfeld F Beobachtungen uber die Acentonurieund das Coma diabeticum Z klin Med 189528176 ndash209

3 Peters JP Van Slyke DD Quantitative ClinicalChemistry Interpretations Volume 1 2nd editionBaltimore Williams amp Wilkins 1946439

4 Cohen PP Studies in ketogenesis J Biol Chem1937119333ndash346

5 Cohen PP Stark IE Hepatic ketogenesis and ketol-ysis in different species J Biol Chem 193812697ndash107

6 Jowett M Quastel JH Fat metabolism III The for-mation and breakdown of acetoacetic acid in animaltissues Biochem J 1935292181ndash2191

7 Chaikoff IL Soskin S The utilization of acetoaceticacid by normal and diabetic dogs before and afterevisceration Am J Physiol 19288758 ndash72

8 Barnes RH Drury DR Greeley PO Wick AN Utili-zation of the ketone bodies in normal animals and inthose with ketosis Am J Physiol 1940130144 ndash150

9 Mirsky IA Broh-Kahn RH The inuence of in-creased metabolism on b-hydroxybutyric acid utili-zation Am J Physiol 1937120446 ndash 450

10 Waters ET Fletcher JP Mirsky IA The relationbetween carbohydrate and b-hydroxybutyric acid

Table 3 Relationship of Metabolic State and Diet Composition to Plasma Levels of Acetoacetate (AcAc)and b-Hydroxybutyrate (BHB) in Normal-weight Overweight and Obese Adults

Metabolic StateDiet

Composition CommentsAcAc

(mML)BHB

(mML) References

Postabsorptive Malesn 5 11 Normal weight 005 007 Hall et al18

n 5 12 Normal weight 008 Sharman et al87

n 5 10 Obese 015 027 Hall et al18

n 5 6 Overweight 001 002 Cahill et al39

Ketogenic diets Males (normal weight)(eucaloric) 3 weeks on diet (n 5 12) 040 Sharman et al87

C8 P30 F61 6 weeks on diet 028 Sharman et al87

Female (normal weight)C2 P8 F90 2 weeks on diet (n 5 1) 070 602 VanItallie Nonas amp Heyms elddagger

Fasting Malesn 5 6 (overweight)

1-week fast 095 358 Cahill et al39

n 5 6 (obese)1ndash2-week fast 104 597 Hall et al18

Distribution (percent total calories) of carbohydrate (C) protein (P) and fat (F)daggerUnpublished study (St Lukersquos-Roosevelt Hospital 2003)Where n 1 all values are means

338 Nutrition Reviews Vol 61 No 10

utilization by the heart-lung preparation Am JPhysiol 1938122542ndash546

11 Johnson RE Sargent F II Passmore R Normalvariations in total ketone bodies in serum and urineof healthy young men Q J Exp Physiol 195843339 ndash344

12 Cahill GF Jr Starvation Trans Am Clin ClimatolAssoc 1982941ndash21

13 Foster DW Diabetes mellitus In Wilson JD Braun-wald E Isselbacher KJ eds Harrisonrsquos Principles ofInternal Medicine 12th edition New York McGraw-Hill 19911750

14 Crandall DI Gurin S Studies of acetoacetate for-mation with labeled carbon I Experiments withpyruvate acetate and fatty acids in washed liverhomogenates J Biol Chem 1949181829 ndash 843

15 Lynen F Henning U Bublitz C Sorbo B Kroplin-Rueff L Der chemische Mechanismus der Acetes-sigsaurebildung in der Leber Biochem Ztschr1958330269 ndash295

16 Robinson AM Williamson DH Physiological roles ofketone bodies as substrates and signals in mam-malian tissues Physiol Rev 198060143ndash187

17 Wastney ME Hall SEH Berman M Ketone bodykinetics in humans a mathematical model J LipidRes 198425160 ndash174

18 Hall SEH Wastney ME Bolton TM Braaten JTBerman M Ketone body kinetics in humans theeffects of insulin-dependent diabetes obesity andstarvation J Lipid Res 1984251184 ndash1194

19 Williamson DH Mellanby J Krebs HA Enzymicdetermination of D(-)b-hydroxybutyric acid and ace-toacetic acid in blood Biochem J 19628290 ndash96

20 Krebs HA The physiological role of the ketonebodies Biochem J 196180225ndash233

21 Cahill GF Jr Starvation in man N Engl J Med1970282668 ndash 675

22 Laffel L Ketone bodies a review of physiologypathophysiology and application of monitoring todiabetes Diabetes Metab Res Rev 199915412ndash426

23 Courtice FC Douglas CG The effects of prolongedmuscular exercise on the metabolism Proc R Soc1936119B381ndash 439

24 Owen OE Reichard GAJ Human forearm metabo-lism during progressive starvation J Clin Invest1971501538 ndash1545

25 Cahill GF Jr Starvation in man Clin EndocrinolMetab 19755397ndash 417

26 Behre JA Studies in ketone body excretion I Dailyvariations in the ketone bodies of normal urine andthe ketonuria of short fasts with a note on diabeticketonuria during insulin treatment J Biol Chem193192679 ndash 697

27 Kartin BL Man EB Winkler AW Peters JP Bloodketones and serum lipids in starvation and waterdeprivation J Clin Invest 194423824 ndash835

28 Johnson RE Passmore R Sargent F Multiple fac-tors in experimental human ketosis Arch InternMed 196110743ndash50

29 Bach SJ Hibbitt KG Biochemical aspects of bovineketosis Biochem J 19597287ndash92

30 Paterson P Sheath J Taft P Wood C Maternal andfoetal ketone concentrations in plasma and urineLancet 19671862ndash865

31 Aoki TT Muller WA Brennan MR Cahill GF Jr Themetabolic effects of glucose in brief and prolongedfasted man Am J Clin Nutr 197528507ndash511

32 Heinbecker P Studies on the metabolism of Eski-mos J Biol Chem 192880461ndash 475

33 Astrand P-O Rodahl K Textbook of Work Physiol-ogy Physiological Bases of Exercise 3rd editionNew York McGraw-Hill 1986317

34 McClellan WS DuBois EF Clinical calorimetry XLVProlonged meat diets with a study of kidney func-tion and ketosis J Biol Chem 193087651ndash 667

35 Krebs HA The metabolic fate of amino acids InMunro HN Allison JB eds Mammalian Protein Me-tabolism New York Academic Press 1964126 ndash176

36 Jungas RL Halperin ML Brosnan JT Quantitativeanalysis of amino acid oxidation and related glu-coneogenesis in humans Physiol Rev 199272419 ndash 448

37 Swink TD Vining EPG Freeman JM The ketogenicdiet 1997 Adv Pediatr 199744297ndash329

38 VanItallie TB Bergen SS Jr Ketogenesis and hyper-ketonemia Am J Med 196131909 ndash918

39 Cahill GF Jr Herrera G Morgan AP et al Hormone-fuel interrelationships during fasting J Clin Invest1966451751ndash1769

40 Sokoloff L Metabolism of the central nervous sys-tem in vivo In Field J ed Handbook of PhysiologySection I Neurophysiology Baltimore AmericanPhysiological Society Williams and Wilkins 19601843ndash1864

41 Benedict FG A study of prolonged fasting Wash-ington DC The Carnegie Institute 1915Publication203

42 Gamble JL Ross GS Tisdall FF The metabolism ofxed base during fasting J Biol Chem 192357633ndash695

43 Leiter LA Marliss EB Survival during fasting maydepend on fat as well as protein stores JAMA19822482306 ndash2307

44 Owen OE Morgan AP Kemp HG et al Brainmetabolism during fasting J Clin Invest 1967461589 ndash1595

45 Lardy HA Hansen RG Phillips PH The metabolismof bovine epididymal spermatozoa Arch Biochem1945641ndash51

46 Lardy HA Phillips PH Studies of fat and carbohy-drate oxidation in mammalian spermatozoa ArchBiochem 1945653ndash61

47 Kashiwaya Y Sato K Tsuchiya S et al Control ofglucose utilization in working perfused rat heartJ Biol Chem 199426925502ndash25514

48 Sato K Kashiwaya Y Keon CA et al Insulin ketonebodies and mitochondrial energy transductionFASEB J 19959651ndash 658

49 Kashiwaya Y King MT Veech RL Substrate signal-ing by insulin a ketone bodies ratio mimics insulinaction in heart Am J Cardiol 19978050Andash 64A

50 Veech RL Chance B Kashiwaya Y Lardy HA CahillGF Jr Ketone bodies potential therapeutic usesIUBMB Life 200151241ndash247

51 McHenry LC Jr Cerebral blood ow N Engl J Med196627482ndash91

52 Woodyatt RT Objects and method of diet adjust-

339Nutrition Reviews Vol 61 No 10

ment in diabetics Arch Intern Med 192128125ndash141

53 Wilder RM The effect of ketonemia on the course ofepilepsy Mayo Clin Bull 19212307ndash8

54 Peterman MG The ketogenic diet in the treatmentof epilepsy A preliminary report Am J Dis Child19242828 ndash33

55 Huttenlocher PR Wilbourn AJ Signore JM Medium-chain triglycerides as a therapy for intractable child-hood epilepsy Neurology 1971211097ndash1103

56 Gordon N Medium-chain triglycerides in a keto-genic diet Dev Med Child Neurol 197719535ndash544

57 Bergen SS Jr Hashim SA VanItallie TB Hyperk-etonemia induced in man by medium-chain triglyc-eride Diabetes 196615723ndash725

58 Schwartz RH Eaton J Bower BD Aynsley-Green AKetogenic diets in the treatment of epilepsy Short-term clinical effects Dev Med Child Neurol 198931145ndash151

59 Al-Mudallal AS LaManna JC Lust WD Harik SLDiet-induced ketosis does not cause cerebral aci-dosis Epilepsia 199637258 ndash261

60 Uhleman ER Neims AH Anticonvulsant propertiesof the ketogenic diet in mice J Pharmacol Exp Ther1972180231ndash238

61 DeVivo DC Malas KL Leckie MP Starvation andseizures Arch Neurol 197432755ndash760

62 Pan JW Bebin EM Chu WJ Hetherington HPKetosis and epilepsy 31P spectroscopic imaging at41 T Epilepsia 199940703ndash707

63 Thavendiranathan P Chow C Cunnane S McIntyreBW The effect of the lsquoclassicrsquo ketogenic diet onanimal seizure models Brain Res 2003959206 ndash213

64 Janigro D Blood-brain barrier ion homeostasis andepilepsy possible implications towards the under-standing of ketogenic diet mechanisms EpilepsyRes 199937223ndash232

65 Leino RL Gerhart DZ Duelli R Enerson BE DrewesLR Diet-induced ketosis increases monocarboxy-late transporter (MCT1) levels in rat brain Neuro-chem Int 200138519 ndash527

66 Seidner G Garcia Alvarez M Yeh J-I et al GLUT-1deciency syndrome caused by haploinsufciencyof the blood-brain-barrier hexose carrier Nat Genet199818166 ndash191

67 Klepper J Voit T Facilitated glucose transporterprotein type 1 (GLUT1) deciency syndrome im-paired glucose transport into brainmdasha review EurJ Pediatr 2002161295ndash304

68 Wexler ID Hemalatha SG McConnell J et alOutcome of pyruvate dehydrogenase deciencytreated with ketogenic diets Studies in patientswith identical mutations Neurology 1997491655ndash1661

69 Wijburg FA Barth PG Bindoff LA et al Leigh syn-drome associated with a deciency of the pyruvatedehydrogenase complex results of treatment with aketogenic diet Neuropediatrics 199223147ndash152

70 Buckley BM Williamson DH Acetoacetate andbrain lipogenesis developmental pattern of aceto-acetyl-coenzyme A synthetase in the soluble frac-tion of rat brain Biochem J 1973132653ndash656

71 Patel MS Owen OE Lipogenesis from ketone bod-

ies in rat brain Evidence for conversion of aceto-acetate into acetyl-coenzyme A in the cytosol Bio-chem J 1976156603ndash607

72 Koper JW Zeinstra EC Lopes-Cardozo M vanGolde LM Acetoacetate and glucose as substratesfor lipid synthesis by rat brain oligodendrocytes andastrocytes in serum-free culture Biochim BiophysActa 198479620 ndash26

73 Minassian BA Lee JR Herbrick JA Huizenga J etal Mutations in a gene encoding a novel proteintyrosine phosphatase cause progressive myoclonicepilepsy Nat Genet 119820171ndash174

74 Calabrese V Scapagnini G Giuffrida Stella AMBates TE Clark JB Mitochondrial involvement inbrain function and dysfunction relevance to agingneurodegenerative disorders and longevity Neuro-chem Res 200126739 ndash764

75 DiMauro S Schon EA Mitochondrial respiratory-chain diseases N Engl J Med 20033482656 ndash2668

76 Schapira AH Causes of neuronal death in Parkin-sonrsquos disease Adv Neurol 200186155ndash162

77 Finsterer J Parkinsonrsquos disease as a manifestationof mitochondriopathy Acta Neurol Scand 2002105384 ndash389

78 Veech GA Dennis J Keeney PM et al Disruptedmitochondrial electron transport function increasesexpression of anti-apoptotic bcl-2 and bcl-XL pro-teins in SH-SY5Y neuroblastoma and in Parkinsonrsquosdisease cybrid cells through oxidative stress J Neu-rosci Res 200061693ndash700

79 Kashiwaya Y Takeshima T Mori N Nakashima KClarke K Veech RL D-b-Hydroxybutyrate protectsneurons in models of Alzheimerrsquos and Parkinsonrsquosdisease Proc Natl Acad Sci U S A 2000975440 ndash5444

80 Schon EA Manfredi G Neuronal degeneration andmitochondrial dysfunction J Clin Invest 2003111303ndash312

81 Ogawa M Fukuyama H Ouchi Y Yamauchi HKimura J Altered energy metabolism in Alzheimerrsquosdisease J Neurol Sci 199613978 ndash 82

82 Watson GS Craft S The role of insulin resistance inthe pathogenesis of Alzheimerrsquos disease implica-tions for treatment CNS Drugs 20031727ndash45

83 Molina JA Jimenez-Jimenez FJ Vargas C et alCerebrospinal uid levels of insulin in patients withAlzheimerrsquos disease Acta Neurol Scand 2002106347ndash350

84 Hoshi M Takashima K Noguchi M et al Regulationof mitochondrial pyruvate dehydrogenase activityby tau protein kinase Iglycogen synthase kinase3beta in brain Proc Natl Acad Sci U S A 1996932719 ndash2723

85 Hoshi M Takashima M Murayama K et al Non-toxic amyloid beta peptide 1ndash42 suppresses ace-tylcholine synthesis J Biol Chem 19972722038 ndash2041

86 Casley CS Canevari L Land JM Clark JB SharpeMA Beta-amyloid inhibits integrated mitochondrialrespiration and key enzyme activities J Neurochem20028091ndash100

87 Sharman MJ Kraemer WJ Love DM et al A keto-genic diet favorably affects serum biomarkers for

340 Nutrition Reviews Vol 61 No 10

cardiovascular disease in normal-weight youngmen J Nutr 20021321879 ndash1885

88 Furth SL Casey JC Pyzik PL et al Risk factors forurolithiasis in children on the ketogenic diet PediatrNephrol 200015125ndash128

89 Berry-Kravis E Booth G Taylor A Valentino LA

Bruising and the ketogenic diet evidence for diet-induced changes in platelet function Ann Neurol20014998 ndash103

90 Best TH Franz DN Gilbert DL Nelson DP EpsteinMR Cardiac complications in pediatric patients onthe ketogenic diet Neurology 2000542328 ndash2330

341Nutrition Reviews Vol 61 No 10