Copyright © 1997, Elsevier Science Ltd. All rights reserved. 0166 - 2236/97/$17.00 PII: S0166-2236(96)01044-2 TINS Vol. 20, No. 7, 1997 303

46 Vaid, R.R. et al. (1996) Brain Res. 733, 31–4047 Gonzalez, H.T., De la Cruz, M. and Mantolan, S.B. (1996)

J. Histochem. Cytochem. 44, 1399–141348 O’Dell, T.J. et al. (1994) Science 265, 542–54649 Dinerman, J.L. et al. (1994) Proc. Natl. Acad. Sci. U. S. A. 91,

2414–241850 Murphy, K.P.S.J. et al. (1994) Neuropharmacology 33, 1375–138551 Bennett, M.R. (1994) Gen. Pharmacol. 25, 1541–155152 Hawkins, R.D., Zhuo, M. and Arancio, O. (1994) J. Neurobiol.

25, 652–66553 Schuman, E.M., Meffert, M.K. and Schulman, H. (1994)

Proc. Natl. Acad. Sci. U. S. A. 91, 11958–1196254 Cummings, J.A., Nicola, S.M. and Malenka, R.C. (1994)

Neurosci. Lett. 176, 110–11455 Hartell, N.A. (1996) Neuron 16, 601–61056 Son, H. et al. (1996) Cell 87, 1015–102357 Kantor, D.B. et al. (1996) Science 274, 1744–174858 Norris, P.J., Faull, R.L.M. and Emson, P.C. (1996) Mol. Brain

Res. 41, 36–4959 Iga, Y. et al. (1993) Eur. J. Pharmacol. 238, 395–39860 Mizutani, A., Saito, H. and Abe, K. (1993) Brain Res. 605,

309–31161 Bannerman, D.M., Butcher, S.P. and Morris, R.G.M. (1994)

Neuropharmacology 33, 1387–139762 Buxton, L.O. et al. (1993) Circ. Res. 72, 387–39563 Macrae, I.M. et al. (1993) J. Cereb. Blood Flow Metab. 13,

985–99264 Babbedge, R.C. et al. (1993) Br. J. Pharmacol. 110, 225–22865 Moore, P.K. et al. (1993) Br. J. Pharmacol. 108, 296–29766 Hara, H. et al. (1996) Neuroscience 75, 881–89067 Doyle, C. et al. (1996) J. Neurosci. 16, 418–42668 Wu, J. et al. (in press)

69 Chapman, P.F. et al. (1992) NeuroReport 3, 567–57070 Hölscher, C. and Rose, S.P.R. (1992) Neurosci. Lett. 145, 165–16771 Hölscher, C. and Rose, S.P.R. (1993) Brain Res. 619, 189–19472 Estall, L.B., Grant, S.J. and Cicala, G.A. (1993) Pharmacol.

Biochem. Behav. 46, 659–96273 Huang, A-M. and Lee, E.H.Y. (1995) Pharmacol. Biochem. Behav.

50, 327–33274 Bernabeu, R. et al. (1995) NeuroReport 6, 1498–150075 Toyoda, M., Saito, H. and Matsuki, N. (1996) Jpn. J. Pharmacol.

71, 205–21176 Ohno, M., Yamamoto, T. and Watanabe, S. (1993) Brain Res.

623, 36–4077 Yamada, K. et al. (1996) Neuroscience 74, 365–37478 Böhme, G.A. et al. (1993) Proc. Natl. Acad. Sci. U. S. A. 90,

9191–919479 Bannerman, D.M. et al. (1994) J. Neurosci. 14, 7404–741480 Saucier, D. and Cain, D.P. (1995) Nature 378, 186–18981 Hargreaves, E.L. and Cain, D.P. (1992) Behav. Brain Res. 47,

23–3382 Cain, D.P. et al. (1996) Behav. Neurosci. 110, 86–10283 Bannerman, D.M. et al. (1995) Nature 378, 182–18684 Cain, D.P. Psychopharmacology (in press)85 Pontecorvo, M.J. et al. (1991) Behav. Neurosci. 105, 521–53586 Hölscher, C. (1995) Learn. Mem. 1, 9–1187 Hölscher, C. et al. (1996) Learn. Mem. 2, 267–27888 Gerlai, R. (1996) Trends Neurosci. 19, 177–18089 Handy, R.L.C. et al. (1995) Br. J. Pharmacol. 116, 2349–235090 Barnes, C.A. (1995) Neuron 15, 751–75491 Nosten-Bertrand, M. et al. (1996) Nature 379, 826–82992 Saucier, D. et al. (1996) Behav. Neurosci. 110, 103–11693 Hölscher, C., Anwyl, R. and Rowan, M. (1997) NeuroReport 8,

451–454

C. Hölscher – NO – the enigmatic messenger RE V I E W

AcknowledgementsThe work wassupported by theIrish HealthResearch Board. Thegenerous supportand help of M.J. Rowan isgratefullyacknowledged, aswell as the help ofH.V. Budgey inediting themanuscript.

Maturation of the mammalian dorsal rootentry zone – from entry to no entry Jon Golding, Derryck Shewan and James Cohen

Interfaces between glial cell precursors of the PNS and CNS are established early in developmentand form the sites where sensory axons enter and motor axons exit the developing CNS.Themolecular and cellular interactions that lead to the formation of these glial interfaces are only nowbecoming apparent. New in-vitro techniques are providing clues as to how the maturation ofPNS–CNS glial interfaces generates barriers to regenerating axons.Trends Neurosci. (1997) 20, 303–308

THE INABILITY OF axons of the mature mammalianCNS to regenerate after injury is believed to be

due in large part to the combined inhibitory effects of its major macroglial elements: oligodendrocytes andastrocytes. Whilst there is now substantial evidencefor an inhibitory role for myelin proteins, synthesizedby oligodendrocytes1, the role of astrocytes remainsrelatively obscure. Part of the problem is that astro-cytes are heterogeneous, both in their morphology andtheir interactions with other cell types of the CNS(Ref. 2). Moreover, their antigenic phenotype, es-pecially after injury, has been poorly characterized.Together these problems have made it difficult tostudy the role of astrocytes in nerve injury. In contrastto the grey and white matter of the CNS, the anatomyof the PNS–CNS interfaces between mammalian spinalnerve roots and the spinal cord is comparatively simple. These sites, where motor and sensory axons,respectively, exit and enter the CNS, contain a uniquecellular apposition between astrocytes and Schwanncells. In the dorsal roots, this interface between the

CNS and PNS is known as the dorsal root entry zone(DREZ). This region has previously been the focus ofstudies related to the failure of lesioned sensory axonsto reconnect with the spinal cord after injury. Thus, ifthe dorsal root is damaged in mature animals, theaxons of primary afferents regenerate successfully inthe Schwann cell-containing PNS portion of theinjured roots, but stop abruptly on contacting theDREZ astrocytes. Two important aspects of the DREZmake it an attractive region for studying the role ofastrocytes in the failure of nerve repair and regener-ation in the CNS. First, the simplicity of the cellularcomposition of the DREZ has facilitated studies ofinteractions between regenerating axons and astro-cytes. Second, Carlstedt showed that, within the firstpostnatal week in rats, injured sensory axons canregenerate through the DREZ and reconnect within thespinal cord3,4. This suggests that development of theDREZ deserves closer scrutiny, since it raises the possi-bility that DREZ astrocytes undergo a developmentaltransition soon after birth in rats and begin to express

Jon Golding,Derryck Shewanand James Cohenare at the Dept ofDevelopmentalNeurobiology,UMDS-Guy’sHospital, London,UK SE1 9RT.

304 TINS Vol. 20, No. 7, 1997

molecules that repel growing axons. What is the basisfor the change in the properties of PNS–CNS glialinterfaces, from a conduit for growing axons in devel-opment, to a ‘barrier’ to their reconnection to themature spinal cord after injury? Here, we first describewhat is known of the origins of glial interfaces in earlydevelopment, and their influence on the formation ofaxon exit and entry points along the neuraxis. Wethen review recent progress in identifying the cellularand molecular components of the DREZ that mightconfer its barrier properties.

The formation of PNS–CNS glial interfaces

Astrocytes and Schwann cells have different embryo-logical origins in vertebrate development: from neuraltube and neural crest, respectively. PNS–CNS glialinterfaces arise at the surface of the neural tube,potentially by a tripartite interaction between cellsderived from the neural crest, the processes of neuro-epithelial cells, and the intervening basal lamina ofthe neural tube. Of these, the contribution of neural-crest cells has been the best characterized. In thechick, Le Douarin et al.5 have shown that a subset ofneural-crest cells, migrating in the ventrolateral path-way alongside the neural tube, take part in the formationof exit–entry points (Fig. 1A,B). This subset is a late-migrating population that selectively expresses c-cad7,a member of the cadherin family of cell-adhesion mol-ecules, initially in the dorsal midbrain at stage 10, andthen progressively further caudally along the neur-axis6. These c-cad7-positive cells contribute to the formation of ‘boundary caps’, delineating the sites ofpresumptive exit–entry points (Fig. 1C). Once theyadhere, neural-crest cells might breach the basal lam-ina of the neural tube by secreting proteases7,8, sincethey are thought to degrade the basal lamina of ecto-derm in this way9. Although these neural-crest cellsare the earliest reported markers of developingexit–entry points, it is possible that restricted domainsof the basal lamina of the neural tube define the siteswhere the migrating c-cad7-expressing neural-crestcells arrest. Thus, the function of c-cad7 might be toaggregate neural-crest cells, ensuring coherent migr-ation to the same loci, but the signals to stop migrationat specific points on the neural tube are possibly medi-ated by distinct adhesive mechanisms10. A furtherpossibility is that specific groups of neuroepithelialcells differentiate at the presumptive exit–entry pointsand degrade the basal lamina by secreting proteasesthemselves or by inducing the attachment of neural-crest cells, or both. Thus, a population of chick neuro-epithelial cells at the prospective ventral exit pointspenetrate the basal lamina of the neural tube at stage17, when the first motor axons emerge from the ventral exit points11. In particular, matrix metallopro-teinases (MMPs) specific for glycoprotein substrates ofthe extracellular matrix (ECM) might be involved,such as the MMP Stromelysin-1, which is expressed byneuroepithelial cells in the chick12.

A recent study further implicates neuroepitheliumin determining positional specification of exit–entrypoints. Niederländer and Lumsden13 excised neuralcrest from rhombomere (r)3, in which an exit–entrypoint does not arise, in a stage-10–11 chick host, andtransplanted age-matched quail neural crest from r4,in which the exit–entry point of the facial nerve nor-mally develops. However, this failed to generate an

inappropriate exit–entry point in r3 of the host,implying that the initial signals for exit–entry pointformation are derived from neuroepithelium and notneural crest.

Glial interfaces, generated as a result of such cellu-lar interactions, can be grouped into three categories.Most Schwann cell–CNS glial interfaces are segregatedinto distinct dorsal (sensory-axon) entry points andventral (motor-axon) exit points (Fig. 1D). However,in some regions of the hindbrain, motor and sensoryaxons exit or enter the CNS at common sites14 (Fig. 1E),whilst in others, distinct exit points for ventral motoraxons are also produced (for example, cranial nervesVI and XII) (Fig. 1F). This pattern is also evident in thecervical spinal cord, where common dorsal exit–entrypoints are produced transiently during development,while distinct ventral motor-axon exit points alsoarise15,16 (Fig. 1F).

Development of PNS–CNS glial interfaces andinteractions with axons

Although the changes in antigenic phenotype asso-ciated with the maturation of CNS glial cells andSchwann cell precursors have been well documented,few studies have focused specifically on developingglia at PNS–CNS interfaces. Boundary-cap cells in thetrunk region of the mouse are the first neural-crestcells at this axial level to express the transcription fac-tor Krox20 [at embryonic day (E)10.5] and subse-quently to become positive for S-100, a marker of dif-ferentiated Schwann cells, by E12.5 (Ref. 17). Earlyphenotypic changes specific to CNS glia at such inter-faces have not been reported. However, in the ratspinal cord, the only definitive astrocytic marker, thecytoskeletal protein GFAP, is first detected in the distal processes of radial glia at the margins of the ventral, then dorsal spinal cord, at E16 and E19 re-spectively18, coincident with the growth of the finalcohort of axons through the glial interfaces in theseregions. Could these studies imply that the Schwanncells and CNS glia that populate the interfaces maturebefore glia elsewhere in the nervous system?Precocious development might be required for appro-priate interactions to take place between the earliestarriving axons and the glial cells at PNS–CNS inter-faces. Thus, it is known that in the case of primaryafferents, there occurs a protracted ‘waiting period’ at the surface of the dorsal grey matter, in the vicinityof the DREZ. Recent work suggests that this stalling of axons is regulated by the local expression of SemIII (D), a member of the semaphorin family19–21. Inthe mouse, neurotrophin 3 (NT3)-dependent musclesensory afferents are the first to grow into the greymatter of the spinal cord at E14.5, to reach their ventral motoneurone targets, coinciding with the ageat which the growth of their axons becomes insensi-tive to the repulsive effects of SemD in tissue culture.In contrast, NGF-dependent small-diameter afferents,whose growth continues to be inhibited by SemD in vitro, grow into their target fields within the super-ficial laminae of the dorsal horn at E17.5, only afterexpression of SemD mRNA in the spinal cord hasreceded ventrally. In addition, another semaphorin,SemA, is expressed in the nerve roots of the mousetrunk between E12.5 and E14.5, and thus by analogy,might also be involved in confining the growth of sensory afferents22.

J. Golding et al. – Axon growth and regeneration across PNS–CNS glial interfacesRE V I E W

TINS Vol. 20, No. 7, 1997 305

Postnatal changes at the DREZ generate a barrierto axon growth

Shortly after birth, changes occur in the organiz-ation of the mammalian DREZ that might be correlatedwith the inability of regenerating primary sensoryaxons to re-enter the spinal cord after injury.Astrocytes extend processes up to 100 mm into thedorsal roots between basal lamina tubes of Schwanncells, and gaps are present in this intervening basallamina at the DREZ (Ref. 23). This organization is notonly thought to confer mechanical strength on theDREZ (Ref. 24), but also increases the surface area ofdirect contact between Schwann cells and astrocytes,generating a unique environment at the interface25.Moreover, this cellular organization ensures thatastrocytic processes are the first CNS elements that areencountered by regenerating primary sensory axons.

In key experiments carried out by Carlstedt3,4 on theinfluence of age on the ability of injured rat sensoryafferents to reconnect with the spinal cord, a ‘criticalperiod’ was identified, between birth and one week,when significant numbers of injured axons were ableto regenerate into the cord. In older animals, regener-ating labelled axons were observed stopping at, orturning back from, the DREZ. Electron microscopicstudies in adult rats have shown that regeneratingaxons stop growing precisely at the astrocyticprocesses within the DREZ (Ref. 26). Growth conesthat contact DREZ astrocytes exhibit ultrastructuralfeatures that are more reminiscent of presynaptic end-ings than advancing growth cones26. These studiessuggested that contact with mature DREZ astrocytesmight activate a ‘physiological stop pathway’26 withinthe neurones, which is proposed to be analogous tothe mechanisms whereby growth cones stop at theirappropriate targets27,28.

Further evidence for the role of mature astrocytes inpreventing regenerating axons (and possibly Schwanncells) from crossing the DREZ is also provided byexperiments in which the dorsal spinal cord of the ratis depleted of glia by X-irradiation soon after birth.This treatment generates gaps in the glial limitansthrough which Schwann cells migrate ectopically intothe spinal cord along astrocyte- and oligodendrocyte-free sensory afferents29. Invasion by Schwann cells islimited to the glia-depleted areas of the CNS and thefew remaining astrocytes appear to block the furthermigration of Schwann cells. Examination of dorsalroot lesions that had been generated two weeks after X-irradiation revealed that several regeneratingsensory axons entered the spinal cord through theastrocyte-free DREZ (Ref. 30).

What is known of the identity of the molecules thatmight contribute to the DREZ barrier? Tenascin andsulphated proteoglycans have been implicated asinhibitors of axon growth that are expressed by DREZastrocytes31 and in other regions of the developingCNS (Refs 32,33). Thus, Silver et al. reported that bothtenascin and sulphated proteoglycans become con-centrated at the CNS side of the rat DREZ towards theend of the critical period31, although a later study indi-cated that tenascin mRNA and protein are stronglyexpressed by rat DREZ astrocytes from birth34.Following injury to the dorsal root in rats older thanthe critical age, these molecules become highly con-centrated at the DREZ and within proximal regions ofthe dorsal horn31, in association with extensively

branched ‘reactive’ astrocytes35. Reactive astrocytes area major component of the glial scars that are formedaround CNS lesions in adult rats, and through whichaxons fail to regenerate1,36.

In vitro, myelin-free plasma membranes isolatedfrom glial scars in lesioned brains of adult rats havebeen shown to inhibit the growth of neurites fromdorsal root ganglia, and septal and hippocampal

J. Golding et al. – Axon growth and regeneration across PNS–CNS glial interfaces RE V I E W

A

C

D E F

B

Neuraltube

Basal lamina

Developingsensoryganglion

Developingexit points

Neural-crest cellc-cad7-expressingneural-crest cellNeuroepithelial cellMotor neurone

Hindbrain or cervicalspinal cord

HindbrainTrunkspinal cord

Fig. 1. Development of PNS–CNS glial interfaces. (A,B) Schematictransverse sections through the neural tube that illustrate the formationof the PNS–CNS glial interface. The earliest known marker of prospectivePNS–CNS glial interfaces is the cadherin c-cad7. A subpopulation of c-cad7-expressing neural-crest cells migrate as a group from the dorsalsurface of the neural tube (A) to form boundary caps, the sites wherePNS–CNS glial interfaces ultimately develop (B). Neural-crest cells, neuroepithelium, the basal lamina of the neural tube and axons (someprojecting only transiently, dotted line) can all contribute to the formation of glial interfaces, as summarized in B. (C) In a transversesection through the chick hindbrain at stage 16, in-situ hybridizationidentifies a c-cad7-expressing subpopulation of neural-crest cells thatare forming boundary caps. Scale bar, 50 mm. Photo courtesy of C. Niederländer. (D) In the trunk regions, glial interfaces are segre-gated into distinct dorsal (sensory-axon) entry and ventral (motor-axon) exit points. (E,F) However, at some levels of the hindbrain, common dorsal exit–entry points for branchial motor and sensory axonsare formed (E) whilst at other hindbrain levels and in the cervical spinalcord, common dorsal exit–entry points as well as ventral exit points areformed (F).

306 TINS Vol. 20, No. 7, 1997

neurones of embryonic rat, whilst membranes isolatedfrom uninjured brains supported neurite outgrowth37.This growth-inhibitory effect could be removed bypretreatment with proteoglycan-degrading enzymes37.Furthermore, purified chondroitin-sulphate proteo-glycans38 and tenascin39 have been shown to act asbarriers to growth of CNS neurites in vitro when pre-sented as sharp substrate boundaries. However, theincreased expression of inhibitory molecules is onlyone possible mechanism that might account for thefailure of axon growth at the mature DREZ. Anotherpossibility is that maturation of astrocytes, both invivo and in vitro, might downregulate the productionof cell-adhesion molecules that promote axongrowth40, in tandem with the upregulation of barriermolecules.This raises the question of whether matu-rational changes that are intrinsic to normal astro-cytes are themselves responsible for the acquisition of barrier properties at the DREZ.

Culture model systems enable the interaction ofaxons with the DREZ to be studied in novel ways

To test whether maturing astrocytes acquire theability to inhibit axon growth, it is necessary to con-front growing primary sensory axons with the unin-jured DREZ, a scenario that is impossible in vivo.However, this has been made possible by adapting anin-vitro cryoculture approach41–43. In this technique,sensory neurones are cultured on thin cryostat sec-tions of nerve tissue, an environment that closely rep-

resents the one that they wouldencounter in vivo. The ECM andcell-surface molecules are pre-served, whilst cells within thefrozen tissue are non-viable and areno longer able to secrete solublefactors. This allows the differentialeffects of substrate and soluble factors to be studied. By preparinglongitudinal cryostat sections ofthe dorsal spinal cord of the rat, ithas been possible to incorporatethe DREZ and attached dorsal rootsand use these as substrates for cultures of dissociated neurones44

of the dorsal root ganglia (Fig. 2).Neurones that adhere to the dorsalroots extend neurites along thebasal lamina tubes of Schwanncells towards the DREZ, as theywould in vivo, where they mightgrow across to the spinal cord,stop, or turn back along the dorsalroot. A major advantage of thisapproach is that DREZ from various developmental stages, bothbefore and after the hypothesizedcritical period, can be employed assubstrates, enabling us to studymore closely how the developmentof this region influences neuritegrowth. Thus, we have found thatneurites growing from neurones ofneonatal dorsal root ganglia crossthe newborn (postnatal day 0, P0)DREZ more readily than P6 or adultDREZ (Fig. 3A,B). This supports the

idea that inhibitors of axon growth appear at theDREZ during a critical period within the first postnatalweek3,4 that is independent of injury-inducedresponses by glial cells. This might indicate that thesulphated proteoglycans that accumulate at the DREZby P6 are sufficient to halt growing primary sensoryaxons, but the possibility remains that other, as yetuncharacterized, molecules might also be involved.Significantly, neurite outgrowth on the spinal cord ordorsal root, immediately central or peripheral to theDREZ, was similar at different ages that encompassedthe critical period, suggesting that changes intrinsic tothe DREZ are initially responsible for developing a barrier to regenerating axons. By adulthood, both theDREZ and the CNS adjacent to it were poor substratesfor outgrowth of neurites of the neonatal dorsal rootganglia, suggesting that after the critical period ad-ditional inhibitory molecules become expressed generally within the CNS.

Influence of age of neurones on their interactionswith the DREZ

By testing a range of ages of neurones in cryocul-ture, we were also able to analyse separately the influ-ence of development of neurones of the dorsal rootganglia on the ability of neurites to cross the DREZ.We found that neurites from early embryonic neur-ones were less sensitive to the P6 DREZ inhibitors44

(Fig. 3C) than neurites from more mature neurones,suggesting that, in tandem with changes at the DREZ

J. Golding et al. – Axon growth and regeneration across PNS–CNS glial interfacesRE V I E W

A Cryoculture B

C

D

CryosectionPlane ofsection

Dorsalroots

Dorsal rootentry zone

(DREZ)

DissociatedDRG neurones

Cryosection

Neonatal DRG neuroneon adult DREZ cryosection after immunolabelling

Rostral

Grey matter Ventral root

Fig. 2. Axon interactions with the dorsal root entry zone (DREZ) can be modelled in vitro, using cryoculture. In thistechnique, uninjured spinal cord, taken from rats of various ages (A), is cut into thin (8–10 mm) longitudinal sectionson a cryostat (B). These sections are transferred to sterile glass coverslips and are used as substrates for the growth ofneurites from dissociated neurones of rat dorsal root ganglia (DRG) (C). After 18 h, the cultures are fixed and stained(D) with antibodies against: laminin to label the basal lamina tubes of the Schwann cells of the dorsal roots (in green),GFAP to label astrocytes within the CNS (in blue) and GAP-43 to label growing neurites and their cell bodies (in red).Neurites from neurones of postnatal DRG grow well along the dorsal roots, but stop at the PNS–CNS interface of theadult DREZ, mimicking closely the behaviour of regenerating axons after a prior lesion in vivo. Abbreviation: P0, new-born, postnatal day 0.

TINS Vol. 20, No. 7, 1997 307

during the critical period, there are correspondingmaturational changes in the expression of axonalreceptors for ligands that influence growth. A parallel

approach in vivo involves transplanting allografts ofembryonic dorsal root ganglia into adult rats. In theseanimals some immature axons were found to haveentered the spinal cord of the adult host45, although itis unclear whether they actually grew through theDREZ. Both studies are consistent with an emerginggeneral principle that immature neurones are betterable to extend axons within the mature CNS environ-ment46–50, possibly as a result of their lack of receptorsthat recognize inhibitory ligands. The complementaryfindings that have been obtained from both in-vivoand in-vitro studies on maturation of the DREZ aresummarized in Fig. 4.

J. Golding et al. – Axon growth and regeneration across PNS–CNS glial interfaces RE V I E W

DR injurysite

Reactivegliosis

DREZ

Some injured DRGaxons regeneratethrough the DREZ

Regenerating DRGaxons fail to cross

the DREZ

In vivo

Cryoculture model

Neonatal rhizotomy

DR injurysite

Mature rhizotomy

P P

E E

Oligodendrocytes

Postnatal (P)or embryonic (E)DRG neurones

CNS DRDREZ CNS DRDREZ

Neonatal cryosection Mature cryosection

Astrocyte(growth permissive)

Astrocyte processes extend into DR,increasing the width of the DREZ,

and become growth inhibitory

DRG DRG

Fig. 3. An in-vitro model of the dorsal root entry zone (DREZ) repli-cates in-vivo observations. The critical period for regeneration of pri-mary sensory axons across the DREZ in vivo might be dependent upondevelopmental changes that occur both at the DREZ and within neur-ones of dorsal root ganglia (DRG). Cryosections of uninjured spinalcord, incorporating dorsal roots and DREZ from newborn (P0) (A) orpostnatal day 6 (P6) (B) rats were used as substrates for the growth ofneurites from DRG neurones of the P0 rat. Neurites (labelled red withanti-GAP-43 antibody and arrowed in A) consistently cross the P0 DREZmore successfully than the P6 DREZ (dotted line, between anti-laminin-stained endoneural tubes of Schwann cells and the spinal cord), whiletheir growth on P0 and P6 spinal cord adjacent to the DREZ is similar.Because the DREZ substrates are taken from uninjured animals, thissuggests that during the first postnatal week, changes intrinsic to theDREZ generate an environment that is inhibitory for axon growth.Although the P6 DREZ inhibits neurite crossing by postnatal DRG neur-ones, neurites from embryonic (E15) DRG neurones are less sensitive to thisbarrier (arrow in C), suggesting that developmental changes in neuronalreceptors are also important for the generation of a barrier to growingaxons at the DREZ. Scale bars, 70 mm (A and B), 50 mm (C).

Fig. 4. In vivo and in vitro experimental models of the rat dorsal root entry zone (DREZ).Both approaches demonstrate a critical neonatal period, between birth and 1 week post-natally, during which regenerating primary sensory axons of dorsal root ganglia (DRG) can cross the DREZ. This might be due to maturational or injury-induced changes at the DREZ,or both. As the rat matures postnatally, astrocytic processes (but not oligodendrocytes) extendinto the dorsal root (DR) (bottom panels), expanding the region of contact between astrocytesand Schwann cells (which defines the DREZ). Molecules that inhibit axon growth mightbecome concentrated at these contact sites within the DREZ, as indicated by the change in theDREZ astrocyte surface from green (growth permissive) to red (growth inhibitory). A conse-quence of DR injury (rhizotomy) after the critical age is a marked gliosis at the DREZ. Moleculesthat inhibit axon growth might become associated with these reactive glia (top panels). Byusing uninjured spinal cord cryosections to model the DREZ in vitro, it is possible to examinehow DRG axons behave at the neonatal and adult DREZ in the absence of reactive gliosis. Aswith the in-vivo studies, postnatal DRG neurones (P) fail to grow neurites across the matureDREZ, suggesting that maturational changes are primarily involved in generating a barrier toaxon growth at the end of the critical period. When embryonic DRG neurones (E) are appliedto the cryosections, they are able to extend neurites across both the neonatal and matureDREZ, suggesting that they have yet to acquire functional receptors for the inhibitory DREZ ligands.

308 TINS Vol. 20, No. 7, 1997

Future prospectsCurrent knowledge on the provenance of exit and

entry points of the neural tube is limited, but recentwork suggests that cells of neuroepithelial origin,rather than of the neural crest, determine the siteswhere these points form in development. Furtherinformation on the molecular and structural proper-ties of these immature glial environments shouldenlighten us as to the optimal conditions for axongrowth across such interfaces.

The mechanisms that underlie the failure of injuredaxons to regenerate across mature interfaces remainequally elusive. Complementary in-vivo and in-vitrostudies that focus on the DREZ offer the prospect ofcircumventing some of the complexity encounteredelsewhere in the CNS, and provide new leads as to theidentity of the molecules responsible. Until recently,the most likely mechanism involved changes in thecomposition of the ECM in the vicinity of the DREZ,which were proposed to be instrumental in blockingthe reconnection of lesioned primary afferent axonswith the spinal cord. However, the spatiotemporalpattern of expression of SemD in the developing spinalcord, and its selective repulsive effects on ingrowth ofsensory afferents, have been invoked more recently to explain the patterning of primary afferent inner-vation20. The effects of injuries to the mature CNS onpossible re-expression of SemD within the DREZ anddorsal horn, and of its cognate receptor on sensoryaxons, remain to be determined, but might also belinked to the failure of afferent re-innervation. Similarpossibilities are raised by the recent demonstration ofthe key role played by the Eph-receptor family andtheir ligands in regulating the establishment of topo-graphic maps in CNS development51, again mediatedby selective repulsive effects on growing axons. Thecontinuing elucidation of complementary expressionpatterns of Eph receptors and their ligands through-out the nervous system52,53 might help to providesome of the answers as to why PNS–CNS interfaceschange from entry to no-entry zones.

Selected references1 Schwab, M.E. and Bartholdi, D. (1996) Physiol. Rev. 76, 319–3702 Wilkin, G.P., Marriot, D.R. and Cholewinski, A.J. (1990)

Trends Neurosci. 13, 43–463 Carlstedt, T. et al. (1987) Neurosci. Lett. 74, 14–184 Carlstedt, T. (1988) J. Neurocytol. 17, 335–3505 Le Douarin, N.M. et al. (1992) in Sensory Neurons: Diversity,

Development, and Plasticity (Scott, S.A., ed.), pp. 143–170, OxfordUniversity Press

6 Nakagawa, S. and Takeichi, M. (1995) Development 121,1321–1332

7 Carroll, P.M. (1994) Development 120, 3173–31838 Valinsky, J.E. and Le Douarin, N.M. (1985) EMBO J. 4,

1403–14069 Erickson, C.A. et al. (1992) Dev. Biol. 151, 251–272

10 Lallier, T. et al. (1992) Development 116, 531–54111 Lunn, E.R. et al. (1987) Development 101, 247–25412 Nordström, L.A. et al. (1995) Mol. Cell. Neurosci. 6, 56–6813 Niederländer, C. and Lumsden, A. (1996) Development 122,

2367–237414 Lumsden, A. and Keynes, R. (1989) Nature 337, 424–42815 Quebada, P. and Ignatius, M.J. (1995) Soc. Neurosci. Abstr. 21,

150916 Tanaka, H. (1991) J. Comp. Neurol. 303, 329–33717 Murphy, P. et al. (1996) Development 122, 2847–285718 Yang, H-Y. et al. (1993) J. Neurocytol. 22, 558–57119 Messersmith, E.K. et al. (1995) Neuron 14, 949–95920 Püschel, A.W., Adams, R.H. and Betz, H. (1996) Mol. Cell.

Neurosci. 7, 419–43121 Wright, D.E. et al. (1995) J. Comp. Neurol. 361, 321–33322 Püschel, A.W., Adams, R.H. and Betz, H. (1995) Neuron 14,

941–94823 Berthold, C-H. and Carlstedt, T. (1977) Acta Physiol. Scand.

Suppl. 446, 23–4224 Livesey, F.J. and Fraher, J.P. (1992) Neuropathol. Appl.

Neurobiol. 18, 376–38625 Ghirnikar, R.S. and Eng, L.F. (1995) Glia 14, 145–15226 Liuzzi, F.J. and Lasek, R.J. (1987) Science 237, 642–64527 Liuzzi, F.J. and Tedeschi, B. (1992) J. Neurosci. 12, 4783–479228 Liuzzi, F.J. (1990) Brain Res. 512, 277–28329 Sims, T.J. and Gilmore, S.A. (1983) Brain Res. 276, 17–3030 Sims, T.J. and Gilmore, S.A. (1994) Brain Res. 634, 113–12631 Pindzola, R.R., Doller, C. and Silver, J. (1993) Dev. Biol. 156,

34–4832 Steindler, D.A. et al. (1989) Dev. Biol. 131, 243–26033 Snow, D.M., Steindler, D.A. and Silver, J. (1990) Dev. Biol. 138,

359–37634 Zhang, Y. et al. (1995) J. Neurocytol. 24, 585–60135 Bignami, A., Chi, N.H. and Dahl, D. (1984) Exp. Neurol. 85,

426–43636 Reier, P.J., Stensaas, L.J. and Guth, L. (1983) in Spinal Cord

Reconstruction (Kao, C.C., ed.), pp. 163–195, Raven Press37 Bovolenta, P., Wandosell, F. and Nieto-Sampedro, M. (1993)

Eur. J. Neurosci. 5, 454–46538 Snow, D.M. et al. (1991) Development 113, 1473–148539 Taylor, J., Pesheva, P. and Schachner, M. (1993) J. Neurosci.

Res. 35, 347–36240 Smith, G.M., Jacobberger, J.W. and Miller, R.H. (1993)

J. Neurochem. 60, 1453–146641 Carbonetto, S., Evans, D. and Cochard, P. (1987) J. Neurosci. 7,

610–62042 Sandrock, A.W., Jr and Matthew, W.D. (1987) Proc. Natl. Acad.

Sci. U. S. A. 86, 6934–693843 Shewan, D. et al. (1994) Neuroprotocols 4, 142–14544 Golding, J.P. et al. (1996) Mol. Cell. Neurosci. 7, 191–20345 Rosario, C.M. et al. (1993) Exp. Neurol. 120, 16–3146 Davies, S.J., Field, P.M. and Raisman, G. (1993) Eur. J. Neurosci.

5, 95–10647 Li, Y. and Raisman, G. (1993) Brain Res. 629, 115–12748 Shewan, D., Berry, M. and Cohen, J. (1995) J. Neurosci. 15,

2057–206249 Wictorin, K. et al. (1990) Neuroscience 37, 301–31550 Wictorin, K. and Björklund, A. (1992) NeuroReport 3,

1045–104851 Cheng, H-J. et al. (1994) Cell 82, 371–38152 Gale, N.W. et al. (1996) Neuron 17, 9–1953 Kolodkin, A.L. (1996) Trends Neurosci. 19, 507–513

J. Golding et al. – Axon growth and regeneration across PNS–CNS glial interfacesRE V I E W

AcknowledgementsWe are grateful to

ChristianneNiederländer for

Fig. 1C and thankAnthony Graham

and Kuldip Bedi fortheir helpful

comments on themanuscript. This

work was supportedby Action Research

and the MedicalResearch Council.

Letters to the EditorLetters to the Editor concerning articles published recentlyin TINS are welcome. Please mark clearly whether they areintended for publication; the authors of the article referredto are given an opportunity to respond to any of the pointsmade in the Letter. Maximum length: 500 words. The Editorreserves the right to edit Letters for publication.

Please address Letters to:Dr Gavin Swanson, Editor, Trends in Neurosciences,

68 Hills Road, Cambridge, UK CB2 1LA.

Book ReviewsPublishers

Trends in Neurosciences welcomes books for review.Please send books or book details to:

Dr Gavin Swanson, Editor, Trends in Neurosciences68 Hills Road, Cambridge, UK CB2 1LA.

ReviewersIf you are interested in reviewing books for TINS, please con-tact the Editor, Dr Gavin Swanson, with your suggestions.

Tel: +44 1223 315961; Fax: +44 1223 464430.