amino acid requirements for formation of the tgf-β-latent tgf-β binding protein complexes

doi:10.1016/j.jmb.2004.10.039 J. Mol. Biol. (2005) 345, 175–186

Amino Acid Requirements for Formation of theTGF-b-Latent TGF-b Binding Protein Complexes

Yan Chen1†, Tariq Ali2†, Vesna Todorovic1, Joanne M. O’Leary2

A. Kristina Downing2 and Daniel B. Rifkin1,3*

1Department of Cell BiologyNew York University School ofMedicine, 550 First AvenueNew York, NY 10016, USA

2Division of Structural BiologyDepartment of BiochemistryUniversity of Oxford, SouthParks Road, Oxford OX1 3QUUK

3Department of MedicineNew York University School ofMedicine, 550 First AvenueNew York, NY 10016, USA

0022-2836/$ - see front matter q 2004 E

† Y.C. and T.A. contributed equalAbbreviations used: TGF-b1, tran

factor b1; LLC, large latent complexbinding protein; LAP, latency assocfluorescence resonance energy transE-mail address of the correspond

[email protected]

Transforming growth factor b (TGF-b) is secreted primarily as a latentcomplex consisting of the TGF-b homodimer, the TGF-b propeptides(called the latency-associated protein or LAP) and the latent TGF-b bindingprotein (LTBP). Mature TGF-b remains associated with LAP by non-covalent interactions that block TGF-b from binding to its receptor.Complex formation between LAP and LTBP is mediated by an intra-molecular disulfide exchange between the third 8-cysteine (8-Cys3) domainof LTBP with a pair of cysteine residues in LAP. Only the third 8-Cysdomains of LTBP-1, -3, and -4 bind LAP. From comparison of the 8-Cys3LTBP-1 structure with that of the non-TGF-b-binding 8-Cys6fibrillin-1, weobserved that a two-residue insertion in 8-Cys3LTBP-1 increased thepotential for disulfide exchange of the 2–6 disulfide bond. We furtherproposed that five negatively charged amino acid residues surroundingthis bond mediate initial protein–protein association.

To validate this hypothesis, we monitored binding by fluorescenceresonance energy transfer (FRET) analysis and co-expression assays withTGF-b1 LAP (LAP-1) and wild-type and mutant 8-Cys3 domains. FRETexperiments demonstrated ionic interactions between LAP-1 and 8-Cys3.Mutation of the five amino acid residues revealed that efficient complexformation is most dependent on two of these residues.

Although 8-Cys3LTBP-1 binds proTGF-bs effectively, the domain fromLTBP-4 does so poorly. We speculated that this difference was due to thesubstitution of three acidic residues by alanine, serine, and arginine in theLTBP-4 sequence. Additional experiments with 8-Cys3LTBP-4 indicated thatenhanced binding of LAP to 8-Cys3LTBP-4 is achieved if the residues A, S,and R are changed to those in 8-Cys3LTBP1 (D, D, and E) and the QQdipeptide insertion of LTBP-4 is changed to the FP in 8-Cys3LTBP-1. Thesestudies identify surface residues that contribute to the interactions of8-Cys3 and LAP-1 and may yield information germane to the interaction of8-Cys domains and additional TGF-b superfamily propeptides, anemerging paradigm for growth factor regulation.

q 2004 Elsevier Ltd. All rights reserved.

Keywords: transforming growth factor-b1; large latent complex; latenttransforming growth factor-b binding protein; 8-cysteine/CR/TB domain;nuclear magnetic resonance
*Corresponding author
lsevier Ltd. All rights reserve

ly to this work.sforming growth; LTBP, latent TGF-biated protein; FRET,fer; CR, cysteine-rich.ing author:

Introduction

Growth factor action in the extracellular milieu isregulated by multiple mechanisms that includesequestration, transmembrane attachment, inter-action with binding proteins, and release in anunprocessed pro form.1 For the three transforminggrowth factor b (TGF-b) isoforms, TGF-b1, -b2,and -b3, release as inactive complexes is critical toinsure proper action.2 Most latent TGF-b is secreted

d.

176 Residues Involved in Large Latent Complex Assembly

as a large latent complex (LLC) consisting of themature TGF-b dimer, the TGF-b dimeric propeptideand a second gene product called the latent TGF-bbinding protein (LTBP)2,3 (Figure 1(A)). Althoughthe peptide bond between TGF-b and its propeptideis cleaved within the cell, the TGF-b propeptideremains bound to TGF-b by non-covalent inter-actions that prevent TGF-b from binding to itshigh affinity plasma membrane receptor.4–6 For thisreason, the TGF-b propeptide is also called thelatency-associated protein (LAP). The third com-ponent of the LLC, LTBP, is covalently disulfidebound to LAP via a pair of cysteine residues in theN-terminal region of LAP and a cysteine pair inLTBP (Figure 1(A)).7,8

The significance of LLC formation in regulatingthe action of TGF-b has been demonstrated byseveral approaches. For example, transgenic micethat constitutively express active TGF-b in theirlungs develop severe pulmonary fibrosis, whereascompanion mice that express wild-type proTGF-bunder the control of the same lung-specific pro-moter develop no fibrosis.9 In addition, mice withnull (LTBP-3) or hypomorphic (LTBP-4) mutations

Figure 1. Schematic illustration of the LLC and amino acid sand -4. In (A) the domain organization of LTBP-1 is shown, wiin the key. The LLC comprises three molecules: an LTBP, a LAis stabilized by covalent disulfide cross-links. In addition, dishown. In (B), the positions of amino acid residues predicted tand LAP based on the structure of 8-Cys3LTBP-1

19 are numdipeptide insertion required for LTBP–LAP complex formatioSequences were aligned using CLUSTAL_X with default par

in LTBP genes display phenotypes previouslydescribed for transgenic animals or humans thathavedecreasedTGF-b signaling in bone or colon.10–14

Finally, LTBP binding to LAP appears to be requiredfor effective latent TGF-b secretion and activationby certain cultured cells.15,16

The four LTBP isoforms (1–4) vary in their ability tocomplexwith LAP. LTBP-2 does not form complexes;LTBP-4 forms complexes poorly; and LTBP-1 and -3readily form complexes (see Table 1). The LTBPisoforms are similar in structural organization asthey contain multiple EGF-like repeats and signa-ture domains called 8-cysteine (8-Cys) domains.17

These later domains are also calledCR (cysteine-rich)orTB (latentTGF-bbindingprotein-likedomains).Asthe name suggests, 8-Cys domains contain eightcysteine residues that are intramolecularly linked.The LTBPs contain four 8-Cys domains, although theN-terminal 8-Cys is divergent in sequence and mayhave evolved from the splicing of an EGF-like and an8-Cys domain.3 8-Cys domains also occur in thefibrillins, which are similar to the LTBPs in domainorganization, but are larger.3 However, like LTBP-2,the fibrillins do not bind LAP.18

equence alignment of 8-Cys3 domains from LTBP-1, -2, -3th symbols corresponding to each protein module definedP homodimer, and a TGF-b homodimer. Each homodimersulfide bonds form between the LTBP and LAP dimer aso be important for disulfide bond exchange between LTBPbered and indicated by red arrows. The position of then is highlighted by white lettering on a black background.ameters.37

Table 1. LTBP1-4 comparison

Residuenumber 5 12 17 39 42 44–45

TGF-b1binding

LTBP-1: E D D D E FP CCCLTBP-2: D G S D D – –LTBP-3: E D D D E YP CCCLTBP-4: E A D S R QQ K or C

Results in this Table are based on previously published observations.18

Residues Involved in Large Latent Complex Assembly 177

We recently solved the solution structure ofthe LAP-binding 8-Cys domain 3 (8-Cys3) fromLTBP-1.19 This domain has a globular structure andcontains six b-strands and two a-helices. The eightcysteine residues bond in a 1–3, 2–6, 4–7, 5–8pattern, and these disulfide bonds stabilize theoverall domain conformation. The structure of the8-Cys3LTBP-1 domain is similar to the structure ofthe 8-Cys6fibrillin-1,

20 with one major exception. Theinsertion of an FP dipeptide in the 8-Cys3LTBP-1sequence between cysteine residues 6 and 7displaces the 2–6 cysteine pair into the solution byprojecting one of the b-strands.19 We suggested thatdisulfide bond formation between LAP and 8-Cys3domains involves the 2–6 disulfide pair based uponits increased surface accessibility.

The solution structure of 8-Cys3 also revealed aring of five negatively charged amino acid residuessurrounding the proposed reactive 2–6 disulfide, aswell as two flanking regions with hydrophobicsurface character.19 These observations suggestedthat the initial bonding between LAP and 8-Cys3domains is mediated by electrostatic interactionsand that this association is followed by intermole-cular disulfide bond formation.

Modeling studies of 8-Cys3 from LTBP-2, -3, and-4 using the solved 8-Cys3LTBP-1 and 8-Cys6fibrillin-1structures indicated that the TGF-b-bindingdomains of both LTBP-1 and LTBP-3 have fivenegatively charged residues surrounding the 2–6disulfide bond, whereas the 2–6 disulfide bondfrom the weakly binding 8-Cys3LTBP-4 is sur-rounded by only three acidic amino acid residues.19

Thus, the ability of the 8-Cys3 domains to complexwith LAP correlates with the amount of chargesurrounding the predicted labile disulfide. Thepositions of the important acidic residues arehighlighted in the multiple sequence alignmentshown in Figure 1(B). For simplicity, here aminoacid residue numbering is described with respect tothis sequence alignment. With reference to theamino acid residue numbering of human LTBP-1,21

E5 corresponds to E1021, and for human LTBP-4, E5corresponds to E1069.22

To better define the requirements for interactionof LAP and the LTBPs, we conducted a series ofmutagenesis and binding experiments based uponthe solution structure of 8-Cys3LTBP-1. Here wepresent data from fluorescence resonance energytransfer (FRET) analysis proving that non-covalentionic interactions exist between 8-Cys3LTBP-1 andTGF-b1 LAP (LAP-1) in solution. We further

describe mutation and co-expression assays thatdelineate the relative contributions of the negativelycharged amino acid residues that surround the 2–6disulfide bond to the efficiency of LAP-1 binding to8-Cys3LTBP-1. Finally, we demonstrate how alteringthese charged amino acid residues modulates thereactivity of the 8-Cys3LTBP-4 isoform bonding toLAP-1. On the basis of these results, we speculateon how protein–protein interactions of the pro-peptides of other TGF-b superfamily members with8-Cys may contribute to the regulation of theiractivity.

Results

Non-covalent association of 8-Cys3LTBP-1 andLAP-1 is electrostatically mediated

To examine non-covalent interactions of between8-Cys3LTBP-1 and LAP-1C33S, FRET from the trypto-phan in 8-Cys3LTBP-1 to Alexa Fluor-350-labeledlysine residues in LAP-1C33S was utilized. FRETwasemployed because it provides information onmolecular interactions occurring at distances of10–100 A. The C33S LAP mutation was usedbecause this substitution removes the cysteineresidues involved in the covalent interactionLAP-1 with 8-Cys3LTBP-1. This mutation was intro-duced to limit our observations to non-covalentinteractions and to improve the yield of recombin-ant LAP-1C33S. The C33S mutation does not affectthe biological activity of LAP-1 based on the abilityof the mutant protein to inhibit TGF-b function.23

8-Cys3LTBP-1, which contains a single tryptophanresidue, was labeled with the dye, and emissionspectra were recorded with the concentrations ofNaCl, LAP-1C33S and u8-Cys3LTBP-1 (unlabeled byAlexa Fluor-350) treated as adjustable parameters.A subset of the data is summarized in Figure 2,illustrating the following points. The intensity ofthe FRET emission is enhanced by the presence ofLAP-1C33S. The change in intensity demonstrates abinding interaction between the two molecules.This interaction is effectively competed by theaddition of NaCl. Data are shown only for theaddition of 1 M NaCl; however, NaCl concen-trations as low as 50 mM had the same effect.The addition of 1 M NaCl to either 8-Cys3LTBP-1 orLAP-1C33S does not alter their respective emissionspectra (data not shown). The interaction also canbe competed by the addition of excess unlabeled

Figure 2. FRETanalysis of 8-Cys3LTBP-1 interaction withLAP-1C33S. Fluorescence intensity (measured at themaxima, lZ439 nm, nZ4) corresponding to fluorescenceenergy transfer from the tryptophan of 8-Cys3LTBP-1 toAlexa Fluor-350 attached on lysine residues of LAP-1C33S.An increase in intensity corresponding to protein–proteininteraction is observed upon the addition of LAP-1C33S to8-Cys3LTBP-1. This interaction is competed by the additionof NaCl or unlabeled 8-Cys3LTBP-1, indicating that theinteraction is electrostatic and heteromeric in nature. Keyto experiment numbers: 1, 8-Cys3LTBP-1; 2, LAP-1; 3, 8-Cys3LTBP-1 plus LAP-1; 4, 8-Cys3LTBP-1 plus LAP-1 plus1 M NaCl; 5, 8-Cys3LTBP-1 plus LAP-1 plus unlabeled8-Cys3LTBP-1; 6, 8-Cys3LTBP-1 plus 1 MNaCl; 7, LAP-1 plus1 M NaCl.


8-Cys3LTBP-1. The sum of these results indicatesthat there exists a non-covalent association between8-CysLTBP-1 and LAP-1C33S that contains an essen-tial electrostatic component. This interaction mayprecede the formation of disulfide bonds between8-Cys3 and LAP-1.

NMR analysis of 8-Cys3LTBP-1 mutants

To determine the relative importance of each ofthe negatively charged residues surrounding the2–6 disulfide bond for interaction with LAP-1, aseries of mutant 8-Cys3LTBP-1 constructs, designatedMUT1–MUT12 (Table 2), were produced in whichone or more of the five negatively charged aminoacid residues were replaced with neutral orpositively charged residues, respectively. Themutations were introduced into the constructEGF-8Cys3-EGF-2HA (E8CYS3EHA)† in whichthe flanking EGF groups enhance the productionof 8-Cys3LTBP-1. However, before these constructswere utilized in co-expression assays with proTGF-b1, we probed by NMR the in vitro refoldingproperties of selected mutant domains relative to

† Proteins are referred to as E8-Cys3E, and nucleic acidconstructs are referred to as E8CYS3E. Unless otherwisespecified, these constructs are derived from the LTBP-1isoform.

the wild-type domain. MUT8 and MUT12 werechosen as representative, because they contain themost extreme mutagenesis, in which all five acidicresidues are changed to alanine or lysine, respec-tively. MUT11 was later included based uponresults for MUT12. The mutant proteins wereexpressed as isolated 8-Cys3LTBP-1 domains inhigh yield and the structures were assessed byNMR. Both MUT8 and MUT11 spectra retained aproportion of well-dispersed peaks in similarpositions to those in the spectrum of wild-typeprotein (Figure 3). MUT12 did not refold in vitrobased on NMR analysis (data not shown), and amuch lower level of MUT12 E8-Cys3EHA wasdetected by anti-HA, suggesting a problem withsecretion (Figure 5, lane 9). Therefore, MUT12 wasnot studied further. Because MUT8 and MUT11retained the ability to refold into a native-likeconformation, we assumed in subsequent experi-ments that the reduced levels of complex formed byall of the constructs apart from MUT12 relateddirectly to changes in electrostatic potential sur-rounding the 2–6 disulfide bonds invoked by themutagenesis. This conclusion is also supported bythe general expression levels of the mutant con-structs detected using anti-HA (Figures 4–7).

Mutation of 8-Cys3LTBP-1 residue D39 inhibitsdisulfide bonding to LAP-1

The contribution of the five negatively chargedamino acid residues E5, D12, D17, D39, E42 (seeFigure 1(B)) in 8-Cys3LTBP-1 to complex formationwith LAP-1 was explored using co-expression.Neutral single amino acid mutant constructs ofE8CYS3EHA (MUT1-5; Table 2) were co-expressedwith proTGF-b1. The conditioned media fromthe transfected cells were analyzed by Westernblotting utilizing either anti-HA antibody to revealE8-Cys3EHA or monoclonal antibody VB3A9 toreveal LAP. Preliminary experiments in which theamount of 8-Cys3 DNA used for transfections wasvaried demonstrated that at the DNA concen-trations used, 8-Cys3 protein was always in excessover proTGF-b (data not shown). Therefore,changes in the intensity of the product band wereused to monitor differences in the ability of eachconstruct to form complex. The results indicate thatall of the E8-Cys3EHA proteins with single alaninesubstitutions retained the ability to form covalentcomplexes with LAP visualized by both anti-HA(Figure 4(A)) and anti-LAP-1 (Figure 4(B)) staining.The position of wild-type E8-Cys3EHA and itscomplex with LAP-1 was equivalent to thatobserved with the E5A, D12A, D17A, and E42Amutants (Figure 4(B)). The D39A protein (MUT4,lane 5) complexed less readily than did the E5A,D12A, D17A and E42A mutant proteins, and E42Aalso formed complexes slightly less efficiently thandid the wild-type protein. The reduced binding ofD39A to LAP-1 was also reflected by an increasedamount of free E8-Cys3EHA in the medium (MUT4;lane 5) compared to that found in conditioned

Table 2. E8-Cys3EHA mutations

Residuenumber 5 12 17 39 42 44–45

TGF-b1binding

A. LTBP-1 mutantsMUT1 E5A CCCMUT2 D12A CCCMUT3 D17A CCCMUT4 D39A CCMUT5 E42A CC Or

CCC

MUT6 E5A D12A D17A CC OrCCC

MUT7 E5A D39A E42A CMUT8 E5A D12A D17A D39A E42A –MUT9 E5K CCCMUT10 E5K D12K D17K CCMUT11 E5K D39K E42K –MUT12 E5K D12K D17K D39K E42K Unstable

MUT22 F44Q CCCMUT23 P45Q CCCMUT24 FP/QQ CCCMUT25 E42R FP/QQ CMUT26 D39S E42R FP/QQ –MUT27 D12A D39S E42R FP/QQ –

B. LTBP-4 mutantsMUT13 A12D S39D R42E –MUT14 QQ/FP –

MUT15 A12D QQ/FP –MUT16 S39D QQ/FP –MUT17 R42E QQ/FP –MUT18 A12D S39D QQ/FP –MUT19 A12D R42E QQ/FP –MUT20 S39D R42E QQ/FP CMUT21 A12D S39D R42E QQ/FP CC


media from cells expressing any of the other fourmutant constructs. The D39A E8CYS3EHA con-struct produced a protein doublet at the position ofthe complex (MUT4; lanes 5 and 12). Both bandsappear to represent a LAP-1-E8–Cys3EHA complex.

Figure 3. Representative regions of two-dimensional NOES(C) MUT11 samples. Whereas each mutant spectrum shows gthe central portion of the amide region, a number of NOE crosthree of the spectra. These results indicate that the MUT8 andsimilar, although not identical, to wild-type.

These bands may represent different glycosylationproducts; however, analysis of these bands was notpursued.We next studied additional mutant E8CYS3EHA

constructs in which E5, D12, D17, D39, E42 were

Y spectra for 8-Cys3LTBP-1. (A) Wild-type, (B) MUT8, andreater line broadening relative to wild-type, especially ins-peaks may be identified in highly similar positions in allMUT11 samples refold in vitro to a conformation that is

Figure 4. Association of LAP-1 with E8-Cys3EHALTPB-1: amino acid point mutations (MUT1-5). Each of the negativelycharged residues (E5, D12, D17, D39 and E42) in the 8-Cys3 domain of the LTBP-1 construct E8CYS3EHAwasmutated toalanine in turn. These mutant constructs (MUT1–MUT5) were co-expressed with proTGF-b1 in 293T cells. Conditionedmedia collected 48 hours after transfection were separated by SDS-PAGE under non-reducing conditions, transferred tonitrocellulose membrane and probed with either antibody against the HA tag (A) or antibody against LAP-1 (VB3A9)(B).


changed toeitherKorAsinglyor ingroupsof threeorfive residues (MUT-1, MUT6-12; Table 2). As isapparent from Figure 5, mutation to A at position 5(MUT1; lane 2) or at positions 5, 12 and 17 (MUT6;lane 3) had little effect on LAP-1–E8-Cys3EHAcomplex formation. A triple-substitution of A at thepositions 5, 39 and 42 (MUT7; lane 4) severelyimpeded complex formation, and the substitution ofA at all five positions (MUT8; lane 5) abolishedcovalent complex formation. Control experimentsanalyzing larger amounts ofmediumexposed toECLfor longer times failed to demonstrate a complexbetween MUT8 and LAP (data not shown). Substi-tution of K at position 5 (MUT9; lane 6) or positions 5,12 and 17 (MUT10; lane 7) had a small but detectableeffect on complex formation, whereas substitution atpositions 5, 39 and 42 (MUT11; lane 8) abolishedcomplex formation. Substitution of all five of the

targeted positions by K also yielded no complex(MUT12; lane 9). These mutations (MUT12) presum-ably rendered the E8-Cys3EHA unstable as no freeprotein was observed in the conditioned medium,and the in vitro refolded MUT12 8-Cys3LTBP-1 wasshown to be unstructured as monitored by NMRanalysis. Similar results were observedwhen the gelswere probed with the anti-LAP antibody VB3A9(data not shown). The free E8-Cys3EHA mutantproteins displayed slight differences in mobility(Figure 5). The reasons for this are not clear but mayrelate to the effects of total charge on the knownabnormal mobility of 8-Cys3 domains in SDS.7 Theresults indicate that residues D39 and, to a lesserextent, E42 are themost critical in LAP–E8-Cys3EHAcomplex formation, most likely because they provideimportant negative charge for the electrostaticinteraction.

Figure 5. Association of LAP-1with E8-Cys3EHALTBP-1: multipleamino acid mutations (MUT1,MUT6-13). One, three or five ofthe negatively charged residues inthe 8-Cys3 domain of the LTBP-1construct E8CYS3EHA weremutated to alanine (MUT-1,MUT6–MUT8) or lysine (MUT9–MUT12), respectively (see Table 2).Wild-type E8CYS3EHALTBP-1 andthe mutant constructs were co-expressed with proTGF-b1 in 293Tcells. Conditioned media were col-lected 48 hours after transfectionand were separated by SDS-PAGEunder non-reducing conditions,transferred to a nitrocellulosemembrane and probed with anti-body against the HA tag.

Figure 6.Amino acid requirements for 8-Cys3LTBP-4 isoform binding to LAP-1. (A)Wild-type E8CYS3EHALTBP-1, wild-type E8CYS3EHALTBP-4, E8CYS3EHALTBP-4 triple mutant (A12D, S39D, R42E) (MUT13) and E8CYS3EHALTBP-4 QQ-FPmutant (MUT14) constructs were co-expressed with proTGF-b1 in 293T cells. Conditioned media, collected 48 hoursafter transfection, were separated by SDS-PAGE under non-reducing conditions, transferred to a nitrocellulosemembrane and blotted with antibody against the HA tag. (B) Residues A12, S39 and R42 in the 8-Cys3 domain of theconstruct E8CYS3EHALTBP-4 were mutated to D, D, E, respectively, one, two or three at a time, together with the QQ/FPmutation (MUT15–MUT21; see Table 2). Wild-type E8CYS3EHALTBP-1, wild-type E8CYS3EHALTBP-4, and mutantconstructs were co-expressed with proTGF-b1 and complexes detected with antibody against the HA tag as describedabove.


Specificity of LTBP isoform binding to LAP-1

The binding of LAP-1 to LTBPs shows isoformspecificity as LTBP-1 binds LAP-1 well, whereasLTBP-4 does so poorly.18 Our original modelingexperiments indicated that the enhanced projectioninto the solvent of the 2–6 disulfide loop of 8-Cys3LTBP-1 versus the comparable loop in the nonbinding 8-Cys6fibrillin-1 was a result of the insertionof the dipeptide at positions 44 and 45.19 Theresulting sequence, EIFP, moves the disulfide bond

Figure 7. Role of FP in the E8-Cys3E isoform complexformation with LAP-1. Single amino acid substitutions ofFP to QQ in E8CYS3EHALTBP-1 were prepared. Inaddition, the amino acid residues at positions 12, 39 and42 of E8CYS3EHALTBP-1 were mutated to resemble thoseof 8-Cys3LTBP-4. The E8CY3EHA constructs were co-expressed with proTGF-b1 in 293T cells. Media collected48 hours after transfection were separated by SDS-PAGEunder non-reducing conditions, transfected to a nitro-cellulose membrane and blotted with antibody againstthe HA tag.

into the solvent causing the bond to be accessible,whereas the sequences in 8-Cys6fibrillin-1 and 8-Cys3LTBP-2, containing only two residues instead offour, do not allow solvent access to the 2–6 cysteinepair (see Figure 1(B)). The presence of four residuesbetween cysteine residues 6 and 7 in the compar-able 8-Cys3LTBP-4 sequence yields a predictedstructure in which the 2–6 cysteine pair is moresolvent exposed than that of LTBP-2. However, in8-Cys3LTBP-4 the D12, D39 and E42 residues foundin LTBP-1 8-Cys3 are replaced by A, S, and R,respectively, thereby reducing the negative chargeby four units, resulting, perhaps, in the observeddecreased complex formation with LAP.To strengthen our hypothesis that the negative

charges at positions 39 and 42 are important forLTBP–LAP-1 interactions and may regulate bothbinding and LTBP isoform specificity, we per-formed a series of mutagenesis experiments inwhich the aforementioned three residues in 8-Cys3LTBP-4 were mutated to resemble those ofLTBP-1 (MUT13-21; Table 2). In addition, the QQsequence in LTBP-4 was altered to FP because of theimportance of the insertion in projecting the 2–6disulfide bond into solution. The results of co-expression of these mutant constructs withproTGF-b1 are illustrated in Figure 6. Mutation of8-Cys3LTBP-4 residues A12, S39 and R42 to the com-parable amino acid residues found in LTBP-1 (D, D,and E) does not potentiate the binding of E8-Cys3EHALTBP-4/1 to LAP-1 (MUT13; Figure 6(A),lane 3). After long ECL exposure times, however, asmall amount of complex was observed (data notshown). Mutation of the QQ sequence to FP alsohad no effect on the ability of E8-Cys3EHALTBP-4 tocomplex with LAP-1 (MUT14; lane 4). Therefore,neither the change to acidic residues at positions 12,


39 and 42 nor the conversion of QQ to FP, bythemselves, was sufficient to enhance complexformation of E8-Cys3EHALTBP-4 with LAP-1. Wenext combined mutation of QQ to FP with substi-tution of the negatively charged amino acidresidues from LTBP-1 for the equivalent residuesof LTBP-4 in E8-Cys3EHA and tested for complexformation (Figure 6(B)). Single, A12D, S39D orR42E, or double, A12D plus S39D or A12D plusR42E, amino acid substitutions together with theQQ to FP mutation fail to convert E8-CysEHALTBP-4

to a binding molecule for LAP-1 (MUT15-19; lanes7–11). When the S39D and R42E mutations wereintroduced together with FP, some complex wasobserved after Western blotting of the conditionedmedium (MUT20; lane 12). Further enhancement ofcomplex formation was achieved with the substi-tutions of A12D plus the previous changes (MUT21;lane 13).

To investigate whether the efficiency of complexformation always involves both the charged aminoacid residues and the amino acid residues atpositions 44 and 45, an additional set of mutationswas prepared and tested (Figure 7). These con-structs (MUT 22–27; Table 2) represented pointmutations to residues FP and the charged residuesat positions 12, 39 and 42 of 8-Cys3LTBP-1 introducedto resemble the QQ and the uncharged residues in8-Cys3LTBP-4. Intriguingly, co-expression of theseconstructs and proTGF-b1 revealed that single pointmutations of F44Q or P45Q (MUT22; lane 2 andMUT23; lane 3) or the double mutation of FP to QQ(MUT24; lane 4) do not impair complex formationof E8-Cys3EHALTBP-1 with LAP-1, as would beexpected based on the requirement of FP at posi-tions 44–45 for 8-Cys3LTBP-4 binding (Figure 6).However, mutation of the 8-Cys3LTBP-1 sequence toQQ and E42R (MUT25; lane 5) significantlydecreased complex formation. Additional changesat positions 12 and 39 to resemble the LTBP-4sequence (MUT26 and 27; lanes 6 and 7) abolishedcomplex formation. These results further confirmthe essential requirement of negatively chargedresidues at positions 39 and 42 for efficient LAP-1binding.

Discussion

The interaction of the TGF-b LAP with LTBP iscritical for the proper function of the cytokine.15,24,25

Animals with null or hypomorphic mutations havemultiple phenotypes that are related to diminishedTGF-b signaling.24,25 These phenotypes may reflectthe ability of LTBP to facilitate TGF-b secretion aswell as the necessity for LTBP to localize the latentTGF-b complex at the proper site for activation.

Based upon the solution structure of 8-Cys3LTBP-1,we suggested that the five negatively chargedamino acid residues (E5, D12, D17, D39 and E42;see Figure 1(B)) surrounding the reactive disulfidebond might influence the interactions of LAP and8-Cys3.19 In the current research we have shown,

using FRET and site-directed mutagenesisapproaches, that electrostatic interactions betweenthe 8-Cys3 domains of either LTBP-1 or LTBP-4 andLAP-1 are a pre-requisite for intermolecular di-sulfide bond formation. Hence, we speculate thatthe generation of the native LLC occurs in a two-step manner in which the attraction of complimen-tary charges is followed by disulfide exchangebetween 8-Cys3 and LAP. We further elucidate that,of the charged amino acid residues previouslyidentified, D39 and E42 are the most important inmediating complex formation for both the LTBP-1and LTBP-4 isoforms as mutation at these positionsblocked or severely impaired complex formationwith LAP.

Our results delineate further contributions fromcharged residues at positions 5, 12 and 17, as well asthe amino acid composition of the dipeptideinsertion, mediating LTBP–LAP-1 complex forma-tion. The specific contribution made by the aminoacid side-chain at each of these sites is isoformspecific, i.e. dependent on the LTBP-1 or LTBP-4core protein. For example, some pairs of LTBP-1 andLTBP-4 mutants, i.e. MUT2/MUT20 and MUT4/MUT19 have identical residues at the positions 12,17, 39, 42 and 44–45; however, the two members ofeach pair do not show the same ability to formcomplexes. It is not until all of the positions arechanged to resemble one or the other LTBP thateither binding or lack of binding to LAP-1 changes.Changes at these seven positions can be used toconvert the binding of one LTBP isoform to another,or a bindingmolecule to a non-binder, and thereforewe have identified a binding epitope that definesthe amino acid requirements for LTBP associationwith LAP-1. The positions of these seven aminoacid residues that are important for either LTBP-1 orLTBP-4 binding to LAP-1 are mapped onto a surfacerepresentation of 8-Cys3LTBP-1 in Figure 8. Weconclude that electrostatic interactions contributemost to LTBP–LAP-1 binding efficiency. Althoughthe enhanced binding observed upon substitutionof the dipeptide insertion QQ by FP for 8-Cys3LTBP-4suggests that either hydrophobic or thermodyn-amic properties (e.g. structural stability) can alsoplay a role.

The association of 8-Cys domains and LAP basedon ionic interactions followed by covalent bondformation has potential significance for how theactivities of other members of the TGF-b super-family might be regulated. Although the propep-tides of these additional signaling molecules areunlikely to form disulfide bonds with LTBPsbecause of the absence of unpaired cysteineresidues, non-covalent interactions between theirpropeptides and LTBPs may exist, and theseinteractions may regulate growth factor action.26

Indeed, we have reported that in the early Xenopusembryo the dorsalizing activity of the TGF-bsuperfamily member activin is potentiated byLTBP-1.27 Because the interaction of these twoproteins cannot occur via disulfide bond exchange,it is appealing to suggest the existence of ionic

Figure 8. Surface representation of an 8-Cys3LTBP-1domain showing the complete binding epitope for LTBP-1and LTBP-4 defined by residues 5, 12, 17, 39, 42 and 44-45.The cysteine residues of the 2,6 disulfide bond, predictedto participate in LTBP–LAP complex formation, arecolored in yellow. The two most important residues forcomplex formation for both isoforms, D39 and E42, areshown in red. Charged residues identified as playing aless important role are shaded pink. The two residues ofthe dipeptide insertion that mediates complex formationare shown in green.


interactions between LTBP-1 8-Cys domains andactivin or its propeptide. In addition, several otherTGF-b superfamily members, such as myostatin28

and BMP-7,29 are found in the extracellular milieuas latent complexes of cytokine and propeptideanalogous to the TGF-b LAP complex. How theseother cytokines are eventually released from theirLAPs is not clear; nor is it clear how latency of theseproteins relates to their biological function. More-over, the significance of continued association ofmature cytokine with the processed propeptides isunknown for myostatin and BMP-7.

The interactions of latent complexes with LTBPsor fibrillins may serve to modify growth factoraction in several ways, e.g. growth factor gradientsmay be stabilized and active cytokine concen-trations may be enhanced in the local environmentby binding to LTBPs, growth factor may be shieldedinteraction with soluble inhibitors, growth factorinteraction with signaling receptors may be modi-fied, and latent growth factor activation may befacilitated.30 It is interesting to speculate that theearly embryonic lethality of the LTBP-2 nullmouse31 illustrates the effects of LTBP-2 for a cyto-kine, other than TGF-b, whose activity is required inearly development and that binds to an 8-CysLTBP-2domain. Likewise, it is appealing to consider thatthe abnormal growth and differentiation of thebones and lungs in some individuals with Marfansyndrome results not from a defect in fibrillin-1structural activities but through changes in thelevels of growth mediators as a consequence ofdefective or decreased levels of fibrillin-1.26,30,32,33

In the absence of proper extracellular localization,these growth factors induce improper tissue growtheventually resulting in the observed phenotype.Thus, 8-Cys interactions with signaling moleculesmay extend beyond interaction with LAP. Ourresults will contribute towards understandingthese homologous interactions by enabling the useof structural modeling to assess the electrostaticproperties of 8-Cys domains in the LTBPs andfibrillins to produce testable hypotheses regardingmolecular recognition.

Materials and Methods

FRET studies of LAP-1C33S binding to 8-Cys3LTBP-1

The production of LAP-1C33S and 8-Cys3LTBP-1 havebeen described.19,34 LAP-1C33S was buffer-exchanged into20 mM Tris–HCl (pH 7.4) using a 0.5 ml centrifuge filter(Millipore, Billerica, MA) and diluted to a concentrationof 1 mM, which was estimated spectrophotometrically.NAPe-10 Sephadex G-25 desalting columns (5 ml;

Amersham Pharmacia Biotech, Piscataway, NJ) werewashed with carbonate buffer (pH 9.2). Then 200 mlof 8-Cys3LTBP-1 (200 mM) was loaded onto the columnfollowed by 20 aliquots (200 ml) of buffer, and 200 mlfractions were collected. Fractions containing protein(based on A280 absorbance readings) were pooled.Labeling of the proteins was achieved by adding an

estimated molar ratio of 3 : 1 of dye (Alexa Fluor-350carboxylic acid, succinimidyl ester; Molecular Probes,Eugene, OR) to protein and incubating at room tempera-ture for one hour. Alexa Fluor-350 was chosen, since itsexcitation spectrum overlaps with the emission spectrumof tryptophan. The 8-Cys3LTBP-1 amino acid sequencecontains a single tryptophan residue. The labelingreactions were quenched by addition of Tris–HCl (pH7.4) to a final concentration 100 mM and passed through aNAPe-10 column that was pre-equilibrated with 20 mMTris–HCl (pH 7.4). Detection and separation of AlexaFluor-350 labeled proteins was achieved by viewingunder UV light. Labeled protein fractions were pooledand used in FRET binding experiments. The concen-tration of labeled 8-Cys3LTBP-1 molecule was calculatedusing methods described in the Molecular Probes hand-book. An average of three labels per 8-Cys3LTBP-1 wasestimated from mass spectrometry data.Alexa-Fluor 350 emission spectra in the range of >400–

500 nm were recorded at 20 8C in 20 mM Tris–HCl (pH7.4) for 10 mM 8-Cys3LTBP-1C50, 100, 250, 500 mM and1 M NaCl, 1 mM LAP-1C33S, 1 mM LAP-1C33SC1 NaCl,10 mM8-Cys3LTBP-1C1 mMLAP-1C33S, 10 mM8-Cys3LTBP-1C1 mM LAP-1C33SC40 mM u8-Cys3LTBP-1 (unlabeled). Eachexperiment was performed four times and data at theemission maxima of 439 nm for LAP Alexa-Fluor 350 areshown. Spectra were recorded in a quartz cuvette(Hellma, Mullheim, Germany; 3 mm light path) usingan Aminco-Bowmanw (Thermo Spectronic, Rochester,NY) series 2 fluorimeter. The excitation wavelength wasset to 280 nm and the excitation and emission bandwidthswere set to 4 nm.

NMR analysis of 8-Cys3LTBP-1 mutants

NMR experiments were performed using a home-built/GE Omega spectrometer operating at 600 MHz


proton frequency having a triple resonance probe withself-shielded pulsed field gradients. The in vitro refoldingproperties of 8-Cys3LTBP-1 mutants MUT8, MUT11 andMUT12 (Table 2) were assessed based on one-dimen-sional 1H NMR data acquired at pH 6.0 and 25 8C (datanot shown). For those mutants that exhibited spectralcharacteristics associated with folded protein, two-dimensional 1H nuclear Overhauser enhancementspectroscopy (NOESY) spectra35,36 were also acquiredwith a mixing time of 150 ms. Acquisition times were102 ms and 25.6 ms in the direct and indirect dimensions,respectively. Pulsed field gradients were used for watersuppression. The two-dimensional NOESY data wereprocessed with using Felix 97 (MSI, Inc, San Diego, CA)using a Gaussian line-broadening window function in theacquisition dimension, with K15 and 0.15 line broaden-ing and Gaussian parameters, respectively, and a 708shifted squared sine-bell window function in the indirectdimension. The sample concentrations used in theseexperiments were: wild-type, 900 mM; MUT8, 740 mM;MUT11, 265 mM; MUT12, 550 mM.

Construction of LTBP-1 cbEGF-8Cys3-cbEGF-2HA(E8CYS3EHA) and MUT1-13 mutant expression vectors

The sequences spanning EGF-like domain 13, 8-Cys3,and EGF-like domain 14 (E8CYS3E) were amplified fromthe LTBP-1S cDNA by PCR using primers 50 CGGGGATCCACTAGTGGATGTGAATGAATGTGAACT (sense) and50 AACAAGCA CTGCAGTTTCACAGG (anti-sense). ThePCR product was subsequently fused behind the BM40signal sequence (from SPARC) of pRcCMV/Ac7 vector(a gift from Dr Rupert Timpl, Max-Planck Institut furBiochemie, Martinsried, Germany). Two copies of an HAtag sequence were added at the 3 0 end of the E8CYS3Esequence by digesting pRcCMV/Ac7-E8CYS3E withHpaI and XbaI, and introducing an adaptor cassette thatcontains the sequence 5 0 AACCTACCCCTACGACGTGCCCGACTACGCCTACCCCTACGACGTGCCCGACTAGCCTGAAGATCTTGATTGGAATTCCGGCCGT. The EagI-EagI fragment of pRcCMV/Ac7-E8CYS3EHA was sub-cloned into pKN185 (a gift from Dr Y. Yamada, NationalInstitutes of Health, Bethesda, MD) to yield pKN-E8CYS3EHA. The HindIII-BamHI fragment of pKN-ETB3E-2HA containing ETB3E-2HA was subcloned intopBluescript II SK (C) (Stratagene) or pcDNA3 (InvitrogenCorp., Carlsbad, CA) to yield pBlu-E8CYS3EHA andpcDNA3 E8CYS3EHA. Site-directed mutagenesis wasperformed in pBlu-E8CYS3EHA using the QuikChangew

Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA)to make different mutations in MUT1–MUT12 andMUT22-27 (see Table 2). The HindIII-BamHI MUT1–MUT13 and MUT22–MUT27 fragments from pBlu-E8CYS3EHA were subcloned into pcDNA3 eukaryoticexpression vector. All mutagenesis of plasmid DNAconstructs described here was confirmed by automatedDNA sequencing using an ABI 3730 prism sequencer(Applied Biosystems).

Construction of pcDNA3 L4 E8CYS3EHA and its mutantexpression vectors

The sequence encoding the 8-Cys3 and its two flankingEGF-like domains of LTBP-4 was amplified by PCRusing LTBP-4 cDNA as the template and primers: 5 0

TCCACAGAAGCTTATGTGAACGAGTG-3 0 (sense) and5 0 ATATCTGCAGAATTGTCTTGACCCTTAGGCGTAGTCGGGCACGTCGTAGGGGTAGGTTAACTCGTCATTGTCG-3 0 (anti-sense). An HA tag was included in the 3 0

primer. The L4.E8CYS3EHA PCR fragment was clonedinto pSecTag2C vector (Invitrogen Corp., Carlsbad, CA)in frame with the Ig k-chain leader sequence. The L4E8CYS3EHA insert including the secretion signal and theHA tag was amplified by PCR by using pSecTag2CL4.E8CYS3E as a template and primers: 5 0-ATCGCGGTCCAGCTGGCTAGCCACCATGG-3 0 and 5 0-TCTAGTCTAGACTAGAAGGCACAGTCGAGGC-3 0. This PCR pro-duct was digested with BamHI and XbaI and transferredinto pcDNA3. Site-directed mutagenesis was performedusing the QuikChangew Site-Directed Mutagenesis Kit(Stratagene, La Jolla, CA) on pcDNA3 L4E8CYS3EHA toget MUT13–MUT21 (see Table 2).

Transfection and Western blotting assays

The 293T Cells were plated in 35 mm diameter tissueculture dishes at a density of 5!105 cells per dish the daybefore transfection and were transfected with 1 mg of thepcDNA3 E8CYS3EHA DNA or mutant constructstogether with 1 mg of pcDNA3 TGF-b1 DNA. Transfec-tions were carried out using a LipofectAMINE PLUStransfection kit (Invitrogen) in Opti-MEM I (GIBCO BRL,Grand Island, NY) following the manufacturer’s instruc-tions. The transfected cells were washed with Opti-MEMI medium three hours after transfection, and theconditioned medium was collected 48 hours aftertransfection. Proteins in the conditioned media wereseparated by SDS-PAGE under non-reducing conditionsand electrophoretically transferred to nitrocellulose mem-branes. Membranes were blocked with TBS-T (TBS with0.1% Triton X-100) with 5% (w/v) non-fat dry milk,incubated with primary antibody (anti-HA; HA.11,BAbCO (Denver, PA) or anti-TGFb-1 LAP (VB3A933) forone hour at room temperature, washed three times withTBS-T, and incubated with an appropriate secondaryantibody conjugated with horseradish peroxidase(Amersham Pharmacia Biotech). After washing, immuno-reactive bands were revealed by processing with ECLWestern blotting detection reagents (AmershamPharmaciaBiotech) following the manufacturer’s instructions.Samples were normalized, when possible, to totalsecreted E-8Cys3E determined after reducing SDS-PAGE and Western blotting. These experiments wereperformed three times with similar results.

Acknowledgements

The authors thank J. Annes for originally pointingout the potential significance of E42 in distinguish-ing reactivities of LTBP-1 and -4 with LAP. Thiswork was supported by grants CA34282 and CA78422 from the NIH to D.B.R. and 070417 from theWellcome Trust (to A.K.D.). A.K.D. is a WellcomeTrust Senior Research Fellow (057725).

References

1. Flaumenhaft, R. & Rifkin, D. B. (1992). The extra-cellular regulation of growth factor action. Mol. Biol.Cell, 3, 1057–1065.

2. Annes, J. P., Munger, J. S. & Rifkin, D. B. (2003).Making sense of latent TGF-beta activation. J. Cell Sci.116, 217–224.


3. Koli, K., Saharinen, J., Hyytiainen, M., Penttinen, C. &Keski-Oja, J. (2001). Latency, activation, and bindingproteins of TGF-beta. Microsc. Res. Tech. 52, 354–362.

4. Dubois, C. M., Laprise, M. H., Blanchette, F., Gentry,L. E. & Leduc, R. (1995). Processing of transforminggrowth factor beta 1 precursor by human furinconvertase. J. Biol. Chem. 270, 10618–10624.

5. Lawrence, D. A., Pircher, R., Kryceve-Martinerie, C. &Jullien, P. (1984). Normal embryo fibroblasts releasetransforming growth factors in a latent form. J. CellPhysiol. 121, 184–188.

6. Lawrence, D. A., Pircher, R. & Jullien, P. (1985).Conversion of a high molecular weight latent beta-TGF from chicken embryo fibroblasts into a lowmolecular weight active beta-TGF under acidic condi-tions. Biochem. Biophys. Res. Commun. 133, 1026–1034.

7. Gleizes, P. E., Beavis, R. C., Mazzieri, R., Shen, B. &Rifkin, D. B. (1996). Identification and characteriz-ation of an eight-cysteine repeat of the latenttransforming growth factor-beta binding protein-1that mediates bonding to the latent transforminggrowth factor-beta1. J. Biol. Chem. 271, 29891–29896.

8. Saharinen, J., Taipale, J. & Keski-Oja, J. (1996).Association of the small latent transforming growthfactor-beta with an eight cysteine repeat of its bindingprotein LTBP-1. EMBO J. 15, 245–253.

9. Sime, P. J., Xing, Z., Graham, F. L., Csaky, K. G. &Gauldie, J. (1997). Adenovector-mediated gene trans-fer of active transforming growth factor-beta1 inducesprolonged severe fibrosis in rat lung. J. Clin. Invest.100, 768–776.

10. Filvaroff, E., Erlebacher, A., Ye, J., Gitelman, S. E.,Lotz, J., Heillman, M. & Derynck, R. (1999). Inhibitionof TGF-beta receptor signaling in osteoblasts leads todecreased bone remodeling and increased trabecularbone mass. Development, 126, 4267–4279.

11. Yang, X., Chen, L., Xu, X., Li, C., Huang, C. & Deng,C.-X. (2001). TGF-Beta/Smad3 signals repress chon-drocyte hypertrophic differentiation and are requiredfor maintaining articular cartilage. J. Cell Biol. 153,35–46.

12. Markowitz, S., Wang, J., Myeroff, L., Parsons, R., Sun,L., Lutterbaugh, J. et al. (1995). Inactivation of the typeII TGF-beta receptor in colon cancer cells withmicrosatellite instability. Science, 268, 1336–1338.

13. Riggins, G. J., Kinzler, K. W., Vogelstein, B. &Thiagalingam, S. (1997). Frequency of Smad genemutations in human cancers. Cancer Res. 57, 2578–2580.

14. Zhou, S., Buckhaults, P., Zawel, L., Bunz, F., Riggins,G., Dai, J. L. et al. (1998). Targeted deletion of Smad4shows it is required for transforming growth factorbeta and activin signaling in colorectal cancer cells.Proc. Natl Acad. Sci. USA, 95, 2412–2416.

15. Annes, J. P., Chen, Y., Munger, J. S. & Rifkin, D. B.(2004). Integrin alphaVbeta6-mediated activation oflatent TGF-beta requires the latent TGF-beta bindingprotein-1. J. Cell Biol. 165, 723–734.

16. Miyazono, K., Olofsson, A., Colosetti, P. & Heldin,C. H. (1991). A role of the latent TGF-beta 1-bindingprotein in the assembly and secretion of TGF-beta 1.EMBO J. 10, 1091–1101.

17. Saharinen, J., Hyytiainen, M., Taipale, J. & Keski-Oja,J. (1999). Latent transforming growth factor-betabinding proteins (LTBPs)—structural extracellularmatrix proteins for targeting TGF-beta action. Cyto-kine Growth Factor Rev. 10, 99–117.

18. Saharinen, J. & Keski-Oja, J. (2000). Specific sequencemotif of 8-Cys repeats of TGF-beta binding proteins,

LTBPs, creates a hydrophobic interaction surface forbinding of small latent TGF-beta. Mol. Biol. Cell, 11,2691–2704.

19. Lack, J., O’Leary, J. M., Knott, V., Yuan, X., Rifkin,D. B., Handford, P. A. & Downing, A. K. (2003).Solution structure of the third TB domain from LTBP1provides insight into assembly of the large latentcomplex that sequesters latent TGF-beta. J. Mol. Biol.334, 281–291.

20. Yuan, X., Downing, A. K., Knott, V. & Handford, P. A.(1997). Solution structure of the transforming growthfactor b-binding protein-like module, a domainassociated with matrix fibrils. EMBO J. 16, 6659–6666.

21. Kanzaki, T., Olofsson, A., Moren, A., Wernstedt, C.,Hellman, U., Miyazono, K. et al. (1990). TGF-beta 1binding protein: a component of the large latentcomplex of TGF-beta 1 with multiple repeatsequences. Cell, 61, 1051–1061.

22. Saharinen, J., Taipale, J., Monni, O. & Keski-Oja, J.(1998). Identification and characterization of a newlatent transforming growth factor-beta-binding pro-tein, LTBP-4. J. Biol. Chem. 273, 18459–18469.

23. Bottinger, E. P., Factor, V. M., Tsang, M. L.,Weatherbee, J. A., Kopp, J. B., Qian, S. W. et al.(1996). The recombinant proregion of transforminggrowth factor beta1 (latency-associated peptide) inhi-bits active transforming growth factor beta1 in trans-genic mice. Proc. Natl Acad. Sci. USA, 93, 5877–5882.

24. Dabovic, B., Chen, Y., Colarossi, C., Obata, H.,Zambuto, L., Perle, M. A. & Rifkin, D. B. (2002).Bone abnormalities in latent TGF-[beta] bindingprotein (Ltbp)-3-null mice indicate a role for Ltbp-3in modulating TGF-[beta] bioavailability. J. Cell Biol.156, 227–232.

25. Sterner-Kock, A., Thorey, I. S., Koli, K., Wempe, F.,Otte, J., Bangsow, T. et al. (2002). Disruption of thegene encoding the latent transforming growth factor-beta binding protein 4 (LTBP-4) causes abnormal lungdevelopment, cardiomyopathy, and colorectal cancer.Genes Dev. 16, 2264–2273.

26. Charbonneau, N. L., Ono, R. N., Corson, G. M., Keene,D. R. & Sakai, L. Y. (2004). Fine tuning of growth factorsignals depends on fibrillin microfibril networks.Birth Defects Res. Part C Embryo Today, 72, 37–50.

27. Altmann, C. R., Chang, C., Munoz-Sanjuan, I., Bell, E.,Heke, M., Rifkin, D. B. & Brivanlou, A. H. (2002). Thelatent-TGFbeta-binding-protein-1 (LTBP-1) is expressedin the organizer and regulates nodal and activinsignaling. Dev. Biol. 248, 118–127.

28. Lee, S. J. & McPherron, A. C. (2001). Regulation ofmyostatin activity and muscle growth. Proc. NatlAcad. Sci. USA, 98, 9306–9311.

29. Sampath, T. K., Maliakal, J. C., Hauschka, P. V., Jones,W. K., Sasak, H., Tucker, R. F. et al. (1992). Recombin-ant human osteogenic protein-1 (hOP-1) induces newbone formation in vivo with a specific activity com-parable with natural bovine osteogenic protein andstimulates osteoblast proliferation and differentiationin vitro. J. Biol. Chem. 267, 20352–20362.

30. Ramirez, F. & Rifkin, D. B. (2003). Cell signaling events:a view from the matrix.Matrix Biol. 22, 101–107.

31. Shipley, J. M., Mecham, R. P., Maus, E., Bonadio, J.,Rosenbloom, J., McCarthy, R. T. et al. (2000). Develop-mental expression of latent transforming growthfactor beta binding protein 2 and its requirementearly in mouse development. Mol. Cell. Biol. 20,4879–4887.

32. Neptune, E. R., Frischmeyer, P. A., Arking, D. E.,Myers, L., Bunton, T. E., Gayraud, B. et al. (2003).


Dysregulation of TGF-beta activation contributes topathogenesis in Marfan syndrome. Nature Genet. 33,407–411.

33. Judge, D. P., Biery, N. J., Keene, D. R., Geubtner, J.,Myers, L., Huso, D. L. et al. (2004). Evidence for acritical contribution of haploinsufficiency in thecomplex pathogenesis of Marfan syndrome. J. Clin.Invest. 114, 172–181.

34. Munger, J. S., Harpel, J. G., Giancotti, F. G. & Rifkin,D. B. (1998). Interactions between growth factors andintegrins: latent forms of transforming growth factor-beta are ligands for the integrin alphavbeta1. Mol.Biol. Cell, 9, 2627–2638.

35. Macura, S., Huang, Y., Suter, D. & Ernst, R. R. (1981). 2-Dimensional chemical exchange and cross-relaxationspectroscopy of coupled nuclear spins. J. Magn. Reson.43, 259–281.

36. Jeener, J., Meier, B. H., Bachmann, P. & Ernst, R. R.(1979). Investigation of exchange processes by two-dimensional NMR spectroscopy. J. Chem. Phys. 71,4546–4553.

37. Thompson, J.D.,Gibson,T. J., Plewniak, F., Jeanmougin,F. & Higgins, D. G. (1997). The CLUSTAL_X windowsinterface: flexible strategies for multiple sequencealignment aided by quality analysis tools. Nucl. AcidsRes. 25, 4876–4882.

Edited by P. Wright

(Received 24 September 2004; accepted 14 October 2004)

amino acid requirements for formation of the tgf-β-latent tgf-β binding protein complexes

Documents