Three Key Residues Underlie the Differential Affinity of the TGFβ Isoforms for the TGFβ Type II Receptor

doi:10.1016/j.jmb.2005.10.022 J. Mol. Biol. (2006) 355, 47–62

Three Key Residues Underlie the Differential Affinityof the TGFb Isoforms for the TGFb Type II Receptor

Gregory De Crescenzo3†, Cynthia S. Hinck1†, Zhanyong Shu1

Jorge Zuniga1, Junhua Yang2, Yuping Tang2, Jason Baardsnes3

Valentın Mendoza4, LuZhe Sun2, Fernando Lopez-Casillas4

Maureen O’Connor-McCourt3 and Andrew P. Hinck1*

1Department of BiochemistryUniversity of Texas HealthScience Center at San AntonioSan Antonio, TX 78229, USA

2Department of Cellular andStructural Biology, Universityof Texas Health Science Centerat San Antonio, San AntonioTX 78229, USA

3Biotechnology ResearchInstitute, National ResearchCouncil, Montreal,Que. Canada H4P2R2

4Instituto de Fisiologıa CelularUniversidad NacionalAutonoma de Mexico,Ciudad de Mexico,D.F. 04510, Mexico

0022-2836/$ - see front matter q 2005 E

† G. De C. & C.S.H. contributed ePresent addresses: G. De Crescen

Polytechnique de Montreal, Departchimique, Montreal, Que. Canada HDepartment of Biological ChemistryCalifornia, Irvine, CA 92697, USA.

Abbreviations used: TGFb, transfb; ED, extracellular domain; SPR, suresonance; FBHE, fetal bovine heart

E-mail address of the [email protected]

TGFb1, b2, and b3 are 25 kDa homodimeric polypeptides that play crucialnon-overlapping roles in development, tumor suppression, and woundhealing. They exhibit 70–82% sequence identity and transduce their signalsby binding and bringing together the TGFb type I and type II receptors,TbRI and TbRII. TGFb2 differs from the other isoforms in that it binds TbRIIweakly and is dependent upon the co-receptor betaglycan for function. Toexplore the physicochemical basis underlying these differences, wegenerated a series of single amino acid TbRII variants based on the crystalstructure of the TbRII:TGFb3 complex and examined these in terms of theirTGFb isoform binding affinity and their equilibrium stability. The resultsshowed that TbRII Ile53 and Glu119, which contact TGFb3 Val92 andArg25, respectively, together with TbRII Asp32, Glu55, and Glu75, whichcontact TGFb3 Arg94, each contribute significantly, between 1 kcal molK1

to 1.5 kcal molK1, to ligand binding affinities. These contacts likely underliethe estimated 4.1 kcal molK1 lower affinity with which TbRII binds TGFb2as these three ligand residues are unchanged in TGFb1 but areconservatively substituted in TGFb2 (Lys25, Ile92, and Lys94). To test thishypothesis, a TGFb2 variant was generated in which these three residueswere changed to those in TGFbs 1 and 3. This variant exhibited receptorbinding affinities comparable to those of TGFbs 1 and 3. Together, theseresults show that these three residues underlie the lowered affinity ofTGFb2 for TbRII and that all isoforms likely induce assembly of the TGFbsignaling receptors in the same overall manner.

q 2005 Elsevier Ltd. All rights reserved.

Keywords: TGFb; TbRII; betaglycan; ligand–receptor; alanine scanning
*Corresponding author
Introduction

Transforming growth factor-beta (TGFb) isoformsregulate cell proliferation, cell differentiation, and

lsevier Ltd. All rights reserve

qually to this work.zo, Ecoleement de genie

3C3A7; Z. Shu,, University of

orming growth factorrface plasmonendothelial.

ing author:

expression of extracellular matrix proteins and arethe founding members of a highly diversifiedsuperfamily of w25 kDa signal ligands.1 Theseisoforms, and other structurally related proteins ofthe TGFb superfamily, such as activins, bonemorphogenetic proteins (BMPs), and growth anddifferentiation factors (GDFs), exert their biologicaleffects by binding and bringing together two pairsof structurally similar, single-pass transmembranereceptors, classified as types I and II.2 The ligand-mediated assembly of these two receptor typestriggers an intracellular phosphorylation cascade,initiated by the serine-threonine kinase domain ofthe type II receptor, which transphosphorylates theadjacent type I kinase.3 The type I kinase in turnphosphorylates the nuclear-translocating Smad

d.

48 Differential Affinity of TbRII for TGFb Isoforms

proteins, which together with various transcrip-tional coactivators and corepressors,4 regulatetranscription of target genes.

There are three TGFb isoforms in mammals. Theisoforms, termed TGFbs 1–3, are each encoded by adistinct gene, share between 70% and 82% sequenceidentity, and are expressed in a developmentallyregulated tissue-specific fashion.1 They exhibitoverlapping biological activities in vitro althoughthe phenotypes of the isoform-specific null mice arenon-overlapping. TGFb1 null mice have an auto-immune-like inflammatory disease,5 TGFb2 nullmice exhibit perinatal mortality and developmentaldefects,6 and TGFb3 null mice have cleft palate andare defective in lung development.7 This indicatesthat each isoform fulfills a distinct role in vivo.

The three TGFb isoforms have been shown totransduce their signals by binding the TGFb type Iand type II receptors, TbRI and TbRII, respectively.8

TGFb2 differs from TGFbs 1 and 3, however, in thatit binds TbRII with an affinity about 1000-foldweaker (corresponds to a reduction of its bindingfree energy of about 4.1 kcal molK1).9–11 Consistentwith this altered pattern of TbRII binding, cellresponsiveness to TGFb2 has been shown to bedependent upon a third abundant TGFb bindingprotein, known as betaglycan.12

The assembly on the cell surface of the complexbetween TbRI and TbRII and the TGFb isoforms hasbeen shown through experiments with differen-tially tagged receptors to occur with a 2:2:1stoichiometry.13,14 This stoichiometry has beenconfirmed in part through the determination ofthe structure of the TbRII extracellular domain(TbRII-ED) bound to the TGFb3 dimer.15 Thisshowed that TbRII-ED binds by wedging itselfbetween the fingers on the distal ends of thedimer. There is no contact between the two boundTbRII-ED and neither TbRII-ED nor TGFbundergo significant structural rearrangementsupon binding.15–17 There is at present no structuralinformation available for the TbRI extracellulardomain (TbRI-ED), although based on the crystalstructure of the 2:1 BMP type Ia receptor extracellulardomain (BMPRIa-ED) complexed to BMP-2,18 it hasbeen proposed that all type I receptors of the TGFbsuperfamily bind their cognate growth factors bybinding at the dimer interface in a manner similarto that of BMPRIa.18

The objective of the studies reported here was toidentify interactions that contribute to the affinityand specificity of TbRII:TGFb isoform binding. Thiswas accomplished by generating a series of singleamino acid TbRII-ED variants. These were charac-terized in terms of their stability using fluorescenceand their TGFb isoform binding properties usingnative gels, surface plasmon resonance, and cell-based Smad phosphorylation assays. The resultsshowed that the effects of the substitutions onaffinity are largely due to alterations in ligandcontacts, not stability, and that complex formationis mediated by a linear array of hydrophobicresidues that bind in a hydrophobic cleft and by

two flanking hydrogen-bonded ion pairs. TbRIIIle53 and Glu119, which contact TGFb3 Val92 andArg25, respectively, together with TbRII Asp32,Glu55, and Glu75, which contact TGFb3 Arg94,were each shown to contribute significantly,between 1 kcal molK1 to 1.5 kcal molK1, to ligandbinding affinities. These contacts likely underlie the4.1 kcal molK1 lower affinity with which TbRIIbinds TGFb2 as these three ligand residues areunchanged in TGFb1 but are conservatively sub-stituted in TGFb2 (Lys25, Ile92, and Lys94). To testthis hypothesis, a TGFb2 variant was generated inwhich these three residues were changed to those inTGFbs 1 and 3. This variant exhibited receptorbinding affinities comparable to those of TGFbs 1and 3. Together, these results show that these threeresidues underlie the lowered affinity of TGFb2 forTbRII and that all isoforms induce the assembly ofthe TGFb signaling receptors in the same overallmanner.

Results

Variant TbRII-EDs

The initial objective of this study was todetermine the contribution of individual TbRII-ED residues toward TGFb isoform binding. Theresidues chosen for study were those that contactTGFb3 in the crystal structure of the TbRII-ED:TGFb3 complex.15 These, as shown in Figure 1,vary in contact area from 4 A2 to 90 A2 and wereeach initially substituted with alanine. Threeadditional interfacial variants were also chosenfor study, D32N, S52L, and E119Q. S52L waschosen as Ser52 is changed to alanine in othermammalian species (Figure 1(b)). D32N andE119Q were chosen to assess the extent towhich electrostatic effects play a role in theinteraction between TbRII Asp32:TGFb1/b3Arg94 and between TbRII Glu119:TGFb1/b3Arg25 (Figure 1(a)).

Seven non-interfacial variants, N47A, V62A,E75A, H79A, F110A, M112A, and I125A, were alsochosen for study. E75A was chosen as Glu75 forms awater-mediated hydrogen-bond with Arg94 ofTGFb3 in the TbRII-ED:TGFb3 crystal structure(Figure 1(a)). Substitution of this residue waspredicted to affect TGFb isoform binding affinities.V62A, H79A, and M112A were chosen by virtue oftheir structural homology to critical ligand contactresidues, Phe42, Trp60, and Phe83, identified inActRIIa.19,20 Substitution of these residues waspredicted to have little if any effect on TGFb isoformbinding affinities as TbRII and ActRIIa have beenshown to interact with their ligands throughinterfaces that are entirely distinct.15,20 N47A,V62A, F110A, and I125A were chosen as thesecorrespond to surface-exposed residues that lieoutside the two possible binding interfaces. Substi-tution of these residues was also predicted to havelittle if any effect on TGFb isoform binding affinities.

Figure 1. TbRII-TGFb3 interactions identified through analysis of the binding interface in the TbRII:TGFb3 crystalstructure. (a) Stereo diagram of the TbRII:TGFb3 interface. Backbones are gray, disulfide bonds are black, and contactresidues of TbRII and TGFb3 are magenta and green, respectively. Interfacial hydrogen bonds are depicted by brokenlines and ordered interfacial water molecules are shown as red spheres. (b) Major interactions evident in thecrystallographic interface are indicated by double-headed arrows. Classification of the interactions, and the interfacialcontact areas, are indicated on the left.

Differential Affinity of TbRII for TGFb Isoforms 49

TbRII-ED:TGFb isoform binding as detectedby native gel electrophoresis

The binding of TGFb isoforms by TbRII-ED canbe monitored using native gels. This is illustrated bythe data presented in Figure 2(a). TbRII-ED, whichhas a molecular mass of about 15 kDa and is highlysoluble at the pH of the gel, migrates rapidly andappears as a sharp band near the gel front(Figure 2(a), all subpanels, lane 1). TGFb isoforms,which have molecular masses of about 25 kDa andare insoluble at the pH of the gel (pH 6.8), exhibit nodetectable bands (Figure 2(a), all subpanels, lane 2).The addition of increasing amounts of TbRII-ED toa fixed amount of TGFb1 or TGFb3 homodimerleads to a complex band that migrates to a positionmidway along the length of the gel. This increases

in intensity up to the point at which two equivalentsof TbRII-ED have been added (Figure 2(a), first andthird subpanels from the left, lanes 3–6), consistentwith the expected TGFb:TbRII-ED 1:2 stochiometry.

The results obtained with TGFb2 differ as thecomplex band was absent at all TbRII-ED:TGFb2ratios examined (Figure 2(a), second subpanel fromthe left, lanes 3–8). The fact that TGFb1 and TGFb3form complex bands with TbRII-ED, but TGFb2does not, is attributed to the reduced affinity ofTbRII-ED for this isoform. To rule out the possibilitythat the absence of a complex band was due toaltered mobility of the TbRII-ED:TGFb2 complex,the native gel was carried out at a lower tempera-ture (2 8C instead of 22 8C; Figure 2(a), rightmostsubpanel). This resulted in the appearance of acomplex band at the expected position, indicating

Figure 2. Qualitative analysis of wild-type and variant TbRII-ED:TGFb isoform binding by native gel electrophoresis.(a) Titration of a fixed amount (1.5 mg) of TGFbs 1, 2, and 3 with increasing amounts of wild-type TbRII (0–3.0 molarequivalents) are shown in the three leftmost subpanels, respectively. The gels were each run for about 1 h at 22 8C andstained with Coomassie blue. Titration of a fixed amount of TGFb2 (1.5 mg) with increasing amounts of wild-type TbRII isshown in the rightmost subpanel. This titration was performed as above, but at a lower temperature (2 8C, rather than22 8C). TbRII:TGFb isoform complexes are indicated by horizontal arrows. (b) and (c) The effect of a representative set ofTbRII substitutions on complex formation with TGFbs 1 and 3 (native gel data showing the full panel of 22 TbRII-EDvariants is provided as Supplementary Data). The data are presented in pairs, with two equivalents of receptor alone(2.0 mg) in the left lane and two equivalents of receptor (2.0 mg) mixed with one equivalent of TGFb1 or TGFb3homodimer (1.8 mg) in the right lane. The individual receptor variants were analyzed on separate gels and were alignedrelative to one another based on the migration of the loading dye. The migration of some of the receptor variants, such asE119A, was retarded relative to wild-type. This was attributed to a decrease in the net overall charge of the receptor at thepH of the gel (8.8). This follows as such behavior was only observed for receptor variants in which the net overall chargewas changed (see Supplementary Data).


that the absence of the complex band at the highertemperature is not caused by the altered mobility ofthe complex. The fact that the complex bandappears at the lower temperature, but not thehigher temperature, is probably caused by areduction in the dissociation rate of the complex,koff, at the lower temperature. The fact that thecomplex band with TGFb2 was more diffuse andfree TbRII-ED was present, even when TGFb2was in excess, suggests that the reduction in koff

caused by the lower temperature is not sufficientto fully suppress dissociation of the complexduring the course of electrophoresis. These resultsdemonstrate that the native gel binding assayreports in a qualitative way on TbRII:TGFb isoformbinding in solution.

Qualitative assessment of TbRII-EDsubstitutions on binding affinity

The contribution of individual TbRII-ED residuesto TGFb1/b3 binding was assessed using nativegels. This was accomplished by combining twoequivalents of each TbRII-ED variant with oneequivalent of TGFb1 or TGFb3 homodimer. Thesemixtures, together with TbRII-ED alone controls,were then electrophoresed through 12% native gels.To reduce problems encountered with proteolysis ofthe N-terminal segment, all substitutions were

made in the background of TbRII-D14-ED, a formof the receptor lacking the poorly conserved 14N-terminal residues. To demonstrate that TbRII-D14-ED retained high affinity binding to TGFb1and TGFb3, as previously shown,21 complexes wereformed with TGFb1 or TGFb3 and then electro-phoresed on native gels as before. The results,shown in Figure 2(b) and (c), revealed well-definedcomplex bands regardless of whether thefull-length (subpanel labeled RII) or truncated(subpanel labeled RII-D14) receptorwas used, confirming that the N-terminal trunca-tion indeed has little or no effect on TGFb isoformbinding.

Representative TGFb1 and TGFb3 native gelbinding data for the 15 interfacial and sevennon-interfacial TbRII-D14-ED variants is shown inFigure 2(b) and (c), respectively (native gel bindingdata for all other variants is provided asSupplementary Data). For 11 interfacial variants,including L27A, F30A, D32A, D32N, I50A, T51A,S52L, I53A, E55A, E119A, and E119Q, complexformation with TGFb1 and TGFb3 was found to beeither completely eliminated (see, as an example,F30A and E119A) or significantly diminished (see,as an example, I53A) (Figure 2(b) and (c), respect-ively). For the remainder of the interfacial variants,including S49A, S52A, V77A, and D118A, complexformation with TGFb1 and TGFb3 was unaffected(see, as an example, S52A) (Figure 2(b) and (c),


respectively). For the non-interfacial variants, theresults were much different, in that all, with theexception of E75A, had no effect on complexformation with TGFb1 and TGFb3 (see, as anexample, V62A, H79A, and M112A) (Figure 2(b)and (c), respectively).

These results, even though qualitative, areconsistent with expectations based on the TbRII-ED:TGFb3 crystal structure. They show, forexample, that the receptor substitutions each yieldsimilar results when tested with either TGFb1 orTGFb3, consistent with the observation that resi-dues of TGFb3 contacted by TbRII in the TbRII-ED:TGFb3 crystal structure are identical to those inTGFb1.15 They also show that the interfacialvariants that affect complex formation are, for themost part, of residues that exhibit favorableelectrostatic, hydrogen-bonding, or hydrophobiccontacts in the crystal structure of the TbRII-ED:TGFb3 complex (Figure 1). Those that do notare either substituted with alanine in other species(Ser52) or lie on the periphery of the bindinginterface (S49A, V77A, and D118A) (Figure 1). Thefact that the non-interfacial variant, E75A, affectedcomplex formation, is consistent with the favorableelectrostatic and hydrogen-bonding interactionsthat it exhibits in the crystal structure (Figure 1).The fact that none of the other interfacial variantsaffected complex formation is consistent withexpectations from the structures that have beenreported;15,20 V62A, H79A, and M112A areespecially important in this regard as these showthat TbRII does not bind TGFb isoforms with theconcave surface of the receptor, as has been shownfor ActRIIa and ActRIIb.20,22

Quantitative assessment of TbRII-EDsubstitutions on binding affinities

The assessment of binding affinities describedabove provides valuable information as tothe identity of residues critical for binding. Thequalitative nature of these data however limits theconclusions that one can draw concerning theirrelative importance. To obtain further information,the effects of the receptor substitutions on ligand(TGFb1 and TGFb3) binding were quantifiedusing surface plasmon resonance (SPR). This wasaccomplished by immobilizing TGFb1 or TGFb3 andby recording the steady-state SPR response as varyingconcentrations of TbRII-ED variants were injectedover these surfaces (Figure 3(a)). Scatchard analysis ofthe maximal SPR response when steady state hadbeen reached as a function of receptor concentrationwas then used to obtain an equilibrium dissociationconstant, Kd, for each variant (Figure 3(b)). Such anapproach yielded Kd values of 200 and 290 nM forwild-type TbRII-ED interactions with TGFbs 1 and 3,respectively. These values are in good agreement withthose previously determined using an in-depthkinetic analysis of TbRII-ED binding to TGFb1(KdZ160 nM),23 thus validating our experimentalapproach.

The fitted Kd values obtained from the SPRanalysis are listed in Table 1. There were found tolie within the TbRII-ED concentration ranges usedfor all variants, except S52L, thus justifying the useof Scatchard plots for analysis of the variants aswell. For the S52L variant, weak binding togetherwith a limiting amount of available receptor,precluded us from flowing this TbRII-ED variantat concentrations above the apparent Kd value. Kd

values for this variant are therefore approximate,although it is nevertheless clear that S52L exhibitedthe lowest affinity among the variants tested.

The effects of the substitutions on bindingaffinities were assessed by calculating DDGbind

values, where DDGbindZDGwildtypeKDGvariant

where DGwildtype and DGvariant correspond to thefree energies for disassociation of ligand–receptorcomplexes with wild-type and variant receptor,respectively. These DDGbind values are presentedgraphically in Figure 4. This shows that thesubstitutions that led to the largest DDGbind valuesare those that also led to disruption of complexband formation on native gels (broken boxes). Thecorrelation in this regard is essentially perfect, asthe 12 variants (11 interfacial; one non-interfacial)that exhibited disrupted binding by native gels allhave DDGbind values above thresholds of 1.2 and1.4 kcal molK1 for TGFb1 and TGFb3, respectively,while all other variants had DDGbind values belowthese thresholds (Figure 4).

The 12 substitutions that have the largesteffects on affinity are in residues that assumecharacteristic positions within the binding inter-face. Five of these are hydrophobic residues,which form a linear array, Ile50-Phe30-Thr51-Ile53-Leu27, and which occupy a long butshallow hydrophobic cleft between the connectingloops at the tips of the TGFb fingers (Figure 1).The unique position of these residues within theinterface, together with the fact that theirsubstitution to alanine would diminish theirhydrophobic contact area with TGFb, supportsthe notion that these receptor residues contributeto affinity through hydrophobic contact. Thisnotion is further supported by the resultsobtained with Ser52. The substitution of thisresidue with alanine had only minor effects onaffinity (0.2–0.4 kcal molK1), whereas substitutionwith leucine drastically diminished affinity, thelargest change of any variant characterized(Figure 4). This is indicative of tight packing atthe interface as Ser52 is fully buried and lies atthe point of closest approach between thebackbones of TbRII and TGFb3 (5.5 A betweenthe Ca carbon of TbRII Ser52 and the backboneCa carbon of TGFb3 V92 and 3.8 A between theCa carbon of TbRII Ser52 and the carbonyl carbonof TGFb3 Tyr91). The methyl group of the alanineat this position is likely accommodated, whereasthe isopropyl group of the leucine, owing to itslarger volume, likely introduces steric overlapthat prevents the receptor and ligand from fullyengaging one another.

Figure 3. Quantitative analysis of TbRII-ED single amino substitutions on TGFb isoform binding affinities by surfaceplasmon resonance. (a) Typical binding curves obtained as increasing concentrations of a TbRII-D14-ED variant (F30A;left-to-right 0.0, 6, 15, 31, 46, and 60 mM) were injected over TGFb1 (red), TGFb3 (blue), and control (black) surfaces.(b) Scatchard plot used to determine the binding affinity based on the resonance units obtained at equilibrium, Req, forthe SPR data shown in (a). The data points shown correspond to the difference between the Req values obtained from (a)for the TGFb isoform and control surfaces, respectively. The continuous line corresponds to the best-fit line and has theform Req=½Receptor�ZK14;173!Req C4:0!106 ðr2 Z0:99Þ and Req=½Receptor�ZK6785!Req C2:0!106 ðr2Z0:99Þ forthe data shown in the left (TGFb1) and right panels (TGFb3), respectively. The absolute value of the slope corresponds tothe equilibrium binding constant, Ka or KK1

d , in units of MK1.


The remainder of the substitutions that effectaffinity are of charged residues, Asp32, Glu55,Glu75, and Glu119. In the crystal structure of theTbRII:TGFb3 complex, Asp32 and Glu119 lie onopposite sides of the hydrophobic array and engage,via hydrogen-bonded ion pairs, the guanidiniumgroups of the positively charged arginine residueslocated on the tips of the loops connecting the fingersof the ligand (Arg94 and Arg25, respectively).Conservative (Asp32 substituted with asparagine orGlu119 substituted with glutamine) or non-conserva-tive (Asp32 or Glu119 substituted with alanine)substitutions of either lead to comparableand significant disruption of binding affinity(1.4–2.0 kcal molK1) (Figure 4). This indicates thatAsp32:Arg94 and Glu119:Arg25 interact boththrough electrostatic attraction and through hydro-gen-bonding. This follows as substitution of thecarboxylate group of Asp32 or Glu119 with either

a methyl group (alanine) or an amide group(asparagine or glutamine) would be expected tofully disrupt the charge-charge interaction. The full orpartial disruption of the hydrogen-bonding interac-tions is also anticipated as the two oxygen atoms ofthe side-chain carboxylate of the parent residues areeach shown to function as hydrogen-bond acceptorsin the crystal structure of the TbRII:TGFb3 complex(hydrogen-bond donors in each case are the H3 andthe Hz atoms of the side-chain guanidinium groups ofthe two respective arginine residues). The substitutedmethyl group of alanine cannot obviously functionin this capacity, although neither can the amide groupof asparagine or glutamine, as this functionalityhas only one acceptor group, not two as in aspartateor glutamate.

Two other charged residues that contribute toaffinity, but to a lesser extent, are Glu55 andGlu75. Glu55, which affected the binding affinity

Table 1. Apparent dissociation constants for the interaction of wild-type and single amino acid human TbRII variantswith human TGFb1 and TGFb3

Protein variant Kd TGFb1 (mM)aDDGbind TGFb1

(kcal/mol)b Kd TGFb3 (mM)aDDGbind TGFb3

(kcal/mol)b

Wild-typeTbRII 0.20 (0.04) K0.3 0.29 (0.10) K0.5TbRII-D14 0.32 (0.01) 0.0 0.68 (0.23) 0.0

Single amino variants at interfacial positionsTbRII-D14 L27A 7.97 (2.12) 1.9 14.8 (0.6) 1.8TbRII-D14 F30A 44.0 (37.0) 2.9 104 (61) 3.0TbRII-D14 D32A 3.52 (0.46) 1.4 8.87 (0.41) 1.5TbRII-D14 D32N 5.90 (0.05) 1.7 19.9 (0.8) 2.0TbRII-D14 S49A 0.79 (0.14) 0.5 1.18 (0.12) 0.3TbRII-D14 I50A 7.96 (1.04) 1.9 16.7 (2.1) 1.9TbRII-D14 T51A 2.97 (0.30) 1.3 8.75 (1.28) 1.5TbRII-D14 S52A 0.60 (0.08) 0.4 0.98 (0.10) 0.2TbRII-D14 S52L 154 3.7 620 4.0TbRII-D14 I53A 2.49 (0.31) 1.2 6.87 (0.47) 1.4TbRII-D14 E55A 2.99 (0.28) 1.3 5.30 (0.88) 1.2TbRII-D14 V77A 0.69 (0.10) 0.5 1.37 (0.15) 0.4TbRII-D14 D118A 1.51 (0.21) 0.9 2.69 (0.19) 0.8TbRII-D14 E119A 3.35 (0.36) 1.4 8.47 (0.40) 1.5TbRII-D14 E119Q 5.95 (1.95) 1.7 10.6 (1.0) 1.6

Single amino acid variants at non-interfacial positionsTbRII-D14 N47A 0.67 (0.02) 0.3 1.10 (0.24) 0.3TbRII-D14 V62A 0.95 (0.12) 0.7 2.03 (0.05) 0.7TbRII-D14 E75A 1.64 (0.42) 1.0 4.21 (0.79) 1.1TbRII-D14 H79A 0.66 (0.07) 0.4 1.12 (0.22) 0.3TbRII-D14 F110A 1.51 (0.11) 0.9 3.28 (0.53) 0.9TbRII-D14 M112A 1.16 (0.09) 0.8 2.96 (0.69) 0.9TbRII-D14 I125A 0.77 (0.11) 0.5 1.69 (0.30) 0.5

a Measurements were made at 25 8C in buffer consisting of 20 mM Hepes containing 150 mM NaCl, 3.4 mM EDTA, and 0.005%Tween-20 at pH 7.4. The values indicated in parenthesis correspond to the standard deviation among three independentdeterminations via Scatchard analysis. The one exception to this is the TbRII-D14 S52L mutant for which one determination of the Kd

was performed.b DDGbindZDGwildtypeKDGvariant where DGwildtype and DGvariant correspond to the free energies for the disassociation of TGFb

isoform complexes (i.e. KRTln(Kd)) with wild-type TbRII-D14-ED and variants, respectively.


by 1.2–1.3 kcal molK1 when substituted withalanine, lies adjacent to Asp32 and forms along-range (ca 7 A) electrostatic interaction withArg94 of TGFb1/b3. Glu75, which affected thebinding affinity by 1.0–1.1 kcal molK1 when sub-stituted with alanine, appears to stabilize thecomplex by interacting with Arg94 of TGFb1/b3,although in this case indirectly via a water-mediated hydrogen-bond.

Assessment of TbRII-ED substitutionson antagonist activity in cultured cells

Previous results have shown that TbRII-ED canfunction to antagonize the biological activity ofTGFbs 1 and 3.24,25 This antagonistic activitypresumably results from the ability of TbRII-ED tobind and sequester the TGFb isoforms. Therefore,antagonistic potency and binding affinity shouldcorrelate. To test this, the antagonistic potency of thewild-type receptor and variants were measured byquantifying the level of Smad2 phosphorylation incultured MCF-10A cells as induced by 1 ng mlK1

TGFb3, either alone or in the presence of wild-typeor one of the 22 TbRII-ED-D14 variants. Theconcentrations of TbRII-D14-EDs used in thisexperiment, 50 mg mlK1, are somewhat higher

than used in previous studies,26 but were necessarybecause most of the variants, whether they carried asubstitution at an interfacial position or not,exhibited only weak antagonism at the usualconcentration (5 mg mlK1). This reduction inpotency is likely related to two factors: one is thatthe N-terminal truncation causes a small increase(ca factor of two) in the Kd values (Table 1); anotheris that most of the substitutions, whether interfacialor not, also cause at least a factor of two increase inTGFb isoform Kd values (Table 1). The latter effectmost likely arises due to minor perturbations instructure, which in turn, indirectly affect affinity.

The results of the Smad2 phosphorylation experi-ment are shown in Figure 5. This shows that, exceptfor E75A, all of the non-interfacial variantsantagonize the induction of Smad2 phosphoryl-ation (Figure 5, broken boxes). The fact that E75Adoes not antagonize is consistent with the results ofthe in vitro binding studies that showed that thisvariant exhibited weak complexes with TGFbs 1and 3 on native gels and a perturbation in itsbinding free energy (DDGbind) above the thresholdvalues of 1.2 and 1.4 kcal molK1 with theseisoforms, respectively. The 11 interfacial variantsthat disrupted complex formation on native gels(designated by a filled box) and whose DDGbind

Figure 4. Graphical representation of changes in TGFb isoform binding free energies (DDGbind) for the panel of 22single amino acid TbRII-D14-ED variants. DDGbind values were calculated from DDGbindZDGwildtypeKDGvariant whereDGwildtype and DGvariant correspond to the free energies for the disassociation of TGFb isoform complexes (i.e. KRTln(Kd)) with wild-type and variant TbRII-D14-ED, respectively. Error bars represent the standard deviation amongtriplicate determinations. Dissociation constant for the S52L mutant was measured only once and thus error estimatesare not available. Dotted lines mark thresholds that separate the variants that either did (unboxed) or did not (boxed)form complexes with TGFbs 1 and 3 in native gels.

Figure 5. Antagonistic potencies of TbRII-ED variants in a cell-based bioassay. Antagonistic potencies of the 22 singleamino acid TbRII-D14-ED variants were assessed by treating cultured MCF-10A cells with 1 ng mlK1 TGFb3 alone or inthe presence of one of the variants at 50 mg mlK1 for 30 min. The cells were then lysed and 20 mg of protein was thenanalyzed by Western blotting using either an anti-phospho-Smad2 (Ser465/Ser467) antibody (top panel) or an antibodyto assess equal loading (lower panel). The data shown were collected in two separate experiments; the first experimentincludes the controls and receptor variants included in lanes 1–20; the second includes the controls and receptor variantsincluded in lanes 21–28. The loading control antibody used for the first experiment was anti-GAPDH; the loading controlantibody used for the second experiment was MSH2. Continuous and broken boxes designate interfacial and non-interfacial variants, respectively. Asterisks designate variants that did not form complexes in native gels with eitherTGFbs 1 or 3.


values were higher than the threshold values of1.2 and 1.4 kcal molK1 with TGFbs 1 and 3,respectively, similarly did not antagonize TGFb3or did so very poorly. The remainder of theinterfacial variants, S49A, S52A, V77A, andD118A, that did exhibit complex formation onnative gels and whose DDGbind values werebelow the threshold values of 1.2 and1.4 kcal molK1 with TGFbs 1 and 3, respectively,did antagonize. These results show that themeasurements of binding affinity made in vitrocorrelate well with the potency of the TbRII-EDvariants as TGFb antagonists in cell-basedassays. This confirms that the mode of bindingdefined in vitro using native gels and SPRanalysis corresponds with the ability of the

receptors to bind and sequester TGFb3 from thecell surface TGFb receptors.

Assessment of the effects of the pointsubstitutions on stability

The differential effects of the interfacial and non-interfacial substitutions on binding affinitiessuggest that their effects are principally throughalterations in ligand contacts, not through effects onprotein stability. To examine this experimentally, theequilibrium stability of all TbRII-ED variants weremeasured using a GdmCl unfolding assay. This wasaccomplished by preparing multiple samples con-taining increasing concentrations of GdmCl and bymeasuring the steady-state fluorescence emission

Figure 6. Stability measurement of wild-type TbRII-EDand variants by guanidinium chloride unfolding. Thedata are presented as plots of the normalized intrinsicfluorescence emission intensity at 350 nm (ordinate)versus the concentration of GdmCl (abscissa). Individualsamples, used to obtain the corresponding data points inthe Figure for TbRII-D14-ED (circles), F30A TbRII-D14-ED, (triangles), V62A TbRII-D14-ED (squares), or E75ATbRII-D14-ED (diamonds) were prepared by eitheradding a stock solution of GdmCl to a sample of proteinin native buffer (open points, buffer comprised of 25 mMsodium phosphate, 0.10 M NaCl, pH 7.0) or by addingnative buffer to denatured protein (filled points, proteindissolved initially in 25 mM sodium phosphate, 0.10 MNaCl, 4.0 M GdmCl, pH 7.0). The data were then fit to atwo-state unfolding model (continuous lines) using theprogram ProFit (QuantumSoft, Zurich, Switerzland). Theparameters for the fits, along with error estimates, arelisted in Table 2.


from the single tryptophan residue, Trp65. Toshow that unfolding was reversible, samples wereprepared both by the addition of increasingconcentrations of denaturant to native protein orby the addition of non-denaturing buffer todenatured protein.

Typical data, obtained for the wild-type proteinand three variants, are shown in Figure 6. The“unfolding” (open) and “refolding” (filled)points superimpose, indicating that each undergoreversible transitions. The fact that the curves areS-shaped, even for highly destabilized variants suchas E75A, indicates that all of the variants are foldedand adopt stable tertiary structures in solution inthe absence of denaturant. The unfolding curveswere fitted to a two-state unfolding model toquantify alterations in stability. The fitted para-meters, along with error estimates, are listed inTable 2.

The 22 TbRII-ED variants exhibit a wide range ofunfolding free energies (DGU). Roughly one quarterof the variants, including F30A, D32N, S52A,S52L, and E55A, are more stable than wild-type(DDGU ! K0.3 kcal molK1). Another quarter of thevariants, including L27A, D32A, I53A, V77A, E119A,E119Q, and H79A, are not substantially alteredrelative to wild type (K0.3 kcal molK1!DDGU

!0.3 kcal molK1). The last two quarters, including

N47A, S49A, I50A, T51A, D118A, N47A, V62A, E75A,F110A, M112A, D118A, and I125A are less stable thanwild-type (DDGU O 0.3 kcal molK1) (Table 2).

The simplest way by which changes in stabilitymight alter binding affinity is by increasingthe proportion of the denatured form of the proteinpresent in solution at equilibrium. Thus, if themeasured affinities were determined principally bythe effects of the substitutions on stability, one wouldpredict that the variants that led to an increase instability, or no change in stability, would largely beunaffected in terms of their ligand binding affinities,whereas those variants that led to the largest changesin stability, would be most strongly affected. Thisclearly is not the case for the set of variants presentedhere as most of the variants that had the largest effectson affinity, such as L27A, F30A, D32A, D32N, T51A,S52L, E119A, and E119Q (Figure 4), were either ofcomparable or greater stability than wild-type(Table 2). Conversely, most of the variants that led toonly modest changes in affinity, such as S49A, N47A,F110A, M112A, and I125A (Figure 4), were amongthose that exhibited the largest decreases in stability(Table 2). This shows that the lowered affinities are notgenerally caused by lowered stability, although itdoes not preclude the possibility that such indirecteffects diminish affinities for some of the variants,especially those that are highly destabilized, such asI50A and E75A.

The other aspect of the stability data worth noting isthat there appears to be no systematic patternbetween the location of residues in the three-dimensional structure and their reduction in equili-brium stability. Thus, among the residues that aremost stabilized, one, I50, is exposed and lies near thecenter of the ligand binding site, two others, N47 andS49, are surface-exposed and lie adjacent to theligand binding site, while three others, F110A,M112A, and I125A, are buried and lie on the concavesurface of the protein opposite the ligand binding site.The lack of any type of consistent pattern providesfurther support for the notion that the effects ofthe substitutions on binding affinity are by and largedirect and not the consequence of a systematicalteration in structure.

TGFb2 K25R V92I K94R exhibits TGFb1/b3-likeaffinity for TbRII

The crystal structure of the 2:1 TbRII-ED:TGFb3complex revealed that the TGFb3 residues thatcontact TbRII are identical to those in the highaffinity ligand TGFb1, but are conservativelysubstituted at three positions (Arg25 to Lys, Val92to Ile, and Arg94 to Lys) in the low affinity ligandTGFb2.15 The results presented here have furthershown that substitution of the correspondingcontact residues in TbRII-ED, Glu119 which con-tacts Arg25 of TGFb1/b3, Ile53 which contactsVal92 of TGFb1/b3, and Asp32, Glu55, and Glu75which contact Arg94 of TGFb1/b3, each lead toperturbations in binding free energies (DDGbind) for

Table 2. Parameters for the guanidine hydrochloride unfolding of wild-type and mutant forms of TbRII

Protein varianta Cm (M) m (kcal molK1 MK1) DGU (kcal molK1) DDGU (kcal molK1)bFraction folded

(%)c

Wild-typeTbRII 1.39 (0.11) 2.82 (0.11) 3.91 (0.16) 99.9TbRII-D14 1.39 (0.20) 2.82 (0.18) 3.92 (0.28) 99.9

Single amino variants at interfacial positionsTbRII-D14 L27A 1.21 (0.11) 3.13 (0.14) 3.78 (0.19) 0.14 99.8TbRII-D14 F30A 1.35 (0.28) 3.56 (0.35) 4.82 (0.50) K0.90 100.0TbRII-D14 D32A 1.13 (0.72) 3.19 (0.87) 3.62 (1.14) 0.30 99.8TbRII-D14 D32N 1.07 (0.16) 4.76 (0.33) 5.10 (0.37) K1.18 100.0TbRII-D14 S49A 0.98 (0.27) 2.19 (0.23) 2.16 (0.35) 1.76 97.5TbRII-D14 I50A 0.56 (0.16) 2.64 (0.22) 1.47 (0.31) 2.45 91.9TbRII-D14 T51A 0.93 (0.15) 3.50 (0.25) 3.25 (0.29) 0.67 99.6TbRII-D14 S52A 1.39 (0.23) 3.51 (0.28) 4.87 (0.42) K0.95 100.0TbRII-D14 S52L 1.31 (0.10) 4.21 (0.16) 5.51 (0.22) K1.59 100.0TbRII-D14 I53A 1.33 (0.21) 2.78 (0.20) 3.71 (0.31) 0.21 99.8TbRII-D14 E55A 1.39 (0.25) 3.51 (0.31) 4.90 (0.45) K0.98 100.0TbRII-D14 V77A 1.36 (0.17) 2.98 (0.18) 4.06 (0.27) K0.14 99.9TbRII-D14 D118A 1.03 (0.24) 3.19 (0.33) 3.27 (0.42) 0.65 99.6TbRII-D14 E119A 1.08 (0.26) 3.36 (0.37) 3.64 (0.47) 0.28 99.8TbRII-D14 E119Q 1.31 (0.23) 2.82 (0.22) 3.69 (0.33) 0.23 99.8

Single amino acid variants at non-interfacial positionsTbRII-D14 N47A 1.01 (0.28) 2.74 (0.32) 2.76 (0.43) 1.16 99.1TbRII-D14 V62A 1.13 (0.11) 3.12 (0.14) 3.54 (0.19) 0.38 99.5TbRII-D14 E75A 0.60 (0.11) 2.42 (0.12) 1.46 (0.19) 2.46 91.8TbRII-D14 H79A 1.29 (0.10) 2.90 (0.11) 3.73 (0.16) 0.19 99.8TbRII-D14 F110A 0.35 (0.85) 2.67 (0.73) 0.92 (1.85) 3.00 79.3TbRII-D14 M112A 0.67 (0.12) 2.96 (0.18) 2.00 (0.22) 1.92 96.7TbRII-D14 I125A 0.81 (0.14) 2.86 (0.19) 2.32 (0.24) 1.60 98.1

a Samples were dissolved in 25 mM monobasic sodium phosphate containing 100 mM NaCl at pH 7.0. The unfolding experimentswere conducted at 21 8C.

b DDGUZDGU,WTKDGU,MUT, where DGU,WT and DGU,MUT are the free energies for the unfolding of TbRII-ED-D14 and TbRII-ED-D14variants, respectively, in the absence of guanidine hydrochloride.

c Fraction folded corresponds to the fraction of the protein that is folded in the absence of denaturant at a temperature of 21 8C.


TGFb1 or TGFb3 of 1.0 kcal molK1, or more(Table 1). This suggests that these three conservativesubstitutions in TGFb2 may indeed underlie the4.1 kcal molK1 lower affinity with which TbRIIbinds TGFb2.9–11 To evaluate this hypothesis,a TGFb2 variant, TGFb2 K25R I92V K94R orTGFb2-TM, was prepared and examined in vitrousing native gels and SPR.

The results of the native gel binding assay showthat when TbRII-ED and TGFb2-TM are mixedtogether and electrophoresed, a well-defined TbRII-ED:TGFb2-TM complex band is formed (Figure 7(a),lane 7). The complex band is also observed withTGFb3 (lane 5) as expected, but not with TGFb2 (lane6), demonstrating that the three substitutions restorehigh affinity binding to TGFb2. To quantify thischange, the affinity of TbRII-ED and TbRII-D14-EDfor TGFb2-TM was measured using SPR. This wasaccomplished in the same overall manner as beforeand yielded apparent Kd values of 0.48 and 1.3 mM forbinding of TGFb2-TM to TbRII-ED and TbRII-D14-ED, respectively. These Kd values are higher thanthose measured for TGFb1 by a factor of three to fourand are higher than those measured for TGFb3 by afactor of approximately two (Table 1). These threesubstitutions therefore confer TGFb2 with a bindingaffinity for TbRII that is only slightly reduced relativeto that of TGFbs 1 and 3.

Receptor binding and biological activity ofTGFb2 K25R V92I K94R in cultured cells

There exist a number of cell lines that have beenshown to exhibit a diminished biological responseto TGFb2.9,12,27–29 This diminished response hasbeen correlated with the absence of betaglycanexpression and has been attributed to the dimin-ished affinity of TbRII for this isoform.9,12 Therefore,in such cell lines, TGFb2-TM should elicit itsresponse with a potency that more closely matchesthat of TGFb3. To examine this experimentally, thegrowth inhibitory activity of TGFb2-TM wascompared with that of TGFb2 and TGFb3 incultured fetal bovine heart endothelial (FBHE)cells. This particular assay is one of the most widelyused to assess the biological activity of TGFb and isespecially appropriate for the present application asthese cells do not express betaglycan9 and havebeen shown to respond to TGFb3 at concentrationsthat are two to three orders of magnitude lowerthan that of TGFb2.9,30

The growth inhibition experiments were carriedout by culturing FBHE cells to mid log phase and bytreating them with varying concentrations ofTGFb3, TGFb2-TM, or TGFb2 in the presence ofan added radiolabeled nucleotide precursor(5-[125I]iodo-2 0-deoxyuridine). The results, shown

(b)(a)

(c)

Figure 7. Binding properties and growth inhibitory potency of TGFb2 K25R I92V K94R (TGFb2-TM). (a) TGFb2-TMbinary complex formation with TbRII-ED as assessed using native gel electrophoresis. The various ligands werecombined with TbRII-ED in the molar ratios shown, electrophoresed through a 12% native gel, and Coomassie stained(1 equivalent TGFb2Z1.5 mg; 1 equivalent TGFb3Z1.5 mg; 1 equivalent TGFb2-TMZ1.5 mg; 1 equivalent TbRII-EDZ0.9 mg). The positions of the TbRII:TGFb3 and TbRII:TGFb2-TM complexes are indicated with labels along the left side ofthe gel. (b) Growth inhibition of cultured FBHE cells in medium containing 10% fetal bovine serum and 5-[125I]iodo-2 0-deoxyuridine incubated for 24 h with the indicated concentrations of TGFb2 (filled circles), TGFb3 (filled squares), orTGFb2-TM (open squares). Incorporation of radiolabeled iodine into DNA was then measured and is expressed as thepercent decrease relative to the value in cultures not treated with TGFb isoforms. The data points and error barsrepresent the mean and standard deviation of four independent measurements. (c) Affinity labeling of cell-surface TbRIand TbRII by 125I-labeled TGFb1, TGFb2, TGFb2-TM, and TGFb3. L6E9 myoblast monolayers were incubated with theindicated concentrations of iodinated TGFb isoforms labeled with 125I to uniform specific activity (5.53!106 cpm/pmol)for 3 h at 4 8C, unbound ligand was removed by washing, and the cells were then treated with the crosslinking reageantdisuccinimidyl suberate. Cell lysates were then prepared and analyzed by SDS-PAGE and autoradiography.


in Figure 7(b), revealed midpoint inhibitory poten-cies of 8, 60, and 500 pM for TGFb3, TGFb2-TM, andTGFb2, respectively. TGFb2-TM, therefore, func-tions with a potency about tenfold greater thanTGFb2 and about tenfold less than TGFb3. Thetwofold lower affinity that TbRII exhibits forTGFb2-TM compared to TGFb3, may account forpart of the reduced growth inhibitory potency ofTGFb2-TM.

To further examine TGFb2-TM with regard to thisgain-of-function, affinity labeling experiments werecarried out with iodinated TGFb isoforms labeled tothe same specific activity and rat L6E9 myoblasts,

which lack betaglycan.12 Figure 7(c) shows, asreported before,12 that TGFb2 marginally bindsTbRII and TbRI (upper right), even at concen-trations of 250 pM, which are sufficient to produce aclear and distinct labeling with TGFb1 or TGFb3(upper and lower left, respectively). Importantly,the labeling of TbRI and TbRII by TGFb2-TM isclearly detectable at a concentration of 25 pM(lower left), comparable to that observed withTGFbs 1 and 3 (upper left and lower right,respectively). In spite of the qualitative nature ofthis assay, these results clearly affirm that the threeconservative substitutions in TGFb2-TM confer this


variant with high affinity binding, like that ofTGFbs 1 and 3, to the cell-surface TGFb receptors.

Discussion

The differential affinity of TbRII for the TGFbisoforms has been well documented in the litera-ture.9,31,32 The differences in affinity have recentlybeen quantified using mimicry of TbRII:TGFb cellsurface interactions, coupled with SPR biosensoranalysis.10,11 These studies have shown that the Kd

of TbRII for TGFb1 and TGFb3 (5–30 pM) isapproximately 1000-fold lower than that forTGFb2 (5 nM). This corresponds to a reduction inthe binding free energy of the receptor for TGFb2 ofabout 4.1 kcal molK1.

The Kd values measured using these approaches,in which the TbRII extracellular domain is dimer-ized and immobilized on the biosensor surface,10,11

are in general much lower than those measuredwhen the TbRII extracellular domain is monomericand in solution,23,33 as in this study. This difference,which is some four orders of magnitude for TGFb1and TGFb3, and undetermined for TGFb2 due toweak binding, is attributed to the fact that TGFbisoforms have two equivalent non-interacting sitesfor TbRII, and that avidity effects contributesignificantly to binding.10,11

The crystal structure of the 2:1 TbRII-ED:TGFb3complex has shown that the ten TGFb3 residuesthat contact TbRII are identical to those in TGFb1,but are substituted at three positions in the lowaffinity ligand, TGFb2.15 The three TGFb1/b3 toTGFb2 substitutions are Arg25 to lysine, Val92 toisoleucine, and Arg94 to lysine. Through studies ofTGFb isoform chimeras, it has been shown thatreplacement of Val92-Gly93-Arg94-Lys95 in TGFb1with Ile92-Gly93-Lys94-Thr95 from TGFb2 weakensbinding to the TbRII extracellular domain anddecreases potency in a LS513 colon cancer cellgrowth inhibition assay.32 Taken together, thesefindings suggest that the three substituted residues,in spite of their similar nature, might underlie thediminished affinity of TGFb2 for TbRII.15

To investigate this hypothesis, mutagenesis wasused to identify residues within the TbRII extra-cellular domain that contribute significantly toTGFb isoform binding. These studies were carriedout by generating single amino acid substitutionswithin the TbRII extracellular domain, and byassessing stabilities by denaturation with GdmCland affinities for binding the high affinity TGFbisoforms, TGFbs 1 and 3, using native gel electro-phoresis and SPR (such methods could not be usedto assess affinities for TGFb2 due to weak binding).The affinities determined by the two techniqueswere fully consistent with one another and withrelative potencies determined in a cell-based Smadphosphorylation assay. The substitutions did lowerthe stability of several of the receptor variants, butfor the most part, these effects had only negligibleeffects on the measured affinities.

The receptor residues that contact the threevariable residues of the ligand were found to beamong those that contributed greatest to TGFbisoform binding. Thus, Ile53 of TbRII, which is theprimary contact with Val92 of TGFb1/b3, led to adecrease in affinity of 1.2–1.4 kcal molK1 whensubstituted with alanine; Glu119 of TbRII, whichis the primary contact with Arg25 of TGFb1/b3, ledto a decrease in affinity of 1.4–1.7 kcal molK1 whensubstituted with alanine or glutamine; Asp32,which is in primary contact with Arg94 of TGFb1/b3, led to a decrease in affinity of 1.4–2.0 kcal molK1

when substituted with alanine or asparagine; Glu55and Glu75 of TbRII, which are the secondary contactswith Arg94 of TGFb1/b3, each led to a decrease inaffinity of 1.0–1.3 kcal molK1 when substituted withalanine (Table 1). These findings are important,because through additive effects, they provide aplausible mechanism to account for the estimated4.1 kcal molK1 lower affinity with which TbRII bindsTGFb2.

The studies of the receptor variants also providedinformation regarding the specificity of the inter-actions. Thus, comparable and significant effects onaffinity were observed when Asp32 and Glu119 ofthe Asp32 TbRII-Arg94 TGFb1/b3 and Glu119TbRII-Arg25 TGFb1/b3 hydrogen-bonded ionpairs, respectively, were conservatively (Asp32substituted to asparagine and Glu119 substitutedto glutamine) or non-conservatively (Asp32 andGlu119 both substituted with alanine) substituted.Thus, substitution of Asp32 with either alanine orasparagine led to a decrease in binding affinity of1.4–1.7 kcal molK1 and 1.5–2.0 kcal molK1 forTGFb1 and TGFb3, respectively (Table 1). Thesubstitution of Glu119 with either alanine orglutamine led to a decrease in binding affinity of1.4–1.7 kcal molK1 and 1.5–1.6 kcal molK1 forTGFb1 and TGFb3, respectively (Table 1). This, asarticulated in Results, is likely a consequence of thefact that both the electrostatic and hydrogen-bonding components contribute to the interaction.This conclusion is important because it provides aplausible explanation as to why the conservativesubstitution of arginine residues (Arg25, Arg94)with lysine on the ligand side also leads to asignificant disruption of binding affinity, assuggested by the results reported by Burmesterand co-workers.32

To further evaluate the overall hypothesis, thethree variable receptor contact residues in TGFb2were changed to match those in TGFbs 1 and 3. ThisTGFb2 variant, designated TGFb2-TM, was shownthrough native gel electrophoresis and SPR to bindTbRII-ED with an affinity just twofold lower thanthat of TGFb3 and three to fourfold lower thanTGFb1. This shows unequivocally that the threesubstitutions underlie the diminished affinity ofTGFb2 for TbRII. The underlying source of theremaining two to fourfold lower affinity is notknown at present, although possible sourcesinclude slight differences in the structural frame-work of TGFb2 relative to TGFbs 1 and 3,34,35


and/or the possible influence of non-interfacialresidues that differ between TGFb2 and the otherisoforms. To examine whether the enhanced affinitybetween TGFb2-TM and TbRII-ED translated intoenhanced biological activity, the potency of thisvariant was compared to that of TGFb2 and TGFb3in a fetal bovine heart endothelial (FBHE) cellgrowth inhibition assay. This experiment showedthat TGFb2-TM exhibited a growth inhibitorypotency intermediate between that of TGFb2 andTGFb3. This increase in potency, which was highlyreproducible and well beyond experimentalerror of that measured for TGFb2, is consistentwith the binding data with TbRII-ED as well as theefficient crosslinking observed between TGFb2-TMand cell-surface TbRII. The fact that the growthinhibitory potency of TGFb2-TM was below thatmeasured for TGFb3 is attributed in part to thetwofold lower affinity with which TGFb2-TM bindsTbRII-ED.

The results presented here extend the previousinformation available concerning the structuralbasis of TGFb receptor assembly10,11,15,32,36 byidentifying critical contacts used by TbRII tocontact the ligand and by demonstrating that thethree conservatively substituted residues inTGFb2 indeed underlie its lowered affinity forTbRII. These findings, together with previousfindings reported by others,10,11,15,32,36 suggestthat TGFb2 is no different from TGFb1 orTGFb3 in either its overall structure or its modeof interaction with the TGFb signaling receptors.TGFb2 has however apparently co-evolved withTbRII in such a way that its affinity for the typeII receptor has dropped by about 4.1 kcal molK1

so that it no longer binds with sufficient affinityto enable responses at biologically relevantconcentrations (picomolar to subpicomolar). Theaffinity, however, has evidently not dropped somuch that it cannot be compensated in a yet-to-be-defined manner by the co-receptor betaglycan.These differences in receptor binding propertiesare likely significant, as this has presumablyenabled further diversification of functionamongst the TGFb isoforms through cell-specificexpression of the co-receptor betaglycan.

Materials and Methods

Expression and purification of TbRII-ED and variants

Two different wild-type TbRII extracellular domainconstructs, designated TbRII-ED and TbRII-D14-ED,were prepared for the current study. These correspondto either the entire extracellular domain of the humanreceptor or a truncated variant lacking 14 residues onthe N terminus.37 The former is 136 residues in length,with the first residue designated as 1, while the latteris 122 residues in length, with the first residuedesignated as 15. All variants were constructed in theTbRII-D14-ED background by PCR mutagenesis andwere expressed and purified in the manner described,38

except 1 mM EDTA, 1 mM leupeptin, 1 mM PMSF,

1 mM benzamidine, and 0.5 mM N-a-tosyl-L-lysyl-chloromethyl ketone were included in the finalpurification step. The identity of wild-type TbRII-EDand variants were verified by dialyzing small aliquotsinto distilled water and by determining the mass usingan electrospray mass spectrometer (LCQ, ThermoFinnigan, San Jose, CA).

Native gel binding assay

Two equivalents of TbRII-ED in 25 mM Mes (pH 6.0)was mixed with one equivalent of human TGFb1, TGFb2,or TGFb3 in 40 mM acetic acid. TGFb1 was produced byCHO cell expression and was purchased from R&DSystems (Minneapolis, MN). TGFb2, TGFb2 K25R I92VK94R, and TGFb3 were produced by Escherichia coliexpression, oxidatively folded to the disulfide-linkedhomodimer, and HPLC purified as described.39 Proteinconcentrations were determined using the correspondingcalculated extinction coefficients40 and the measured UVabsorbance at 280 nm. Receptor-ligand mixtures, alongwith receptor and ligand alone controls, were analyzed byaddition of an equal volume of 2x native gel sample buffer(50 mM Tris–HCl (pH 8.45), 20% (v/v) glycerol) and byelectrophoresing at 22 8C for approximately 1 h through12% (w/v) native polyacrylamide gels containing a short(1 cm) stacking gel buffered with 0.13 M Tris–HCl atpH 6.8 and a longer (6 cm) running gel buffered with0.38 M Tris–HCl at pH 8.8.

Surface plasmon resonance measurements

CM5 sensor chips, N-hydroxysuccinimide(NHS), N-ethyl-N 0-(3-diethylaminopropyl) carbodiimidehydrochloride (EDC), and 1 M ethanolamine (pH 8.5)were purchased from Biacore Incorporated (Piscataway,NJ). Recombinant human TGFb1 and human TGFb3expressed in CHO and Sf21 cells, respectively, werepurchased from R&D Systems (Minnepolis, MN).Recombinant human TGFb2 K25R I92V K94R wasexpressed in E. coli, refolded, and purified as describedabove.

TGFb isoforms were coupled to CM-5 sensor chipsurfaces through the standard amine coupling pro-cedure at a flow rate of 5 ml minK1. Sequentialinjections consisted of a 0.05 M NHS and 0.2 M-EDCmixture (30 ml) followed by solutions of TGFb isoforms(0.8–1.6 mg/ml) in 10 mM formic acid (pH 4.0) untilthe amount of coupled TGFb yielded between 2.5!103

to 5.0!103 resonance units (RUs). A solution of 0.1 Methanolamine was then used to block the remainingactivated sites. Control dextran surfaces were generatedby replacing the TGFb solutions with formic acidalone.

All SPR measurements were made at 25 8C at a flowrate of 5 ml minK1. Running buffer (20 mM Hepes(pH 7.4), 150 mM NaCl, 3.4 mM EDTA, 0.005% Tween-20; HBS), was used to dilute all TbRII-ED variantsinjected on the TGFb isoform surfaces. Binding proper-ties were determined by injecting buffer solutions,together with five different concentrations of the TbRII-ED variants, over the TGFb isoform surfaces as well asover the control surfaces for 300 s, after which time theTbRII-ED solutions were replaced by HBS buffer for 50 s.The concentrations of the TbRII-ED variants ranged from0.6 mM to 6 mM for N47A, S49A, S52A, I53A, T51A,V77A, and M112A, from 1 mM to 12 mM for D32A, I50A,E55A, V62A, E75A, F110A, and I125A, from 0.9 mM to


9 mM for E119A, from 3 mM to 30 mM for TbRII-D14, from2 mM to 20 mM for L27A and E119Q, from 4 mM to 25 mMfor D32N, from 5 mM to 40 mM for S52L, from 6 mM to60 mM for F30A and H79A, and from 0.065 mM to 0.5 mMfor full-length wild-type TbRII-ED. The TGFb isoformsurfaces were regenerated between each series ofinjections using two 50 ml pulses of 20 mM HCl at aflow rate of 100 ml minK1, followed by an EXTRACLEANand a RINSE procedure carried out according to theBIACORE manual.

Resonance unit values at steady-state, Req, for eachTbRII-ED variant were determined by subtracting thevalue obtained using the TGFb isoform surface from thecorresponding value obtained using the control surface.The apparent dissociation constants, Kd, were thendetermined by Scatchard analyses. The final set of Kd

values and errors were determined through analyses ofall variants in triplicate. The one exception was the S52Lvariant, whose Kd value was determined through analysisof a single Scatchard plot. This was necessary as thisvariant bound weakly and as a consequence significantlygreater quantities of receptor were required to completethe measurements.

Stability measurements

TbRII-ED stabilities were measured by monitoring theintrinsic tryptophan fluorescence as a function of theguanidinium chloride (GdmCl) concentration. This wasaccomplished by preparing a total of 20–30 3.00 mlsamples, each consisting of a fixed amount of protein(60 mg) mixed with a known concentration of GdmCl in25 mM NaH2PO4 (pH 7.0), 100 mM NaCl. The reversi-bility of the folding-unfolding transition was investigatedby adding protein in non-denaturing buffer to solutionscontaining increasing concentrations of GdmCl, or byadding protein in 4.0 M GdmCl to solutions of non-denaturing buffer with decreasing concentrations ofGdmCl. The fluorescence intensities were then measuredat 21 8C with an SLM 500C fluorimeter equipped with athermostated cell holder. Excitation was by means of atungsten lamp with a 2 nm bandwidth at 285 nm. Theemission wavelength was 350 nm with a 5 nm band-width. The raw fluorescence intensities were fitted in anon-linear fashion to a two-state unfolding model withlinear baselines to yield DGU, the free energy of unfoldingin the absence of denaturant, and the m-value, defined asdðDGUÞ=d ½GdmCl�ð Þ.41 The concentration of the GdmCl atthe midpoint of the unfolding transition, Cm, wascalculated from the ratio of DGU to the m-value.41

Smad phosphorylation assays

Smad2 phosphorylation assays26 were performed bytreating exponentially growing human mammaryepithelial MCF-10A cells with 1 ng mlK1 TGFb3 and50 mg mlK1 wild-type TbRII-D14-ED or variant for30 min. The cells were rinsed twice with ice-cold PBSand lysed in 50 mM Tris–HCl (pH 7.4), 150 mM NaCl,1% Nonidet P-40 containing protease (Roche) andphosphatase (1 mM NaVO3 and 1 mM NaF) inhibitors.Twenty micrograms of total protein were then separa-ted with SDS-PAGE and transferred to a HyBond-ECLnitrocellulose membrane (Amersham Biosciences,Piscataway, NJ). The membrane was blocked withTBST (100 mM Tris–HCl (pH 8.0), 150 mM NaCl,0.05% Tween-20) containing 5% (w/v) non-fat driedmilk and then incubated with a rabbit polyclonal anti-

phospho-Smad2 (Ser-465/467) antibody (Upstate Bio-technology) overnight or an anti-GAPDH antibody(Ambion) for 1 h. After three washes with TBST, themembrane was incubated with HRP-linked anti-rabbitor anti-mouse antibody (Santa Cruz) for 1 h andwashed again. Bound complexes were visualizedusing chemiluminescence procedures (NEN LifeScience Products, Boston, MA).

Growth inhibition assays

The effect of TGFb isoforms on the proliferation ofcultured fetal bovine heart endothelial (FBHE) cells wastested as described by Cheifetz et al.9 Briefly, FBHE cellswere seeded in 24-well plates and cultured withDulbecco’s modified Eagle’s medium (DMEM) supple-mented with 10% (v/v) Fetal Bovine Serum (Gibco-BRL,Carlsbad, CA) and 100 mg mlK1 endothelial cell mitogen(Biomedical Technologies, Stoughton, MA). Once FBHEcells reached a density of 50,000 cells per well, TGFbisoforms were added at the indicated concentrations infresh media without endothelial mitogen, and the cellswere then incubated for another 48 h. DNA synthesis wasmeasured during the last 8 h of the TGFb isoformtreatment by pulsing the cells with 5-[125I]iodo-2 0-deoxyuridine (Amersham, Piscataway, NJ) at 1 mCi mlK1.

Affinity labeling

TGFb1, TGFb2, TGFb2 K25R I92V K94R, and TGFb3were labeled with 125I as described.9 Iodinated TGFbs,added with sufficient unlabeled ligand to have them atthe same specific radioactivity (5.53!106 cpm/pmol),were used to affinity label L6E9 rat skeletal myoblasts asdescribed.42 Briefly, confluent monolayers of L6E9 cellswere incubated with the radiolabeled TGFb isoforms atconcentrations of 5, 25, 50, or 250 pM for 3 h at 4 8C. Afterextensive washing of the unbound ligand, disuccinimidylsuberate (Pierce, Rockford, IL) was added to cross-linkthe receptor-bound ligand complexes, which weresolubilized in the presence of Triton X-100. Thesecomplexes were displayed by subjecting the TritonX-100 cell extracts to SDS-PAGE followed by phosphor-imager autoradiography.

Acknowledgements

Dr Susan Weintraub is thanked for performingthe electrospray mass spectrometry measure-ments. Dr Jay Groppe and Dr John Lee arethanked for critically reading the manuscript.Financial support was provided by the NIH(GM58670) and the Robert A. Welch Foundation(AQ1431 to A.P.H.). Additional support wasprovided by the NIH (RR13879 to A.P.H.;CA79683 to L.-Z.S.; CA54174 to the Macromol-ecular Structure and Mass Spectrometry SharedResources of the San Antonio Cancer Institute),the Howard Hughes Medical Institute (Inter-national Research Scholar Grant to F.L.-C.), andthe Consejo National de Ciencia y Tecnologıa(37749N to F.L.-C.).


Supplementary Data

Supplementary data associated with this articlecan be found, in the online version, at doi:10.1016/j.jmb.2005.10.022

References

1. Roberts, A. B. & Sporn, M. B. (1990). The transform-ing growth factor-betas. In Peptide Growth Factors andtheir Receptors (Roberts, A. B. & Sporn, M. B., eds),pp. 421–472, Springer, Heidelberg, Germany.

2. Derynck, R. (1994). TGF-b-receptor-mediated signal-ing. Trends Biochem. Sci. 19, 548–553.

3. Massague, J. (1998). TGF-b signal transduction. Annu.Rev. Biochem. 67, 753–791.

4. Massague, J., Blain, S. W. & Lo, R. S. (2000). TGFbsignaling in growth control, cancer, and heritabledisorders. Cell, 103, 295–309.

5. Shull, M. M., Ormsby, I., Kier, A. B., Pawlowski, S.,Diebold, R. J., Yin, M. Y. et al. (1992). Targeteddisruption of the mouse transforming growth factor-b-1 gene in multifocal inflammatory disease. Nature,359, 693–699.

6. Sanford, L. P., Ormsby, I., Gittenberger-de Groot, A. C.,Sariola, H., Friedman, R., Boivin, G. P. et al. (1997).TGFbeta2 knockout mice have multiple develop-mental defects that are non-overlapping with otherTGFbeta knockout phenotypes. Development, 124,2659–2670.

7. Proetzel, G., Pawlowski, S. A., Wiles, M. V., Yin, M.,Boivin, G. P., Howles, P. N. et al. (1995). Transforminggrowth factor-b3 is required for secondary palatefusion. Nature Genet. 11, 409–414.

8. Wrana, J. L., Attisano, L., Carcamo, J., Zentella, A.,Doody, J., Laiho, M. et al. (1992). TGF-b signalsthrough a heteromeric protein kinase receptorcomplex. Cell, 71, 1003–1014.

9. Cheifetz, S., Hernandez, H., Laiho, M., ten Dijke, P.,Iwata, K. K. & Massague, J. (1990). Distincttransforming growth factor-beta (TGF-b) receptorsubsets as determinants of cellular responsivenessto three TGF-b isoforms. J. Biol. Chem. 265,20533–20538.

10. De Crescenzo, G., Pham, P. L., Durocher, Y., Chao, H.& O’Connor-McCourt, M. D. (2004). Enhancement ofthe antagonistic potency of transforming growthfactor-b receptor extracellular domains by coiledcoil-induced homo- and heterodimerization. J. Biol.Chem. 279, 26013–26018.

11. De Crescenzo, G., Pham, P. L., Durocher, Y. &O’Connor-McCourt, M. D. (2003). Transforminggrowth factor-beta (TGF-b) binding to the extracellu-lar domain of the type II TGF-beta receptor: receptorcapture on a biosensor surface using a new coiled-coilcapture system demonstrates that avidity contributessignificantly to high affinity binding. J. Mol. Biol. 328,1173–1183.

12. Lopez-Casillas, F., Wrana, J. L. & Massague, J. (1993).Betaglycan presents ligand to the TGF b signalingreceptor. Cell, 73, 1435–1444.

13. Yamashita, H., ten Dijke, P., Franzen, P., Miyazono, K.& Heldin, C.-H. (1994). Formation of hetero-oligo-meric complexes of type I and type II receptors fortransforming growth factor-b. J. Biol. Chem. 269,20172–20178.

14. Gilboa, L., Wells, R. G., Lodish, H. F. & Henis, Y. I.(1998). Oligomeric structure of type I and type IItransforming growth factor b receptors: homodimersform in the ER and persist at the plasma membrane.J. Cell Biol. 140, 767–777.

15. Hart, P. J., Deep, S., Taylor, A. B., Shu, Z., Hinck, C. S.& Hinck, A. P. (2002). Crystal structure of the humanTbR2 ectodomain - TGF-b3 complex. Nature Struct.Biol. 9, 203–208.

16. Bocharov, E. V., Korzhnev, D. M., Blommers, M. J.,Arvinte, T., Orekhov, V. Y., Billeter, M. & Arseniev,A. S. (2002). Dynamics-modulated biological activityof transforming growth factor beta 3. J. Biol. Chem.277, 46273–46279.

17. Deep, S., Walker, K. P., 3rd, Shu, Z. & Hinck, A. P.(2003). Solution structure and backbone dynamics ofthe TGFb type II receptor extracellular domain.Biochemistry, 42, 10126–10139.

18. Kirsch, T., Sebald, W. & Dreyer, M. K. (2000). Crystalstructure of the BMP-2-BRIA ectodomain complex.Nature Struct. Biol. 7, 492–496.

19. Gray, P. C., Greenwald, J., Blount, A. L., Kunitake,K. S., Donaldson, C. J., Choe, S. & Vale, W. (2000).Identification of a binding site on the type II activinreceptor for activin and inhibin. J. Biol. Chem. 275,3206–3212.

20. Greenwald, J., Groppe, J., Gray, P., Wiater, E.,Kwiatkowski, W., Vale, W. & Choe, S. (2003). TheBMP7/ActRII extracellular domain complex providesnew insights into the cooperative nature of receptorassembly. Mol. Cell. 11, 605–617.

21. Guimond, A., Sulea, T., Pepin, M. C. & O’Connor-McCourt, M. D. (1999). Mapping of putative bindingsites on the ectodomain of the type II TGF-b receptorby scanning-deletion mutagenesis and knowledge-based modeling. FEBS Letters, 456, 79–84.

22. Thompson, T. B., Woodruff, T. K. & Jardetzky, T. S.(2003). Structures of an ActRIIB:activin A complexreveal a novel binding mode for TGF-b ligand:receptor interactions. EMBO J. 22, 1555–1566.

23. De Crescenzo, G., Grothe, S., Zwaagstra, J., Tsang, M.& O’Connor-McCourt, M. D. (2001). Real-time moni-toring of the interactions of transforming growthfactor-beta (TGF-b) isoforms with latency-associatedprotein and the ectodomains of the TGF-beta type IIand III receptors reveals different kinetic models andstoichiometries of binding. J. Biol. Chem. 276,29632–29643.

24. Tsang, M., Zhou, L., Zheng, B., Wenker, J., Fransen, G.,Humphrey, J. et al. (1995). Characterization ofrecombinant soluble human transforming growthfactor-b receptor type II (rhTGF-b sRII). Cytokine, 7,389–397.

25. Lin, H., Moustakas, A., Knaus, P., Wells, R., Henis, Y.& Lodish, H. (1995). The soluble exoplasmic domainof the type II transforming growth factor (TGF)-betareceptor. A heterogenously glycosylated protein witha high affinity and selectivity for TGF-beta ligands.J. Biol. Chem. 270, 2747–2754.

26. Lei, X., Bandyopadhyay, A., Le, T. & Sun, L. (2002).Autocrine TGFb supports growth and survival ofhuman breast cancer MDA-MB-231 cells. Oncogene,21, 7514–7523.

27. Rosa, F., Roberts, A. B., Danielpour, D., Dart, L. L.,Sporn, M. B. & Dawid, I. B. (1988). Mesoderminduction in amphibians: the role of TGF-b 2-likefactors. Science, 239, 783–785.

http://dx.doi.org/doi:10.1016/j.jmb.2005.10.022

http://dx.doi.org/doi:10.1016/j.jmb.2005.10.022


28. Tsunawaki, S., Sporn, M., Ding, A. & Nathan, C.(1988). Deactivation of macrophages by transforminggrowth factor-b. Nature, 334, 260–262.

29. Jennings, J. C., Mohan, S., Linkhart, T. A., Widstrom, R.& Baylink, D. J. (1988). Comparison of the biologicalactions of TGF-b-1 and TGF-b-2: differential activity inendothelial cells. J. Cell Physiol. 137, 167–172.

30. Qian, S. W., Burmester, J. K., Sun, P. D., Huang, A.,Ohlson, D. J., Suardet, L. et al. (1994). Characterizationof mutated transforming growth factor-bs whichpossess unique biological properties. Biochemistry,33, 12298–12304.

31. Qian, S. W., Burmester, J. K., Tsang, M. L.-S.,Weatherbee, J. A., Hinck, A. P., Ohlsen, D. J. et al.(1996). Binding affinity of transforming growthfactor-b to the type II receptor is determined by theC-terminal region of the molecule. J. Biol. Chem. 271,30656–30662.

32. Burmester, J. K., Qian, S. W., Ohlsen, D., Phan, S.,Sporn, M. B. & Roberts, A. B. (1998). Mutationalanalysis of a transforming growth factor-b receptorbinding site. Growth Factors, 15, 231–242.

33. O’Connor-McCourt, M. D., Segarini, P., Grothe, S.,Tsang, M. L.-S. & Weatherbee, J. A. (1995). Analysis ofthe interaction between two TGF-b-binding proteinsand three TGF-b isoforms using surface plasmonresonance. Ann. NY Acad. Sci. 766, 300–302.

34. Hinck, A. P., Archer, S. J., Qian, S. W., Roberts, A. B.,Sporn, M. B., Weatherbee, J. A. et al. (1996).Transforming growth factor b1: three-dimensional

structure in solution and comparison with the X-raystructure of transforming growth factor b2. Biochem-istry, 35, 8517–8534.

35. Mittl, P. R. E., Priestle, J. P., Cox, D. A., McMaster, G.,Cerletti, N. & Grutter, M. G. (1996). The crystalstructure of TGF-b3 and comparison to TGF-b2:Implications for receptor binding. Protein Sci. 5,1261–1271.

36. Wrana, J. L., Attisano, L., Wiesner, R., Ventura, F. &Massague, J. (1994). Mechanism of activation of theTGF-b receptor. Nature, 370, 341–347.

37. Lin, H. Y., Wang, X.-F., Elinor, N.-E., Weinberg, R. A. &Lodish, H. F. (1992). Expression cloning of the TGF-btype II receptor, a functional transmembrane serine/threonine kinase. Cell, 68, 775–785.

38. Hinck, A. P., Walker, K. P., 3rd, Martin, N. R., Deep, S.,Hinck, C. S. & Freedberg, D. I. (2000). Sequentialresonance assignments of the extracellular ligandbinding domain of the human TGF-b type II receptor.J. Biomol. NMR, 18, 369–370.

39. Cerletti, N. (2000). Process for the production ofbiologically active dimeric protein, US Patent6057430.

40. Gill, S. C. & von Hippel, P. H. (1989). Calculation ofprotein extinction coefficients from amino acidsequence data. Anal. Biochem. 182, 319–326.

41. Pace, C. N. (1975). The stability of globular proteins.CRC Crit. Rev. Biochem. 3, 1–43.

42. Massague, J. (1987). Identification of receptors fortype-b transforming growth factor. Methods Enzymol.146, 174–195.

Edited by P. T. Lansbury Jr

(Received 30 April 2005; received in revised form 16 August 2005; accepted 6 October 2005)Available online 2 November 2005

Three Key Residues Underlie the Differential Affinity of the TGFβ Isoforms for the TGFβ Type II Receptor

Documents