ScFv multimers of the anti-neuraminidase antibody NC10: shortening of the linker in single-chain Fv fragment assembled in VL to VH orientation drives the formation of dimers, trimers, tetramers and higher molecular mass multimers

Synthetic genes encoding single-chain variable fragments (scFvs)of NC10 anti-neuraminidase antibody were constructed by joiningthe V_L and V_H domains with linkers of fifteen, five, four, three,two, one and zero residues. These V_L–V_H constructs wereexpressed in Escherichia coli and the resulting proteins werecharacterized and compared with the previously characterizedNC10 scFv proteins assembled in V_H–V_L orientation. Size-exclusionchromatography and electron microscope images of complexes formedbetween various NC10 scFvs and anti-idiotype Fab' were usedto analyse the oligomeric status of these scFvs. The resultshowed that as the linker length between V_L and V_H was reduced,different patterns of oligomerization were observed comparedwith those with V_H–V_L isomers. As was the case for V_H–V_Lorientation, the scFv-15 V_L–V_H protein existed mainlyas a monomer whereas dimer (diabody) was a predominant conformationfor the scFv-5, scFv-4 and scFv-3 V_L–V_H proteins. In contrastto the V_H–V_L isomer, direct ligation of V_L to V_H led tothe formation of predominantly a tetramer (tetrabody) ratherthan to an expected trimer (triabody). Furthermore, the transitionbetween dimers and higher order oligomers was not as distinctas for V_H–V_L. Thus reducing the linker length in V_L–V_Hfrom three to two residues did not precisely dictate a transitionbetween dimers and tetramers. Instead, two-residue as well asone-residue linked scFvs formed a mixture of dimers, trimersand tetramers.     

scFv multimers of the anti-neuraminidase antibody NC10: length of the linker between VH and VL domains dictates precisely the transition between diabodies and triabodies

Single-chain Fv antibody fragments (scFvs) incorporate a polypeptidelinker to tether the V_H and V_L domains together. An scFv moleculewith a linker 5–12 residues long cannot fold into a functionalFv domain and instead associates with a second scFv moleculeto form a bivalent dimer (diabody). Direct ligation of V_H andV_L domains further restricts association and forces three scFvmolecules to associate into a trivalent trimer (triabody). Wehave defined the effect of linker length on scFv associationby constructing a series of scFvs from anti-neuraminidase antibodyNC10 in which the linker varied from one to four glycine residues.NC10 scFv molecules containing linkers of three and four residuesshowed a strong preference for dimer formation (diabodies),whereas a linker length of one or two glycine residues preventedthe formation of diabodies and directed scFv association intotrimers (triabodies). The data suggest a relatively strict transitionfrom dimer (diabody) to trimer (triabody) upon reduction ofthe linker length from three to two glycine residues. Modellingstudies are consistent with three residues as the minimum linkerlength compatible with diabody formation. Electron microscopeimages of complexes formed between the NC10 scFv multimers andan anti-idiotype Fab' showed that the dimer was bivalent forantigen binding and the trimer was trivalent.     

Properties of a single-chain antibody containing different linker peptides

Single-chain antibodies were constructed using six differentlinker peptides to join the V_H and V_L domains of an anti-2-phenyloxazolone(Ox) antibody. Four of the linker peptides originated from theinterdomain linker region of the fungal cellulase CBHI and consistedof 28, 11, six and two amino acid residues. The two other linkerpeptides used were the (GGGGS)₃ linker with 15 amino acid residuesand a modified IgG2b hinge peptide with 22 residues. Proteolyticstability and Ox binding properties of the six different scFvderivatives produced in Escherichia coli were investigated andcompared with those of the corresponding Fv fragment containingno joining peptide between the V domains. The hapten bindingproperties of different antibody fragments were studied by ELISAand BIAcore^TM. The interdomain linker peptide improved the haptenbinding properties of the antibody fragment when compared withFv fragment, but slightly increased its susceptibility to proteases.Single-chain antibodies with short CBHI linkers of 11, six andtwo residues had a tendency to form multimers which led to ahigher apparent affinity. The fragments with linkers longerthan 11 residues remained monomeric.     

Humanization of mouse anti-human IL-2 receptor antibody B-B10

Mouse monoclonal anti–human IL–2 receptor antibody(BB10) inhibits EL–2–dependent human T–cellproliferation. It has been used in clinical trials in the transplantationfield and promising results are being accumulated. Mouse B–B10antibody was humanized by grafting all CDRs and some frameworkamino add residues onto human antibodies, KAS for V_H and PAYfor V_x. Nine humanized B–BlOs with differently graftedframework residues were constructed and assessed for their biologicalactivities. One of these humanized B–B10, M5, showed nearlythe same activity as the mouse B–B10. The 49th residueof V_x was demonstrated to play a crucial role in the antigen–antibodyinteraction by 3–D structure analysis with a computermodeling system.     

Single-chain Fv multimers of the anti-neuraminidase antibody NC10: the residue at position 15 in the VL domain of the scFv-0 (VL-VH) molecule is primarily responsible for formation of a tetramer-trimer equilibrium

Single-chain variable fragment of the murine monoclonal antibodyNC10 specific to influenza virus N9 neuraminidase, joined directlyin the V_L to V_H orientation (scFv-0), forms an equilibrium mixtureof tetramer and trimer with the tetramer as the preferred multimericspecies. In contrast, the V_H-V_L isomer was previously shownto exist exclusively as a trimer. Computer-generated trimericand tetrameric scFv models, based on the refined crystal structurefor NC10 Fv domain, were constructed and used to evaluate factorsinfluencing the transition between V_L-V_H trimer and tetramer.These model structures indicated that steric restrictions betweenloops spanning amino acid residues L55–L59 and L13–L17from the two adjacent V_L domains within the V_L-V_H trimer wereresponsible for four scFv-0 molecules assembling to form a tetramer.In particular, leucine at position L15 and glutamate at positionL57 appeared to interfere significantly with each other. Tominimize this steric interference, the site-directed mutagenesistechnique was used to construct several NC10 scFv-0 clones withmutations at these positions. Size-exclusion chromatographicanalyses revealed that several of these mutations resulted inthe production of NC10 scFv-0 proteins with significantly alteredtetramer–trimer equilibrium ratios. In particular, introductionof a polar residue, such as asparagine or threonine, at positionL15 generated a highly stable NC10 scFv-0 trimer.     

Fluorolabeling of antibody variable domains with green fluorescent protein variants: application to an energy transfer-based homogeneous immunoassay

A site-specific and efficient fluorolabeling of antibody variableregions with green fluorescent protein (GFP) variants and itsapplication to an energy transfer-based homogeneous fluoroimmunoassay(open sandwich FIA) were attempted. Two chimeric proteins, Trx–V_H–EBFPand Trx–V_L–EGFP, consisting of V_H and V_L fragmentsof anti-hen egg lysozyme (HEL) antibody HyHEL-10 and two GFPcolor variants, EBFP and EGFP, respectively, were designed tobe expressed in cytoplasm of trxB – mutant Escherichiacoli as fusions with thioredoxin from E.coli The mixture oftwo proteins could be purified with HEL-affinity chromatography,retaining sufficient intrinsic fluorescence and binding activityto HEL. A significant increase in fluorescence resonance energytransfer (FRET) dependent on HEL concentration was observed,indicating the reassociation of the V_H and V_L domains of thesechimeric proteins due to co-existing antigen. With this opensandwich FIA, an HEL concentration of 1–100 µg/mlcould be non-competitively determined. The assay could be performedin a microplate format and took only a few minutes to obtaina sufficient signal after simple mixing of the chimeric proteinswith samples. This represents the first demonstration that theFRET between GFP variants is applicable to homogeneous immunoassay.     

Multimerization behaviour of single chain Fv variants for the tumour-binding antibody B72.3

A systematic study has been performed on the relationship betweenlinker length, relative orientation of variable domains, multimerizationbehaviour and antigen binding activity for single chain Fvs(scFvs) of the tumour-binding antibody B72.3. Thirteen scFvvariants with linkers comprising up to six repeats of the motifGly-Gly-Gry-Gly-Ser were studied. All these scFvs showed a tendencyto form dimers or higher molecular weight species, and thistendency decreased with increasing linker length. The dimersand higher molecular weight forms may arise from head to tailintermolecular association of V_H and V_L domains. For each linkerlength, scFvs with the organization VL-linker-VH showed greaterbinding activity than those with the organization VH-linker-VL.In fact, for the latter organization only the variant with a30 amino acid linker showed good binding activity, suggestingthat (0 for B72.3 the C-tenninus of VH or the N-tenninus ofVL makes a structural contribution to antigen binding, and (ii)shorter linkers interfere with this contribution. Antigen bindingstudies on scFvs should be interpreted with caution becauseof their tendency to multlmerize. Such multimerization can beminimized by using linkers longer than those in common use     

Multimerization of a chimeric anti-CD20 single-chain Fv-Fc fusion protein is mediated through variable domain exchange

A series of single-chain anti-CD20 antibodies was produced byfusing single-chain Fv (scFv) with human IgG1 hinge and Fc regions,designated scFv-Fc. The initial scFv-Fc construct was assembledusing an 18 amino acid (aa) linker between the antibody light-and heavy-chain variable regions, with the Cys residue in theupper hinge region (Kabat 233) mutagenized to Ser. Anti-CD20scFv-Fc retained specific binding to CD20-positive cells andwas active in mediating complement-dependent cytolysis. Size-exclusionHPLC analysis revealed that the purified scFv-Fc included multimericas well as monomeric components. Variant scFv-Fcs were constructedincorporating four different hinges between the scFv and Fcregions, or three different linkers in the scFv domain. Allformed multimers, with the highest level of multimerizationfound in the scFv-Fc with the shortest linker (8 aa). Eliminationof an unusual salt bridge between residues L38 and H89 in theV_L-V_H domain interface failed to reduce the formation of higherorder forms. Structural analysis of the scFv-Fc constructedwith 18 or 8 aa linkers by pepsin or papain cleavage suggestedthe proteins contained a form in which scFv units had cross-pairedto form a `diabody'. Thus, domain exchange or cross-pairingappears to be the basis of the observed multimerization.     

Engineering interchain disulfide bonds into conserved framework regions of Fv fragments: improved biochemical characteristics of recombinant immunotoxins containing disulfide-stabilized Fv

Using molecular modeling technology, we have recently identifiedtwo positions in conserved framework regions of antibody Fvfragments (Fvs) that are distant from CDRs, and potentiallycan be used to make recombinant Fv fragments in which the unstableV_H and V_L heterodimer is stabilized by an interchain disulfidebond inserted between structurally conserved framework positions.A disulfide bond has been introduced at one of these positions,V_H44-V_L105, and shown to stabilize various Fvs that retain fullbinding and specificity. Recombinant immunotoxias, e.g. B3(dsFv)-PE38KDELin which this disulfide-stabilized Fv moiety is connected toa truncated form of Pseudomonas exotoxin (PE; PE38KDEL) whichcontains the translocation and ADP ribosylation domains, areindistinguishable in binding and specificity from its single-chainimmunotoxin counterparts. We have now analyzed the alternativeposition, (V_H111-V_L48), predicted by the modeling methodology,for disulfide stabilization of mAb B3(Fv) by producing a recombinantimmunotoxin with such disulfide-stabilized (ds) Fv. This immunotoxinwas also very active and retained full specificity to B3 antigen-positivecells. However, it was 2- to 3-fold less active than the V_H44-V_L105dsFv-molecule. We also tested various biochemical features ofV_H44-V_L105 and V_H111-V_L48 dsFv immunotoxins and compared themwith the corresponding single-chain immunotoxin. We found thedsFv immunotoxins were more stable in human serum and more resistantto thermal and chemical denaturation than the single chain (sc)Fv immunotoxin. Because dsFv immunotoxins and dsFvs have fullactivity and specificity and improved stability, they may bemore useful than scFv immunotoxins as therapeutic and diagnosticagents.     

Disulfide stabilization of antibody Fv: computer predictions and experimental evaluation

Using molecular modeling technology we have recently identifiedpositions in conserved framework regions of Fvs which can beused to stabilize antibody Fvs by an interchain disulfide bondengineered in between the structurally conserved framework positionsof the variable domains of heavy (V_H) and light (V_L) immunoglobulinchains (disulfide-stabilized Fv; dsFv). The computer model indicatedthe existence of other potential sites in the framework regionsthat might be suitable for disulfide bond formation betweenV_H and V_L. The possibility of obtaining dsFvs using these positionsis evaluated here experimentally by constructing dsFv immunotoxinsin which the Fv moiety is fused to a truncated form of Pseudomonasexotoxin. We analyzed the extent of dsFv formation and the activityof the resulting dsFv immunotoxins, and compared various dsFvmolecules with the scFv immunotoxin. Our results demonstratethat position H44-L105 is the only one which gives high productionyields of active dsFv. All other positions gave either low yieldsand activity or completely failed to produce active dsFv. Withone exception, the formation and activities of the dsFvs correspondedto the C-C distance between the VH and VL positions, with anoptimal distance of 5.7 Å producing the best dsFv. Distancesof 6.0–6.9 Å resulted in a' low yield of proteinthat was still capable of binding antigen, whereas distances>7.0 Å resulted in molecules in which dsFv formationwas not obtained.     

The random peptide library-assisted engineering of a C-terminal affinity peptide, useful for the detection and purification of a functional Ig Fv fragment

The facile detection and purification of a recombinant proteinwithout detailed knowledge about its individual biochemicalproperties constitutes a problem of general interest in proteinengineering. The use of a novel kind of random peptide libraryfor the stepwise engineering of a C-terminal fusion peptidewhich confers binding activity towards streptavidin is describedin this study. Because of its widespread use as part of a varietyof conjugates and other affinity reagents, streptavidin constitutesthe binding partner of choice both for detection and purificationpurposes. The streptavidin-affinity tag was engineered at theC-terminus of the V_H domain as part of the D1.3 Fv fragmentwhich was functionally expressed in Escherichia coli. Irrespectiveof whether it was displayed by the V_H or the V_L domain, theoptimized version of the affinity peptide termed ‘Strep-tag’allowed the detection of the Fv fragment both on Western blotsand in ELISAs by a streptavidin–alkaline phosphatase conjugate.In addition, the one-step purification of the intact Fv fragmentcarrying a single Strep-tag at the C-terminus of only one ofits domains was achieved by affinity chromatography with streptavidin-agaroseusing very mild elution conditions.     

Modular construction of extended DNA recognition surfaces: mutant DNA-binding domains of the 434 repressor as building blocks

Single-chain derivatives of the 434 repressor containing onewild-type and one mutant DNA-binding domain recognize the generaloperator ACAA–6 base pairs–NNNN, where the ACAAoperator subsite is contacted by the wild-type and the NNNNtetramer by the mutant domain. The DNA-binding specificitiesof several single-chain mutants were studied in detail and theoptimal subsites of the mutant domains were determined. Thecharacterized mutant domains were used as building units toobtain homo- and heterodimeric single-chain derivatives. TheDNA-binding properties of these domain-shuffled derivativeswere tested with a series of designed operators of NNNN–6base pairs–NNNN type. It was found that the binding specificitiesof the mutant domains were generally maintained in the new environmentsand the binding affinities for the optimal DNA ligands werehigh (with K_d values in the range of 10^–11–10^–10M). Considering that only certain sequence motifs in place ofthe six base pair spacer can support optimal contacts betweenthe mutant domains and their subsites, the single-chain 434repressor mutants are highly specific for a limited subset of14 base pair long DNA targets.     

Sequence and structure of VH domain from naturally occurring camel heavy chain immunoglobulins lacking light chains

We cloned 17 different PCR fragments encoding V_H genes of camel(Camelus dromedarius). These clones were derived from the camelheavy chain immunoglobulins lacking the light chain counterpartof normal immunoglobulins. Insight into the camel V_H sequencesand structure may help the development of single domain antibodies.The most remarkable difference in the camel V_H, consistent withthe absence of the V_L interaction, is the substitution of theconserved Leu45 by an Arg or Cys. Another noteworthy substitutionis the Leu11 to Ser. This amino acid normally interacts withthe C_H1 domain, a domain missing in the camel heavy chain immunoglobulins.The nature of these substitutions agrees with the increasedsolubility behavior of an isolated camel V_H domain. The V_H domainsof the camels are also characterized by a long CDR3, possiblycompensating for the absence of the V_L contacts with the antigen.The CDR3 lacks the salt bridge between Arg94 and Asp101. However,the frequent occurrence of additional Cys residues in both theCDR1 and CDR3 might lead to the formation of a second internaldisulfide bridge, thereby stabilizing the CDR structure as inthe DAW antibody. Within CDRs of the camel V_H domains we observea broad size distribution and a different amino acid patterncompared with the mouse or human V_H. Therefore the camel hypervariableregions might adopt structures which differ substantially fromthe known canonical structures, thereby increasing the repertoireof the camel antigen binding sites within a V_H.     

'Knobs-into-holes' engineering of antibody CH3 domains for heavy chain heterodimerization

‘Knobs-into-holes’ was originally proposed by Crickin 1952 as a model for the packing of amino acid side chainsbetween adjacent -helices. ‘Knobs-into-holes’ isdemonstrated here as a novel and effective design strategy forengineering antibody heavy chain homodimers for heterodimerization.In this approach a ‘knob’ variant was first obtainedby replacement of a small amino acid with a larger one in theC_H3 domain of a CD4-IgG immuno-adhesin: T366Y. The knob wasdesigned to insert into a ‘hole’ in the C_H3 domainof a humanized anti-CD3 antibody created by judicious replacementof a large residue with a smaller one: Y407T. The anti-CD3/CD4-IgGhybrid represents up to 92% of the protein A purified proteinpool following co-expression of these two different heavy chainstogether with the anti-CD3 light chain. In contrast, only upto 57% of the anti-CD3/CD4-IgG hybrid is recovered followingco-expression in which heavy chains contained wild-type C_H3domains. Thus knobs-into-holes engineering facilitates the constructionof an antibody/ immunoadhesin hybrid and likely other Fc-containingbifunctional therapeutics including bispecific immuno-adhesinsand bispecific antibodies.     

Effect of linker sequences between the antibody variable domains on the formation, stability and biological activity of a bispecific tandem diabody

Bispecific single-chain Fv antibodies comprise four covalently linked immunoglobulin variable (V(H) and V(L)) domains of two different specificities connected by three linkers. When assembled in the order V(H)(A)-linker(1)-V(L)(B)-linker(2)-V(H)(B)-linker(3)-V(L)(A), the single-chain molecule either folds head-to-tail with the formation of a diabody-like structure, a so-called bispecific single-chain diabody, or forms a homodimer that is twice as large, a so-called tandem diabody. The formation of the tandem diabody is determined by the association of complementary V(H) and V(L) domains located on different polypeptide chains, and depends on the length and probably the amino acid composition of the three linkers joining the variable domains. We generated a number of single-chain constructs using four V(H) and V(L) domains specific either for human CD3, a component of T-cell receptor (TCR) complex, or for CD19, a human B-cell antigen, separated by different rationally designed peptide linkers of 6-27 amino acid residues. The generated bispecific constructs were expressed in bacterial periplasm and their molecular forms, antigen-binding properties, stability, and T-cell proliferative and anti-tumor activities were compared. Using peripheral blood mononuclear cell cultures from patients suffering from B-cell chronic lymphocytic leukemia, we demonstrated that the tandab-mediated activation of autologous T cells and depletion of malignant cells correlates with the stability of the recombinant molecule and with the distance between the CD19 and CD3 binding sites.     

Linker modification introduces useful molecular instability in a single chain antibody

SCA 4-4-20/212, is a recombinant single chain antibody directedagainst fluorescein (Fl) composed of the variable light (V_L)and variable heavy (V_H) domains of the monoclonal antibody 4-4-20,tethered by a 14 amino acid linker. Binding of SCA 4-4-20/212to Fl quenches its fluorescence, thus enabling the distinctionbetween bound and free Fl. This was used to follow antibodydenaturation which followed a two-step process: rapid selectedand restricted denaturation followed by slow and progressivedenaturation. This two-phase phenomenon might reflect selectivesusceptibility of the CDR loops to denaturation. Furthermore,a new SCA, SCA 4-4-20/9, was constructed by site-directed mutagenesisof SCA 4-4-20/212 using PCR methodology. SCA 4-4-20/9 was similarto SCA 4-4-20/212, but for a nine residue linker. The two SCAswere compared for Fl binding, heat stability, the effect ofdenaturing agents and susceptibility to proteolysis. The modificationof the linker caused a general conformational rearrangementin the SCA molecule, rendering it more sensitive to denaturationand proteolysis. This molecular instability may find utilityin the application of SCAs in analytical systems or as the recognitioncomponent in biosensors.     

Restructuring an interdomain linker in the dihydrolipoamide acetyltransferase component of the pyruvate dehydrogenase complex of Escherichia coli

The lipoyl, subunit-binding and catalytic domains of the dihydrolipoamideacetyltransferase subunits (E2p) of the Escherichia coli pyruvatedehydrogenase complex are connected by linker sequences whichare characteristically rich in alanine and proline residues.By facilitating domain movement these linkers are thought topromote interactions between the three types of active sitethat participate in the catalytic cycle of the complex. To investigatefunctional constraints associated with linker composition andsequence, the natural linker of an E2p subunit containing onelipoyl domain was replaced by shorter sequences containing:mixtures of alanine plus proline residues; mainly alanine; mainlyproline; and mainly charged residues. Each artificial linkerpossessed a central histidine residue for assessing linker flexibilityby ¹H-NMR spectroscopy. The resultant complexes exhibited 181%(proline), 74–79% (alanine plus proline), 63% (alanine)and 7% (charged residues) of parental activity compared witha value of 75% expected for a complex with a comparably shortenedlinker. The ¹H-NMR spectra showed that the alanine plus prolinelinkers are flexible but the alanine linker and the prolinelinker are relatively inflexible. Substantial variations inlinker sequence and composition were tolerated without lossof function, and the enhanced activity conferred by the prolinelinker was attributed to the combined effects of length andrelative inflexibility.     

Production and characterization of anti-human interferon {gamma} receptor antibody fragments that inhibit cytokine binding to the receptor

Three single-chain antibody fragments that recognize the extracellularhuman interferon receptor -chain (IFNR), and inhibit the bindingof human IFN, have been produced in Escherichia coli. Thesefragments are derived from murine anti-receptor monoclonal antibodies,and comprise the variable heavy (V_H) domain linked to the variablelight (V_L) chain through a 15 amino acid linker [(GGGGS)₃].Using surface plasmon resonance technology (BIAcore), the solubleproteins were shown to retain a high affinity for recombinantIFNR, and by radioimmunoassay to possess high inhibitory activitytowards IFN-binding to human Raji cells. The antibody fragmentsmost likely recognize epitopes that overlap the cytokine bindingsite on the receptor surface. Attempts to dissect further theantibodies to isolated V_H- and V_L-chains and to synthetic linearand cyclic peptides derived from the individual complementaritydetermining regions failed to afford fragments with significantIFNR binding affinity. Nevertheless, these native-like variableregion fragments and petidomimetics derived from them are ofinterest in the design of novel IFNR antagonists.     

Generation of a Ni(II) binding site by introduction of a histidine cluster in the structure of human glutathione transferase Al-1

Two mutant forms of human glutathione transferase (GST) Al–1with affinity for metal ions were constructed by introductionof His residues by site–directed mutagenesis. A mutant,2–His, contained the mutations Lys84Gln, Asp85His andGlu88His, and another, 5–His, contained the mutationsTyr79His, Asn80His, Lys84His, Asp85His and Glu88HLs. The mutantproteins were obtained in good yields (40–150 mg per 3I culture) by heterologous expression in Escherichia coli. Themutant enzymes possessed novel binding affinities for Ni(II)and Zn(II) ions, as demonstrated by immobilized metal ion affinitychromatography. The mutant with two novel His residues (2–Hismutant) did not bind as tightly to immobilized Nifll) as didthe mutant with five novel His residues (5–His mutant).When tested for affinity to immobilized Zn(II), only the 5–Hismutant remained bound to the column. The affinity of the 5–Hismutant for Ni(II) ions in solution was determined by bindingexperiments in an aqueous polymeric two–phase system.Analysis of the binding curve showed two binding sites per enzymesubunit and a dissociation constant of 6.7 1 . 6 M. The kineticconstants k_cat, K_m and k_cat/k_m for the reaction with glutathioneand l–chloro–2,4–dinitrobenzene were determinedby steady–state kinetic analysis and the parameter valuesfor the mutant forms were found to be indistinguishable fromthose obtained for the wild–type GST Al–1. The differencesin surface charge in the mutant proteins as compared with thewild–type enzyme did not alter the pH dependence of k_cat.The results provide an alternative method for purification offully active recombinant GST Al–1 by the introductionof novel metal binding sites. The data also showed that twoHis residues are sufficient for Ni(II) binding.     

Evolutionary trace analysis of TGF-{beta} and related growth factors: implications for site-directed mutagenesis

The TGF-ß family of growth factors contains a largenumber of homologous proteins, grouped in several subfamilieson the basis of sequence identity. These subgroups can be combinedinto three broader groups of related cytokines, with markedspecificities for their cellular receptors: the TGF-ßs,the activins and the BMPs/GDFs. Although structural informationis available for some members of the TGF-ß family,very little is known about the way in which these growth factorsinteract with the extra-cellular domains of their multiple cellsurface receptors or with the specific protein inhibitors thoughtto modulate their activity. In this paper, we use the evolutionarytrace method [Lichtarge et al. (1996) J. Mol. Biol., 257, 342–358]to locate two functional patches on the surface of TGF-ß-likegrowth factors. The first of these is centred on a conservedproline (P₃₆ in TGF-ßs 1–3) and contains twoamino acids which could account for the receptor specificityof TGF-ßs (H₃₄ and E₃₅). The second patch is locatedon the other side of the growth factor protomer and surroundsa hydrophobic cavity, large enough to accommodate the side chainof an aromatic residue. In addition to two conserved tryptophansat positions 30 and 32, the main protagonists in this potentialbinding interface are found at positions 31, 92, 93 and 98.Several mutagenesis studies have highlighted the importanceof the C-terminal region of the growth factor molecule in TGF-ßsand of residues in activin A equivalent to positions 31 and94 of the TGF-ßs for the binding of type II receptorsto these ligands. These data, together with our improved knowledgeof possible functional residues, can be used in future structure–functionanalysis experiments.