Protein Engineering vol.5 no.4 pp.343-350, 1992

Molecular modelling and site-directed mutagenesis on a bovineanti-testosterone monoclonal antibody

Terry Jackson1'5, Brian A.Morris2,Andrew C.R.Martin4'6, David F.V.Lewis3 andPeter G.Sanders1'7

'Molecular Microbiology Group, 2Heterohybridoma Antibody Group and3Molecular Toxicology Group, School of Biological Sciences, University ofSurrey, Guildford, Surrey, GU2 5XH, and 4Laboratory of MolecularBiophysics, The Rex Richards Building, University of Oxford, South ParksRoad, Oxford 0X1 3QU, UK5Present address: Institute for Animal Health, Pirbright, Surrey, UK

'Present address: SciTech Software, 23 Stag Leys, Ashtead,Surrey KT21 2TD, UK7To whom correspondence should be addressed

A three-dimensional (3D) molecular model of the antigen-combining site of a bovine anti-testosterone monoclonalantibody has been constructed. In the model, the CDRs, anda single heavy chain framework region residue (Trp47),associate to form a hydrophobic cavity large enough toaccommodate a single molecule of testosterone. Tyr97 ofCDR-H3 lies at the bottom of the cavity with its hydroxylgroup exposed to solvent. Using the model and data frombinding studies, we predicted that the cavity forms theantibody's paratope and on binding testosterone a hydrogenbond is formed between Tyr97 of CDR-H3 and the hydroxylgroup on the D-ring of testosterone. This prediction has subse-quently been tested by site-directed mutagenesis. An antibodywith phenylalanine in place of tyrosine at position 97 inCDR-H3 has its affinity reduced by ~ 800 fold. The reductionin binding energy associated with the reduced affinity has beencalculated to be 3.9 kcal/mol which is within the range(0.5-4.0 kcal/mol) expected for the loss of a single hydrogenbond. The model has been used to suggest ways of increasingthe antibody's affinity for testosterone.Key words: antibody/testosterone/molecular modelling

IntroductionTestosterone is a key hormone in the ruminant oestrous cyclehaving a negative feedback effect on the secretion of gonado-trophin hormones and, therefore, has been chosen as a meansof immunoregulating the oestrous cycle of sheep and cattle(Scaramuzzi et al, 1977; Land et al. ,1982; Webb et al, 1984;Price et al, 1987; Sreenan et al, 1987; Morris et al, 1988).In order to obtain consistent regulation of the oestrous cycle incatde, a heterohybridoma, B/MT.4A.17.H5.A5, secreting bovineanti-testosterone (IgGl) monoclonal antibody, has been produced(Groves et al., 1987a). The antibody secreted by B/MT.4A. 17.H5.A5 has a relatively high affinity (Kd = 2.5 x 10"" M)(Groves et al., 1987a). However, the affinity of this antibodyis only one to two orders of magnitude higher than that of recep-tors for steroid hormones (Chang et al., 1988). Therefore, inorder to improve the immunoneutralizing properties of theantibody secreted by clone B/MT.4A.17.H5.A5, we would liketo increase its affinity for testosterone by at least another orderof magnitude.

© Oxford University Press

The affinity of the antibody secreted by clone B/MT.4A.17.H5.A5 is already relatively high; however, an ovine monoclonalantibody with even greater affinity for testosterone (ATd = 7.63X 10"12 M) has been produced (Groves et al, 1987b),demonstrating that antibodies with a greater affinity fortestosterone can be obtained. In addition, ovine polyclonal antiseraraised against oestradiol and fluorescein isothiocyanate withavidities of Kd = 10"14 M have also been produced(B.A.Morris, unpublished), which suggests that ruminantmonoclonal antibodies with affinities in the Kd = 10"l5 M rangeare possible.

A bovine antibody with greater affinity for testosterone maybe achieved either by producing more heterohybridomas andselecting clones secreting higher affinity antibody, or byattempting to increase the affinity of the current antibody bysite-directed mutagenesis of the antigen-combining site. Thesecond route may provide important information for the designof high affinity antibodies generally (Roberts et al, 1987) andhas, therefore, been selected as our approach. To aid the designof specific amino acid substitutions we have constructed athree-dimensional (3D) molecular model of the antigen-combiningsite of the bovine anti-testosterone monoclonal antibody. Antigen-combining sites are located on the Fv modules (VH:VL domainpairs) of antibodies and formed by the complementaritydetermining regions (CDRs) or hypervariable loops (three on theVH domain and three on the VL domain) (Amit et al., 1986;Colman et al., 1987; Sheriff et al., 1987; Padlan et al., 1989).The CDRs form exposed loops at one end of each variabledomain, and are supported by structurally conserved frameworkregions. The high degree of structural homology between variabledomains, and their packing to form Fv modules (Chothia et al,1985; Novotny and Haber, 1985) has enabled molecular modelsof antigen-combining sites to be made by replacing the CDRsin a known antibody crystal structure with loops of modelledconformation (de la Paz et al., 1989; Chothia and Lesk, 1987;Chothia et al, 1989; Martin et al, 1989). For five of the CDRs(LI, L2, L3, HI and H2), the main chain conformation of themodelled loop may normally be selected from the database ofantibody crystal structures as only a small number of main chainconformations are available to these CDRs (termed canonicalfamilies) (Chothia and Lesk, 1987; Chothia et al, 1989). Theconformation adopted by these loops is determined by the natureof the amino acid at a few conserved 'key' interacting positionsfound in both the CDRs and the framework regions (Chothia andLesk, 1987; Chothia et al, 1989). CDR-H3 shows a muchgreater variability and in all the antibody crystal structures themain chain conformations and packing are different. This CDRcannot, therefore, be modelled using the above approach.However, accurate models of all six CDRs, including CDR-H3,have been constructed using conformational searching of templateloops selected from the complete Brookhaven Protein Databank(Bernstein et al., 1977) using distance constraints derived fromthe CDRs of known antibody structures (Martin et al, 1989).The methods of Chothia et al (1989) and Martin et al (1989)represent the state of the art in antibody modelling and have been

343

at The U

niversity of British C

olombia L

ibrary on Novem

ber 24, 2014http://peds.oxfordjournals.org/

Dow

nloaded from

T.Jackson et al.

used here to model the antigen-combining site of the bovineantibody.

The model has suggested which CDR residues are likely toform the antibody paratope. This information, together with thecross-reactivities of the antibody for structurally related steroidhormones and the conjugated state of testosterone when used toimmunize cattle for the construction of clone B/MT.4A.17.H5.A5 (Groves et al., 1987a), have been used to predict thenature of the likely interactions between the antibody andtestosterone. The major prediction made has been tested by site-directed mutagenesis.

Materials and methodsModelling the antigen-combining siteThe nucleotide and derived protein sequences of the heavy andlight chains of the antibody secreted by clone B/MT.4A.17.H5.A5 have been published elsewhere (Jackson et al., 1992).The model was constructed using the methods of Chothia andLesk (1987), Chothia et al. (1989) and Martin et al. (1989) andvisualized using an Evans and Sutherland ESV20 Workstationrunning the Sybyl molecular modelling package (TriposAssociates). The atomic coordinates of the crystal structures usedto construct the model were obtained from the BrookhavenProtein Databank (Bernstein et al., 1977). The completed model,with all the loops in place, was energy minimized using the fullGROMOS potential in vacuo (Aqvist et al., 1985).

The framework regionsThe framework regions (FRs) were taken from the VL and VH

domains of the crystal structures KOL (Marquart et al., 1980)and Dl .3 (Amit et al., 1986) respectively. These domains were

selected so that the packing of 'key' interacting residues whichare predicted to determine the main chain conformations of theCDRs in the bovine antibody (see below) are also present in themodel. The above domains were paired to form an Fv moduleby least squares fitting the main chain atoms of the frameworkregions (FR1, FR2 and FR3) of the VH domains of KOL andD 1.3 and then deleting VL of D1.3 and VH of KOL.

The complementarity determining regionsUsing the methods of Chothia and Lesk (1987) and Chothia et al.(1989), CDRs LI, L2, L3, HI and H2 were all predicted to sharethe same, or a similar main chain conformation to a CDR ofknown structure. For the CDRs where more than one knownstructure fell into the canonical group selected, the loop with thegreatest sequence homology to the bovine antibody was chosen.The CDR sequences were corrected to those of the bovineantibody by using the maximum overlap protocol as implementedby the REPLACE and REFI options (Hermans and McQueen,1974) of the molecular graphics program FRODO (Jones, 1978).

CDR-L1The length of the CDR-L1 in the antibody KOL (Marquart et al.,1980) and the light chain dimer RHE (Furey et al, 1983) arethe same and their main chain conformations are virtuallyidentical (de la Paz et al., 1986). The key structural residuesresponsible for the observed conformation have been identifiedas CDR-L1 residues Gly25, Ile28 and Val33, and frameworkresidue Ala71 (Chothia and Lesk, 1987). CDR-L1 in the bovineantibody is the same length as that of KOL and RHE and includesidentical key residues (Table I). CDR-L1 was, therefore,predicted to share the same main chain conformation as CDR-Ll of KOL and RHE and was modelled using KOL.

Table I. Residues

Light chain

ResKOLBOV

ResKOLBOV

ResKOLBOV

Heavy chain

ResD1.3BOV

ResD1.3BOV

ResBOVPGK

at key positions

CDR-L124TSCDR-L250RGCDR-L389AA

FR26*GGCDR-H250MG

25*GG

51DS

90SA

27*FF

5111

C D R - H 395SE

96TG

for model

26TS

52AN

91WG

28S

s52WT

97YS

construction

27SS

53MS

92ND

29*LL

53GS

98GR

27aSS

54RR

93SS

54DG

99EK

27bNN

55PP

94SS

CDR-H131

GS

55*GG

100VV

28*II

56SS

95DS

32YY

56NT

AGD

29GGFR48*1I

95a*NR

33GA

57TT

BDG

30ST

64*GG

95bSG

34*VL

58DY

CGQ

31IY

96YA

35NT

DAK

32TG

97VV

FR71*KK

EIV

33*VV

FR94*RR

FAK

34NE

101DA

FR71*

AA

102AS

Comparison of the CDRs and residues at key interacting positions in the bovine antibody and the crystal structures used to construct the model (VL of KOLand VH of D1.3). The residues that determine the main chain conformation of the CDRs are highlighted by *. Res = residue number (Kabat definition);BOV = bovine antibody; PGK = phosphoglycerate kinase (residues 127-140).

344

at The U

niversity of British C

olombia L

ibrary on Novem

ber 24, 2014http://peds.oxfordjournals.org/

Dow

nloaded from

Molecular modelling of testosterone antibody

CDR-L2The length of CDR-L2 is the same in the immunoglobulin crystalstructures and all share virtually the same main chain conforma-tion despite significant differences in their amino acid sequences(de la Paz et al., 1986; Chothia and Lesk, 1987; Chothia et al.,1989). The main chain conformation is dependent uponframework residues De48 and Gly64 (Chothia et al., 1989), whichare highly conserved in VL domains (Kabat et al., 1987). BothIle48 and Gly64 are present in the bovine light chain (Table I)and CDR-L2 was, therefore, predicted to have the same mainchain conformation as in the antibody crystal structures and wasmodelled using KOL.

CDR-L3The lengths of CDR-L3 in KOL and RHE are the same and theirmain chain conformations are very similar with residues 93, 94,95 and 95a forming a four residue turn (Chothia and Lesk, 1987).The major structural determinant responsible for the conforma-tion has been identified by Chothia and Lesk (1987) as residue95a, which is hydrophilic in both KOL (asparagine) and RHE(aspartate). In both cases, this hydrophilic residue is on the proteinsurface and is exposed to solvent. CDR-L3 of the bovine antibodyhas an arginine at this position (Table I) and was thus predictedto have a similar main chain conformation to CDR-L3 of KOLand RHE. Although this represents a change of charge, theexposed nature of this residue should allow this with minimalstructural consequences. KOL was therefore used to constructthe model.

CDR-H1CDR-H1 is the same length in all the available antibody crystalstructures and their main chain conformations are divided intotwo similar canonical families (Chothia et al., 1989). The residuesresponsible for the observed conformation are frameworkresidues 26, 27, 29 and 94, and CDR-H1 residues 34 (Chothiaet al., 1989). In the bovine antibody, the amino acids at theseresidues are identical to those in the antibody D1.3, with theexception of residue 34, which is leucine in the bovine antibodyand valine in D1.3 (Table I). CDR-H1 of D 1.3 belongs to acanonical family which permits valine, methionine or isoleucineat residue 34. Leucine is a similar amino acid and we have,therefore, predicted that the main chain conformation of CDR-H1of the bovine antibody will be similar to that of D1.3 which wastherefore used to construct the model.

CDR-H2CDR-H2 is the same length in the antibodies NEW (Saul et al.,1978), D1.3 (Amit et al., 1986) and HyHEL-10 (Padlan et al.,1989) which all belong to the same canonical family (Chothiaet al., 1989). The main chain conformation of this family isdetermined by the presence of glycine at residue 55, whichoccupies the fourth position of a seven residue turn. CDR-H2in the bovine antibody is the same length as the above loops andalso includes glycine at residue 55 (Table I). Therefore, we havepredicted that the main chain conformation of CDR-H2 in thebovine antibody will be the same as the above loops and D1.3was used to construct the model.

CDR-H3The conformation of CDR-H3 was modelled using the methodsof Martin et al. (1989) in the presence of the other loops builtas described above. The Brookhaven Protein Databank wassearched for loops of conformation similar to the knownCDR-H3s using inter-Ca distance constraints applied to the fourresidues at either end of the loop. Five residues from the middle

of each database loop were deleted and reconstructed using theCONGEN conformational search program (Bruccoleri et al.,1988). Energies were calculated for the conformations generatedusing a solvent modified version of the GROMOS potential(Aquist et al., 1985). The selected conformation originated fromresidues 127-140 of yeast phosphoglycerate kinase (Watsonetal, 1982).

COS-1 cell transfections and ELISAThe wild-type and mutated antibodies were expressed in COS-1cells (Gluzman, 1981). cDNAs encoding the wild-type heavy andlight chains and the mutated heavy chains were ligated separatelyinto vector pUSlOOO (Jackson et al., 1992) to produce vectorspUSlOOl (wild-type heavy chain), pUS1002 (wild-type lightchain), (Jackson et al., 1992), and pUS1003 (H3 Phe97 mutation)and pUS1004 (H3 Glu97 mutation). Co-transfections wereperformed using the light chain expression vector with a vectorexpressing either the wild-type or a mutated heavy chain. VectorDNA was transfected into COS-1 cells using DEAE-Dextranas described previously (Jackson etal., 1992). Control transfec-tions were carried out in parallel on cells in the absence ofexogenous DNA (cell control) and using the vector (pUSlOOO)without antibody cDNA (vector control). In addition, controltransfections were performed using vectors expressing the heavy(including the mutated heavy chains) and light chains in individualexperiments. The presence of anti-testosterone antibodies intransfected cell culture media, including the controls, was detectedusing an ELISA specific for bovine anti-testosterone IgG (Groveset al., 1987a). A positive control of the antibody secreted by cloneB/MT.4A.17.H5.A5 was included on each ELISA plate.

Site-directed mutagenesisSite-directed mutagenesis was carried out using an Amershamoligonucleotide-directed in vitro mutagenesis system. Oligo-nucleotides containing the mutations were made on an AppliedBiosystems DNA synthesizer model 381A according to themanufacturer's instructions.

cDNA for the wild-type heavy chain cDNA (Jackson et al.,1992) was ligated into the £coRI site of M13mp 18 and the singlestranded form used as a template for mutagenesis. Heavy chainsincluding the introduced mutation(s) were identified directly byDNA sequencing through the mutated region. The completecoding sequence of the mutated VH domain was sequenced toverify that additional unwanted changes had not been introduced.

DNA sequencingDNA sequencing was carried out on 1—2 ng of recombinantM13mpl8 single stranded DNA templates using Sequenase™(Cambridge BioScience).Radioimmunoassay (RIA)The equilibrium binding dissociation constant (Kd) of the wild-type and mutated antibodies when expressed by transfectedCOS-1 cells were measured at 4°C using a competitive radio-immunoassay (Groves etal., 1987a). Briefly, 100 fd of culturemedium containing anti-testosterone antibodies was incubatedovernight in a volume of 300 /tl, containing 0.04 M phosphatebuffer (pH 7.4) with a fixed amount of tritiated testosterone, inthe presence of increasing amounts of unlabelled testosterone(0.0078-1000 pmol/tube). Antibody-bound testosterone wasseparated from free testosterone using a dextran-coated charcoalsuspension in phosphate buffer, and the amount of antibody-boundradioactivity measured by scintillation counting. Data from thesestudies were analysed by the method of Muller (1980) (seeFigure 3).

345

at The U

niversity of British C

olombia L

ibrary on Novem

ber 24, 2014http://peds.oxfordjournals.org/

Dow

nloaded from

T.Jackson et al.

Results and discussion

A 3D molecular model of the antigen-combining site of a bovineanti-testosterone monoclonal antibody has been constructed toaid the design of a higher affinity antibody. The major structuralfeature described by the model is a hydrophobic cavity whichis formed by residues from five of the CDRs and a single residuefrom the heavy chain framework region (Trp47) (Figure 1A andTable II). The cavity is estimated to be 10 A deep and 15 A wide(at the mouth) and, therefore, able to accommodate a singlemolecule of testosterone (Figure 2). The walls of the cavity areincomplete as part of the rim is missing (Figure 1A). CDR-H3Tyr97 lies at the bottom of the cavity with its hydroxyl group

positioned so that it could interact with solvent molecules (FigureIB).

Several lines of indirect evidence support the overall findingsof the model, i.e. the general shape of the antigen-combiningsite. Four of the CDRs (LI, L2, H2 and H3) were modelled usingmethods that have been vigorously tested by applying them toCDRs of known structure and have given accurate reproducibleresults (Chothia and Lesk, 1987; Chothia et al., 1989; Martinet al., 1989). CDRs HI and L3 were modelled using the Chothiamethod, although these loops do not match exactly with any ofthe known canonical families for these CDRs (Chothia and Lesk,1987; Chothia et al., 1989); however, both CDR-H1 and CDR-L3 of the bovine antibody have a different amino acid at only

Fig. 1. Structure of the modelled antigen-combining site. To construct the model, residues of the VL and VH domains were numbered sequentially (VL

1 — 110; VH 111 —229). In the text, residues are numbered according to the Kabat numbering scheme (Kabat el al.. 1987). In the figure legends, residues areidentified by their Kabat number followed by the model number in brackets. (A) The modelled antigen-combining site viewed from above. The light chainCDRs are shown in red and the heavy chain CDRs in green. The CDRs associate to form a cavity. (B) The cavity in cross-section showing the positions ofCDR-H3 Tyr97 (Tyr210) and the heavy chain framework residue Trp47 (shown in blue at the bottom).

346

at The U

niversity of British C

olombia L

ibrary on Novem

ber 24, 2014http://peds.oxfordjournals.org/

Dow

nloaded from

Molecular modelling of testosterone antibody

a single key residue compared with the CDRs used to constructthe model (residues 34 and 95a in the heavy and light chainsrespectively). The canonical families to which CDR-H1 of D1.3and CDR-L3 of KOL belong permit varying amino acids at these

Table II. Residues that form the cavity walls in the modelled antigen-combining site

Heavy chain

CDR-H1

CDR-H2

CDR-H3

FR

Ala33Leu34Thr35Gly50Ile51Thr52Thr56Tyr58Ser95Thr96Tyr97Trp47

Light chain

CDR-L1CDR-L3

Tyr32Gly91Asp92Ser93Arg95aGly95bAla96

OHCH3

OHC) DmYDROTESTOSTERONE CH3

Fig. 2. Structure of testosterone and related steroid hormones. When usedfor immunization, testosterone (A) was conjugated to a carrier proteinthrough the A-ring carbonyl group Testosterone (A) and progesterone (B)differ in the side group attached to C17 on the D-ring, whereas testosterone(A) and dihydrotestosterone (C) differ in structure at the A-ring end

positions. In the CDR-H1 D1.3 canonical family, residue 34 canbe methionine, isoleucine or valine while in the bovine antibodyit is leucine, a conservative substitution. In the CDR-L3 KOLcanonical family, examples of aspartate and asparagine are seenat residue 95a. The bovine antibody has arginine at this position.As in the known examples of this canonical group, arginine ishydrophilic and the solvent-exposed nature of this residue suggeststhat the positive charge of arginine should be accepted withoutany major conformational change in the backbone of the protein.

A correlation between the length and volume occupied byCDRs LI and H3 with the type of surface formed by the CDRshas been observed (de la Paz et al., 1986). Antigen-combiningsites which include short CDRs LI and H3 are relatively flat,whereas those which include long CDRs LI and H3 form groovesor cavities (Table III). In the bovine antibody, CDRs LI and H3are both long (13 and 14 residues respectively) which is consistentwith the formation of a cavity. In addition, cavity formationappears to be enhanced by CDR residues with small side chainswhich effectively 'open up' the cavity (de la Paz et al., 1986).In the model, the residues that line the cavity walls arepredominantly small which is also consistent with the formationof a cavity.

The predicted structure of the antibody—testosterone complexTestosterone is a relatively small, hydrophobic molecule (Figure2) and the high affinity of the antibody secreted by cloneB/MT.4A.17.H5.A5 implies that a large proportion of testoster-one will be in contact with the antibody on association. Thissituation would be best served by testosterone binding into a cavityor groove, as is known to be the case for antibodies binding toother small ligands, such as phosphorylcholine (Satow et al.,1986) and fluorescein (Herron etal., 1989). Testosterone bindinginto a cavity must also satisfy the conjugated state of the moleculewhen presented to the animal on immunization (Groves et al.,1987a) i.e. conjugated through the A-ring carbonyl oxygen atom.Therefore, if testosterone does bind in the cavity, it follows thatit is more likely to do so with the D-ring at the bottom of thecavity and the A-ring out of the combining site. The involvementof the D-ring in binding to the antibody is further implicated bythe cross-reactivities of the antibody for structurally related steroidhormones (Groves et al., 1987a). The antibody displays 87%cross-reactivity with dihydrotestosterone which differs in structurefrom testosterone only on the A-ring (Figure 2), but fails to bind

Table III. Relationships between the lengths of CDRs LI and H3, andconformation of the antigen-combining site

Crystal structure

D1.3 (Amit et ai, 1986)HyHEL-5 (Sheriff et ai, 1987)HyHEL-10 (Padlan et al, 1989)Gloop-2 (Jeffrey, 1989)

NEW (Saul et al., 1978)4.4.20 (Herron et ai, 1989)

McPC603 (Satow et al., 1986)KOL (Marquart et ai, 1980)R19.9 (Lascombe et ai, 1989)

Antigen-combining sites which include relatively long CDRs LI and H3form cavities, whereas those which include relatively short CDRs LI andH3 form flat or slightly concave surfaces. CDRs LI and H3 of the bovineantibody are 13 and 14 residues long respectively.

CDR

LI

11101111

1416

121311

H3

8754

97

171712

Surfaceconformation

Flat

Slot or groove

Cavity

347

at The U

niversity of British C

olombia L

ibrary on Novem

ber 24, 2014http://peds.oxfordjournals.org/

Dow

nloaded from

T.Jackson et al.

60

50 •

40 -

1 2

cAntibody

WT-COS

*WT-C0S

•Y97F

Y97F

Y97F

5 10 20 50 100 200 500 1.000

Antibody dilution

[It]10-9

0.433

0.086

43.29

29.97

153.18

[Tt]10 •»

0.309

0.047

0.047

0.050

0.045

b

0.52

0.55

0.50

0.44

0.40

Kd

4.4x10'"

1.3x10'"

1.6x10*

1.3X10"8

7.1 x10"*

30 •

20

10

b

A1

-

1 1

X>

•

\

\\\

i\\\\

A*t

\\\\

\\A

t i

WT-COS

\\

\\

Y97F'A-

1

0.0001 0.001 0.01 0.1 1 10 100 1.000 10.000

Log unlabelled testosterone (pmol/tube)

Fig. 3. Testosterone binding to wild-type and mutant antibodies, (a) ELISA results. Both of the mutated antibodies (Y97F and Y97E) retain the ability to bindtestosterone. Key: WT-BMT = wild-type antibody secreted by heterohybridoma cells of clone B/MT.4A.17.H5.A5; WT-COS = wild-type antibody expressedby COS-1 cells; Y97F = mutant antibody Y97F expressed by COS-1 cells; Y97E = mutant antibody Y97E expressed by COS-1 cells; NC = culture mediafrom negative control transfections (cell control, vector control and separate heavy and light chain controls). The negative controls gave near identical resultsand are shown in the figure as a single curve, (b) R1A standard curve. The ATds for both the wild-type antibody and Y97F when expressed by COS-1 cellswere calculated from plots of percentage of the total added radiolabel bound to antibody against unlabelled testosterone (pmol/tube). (c) Muller analysis. Datafrom the standard curves were analysed using the Muller equation, Kd = ([It] - [Tt]) ((1-1.56) + (0.5b2) (Muller, 1980), where [Tt] = molarconcentration of labelled testosterone; [It] = molar concentration of unlabelled testosterone that gives 50% inhibition of antibody binding to label;b = percentage of the total added radiolabel bound by antibody in the absence of unlabelled testosterone [* indicates values derived from the standard curveshown in (b)]. Each Kd value was derived from a similar curve but only two representative graphs are shown. Each KA value was obtained from a differenttransfection experiment.

progesterone which is structurally different to testosterone onlyin the group attached to C-17 on the D-ring (Figure 2). CDR-H3 residue Tyr97 lies at the bottom of the cavity and it isattractive to speculate that the hydroxyl group of this residueforms a hydrogen bond with the hydroxyl group on the D-ringof testosterone. In addition, the hydrophobic nature of the cavitycould provide a hydrophobic environment for testosterone whenhydrogen-bonded to Tyr97. Based on the above observations,we predicted that on formation of an antibody—testosteronecomplex, testosterone will bind into the cavity and a hydrogenbond will be established between the hydroxyl groups on theD-ring of testosterone and CDR-H3 Tyr97. To test this predic-tion, Tyr97 of CDR-H3 was selected for mutagenesis.

Construction and expression of mutated antibodiesAntibodies, with phenylalanine (Y97F) or glutamate (Y97E) inplace of H3 Tyr97 were constructed by co-expression of mutatedheavy chain cDNAs with wild-type light chain cDNAs in COS-1

348

cells. The ability of each mutant antibody to bind testosteronewas determined using an ELISA specific for bovine anti-testosterone IgG (Groves et al., 1987a). The affinities of themutated antibodies for testosterone were measured by calculatingtheir binding equilibrium dissociation constants (Kd). The KA ofthe wild-type antibody when expressed by COS-1 cells wascalculated to allow a direct comparison to be made between theaffinities of the wild-type and mutated antibodies when expressedin the same system.

Figure 3a shows the ELISA results and indicates that both ofthe mutated antibodies retain some ability to bind testosterone.

The Kd of the wild-type antibody when expressed by COS-1cells was found to be virtually identical [Kd = 2.8 X 10"" M(the mean of two transfections; Figures 3b and c)] to the Kd ofthe antibody secreted by heterohybridoma cells of cloneB/MT.4A.17.H5.A5 (Kd = 2.5 X 10"" M, Groves et al,1987a). The affinity of Y97F was found to be reduced by - 800fold (Kd = 3.4 X 10~8 M, the mean of three assays, Figure

at The U

niversity of British C

olombia L

ibrary on Novem

ber 24, 2014http://peds.oxfordjournals.org/

Dow

nloaded from

Molecular modelling of testosterone antibody

3c), whereas the Kd of Y97E could not be measured under theassay conditions used to measure the affinity of the otherantibodies. The most likely explanation for this observation isthat the affinity of Y97E for testosterone is so greatly reducedthat when free testosterone is removed with the dextran-coatedcharcoal at phase separation, the antibody—testosterone complexdissociates resulting in the loss of antibody-bound label byabsorption to the charcoal.

The change in binding energy associated with the tyrosine tophenylalanine substitution can be calculated from the Kd valuesof the wild-type and mutated antibody using the equation:

AG = (-RT In KA Y97F) - (-RT\n KA wild-type antibody),

(where AG = the Gibbs free energy change (J/mol), R = theideal gas constant (8.314 J/K/mol) and T = 277 Kelvin). Usingthis equation AG = 16.3 kJ/mol or 3.9 kcal/mol.

Definitive statements as to how the mutations at CDR-H3Tyr97 effect the conformation of the antibody and its interactionswith testosterone cannot be made since the structures of the wild-type and mutated antibodies when bound to testosterone have notbeen determined; however, the reduction in binding energy ofY97F (3.9 kcal/mol) is consistent with that expected for the lossof a single hydrogen bond (4.0-0.5 kcal/mol) between the typesof groups predicted to be involved (Fersht et al., 1985; Bartlettand Marlowe, 1987; Tronrud et al., 1987). The strength of ahydrogen bond appears to reflect the degree by which thehydrogen bonding components are immobilized before the bondis established. The energy of a hydrogen bond formed betweentwo uncharged components held in the same orientation in theabsence and presence of the hydrogen bond has been calculatedto be 4.0 kcal/mol (Bartlett and Marlowe, 1987; Tronrud et al.,1987). This value is greater than that determined for a hydrogenbond between two uncharged components which are on flexibleamino acid side chains and have rotational and translationalenergy before the hydrogen bond is established (0.5-1.5kcal/mol) (Fersht et al., 1985). The hydrogen bond predictedto exist between the antibody and testosterone is formed betweentwo relatively inflexible groups. Therefore, the binding energyfor this bond would be expected to be intermediate between theabove examples.

The possibility that CDR-H3 Tyr97 does not make directcontact with testosterone, but that the reduction in affinityassociated with the tyrosine to phenylalanine substitution resultsfrom a distant conformational change at residues which makecontact with testosterone should be considered. However, theability of Y97F to bind testosterone with reasonable affinity (Kd

3.4 x 10~8 M) implies that gross conformational changes in theantigen-combining site have not taken place and, therefore, thatany conformational change(s) that may have occurred are likelyto be small and localized around CDR-H3 Tyr97.

It is worth considering the tyrosine to phenylalanine mutationin the reverse direction. The antibody with phenylalanine atCDR-H3 residue 97 has a reasonable affinity for testosterone(Kd = 3.4 x 10~8 M). Affinities in the range of Kd =10~8-10~9 M are often obtained for rodent antibodies. Thetherapeutic importance of chimeric and 'humanized' rodentmonoclonal antibodies has stimulated much interest in improvingthe affinity of such antibodies by site-directed mutagenesis(Morrison et al., 1984; Riechmann et al., 1988). The reversemutation demonstrates that an increase in affinity of ~ 800-foldcan be achieved by the introduction of a single conservativesubstitution at a CDR residue.

Increasing the affinity of the bovine antibodyAn antibody with greater affinity for its antigen could be obtainedby: (i) increasing the number of van der Waals and hydrophobiccontacts made by increasing the buried surface area at theantibody—antigen interface; (ii) introducing mutations whichresult in the antigen and antibody moving closer together, therebyincreasing the strength of both non-specific and specific inter-actions (Roberts et al., 1987); (iii) increasing the number ofspecific interactions at the interface (Bratlett and Marlowe, 1987;Tronrud et al., 1987).

The only site on testosterone available for specific interactions(i.e. the hydroxyl group on the D-ring) has been stronglyimplicated to form a hydrogen bond with CDR-H3 Tyr97.However, the model has revealed a structural feature of theantibody that may allow for the construction of an antibody withgreater affinity: part of the cavity rim is missing (Figure 1A).This gap may be required by the antibody to accommodate thebridge on testosterone when the steroid is conjugated to its carrierprotein for immunization. However, when the antibody bindsnon-conjugated testosterone the gap may expose a region oftestosterone to solvent molecules. Therefore, filling in thisgap may allow for the buried surface area at the antibody-testosterone interface to be increased. CDR-H2 Gly50 andCDR-H1 Ala33 are positioned in the lower region of the cavityand orientated so that their side chains point towards the gap inthe cavity wall. Therefore, replacing these residues with aminoacids with larger non-polar side chains could result in a greatercontacting surface area. In addition, benefit may be gained byreplacing these residues with aromatic amino acids as the 'fused'carbon atoms of the aromatic ring may reduce the conformationalentropy lost on forming an antibody-testosterone complex(Padlan, 1990). However, it is also possible that the 'gap' in thecavity wall may be a structural requirement necessary for theentry of testosterone into the cavity and mutations which reduceits size may preclude binding.

The model has also revealed a second possible route toobtaining an antibody with greater affinity. The side chains ofthe residues immediately adjacent to CDR-H3 Tyr97 (CDR-L3Ala96 and CDR-H3 Thr96) appeared to protrude further into thecavity than the side chains of the residues in the generalsurrounding area. Reducing the volume occupied by theseresidues may allow for closer contacts to be made between theantibody and testosterone and may increase the strength of bothnon-specific hydrophobic interactions and the predicted hydrogenbond.

The model may be tested further by the introduction of specificamino acid substitutions at other CDR residues. For example,in the model, CDR-L2 does not participate in forming the cavitywalls and we would, therefore, predict that residues of this CDRdo not make contact with testosterone. The introduction ofspecific mutations into CDR-L2, away from key residuepositions, should, therefore, have little or no effect on the affinityof the mutant antibody for testosterone. The existence of thepredicted hydrogen bond could be tested by measuring the Kd

of the wild-type antibody for a modified testosterone moleculein which the D-ring hydroxyl group has been replaced by ahydrogen atom. The mutant antibody, Y97F, could also be usedto determine the Kds for various C-17 substituted steroids. TheY97F antibody should be less sensitive to substitutions at thisposition because there is no hydrogen bond to lose. In addition,the mutations discussed earlier which are designed to increasethe affinity of the antibody for testosterone, could be made andanalysed.

349

at The U

niversity of British C

olombia L

ibrary on Novem

ber 24, 2014http://peds.oxfordjournals.org/

Dow

nloaded from

T.Jackson et al.

In conclusion, a combination of molecular modelling and site-directed mutagenesis has been used to identify a residue (CDR-H3 Tyr97) which is of critical importance for the high affinityof a bovine anti-testosterone monoclonal antibody. Several linesof evidence which support the model have been discussed.Together with the mutagenesis data, the accumulated evidencestrongly suggests that the overall shape of the modelled antigen-combining site is correct and that the residues that line the cavitywalls are those most likely to form the antibody paratope.

AcknowledgementsT.Jackson was the recipient of a MAFF postgraduate training award.

ReferencesAmit.G.A., Mariuzza.R.A., Phillips.S.E.V. and Poljak.R.J. (1986) Science, 233,

747-753.Aqvist.J., van Gunsteren,W.F., Leijonmarck.M. and Tapia,O. (1985) 7. Mol.

Biol., 183, 461-477.Bartlett.P.A. and Marlowe,C.K. (1987) Science, 235, 569-571.Bernstein,F.C., Koetzle,T.F., Williams,G.J.B., Meyer.E.F., Brice,M.D.,

Rodgers.J.R., Kennard,O., Shimanouchi.T. and Tasumi.M. (1977)7. Mol.Biol., 112, 535-542.

Bruccoleri,R.E., Haber.E. and Novotny,J. (1988) Nature, 335, 564-568.Chang.C, Kokontis.J. and Liao,S. (1988) Science, 240, 324-326.Chothia,C. and Lesk,M. (1987) 7. Mol. Biol., 196, 901-917.Chothia,C, Novotny.J., Bruccoleri,R. and Karplus.M. (1985)7. Mol. Biol., 186,

651-663.Chothia.C, Lesk,A.M., Tramontano.A., Levitt.M., Smith-Gill,S.J., Air,G.,

Sheriff.S., Padlan.E.A., Davies.D., Tulip,W.R., Colman.P.M., Spinelli,S.,Alzari.P.M. and Poljak,R.J. (1989) Nature, 342, 877-883.

Colman,P.M., Laver,W.G., Varghese,J.N., Baker.A.T., Tulloch,P.A., Air.G.M.and Webster.R.G. (1987) Nature, 326, 358-363.

de la Paz,P., Sutton,B.J., Darsley,M.J. and Rees.A.R. (1986) EMBO J., 5,415-425.

Fersht,A.R., Shi,J.-P., Knill-Jones,J., Lowe.D.M., Wilkinson.A.J., Blow.D.M.,Brick,P., Carter.P., Waye,M.M.Y. and Winter.G. (1985) Nature, 314,235-238.

Furey.W., Wang,B.C, Yoo.C.S. and Sax.M. (1983) 7. Mol. Biol., 167,661-692.

Gluzman.Y. (1981) Cell, 23, 175-182.Groves.D.J., Morris.B.A. and Clayton,J. (1987a) Res. Vet. Sci., 43, 253-256.Groves.D.J., Morris,B.A., Tan,K., De Silva.M. and Clayton,J. (1987b)

Hybridoma, 6, 71—76.Hermans,.!, and McQueen.J.E. (1974) Ada Cryst., A30, 730-739.Herron,J.N., He.X., Mason,M.L., Voss.E.W. and Edmundson,A.B. (1989)

Proteins: Struct. Fund. Genet., 5, 271-280.Jackson.T., Morris.B.A. and Sanders,P.G. (1992) Mol. Immunol., in press.Jeffrey,P.D. (1989) The Structure and Specificity of Immunoglobulins. D. Phil.

Thesis, University of Oxford.Jones.T.A. (1978)7. Appl. Cryst., 11, 268-272.Kabat,E.A, Wu,T.T., Reid-Miller,M'., Perry,H.M. and Gottsman.K.S. (1987)

Sequences of Proteins of Immunological Interest. Fourth edn. US Departmentof Health and Human Services, US Government Printing Office.

Land.R.B., Morris.B.A., Baxter.G., Fordyce.M. and ForsterJ. (1982)7. Reprod.Fertii, 66, 625.

Lascombe.M.B., Alzari.P.M., Boulot.G., Saludjian.P., Tougard,P., Berek.C,Haba.S., Rosen.E.M., Nisonoff.A. and Poljak.R.J. (1989) Proc. Natl Acad.Sci. USA, 86, 607-611.

Marquart.M., Deisenhofer.J., Huber.R. and Palm.W. (1980)7. Mol. Biol, 141,369-391.

Martin.A.C.R., Cheetham,J.C. and Rees.A.R. (1989) Proc. Natl Acad. Sci. USA.86, 9268-9272.

Morris.B.A., Rhind.S.M., Clayton.J., Price.C.A. and Webb.R. (1988)Proceedings of the 11th International Congress Reproduction and A.I. (Dublin),4, 498.

Morrison.S.L., Johnson,M.J., Herzenberg,L.A. and Oi.V.T. (1984) Proc. NatlAcad. Sci. USA, 81, 6851-6855.

Muller,R. (1980) 7. Immunol. Methods. 34, 345-353.Novotny,J. and Haber.E. (1985) Proc. Natl Acad. Sci. USA, 82, 4592-4596.Padlan,E.A. (1990) Proteins: Struct. Fund. Genet., 7, 112-124.Padlan.E.A., Silverton.E.W., Shenff.S , Cohen,G.H.. Smith-GilGill.S J. and

Davies.D.R. (1989) Proc. Natl Acad. Sci. USA, 86, 5938-5942.Price.C.A . Morris.B.A and Webb.R. (1987) 7 Reprod. Fertii. 81. 149.

Riechmann.L., Clarke.M., Waldmann.H. and Winter,G. (1988) Nature, 332,323-327.

Roberts.S., Cheetam.J.C. and Rees.A.R. (1987) Nature, 328, 731-734.Satow.Y., Cohen,G.H., Padlan,E.A. and Davies.D.R. (1986)7. Mol. Biol, 190,

593-604.Saul,F.A., Amzel,L.M. and Poljak.R.J (1978)7. Biol. Chem., 253, 585-597.Scaramuzzi.R.J., Davidson,W.G. and van Look,P.F.A. (1977) Nature, 269,

817-818.Sheriff,S., Silverton.E.W., Padlan.E.A., Cohen.G.H., Smith-Gill.S.J.,

Finzel.B.C. and Davies.D.R. (1987) Proc. Natl Acad. Sci. USA, 84,8075-8079.

Sreenan,J.M., Morns.D., Tait.A. and Diskin.M.G. (1987) In Roche.J.F. andO'Callaghan.D. (eds), Follicular Growth and Ovulation Rate in Farm Animals.Martinus Nijhoff, Dordrecht, pp. 73-86.

Tronrud,D.E., Holden.H.M. and Matthews.B.W. (1987) Science, 235, 571-574.Watson.H.C, Walker.N.P.C, Shaw.P.J., Bryant.T.N., Wendell.P.L.,

Fothergill.L.A., Perkins.R.E., Conroy,S.C, Dobson.M.J., Tuite.M.F.,Kingsman.A.J. and Kingsman.S.M. (1982) EMBO J., 1, 1635-1640.

Webb.R., Land.R.B., Pathiraja.N. and Morris.B.A. (1984) In Chrichton.D.B(ed.), Immunological Aspects of Reproduction in Mammals. Butterworth,London, p. 475.

Received on December 23, 1991, revised and accepted on March 18, 1992

350

at The U

niversity of British C

olombia L

ibrary on Novem

ber 24, 2014http://peds.oxfordjournals.org/

Dow

nloaded from