160 lm111111wlogy Toda.Y. uo/. J, No. 6, /fJ,~2

The three-dimensional structure of antibodies

Markus Marquart and Johann Deisenhof er Max-Planck l nsritul fi.ir Biochemie, Abtc ilung S tr ukru rforschung IL D-8033 Marrinsried, f.R.G.

Antibody molecules are glycoproteins which occur in vertebrate species. T hey recognize and bind an enor­mous variety of foreign substances (antigens) a nd sub­sequently trigger further defense mechanisms at the molecular or cellular level. Specific recognition requires surface structures complementa ry to the antigen and hence a h uge v;:iriety of ant ibody molecules. In contrast the effector functions need identica l interaction sites in all an tibody molecules.

The determination of the primary structure of imrnunoglobulins1-3 and the X-ray crystallographic studies of several an tibody molecu les and frag­ments•.s.7.rn.iz-rs led to an advanced understanding of the way in which ant ibodies meet these opposing requirements.

IGG

F ig. I Schematic represcnlalion of an IgGl immunuglobulio molecule. The a rms ol' the Y-shaped molecule arc limned by 1he Fab pans, the s1em is made up by the Fe pan. The ligh1 c hains are linked to

1he heavy c hains by a disulphide bridge close to the C-tcrmi nus. The two heavy chains arc connected via two disu lphide linkages i11 the hinge regio n.

• £1$evl.;f' KM>m('"(l1nil P11'» 1\1~2 OJC.7-4Q J<J/82/ 0QOo..oo(l0/$2 l!J

Fig. I is a schematic drawing of a n anLibody molecu le of class IgG I. IL is composed of two identical heavy chains and two identical light cha ins with rnol. wts of 50,000 and 25,000, respectively. Both types of polypeptide chain are folded into domains: the four domains of the heavy chain are VH, C H I, CH2, and C H 3; the light cha in consists of the two domains VL and C~. All doma ins except CH2 are arranged in pairs which a rc held together· by non-cova lent forces. Inter-chain disulfide bridges provide further stabil ity.

Among antibody molecules of a given class and species, the V-domains differ considerably in amino acid sequence, whereas the C-domains have identical sequences. The V-domains are composed of about 110 amino acid residues at. the N-tcr minal end of heavy a n<l light chains. The VH-VL pair together forms the antigen b inding site; differem a nt ibody specificities are the result of different amino acid sequences of the V-domains. T he sequence variability in V-domains is most pronounced in a few hypervariablc regions. On t he other ha nd the framework res id ues are well con­served . The constant domains C I 12 and Cl-13 a re involved in effector fu nctions such as compl.ement activation and b inding to receptors on certain cell types. There is significant homology between the amino acid sequences of all C-domains, and of 1hc framework residues of V -domains.

Proteolytic cleavage at the hinge reg ion yields stable a nd functional fragments: t he antigen-bi nding rrag­rnent P'ab, and the Fe fragment (Fe was the first ant i­body fragment obtained in crys talline form)6 •

A

H

B F E

G c D ARltANGEMENT OF snt.ANDS I N BWHKlGLOSUU H DOMA I NS

X N-lfRMJ NUS UP. • C-HRMI NUS IJP

x

Fig. 2 Schc rna1ic drawing of the strand topology in a V­domain viewed parallel to the strands. (x ) and (• ) indi«<lle N- and C-terminal e nd s of the slrnnd~ poin1-ing LOw<!rdS I he Observer.

1 of 7 Celltrion, Inc., Exhibit 1082

!111mu11olugy Today, wl. 3, Nu. Ii, 1982

Besides IgC I, several other classes (IgM, lg!\, lgD, IgE) and subclasses of immunoglobulins have been identified; the differences between these are located in the constant region of the heavy chain. The two types of light chain (kappa, lambda) can combine with heavy chains of any class.

Domain folding The general folding pattern in a ll immunoglobulin

domains is very similar. It is shown schematically in Fig. 2 for a V-domain. The folding is characterized by two pleated sheets connected by an internal di ­sulphide bridge linking st rands 13 and C. The Lwo sheets cover a large number of hydrophobic amino acid side chains.

Despite that gross similarity there exist substanlial differences when one compares V- and C- domains: C-domains lack strand X, strand D is very short (2-3 amino acids) and connected to strand E. In addition the length of the loop regions in C-domains is different from V-domains, thus changing the overall shape con­siderably.

VH and V L, on the other hand, show only minor differences when compared with each other (except in the hypervariable regions) as do CL, C H I and CH3.

CH 2 represents yet a third type of domain, differentiated from the other C -domains mainly by the branched carbohydrate chain linked to it. It will be discussed in more detai l below.

Do main-do main interactio n Two kinds of domain interactions occur in immuno­

globulins: lateral (or trans) interactions and longi­tudinal (or cis) interactions.

In lateral interactions immunoglobu lin domains other than CH2 strongly associate to form modules VL-VH, CL-CHI, CH3-CH3. In V modules VH may be replaced by VL lo form light chain V dimers as seen in the Bence-Jones protein fragments Rei or Au7- 9 • In Bence-Jones proteins, which are light chain dimers, one of the light chains simulates the Fab pans of the heavy chain, as described for Mcg1 ~.

V modules associate in a differenL way than C modules do. In V modules MGCD faces (see Fig. 2) of the domains get into contact, in C modules the /\ BFE. faces are involved.

/\ considerable loss of accessible surl'ace area11 i~ connected with contact formation of the immuno­globulin domains. It amounts to 1760 A2, 1923 f..2 and 21801\2 for VL-VH, CL-CH I modules of IgC KoJIW and the C I l3-CH3 module of an human Fe fragmem '·'·11 respectively. In VL- VH association both framework residues and amino acids from hyper­variable segments <:u-e involved. I\ comparison of V­domain amino acid sequences of different animal species shows that the contacting framework residues are highly conserved. /\lso the consume dom;iin residues participating in lateral conlact arc either invariant or replaced by homologous residues in

161

different immunoglobulin chains. This low degree of sequence variability for the residues importan1 for lateral comacc formation provides an explanation for the fact that differenl L-chains c.;an associate wi th different f I-chains to give intact irnmunoglobulins.

In addition to the extensive Van dcr Waals cornacts , there exist a few trans hydrogen bonds, in which ma inly pol<1r side c;h;iin groups are involved. There <1re 1.wo s;i lt linkages in Kol CL-CH I contact: Glu 125 light chain - Lys 214 heavy chain, G lu 126 light chain - .Lys 148 heavy chain, which have their analgon in CH3 - CT-13 pairing: Glu 356- Lys 439, Glu 357 - Lys 370.

CJ 12 is an exception, as it forms a single unit without lateral domain interactions (see Fig. 3)*. Instead it interacts with bound carbohydrate, which is attached to Asn 297. The CH2 residues that are involved in carbohydrate contact are, with a few exceptions, structurally in the same positions as the residues that form the CH3-CH 3 contact (face ABFE in Fig. 2). This demonstrates that the carbohydrate in CH2 provides a substitute f'or the C-C con­tact and presumably helps to stabilize the CH2-domain. The branched carbohydrate forms a few hydrogen bonds with the CH2-domain, but the dom­inant interactions are hydrophobic in nature. The carbohydrate covers a hydrophobic pat.ch of the protein made up of Phe 241, 243, Val 262, 264, Tyr 296, Thr 260, Arg 301, which would otherwise be exposed to the solvent. The loss of accessible surface area of one C l-1 2 domain is 522 A2 , which is only about ha lf as much covered surface area as seen in CII3-CH3 contact ( 1080 A2). This observation could explain the apparent 'softness' of those pans of the CH2-domain, as seen in the crysta l structureu·1\

which are most remote from the C H3-CH2 interface. The functional re levance of ca rbohydra te in ami­

bodies is unclear. It might be involved in intracellular movements of the glycoproteins and in secretion lf•. 1•. It may well be that the origin of t he a ltered funct ional properties of carbohydrate-free antibody variants is structural destabilization.

ln contrast to the extensive lateral interac.:tions, nonbondcd longitudinal interactions a long the hcav} chain or light chain are much weaker or do not exist al

all. However, they arc interes1ing because con­formational changes in antibodies affect those inter­actions.

Fig. 3, which represents the Fe parr ol' .a n lg(; I molecule shows the CH2-CH3 interaction. With a loss in accessible surface area of 778 J..z this wntacl has rough ly one l hi rd of the size of C H3-CH3 cont11ct. The residues that participate in CI-12-CI-13 contact arc highly conserved in a ll lg classes, suggesting that this contact is likely LO be found in IgC a nd lg/\ 11nd as CH3-CH4 contact in lgE and IgM.

•Most t'<:adcrs will need a sicrco viewer (commcrciallr available ) 10 see in three dimensions the s tructures shown in 1hc paired diagrams on pages J 62, 163 and t 66.

2 of 7 Celltrion, Inc., Exhibit 1082

fig. 5 Amino acid comparison of residues 98-119 (Eu num­bering) of M603, New, Kol and Eu heavy chains. The underlined residues were left out in Fig. 6c.

End ofVH 98

M603: Cys Ala Arg Asn Tyr Tyr New Cys Ala Arg Asn Leu lie Kol Cys Ala Arg Asp Gly Gly Eu Cys Ala Gly Gly Tyr Gly

Gly Ala His lie

Ser Gly Gly Tyr

Thr

Fig . .3 Stereo drawing of a space fi ll ing model of human Fe-frag­ment. Th~ molecule is bui lt from two identical polypeptide ch;i ins (chain I, chain 2), and identical carbohydrate g roups. Both halves arc related by approximate diads.

Fig. 4 lgG l molecule Kol. The Fab parts and the hinge segrnen1 are well ordered in the Kol crystals, the Fe part is disordered and not visible.

PLEASE NOTE

We regret that for tech nical reasons it has not been possible to reproduce Figs 3, 4, 6, 7 and 8 with che colou r coding that a llows different parts of the molec­ules to be distinguished.

The rull-colou r diagrams, with explanatory legends, can be found in the personal monthly edition of /mm1111-ology Today dated June 1982.

D segment

Cys lie Phe Cys Ser Ser Ala Ser Cys Ser

Phe

3 of 7 Celltrion, Inc., Exhibit 1082

110 Try Tyr Phe Asp Val Try Gly

Asp Val Try Gly Gly Pro Asp Tyr Try Gly

Pro Glu Glu Tyr

J segment

Ala Gly Thr Thr Gin Gly Ser Leu Gin Gly Thr Pro Asn Gly Gly Leu

Fig. 6 Antigen binding region of lgGI Kol. (a) The extended third hyper­variable loop of the heavy chain folds into the putative antigen binding pocket. (b) C a backbone and sidcchains of Kol antigen binding pocket. (c) Artificial deletion of nine resid ues in the third hyper­variable segmem of Kol, which makes it of equal length with lgG I Eu'", reveals a deep curved cleft.

119 Val Thr Val Ser Ser Val Thr Val Ser Ser Val Thr Val Ser Ser Val Thr Val Ser Ser

4 of 7 Celltrion, Inc., Exhibit 1082

IM

The C H2-CH 3 orientation is found LO be somc::whaL variable and influe::nced by external forces. In the F<' fragment crystals t he 1 wo chemically identical chains a re in a different environment. As a conse::quencc the CH2-CH3 orientation varies by about 6°. In Fe-Protein A complex crystals this arrangement differs slightly from that of Fe crys1als1s.

More drastic changes are observed in VI-I-Cl 11 and VL-CL longitudinal contacts, when chemically different Fab fragments are compared. These differences in longitudinal arrangement are most con­veniently described by an elbow a ngle, which is enclosed by the pseudo diads rela ting VL to VH and CH1 to CL respectively. The elbow a ngle may vary from more than 170° to 135° when we compare Kol Fab with M cPc Fab12-1>. i9.zo.

In two cases the elbow angles of the same molecule in two different crystal lall ices were compared aud fou nd to differ by 8° and 17° respectively19.21 . Tn Fab New, with an elbow angle of approximately 137°, there exist a few longitudinal contacts between VL and C L and VI I and C l 11 22•21, whereas there are no non-bonded longitudinal contacts in intact Kol and Fab Kol (see Fig. 4), which are characterized by an open elbow angle. We interpret these observations to mean that in Fab Kol the V- C arrangement is flexible in soluLion. In the crys1al the molecule is stablized by packing interactions; these wi ll be discussed from a different poinl or view later.

The a n tigen-binding area Comparison or amino acid sequences of variable

parts has demonstrated the hypcrvariabil ity of some segments. These were considered to be involved in antigen bindingH. Indeed, crystal structure analyses of lg fragment-haptcn complexes show that haptens bind in a cleft or depression formed by the hyper­variablc segments.

The VL d imer of Rei7•9 may serve as an illustrat ive examplt:. The symmetrica lly arranged hyperva riablc regions form a deep slit- like pocket around the diad relating lhe two VL monomers. The walls of t he sliL are li ned by tyrps incs 49, 91, 96, /\sn 34 and G in 89; the bottom of the pocket is formed by T yr 36 and G in 89. A trinitrophcnyl group binds to the Rei fragment and fills the binding pocke1 completely.

Another example of an IgG fragment haptcn complex is Fab Ne", which is known to bind among other ligands a hydroxy derivative of vitamin K, 2\ .

The hypervariable segments of l'\cw form a ~hallow groove wit~ approxima1c dimensions of 16 x 7 A O:Jnd a depth or 6 I\.

McPc 603, a mouse lg/\ (K) Fab fragme11t 211 u inds phosphorylcholine. The site or hapten binding is ~· large wedge shape~ cavity, with dimensions 15 x 20 A and a depth of' 12 A. Only five of the six hypcrvariablc regions contribuLe to the forma tion of the cavity: L­chain hypervariable regions one and three, and all three H-chain hypervariable regions. The second hypervariablc region of L-chain is screened from 1hc

/rmm111ology Today, ~of. 3, No. fi, IY/i2

caviLy by the first hypervariable loop or L-chain and the third hypervariable loop of H-ehain. The deeper cavity in McPc603, as compared to Fab New, is due to longer hypervariable loops. The firs l hypervariablc region of L-cha in and the thi rd hypervariable region of I I-chain is three residues and the second hypcr­variable loop of the H-chain is two residues longer in Mcl'c603 than in r\ew.

Phosphorylcholine occupies only a small part of the cavity and interacts via Van der Waals forces, electro­static interaciions, and hydrogen bonds with the p ro1ein.

In contrast LO the above examples lgG Kol shows no cleft or depression in 1hc antigen-binding region. In lgG Kol the heavy chain has a rather long third hyper­variable loop, which contains six residues more than M603 and eight more residues t h;111 Fab New. The amino acid sequences of the third hypervariable regions of M60326, NcwH, Kol2' and Eu28 a rc com­pared in Pig. 5. T he sequence alignmc111 and classifi­cation in VH, I) and J segment26.i• is somewhat arbicrary, especially for the beginning ofthe.J segment as a nucleotide sequence has been determined only for M6Q326 . The additional residues in Kol with the nearly palindromic amino acid sequence -Gly-Phc­Cys-Ser-Ser-Ala-Scr-Cys-Phc-Gly fold into the puta­tive antigen binding site and fill it completely (sec r ig. 6a,b). The two cystcins arr disulphide bridged and form the start and endpoints of a shon ;1ntiparalle l ~­sheet, comprising residues -Cy~-Scr-Ser-Ala-Scr-Cys - . If in a model building experiment nine residues arc cut from the third hypcrvariablc region of the Kol heavy chain, thus making it of equal length with lgG I Eu2",

a deep curved cleft appears (Fig. 6c), which easily could accommodate haptcn~. With respect to the anti­gen binding area lgG Kol 1hus looks as if it carried its own haptcn in form of an extended th ird hype::r­variablc loop. Another peculiarity of lgG Kol mighi be of interes1 in that context. In the Kol crystal lattice the hypcrvariablc parts of one molecule !Ouch the hinge and spa1ially ac\jacent segments of a symmetrically related molecule. This contact consists of three salt linkages (Arg 49 light chain-COOH light chain, Asp 50 light chain-Arg 215 heavy chain, /\sp 53 heavy ch;1in-Lys 134 heavy chain). a few hydrogen bonds and extensive Van der \Vaals interactions. Thus, che lauice contact found in Kol crystals might give an instructive model for antibody- an1igen interaction, as antigens arc usually macromolecules which cover a much larger part of the antibody than haptcns do.

T he h inge segm en t The hinge segment which cova lcndy links Pab and

Fe parts, has a unique primnry and spatial structure. Its centra l region consists of two para llel disulphide­li nked poly L-p rolinc hel ices wirh an a mino acid sequence -Cys-Pro-Pro-Cys-1l.I '. In the lgG I subclass represented by the Kol molecule the poly-proline double helix is short (Fi~. 7). However, in lgG3 the hinge sequence is quadruplicated'" and model build-

5 of 7 Celltrion, Inc., Exhibit 1082

Immunology Today, uol. 3, Nv. 6, 1982

ing suggests that the poly-prolinc segment of this molecule may be more than 100 A long.

The poly-proline segment, a relatively rigid struc­rurc, is Aanked on both sides by Acxible segments: The segment on the N-te rminal side is well defined in the crystal lattice or Kol due LO crystal packing inter­act ions, but it lacks interna l interactions, that would provide stability in solution. The C terminal segment is di~ordcrcd and Acxiblc in Kol crystals and in 1 he Fe crystal structurc1 1.1~. The rigid hinge segment allows independent movement of the Fab arms and the Fe part. There is direct evidence for flexibility in the crystal lattice of Kol''-19 and Zie3 ' . This is in contrast to the abnormal lgG protein Dob, which lacks a hinge rcgion32• The significance of the hinge for Fab-Fc flexib il ity is obvious.

Complement binding The binding of the Clq component of the Cl

complex to antigcn-ancibody complexes is t he first step in the classical pathway of complement ac tivation 13·3 '· The Clq head pieces bind to the CH2 domains of antibodies3·1•3". Protein /I. , a constituent of the cell wait of Staphylococcus aureus, binds LO t he Fc­part of antibody molecules of certain classes and sub­classes, but does not interfere with complement binding. The determination of the crys tal st ructure of the complex between FB (one of the four Fe-binding domains of protein AH) and Fe-fragment showed that protein A b inds al the Cl 12-CH3 contact•S.Js. Fig. 8 shows a space-filling model of the FB-Fc complex. T he area of CH2 not covered by FB must contain the Clq binding site. In view of 1hc size of the Clq he<1d pieces (mo!. wt 50,000) it appears unlike ly that t hey can bind at the inner sides of C H2, i.e. nea r the carbo­hydrate. T he most plausible binding site is therefore near the tip of CH2 on t he outer side of the domain. It is worth mentioning that this region is disordered in crystals of' the FB-Fc complex which indicates th<tt this part of the C H2 domain is flexible. Possibly, flexibility is required for antibody C lq interaction.

Summary a nd perspectives Investigations of the three-dimensional architecture

of ;mtibodies h11ve elucidated the folding of the polypept ide chains into domains, and the spatial arrangement of the domains. The structural basis for understanding antibody specificity and ant ibody Aexibility was obtained. Segmental flexibi lity is a n important property of amibodics: Flexible segments of the polypeptide chains at the switch and hinge regions allow the !"ab fragments to change their shape and their relat ive orientation. Conformationill changes of this kind are necessary to meet the geometric require­ments which arise on binding of ant ibodies to multi­va lent antigens.

The undersrnnding of the effector functions of anti­body molecules is much less complete. One of the central problems is the explan:-ition of the strong e nh:-incement of C lq binding to antigen-antibody

165

complexes as compared to free ant ibody molecules. Two mechanisms have been considered (for a review see Ref. 39): since Clq is multimeric with at least six antibody binding sites, binding may be enhanced by the form<1tion of a ntigen-antibody aggregates through crosslinking. Alternatively, antigen binding might induce a conformational change in the Fe-part which enhances affi nity for Clq.

There is strong evidence for the importance or aggregation, but a mixed mechanism which involves aggregation and a conformational change cannot be ruled out.

The studies described here were almost exclusively carried out with myeloma or Hence-Jones proteins because these were 1he only homogeneous irnmuno­globulins which could be obtained in sufficient quant ity. However, in most cases the specificities of such molecules is unknown. Recent ly, large amounts or homogeneous antibodies elicited against strepto­coccal or pneumococcal polysaccharides became available from cen a in rabbit and mouse st ra ins"'·"'. These sources, and the use of hybrids obtained from myeloma and spleen cells have made it possible to obtain homogeneous antibodies of defined speci­fic ity·•?.•~. Structural studies of 'natural' antigen-anti­body complexes can be expected to lead to a more complete understanding of antibody function . Crystallographic work on a specific ant ibody and of its antigen is already in p rogress44 •

Acknowledgements We thank Prof. R. Huber for helpful discussions.

Refer ences I Edelman, G. M . ( 1970) Sti. :Im. /\ugusL, 8 t-S7 2 Porter, R. R . ( 1976) Sri. Am. Oc1obcr, 8 1- 87 3 Hi lschmann, N. (1%9).V11111rwi.r.1msdlflflm56, 19S-205 4 F.dmundson, /\. B., Ely, K. R. and Ahola, E. £. ( 1978) 01111.

To/1 . .lfnl. b1111111111Jf. 7, 95- 118 5 1\m7.el , L. M . and Polj ak. R. J. ( 1?79) Ann. Rn·. l/i1Jr/1m1. 48,

96 1- 997 6 Porter. R. R. (1958)..V11l1ur( / ,md1m) 182, 670-67 1 7 Epp, O .. Colman. P. M ., Fchlh<1mmcr, H., Bode, W., Schiffer,

:--~ .. Huber, R. and P<ilm. W. ( 1974) J\11r. ]. /Jinrhm1. 45, 513- 524

8 Fehlhammer, t I. , Schiffer, M., Epp, 0., Colman. P. M .. l.auman, E. E., Schwager, I'., Stcigcmann, W. and Schramm, 1 l..J. ( 1975) /Jin/1/i) '·'· Strn/'/ . .lfrd11111i.wu I, 139-146

9 Epp, 0., Laurnan, I~. E., Schiffer, M ., Huber, R. and Palm. W. ( 1975) 8i1Jdmnistn• 14, 4943-4?52

10 Edmundson, /\. B., Ely, K. R., /\bola, R. R., Schiffe r, M. and Paniagia1npou los. N. (t975) llinc"hmn<I')' 14, 3953-J961

11 Lee, B. mid Ri<:hards, F. M . ( 1970)]. .\In/. Jim/. 55, 379-400 t2 Colman, I'. M .. Deisenhofcr, .J .. Huber, R . and Pa lm. W .

(1976)] . . lfo/. /lfo/. 100. 257-282 13 Marquart, M ., Deiscnhofcr, J., Huber, R. and P~lm. W.

(I 980)]. ,\/11/. /Jin/. 14 1. 369-392 14 Dciscnhofcr, .J., Colman, P. M ., F.pp, 0 . and Huber-, R. (1976)

H11f'lir-.\r.rifr\ ,(. l'h1•.<i11f. 01r111. 357, 1421- 1434 I 5 Dei~cn hofcr,J. ( 198 1) Hi1H-lll'1ui.<lr.v 20, 236 1-2370 16 !'l'lclchers, F. ( 1?73) lli11rhr11111/TJ' 12, 1471-t47(1 17 Weitzman, S. and Scharf1, ~I. D. ( 1976)]. .lfol. /Jwl. 102 ,

237- 252 II! Hickman,!:>., Kukzycki, /\ . .Jr, 1.ynch, R. G. and Kornfeld, S .

c 1977)]. 11;"' · u,,111. 2s2. 4402-4408

6 of 7 Celltrion, Inc., Exhibit 1082

Fig. 8 Space filling model of th e FS ( pro te in A ) - F e co mplex .

19 Matsushima, M., Marquart, M., J ones, T. A., Colman, P. M ., Bartels, K., Huber, R. nnrl Palm, W. (1978) J. Mn/. BitJI. 121, 44 1-459

20 Segal, I). M., Padlan, E. /\ ., Cohen, G. H., RudikoIT. S., Potter, M. and Davies, D. R. ( 1974) Pmc. Nall Acarl. Sci. U.S.A. 71,4298-4302

21 /\bola, E. E., l::ly, K. R. and Edmundson, A. B. ( 1980) B1'telmmslry 19, 432-439

22 Poljak, R. J., Anncl. L. M., Chen, B. (,., Phiackerley. R. P. and Saul, F. (1974) !'me. Nntl Acnd. Sci. U.S.A. 71, 3440-3444

23 Saul, F., Arnzcl, l, . M. and Poljak, R. J. ( 1978)]. I/in/. Clum. 253, 585- 597

24 Wu, T. T. and Kabat, E. /\. ( 1970)]. /lrp. Mttl 132, 21 1-250 25 /\mzel, L. M ., Poljak, R. j.. Saul, F . . Varga, J. M . and

Richards. F. F. ( I 974) Proc. Nall Acad. Sci. U.S.A. 71, 1427-1430

26 Eady, P., Hua ng, H., Oavis, M., Calame, K. and Hood, L. (1980) Cell 19, 98 1-992

27 Schmidt, W., Jung, H. D., Palm, W. and Hi lschmann, N. (1981) p rivate communication

28 Cunningham, B. /\., Rutishauser, U., Ga ll, W. E., Cottlicb,P. D., Waxdal, M. J. and Edelman, G. M. ( 1970) /Jiochenustry '>, 3161-3170

29 Sakano, H., Maki, R., Kurosawa, Y., Roeder, W. and Toncgawa, S. (1980) N11ture ( l,ondo11) 286, 676-683

30 Michaelson, T . E., Frangione, B. and Franklin, E. C. (1977) ] . Ri11l. (.'hm1. 252, 883- 889

Fig . 7 Confor ma tion of the hinge regio n as seen in IgCI Kol.

3 I Ely, K. R., Colman, P. M., A bola , E. E., Hess, 1\. C., l'cabody, D. S., Parr, D. M ., Connell, G. E., Lauschinger, C. /\. and Edmundson , A . B. ( I 978) 8i.11chm1iflry 17. 820-823

32 Si lverton, E: W ., Navia, M.A. and Davies, D.R. (1977) Pwr. Nall Arorl. Sci. U.S.A. 74, 5140-5 I 44

33 Mueller-Eberhard, H. J. ( 1975) 111111. Rev. Bfocllrm. 44, 697- 724 34 .Porter, R. R. and Reid, K. 13. M. ( 1979) Adv. Pmt. Cltem. 33,

1- 71 35 Connell, C. E. a nd Porter, R. R (197 I) llwchm1. ]. 124, 53P 36 Yasmeen, D., Ellerson, J. R., Dorrington, I<. J. and Paimer,R.

H. (1976}]. lmmtmt1/. 116, 518- 526 37 Sjocdahl, J. (1977) far.J. lJiochem. 78, 47 1-490 38 Oeisenhofcr, J., Jones, T. /\., Huber, R., Sjoedahl, .J. and Sjoe·

quist,J. (1978) z. !'hysio/. (:hem. 359, 975-985 39 Metzger, J-1. (1978) Ctmt. Top. Mo/. lmrmmol. 7, 119-148 40 Jaton, J.-C., Huser, H., Braun, D. C. , C i vol, D. , P~dll,J. ;u1d

Schlessingcr,J. C. (1975) Bif)(/tr.mi.1try 14, 53 12- 5315 41 Braun, D. G. a nd Huser, H. ( 1977) in f>mgress in lmmwwfflgy If/

(Mandel,'!'. E., Cheers, C. H., Hosking, C. S .. McKenzie,[. F. C. and Nossa!, G. J. V., eds) pp. 255-264, l::lsevie r North· Holland, Amsterdam, New York, Oxford

42 Kuchler, G. and M ils1cin, C. (1975) Na11m ( l.1mdrm) 256, 495-497

43 Melchers, F., Potter, B. M. and Bethesda, N. W. (eds) (1978) C11rr. Top. Microbfol. lmmwlfll. 8 1

44 Colman, P. M ., Gough, K. H., Lilley, G. G., Ulagrove, R . .J., Webster, R. C. and Laver, W. G. (198 1) J. Mo/. Bini. 152, 609--014

7 of 7 Celltrion, Inc., Exhibit 1082