`
`Domain Association in Immunoglobulin Molecules
`The Packing of Variable Domains
`
`Cyrus Chothia, Jiri Novotny, Robert Bruccoleri
`and Martin Karplus
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1063 Page 1 of 14
`
`
`
`J. Mijl, Biol. (1985) 186, 651-663
`
`Domain Association in Immunoglobulin Molecules
`The Packing of Variable Domains
`
`Cyrus Chothia
`
`MRC Laboratory of Molecular Biology
`Hills Road, Cambridge CB2 2QH
`
`and
`
`Ingold Laboratories
`Christopher
`Department of Chemistry, University College London
`20 Gordon Street, London WCIH OAJ, England
`
`Jiri Novotny, Robert Bruccoleri
`
`Molecular dh Cellular Laboratory
`Massachusetts General Hospital
`
`and
`
`Harvard Medical School, Boston MA 02114, U.S.A.
`
`and
`Martin Karplus
`
`Department of Chemistry
`Harvard University, Cambridge MA 02138, U.S.A.
`
`(Received 17 July 1984, and in revised form 19 July 1985)
`
`We have analyzed the structure of the interface between VL and VH domains in three
`immunoglobulin fragments: Fab KOL, Fab NEW and Fab MCPC 603. About 1800 A^ of
`protein surface is buried between the domains. Approximately three quarters of this
`interface
`is formed by
`the packing of the VL and
`\'H
`j8-sheets in the conserved
`"framework" and one quarter from contacts between the hypervariable regions. The
`jS-sheets that form the interface have edge strands that are strongly twisted (coiled) by
`^-bulges. As a result, the edge strands fold back over their own P-sheei at two diagonally
`opposite corners. When the VL and VH domains pack together, residues from these edge
`strands form the central part of the interface and give what we call a three-layer packing;
`i.e. there is a third layer composed of side-chains inserted between the two backbone side-
`chain layers that are usually in contact. This three-layer packing is different
`from
`previously described ^-sheet packings. The 12 residues that forin the central part of the
`three observed VL-VH packings are absolutely or very strongly conserved
`in all
`immunoglobulin sequences. This strongly suggests that the structure described here is a
`general model for the association of VL and VH domains and that the three-layer packing
`plays a central role in forming the antibody combining site.
`
`1. Introduction
`
`Immunoglobulins are the best-studied examples
`of a large and ancient family of proteins, which also
`includes )5-microglobulins. Thy-1 antigens, major
`
`(i.e. class II) histo(cid:173)
`I) and minor
`(i.e. class
`compatibility antigens and cell surface receptors.
`Functionally, all these structures are involved in
`cell recognition processes (Jensenius & Williams,
`1982), either actively as vehicles endowed with
`
`(M)i>2-2836/85/230651-13 $03.00/0
`
`651
`
`© 1985 Academic Press Inc. (London) Ltd.
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1063 Page 2 of 14
`
`
`
`652
`
`C. Chothia, J. Novotny, R. Bruccoleri and M. Karplus
`
`recognition specificity (antigen-combing antibodies)
`or passively as surface structures
`t h at are being
`recognized (histocompatibility antigens). Only
`the
`immunoglobulin
`tertiary structures are known
`to
`date (Schiflfer el al, 1983; E pp et al., 1974; Saul et
`al., 1978; Segal et al, 1974; Marquart et al, 1980;
`Deisenhofer,
`1981; Phizackerley
`et al,
`1979).
`However, the homology among primary structures
`of immunoglobulin, /^-microglobulin, Thy-1 antigen,
`some of the histocompatibility antigen domains,
`T-cell
`receptor P chain and
`the
`transepithelial
`"secretory component" has been
`interpreted as
`evidence for a common fold (Cunningham el al,
`1973; Orr et al, 1979; Feinstein, 1979; Cohen et al,
`1980, 1981a; Novotny & Auffray, 1984; Yangai et
`al, 1984; Hedrick et al, 1984; Mostov et al, 1984).
`
`A typical antibody molecule (IgGl) consists of
`two pairs of light chains [M, 25,000) and two pairs
`of heavy chains (M, 50,000), each of the chains
`being
`composed
`of
`domains made
`up
`of
`approximately
`100
`amino
`acid
`residues. The
`domains are autonomous folding units; it has been
`demonstrated
`experimentally
`(Hochman
`et
`al,
`1973; Goto & Hamaguchi, 1982) t h at a polypeptide
`chain segment corresponding to a single domain can
`be
`refolded
`independently of
`the
`rest of
`the
`polypeptide chain. All the immunoglobulin domains
`are formed by two jS-sheets packed face-to-face and
`covalently connected together by a disulfide bridge.
`The topology of the N-terminal, variable domains
`in both the light and heavy chains differs from t h at
`of the C-proximal constant domains. While the two
`variable domain sheets consist of five and
`four
`strands, respectively, the constant domain sheets
`are
`three- and four-stranded
`(Fig. 1). The
`four-
`stranded
`jS-sheets of
`the
`two domain
`types are
`homologous; the five- or four- stranded ^-sheet of
`the variable domains derives from the three-strand
`sheet of the constant domains by the addition, at
`one side, of a two-stranded ^-hairpin or a single
`jS-strand, respectively.
`In a complete immunoglobulin molecule, domains
`t h at correspond
`to different polypeptide
`chains
`associate to form domain dimers V L - V H, C L - C Hl
`and CH3-CH3. Edmundson et al. (1975) were the
`first
`to
`note
`the
`phenomenon
`of
`rotational
`allomerism between
`the variable and
`constant
`domain dimers, t h at
`is. whereas the C-C dimers
`interact via a close packing of their
`four-strand
`sheets, the V-V dimers pack "inside out", with the
`five-stranded
`sheets oriented
`face-to-face. The
`reversal of domain-domain
`interaction is reflected
`in the amino acid sequence homology between, and
`among,
`the
`constant
`and
`variable
`domains
`(Novotny & Franek, 1975; Beale & Feinstein, 1976;
`Novotny et al, 1977).
`same
`the
`in
`Different
`antibody molecules
`organism bind different antigenic structures. The
`variation
`in s])ecificity
`is produced by
`several
`mechanisms: mutations, deletions and insertions in
`the binding regions of the VL and VH domains; and
`the association of different light and heavy chains.
`.Aspects of the second mechanism are analyzed in
`
`this paper. In particular, the n a t u re of the interface
`between VL and VH domains
`is examined by
`comparing the F ab fragments of K O L, N EW and
`MCPC
`603 myeloma
`proteins whose X-ray
`structures are known. T he relative contributions to
`the buried surface between the domains from the
`conserved
`framework
`residues
`and
`the hyper(cid:173)
`variable
`regions
`are determined. Attention
`is
`focused on the unique packing of the interfaces and
`the reasons for this packing are examined.
`
`2. Materials and M e t h o ds
`
`(a) Fab fragment co-ordinala
`
`Cartesian co-ordinates for Fab fragments KOL. XEW
`and MCPC 603 were obtained from the Brookhaven Data
`Bank (Bernstein et al, 1977). Table 1 lists the domain
`classification, the nominal resolutions and the crystalio-
`graphic residuals (R factors) for the 3 Fab fragments. To
`facilitate comparisons of the 3 structures, their residue
`numbering was changed from that used in the original
`descriptions to that used by Kabat et al. (1983). Thus, in
`this paper residues that are structurally homologous have
`the same sequence number.
`To obtain consistent sets of atomic co-ordinates, the
`original co-ordinates were dissected into individual VI^
`
`(a)
`
`(b)
`immunoglobulin
`typical
`in
`Figure 1. The /J-sheets
`domams. Wrtices represent the position of Ca atoms:
`those
`in
`/J sheets are
`linked bv ribbons; and
`those
`between strands by lines, (a) The VL domain of KOL: the
`^-sheet involved
`in \'L-VH contacts is closer to the
`viewer (unbroken line), (b) The same \h domain rotated
`by appro.ximately 90°. Note that the
`interface-forminj;
`^-sheet IS str,,ngiy twisted at diagonallv opposite corners
`(drawmg by A. M. Lesk).
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1063 Page 3 of 14
`
`
`
`Packing of Immunoglobulin Variable Domains
`
`653
`
`Table 1
`Summary of X-ray crystallographic data
`
`L and H
`chain
`types
`
`Al, vIII
`
`XI, yll
`
`K, yl
`
`X-ray
`
`data
`
`Minimized
`
`Resolution
`(A)
`
`R factor
`(%)
`
`Energy
`(kJ)
`
`r.m.s. shift
`(A)
`
`Reference
`
`1-9
`
`20
`
`2-7
`
`26
`
`19
`
`24
`
`-3010
`
`-2592
`
`- 3 7 03
`
`—
`
`0-21
`
`0-26
`
`Marquart et al. (1980)
`
`Saul et al. (1978)
`
`Segal el al. (1974)
`
`Protein
`
`Fab KOL
`human
`Fab XEW
`human
`Fab MCPC 603
`mouse
`
`The energy given for Fab KOL is that of the unminimized crystallographic data.
`
`VH domain dimers. The structures were subjected to 100
`cycles of constrained energy minimization with
`the
`program CHARMM version 16 using the adopted-basis
`Newton-Raphson procedure (Brooks el al, 1983) with
`constraints of 41-8kJ (lOkcal) present on all the atoms
`(Bruccoleri & Karplus, unpublished results). Typically,
`the constrained minimization converged from original
`positive values of potential energy to values of about
`— 2TkJ/atom ( —0-50 kcal/atom) with an average root-
`mean-square co-ordinate different from the original X-ray
`structure of 0-3 A (see Table 1). The results indicate that
`the crystallographic structures were satisfactory and that
`acceptable values of potential energy can be achieved by
`small adjustments of the co-ordinates. Thus, both energy
`minized structures and the crystallographic co-ordinates
`were used in the present study: essentially
`identical
`results were obtained from the 2 types of co-ordinates
`sets.
`
`(b) Computation of solvent-accessible surfaces
`and contact areas
`Solvent-accessible surfaces (Lee & Richards, 1971) were
`computed with programs written by A. M. Lesk using the
`method of Shrake & Rupley (1973) and by T. Richmond
`using
`the methods of Lee & Richards
`(1971) and
`Richmond & Richards (1978). The latter program was
`obtained from Yale University. The water probe radius
`used was 1-4 A and the section interval along the Z axis
`was 0-05 A; the atom van der Waals' radii used were 2 A
`for all the (extended) tetrahedral carbon atoms, 1-85 A
`for all the planar (sp2 hybridized) carbons, 1-4 A and
`1-6 A for carbonyl and hydroxyl oxygens, respectively,
`1-5 A for a carbonyl OH group, 20 A for all
`the
`(extended) tetrahedral nitrogen atoms, 1-5 A, 1-7 A and
`1-8 A for «p2-hybridized nitrogen atoms carrying no
`hydrogen, 1 and 2 hydrogen atoms, respectively, 20 A for
`a sulfhydryl group and 1-85 A for a divalent sulfur atom
`with no hydrogens.
`
`(c) pStrands and ji-sheets
`
`Protein structures were analyzed using the CHARMM
`program (Brooks et al. 1983) in the so-called explicit
`hydrogen atom representation: aliphatic hydrogens were
`combined together with their heavy atoms into"extended
`atoms" whereas hydrogens bound to polar atoms and
`possibly
`involved
`in hydrogen bonds were explicitly
`present. The ^-strands and ^-sheets were defined by
`their inter-strand backbone (C = 0 . . . H-N) hydrogen-
`
`bonding pattern. A hydrogen bond list was generated in
`CHARMM for all the polypeptide chain segments under
`consideration and amino acids with hydrogen bonds of
`nearly optimal geometry (energy of —4T8kJ/bond or
`less) were taken to be parts of the ;8-sheets (cf Fig. 3 of
`Novotny et al, 1983). This method of defining ^-strand
`boundaries gives results essentially identical to those
`obtained by visual inspection of crystallographic models,
`although it tends to be somewhat more restrictive (the 2
`methods sometimes differ
`in inclusion of the N- or
`C-terminal ^-strand residues). Ambiguities arise in cases
`of edge /3-strands that start and end with
`irregular
`conformations ()S-bulges); such cases are discussed in
`more detail below.
`
`(d) ^-Strand conformation
`
`In a typical extended polypeptide chain segment, the
`dihedral angle between the 2 consecutive side-chains is
`not 180° as in the ideal ^-sheet (Pauling et al, 1951) but
`closer
`to —160°; that
`is, the ^-strands are
`twisted
`(Chothia,
`1973).
`The
`out-of-planarity
`angle
`(180°-160°) = 20' can be obtained explicitly from the
`values of the principal backbone torsion angles tp, \ii and
`0) (see. e.g. Chou el al, 1982). We define the local
`backbone twist for 2 consecutive residues as:
`
`9=
`
`-
`
`(180-
`
`where T is the torsion angle C/J-Ca-C"a-C'jS and |T| denotes
`its magnitude. When glycine residues that lack C;8 atoms
`are encountered, the torsion angle 9 is measured with
`respect to the C'jS atom following the glycine. Thus,
`glycine residues contribute to the local backbone twist
`indirectly, by being included in the virtual bond Ca-Ca
`that spans from the residue preceding the glycine to that
`which follows it.
`Backbone twist profiles (plots of 9 as a function of the
`amino acid residue) serve to characterize polypeptide
`chain conformations. Certain conformational character(cid:173)
`istics of polypeptides are more clearly seen using 8 values
`instead of the cp^ values for individual residues. In our
`plots, the value of the torsion angle Ca-C;8-C'a-C'^ is
`assigned to the second (C) residue. The angle 9 is related
`to "the amount of twist per 2 residues", defined as 8 by
`Chou et al. (1982); in fact, 9 = \5. It thus follows that 9
`can be obtained from the helical parameters n (number of
`residues per
`turn), h (the rise per residue) and T
`(T = 360°/») in a corresponding way to that descrilied for
`,5bvChou Wr//. (1982).
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1063 Page 4 of 14
`
`
`
`654
`
`C. Chothia. J. Novotny, R. Bruccoleri and M. Karplus
`
`3. Results
`(a) Domain-domain cotilarl .Kurfncen
`We identified the residues that form the interface
`between \'L and VH by calculation of the solvent-
`accessible surface of the domains, first in isolation
`and second when associated. Any residue that lost
`surface on the association of VL and VH was taken
`as part of the interface between them. We also
`determined which residues form van der Waals'
`contacts across the interface (distance cutoff 41 A).
`The hsts of residues obtained by the two methods
`were very similar. Thus, exce])t for a few marginal
`cases, the residues that
`lose surface in domain-
`domain contacts also have van der Waals' inter(cid:173)
`actions between the domains, indicating that the
`\'L-VH interface is tightly packed.
`The total surface areas of the separated VL and
`VH domains and that buried on the association is
`shown in Table 2. The values for the buried surface
`area (between 1700 and 1900 A^) and the fraction of
`the buried surface that is composed of polar atoms
`are similar to those found in other cases (Chothia &
`Janin, 1975). For the bovine pancreatic
`trypsin
`inhibitor and trypsin it is known that the structure
`of
`the
`isolated
`proteins
`does
`not
`change
`significantly on association. In most cases, as for
`the VL and VH domains considered here, there are
`no
`data
`concerning
`the
`structure
`of
`the
`unassociated domains.
`Of the total area buried between the VL-VH
`dimers about one quarter comes from residues in
`the hypervariable regions and about three quarters
`from
`residues
`in ^-sheets. Figure 2 shows
`the
`residues that form the interfaces and the areas that
`are buried for the three VH-VL packings. Two
`important features are evident in this Figure. First,
`homologous residues form the interface in the three
`structures. Second,
`the pattern formed by
`the
`contact residues is most unusual. The contacts of
`residues on the edge strands of the /8-sheets are
`more extensive than those of residues on the innei-
`strands. This is the opposite of the behavior found
`in previously described jS-sheet jiackings, where it
`is the central strands that have the largest contact.
`
`For example, for packing of j9-sheets in the same
`domain, the region of maximal contact generally
`runs diagonally across the sheets at 45° with respect
`to the )?-strands (Cohen et al, 19816; Chothia &
`Janin, 1981). The point is clearly illustrated in the
`Ca backbone plot in Figure 2(c); here, for each of
`the Ca atoms a circle is displayed, the area of which
`is proportional to the total contact area made by
`the residue with the other sheet. As we describe
`below, the unusual packing is a direct consequence
`of the distortions present in this type of ^-sheet.
`
`(b) Conformation of interface P-sheets
`the
`The deviation of
`the conformations of
`jS-sheets that form the interface between VL and
`VH from the idealized flat structure (i.e. twisting,
`coiling and bending) can be characterized by the
`variations in the twist angle 9 (see Materials and
`Methods). On such twist profiles, regular twisted
`P-sheets correspond
`to horizontal
`lines with an
`average 9 = -1-20°, right-handed a helices to lines of
`9 = — 110° and tight reverse turns as triplets of
`points of approximately the same magnitude and
`alternating sign. The insertion of an additional
`residue in an edge strand of a ^-sheet, so that two
`edge residues face one another on an inner strand,
`forms what has been called a jS-bulge (Richardson et
`al, 1978). Such insertions can have a variety of
`conformational effects depending upon the exact
`(pi// values of the inserted residue and those of its
`neighbors. Usually a sharp bend or local coiling is
`produced in the edge strand; this gives rise to a
`single- oi' double-point peak or trough in the 9
`values.
`In Figure 3 we show the 9 values for the VL-VH
`interface segments (jS-strands with
`the adjacent
`hypervariable loops) in KOL, NEW and MCPC 603.
`Two
`important
`features of
`these
`jS-sheets are
`evident
`from
`the Figure. First, most of
`the
`individual values of 9, and the patterns formed by
`the variations in 9 angles, are very similar in the
`different sheets, particularly in the inner )?-strands
`(;81, i93, P5 and )38 of Fig. 3) and in the i?-bulges; the
`edge jS-strands (;S2, )S4, ^36 and jS9 of Fig. 3) have
`
`Table 2
`Accessible surfaces and those lost on VL-VH association
`
`(A')
`
`Domain pair
`
`KOL VL domain
`KOL VH domain
`VL-VH in KOL
`XKW VL domain
`XP:\V VH domain
`\'L VH in XEW
`.MCPC 603 VL domain
`M(;PC 603 VH domain
`Vl^VH in MCPC 603
`\'L-VH averaj^e
`
`Isolated surface
`
`Contact
`
`surface
`
`Hydrophobic
`
`Polar
`
`Total
`
`Hydrophobic
`
`Polar
`
`Total
`
`1)21
`1216
`2337
`1233
`1186
`2419
`1082
`1156
`2238
`2331
`
`658
`700
`1358
`744
`801
`1545
`689
`760
`1449
`1714
`
`1779
`1926
`3705
`1977
`1985
`3962
`1771
`1916
`3687
`3785
`
`580
`615
`1195
`529
`506
`1035
`676
`619
`1295
`1195
`
`311
`250
`.561
`387
`386
`773
`299
`324
`623
`652
`
`891
`865
`1756
`916
`892
`1808
`975
`943
`1918
`1827
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1063 Page 5 of 14
`
`
`
`Packing of Immunoglobulin
`
`Variable Domains
`
`655
`
`I
`I S23
`95L76::::92
`I P36
`I
`I Y9I
`I WI06
`96R143
`9IY39
`I LS4
`I D64
`
`31
`
`I TI5
`32H24
`I F55
`
`9 7:
`
`; 90
`
`33
`
`I F96
`98F9S
`I FIIO
`
`I N24
`I AO
`8 9 Q 6 : : ; : 3 4 K 30
`I Q9
`I A9
`
`I P50
`l o om
`I F62
`I D27
`IOID52
`I DI8
`
`I
`96
`I
`I
`I DO
`95N3I_.,.33
`I N20
`I
`
`I 0 2.
`
`-94
`
`34
`
`I W75
`I03W7I
`I W73
`
`I YI2
`I AO
`9 3 A o : : : : 3 5 To
`I AO
`I E9
`
`^ 9 9-
`T7
`GI4
`A32
`
`lOO
`
`lOI
`
`88
`
`35
`
`I Y41
`1 Y48
`87Y5i::::36Y4i
`I Y44
`I Y43
`
`i L45
`46L57
`I L50
`
`^I04"
`I05 QI4
`025
`AI7
`
`I06
`
`92
`
`36
`
`I V6
`I F32
`9IY29::::37V6
`V5
`Y33
`
`I was
`47W88
`wes
`
`::-V-V.9o
`I07
`
`1
`
`8 9 '"
`
`I 08
`
`3 8 : :: ::::.^
`46
`
`Q48
`9Q20
`•3
`Q32
`
`V. Lies
`45 LI04
`I L88
`I G30
`44GI3
`•' R96
`
`1 0 2 : : : ; : :: 86
`
`3 7:
`
`-45.
`
`103
`
`8 5:
`
`I Q29
`. 38 038
`I Q35
`
`104
`
`84
`
`39"
`
`"42
`
`X P69
`44P67
`I P70
`
`I A52
`43A63
`P6I
`
`'
`
`VL
`
`(a)
`
`: : : : : ; : 88
`I09:
`
`40
`
`4
`43
`
`VH
`
`(b)
`
`( c)
`Figure 2. fi-Sheet residues that form the VL-VH interface in the Tabs KOL. NEW and MCPC 603^ Residue numbers
`are those of Kabat ./ al
`(1983). (a) VL interface-forming ^-sheet; (b) VH interface-formmg ^-sheet^ Broken Imes
`indicate hydrogen bonds. At each position where a residue forms part of the interface, we g.ve the residue -dentity in
`KOL. NEW and MCPC 603, and the accessible surface of the residue that i^^uned in the VL-\ H mterface^ Xote^^^^^^
`^-bulges in the edge strands at positions 43. 44 and 100, 101 in VL and 44. 45 and 105. 106 in the \ H. (c) The J-sheet
`from KOL VH domain. Residues making contacts to the VL domain across the domain-domain •"*« f^;^ ^'^-^'f^^
`The main-chain atoms are displayed. The circles associated with each Ca atom have an area P ^ P "^ " " t he e ^S
`accessible surface area lost when the VL-X'H dimer forms. Note the large areas associated with residues in the edge
`strands of the /S-sheet.
`
`25
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1063 Page 6 of 14
`
`
`
`656
`
`('. Chothia, J. Novotny, R. Bruccoleri and M. Karplus
`
`150
`
`-
`
`100
`
`<u a o
`o
`n
`
`- 50
`
`-
`
`- 1 00
`
`- 1 50
`
`-
`
`150 -
`
`100
`
`-
`
`50 -
`
`0)
`
`a
`o
`M o
`n -50
`
`-100 -
`
`-150 -
`
`Sequence number
`
`(a)
`
`85
`
`90
`
`95
`
`100
`
`105
`
`Sequence number
`
`(b)
`Figure 3. The backbone twist (9) profiles of VL-VH interface-forming segments. The segments shown include the
`hypervariable loops (LI, L2. L3, HI, H2 and H3) and the )?-strands. The j8-strands are indicated by bars at the bottom
`of the plots and labeled ^1 through ^9 according to Novotny et al. (1983). /S-Bulges are denoted by open boxes. Seciuence
`numbers correspond to the Kabat et al. (1983) numbering system and are the same as in Fig. 2. (a) and (b) The 2
`interface-forminfi sefiuients of the \'L domain; (c) and (d) the 2 interface-forming segments of the VH domain.
`(O) KOL; (D) NEW; (A) M('I><' 603.
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1063 Page 7 of 14
`
`
`
`Packing of Immunoglobulin Variable Domains
`
`657
`
`Sequence number
`(c)
`
`150 -
`
`100 -
`
`d
`o
`
`a
`«
`
`-50 -
`
`-100
`
`-150
`
`150 -
`
`100 -
`
`50 -
`
`01
`a
`o ,a
`a
`a
`m
`
`-50 -
`
`-100 -
`
`-150 -
`
`120
`
`j9-bulge
`of
`differences. Conservation
`greater
`conformations is especially striking and implies that
`they are important architecturally, as previously
`suggested
`by
`Richardson
`(1981).
`The
`correspondence in the ^-sheets is made even more
`
`the
`in behavior of
`the difference
`evident by
`hvfiervariable
`loops. The overall similarity of
`/?-sbeet geometries is confirmed by a least-squares
`fit of their atomic co-ordinates. Fits of the main-
`chain atoms of the three VL j8-sheets to each other
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1063 Page 8 of 14
`
`
`
`658
`
`C. Chothia, J. Novotny, R. Bruccoleri and M. Karplus
`
`Pro
`lOO
`
`I04
`
`84 39
`
`VL
`Figure 4. The key residues in the edge strands involved in VL-VH packings (Fab KOL). Note how in (a) Pro44,
`Tyr96 and Phe98 in VL and in (b) Leu45, ProlOO and Trpl03 in VH fold over the central strands of their ^-sheets and
`so in (c) form the core of the VL-VH packing (see also the position of these residues in Fig. 5).
`
`VH
`
`VL-VH
`
`(30 residues), and of the three VH )3-sheets to each
`other
`(32
`residues)
`give
`root-mean-square
`differences in atomic positions of between 0-73 and
`1-23 A (see Table 3). If a few peripheral residues are
`removed from
`the fits the r.m.s.f differences are
`reduced to 0-55 to 0-87 A. Table 3 also reports the
`results of least-squares fits of the VL P-sheeis
`to the
`VH ^-sheets. The r.m.s. differences are only a little
`greater than for the fits of the VL or VH /S-sheets to
`each other, 0-70 to 1-11 A. Thus, the six regions of
`P-sheet
`t h at form
`the V L - VH
`interface in K O L,
`N EW and MCPC 603 have very similar structures.
`In fact, the least-squares superposition of the two
`sheets can be achieved as a 2-fold
`symmetry
`operation,
`i.e.
`rotation around an axis passing
`through the centroid of the interface.
`The second feature of the |8-sheets illustrated in
`Figure 3 is the different amounts of twist found in
`the edge and inner strands. The two central strands
`in both VL and VH have 3 values in the range t h at
`indicate a degree of
`twist commonly
`found
`in
`P-sheets. The average 9 value tends to be the same
`for both the inner and the edge strands, b ut
`the
`twist of the edge strands is dominated by j8-bulges
`(Figs 2 and 3) with characteristic 9 values ± 7 0. I ts
`effect is to fold the ends of the edge strands over
`central strands. This occurs at
`two diagonally
`opposite corners of
`the P-sheets.
`Side-chains of
`residues 44 (Pro), 96 (Tyr, Arg, Leu) and 98 (Pro) in
`VL and 45 (Leu), 100 (Pro, He, Phe) and 103 (Trp)
`cover residues in the inner strands (Fig. 4(a) and
`(b)). The other parts of the edge strand residues,
`45-46 and 101-104 in VL, 46-48 and 106-109 in
`VH,
`lie next
`to the
`inner strand
`in the normal
`manner.
`
`t Abbreviation used: r.m.s., root-mean-square.
`
`(c) Packing of the P-sheets at the VL-VH
`
`inteiface
`
`As noted above, the strong twists t h at occur in
`the edge strands of VL and VH means t h at residues
`at two diagonally opposite corners fold over the
`^-sheets: 44, 96 and 98 in VL (Fig. 4(a)), and 45.
`100 and 103 in VH (Fig. 4(b)). Figure 4(c) shows
`t h at when the VL and VH domains pack together
`
`Table 3
`The fit of P-sheets forming VL- VH
`A. Fit.'! of individual P-sheets
`
`V Lt
`
`interfaces
`
`VHJ
`
`KOL NEW
`
`MCPC
`
`KOL
`
`NEW
`
`MCPC
`
`0-76
`—
`— —
`
`0-55
`0-82
`—
`
`VLt KOL
`NEW
`MCPC
`V H | K OL
`XEW
`MCPC
`
`0-88
`0-96
`0-70
`—
`
`111
`105
`100
`0-87
`—
`
`0-94
`0-97
`0-84
`0-65
`0-87
`—
`
`B. FiU of both (i-sheel regions of the VI^VH interfaces^
`
`KOL NEW MCPC
`
`0-87
`
`0-70
`0-87
`
`KOL
`NEW
`MCPC
`
`The Table i;ive.s r.m.s. differences in position of the main chain
`atoms
`following
`least-squares
`fits of
`their co-ordinates.
`Differences are given in A.
`t VL residues used to determine fits and r.m.s. differences
`33-39, 43-47. 84-90 and 98-104.
`I VH residues used to determine fits and r.m.s. differences
`33-40, 44-48, 88-94 and 102-109.
`§ Residues used in fits 33-39, 43-47, 84-90 and 98-104 of VL
`and 34-40, 44^8. 88-94 and 103-109 of VH.
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1063 Page 9 of 14
`
`
`
`Packing of Immunoglobulin
`
`Variable Domains
`
`659
`
`VL
`
`90
`i\ Tyr
`
`/..
`
`(a)
`
`I03
`f 102 ^ Trp
`\ Tyr
`/\
`
`y-Jz^^i.^^
`
`-«'' -"
`
`\'ri7% K
`i ' V Q V\ ^/ „
`38 r'
`\f
`\ Arg ; ^ - -^
`V ^ v_
`
`V.
`
`'
`
`VH
`
`(b)
`
`47
`
`VL
`
`35
`Trp
`
`46
`Leu
`
`,Asr
`
`89
`Ala
`
`/
`
`1 U Pro
`X
`W-v
`M A la A V - - ^ T y r y «'
`n','^36 <_V_Jrp
`
`VH
`
`97
`Val
`
`) o
`
`VH
`
`(c)
`
`(d)
`
`Figure 5. Residue packing at the KOL VL-VH interface.. This Figure shows superimposed serial sections cut through
`a space-filling model of the interface. VH residues are shown by broken lines and VL residues by continuous lines. The
`p.seudo 2-fold axis that relates VL to VH is perpendicular to the page. Each part of the Figure shows 4 sections.
`separated by 1 A. superimposed, (a) Sections 0 to 3 A; (b) sections 4 to 7 A; (c) sections 8 to 11 A; and (d) Sections 12 to
`ISA.
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1063 Page 10 of 14
`
`
`
`660
`
`('. Chothia. J. Novotny, R. Bruccoleri and M. Karplus
`
`Table 4
`VI^VH
`Re.sid '»e.s b
`urietl
`in
`
`interfaces
`
`Residue at this
`in
`
`position
`
`surface
`Accessible
`Df residue (A )
`area <
`
`Domain
`
`Residue
`No.
`
`KOL
`
`NEW
`
`MCPC
`
`KOL
`
`XEV\'
`
`MCPC
`
`No. of sequences known that
`include this positiont
`
`Principal residues found
`at this nositiont
`(identity and number of cases)
`
`\L
`
`VH
`
`34
`36
`38
`44
`46
`87
`89
`91
`96
`98
`35
`37
`39
`45
`47
`91
`93
`95
`100
`103
`
`Asn
`Tyr
`(iln
`Pro
`Leu
`Tvr
`Ala
`Trp
`Tyr
`Phe
`Tvr
`Val
`Gin
`Leu
`Trp
`Phe
`Ala
`.Asj)
`Pro
`Trp
`
`L\'s
`Tvr
`Gin
`Pro
`Leu
`Tyr
`Gin
`Tyr
`Arg
`Phe
`Thr
`Val
`Gin
`Leu
`Trp
`Tyr
`Ala
`Asn
`He
`Trp
`
`Ala
`Tyr
`Gin
`Pro
`Leu
`Tyr
`Gin
`Asp
`Leu
`Phe
`Gin
`Val
`Gin
`Leu
`Trp
`Tvr
`Ala
`Asn
`Phe
`Trp
`
`2
`1)
`2
`8
`17
`9
`0
`3
`6
`10
`0
`0
`8
`10
`11
`1)
`0
`0
`0
`27
`
`39
`0
`7
`5
`35
`1
`1
`12
`5
`9
`2
`3
`20
`6
`6
`8
`0
`0
`32
`28
`
`0
`1
`17
`5
`8
`U
`0
`0
`3
`2
`0
`1
`21
`3
`4
`11
`1
`5
`0
`26
`
`t Data taken from Kabat et al. (1983).
`
`362
`318
`302
`238
`235
`227
`217
`211
`199
`206
`217
`200
`183
`163
`157
`1.59
`161
`131
`113
`125
`
`Alall7, Asn92, His51. Ser37
`Tvr243. Phe40. Val28
`G'ln279
`Prol90. Phe29. \ a l l4
`Leul.57, Gly32. Prol9. Vall3
`Tyr 160, Phe65
`Gin 128, Ala35
`Trp.59, TyrSl, Ser27
`Trp46, Tvr31. 126. R20
`Phe203
`Gln53. Asn42, Ser34. Lvs22
`Vall78. Ilel9
`Gin 176
`Leu160
`TrplSl
`Tvr 128, PheSO
`Ala 146
`Asp53, (;iyl8
`Phe76. MetU. Leu6
`Trp 118
`
`these six residues form the center of the interface.
`They are in contact with each other in pairs and
`make a herringbone pattern.
`the VI^-VH
`Details of how residues pack at
`interface can be seen in sections cut through space(cid:173)
`filling models. Figure 5 shows sections of the KOL
`VL-VH interface. The central role played by the
`three pairs of edge residues, Tyr96 and Trp 103, and
`Leu45 and Pro44 are seen in parts (b), (c) and (d) of
`the Figure. The inner strands of the j8-sheets, 32-39
`and 84-92 in VL and 33-40 and 88-95 in VH, only
`make
`interdomain contacts at one end of
`the
`interface where the side-chains of Gln38 and Gln39
`hydrogen bond
`to each other
`(Fig. 5(a)). The
`structures of the VL-VH interfaces in NEW and
`MCPC 603 are very similar
`to
`that of KOL
`illustrated here. This is demonstrated by graphical
`inspection of their packing and by the fits of the co(cid:173)
`ordinates of the main-chain atoms forming the VL-
`VH interfaces described above (Table 3).
`The packing of the ^-sheets at the three \'L-\'H
`interfaces can be described in terms of a three-layer
`structure: an inner layer consisting of large side-
`chains from strongly
`twisted ends of the edge
`strands; and two outer layers formed by the main
`and side-chains of the inner )?-strands and
`the
`middle part of the edge strands (Figs 4 and 5).
`
`(d) Three-layer pncking n.s a general model for
`VL-VH
`ns.'^ocialinns
`Ten years ago Poljak et al (1975) examined their
`Fab XEW structure and noted that the residues
`
`that form the \T.,-VH interface were conserved in
`the other immunoglobulin sequences then known.
`They predicted that
`the mode of association of
`other VL-VH dimers would be the same as that
`found
`in Fab NEW. The structures and many
`sequences determined since then, and
`the work
`reported here, confirm their prediction.
`The three structures studied here include a wide
`range of immunoglobulins: human X and y to mouse
`K and y (Table 1). In KOL, NEW and MCPC 603
`residues at about ten positions in VL and in VH are
`buried in the interface between the domains. The
`amino acid sequences of many other
`immuno(cid:173)
`globulins have been determined and a tabulation
`published by Kabat et al. (1983). We examined the
`tables of VL and VH sequences
`to find what
`residues occur at positions homologous to the 20
`buried in the \'L-VH interfaces studied here. The
`results of this survey are given in Table 4 and
`Figure 6.
`At 12 of the 20 positions residue identity is
`absolutely, or very strongly, conserved: in VL.
`residues 36. 38, 44." 87 and 98; in VH, residues 37,
`39, 45, 47, 91, 93 and 103. As shown in Figure 6,
`these residues form the whole of the central and
`lower regions of the interface. The eight positions
`that have some variation in residue identity are all
`in the upper part of the interfaces where they are
`adjacent
`to and partially buried by the hyper(cid:173)
`variable regions. The three structures studied have
`a range of residues at these positions that is fairly
`representative of those found
`in other sequences
`(Table 4). Inspection of the three structures shows
`
`PETITIONER'S EXHIBITS
`
`Exhibit 1063 Page 11 of 14
`
`
`
`Packing of Immunoglobulin
`
`Variable Domains
`
`661
`
`96 0-23 W
`0-16 Y
`
`91 0-28 W
`0-25 Y
`
`VL
`
`98 0-99 F
`
`89 0-60 Q
`0-16 A
`
`34 0-32 A
`0-25 N
`
`X = + I8 A
`
`99
`
`I -00 G
`
`101
`I
`
`l-OOG
`
`87 0'99F/Y
`I
`
`36 0-90 Y/F
`
`I
`46 0-67 L
`0-14G
`
`38 0-92 Q
`I
`
`44 0-80 P
`I 0-12 F
`
`x= OA
`
`100 0-73 F
`0-10 L
`
`95 0 - 4 0D
`0 - I 4G
`
`VH
`
`103 0-94 W
`
`93 0-91 A
`
`35 0-23E
`0 - I 9N
`
`104 0-98 G
`
`91 0-99 Y/F
`
`37 0-89 V
`0-10 I
`
`47 0-93 W
`
`106 0-98 G
`I
`
`45 0-98 L
`
`39 0-96 Q
`I
`Figure 6. The conservation of residues that form VL-
`VH interfaces. On a plan of the VL and \'H )S-sheets we
`show the principal residues found at sites buried in the
`interface and at sites involved in the formation of the
`;8-bulges. At each site we note the proportion of known
`sequences that contain the given residue, for example,
`0-99 of known VL sequences have Phe at position 98 (see
`Table 4). The one-letter code for amino acids is used.
`
`I 04
`LEU
`
`/
`
`f R/
`
`' ; P HE If:.
`
`'^C^^'--
`" ^"^v
`
`VAL
`
`V^^.ARG/'l
`
`;if = OA
`
`residues are accommodated by
`that the different
`small conformational changes in the ends of
`the
`)?-strands and
`somewhat
`larger changes
`in
`the
`hypervariable
`regions. The G l y - X - G ly
`sequence
`that produces the )S-bulge at residues 99-101 in VL
`and 104-106 in VH is absolutely conserved in VL
`and VH sequences (Fig. 6). Thus
`the pattern of
`conserved
`residues
`in VL
`and VH
`sequences
`suggests t h at the three-layer packing found for the
`structures studied here is a general model for V L-
`VH associations.
`
`(e) Three-layer