THK JOURNAL. OF BiOLOGICAL CHEMISTRY
`Vol. 253, No. 2, Issue of January 25, pp. 585-597, 1978
`Printed in U.S.A.
`
`Preliminary Refinement and Structural Analysis of the Fab
`Fragment from Human Immunoglobulin New at 2.0 A
`Resolution*
`
`(Received for publication, July 7, 1977)
`
`FREDERIC!< A. SAUL, L. MARIO AMZEL, AND ROBERTO J. POLJAKt
`From the Department of Biophysics, The Johns Hopkins University School of Medicine, Baltimore,
`Maryland 21205
`
`The three-dimensional structure of the Fab fragment
`from human myeloma lgG New has been refined using
`"model building" and " real space" procedures. By these
`techniques, the correlation between the amino acid se(cid:173)
`quences and the 2.0 A resolution multiple isomorphous
`replacement Fourier map has been optimized. The average
`shift of all atoms during real space refinement was 0.62 A. A
`list of the refined atomic coordinates for the 440 amino acid
`residues in the structure is given. Ramachandran plots
`prepared using the refined coordinates show a distribution
`of q,, 1/.1 angular values which corresp()nds to the predomi(cid:173)
`nant JI-pleated sheet conformation present in the structure .
`The structures of the homology subunits V11, VL, Cnl, and
`CL were superimPosed by pairs and quantitatively com(cid:173)
`pared. The closest similarities were @bserved between V H
`and VL and between Ciil and CL. Amino acid sequence
`alignments obtained from this structural superposition are
`given. The closest sequence homology in Fab New is ob(cid:173)
`served between CHI(')' heavy chain) and C1, (A. light chain).
`In addit ion, there is considerable homology between the
`variable and constant regions.
`The distances of close contacts between the homology
`subunits of F ab New have been determined. The closer
`contacts, those between atoms at a distance s; 1.2 times their
`van der Waals radii, are analyzed in relation to the constant,
`variable, and hypervariable nature of the immunoglobulin
`sequence positions at which they occur. Most of the residues
`which determine the closer contacts between V11 a nd VL and
`between CH 1 and Ci. are structurally homologous and highly
`conserved or conservatively replaced in immunoglobulin
`sequences.
`The relation between idiotypic determinants, antigen
`combining site and hypervariable regions, is discussed in
`terms of the refined model.
`
`In this paper we present the results of a preliminary
`crystallographic refinement and a list of atomic coordinates of
`* This work was supported by Research Grant AI 08202 from the
`National Institutes of Health and by Researeh Grants NP-141B and
`IM-105C from the American Cancer Society. The costs of publication
`of this article were defrayed in part by the payment of page charges.
`This article must therefore be hereby marked "advertisement" in
`accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
`t Recipient of United States Public Health Service Research
`Career Development Award AI-70091.
`
`Fab New. 1 Some features of the refined structure are discussed
`in relation to the genetic control and physiological function of
`immunoglobulins.
`It is generally accepted (see review in Ref. 1) that electron
`density maps calculated by multiple isomorphous replacement
`techniques contain significant errors which may lead to impre(cid:173)
`cise determination of structural details such as the location of
`amino acid side chain atoms, bond angles, r/l and ijl values, cis
`or trans character of proline residues, etc. This refinement
`project was undertaken w1th the aim of obtaining more
`accurate coordinates which can be appHed to structural studies
`of other immunoglobulins and Fab · hapten complexes (2). The
`starting atomic coordinates were those of the structure previ(cid:173)
`ously described (3, 4) obtained using multiple isomorphous
`heavy atom replacements. Since the complete refinement of
`the structure of Fab New is a complex undertaking, we
`present here initial results obtained after application of two
`consecutive refinement techniques. In the first step we have
`applied a "model building'' procedure (5) in which the mea(cid:173)
`sured atomic coordinates were adjus.ted to impose standard
`bond lengths and bond angles. In a second step a "real space"
`procedure (6, 7) was used to optimize the correlation between
`the Fab New model and the multiple isomorphous replace(cid:173)
`ment, electron density Fourier map.
`The coordinates obtained by these P'rocQdures have provided
`an improved model which has been used to compare the
`tertiary structure of the homology subunits, to calculate
`interatomic contacts that define the quaternary structure of
`Fab, and to re-examine the conformation of the combining
`site.
`
`METHOOS
`Measurement of Model Coordinates -Atomic coordinates were
`measured on the 2 A (nominal) resolut ion model previously de(cid:173)
`scribed (4). A two-pointer device was used for this purpose: a
`horizontal pointer (50 inches long) was brought to touch atom
`centers by displacing the device on the base of the model, adjusting
`the height of the pointer aloni;r a i;rraduated scale (z coordinate)
`while a second, parallel, fixed poin~r of equal length gave the x, y
`coordinates on a grid at the base of the model. The use of a level and
`leveling screws at the base of the two-pointer device was essential
`for obtaining reproducible coordinates. While these measurements
`were made, the image of the model and the atom centers were
`
`' The abbreviations used for immunoglobulins, their polypeptide
`chains, and fragments are as recommended in (1964) Bull. W H 0
`30, 447.
`
`585
`
`1 of 14
`
`Celltrion, Inc., Exhibit 1083
`
`

`

`Structural Refinement of Fab New
`
`586
`projected on the corresponding sections of the Fourier map using an
`optical comparator (8) to verify that their location and the coordinate
`values corresponded with the Fourier map.
`Model Building Procedure - The set of measured coordinates for
`the 3185 non-hydrogen atoms in the structure provided the starting
`point for this procedure. In general, these coordinates are subject to
`errors due to measurement uncertainty and to mechanical deforma(cid:173)
`tions of t he skeletal brass model. Consequently, the mathematical
`model building procedure of Diamond (5) programmed for a digital
`computer was used in order to impose standard bond lengths and
`bond angles in the model. The measured coordinates were used to
`provide a guide point for each (non-hydrogen) atom in the structure.
`Some conditions used in this part of the refinement process are given
`in Table I. All the varied angles are dihedral; the -rJ..r,-c..-<:) angle
`was allowed to vary since this condition gave a much closer correla(cid:173)
`tion with input coordinates without introducing large distortions in
`the idealized, modd-1..ouilt f:St:Omt:tq of the molecule. Residues for
`which the model-built coordinates differed considerably from the
`input coordinates or which had an abnormal r, value were remeas(cid:173)
`ured and checked for correspondence with the Fourier map. These
`discrepancies could always be traced to errors in the measurement
`of coordinates or to distortions of the brass model. When necessary
`the coordinates were measured again after rebuilding distorted
`regions and the remeasured coordinates were submitted to the model
`building procedure as guide points. The final average value of
`&- 0(&-0 = l-r0
`109.3°1) for the 440 residues in the model-built
`structure was 5.43°. Pro 151 in C.. l was built and refined in the cis
`conformation. The root mean square shift from the initial coordi(cid:173)
`nates for all non-hydrogen atoms in the structure was 0.2 A.
`Real Space Refinement-The model-built coordinates were used
`as a starting set for real space refinement (6). The 2.0 A electron
`density map used for the automatic fitting of the atomic coordinates
`was calculated using multiple isomorphous replacement phases as
`described before (3, 4). The electron density function was calculated
`at intervals of 1/160 along x, (a = lll.43 A), 1/80 along y (b = 56.68
`A), and 1/130 along z (c = 90.30 A) in sections of constant y . A
`compute-r program incorporating a fast-Fourier transform algorithm
`was used for this calculation. Five cycles of real space refinement
`were carried out using the conditions defined in Table II. Values of
`atomic radii for carbon, nitrogen, oxygen, and sulfur that gave
`fastest convergence in trials using a small part of the structure were
`adopted and kept constant during refinement. Progress in the
`refinement process was followed by inspection of the root mean
`square shifts in coordinates, shifts in the coordinates of individual
`atoms, and adjustments of amino acid scale factors. Unusually large
`shifts in coordinates were checked using the optical comparator and
`the Fourier map. All refinement calculations were carried out using
`computer programs implemented at the Brookhaven National Lab(cid:173)
`oratories.
`
`-
`
`RESUJ.: ns A~ll lJISl,;USSlON
`
`The root mean square shifts of atomic coordinates after five
`cycles of real space refinement converged to an average value
`of0.09 A. The values after each cycle were: 0.46 A, 0.20 A, 0.13
`A, 0.10 A, and 0.09 A. The average shift of all atoms during
`real space refinement was 0.62 A. In most parts of the
`molecule, the coordinates after refinement are in very good
`agreement with the features of the electron density map (Fig.
`
`TABLE I
`Conditions of model building refinement•
`Parameters varied were: <I>, o/J, x, T (N Ca C). Flexible proline
`residues were used. In addition, T(Ca Cfj Cy) was allowed to vary in
`Cys, His, Phe, Trp, and Tyr residues. A list of the sources of the
`amino acid groups used in model building is given in Table I of
`Diamond (7).
`
`TABLE 11
`Conditions of real space refinement"
`5
`Zone length:
`6
`Margin width:
`Fixed atomic radii: 1.4 A
`Relative atomic weights; C:6, 0:7, N :8, 8:16
`Relative softne-ss of angular parameters that were allowed to vary:
`</>.iii: 4_0
`x: 3.2
`Filter levels:
`
`0.01
`0_0001
`Scale factor and background
`0.01
`0.001
`Translational refinement
`0.001
`0.001
`Rotat ional refinement•
`Electron density map grid; 111.43/160, 56.68/80, 90.30/130 along cell
`edgl!s
`• See Diamond (6, 7 ) for definition of terms.
`• Nonlinear constraints were used to preserve chain continuity.
`
`1). Many carbonyl groups of the main polypeptide chain can
`be reliably positioned (Fig. 2). In general, coordinates of atoms
`in regions of strong, well defined electron density converged
`rapidly in the first cycle of refinement and moved very little
`in subsequent cycles. Atoms in regions of lower density
`converged more slowly. A few residues in poorly defined
`regions of low density showed little convergence, although
`their total shift from the starling coordinates was small. The
`movement of main chain atoms tended to be smaller than
`those of side chains, presumably due to their better defined
`electron density and to greater constraints on their positions.
`The progress of refinement was checked after each cycle by
`inspect ing the fit of atoms whose shifts were substantially
`greater than the average. Some side chain groups which had
`been shifted by the Diamond model building procedure were
`moved back to their original positions by real space refine(cid:173)
`ment. In the fifth cycle of refinement, an average shift greater
`than 1 A occurred for three consecutive amino acid residues
`(Gly 166, Val 167, and His 168) in the C"l region. Inspection
`of the position of these atoms showed that they had moved to
`a conformation that appears to be in better agreement with
`the electron density map than the original model. The coordi(cid:173)
`nates for all atoms of Fab New after relinement, given in the
`"Appendix," are filed with the Protein Data Bank at Brook(cid:173)
`haven National Laboratory. No major features of the map
`rem<iin unexplained, although a !'lUIDbcr of pQS$iblc solvent
`molecules are found on the surface of the molecule. The
`conventional R factor, 2 based on Fe obtained with the coordi(cid:173)
`nates i n "Appendix" and an overall temperature factor (B =
`18.0), is R = 0.46. This value is reasonable given the refine(cid:173)
`ment approximations outlined in Table 11 and the fact that no
`solvent atoms were included in the model. Further refinement
`using observed structure factors and calculated phases is
`currently underway.
`The S- S distances in the five Fab disulfide bonds were
`allowed to vary without constraints. At the end of the refine(cid:173)
`ment these distances were found to be: V "' 2.00 A; V 1,.1 2.46 A;
`CHI, 2.30 A; c ... , 2.30 A; C11l -C1,., 2.43 A.
`Ramachandran Plots
`The Ramachandran plots of the V 1,. and C1,. homology sub-
`
`- Fe I / IF0 , where F0 is observed structure amplitude
`l I F0
`' R =
`and Fe is calculated structure amplitude.
`3 The isotropic temperature factor (8) used in the expression exp
`{-8 sin 2 91>- 2) .
`
`Probe
`
`Len£h
`(resi ues)
`
`0.1
`1
`0.1
`2
`2
`0.1
`3
`3
`• See Diamond (5) for definition of terms.
`
`C,
`
`Filter constants
`c.
`lo-•
`10- •
`10 .....
`
`2 of 14
`
`Celltrion, Inc., Exhibit 1083
`
`

`

`Structural R efinement of Fab New
`
`587
`
`F1G. 1 (upper) . Stereo view of some main chain and side chain
`groups of Fab New after five cycles of real space refinement super(cid:173)
`imposed on the corresponding electron density of the multiple
`isomorphous replacement 2 A resolution map.
`
`FIG. 2 (lower). Stereo view of some amino acid residues of Fab
`New superimposed on the corresponding density of the multiple
`isomorphous replacement map. Carbonyl groups of the main poly(cid:173)
`peptide chain are clearly seen.
`
`360 r-~~-.~.~,~.--.-~ ....... -, ~~--.~~~~~~~ ...... ~~ ......
`..
`•• • ., ..
`<t-
`-~~ ... _ o:
`•": •
`I
`•
`I
`• •• •
`>'
`
`. .. , ... :
`
`/
`
`•
`
`I
`I
`I
`I
`1 _____ _ _ 1
`I •
`
`/ 1
`I
`f •
`I
`I
`I
`,
`I
`I
`,_,
`(
`I
`
`0
`
`0
`
`/J .
`lo
`/
`I
`[
`I
`I
`'l
`I
`I
`I
`' ..... 1
`
`.
`
`. . . . \
`·: .
`. .
`.. ~.,t:
`• •.•.
`•
`I
`-4\~ • •• • I
`• •'1~
`I
`. ,
`••
`'
`• 1 e
`.
`'-------··
`/---.-,
`/ . .,
`_,,,, .
`'
`'• .,
`
`,,,,, - •
`
`• •
`0 ·~ /
`f
`I
`I
`I
`I
`
`•
`
`0
`
`f
`
`~
`
`360
`
`300
`
`240
`
`180
`
`120
`
`60
`
`'
`
`•
`
`o ~-----~---_-~ ..... ~--·~---~'"..-~~-"~~~ ...... ~~~ ..... ~~ ........
`
`0
`
`60
`
`120
`
`180
`
`240
`
`300
`
`360
`
`300
`
`2.40
`
`180
`
`'
`
`, ----
`
`/
`-.._ __
`
`/
`
`••
`
`I
`I
`I
`
`. .
`
`120
`
`I
`- - - - - - - - _ _ J
`
`60
`
`.
`
`- - - - - - , ,
`
`0 -
`
`- .- -
`
`0
`
`60
`
`120
`
`180
`
`240
`
`300
`
`360
`
`FJC. 3. Ramachandran plots of the VL and Q homology subunits ofFab New. The distribution of<f:> , o/J angles indicates the predominant
`antiparallel P-pleated sheet structure of the subunits. Glycine residues are indicated by 0 .
`
`units (see Fig. 3 ) show a distribution of <f>, l/J angles which
`corresponds with the predominant antiparallel ,8-pleated sheet
`structure present in V H• V L• CHl, and c ... As observed in
`sever a l other protem structures, the </>, tfJ angles for glycine
`
`residues are scattered in the plot, frequently appearing in
`nonallowed regions for L-amino acids. Several other residues
`which occur outside areas of allowed conformation are found
`in bends of the polypeptide chains; it is expected that further
`
`3 of 14
`
`Celltrion, Inc., Exhibit 1083
`
`

`

`588
`
`Structural Refinement of Fab New
`
`refinement will improve the angular values observed for these
`residues.
`
`Structural Comparison of Homology Subunits
`The structural homology of V H• VL, C8 1, and CL has been
`described before (3, 4). A quantitative analysis of this homol(cid:173)
`ogy using the method of Rao and Rossmann (9) is presented
`here. A $imil!lr !ln'1lysis h!l$ been m!lde by Rich<irdson et r;rl.
`(10) comparing the structures of superoxide dismut ase and
`the murine Fab McPC 603 fragment.
`Initial matrices relating the Ca coordinates of the homology
`subunits were obtained from a small number of structurally
`equivalent amino acids. The num.ber of equivalences was
`then extended by an automatic search for stretches of poly(cid:173)
`peptide chain for which the distances between putatively
`equivalent Cas was smaller than 3.8 A. Based on the extended
`eq uivalences new matrices were calculated and the process
`was iterated until no changes in equivalences were observed.
`A summary of the results is presented in Table Ill which lists
`the number of Cas occurring at distances of less than 1.5 A
`and less than 3.0 A, for all the six possible pairings of
`subunits which were superimposed and compared by this
`process. The average value of the minimum base change
`necessary to exchange the codons of the structurally equiva(cid:173)
`lent amino acids is also given in Table lll.
`As can be seen in Table Ill, there is an even closer structural
`homology between VH, VL, C81, and CL in Fab New than that
`observed for McPC 603 Fab (10), probably reflecting the
`higher resolution of the Fab New model. Presumably the Ca
`distances given in Table IIJ could become smaller with further
`crystallographic refinement. The number of Cas which super(cid:173)
`impose with distances shorter than 1.5 A and 3.0 A is larger
`wh en comparing V H to V L and CHl to CL. Also, there is good
`(inverse! correlation between the number of Cas t hat are
`structurally equivalent and the average minimum base
`change per codon. Furthermore, when a restrictive condition
`for structural equivalence is imposed ( d c"-<.A 5 1.5 A) the
`average base change per codon becomes smaller, reflecting a
`higher degree of conservation of amino acid sequences.
`It should be emphasized here that amino acid sequence
`information is not used in the quantitative three-dimensional
`alignment procedure described above. However, this proce·
`dure leads to amino acid sequence alignments that clearly
`reflect the well established homologies between the V8 and
`Vi.., and between the Cii and Ct regions of immunoglobulins
`(see Figs. 4 and 5). The closest sequence similarity in Fab
`New occurs between ~l(y) and Ct(X), although, as shown in
`Table III, the structural similarity between V8 and v .. is close
`to that between ~ 1 and Ci. (see Figs. 6, 7, and 8). In addition,
`
`TABLE Ill
`Alignment of a-carbon coordinates of four homology subu nics of Fab
`(New) using m ethod of Rao and Rossmann (9)
`Number of
`Average
`Number of
`Av erage
`c. pairs
`c. pairs
`minimum
`minimum
`equiva·
`base change
`base change
`equiva·
`lenced with per codon for
`lenced with per codon for
`deo.,,.A s. 3. o dc.-eoAs. 3.o
`d c.-ca{ ' I. 5 d c.-eoA'5. 1. 5
`
`Subunits
`
`VwV1.
`C11i-CL
`c .. -v ..
`c .. -vH
`CHI-VI.
`c .. 1-v11
`
`56
`60
`40
`29
`27
`25
`
`0.98
`0.71
`1.03
`1.04
`1.04
`J.29
`
`81
`82
`66
`59
`58
`49
`
`0.97
`0.80
`1.23
`1.28
`1.24
`1.40
`
`Table III shows that there is considerable homology between
`the V and C regions. These results can be interpreted to
`indicate that all the homology regions contain a basic core of
`amino acid residues with highly preserved three-dimensional
`structure. The chemical nature of these residues is also
`preserved as indicated by the correspondingly lower values of
`the average base change per codon. As stated before (3) these
`findings strongly support the postulate (15) of a gene duplica(cid:173)
`tion m echa11i:;rn whida gave ri:;e w thi:! different humulugy
`regions of immunoglobulins.
`The Cas of the homologous sequences, -Phe-Gly-Gly-Gly(cid:173)
`(99 t.o 102) in Vi, and -Trp-Gly-Gln-Gly- (107 t.o 110) in VH, can
`be closely superimposed as can Ca atoms immediately preced(cid:173)
`ing and following those residues. This conserved conformation
`gives no evidence supporting the postulate (16) that the
`glycine residues could serve as a pivot to allow for optimal
`contacts between an antibody and its ligands. An alternative
`explanation for these constant, homologous V11 and Vi, se(cid:173)
`quences has been proposed (4) in terms ofintersubunit (Vu to
`V1, , see below) and intrasubunit contacts.
`The limits between the V and C homology regions can be
`defined from the model. The sequence -Val-Ser-Ser- (115 to
`117, Fig. 4) which is shared by y and µ.human H chains marks
`the COOH terminus of Vu, and following a sharp bend in the
`polypeptide chain the sequence -Ala-Ser-Thr (118 to 120)
`marks the NH2 terminus of~l. The sequence -Thr-Val-Leu(cid:173)
`(106 t.o 108) corresponds to the COOH terminus of Vi.. and the
`residues -Gln-Pro-Lys- (110 to 112) constitute the NH2 termi(cid:173)
`nus of Ct. Thus, in the three-dimensional model Arg 109
`(usually assigned to VL) could be considered either as the end
`of VL or as the beginning residue of Ct. By the structural
`alignment described here however, Arg 109 can be properly
`considered as the COOH terminus of Vi, .
`In agreement with the Gm(4-) serological specificity oflgG
`New (17), a lysine residue is placed at position 214 in Cu 1.
`This residue, corresponding to the Gm(l 7) allotype provides a
`better fit with the Fourier map than an arginine residue
`(which correlates with Gm(4)) at that position. This interpre(cid:173)
`tation is reinforcR.d hy thP. Gm(l +) ."-P.rolngic.sil spe.r.ifir.it.y of
`IgG New (17) which has been verified by amino acid sequence
`analysis (12).
`
`Quaternary Structure
`Contacts between Homology Subunits -The closer contacts
`between the homology subunits of Fab New are diagrammat(cid:173)
`ically represented in Fig. 9 by lines joining Ca atoms sepa(cid:173)
`rated by a distance of 8 A or less. This figure provides a
`description of regions of VH, Vi., ~ 1, and C.. in which there
`are higher density of contacts. Inspection of Figs. 8 and 9
`indicates that the interactions between Vu and VL and be(cid:173)
`tween ~ 1 and C.. are more extensive than those between Vu
`and C,,, 1 and those between Vi.. and Q. The fact that the V11
`and Ci.i 1 subunits (whose major axes make an angle smaller
`than 90°) interact more extensively than V1, and Ct (whose
`major axes makes an angle larger than 90°) is also reflected in
`Figs. 8 and 9.
`Intersubunit contacts between side chain and main chain
`atom~ situated at a distance not larger than 1.2 times their
`van der Waals radii are given in Table IV. This table lists
`contacting residues and the number of close contacts that
`atoms from a given residue make with atoms of other residues.
`Evidently, amino acids with larger side chains have a poten(cid:173)
`tial to make more contacts with other amino acids, thus for
`example, VH Trp 107 makes 29 intersubunit contacts, Trp 47
`
`4 of 14
`
`Celltrion, Inc., Exhibit 1083
`
`

`

`Structural Refinement of Fab New
`
`589
`
`•o
`30
`z7abc
`20
`10
`~o
`----SSVL~OPtSVSGAt•G0kVTI8CTGSSSMIGAGRBVKWYQQL•GrAPI-~LlP•••Aa••••
`
`120
`110
`lSO
`1'0
`UO
`Q1KAAPSVTLPPPSS&ILQAIEATLVCLISDPYPGAV•TVAIK••AD&S•••••••••••••••••
`
`120
`160
`150
`100
`130
`ASTKGPSV7PLAPSSKSTSGGTAALGCLV~OYPPBPV-TVSW•---SG-•••--------------
`
`70
`lot
`100
`90
`80
`61
`·---rsvsssG----------SSATLAITGLQAIDllDYYCQSYOlSLt••·VPGGGTJLTVLl
`
`110
`70
`117
`100
`90
`80
`TPLRSRYTMLVWT-s-------&•QP SL•LSSVT AADTAVYICARMLIAG•CI DV WGQG SLVTVS a
`
`160
`
`170
`
`180
`
`190
`
`200
`
`210
`
`11•
`
`220
`210
`190
`180
`170
`200
`•ALTS••GV8TPPAVLQSSGLY SL88VVTVP888LGT•QT YICIVIBKPllTK•V DKJ•V IPllC
`Fie. 4. Amino acid sequences of the V,, , C.,, V,,, and C,, 1 homology regions of Fab New aligned by comparison of their three-dimensional
`structures. - - - indicate gaps introduced to maximize alignment of the three-dimensional structures. * indicate deletions in the VL
`sequence. See Ref. 11 for the VL and C.. sequences, Ref. 12 for VH , and Ref. 13 for C,.l. Abbreviations for amino acids are as given in Ref. 14.
`
`FIG. 5. Diagram of hydrogen bond(cid:173)
`ing (broken lines) between main chain
`atoms for the VL, C.,, V,,, and C.. l
`homology regions of Fab New. The
`hydrogen-bonded clusters correspond
`to the two ,S-sheet structures of each
`subunit. Cysteine residues that partic(cid:173)
`ipate in intrachain and interchain di(cid:173)
`sulfide bonds are underlined.
`
`makes 28 contacts and Arg 43 makes 24 contacts.
`The contacts between Vu and Vi. are of particular interest
`in view of the fact that different H and L immunoglobulin
`chains can form structurally viable pairs. Three types of Vu -
`
`V1, contacts will be considered in this discussion: first, the
`contacts which are at the core of the contacting region, made
`by residues which are invariant or semi-invariant in Vu and
`Vt. sequences; second, the contacts made by invariant or semi-
`
`5 of 14
`
`Celltrion, Inc., Exhibit 1083
`
`

`

`590
`
`Structural Refinement of Fab New
`
`Flo. 6. Stereo pair drawings of the
`a carbon backbones of the VL ( wp) and
`VH (botwm) subunits. The subunits are
`viewed here in similar orientations.
`
`F10. 7. Stereo pair drawing of the a
`carbon backbones of the Q (top) and
`C,, 1 (bot.I-Om) subunits viewed in simi(cid:173)
`lar orientations.
`
`6 of 14
`
`Celltrion, Inc., Exhibit 1083
`
`

`

`Structural Refinement of Fab New
`
`591
`
`Fie. 8. Stereo pair drawing of the a
`carbon backbone of Fab New.
`
`invariant residues with hypervariable residues; and finnlly,
`those made between hypervariable residues.
`The core of the V11 -VL contacting region can be described as
`determined by residues Val 37, Gin 39, Leu 45, Tyr 94, and
`Trp 107 in VH and by residues Tyr 35, Gin 37, Ala 42, Pro 43,
`Tyr 86, and Phe 99 in V1• • These residues are structurally
`homologous with the exception that v .. Ala 42 has no clear
`correspondence in V11 due to a structural "insertion" (see Fig.
`4). These homologous V11 and V1. residues make numerous
`contacts with each other (about 50% of those listed in Table
`IV) or with other, nonhypervariable residues. The rings ofTrp
`107 (V11 ) and Pro 43 (Vi), at the center of the VL -V11 contacting
`region, are nearly parallel and stacked on each other. The
`contact residues listed above are invariant or are replaced by
`homologous residues in VL (/< and X) and V11 sequences from
`different animal species. For example, Tyr 35, Gin 37, Pro 43,
`and Phe 99 appear constant in human L chains (K or>..), and
`Gin 39, Tyr 94 (replaced by Phe in a very few cases), and Trp
`107 (replaced by Phe or Tyr in a very few cases) appear nearly
`constant in human H chains. Ala 42, Tyr 86 in VL and Val 37
`in V11 are more frequently replaced by homologous residues:
`Ser 42, Phe 86, and Ile 37. The invariant or nearly invariant
`nature of these residues of the main VwVL contacting area
`provides a structural basis (together with interactions be(cid:173)
`tween Ci 1 and C.., see below) for the property of different H
`and L chains to recombine into new immunoglobulin mole(cid:173)
`cules (see References 18, 19, and in particular 20, for a recent
`review and experimental data on this topic).
`A second type of contact listed ln Table IV is mad~ ~tw~tm
`constant or nonhypervariable residues and hypervariable res·
`idues. For example, the side chain atoms of V,. Trp 47. a
`constant residue in human, mouse, guinea pig, and in most
`rabbit immunoglobulin sequences, make close contacts with
`Ser 93, Leu 94, and Arg 95 in the third hypervariable region of
`V1,. However, a large number of these contacts involve the
`peptide chain atoms of the V,, residues. Replacements in the
`V1, side chains will not necessarily a lter the nature of these
`contacts. Similar contacts appear to be made by Vi. Leu 45
`(invariant or semi-invariant in human L chains) with the
`peptide chain at VH hypervariable position 104. Contacts of
`this type could also be made from V,, Tyr 35 to the peptide
`chain atoms of the fourth hypervariable region of Vu in chains
`of different length than Vu New.
`The third type of contact to be discussed here is that made
`between hypervariable residues, such as those made between
`V11 Asn 98 and V,, Arg 95. These contacts are more difficult to
`evaluate in general terms (a) because the location of some of
`the residues involved might be changed by further refinement
`to a larger extent than those of most other residues in the
`sequence, and (b) because it is possible that in other immu(cid:173)
`noglobulins, replacements by different amino acid side chains
`
`nt these positions could be accommodated by small displac.
`ments of the hypervariable peptide loops. These "idiotypic"
`interactions are consequently more difficult to assess. How(cid:173)
`ever, they could perhaps explain the preferred reassociation
`observed between complementary H and L chains derived
`from a single immunoglobulin molecule (20). Most of the
`contacts discussed above consist of van der Waals interactions
`between hydrophobic side chains. However, a few hydrogen
`bonds can be indicated: V11 Gin 39 to V,. Gin 37, and V,, Asn 98
`to VL Tyr 90 and/or V.. Arg 95. Also, an ion pair is formed
`between V11 Arg 43 and VL Asp 84.
`In the Fab New model the contacts between V" and V1• are
`very close (Table IV}, giving rise to a compact dimer. No
`haptens or even solvent molecules can be accommodated
`between V11 and VL beyond the combining site, a situation
`which is different from that described for an L-chain dimer
`(21).
`As shown in Table IV the interactions between C,, 1 and C,,
`are extensive. The core of the contact area between C,11 and
`C.. is defined by ~ 1 residues Leu 128, Ala 129, Gly 143, and
`Leu 145 and the structurally homologous C.. residues Phe 120,
`Pro 121, Val 135, and Leu 137. These residues appear to be
`invariant or nearly invariant in the H and L sequences from
`different animal species. Most of the other contact residues
`such as ~1: Phe 126, Pro 127, Thr 139, Lys 147, Phe 170, Pro
`171, Val 173, Gin 175, Ser 181, Val 185, Lys 218, and C,,: Thr
`118, Ser 123, Glu 125, Glu 126, Lys 131, Thr 133, Thr 164, Ser
`177, Tyr 179, Lys 206, are also invariant or replaced by
`humvlugvus re:si<lues in ihe inID1w1oglobulin chains from dif·
`ferent animal species. In the contact area the central location
`of C,, 1 Leu 128 and C,_ Phe 120 is reflected in the large number
`of contacts (20 contacts) they make with each other and with
`many other residues (see Table IV). As pointed out by Novotny
`and Franek (22) the amino acid sequence of the four-stranded
`,8-pleated sheet is more conserved than the rest of the C\
`regions in different animal species, leading to a dendrogram
`(or genealogic tree) of distorted evolutionary distances. This
`observation can be analyred in terms of the structural model
`presented here as follows. The four-stranded ~ sheets of C,11
`and C,, contain side chains which make intrasubunit contacts
`and in particular, they contain all or nearly all of the contact
`residues between C.il and C.. (discussed above). Evjdently,
`mutational events leading to amino acid replacements at
`t.hese positions would have to occur in a complementary
`pattern in both C., 1 and C,, in order to preserve tertiary and
`quaternary immunoglobulin structure, and consequently they
`would be expected to occur at a slower rate than mutations in
`other regions of~ 1 and C,,.
`As can be seen in Table IV the region immediately preced(cid:173)
`ing the interchain disulfide bond does not provide close con(cid:173)
`tacts between 4 11 and C. .. In addition, the two strands of
`
`7 of 14
`
`Celltrion, Inc., Exhibit 1083
`
`

`

`592
`
`VL
`
`Structural Refinement of Fab New
`
`VH
`
`CL
`
`116
`117
`118
`119
`
`120
`121
`122
`123
`124
`
`125
`128
`
`182
`163
`114
`185
`
`198
`117
`
`175
`171
`1n
`
`212
`213
`214
`
`44
`
`45 ...
`47 ...
`
`49
`
`51
`
`110
`
`CH1
`
`CL
`
`187
`
`188
`
`1119
`
`171
`
`1n
`
`161
`
`162
`
`163
`
`1.•
`
`1.0
`
`101
`
`102
`
`VH
`9
`
`10
`
`11
`
`112
`
`113
`
`CH1
`
`128
`127
`128
`129
`130
`131
`132
`133
`
`140
`141
`142
`1'3
`
`170
`171
`
`219
`:no
`
`VL
`
`38
`
`JI
`
`40
`
`79
`
`90
`
`12
`
`114
`FIG. 9. Intersubunit a carbon contacts at distances of 8 A or less. Contacts are indicated by lines j oining the corresponding amino acid
`residue numbers. Numbers on the lines indicate the contact distance (in Angstroms). Note the extensive V1• -VH and C..-C.1 I interactions.
`
`polypeptide chain that come together at the interchain disul(cid:173)
`fide bond do not closely int.eract with the rest of C... l or Ci ..
`This region can be described as having a loose conformation,
`with a lower electron density in the Fourier map. These
`structural features are in agreement with the notion of seg-
`
`mental flexibility residing around this part of the immuno(cid:173)
`globulin structure and in the immediat.ely adjacent hinge
`region of the H chain.
`Hyperuariable Regions, Idiotypes, and Combining Site(cid:173)
`The results of several experimental approaches (see Chapter II
`
`8 of 14
`
`Celltrion, Inc., Exhibit 1083
`
`

`

`Structural Refinement of Fab New
`
`593
`
`in Ref. 23 for a comprehensive review) strongly suggested that
`hypervariable residues in the amino acid sequences of H and
`L chains, idiotypic deter