`Vol. 76, No. 8, pp. 4071-4074, August 1979
`Immunology
`
`Crystal structure of galactan-binding mouse immunoglobulin J539
`Fab at 4.5-A resolution
`(x-ray diffraction/pattern recognition/antibody structure)
`
`MANUEL A. NAVIA*, DAVID M. SEGAL*t, EDUARDO A. PADLAN*l, DAVID R. DAVIES*, NARAYANA RAO§,
`STUART RUDIKOFF§, AND MICHAEL POTTER§
`
`"Laboratory of Molecular Biology, National Institute of Arthritis, Metabolism and Digestive Diseases; and §Laboratory of Cell Biology, National Cancer Institute,
`National Institutes of Health, Bethesda, Maryland 20005
`
`Contributed by David R. Davies, May 11, 1979
`
`An electron-density map of the mouse galac(cid:173)
`ABSTRACT
`tan-binding immunoglobulin 1539 (lgA2,K) Fab has been cal(cid:173)
`culated to a resolution of 4.5
`by the method of heavy atom
`isomorphous replacement with four derivatives. The map has
`been interpreted with the aid of a computer program which
`systematically searched for the best fit between the electron(cid:173)
`density map and the known coordinates of individual immu(cid:173)
`noglobulin domains. The quaternary structure of }539 Fab at
`this resolution appears similar to that of another mouse immu(cid:173)
`noglobulin, lgA2,K Fab, McPC603. The model coordinates for
`}539 Fab should allow us to proceed directly to a high-resolution
`structure determination without further heavy atom isomor(cid:173)
`phous replacement.
`
`The crystal structures of three immunoglobulin Fabs have been
`reported (l-3). One of these, the mouse myeloma protein
`McPC603, belongs to a class of immunoglobulins that bind
`specifically to phosphocholine. In this paper, we present the
`results of.a structural analysis at 4.5-A resolution of a second
`mouse Fab, the galactan-binding immunoglobulin ]539. A
`molecular model has been constructed with the aid of a com(cid:173)
`puter search procedure used previously in the interpretation
`of the electron-density map of an intact immunoglobulin (4).
`The model allows us to describe the quaternary structural in(cid:173)
`teractions between the constituent domains and domain pairs
`of the ]539 Fab and reveals a general similarity to
`McPC603.
`
`MATERIALS AND METHODS
`The immunoglobulin ]539 [a mouse lgA2,K; anti-(l-+6)(cid:173)
`,8-D-galactan] was isolated and purified as described (5), and
`pepsin fragments were obtained by using a published procedure
`(6).
`Crystals of ]539 Fab were grown at room temperature from
`35% saturated solutions of ammonium sulfate containing 0.07
`M imidazole and 0.03 M zinc sulfate at pH 6.8. The crystals
`exhibit the symmetry of space group P212121 with unit cell
`dimensions: a = 54.l A, b = 74.2, A, and c = 130.8 A. They
`diffract to a resolution of at least 2.5 A. Prior to further work,
`the crystals were transferred into stabilizing solution (45%
`saturated ammonium sulfate/0.07 M imidazole/0.03 M zinc
`sulfate, pH 6.8). On rare occasions, a second crystal form was
`observed in the space group P212121 with unit cell dimensions:
`a = 55. 7 A, b = 7 4.1 A, and c = 125. 7 A. Cocrystallization in
`the presence of the haptens (l-+6)-,8-D-galactodiose (Gal2) and
`(l-+6)-,8-D-galactotetraose (Gal4), both kindly provided by
`C.P.J. Glaudemans, yielded a needle-like crystal form which
`remains uncharacterized.
`
`0.500
`
`\ ._, ·c
`
`u :0.000
`.. -·.
`f./
`.-r v
`
`.::>
`
`0.500
`
`'
`
`u: 0.500
`
`)
`
`(•
`
`0
`
`w
`I
`
`w
`
`B
`,_.
`
`-
`
`"~
`: ,_
`
`1
`
`FIG. 1. Sections through the difference Patterson map for the
`K2Pt(CNS)6 derivative, showing the Harker peaks ·(A, B, and C).
`Contours are at 1/40ths of origin peak; zero and negative contours are
`omitted.
`
`Useful heavy atom derivatives were obtained by soaking ]539
`Fab crystals in stabilizing solutions with 0.25 mM K2Pt(CNS)s
`for 2 weeks, 1 mM K2PtCl4 for l week, 0.3 mM K2Hgl4 for 2
`weeks, and 50 mM KI/ l mM chloramine-T (Eastman Kodak)
`for 1 week. The iodine-derivatized crystals were washed with
`fresh stabilizing solution after the soaking ~riod. Unique sets
`of three-dimensional reflection data to 4.5-A spacings were
`collected from native and derivative crystals with a Picker
`FA CS-1 diffractometer operated in the w-scan mode and using
`Cu Ka radiation. The various data sets were placed on the same
`relative scale, and pseudo temperature factors were applied to
`correct for differences in intensity fall-off. The KzPt(CNS)s
`derivative was analyzed first, and difference Patterson maps,
`readily interpretable, revealed one site of substitution (Fig. 1).
`After preliminary refinement with projection data only, the
`coordinates and occupancy of the K2Pt(CNS)s site were used
`to determine the signs of the centric native protein reflections.
`These signs were used to examine the K2PtCl4 and K2Hgl4
`derivatives by difference Fourier analysis. The variables for
`these three heavy atom derivatives were then subjected to al(cid:173)
`ternating cycles of least squares refinement and phase deter(cid:173)
`mination with the three-dimensional data. The resulting phases
`
`The publication costs of this article were defrayed in part by page
`charge payment. This article must therefore be hereby marked "ad(cid:173)
`vertisement" in accordance with 18 U. S. C. §1734 solely to indicate
`this fact.
`
`t Present address: Immunology Branch, National Cancer Institute.
`t Present address: Biophysics Department, Johns Hopkins University,
`Baltimore, MD.
`
`4071
`
`1 of 4
`
`Celltrion, Inc., Exhibit 1084
`
`
`
`'tVI~
`
`uu1uuuu1u~y: nav1a ta ai.
`
`rroc. I'llau. iicaa . .)et. lJ.)f\ ro \l:JHJJ
`
`Table 1. Heavy atom variables after refinement
`Isotropic
`temperature
`factor,
`J...2
`
`Rel.
`Site occupancy•
`
`Derivative
`
`K2Pt(CNS)s
`K2PtCl4
`
`K2Hgl4
`Iodine
`
`Fractional coordinates
`x
`z
`y
`22.1
`0.0437
`0.2320
`0.0624
`1.00
`1
`0.51
`1
`0.0477
`0.2304
`0.0627
`18.8
`0.51
`2
`25.1
`0.0333
`0.6H3
`0.3444
`0.0550
`0.6346
`0.62
`1
`0.3477
`25.6
`25.6
`0.4514
`0.5425
`0.5867
`0.62
`1
`2
`13.0
`0.8920
`0.6349
`0.4008
`0.96
`0.32
`22.2
`0.1535
`0.2927
`0.3469
`3
`• The site occupancies are on an arbitrary scale with the K2Pt(CNS)6 site assigned an occupancy of 1.00.
`t Re= :Zl IFPH - Fpl - FHll~IFPH - Fpl, in which the sums are over centric reflections only.
`I Ri.. = :Zl IFP.HI -
`IFP + FHI ll~IFPHI·
`§ E is the root mean square lack of closure error. The average protein structure amplitude is 10.3.
`
`R.t
`
`0.507
`0.590
`
`Q.606
`0.587
`
`R1rl
`
`0.068
`0.068
`
`0.070
`0.116
`
`Et
`
`0.95
`0.89
`
`0.92
`1.62
`
`were subsequently used to analyze the iodine derivative. The
`refined variables of the four heavy atom derivatives are listed
`in Table 1. The mean figure-of-merit for the 1895 phased re(cid:173)
`flections is 0.82. "Best" (7) Fourier maps were computed by
`using these phases.
`Two methods w~re used to interpret this map.
`(i) Visual examination of the map, making use of the known
`structure of McPC603. In addition, because K2Pt(CNS)6 bound
`also to McPC603 where it was attached to the constant region
`of the molecule (8) and because the constant regions of
`McPC603 and J539 Fab are most probably identical, it was
`assumed that Pt( CNS)~- bound to the same site in both mole(cid:173)
`cules, thus providing a known location in J539 from which to
`start the interpretation.
`(ii) The overall three-dimensional structures of the domains
`of McPC603 were assumed to be similar to those of J539. A
`computer program was accordingly used to rotate and translate
`exhaustively the known coordinates of the domains of McPC603
`Fab (9) into the electron-density map of J539; after each such
`transformation, they were compared with the density. The
`criterion of fit was the number of overlaps of transformed
`
`280
`
`260
`
`240
`
`220
`
`8.200
`"' -.::
`~
`0 180
`
`100.._~..__~..__~_._~...._ ..... ......_~_._~_._~_._~..,.....
`0
`40
`80 120 160 200 240 280 320 360
`·increments, degree~
`
`FIG. 2. Overlap of domain pair CHl-CL on the J539 4.5-A elec(cid:173)
`tron-density map as a function of the single Eulerian angle 81. The
`angle was systematically varied in 1° intervals over its full range with
`all other rotation and translation variables fixed at their maximal
`overlap values.
`
`coordinates on the electron density. This program was originally
`used·in the interpretation of the electron-density map of the
`intact immunoglobulin "Doh" (4) and will be described in detail
`elsewhere. The translations of the domains were conducted
`along the three orthogonal axes with increments of 2 A at 6.(}.A
`resolution and of 1.5 A at 4.5-A resolution. The rotations were
`eomputed by varying the three Eulerian angles with increments
`of 30° at 6.(}.A resolution and of 15° at 4.5-A resolution (these
`increments were decreased as the final fit solution was ap(cid:173)
`proached). To further limit the scope of the search, the coor(cid:173)
`dinates of McPC603 heavy chain residue 148, which in McPC
`603 was near the K2Pt(CNS)6 binding $ite, was tethered to
`within.20 A of the coordinates of .the platinum binding site in
`the J539 electron-density map, again on the assumption that
`both molecules have the same Pt binding site.
`For the initial search at 6-A resolution, the constant region
`domain pair (CHl-CL) of McPC603 Fab was used as a unit.
`Afterwards, the variable region domain pair (V H-V L) was used
`as a unit, with the hypervariable r~dues (10) for the light chain,
`26-34, 50-55, and 92-96, and for the heavy c~ain, 31-35,
`51-59, and 97-100, omitted because they could be expected
`to differ between McPC603 and J539. At this stage of the search,
`only a-carbon coordinates were used in order to minimize
`computer time and the amount of memory used. All search.es
`were conducted on a PDP 11/70 computer which has an ad-
`dressing limit of 32,000 words.
`·
`The 6-A resolution results were refined by searches of the
`4.5-A resolution J539 map. This time, however~ the individual
`McPC603 domains (CHl. CL. VH, VL) were fitted separately.
`An idealized (11) polyalanine structure, generated from the
`McPC603 backbone coordina~es and again excluding hyper(cid:173)
`variable residues, was used in the search. Typical results for the
`two searches are presented in Table 2 and Fig. 2. An indication
`of the precision of the fit obtained by this procedure can be
`
`Table 2. Comparison of search results using the CHI- CL domain
`pair at two resolutions
`
`Peak height
`above mean,
`SD
`
`Over la~,• no.
`Resolution, No. of atoms
`A.
`used in search Maximum Mean
`3.9
`46
`93
`208
`6
`4.8
`305
`503
`1020
`4.5
`The Euler angle 81 was systematically varied about the final
`CHl-CL domain-pair solution; all other rotation/translation variables
`were kept fixed.
`• "Overlaps" refer to the number of coordinates found to superimpose
`on electron density above a specified level (see text).
`
`2 of 4
`
`Celltrion, Inc., Exhibit 1084
`
`
`
`Immunology: Navia et al.
`
`Proc. Natl. Acad. Sci. USA 76 (1979)
`
`4073
`
`FIG. 3. Sections through the 4.5-A electron-density map of J539 perpendicular to the Y (74.2 A) axis, with model polyalanine coordinates
`superimposed. The sections shown are from y = 0.06 toy = 0.16. Solid and open circles, heavy and light chains, respectively; arrows, ends of
`the independently fitted VH and CHI domains that have to be joined at the switch region.
`
`obtained from the Euler angle solutions presented in Table 3.
`Another estimate of the precision of fit might be obtained from
`the width of the curve in Fig. 2. A quantitative estimate is dif(cid:173)
`ficult, however, because the error cannot be demonstrated to
`have a normal distribution but seems to depend rather on the
`form and pseudosymmetry of the fitted molecule. The solution
`peak for the 6-A search was broader than that for the 4.5-A
`search, thus justifying the use of a coarser and more economical
`sampling scheme at the lower resolution.
`
`RESULTS AND DISCUSSION
`Examination of more than 50 heavy atom compounds yielded
`only the four usable derivatives of Table l. For reasons that
`were not apparent, these derivatives were of significantly poorer
`quality than those used for the comparable McPC603 analysis
`(8), and their use led to an electron-density map in which the
`contrast between protein and solvent regions was low, so that
`the molecular boundary was not immediately apparent. An
`attempt to interpret this map visually based on the positions of
`the heavy atom derivatives, particularly K2Pt(CNS)6. and the
`known structure of McPC603 Fab resulted in a tentative de(cid:173)
`lineation of the molecular boundary. However, interpretation
`of the electron-density map in specific molecular terms ne(cid:173)
`cessitated a more quantitative approach.
`
`Table 3. "Best-fit" Euler angles at two resolutions and for the
`constant domains taken separately and in pairs
`CHl-CL
`CHl
`CL
`81 82 83 81 82 83 81 82 83
`
`Resolution,
`A
`
`99
`105 336
`6
`99 332 113 100 334 111 106 328 110
`4.5
`The electron-density maps used in the search procedure were
`sampled at one-third the resolution indicated.
`
`The search procedure described above was carried out in(cid:173)
`dependently of the visual interpretation and ultimately led to
`a best fit solution (Figs. 3 and 4). We have calculated structure
`factors and crystallographic R factors for the molecular models
`resulting from both the high- and low-resolution searches (Table
`4). These numbers compare favorably with the R factors of
`other structures at similar stages of their solution (12-15).
`In the search procedure at 4.5-A resolution, the individual
`domains were fitted separately and their final positions in the
`model correspond to the highest peak of overlap. Nothing in
`this search procedure constrains the separate domains to a
`well-behaved and consistent solution. It is therefore reassuring
`that the domains do not interpenetrate. Also, the packing of the
`molecules in the unit cell, taking into account the symmetry
`elements of the space group P212121, provides an arrangement
`in which neighboring molecules do not share density but nev(cid:173)
`ertheless appear to make sufficient intermolecular contacts to
`sustain a crystal lattice.
`
`FIG. 4. Stereo view of the Ca skeleton of the J539 Fab structure.
`Open circles, McPC603 hypervariable residues that were excluded
`from the search.
`
`3 of 4
`
`Celltrion, Inc., Exhibit 1084
`
`
`
`4074
`
`Immunology: Navia et al.
`
`Proc. Natl. Acad. Sci. USA 76 (1979)
`
`Table4. Crystallographic R factors for the J539 models
`Temperature
`factor,
`A2
`
`R factor,
`%
`
`Resolution,
`A
`
`Coordinates Reflections
`
`6
`4.5
`
`a-Carbon
`Polyalanine
`
`934
`2167
`
`50
`25
`
`44.7
`46.5
`
`The final configuration of the model is such that natural
`connections can be made easily between the variable and
`constant halves of the heavy and light chains, respectively. The
`polypeptide chains at this "switch region" (16) of the molecule
`can readily be linked to provide a reasonable connection (Fig.
`3), there being no constraints that would neces&tate an addition
`or deletion of residues.
`The elbow bend, the angle between the pseudo 2-fold sym(cid:173)
`metry 'xes that skewer the constant and variable domain pairs
`(16, 17), provides a measure of the quaternary structure of Fabs.
`All knqwn Fab elbow bend angles are given in Table 5. The
`J539 F~b elbow bend is 136°, which is indistinguishable from
`that of McPC603. This appears to be an extremum in the range
`of allowed values for Fab elbow bends, probably as a result of
`contacts between residues of the constant and variable domain
`pairs.
`The J539 and McPC603 Fabs are simila.r in their quatema.ry
`structures, as suggested by the similarities in their elbow bends.
`The principal difference lies in a small torsion about the local
`dyad of the variable region with respect to the constant region.
`Note that the resolution of the electron-density map is too low
`to provide any information about the conformations of the
`various hypervariable regions of J539; these should be revealed
`with the future high-resolution refinement of the structure.
`The model can provide a plausible explanation for the ob(cid:173)
`servation that soaking of the haptens Gal2 and Gal4 into J539
`crystals did not yield significant intensity changes indicative
`of binding. Examination of the molecular packing shows that
`
`Table 5. Published Fab elbow bend angles
`
`Protein (ref.)
`
`Angle,
`degrees
`
`Newrn (2)
`131*
`134
`McPC603 (l)
`J539 (this work)
`136
`Doh lgG (4)
`147
`Kol Fab (3)
`166
`Kol IgG (17)
`174
`• This elbow bend angle was computed from coordinates in AMSOM
`(9).
`
`the binding cavity in the variable region of J539 may in fact be
`blocked, at least in part, by the constant region of a neighboring,
`symmetry-related molecule.
`Finally, the limited number of degrees of freedom in the
`searches, coupled with the global constraints imposed by the
`electron-density map, confer a higher precision on our model
`than would be expected had the model been constructed from
`the map de novo. This added precision should allow us to pro(cid:173)
`ceed directly to a high-resolution structure determination and
`to the refinement of J539 Fab without need of additional
`high-resolution heavy atom derivatives.
`
`The authors are grateful to Dr. Daniel Mercola for help in the search
`for heavy atom derivatives. We thank Drs. G. H. Cohen and E. W.
`Silverton for valuable discussion and advice and Dr. C. P. J. Glaude(cid:173)
`mans for the haptens used in this study.
`
`5.
`
`l. Segal, D. M., Padlan, E. A., Cohen, G. H., Rudikoff, S., Potter,
`M. & Davies, D. R. (1974) Proc. Natl. Acad. Set USA 71,
`429~.
`2. Poljak, R. J .• Amul, L M., Avey, H.P., Chen, B. L., Phizackerly,
`R. P. & Saul, F. (1973) Proc. Natl. Acad. Sci. USA 70, 330.S-
`3310.
`3. Matsushima, M., Marquart, M., Jones, T. A., Colman, P. M.,
`Bartels, K., Huber, R. & Palm, W. (1978) J. Mol. Biol. 121,
`441-459.
`4. Silverton, E. W., Navia, M. A. & Davies, D. R. (1977) Proc. Natl.
`Acad. Sci. USA 74, 5140-5144.
`Jolley, M. E .• Rudikoff, S., Potter, M. & Glaudemans, C. P. J.
`(1973) Biochemistry 12, 3039-3044.
`6. Rudikoff, S., Potter, M., Segal, D. M., Padlan, E. A. & Davies, D.
`R. (1972) Proc. Natl. Acad. Sci. USA 69, 3689-3692.
`7. Blow, D. M. & Crick, F. H. C. (1959) Acta Crystallogr. 12,
`794- 802.
`8. Padlan, E. A., Segal, D. M., Spande, T. F., Davies, D. R., Rudikoff,
`S. & Potter, M. (1973) Nature (London) New Biol. 145, 165-
`167.
`9. Feldmann, R. J .• ed. (1976) AM SOM -Atlas of Macromolecular
`Structure on Microfiche (Tracor Jitco, Inc., Rockville, MD).
`10. Kabat, E. A., Wu, T. T. & Bilofsky, H. (1976) in Variable Regions
`of lmmunoglobulin Chains (Medical Computer Systems, Bolt,
`Beranek & Newman, Cambridge, MA).
`• 11. Diamond, R. (1966) Acta Crystallogr. 21, 253- 266.
`12. Schmid, M. F., Herriott, J. R. & Lattman, E. A. (1974) J. Mol.
`Biol. 84, 97- 101.
`13. Wishner, B. C., Ward, K. B., Lattman, E. A. & Love, W. E. (1975)
`]. Mol. Biol. 98, 179-194.
`14. Ward, K. B., Hendrickson, W. A. & Klippenstein, G. L. (1975)
`Nature (London) 257, 818-821.
`15. Sarma, R. & Bott, R. (1977) J. Mol. Biol. 113, 555-565.
`16. Davies, D. R., Padlan, E. A. & Segal, D. M. (1975) in Contem(cid:173)
`porary Topics tn Molecular Immunology, eds. Inman, F. P. &
`Mandy, W. J. (Plenum, New York), Vol. 4, pp. 127-155.
`17. Colman, P. M .• Deisenhofer, J., Huber, R. & Palm, W. (1976)].
`Mol. Biol. 100, 257- 282.
`
`4 of 4
`
`Celltrion, Inc., Exhibit 1084
`
`