Proc. Natl. Acad. Sci. USA
`Vol. 90, pp. 5167-5171, June 1993
`Biochemistry
`
`The structure of granulocyte-colony-stimulating factor and its
`relationship to other growth factors
`(protein structure/x-ray crystallography/cytokine structure/colony-stimulating factor 3)
`CHRISTOPHER P. HILL*, TIMOTHY D. OSSLUNDt, AND DAVID EISENBERG
`Molecular Biology Institute and Department of Chemistry and Biochemistry, University of California, Los Angeles, California 90024-1569
`Contributed by David Eisenberg, December 8, 1992
`
`ABSTRACT
`We have determined the three-dimensional
`structure of recombinant human granulocyte-colony-
`stimulating factor by x-ray crystallography. Phases were ini-
`tially obtained at 3.0-A resolution by multiple isomorphous
`replacement and were refined by solvent flattening and by
`averaging of the electron density of the three molecules in the
`asymmetric unit. The current R factor is 21.5% for all data
`between 6.0- and 2.2-A resolution. The structure is predom-
`inantly helical, with 104 of the 175 residues forming a four-a-
`helix bundle. The only other secondary structure is also helical.
`In the loop between the first two long helices a four-residue
`310-helix is immediately followed by a 6-residue a-helix. Three
`residues in the short connection between the second and third
`bundle helices form almost one turn of left-handed helix. The
`up-up-down-down connectivity with two long crossover con-
`nections has been reported previously for five other proteins,
`which like granulocyte-colony-stimulating factor are all signal-
`ing ligands: growth hormone, granulocyte/macrophage-
`colony-stimulating factor, interferon 13, interleukin 2, and
`interleukin 4. Structural similarity among these growth factors
`occurs despite the absence of similarity in their amino acid
`sequences. Conservation of this tertiary structure suggests that
`these different growth factors might all bind to their respective
`sequence-related receptors in an equivalent manner.
`
`Growth and differentiation of various blood cell lines from
`progenitor stem cells are regulated by a group of proteins
`known as hematopoietins (1). These proteins include the
`interleukins (ILs), erythropoietin, macrophage-colony-
`stimulating factor, granulocyte/macrophage-colony-stimu-
`lating factor (GM-CSF), and granulocyte-colony-stimulating
`factor (G-CSF).
`G-CSF is a 19.6-kDa glycoprotein consisting of 174 amino
`acid residues (2). In G-CSF from human blood, there is one
`0-linked glycosyl group at Thr133 (3), which protects the
`molecule from aggregation but does not appear to influence
`receptor binding directly (4). G-CSF, produced mainly by
`macrophages, induces proliferation of neutrophil colonies
`and differentiation of precursor cells to neutrophils, and it
`stimulates the activity of mature neutrophils (5).
`G-CSF belongs to a group ofgrowth factors that have been
`predicted to share a common architecture, despite very low
`sequence similarity (6, 7). The structures offive ofthese have
`been determined-namely, growth hormone (GH) (8, 9),
`GM-CSF (10, 36), interferon ,B (IFN-,8) (11), IL-2 (12, 13), and
`IL-4 (14-16); they all have the same four-a-helix bundle motif
`with up-up-down-down connectivity. Other signaling li-
`gands that are predicted (6, 7) to share this fold include
`prolactin; erythropoietin; IL-3, -5, -6, and -7; myelomono-
`cytic growth factor (MGF); cholinergic differentiation factor;
`ciliary neurotrophic factor; and oncostatin M. Only MGF and
`
`The publication costs of this article were defrayed in part by page charge
`payment. This article must therefore be hereby marked "advertisement"
`in accordance with 18 U.S.C. §1734 solely to indicate this fact.
`
`IL-6 show any significant sequence similarity to G-CSF, with
`37% and 32% conservation of sequence identity, respectively
`(7).
`We have determined the crystal structure of recombinant
`human G-CSF (rhG-CSF),* which is expressed in Esche-
`richia coli, is not glycosylated and retains the amino-terminal
`fMet residue; it is, however, biologically active (2). We refer
`to the fMet as residue number -1.
`
`MATERIALS AND METHODS
`The rhG-CSF used in this study was provided by Amgen
`Biologicals. Crystals were grown (T.D.O., R. Luthy, D.
`Cascio, and D.E., unpublished work) in hanging drops at pH
`5.8 over a reservoir of 8% (wt/vol) PEG 8000/380 mM
`MgSO4/220 mM LiCl. The space group is P212121 (a = 91.2
`A, b = 110.3 A, c = 49.5 A). There are three molecules in the
`asymmetric unit and the Matthews parameter Vm (17) is 2.1
`A3/Da.
`Soaking experiments were performed by dissolving the
`heavy-atom compound in reservoir solution, a small volume
`of which was added to drops containing the crystals. Data
`from native and heavy-atom-soaked crystals were collected
`to 3.0-A resolution on a San Diego Multiwire Systems area
`detector. Data were processed using the program FS (18), and
`two heavy-atom derivatives, thimerosal and praseodymium
`acetate, were used to initiate determination of the protein
`structure (details are given in Tables 1 and 2).
`The solvent-flattened (21) multiple isomorphous replace-
`ment (MIR) map clearly showed the presence of four helices
`corresponding to each of the three molecules in the asym-
`metric unit. These three molecules are not related to each
`other by a proper rotation axis. Noncrystallographic sym-
`metry operators were derived from a partial atomic model,
`and averaging of the electron density led to a significant
`improvement in the quality of the map; revealing longer
`helices, side-chain density, and parts ofthe connecting loops.
`This map was interpreted conservatively as 100 alanine
`residues in the helical regions only. Positional refinement
`with PROLSQ (22) gave an R factor of 38%. Combination of
`MIR and polyalanine model phases prior to solvent leveling
`and averaging gave a map that in most regions was of good
`quality, and into which it was easy to fit the known amino acid
`sequence (23).
`
`Abbreviations: G-CSF, granulocyte-colony-stimulating factor; rhG-
`CSF, recombinant human G-CSF; GM-CSF, granulocyte/
`macrophage-colony-stimulating factor; IL, interleukin; IFN, inter-
`feron; GH, growth hormone; hGH, human GH; pGH, porcine GH;
`MIR, multiple isomorphous replacement.
`*Present address: Department of Biochemistry, University of Utah,
`Salt Lake City, UT 84132.
`tPresent address: Amgen, Molecular Structure Lab, Building 2,
`Thousand Oaks, CA 91360.
`tThe atomic coordinates and structure factors have been deposited in
`the Protein Data Bank, Chemistry Department, Brookhaven Na-
`tional Laboratory, Upton, NY 11973 (reference 1RHG, R1RHGSF).
`
`5167
`
`Page 1
`
`KASHIV EXHIBIT 1054
`IPR2019-00791
`
`

`

`5168
`
`Biochemistry: Hill et al.
`
`Proc. Natl. Acad. Sci. USA 90 (1993)
`
`Table 1.
`
`Soaking
`
`Data used to solve the structure of rhG-CSF
`Pr(OAc)3
`Thimerosal
`Native
`0.1 mM,
`0.1 mM,
`48 hr
`48 hr
`No. of sites
`4
`3
`Riso*
`0.24
`0.21
`3.o A
`3.o A
`2.9 A
`Resolution
`43,495
`No. of observations
`46,840
`50,868
`5,970
`1,565
`No. of rejects
`1,281
`8,815
`10,031
`No. unique
`10,735
`% completeness
`84.4
`96.1
`93.3
`% ofIF > 2oF
`96.7
`97.0
`95.5
`Rsym (no rejects)t
`0.071
`0.057
`0.056
`Unless otherwise stated, crystallographic programs used were
`from the CCP4 suite (19). Pr(OAc)3, praseodymium acetate.
`*Riso = XlFderivativel - Fnativell/IlFnativei.
`tRsym =IE - IavI/:Iav.
`
`Most of the structure, especially the helical regions, was
`clearly defined at this stage. Other sections were less pre-
`cisely determined; in particular, residues -1 to 9, 65 to 70,
`127 to 136, and 173 to 174 did not have convincing density.
`In general the segments that could not be located in MIR/
`averaged maps are close to neighboring molecules in the
`crystal, suggesting that these regions of poor electron density
`may be due to averaging out features that differ among the
`three molecules. Consequently, a mask was manually defined
`to include only regions that were known to obey the non-
`crystallographic symmetry (i.e., were well defined in the
`averaged map). Another mask was defined for regions that
`were confidently expected to be solvent (i.e., well removed
`from the missing residues). An iterative procedure of map
`calculation, modification, and back-transformation was em-
`ployed, in which regions outside the masks were not modified
`explicitly. Several different strategies were tried, all with
`similar results (Fig. 1). This gave a noticeable improvement
`in the map quality; however, it was still not possible to locate
`the missing residues, which remain undefined even after
`refinement.
`The structure has been refined with XPLOR (25) against
`2.2-A data collected on an R-axis imaging plate detector. The
`current R factor is 21.5% and the rms deviation from ideality
`of covalent bond lengths is 0.017 A; further details are given
`in Tables 3 and 4. The unusually large average atomic B factor
`for protein atoms, 44 A2, is in very good agreement with the
`value estimated from a Wilson plot (26), 45 A2. The accuracy
`of our model is supported by the free R factor (27), which is
`34.4%, and by three-dimensional profile assessment (28).
`
`MIR/averaged density for helix D of rhG-CSF. This map
`FIG. 1.
`was computed from Fo coefficients with MIR phases that were
`refined by 10 cycles of averaging, Fourier inversion, and map
`calculation with combined MIR/averaged phases. Solvent leveling
`was not applied in this case. Map averaging was performed using
`Bricogne's programs (24). The helical main chain is well defined and
`in general the side chains have reasonable density. Side chains with
`weak or absent density are usually fully exposed to solvent-for
`example, Glu'62, Tyr165, and Arg169.
`
`RESULTS
`rhG-CSF is an antiparallel four-a-helix bundle with a left-
`handed twist, and with overall dimensions of 45 A x 26 A x
`26 A (Fig. 2). The four helices within the bundle are referred
`to as helices A-D; their connecting loops are known as the
`AB, BC, and CD loops. The AB and CD loops are long
`overhand connections; only the BC loop is of the more usual
`short hairpin type (31).
`The rhG-CSF bundle is regular with helix crossing angles
`that range between -167° and -159°. The average crossing
`angle (-162.5°) is very close to that expected (-161°) for an
`ideal left-handed antiparallel four-a-helix bundle (32). Heli-
`ces A, B, and C are straight, whereas helix D bends towards
`the shorter helix B. The change in axial direction between the
`ends of helix D is 350, with the greatest changes centered on
`Gly149 and Ser159. The longest straight portion of helix D
`(residues 159-173) makes the most extensive interactions
`with the A, B, and C helices, and this is the section of helix
`D used above to define the crossing angles within the bundle.
`In addition to the four major helices that comprise the
`bundle, there is a shorter helical section within the AB loop.
`
`Table 2.
`
`Heavy-atom phasing statistics
`Resolution
`9.91
`Thimerosal
`fh(rms)/E(ms)(centrics)
`fh(rms)/E(rms)(acentrics)
`Rcullis
`Pr(OAc)3
`fh(rms)/E(rms)(centrics)
`fh(rms)/E(rms)(acentrics)
`Rcullis
`Mean figure of merit = 0.59
`Heavy-atom parameters were refined by the method of correlating origin-removed Patterson functions (20). The phasing power of the
`praseodymium acetate [Pr(OAc)3] derivative, 0.84, would normally indicate a useless derivative. In this case, however, the density maps were
`clearly improved when anomalous scattering from Pr(OAc)3 was included in the phase calculation. We assume that this results from the large
`anomalous signal (fi' = 10.5e) of praseodymium. The heavy-atom binding sites are chemically reasonable. Each of the three thimerosal mercury
`atoms is bound to the single free thiol, Cys17, in each of the three molecules in the asymmetric unit. In contrast, praseodymium sites do not obey
`the noncrystallographic symmetry. They are located near clusters of at least three carboxylate side chains, which in every case come from at least
`two different rhG-CSF molecules. fh = heavy-atom structure factor; E = residual lack of closure; Rcullis = XIIFPH - FPI - Ifh(cac)II/jIFPH - FpI.
`
`6.56
`
`1.04
`1.35
`0.57
`
`0.94
`1.38
`0.59
`
`5.22
`
`0.74
`1.03
`0.75
`
`0.75
`0.92
`0.67
`
`4.46
`
`0.62
`0.83
`0.71
`
`0.43
`0.64
`0.72
`
`3.96
`
`0.64
`0.84
`0.68
`
`0.44
`0.70
`0.77
`
`0.93
`1.13
`0.59
`
`1.13
`1.31
`0.52
`
`3.60
`
`0.60
`0.85
`0.70
`
`0.61
`0.79
`0.69
`
`3.32
`
`0.67
`0.94
`0.72
`
`0.53
`0.77
`0.75
`
`3.09
`
`1.00
`1.08
`0.57
`
`0.60
`0.80
`0.72
`
`Total
`
`0.77
`0.98
`0.66
`
`0.66
`0.84
`0.67
`
`Page 2
`
`

`

`Biochemistry: Hill et al.
`
`Table 3.
`
`Refinement of rhG-CSF
`
`Proc. Natl. Acad. Sci. USA 90 (1993)
`
`5169
`
`D
`
`Resolution, A
`6.00-3.99
`3.99-3.34
`3.34-2.97
`2.97-2.73
`2.73-2.55
`2.55-2.41
`2.41-2.29
`2.29-2.20
`Overall
`
`R factor
`16.3
`17.7
`22.4
`26.4
`28.2
`30.0
`31.3
`32.7
`21.5
`
`Free
`R factor
`34.2
`33.1
`32.7
`35.6
`36.5
`34.8
`38.5
`37.2
`34.4
`
`The protein chain makes a sharp turn out of helix A and 5
`residues later goes into a 4-residue 310-helix. At Leu47 there
`is a shift in the chain direction and the 310-helix leads
`immediately into a 6-residue a-helix. The 45° angle between
`these short helices wraps them around the N-terminal end of
`helix D. They are relatively exposed and protrude from the
`main body of the structure. Residues from this region overlap
`with epitopes recognized by neutralizing monoclonal anti-
`bodies (33).
`All of the non-glycine residues fall into the allowed region
`of 0,4i space, except for Lys4O, Glu93, and Ile95. Lys4O lies at
`the C terminus of helix A, a context that is known to favor the
`aL conformation (34). Glu93, Gly94, and Ile95, located in the
`short BC loop, all have positive
`angles and together form
`almost one turn reminiscent of a left-handed helix. The
`conformation of these residues is clear in the MIR/averaged
`map.
`The two disulfide bonds in rhG-CSF, Cys36-Cys42 and
`Cys64_Cys74, are both required for activity (35). They are
`located at opposite ends ofthe long AB loop, where they form
`short loops to the C-terminal end of helix A and the N-ter-
`minal end ofhelix B (Figs. 2 and 3). The Cys36-Cys42 disulfide
`forms the major part of a neutralizing antibody epitope (33).
`Circular dichroism measurements show that in the absence of
`Cys64-Cys74 only about half of the native structure a-helix is
`formed (35).
`
`DISCUSSION
`rhG-CSF belongs to a distinct structural class of growth
`factors. Comparison with GM-CSF (10, 36), GH (8, 9), IFN-j3
`(11), IL-2 (12), and IL-4 (14-16) reveals a common motif of
`a four-a-helix bundle with two long crossover connections.
`This similarity is evident from connectivity diagrams (Fig. 3),
`and it exists despite little sequence similarity. Differences
`include the number and position of disulfide bonds. To our
`knowledge no other proteins are yet known to share this
`architecture.
`
`9-61, 71-126, 137-172
`209-264, 271-322, 337-372
`409-461, 471-526, 537-572
`
`Table 4. The final model
`Residues included
`Molecule A
`Molecule B
`Molecule C
`No. of water molecules
`rms deviations
`40
`0.017 A
`Bonds
`3.3540
`Angles
`20.8050
`Dihedrals
`1.4370
`Impropers
`1.63 A
`B (main-chain bonds)
`The residue numbers of molecules B and C have been incremented
`by 200 and 400, respectively. Nine residues that lack clear side-chain
`density have been truncated to Ala.
`
`C
`
`A
`
`Ribbon diagram of rhG-CSF. This figure was prepared
`FIG. 2.
`using the program MOLSCRIPT (29). Secondary structure was defined
`using DSSP (30). The main bundle helices A (residues 11-39), B
`(71-91), C (100-123), and D (143-172) are labeled near their N
`termini. The short 310 (44-47) and a (48-53) helices are also indi-
`cated, as are residues 93-95, which form almost one turn reminiscent
`of a left-handed helix. Residues - 1 to 8 and 173 to 174 are not visible
`in Fourier maps and are not included in this figure. The approximate
`positions ofthe other missing residues, 65-70 and 127-136, have been
`drawn with thin lines to indicate connectivity.
`
`These growth factors differ in the local conformations of
`their loops. For example the BC loops of IL-4 and hGH
`contain short a-helices, while pGH has an co loop (37), and
`rhG-CSF has almost one turn of a left-handed helix in this
`position. The AB and CD loops of GM-CSF and IL-4 form
`two-stranded antiparallel
`3-sheets that run approximately
`parallel to the bundle axis; the CD loop of IFN-f3 contains a
`17-residue a-helix that is aligned with the bundle axis and
`packs against the helices equivalent to B and D of rhG-CSF.
`Only hGH, pGH, and IL-2 appear to have helices that are
`even approximately close to the short 310- and a-helices
`found in the AB loop of rhG-CSF.
`There are also differences in the bundle geometries. The
`bundle helices of rhG-CSF, pGH, and hGH have almost
`twice as many residues as in GM-CSF. The relative lengths
`of bundle helices vary (Fig. 3): for example, in rhG-CSF,
`pGH, and hGH the A and D helices are longer than the B and
`C helices, whereas for IL-4 the reverse is true. The helix
`crossing angles also differ. rhG-CSF, which has one of the
`longest bundles, has crossing angles very close to those
`expected for packing of i + 4 ridges into i + 3 grooves
`(-160°). At the other extreme, GM-CSF, which has the
`smallest bundle, has wide crossing angles. IL-4, whose
`bundle is intermediate in size, has crossing angles that are
`intermediate between those of rhG-CSF and GM-CSF.
`Despite their differences, these growth factors, which all
`function by binding to cell surface receptors, clearly share the
`same basic architecture. The extracellular portion of their
`receptors includes a conserved "cytokine-binding" domain
`of :210 residues (38-42). This suggests that these signaling
`ligands might all bind to their respective receptors with
`equivalent geometries. Thus far, hGH is the only one ofthese
`ligands whose receptor-bound structure has been reported
`(9). A single molecule of hGH facilitates receptor dimeriza-
`tion by simultaneously binding to two different receptor
`
`Page 3
`
`

`

`5170
`
`Biochemistry: Hill et al.
`
`Proc. Natl. Acad. Sci. USA 90 (1993)
`
`rhG-CSF
`
`hGH
`
`pGH
`
`1A
`
`171'
`
`174
`
`GM-CSF
`
`INF-13
`
`103
`
`t
`
`4139271
`
`8
`
`1
`
`llG16 ALDS 43
`5 4J196 155
`
`13
`
`1
`
`A
`
`C
`
`6
`
`13
`
`77
`
`6
`
`IL-4
`69
`
`67
`
`127 C
`
`N
`
`114
`
`IL-2
`
`36
`
`A
`
`C
`
`96
`
`13352S
`
`N
`
`8010
`
`13 24
`113
`
`12936
`
`~~~~2
`
`30
`
`~~31
`
`0
`
`112
`
`10
`
`FIG. 3.
`Connectivity diagrams of rhG-CSF, human GH (hGH; ref. 9), porcine GH (pGH; ref. 8), GM-CSF (10, 36), IFN-,B (11), IL-2 (12),
`and IL-4 (14, 15). These diagrams are based on inspection of cited references. The lengths of secondary structural elements are drawn in
`proportion to the number of residues. There are some minor differences in the secondary structure assignments reported by the various groups
`that have independently determined the structures of GM-CSF and of IL-4. A, B, C, and D helices are labeled according to the scheme used
`in this paper for rhG-CSF. For IFN-,3 the original labeling of helices is indicated in parentheses.
`molecules. The first hGH receptor-binding surface is primar-
`ily composed ofresidues from helix A, the AB loop, and helix
`D; the residues of the second binding surface are from helix
`A, the BC loop, and helix C. It remains to be seen whether
`the structural similarity shown within this class of growth
`factors extends to their mode of receptor recognition.
`
`5.
`
`6.
`7.
`8.
`
`Note Added in Proof. The recently reported crystal structure reveals
`that macrophage-colony-stimulating factor (M-CSF) (43) belongs to
`the same structural family as G-CSF, although M-CSF is unique in
`consisting of disulfide-linked dimers. In this regard we note that
`IFN-yis a dimer in which interpenetrating helices form two domains,
`each of which is reminiscent of IFN-,B (44).
`
`We thank the National Institutes of Health for support and the San
`Diego Super Computer for computer time.
`
`1.
`2.
`
`3.
`
`4.
`
`Nicola, N. A. (1989) Annu. Rev. Biochem. 58, 45-77.
`Souza, L. M., Boone, T. C., Gabriloe, J., Lai, P. H., Zsebo,
`K. M., Murdock, D. C., Chazin, V. R., Bruszewski, J., Lu,
`H., Chen, K. K., Barendt, J., Platzer, E., Moore, M. A. S.,
`Mertelsmann, R. & Welte, K. (1986) Science 232, 61-65.
`Kubota, N., Orita, T., Hattori, K., Oh-eda, M., Ochi, N. &
`Yamazaki, T. (1990) J. Biochem. (Tokyo) 107, 486-492.
`Oh-eda, M., Hasegawa, M., Hattori, K., Kuboniwa, H., Ko-
`jima, T., Orita, T., Tomonou, K., Yamazaki, T. & Ochi, N.
`(1990) J. Biol. Chem. 265, 11432-11435.
`
`9.
`
`10.
`
`11.
`
`15.
`
`16.
`
`17.
`
`Nagata, S. (1990) in Handbook ofExperimental Pharmacology:
`Vol 95/I: Peptide Growth Factors and Their Receptors, eds.
`Sporn, M. B. & Roberts, A. B. (Springer-Verlag, New York),
`pp. 699-722.
`Bazan, J. F. (1990) Immunol. Today 11, 350-354.
`Bazan, J. F. (1991) Neuron 7, 197-208.
`Abdel-Meguid, S. S., Shieh, H.-S., Smith, W. W., Dayringer,
`H. E., Violand, B. N. & Bentle, L. A. (1987) Proc. Natl. Acad.
`Sci. USA 84, 6434-6437.
`Vos, A. M., Ultsch, M. & Kossiakoff, A. A. (1992) Science
`255, 306-312.
`Diederichs, K., Boone, T. & Karplus, P. A. (1991) Science 254,
`1779-1782.
`Senda, T., Shimazu, T., Matsuda, S., Kawano, G., Shimizu,
`H., Nakamura, K. T. & Mitsui, Y. (1992) EMBO J. 11, 3193-
`3201.
`12. McKay, D. B. (1992) Science 257, 412-413.
`Bazan, J. F. (1992) Science 257, 410-412.
`13.
`Powers, R., Garrett, D. S., March, C. J., Frieden, E. A.,
`14.
`Gronenborn, A. M. & Clore, G. M. (1992) Science 256, 1673-
`1677.
`Smith, L. J., Redfield, C., Boyd, J., Lawrence, G. M. P.,
`Edwards, R. G., Smith, R. A. G. & Dobson, C. M. (1992) J.
`Mol. Biol. 224, 899-904.
`Wlodawer, A., Pavlovsky, A. & Gustchina, A. (1992) FEBS
`Lett. 309, 59-64.
`Matthews, B. W. (1968) J. Mol. Biol. 33, 491-497.
`
`Page 4
`
`

`

`Biochemistry: HUI et al.
`
`Proc. Natl. Acad. Sci. USA 90 (1993)
`
`5171
`
`18.
`
`20.
`
`Weissman, L. J. (1979) Ph.D. dissertation (Univ. of California,
`Los Angeles).
`19. The SERC Collaborative Computing Project No. 4 (1979) A
`Suite of Programs for Protein Crystallography (Daresbury
`Lab., Warrington, U.K.).
`Terwilliger, T. C. & Eisenberg, D. (1983) Acta Crystallogr.
`Sect. A 39, 813-817.
`21. Wang, B.-C. (1985) Methods Enzymol. 115, 90-112.
`Hendrickson, W. A. (1985) Methods Enzymol. 115, 252-270.
`22.
`Jones, T. A. (1985) Methods Enzymol. 115, 157-171.
`23.
`Bricogne, G. (1976) Acta Crystallogr. Sect. A A32, 832-847.
`24.
`Brunger, A. T. (1992) X-PLOR, A System for Crystallography
`25.
`and NMR (Yale Univ., New Haven, CT), Version 3.0.
`Wilson, A. J. C. (1949) Acta Crystallogr. 2, 318-321.
`Brunger, A. T. (1992) Nature (London) 355, 472-475.
`Luthy, R., Bowie, J. U. & Eisenberg, D. (1992) Nature (Lon-
`don) 356, 83-85.
`Kraulis, P. J. (1991) J. Appl. Crystallogr. 24, 946-950.
`Kabsch, W. & Sander, C. (1983) Biopolymers 22, 2577-2637.
`Presnell, S. R. & Cohen, F. E. (1989) Proc. Natl. Acad. Sci.
`USA 86, 6592-6596.
`Chothia, C., Levitt, M. & Richardson, D. (1977) Proc. Natl.
`Acad. Sci. USA 74, 4130-4134.
`Layton, J. E., Morstyn, G., Fabri, L. J., Reid, G. E., Burgess,
`A. W., Simpson, R. J. & Nice, E. C. (1991) J. Biol. Chem. 266,
`23815-23823.
`
`26.
`27.
`28.
`
`29.
`30.
`31.
`
`32.
`
`33.
`
`34.
`
`35.
`
`36.
`
`Richardson, J. S. & Richardson, D. C. (1988) Science 240,
`1648-1652.
`Lu, H. S., Clogston, C. L., Narhi, L. O., Merewether, L. A.,
`Pearl, W. R. & Boone, T. C. (1992) J. Biol. Chem. 267,
`8770-8777.
`Walter, M. R., Cook, W. J., Ealick, S. E., Nagabhushan,
`T. L., Trotta, P. P. & Bugg, C. E. (1992) J. Mol. Biol. 224,
`1075-1085.
`Leszczynski, J. F. & Rose, G. D. (1986) Science 234, 849-855.
`37.
`38. Cosman, D., Lyman, S. D., Idzerda, R. L., Beckmann, M. P.,
`Park, L. S., Goodwin, R. G. & March, C. J. (1990) Trends
`Biochem. Sci. 15, 265-270.
`Arai, K.-I., Lee, F., Miyajima, A., Miyatake, S., Arai, N. &
`Yokota, T. (1990) Annu. Rev. Biochem. 59, 783-836.
`Bazan, J. F. (1990) Proc. Natl. Acad. Sci. USA 87, 6934-6938.
`Davis, S., Aldrich, T. H., Valenzuela, D. M., Wong, V., Furth,
`M. E., Squinto, S. P. & Yancopoulos, G. D. (1991) Science
`253, 59-63.
`Gearing, D. P., Thut, C. J., Vandenbos, T., Gimpel, S. D.,
`Delaney, P. B., King, J., Price, V., Cosman, D. & Beckmann,
`M. P. (1991) EMBO J. 10, 2839-2848.
`Pandit, J., Bohm, A., Jancarik, J., Halenbeck, R., Koths, K. &
`Kim, S.-H. (1992) Science 258, 1358-1362.
`Ealick, S. E., Cook, W. J., Vijay-Kumar, S., Carson, M.,
`Nagabhushan, T. L., Trotta, P. P. & Bugg, C. E. (1991) Sci-
`ence 252, 698-702.
`
`39.
`
`40.
`41.
`
`42.
`
`43.
`
`44.
`
`Page 5
`
`