Human Growth Hormone and
`Extracellular Domain of Its Receptor:
`Crystal Structure of the Complex
`
`ABRAHAM M. DE VOS,* MARK ULTSCH, ANTHONY A. KOSSLAKOFF
`
`Downloaded from
`
`http://science.sciencemag.org/
`
`
`
`on September 17, 2020
`
`Binding ofhuman growth hormone (hGH) to its receptor
`is required for regulation of normal human growth and
`development. Examination of the 2.8 angstrom crystal
`structure of the complex between the hormone and the
`extracellular domain ofits receptor (hGHbp) showed that
`the complex consists of one molecule ofgrowth hormone
`per two molecules of receptor. The hormone is a four-
`helix bundle with an unusual topology. The binding
`protein contains two distinct domains, similar in some
`respects to unmunoglobulin domains. The relative orien-
`tation of these domains differs from that found between
`constant and variable domains in immunoglobulin Fab
`fragments. Both hGHbp domains contribute residues
`that participate in hGH binding. In the complex both
`receptors donate essentially the same residues to interact
`with the hormone, even though the two binding sites on
`hGH have no structural similarity. Generally, the hor-
`mone-receptor interfaces match those identified by previ-
`ous mutational analyses. In addition to the hormone-
`receptor interfaces, there is also a substantial contact
`surface between the carboxyl-terminal domains of the
`receptors. The relative extents of the contact areas sup-
`port a sequential mechanism for dimerization that may be
`crucial for signal transduction.
`
`T HE GROW1TH HORMONE RECEPTOR IS ACTIVATED ON BIND-
`ing of growth hormone to stimulate the growth and metab-
`olism ofmuscle, bone, and cartilage cells (1). This receptor is
`a member ofa group ofreceptors that are found on various cell types
`and are generally involved in cell growth and differentiation. It has
`been recognized that a structural relationship exists between the
`extracellular domain of the endocrine hormone receptors and the
`extracellular domains of a group of cytokine receptors, including
`those for interleukins 2, 3, 4, 6, and 7, granulocyte and granulocyte-
`macrophage colony-stimulating factors, and erythropoietin (2, 3).
`Also, there is a more distant relationship with the extracellular
`domain of the receptors for tissue factor and the interferons (3). All
`these receptors are grouped together in the hematopoietic super-
`family (2, 3). A recent addition to this superfamily is the receptor for
`ciliary neutrophic factor, which is involved in neuropoiesis (4).
`Like the receptor tyrosine kinases (5), members of the hemato-
`
`The authors are in the Department of Protein Engineering, Genentech, Inc., 460 Point
`San Bruno Boulevard, South San Francisco, CA 94080.
`*To whom correspondence should be addressed.
`
`306
`
`poietic receptor superfamily have a three-domain organization com-
`prising an extracellular ligand binding domain, a single transmem-
`brane segment, and an intracellular domain of unknown function,
`which within the family is not homologous. Beyond this, there is
`virtually no direct structural information bearing on possible mech-
`anisms of activition or on details of molecular contacts. In analogy
`to receptor tyrosine kinases, the mechanism through which infor-
`mation from the ligand binding event is transmitted through the
`membrane by the activated receptor is assumed to involve some type
`ofaggregation. However, the molecular details ofaggregation ofthe
`ligand-bound receptors are not understood; most proposed models
`for receptor aggregation postulate complexes of ligand-receptor
`pairs, that is, a stoichiometry of two ligands and two receptors.
`The extracellular domain of the human growth hormone (hGH)
`receptor (residues 1 to 246) occurs naturally in serum in the form of
`a hormone binding protein, which binds hGH with approximately
`the same affinity as the intact receptor (6) and which may play a
`physiological role in the regulation of hormone clearance. The
`complex between hGH and a slightly truncated form of this binding
`protein (hGHbp, residues 1 to 238) consists of one molecule of
`hGH and two molecules of hGHbp hGH-(hGHbp)2 (7, 8). This
`was surprising because it was known from the structure of the
`porcine growth hormone (9) that there was no evidence for even
`pseudo-symmetrical binding surfaces that would support binding
`for two receptors simultaneously. This raised the possibility that
`either the two hormone binding sites interfaced with different
`regions of the receptor, or that the receptor binding surface could
`reconfigure to bind tightly a second set of hormone binding
`determinants.
`Here, we report the structure of the hGH.(hGHbp)2 complex
`which shows the novel manner in which a single monomeric protein
`molecule binds and brings together two receptor molecules. No
`other structures of protein-receptor complexes are known, although
`crystals of other such complexes have been reported (10). Interac-
`tions between receptors and ligands and between antibodies and
`antigens are examples of molecular recognition. However, unlike
`the antibody binding diversity that is expressed by changes in
`sequence of a limited number of residues on a relatively constant
`structural scaffold, the hormone-binding determinants of the hGH
`receptor as seen in the structure that we describe depend on
`conformational diversity in the presence of conserved sequence.
`Although the growth hormone system differs in detail from other
`hormone-receptor complexes in the hematopoietic superfamily, the
`general theme as to how receptors aggregate is likely to be a
`relatively common feature of the family as a whole.
`Structure of the hormone and the binding proteins. The hGH
`binding protein (hGHbp, residues 1 to 238) was produced as a
`soluble protein from Escherwhia coli (6). Purification of the binding
`
`SCIENCE, VOL. 255
`
`Novo Nordisk Ex. 2054, P. 1
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Downloaded from
`
`http://science.sciencemag.org/
`
`
`
`on September 17, 2020
`
`Fig. 1. Electron density for part of the hGH-
`hGHbp I interface. The current refined model in
`interface region I is superimposed on (A) the
`solvent flattened MIR map, and (B) the 2FO-F,
`map, with phases calculated from the final model.
`The hGH atoms are green and receptor atoms are
`orange.
`
`helix. Alignment of the sequences to the
`density was straightforward, as there was
`good density for all the expected disulfide
`bonds and for almost all large side chains.
`Electron density was weak or absent for all
`termini, for part ofone loop in hGH and for
`two loops in each receptor, both in the MIR
`map and in the solvent flattened map. The
`structure was refined to an R factor of 0.204
`(10 to 2.8 A) (Table 1).
`The major structural feature of the hGH
`molecule is a four-helical bundle (Fig. 2)
`with unusual connectivity, which was de-
`scribed first for the structure of porcine
`growth hormone (9); the helices run up-up-
`down-down, in contrast to the more usual
`up-down-up-down case. The NH2- and
`COOH-terminal helices (helices 1 and 4)
`are longer than the other two (26 and 30
`residues compared to 21 and 23 residues),
`and helix 2 is kinked at Pro89. A long
`crossover connection, consisting of residues
`35 to 71, links helix 1 to helix 2, and a
`similar connection (residues 129 to 154) is
`found between helices 3 and 4. The first
`connection is disulfide-bonded to helix 4
`through Cys53 and Cys`65. In contrast, helix
`2 is linked to helix 3 by a much shorter
`segment (residues 93 to 105). In addition to
`the four helices in the core, three much
`shorter segments of helix are found in the
`connecting loops: one each at the beginning
`and end of the connection between helices 1
`and 2 (residues 38 to 47 and 64 to 70,
`respectively), and one in the short connec-
`tion between helices 2 and 3 (residues 94 to
`100). The NH2-terminal eight residues ex-
`tend away from the remainder of the mole-
`cule, whereas the COOH-terminus is linked
`to helix 4 with a disulfide bond between
`Cys'82 and Cys'89
`The topography of the hormone appears
`to be similar to that described for porcine growth hormone (pGH)
`(9). Exceptions are the two short helices in the connecting segment
`between helix 1 and 2, which were not described for pGH; since
`they are involved in contacts between hormone and receptor
`(below), they may represent conformational changes in the hor-
`mone upon receptor binding. In addition, the connection between
`helices 2 and 3 has an omega-loop conformation in the porcine
`hormone (9). Since this connection does not participate in receptor
`binding (below), the difference in loop conformation represents a
`structural difference between hGH and pGH. The residues on the
`hormone that are color coded in Fig. 2 are directly involved
`receptor binding.
`The core of the four-helix bundle is made up of mostly hydro-
`phobic residues (Fig. 2) with the exceptions of Ser79 and Asp169.
`The Oy of Ser79 in helix 2 hydrogen-bonds back to the carbonyl
`
`protein, formation and characterization of the complex, and crystal-
`lization procedures have been described (7). Crystals with cell
`parameters a = 145.8 A, b = 68.6 A, c = 76.0 A were in space group
`P21212. Before the data were collected, the crystals were stabilized
`in 40 percent saturated ammonium sulfate and 0.1 M sodium
`acetate, pH 5.5. The crystals contain a mixture ofhGH and hGHbp
`(1:2) in the asymmetric unit (7, 8), and this is also the stoichiometry
`of the complex in solution (8). Phases for the observed intensities
`were determined by multiple isomorphous replacement with two
`heavy atom derivatives, combined with solvent flattening. The
`overall quality of the electron density maps was quite good (Fig. 1)
`(11), and the outline of the molecules and the individual domains
`was obvious. The electron density for the hormone was easily
`recognizable because of its four-helix bundle structure, whereas the
`density assigned to the binding protein did not contain any obvious
`
`17 JANUARY 1992
`
`RESEARCH ARTICLE
`
`307
`
`Novo Nordisk Ex. 2054, P. 2
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Downloaded from
`
`http://science.sciencemag.org/
`
`
`
`on September 17, 2020
`
`Leu75 and I1e78 in helix 2, and with Leu'57, Tyr160, Tyr164, Cys165,
`and Phe`76 in helix 4; but Leu93, Val96, and Phe97 in the short
`segment between helices 2 and 3 interact with Phe31 of helix 1 and
`with Leu'62 and Leu'63 of helix 4.
`The extracellular part of the receptor consists of two domains
`(residues 1 to 123 and 128 to 238, respectively), linked by a single
`four-residue segment of polypeptide chain (Fig. 3A). Each domain
`contains seven P strands (Fig. 3B) that together form a sandwich of
`two antiparallel P3 sheets, one with four strands and one with three,
`with the same topology in each domain. The two-domain structure
`and the presence in each domain of two P3 sheets were predicted by
`Bazan (3). He also proposed that the topology of the sandwich
`might be that of immunoglobulin constant domains. Instead, the
`topology of the hGHbp domains is identical to that of domain D2
`of CD4 (14) and domain D2 of chaperone protein PapD (15),
`which differs from immunoglobulin constant domains in that "sheet
`switching" has taken place (14), with strand C' as part of the sheet
`formed by strands C, F, and G rather than of the other sheet. Strand
`G in the COOH-terminal domain is preceded by a stretch of
`irregular extended structure between Tyr222 and Ser226, with a
`bulge at Gly223 to Glu224. As a result, the side chains of Tyr222 and
`Phe225 both point into the solvent, whereas Oy of Ser226 forms a
`hydrogen bond to the main chain amine of Val212 in the neighbor-
`ing strand.
`The NH2-terminal 30 residues of both receptor molecules in the
`complex were not apparent in the electron density map and are not
`part of our model. Therefore, the ordered structure of the NH2-
`terminal domain is smaller and more compact than that of the
`COOH-terminal domain. Superposition of the domains shows that
`
`oxygen of Leu75 (2.9 A). The 081 of Asp`69 in helix 4 hydrogen
`bonds to the Q-y of Ser55 (3.0 A) as well as to the NE ofTrp86 (2.9
`A), as proposed on the basis of absorption spectroscopy (12)
`combined with mutagenesis (13). 082 ofAsp'69 is pointed outward
`from the core and appears to interact with N; of Lys 72 (4.1 A).
`Other hydrophobic clusters can be found between the four-helix
`core and the connecting segments. Thus, lle36, Phe44, Cys53, Phe54,
`and Ile58 in the connection between helices 1 and 2 interact with
`
`Crystallographic statistics. Data were collected on an Enraf-
`Table 1.
`Nonius FAST area detector, mounted on a Rigaku RU200 rotating anode
`generator operated at 45 kV, 110 mA. Crystals were mounted with the b*
`axis parallel to the rotation axis, and two crystal settings were used to
`produce complete data sets. Processing was done with MADNES (25) and
`PROCOR (26). Two native data sets were collected to a resolution of 2.8
`A, and when combined gave 95 percent completeness [Rmcrg(I) = 0.13,
`all reflections between 15 and 2.8 A with F > 0]. For derivatives, crystals
`were soaked in heavy atom compounds dissolved in stabilization solution.
`Both K2PtCI4 and K2AuCI4 gave a highly occupied single-site derivative.
`with
`refinement
`phase
`during
`used
`were
`differences
`Anomalous
`PROTEIN (27). The final figure of merit was 0.55 (15 to 3.0 A, 14,787
`reflections). Solvent flattening (28) increased the figure of merit to 0.76.
`The resulting solvent flattened map was used for chain tracing and model
`building with the original MIR map as a reference. The starting model for
`refinement consisted of hGH residues 3 to 134 and 154 to 189, residues
`33 to 51, 65 to 70, and 79 to 231 for the first receptor, and residues 35
`to 51, 65 to 69, and 80 to 235 for the second receptor. Of these 516
`amino acids (out of 667), 52 side chains were trimmed back to alanine.
`Crystallographic refinement was done with XPLOR (29). The starting R
`factor was 0.47 (10 to 3.0 A); conventional positional refinement
`decreased the R factor to 0.32, and one cycle of simulated annealing to
`0.27. The resolution was extended to 2.8 A, and combination of map
`fitting and refinement resulted in R = 0.249 (10 to 2.8 A, 17,985
`reflections, or 95 percent of the possible number). At this stage, tightly
`restrained individual temperature factors were refined. The final model
`consisted of residues 3 to 146 and 154 to 190 of hGH, residues 29 to 54,
`59 to 72, and 79 to 234 of the first receptor, and residues 31 to 53, 61 to
`72, and 76 to 238 of the second receptor. No water molecules were
`added to the model.
`
`Diffraction data
`
`Sample
`
`Native 1
`Native 2
`K2PtCI4
`K2AuCI4
`
`Reso-
`lution
`(A)
`2.8
`2.8
`3.0
`3.0
`
`Measure-
`ments
`(No.)
`48635
`47414
`25316
`42964
`
`Reflec-
`tions
`(No.)
`17302
`18368
`14794
`14482
`
`Data
`cover-
`age (%)
`89
`95
`94
`92
`
`R.
`sym
`(on I)
`0.063
`0.061
`0.077
`0.067
`
`Phase refinement at resolution (A):
`3.0 Overall
`3.3
`3.7
`4.3
`5.0
`6.0
`
`10.0 7.5
`
`Native
`0.51
`0.79 0.79 0.74 0.65 0.58 0.47 0.43 0.39
`Figure of merit
`Reflections (No.) 316 601 976 1414 1916 2484 3165 3915 14787
`K2PtCl4
`0.71
`0.61 0.61 0.60 0.67 0.66 0.74 0.71 0.76
`Rcunlis*
`1.20
`0.93 1.22 1.46 1.30 1.21 1.19 1.16 1.10
`Phasing powert
`K2AuCI4
`RcunIS*
`Phasing powert
`
`0.51 0.56 0.51 0.62 0.68 0.78 0.77 0.72
`1.61 1.85 2.13 1.63 1.27 1.26 1.32 1.41
`
`0.66
`1.56
`
`R (I > 0)
`
`Resolution
`(A)
`
`Crystallographic refinement
`A(angle)
`R (I > 2ao)
`A(B
`A(bond)
`(A)
`(2
`(0)
`2.0
`3.6
`0.204 (15632)
`0.015
`0.228 (17985)
`10-2.8
`*RCRi,1: Cullis R factor for centric reflections.
`tPhasing power: mean value of
`heavy atom structure factor amplitude divided by residual lack of closure error.
`
`308
`
`Fig. 2. Ribbon representation of the structure of hGH, viewed as perpen-
`dicular to the four-helix bundle. The NH2-terminus is marked N, the
`COOH-terminus, C. Residues in the interfaces between the hormone and
`the two receptors are colored green (interface I) and blue (interface II),
`respectively, and selected interface residues are labeled; helix 1, 9 to 34; helix
`2, 72 to 92; helix 3, 106 to 128; and helix 4, 155 to 184. Additional short
`helical segments are 38 to 47, 64 to 70, and 94 to 100. The core of the
`four-helix bundle is formed by the side chains ofPhe', Ala'3, Ala'7, Leu20
`and Ala24 of helix 1; Leu76, Ser79, Be83, Trp86, and Val' of helix 2; Val"0,
`Leu"4, Leu"17, Ile'21, and Leu'24 of helix 3; and Phe'16, Asp'69, Met'70,
`Val'73, Leul77, and Vall'80 of helix 4. (Residues 1 and 2, 147 to 153, and
`191 are not visible in the electron density map and are not included in the
`model).
`
`SCIENCE, VOL. 255
`
`Novo Nordisk Ex. 2054, P. 3
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Downloaded from
`
`http://science.sciencemag.org/
`
`
`
`on September 17, 2020
`
`Fig. 4. Backbone structure of the hGH.(hGHbp)2 complex. The hormone is
`shown as yellow cylinders representing the helices connected by red tubes.
`The ,B strands of the binding proteins are shown in brown, the loops are
`green (hGHbp I) and blue (hGHbp II). The viewing direction is approxi-
`mately down the four-helix bundle of hGH. In this orientation, the
`COOH-termini of the extracellular domains, and therefore the cell mem-
`brane, are at the bottom. A rotation of 159°, followed by a translation of 8
`A, superimposes the two receptor molecules with an rms difference in Ccx of
`1.0 A (179 atoms). Superposition of the individual domains gives rms
`differences of 0.7 A for the NH2-terminal domain (74 atoms), and 0.9 A for
`the COOH-terminal domain (93 atoms).
`
`each having its COOH-terminus pointing away from the hormone
`in the direction where the membrane surface would presumably be.
`Intact receptors would have an additional eight residues between the
`COOH-terminus at the end of strand G of the hGHbp and the
`putative membrane-spanning helix. The structure suggests a model
`in which this eight-residue segment provides the flexibility and
`freedom of orientation needed for the hormone to bring together
`efficiently the extracellular domains.
`As a result of complex formation, some of the surface area is
`buried in the interfaces between hormone and receptor (Fig. 5). The
`receptor-binding sites on hGH (Figs. 2, 5, A and B, 6) are located
`on the faces of opposite sides of the four-helical bundle. The first
`binding site on hGH for the hGHbp (site I; color coded green in
`Fig. 2) has a concave character. It is formed by residues on exposed
`faces of mainly helix 4 but also of helix 1, of the four-helix bundle,
`together with residues in the connecting region between helices 1
`and 2. The total surface buried by the hormone on the receptor in
`this interface is about 1230 A2. The second binding site on hGH
`(site II) (Fig. 2) is made up of the exposed sides of helices 1 and
`3 and, in contrast to the concave character of site I, it is relatively
`flat. The NH2-terminal tail of hGH is extended, pointing away from
`the helical bundle, and contributes to site II (Fig. 2). The total
`surface buried in this interface is approximately 900 A2, and thus
`smaller by about 25 percent compared to interface I. A third region
`contributing to the stabilization of the complex is the contact surface
`between the membrane-proximal halves of the COOH-terminal
`domains of the receptors, which buries about 500 A2 on each
`receptor (see below). The ratio of the polar to the nonpolar atoms
`buried in the interfaces between hormone and receptors shows a
`small excess of polar surface, whereas the interface between the two
`receptors is more apolar (16).
`Although the overall shapes of the two binding sites on the
`309
`RESEARCH ARTICLE
`
`they are similar in their core, with a root-mean-square (rms)
`difference between corresponding Ca atoms of 1.1 A (41 Ca
`positions were examined).
`The NH2-terminal domain of the receptor contains three disulfide
`bridges (Fig. 3A), and the disulfide connections observed in the
`structure confirm the previous assignments made on the basis of
`chemical methods (6). Two of the disulfide bonds link neighboring
`strands. Thus, Cys38 in strand A is bridged to Cys48 in strand B with
`the disulfide packed in the interior between the two sheets, while
`strands F and G of the other sheet are linked by Cys'08 and Cys122,
`the disulfide in this case being exposed on the solvent-accessible side
`of the barrel. The third disulfide cross-links the two sheets of the
`sandwich, thereby connecting Cys83 in strand C' to Cys94 of strand
`E (Fig. 3). The loops between the strands that are disulfide-linked
`are relatively short (only 3 to 6 residues), whereas the other
`connections are longer (9 to 14 residues). Although two of the
`disulfides are part of the hydrophobic core of the NH2-terminal
`domain, their presence is apparently not required for the observed
`fold; the COOH-terminal domain, and domain D2 of PapD (15) do
`not have any disulfides, and domain D2 of CD4 has only one (14).
`The two domains of the hGHbp are linked by a four-residue
`segment that immediately follows strand G of the NH2-terminal
`domain. The main-chain torsion angles of these four residues are
`unusual for a linker between immunoglobulin-like domains in that
`they generate a helical turn (Vall25 and Asp126 have (p,j
`-70°, -200; Glu127 and Ile'28 have (p,1
`= -115°, 100). The result
`of this is that the relative orientation of the two domains is
`completely different from that found between the constant and
`variable domains of immunoglobulins. A salt bridge (2.9 A) be-
`tween Arg39 in the NH2-terminal domain and Asp'32 in the
`COOH-terminal domain may participate in stabilization of the
`relative orientation between the domains.
`Structure of the complex. The two receptor molecules in the
`hGH-(hGHbp)2 complex show apparent twofold symmetry about
`an axis approximately perpendicular to the helical axes of the hGH
`bundle (Fig. 4). The COOH-terminal domains are closely parallel,
`
`=
`
`BA
`
`G
`
`F CC
`
`B E
`
`N<~W10
`
`A
`
`,2 W
`
`1
`
`1
`
`6~~~~~~19
`
`N1443
`
`of the
`Structure
`Fig.
`3.
`hGHbp. (A) Ribbon repre-
`sentation of the backbone
`structure of the hGHbp. The
`termini are marked N and C.
`Both the NH2-terminal and
`the COOH-terminal domains
`contain seven P strands, divid-
`ed into two sheets. Residues
`involved in hormone binding
`are blue. Residues in the inter-
`face between the hGHbp I
`and hGHbp II are green. Se-
`lected side chains in the inter-
`faces are labeled. The position
`of the characteristic Trp-Ser-
`X-Trp-Ser pattern occurring in other members of the superfamily is gray. (B)
`Topology diagram of the domains of the hGHbp. Strands are labeled as
`described (14). A, B, and E belong to one sheet; C, C, F, and G to the other
`sheet. C' is significantly shorter than the other strands. (Amino acids not
`visible in the electron density map and not included in the current model are
`residues 1 to 28, 55 to 58, 73 to 78, and 235 to 238 of hGHbp I; and
`residues 1 to 30, 54 to 60, and 73 to 75 of hGHbp II.)
`
`W
`
`17 JANUARY 1992
`
`Novo Nordisk Ex. 2054, P. 4
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Downloaded from
`
`http://science.sciencemag.org/
`
`
`
`on September 17, 2020
`
`Table 2.
`
`Salt bridges and hydrogen bonds in intermolecular contact areas
`
`hGH-hGHbp I interface
`
`;hGH-hGHbp II interface
`
`hGHbp I-hGHbp H interface
`
`hGH
`atom
`Lys41 N;
`Gln'Ns2
`Pro61O
`Arg167Nn1
`Arg'67N'q2
`Lys168N;
`Asp171O82
`Thr"7UO^eyl
`Arg'78Nii2
`
`hGHbp
`atom
`Glu1270o2
`Glu1200,2
`Ile'03N
`Glu1270e1
`Glu'270 1
`Trp O4
`Arg43NTr2
`Arg43Ni 1
`ne'650
`
`Distance
`(A)
`2.9
`3.3
`2.9
`3.2
`2.9
`3.1
`3.1
`3.2
`2.9
`
`hGH
`atom
`Asn'2081
`Asn'2N82
`Arg16N 1
`Argl9Nii2
`
`hGHbp
`atom
`Arg43Nvi2
`Asp126082
`Glu4Os2
`Gln1660E1
`
`Distance
`(A)
`2.9
`3.0
`3.1
`3.0
`
`hGHbp I
`atom
`Ser'450y
`Leu146N
`Thrl470 y
`His's5Ne2
`Asp152082
`Ser2010y
`
`hGHbp II
`atom
`Asp152082
`Se9201Oy
`Asp'52O&1
`Asn143081
`Tyr200Oh
`Tyr000'i
`
`Distance
`(A)
`3.0
`3.1
`2.7
`2.9
`2.7
`3.3
`
`hormone are quite different, the residues on both receptors that
`interact with these sites are largely the same (Fig. 5, C and D). On
`both receptors, binding determinants in the NH2-terminal domain
`include Arg43 (on the loop between strands A and B), Trp'` (on
`the loop between strands E and F), and some residues on strand G
`immediately preceding the linker between the two domains. The
`Glu127 in the linker is part of the interface, as is the loop between
`strands B and C (notably Trp`69) in the COOH-terminal domain.
`The only receptor determinant that is different in both interfaces
`between hormone and receptors is Asn218 in interface I on the loop
`between strands F and G of the COOH-terminal domain of the
`hGHbp (Fig. 5B).
`Not only are the binding determinants on both receptors largely
`the same, but their structures are similar, as shown by an rms
`difference in CGa after superposition of 1.0 A (179 atoms).
`Because, overall, the receptors superimpose so well, it is possible
`that the linker between the NH2- and COOH-terminal domains is
`fairly rigid and confers a special orientation between them. The
`similarity in structure extends to the backbone of most of the
`binding determinants, and is even observed for the side chain
`conformations of many of the residues involved in interactions with
`the hormone, such as Arg43, Glu127, Trp'69, and Asn218. Excep-
`tions are the conformations of Trp1'
`and of the loop comprising
`residues 163 to 168. The difference in Cet position ofTrp'` is 2.8
`A, and the side chain orientation differs in the two receptors. Loop
`163 to 168 also takes on a different conformation, resulting in
`
`differences in Ca positions after superposition of 2 to 4 A.
`Many of the interactions in the binding sites are apolar; most of
`the hGH side chains that have binding functionality interact
`primarily through hydrophobic contacts. Examples are the van der
`Waals contacts between the methylene groups of Lys'68 and
`Lys172 of hGH with the side chain ofTrp`04 ofhGHbp I. In both
`interfaces, Trp104 of the receptors buries most surface area with a
`decrease in solvent accessibility of 170 A2 in site I and of more
`than 210 A2 in site II.
`The hydrogen bonds and salt bridges in the three intermolecular
`interfaces in the complex are shown in Table 2. The side chain of
`Arg43 of the hGHbp is involved in specific hydrogen-bonding
`interactions in both hormone-receptor interfaces (Table 2). It
`participates in a network of H bonds in site I (Figs. 1 and 6A) that
`includes Trp`0 ofhGHbp I and Asp'7' and Thr'75 of hGH. In site
`II, the cluster consists of Arg43 and Asp'26 of hGHbp II and Asn`2
`of hGH (Fig. 6B). Another residue with multiple interactions is
`Glu 27 of hGHbp I, which forms salt bridges to Lys4' and Arg167
`of hGH (Table 2). The total number of possible intermolecular salt
`bridges and hydrogen bonds in binding site I is 9, compared to only
`4 in binding site II (Table 2).
`The structure shows that hormone binding to the extracellular
`part of the receptor promotes association at the base of the
`COOH-terminal receptor domain, which is adjacent to the mem-
`brane. The contact area involved is between the three-stranded
`sheets of the COOH-terminal domains (Fig. 3A). Because of the
`
`A
`
`80J
`
`Y42
`
`hGH site l
`
`C189
`
`1
`
`B
`
`H1B
`
`II
`
`*
`~~~~~~~~~R178I
`15
`iii ~10
`
`D171
`
`10
`
`0
`
`50
`
`100
`
`150
`
`Residue number
`
`I
`
`Id
`
`60-
`40-
`20
`
`0
`
`20
`
`40
`
`60
`
`80
`
`310
`
`W104
`
`hGHbp I
`
`N218
`
`E127
`
`W169
`W8
`
`--I ~
`
`~
`
`~
`
`--
`
`---- ||-|1
`
`E127
`
`LW169
`
`K167
`
`hGHbp 11
`
`150
`
`200
`
`W104
`
`100
`
`Fig. 5. Decrease in solvent
`accessibility on complex for-
`mation. (A) Residues on the
`hormone: top, site I; and
`bottom, site II. (B) Resi-
`dues on the receptors: top,
`hGHbp I; bottom, hGHbp
`H. Solvent accessibility was
`calculated with the program
`written by Lee and Richards
`(24); a probe radius of 1.4 A
`was used.
`
`SCIENCE, VOL. 255
`
`Novo Nordisk Ex. 2054, P. 5
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Downloaded from
`
`http://science.sciencemag.org/
`
`
`
`on September 17, 2020
`
`tion of Pro"0, whereas other variants were affected much less (less
`than eight times). Overall, the results are again in good agreement
`with the interactions seen in the crystal structure of the complex. A
`notable exception is Arg43, mutation of which to alanine had little
`effect on binding of hGHbp I. Considering the network of interac-
`tions in which this residue pointing out (Figs. 1 and 6), it is hard to
`reconcile the differences in this instance, pointing out the difficulties
`in cross-referencing hormone and receptor binding determinants on
`the basis of mutational analysis.
`Signal transduction by the growth hormone receptor. Analysis
`ofthe composition of our crystals (7), biophysical measurements on
`the complex in solution (8), and mutational alterations for mapping
`the second receptor binding site on hGH (8) show that the growth
`
`approximate twofold symmetry in the complex, this interface is
`formed by the same residues of each receptor (Table 2). The
`segments buried in the interface are the very end of strand A, most
`of the loop between A and B, some residues on strand B, part of
`the loop between D and E, and three or four residues on strand E.
`In both cases, on the basis of surface area buried, Tyr200 contrib-
`utes most. Only about half of the side chains buried in the
`interface are hydrophobic; examples are Leul' and Ie'49 of
`hGHbp I, and Leul42 and Pro'98 of hGHbp II. Most of the
`hydrophilic side chains are involved in specific interactions; for
`example, Asp'52 of hGHbp II interacts with Ser'"4 (3.0 A) and
`Thr'47 (2.7 A) of hGHbp I; Asp'52 ofhGHbp I is close to Tyr200
`of hGHbp II (3.0 A).
`Comparison with mutational studies. The
`receptor binding determinants on hGH for site
`I have been mapped by means ofhomolog- and
`alanine-scanning mutagenesis (17). Binding site
`I was identified as a patch consisting of three
`discontinuous segments of hGH, the loop be-
`tween residues 54 and 74, the COOH-terminal
`half of helix 4 and, to a lesser extent, the
`NH2-terminal region of helix 1. Subsequent to
`that work, analysis of our crystals of the com-
`plex revealed the presence of the second
`hGHbp (7, 8). Mutational analysis was again
`used to identify this second binding site, show-
`ing it to consist of residues near the NH2-
`terminus and on the hydrophilic faces ofhelices
`1 and 3 (8). The three-dimensional structure of
`the complex confirms this interface region (Fig.
`2). From the structure, there is one additional
`segment of polypeptide chain that is part ofthe
`interface in binding site I, namely, the small
`piece of helix (residues 38 to 47) at the begin-
`ning ofloop 1 (Fig. 5). Since mutation ofthese
`residues did not have significant effts on
`binding ofhGHbp I, the interface in this region
`may not contribute significantly to the binding
`energy, or may be able to adjust to different side
`chains. On a residue by residue basis, the
`correspondence between the structure and the
`mutagenesis mapping is also good. Most ofthe
`residues identified by alanine scanning can be
`classified as direct binding determinants in that
`they are found in the hormone-receptor inter-
`face; the structure also shows that some muta-
`tions resulting in decreased binding probably
`interfere with the proper folding of the hor-
`mone (Phe'0, Phe54, Ie58, and Phe'76 in bind-
`ing site I). Changing Phel in binding site II to
`alanine reduced the binding affinity by a factor
`of 5 (8). From the structure, however, it is
`undear what the role of this amino acid side
`chain is since the NH2-terminal two residues
`cannot be seen in the electron density map.
`involving
`analysis
`A similar mutational
`changes of charged residues or selected tryp-
`tophans to alanine was applied to the hGHbp
`(18). By far, the largest decrease (2500 times)
`in hGH binding was observed for the change of
`Trp104 to alanine, while even the more con-
`served substitution to phenylalanine resulted 'm
`a large reduction (110 times) in binding. The
`next largest effect (84 times) was on substitu-
`
`interfaces between hormone and receptors. (A) Binding site I; (B) binding site
`Fig. 6. Close-up of
`presented by a space filling model, the receptors by a stick model. The hGH
`II. The hGH is rei
`cyan, side chain carbons are white, and side chain oxygens and nitrogens are red
`backbone atoms are
`and blue, respectiv
`rely. The receptor carbon atoms are in yellow, with red oxygens and blue
`residues are labeled.
`nitrogens. Selected
`
`17 JANUARY 1992
`
`RESEARCH ARTICLE
`
`311
`
`Novo Nordisk Ex. 2054, P. 6
`Mylan Institutional v. Novo Nordisk
`IPR2020-00324
`
`

`

`Downloaded from
`
`http://science.sciencemag.org/
`
`
`
`on September 17, 2020
`
`hormone-receptor complex has the form hGH.(hGHbp)2. The crystal
`structure ofthe complex reveals how the hormone, a nonsymmetrical
`molecule, binds two copies ofthe receptor that use essentially the same
`binding determinants. The difference in surface area between interfac-
`es I and II supports the sequential mechanism for receptor dimeriza-
`tion proposed by Cunningham et al. (8), who showed that the second
`receptor can only bind to hGH if the first receptor is already bound.
`This is consistent with the observation that the contact surface
`between receptor I and the hormone (1230 A2) is significan