Proc. Nadl. Acad. Sci. USA Vol. 88, pp. 4498-4502, May 1991 Biochemistry

A systematic mutational analysis of hormone-binding determinants in the human growth hormone receptor (hormone-receptor interactions/protein-protein interactions/site-directed mutagenesis)

STEVEN H. BASS, MICHAEL G. MULKERRIN, AND JAMES A. WELLS* Department of Protein Engineering, Genentech, Inc., 460 Pt. San Bruno Boulevard, South San Francisco, CA 94080

Communicated by George R. Stark, January 31, 1991

mutational strategy should be applicable to probing other hormone-receptor and protein-protein interactions for which little or no structural information is available.

A mutational strategy is presented that alABSTRACT lowed us to identify hormone-binding determinants in the extracellular portion of the human growth hormone receptor (hGHbp), a 238-residue protein with sequence homology to a number of cytokine receptors. By systematically replacing side chains with alanine we probed the importance of charged residues (49 total, typically located on the surface), aromatic residues (9 total), and neighbors of these (26 total). The alanine substitutions that were most disruptive to hormone binding are located predominantly in four segments of a cysteine-rich domain in the hGHbp, and collectively they form a patch when mapped upon a structural model proposed for cytokine receptors. Control experiments with monoclonal antibodies confirmed that most of these alanine substitutions do not disrupt the overall antigenic structure of the hGHbp. This highresolution functional analysis will complement structural studies and provides a powerful basis for evaluating and engineering the energetics of hormone-receptor interactions. Moreover, the hormone-binding determinants identified here may be similarly located in other, homologous, receptors.

MATERIALS AND METHODS Restriction enzymes, DNA polymerase I large fragment, T7 DNA polymerase, T4 DNA polymerase, and polynucleotide kinase were from New England Biolabs, Bethesda Research Laboratories, or United States Biochemical. Genentech provided hGH, synthetic oligonucleotides, and monoclonal antibodies (mAbs), except mAbs 5 and 265, which were purchased from Agen Biomedical (Parsippany, NJ). The plasmid phGHbp(1-238) (ref. 12), coding for residues 1-238 of the wild-type hGHbp sequence (7), contains the fl origin to allow preparation of single-stranded DNA for mutagenesis (13) and sequencing (14). Mutants of the hGHbp were expressed in E. coli and purified as described for the wild-type hGHbp (12). Briefly, cultures of E. coli KS330 (15) carrying plasmid phGHbp(1-238) secreted properly folded and processed hGHbp into the periplasm. Periplasmic proteins were released from the cells by a freeze-thaw and extraction into hypotonic buffer (10 mM Tris HCI, pH 7.5) containing 2 mM EDTA and 1 mM phenylmethylsulfonyl fluoride to inhibit metallo- and serine proteases, respectively. The hGHbp was precipitated with ammonium sulfate (45% saturation), resuspended in the same buffer, and clarified by centrifugation. At this point the affinity of each hGHbp mutant for hGH was determined. Controls show that contaminating proteins from E. coli do not interfere with the hGH binding assay (5). Mutant binding proteins with decreased binding affinity to hGH were purified to homogeneity as previously described (12) except for W104A, which was purified by Q-Sepharose chromatography (Pharmacia) followed by a mono-Q column (Pharmacia). Purified binding proteins were dialyzed against 10 mM Tris HCI, pH 7.5/100 mM NaCI for circular dichroic spectra. Protein concentrations were determined from the absorbance spectrum using A`1t° = 2.35 for hGHbp (12) and AO-'% = 2.11 for W104A and W104F. Spectra were obtained on an Aviv Associates (Lakewood, NJ) Cary 60 spectropolarimeter in the near-UV (320-250 nm) at a 0.5-nm interval and in the far-UV (250-190 nm) at a 0.2-nm interval. All spectra were averaged over 5 scans with a 2-sec averaging time for each datum. Binding constants were determined from competitive displacement assays using 125I-labeled hGH as a tracer (16). An anti-receptor mAb (mnAb 263; ref. 17) and 15% (vol/vol)

Human growth hormone (hGH) is homologous to a large family of hormones that includes prolactins, placental lactogens, and proliferins (for review see ref. 1). Collectively, these hormones regulate a vast array of physiological effects, including growth, lactation, differentiation, and electrolyte balance (for reviews see refs. 2 and 3). These biological effects begin with the binding of hormone to specific cellular receptors. Systematic mutational studies have revealed functionally important residues in hGH for binding to the hGH receptor (4-6). In contrast, virtually nothing is known of the hormone-binding determinants in the hGH receptor. The hGH receptor cloned from liver (7) consists of a single polypeptide chain (620 residues total) containing an extracellular hormone-binding segment (246 residues), a single transmembrane region (23 residues), and a cytoplasmic segment (351 residues). The extracellular portion of the hGH receptor is found naturally in the blood stream (8) as an hGH-binding protein (hGHbp). Recent comparative sequence analyses suggest that the hGHbp is structurally related to a large family of cytokine receptors (9-11). The hGHbp (containing residues 1-238) has been expressed in Escherichia coli in large quantities (12), and it retains the same high affinity for hGH as its natural glycosylated counterpart. Here, using a combination of mutational and biophysical analysis, we have identified important hormone-binding determinants in the hGHbp. These determinants are chemically complementary to those previously identified in hGH, and they map predominantly to a cysteinerich region of the receptor and to loop regions on one side of a structural model proposed for cytokine receptors (10). This

Abbreviations: hGH, human growth hormone; hGHbp, human growth hormone-binding protein; mAb, monoclonal antibody. Single mutants are designated by the wild-type residue (single-letter amino acid code) followed by its sequence position and the mutant residue. For example, F96S indicates a mutation in which Phe-% is converted

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.

to Ser.

*To whom reprint requests should be addressed.

4498

Biochemistry: Bass et al. polyethylene glycol were used to precipitate the binding protein hormone complex after equilibration overnight at 250C (12). Under these conditions hGH and the hGHbp form a one-to-one complex.

RESULTS Binding of Natural Variants of the hGHbp to hGH. Northern analysis and cDNA cloning from human hepatocytes producing the hGH receptor revealed a messenger RNA that is lacking the exon 3 sequence encoding residues 7-28 (ref. 18; W. Wood, personal communication). We constructed a binding protein missing residues 7-28. This protein has the same affinity for 125I-labeled hGH as does full-length hGHbp (Table 1), suggesting these amino-terminal residues are not essential for hGH binding. Goossens and coworkers (19) sequenced the coding region of the hGH receptor from one family with Laron-type dwarfism and found a single phenylalanine-to-serine substitution at position 96 (F96S) in the extracellular domain of the hGH receptor. They postulated this mutation resulted in production of a receptor defective for hGH binding. We produced the F96S mutant hGHbp in E. coli and found it to have a Kd for hGH that was identical to that of the wild-type hGHbp (ref. 20; Table 1), suggesting that this defect does not affect hGH binding directly. Scanning-Mutational Analysis of Charge Clusters. To determine if charged residues of the hGHbp are involved in hGH binding, site-directed mutagenesis was used to replace systematically all arginine, lysine, aspartic, and glutamic residues with alanine. We substituted charged side chains with alanine because in the absence of other structural information we anticipated charged residues to be exposed to solvent and the alanine substitutions to cause minimal perturbation in the folding of the protein. Alanine, the most common amino acid (21), is small and is found in both buried and exposed positions in proteins (for reviews see refs. 22 and 23). The charged residues were mutated in clusters of 2-5 residues to maximize the efficiency of our analysis. Fourteen clustered charge-to-alanine mutants were prepared that collectively mutated all of the charged residues (except Lys-110) from position 31 to 217 in the hGHbp (Table 1). Eight of these multiple mutants expressed sufficient quantities in E. coli to be analyzed, and two of them (RRE 70-75 and KDKEE 203-209) showed reductions in binding affinity to hGH of about 6-fold. We next dissected the six nonexpressing and two disruptive binding variants by producing single-alanine mutants (30 total) in these segments (Table 1). Of these mutants, only one (D85A) could not be expressed in quantities that allowed analysis. Nine of the single-alanine mutants, E42A, E44A, R70A, R71A, E82A, D126A, E127A, D132A, and K215A, caused reductions in binding affinity to hGH of 2- to 8-fold. The disruptive effects in the RRE 70-75 mutant can be accounted for by R70A and R71A. In contrast, dissection of the pentamutant KDKEE 203-209 did not single out a single residue prominent in disrupting binding, although slight effects were observed for K203A and K206A. We interpret these data to mean that side chains in this charged segment are not substantially involved in binding to hGH. Alanine-Scanning of Tryptophans and Tyrosines in the Cysteine-Rich Domain. Fluorescence experiments show that binding of hGH to the hGHbp causes changes in the tryptophan fluorescence spectra of the hGHbp (M.G.M., unpublished results). The charged-to-alanine-scan data indicate that the strongest binding interactions with hGH are located in the cysteine-rich region of the hGHbp (Fig. 1). Of the four tryptophan residues in this region (Trp-50, Trp-76, Trp-80, and Trp-104), only the WSOA mutant could not be expressed

Proc. Natl. Acad. Sci. USA 88 (1991)

4499

Table 1. Dissociation constants for binding of mutants of the hGHbp to hGH hGHbp hGHbp Kd, nM mutant mutant Kd, nM Scan of Natural variants Trp and Wild type 0.40 ± 0.05 Tyr 0.42 ± 0.11 A7-28* NE WSOA 0.32 ± 0.05 F96S 1.03 ± 0.18 W76A Clustered charge0.53 ± 0.08 W80A to-Ala scant >1000 W104A 0.60 ± 0.04 KEKK 31-37 44 ± 0.4 NE W104F RERE 39-44 NE Y68A 0.56 ± 0.13 DEK 52-59 NE Y86A 2.03 ± 0.10 RRE 70-75 0.56 ± 0.06 Y9SA NE EKE 79-82 NE Y107A NE DE 85-91 0.57 ± 0.25 DEK 119-121 NE Scan near DED 126-132 charged DRE 152-158 0.68 ± 0.13 and Trp RDK 161-167 0.36 ± 0.06 determinants NE EEKE 173-180 NE S40A EKKD 183-190 0.49 ± 0.15 0.62 ± 0.12 P41A KDKEE 203-209 2.65 ± 1.07 NE T45A NE RRKR 211-217 NE F46A Charge-to-Ala scan S47A 0.39 ± 0.07 R39A 0.63 ± 0.03 E42A 3.20 ± 0.01 0.45 ± 0.06 T5lAt 0.64 ± 0.04 R43A 1.20 ± 0.13 E44A 0.19 ± 0.06 N72A 0.81 ± 0.10 R70A 0.19 ± 0.04 T73A 1.10 ± 0.04 R71A 0.15 ± 0.03 Q74A 0.35 ± 0.04 E75A 0.26 ± 0.03 T77A 0.35 ± 0.04 E79A 0.20 ± 0.03 0.51 ± 0.05 K81A Q78A 1.50 ± 0.09 E82A NE F96A NE D85A 0.26 ± 0.04 N97A 0.54 ± 0.06 E91A 0.23 ± 0.04 S98A 0.43 ± 0.01 K11OA 0.17 ± 0.02 S99A 2.10 ± 0.37 D126A F100A 0.39 ± 0.02 0.80 ± 0.15 E127A 2.5 ± 0.17 2.50 ± 0.08 TlOlA D132A 0.88 ± 0.02 S102A 0.46 ± 0.14 E173A 0.83 ± 0.07 1103A 0.35 ± 0.14 E175A 0.50 ± 0.14 I1OSA 0.40 ± 0.02 K179A 33.7 ± 16 P106A 0.56 ± 0.05 E180A 0.60 ± 0.20 K203A NE S124A 0.37 ± 0.12 D205A 2.75 ± 0.04 V125A K206A 0.56 ± 0.02 1.08 ± 0.13 1128A 0.35 ± 0.04 E207A 0.37 ± 0.01 V129A 0.36 ± 0.04 E209A 0.44 ± 0.09 Q130A 0.38 ± 0.05 R211A R213A 0.29 ± 0.06 0.93 ± 0.14 K215A K217A 0.70 ± 0.03 0.48 ± 0.01 E224A Mutants were prepared by site-directed mutagenesis (13), expressed in E. coli, and assayed as described in Materials and Methods. Results are presented as mean ± SEM. NE, not expressed in levels suitable for analysis. *A7-28 indicates a deletion mutant of exon 3 in which residues from Thr-7 to Asn-28 were deleted. tThe KEKK 31-37 mutant, for example, is a tetraalanine mutant in which Lys-31, Glu-32, Lys-34, and Lys-37 have all been mutated to Ala. This nomenclature applies to all of the clustered charge-toalanine variants. tTaken from ref. 12.

in amounts sufficient for binding analysis (Table 1). The W80A mutant caused no significant reduction in binding affinity, whereas W76A caused a 3-fold reduction in affinity.

Biochemistry: Bass et al.

4500

Proc. Natl. Acad. Sci. USA 88 (1991)

8-

D6(D 6

E

4

oI

c

-

-

2-

-1 ----

1 -----------A

DI

-

20

10

30

>2,500-*jj

-I

40

e5

I

m

-N

----- ---------

50

60

w --------0I 70

-84

8-

-D(

4-

sCc

-02-

E-

-

Iac

L IA ------ II ----- 11 ----------------a ----- OMM -1 W KECPDYVSAGENSCYFNSSFTSIWIPYCIKLTSNGGT'VDEKC FSVDEIVOPDPPIALNWTLLNVSLTGIHADIOVRWE Al-

----------

0v

80

E0 EIc:I 0

100

Fe

110

130

S-

140

150

84-6

-§ 2APRNADIOKGWMVLEYELOYKEVNETKWKMMDPILTTSVPVYSLKVDKEYEVRVRSKORNSGNYGEFSEVLYVTLPOMSO 160

170

180

190

200

210

220

230

Residue

In contrast, we could not detect any binding to the W104A mutant (reduced more than 2500-fold) and a more conservative variant (W104F) was reduced in binding by 110-fold. These reductions in binding affinity for mutating Trp-104 are by far the largest in the entire set. Such a large change in binding could be a result of structural change in hGHbp. To assess the structural effects of these mutants we determined the circular dichroic (CD) spectra for the mutants W104A and W104F and wild-type hGHbp (Fig. 2). Both the near- and the far-UV CD spectra are similar for each of these proteins. A comparison of the near-UV CD spectra of the wild type with the spectra of both mutants indicates that the substitution of the tryptophan results in virtually no change except for an expected decrease in the intensity of the tryptophan spectrum. The far-UV CD shows some alterations in intensity but the overall shape appears to be the same for each of the proteins, indicating they are not grossly misfolded. lodination of the hGHbp disrupts binding to hGH (J.A.W., unpublished results). We therefore mutated the four tyrosines to alanine in the cysteine-rich domain including Tyr-68, Tyr-86, Tyr-95, and Tyr-107. However, only Y95A could be expressed in amounts sufficient for analysis, and it yielded a dissociation constant essentially the same as that of the wild-type hGHbp (Table 1). Although Tyr-95 does not appear to be involved in binding to hGH, we do not know about the importance of the others. Alanine-Scanning of Residues Flanking Disruptive Charged and Tryptophan Residues. Protein-protein interfaces often contain several binding segments each contributing more than one side-chain determinant of hydrophobic or hydrophilic character (for reviews see refs. 24 and 25). Thus, we replaced with alanine the neutral and hydrophobic residues that closely flank those charged and tryptophan residues found to be most disruptive to binding (Table 1). Within the first disulfide loop no additional residues were identified beyond Glu-42 and Glu-44 that disrupted binding; however, we were unable to express S40A, T45A, and F46A in quantities suitable for analysis. In the region around Arg-70 and Trp-76, no other disruptive alanine mutants were identified. However, we identified four mutants, N72A, T73A, Q74A,

FIG. 1. Histogram showing the effect upon binding to hGH for alanine substitutions scanned over the hGHbp (residues 31-224). Values of Kd (mutant)/Kd (wild-type hGHbp) were calculated from data in Table 1. Sites of alanine substitutions causing effects ranging from a 2-fold to >2500-fold decrease in affinity cluster in four segments in the cysteinerich domain. The disulfide pairings (-S-S-) are taken from Fuh et al. (12). Single alanine replacements are indicated by single bars, and clustered alanine variants and the A7-28 mutant are indicated by linked bars.

and Q78A, that slightly increased affinity for hGH. The region flanking Trp-104 contained four other residues, Thr101, Ser-102, Ile-103, and Pro-106, where alanine substitutions caused a loss in binding affinity of 2- to 85-fold, and one (S99A) that slightly increased affinity, about 2-fold. Finally, the region from Asp-126 to Asp-132 contained two additional disruptive alanine mutants, V125A and I128A. Binding of mAbs to Mutants of hGHbp. To evaluate the structural integrity of hGHbp mutants that were reduced in affinity for hGH, we analyzed their binding properties with a panel of six different anti-hGHbp mAbs, whose affinities ranged from about 0.5 to 6 nM (Table 2). These mAbs were prepared by immunizing with the natural rabbit or rat, growth hormone receptors (mAb 5, ref. 26; and mAb 263, ref. 17), or

Cf)

O x

0 -2 -4 -6 -8

hA 200

210

220

230

2 240

50 0

3:

-50

2 -100

-150

-200 260 270 280 290 300 310 Wavelength (nm)

) and the mutants FIG. 2. CD spectra of wild-type hGHbp ( W104A (-.-----) and W104F ( ). CD spectra were obtained on solutions of -1 mg/ml in 10 mM Tris-HCI, pH 7.5/100 mM NaCl at 20°C. (A) Far-UV CD spectra of proteins in a 0.01-cm cell. (B) Near-UV CD spectra in a 1.0-cm cell. The units of mean residue ellipticity, [O]MRw, are degrees cm2-decimole-1.

Biochemistry: Bass et al.

Proc. Natl. Acad. Sci. USA 88 (1991)

4501

Table 2. EC50 for binding of six anti-hGHbp mAbs to the wild-type and variants of the hGHbp that disrupt binding of hGH mAb EC50, nM 5 hGHbp mutant 263 3B7 3D9 12B8 13E1 Wild type 0.7 ± 0.2 0.6 ± 0.1 0.5 ± 0.4 2.2 ± 1.0 6.1 ± 1.8 3.2 ± 1.7 E42A 0.5 ± 0.1 0.5 ± 0.3 0.2 ± 0 1.7 ± 0.1 2.5 ± 1.8 0.8 ± 0.3 E44A 1.7 ± 1.0 0.7 ± 0.3 0.8 ± 0.3 3.3 ± 2.2 3.3 ± 0.9 2.7 ± 0.5 E82A 1.0 ± 0.6 5.3 ± 1.9 0.7 ± 0.4 2.9 ± 2.1 1.5 ± 0.4 0.9 ± 0.2 T1OlA 1.7 ± 0.4 0.7 ± 0.06 >100 >100 14.0 ± 2.1 8.2 ± 2.0 W104A 1.3 ± 0.3 0.6 ± 0.07 0.7 ± 0.07 2.2 ± 0.07 5.4 ± 1.6 3.1 ± 0.6 P106A 2.8 ± 0.8 0.8 ± 0.4 1.0 ± 0.2 8.5 ± 0.9 33.4 ± 21 7.8 ± 0.3 V125A 0.8 ± 0.2 0.4 ± 0.06 0.5 ± 0.1 3.2 ± 1.2 2.4 ± 0.3 0.9 ± 0.3 D126A 1.0 ± 0.1 0.6 ± 0.1 0.7 ± 0.1 3.9 ± 1.6 1.7 ± 0.9 3.9 ± 0.4 1.4 ± 0.1 1128A 0.4 ± 0.1 0.7 ± 0.1 6.4 ± 0.8 5.7 ± 0.7 0.9 ± 0.1 D132A 1.5 ± 0.5 0.7 ± 0.06 0.6 ± 0.06 4.2 ± 1.8 15.5 ± 7.4 5.1 ± 1.9 Blocks hGH binding No No No No Yes Yes An ELISA format was used in which microtiter plates were coated with mAb, then incubated with 3-fold serial dilutions of the purified hGHbp mutant. The extent of binding was quantified by incubating with purified rabbit polyclonal anti-hGHbp antibody, then a horseradish peroxidase-conjugated goat anti-rabbit IgG antibody. Results are mean ± SEM.

with the native hGHbp expressed and purified from E. coli (mAbs 3B7, 3D9, 12B8, and 13E1). Most of the hGHbp mutants reacted with all of the mAbs as well as the wild-type hGHbp (Table 2). However, the TlOlA variant was reduced in binding by >200- and 45-fold to mAbs 3B7 and 3D9, respectively, and by about 2-fold with three out of the remaining four mAbs. Similarly, the binding affinity of the P106A mutant to the same five mAbs is lowered by 2- to 5-fold. It is unlikely that all of these effects are mediated by direct effects that the TiOlA and P106A mutations have on the antibody binding epitopes. Typically, direct effects show up as a highly selective disruption pattern (4, 5). For example, the E82A variant is reduced in binding affinity to mAb 263 by 9-fold, yet it is virtually unchanged in binding to the other five mAbs. Of the mAbs tested 13E1 and 12B8 completely block binding of hGH to the hGHbp. The fact that most of the alanine mutants that disrupt receptor binding do not disrupt binding of the mAbs that block hGH binding is consistent with these being overlapping but nonidentical, as was typically seen for hGH (4, 5).

from these putative loop regions have virtually no effect upon binding. Our data do not support the hormone-binding site that was predicted (10) to span the G-,8 strand of the N-terminal domain (residues 119-125) and the D-,3 strand in the C-terminal domain (residues 184-195). Moreover, the CD spectra (Fig. 2) do not substantiate the extensive (3-sheet structure predicted for the hGHbp. The far-UV CD spectrum (Fig. 2A) is similar to the spectrum of the kringle of tissue plasminogen activator (27), which is composed mainly of turns or loops that are held together by disulfide crosslinks. The CD spectrum of the hGHbp is different from that of the T-cell receptor, CD4 (28), which has a (-sandwich type fold exhibiting -50%,p-sheet structure overall (29, 30). However, we note that p-sheet content is more difficult to quantify accurately than a-helix by CD measurements and that the aroP

Q~P7

-

.P T

DISCUSSION We have presented a scheme to systematically identify hormone-binding determinants where little structural information was available. This analysis has identified four segments located in the cysteine-rich domain of the hGHbp that are important to binding of hGH (Fig. 1). In order of decreasing importance to binding these segments include residues Thr-101 to Pro-106 >> Val-125 to Asp-132 > Arg-70 to Glu-82 Glu-42 to Glu-44. The segments Thr-101 to Pro-106 and Val-125 to Asp-132 should be in close proximity because they are bridged by the Cys-108 Cys-122 disulfide. Recent sequence analyses suggest that the extracellular binding domains of the growth hormone and prolactin receptors are structurally homologous to a large number of cytokine and other receptors including the interleukin (IL)-2, IL-3, IL-4, IL-6, IL-7, erythropoietin, and granulocyte-macrophage colony-stimulating factor (GM-CSF) receptors (refs. 9-11 and references therein). Bazan (10) predicted that the extracellular portions of these receptors contain two immunoglobin-like domains, each consisting of seven P-strands and connecting loops. When the alanine mutants that cause a 2-fold or greater effect upon binding to hGH are placed upon the topographic map predicted for the cysteine-rich domain of the hGHbp (Fig. 3) they form a patch in the four loops at one end of the model. By comparison, residues altered away

100;

s

y

F

.

40 .S

F.

,S

,Q

F

Y.

R

S

.S

,E

L

C-S-S-C

C -S-S- C

N

W80 K

Q

K

H

F

K K 110 ,E 120

T

oY W 50

P

L

F

T

C -S-S- C

G

T

V

K

D

S

P

L

S

T

N

:D~.,N 60

P

E -

3' K

N

-

_.

.EE.,H 4

A

B

E

N> G

.

G

Y- T

go Go

A

.

..

S

D

C

F

G

FIG. 3. Location of residues in the hGHbp causing a 2-fold or greater effect upon binding to hGH mapped upon an immunoglobulin-like folding diagram predicted for the cysteine-rich domain (10). Seven antiparallel 13-strands (A-G) compose a ,3-sandwich by folding the G strand around the back until it hydrogen bonds to the A strand. Residues causing less than a 2-fold reduction (0), a 2- to 4-fold reduction (0), a 4- to 10-fold reduction (0), a 10- to 100-fold reduction (0), and >100-fold reduction (*) in binding are indicated. Those residues that gave low expression yields are indicated by C). Disulfides are shown by -S-S- connecting cysteine residues, and residues are numbered by tens.

4502

Biochemistry: Bass et al.

matic residues and disulfides can contribute substantially to the far-UV CD spectrum. Thus, our data are consistent with the positioning of the loops in the structure predicted for cytokine receptors (10) but cannot confirm the degree of

,a-sheet structure predicted.

Several lines of evidence suggest that the changes in binding affinity for the hGHbp mutants are not a result of the proteins being grossly misfolded. First, these proteins can be expressed at levels comparable to wild-type hGHbp. Poor expression can be correlated with an unstable or misfolded protein (31). Second, mAb binding to most of the mutant proteins parallels the binding found for wild-type hGHbp. We do not know that the mAbs raised to native hGHbp react only with the native hGHbp, because denatured or reduced hGHbp aggregates in solution (data not shown). However, structural analyses of antibody-antigen complexes (for review see ref. 32) and epitope mapping of hGH (4) and CD4 (29, 30) show that virtually all of these epitopes are highly discontinuous. Third, the CD spectra in both the near- and far-UV of the W104A and the W104F mutant are similar to the spectrum of wild-type hGH (Fig. 2). The fact that the W104F mutation is reduced in affinity by 100-fold, whereas the W104A mutant is reduced >2500-fold, demonstrates the keen requirement for a large hydrophobic side chain at position 104 for tight binding to hGH. Although we do not believe the mutations cause enormous changes in the folding of the mutant proteins, it is possible that for some of these mutants (perhaps T1OlA or P106A) the reduction in binding affinity results from indirect structural effects as opposed to making direct side-chain contacts with hGH. It is notable that neither of these later mutants cause substantial changes in the CD spectra (data not shown), suggesting the CD in this case is less sensitive than the mAb binding analysis. The combined mutational analyses of hGH (5) and the hGHbp show there is a great deal of functional complementarity at the interface. Of the 10 charged-to-alanine substitutions in the hGHbp causing a 2-fold or greater reduction in hormone binding affinity, 7 are acidic residues; any one of the top 5 of these (E42A, E44A, E82A, D126A, and D132A) is more disruptive than any of the three disruptive basic residues (R70A, R71A, and K215A). By comparison, alaninescanning mutagenesis of hGH has revealed that 3 of the top 5 disruptive binding mutants are basic residues (R64A, K172A, R178A) and none are acidic residues. Thus, there is electrostatic complementarity between an electropositive binding epitope on hGH and an electronegative epitope on the hGHbp. Moreover, both epitopes contain important hydrophobic determinants as well. The distribution of strong and weak binding determinants is not identical between the epitope on hGH as compared to the hGHbp. For example, of the 17 disruptive alanine mutants in the hGHbp, 10 cause 2- to 4-fold reductions in binding, 5 cause 4- to 10-fold reductions, 1 residue causes a 10- to 100-fold disruption, and 1 causes greater than 1000-fold reduction in affinity for hGH. By comparison, of the 20 disruptive alanine mutants in hGH, 8 cause 2- to 4-fold reductions, 7 cause 4- to 10-fold reductions, 5 cause 10- to 100-fold reductions, and none cause over 100-fold reduction in binding to the hGHbp (5, 6). These data suggest that there is not a simple one-to-one side-chain to side-chain interaction between hGH and the hGHbp. In conclusion, the mutational analyses provide a functional map of side chains in the hGHbp important for binding of hGH. We cannot be sure that additional side-chain binding determinants are present because not all residues were tested. A high-resolution structure of the hormone-receptor complex will be necessary to determine if these residues

Proc. Natl. Acad. Sci. USA 88 (1991)

modulate binding by direct or indirect effects and to define main-chain interactions. The functional data provided here, when combined with high-resolution structural information, should provide a solid basis from which one may begin to rationally design small-molecule mimics of hGH. We are grateful to Dr. Brian Fendley and Marcie Winget for providing mAbs; to Peter Ng, Parkash Jhurani, and Mark Vasser for synthesis of oligonucleotides; to Dr. Brad Snedecor, Mike Brochier, and Michael Covarrubias for fermentations; to Dr. William Wood for helpful suggestions; and to Wayne Anstine for graphics. 1. Nicoll, C. S., Mayer, G. L. & Russell, S. M. (1986) Endocrine Rev. 7, 169-203. 2. Chawla, R. K., Parks, J. S. & Rudman, D. (1983) Annu. Rev. Med. 34, 519-547. 3. Isaksson, O., Eclen, S. & Jansson, J. 0. (1985) Annu. Rev. Physiol. 47, 483-499. 4. Cunningham, B. C., Jhurani, P., Ng, P. & Wells, J. A. (1989) Science 243, 1330-1335. 5. Cunningham, B. C. & Wells, J. A. (1989) Science 244, 1081-1085. 6. Cunningham, B. C., Henner, D. J. & Wells, J. A. (1990) Science 247, 1461-1465. 7. Leung, D. W., Spencer, S. A., Cachianes, G., Hammonds, R. G., Collins, C., Henzel, W. J., Barnard, R., Waters, M. J. & Wood, W. I. (1987) Nature (London) 330, 537-543. 8. Baumann, G., Stolar, M. W., Ambara, K., Barsano, C. P. & DeVries, B. C. (1986) J. Clin. Endocrinol. Metab. 62, 134-141. 9. 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. 10. Bazan, J. F. (1990) Proc. Natl. Acad. Sci. USA 87, 6934-6938. 11. Patthy, L. (1990) Cell 61, 13-19. 12. Fuh, G., Mulkerrin, M. G., Bass, S., McFarland, N., Brochier, M., Bourell, J. H., Light, D. R. & Wells, J. A. (1990) J. Biol. Chem.

265, 3111-3115. 13. Kunkel, T. A., Roberts, J. D. & Zakour, R. A. (1987) Methods Enzymol. 154, 367-382. 14. Sanger, F., Nicklen, S. & Coulson, A. R. (1977) Proc. Natl. Acad. Sci. USA 74, 5463-5467. 15. Strauch, K. L. & Beckwith, J. (1988) Proc. Nat!. Acad. Sci. USA 85, 1576-1580. 16. Spencer, S. A., Hammonds, R. G., Henzel, W. J., Rodriguez, H., Waters, M. J. & Wood, W. I. (1988) J. Biol. Chem. 263,7862-7867. 17. Barnard, R., Bundesen, P. G., Rylatt, D. B. & Waters, M. J. (1985) Biochem. J. 231, 459-468. 18. Godowski, P. J., Leung, D. W., Meacham, L. R., Galgani, J. P., Hellmiss, R., Keret, R., Rotwein, P. S., Parks, J. S., Laron, Z. & Wood, W. I. (1989) Proc. Natl. Acad. Sci. USA 86, 8083-8087. 19. Amselem, S., Duquesnoy, P., Attrec, O., Novelli, G., Bousnina, S., Postel-Vinay, M.-C. & Goossens, M. (1989) N. Engl. J. Med. 321, 989-995.

20. Bass, S. & Wells, J. A. (1990) N. Engl. J. Med. 322, 834. 21. Klapper, M. H. (1977) Biochem. Biophys. Res. Commun. 78, 10181024. 22. Chothia, C. (1976) J. Mol. Biol. 105, 1-14. 23. Rose, G. D., Geselowitz, A. R., Lesser, G. L., Lee, R. H. & Zehfus, M. H. (1985) Science 229, 834-838. 24. Argos, P. (1988) Protein Eng. 2, 101-113. 25. Janin, J., Miller, S. & Chothia, C. (1988) J. Mol. Biol. 204,155-164. 26. Barnard, R., Bundeson, P. G., Rylatt, D. B. & Waters, M. J. (1984) Endocrinology 115, 1805-1813. 27. Cleary, S., Mulkerrin, M. G. & Kelley, R. F. (1989) Biochemistry 28, 1884-1891. 28. Chamow, S. M., Peers, D. H., Bryon, R. A., Mulkerrin, M. G., Harris, R. J., Wang, W.-C., Bjorkman, P. J., Capon, D. J. & Ashkenazi, A. (1990) Biochemistry 29, 9885-9891. 29. Wang, J., Yan, Y., Garrett, T. P. J., Liu, J., Rodgers, D. W., Garlick, R. L., Tarr, G. E., Husain, Y., Reinherz, E. L. & Harrison, S. C. (1990) Nature (London) 348, 411-418. 30. Ryu, S.-E., Kwong, P. D., Truneh, A., Porter, T. G., Arthos, J., Rosenberg, M., Dai, X., Xuong, N., Axel, R., Sweet, R. & Hendrickson, W. A. (1990) Nature (London) 348, 419-426. 31. Pakula, A. A., Young, V. B. & Sauer, R. T. (1986) Proc. Natl. Acad. Sci. USA 83, 8829-8833. 32. Davies, D. A., Padlan, E. A. & Sheriff, S. (1990) Annu. Rev. Biochem. 59, 439-473.