Influence of subunit interactions on lutropin specificity. Implications for studies of glycoprotein hormone function.

Bovine lutropin (bLH) and human chorionic gonadotropin (hCG) are heterodimeric glycoprotein hormones required for reproduction. Both bind rat LH receptors (rLHRs), but hCG binds human LH receptors (hLHRs) 1000-10,000 fold better than bLH. We tested the premise that this difference in affinity could be used to identify lutropin receptor contacts. Heterodimers containing hCG/bLH α- or β-subunit chimeras that bound hLHR like hCG (or bLH) were expected to have hCG (or bLH) residues at the receptor contact sites. Analogs containing one subunit derived from hCG bound hLHR much more like hCG than bLH, indicating that each bLH subunit contains all the residues sufficient for high affinity hLHR binding. Indeed, the presence of bovine α-subunit residues increased the activities of some hCG analogs. The low hLHR activity of bLH was due primarily to an interaction between its α-subunit and β-subunit residue Leu95. Leu95 does not appear to contact the hLHR since it did not influence the hLHR activity of heterodimers containing human α-subunit. These observations show that interactions within and between the subunits can significantly influence the activities of lutropins, thereby confounding efforts to identify ligand residues that contact these receptors.

The gonadotropins human lutropin (hLH), 1 human chorionic gonadotropin (hCG), and human follitropin (hFSH) are essential for reproduction and have been used for many years to enhance human fertility. Development of clinically useful agonist and antagonist analogs would be facilitated by knowledge of how these ligands interacted with their receptors. Two radically different models of gonadotropin-receptor interaction have been proposed (1, 2) based on the crystal structures of hCG (3,4) and ribonuclease inhibitor (5, 6), a protein containing a leucine-rich repeat motif thought to be similar to those in the glycoprotein hormone receptors. These models could be readily distinguished if the portions of the hormone that contacted their receptors were known.
Like other glycoprotein hormones, the gonadotropins are heterodimers that contain a conserved ␣-subunit and a hormone-specific ␤-subunit (7). Each subunit is divided into three large loops by a cysteine knot (3,4), and the heterodimer is stabilized by a portion of the ␤-subunit termed the "seat belt" (3) that is wrapped around ␣-subunit loop 2. Based on the activities of chemically and enzymatically modified hormones (summarized by Pierce and Parsons (7)), synthetic hormone fragments (8 -12), and analogs prepared by site-directed mutagenesis (13)(14)(15)(16)(17)(18)(19), residues throughout both hormone subunits have been suggested to participate in essential high affinity hormone receptor contacts. Surprisingly, some hCG residues proposed to contact LH receptors are in regions recognized by monoclonal antibodies that bind to hCG-receptor complexes (1,20). Others thought to be essential for receptor contacts can be replaced without disrupting receptor binding. For example, replacing hCG seat belt residues 101-109 with their hFSH counterparts led to analogs that bound FSH receptors with high affinity (21,22), even though they contained hCG ␤-subunit residues at positions thought to interact with the FSH receptor. These findings raise signficant questions about the identities of residues that have been proposed to be responsible for high affinity receptor contacts.
Studies described in this report were initiated with the goal of identifying residues in hCG that are essential for high affinity contacts with the human LH receptor. The heterodimeric nature of the glycoprotein hormones confounds efforts to identify residues needed for essential high affinity receptor contacts, and with few exceptions (23), neither free subunit has significant biological activity (7). Thus, any modification of the hormone that distorts the interaction between its subunits has the potential to disrupt hormone activity, a problem that has been recognized for many years (7) but largely ignored. In the experiments reported here, we sought to minimize this problem by limiting our studies to lutropins such as hCG and bLH that bind the rat lutropin receptor well, an indication that their overall conformations are very similar. We anticipated that the dramatic differences in the affinities of these hormones for the human lutropin receptor (24,25) would be caused by a few key residues in the primary receptor contact site that could be identified by comparing the abilities of hCG/bLH chimeras to bind the human and rat receptors. Chimeras that had hCG residues in the contact site were expected to bind the hLHR like hCG; those that had bLH residues in this site were expected to bind hLHR like bLH. As we show here, the low affinity of bLH for hLHR appears to involve an interaction between its subunits that either distorts the region of the high affinity contact and/or creates a steric interaction between the hormone and the receptor. This implies that subtle changes in hormone conformation can exert a much more dramatic influence on ligand binding than commonly perceived.

MATERIALS AND METHODS
Purified hCG was obtained from Drs. Robert Canfield and Steven Birken (Columbia University, NY). Purified bLH was obtained from Dr. John Pierce (University of California at Los Angeles, CA). Antibodies were obtained from Dr. Canfield, Drs. Glenn Armstrong and Robert Wolfert (Hybritech Inc., CA), and Dr. Janet Roser (University of California at Davis, CA) as noted previously (1,20). The hLHR cDNA was obtained from Dr. Aaron Hsueh (Stanford University, CA) in a vector that was used without modification to make stable CHO cell lines that express the human receptor. These lines were prepared by co-transfecting CHO cells with the hLHR vector and pSV2-Neo, a vector encoding aminoglycoside phosphotransferase downstream of the SV40 early promoter (26), selecting stable transformants in the presence of 500 g G418/ml, and identifying receptor-expressing cell lines based on their abilities to bind 125 I-labeled hCG. A similar strategy was used to select CHO cell lines that express the rat LH receptor. The rLHR cDNA (27) was inserted downstream of the metallothionein promoter in a vector (pLEN) kindly supplied by Dr. Peter Kushner (University of California at San Francisco, CA). This was accomplished by cloning the XhoI-BamHI fragment of pSVL-hCG␤Ј (21) into the unique XhoI-BamHI sites of pLEN to create a vector (pLEN-hCG␤Ј) that has a unique XbaI site immediately downstream of the XhoI site. The XbaI-BamHI fragment containing the coding region of hCG␤Ј was excised and replaced with an XbaI-BamHI fragment containing the entire rLH receptor cDNA prepared as described earlier (27).
Figs. 1 and 2 illustrate the amino acid sequences of the hormones and analogs used in this study relative to the overall structures of the ␣and ␤-subunits. Characterization of the immunological properties of each ␣-subunit analog has been described (1). All but two of the ␤-subunit analogs have also been described (20,28). CLC89 -96, an hLH/hCG ␤-subunit chimera having hCG ␤-subunit residues 89 -92 replaced by their hLH counterparts, was made by swapping parts of the cDNA that encode hCG and CLC77-96 (28) at PvuII restriction sites common to both. These are located in the codons for Ala 86 -Ser 87 -Cys 88 and in the pSVL vector (Pharmacia Biotech Inc.). Specifically, the small fragment prepared by PvuII digestion of pSVL-hCG␤Ј (21) was ligated to the large fragment created by PvuII digestion of pSVL-hCLC␤77-96 (28), and a clone having the proper orientation of the insert was selected by endonuclease restriction mapping. The vector encoding hCG ␤-subunit in which Arg 95 was converted to Leu, pSVL-hCG␤R95L, was made by simultaneously ligating three fragments including the large XhoI-PpuMI fragment of the vector pKBM-hCG␤Ј (21) that contained the vector and codons for hCG ␤-subunit residues 103-145, an XhoI-BssHII fragment that contained the codons for the hCG ␤-subunit signal sequence and amino acids 1-92, and a synthetic oligonucleotide cassette that had BssHII and PpuMI overhangs and encoded residues 93-102. This restored all hCG ␤-subunit codons except that for Arg 95 , which was changed to Leu during synthesis of the DNA cassette. The coding sequence was confirmed by standard dideoxy methods and the XhoI-BamHI fragment was cloned into pSVL for expression in COS-7 cells.
Glycoprotein hormone heterodimers were prepared by transient cotransfection of COS-7 cells with vectors encoding ␣and ␤-subunits as described (21). Hormone and analog heterodimers released into the medium were quantified using a sandwich immunoassay similar to one we reported (29) except that different antibodies and standards were used, depending on the hormone or analog being measured. To quantify heterodimers consisting of human/bovine ␣-subunit chimeras and hCG ␤-subunits, we captured the analogs with an anti-␤-subunit antibody (B112) having high affinity for an hCG ␤-subunit epitope that included Asn 77 (28). We detected the complex using a radiolabeled antibody (B109) that recognized a ␤-subunit epitope specific to the heterodimer (28). Highly purified urinary hCG was used as the standard. This dimer-specific assay has been shown to be unaffected by the presence of bovine ␣-subunit residues (1). Heterodimers containing hCG ␣-subunit and bLH/hCG ␤-subunit chimeras were assayed using an ␣-subunit antibody (A113) as a capture agent and a labeled antibody (B410) that has high affinity for nearly all mammalian LH ␤-subunits, including those from hCG and bLH (30). Again, we used highly purified urinary hCG as a standard. We also used this same assay to quantify heterodimers containing bovine hCG/bLH ␤-subunit chimeras. However, because A113 has lower affinity for the ␣-subunit from bLH than that from hCG (1), we used highly purified bLH as the standard in these latter assays. Procedures for preparing radiolabeled hCG and antibodies using IODO-GEN (Pierce) to specific activities of approximately 50 Ci/g have been described (31). Methods for measuring the abilities of hCG, bLH, and analogs to inhibit binding of 125 I-labeled hCG to CHO cells expressing lutropin receptors have also been cited (22). Briefly, we mixed the hormones and analogs with 125 I-labeled hCG (100,000 cpm, approximately 1.5 ng) and then added cells expressing the receptors. Following 1 h at 37°C, we added 2 ml of 0.9% NaCl solution containing 2 mg bovine serum albumin, sedimented the cells at 1000 ϫ g (10 min), aspirated the supernatant, and measured the radioactivity in the pellet using a ␥ counter.

RESULTS
The Small Influence of the bLH ␣-Subunit on Receptor Binding Is Not Sufficient to Account for the Inability of bLH to Bind hLHR-Consistent with earlier reports (32), we found that the heterodimer consisting of the bovine ␣-subunit and hCG ␤-subunit bound rLHR well (Table I). This analog bound hLHR about half as well as hCG (Table I), but more than 1000-fold better than bLH, indicating that differences between the human and bovine ␣-subunits per se were not responsible for the very low activity of bLH in hLHR assays. To localize residues for the small loss in binding to the hLHR, we compared the activities of heterodimers containing hCG ␤-subunit and human/bovine ␣-subunit chimeras ( Figs. 1 and 2). Many of these were more active than those containing either the human or bovine ␣-subunits (Table I). While some of the most active analogs (e.g. bH1-6/41-81) contained bovine residues in loop 1 and human residues in all or parts of loop 3, the presence of bovine residues in loop 1 was not required for increased activity. Other analogs (e.g. bH11-26/64 -68, bH11-26/73-75, and bH11-26/81) were more active than hCG even though they contained human residues in loop 1 and bovine residues in parts of loop 3 (Table  I). This suggested that the increased activities of some chimeras was caused by interactions between residues in the regions of the ␣-subunit that differ most in the human and bovine proteins (i.e. the N terminus, loop 1, and loop 3) rather than by interactions between bovine-specific ␣-subunit residues and either of the receptors.
A Portion of the bLH ␤-Subunit in Loops Two and/or Three containing hCG ␤-subunits The analogs illustrated here were prepared as described (1). The block diagram refers to the locations of codons that were changed by mutagenesis (i.e. the results of restriction endonuclease swapping or PCR site-directed changes). Several of these, particularly those in loop 2, encode the same amino acid in hCG and bLH ␣-subunits. The analog name refers to the portions of each block that contain hCG-specific residues. Values refer to the overall average potency determined in three (rLHR) or four separate experiments (hLHR). The potency of hCG in both assays was defined as unity. Note, at least 1000-fold more bLH than hCG was needed to inhibit binding of 125 I-hCG to hLH receptors by 50%.
Had a Small Influence on Binding to hLHR, but Not Enough to Account for the Inability of bLH to Bind to hLHR-Heterodimers containing the human ␣-subunit and bLH/hCG ␤-subunit chimeras bound hLHR at least 100-fold better than bLH, suggesting that residues in the bLH ␤-subunit per se accounted for only a fraction of the low affinity of bLH for the hLHR. The design of the chimeras permitted an assessment of differences in ␤-subunit loop 1, loops 2-3, and the seat belt (Table II). The presence of bLH residues in loops 2-3 had a greater influence than those in loop 1 or the seat belt (Fig. 3, top, Table II, column labeled "human ␣-subunit") as seen by comparing the relative activity of an analog containing bLH ␤-subunit residues only in loop 1 (i.e. CbL36 -145) with that of an analog containing bLH ␤-subunit residues in loops 1-3 (i.e. CbL88 -145). The influence of loops 2-3 can also be seen by comparing the activity of an analog containing bLH seat belt residues (i.e. CbL1-87) with that of an analog containing the bLH seat belt and ␤-subunit residues derived from loops 2-3 (i.e. CbL1-35). Heterodimers containing human ␣-subunit and bLH ␤-subunit residues only in loop 1 (i.e. CbL36 -145) or the seat belt (i.e. CbL1-87) were about as active as hCG.
The Low Potency of bLH in the hLHR Binding Assay Is Due to a Combined Influence of the Bovine ␣-Subunit and the Seat Belt Portion of the bLH ␤-Subunit-To test the idea that the inactivity of bLH was due to the combined effects of bovine residues in both subunits, we assayed heterodimers that contained bovine residues in parts of both the ␣and ␤-subunits. We anticipated that the least active analogs would contain the bovine ␣-subunit and portions of the ␤-subunit in which loops 2-3 originated from bLH. Although analogs containing the bovine ␣-subunit and these bLH ␤-subunit residues were less active than those containing the human ␣-subunit (Table II), they were more active than bLH in hLHR assays. For example, the hLHR activities of heterodimers containing the entire bovine ␣-subunit and bLH ␤-subunit residues solely in loop 1 (CbL36 -145) or loops 1-3 (CbL88 -145) were reduced 50 -80% relative to those that contained the human ␣-subunit (Fig. 3, top and bottom, and Table II), yet they remained at least 100-fold greater than bLH.
The least active chimeras contained the bovine ␣-subunit and the bLH seat belt. This can be seen by comparing the activities of heterodimers containing the human or bovine ␣-subunits and CbL1-87, a ␤-subunit chimera containing hCG loops 1-3 and the bLH seat belt (Figs. 3 and 4, Table II). The heterodimer containing the human ␣-subunit had nearly the same potency as hCG; that containing the bovine ␣-subunit was inactive at the highest concentration available for testing.
The relative influences of ␤-subunit loops 1-3 and the seat FIG. 1. Sequences of the ␣-subunit chimeras used in these studies. The residues listed in single letter code are those that are unique to the human ␣-subunit. A dot represents a residue that is identical to that found in the bovine ␣-subunit. The dash refers to four residues that are not found in the human protein. Numbers at the top of the sequence refer to human ␣-subunit residues and that differ in the two proteins.

FIG. 2. Comparison of the hCG, bLH, hLH, and chimeric ␤-subunit sequences used in this study.
Residues that differ from those of hCG ␤-subunit are illustrated here. The positions of the dots refers to residues that are identical to those in hCG. belt can be seen by comparing the activities of heterodimers containing the CbL1-87 and CbL88 -145 ␤-subunits. That containing the bovine ␣-subunit, bLH ␤-subunit loops 1-3, and the hCG seat belt (i.e. CbL88 -145) was at least 100-fold more active than that containing the bovine ␣-subunit, hCG ␤-subunit loops 1-3, and the bLH seat belt (i.e. CbL1-87). These observations showed that the effect of the interactions between bLH-specific residues in the ␣-subunit and seat belt was considerably greater than those in the ␣-subunit and ␤-subunit loops 1-3.
Identification of Seat Belt Residues That Influence Binding to hLHR-Previous studies showed that seat belt residues influence the receptor binding specificity of hCG (21). Those between Cys 93 and Cys 100 had a greater influence on binding to LH receptors than those between Cys 100 and Cys 110 (22). Residues in the N-terminal half of the bLH seat belt appeared to exert a greater influence on hLHR binding than those in the C-terminal half. This can be seen by comparing the hLHR binding activities of heterodimers containing the bovine ␣-subunit and ␤-subunit chimeras CbL88 -145 or CbL103-145, analogs that differ only by the presence of hCG and bLH residues in the N-terminal halves of the seat belt, respectively (Table II). That containing hCG residues in the N-terminal half of the seat belt (i.e. CbL88 -145) was much more active than that containing bLH residues in this region (i.e. CbL103-145).
The amino acid sequences of the entire N-terminal halves of the seat belts of hCG, hLH, and bLH are C 90 ALCRRSTTDC 100 , C 90 GPCRRSTSDC 100 , and C 90 GPCRLSSTDC 100 , respectively. Residues Gly 91 , Pro 92 , Leu 95 , and Ser 97 differ in this region of the hCG and bLH ␤-subunits. Although residues Gly 91 and Pro 92 are found in the seat belt of hLH, a hormone that binds hLHR well, we considered it possible that an interaction between the bovine ␣-subunit and Gly 91 /Pro 92 could lead to a reduction in hLHR binding. To test this, we compared the activities of heterodimers containing either the human or bovine ␣-subunit and an hCG/hLH chimeric ␤-subunit (CLC89 -92) containing Gly 91 and Pro 92 . In rLHR assays, the bovine ␣-subunit/CLC89 -92 ␤-subunit heterodimer was twice as active as that containing the human ␣-subunit (Fig. 5), an increase similar to what was observed when the bovine ␣-subunit was expressed with the ␤-subunit of hCG (Table II). In hLHR assays, the bovine ␣-subunit/CLC89 -92 ␤-subunit heterodimer was similar in activity to that containing the bovine ␣-subunit and the hCG ␤-subunit. Both had approximately half the activity of hCG (Table II). Thus, the differences at ␤-subunit residues 91 and 92 were not responsible for the very low ability of bLH to bind to the hLHR.
Finally, we studied the role of Leu 95 , the residue that was the least conserved in this region of the hCG and bLH seat belts. Heterodimers containing the human ␣-subunit and an hCG ␤-subunit analog in which Arg 95 was converted to Leu (i.e.

TABLE II
Summary of ␤-subunit mutations on the activities of heterodimers in hLHR assays This table illustrates the activities of heterodimers containing either the human or bovine ␣-subunit and the ␤-subunit analogs. All values are potencies relative to that of hCG and are the means Ϯ S.E. of at least two experiments (n). The block diagram refers to the structures of the chimeras prepared as described in the text or in an earlier publication (20). Loop 1 refers to ␤-subunit residues 9 -34. Loops 2 and 3 refer to ␤-subunit residues 38 -88. SB refers to the seat belt ␤-subunit residues residues 91-110. Note, residues in the N-and C-terminal halves of the seat belt have been shown to influence the abilities of hCG to bind to LH and FSH receptors, respectively (1). NT refers to not tested. Note also, CbL103-145 refers to an analog that has the Nterminal half of the bLH seat belt and the C-terminal half of the hCG seat belt. *Denotes that the activity of the bovine ␣-subunit/hCG␤R95L heterodimer was estimated from the ED 50 obtained in two experiments, one of which is shown in Fig. 7. This value appears to be a maximum estimate since the displacement of 125 I-hCG by the analog was not parallel to that of hCG. Much higher amounts of this analog relative to those of hCG would have been required to inhibit binding of the radiolabel to an equivalent extent. hCG␤R95L) bound hLHR like hCG (Fig. 6), suggesting that neither Arg nor Leu at this position participated in essential hLHR contacts needed for high affinity ligand binding. In contrast, the heterodimer containing the bovine ␣-subunit and the hCG␤R95L ␤-subunit had much less ability to bind hLHR since concentrations of more than 10 Ϫ7 M were needed to observe 50% inhibition of binding. While this heterodimer was more active than bLH, it was clear that the presence of a single amino acid change, namely Leu for Arg 95 , accounted for much of the influence of the bLH seat belt. DISCUSSION Several mutagenesis strategies are available for identifying ligand residues that may be involved in receptor contacts. For example, to identify interactions between lutropins and their receptors, one might begin with hCG and change conserved residues to search for those that disrupt its ability to interact with receptors. Several of these have been found (13)(14)(15)(16)34). We assumed that it would be more efficient to distinguish receptor contacts using a homolog scanning approach in which we characterized the hLHR binding activities of chimeras containing parts of lutropins that had both high and low affinities for the hLHR. In principle, this would enable us to identify clusters of amino acids that would contain residues responsible for the inactivity of bLH. Subsequent mutagenesis would need focus only on these smaller regions. This strategy has been shown to be highly efficient (35).
When these studies were initiated, we expected residues, which differed in hCG and bLH in one or both subunits, would participate in high affinity hLHR contacts and be recognized by contributions they made to the overall affinity of the hormone for the receptor. Given that bLH is virtually inactive in hLHR assays, we were surprised to find that heterodimers containing one hCG subunit were highly active. Those containing the hCG ␤-subunit and a chimeric ␣-subunit were at least half as active as hCG; some were even more active than hCG in both rLHR and hLHR assays. The activities of heterodimers containing the hCG ␣-subunit and bLH-specific residues in ␤-subunit loops 2-3 were reduced only 5-10 fold. Even when present throughout loops 1-3 of both subunits, bLH-specific residues reduced binding only 10 -20 fold. Taken together, this suggests that residues that differ in hCG and bLH are probably not essential for receptor binding. In marked contrast to the effects of residues in ␤-subunit loops 1-3, the combined influence of a single residue (Leu 95 ) in the bLH seat belt and the bLH ␣-subunit accounted for most of the inactivity of bLH in hLHR assays. Thus, while neither the entire bLH seat belt nor Leu 95 had a significant influence on the hLHR binding activities of heterodimers containing the human ␣-subunit, the activities of heterodimers containing these residues and the bLH ␣-subunit were reduced 1000-fold, nearly to the level of bLH. This extreme synergism between the bLH ␣-subunit and the bLH seat belt suggested that differences in the abilities of hCG and bLH to bind hLHR are due primarily to an influence of the seat belt on subunit interaction.
These observations have important implications for efforts to identify residues responsible for high affinity glycoprotein-hormone receptor interactions. Because there is no crystal structure of any glycoprotein hormone-receptor complex, most efforts to decipher portions of the hormone that might interact with the receptor are based on monitoring changes in receptor binding or function following chemical or enzymatic modification (7), site-directed mutagenesis (13)(14)(15)(16)(17)(18)(19), chemical cross-linking (36 -38), or competition with hormone fragments (8 -12). Due to the unusually high proportion of residues in the subunit FIG. 4. Influence of the bLH seat belt on rLHR and hLHR Binding. COS-7 cells were transfected with vectors expressing the human or bovine ␣-subunits and CbL1-87 ␤-subunit, an hCG-based analog that contains the bLH seat belt. Binding was monitored to hLHR (h) or rLHR (r). Values are means of triplicates, and the vertical bars extend to the limits of the S.E.
FIG. 5. Influence of human LH ␤-subunit residues 89 -92 on hLHR and rLHR binding. COS-7 cells were transfected with vectors encoding the human or bovine ␣-subunits and a vector encoding CLC89 -92 ␤-subunit, an hCG/hLH chimera in which residues 89 -92 are derived from hLH. The abilities of these heterodimers to inhibit binding of 125 I-labeled hCG to rLHR (r) or hLHR (h) were determined in triplicate. Vertical bars extend to the limits of the S.E.
FIG. 6. Influence of replacing hCG ␤-subunit residue Arg 95 with Leu on binding to hLHR and rLHR. COS-7 cells were transfected with vectors encoding the human or bovine ␣-subunits and a vector encoding hCG␤R95L ␤-subunit, an hCG ␤-subunit analog in which Arg 95 is replaced by leucine. The abilities of these heterodimers to inhibit binding of 125 I-labeled hCG to rLHR (r) and hLHR (h) were determined and are illustrated here in triplicate. Vertical bars extend to the limits of the S.E. interface (39), many mutations distant from the receptor contact sites may influence subunit interaction and thereby alter receptor binding and/or signal transduction. This would account for the observations noted earlier that some parts of the hormones previously suggested to make important receptor contacts appear to be exposed in the hormone-receptor complex or can be deleted without disrupting hormone activity. The influence of small changes in gonadotropin conformation on receptor binding is readily explained by the model in which the hormone binds to the concave surface of a horseshoe-shaped receptor extracellular domain and contacts it at two distinct sites (1). A small change in the conformation of a heterodimer that interferes with either of these contacts would be expected to reduce binding in a fashion similar to that observed between the interaction between the bovine ␣-subunit and the bLH seat belt.