Site-directed Mutagenesis Defines the Individual Roles of the Glycosylation Sites on Follicle-stimulating Hormone"

To determine the specific role of each follicle -s t' 1mu- lating hormone (FSH) oligosaccharide, we mutated Asn to Gln at each glycosylation site (ruGlnS2, aGln7', aGlnS2", PGln7, PGlnZ4, and pGln7") to selectively inhibit oligosaccharide attachment. For wild-type and mutant FSH, we determined the binding affinity to homogenized rat Sertoli cells and the signal-transducing activity in cultured rat granulosa cells. The binding affinity of FSH lacking any one of the oligosaccharides was increased over wild-type FSH, while the signal-transducing activ- ity of FSH lacking the oligosaccharide at &ns2 (aGlnS2 FSH) was markedly reduced, and that of FSH lacking either p oligosaccharide (pGln' and PGlnZ4 FSH) was slightly reduced. At each FSHP glycosylation site, we made a second amino acid substitution to inhibit glycosylation (PTyrS and PTyP) and an amino acid substitution that preserved glycosylation (pSe# and pSef16). The amino acid sequence of the second p subunit glycosylation site was important for signal transduction, regard- less of the presense or absence of the oligosaccharide. Thus, while each FSH oligosaccharide

To determine the specific role of each follicle -s t' 1mulating hormone (FSH) oligosaccharide, we mutated Asn to Gln at each glycosylation site (ruGlnS2, aGln7', aGlnS2", PGln7, PGlnZ4, and pGln7") to selectively inhibit oligosaccharide attachment. For wild-type and mutant FSH, we determined the binding affinity to homogenized rat Sertoli cells and the signal-transducing activity in cultured rat granulosa cells. The binding affinity of FSH lacking any one of the oligosaccharides was increased over wild-type FSH, while the signal-transducing activity of FSH lacking the oligosaccharide at &ns2 (aGlnS2 FSH) was markedly reduced, and that of FSH lacking either p oligosaccharide (pGln' and PGlnZ4 FSH) was slightly reduced. At each FSHP glycosylation site, we made a second amino acid substitution to inhibit glycosylation (PTyrS and PTyP) and an amino acid substitution that preserved glycosylation (pSe# and pSef16). The amino acid sequence of the second p subunit glycosylation site was important for signal transduction, regardless of the presense or absence of the oligosaccharide. Thus, while each FSH oligosaccharide has a similar impact on binding affinity, the ruS2 oligosaccharide has a disproportionate role in signal transduction, and the amino acid sequence at pAsna4 functions in both binding and signal transduction.
FSH,' luteinizing hormone, thyroid-stimulating hormone, and CG, make up a family of dimeric glycoprotein hormones with a common (Y subunit and a /3 subunit that confers biologic specificity. The (Y subunit has two N-linked oligosaccharides, and the /3 subunit has either one (in the case of luteinizing hormone and thyroid-stimulating hormone) or two (in the case of CG and FSH) N-linked oligosaccharides (1-3). Removal of these oligosaccharides by enzymatic or chemical methods decreases adenylyl cyclase-stimulating activity, suggesting that in general, the oligosaccharides are required for efficient signal transduction (4-11). Although removal of all four FSH oligosaccharides reduces biologic activity, the role of the individual oligosaccharides remains to be elucidated (12,131. Site-directed mutagenesis can be used to examine the role of individual glycosylation sites by selectively inhibiting oligosaccharide attachment at each site. Using this approach, Matzuk and coworkers (14) identified the oligosaccharide at position 52 on the (Y subunit as having a greater role in CG activity than the other * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. $ To whom correspondence and reprint requests should be addressed: Bldg. 10 oligosaccharides. To our knowledge, this type of detailed analysis has not been previously reported for FSH. Although site-directed mutagenesis has advantages over chemical deglycosylation, it has some limitations that have not been fully appreciated. The first is that deductions made about mutant hormones have been based on immunologic assessments of their concentration. Although this can provide some useful information, it assumes that the epitopeb) involved in immunoreactivity have not been altered by mutagenesis. A more straightforward, non-immunologic approach would be to determine the binding activity in wild-type or mutant media and then measure the steroidogenic activity of equivalent amounts of receptor binding activity (ie. shifts in potency reflect differences in signal-transducing activity). The second assumption is that differences in mutant hormone activity are due to inhibition of glycosylation rather than to changing the amino acid sequence. Some investigators have used two separate amino acid substitutions to prevent glycosylation a t a single site (14), but important additional information could be gained from mutating the amino acid sequence of the region without inhibiting glycosylation.
In the present study, we used site-directed mutagenesis to examine the role of each glycosylation site on FSH. We substituted Gln for Asn at each glycosylation site to inhibit oligosaccharide attachment. We examined the immunologic activity, apparent receptor affinity, and signal-transducing activity of the mutant FSH analogues. At the FSHp glycosylation sites, we inhibited glycosylation by a second mutation (Thr to Tyr) and included a third amino acid substitution at each site (Thr to Ser) that would not inhibit glycosylation, to further delineate the effects of changing the amino acid sequence from those of inhibiting glycosylation.
EXPERIMENTAL PROCEDURES Materials-Enzymes for cloning, sequencing, and polymerase chain reaction mutagenesis were obtained from Bethesda Research Laboratories and Perkin-Elmer. The expression vectors, pAV2 and pAv2-(1, were gifts of F. Wondisford (NIDDK, NIH). Tissue culture reagents were obtained from Life Technologies, Inc., Biofluids, and Sigma. IRMA kits were obtained from Serono Diagnostics. Human pituitary FSH AFP8417B for radioimmunoassay was obtained from the National Hormone Pituitary Program. Human pituitary FSH F0614 for iodination was obtained from The Scripps Institute.
Mutagenesis and Dansfection-Mutagenesis was performed by overlap extension in the polymerase chain reaction using sense and antisense mutant oligonucleotide primers and an a subunit cDNA template or a FSHp cDNA template constructed by overlap extension from a FSHp genomic clone (gift of Larry Jameson, MGH, Boston, MA). Polymerase chain reaction products were sequenced to confirm the presence of the intended mutation and the absence of any unintended Tuq polymerase mutations. Wild-type and mutant FSHp cDNAs were cloned into pAV2 and co-transfected with pAV2 containing the wild-type ( I subunit cDNA using calcium phosphate precipitation into 293 human embryonal kidney cells (gift of Barrie Carter, NIADDK, NIH). Since 293 cells were found to secrete an ( I subunit that cross-reacted in our ( I subunit radioimmunoassays, mutant ( I subunit cDNAs were cloned into 14016 FSHP Giycosyiation Sites  pSVL and co-transfected with pSVL containing the wild-type FSHp cDNA into cos7 monkey kidney cells (ATCC). Cell media were collected 96 h after removal of the precipitate, centrifuged, filtered to remove cell debris, and concentrated 3-4-fold with Amicon Centripreps.
Column Chromatography-Media from transfected cells were chromatographed on two G-75 Sephadex columns (1.6 x 80 cm) in tandem eluted with 0.2 M ammonium acetate (pH 7.4). Media were chromatographed in separate runs with lZ5I-labeled hCG co-chromatographed as by FSH IRMA. an internal marker in each run. FSH immunoreactivity was monitored Immunologic Assays-The immunologic activity of the mutant media was determined using three different FSH immunoassay systems, including a two-site IRMA and two different radioimmunoassays. The first radioimmunoassay employed a rabbit anti-ovine FSH polyclonal antiserum designated H-31 (151, and the second radioimmunoassay employed a rabbit anti-human FSH polyclonal antiserum designated DDBB1001, obtained from Immunosearch (New Brunswick, NJ).
Radio Receptor Assay-The FSH receptor binding affinities of wildtype and mutant FSH media were measured in a radio receptor assay using rat testes homogenate and radiolabeled hFSH, as described previously (16). The ED,, of each mutant FSH media relative to wildtype FSH media was determined using the computer program Allfit (171, which was used to calculate the relative binding potency. Statistical analysis of the differences in the relative binding potency of wildtype and mutant FSH media was performed using the Mann-Whitney test. Determination of Signal-transducing Actiuity-We examined the signal-transducing activity of wild-type and mutant media by comparing their steroidogenic response at the same level of receptor occupancy. We used the rat testis FSH radio receptor assay to determine the binding activities of wild-type and mutant media. The binding activities were then dosed out in the rat granulosa cell bioassay, performed as detailed elsewhere (18,19). Full dose-response curves for wild-type media and each mutant media were generated in at least two granulosa cell bioassays. Dose-response curves for wild-type and mutant FSH were compared using the computer program Allfit to determine the ED,, and R,, of each curve. Statistical analysis of the difference between the ED,, and R,, of wild-type and mutant media was performed using the Mann-Whitney test.

FIG. 2. G-75 Sephadex elution profiles of wild-type and pSe#' FSH.
Media from cells co-transfected with the wild-type a subunit cDNA and either wild-type or mutant FSHp cDNA were chromatographed on two G-75 Sephadex columns (1.6 x 80 cm) in tandem eluted with 0.2 M ammonium acetate (pH 7.4). Media were chromatographed in separate runs with lZ5I-labeled hCG co-chromatographed as an internal marker in each run. FSH immunoreactivity was monitored by FSH IRMA in terms of human pituitary FSH standard (AFP8417B).

RESULTS AND DISCUSSION
hFSH Glycosylation Mutants-The tripeptide required for oligosaccharide attachment is Asn-X-Thr/Ser. The glycosylation sites on the a subunit are at amino acids 52-54 (Am-Val-Thr) and amino acids 78-80 (Asn-His-Thr). The glycosylation sites on the p subunit are at amino acids 7-9 (Asn-Thr-Thr) and 24-26 (Am-Ile-Thr). At each site individually, and both sites in tandem, we mutated Asn to Gln (aGlnSZ, ~x G l n~~, pGln7, pGlnZ4, and pGln7-24) to prevent attachment of the oligosaccharide at one or both glycosylation site(s). The oligosaccharides make up approximately 30% of the molecular weight of FSH, contributing substantially to its apparent molecular weight as determined by Sephadex chromatography. Thus, the progressive shift to the right of the G-75 Sephadex elution profiles of PGln7, PGlnZ4, and pGln7-24 FSH (Fig. 1) is consistent with the lack of one or both p subunit oligosaccharides secondary to mutations in their respective glycosylation sites. In addition to the Gln mutations, at each p subunit glycosylation site, we mutated Thr to either Tyr ( p T p s and pTyrZ6) or Ser (pSerQ and pSerZ6). The Tyr substitution provides an additional amino acid substitution that prevents oligosaccharide attachment, while the Ser substitution changes the amino acid sequence but maintains a functional glycosylation site. We have previously shown, using radiolabeled glucosamine, that substitution of Ser for Thr at the analogous site on p thyroid-stimulating hormone does not impair glycosylation (20). Consistent with the presense of a functional glycosylation site, we observed no shift in the G-75 Sephadex elution profile of /3Se?6 FSH compared with wild-type FSH (Fig. 2). Small changes in the amount or type of oligosaccharide present on j3Se?6 FSH, however, would not be detected by G-75 chromatography.  (Fig. 3), indicating a distinct change in the conformation of this FSH analogue. This change can be attributed to the amino acid substitution at this site, rather than the absence of the oligosaccharide per se, since the Gln substitution, which also inhibited glycosylation, did not result in a change in FSH conformation that could be detected by our immunoassay systems (Fig. 3). The finding of a conformational change for /3w6 FSH, when none of the other mutations produced a detectable change in conformation, suggests that this amino acid is particularly important for the tertiary structure of FSH. The proximity of amino acid 26 to the potential cysteine loop between residues 28 and 32 and to the large Trp at position 27 may make this site vulnerable to conformational changes due to substitution of a bulky Tyr for the smaller Thr. In addition, this region has been identified as a site of interaction between the (Y and /3 subunits (21), which may contribute to the conformational changes in the resulting FSH dimer.
Receptor Binding Affinity of the FSH Glycosylation Mutants-The dose-response curves of ctGln5', (~G l n~~, and aGln52-78 (Fig. 4A), and /3Gln7, /3GlnZ4, and pGln7-24 FSH (Fig.  4B) were shifted to the left of that for wild-type FSH in the radio receptor assay, suggesting that inhibition of oligosaccharide attachment at either or both glycosylation sites on either subunit enhances hormone binding affinity. Indeed, the apparent binding affinities of all the non-glycosylated FSH analogues, except FSH, were greater than that of wild-type FSH (Table I). Previous chemical deglycosylation studies of the glycoprotein hormones indicate that in general, deglycosylation does not interfere with receptor binding but rather enhances binding affinity (5-13). These studies are limited, however, by the fact that chemical deglycosylation can only remove about 80% of the oligosaccharide. In contrast, site-directed mutagenesis can completely inhibit oligosaccharide attachment at one or more sites on the hormone. A previous site-directed mutagenesis study of CG indicated no change in the binding affinity of non-glycosylated hormone (141, suggesting that the increased binding affinity of chemically deglycosylated hCG   may have been an artifact of the deglycosylation procedure. In contrast, our findings for FSH indicate that lack of the oligosaccharides enhances binding affinity. Furthermore, our findings extend this principle to suggest that not only is binding affinity enhanced by removal of all the oligosaccharides, but it occurs with FSH lacking just one oligosaccharide. The reason for the differences in our findings compared with those for hCG may relate to differences in protein conformation or to differences in the terminal oligosaccharides of FSH made in 293 and cos7 cells compared with hCG made in Chinese hamster ovary cells.

FSHp Glycos
At the first FSHp glycosylation site (Am7), mutations that inhibit glycosylation (pGln7 and pTyrg) enhanced binding affinity, while a mutation that preserves glycosylation (j3Serg) had no effect (Fig. 5 4 ) . Thus, increases in binding affinity due to mutations of the first p subunit glycosylation site appear to be related to the absence of the oligosaccharide rather than the amino acid substitution per se. At the second FSHp glycosylation site ( A m z 4 ) , however, both the Gln substitution (pGlnZ4) that inhibits glycosylation and the Ser substitution (pSerZ6) that preserves glycosylation enhanced binding affinity (Fig. 5B). Thus, increases in binding affinity due to mutations of the second p subunit glycosylation site can be related to changes in the amino acid sequence without inhibition of glycosylation. In contrast to the findings for the other mutants, substitution of Tyr for Thr at position 26 (p"yP) dramatically reduced binding affinity (Fig. 5B). Apparently, the conformational change induced by this amino acid substitution affects not only its interaction with certain antibodies but also its interaction with the FSH receptor. Amino acid substitutions at position 26 could affect the nearby cysteine loop between residues 32 and 51 that has been identified by previous peptide studies as a receptor binding region (22, 23). It is also possible, however, that the conformational change induced by this amino acid substitution altered more distant receptor binding sites (23, 24).
Signal-transducing Activity of the FSH Glycosylation Mutants-Determination of signal-transducing activity does not require accurate immunologic assessments of hormone con-  . 6. Dose-response curves of wild-type FSH and FSH composed of mutant a subunit in the rat granulosa cell bioassay. The dose in each case is the receptor binding activity of wild-type or mutant media determined in a radio receptor assay using rat testes homogenate. The response is estradiol concentration in the media of cultured rat granulosa cells after a 72-h incubation.

Signal-transducing activities of the various mutant FSHp analogues relative to wild-type recombinant FSHp
Glycosylation status is indicated by the + andsymbols. The receptor binding activity was determined in the rat testis radio receptor assay in terms of pituitary FSH standard (AFP8417B). The binding activities were dosed out in the rat granulosa cell bioassay. The ED,,, used to determine potency, and the R,,, used to determine intrinsic activity, were determined from full dose-response curves in at least two granulosa cell bioassays using the computer program Allfit. centration but involves the straightforward use of a radio receptor assay to determine FSH binding activity in wild-type and mutant media. When equivalent amounts of FSH binding activity elicit different steroidogenic responses, wild-type and mutant FSH must differ in their signal-transducing activity. Alterations in signal-transducing activity, therefore, were deduced by comparing the ability of equivalent amounts of wildtype and mutant FSH binding activity to stimulate estradiol production in cultured rat granulosa cells.

Signal-transducing Activity of FSH Containing Mutant a
Subunit-The dose-response curve of the signal-transducing activity of aGlnS2 FSH was shifted well to the right of wild-type FSH, while that of aGln78 was similar to that of wild-type FSH, and that of aGln52-78 was essentially flat (Fig. 6). The maximum response and potency obtained with ~G l n~~ FSH was 24 and 2%, respectively, of wild-type FSH, while that of c~Gln'~ was indistinguishable from wild-type FSH, and that of aGln52-78 FSH was nil (Table 11). Thus, the oligosaccharide at position 52 of the a subunit is critical for FSH signal transduction. Our findings are consistent with previous studies, using chemically deglycosylated FSH subunits, that suggested the a subunit oligosaccharides were critical for FSH signal transduction (13).
Using site-directed mutagenesis, however, we have identified the oligosaccharide at position 52 as having a critical role in FSH signal transduction that the oligosaccharide a t position 78 does not have. This distinction between single oligosaccharides is not possible with chemical deglycosylation methods. Despite the disparate biologic functions of FSH and CG, our findings for aGlnS2 FSH are similar t o previous findings for CG lacking the a52 oligosaccharide (14). Although the amino acid sequences of the a subunits of CG and FSH are identical, there is evidence that the conformation of the a subunit changes upon interaction with different /3 subunits (24). Nonetheless, the oligosaccharide at position 52 appears to be critical for signal transduction of both hormones. In fact, the reduction in the signal-transducing potency of aGln52 FSH was 10-fold greater than that of CG lacking the a52 oligosaccharide (14). Thus, if there are differences in the a subunit conformation of FSH compared with CG, they appear to accentuate the need for the a52 oligosaccharide.
Recent cassette mutagenesis studies have indicated that a discrete amino acid region in the p subunits of FSH (p87-94) and CG (093-100) determines the specificity of their interaction with their respective receptors (25). Our findings for FSH, as well as the previous findings for CG, suggest that the a52 oligosaccharide is an additional site of receptor interaction distinct from this p subunit site. This would be consistent with recent three-dimensional models of CG and FSH that place these a and p subunit regions a t distinct sites on the same surface of the molecule (26, 27). The p subunit site varies between the glycoprotein hormones and appears to determine binding specificity, while the a subunit site of interaction is similar between glycoprotein hormones and is critical for their signal-transducing activity.

Signal-transducing Activity of FSH Containing Mutant p
Subunit-The dose-response curves of the signal-transducing activity of PGln7, pGln24, and pGln7-24 FSHp were shifted to the right of wild-type FSH in the rat granulosa cell bioassay (Fig.  7, A-C). Although there was no difference in the maximum response of pGln7, PGlnZ4, and pGln7-24 FSH, there was a 2-fold decrease in the their signal-transducing potency (Table 11).
Thus, the oligosaccharides on the p subunit of FSH appear to influence FSH signal transduction. These data help resolve some of the controversy regarding the relative role of the a and p subunit oligosaccharides in FSH activity. Although some studies have indicated that the a subunit oligosaccharides are essential for FSH activity (13), other studies, using chemically deglycosylated a subunit, have indicated that only the p subunit oligosaccharides are required for signal transduction (12).
It has been difficult to examine the role of the p subunit oligosaccharides directly, due to the limited quantities of purified FSHp available for chemical deglycosylation. Using recombinant technology, however, we have circumvented this problem and directly examined the properties of FSH lacking one or both p subunit oligosaccharides. Our findings suggest that in contrast to the site-directed mutagenesis findings for CG, indicating no role for the p subunit oligosaccharides, the FSHp oligosaccharides have a role in signal transduction, albeit a much smaller role than the a52 oligosaccharide. Differences in the conformation of FSH, compared with CG, may require the presence of the / 3 subunit oligosaccharides for the most efficient FSH signal transduction. The above changes in signal-transducing activity observed for the FSHp mutants can be attributed either to the lack of the oligosaccharide or to the amino acid substitution used to inhibit glycosylation. At the first FSHp glycosylation site, amino acid substitutions that inhibit glycosylation (pGln7 and pTyrg) decreased signal-transducing activity, while a mutation that preserves glycosylation (pSerg) had signal-transducing activity equivalent to wild-type FSH (Fig. 7 , A, D, and F ) (Table 11).
This suggests that at the first FSHp glycosylation site, the FSHP Glycosylation Sites oligosaccharide is the crucial feature for efficient signal transduction.
In contrast, at the second FSHp glycosylation site, a mutation that preserves glycosylation (pSe9'j) had a marked reduction in signal-transducing activity (Fig. 7E), suggesting that the amino acid sequence influences efficient signal transduction independent of the glycosylation status. These findings have important implications for site-directed mutagenesis studies, because they suggest that in some cases, changes in hormone activity that have been attributed to the lack of the oligosaccharide may actually be due to the amino acid substitutions employed to inhibit glycosylation. We cannot entirely rule out the possibility, however, that the Ser substitution resulted in incomplete or aberrant glycosylation at Amz4 that may account for the observed changes in signal-transducing activity.
Substitution of Tyr for Thr at position 26 (PTy?') caused a dramatic decrease in signal-transducing activity (Fig. 7G), affecting not only the potency but also the R,, of signal transduction (Table I). By definition, this reduction occurs at equivalent levels of receptor binding activity; thus, the Ty?'j substitution not only decreased the binding affinity of FSH for its receptor, it also reduced the efficiency of FSH signal transduction by the hormone-receptor complex. Apparently, the amino acid residue at position 26 is critical for FSH to achieve a conformation compatible with both tight receptor binding and efficient signal transduction. Although it is possible that substitutions at residue 26 could impact on distant FSH regions, it is of interest that nearby residues 34-37 (Thr-Arg-Asp-Leu) have been identified by peptide studies as having a direct role in FSH action (26). Conformational changes brought about by insertion of a Tyr at position 26 might limit the accessibility of this critical region. Thus, our findings indicate that the amino acid sequence near the second FSHp glycosylation site represents a third region of receptor interaction, in addition to p87-94 and a52. This region appears to be critical for both binding and signal transduction.
With the exception of pTyrZ6 FSH above, our findings generally agree with previous deglycosylation studies of the glycoprotein hormones, indicating that alterations that enhance binding affinity also decrease signal transduction (5-13, 28,29). This inverse relationship between binding affinity and signal transduction accounts for the fact that deglycosylated glycoprotein hormones can act as potent antagonists of native hormone (5-11, 28, 29). A model that could explain these findings would be that the oligosaccharides either directly, by lectin-like binding to the receptor (28), or indirectly, by effects on protein conformation (29), promote a hormone-receptor complex with enhanced signal transduction at the expense of some binding energy. This model is supported by epitope-mapping studies indicating a different conformation for deglycosylated hormone-receptor complex than for native hormone-receptor complex (30) and by studies indicating that conformational changes associated with efficient signal transduction reduce binding affinity (28,29). Our findings suggest that this inverse relationship is not a strict one, however, so that while all the FSH oligosaccharides impact equally on binding affinity, the a52 oligosaccharide has a disproportionate role in signal transduction. This supports a model of FSH action, suggested by previous peptide studies (24), in which there are multiple sites of hormone-receptor interaction; some are involved in binding and others in signal transduction. The site(s) related to binding affinity are equally affected by each of the FSH oligosaccharides, but the site(s) of signal transduction are concentrated at the a52 oligosaccharide. Furthermore, there may be additional sites, such as PThP, that are important for both binding and signal transduction.