Follicle-Stimulating Hormone Glycobiology

Abstract

FSH glycosylation varies in two functionally important aspects: microheterogeneity, resulting from oligosaccharide structure variation, and macroheterogeneity, arising from partial FSHβ subunit glycosylation. Although advances in mass spectrometry permit extensive characterization of FSH glycan populations, microheterogeneity remains difficult to illustrate, and comparisons between different studies are challenging because no standard format exists for rendering oligosaccharide structures. FSH microheterogeneity is illustrated using a consistent glycan diagram format to illustrate the large array of structures associated with one hormone. This is extended to commercially available recombinant FSH preparations, which exhibit greatly reduced microheterogeneity at three of four glycosylation sites. Macroheterogeneity is demonstrated by electrophoretic mobility shifts due to the absence of FSHβ glycans that can be assessed by Western blotting of immunopurified FSH. Initially, macroheterogeneity was hoped to matter more than microheterogeneity. However, it now appears that both forms of carbohydrate heterogeneity have to be taken into consideration. FSH glycosylation can reduce its apparent affinity for its cognate receptor by delaying initial interaction with the receptor and limiting access to all of the available binding sites. This is followed by impaired cellular signaling responses that may be related to reduced receptor occupancy or biased signaling. To resolve these alternatives, well-characterized FSH glycoform preparations are necessary.

FSH is a pituitary gonadotropin that plays a central role in reproduction. In females, FSH stimulates antrum formation in secondary follicles, growth and maturation in antral follicles, and it prepares the latter for ovulation in response to the LH surge (1). In males, FSH stimulates Sertoli cell proliferation during testicular development and maintains Sertoli cell function in the mature testis (2). Whereas FSH is essential for female fertility, in some species, such as mice, FSH is clearly not essential in the male, as Fshb-null males retain complete fertility (3). In other species, such as humans, FSH may be required for male fertility, as some men deficient in FSH or in FSH receptor (FSHR) function are azoospermatic (4). However, others produce sperm, making the necessity for FSH action in the human testis uncertain (5). In recent years FSH has been suggested to contribute to bone loss in rodent models (6) and perhaps in humans. Recently, it has been proposed to play a role in promoting obesity (7). FSH involvement in these nontraditional targets is controversial (8), yet may be affected by FSH glycosylation.

FSH biological activity in the gonads is dependent on glycosylation, not only as a critical factor in determining survival in the circulation (9), but also in stimulating cellular signaling via FSHR activation, particularly when signaling through the Gα_s/cAMP/PKA pathway (10, 11). Very few data exist for the impact of FSH glycosylation on other, more recently identified FSHR-activated pathways (12, 13). FSH glycosylation appears to change in response to changing physiological states, including sex (14), puberty (15, 16), reproductive cycles (17, 18), as well as with age (14, 19, 20). Accordingly, the biologic activity of circulating FSH can vary under changing physiological conditions (21–26). The mechanisms underlying carbohydrate modulation of FSH activity are poorly understood. A major challenge is the heterogeneous population of oligosaccharides found in FSH preparations. The lack of a consensus format for representing the patterns of branching and branch termination, which are the two major sources of oligosaccharide structural variation, makes it difficult to compare one study with another. Glycan structural diagrams combining monosaccharide symbols proposed by the Consortium for Functional Glycomics (27) with solid and dashed lines at defined angles representing linkage as employed by the Oxford Glycobiology Institute (28, 29) will be used throughout. This hybrid format is an accepted alternative to the Consortium for Functional Glycomics system (27).

Classic FSH isoforms were thought to reflect only oligosaccharide negatively charged microheterogeneity reflecting sialic acid content (30–32). However, sulfate and phosphate provide alternative negatively charged groups, although sialic acid predominates. Moreover, several oligosaccharide structures can possess the same negative charge. Macroheterogeneity, resulting from the absence of one FSHβ N-glycan, provides another mechanism for significantly altering FSH glycosylation (33, 34). FSH variants resulting from macroheterogeneity are designated glycoforms (35, 36). The challenge provided by glycoforms is that there is no direct correspondence between them and classic charge-based FSH isoforms (32).

FSH Peptide and Carbohydrate Moieties

FSH is a heterodimeric glycoprotein comprised of a common α-subunit noncovalently associated with a hormone-specific FSHβ subunit (37). Both subunits are N-glycosylated and, similar to other cystine knot proteins (38), their tertiary structure is dominated by three loops defined by the three Cys knot motif disulfide bonds (Fig. 1). Two loops, designated L1 and L3, are located on one side of the Cys knot, whereas the middle loop, L2, is found on the opposite side. FSHβ also possesses a fourth loop, the so-called seatbelt loop, which is involved in heterodimer formation. The FSH heterodimer comprises antiparallel subunits stabilized by the FSHβ seatbelt loop embracing the FSHα L2 loop. Both α-subunit N-glycans are found on different loops, αL2 and αL3, in the C-terminal half of the primary structure (Fig. 1A), but at opposite ends of the tertiary structure long axis. As these glycans extend the long axis of the elliptically shaped FSH heterodimer (38), their orientation along this axis may account for the nonsignificant increases in metabolic clearance when either one or both are deleted from FSH derivatives (9). Both FSHβ N-glycans are located in the βL1 loop in the N-terminal quarter of the primary structure and side by side in the tertiary structure (Fig. 1B), where they more than double the narrow axis of FSH. Each FSHβ oligosaccharide contributes significantly to the circulatory survival of FSH, and the combined effects are synergistic (9). Although the α-subunit polypeptide moieties derived from the common CGA gene are identical for FSH and the other glycoprotein hormones, LH, TSH, and chorionic gonadotropin (CG), the oligosaccharide populations decorating both N-glycosylation sites distinguish FSHα from α-subunits derived from these other glycoprotein hormones as well as the free α-subunit (39, 40).

Figure 1.

FSH subunit primary structures. (A) FSHα primary and tertiary structures. (B) FSHβ primary and tertiary structures. Amino acid sequences of FSHα and FSHβ subunits indicated by single-letter amino acid code. Cystine knot loops in primary and tertiary structures are indicated by yellow highlighting for loops L1 and L3 and green for loop L2. The seatbelt loop is colored cyan. Lines above the subunit sequences connecting Cys residues (C) indicate disulfide bonds. Cys residue numbers are shown below. Cys knot disulfides are indicated by thick lines. N-glycosylation site sequons of the type NXT (N, Asn; X, any residue except Pro; T, Thr) are underlined, and oligosaccharide structures are shown below each subunit. Key to oligosaccharide structures can be found in Fig. 3. FSH subunit three-dimensional structures were extracted from pdb file 4ay9 using PyMol. Known FSH glycans were added to the FSH polypeptide backbone using GLYCAM.

FSH Glycan Functions Revealed by Deglycosylation or Mutation

Functions of individual FSH glycans at each of four N-glycosylation sites are summarized in Fig. 2. FSHα subunit oligosaccharides are essential for α-subunit folding (41), α-subunit stability (42), heterodimer stability (43), full FSHR activation (10, 44, 45), and at least one may modulate FSHR binding (10, 44). FSHβ subunit N-glycans both contribute significantly to metabolic clearance (9). Conflicting reports exist for the effects of these glycans on FSHR binding and signaling (9–11, 44).

Figure 2.

Functions of specific Asn-linked oligosaccharides in FSH. Polypeptide moiety of FSH extracted from pdb file 1ay9 and the heterodimer rendered as cartoon using PyMol. The subunits are colored as in Fig. 1 (yellow indicates cystine knot loops L1 and L3, cyan indicates the seatbelt loop, green indicates remaining FSHα and FSHβ). Oligosaccharides are rendered as sticks colored green for αAsn⁵² and αAsn⁷⁸, cyan for βAsn⁷ and βAsn²⁴.

FSH Glycosylation Microheterogeneity

FSH microheterogeneity results in size and charge variants of the hormone called FSH isoforms (46–49). Early FSH oligosaccharide structure reports described 11 to 30, largely complex-type glycans, possessing two, three, or four branches, some of which lacked terminal sialic acid residues, and most of which existed as both fucosylated and nonfucosylated forms (50–52). Some low-abundance glycans were hybrid types, which are more commonly associated with LH and TSH (51, 52).

Pituitary FSH, pituitary FSH glycoform, and urinary FSH glycans released from reduced, carboxymethylated samples by PNGaseF digestion, analyzed by nano-electrospray ionization mass spectrometry, revealed an even greater variety of human FSH (hFSH) glycan structures (53, 54). However, the predominance of complex-type glycans, representing 85% to 97% of each population, was confirmed [Table 1 (53, 55–59)]. At least 230 oligosaccharide structures were proposed, including sialylated and sulfated variants of >80 neutral core glycans (Fig. 3). The actual neutral core structures of biantennary or triantennary glycans were detected, albeit in low abundance, but no asialo-tetra-antennary glycans were encountered. Because several oligosaccharide structures can possess the same combination of monosaccharides, there were fewer glycan ions in the spectra than proposed structures. Fragmentation analysis used to confirm ∼70% of the oligosaccharide structures revealed that many of the alternative core glycan structures existed in one FSH glycan preparation or another. Charged glycan variants averaged more than three per core structure, based largely on the number of negatively charged, terminal sialic acid and sulfate groups. Additional heterogeneity arises from N-acetyl-neuraminic acid (Neu5Ac) residue attachment by two different linkages, either α2–3 or α2–6. The latter predominates in pituitary hFSH (52). Sialic acid linkage has been implicated in FSH clearance in rodents, as α2–3-linked Neu5Ac prevents the liver asialoglycoprotein receptor from binding Gal residues in glycans, whereas α2–6-linked Neu5Ac residue does not (60, 61).

Table 1.

Macroheterogeneity and Microheterogeneity in Pituitary, Urinary, and Recombinant hFSH Preparations

FSH Preparation	Pituitary hFSH (53)	Urinary hFSH (53)	Urinary hFSH (59)	Recombinant GH3-hFSH (55)	Recombinant CHO hFSH (58)	GonalF (56)	GonalF (57)	(Bemfola) (56)	Bemfola (57)	Puregon-HP (59)
Glycoform Abundance, %
FSH24	77	86	+a	45	—b	78.1	—	—	—	—
FSH21	23	14	+	55	—	15.8	—	—	—	—
FSH18	—	—	—	—	—	6.1	—	—	—	—
FSH15	—	—	—	—	—	—	—	—	—	—
No. oligosaccharide structures	80	67	20	33	10	23	38	27	39	17
Oligosaccharide Type Abundance, %
Oligomannose	0.3	0	—	0	—	0	—	—	—	—
Complex	94.2	95.9	82.6	85.3	91	96.7	93.9	93.9	90	94.6
Biantennary	38.2	37.2	56.4	55.5	53	63.7	60.4	53.9	47.8	72.1
Triantennary(3)c	41.0	44.0	26.1	—	18	21.0	22.7	23.4	26.9	12.3
Triantennary(6)d	—	—	—	29.7	8	—	—		—	—
Tetra-antennary	15.0	14.8	—	0	12	12.0	10.8	16.6	15.3	—
Neutral	0.3	2.2	—	12.3	—	1.5	5.1		2.0	10.2
Sialylated	99.1	97.5	100	87.7	91	96.7	94.4	96.7	97.8	89.8
O-Ac-Neu5Ac or Neu5Gc	—	—	—	—	—	—	0.1	—	—	1.5
Sulfated	6.5	4.2	—	—	—	—	—	—	—	3.7
Sialylated/sulfated	5.9	3.9	—	—	—	—	—	—	—	—
Core fucose	43.0	23.9	17.0	50.6	29	32.9	36.4	30.5	36.9	13.7
Antenna fucose	0.3	—	3.6	19.9	—	—	2.4	—	3.0	5.1
Bisect GlcNAc	32.6	23.9	13.1	47.0	—	—	—	—	—	—
GalNAc	2.8	1.7	—	10.5	—	—	—	—	—	—
Lactosamine	2.1	1.3	—	3.9	5	3.0	10.0	10.1	12.6	—

^aQualitative detection only.

^bEither not detected or not reported.

^cThird antenna on Man3 branch.

^dThird antenna on Man6 branch.

Figure 3.

Pituitary FSH glycans. (A) Enzymatically released human FSH oligosaccharides characterized by nano-electrospray mass spectrometry (53, 54). Only the nonfucosylated forms are shown. Of these, ∼90 have fucosylated counterparts (see Fig. 5C). (B) Chemically released, nonfucosylated human FSH oligosaccharide core structures (50). (C) Enzymatically released human FSH oligosaccharides characterized by nano-electrospray mass spectrometry (53, 54). A total of 91 structures were identified, most of which had nonfucosylated counterparts (Fig. 5A). (D) Chemically released, fucosylated human FSH oligosaccharides (50).

Of the 230 FSH glycans proposed, 139 lacked core-linked fucose (Fig. 3A), whereas 91 were core-fucosylated (Fig. 3C). Virtually all of the core-fucosylated oligosaccharides had nonfucosylated counterparts. There were 100 derivatives of the 11 desialylated FSH oligosaccharide structures initially reported (50). Some glycans exhibited partially synthesized antennae consisting of either unterminated Gal residues, or just the branch initiating GlcNAc residue. Oligosaccharides not reported before included oligomannose-type, either not encountered (50) or present in uncharacterized neutral oligosaccharide fractions (51, 52). These were well characterized owing to their high abundance in the partially glycosylated hFSH variant isolated from purified human LH (hLH) preparations (54) and represent intermediates between the original Glc₃Man₉GlcNAc₂ oligosaccharide transferred en bloc to FSH subunits in the endoplasmic reticulum (ER) and complex glycans assembled in the Golgi cisternae. Hybrid-type glycans fell into two groups, one possessing GalNAc in the complex branch, whereas the others possessed Gal. Complex-type biantennary, triantennary, and tetra-antennary glycans possessing one or more GalNAc residues instead of Gal in at least one antenna were novel. Biantennary and triantennary GalNAc glycans were found in both nonfucosylated and core-fucosylated structures. Tetraantennary GalNAc-including glycans all lacked fucose. The biantennary group included those with GalNAc on both antennae, or GalNAc on one, usually Man3, with Gal on the Man6. The latter group were found in the population possessing core-linked fucose (Fig. 3C), but not in the group lacking core fucose (Fig. 3A). Triantennary glycans in both groups possessed a single GalNAc residue either Man3 or Man6 linked. Tetraantennary glycans were only found in the population lacking core fucose and possessed either one or two GalNAc residues. Lactosamine repeats are often attached to the GlcNAc β1–6 linked to the Man6 branch (62). FSH lactosamine repeats, found on some of the triantennary and tetra-antennary glycans, were located on either Man3 or Man6 branches. Triantennary glycans with the third antenna on the Man6 branch were rare. In contrast, these were the only triantennary glycans observed in recombinant hFSH produced by the rat GH₃ cell line [Table 1 and Bousfield et al. (54) and Butnev et al. (55)]. Curiously, sulfate was observed in a few oligosaccharides possessing Gal instead of GalNAc. Additionally, several oligosaccharides possessed one more Neu5Ac residue than did terminal Gal or GalNAc residues in complex branches. Perhaps additional sialic acid residues were linked to GlcNAc residues, as in bovine FSH (51).

Although mass spectrometry greatly expanded the extent of microheterogeneity associated with FSH oligosaccharides, positional information was completely absent. Core fucosylation is typical of >95% of oligosaccharides released from FSHβ, yet present in <5% of those from FSHα. More precise positional information has been obtained by analysis of glycopeptides containing each of the N-glycosylation sites and their attached oligosaccharides. Proteinase K digestion released three– to four–amino acid residue glycopeptides from reduced, carboxymethylated FSH samples (63). Gel filtration eliminated peptides that interfered with mass spectrometry analysis (63). All but Asn-His-Thr were detected by electrospray mass spectrometry. Only a glycosylated tetrapeptide, Glu-Asn-His-Thr, was detected in a subsequent study (32). Overall, 29 glycans were localized: 4 on βAsn⁷, 12 on βAsn²⁴, 11 on αAsn⁵², and 2 on αAsn⁷⁸ (Fig. 4). Twenty-four unique structures were observed, as compared with 230 in the oligosaccharide mass spectra. The uneven yield of site-specific glycans made general conclusions rather tentative, however, with one exception, that is, the restriction of core fucose to FSHβ glycans was supported. The association of greater branching at βAsn⁷ was suggested by the presence of tetra-antennary glycans, which were absent from the largely biantennary and triantennary βAsn²⁴ and αAsn⁵² oligosaccharide populations. Proteinase K digestion proved inadequate for LH and CG β-subunits, which possess a high Pro content, thereby making them resistant to proteolysis (63, 64). Better coverage may be obtained from digesting reduced, carboxamidomethylated FSH with chymotrypsin (56) or mixtures of specific and nonspecific proteases (65).

Figure 4.

Site-specific analysis of pituitary FSH N-glycosylation by glycopeptide mass spectrometry. Reduced carboxymethylated hFSH was exhaustively digested with 10% (w/w) proteinase K, and glycopeptides were isolated by gel filtration and analyzed by electrospray mass spectrometry. Composite results are from an analysis of purified hFSH and purified hFSH isoforms separated by chromatofocusing (32, 63).

Recombinant hFSH preparations are currently more readily available than pituitary hFSH and exhibit greater biological activities (14,403 IU/mg) (57) than the latter (National Hormone and Pituitary Program, 8560 IU/mg). This appears partly due to greater FSH content in recombinant FSH preparations as compared with pituitary FSH preparations (55), but it may also reflect reduced glycan branching in the former (Table 1).

PNGaseF-released oligosaccharides from Chinese hamster ovary (CHO) cell–expressed, recombinant hFSH provided by Organon, were characterized by nuclear magnetic resonance (NMR) (58). Biantennary, triantennary (both Man3 and Man6 types), and tetra-antennary oligosaccharide structures possessing α2–3-linked Neu5Ac comprised 91% of the carbohydrate recovered (Fig. 5A). The α2–6-linked Neu5Ac, which comprised the majority in pituitary (52) and urinary hFSH oligosaccharides (59), was not detected. A disialylated, biantennary glycan was the most abundant nonfucosylated structure. The absence of bisecting GlcNAc residues indicated no GlcNAc transferase III activity (66). The relative abundance of nonfucosylated oligosaccharides was 62%, more than twice that of the 29% core-fucosylated structures. Subsequent FSHβ glycopeptide mass spectrometry revealed that βAsn⁷ possessed both types of glycans [Fig. 5B and Grass et al. (56) and Mastrangeli et al. (57)]. One or more lactosamine repeats were associated with triantennary or tetra-antennary oligosaccharides comprising 4% of the glycans; however, structures were not proposed, and the diagrams in Fig. 5A merely illustrate the glycan types incorporating lactosamine repeats.

Figure 5.

Analysis of recombinant CHO-expressed hFSH glycosylation by NMR and glycopeptide mass spectrometry. (A) PNGaseF-released oligosaccharides from recombinant FSH provided by Organon. Structures were determined by NMR (58). Glycans with lactosamine repeats were detected, but complete structures were not defined. The bar graph shows relative glycan abundance determined by NMR. (B–E) Reduced, carboxamidomethylated recombinant hFSH samples were digested with 5% (w/w) chymotrypsin and glycopeptides analyzed by ethylene bridged hybrid amide chromatography–electrospray ionization mass spectrometry (57). Oligosaccharide structural diagrams are shown, and the adjacent graphs show relative abundance of each. The bars are means of either duplicate (Bemfola) or triplicate (GonalF) determinations (±SD). (B) GonalF and Bemfola βAsn⁷ glycosylation. (C) βAsn²⁴ glycosylation. (D) αAsn⁵² glycosylation. (E) αAsn⁷⁸ glycosylation. The bracketed oligosaccharides were found in Bemfola but not in GonalF. The remainder were found in both.

Mass spectrometry has been employed to characterize total glycans released from recombinant FSH (67), recombinant FSHα and FSHβ subunits (68), and chymotryptic peptides derived from reduced, carboxamidomethylated preparations (56, 57, 59). Two of the latter studies evaluated glycopeptides in three batches of GonalF and one or two batches of the biosimilar Bemfola (56, 57). The results of the second study are summarized in Fig. 5B–5E. The glycan populations of GonalF and Bemfola were qualitatively similar in that the latter possessed only three glycans not found in GonalF (Fig. 5C–5E). Quantitatively, Bemfola possessed more triantennary and tetra-antennary glycans than did GonalF (Table 1). By far, the most abundant oligosaccharides attached to αAsn⁵², αAsn⁷⁸, and βAsn²⁴ were disialylated, biantennary, differing only in core fucose, which the FSHβ glycan possessed and both FSHα glycans lacked (Fig. 5B–5E). Only 9 glycan structures were found on FSHα, whereas βAsn²⁴ possessed 12 glycans and βAsn⁷ had 29. The latter included significant amounts of tetra-antennary glycans, particularly Bemfola, which exhibited a reduction in biantennary glycans as a result of being better endowed with triantennary and tetra-antennary glycans. Triantennary glycans possessing a lactosamine repeat, which had been identified in an earlier FSH study by fast atom bombardment mass spectrometry fragment ions at m/z 1274 and 913 representing sialic acid_1–hexose_2–N-acetylhexosamine₂ and hexose_2–N-acetylhexosamine₂ fragments (68), were localized to βAsn⁷. Tetra-antennary glycans possessing lactosamine repeats were also identified (56, 57). Missing in the mass spectrometry studies were triantennary glycans in which the third antenna was on the Man6 branch, rather than Man3, as indicated by NMR [Fig. 5A and Hard et al. (58)].

Another study of chymotryptic glycopeptides derived from Puregon, a CHO cell–expressed recombinant FSH preparation, reported 22 to 26 FSHβ glycans at each glycosylation site and 15 to 18 glycans at each FSHα glycosylation site (59). However, several proposed structures were unusual; for example, glycans possessing two antenna fucose residues on the same antenna, or sulfated sialic acids. Caution is indicated as the authors appear to have deduced the structures of the N-linked glycans mainly by comparing glucose unit values with those available in GlycoBase (69) and with some very poor-quality fragmentation spectra, which do not contain sufficient information to assign the structures shown. The problem with GlycoBase is that it only contains a limited number of glycans, whose structures were mainly deduced by exoglycosidase digestions and HPLC retention times. The latter, even when expressed as glucose unit values, are not absolutely reproducible. Exoglycosidase digestions do not differentiate isomers, and it is difficult to confirm the presence of a bisecting GlcNAc because of the lack of a suitable enzyme. For urinary hFSH, fewer glycans were reported, 10 to 25 at each site, even though total oligosaccharide analyses in other studies indicate greater heterogeneity in urinary hFSH than in recombinant hFSH (53–55).

Defining combinations of the three to four N-glycans in FSH preparations is important for developing a complete understanding of FSH glycosylation. Currently, only subunit data are available. Pairs of RCM-FSHα glycans were predicted from its mass spectrum, whereas the RCM-FSHβ spectrum proved unsuitable (68). Reduction and carboxamidomethylation yielded both FSH subunit derivatives that permitted prediction from subunit mass spectra and knowledge of glycan populations (56, 57, 59). A recent study (57) reported seven FSHα masses indicating three variants with two biantennary glycans, two with one biantennary and one triantennary glycan, and two with two triantennary glycans (Fig. 6). The FSHβ subunit ions indicated a pair of biantennary glycans, three with one biantennary and one triantennary glycan, two with both triantennary glycans, and three with one triantennary glycan paired with either a lactosamine repeat–containing triantennary or tetra-antennary glycan. Low pH conditions for liquid chromatography–mass spectrometry (LC-MS) dissociate FSH into subunits, creating a major technical barrier that may be overcome by chemical crosslinking, as has been applied to ovine LH (70).

Figure 6.

FSH subunit glycan pairs. LC-MS of reduced, carboxamidomethyl GonalF provided several pairs of FSH subunit glycan masses (56). In contrast to the original analysis of subunit glycosylation, ions consistent with two triantennary glycans were observed for both subunits.

Macroheterogeneity in the FSHβ Subunit

Macroheterogeneity, the absence of one or more N-glycans, was initially reported for human CG (hCG)β lacking either Asn¹³ or Asn³⁰ carbohydrates, which exhibited altered electrophoretic mobility (71). The first report of partially glycosylated FSH involved recombinant bovine FSHβ (72). Following SDS-PAGE, FSH subunits appear as a very broad band when stained with Coomassie brilliant blue, thereby obscuring FSHβ mobility shifts (33). Matrix-assisted laser desorption ionization–mass spectrometry (MALDI-MS) and SDS-PAGE revealed a partially glycosylated eFSHβ subunit in purified β-subunit preparations (73). Western blotting with monoclonal antibody RFSH20 (74) revealed two human FSHβ subunit variants, a 24,000 relative molecular mass (M_r) band (24k-FSHβ) and a 21,000 M_r band (21k-FSHβ), while the FSHα subunit migrated as an intermediate sized band (33).

Four variants of human FSHβ have been reported: the 24k-FSHβ, mentioned above, which possesses both Asn⁷ and Asn²⁴ glycans (33), the 21k-FSHβ, which possesses only Asn⁷ glycans (34, 55), an 18,000 M_r FSHβ band that possesses only Asn²⁴ glycans (34, 55), and a 15,000 M_r FSHβ band (15k-FSHβ) that lacks both glycan populations (33, 75). We identify the corresponding FSH glycoforms by the FSHβ variant that each possesses, thus fully glycosylated FSH24 or hypoglycosylated FSH21, FSH18, and FSH15, respectively (35). FSH24 and FSH21 appear to be the most abundant and are readily detected in purified, pooled pituitary FSH preparations (33), individual pituitary FSH (53), pooled postmenopausal urinary gonadotropin preparations (53), and individual urine sample FSH (53). FSH18 appears when FSH21 is purified, and hypoglycosylated FSH preparations consisting of both glycoforms are designated as FSH21/18 (34). MALDI-MS analysis of two FSHβ preparations revealed nonglycosylated 15k-FSHβ. Western blotting detected only 21k-FSHβ in one preparation, whereas MALDI-MS revealed only 15k-FSHβ. In a 24k-FSHβ preparation, a small amount of nonglycosylated 15k-FSHβ was detected by MALDI-MS analysis of a 24k-FSHβ preparation (33). The 15k-FSHβ subunit has recently been observed in pituitaries from transgenic mice expressing a double glycosylation site mutant FSHB transgene (75).

In principle, mass spectrometry is a much more sensitive approach for measuring FSH glycoform abundance than Western blots. However, MALDI-MS of pituitary and urinary FSH is confounded by subunit dissociation. FSHα ions partially obscure pituitary FSHβ glycoform ions (59, 76) and are coincident with those of recombinant FSHβ (59). FSH subunits purified by reverse-phase HPLC (68) produce only fully glycosylated 21k-FSHβ (33). Fast atom bombardment mass spectrometry of PNGaseF-treated tryptic FSHβ peptides supported the quantitative glycosylation of both FSHβ Asn residues 7 and 24. Methyl esterification revealed that the 19-35 peptide possessed one more carboxyl group than predicted by its sequence, and the masses of underivatized N-terminal peptides 1-18 and 3-18 were one mass unit larger due to conversion of glycosylated Asn resides to Asp by PNGaseF (68, 77). In the course of chymotryptic glycopeptide analysis, 4.5% to 7.5% nonglycosylated peptides bearing Asn⁷ and 15.8% bearing Asn²⁴ were encountered, suggesting that as much as 20% to 23% GonalF was hypoglycosylated (61). This was close to the 21% hypoglycosylated FSHβ reported by Western blotting of GonalF (55). A subsequent study of reduced, carboxamidomethylated GonalF chymotryptic digests did not report nonglycosylated glycopeptides (57). A third study reported an unaltered FSHβ chymotryptic peptide following PNGaseF digestion, indicating a nonglycosylated Asn²⁴ in a urinary FSH preparation, but not in Puregon (59). The relative abundance of the unmodified peptide was not reported.

Three groups reported expression of recombinant, glycosylation site mutant versions of two to three hypoglycosylated FSH glycoforms, FSH21, FSH18, and FSH15 (10, 44, 78). Although quantitative data were not reported, the FSHα glycosylation site mutations reduced recombinant hFSH yield more than did the FSHβ mutations (10). Based on FSH immunoreactivity measured in Sephadex G-75 fractions of medium conditioned by Cos7 cells expressing either wild-type (wt) or one of the FSHβ glycosylation site mutants, βN7Q, βN24Q, or βN7,24Q, yields of mutant recombinant FSH were 45%, 69%, and 73% that of the wt, respectively (44). In contrast to expression in transformed cell lines, mouse gonadotropes did not secrete much, if any, FSH15 (75). Accordingly, FSH24, FSH21, and FSH18 are probably the physiologically relevant FSH glycoforms (36).

The cellular mechanism that yields FSH glycoforms seems to be the asparagine (N)-linked glycosylation site skipping by oligosaccharyl transferase (OST). Automated Edman degradation of hypoglycosylated FSHβ preparations (mixtures of 18k- and 21k-FSHβ) revealed phenylthiohydantoin (PhNCS)-Asn at cycles corresponding to Asn⁷ and Asn²⁴ (33, 55). No PhNCS derivative was observed at Edman cycles corresponding to these residues in 24k-FSHβ, because glycosylated Asn residues are not extracted from the Edman chemistry cartridge of a typical automated gas-phase or pulsed liquid sequencer due to excess hydrophilic character conferred by the oligosaccharide (33). Solid-phase Edman degradation produced a cluster of PhNCS-Asn/carbohydrate peaks when N-glycosylation sites were encountered owing to a more robust extraction with trifluoroacetic acid (39). The presence of PhNCS-Asn at cycles corresponding to N-glycosylation sites indicated that OST failed to transfer the preformed oligosaccharide precursor, Glc₃Man₉GlcNAc₂, from dolichol pyrophosphate to FSHβ at either one of these Asn residues (79, 80). Transient glycosylation followed by enzymatic deglycosylation, as reported for concanavalin A (81), produces a PhNCS-Asp derivative at Edman cycles corresponding to deglycosylated Asn residues (77). It is an unlikely mechanism because folded FSHβ subunits are insensitive to PNGaseF digestion (82), and PhNCS-Asp, although readily differentiated from PhNCS-Asn (39), is not above background when sequencing partially glycosylated FSHβ variants (55).

Glycosylation and FSH Biosynthesis

OST is found in all domains of life. Both bacterial PglB and archaeal AglB consist of single catalytic subunit OSTs (83, 84). The crystal structure of PglB revealed a requirement for unfolded protein structure, as the glycopeptide precursor makes a sharp turn while being scanned in the OST active site (83). This feature has been confirmed in the structure of multisubunit eukaryotic OST (85). Mammals possess two major OST complexes consisting of eight to nine subunits encoded by a dozen genes (85). Six noncatalytic subunits are common to both OST complexes: essential common subunits, ribophorin I/II, DAD1, DDOST, and nonessential common subunits OST4 and TMEM258. The OST complexes are largely differentiated by their catalytic subunit, staurosporine and temperature sensitivity 3 (STT3). The Sec61 translocon-associated, glycosylation site skip-prone isoform is designated STT3A, whereas the other, more versatile variant, is STT3B (86). The former performs N-glycosylation only cotranslationally, whereas the latter can perform N-glycosylation both cotranslationally and posttranslationally (87). Moreover, STT3B often glycosylates Asn residues skipped by STT3A (87). Two subunits, KRTCAP2 and OSTC, are associated only with STT3A. OSTC couples the OST complex to the Sec61 translocon (85). Two oxidoreductase subunits, TUSC3 or MAGT1, are associated with the STT3B-containing OST complex in a mutually exclusive manner (88, 89).

Studies with model glycoproteins have identified several structural characteristics associated with STT3A skipping potential N-glycosylation sites. However, glycosylation sites skipped in FSHβ subunit variants are not consistent with partial N-glycosylation of model glycoproteins (86). For example, the distance between the P site in the ribosome and the active site in STT3A (the Sec61 translocon-associated OST) is estimated to be 65 residues (90). Glycosylation sites in the last 75 C-terminal residues are frequently skipped by STT3A, with a steep drop-off in glycosylation efficiency beginning 50 residues from the C terminus (79). FSHα glycosylation sites at Asn⁵² and Asn⁷⁸ are only 40 and 12 residues from the C terminus, respectively (Fig. 7), yet they are always N-glycosylated in hormone preparations derived from natural sources. The partially glycosylated Asn⁷ and Asn²⁴ residues in FSHβ are 105 and 87 residues from the FSHβ C terminus, respectively. FSHβ Asn⁷ is fairly close to the human FSHβ signal peptide cleavage sites [Cys⁻¹/Asn¹ and the preferred Ser²/Cys³ site (33)], which is another region where glycosylation sites are skipped by STT3A (86). Nevertheless, Asn⁷ is more frequently glycosylated than Asn²⁴, as indicated by the greater abundance of FSH21 (34). Both FSHβ glycosylation sites are located four residues from Cys residues involved in disulfide bond formation (Fig. 7), which is a third protein structural feature that promotes STT3A glycosylation site skipping (88). In contrast, FSHα glycosylation sites are not as close to disulfide bond–associated Cys residues. As both LH and FSH are often synthesized in the same cell (91), a glaring gap in our knowledge is how LH subunits and FSHα are quantitatively N-glycosylated whereas FSHβ is partially glycosylated at the same time in the same cellular compartment.

Figure 7.

Cotranslational and posttranslational glycosylation of FSH subunits. (A) Distance of subunit C terminus from N-glycosylation sites. The 65-residue bar indicates the distance from the ribosomal P site to the Asn residue of the FSH subunit sequon (90). Prepeptide numbering is used for each subunit diagram. (B) Potential cotranslational glycosylation of FSHβ and posttranslational glycosylation of FSHα. OSTA indicates isoform possessing STT3A associated with Sec61 translocon. OSTB indicates isoform possessing STT3B not associated with Sec61 translocon that can perform posttranslational N-glycosylation. For the oligosaccharide diagrams, the blue circles represent Glc residues, green circles represent mannose, and blue squares represent GlcNAc. Connecting lines show linkages as in Fig. 6 key.

Studies involving OST employing either small interfering RNA suppression or CRISPR/Cas-directed knockout procedures to inhibit selected OST subunits show altered OST complex activities toward model proteins (88, 92, 93). This could be a mechanism by which steroid hormones, such as estrogen, progesterone, and testosterone, modulate OST activity. Although there is a report of progesterone regulating OST activity in hybridoma cells (94), it is not known whether expression of any OST subunit genes is regulated directly by steroid hormones. Application of RNA sequencing to fluorescence-labeled transgenic gonadotropes (95), or single-cell RNA sequencing to immortalized gonadotropes (96), or unmodified pituitary cells may help resolve this question. The effects of gonadal steroids on glycoform abundance may be mediated by GnRH, perhaps via the PKC pathway, as yeast PKC was reported to increase Stt3 activity (97). In mammalian cells, this might involve posttranslational glycosylation of skipped sites, as only STT3B possesses potential phosphorylation sites (98).

Other than the existence of partially glycosylated FSH variants in pituitary extracts, urinary gonadotropin preparations, and individual urine samples, evidence for physiological regulation of OST activity producing these variants is difficult to obtain. An apparent age-related reduction in the relative abundance of FSH21 in human pituitaries was revealed by Western blotting (Fig. 8) of anti-FSHβ monoclonal antibody 46.3H6.B7-immunopurified FSH samples (99). Extension to serum or urinary FSH has been hampered by the need to purify FSH and the low sensitivity of Western blotting procedures. Differential secretion of serum FSH glycoforms identified by electrophoretic separation and radioimmunoassay has been reported for women 21 to 40 years of age (100). However, these resulted from charge-based isoform measurements, and there does not seem to be a direct relationship between glycoforms and isoforms (32). Indeed, subsequent reports involving partially glycosylated FSH, LH, and TSH added size-based separation by gel filtration (24). Ovariectomy of female rhesus monkeys was attended by increased abundance of high–molecular weight forms of recombinant hFSH immunoreactivity, as assessed by Sephadex G-100 gel filtration (46). Western blots of ovariectomized rhesus pituitary FSH samples revealed the presence of both 24k-FSHβ and 21k-FSHβ (76). However, although Superdex 75 gel filtration chromatograms of the immunopurified rhesus FSH samples were very broad, no high–molecular weight peak was obvious, and the FSHα subunit obscured FSHβ glycoforms in MALDI-MS spectra. Estrogen replacement therapy was shown to eliminate the high–molecular weight pituitary recombinant hFSH found in ovariectomized monkey pituitaries, suggesting that negative feedback in the hypothalamic–pituitary axis can modulate OST activity toward FSHβ (101). More subtle shifts in rat pituitary FSH size following gonadectomy were altered by estrogen replacement in females and testosterone treatment in males (47). Again, because steroid hormones can act at both the pituitary and hypothalamic levels it is not known whether changes in FSH glycosylation are a direct effect on the gonadotrope or an indirect effect via hypothalamic stimulation. For example, GnRH stimulated secretion of less acidic FSH from human pituitaries (102–104).

Figure 8.

Age-related decline in pituitary FSH21 abundance. (A) Representative Western blots of pituitary hFSH isolated from individual human pituitary glands by immunoaffinity followed by gel filtration chromatography. The immunoaffinity column employed anti-FSHβ monoclonal antibody 46.3H6.B7 (99), and the Western blot primary antibody was anti-FSHβ monoclonal RFSH20 (74). The age at autopsy is indicated. Each FSH sample was evaluated in three to four independent Western blots. Postmenopausal hFSH derived from three lots of Pergonal is shown. (B) Linear regression of FSH21 abundance based on relative density of 21k-FSHβ bands. The dotted line indicates the average relative abundance of FSH21 in the Pergonal lots. [Reproduced from Bousfield GR, Butnev VY, Rueda-Santos MA, Brown A, Smalter Hall A, Harvey DJ. Macro and micro heterogeneity in pituitary and urinary follicle-stimulating hormone glycosylation. J Glycomics Lipidomics. 2014;4:125. Copyright: © 2014 Bousfield GR, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.]

N-linked oligosaccharides play important roles in glycoprotein biosynthesis, as they recruit lectin domain chaperones, calnexin or calreticulin, to assist in folding (105). Although there are no reports of specific chaperones interacting with FSH subunits, chaperone-mediated folding is likely important for both FSH subunits. For example, the common α-subunit can fold on its own (106–109). Bovine LHα subunit folding requires 75 minutes to achieve near native conformation, as indicated by circular dichroism in vitro (106). Meanwhile, human α-subunit disulfide bond formation is complete in less than 2 minutes in vivo (110). Accordingly, chaperones accelerate even normally spontaneous reactions. Studies with hCGα implicate αAsn⁷⁸ oligosaccharides with α-subunit folding, as mutation to eliminate this site resulted in <20% α-subunit secreted, with the remainder degraded (41). NMR studies of selectively αAsn⁵²-deglycosylated hCGα revealed that both reducing terminal αAsn⁷⁸ GlcNAc residues formed hydrogen bonds with αL1 and αL3 loop amino acid residue side chains (111). Reduced mobilities of the Man3 and Man6 residues suggested that they also interacted with the α-subunit loops (42). All of these monosaccharide residues are found in N-glycans regardless of glycan type, antenna number, or structure. The NMR three-dimensional structure of Asn⁵²-deglycosylated hCGα demonstrated that Leu¹² and Leu²⁶ in loop αL1 and Tyr⁶⁵, Val⁷⁶, His⁷⁹, and Thr⁸⁰ in αL3 specifically interact with αAsn⁷⁸ oligosaccharide (112). The corresponding cystine knot loops βL1 and βL3 in FSHβ are stabilized by a covalent disulfide bond between Cys residues 17 and 66 (38). Heterodimer stability has been associated with αAsn⁵² glycan because <30% of the mutant hCG dimer lacking this glycan was secreted (41). Moreover, deglycosylation of the homologous αAsn⁵⁶ in equine FSH reduced the time to reach 50% subunit dissociation at 37°C from 72 hours to <2 hours (43). Three laboratories mutated all four FSH N-glycosylation sites and reported a variety of effects on FSH synthesis and secretion (10, 44, 45). CHO cells expressing single FSHα glycosylation site mutants exhibited reduced secretion rates for FSH heterodimer (10). Mutating βAsn⁷, βAsn²⁴, or both glycosylation sites to Gln and transient transfection of Cos7 cells resulted in 31% to 55% reduction of hFSH immunoactivity in 96-hour conditioned medium samples, based on Sephadex G-75 peak areas (44). As long as a single glycosylation site was maintained, FSH triple mutants were secreted from CHO cells (10). In contrast, gonadotropes of transgenic mice expressing N7Q/N24Q FSHβ double mutants failed to secrete sufficient FSH15 to restore fertility to transgenic females lacking a functional mouse Fshb gene (75). Real-time quantitative PCR indicated that both wt-FSHB and N7,24Q-FSHB transcripts were expressed at the same levels. Confocal microscopy confirmed expression of FSHβ immunoreactivity in gonadotropes; however, pituitary extract Western blots revealed significantly reduced amounts of 15k-FSHβ protein under reducing conditions and little FSH15 heterodimer under nonreducing conditions. Mouse serum contained very little FSH immunoactivity, and no FSH was detected in pituitary organ culture medium. Thus, for FSHβ subunit folding, oligosaccharide-recruited chaperones appear to be necessary. Eliminating single FSH glycosylation sites appeared to slow biosynthesis of the heterodimer, no matter which subunit was mutated. Asn to Gln mutations at all four glycosylation sites produced no detectable secreted hormone from CHO cells, and that retained intracellularly was barely detectable (10). Mutation of all four Asn resides to Asp produced nonglycosylated FSH, although secretion was only ∼10% that of wt-FSH (45).

Glycosylation and FSH Secretion

An appreciation of the complexities of gonadotropin synthesis, storage, and release can be obtained in reviews focused on sheep and rat gonadotropes (91, 113). It is generally accepted that FSH and LH are differentially secreted, with FSH secretion primarily via the constitutive secretory pathway, whereas LH secretion primarily involves the regulated secretory pathway (113–116). In sheep pituitaries, FSH secretion was estimated to involve 60% to 80% of the pituitary content per 24 hours, but only 1% to 8% of LH (117). Synthesis and release of most pituitary FSH in contrast to storage of the bulk of LH is consistent with relative yields of both gonadotropins in biochemical studies. For sheep, pigs, and cattle, purified FSH yields are low, ranging from 2 to 3 mg/kg pituitary tissue, whereas 100 to 300 mg/kg recoveries of purified LH are readily obtained (118, 119). This pattern is not universal, as horse pituitaries yield 60 mg/kg purified equine FSH and 110 mg/kg purified equine LH (120).

In the rat pituitary, dual LH and FSH immunostaining revealed three populations of gonadotropes with regard to gonadotropin content, 21% to 35% for LH-only, 10% to 22% for FSH-only monohormonal cells, and 45% to 69% for bihormonal cells (114). The relative proportions are dynamic in cycling females, with the proportion of bihormonal cells reaching a maximum just before the preovulatory gonadotropin surge (114). Sheep gonadotropes fall into two groups, LH-only and bihormonal, with only the occasional FSH-only cell detected at the end of the preovulatory surge (121). At the light microscope level, FSH and LH appear differentially packaged in bihormonal gonadotropes, as they respond differentially to GnRH stimulation by aggregating in different regions of the cell (122). Immunogold labeling of rat gonadotropes detected both LH and FSH in secretory vesicles during all stages of the estrus cycle (114). Monohormonal LH vesicles were also observed and reached a maximum prior to the LH surge (114), whereas FSH-only vesicle abundance increased prior to rises in serum FSH (114). In contrast, bihormonal sheep gonadotrope secretory granules were either LH-only or both LH and FSH (123). The latter are consistent with GnRH stimulation of simultaneous LH and FSH pulses observed in sheep portal blood (124). The nearly complete absence of FSH-only granules is inconsistent with GnRH-independent FSH pulses lacking a corresponding LH pulse in the same study, but it may reflect the use of intact animals in one study (123) vs ovariectomized animals in the other (124). Constitutive FSH release may render FSH-only granules very rare. The immediacy of FSH synthesis and secretion may permit changes in glycosylation to influence current physiological change, whereas storage of LH may result in glycosylation patterns that differ from conditions prevailing at the time of release.

Discovery of sulfated oligosaccharides on LH led to studies regarding carbohydrate-mediated intracellular targeting; however, these studies showed no carbohydrate-dependent sorting mechanisms (125). LH secretion by the regulated secretory pathway appears dependent on a C-terminal heptapeptide, Leu-Ser-Gly-Leu-Leu-Phe-Leu (126). This sequence was not detected during sequence determination by Edman degradation, as the C-terminal sequence was poorly defined beyond the last Cys residue (127–129). First predicted by sequencing the LHB gene (130), evaluation of LH peptides by mass spectrometry identified this peptide in LHβ derived from recombinant LH (68). LH and FSH appear to diverge early in the secretory pathway, rather than later in the trans-Golgi network, where much protein sorting is accomplished. Expressed in the absence of α-subunit, FSHβ appeared in peripheral ER, where its distribution resembled that of calnexin (131). LHβ was sequestered in a perinuclear ER compartment where it was significantly colocalized with the HSP70 family chaperone BiP (immunological heavy chain–binding protein, or HSPA5). Coexpression of the α-subunit with LHβ resulted in displacement of the latter to the peripheral ER, most likely as a heterodimer. As LH and FSH glycan populations are quite different in their branching patterns, their paths may remain separate. Their exposure in the medial Golgi to oligosaccharide branch-initiating GlcNAc transferases I to VI (132) appears to differ, resulting in more extensive branching of FSH glycans than LH (133). Adding the seven-residue, C-terminal LHβ peptide to the FSHβ C terminus directed the mutant FSH to the regulated secretory pathway (134). How the mutant FSH N-glycan population compares with that of wt-FSH is unknown.

Role of Glycosylation on FSH Clearance

Effects of individual FSH glycosylation sites on FSH clearance have been tested with recombinant hFSH expressed in CHO cells (9). Although loss of either one or both α-subunit oligosaccharides appeared to accelerate clearance, the results were not significant. Furthermore, there was no significant loss of in vivo biological activity in the Steelman–Pohley assay for either mutant FSH derivative as compared with wt-FSH. Each FSHβ N-glycan significantly increased clearance and reduced in vivo biologic activity, whereas mutation of both sites dramatically increased clearance and appeared to eliminate FSH activity in the Steelman–Pohley assay.

Recombinant FSH expression by CHO cell lines deficient for GlcNAc transferase I (Man₅GlcNAc₂ oligosaccharides) or sialic acid (complex glycans terminated with Gal) resulted in rapid clearance from serum, effectively eliminating in vivo biological activity as measured by ovarian aromatase expression (78). In vitro steroidogenic activity was unaffected by either change in oligosaccharide structure. Selective inhibition of ST3 β-galactoside α2–3-sialyltransferase (ST3GAL) expression by estrogen has been demonstrated in female rat pituitary glands (135). In contrast, ST6 β-galactoside α2–6-sialyltransferase (ST6GAL) expression remained unaltered. The reduction in ST3GAL may shift the ratio of α2–6 to α2–3 linkages in favor of α2–6, thereby increasing clearance rates, as α2–3-linked sialic acid appears more effective than α2–6-linked sialic acid at preventing FSH clearance in the liver (61). A caveat in humans is the reduced impact of the liver Ashwell asialoglycoprotein receptor on glycoprotein clearance as compared with rodents (136).

FSHβ Carbohydrate Effects on FSHR Binding

Chemical deglycosylation of FSH showed little or no impact on receptor binding (137, 138). FSH scarcity precluded studying the effects of deglycosylation on each subunit. Site-directed mutagenesis provided FSH glycoform models, but it also provided a cautionary tale, as mutations near the βAsn²⁴ site appeared to alter FSH conformation (44). Individual mutation of either the βAsn⁷ or βAsn²⁴ glycosylation sites increased receptor binding affinity in one study (44) and left it unchanged in another (10) Three groups reported mutating both FSHβ glycosylation sites increased FSHR binding (10, 44, 45). A report involving Cos7-expressed FSH glycosylation mutants and human FSHR expressed in Y-1 cells found no increase in FSHR binding by elimination of αAsn⁷⁸, βAsn⁷, or βAsn²⁴ oligosaccharides, but a 2.4-fold increase following elimination of glycosylation at αAsn⁵² (11). In FSH glycoforms, partial hFSHβ glycosylation increased FSHR binding 3- to 20-fold (34, 55). The biggest difference in FSHR binding was associated with major changes in oligosaccharide structure, as the FSH21/18 preparation was decorated with oligomannose glycans (54). Although it is not known whether fucosylation is restricted to the FSHβ subunit in this preparation, as it appears to be in complex glycan-decorated FSH preparations, it is tempting to speculate a site-specific location for these glycans on αAsn⁵², implying altered binding associated with a specific glycan structure. Alternatively, recombinant FSH expressed by a CHO cell line lacking GlcNAc transferase I, thereby decorated with oligomannose type glycans at all four sites, retained equivalent FSHR binding activity as wt-recombinant FSH (78).

During kinetic receptor-binding studies involving FSH glycoforms, FSH21 engaged the FSHR immediately, occupied more binding sites, and reached maximum binding an hour before FSH24 (34). The latter exhibited a >30-minute lag before significant FSHR binding was observed, and then binding proceeded at the same rate as FSH21, but required a total of 4 hours before reaching a submaximal level as compared with that which FSH21 attained in 3 hours. Saturation binding studies revealed that FSH21 occupied twofold to threefold more receptors than did FSH24. The mechanisms involved in carbohydrate inhibition of FSHR binding are unknown. It is unlikely that FSH oligosaccharides influence binding to the high-affinity site in the FSHR extracellular domain (FSHR_ECD), as both FSH/FSHR_ECD structures show the FSH polypeptide moiety between the receptor and the carbohydrate (139, 140). Neighboring FSHRs may be involved instead.

Cocrystallization of FSH with the entire FSHR_ECD produced FSH/FSHR_ECD trimers (140). The endoglycosidase F-digested FSH and FSHR_ECD preparations employed in the crystallization studies possessed only single GlcNAc residues at each glycosylation site. The organization of the FSHR trimers in the model suggested that a biantennary αAsn⁵² oligosaccharide would prevent more than one FSH ligand binding to an FSHR trimer at a time. Support for this concept was provided by saturation binding studies with wt-recombinant hFSH and αAsn⁵² glycosylation site mutant hFSH, as the latter occupied threefold as many sites as did wt-hFSH (141). It was noted that FSHR allosteric modulators increased FSH binding threefold (142). The nature of the allosteric modulator did not matter, as a negative allosteric modulator (143), positive allosteric modulator (144), and nonfunctional allosteric modulator [an LH allosteric modulator that also bound FSHR, but did not affect FSHR activity (145)] all increased ¹²⁵I-hFSH binding threefold. In a subsequent report, an allosteric FSH modulator increased FSHR binding to β-arrestin threefold (141). These data suggested a common allosteric effect, dissociating FSHR trimers, thereby making more monomeric FSHR available for glycosylated FSH to bind. Although it is well-established that FSHR exits as at least dimers (146), studies of negative cooperativity involving glycoprotein hormone receptors suggest that at least a portion of FSHRs are monomeric (147). Application of super-resolution light microscopic procedures, such as photo-activated dyes and localization microscopy used to quantify LH receptor oligomerization in the cell membrane (148), may be useful to clarify FSHR oligomer patterns.

Glycosylation and FSH Signaling

Studies of signaling by glycosylation-impaired FSH have largely been restricted to the G_s–cAMP–PKA–CREB steroidogenesis pathway. The primary role of FSH αAsn⁵² in FSH-mediated steroidogenesis, initially recognized in hCG (149), has been consistently demonstrated for FSH in rat gonadal cells, whereas αAsn⁷⁸ glycosylation appeared to have no impact on FSH biological activity when each glycan was eliminated by itself and tested against rat FSH target cells (10, 44, 45). In Y-1 cells that expressed human FSHR, only elimination of αAsn⁵² oligosaccharide in Cos7-expressed recombinant hFSH reduced steroidogenesis when FSH concentrations yielding equivalent FSHR binding were employed (11). For FSHβ glycosylation sites, elimination of either one or both Asn⁷ and Asn²⁴ glycans increased steroidogenic potency in rat granulosa cells (44). Experiments involving βAsn⁷ were fairly straightforward, as Thr⁹ mutation to Ser preserved steroidogenic activity along with N-glycosylation. Mutation at Thr²⁶ reduced signaling regardless of whether the mutation preserved glycosylation of Asn²⁴ (44). These results suggest naturally occurring FSH21 is preferable to βAsn²⁴ glycosylation site mutants. In Y-1 cells, there was no effect of eliminating either βAsn⁷ or βAsn²⁴ glycosylation on steroidogenesis for single glycosylation site mutants, although double mutants were not tested (11). A potentially significant difference in the responsiveness of gonadal granulosa cells and Sertoli cells, as compared with Y-1 cells, is the greater number of receptors expressed by Y-1 cells as compared with physiological target cells. In a human ovarian granulosa cell line, recombinant FSH21/18 was more effective than FSH24 in stimulating cAMP accumulation, PKA activation as indicated by PKA substrate phosphorylation, and CREB phosphorylation (150). FSH21/18 also stimulated more CRE-mediated transcriptional activity, androgen conversion to estradiol by aromatase, and progesterone biosynthesis. In Fshb-null mice, FSH21/18 exhibited significant in vivo biological activity, often exceeding that of FSH24 (151).

FSHR has been demonstrated to activate additional cellular signal pathways via other trimeric G proteins, such as G_i and G_q, by β-arrestin, as modulated by GRK5/6 or GRK2/3, and by pleckstrin homology domain, phosphotyrosine-binding domain, and leucine zipper motif isoform 1 (APPL1) (152, 153). Whereas pituitary and recombinant hFSH preparations provoked the usual sigmoidal dose-response curves for cAMP accumulation in a CHO cell line stably transfected with human FSHR, chemically deglycosylated and baculovirus-expressed (oligomannose glycans only) FSH preparations provoked biphasic responses (13). The increases in cAMP occurred at the same concentrations as those for both pituitary and recombinant FSH, whereas the decreases occurred at 1000-fold higher FSH concentrations. Stimulation of G_s with cholera toxin increased cAMP accumulation over that stimulated by a fixed concentration of pituitary FSH, but it did not prevent its decrease in the face of high concentrations of baculovirus-derived FSH. Inhibition of G_i with pertussis toxin did not affect the pituitary FSH-stimulated cAMP accumulation, but it prevented the cAMP decrease at elevated baculovirus FSH concentrations, indicating G_i involvement, but only at higher FSH concentrations. In a separate study, baculovirus recombinant hFSH provoked a biphasic progesterone response in rat granulosa cells, but an αAsn⁵⁶-deglycosylated eFSH derivative provoked a sigmoidal response (154). Impaired G_s signaling was indicated by the 10-fold greater FSH derivative concentration needed to stimulate progesterone synthesis and submaximal steroid accumulation. The absence of biphasic response suggested impaired G_i activation as well as impaired G_s activation by the partially glycosylated FSH derivative.

Biased signaling via the FSHR was reported for an equine LH (eLH) derivative that modified both O- and N-glycosylation. Removal of the eLHβ C-terminal 25 residues and eLHα Asn⁵⁶ oligosaccharide (eLHβt) resulted in maintenance of FSHR binding, but complete loss of steroidogenic activity (155). Both progesterone and estrogen synthesis by DES-primed rat granulosa cells were completely eliminated by these two modifications to eLH (155). However, significant FSHR internalization was noted, despite that the FSHR activation of the PKA pathway appeared completely compromised (12). Recruitment of β-arrestin by eLHβt/FSHR binding in a cAMP-independent manner in turn activated ERK (12).

Studies using microarrays to evaluate global granulosa cell gene expression changes in response to FSH employed either recombinant hFSH isoforms (156) or compared recombinant hFSH (GonalF) with menopausal gonadotropins (urinary hFSH, hLH, and hCG) (157). It is likely that isoelectric focusing produced high and low sialic acid content FSH preparations only at βAsn⁷ because of the lack of heterogeneity at the other three glycosylation sites (Fig. 5). However, the hybrid/oligomannose population derived from concanavalin A lectin chromatography cannot be defined, as these glycans were missing from glycopeptide analyses. Accordingly, their structures and locations are unknown. Accordingly, significant differences in KGN granulosa cell gene expression cannot be related to glycosylation patterns (156). Comparison of the differences in branching between recombinant hFSH, once again, GonalF, and urinary FSH in menopausal gonadotropin stimulation of human granulosa cells derived from in vitro fertilization overlooked the contributions of hLH and hCG, as these cells express both FSH and LH receptors (157).

Conclusions

Significant progress has been made in defining FSH microheterogeneity using mass spectrometry, which offers both the high resolution and sensitivity needed to characterize the many glycans attached to FSH preparations. The addition of quantitative measures of glycan abundance helps measure changes in FSH glycosylation that could only be guessed based on differences in charge. Challenges still remain such as the structural ambiguity regarding position and linkage in the many partially sialylated glycans that decorate FSH. Subunit dissociation and preferential ionization of the FSHα subunit provide major challenges for using mass spectrometry to measure FSH glycoform abundance. Although conflicting data exist, FSH oligosaccharides appear to hinder high-affinity binding to FSHR; however, the nature of this interference remains to be determined. Structure determination for the entire FSHR with and without FSH ligand may reveal as yet unknown interactions between carbohydrates and the transmembrane domain of the receptor. FSHR oligomerization may be responsible for glycan inhibition of binding, which will require structures of FSHR complexes rather than individual receptors. Additional FSHR-activated signaling pathways remain to be investigated to evaluate the effects of FSH glycosylation on them. These studies will require production of well-characterized FSH glycoform preparations, which remains a major challenge.

Glossary

Abbreviations:

15k-FSHβ

15,000 relative molecular mass FSHβ band

21k-FSHβ

21,000 relative molecular mass FSHβ band

24k-FSHβ

24,000 relative molecular mass FSHβ band

APPL1

adapter protein, phosphotyrosine interacting with PH domain and leucine zipper motif isoform 1

CG

chorionic gonadotropin

CHO

Chinese hamster ovary

eLH

equine LH

ER

endoplasmic reticulum

FSHR

FSH receptor

FSHR_ECD

FSH receptor extracellular domain

hCG

human chorionic gonadotropin

hFSH

human FSH

hLH

human LH

LC-MS

liquid chromatography–mass spectrometry

MALDI-MS

matrix-assisted laser desorption ionization–mass spectrometry

M_r

relative molecular mass

Neu5Ac

N-acetyl-neuraminic acid

Neu5Gc

N-glycolyl-neuraminic acid

NMR

nuclear magnetic resonance

O-Ac-Neu5Ac

O-acetyl-N-acetyl-neuraminic acid

OST

oligosaccharyl transferase

PhNCS

phenylthiohydantoin

STT3

staurosporine and temperature sensitivity 3

wt

wild-type