Follicle-stimulating hormone is one of three gonadotropins in the human glycoprotein hormone family. This hormone family is part of the cystine knot growth factor superfamily, a large group of homo- and heterodimeric signaling molecules (1). FSH plays a central role in reproduction, particularly in females. In the ovary, FSH stimulates follicle development and estrogen synthesis. In the testis, FSH maintains Sertoli cell function, which supports spermatogenesis. Although currently controversial (2, 3), FSH has been claimed to play a direct role in osteoporosis by stimulating differentiation of osteoclasts, which are responsible for removing bone (4). The idea put forth is that in the postmenopausal period when FSH levels rise, activation of osteoclasts results in bone loss. Reports of non-gonadal actions of FSH have recently been summarized (5).

Follicle-stimulating hormone is composed of two dissimilar, cystine knot motif glycoprotein subunits: a common α-subunit and hormone-specific β-subunit (Figure 1) (6). The FSHα subunit amino-acid sequence and disulfide bond organization, including a cystine knot motif, are identical to those in the other glycoprotein hormones, luteinizing hormone (LH), thyroid-stimulating hormone (TSH), and chorionic gonadotropin (CG) (7). However, the N-glycan populations at both glycosylated residues, Asn⁵² and Asn⁷⁸, differ from those of the other glycoprotein hormone α-subunits such that these otherwise identical subunits can be distinguished from each other and from free α-subunit by their oligosaccharide populations (8–10). The hormone-specific FSHβ subunit shares 34–40% sequence homology, six conserved disulfide bonds, cystine knot motif, and seatbelt loop with the other human glycoprotein hormone β-subunits (7, 11, 12). While there are two potential N-glycosylation sites in FSHβ, partially glycosylated variants exist that are missing either one of these oligosaccharides (13). These contribute to an unknown degree of charge variation in FSH preparations and result in the classic FSH isoforms (14, 15). The classic interpretation of FSH isoforms was based solely on the notion that variant patterns of negatively charged sialic acid or, to a much lesser extent, sulfate residues terminated oligosaccharide branches, which gave rise to differentially charged isoforms. The observation of hypo-glycosylation further refines our understanding of isoforms, in that net charge may vary, due to presence or absence of entire glycans.

Follicle-stimulating hormone glycosylation exhibits both macro- and microheterogeneity (Table 1). Macroheterogeneity herein refers to the presence or absence of glycosylation at any one potential glycosylation site. Examples of FSH macroheterogeneity involve the absence of either FSHβ Asn⁷ or Asn²⁴ oligosaccharides in a population of fully processed and secreted FSH. Microheterogeneity herein refers to as many as 80 to over 100 unique oligosaccharide structures, which can be detected once released from each of the 3–4 glycan-occupied Asn residues in FSH.

Differences in electrophoretic mobility of FSH subunits, revealed by subunit-specific Western blots, provide a convenient means to distinguish four FSH variants resulting from macroheterogeneity. Fully glycosylated hFSHβ migrates as a 24-kDa band (hereinafter, 24k-FSHβ), desN²⁴glycan-FSHβ migrates as a 21-kDa band (21k-FSHβ), and desN⁷glycan-FSHβ migrates as an 18-kDa band (18k-FSHβ). The FSH heterodimers that incorporate these β-subunit variants are designated, FSH²⁴, FSH²¹, and FSH¹⁸, respectively (19), and are shown in Figure 2. Pituitary extracts also possess a non-glycosylated, 15-kDa FSHβ variant (20). However, the corresponding FSH¹⁵ does not appear to be physiologically relevant, because subunit association is extremely inefficient when both FSHβ glycans are missing, and little, if any, FSH heterodimer is secreted (21). FSH²⁴ and FSH²¹ are detected in FSH derived from human pituitary extracts, as well as from urinary protein preparations (Table 1). When FSH is separated into fully- and hypo-glycosylated fractions, the latter often include FSH¹⁸, which can constitute as much as 40% of the hypo-glycosylated FSH preparation (13). As most hFSH²¹ preparations also possess hFSH¹⁸, and are not easily separated, it has become a convention to abbreviate the mixture of physiologically relevant hypo-glycosylated FSH preparations as hFSH^21/18.

Follicle-stimulating hormone microheterogeneity results from a structurally heterogeneous population of oligosaccharides attached to each glycosylated Asn residue of the four glycosylation sequons in FSH. Microheterogeneity in this hormone has largely been evaluated at the whole hormone level in studies of pituitary and urinary FSH preparations (16, 22–25). Human pituitary FSH oligosaccharides are 85–98% complex-type, 88–99% are sialylated, 36–46% are biantennary, 30–49% are triantennary, 5–15% are tetra-antennary, while only 4–7% are sulfated (Table 1). The low extent of oligosaccharide sulfation appears to be a human-specific characteristic (no data exist for nonhuman primate FSH glycans), as FSH preparations from cattle, pigs, sheep, and horses possess higher levels of sulfated oligosaccharides, ranging from 13 to 58% (23, 26). Accordingly, a major factor in determining hFSH clearance rates is the extent of sialic acid termination at the non-reducing ends of oligosaccharide branches. As compared with naturally occurring hFSH preparations, recombinant hFSH preparation oligosaccharides exhibit a reduced degree of branching, consisting of largely (55%) biantennary glycans. However, the degree of sialylation in these preparations lags that of urinary hFSH to a lesser extent, because the most abundant urinary FSH triantennary and tetra-antennary glycans are one sialic acid residue short of a full complement (16, 25, 27).

As mentioned above, microheterogeneity contributes to charge variation in FSH, and this has been reported to alter FSH biological activity (14, 28, 29). Comparisons of microheterogeneity in early studies were challenged not only by the large number of oligosaccharide structures encountered, but also by the different analytical methods each group employed, as each of these exhibited bias toward or against specific families of oligosaccharides. We recently characterized microheterogeneity in three purified human pituitary FSH glycoform preparations, as well as highly purified pituitary, urinary, and recombinant hFSH preparations using nano-electrospray mass spectrometry (13, 16–18). Because over 33–109 structures were detected in each sample, comparing oligosaccharide populations derived from different FSH preparations proved challenging.

The oligosaccharide structures shown in Figure 3 represent those present in at least 1% relative abundance in at least FSH preparation. Using this criterion, a total of 54 glycans were selected for comparison. The glycans are organized by position in the N-glycan biosynthetic pathway or by the number of complex branches. Within each antennary group, 2-, 3-, or 4-branch glycans, monosaccharide composition is the basis of organization. Structures 1–7 are oligomannose glycan intermediates found in ER and cisGolgi-derived glycoprotein precursors (Figure 3A). In multi-glycosylation site glycoproteins, these can be found in glycoproteins possessing mature glycans at other sites, when glycan processing at individual sites differs (30). Structures 8 and 9 exhibit the beginnings of complex oligosaccharide synthesis (Figure 3A), structures 10–34 are biantennary glycans (Figures 3A–C), structures 35–48 are triantennary glycans (Figures 3C,D), and structures 49–54 are tetra-antennary glycans (Figure 3D). The oligosaccharide populations of fully glycosylated FSH²⁴ and hypo-glycosylated FSH²¹ preparations, F and D, respectively, possessed 51 of the 54 major glycans identified in these studies, and 45 of these, representing 88% of these more abundant glycans, were detected in both preparations. Pituitary and urinary FSH preparations P and U, respectively, both possessed 38 glycans (75%) in common with glycoforms F and D, while the hypo-glycosylated hFSH^21/18 preparation L, possessed 35 glycans (68%) found in glycoform preparations F and D. Recombinant hFSH preparation G, expressed by stably transfected GH₃ cells, displayed the lowest qualitative similarity to FSH²⁴ and FSH²¹, possessing only 28 (55%) of the glycans found in glycoforms F and D. Moreover, the triantennary recombinant hFSH oligosaccharides displayed a different branching pattern.

Raising the cutoff to 4% relative abundance identified four groups of highly abundant glycans. The first group revealed a unique pattern of glycosylation for hFSH^21/18 preparation L, consisting of a series of high mannose oligosaccharide intermediates possessing 9, 8, 7, 6, 5, and 3 mannose residues (structures 1–7, Figure 3A). Taken in isolation, this observation suggests that these glycoforms may not have exited the biosynthetic pathway. However, complex oligosaccharides, identical to those found in all other FSH preparations examined in this study, were also present in hFSH-L, suggesting oligosaccharide processing occurred at least at one glycosylation site in the Golgi. Glycosylation site-specific glycan analysis, when sufficient samples are available, or top–down proteomics for limited samples, have the potential to demonstrate the presence of both oligomannose and complex glycans in the same hypo-glycosylated hFSH molecule to support this hypothesis. Oligosaccharide structures 2–7 were also found in two pituitary glycoform preparations, hFSH²⁴ and hFSH²¹. However, in both cases, these glycans were present in very low abundance, consistent with their being N-glycan biosynthetic intermediates. Moreover, both secreted hFSH preparations, urinary hFSH and recombinant hFSH, were devoid of oligomannose structures 1–7. In the case of urinary hFSH, this could have resulted either from rapid clearance of oligomannose-containing hFSH from the circulation or bias during purification.

As only secreted recombinant hFSH was recovered from conditioned medium, the absence of oligomannose glycans indicated that mature hFSH secreted by the GH₃ cell line possessed only complex N-glycans. Moreover, the antibody used to capture recombinant hFSH appeared to capture all FSH forms, reducing the likelihood of purification biasing the oligosaccharide population (13). The high abundance of biosynthetic intermediate and low abundance of complex glycans in hFSH^21/18 preparation L was notable because it exhibited the highest receptor binding-activity of any hFSH preparation we have studied. This led to the concern that we were studying a physiologically irrelevant glycoform. However, subsequent demonstration of significant biological activity differences between other pituitary and recombinant FSH glycoform preparations eliminated this concern (18, 31, 32).

Three clusters of high-abundance, complex glycans were noted in the other five hFSH preparations comprising oligosaccharide structures 22–23, 31–34, and 38–42. Group 2 structure 23, a disialylated, biantennary glycan possessing one GalNAc substituted for Gal, was highly abundant in all five preparations. This was notable, because the absence of sulfated GalNAc from hFSH N-glycans has been attributed to impaired recognition of a Pro-Leu-Arg motif in the common α-subunit of hFSH by β1,4-N-acetylgalactosaminyltransferase-T3 and -T4 (βGalNAct-T3 and βGalNAc-T4, respectively), as compared with hCG and hLH. The resulting reduction in FSH oligosaccharide sulfation was proposed as a consequence of altered motif access in this hormone, probably due to conformational change (33).

Comparison of Pro-Leu-Arg motifs in both hCG crystal structures, 1hcn (12) and 1hrp (11), with those in the two hFSH structures found in 1fl7 (6) showed positions of the Pro⁴⁰ and Leu⁴¹ residue side chains were very similar in all six possible alignments (Figure 4). The Arg⁴² side chains were closely aligned in only one comparison, u-hCGα2:r-hFSHα1 (Figure 4D), suggesting flexibility in that region of the subunit (6, 34). Indeed, molecular dynamics simulations of FSH bound and unbound to the FSH receptor (FSHR) high-affinity binding site support flexibility in residue 40–47 region as unbound FSH exhibits root mean square fluctuations >1 Å (35). Unbound FSH is the form of the heterodimer recognized by β4GalNAc transferases. When FSH is bound to FSHR, this region loses flexibility, indicating it can achieve a stable conformation when bound to another protein. Thus, pituitary βGalNAc transferases are likely to bind this motif in both hLH and hFSH, consistent with the widespread distribution of GalNAc in hFSH oligosaccharides. The frequent appearance of GalNAc in sulfate-deficient glycans suggests an alternative hypothesis to explain reduced sulfation; human sialyltransferases compete more effectively with sulfotransferase in the human pituitary, leading to preferential addition of Neu5Ac to GalNAc. As N-glycan branches terminated with Neu5Ac-GalNAc were first reported for hLH oligosaccharides, finding this type of glycan is not unprecedented (36).

In fact, hLH possesses the greatest abundance of sialic acid of all characterized mammalian LH preparations (23, 36, 37). Moreover, structure 23 is part of a series of 15 GalNAc-containing, biantennary glycans observed in at least one of the six hFSH preparations (structures 10–25, Figures 3A,B). While two other structures are possible for the m/z 1130.9 ion associated with structure 23 (17), they do not permit addition of the two sialic acid residues associated with this oligosaccharide because the 5th hexosamine in the alternative structures is a bisecting GlcNAc residue and the single antenna possessing a Gal residue provides attachment for only one Neu5Ac residue. Group 3 glycan structures 31–34, are conventional, disialylated, biantennary oligosaccharides in which Neu5Ac residues are attached to Gal residues (Figure 3C). Structures 31 and 32 were the most abundant oligosaccharides derived from recombinant, urinary, and pituitary hFSH (Figure 3C). As 85–100% core-fucosylated glycans are found on the other human pituitary hormone LHβ and TSHβ subunits, structure 31 most likely reflects FSHα subunit glycosylation, while structure 32 reflects FSHβ subunit glycosylation (36, 38). The 4th high abundance glycan cluster, comprising structures 38–42, includes triantennary oligosaccharides possessing only two sialic acid residues. For this group of oligosaccharides, recombinant hFSH differed in the location of the two branch-mannose residues. In pituitary hFSH, GlcNAc transferase IV initiated a third glycan branch on Man (α1–3), while in recombinant hFSH GlcNAc transferase V initiated a third branch on Man (α1–6) (Figure 3C, compare row G with the other five rows). This suggested a difference in the relative activities of GlcNAc transferases IV and V between pituitary gonadotropes and somatotrope-derived GH₃ cells, despite the expression of both transferase genes in GH₃ cells (18). Another feature of recombinant hFSH glycans was antenna-linked fucose residues, such as observed in structure 43, one of the >1% abundance class of oligosaccharides (18).

The FSHR is a G-protein-coupled receptor (GPCR) with a leucine-rich repeat extracellular domain comprising 358 amino-acid residues. This ligand binding domain is connected to a 337-residue, hepta-helical transmembrane domain (39, 40). Crystal structures of the high-affinity FSH binding domain in complex with FSH revealed that the interface of the complex involves contacts exclusively via protein–protein interactions (41, 42). FSH oligosaccharides added by modeling do not appear to interact with the extracellular domain engaged with FSH, as they are located on a face of the hormone, which is oriented away from the hormone receptor interface (Figure 5). Since it is well established that FSH carbohydrate is necessary for full FSHR activation (43–46), it seems reasonable to assume that the carbohydrate affects hormone conformation, which in turn modulates activity. The structure of the entire FSHR (extracellular domain and transmembrane domains) in complex with FSH has yet to be determined, and until then, carbohydrate interaction with the transmembrane domain cannot be ruled out. Alternatively, carbohydrate modulation of FSH conformation may affect the final disposition of FSHR extracellular domain (FSHR_ECD) hinge region putative interactions with extracellular loops of the transmembrane domains (34, 47). Consistent with the absence of FSH carbohydrate interaction with FSHR_ECD, isolated hybrid-type oligosaccharides related to structure 12 in Figure 3 have no effect on FSHR binding (48). Nevertheless, these oligosaccharides significantly inhibit both basal granulosa cell steroidogenesis, as well as FSH-stimulated steroidogenesis (48). The low affinity of carbohydrate–protein interactions requires sufficiently high oligosaccharide concentrations in inhibition studies that hormone contamination can inhibit binding assays. In our hands, a minimum of two purification steps is necessary to eliminate residual hormone assay interference (48). Accordingly, we attributed hormone contamination in the oligosaccharide preparation as the reason for a report that hCG-derived oligosaccharides inhibited both receptor binding and cellular activation (49).

Loss of a single FSHβ oligosaccharide has three effects on FSH binding to its receptor. First, hypo-glycosylated hFSH immediately engages FSHR preparations, whereas fully glycosylated hFSH²⁴ exhibits about a 30-min lag before FSHR binding begins in earnest (13). Second, hypo-glycosylated hFSH^21/18 exhibits a 2.8- to over 14-fold higher apparent affinity for the FSHR as compared with hFSH²⁴ (Table 2). Third, hypo-glycosylated hFSH^21/18 occupies 2- to threefold more FSHRs than FSH²⁴ (13, 18). A glance at the structures of FSH glycoforms bound to the FSHR_ECD immediately raises the question of how loss of either FSHβ N-glycan facilitates FSH association with the receptor, as neither glycan is close to the binding site (Figure 5). This leaves yet to be defined hindrance by the FSHR transmembrane domain or FSHR oligomerization as potential mechanisms.

The crystal structure of the high-affinity binding site of the FSHR_ECD comprised two FSHR domains associated back to back, sandwiched by FSH ligands (41). There was no indication of FSH oligosaccharide interaction with the receptor. The crystal structure of the entire FSHR_ECD with FSH bound revealed a strikingly different FSHR_ECD conformation as trimeric FSHR–FSH complexes (42). To obtain diffractable crystals in both studies, endoglycosidase-F digestion reduced FSH and FSHR_ECD N-glycans to single GlcNAc residues, which eliminated oligosaccharide influence on hormone-receptor binding. The trimeric FSHR crystal structure suggested FSH αAsn⁵² oligosaccharide, when present, would restrict ligand binding to one glycosylated FSH ligand per FSHR trimer as a biantennary glycan attached to this Asn residue would occupy the center of the trimeric complex (47). While no subsequent studies supporting the dimeric FSHR model have been reported, several lines of evidence appear to support the trimeric FSHR_ECD model. Biochemical data in support of the trimeric FSHR model were provided when recombinant-mutant des-αN⁵²-hFSH exhibited threefold greater binding to CHO cells expressing hFSHRs as compared with recombinant wt-hFSH (47). Small molecule allosteric FSHR modulators were reported to increase FSH binding ~threefold, suggesting trimeric FSHR complexes dissociating to form FSHR monomers (50–52). Incorporating a transmembrane domain model to the FSHR_ECD trimer model predicted that only a single β-arrestin could bind to the trimeric FSHR. Addition of an allosteric modulator to β-arrestin binding assays produced a threefold increase in β-arrestin binding, supporting a model that allosteric small molecule FSHR modulators dissociate FSHR trimers into monomers, thereby increasing FSH access (47). However, a superresolution microscopic technique, dual-color photoactivatable dyes, and localization microscopy (PD-PALM) revealed the closely related LHR existed as a variety of oligomeric forms as well as monomers in the cell membrane (53). Docking of complete LHR models in this study provided a variety of conformations of LHR oligomers, including trimeric LHRs. Similar studies with FSHRs would help clarify the relationship of FSHRs.

As greater FSHR occupancy is directly proportional to FSH-stimulated cAMP production by target cells, increased hypo-glycosylated hFSH binding to FSHR is expected to provide a correspondingly greater cellular activation than fully glycosylated hFSH (54). However, since the model of an FSHR trimer can only accommodate one G protein, it is unlikely that the increase in cAMP is due to occupancy alone. Another possibility is that occupancy by hypo-glycosylated FSH fails to engage the GRK/arrestin pathway which would otherwise attenuate the reengagement of G protein subsequent to activation of adenyl cyclase. Another possibility is that hypo-glycosylated FSH creates a more stable complex with FSHR such that during intracellular trafficking, cAMP- and arrestin-mediated persistent signaling (55) is enhanced. Finally, one may also suggest that since the FSH/FSHR complex appears to recycle to the cell surface (56, 57), the high-affinity binding of hypo-glycosylated FSH may have a proclivity for FSHR, thus failing to dissociate upon relocation to the plasma membrane and perhaps reformation of the putative trimeric structures. This could affect the dynamic stoichiometry of the cell surface unoccupied receptor cohort whose ontogeny resets not only with new FSHR synthesis but also by occupancy/recycling engaged by other members of the orchestra¹ of glycoforms.

Biased signaling has underpinned GPCR drug development for years but only recently has the mechanism of this phenomenon been revealed in the GPCR field, including the FSHR (51, 58–60). The realization that one GPCR can activate several effector proteins to activate different pathways has prompted the challenging of previously accepted dogma and may help to explain previously unexplained observations. An example of such dogma is that both FSHR and LH/CGR primarily signal via Gαs leading to the activation of the cAMP/protein kinase A (PKA) pathway and subsequently leading to steroidogenesis (51, 61–64). Alternative pathways, such as phospholipase C/inositol trisphosphate metabolism were first recognized over 25 years ago (65, 66); however, most studies examining the actions of gonadotropin glycosylation variants remain fixed on the primary pathway. The concept of biased signaling predicts that the specificity of signal transduction depends on, at least in part, the structure of the ligand [reviewed in Ref. (58, 59)]. In support of this idea, a partially deglycosylated eLH variant (67) (eLHdg) was found to exhibit biased signaling through the FSHR (68). While incapable of activating the cAMP/PKA pathway and eliciting steroidogenesis in granulosa cells, binding of eLHdg to FSHR recruited β-arrestins and activated ERK MAPK signaling via a cAMP-independent pathway (68).

Another recent study showed that the oligosaccharide complexity of recombinant hFSH preparations differentially affected gene expression and steroidogenesis in human granulosa cells (69). Our own studies with hFSH glycoforms have found evidence for biased signaling, albeit in different cell types. The hFSH^21/18 glycoforms were more active than hFSH²⁴ in activating the cAMP/PKA pathway and phosphorylation of PKA substrates via Gαs in human KGN granulosa cells (31). The actions of FSH^21/18 were 10-fold greater than FSH²⁴ on induction of CYP19A1 and estrogen (31). The obvious next step is to determine if this biased signaling by hFSH²⁴ occurs in gonadal cells, which is an active area of pursuit using both in vitro and in vivo genetic approaches.

GB, JM, JSD, JD, and TK authored individual sections of the review. GB, JM, JD, JD, and TK reviewed the entire manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.