The Emerging Importance of IgG Fab Glycosylation in Immunity

Immunoglobulins are glycoproteins produced by B cells that play a crucial role in protective immunity. Both the Fc tail, responsible for triggering effector functions, and the Fab arms, responsible for Ag binding, may contain glycans (Fig. 1). In addition to a conserved N-linked glycan in the Fc of all Ig classes, N-linked glycans are found in the C_H1 domains of IgM and IgE, and O-linked glycans are found in the hinge regions of IgD, IgA (1), and IgG3 (2). Furthermore, the variable domains of Igs may also contain N-linked glycans. For IgG, the most abundant class of serum Igs, Fab glycosylation is restricted to the variable domains.

Although the existence of IgG Fab glycans has been known for quite some time, their emergence, regulation, and function remain poorly understood. Nevertheless, recent evidence shows that Fab glycans exhibit distinct patterns according to the pathophysiological conditions and have immunomodulatory effects, thus highlighting their potential role in regulating immunity. In this article, we summarize the literature on the structural features of IgG Fab glycans, Fab glycosylation during physiological and pathological conditions, the influence of Fab glycosylation on IgG function, and immune modulation.

N-linked glycosylation of IgG Fab is contingent on the presence of consensus amino acid motifs making up the so-called “N-linked glycosylation sites.” These sites consist of asparagine, followed by any amino acid but proline, and either serine or threonine (N-X-S/T). N-linked glycosylation sites are largely absent in the naive human B cell repertoire because few germline-encoded alleles (IGHV1-8, IGHV4-34, IGHV5-10-1, IGLV3-12, and IGLV5-37) contain such sites (3). Therefore, the relatively high frequency of IgG Fab glycans is primarily the result of somatic hypermutation during Ag-specific immune responses (4). Fab glycosylation sites can appear in both H and L chains, in CDRs, and in framework regions (5).

Although a glycosylation site is required, it is not sufficient for the addition of a glycan. For instance, for IGHV4-34, one of the more frequently used human V_H alleles, the germline-encoded glycosylation site is usually unoccupied (6). Similarly, the mouse IGKV5-45 allele contains a germline-encoded glycosylation site that remains unoccupied in both the therapeutic chimeric Abs infliximab (7) and cetuximab (8, 9). In contrast, the mouse IGHV2-2 germline-encoded glycosylation site in cetuximab carries glycans (8–10).

Estimates of the percentage of Fab-glycosylated IgGs in healthy individuals range from ∼15 to 25%, depending on the experimental approach used. About 20% of IgG variable domains from Ig sequence database entries contain N-linked glycosylation sites (11). Using sialic acid–binding lectin-affinity chromatography, 25% of human serum IgGs were reported to contain Fab glycans (12), calculated as the percentage of sialylated Fab arms (10–12%) (13) divided by the percentage of Fab glycans containing sialic acid residues (46%) (12, 14). However, high-resolution analytical techniques yield a higher degree of sialylation (79–93%) (15–17), which equates to ∼15% IgG Fab glycosylation. Furthermore, using high-resolution HPLC with 2-aminobenzoic acid labeling, ∼14% Fab glycosylation was found (17).

N-linked glycans at both the Fc and Fab of polyclonal serum IgG are mainly complex-type biantennary N-linked glycans (Fig. 1). The core heptasaccharide consists of two N-acetylglucosamine (GlcNAc) residues, three mannose residues, and two more GlcNAc residues. In addition, fucose, bisecting GlcNAc, galactose, and sialic acid residues can be present in the glycan. Compared with IgG Fc glycans, IgG Fab glycans contain high percentages of bisecting GlcNAc, galactose, and sialic acid and low percentages of core fucose (15–17) (Fig. 1). Furthermore, although only 4% of polyclonal IgG Fab glycans are high-mannose glycoforms (versus 0% for Fc glycans), these structures can predominate on certain mAbs, depending on their precise position (see next section) (18). The high degree of sialylation of Fab glycans of serum IgG might stem, in part, from selective removal of nonsialylated structures via the hepatic asialoglycoprotein receptor (see “Ab Half-Life” below). Furthermore, Fab glycans are presumably more accessible for glycosyltransferases, resulting in more processing compared with Fc glycans that are spatially localized at the inner face of the C_H2 domains. Indeed, lectin- and Ab-binding data indicate that Fab glycans are highly exposed, whereas Fc glycans are (partially) shielded (19, 20). Nonetheless, enzymatic-digestion experiments suggested that Fc glycans are more accessible, at least to some enzymes, than Fab glycans (17, 21).

Several factors modulate the presence and structure of IgG Fab glycans. First, the glycoprofile varies, depending on the position of the glycosylation site (22, 23) and amino acid residues surrounding it (24–27). For instance, introducing glycosylation sites in the CDR2 of otherwise identical Abs resulted in complex-type biantennary glycans at asparagines 54 and 58 but high-mannose structures at asparagine 60 (22). Furthermore, Fab glycosylation may be regulated by the mode of B cell activation. For instance, IL-6 and progesterone can enhance the expression of oligosaccharyltransferase, which catalyzes attachment of the N-linked glycan precursor to the polypeptide chain in the lumen of the endoplasmic reticulum, resulting in increased IgG Fab glycosylation with a proportional increase in high-mannose structures up to 30% (28–30). T cell signaling, known to influence the structure of Fc glycans (31), may also affect Fab glycosylation, but this has not been studied. In addition, endogenous lectins may serve as surrogate B cell Ags via binding to Fab glycans, as described for certain B cell lymphomas (see “Malignancies” below).

The amounts and types of Fab glycans can vary during certain physiological and pathological conditions, suggesting that they might contribute to immune suppression or pathophysiology or potentially serve as a biomarker, as discussed below.

Margni and Binaghi (32) investigated the properties of IgG Abs that are retained on a ConA affinity resin (5–15%) and concluded that their Fab arms contain high-mannose glycans. However, ConA also retains complex-type biantennary N-linked glycans without bisecting GlcNAc; most likely, the retained IgG Abs contain both types of Fab glycans (33, 34). Interestingly, levels of ConA-binding Abs are increased during the second trimester of pregnancy. Moreover, reduced serum levels of ConA-binding Abs were found during the first and second trimester of pregnancy in women who subsequently suffered from spontaneous abortions, thus serving as a potential early marker for diagnosing a threatened pregnancy (35–37). Consistent with these results (given that ConA interacts preferentially with nonbisected complex-type biantennary N-linked glycans), mass spectrometry data show that IgG Fab glycans contain more sialic acid and less bisecting GlcNAc during pregnancy compared with after delivery (15). Similar shifts in sialic acid content are observed for Fc glycans.

Intriguingly, it was proposed that only one Fab arm of the ConA-binding Abs carries a glycan, which might minimize elimination of paternal (fetal) Ags by maternal Abs during pregnancy (see “Ab Aggregation and Immune Complex Formation” below) (32). In line with this, the majority of Fab-glycosylated IgG Abs from the mother is directed against paternal Ags (38). Although firm evidence is lacking, the hypothesis that Fab glycosylation can modulate the Ab repertoire to minimize unwanted reactivity toward the fetus is worth exploring.

In contrast to the supposedly protective role of Fab glycosylation during pregnancy, enhanced IgG Fab glycosylation is associated with several autoimmune diseases. Using a small number of individuals/patients (n = 7), Youings et al. (39) reported that the Fab of IgGs from rheumatoid arthritis (RA) patients carry three times more oligosaccharides than do those from healthy individuals. Moreover, they observed an increase in monogalactosylated glycoforms containing bisecting GlcNAc and fucose residues in the RA patients (39). Furthermore, we recently found that anti-citrullinated protein Abs, a diagnostic and prognostic biomarker in RA, have a higher m.w. compared with other IgG (auto)antibodies as the result of an increase in IgG Fab glycans (40). Nevertheless, although IgG Fc-glycosylation patterns were associated with disease activity/severity and outcome (41, 42), such associations for Fab glycans still need to be established. Other studies suggested that differences in Fab glycosylation may also characterize other types of autoantibodies. For instance, enhanced and reduced levels of Fab glycosylation were described for anti-neutrophil cytoplasmic Abs and anti-glomerular basement membrane Abs, respectively (43). Furthermore, primary Sjögren’s syndrome patients show a higher prevalence of IgG B cell sequences with N-linked glycosylation sites than do healthy controls (44). It was suggested that B cell hyperproliferation and selection within the parotid glands may result from Ag-independent interactions of glycosylated BCRs with microbial lectins, as described for lymphoma B cells (see below).

B cell malignancies are sometimes characterized by aberrant Fab glycosylation. Increased frequencies of N-linked Fab-glycosylation sites in IgG and/or BCR were described for follicular lymphoma, diffuse large B cell lymphoma, and Burkitt’s lymphoma B cells, whereas mutated chronic lymphocytic leukemia, multiple myeloma, and MALT lymphoma B cells were comparable with normal B cells (5, 45, 46). These tumor-associated Fab glycans are high-mannose structures, whereas the corresponding Fc glycans are complex-type biantennary glycans, confirming that the normal N-linked glycan-processing pathway is intact (6, 47). Signaling through interactions between N-linked glycans on follicular lymphoma BCRs and lectins, such as DC-SIGN or mannose-binding lectin, at the surface of macrophages and dendritic cells may free these cells from dependence on Ag and may contribute to tumor cell persistence or growth (48, 49). Some lectins from opportunistic pathogens, such as Pseudomonas aeruginosa and Burkholderia cenocepacia, can also interact with high-mannose BCR Fab glycans (50) (Fig. 2A). Increased IgG Fab glycosylation also was observed in myeloma patients, accompanied by an increased proportion of sialic acid, bisecting GlcNAc, and fucose (51). Furthermore, variable domain glycosylation of Bence Jones proteins (free Ig L chains) was 4-fold higher in multiple myeloma patients with amyloidosis than those without amyloidosis, suggesting a role for variable domain glycans in pathogenesis (see “Ab Aggregation and Immune Complex Formation” below) (52).

In addition to B cell malignancies, it was shown that, during the progression of malignant melanoma, autoantibodies to GRP78 are expressed, and changes in their oligosaccharide chains (putative increases in Fab glycosylation) stimulate melanoma cell growth and survival (53).

Several studies showed that the presence of N-linked glycosylation sites in the variable domains can increase (18, 23, 54–56) or decrease (18, 23, 50) Ag-binding affinity. For instance, the anti–α(1→6)-dextran Ab 14.6b.1 has a 10-fold higher affinity compared with a single amino acid mutant that lacks the N-linked glycosylation site (54), presumably due to additional hydrophilic interactions of the Fab glycans with the carbohydrate Ag (Fig. 3). Interestingly, insertion of alternative N-linked glycosylation sites by mutagenesis around the pre-existing N-linked glycosylation site differentially affects the Ag binding, depending on the location of Fab glycans. Furthermore, a 100-fold reduction in binding to tetanus toxoid, diphtheria toxoid, and dsDNA was observed for the polyreactive human mAb CBGA1 upon blocking N-linked glycosylation during production by tunicamycin (56). Decreased binding was explained by a conformational change in the variable domains in the absence of Fab glycans. In addition, the removal of sialic acid residues from Fab glycans can decrease Ag-binding affinity (57). In contrast, insertion of follicular lymphoma–specific N-linked glycosylation sites into the model BCRs B1-8 and HyHEL10 strongly impairs the binding to their respective Ags (50), and the presence of N-linked glycans in the Fab of an anti-CD33 Ab decreases its binding to CD33 by 3–8-fold (58). Fab glycans may also leave Ag-binding affinity essentially unaffected (23).

Fab glycans may also alter Ab activity without affecting Ag binding. LE2E9, an anti-factor VIII (FVIII) Ab derived from a patient with mild hemophilia A, neutralizes FVIII activity up to 90% and prevents FVIII binding to von Willebrand factor (vWF). Removal of the Fab glycans from the CDR1 by enzymatic digestion or mutagenesis restored FVIII–vWF complex formation and rendered the Ab 40% less efficient in inhibiting FVIII activity, without altering the binding affinity of LE2E9 to FVIII. It was hypothesized that, via steric hindrance, the oligosaccharide prevents binding of FVIII to vWF, as well as the assembly of activated FVIII, activated FIX, and FX protein complexes (59) (Fig. 3).

The serum half-life of glycoproteins is modulated by the structure of the glycans. These can interact with glycoreceptors of liver cells, including the asialoglycoprotein receptor, which recognizes terminal galactose and N-acetylgalactosamine residues (60). Thus, the absence of terminal sialic acid residues can result in a decreased half-life of a glycoprotein (61).

Structural variation in conventional biantennary Fc glycans does not significantly influence the Ab half-life (62), probably because of the inaccessibility of these glycans to glycoreceptors (see “Structure of IgG Fab Glycans” above), although “nonnatural” high-mannose glycans can increase the clearance rate (63). In contrast, Fab glycan sialylation can enhance the serum half-life of mAbs, as shown for the anti-EGFR Ab cetuximab and for the anti–TA-MUC1 Ab PankoMab (64). Furthermore, depending on their position, Fab glycans enhanced or decreased the half-life of anti–α(1→6)-dextran Abs in mice (7 d) by ± 3 d (23). Interestingly, the rapidly and slowly clearing fractions were found to accumulate predominantly in the liver and kidney or in the blood and spleen, respectively (23), indicating that Fab glycans may also influence the accumulation of Abs in different organs.

Variable domain glycosylation may affect the propensity of Abs and L chains to aggregate and/or precipitate. One example is cryoimmunoglobulins, which precipitate at low temperatures and are associated with human diseases like RA and systemic lupus erythematosus (65). Sialic acid residues in the variable domains of the cryoimmunoglobulin “Ger” were found to provide additional electrostatic contacts required for its cryoprecipitation (66) (Fig. 3). Furthermore, variable domain glycans might enhance aggregation of Bence Jones proteins, given that such glycans were reported more often in multiple myeloma patients with amyloidosis (52). It was proposed that glycosylation can induce a conformational change in these proteins and trigger fibril formation (67), thereby contributing to amyloidosis; however, this hypothesis needs further research to substantiate. In contrast, a recent study described a decreasing effect of Fab glycans on Ab aggregation (68).

Fab glycosylation also was implicated in the modulation of immune complex formation. IgG Abs that bind ConA exhibited distinct Ag-binding profiles, which were attributed to the presence of glycans in only one Fab arm (69). If true, the glycosylated Fab arms may not bind Ag, and the Ab would be effectively univalent (“asymmetrical”), precluding formation of precipitating immune complexes and triggering effector mechanisms, analogously to IgG4 (70). Of course, in case of partial occupancy of a glycosylation site, one would expect B cells to secrete asymmetrically glycosylated Abs, as well as Abs either without Fab glycans or with a glycan in both Fab arms. However, it is unclear whether and how this asymmetry might be specifically imposed upon the Abs.

Instead of interfering with Ag binding (see “Ag Binding” above), the effect of introducing a Fab glycan may be more subtle, not affecting binding to cognate Ag but shaping the Ab reactivity in such a way that autoreactivity is minimized. One study described that, in nonimmunized rats ∼10–20% of IgGs were ConA-binding Abs, of which 78% exhibited autoreactivity and 14% reactivity against intestinal bacterial Ags. In contrast, only 40% of the “conventional” rat IgG was found to be autoreactive, suggesting that at least part of the Abs containing a Fab glycan confers a higher autoreactivity (71). In another study, following immunizations in mice, N-linked glycosylation sites were introduced in the variable domains of the BCR of anergic B cells that presumably allowed these cells to move away from autoreactivity, while allowing them to react against cross-reacting foreign Ags (25). However, in another study, autoantigen binding of follicular lymphoma–derived Abs was unaffected upon expression in the presence of tunicamycin to suppress N-linked (Fab) glycosylation (46).

IVIg treatment successfully ameliorates the symptoms of a number of autoimmune diseases, like systemic lupus erythematosus, RA, and idiopathic thrombocytopenic purpura. A variety of possible mechanisms was suggested (72). Recent studies emphasized the importance of glycosylation in the therapeutic effects of IVIg (73). In certain mouse models (74, 75), but not in others (13, 76–78), the anti-inflammatory activity of IVIg depends mainly on a fraction of sialylated Abs. These sialylated Abs may interact with C-type lectins (like DC-SIGN, DCIR) and sialic acid–binding Ig-type lectins, such as CD22, at the surface of immune cells (79, 80). It was suggested that these lectins interact with Fc-bound sialic acid (74, 75), which was further supported by a study that showed that IVIg with fully sialylated Fc glycans has enhanced anti-inflammatory activity compared with IVIg (81). However, a number of studies [including (74)] point to a role for Fab glycosylation of IVIg in modulating the immune system (Fig. 2B). These studies used Sambucus nigra agglutinin (SNA) lectin-affinity chromatography to fractionate IVIg, thereby predominantly enriching for Fab sialylation rather than Fc sialylation (13, 76). Fab-sialylated IVIg downregulated CD54 expression and the MCP-1 secretion of monocytes, whereas neither IVIg without sialylated Fab nor a highly sialylated Fc could achieve these effects (13). SNA-enriched IVIg also was shown to induce upregulation of PGE₂ by monocytes, thereby enabling inhibition of IFN-α production by plasmacytoid dendritic cells (82). In addition, the treatment of dendritic cells with IVIg Fab downregulated ERK1/2 phosphorylation after TLR4 stimulation (83). Furthermore, the interaction of IVIg with DC-SIGN also induced COX-2 expression and PGE₂ production, as well as the expansion of regulatory T cells (84). Silencing of human B cells via apoptosis could be induced by SNA-enriched IVIg through binding to CD22 (80, 85). Altogether, the available literature suggests that the anti-inflammatory effects of IVIg are mediated, at least in part, by its Fab arms, with a potential role for Ig variable domain sialylation (73). Differential effects by SNA-enriched and -depleted fractions might reflect, in part, differences in their respective Ab specificities and not be a direct result of the presence or absence of a glycan (13). However, the fact that removal of sialic acid using neuraminidase can ameliorate the effects of IVIg suggests otherwise (75). It would be interesting to investigate whether Fab-glycosylated IgG from pregnant women (see “Physiological Changes” above) shares similar, possibly immunomodulatory, properties with SNA-enriched IVIg.

IgG Fab glycosylation appears to be an important process in immunity, with demonstrated effects on stability, half-life, and binding characteristics of Abs and BCRs. Moreover, in addition to the potential to affect Ag binding, an increasing number of studies hint at the involvement of (endogenous or exogenous) lectins that can interact with these glycans, thereby potentially exerting immunomodulatory effects. At the same time, many unanswered questions remain (Table I). For instance, it is unclear why increased Fab glycosylation is observed during situations of diminished immunity (pregnancy), as well as in certain autoimmune settings, and what the consequences are. Also, why are glycosylation sites largely absent in the naive B cell repertoire? And how is their emergence and structure regulated during B cell activation? All in all, the exact role of Fab glycosylation in immunity remains poorly understood.

The authors have no financial conflicts of interest.