Hypogalactosylation of serum N-glycans fails to predict clinical response to methotrexate and TNF inhibition in rheumatoid arthritis

Human immunoglobulin G (IgG) is a glycoprotein with a biantennary (that is, two-armed) oligosaccharide attached to a canonical asparagine (N) in each heavy chain. These N-glycans are unusual because they do not decorate the protein surface. Instead, they are largely enclosed within the Fc region, where they help to maintain its spatial conformation. Variations in glycan structure "fine-tune" the effector activity of the antibody, modulating its capacity to fix complement and engage Fc receptors [1, 2]. Certain Fc glycan variants enriched for terminal sialic acid render IgG overtly anti-inflammatory, accounting in part for the action of high-dose intravenous Ig [3, 4].

Interestingly, rheumatoid arthritis (RA) is characterized by alterations in IgG glycosylation [5‐8]. Patients with RA exhibit an elevated proportion of IgG in which neither of the two glycan arms bears a terminal galactose (thus termed "G0"). This conformation enables binding of mannose-binding lectin, resulting in an enhanced propensity to fix complement, and animal studies suggest that G0 IgG may be especially arthritogenic [9‐11]. Recently, we and others have confirmed this hypogalactosyl phenotype in large cohorts, demonstrating additionally that change in IgG glycosylation predates the diagnosis of RA by an average of more than 3 years, is enriched in antibodies directed against citrullinated peptides (ACPAs) and correlates with disease activity [12‐16]. Thus multiple lines of evidence point to a role for IgG glycans in the pathogenesis of RA.

Although RA patients as a group exhibit altered IgG glycans, there remains substantial heterogeneity within this population [15]. We wished to understand whether pretreatment glycan status could predict response to therapy and whether disease-modifying antirheumatic drugs (DMARDs) might affect glycans differently, potentially hinting at an unexplored mode of action. Furthermore, we wished to determine whether ACPA positivity correlated with IgG glycoform aberrancy. We therefore performed an analysis of whole-serum N-glycan galactosylation, previously noted to correlate highly with galactosylation of IgG N-glycans [15, 17], on serial samples collected prospectively from patients with RA before and after treatment with MTX and anti-TNF agents.

Abnormal glycosylation of IgG is a well-established phenotype in RA, yet its clinical relevance is uncertain. In this study, we employed a proxy measurement for IgG galactosylation to address two key questions within the glycobiology of RA: (1) whether IgG glycan aberrancy predicts response to treatment and (2) whether DMARDs exert differential impacts upon IgG glycosylation, thereby potentially contributing to their clinical effectiveness. Using a well-characterized high-throughput NP-HPLC technique, we examined total serum N-glycans in two cohorts of RA patients for whom serial blood samples and prospectively collected clinical data were available. Because most biantennary serum N-glycans originate from IgG, our data support prior work indicating that IgG is hypogalactosylated in patients with RA and that a direct correlation between clinical and glycan improvement can be observed at the level of the individual patient. However, we found no evidence that (1) baseline sG0/G1 or change in sG0/G1 two weeks after initiation of treatment could predict therapeutic response to TNF inhibition or MTX or (2) these agents differentially influenced N-glycan galactosylation. The latter result suggests either that these agents exert similar effects on IgG glycans or, perhaps more likely, that altered IgG galactosylation arises via another route in patients experiencing clinical improvement.

By contrast, herein we show for the first time a clear correlation between ACPA positivity and abnormal galactosylation. This correlation had not previously been observed, likely because of confounding by differential disease activity. Our result therefore corroborates the observation of several groups that IgG ACPAs may exhibit glycosylation that is more aberrant than that of the IgG pool as a whole [15, 16].

These results highlight gaps in the understanding of IgG glycobiology in RA. The glycosylation pattern of IgG reflects multiple factors, perhaps most directly the cellular complement of specific enzymes that catalyze each step in glycan assembly [28‐30]. Patients with RA show reduced peripheral blood B-cell galactosyltransferase activity, though the biological basis for this difference and its relevance for IgG production by plasma cells has not been defined [31‐33]. How disease activity is associated with altered glycosylation, why this relationship differs from patient to patient (see, for example, Figure 2) and how antibodies of different specificities can display distinct glycan profiles remain to be defined. In this context, recent work illustrating particular glycan aberrancy of ACPAs from synovial fluid is of interest because it suggests that hypogalactosylated IgG may be generated directly within the rheumatoid synovium, accounting perhaps for the decline of sG0/G1 as this microenvironment normalizes with treatment [16].

Our study has several limitations. As noted in the Materials and methods section, the sG0/G1 index provides an imperfect estimate of IgG G0/G1. We cannot exclude the possibility that some of the changes observed reflect serum glycoproteins beyond IgG, though we are reassured in this respect by the work of other authors who have observed similar changes in IgG purified from RA patients [12‐14]. In particular, because IgG is the major serum source of G0 and G1 glycans, we doubt that we would have missed a clinically important predictive role of IgG galactosylation while using our assessment technique, though formal confirmation of this conclusion would require analysis of purified IgG.

Furthermore, we recognize that our method captures only part of the complexity of IgG Fc glycans. In addition to galactosylation, IgG glycans can vary in core fucosylation, the presence of an additional N-acetylglucosamine bisecting the two glycan arms and the extent of terminal sialylation. Each of these variants can have functional implications, modulating the ability to fix complement, engage Fc receptors and serve as a ligand for the myeloid lectin receptor dendritic cell-specific intercellular adhesion molecule 3-grabbing nonintegrin, or DC-SIGN [1, 2]. Therefore, our data do not exclude the possibility that further examination of individual IgG glycoforms, perhaps together with other clinical or genetic markers, might still uncover associations of diagnostic or therapeutic importance in RA. Because patients receiving MTX alone were enrolled only in the smaller cohort (Nested I), relatively few were available for analysis, thus limiting the power of our data to exclude definitively a difference between MTX and TNF inhibitors with respect to the correlation between changes in G0/G1 and in DAS (Figure 4).

These results represent the first exploration of the potential utility of IgG glycosylation as a predictor of therapeutic response in RA. Our results confirm that hypogalactosylation of IgG (assessed in our present study via the proxy measure sG0/G1) is prominent in patients with RA and improves with therapy, but they neither support the use of the sG0/G1 index as a guide to choosing between MTX and TNF inhibitors nor suggest a differential effect of these agents on IgG galactosylation. However, we have identified for the first time an association between ACPA positivity and aberrant IgG glycosylation, further extending the understanding the relationship between antibody glycans and RA.