Fab glycosylation of immunoglobulin G does not associate with improvement of rheumatoid arthritis during pregnancy

Background Changes in immunoglobulin G (IgG) constant domain (Fc) glycosylation are associated with changes in rheumatoid arthritis (RA) disease activity in response to pregnancy. Here, we sought to determine whether the same holds true for variable domain (Fab) glycosylation. Methods IgGs were captured from RA and control sera obtained before (RA only), during and after pregnancy, followed by Fc and Fab separation, glycan release, and mass spectrometric detection. In parallel, glycans from intact IgG were analysed. The data was used to calculate glycosylation traits, and to estimate the level of Fab glycosylation. Results The overall level of Fab glycosylation was increased in RA patients compared to controls, while no differences in Fab glycosylation patterns were found. For the Fc and intact IgG (Total) previously observed differences in galactosylation and bisection were confirmed. Furthermore, increased galactosylation of Fc and Total were associated with lower disease activity and autoantibody positivity. In addition, the change in Fc galactosylation associated with the change in disease activity during pregnancy and after delivery, while this was not the case for Fab. Conclusions In contrast to changes in Fc glycosylation, changes in Fab glycosylation are not associated with improvement of RA during pregnancy and arthritis flare after delivery. Electronic supplementary material The online version of this article (doi:10.1186/s13075-016-1172-1) contains supplementary material, which is available to authorized users.


Background
Rheumatoid arthritis (RA) is an autoimmune disease, for which it is well known that patients may improve during pregnancy [1]. The immunoglobulin G (IgG) fragment crystallisable (Fc) N-glycan compositions, specifically the levels of galactosylation and sialylation ( Fig. 1), have been recognised to be different in RA patients compared to healthy controls and to be associated with RA disease activity and its improvement during pregnancy [2][3][4].
In addition to the Fc moiety, that always bears N-glycans, the hypervariable region of the antigenbinding fragment (Fab) may harbour N-glycans [5]. These glycans are generally present on approximately 15-25% of the Fab portions, and have been found to alter the binding properties of antibodies to their antigen, as well as effector functions, and are known to change during pregnancy [6][7][8][9][10]. Interestingly, for rheumatoid arthritisspecific autoantibodies (anti-citrullinated protein antibodies (ACPA)) the vast majority has been shown to carry these Fab glycans [11].
In view of the known differences between the Fc N-glycans of RA patients and healthy controls, and its association with disease activity and improvement during pregnancy, we aimed to determine whether similar changes and associations can be found regarding the Fab N-glycans. In addition, the level of Fab glycosylation was investigated.

Study population and data collection
For the current study we used sera from the Pregnancyinduced Amelioration of Rheumatoid Arthritis (PARA) study, a prospective cohort study on pregnancy and RA [12]. At all time points, disease activity (DAS28; based upon a count of 28 swollen and tender joints and Creactive protein (CRP)) and medication was recorded. For the analysis we selected a subset of RA patients (n = 33) with the most pronounced (> = 0.6) or no change in disease activity, both during and after pregnancy, in search of maximum contrast. Only sera obtained before pregnancy, during the third trimester of pregnancy, and 26 weeks postpartum were analysed. In addition sera obtained from 32 healthy pregnant volunteers were analysed (third trimester and 6 months after delivery only). All patients fulfil the 1987 American College of Rheumatology criteria for RA. The study was in compliance with the Helsinki Declaration and was approved by the Ethics Review Board at the Erasmus University Medical Center, Rotterdam, The Netherlands.

IgG capturing and N-glycan release
IgGs were affinity-captured in duplicate from human serum in a 96-well format, as described before, with minor modifications [9]. One set was used for IdeS digestion to separate the Fab and Fc portions, whereas the other set was kept intact to analyse the total IgG glycans. Dried samples were reconstituted in 5 μL phosphate-buffered saline and 10 μL 2% (w/v) sodium dodecyl sulfate solution, instead of 10 and 20, respectively. Similarly, PNGase F was added in 10 μL.

Sialic acid derivatization and HILIC enrichment
To prevent the loss of sialic acids in the mass spectrometric detection of released glycans a derivatization protocol was applied as described before, with minor modifications [9,13]. Briefly, 2 μL released glycans were added to 20 μL ethyl esterification reagent in a V-bottom 96-well plate, and incubated at 37°C for 1 h. After the incubation 20 μL acetonitrile was added. Ethyl esterified N-glycans were purified from the reaction mixture using hydrophilic interaction liquid chromatography (HILIC) solid phase extraction (SPE) with a piece of cotton thread as the solid phase in a micro tip as described before [9].

MALDI-TOF-MS and data processing
Five microliter of eluate was mixed on the spot with sodium hydroxide-spiked Super-DHB matrix (Sigma-Aldrich, St. Louis, MO, USA) on an AnchorChip plate. Automated matrix-assisted laser desorption/ionization (MALDI)-time of flight (TOF) mass spectrometry (MS) measurements were performed summing 10,000 shots per spot with 250 shot steps and a full spot random walk. Details on data processing, extraction, and the calculation of glycosylation traits are described in Additional file 1.

Statistical analysis
Statistical analysis was performed in Stata 13SE (Stata-Corp, College Station, TX, USA). Bonferroni correction for multiple testing was performed throughout, with final significance thresholds depicted in the tables with results. First, a Wilcoxon rank-sum test was performed to explore the differences in glycosylation between RA patients and healthy controls outside pregnancy, at ≥26 weeks after delivery. Next we investigated which clinical covariates (use of medication, autoantibody (AAb) positivity, age at delivery and disease activity) were associated with the glycosylation in RA patients using linear regression analysis.
To study changes in glycosylation over time the Wilcoxon matched-pairs signed-rank test was used, comparing preconception with the third trimester for the patients, and third trimester with 26 weeks postpartum for both patients and controls. Association of the changes in glycosylation over time and the accompanying changes in disease activity were explored using Spearman correlation.

Response during pregnancy and flare after delivery
Clinical characteristics of the patients and controls are given in Additional file 1: Table S1. The patients who improved (n = 14) during pregnancy showed a mean decrease in DAS28 (delta DAS28; ΔDAS) of -1.7 (SD 0.5), whereas the nine patients that worsened showed a ΔDAS of +1.6 (SD 0.5). Ten patients did not change during pregnancy (ΔDAS = 0.0; SD 0.4). After delivery the 12 patients that flared had a mean ΔDAS of +1.5 (SD 0.8), while the 12 improving patients showed a ΔDAS of −1.5 (SD 0.6). Nine patients without a change in disease activity between the third trimester and 6 months postpartum (ΔDAS = −0.1; SD 0.4) were selected.  (Tables 1, 2 and 3). However, no differences between the levels of galactosylation, sialylation, fucosylation or presence of bisecting N-acetylglucosamine were observed for the Fab glycosylation (Tables 1, 2 and 3) compared between healthy controls and RA patients. However, for both Fc and total IgG (Total) there was a lower level of galactosylation and higher level of bisection in RA patients compared to the healthy controls.

Galactosylation of Fc and intact IgG, but not Fab, associate with disease activity and autoantibody positivity
In addition, it was explored which clinical covariates were associated with the levels of the calculated glycosylation traits at 26 weeks postpartum. Covariates were first tested univariate, and those with p <0.2 were included in the multivariate analysis. No Fab glycosylation models Abbreviations: trim trimester, wkpp weeks postpartum, IQR interquartile range remained significant after Bonferroni correction. However, for galactosylation of both Fc and Total multivariate models could be composed with R 2 values of 0.4 and 0.5, respectively. For these models the main contributing factors were disease activity (beta = −0.5) and autoantibody positivity (beta = +0.4). The results for all multivariate models are shown in Additional file 1: Table S2.
Fab glycosylation changes during pregnancy and after delivery are not associated with changes in disease activity Galactosylation and sialylation of Fab, Fc and Total increased between pre-conception and the third trimester of pregnancy in RA patients (Tables 1, 2 and 3; Fig. 2). The opposite was observed after delivery in both patients and healthy controls, except for the Fab portion where only a decrease in sialylation was observed. For fucosylation and bisection a decrease during pregnancy (RA only) and increase after delivery was observed for Fab, Fc and Total, in both patients and controls (Tables 1, 2 and 3). In addition, during pregnancy an increase in the level of Fab glycosylation was found in RA patients. Of note, changes in glycosylation traits that were high, e.g. Fab galactosylation and Fc fucosylation, were often not significant after Bonferroni correction (Tables 1, 2 and 3; Fig. 2). The Fc galactosylation change from pre-conception to the third trimester of pregnancy was negatively associated with the change in disease activity in that time span (rho = −0.39, p = 0.029; Additional file 1: Table S3, Additional file 1: Figure S1). The opposite was observed for the change Fc and Total galactosylation from third trimester to 26 weeks postpartum (rho = 0.48 and 0.62, p = 0.006 and 0.001, respectively). For neither the Fab glycosylation traits nor the Fab glycosylation level these associations were observed.

Discussion
There is a strong association of IgG Fc galactosylation with RA disease activity and the pregnancy-associated improvement thereof [2,3]. For the Fab portion of IgG this has not yet been investigated. Therefore, we applied our recently developed method for the analysis of Fab glycosylation on samples from a set of RA and healthy individuals in the context of pregnancy. Interestingly, we found no qualitative differences in Fab glycosylation between cases and controls nor an association with changes in disease activity when we compared several glycosylation traits. This may in part be caused by the fact that Fab galactosylation and sialylation levels are already high in the non-pregnant state, leaving less room for a potential pregnancy-associated increase.
Since the Fab portion of IgGin contrast to the Fc portionis not always glycosylated, the association of changes in Fab glycosylation with disease activity could be quantitative rather than qualitative. In line with this, it has recently been published that the RAspecific ACPA autoantibodies express high levels of Fab glycosylation [11]. Indeed, in this study we did observe increased levels of Fab glycosylation. However, even though the levels are higher in RA patients compared to controls, they were not associated with disease activity.
In the current study we chose to include only a limited number of patients and time points. Therefore minor associations between changes in RA disease activity and changes in Fab glycosylation could have been missed.

Conclusions
In conclusion, this study clearly demonstrates that, in contrast to changes in the glycosylation of the IgG Fc domain, changes in the glycosylation of the IgG Fab domain do not seem to play a major role in the pregnancy-associated improvement of RA.  2 Fc and Fab galactosylation and sialylation in RA and healthy controls. The percentages of galactosylation (yellow circle) and sialylation (purple diamond) on the Fc show a clear increase during pregnancy (from pre-conception to the 3rd trimester; RA only) and decrease after delivery (from 3rd trim to 26 weeks postpartum), as well as differences between cases and controls. No differences between cases and controls, and only minor pregnancy-associated changes were observed for the Fab glycosylation. Abbreviations: pre-conc pre-conception, trim trimester, wkpp weeks postpartum