The study was conducted in accordance with the Declaration of Helsinki and approved by the Bioethics Committee of the Jagiellonian University in Krakow, Poland (1072.6120.99.2021). All individuals donating blood for the experiment signed informed consent. Blood samples were collected between May 2014 and March 2021.

The study was conducted in cooperation with the Department of Endocrinology of the Jagiellonian University Hospital in Krakow. Recruitment of adult patients with AITDs without any other concomitant autoimmune diseases, and healthy volunteers was performed by an endocrinologist based on thyrotropic hormone (TSH) and thyroid autoantibody levels, thyroid gland ultrasonography and medical history of donors. Patients with AITD affected by other diseases, taking other drugs than those associated with AITD, donors suffering from alcoholism and taking other stimulants as well as pregnant women were excluded.

Human serum samples were collected from donors with the elevated IgG anti-thyroglobulin (TgAb) and/or anti-thyroperoxidase (anti-TPO) levels without symptoms of hypothyroidism (HT1, n=68), from Hashimoto’s thyroiditis patients characterized by the high autoantibody titers, advanced thyroid destruction and with the stabilized TSH level as a result of L-thyroxine treatment (HT2, n=113), GD patients with the elevated thyrotropic hormone receptor IgG (TRAb) levels before (GD, n=62) and after (GD/T, n=47) stabilization of TSH level as a result of the treatment with methimazole (1-methyl-2-mercaptoimidazole), an anti-thyroid agent with immunosuppressive activity. Healthy donors with the serum titers of TSH and autoantibodies within the normal range constituted a control group (CTR, n=90). The control group was age-matched to the study groups. Table 1 contains the demographic and clinical characteristics of donors with AITDs and healthy volunteers, and Figure 1 shows a graphical comparison of TSH and autoantibody (anti-TPO, TgAb, and TRAb) levels between the groups.

Blood samples were collected by venipuncture into the tubes containing a clotting activator (S-Monovette, Sarstedt, 04.1934.001), left for 5 h at room temperature (RT) for blood clotting, and centrifuged at 2500 rpm for 10 min at 4°C (Heraeus). Serum samples were stored at -80°C until IgG isolation.

IgGs were isolated by affinity chromatography with G-protein conjugated to the deposit grains. Serum samples (100 µl) were diluted 1:1 with a binding buffer (0.1 M Na₂HPO₄, 0.15 M NaCl, pH 7.2), applied to 96-well Protein G Spin Plates (Thermo Fisher Scientific, 45204), and incubated for 30 min at RT on an orbital shaker (Biosan). After incubation, the plates were centrifuged (1000 rpm, 1 min, RT) and the fraction containing IgG-depleted serum was discarded. Agarose resin was washed with the binding buffer until the ballast proteins were removed which was monitored by measuring the protein absorbance at 280 nm against the binding buffer (NanoDrop 2000, Thermo Fisher Scientific). Then IgGs were eluted with 200 µl of 0.1 M glycine, pH 2.5 into the collection plates containing 20 µl per well of a neutralization buffer (1 M Tris). IgG concentration was determined at 280 nm against a reagent blank (NanoDrop 2000, Thermo Fisher Scientific). IgG samples were stored at -80°C for further analyses.

The efficiency of IgG isolation from human serum was evaluated by SDS-PAGE. Purified IgG (10 µl) was separated on 10% gels under reducing conditions and then stained with colloidal Coomassie Brilliant Blue G-250 (CBB) (Sigma-Aldrich, B2025) according to the manufacturer’s protocol. The molecular masses of the light (25 kDa) and heavy chains (50 kDa) were verified using PageRuler™ Prestained Protein Ladder (Thermo Fisher Scientific, 26616).

Based on the relative intensity of the glycan peaks (GPs), expressed as percent (%) of the total area of all GPs, statistical analysis was performed using the Kruskal-Wallis nonparametric test with significance at p<0.05. All statistical interpretations of the results were performed in Origin Pro 2021b software (Origin Lab).

To assess the catalytic activity of serum ST6Gal1, desialylated IgG heavy chains immobilized on PVDF membrane were incubated with human sera from control and study groups according to the protocol by Jones et al. (47) with the minor modifications. The purified IgGs were digested with Arthrobacter ureafaciens sialidase with a wide specificity (ABS, 1 U in 100 mL, Roche, 10269611001) to remove sialic acid (SA) from the antibody N-glycans. IgG samples (0.5 μg) were incubated with 1 μl of ABS in 8 μl of 50 mM sodium acetate, pH 5.2 overnight at 37°C (Biosan thermoblock). Desialylated and untreated IgG samples were separated under reducing conditions on 10% gels in SDS-PAGE, electrotransferred onto a PVDF membrane (Millipore, 88518), which was cut into small pieces containing IgG heavy chains based on PageRuler™ Prestained Protein Ladder (Thermo Fisher Scientific, 26616), and blocked overnight in 1% BSA (Sigma-Aldrich, A7906) at 4°C. The selected PVDF sections were incubated with human sera from each study and control group diluted 1:1 with TBST (50 mM Tris-HCl, 150 mM NaCl, and 10% Tween) overnight at 4°C. Then the membranes were probed 1 h at RT with biotinylated Sambucus nigra agglutinin (SNA, Vector Lab., B-1305) diluted 1:4000 in TBS containing 0.1 mM CaCl₂ and 0.01 mM MgCl₂ or with SNA preincubated with 1 M acetic acid as a negative control to check nonspecific binding of SNA via protein domains but not to α2,6-sialylated N-glycans. After washing three times with TBST, alkaline phosphatase (AP)-conjugated avidin (Sigma-Aldrich, E2636) diluted 1:4000 in TBST was applied for 1 h at RT. IgG heavy chain was visualized by a colorimetric reaction using 5-bromo-4-chloro-3-indolylphosphate (BCIP, Roche, 11383221001) and nitro-blue-tetrazolium (NBT, Roche, 11383213001).

ABS re-digestion of IgG N-glycans sialylated by serum ST6Gal1 on PVDF membrane was used as an additional control. Membrane fragments with IgG heavy chains sialylated by serum ST6Gal1 were placed in 4-well plates (Nunc, 176740), and incubated with 3 μl of ABS in 27 μl of 50 mM sodium acetate, pH 5.2 overnight at 37°C (Forma Steri-Cycle i160 CO₂ incubator, Thermo Fisher Scientific). Then the membranes have been subjected to incubation with SNA and visualized by colorimetric reaction as described above.

The statistically significant differences in IgG N-glycan profiles between patients with Graves’ disease before and during immunosuppressive therapy and healthy donors were demonstrated for the twelve out of 25 GP-matched N-glycan structures ( Figure 2 ). A content of agalactosylated structures with proximal fucose (F(6)A1, F(6)A2, F(6)A2B) and oligomannose (M5) increases in GD (all structures) and GD/T (F(6)A2B and M5) compared with CTR. Statistically significant quantitative differences were also found in galactosylated (A2G(4)2) and fucosylated (F(6)A2[6]BG(4)1) N-oligosaccharides with their up-regulation in GD/T patients compared to GD. In contrast to the above-mentioned changes in the case of N-glycans with a bisecting GlcNAc (A2BG(4)2), α1,6-fucosylated (F(6)A2BG(4)2), monosialylated (A2G(4)2S1 and F(6)A2G(4)2S1), and disialylated (A2G(4)2S2) structures we observed their statistically significant down-regulation in GD (all structures) and GD/T (A2G(4)2S2) relatively to CTR.

The effect of N-glycosylation on the development of AITD compared with CTR after UPLC-MS analysis was further verified by performing statistical analysis with the Kruskal-Wallis test for glycan groups (non-sialylated, monosialylated, disialylated, non-fucosylated, fucosylated, nongalactosylated, galactosylated, monoantennary, and diantennary). The results showing statistically significant differences are presented in Figure 4 . Non-sialylated and monoantennary glycans showed an increase in both GD groups compared with healthy volunteers, whereas diantennary structures showed an increasing trend in GD compared with CTR. Statistically significant decreases were shown for monosialylated (in GD and GD/T), and disialylated (in GD) glycans compared with healthy subjects. The relative content of galactosylated structures was lower in GD while in GT/T this group of glycans normalized partly which was related to a decline of agalactosylated structures. The reduced IgG galactosylation was accompanied by an increase of agalactosylated structures in GD ( Figure 4A ).

Statistically significant quantitative differences were also found for 8 N-oligosaccharide structures between patients without hypothyroidism, patients with HT after L-thyroxine treatment, and healthy volunteers ( Figure 4 ). For agalactosylated (F(6)A1, A2, F(6)A2), mono- (A2G(4)2S1) and disialylated (A2BG(4)2S2) structures, an increase was shown in HT2 compared to CTR (A2, A2BG(4)2S2) and HT1 (other structures). In contrast, some fucosylated (F(6)A2G(4)2) and monosialylated (F(6)A2G(4)2S1) structures were increased in HT1 relative to HT2. HT1 also indicated a statistically significant decrease in oligomannose-type N-glycan (M5) compared to CTR.

A statistically quantitative increase in the content of monoantennary oligosaccharides was observed in HT2 compared to HT1. The amount of diantennary sugar structures was significantly higher in HT1 compared to HT2 and CTR. We identified also a relationship between the alterations of galactosylated and agalactosylated N-glycans in HT patients, parallel with the lowering of galactosylation, agalactosylated structures increased in HT2 vs. HT1 donors ( Figure 4B ).

Due to the changes of mono- and disialylated IgG N-glycans in both HT and GD sera detected by UPLC-MS analysis ( Figure 4 ), we evaluated an activity of ST6Gal1 in sera of AITD patients and healthy individuals using SNA lectin blotting method described by Jones et al. (47) with some modifications. We observed a positive reaction of α2,6-sialylated N-glycans on the heavy chain with SNA for IgG isolated from control and AITD sera. Desialylation of IgG using the neuraminidase with a wide specificity resulted in an almost complete attenuation of the reaction with SNA ( Figure 5A ). A weak signal observed for neuraminidase + samples ( Figure 5A ) was a result of non-specific SNA binding to IgG polypeptides as determined by the reaction with SNA preincubated with acetic acid which abolishes SNA binding to SA ( Figure 5D ). Incubation of desialylated IgG heavy chain with human sera restored SNA-positive reaction which resulted from re-sialylation of IgG by ST6Gal1 present in donor sera ( Figure 5B ). To verify reversibility of this sialylation on PVDF membrane-bound IgG heavy chain we used again neuraminidase to remove SA attached by serum ST6Gal1 which led to an almost complete loss of the reaction with SNA ( Figure 5C ).

The detailed characteristic of IgG N-glycome obtained by UPLC-MS showed the statistically significant quantitative differences in sugar structures in the course of immunosuppressive treatment of GD patients, and during the development of thyroiditis. The obtained results allow us to support the previous observation that IgG N-glycosylation is a particularly dynamic process, reflecting the course of the diseases, and conclude that IgG N-glycan alterations accompany the inflammatory processes in AITD and the applied therapy. The changes occur early in the development of the disease when the synthesis of autoantibodies is triggered, and the thyroid structure has not yet undergone the destruction that accompanies chronic inflammation at later stages of the disease. Methimazole therapy significantly changes the structure of the IgG glycome in patients with GD. We have also shown that ST6Gal1, responsible for the attachment of α2,6-SA, is active in the sera of AITD patients and healthy donors and has the potential to modify IgG sialylation independently from the classical pathway of cellular glycosylation.

The effect of immunosuppressive drugs used in the treatment of autoimmune diseases on N-glycosylation of IgG and other serum proteins has been demonstrated in previous studies. This issue has been the most explored in RA, the autoimmune disorders with the best characterized IgG N-glycan alterations. It is well documented that one of the effects of immunosuppressive agents administrated in RA is the altered IgG glycosylation reversing the antibody activity towards the anti-inflammatory response (48–51, 58).

GD patients recruited to our study were treated with methimazole (1-methyl-2-mercaptoimidazole; Polish brand names: Thyrozol or Metizol), which affects thyroid hormone synthesis as thyrostatics. Methimazole was shown also to have an immunosuppressive effect resulting from a decrease of TRAb serum level, triggering of intrathyroidal T cell apoptosis, reduction of HLA class II expression as well as up-regulation of circulating Tregs and reduction of the number of Th cells, NK, and activated intrathyroidal T cells (59–61). TRAb concentration in the sera of methimazole-treated GD patients was partly reduced ( Table 1 and Figure 1A ), which confirms the immunosuppressive activity of this antithyroid agent. Taking into account that the blood was taken in the relatively short term after stabilization of TSH levels, it can be assumed that in the longer term TRAb titer was further stabilized.

The normalisation of TRAb level in GD/T ( Table 1 and Figure 1A ) was accompanied by the altered intensity of the individual UPLC-MS peaks to which agalactosylated (F(6)A2), monogalactosylated (F(6)A2(5)BG(4)1), and digalactosylated (A2G(4)2, F(6)A2G(4)2 and F(6)A2BG(4)2) glycans were assigned. The amount of both digalactosylated N-glycans was up-regulated while the content of agalactosylated structure was reduced as the result of methimazole therapy in GD patients ( Figures 2 , 4A ). This part of our observations is following the previous experiments on IgG N-oligosaccharides from RA patients undergoing treatment with methotrexate and infliximab. In addition to suppressing inflammatory processes confirmed by a reduced level of C-reactive protein (CRP) (58), both antirheumatic drugs were shown to enhance IgG galactosylation (48–51, 58). In turn, in immune thrombocytopenia galactosylation of IgG1 and IgG4 subclasses in the blood of anti-CD20 monoclonal antibody (rituximab)-treated patients was slightly reduced, while IgG1, IgG2/3, and IgG4 showed the higher content of N-glycans with bisected GlcNAc (62). Therefore we can conclude that the effect of immunosuppressive therapy on N-glycosylation depends on the applied therapeutic agent and autoimmune disorder.

Apart from the significantly increased galactosylation of N-glycans in inflammatory arthritis patients treated with anti-TNF IgG, Collins et al. observed an up-regulation of N-glycan core-fucosylation (58), which is well known to contribute to IgG anti-inflammatory activity (6). Our study demonstrated that the amount of IgGs with core-fucosylated structures F(6)A2(6)BG(4)1, F(6)A2G(4)2, and F(6)A2BG(4)2 is higher in GD patients after TSH normalization as the result of methimazole therapy (GD/T) in relation to the untreated donors ( Figure 2 ) which can also contribute to attenuating of the immune response.

According to the literature data, reduced sialylation favors the pro-inflammatory properties of IgG (63, 64). Our UPLC-MS analysis showed that the sialylation status of IgG was not affected by methimazole treatment. The intensity of the sialylated N-glycans was not altered in GD patients during immunosuppressive therapy in relation to the state before this drug implementation (GD/T vs. GD) ( Figure 2 ). However, the entire pool of IgG monosialylated and disialylated N-oligosaccharides was significantly reduced in both GD groups in comparison to healthy donors ( Figure 4A ). Sialic acid is terminally linked to Gal residues in IgG N-glycans, and its attachment requires the presence of galactosylated structures. We presently show that galactosylation of IgG is normalized during methimazole administration ( Figure 4A ) while N-glycans are still undersialylated in relation to healthy donors ( Figure 2 ), which can be an effect of too short drug administration. To resolve this, studies on a long-term treatment cohort would be necessary.

Methimazole therapy partially reversed the changes of IgG N-glycosylation by the up-regulation of galactosylation and reduction of agalactosylation ( Figure 4A ). Down-regulated galactosylation is well known to promote the proinflammatory potential of IgG (34). Without detailed functional studies, we can only speculate that this effect of anti-inflammatory treatment may reduce clinical symptoms of the disease also by affecting galactosylation of antibodies in GD.

Due to the documented activity of serum ST6Gal1 from AITD patients on desialylated IgG N-glycans ( Figure 5 ), we suppose that this serum sialyltransferase may contribute to IgG sialylation in the bloodstream in thyroid autoimmunity. SNA lectin blotting used in our study to assess ST6Gal1 activity provided the qualitative data but did not allow for quantitative analysis and determination of possible differences between groups. A quantitative assessment of the effects of ST6Gal1 activity would be interesting to interpret the UPLC-MS results, and is worth pursuing in future studies. Anti-inflammatory activity of IgG sialylation is also considered as a therapeutic strategy in terms of the usage of intravenous immunoglobulins (IVIGs) for the treatment of autoimmune diseases (65, 66). Pagan et al. used the mechanism of extracellular sialylation to target IgG N-glycan remodeling. They confirmed the attaching of SA by the extracellular enzyme to Fc of IgGs deposited at sites of inflammation. The activity of recombinant human galactosyltransferase B4GALT1 and ST6Gal1 on a single Fc fragment of IgG1 attenuated inflammation in a K/BxN arthritis model. The increased IgG sialylation resulted from in vivo administration of a soluble ST6GAL1 converted IgG activity into anti-inflammatory in autoimmune disease (67).

In the previous study, we have also demonstrated the altered glycosylation on human leukocytes activated in vitro in a two-way mixed leukocyte reaction (MLR) in the presence of two immunosuppressive agents commonly used to induce immune tolerance after organ transplantation: cyclosporin A (CsA), an inhibitor of calcineurin, and rapamycin (Rapa), which blocks mammalian target of rapamycin mTOR. Oligomannose/hybrid-type N-glycans on human leukocytes in MLR model were significantly down-regulated by CsA, while the synergistic action of both immunosuppressive drugs enhanced the amount of these structures on leukocyte proteins (68).

Based on the previous literature data and the presently obtained results, it seems that immunosuppressive agents-triggered changes of glycosylation are a wider phenomenon, which can concern drugs with various mechanisms of action, and different targeted proteins. Functional consequences of the observed changes in IgG N-glycosylation isolated from methimazole-treated patients for Graves’ disease course need further study.

The crucial role of Asn297-linked N-glycans attached to IgG Fc fragment in ADCC and CDC has been demonstrated on various research models (69), and was shown to be important in autoimmune disease development (70, 71), including thyrocyte destruction in Hashimoto’s thyroiditis (72). Forming of immune complexes of IgG autoantibodies with self-antigens gathers innate immune effector cells, like NK and myeloid cells, which express Fcγ receptors (FcγR) and/or recruits complement proteins. Activation of effector cells and complement cascade results in target cell apoptosis and finally tissue damage (34, 73).

Martin et al. detected the reduced level of core Fuc in IgG N-glycans from AITD subjects and depletion of antenna fucosylation in peripheral blood mononuclear cells (PBMCs) isolated from the whole blood of HT patients (40). The current analysis showed that the content of two core-fucosylated IgG structures (F(6)A2G(4)2 and F(6)A2G(4)2S) was decreased in Hashimoto’s thyroiditis patients, while the level of F(6)A2 N-glycan assigned the most abundant UPLC peak was up-regulated in these patients (HT2) compared to the donors with the higher level of anti-Tg/anti-TPO (HT1) ( Figure 3 ). In effect, the group analysis did not show statistically significant changes in the whole pool of core-fucosylated N-glycans between the patients with thyroiditis at different stages of the disease severity and healthy subject ( Figure 4 ). Taking into account the significantly different number of the recruited participants in both studies, not complete reproducibility of the obtained results seems to be understandable.

Core-fucosylation of immune system proteins, including IgG, plays an important role in their activity. Alterations in IgG core-fucosylation are, right after galactosylation, one of the most common modifications detected in inflammatory diseases, including autoimmune diseases (37). Based on the previous results which demonstrated that core-fucosylation impeded FcγRIIIA binding and inhibited ADCC (47), we can speculate that the reduced core-fucosylation of IgG N-glycans in AITD described by Martin et al. (40), and the currently observed decrease of the two core-fucosylated structures in HT2 ( Figure 3 ) may contribute to inflammation and thyroid tissue damage in the course of Hashimoto’s thyroiditis.

Sialic acid is a component of glycans crucially important in the regulation of immune glycoprotein function, due to its negative charge, and terminal localization in oligosaccharide structures. Autoimmunity is usually accompanied by down-regulation of IgG sialylation (74, 75). The reduced sialylation of IgG1 and IgG2 occurs in granulomatosis with vasculitis (GPA) (64). Kemna et al. showed that the level of total IgG sialylation has a prognostic value in GPA relapse, as the SA content of IgG1 in patients with granulomatosis decreased during relapse and remains unchanged in remission (63). Collins et al. demonstrated a decrease in sialylated triantennary N-glycans in inflammatory arthritis patients, strongly correlated with reduced CRP level in the silenced inflammatory process (73).

Our recent study has demonstrated more intensive thyrocyte lysis in the presence of IgG isolated from HT patients than from healthy donors, which resulted from the higher anti‐TPO content in the whole IgG pool of HT donors and the altered IgG N-glycosylation in HT autoimmunity (72). The present analysis demonstrated that the content of F(6)A2G(4)2S1 structure, the most intensely sialylated N-glycan in IgG glycoprofile, was lowered in Hashimoto’s thyroiditis in comparison to HT1 group ( Figure 3 ), which may explain the results obtained in the previous functional analysis performed with IgG from HT patients in the in vitro model (72). The study by Ząbczyńska et al. indicated also that IgGs with desialylated N-glycans were more potent to induce ADCC in human thyrocytes (72).

In summary, N-glycans are involved in fundamental cellular and molecular processes that stimulate and inhibit immune system pathways. Detailed characterization of the IgG N-glycans obtained by UPLC-MS revealed statistically significant quantitative differences in sugar structure during immunosuppressive treatment of GD patients and during the development of Hashimoto’s thyroiditis. The results show that changes in IgG N-oligosaccharides contribute to the development of inflammation in autoimmune thyroid diseases. These changes begin in the early stages of the disease, where autoantibodies are overproduced, but hypothyroidism and thyroid gland destruction, which are associated with later stages of pathology, are not observed. The use of immunosuppressive therapy significantly alters the process of N-glycosylation in patients with Graves’ disease. Further studies are needed to evaluate the changes in IgG N-glycosylation in AITD to see how relevant it is to determine the contribution of altered oligosaccharide content to antibody-mediated autoimmunity in these autoimmune diseases. It is also worth noting, that similarly as in the case of vast majority of glycomic studies of this kind, we compare only relative quantities of glycan structures in tested samples. By its nature, the relative content of a given sugar structure in a sample depends not only on the increase or decrease in its expression, but also on the simultaneous quantitative changes of other glycans. Being aware of these limitations, we used this approach in the described studies and we believe that it is, however, a specific description of the phenotype of the patients and enables comparative analysis. On the one hand, it gives less room for interpretation of the biological contexts and mechanisms causing the observed changes, but on the other hand, when applied according to the same criteria for all analysed samples, it can be a useful parameter for patient stratification.