Decreased IgG core fucosylation, a player in antibody-dependent cell-mediated cytotoxicity, is associated with autoimmune thyroid diseases

Autoimmune thyroid diseases (AITD) are the most common group of autoimmune diseases, associated with lymphocyte infiltration and the production of thyroid autoantibodies, like thyroid peroxidase antibodies (TPOAb), in the thyroid gland. Immunoglobulins (Igs) and cell-surface receptors are glycoproteins with distinctive glycosylation patterns that play a structural role in maintaining and modulating their functions. We investigated associations of total circulating IgG and peripheral blood mononuclear cells (PBMCs) glycosylation with AITD and the influence of genetic background. The study revealed an inverse association of IgG core fucosylation with TPOAb and PBMCs antennary α1,2 fucosylation with AITD, but no shared genetic variance between AITD and glycosylation. These data suggest that the decreased level of IgG core fucosylation is a risk factor for AITD that promotes antibody-dependent cell-mediated cytotoxicity (ADCC) associated with TPOAb levels.


Abstract
Autoimmune thyroid diseases (AITD) are the most common group of autoimmune diseases, associated with lymphocyte infiltration and the production of thyroid autoantibodies, like thyroid peroxidase antibodies (TPOAb), in the thyroid gland.
Immunoglobulins (Igs) and cell-surface receptors are glycoproteins with distinctive glycosylation patterns that play a structural role in maintaining and modulating their functions. We investigated associations of total circulating IgG and peripheral blood mononuclear cells (PBMCs) glycosylation with AITD and the influence of genetic background. The study revealed an inverse association of IgG core fucosylation with TPOAb and PBMCs antennary α1,2 fucosylation with AITD, but no shared genetic variance between AITD and glycosylation. These data suggest that the decreased level of IgG core fucosylation is a risk factor for AITD that promotes antibodydependent cell-mediated cytotoxicity (ADCC) associated with TPOAb levels.

Introduction
Autoimmune thyroid diseases (AITD) are a class of chronic, organ-specific autoimmune disorders that disturb the function of the thyroid gland. They affect close to 5% of the European population (with a gender disparity) and so, represent the most common group of autoimmune diseases 1 . AITD encompass a spectrum of conditions including Hashimoto's thyroiditis (HT) and Graves' disease (GD). One of the features of AITD is the production of autoantibodies against components of thyroid cells that are also detected in the bloodstream.
The autoantibodies are produced against the three core thyroid proteins: thyroid peroxidase (TPO), thyroglobulin (Tg), and the thyroid-stimulating hormone receptor (TSH-R). Except for antibodies against TSH receptors (TSAb), which are known to stimulate the production of thyroid hormones by binding TSH receptors in GD 2 , little is known about the role of two other thyroid autoantibodies, thyroid peroxidase antibodies (TPOAb) and thyroglobulin antibodies (TgAb). Circulating TPOAb is the most common and diagnostically useful marker of AITD, detectable in the serum of most HT (95%) and GD (85%) patients 3 . In recent years, using TPOAb as a marker has been challenged since it appears in approximately 10% of apparently healthy individuals 4 . Even though autoantibodies are often a hallmark of autoimmune disorders, they can appear years before the first symptoms 5 , which poses the question about their causative role. Some evidence exists that autoantibodies can trigger autoimmunity, and IgG isotype seems to be connected with the development of autoimmune diseases potentially through regulation of IgG effector functions by alternative glycosylation 5 . However, it is yet to be determined if anti-thyroid antibodies cause AITD, or whether additional control mechanisms, such as post-translational modifications, are required to trigger the disease onset.
The most abundant and diverse form of post-translational modification is glycosylation, the attachment of sugar moieties to proteins, and various glycans are involved in virtually all physiological processes 6 . Glycans attached to Immunoglobulin G (IgG) are indispensable for its effector function and control of inflammation [7][8][9][10] . There are two glycosylation sites within the fragment crystallizable (Fc) of IgG that affect the 6 molecule's 3D-conformation and affinity for binding to Fcɣ-receptors (FcɣRs) on a range of immune cells 6,11,12 . Additional N-glycosylation sites are present in approximately 20% of IgG fragment antigen binding (Fab) and play a role in immunity, such as the affinity of epitope-binding site [13][14][15][16] . Previous analysis of IgG glycosylation with other autoimmune diseases showed a reduction of IgG galactosylation and sialylation, which trigger inflammatory response [17][18][19][20] . In relation to AITD, two small studies (62 and 146 patients respectively) looked at the TgAb glycosylation and reported differences between AITD, papillary thyroid cancers (PTC), and controls 21,22 .
First, TgAb IgG from HT patients showed higher levels of core fucose than the control group, as well as of terminal sialic acid and mannose 21 . On the other hand, it was observed that among HT, GD, and PTC groups, HT patients had significantly lower core fucose content on TgAb than the other two groups 22 . Furthermore, since recent genome-wide association studies (GWAS) identified novel loci associated with IgG glycosylation, which were known to be strongly associated with autoimmune conditions 23,24 and the heritability of AITD is estimated to 55-75% [25][26][27] , the next logical step was to examine if there was any common genetic background between those features. No data is currently available on common genetic variants associated with IgG glycosylation traits and AITD, and no large study on associations of the glycosylation of total IgG with the level of thyroid autoantibodies or with AITD has been performed.
Our goal was to determine if there are any IgG or PBMC glycan structures associated with the AITD or TPOAb positivity and examine if there are any common heritable factors between AITD and glycan structures. We investigated the association of total serum or plasma IgG glycome composition and PBMC glycosylation in AITD and looked for possible common genetic background in over 3,000 individuals. We found an association between the decreased level of IgG core fucosylation and PBMCs antennary α1,2 fucosylation with TPOAb level as well as with AITD. We observed the association of significantly affected IgG N-glycan traits with FUT8 and IKZF1 genes (responsible for IgG fucosylation), but we could not identify SNPs or a general dysregulation of gene expression in whole blood; suggesting a restricted dysregulation of glycosylation in a subpopulation of B-cells.   Fig. 1a whereas the green line highlights 53 additional IgG N-glycan derived traits. Image created with R package coMET 28 .

Results
TPOAb level and AITD are associated with a decreased level of IgG core fucosylation The presence of autoimmune antibodies is not a definite sign of the AITD, so we wanted to test whether IgG glycosylation status can play a role in active AITD and correlated with the TPOAb level or AITD. We investigated the associations between total plasma IgG glycome composition and peripheral blood TPOAb level and AITD  Table 2), and the amount of glycans in each peak was expressed as a percentage of the total integrated area. Furthermore, we excluded glycan peak GP3 due to the co-elution with the contaminant and combined GP20 and GP21 to get 22 directly measured N-glycan traits (Fig. 1c,d; Supplementary Table 2). As many of these structures share the same features (terminal galactose, terminal sialic acid, core-fucose, bisecting N-acetylglucosamine (GlcNAc)), we calculated 53 additional derived traits that average these features across multiple glycans (Fig. 1c,d; Supplementary Table 2). Latter structural features were found to be more closely related to individual enzymatic activities in different cellular compartments 29 and underlying genetic polymorphisms 23 than the 24 original glycan peaks. As all these structural features partially correlated (Fig. 1b), we estimated only 20 independent glycan traits 23 in our data using the Equation (5) method proposed by Li & Ji (2005) 30 .

9
The P-value for discovery step (0.05/20= 2.5x10 -3 ) was determined using Bonferroni correction for multiple testing with number of independent features 23 . We found eleven directly measured and derived IgG N-glycan traits negatively associated with increased TPOAb level (IGP7, IGP8, IGP15, IGP33, IGP48, IGP56, IGP58, IGP59, IGP60, IGP62 and IGP63) and six directly measured and derived IgG N-glycan traits positively associated (IGP2, IGP21, IGP36, IGP42, IGP45 and IGP46) following Bonferroni correction for multiple testing (P-value <2.5x10 -3 ; circle symbol in Investigating the structural composition of directly measured and derived IgG N-glycan traits associated with AITD and the TPOAb level (Supplementary Table 2), we observed an increase in IgG N-glycan traits without core fucose (IGP2, IGP21, IGP42, IGP46) and structures with bisecting GlcNAc (IGP36, IGP45) whereas a decrease in IgG N-glycan traits containing core fucose and without bisecting GlcNAc (IGP7, IGP8, IGP15, IGP48, IGP62, IGP63) as well as structures with core fucose regardless of bisecting GlcNAc (IGP58, IGP59, IGP60). Additionally, a decrease in IGP33, which refers to the ratio of all fucosylated (regardless of bisecting GlcNAc) mono-and disialylated structures, was observed. Overall, we found 17 IgG N-glycan traits associated with the TPOAb level, which suggests a decrease in abundance of glycan structures containing core fucose, and five of those structures remain associated with AITD status (Fig. 1c, d).  Table   3 -Sheet 5). The Manhattan plot is drawn with colors corresponding to the direction of association (blue=negative, red=positive). The red dashed line corresponds to the level of significance P-value in the replication cohort (0.05). Image created with R package called coMET. c) Reduction (not significant), of the binding of AAL to IgG core fucose in HT patients compared to control healthy individuals from the Polish cohort in the three batches using lectin blotting assay (P-value>0.05; the number of samples per group in each batch is 18,15,14 respectively); Supplementary Table 4). The plot combines a flat violin plot, box plot with whiskers and different data. A violin plot is a hybrid of box plot and kernel density plot.
Box plot with whiskers represents five summary statistics (the median, the first and third quartiles for two hinges and 1.5 times interquantile range from the hinges for two whiskers). Outliers are labeled by the red dot adjacent to the black dot of the measurement. Image created with R package called ggplot2. d) PBMC protein extracts were resolved on 10% SDS-PAGE gels in reducing conditions. After electrophoresis, the proteins were electrotransferred to a PVDF membrane and stained with the UEA I lectin (ncase=9 and ncontrol=10; Supplementary Table 5 HT is associated with the decreased level of IgG core fucose and the PBMC antennary remain decreased (Fig. 2a). In contrast, only one of the 17 IgG N-glycan traits that were significant in the discovery cohort (TPOAb level and AITD status;  Table 3 Table 1) was also examined in three batches by lectin blotting using Aleuria aurantia lectin (AAL), specific for α1,6-linked core fucose. We observed a consistently decreased level of IgG core fucosylation associated with HT status using AAL in the three batches, but none were significant with sample size tested ( Ulex europaeus agglutinin (UEA I). Lectin blotting with UEA l identified a significant reduction of antennary α1,2 fucose in patients with HT ( Fig. 2d) whereas blotting with AAL, which preferentially binds core fucose (α1,6), as well as all other lectin blotting assays, showed no difference in the content of particular glycan species between HT and control group (Supplementary Fig. 1, Supplementary Table 5). Coomassie Brilliant Blue staining control of total protein amount confirmed that there was no significant difference in protein profiles between control and HT samples ( Supplementary Fig. 2). To summarize, we found six significant glycans in the UK HT discovery cohort that were also associated with the TPOAb level (one was replicated in the Polish cohort) and a decrease of PBMC antennary α1,2 fucose associated with HT.

Figure 4 Enrichment in AITD patients of IgG N-glycan traits associated with IKZF1 and FUT8
genes. Correlation network between 75 IgG N-glycan traits, highlighting 17 IgG N-glycan traits associated with AITD and TPOAb level and those associated with SNPs close to HLA, FUT8, IKZF1 genes. Nodes represent different IgG N-glycan traits (the color of nodes is related to the associations performed in discovery and replicated cohorts) and connections show the Pearson correlation between IgG N-glycan traits (only correlations more than 0.75 are visualized, red for negative correlation and blue for positive correlation, the strength of correlation is represented by the width of edge). The three clouds above of nodes cluster the different IgG N-glycan traits previously found to be associated with SNPs within/close to HLA gene (blue cloud), the FUT8 gene (yellow cloud) and IKZF1 gene (red cloud) 23 . Image created with R package qgraph.
Enrichment of IgG N-glycan traits associated with two enzymes, Fut8 and Ikzf1, essential for the core fucosylation of IgGs in TPOAb positive subjects We then focused on the nine groups of genes that were found to be associated with modifications in IgG glycosylation pattern from the previous GWASs 23

Discussion
The role of the autoantibodies in the development of AITD is still unknown. AITD are among the most frequent autoimmune disorders occurring in almost 5% of general population 1 . If individuals harboring positive antibodies against thyroid proteins but without active disease are included, that adds up to 15% of the general population affected by thyroid autoimmunity 4 . Three central thyroid autoantibodies (TPOAb, TgAb, TSAb) were identified in patients with AITD, and their role in AITD is still unclear. Since the presence of TPOAb does not necessarily indicate active AITD, but IgG glycosylation was previously shown to be an essential factor in regulating autoantibody function in autoimmune diseases 18,19 , we examined the IgG glycosylation status in AITD patients and TPOAb positive individuals. This study is the first that investigated the potential association of IgG and PBMC glycosylation with AITD and TPOAb level, genetic background between them and hypothesized about the causality of these relationships.
We observed a decrease in IgG core fucose level, which presented as altered levels However, when we reduced the case group from Croatia to 83 individuals with very high TPOAb (>100 IU/ml in Roche assay, the same criteria to define AITD status in the TwinsUK cohort), the associations for the six replicated IgG N-glycan traits became more significant while the beta values remained similar to the preceding analysis ( Supplementary Table 3 -Sheet 4). Using lectin blotting in the Polish cohort, we validated the reduction of core-fucosylated IgG in HT patients, but due to the lack of data on TPOAb levels for HT patients (i.e. analyzed in the same blood collected for IgG glycosylation analysis), we could not confirm the association with TPOAb level rather than AITD status. Overall, our results suggest that the decreased fraction of core-fucosylated IgG could be more related to the level of TPOAb than the status of AITD.
Several studies showed that 95 % of IgG N-glycan structures in a healthy individual have core fucose and it acts as a "safety switch", attenuating potentially harmful ADCC [44][45][46][47][48] . Using FUT8 knockout Chinese hamsters, one previous study produced defucosylated CD20 antibodies and showed that afucosylated antibodies have a much higher affinity for FcγRIIIa (CD16a) -an immunoglobulin receptor distributed on natural killer (NK) cells, macrophages, and γδ T cells. They enhanced ADCC over 100-fold more than fucosylated CD20 antibodies 49 . Production of antibodies without the core fucose has recently revolutionized antibody therapies, by providing substantially enhanced ADCC 50,51 . Thus, the deficiency of IgG core fucose observed in the people with AITD and TPOAb-positivity in our data and the activities of IgG core fucose on the immune system previously identified, suggest an increase of ADCC in TPOAb positive individuals.
Interestingly, ADCC was previously reported in AITD without restriction to subgroups of patients, however, more patients with HT than with GD presented ADCC activities 52,53 . Although no strong correlation of TPOAb level (total IgG or IgG subclasses) measured by enzyme-linked immunosorbent assay (ELISA) with ADCC activity could be found in previous studies, and other thyroid antigens besides TPO could be involved in ADCC, TPO was described as the primary antigen involved in the thyroid ADCC 52-55 . Deglycosylation of TPOAb was shown to reduce the binding to FcγRs and thus inhibit the apoptosis of thyroid cancer cells via complement-dependent cytotoxicity (CDC) and cytotoxicity via ADCC 54 . Although IgG glycosylation studied in the present paper is of total and not antigen-specific IgG, we found a relationship between TPOAb level and IgG core fucose, and we suggest that the depletion of core fucose observed in TPOAb circulating in blood could enhance their cytotoxicity activities on thyrocytes through ADCC (Fig. 5).
Furthermore, taking advantage of the twin study design of the discovery cohort, we tested shared genetic and environmental effects between different significant glycan features found to be associated with TPOAb positivity and AITD in this study.
Surprisingly, although previous GWASs on IgG N-glycan traits showed several SNPs also associated with autoimmune diseases 23,56 , no shared genetic variance or lead SNPs could be detected between IgG N-glycan traits with AITD status or the TPOAb level. In exploring in LD of different lead SNPs associated with thyroid diseases and IgG N-glycan traits, we found that a SNP associated with IGP14 and IGP54 (rs199442) clustered with hypothyroidism SNP (rs77819282) in the same high LD block around NSF, but since neither IGP14 or IGP54 were found to be associated with TPOAb and AITD, NSF could potentially have a pleiotropic effect on hypothyroidism that we could not address with current data. SNPs rs3094014 and rs3094228 appear to be associated with IGP15 and AITD respectively, are in high LD, are located in the HCP5 gene region and indeed are eQTLs for HCP5 and other genes associated with immune response. These two lead SNPs and four other of the same LD block fall in a regulatory element, more precisely an enhancer and an open-chromatin in B-cell and thyroid tissue. Even though we identified two SNPs significantly associated with the IgG Nglycan structure IGP15, the cross-sectional nature of the cohorts and lack of independence of previous GWASs with the TwinsUK cohort do not allow conclusions to be drawn on the causal relationship between IgG core fucosylation and AITD.
Based on the findings of GWASs performed on the IgG N-glycan traits 23 , we also showed an enrichment of the afucosylated IgG N-glycan traits that are associated with FUT8 and IKZF1 genes. Both genes were considered leading players in the core fucosylation of IgGs, although the mechanism of IKZF1 gene in the fucosylation is still unclear 23 . The IKZF1 gene encodes a transcription factor belonging to the family of zinc-finger DNA-binding proteins associated with chromatin remodeling and regulating lymphocyte differentiation. Interestingly, several SNPs around the IKZF1 gene have been associated with autoimmune diseases based on the GWASs' findings; including systemic lupus erythematosus 57 , Crohn's disease 58 , inflammatory bowel disease 58 , and other diseases such as acute lymphoblastic leukemia 59 . On the other hand, the FUT8 gene encodes fucosyltransferase 8, a known enzyme catalyzing the addition of fucose in α1,6 linkage to the first GlcNAc residue (core fucose), but no SNPs around FUT8 genes have been associated with immune phenotypes other than IgG and plasma N-glycan structures from the previous GWASs 23, 31,32,37 . However, even though these two fucosylation enzymes were found to be associated with the significant Nglycan structures from the present study, we could not find a direct association of the SNPs falling within or near their genomic regions with AITD or TPOAb at the current sample size. Since FUT8 and IKZF1 are essential players in the regulation of the core fucose expression and as we found no identifiable genomic mutations occurring in or near those genes in TPOAb positive individuals or AITD patients, we went on to examine the expression of those genes in whole blood.
The transcriptomic analysis in the whole blood cells performed in the TwinsUK cohort suggested that the general decrease of fucosylation associated with AITD and TPOAb level was not a consequence of dysregulated gene expression of known fucosylation genes in the circulating blood. Furthermore, with the current data, we were also not able to find a dysregulation in PBMC transcriptome associated with AITD that is linked to PBMC fucosylation. We are unable to hypothesize about the functionalities and pathways of production of PBMC antennary α1,2 fucose as up to now, PBMC glycosylation is not well characterized due to the difficulty in the isolation of a sufficient amounts of individual types of PBMC (e.g. B-cells and CD4+ T-cells) for glycan analysis. In comparison, IgG glycan structures are well studied as the IgG comprises approximately 75% of total serum immunoglobulins. Considering we measured IgG glycosylation from whole blood and found the enrichment of afucosylated N-glycan traits, the statement that we found no general dysregulation of the fucosylation pathway in whole blood might seem confusing at first. However, although antibodies circulate and are detected in the blood, most plasma cells that produce circulating IgGs are located in germinal centers of secondary lymphoid organs like the spleen and lymph nodes 60 . In the case of AITD, the production of antibodies against thyroid components was suggested to happen directly in the thyroid gland where formations of germinal centers have occurred in patients with AITD [61][62][63] . Consequently, in addition to the previous detection of germinal centres in the thyroid gland of patients with AITD, our finding from transcriptomic analysis suggests that the IgG glycosylation pattern observed in circulating blood of people with AITD status is not directly associated with immune cells in whole blood, but could be a dysregulation of specific immune celltypes such as thyroid-derived lymphocytes that produce thyroid autoantibodies 64 (Fig.   5).
In this study, we used fixed volumes of total IgG solution and not the same initial mass of IgG for the UPLC analysis of released glycans. Additionally, IgG glycosylation was analyzed on the level of total IgG, not on the level of a specific antibody. This approach was chosen because of the fixed amount of residual salts and the difficulty in obtaining enough material for analysis of IgG glycosylation of a particular antibody (e.g. absence or low concentration of thyroid autoantibodies in the total IgG in healthy individuals). Preparations of reference of TPOAb level were made from a pool of serum from patients with AITD and were prepared and lyophilized 35 years ago and use the international unit per milliliter (IU/mL) as the reference unit. Unfortunately, there is no simple way to convert TPOAb level from IU/mL provided by the assays in the present study to a concentration expressed as microgram per milliliter (µg/ml) and check if the concentration of total IgG is altered by the secretion of thyroid autoantibodies. The concentrations of TPOAb and TgAb were previously measured (up to 1.4 mg/ml for TPOAb and 0.7 mg/ml for TgAb) [65][66][67] . Knowing that normal IgG concentration range is 5.1-15.8 mg/ml, in extreme cases, thyroid autoantibodies could represent up to 50% of total IgGs, and their presence in the blood can also increase the serum levels of total IgGs 68,69 . Thus, the secretion of thyroid antibodies can alter the concentration of total IgG in AITD.
If the modification of IgG glycosylation observed in the current study is the consequence of increased TPOAb concentration in total IgGs, the hypothesis that the depletion of IgG core fucose is a biomarker of TPOAb and its activity, and thereby a risk factor of harmful ADCC in the thyroid cells triggered by TPOAb (Fig. 5), becomes more likely. The cross-sectional nature of cohorts in this study and lack of independence of previous GWASs with the TwinsUK cohort do not allow conclusions on the causal relationship between IgG core fucosylation and AITD. Therefore, further analysis needs to be performed to describe genes and mechanisms playing in IgG and PBMC glycosylation in AITD in more specific tissues and cell-types to test this hypothesis. Studies looking at the expression of the known fucosylation genes in immune cells from thyroid tissue of healthy subjects, TPOAb positive individuals, and AITD patients might be especially valuable in dissecting the role of the fucosylation in the AITD disease genesis. However, to determine the causality between TPOAb level, fucosylation, ADCC and possibly other factors involved in the AITD genesis, prospective longitudinal studies that look at these traits before and after the diagnostic of AITD, as well as Mendelian randomization 70,71 in large independent cohorts would be required.

Conclusion
This study identified for the first time an association of decreased level of IgG corefucosylation and PBMC antennary α1,2 fucosylation with TPOAb level and AITD. This is also the first time that a decrease of fucosylation on IgG and PBMC is associated with one autoimmune disease and one of its biomarkers. Our findings could not be explained by common genetic background or general dysregulation of gene expression in the whole blood. However, drawing from the knowledge generated in a large number of previous studies, we could speculate that this reduction of IgG core fucose could be a consequence of tissue-specific aberrant gene expression. Moreover, we hypothesized that the decreased IgG core fucosylation could be a novel risk factor for potentially harmful ADCC in the thyroid gland, associated with TPOAb and AITD.
Further studies of the glycosylation of thyroid autoantibodies and their interactions with other immune features (e.g., immune cell-types, secreted proteins) and thyroid cells may be helpful to elucidate the potential role of autoantibodies and their glycosylation patterns in the pathogenesis and the treatment of thyroid diseases.