Patients with anti-Jo1 antibodies display a characteristic IgG Fc-glycan profile which is further enhanced in anti-Jo1 autoantibodies

IgG Fc-glycans affect IgG function and are altered in autoimmune diseases and autoantibodies. Anti-histidyl tRNA synthetase autoantibodies (anti-Jo1) are frequent in patients with idiopathic inflammatory myopathies (IIM) and anti-synthetase syndrome (ASS) with associated interstitial lung disease (ILD). Thus, we hypothesized that the total-IgG Fc-glycans from Jo1+ versus Jo1− patients and anti-Jo1-IgG would show characteristic differences, and that particular Fc-glycan features would be associated with specific clinical manifestations. By proteomics based mass spectrometry we observed a high abundance of agalactosylated IgG1 Fc-glycans in ASS/IIM patients (n = 44) compared to healthy age matched controls (n = 24). Using intra-individual normalization of the main agalactosylated glycan (FA2) of IgG1 vs FA2-IgG2, ASS/IIM and controls were distinguished with an area under the curve (AUC) of 79 ± 6%. For Jo1+ patients (n = 19) the AUCs went up to 88 ± 6%. Bisected and afucosylated Fc-glycans were significantly lower in Jo1+ compared to Jo1− patients. Anti-Jo1-IgG enriched from eleven patients contained even significantly lower abundances of bisected, afucosylated and galactosylated forms compared to matched total-IgG. ASS and ILD diagnosis, as well as lysozyme and thrombospondin correlated with Jo1+ characteristic Fc-glycan features. These results suggest that the anti-Jo1+ patient Fc-glycan profile contains phenotype specific features which may underlie the pathogenic role of Jo1 autoantibodies.

Accumulated evidence suggests that anti-Jo1 autoantibodies may be involved in the pathogenesis of ASS and/or IIM (ASS/IIM): serum levels of anti-Jo1 autoantibodies vary with disease activity 5,6 ; and anti-Jo1 positive serum induces interferon (IFN) production from plasmacytoid dendritic cells 7 . However, direct evidence that anti-Jo1 autoantibodies cause or contribute to ASS/IIM pathogenesis is still lacking. One possible mechanism to explore is the effect of IgG Fc-glycans (Fig. 1). These sugars appended to the IgG Fc-region modulate inflammation by affecting the affinity of antibodies and interaction with Fcγ receptors (for example leading to antibody-dependent cellular cytotoxicity), engagement of complement activation, and induction of cytokine secretion 8 . Moreover, IgG Fc-glycosylation has been suggested as a possible disease biomarker, with low levels of galactosylation associated with a high pro-inflammatory status 9,10 . For instance, IgG 1 Fc-galactosylation status predicts methotrexate (MTX) response in early rheumatoid arthritis (RA) patients 10 . Notably, Fc-glycans from anti-citrullinated protein autoantibodies (ACPA, known to induce pain and bone loss) 11,12 , are more agalactosylated and lack sialic acid residues, compared to non-ACPA IgG [13][14][15] . A significant decline in Fc-galactosylation and afucosylation of ACPAs has been reported prior to RA onset, which could suggest that Fc-glycosylation is a disease-driving mechanism 10,16 . In patients with IIM, more agalactosylated forms of IgG have likewise been found compared to healthy siblings and age/sex matched controls 17 . Together, these findings indicate the importance of antibody Fc-glycosylation for mediating immune regulating response and thus a potential role in pathogenicity of autoimmune disorders.
To better understand the potential role of anti-Jo1 autoantibodies in IIM pathogenesis we investigated the Fc-glycans of these autoantibodies as well as total IgG from anti-Jo1 + and anti-Jo1 − ASS/IIM patients. The Fc-glycan profiles were correlated with clinical manifestations and protein information (which was obtained by simultaneous proteomics screening of the IgG enriched samples).

Results
Number of patients, controls and sample types. In total, 24 healthy controls (HC) and 44 patients participated in the study. Of these patients, 15 were diagnosed with IIM but not ASS, 12 were diagnosed with ASS and IIM, one patient was diagnosed with ILD and IIM and 16 patients were diagnosed with ILD, ASS and IIM, ( Table 1). 19 of the patients were Jo1 positive (Jo1 + ) and 25 Jo1 negative (Jo1 − ). For seven Jo1 − patients we had access to enriched sera IgG from two time points, (i.e. T1 and T2, Supplemental Table 1). For the Jo1 + patients we had access to enriched sera from two time points from six patients, three time points from two patients, as well as one patient with seven, one patient with eight and one patient with ten sampling time points (i.e. T1-T10), Supplemental Table 1. Note that the intra-individual glycan profiles of these patients did not change significantly over time (for more details see Supplemental Fig. 1). For this reason, we are focusing on the first sampling time point (T1) for which we have data from all patients. For extra validation both univariate and multivariate data were evaluated by including the remaining data from the other time points. From nine Jo1 + patients, anti-Jo1 auto-immune IgG was further enriched from the total serum IgG from respective patient using either pools from several sampling time points per patient (n = 8), or from one sampling time point (n = 1). Furthermore, we had ) that were shown to affect the Fc-glycan profile of IIM or phenotype specific IIM most prominently. The glycan nomenclature is according to Royle et al. 44 . Blue squares (N-acetyl-glucosamine), red triangle (Fucose), green circles (Mannose), yellow circles (Galactose) and purple diamonds (Sialic Acid).

Total ASS/IIM IgG and anti-Jo1-IgG serum concentration and reactivity. Efficient purification of
total IgG and anti-Jo1-IgG was confirmed by well-defined protein gel bands localized in the antibody heavy and light chain regions of total IgG and anti-Jo1-IgG ( Supplementary Fig. 2C). An ELISA and a dotblot were performed in order to confirm the efficiency of anti-Jo1 IgG enrichment from the total IgG loaded into the Jo1 affinity column (Supplementary Fig. 2C; Supplementary methods). Strong anti-Jo1 reactivity was registered by anti-Jo1-IgG, and no major reactivity against Jo1 was detected in the remaining control flow through IgG (FT; Supplementary Fig. 2B and D). Total IgG concentration (median < 8.6 mg/mL) in serum and IgG percentage among total serum proteins (median < 16.6 mg/mL) at first available sample was similar in patients and controls (Supplementary Table 3).
The median concentration of anti-Jo1-IgG in ASS/IIM was 0.06 mg/mL, reaching a maximum of 0.13 mg/mL. The proportion of anti-Jo1-IgG among total IgG in ASS/IIM serum was 1.58%, with 4 out of 11 patients showing a percentage greater than 2% (Supplementary Table 3). No significant changes of IgG titres were observed in longitudinally collected sera (Supplementary Table 4). ASS/IIM patients display a significantly less galactosylated IgG 1 Fc-glycan profile compared to healthy controls. Eighteen Fc-glycans from IgG 1 , sixteen from IgG 2/(3) and twelve from IgG 3/4 were quantified. The glycoform abundances were normalized to total content (100%) of Fc-glycosylated IgG 1 , total content (100%) of Fc-glycosylated IgG 2/(3) and total content (100%) of Fc-glycosylated IgG 4/(3) peptides, respectively. In Supplementary Table 5, the individual average ± standard deviations for each glycopeptide, as well as p-values (comparing the subgroups listed in Table 1) are given. Comparing ASS/IIM with controls, specifically the ASS/ IIM Fc-glycan profile of IgG 1 was altered with half of the glycans being significantly different (p < 0.05). Of these significant glycans seven (out of nine) indicated a decrease in galactosylation. In contrast, the galactosylation status of IgG 2/(3) and of IgG 3/4 remained relatively stable. For example, compared to the main agalactosylated form FA2 (Fig. 1) from IgG 1 which was significantly increasing (p = 4.5E-4), FA2 of IgG 2/(3) and FA2 of IgG 3/4 were not significantly different between the controls and the ASS/IIM patients (p = 1.9E-1 and p = 6.4E-2, respectively), Supplementary Table 5. Since the galactosylation status is known to change to less galactosylated forms, not only according to inflammatory/disease status, but also according to factors such as age, sex and heritability 18 , we tested if the unchanged FA2 of IgG 2/(3) (FA2_2) could be used as an intra-individual control of the change in FA2 of IgG 1 (FA2_1). Thus, we used log(FA2_1/FA2_2), and tested how well this factor could distinguish patients with ASS/IIM compared to controls. In terms of significance, log(FA2_1/FA2_2) was significantly different (p = 1.4E-5, Table 3 and Fig. 2A). Furthermore, ROC curve analysis of log(FA2_1/FA2_2) generated an area under the curve (AUC) of 79 ± 6% when comparing ASS/IIM to controls. Noteworthy, when selecting and comparing the subgroups of Jo1 + patients to controls and Jo1 − patients to controls separately, the AUC went up to 88 ± 6% for Jo1 + patients and down to 72 ± 8% for Jo1 − patients (Fig. 2B). An AUC of 88 ± 6% was also reached for the subgroup of patients that had an ILD and/or ASS (ILD/ASS) diagnosis (independent on Jo1 reactivity, n = 28, Fig. 2A,B). At a cut off of −0.22, 69% sensitivity and 92% specificity was obtained for all patients (n = 42) or separately 82/92% for all Jo1 + patients as well as for all ILD/ASS diagnosed patients, and 60/92% for all Jo1 − patients.  Table 1. Overview of the sample types and patient sub-groupings. Note that from eleven Jo1 + patients, anti-Jo1 auto-immune IgG was further enriched. Of these 9 were from patients already included in the study by accessible IgG enriched from serum. Additionally, we had access to anti-Jo1 specific IgG from two more patients, (i.e. 9 + 2)). HC; Healthy Control, IIM; idiopathic inflammatory myopathies, ASS; anti-synthetase syndrome, ILD; interstitial lung disease, Jo1 − ; Jo1 negative, Jo1 + ; Jo1 positive. Age, mean years (SD) 58.6 (11.7) 54.5 (11.2) 61.
The Fc-glycan profile of anti-Jo1 specific IgG is an enhanced version of the Fc-glycan profile observed in total IgG from Jo1 + patients. In a next step the Fc-glycans from the disease specific anti-Jo1 autoimmune reactive IgG (enriched from the IgG from the Jo1 positive patients) were interrogated and compared intra-individually to the IgG prior to anti-Jo1 enrichment, ("PE" in Table 3) and to the remaining non-anti-Jo1 specific IgG following enrichment (termed flow through, FT in Table 3). Strikingly, the anti-Jo1 specific Fc-glycans had a similar (but enhanced profile) of what was characteristic for the total-IgG profile of Jo1 + patients. Thus, the Fc-glycan profiles of anti-Jo1 specific IgG were generally the least bisected, least afucosylated and most agalactosylated compared to the total IgG from Jo1 + patients and subsequently compared to the IgG from Jo1 − patients and the controls ( Table 3

Discussion
This study confirms that the Fc-glycan profile of IgG 1 in ASS/IIM patients contain less galactosylated epitopes compared to healthy controls and further shows that the Fc-glycan profile of Jo1 + patients contains less bisected and afucosylated glycans compared to Jo1 − patients. Importantly, the Fc-glycan profile that signified Jo1 + patients (i.e. a profile with more agalactosylated and less afucosylated and bisected epitopes) were even more pronounced in anti-Jo1 specific enriched IgG. It is well known that Fc-glycans contain less galactosylated epitopes in other autoimmune disorder and that these correlate with disease activity and severity 19,20 . Furthermore, a recent study has connected Fc-glycosylation regulation with the axis of IL-23 and T H 17-cell activity and suggest that this is a determining factor for autoimmune disease onset 21 . The Fc-galactosylation status of IgG can regulate binding to complement 1q factor and complement-dependent toxicity (CDC), thereby influencing the inflammatory status. Furthermore, sialylated Fc-glycans (which can only be attached to galactosylated epitopes), have been directly linked to anti-inflammatory activation via SIGN receptor interactions, and reduce CDC efficacy 22,23 .
Since the significantly higher abundance of agalactosylated FA2 (p = 4.5E-4) in ASS/IIM patients is IgG 1 specific we tested if the Fc-galactosylation status of IgG 2 could serve as an internal control used to intra-individually correct for other factors that are also know to affect the galactosylation status of the N-glycan profile such as age, gender and heritability 18,24 . Hence, by using the factor log(FA2_1/FA2_2), ASS/IIM and controls were distinguished with an AUC of 79 ± 6%. This is an improvement to other Fc-galactosylation status measuring factors such as log(FA2/FA2G2) of IgG 1 , which on this data generated an AUC of 71 ± 6% and log(FA2/ (FA2G1 + FA2G2) of IgG 1 which on this data gave an AUC of 73 ± 6%, respectively 10,25 . From investigating the patients with most sampling time points, it was also evident that the log(FA2_1/FA2_2) factor is more stable than log(FA2/FA2G2), Supplemental Fig. 1B. However, whether the log(FA2_1/FA2_2) factor is more specific for IIM than for RA needs to be evaluated further. For example, in contrast to ASS/IIM patients, the galactosylation status of both IgG 1 and of IgG 2 have been shown to be skewed in RA 10 . Noteworthy, both the subgroup of Jo1 + patients and the subgroup of patients diagnosed with ILD/ASS further improved the sensitivity of the analysis (AUC of 88 ± 6%, Fig. 2B). The majority of ILD/ASS diagnosed patients were also Jo1 + (Table 1). However, 10 of 28 patients included in the ILD/ASS group were Jo1 − patients seropositive for other aaRS. Hence, the over-representation of Fc-agalactosylated IgG may suggest a lung-tailored bias, as virtually no other IIM extra-muscular manifestations were associated with this glycan feature.
Our hypothesis is that the agalactosylated features of ASS/IIM IgG (and anti-Jo1-IgG) are enhanced due to underlying lung disease related mechanisms in these patients. This hypothesis is in line with a growing body of Figure 3. 3D-plot of the three factors (agalactosylation, bisection, and afucosylation) that best distinguished the anti-Jo1 specific IgG (red circles) and Jo1 + patients total IgG (black circles) from both total IgG from controls (blue circles), and from Jo1 − patients (green circles). Thus, the anti-Jo1 specific IgG and the Jo1 + patients total IgG has 1) an increase in IgG 1 agalactosylation (described by log(FA2_1/FA2_2) and 2) a decline in bisected forms (described by log Σ[B]) and 3) a decline in afucosylated forms (described by logΣ [aF]). research that indicate a possible immunological role of IgG Fc-glycans in the respiratory tract. For example, the Fc-galactosylation in chronic lung disease (sarcoidosis and severe asthma) is skewed 26 and lung cancer patients display a high degree of Fc-agalactosylation which is inversely correlated with good diagnostic performance 27 . Furthermore, note that RA often is affecting the lungs 28 .
In contrast to the galactosylation status that was more prominently skewed both in Jo1 + and in the ASS/ ILD subgroup, the lower abundance of bisected-and core afucosylated Fc-glycans is more Jo1 + phenotype explicit. This hypothesis was further strengthened by analysis of Jo1 specific IgG which contained even lower abundance of bisected-and core afucosylated Fc-glycans than total IgG from the same patients. The profiling of IgG Fc-glycans of specific disease autoantibody reactivities is well characterized in ACPA from RA patients 15 . Noteworthy, anti-Jo1-IgG and ACPA Fc-glycan profiles have several similarities. In addition to higher content of agalactosylated epitopes ACPA-IgG 1 contains a lower abundance of afucosylated and bisected forms. This might indicate that IgG-Fc-regulatory properties of ACPA and anti-Jo1 are similar and potentially consistent with other autoantibodies. Both bisected-and core afucosylated Fc-glycans increase IgG affinity to FcγRII and III receptors, potentially resulting in more potent antibody dependent cell-mediated cytotoxicity (ADCC) 29,30 . Thus, lower abundance of these epitopes might indicate an inverse effect.
To further evaluate and test how the anti-Jo1 + Fc-glycan profile correlate with clinical manifestations and proteomic information of IgG chain distributions and trace proteins (measured in the IgG enriched samples), MVA-analyses were performed. As expected (and in line with the hypothesis that the anti-Jo1 Fc-glycan characteristic features are inflammatory regulating), several of the trace proteins found to positively correlate with the Jo1 + patients are involved in inflammatory processes (complement-related, lysozyme and IgD, Fig. 4C). In addition, clinical variables such as ASS and ILD, and proteins (plasminogen and thrombospondin) which have previously been linked to myositis were positively correlating with the Jo1 + patients [31][32][33][34] . Noteworthy, both the log(FA2_1/FA2_2) and the log Σ[B] data correlated weakly but significantly (R 2 = 0.13, p = 0.003 and R 2 = 0.09, p = 0.01, respectively) with accessible C reactive Protein (CRP) data linked to 70 of the measuring time points, (from 23 of the Jo1 − and of the 17 Jo1 + patients). However, in contrast to the Fc-glycan profile, the CRP data was not significant between the two ASS/IIM phenotypes (p = 0.8), thus indicating that the Fc-glycan profile give more precise information on the underlying immunological patient status.
In conclusion, our study shows that both total IgG and specifically anti-Jo1-IgG from ASS/IIM patients display IgG Fc-glycans that comprise features associated with pro-inflammation (i.e. agalactosylation), which were also overrepresented in patients diagnosed with ILD/ASS. Furthermore, the lower abundance in bisected and afucosylated forms appears to be directly linked to anti-Jo1 autoimmune IgG. Functional studies with glycan manipulation recurring to anti-Jo1-IgG (and other autoantibodies) will shed insights into the role of these Fc-glycan aberrations in anti-Jo1 autoantibodies.

Methods
Patients. IIM patients (n = 44) fulfilled the Bohan and Peter 35,36 criteria for definite, probable or possible polymyositis (PM) or dermatomyositis (DM). Inclusion Body Myositis (IBM) was diagnosed according to Griggs criteria 37 . Because samples were collected between 1996 and 2016, the new EULAR/ACR classification criteria for adult and juvenile IIM had not been approved and therefore the old classification criteria was applied 38 . The diagnosis of ASS was based on the presence of aaRS antibodies, plus one of the following features: ILD, myositis, arthritis, Raynaud's phenomenon, fever, or mechanic hands 39 . Clinical characteristics of ASS/IIM patients are summarized in Table 2  Jo1 − n = 25, and 24 age/sex-matched healthy controls (HC) was isolated as previously described 40 . Longitudinal sera samples were available from eleven Jo1 + (i.e. T1-T10) and seven Jo1 − (i.e. T1-T2) patients, (Supplemental Table 1). Anti-Jo1-IgG was enriched after pooling IgG from eight of the Jo1 + patients sampled at time points T1-T10 40 . From three Jo1 + patients a larger volume of sample (9-21 mL) was available at one time point for isolation of anti-Jo1-IgG. For details, see Supplemental Fig. 2A IgG sample preparation and LC-MS/MS analysis. Samples were digested with trypsin in duplicates according to previously described protocols 15,25 . Briefly, ten µg of IgG/sample were dissolved to a final volume of 70 µL with a final concentration of 50 mM ammonium bicarbonate (pH 8). The protein was reduced (5 mM dithiothreitol, 30 min, 56 °C) and alkylated (14 mM iodoacetamide, 30 min in darkness). Trypsin was added in a ratio of 1:50 (enzyme:protein) for overnight digestion at 37 °C. Peptides were then desalted using C18 HyperSep Filer Plates (Thermo Fisher Scientific, Waltham, MA). For peptide elution, 60% acetonitrile in 0.1% formic acid was used. Samples were dried using SpeedVac and stored at −20 °C until LC-MS/MS analysis. Samples were randomized both during the sample work up and during the LC-MS/MS analyses. One microgram of digest/sample was analysed using an UltiMate 3000 system connected to an Elite Orbitrap mass spectrometer (both Thermo Fisher Scientific). Reversed phase nano-LC-separation of the peptides was performed on a 15 cm long EASY spray column (PepMap, C18, 2 µm, 100 Å). LC-gradient and MS conditions are described in the supplementary information. Information on run-to-run variation (obtained from a polyclonal IgG standard run in between samples) and from variation between IgG digests are given in Supplemental Table 7 and in Supplemental Fig. 5, respectively.
Fc-glycopeptide analysis. Using in-house developed software (Quanti) 41 , and as previously described 15,25 , glycopeptide ion abundances were integrated and quantified in a label free manner. The program was set to detect and quantify doubly and triply charged ions from Fc-glycopeptides (IgG 1 :EEQYNSTYR, IgG 2 or IgG 3 : EEQFNSTFR and IgG 4 or IgG 3 : EEQFNSTYR/EEQYNSTFR) as well as triply and quadruply charged ions from (IgG 1 :TKPREEQYNSTYR, IgG 2 or IgG 3 : TKPREEQFNSTFR and IgG 4 or IgG 3 : TKPREEQFNSTYR/ TKPREEQYNSTFR) using their accurate monoisotopic masses (within <10 ppm from the theoretical mass values) and within a ± 2 min interval around expected retention times. Note that IgG 3 occurs as both EEQFNSTFR and EEQYNSTFR, with EEQYNSTFR being less frequent in Caucasians (which were predominant in this study) 15,42,43 . In total 63 glycopeptide variants were screened for (i.e. 21 glycoforms substituting the three Fc-peptide variants, Supplemental Table 8). Examples of extracted ion chromatograms of the Fc-glycopeptides from EEQYNSTYR (IgG 1 ) are shown in Supplementary Fig. 6 from one of the healthy controls and one of the Jo1 specific IgG samples, respectively. For each sample, glycoform abundances were normalized to total content (100%) of Fc-glycosylated IgG 1 , total content (100%) of Fc-glycosylated IgG 2/(3) and total content (100%) of Fc-glycosylated IgG 4/(3) peptides, respectively. Glycoforms detected at levels constituting less than 0.05% of the total IgG Fc-glycan profiles were discarded. The sub-grouping of respective glycan feature (shown in Table 3) was done by taking the sum of the individual glycopeptide abundances (%) obtained for respective glycopeptide. Note that this was done independent on other features in the profile. Thus, when taking the sum of all galactosylated forms (containing galactose Σ[G]) this was done independent of other glycan substitution patterns, such as