Serum IgG N-glycosylation profiles distinguish primary Sjögren’s disease, rheumatoid arthritis, and systemic lupus erythematosus and are associated with disease activity in Sjögren’s disease

Abstract

Background

IgG N-glycosylation regulates immune effector functions and inflammation, but disease-specific patterns in primary Sjögren’s disease (SjD), rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE) are insufficiently defined, particularly in SjD.

Objectives

To identify serum IgG N-glycan biomarkers that differentiate SjD, RA, and SLE and reflect disease activity in SjD.

Design

A clinical case-control study.

Methods

Serum IgG N-glycans from 88 SjD patients, 102 RA patients, 86 SLE patients and 91 healthy controls were profiled using ultra-performance liquid chromatography. Discriminative glycans were selected through machine learning-based feature selection and evaluated for their diagnostic performance using cross-validated logistic regression and receiver operating characteristic (ROC) analysis.

Results

All three diseases exhibited decreased galactosylation and increased bisecting N-acetylglucosamine (GlcNAc). SjD showed the lowest sialylation (19.92%) and the highest bisecting GlcNAc levels (19.36%). Shared features included elevated GP1 (FA1) and reduced GP9 (FA2[3]G1). Compared with RA, SjD was characterized by elevated GP3 (A2B, adjusted odds ratio [AOR] 4.27), GP8 (FA2[6]G1, AOR 3.53) and reduced GP23 (FA2G2S2, AOR 0.07). Elevated GP11 (FA2[3]BG1, AOR 11.99) distinguished SjD from SLE. These glycans outperformed conventional autoantibodies in ROC analysis. In SjD, higher disease activity was associated with reduced galactosylation, with GP5 (M5) positively (AOR 1.71) and GP14 (FA2G2) inversely (AOR 0.43) associated with activity. These glycans showed stronger correlations with disease activity than erythrocyte sedimentation rate or serum IgG levels (GP5: r = 0.462; GP14: r = −0.432), and the combined glycan model achieved an AUC of 0.758.

Conclusions

Distinct IgG N-glycan signatures differentiate SjD, RA, and SLE despite their shared clinical and immunologic characteristics and are associated with disease activity in SjD. These findings provide insights into IgG N-glycosylation in these autoimmune diseases and may inform future research into their potential utility as non-invasive biomarkers, particularly in SjD.

Plain language summary

The immune system produces antibodies to help protect the body from infections. Immunoglobulin G (IgG) is the most abundant antibody in the blood and has small sugar structures attached to it. These sugars influence how the immune system functions and regulate inflammation. Changes in these sugar structures have been reported in several autoimmune diseases. However, it has remained unclear whether different autoimmune diseases have distinct IgG sugar patterns, particularly in primary Sjögren’s disease (SjD), an autoimmune disease that mainly affects the salivary and tear glands. In this study, we analyzed blood samples from people with SjD, rheumatoid arthritis (RA), systemic lupus erythematosus (SLE), and healthy individuals. We examined the sugar patterns attached to IgG to determine whether they could help distinguish these diseases and whether they were associated with disease activity in SjD. Using advanced laboratory techniques and statistical analyses, we identified the IgG sugar patterns that best distinguished the disease groups. We found that all three autoimmune diseases showed IgG sugar changes associated with increased inflammatory activity. However, patients with SjD showed the most pronounced changes and distinct sugar patterns that helped distinguish SjD from RA and SLE. These IgG sugar markers also showed better discriminatory performance than commonly used blood tests. In addition, several IgG sugar changes were associated with disease activity in SjD and showed stronger associations than conventional inflammatory markers. Overall, our findings suggest that IgG sugar patterns may serve as useful blood-based biomarkers to help distinguish SjD from related autoimmune diseases and to assess disease activity. These findings support further investigation of IgG sugar patterns as potential non-invasive biomarkers in autoimmune diseases.

Introduction

Primary Sjögren’s disease (SjD), rheumatoid arthritis (RA) and systemic lupus erythematosus (SLE) are representative systemic connective tissue diseases, a major subset of autoimmune disorders. ¹ They are characterized by immune-mediated connective tissue damage, B-cell hyperactivity, autoantibody production and increased pro-inflammatory cytokine levels. ² These conditions overlap in genetic, epigenetic and hormonal regulation, clinical and serological features, and immunopathology. ³ They predominantly affect women and typically involve multiple organ systems with shared manifestations, such as fatigue, arthralgia and lymphadenopathy. ² Serologically, these diseases can share similar autoantibody profiles, such as antinuclear antibodies (ANA), anti-Sjögren’s syndrome-related antigen A (anti-SSA/Ro) antibodies and rheumatoid factor (RF). These similarities increase the risk of misdiagnosis and diagnostic delay. Indeed, patients with SjD may initially be misdiagnosed with RA or SLE due to shared serological and clinical characteristics. ⁴ Such delays can result in irreversible tissue damage, reduced therapeutic efficacy and poorer long-term outcomes. Therefore, there remains a need for complementary, disease-specific biomarkers to aid in the differentiation of these diseases, which may support timely, personalized treatment and potentially improve patient prognosis and quality of life.

Immunoglobulin G (IgG) is a major serum glycoprotein that carries conserved N-glycans essential for its structural stability and effector functions. ⁵ IgG N-glycosylation regulates the balance between pro- and anti-inflammatory responses via interactions with fragment crystallizable receptors (FcRs) and complement proteins. ⁶ The IgG N-glycome is highly diverse and shaped by genetic, environmental and pathological factors.^7–10 Altered IgG N-glycosylation is linked to aging, inflammation, autoimmune disorders and therapeutic antibody efficacy. ¹¹ In autoimmune diseases, abnormal IgG N-glycosylation may impair immune tolerance and promote chronic inflammation. ¹² Recent glycomics studies highlight its diagnostic, prognostic and disease activity monitoring value for autoimmune diseases, particularly for RA and SLE. ¹³ Emerging evidence suggests that IgG N-glycosylation is specifically altered in SjD, although data remain limited.^14,15 Notably, no studies have directly compared IgG N-glycosylation patterns across patients with SjD, RA and SLE. Particularly, the role of IgG N-glycosylation in reflecting disease activity and its associations with clinical and immunological parameters in SjD remains understudied.

This study aims to identify serum IgG N-glycan profiles as potential non-invasive biomarkers to distinguish SjD, RA and SLE, as well as evaluate their relevance to disease activity, clinical features and pro-inflammatory cytokines in SjD. Cytokines were selected based on their established roles in B cell activation and antibody production in SjD, ¹⁶ and their regulatory influence on IgG N-glycosylation. ¹⁷ Given the complexity of IgG N-glycosylation patterns, traditional statistical methods often struggle to identify stable and clinically meaningful features. ¹⁸ To address this challenge, machine learning-based feature selection combined with rigorous cross-validation strategies were applied. ¹⁹ By integrating glycomic profiling with machine learning, this study seeks to identify IgG N-glycan signatures that can differentiate these autoimmune diseases and are associated with disease activity in SjD, thereby supporting their potential utility as complementary diagnostic biomarkers and indicators of disease activity.^5,20

Methods

Patients and study design

This clinical case-control study enrolled 88 patients with SjD, 102 with RA, and 86 with SLE from the First Affiliated Hospital of Shantou University Medical College Shantou, China, between January 2023 and December 2024. Additionally, 91 healthy controls (HC) without sicca symptoms or diagnosed diseases were recruited from the hospital’s Physical Examination Center.

Patient diagnoses were based on the 2016 American College of Rheumatology (ACR)/European Alliance of Associations for Rheumatology (EULAR) classification criteria for SjD, ²¹ the 2010 ACR/EULAR criteria for RA, ²² and the 2019 ACR/EULAR criteria for SLE. ²³ Inclusion criteria were patients fulfilling the corresponding ACR/EULAR classification criteria for SjD, RA, or SLE, with available serum samples and relevant clinical and laboratory data at the time of evaluation. Exclusion criteria included other autoimmune rheumatic diseases, infections, chronic kidney or hepatic diseases, active malignancies, mental disorders, recent intravenous immunoglobulin (IVIG) use (within one month) and pregnancy or lactation.

Sample size was calculated using G*Power 3.1 for a one-way ANOVA with four groups. Based on published data, ¹⁵ Cohen’s d was 0.51, corresponding to an effect size of f = 0.25. With a two-sided α of 0.05 and a power of 0.80, 45 participants per group were required. After accounting for a 15% dropout rate, the target sample size was set at 53 per group, which was exceeded in the final sample.

This study was approved by the Ethics Committee of the First Affiliated Hospital of Shantou University Medical College (Ethics Code: B-2023-244) on January 3, 2023, and by the Edith Cowan University Human Research Ethics Committee (Ethics Code: 2023-04937-CHEN) on January 13, 2023. All participants provided written informed consent prior to enrolment in the study. This research was conducted ethically in accordance with the World Medical Association Declaration of Helsinki.

Data collection and disease activity assessment

Demographic and clinical data were collected, including age, sex, clinical manifestations, laboratory parameters and disease activity. For SjD patients, glandular dryness, systemic organ involvement and fatigue were specifically documented. Glandular dryness was assessed using the American-European Consensus Group (AECG) questions. ²⁴ Systemic organ involvement was defined according to the EULAR Sjögren’s Syndrome Disease Activity Index (ESSDAI) domains. ²⁵ Laboratory tests included erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), complement 3 (C3), and C4, which were measured by standard automated methods. Autoantibody profiling comprised ANA by indirect immunofluorescence assay on HEp-2 cells at a screening dilution of 1:100, RF by immunoturbidimetry, and anti-SSA (anti-Ro52 and anti-Ro60) and anti-SSB antibodies by line immunoassay. Additionally, focal lymphocytic sialadenitis (FLS) and focus score (FS) were assessed in labial salivary gland biopsies of SjD patients. ²⁶

Disease activity was assessed using validated indices specific to each condition. For SjD, objective disease activity was evaluated using the ESSDAI, with patients classified into low activity (ESSDAI < 5) and moderate-to-high activity (ESSDAI ≥ 5) groups. ²⁷ Subjective symptom burden was assessed using the EULAR Sjögren’s Syndrome Patient Reported Index (ESSPRI), with a score < 5 considered to reflect a patient-acceptable symptom state (PASS). ²⁸ Disease activity in SLE was measured by the Systemic Lupus Erythematosus Disease Activity Index 2000 (SLEDAI-2K) ²⁹ and in RA by the Disease Activity Score using 28 joint counts (DAS28). ³⁰

IgG N-glycan profiling and data processing

Serum IgG was isolated and N-linked glycans released following a widely used, validated protocol known for its reproducibility.^31–34 Briefly, 100 μl of serum per participant was diluted (1:7, v/v) in phosphate-buffered saline, and IgG was isolated using protein G monolithic plates (BIA Separations, Slovenia). N-glycans were released using peptide-N-glycosidase F (PNGase F), labeled with 2-aminobenzamide (2AB), and analyzed by ultra-performance liquid chromatography (UPLC; Waters Corporation, USA).^31,35 Samples were randomly allocated across 96-well plates and analytical batches, and pooled quality control (QC) samples were included in each batch to monitor analytical stability and instrument performance. ³⁵ To minimize inter-batch variation, batch effects were corrected using the ComBat method. ³² Labeled IgG N-glycans were separated into 24 directly measured glycan peaks (GPs, GP1-GP24) and quantified using total area normalization, with each peak expressed as a percentage of total GPs (Table S1). Fifty-four derived glycan (DG) traits (DG1-DG54) were calculated based on ratios of the directly measured GPs (Table S2). Ten derived traits representing four key N-glycosylation features: galactosylation, sialylation, bisecting N-acetylglucosamine (GlcNAc) and fucosylation were also derived from the measured GPs (Table S3). ³⁶ Analytical reproducibility was ensured through pooled QC samples across batches, consistent with previously validated UPLC-based glycomics workflows. ³⁵

Cytokine analysis

Plasma concentrations of B-cell activating factor (BAFF), interferon (IFN)-α, IFN-γ, interleukin (IL)-6, IL-17, IL-21 and tumor necrosis factor-α (TNF-α) were measured using ultrasensitive sandwich enzyme-linked immunosorbent assay (ELISA) kits (Cusabio, USA), following the manufacturer’s instructions. Concentrations were determined using standard curves generated on each plate.

Statistical analysis

Normality was assessed using the Kolmogorov-Smirnov test. Due to skewed distribution, continuous variables were reported as medians with interquartile ranges and compared using the Mann-Whitney U test for two groups or the Kruskal-Wallis H test for multiple groups, with Benjamini-Hochberg correction for multiple testing. Categorical variables were presented as frequencies and percentages, and compared using the chi-square or Fisher’s exact test.

All 24 GPs were z-score standardized (mean = 0, standard deviation = 1) prior to feature selection and model development to ensure comparability and prevent scale-related bias. The dataset was randomly stratified split into 70% for the training set and 30% for the testing set using a fixed random seed within each group to preserve the original class distribution and ensure reproducibility. Given the high dimensionality of IgG N-glycan data relative to sample size, strict prevention of data leakage was prioritized. All feature selection and model development procedures were conducted exclusively within the training set, while the test set was kept completely independent and used only for final performance evaluation. Two complementary feature selection approaches, the adaptive least absolute shrinkage and selection operator (AdaLASSO) ³⁷ and multivariate methods with unbiased variable selection in R (MUVR), ³⁸ were applied to the training set. Only features consistently identified by both methods were retained for downstream modeling to ensure robustness. AdaLASSO improves selection consistency and reduces potential bias in high-dimensional settings by assigning adaptive weights to variables, ³⁷ while MUVR employs repeated double cross-validation to minimize bias and overfitting in correlated data. ³⁸ Logistic regression (LR) models were trained and internally validated using repeated five-fold cross-validation within the training set to ensure stability of parameter estimation. Model performance was then evaluated on the independent held-out test set to assess generalization capability, with 95% confidence intervals estimated using bootstrap resampling (1,000 iterations). Classification performance was primarily evaluated by the area under the receiver operating characteristic (ROC) curve (AUC), with secondary metrics including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), accuracy and F1-score. ³⁹ Subgroup analyses were adjusted for sex, age and antibody profiles, and AUCs were compared using the DeLong test. ⁴⁰ Multivariate linear regression was used to analyze the association between IgG N-glycans and disease activity in SjD, adjusting for patient characteristics and relevant clinical features. Correlations between IgG N-glycan traits and clinical or cytokine data were assessed using Spearman’s correlation for continuous variables, or point-biserial correlation for dichotomous variables.

All analyses were performed using SPSS Statistics version 29.0 (IBM Corp., Armonk, NY, USA), GraphPad Prism version 9.5 (GraphPad Software, San Diego, CA, USA) and R version 4.4.1 (R Core Team, Vienna, Austria, 2024).

This study was reported in accordance with the STROBE guidelines for case-control studies. ⁴¹ The completed checklist is provided as a supplementary file.

Results

Participant characteristics

All three diseases (SjD, RA and SLE) predominantly affected females, with proportions of 95.5%, 91.9% and 83.3%, respectively. Median age was higher in SjD (52 years) and RA (52.5 years), compared to SLE (34 years) (Table 1). Serologically, SjD showed higher positivity rates for ANA, anti-Ro52, anti-Ro60 and elevated serum IgG levels compared to RA, with values comparable to SLE (Table 1). Anti-SSB positivity was significantly higher in SjD than in RA and SLE, whereas RF positivity was highest in RA and comparable between SjD and SLE. Serum IgG levels were lowest in RA, with no significant difference between SjD and SLE. CRP and ESR were lowest in SjD, highest in RA and intermediate in SLE. Cytokine concentrations (IL-6, IL-17, IL-21, TNF-α, BAFF, IFN-α and IFN-γ) were elevated in all disease groups relative to healthy controls, with no significant differences among SjD, RA and SLE (Table 1). Collectively, SjD, RA and SLE exhibited largely shared serological and cytokine profiles.

Table 1.

Clinical and laboratory characteristics of the study participants.

Variables	SjD (n=88)	RA (n=102)	SLE (n=86)	HC (n=91)	P-value
General feature
Female, n (%)	84 (95.5%)	85 (83.3%) a, b	79 (91.9%)	86 (94.5%)	0.012
Age, years	52 (44, 59.75)	52.5 (46.75, 59.25)	34 (26.75, 47) a, b, c	53 (43, 60)	<0.001
Autoantibodies, n/N (%)
ANA positivity	84/88 (95.5%)	29/52 (55.8%) b	74/75 (98.7%)	/	<0.001
Anti-Ro60 positivity	79/88 (89.8%)	2/50 (4.0%) b	55/72 (76.4%)	/	<0.001
Anti-Ro52 positivity	67/88 (76.1%)	5/50 (10.0%) b	46/72 (63.9%)	/	<0.001
Anti-SSB positivity	40/88 (45.5%)	2/50 (4.0%) b	13/72 (18.1%) b	/	<0.001
RF positivity	25/56 (44.6%)	67/82 (81.7%) b	11/28 (39.3%) c	/	<0.001
Serological assessment
IgG, g/L	15.80 (12.70, 19.18)	14.20 (11.80, 16.90) b	15.65 (12.60, 19.20) c	/	0.029
CRP, mg/L	2.66 (1.86, 5.98)	11.95 (4.97, 44.33) b	4.77 (2.24, 12.03) b, c	/	<0.001
ESR, mm/hr	21.00 (11.25, 42.75)	35.00 (22.00, 70.00) b	30.50 (13.00, 60.50) b, c	/	<0.001
Disease activity indexes
SjD: ESSDAI	3.00 (1.00, 6.75)	/	/	/	/
SjD: ESSPRI	2.00 (1.33, 3.33)	/	/	/	/
RA: DAS28-CRP	/	3.01 (1.91, 4.44)	/	/	/
SLE: SLEDAI-2K	/	/	6.00 (3.00, 10.25)	/	/
Cytokine, pg/mL
IL-6	191.24 (100.81, 266.32) a	183.03 (111.34, 260.35) a	162.41 (118.88, 291.95) a	33.10 (25.64, 46.15)	<0.001
IL-17	411.79 (198.23, 643.67) a	490.52 (236.76, 768.62) a	369.51 (219.10, 772.91) a	204.74 (146.16, 293.22)	<0.001
IL-21	33.79 (22.50, 44.04) a	35.60 (23.94, 46.73) a	34.02 (26.13, 45.96) a	14.18 (11.88, 15.93)	<0.001
TNF-α	59.68 (44.46, 74.02) a	60.25 (40.42, 77.60) a	68.61 (46.50, 82.46) a	27.88 (22.28, 39.64)	<0.001
BAFF	2.81 (1.76, 4.09) a	3.01 (1.92, 4.29) a	3.79 (2.67, 4.63) a	0.87 (0.60, 1.07)	<0.001
IFN-α	79.84 (53.84, 108.11) a	71.60 (44.58, 106.57) a	62.36 (35.54, 95.55) a	25.81 (19.56, 28.47)	<0.001
IFN-γ	3.17 (2.29, 4.33) a	3.19 (1.89, 4.37) a	3.59 (2.54, 4.45) a	2.14 (1.85, 2.46)	<0.001

^aStatistically significant compared to HC.

^bStatistically significant compared to SjD.

^cStatistically significant compared to RA.

Percentages are based on the number of patients with available data for each variable (n/N).

P < 0.05 was considered statistically significant.

Anti-SSB, anti-Sjögren’s syndrome-related antigen B; ANA, antinuclear antibody; BAFF, B cell activating factor; CRP, C-reactive protein; DAS28, Disease Activity Score using 28 joint counts; ESR, erythrocyte sedimentation rate; EULAR, European League Against Rheumatism; ESSDAI, EULAR Sjögren’s syndrome disease activity index; ESSPRI, EULAR Sjögren’s syndrome patient reported index; HC, healthy controls; IFN, interferon; IgG, immunoglobulin G; IL, interleukin; RA, rheumatoid arthritis; RF, rheumatoid factor; SjD, Sjögren’s disease; SLE, systemic lupus erythematosus; SLEDAI-2K, Systemic Lupus Erythematosus Disease Activity Index 2000; TNF, tumor necrosis factor.

Comparison of IgG N-glycan profiles among SjD, RA, SLE and healthy controls

IgG N-glycosylation differed among SjD, RA, SLE and healthy controls, as reflected by 24 GPs (Table S4), 54 DGs (Table S5) and ten derived traits representing four key glycosylation features (Figure 1, Table S6). SjD showed the highest IgG asialylation (S0, 79.95%, P < 0.01) and the lowest total sialylation (total S, 19.92%, P < 0.05), with no significant differences among RA, SLE and healthy controls. Total galactosylation (total G) was reduced in SjD (47.20%), RA (42.75%) and SLE (48.83%) compared with healthy controls (51.46%) (all P < 0.05). IgG agalactosylation (G0) was elevated in SjD (31.38%) and RA (35.07%) compared with SLE (28.05%) and healthy controls (26.96%) (P < 0.05). Bisecting GlcNAc (B) was higher in SjD (19.36%), RA (17.70%) and SLE (17.94%) than in healthy controls (16.83%), with SjD showing the highest abundance (P < 0.05). Core fucosylation (F) did not differ significantly among groups. Collectively, these results suggested that IgG N-glycosylation alterations were most pronounced in SjD.

Figure 1.

Comparison of the abundances of derived traits between patients with SjD, SLE, RA and healthy controls. Boxes span the interquartile range (IQR, 25th-75th percentiles), with the median indicated by the line inside each box. Whiskers extend to the most extreme observations within 1.5×IQR, and points outside this range are plotted individually as outliers. Significant differences are indicated by asterisks (*P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001). B, bisecting N-acetylglucosamine (GlcNAc); F, core fucosylation; G, galactosylation; G0/G1/G2, a-/mono-/di-galactosylation; HC, healthy controls; IQR, interquartile range; RA, rheumatoid arthritis; S, sialylation; S0/S1/S2, a-/mono-/di-sialylation; SjD, Sjögren’s disease; SLE, systemic lupus erythematosus.

IgG N-glycan biomarkers for classification of three autoimmune diseases

A panel of IgG N-glycans was consistently identified by AdaLASSO and MUVR to distinguish SjD, RA and SLE from each other and from healthy controls (Figure S1). After adjustment for age and sex in multivariable LR models, only significant glycans were retained (Table S7). Compared with healthy controls, all three diseases were associated with increased GP1 (FA1) and decreased GP9 (FA2[3]G1), with SjD additionally associated with reduced GP21 (A2G2S2), RA with elevated GP4 (FA2) and SLE with reduced GP20 (FA2FG2S1). Compared to RA, SjD was associated with increased GP3 (A2B) and GP8 (FA2[6]G1) and decreased GP23 (FA2G2S2), with adjusted odds ratios (AORs) and 95% confidence intervals (CIs) of 4.27 (1.77-10.31), 3.53 (1.68-7.45) and 0.07 (0.02-0.22), respectively. SjD was also associated with a marked elevation in GP11 (FA2[3]BG1) (AOR: 11.99, 95% CI: 4.59-31.27) compared with SLE. SLE demonstrated a significant association with decreased GP4 (FA2) (AOR: 0.23, 95% CI: 0.10-0.53) and GP23 (FA2G2S2) (AOR: 0.05, 95% CI: 0.03-0.15) compared with RA (Figure 2).

Figure 2.

Forest plot showing adjusted odds ratios (AORs) of GPs from logistic regression for pairwise comparisons among patients with SjD, RA, SLE and healthy controls, adjusted for sex and age. AOR, adjusted odds ratio; CI, confidence interval; GP, glycan peak; HC, healthy controls; RA, rheumatoid arthritis; SjD, Sjögren’s disease; SLE, systemic lupus erythematosus.

Glycan-based LR models demonstrated outstanding discrimination, with cross-validated AUCs of 0.919 to 0.958 in training sets and 0.904 to 0.944 in independent testing sets, accurately distinguishing SjD, RA and SLE from healthy controls and from each other (Table S8, Figure 3). The DeLong test confirmed that glycan-based models consistently outperformed sex- and age-only models in AUC for both training and testing sets (P < 0.05) (Table S9-S10), indicating that glycan features add discriminatory value beyond basic demographics. Moreover, the values of all performance metrics, including sensitivity, specificity, PPV, NPV, accuracy and F1-score, exceeded 0.75, demonstrating robust and well-balanced performance of the glycan-based models across different evaluation criteria (Table S8).

Figure 3.

ROC curves for pairwise classification of patients with SjD, RA, SLE and HC based on sex and age, glycan profiles and their combination. Panels show results for the training sets (A, C, E, G, I, K) and testing sets (B, D, F, H, J, L) for the following comparisons: SjD vs. HC, RA vs. HC, SLE vs. HC, SjD vs. RA, SjD vs. SLE and SLE vs. RA, respectively. The “training set” and “testing set” refer to the datasets randomly split for model development and validation, respectively. AUC, area under the curve; CI, confidence interval; HC, healthy controls; SjD, Sjögren’s disease; RA, rheumatoid arthritis; ROC, receiver operating characteristic; SLE, systemic lupus erythematosus.

Subgroup analyses of patients with available antibody data (Tables S11-S13) showed that glycan-based models outperformed conventional markers. In differentiating SjD from RA, the glycan-based model outperformed RF alone, achieving AUCs of 0.938 and 0.706, respectively (DeLong test, P < 0.001; Figure S2A). In distinguishing SjD from SLE, it exceeded the conventional antibody-based model, comprising ANA, anti-Ro52, anti-Ro60 and anti-SSB, with AUCs of 0.921 and 0.691, respectively (DeLong test, P < 0.001; Figure S2B). These findings highlighted the diagnostic potential of IgG N-glycan biomarkers.

IgG N-glycosylation in relation to disease activity, cytokines and clinical phenotypes in SjD

Given the pronounced IgG N-glycosylation changes in SjD and the limited existing research, we investigated their associations with disease activity, cytokine levels and clinical phenotypes based on comprehensive data.

Among 88 SjD patients, 51 had low disease activity (ESSDAI < 5) and 37 had moderate-to-high disease activity (ESSDAI ≥ 5) (Table S14). Comparative analysis revealed increased G0 and decreased total G, monogalactosylation (G1) and digalactosylation (G2) in patients with moderate-to-high disease activity (Table S15-S16, Figure S3).

AdaLASSO and MUVR identified GP5 (M5) and GP14 (FA2G2) as key traits distinguishing moderate-to-high from low disease activity in SjD (Figure S4). In a sex- and age-adjusted LR model, GP5 (M5) was positively associated with moderate-to-high disease activity (AOR: 1.71, 95% CI: 1.01-2.90), whereas GP14 (FA2G2) showed an inverse association (AOR: 0.43, 95% CI: 0.22-0.84) (Table S17).

The glycan-based model achieved an AUC of 0.758 for distinguishing between low and moderate-to-high disease activity in SjD, outperforming the antibody profile (ANA, anti-Ro60, anti-Ro52, anti-SSB) (AUC = 0.564), biological indicators (IgG, C3, C4) (AUC = 0.688), and inflammatory markers, including ESR (AUC = 0.739) and CRP (AUC = 0.654) (Figure 4). Despite significance only against the antibody profile (DeLong test, P = 0.017), its higher AUC suggests potential superiority in assessing disease activity (Table S18).

Figure 4.

ROC curves illustrating the discriminative performance of different models in classifying low versus moderate-to-high disease activity groups in patients with SjD. AUC, area under the curve; C3, complement 3; C4, complement 4; CI, confidence interval; CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; GP, glycan peak; IgG, immunoglobulin G; ROC, receiver operating characteristic; SjD, Sjögren’s disease.

Partial correlation analysis adjusted for age and sex showed that GP5 (M5) correlated positively with ESSDAI (r = 0.462, P < 0.01), whereas GP14 (FA2G2) correlated negatively (r = -0.432, P < 0.01). These correlations were stronger than those for serum IgG (r = 0.194, P > 0.05) and ESR (r = 0.428, P < 0.01) (Figure S5). Multivariate linear regression further showed that GP5 (B = 8.60, P = 0.029) and GP14 (B = -0.34, P = 0.020) were independently associated with disease activity (Table S19).

Correlation analysis revealed exploratory bivariate associations between glycan traits and clinical manifestations in SjD (Figure 5(A)). GP5 (M5) correlated positively with anti-Ro52 positivity, elevated serum IgG levels and lymphadenopathy, articular and multisystem involvement, indicating an association with more severe manifestations. In contrast, GP14 (FA2G2), G2 and total G correlated negatively with lymphadenopathy, pulmonary, constitutional and multisystem involvement. G0 correlated positively with serum IgG, whereas G1, G2 and total G correlated negatively. Double anti-Ro52/Ro60 positivity correlated positively with G2 (including GP14) but negatively with G0, while double anti-Ro60/SSB positivity correlated positively with G1. These patterns suggested that G0 may reflect immune activation and disease severity, whereas higher galactosylation may indicate milder phenotypes, highlighting the potential of IgG N-glycosylation for disease activity stratification in SjD.

Figure 5.

Correlations of IgG N-glycosylation with (A) clinical phenotypes and (B) cytokines in patients with SjD. Heatmaps show correlations between IgG N-glycan features (x-axis) and (A) clinical phenotypes or (B) serum cytokines (y-axis), with the color scale representing point-biserial correlation in (A) and Spearman correlation in (B). Significant correlations are indicated by asterisks (*P < 0.05; **P < 0.01; ***P < 0.001). This analysis is exploratory and no adjustment for multiple comparisons was applied. Anti-SSA, anti-Sjögren’s syndrome antigen A; anti-SSB, anti-Sjögren’s syndrome antigen B; ANA, antinuclear antibody; B, bisecting N-acetylglucosamine (GlcNAc); BAFF, B cell activating factor; C3, complement 3; C4, complement 4; F, core fucosylation; FLS, focal lymphocytic sialadenitis; FS, focus score; G, galactosylation; G0/G1/G2, a-/mono-/di-galactosylation; GP, glycan peak; IFN, interferon; IgG, immunoglobulin G; IL, interleukin; PASS, patient-acceptable symptom state; PNS, peripheral nervous system; S, sialylation; S0/S1/S2, a-/mono-/di-sialylation; SjD, Sjögren’s disease; TNF, tumor necrosis factor.

Similarly, exploratory correlation analysis between glycan traits and cytokines revealed that BAFF positively correlated with GP14 (FA2G2), G2 and total G, but negatively with core fucosylation. Conversely, IFN-α negatively correlated with GP14 (FA2G2), total S and G2, while positively correlated with G1 (Figure 5(B)). These findings suggested that distinct IgG N-glycosylation patterns may be linked to cytokine-mediated alterations observed in SjD.

Discussion

This study provides the first direct comparison of IgG N-glycosylation patterns across SjD, RA, SLE and healthy controls, revealing both shared and disease-specific IgG N-glycan signatures. Using machine learning-based feature selection, we identified robust glycan biomarkers with diagnostic potential despite high-dimensional data. In SjD, we further characterized disease-specific alterations linked to clinical features, disease activity and pro-inflammatory cytokines. From a clinical perspective, these findings may help address key challenges in the differential diagnosis of SjD, particularly in patients with shared autoimmune features, and suggest that IgG N-glycan profiling may have potential as a non-invasive complement to conventional serological markers. These findings deepen the understanding of autoimmune-related IgG N-glycosylation and highlight their potential relevance for future biomarker development in autoimmune diseases.

A shared IgG N-glycosylation signature across these autoimmune diseases included reduced galactosylation and increased bisecting GlcNAc, reflecting a pro-inflammatory profile. ⁶ All three diseases showed elevated agalactosylated GP1 (FA1) and decreased mono-galactosylated GP9 (FA2[3]G1). Consistently, our previous work across multiple rheumatic diseases, including axial spondyloarthritis, RA, SLE and gout, also identified reduced levels of GP9 (FA2[3]G1), ³⁶ highlighting IgG N-glycosylation as a potential generic biomarker of immune-mediated inflammation.

Consistent with our findings, reduced IgG galactosylation, or increased IgG G0, has been reported across many autoimmune diseases, ¹² including SjD,^14,42 suggesting this glycosylation change may be a common feature of autoimmunity and contribute to loss of immune tolerance. Functionally, IgG G0 exposes terminal GlcNAc residues that can bind mannose-binding lectin (MBL) and activate the lectin complement pathway in vitro, although in vivo its pro-inflammatory activity is thought to be mediated primarily through FcγRs rather than MBL.^43,44 In contrast, galactosylated IgG is generally anti-inflammatory, engaging inhibitory receptors such as FcγRIIb, Dectin-1 and Galectin-3 to suppress C5a-mediated responses. ⁴⁵ Increased galactosylation may also competitively block pathogenic autoantibodies from activating FcRs, promoting immune homeostasis. ⁴⁶ Overall, elevated G0 and loss of terminal galactoses are linked to chronic inflammation, though whether these changes are causal or consequential remains unclear.

Elevated bisecting GlcNAc was also common to RA, SLE and SjD, and was most pronounced in SjD. This modification may enhance FcγRIIIa binding and antibody-dependent cellular cytotoxicity (ADCC) by interfering with core fucosylation, even in highly fucosylated IgG, increasing ADCC activity by 7- to 10-fold. ⁴⁷ These findings suggest that bisecting GlcNAc may amplify antibody-mediated effector functions and sustain chronic inflammation, particularly in SjD, though causality remains to be established.

Beyond shared features, SjD was characterized by the lowest IgG sialylation and highest S0 levels, indicating greater disruption of anti-inflammatory glycan structures, consistent with previous reports. ¹⁴ Specific glycan traits, including elevated asialylated glycan GP3 (A2B), GP8 (FA2[6]G1) and GP11 (FA2[3]BG1) and reduced sialylated glycan GP23 (FA2G2S2), showed superior discrimination of SjD from RA and SLE compared with conventional serological markers, suggesting IgG N-glycosylation as a complementary diagnostic biomarker. Mechanistically, IgG S0 enhances ADCC by increasing hinge flexibility and FcγRIIIa binding, whereas sialylated IgG exerts anti-inflammatory effects by reducing FcγRIIIa affinity to suppress ADCC, upregulating inhibitory FcγRIIb to inhibit complement-dependent cytotoxicity (CDC) and engaging type II FcRs to suppress inflammation. ⁴⁸ Thus, reduced sialylation and increased S0 in SjD may reflect disrupted immune homeostasis, though whether these alterations contribute directly to pathogenesis or represent secondary effects remains to be determined.

Age- and hormone-related changes in IgG N-glycosylation may contribute to the female predominance and late onset of SjD. IgG galactosylation declines after age 40, with women showing lower levels than men, a process partially mitigated by estrogen replacement therapy.^10,49 Sialylation shows similar gender- and age-related patterns.^10,49 In genetically predisposed individuals, aging and decreased estrogen shift IgG glycosylation toward pro-inflammatory profiles, promoting chronic inflammation and potentially triggering the transition from preclinical to clinical disease. ⁴⁷ Reduced galactosylation and sialylation have been linked to the onset of other autoimmune diseases^50–52 and are more pronounced in SjD patients than in at-risk individuals with dryness symptoms or anti-SSA/SSB antibodies. ¹⁴ These observations suggest that age- and hormone-related IgG N-glycosylation changes may underlie the higher prevalence of SjD in postmenopausal women, although longitudinal studies are needed to confirm these associations.

IgG N-glycosylation patterns not only discriminate among autoimmune diseases but are also associated with disease activity in SjD. In our study, higher disease activity in SjD was associated with reduced galactosylation and increased G0, both of which were associated with lymphadenopathy and multisystem involvement, consistent with previous reports. ¹⁴ Notably, decreased digalactosylated GP14 (FA2G2) and increased high-mannose GP5 (M5) showed stronger associations with disease activity than conventional indicators such as ESR, serum IgG and autoantibodies, suggesting their potential as complementary biomarkers for disease activity assessment. After adjusting for serum IgG and ESR, GP5 and GP14 remained significantly associated with SjD disease activity, suggesting that these associations are not fully driven by systemic inflammation. In addition to the pro-inflammatory effects of G0, high-mannose glycans may enhance FcγRIIIa binding by lacking core fucose, promoting ADCC, immune activation and possibly sustained inflammation. ⁵³ These findings indicate that elevated G0 and high-mannose IgG glycans are linked to disease activity in SjD and may amplify immune responses, although experimental validation is needed.

Correlation analyses also indicated that BAFF and IFN-α may be associated with IgG N-glycosylation in SjD, particularly galactosylation. In SjD, BAFF promotes B-cell survival and differentiation within ectopic germinal centers, sustaining autoimmune responses, ⁵⁴ while IFN-α induces BAFF production, forming a self-perpetuating inflammatory loop. ⁵⁵ These observations imply potential crosstalk between BAFF, IFN-α and IgG glycosylation in SjD, which warrants further experimental validation. ¹⁷

While the biological mechanisms discussed above may help explain the observed disease- and activity-associated IgG N-glycan patterns, concomitant immunomodulatory treatments should also be considered as a potential source of variation. Across autoimmune diseases, some studies have reported changes in selected IgG N-glycan traits following immunomodulatory treatment, often in parallel with clinical improvement, whereas others have observed little or no significant change. These variable findings suggest that treatment-related influences may depend on the therapeutic agent, disease context, treatment response, and glycan feature examined, and may also be difficult to distinguish from changes accompanying reduced inflammatory activity.^32,49,56 Because comprehensive treatment information was unavailable in the present study, treatment-related influences could not be evaluated separately. Future studies incorporating detailed medication exposure and longitudinal assessment may help clarify whether, and to what extent, specific therapies contribute to variation in IgG N-glycosylation.

Clinically, IgG N-glycan profiling may provide complementary diagnostic information for distinguishing SjD from SLE or RA, which often share similar clinical and serological features. However, as a minimally invasive serum-based approach, the diagnostic signatures identified in this study were internally validated only. External validation in independent, multicenter cohorts, together with further standardization of analytical platforms, is required before clinical implementation. In addition, the observed correlation between glycan patterns and disease activity suggests potential utility for disease activity assessment, although this application requires further validation in prospective, longitudinal, and real-world studies.

Limitation

The high dimensionality of IgG N-glycan data, limited sample size, single-center design, and lack of an independent external validation cohort represent key limitations of this study, which collectively restrict the external validity and generalizability of the findings. While these constraints cannot be fully addressed within the present design, the use of complementary machine learning-based feature selection combined with rigorous cross-validation has mitigated overfitting risk and supported the internal robustness of the identified signatures. Residual confounding from therapies (e.g., immunosuppressive agents and corticosteroids), disease duration, inflammatory markers, and other unmeasured clinical variables related to disease heterogeneity cannot be excluded, as detailed information on these variables was not comprehensively available. The independent discriminative value of these glycan biomarkers in both disease discrimination and disease activity assessment therefore warrants further validation. Correlation analyses between IgG N-glycosylation and clinical phenotypes or cytokines were exploratory and not adjusted for multiple testing, warranting cautious interpretation. As this was a case-control study with cross-sectional assessment of disease activity, the observed associations cannot support conclusions regarding longitudinal disease monitoring, clinical utility, or causal relationships. In addition, this study is based on total IgG N-glycosylation and does not resolve Fc- and fragment antigen-binding (Fab)-specific glycosylation or IgG subclass-specific patterns, which limits mechanistic resolution. Nevertheless, protein G-based UPLC profiling predominantly reflects the conserved Fc glycome, supporting the biological interpretability of the observed systemic glycan signatures. Future studies with larger, multi-center cohorts, comprehensive clinical metadata, longitudinal designs, subclass-resolved liquid chromatography-tandem mass spectrometry glycopeptide profiling and functional validation are warranted to further establish the clinical applicability of IgG N-glycan biomarkers in SjD and related autoimmune diseases.

Conclusions

This study suggests that IgG N-glycosylation profiles may serve as candidate biomarkers for differentiating SjD, RA and SLE, three associated autoimmune diseases. These glycosylation patterns reveal both shared and disease-specific immune dysregulation across these conditions. In SjD, altered IgG N-glycosylation is associated with disease activity, antibody effector functions, cytokine profiles, and age- or hormone-related influences, providing an integrated perspective on immune imbalance. Overall, these findings may inform future studies on the role of IgG N-glycosylation in autoimmune diseases, particularly in SjD.

Appendix.

Abbreviations

2AB

2-aminobenzamide

ACR

American College of Rheumatology

AdaLASSO

Adaptive least absolute shrinkage and selection operator

ADCC

Antibody-dependent cellular cytotoxicity

AECG

American European Consensus Group

ANA

Antinuclear antibody

Anti-SSA

Anti-Sjögren’s syndrome-related antigen A

Anti-SSB

Anti-Sjögren’s syndrome-related antigen B

AOR

Adjusted odds ratio

AUC

Area under the curve

BAFF

B-cell activating factor

B

Bisecting GlcNAc

C3

Complement 3

C4

Complement 4

CDC

Complement-dependent cytotoxicity

CI

Confidence interval

CRP

C-reactive protein

DAS28

Disease activity score using 28 joint counts

DG

Derived glycan

ELISA

Enzyme-linked immunosorbent assay

ESR

Erythrocyte sedimentation rate

ESSDAI

European Alliance of Associations for Rheumatology Sjögren’s Syndrome Disease Activity Index

ESSPRI

European Alliance of Associations for Rheumatology Sjögren’s Syndrome Patient Reported Index

EULAR

European Alliance of Associations for Rheumatology

F

Fucosylation

FcRs

Fragment crystallizable receptors

FLS

Focal lymphocytic sialadenitis

FS

Focus score

G

Galactosylation

G0

Agalactosylation

G1

Monogalactosylation

G2

Digalactosylation

GlcNAc

N-acetylglucosamine

GP

Glycan peak

HC

Healthy controls

IFN

Interferon

IgG

Immunoglobulin G

IL

Interleukin

IVIG

Intravenous immunoglobulin

LR

Logistic regression

MBL

Mannose-binding lectin

MUVR

Multivariate methods with unbiased variable selection in R

NPV

Negative predictive value

OR

Odds ratio

PASS

Patient-acceptable symptom state

PNGase F

Peptide-N-glycosidase F

PPV

Positive predictive value

RA

Rheumatoid arthritis

RF

Rheumatoid factor

ROC

Receiver operating characteristic

S

Sialylation

S0

Asialylation

S1

Monosialylation

S2

Disialylation

SjD

Sjögren’s disease

SLE

Systemic lupus erythematosus

SLEDAI-2K

SLE Disease Activity Index 2000

UPLC

Ultra-performance liquid chromatography

ORCID iDs

Zhixian Chen https://orcid.org/0000-0002-2970-3346

Ling Lin https://orcid.org/0000-0001-5603-0688

Ethical considerations

This study was approved by the Ethics Committee of the First Affiliated Hospital of Shantou University Medical College (Ethics Code: B-2023-244) on January 3, 2023, and by the Edith Cowan University Human Research Ethics Committee (Ethics Code: 2023-04937-CHEN) on January 13, 2023.

Consent to participate

All participants provided written informed consent prior to enrolment in the study. This research was conducted ethically in accordance with the World Medical Association Declaration of Helsinki.

Consent for publication

The manuscript does not contain any identifiable individual data, images, or videos.

Data Availability Statement

The data supporting the findings of this study are available within the article and its supplementary materials.*