This case-control study recruited 188 SLE patients (94 patients with LN and 94 age-matched patients without LN) from the Liaocheng People’s Hospital between July 2020 and September 2021. The inclusion criteria were as follows: (1) Chinese Han ethnicity; (2) female; (3) diagnosis of SLE and (4) SLE patients with LN who met the diagnostic criteria for LN. The exclusion criteria were (1) the coexistence of other autoimmune diseases; (2) primary glomerular diseases or secondary renal diseases; (3) severe trauma, malignant tumors, and severe infections; (4) People who have taken glucocorticoids or immunosuppressants within a month and (5) incomplete patient clinical data. A total of 94 age-matched SLE patients without LN were included in the control group, who newly diagnosed SLE and have not received corresponding treatment. The study was conducted under the supervision of the Ethics Committee of Shandong First Medical University (approval number: 201712). The study was conducted with the written informed consent of all participants.

The current diagnosis of SLE was based on the “American College of Rheumatology (ACR) in 2019” guidelines, which requires a patient to present with 4 out of 11 symptoms/disorders (28), including malar rash, discoid erythema, photic hypersensitiveness, mouth ulcer, joint inflammation, pericarditis or pleurisy, urinary protein > 0.5 g/24 h or +++, seizures or psychosis, hematological diseases, anti-Sm antibody and dsDNA antibody positivity, and antinuclear antibody positivity.

The diagnosis of LN in this study was based on the “Guidelines for the Diagnosis and Treatment of Lupus Nephritis in China” (29). The diagnosis of LN mainly included three criteria: ① persistent urine protein >0.5 g/24 hours, random urine protein +++ or a urine protein-to-creatinine ratio >50 mg/mmol; ② active urinary sediment (excluding urinary tract infection) with >5 white blood cells per high-power field, >5 red blood cells per high-power field, red blood cell casts, or white blood cell casts; ③ cellular casts including red blood cell casts, granular casts, hemoglobin casts, tubular casts, or mixed casts. When SLE patients exhibit any of the above clinical and laboratory abnormalities, a diagnosis of LN was made.

We detected plasma IgG N-glycan profiles using hydrophilic interaction chromatography (HILIC)-ultra-performance liquid chromatography (UPLC) (Waters Corporation, Milford, MA, United States), which was described in a previous study (30). Briefly, the method involved the following steps: (1) plasma IgG was separated with a 96-well Protein G extraction plate (CIM^® r-Protein G 0.2 ml Monolithic 96-well Plate, BIA Separations); (2) IgG Fc N-linked glycans were released by PNGase F enzyme (150 units, Roche); (3) the released IgG N-glycans were labeled with the fluorescent dye 2-aminobenzamide (2-AB, Sigma); and (4) fluorescently labeled N-glycans were detected and analyzed by HILIC-UPLC. The HILIC-UPLC approach initially obtained 24 glycan peaks (GP1-GP24) corresponding to a distinct glycan structure. Composition and structure of IgG initial glycans are available in the Supplementary Table S1 (31). An effective method for normalizing glycan measurements across samples is to divide the peak areas of each glycan by the total areas of the respective chromatograms (32).Typical chromatograms for LN and non-LN samples are provided in the Supplementary Figure S1 (33). From the 24 initial GPs, 17 derived traits were calculated and the calculation formulas can be found in the Supplementary Table S2 (34), which represented the relative abundances of four glycosylation features: core fucosylated, sialylated, bisected GlcNAc, and galactosylated IgG N-glycans (35). Among them, the level of aGal/Gal ratio was calculated from the relative intensity of agalactosylated (G0) vs. monogalactosylated (G1) and digalactosylated (G2) fucosylated biantennary glycans according to the formula of G0/(G1 + G2×2) as we previously described.

A blood sample was obtained from the large antecubital vein of each participant through venipuncture in the morning after an overnight fast. The collected sample was divided into two tubes for further analysis. The sample collected in a vacuum tube containing ethylenediaminetetraacetic acid (EDTA) underwent plasma separation to detect IgG N-glycans, while the sample collected in a tube without EDTA underwent serum separation to determine inflammatory factors and blood biochemical indices.

Face-to-face interviews, a series of clinical examinations and laboratory tests were used to collect basic patient information. Demographic characteristics such as age, sex, and the participants’ history of hypertension, diabetes, and dyslipidemia were collected. Body mass index (BMI) was calculated using measurements of height and weight, with the formula weight (kg)/height² (m²). Hypertension was defined as the use of antihypertensive medication, a self-report of a history of hypertension, and diastolic blood pressure (DBP) ≥ 90 mmHg or systolic blood pressure (SBP) ≥ 140 mmHg. Diabetes was defined as the current use of insulin or oral hypoglycemic agents, a history of diabetes, or a fasting blood glucose level ≥ 7.0 mmol/L (126 mg/dL). Dyslipidemia was defined as the current use of lipid-lowering therapy; a self-report of a history of dyslipidemia; or serum triglycerides (TGs) ≥ 1.7 mmol/L, total cholesterol (TC) ≥ 5.18 mmol/L, high-density lipoprotein cholesterol (HDL-C) < 1.04 mmol/L, or low-density lipoprotein cholesterol (LDL-C) ≥ 3.37 mmol/L, according to the guidelines for the prevention and control of dyslipidemia in adults in China (36).

Continuous variables with normal distributions are expressed as the mean ± standard deviation (SD) and were compared using Student’s t test. Nonnormally distributed data are presented as the median and interquartile range (IQR) and were compared with the Kruskal-Wallis rank-sum test. Categorical variables are presented as n (%) and were evaluated with the chi-square test or Fisher’s exact test. A multivariate logistic regression model was used to evaluate the association of influencing factors with LN, by which the odds ratio (OR) and its 95% confidence interval (CI) were calculated. In logistic regression analysis, the IgG N-glycan measurements were categorized into four groups: Q1 (values ≤ P25), Q2 (P25 < values ≤ P50), Q3 (P50 < values ≤ P75), and Q4 (values > P75). Receiver operating characteristic (ROC) curve analysis was employed to evaluate the performance of the diagnostic models.

The statistical analyses were performed using SPSS 25.0 (IBM, Armonk, NY, United States) and R software (version 3.4.3, R Core Team). All statistical tests were two-sided, and the significance level was set as P < 0.05.

This study included 94 SLE women with LN (mean age 40.10 ± 12.86 years) and 94 age-matched non-LN patients (mean age 41.06 ± 11.83 years). As shown in Table 1, no statistically significant difference in age was detected between the LN group and the control group. Compared with the control group, LN patients had significantly lower serum white blood cell (WBC) counts and significantly higher levels of C3. In addition, the prevalence of anemia was significantly higher in LN patients than in controls.

A total of 24 initial glycan peaks (GP1 - GP24) were obtained from all the chromatograms of UPLC. The distribution of the initial glycans between the LN group and the control group was listed in Figure S2.

The median GP10 level was 3.83 (IQR = 3.27 to 4.68) in the LN group, which was significantly higher than that in the control group (M = 3.37, IQR = 2.87 to 4.07). The level of GP8 containing a galactose in the LN group was 15.66 (IQR = 14.14 to 17.09), which was significantly lower than that in the control group (M = 16.55, IQR = 15.17 to 17.68). The level of GP14 with 2 galactoses decreased significantly when compared to the controls (Table S3; Figures 1, S2). Then, the calculated aGal/Gal ratio, G0 and G2 in the derivatized glycans showed statistically significant differences, and the aGal/Gal ratio and the content of G0 in the LN group were greater than those in the control group; in contrast, the content of G2 in the LN group was less than that in the control group (Table S4).

In the LN group, we observed a statistically significant decrease in all sialylated glycans (Stotal) and in derived trait S1 (Table S4). The level of GP18 in the LN group was 10.29 (IQR = 8.58 to 12.48), which was significantly lower than that in the control group (M = 12.11, IQR= 10.02 to 14.25). The level of GP23 with 2 sialic acids decreased significantly compared with the control group (Table S3; Figures 1, S2).

The differences in glycans that contained core fucosylation were as follows. The levels of GP4, GP6 and GP10 in the LN group were significantly higher than those in the control group. In contrast, the levels of GP8, GP14, GP18 and GP23 in the LN group were significantly lower than those in the control group (Table S3; Figure 1). The decrease was expressed in derived traits related to core fucosylation (F, FS, FG1 and FG2). The derivatized glycan FG0 showed a statistically significant increase compared with the control group (Table S4). At the same time, major glycans that contained bisecting GlcNAc (GP6 and GP10) increased significantly, and the increase was more pronounced in derived traits (B, BN and BS) that measured bisecting GlcNAc between patients and controls (Tables S3, S4; Figures 1, S2).

To further confirm the association between the initial glycans and LN, a multivariable logistic regression analysis was performed. As shown in Figure 2, after adjusting for age, BMI, hypertension, diabetes and dyslipidemia, the relative abundance of GP2, GP4, GP6, and GP10 of LN among SLE women were higher, while the relative abundance of GP8, GP14, GP16, GP18, and GP23 were reduced. In addition, the relative abundance of sialic acid glycans (Stotal, OR = 0.676, 95% CI: 0.515-0.888; S1, OR = 0.666, 95% CI: 0.506-0.875) was decreased in the LN group. In the LN group, the relative abundance of galactosylated glycans (G2, OR = 0.700, 95% CI: 0.534-0.919; FG1, OR = 0.748, 95% CI: 0.572-0.978; FG2, OR = 0.643, 95% CI: 0.488-0.846) was decreased, while the relative abundance of nongalactosylated glycans (G0, OR = 1.525, 95% CI: 1.158-2.007; FG0, OR = 1.475, 95% CI: 1.122-1.941, aGal/Gal ratio, OR = 1.628, 95% CI: 1.232-2.152) was increased. The relative abundance of fucosylated glycans (F, OR = 0.501, 95% CI: 0.373-0.674; FS, OR = 0.642, 95% CI: 0.485-0.849) was decreased, and the relative abundance of bisected GlcNAc glycans (B, OR = 1.480, 95% CI: 1.126-1.947; BN, OR = 1.461, 95% CI: 1.107-1.927; BS, OR = 1.700, 95% CI: 1.280-2.257) was increased (Figure 3).

After adjusting for confounding factors, 9 initial glycans (GP2, GP4, GP6, GP8, GP10, GP14, GP16, GP18, and GP23) showed a significant association with LN (Figure 2). An internal correlation analysis was conducted among these 9 initial glycans, by which significant correlations were observed (P < 0.05) (Figure S3). To address multicollinearity issues, a stepwise logistic regression model was used for variable selection, and three diagnostic models were established: 1) Model 1: GP8, GP10, and GP18 were significant and were selected to develop a identification model for LN; 2) Model 2: anemia and the WBC count, which showed significant differences, were included in the identification model; 3) Model 3: GP8, GP10, GP18, and anemia were included in the identification model (Table S5). We then employed a multiple-factor logistic regression to establish an LN diagnostic model based on the two indicators, uCRE and eGFR: 4) Model 4: GP8, GP10, GP18, uCRE and eGFR; and 5) Model 5: uCRE and eGFR (Table S6). ROC curves were plotted to evaluate the determination performance of the five models (Figures 4, S4). The discriminatory performance, sensitivity, and specificity of the five models in distinguishing between the case group and control group are shown in Table S7. The AUC value of Model 3 was 0.792 (95% CI: 0.727-0.858), which was significantly higher than those of Model 1 (AUC = 0.647, 95% CI: 0.568-0.726), Model 2 (AUC = 0.769, 95% CI: 0.702-0.837), Model 4 (AUC = 0.761, 95% CI: 0.691-0.830), Model 5 (AUC = 0.514, 95% CI: 0.430-0.597).