Altered glycosylation profiles of serum IgG in Takayasu arteritis

All serum samples in this study were collected at Peking Union Medical College Hospital (PUMCH). In total, 392 serum samples collected from 164 patients with TAK, 128 patients with peripheral arterial disease (PAD), and 100 healthy controls (HCs) were used for lectin microarray analysis. The inclusion criteria required every participant in this study to be older than 18 years on the sampling points. Patients with other autoimmune diseases, active infections, pregnancies, cancers, or any severe comorbidities were excluded. For TAK group, patients who met the 1990 American College of Rheumatology classification criteria for TAK were included [16], and patients with incomplete laboratory data on the sampling points were excluded. Patients with TAK were further divided into inactive (n = 82) and active (n = 82) groups with matched age and gender, according to the 1994 National Institutes of Health criteria [17]. Since PAD is also a chronic disease characterized by stenosis or occlusion of arteries but with an atherosclerotic etiology other than vasculitis [18], PAD patients were taken as disease controls (DCs), which includes 75 patients with carotid artery stenosis (CAS) and 53 with chronic limb-threatening ischemia (CLTI). Patients in DC group should meet the 2017 European Society of Cardiology diagnostic criteria for PAD [19], and patients with arterial stenosis/occlusion secondary to non-atherosclerotic conditions (i.e., Buerger's disease) were excluded. For HCs, healthy subjects from the health examination center matched for age and gender to DC group were included.

Specific glycosylation changes were validated by lectin blot in a smaller cohort consisting of 48 patients with TAK (24 inactive and 24 active), 16 DCs (8 CAS and 8 CLTI), and 16 HCs, of which some patients were from the previous study cohort (10 inactive TAK, 10 active TAK, 6 DCs), and the others came from a new set of patients. To reduce the bias caused by age differences in the previous cohort, 16 new HCs were included in the validation cohort to match to the TAK group. The gender distribution among the three groups in the validation cohort was also made consistent. Serum was separated from the whole blood specimen and stored at − 80 °C until use. This study was approved by the Institutional Review Board of PUMCH, Beijing, China (JS-2038). Written informed consent was obtained from each participant and the study was conducted following the Declaration of Helsinki.

To detect the glycosylation patterns of serum IgG, a commercial lectin microarray chip (BCBIO Biotech, Guangzhou, China) with triplicate spots of 56 types of lectins was used (see Additional file 1: Fig. S1 for the layout of the lectin microarray). The 56 types of lectins were known to bind specifically to common glycan structures in mammals (see Additional file 3: Table S1 for the list of lectins). Lectin microarray chips at -80 °C storage were first equilibrated at 4 °C and warmed at room temperature for 30 min before use, and then the chips were incubated in blocking buffer (1% BSA in PBS) for 2 h. After washing with PBS, the microarray chips were dried by centrifugation. The serum from each sample was diluted 1:200 in 0.05% Tween 20 (PBST) and 200 µl of each diluted sample were applied to one well on the chips, and then the chips were incubated at 4 °C overnight. Afterward, the chips were washed with PBST five times and incubated with 3 ml Alexa Fluor 647-labeled goat anti-human IgG antibody (1:1,000; Jackson ImmunoResearch) for 45 min at room temperature protected from light. After washing thoroughly with PBST and ddH2O, the chips were dried by centrifugation and scanned with a GenePix 4000B Microarray Scanner (Molecular Devices, Sunnyvale, CA) at a wavelength of 647 nm. The photomultiplier tube gain was set to 600.

The fluorescent images were converted to a digital format for subsequent analysis by extracting the median foreground and background intensity values of each spot on the chips using the GenePix Pro 6.0 software (Molecular Devices, Sunnyvale, CA). The signal-to-noise ratio (S/N) of each lectin spot, defined as the ratio of the median signal intensity of the foreground to the background, was calculated. To minimize the inter-array bias, we used the method of Silver et al. [20] to normalize the S/N data. Significant differences in lectin-binding levels among groups were determined according to the methods described by Hu et al. [21]. A significant difference was considered to be present when the following two conditions were satisfied: (a) the fold change (FC) between two groups [group1 (S/N)/group2 (S/N)] > 1.3 or < 0.77; (b) P-value < 0.05.

Since the glycans of IgG known to alter the conformation and function are attached to the two heavy chains [22], the IgG heavy chains (50 kDa) are chosen to verify the microarray results in lectin blot [23]. First, serum proteins at 1:100 dilution were separated by 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and transferred onto polyvinylidene fluoride membranes (Millipore, Billerica, MA). After being blocked with blocking buffer (QuickBlock™ Blocking Buffer for Western Blot, Beyotime Biotechnology, Nanjing, China) at room temperature for 30 min, the membranes were incubated with 20 μg/ml of Cy3 (GE Healthcare)-labeled glycine max lectin (SBA; Vector Laboratories Inc., US), Concanavalin A lectin (ConA; Vector Laboratories Inc., US), and Morniga M lectin (MNA-M; EY Laboratories, Inc., US) overnight at 4 °C protected from light. Excess lectins were washed away with TBST. Finally, the membranes were visualized by a fluorescence signal system of Typhoon FLA 9500 (GE Healthcare). Serum proteins from the same healthy subject were used as a reference. In addition, to confirm the location of IgG in immunoblotting and exclude the influence of different IgG concentrations on the measurements of IgG glycosylation, Western blotting of serum IgG was performed. After blocking unspecific binding sites, the membranes were incubated with goat anti-human IgG (H + L) -HRP conjugated antibody (1:15,000; EASYBIO, China) at room temperature for 1 h. Protein bands were visualized by ImageQuant software (GE Healthcare). The signal intensity was quantified for further analysis by ImageJ software (National Institutes of Health, Bethesda, MD, USA). The intensity ratios (lectin/IgG) were calculated.

The basic characteristics of 392 participants included in the lectin microarray analysis are summarized in Table 1. As shown in the table, no significant difference was found in age or gender distribution between the inactive and active TAK groups. However, there were differences in age and gender distribution between the TAK group and the other two groups. The relatively late onset of PAD probably explains the disparity in age distribution. In addition, the levels of all measured inflammatory indicators including hypersensitive C-reactive protein (hsCRP), erythrocyte sedimentation rate (ESR), interleukin-6 (IL-6), and tumor necrosis factor-alpha (TNF-α) were significantly higher in the active TAK group than in the inactive TAK group (p < 0.01). No significant difference was observed in disease duration, ongoing glucocorticoids, and ongoing disease-modifying antirheumatic drugs (DMARDs) and/or biological DMARDs (bDMARDs) between the inactive and active TAK groups. In the lectin blot analysis, the mean age was 31.75 ± 8.43 years for patients with TAK and 33.38 ± 5.66 years for HCs.

To investigate the glycosylation profiles of serum IgG in TAK, lectin microarrays were used, which enabled a high-throughput and ultrasensitive detection of the glycoforms. Serum samples from 164 patients with TAK, 128 DCs, and 100 HCs were tested. After normalizing the signal values, only one lectin showed a significant difference among the three groups. As shown in Fig. 1, there is a significantly higher signal intensity of SBA in the TAK group compared to DCs and HCs (FC = 1.97, p < 0.0001; FC = 1.77, p < 0.0001, respectively), indicating increased glycan levels of N-Acetylgalactosamine (GalNAc) of serum IgG in patients with TAK. No significant difference was found between DCs and HCs (p > 0.99). In addition, decreased mannose levels, as implicated by less binding to ConA (FC = 0.69, p = 0.0002) and MNA-M (FC = 0.69, p = 0.0005), were found in patients with active TAK relative to patients with inactive disease.

As shown in Fig. 2, the intensity ratios (SBA/IgG) in the TAK group were significantly higher than that in DCs and HCs (p = 0.0493 and p < 0.0001, respectively) and no significant difference was found between DCs and HCs (p = 0.1082), which were consistent with the results of the lectin microarray analysis. For TAK subgroups, MNA-M/IgG was significantly lower in the active TAK group than in the inactive TAK group (p = 0.0319). However, no significant difference was observed in ConA/IgG between the inactive and active TAK groups (p = 0.0685), although a similar decreasing trend was shown. Overall, these results confirmed the reliability of the lectin microarray test. Furthermore, correlation analysis showed a significant negative correlation between the signal intensities of MNA-M, as well as ConA, detected by lectin microarrays and hsCRP, ESR, and IL-6 (MNA-M: r = − 0.230, p = 0.003 for hsCRP, r = − 0.221, p = 0.005 for ESR, and r = − 0.184, p = 0.020 for IL-6, respectively; ConA: r = − 0.192, p = 0.014 for hsCRP, r = − 0.155, p = 0.047 for ESR, and r = − 0.242, p = 0.002 for IL-6, respectively). For TNF-α, the correlation was also negative (r = − 0.155 for MNA-M and r = − 0.149 for ConA, respectively), but not significant (p = 0.051 for MNA-M and p = 0.060 for ConA). These indicated a potential relationship between the mannose level of serum IgG and these inflammatory indicators. In addition, the MNA-M signal intensities between patients with and without ongoing glucocorticoids, and DMARD and/or bDMARDs in TAK group, were compared. No significant difference was found in the subgroup analysis (Additional file 2: Fig. S2).

The significant difference in SBA binding levels between patients with TAK and PAD prompted us to further explore the capability of SBA binding levels in the lectin microarray analysis as a potential biomarker to assist the differential diagnosis between TAK and PAD. Thus, ROC curve analysis was performed. As shown in Fig. 3, the AUC was 0.749 (95% CI: 0.693–0.806, p < 0.001), with a sensitivity of 73.8% and a specificity of 71.1% at the optimal cutoff value. These results suggested that the change of SBA affinity for serum IgG might be a promising serological indicator to assist the discrimination between TAK and PAD.