Tetraspanin 7 autoantibodies predict progressive decline of beta cell function in individuals with LADA

Individuals with LADA (n = 173), type 2 diabetes (n = 204) and type 1 diabetes (n = 158) and healthy control individuals (n = 170) were enrolled from the Second Xiangya Hospital of Central South University (Changsha, Hunan, China) from May 2014 to August 2017 for the cross-sectional study. Another 157 individuals with type 1 diabetes (duration >1 year) were recruited to examine the change in the time course of TSPAN7A, GADA, ZnT8A and IA-2A prevalence. Another 53 individuals with LADA were enrolled from the Second Xiangya Hospital of Central South University from October 2010 to December 2013 for a follow-up study in which clinical characteristics were determined annually for 3 years. Sera from another 199 healthy participants were collected from the Second Xiangya Hospital of Central South University to establish the threshold of positivity for TSPAN7A. The criteria of the WHO (version 1999) were applied to the diagnosis of diabetes [11]: fasting plasma glucose (FPG) ≥7 mmol/l or 2 h plasma glucose during 75 g OGTT ≥11.1 mmol/l with at least one obvious diabetic symptom (thirst, polyuria or unexplained weight loss). For asymptomatic individuals, a confirmatory test was conducted on another day. For children, the oral glucose load was based on body weight (1.75 g/kg). Individuals with LADA were recruited based on the criteria used in previous studies [2, 12]: (1) positive for at least one autoantibody (GADA, IA-2A or ZnT8A); (2) age ≥30 years at onset of diabetes and (3) at least 6 months of therapy without insulin. Type 1 diabetes was diagnosed according to the following criteria: acute-onset ketosis or ketoacidosis; reliance on insulin therapy; impaired islet function or islet autoantibody (GADA, IA-2A or ZnT8A) positivity. The individuals with type 1 diabetes presented with diabetes within 1 year. All individuals with type 2 diabetes were negative for GADA. All healthy participants underwent a standard 75 g OGTT to confirm that their blood glucose was normal. The exclusion criteria for healthy participants were: (1) other autoimmune diseases; (2) infectious disease; (3) pregnancy; (4) history of immunosuppressive drugs or steroids >7 days and (5) malignant disease. This study was approved by the ethics committee of the Second Xiangya Hospital of Central South University and all participants or their parents gave written informed consent.

Physical characteristics (sex, age, body height and weight) were recorded by professional researchers. Fasting blood was used for measurement of FPG, HbA_1c and fasting C-peptide (FCP). Postprandial blood samples were used to test 2 h postprandial plasma glucose (PPG) and 2 h postprandial C-peptide (PCP).

FCP and PCP were measured by a chemiluminescence method using the Adiva Centaur Systema kit (Siemens, Munich, Germany). HbA_1c was detected by automated liquid chromatography (VARIANT II Hemoglobin Testing System; Bio-Rad Laboratories, Hercules, CA, USA).

Islet autoantibodies, including GADA, IA-2A and ZnT8A, were detected by radioligand assays in duplicate [13]. Based on the 99th percentile observed in 405 healthy participants, the cut-off values of positivity for GADA and IA-2A were 18 U/ml and 3.3 U/ml of WHO units, respectively. The threshold of antibody index for ZnT8A was 0.011 [14]. For antibody-positive individuals, blood was drawn again within 7 days to exclude false-positive results. In the 2016 islet autoantibody standardization programme (IASP 2016), the sensitivity and specificity in our laboratory were, respectively, 82% and 97.8% for GADA, 76% and 100% for IA-2A and 72% and 100% for ZnT8A.

Luciferase-tagged full-length TSPAN7 was cloned into the pREN-2 vector, which was kindly provided by P. D. Burbelo (National Institute of Dental and Craniofacial Research, National Institutes of Health, MD, USA) [15]. The untagged TSPAN7 expression vector was purchased from Origene (Rockville, MD, USA). Plasmids with full-length TSPAN7 or untagged TSPAN7 were transfected into 293T cells cultured in DMEM supplemented with 10% FBS and 100 U/ml penicillin–streptomycin (Gibco, Carlsbad, CA, USA). After 48 h, cells were lysed in Glo Lysis Buffer (Promega, Madison, WI, USA) for 15 min at 4°C followed by centrifugation (12,000 g at 4°C for 15 min). The supernatant fraction was collected for further testing.

TSPAN7A were assayed in duplicate by a luciferase immunoprecipitation assay (LIPS) as described previously [6, 16] with minor modifications. Briefly, 5 μl serum was incubated at 4°C overnight with 25 μl assay buffer (0.05% Triton X100 and 0.1% Tween 20 in TBS, pH 7.2) containing 5 × 10⁶ counts/s of luciferase-tagged TSPAN7 measured in a MicroBeta LumiJET (PerkinElmer, Waltham, MA, USA). The mixture was then transferred in to a 96-well microfiltration plate (Merck Millipore, Darmstadt, Germany). Then, 15 μl Protein A-agarose (Invitrogen, Carlsbad, CA, USA) and 5 μl assay buffer containing 0.1% IgG-free BSA (Yeasen, Shanghai, China) were added to each well, and the plates were shaken (300 rev/min) at 4°C for 60 min. Plates were washed ten times using 200 μl assay buffer and Renilla luciferase substrate (Promega) was added. For potential positive serum, a competition assay was performed to confirm the binding specificity of the serum antibodies to TSPAN7. Briefly, sera were simultaneously incubated with or without luciferase-free TSPAN7.

The final results were expressed as TSPAN7A index (sample CPS – negative standard CPS) / (positive standard CPS – negative standard CPS), where CPS is counts/s. Serum that was negative for TSPAN7A was obtained from one healthy individual and serum that was positive for TSPAN7A was obtained from one individual newly diagnosed with type 1 diabetes. Based on the 99th percentile of 199 healthy participants, the cut-off value of TSPAN7A was 0.024. Intra-assay and inter-assay CV was in the range 6.3–12.4% and 14.9–35.7%, respectively.

Statistical analysis was performed with SPSS software version 19 (IBM, New York, NY, USA). The data are presented as the mean ± SD or medians (25th–75th percentile). A normality test was performed before the data analysis. For non-normal data, logarithmic transformation was conducted. ANOVA was used to compare normal variables among groups. Non-parametric test was performed by Mann–Whitney assay. A χ² test was used to compare categorical variables. A McNemar test was used to determine whether TSPAN7A could increase the fraction of multiple-antibody-positive individuals. Binary logistic regression was performed to investigate possible risk factors for worsening islet function. A generalised linear mixed model was performed to compare the repeated measures data. There was no imputation on missing values. A p value <0.05 was considered statistically significant.

All previous studies on TSPAN7A have been conducted in white people. However, whether TSPAN7A are valid islet autoantibodies for East Asian populations remains unknown. Therefore, we first evaluated the frequency of TSPAN7A in individuals with LADA, type 1 diabetes and type 2 diabetes and in healthy control individuals. Potential positivity of sera for TSPAN7A was further confirmed in a competitive binding assay (electronic supplementary material [ESM] Fig. 1). Our results showed that the prevalence of TSPAN7A in individuals with LADA, type 1 diabetes and type 2 diabetes and healthy control individuals was 21.4% (37/173), 26% (41/158), 0.5% (1/204) and 1.2% (2/170), respectively (Table 1). The prevalence of TSPAN7A in type 1 diabetes (26%) is comparable with the data reported by Walther et al (27%; p > 0.05) using the same assay [6]. In individuals with LADA, the prevalence of GADA, IA-2A, ZnT8A and TSPAN7A was 95.4%, 17.9%, 16.8% and 21.4%, respectively (Table 1 and Fig. 1a). In individuals with type 1 diabetes, the frequency of GADA, IA-2A, ZnT8A and TSPAN7A was 69%, 43%, 31% and 26%, respectively (Table 1 and Fig. 1b). As for GADA, IA-2A and ZnT8A, positivity for TSPAN7A significantly decreased during disease progression (Fig. 1c). Comparing the prevalence of TSPAN7A in the presence or absence of other autoantibodies showed that there was a significant association of TSPAN7A with IA-2A and ZnT8A in individuals with either LADA or type 1 diabetes (all p < 0.01); there was no association between TSPAN7A and GADA (ESM Fig. 2). All the data strongly suggest that TSPAN7A are valid islet autoantibodies for East Asian populations and might be valuable in the diagnosis of autoimmune diabetes.

We next evaluated whether TSPAN7A could improve the diagnosis of autoimmune diabetes in Chinese individuals with type 1 diabetes. Among 29 individuals with type 1 diabetes who were negative for GADA, IA2A and ZnT8A, four (13.8%) were positive for TSPAN7A. The addition of measurement of TSPAN7A slightly increased the identification of beta cell autoimmunity from 81.6% to 84.2% (p > 0.05) (Fig. 1b). Adding measurement of TSPANA7A also slightly increased the number of individuals with type 1 diabetes found to have multiple-antibody positivity, from 43.7% (69/158) to 46.2% (73/158), an increase of 5.8% (p > 0.05) (Fig. 1b,d). These results were consistent with previous reports showing that TSPAN7A provided minor utility in the diagnosis of type 1 diabetes [6]. In contrast, additional measurement of TSPAN7A significantly increased the percentage of individuals with LADA found to have multiple-antibody positivity, from 22% (38/173) to 32.4% (56/173) (an increase of 47.3%, p < 0.001) (Fig. 1a,d).

Previous studies have suggested that the presence of multiple islet autoantibodies is associated with rapid loss of beta cell function [17]. Therefore, we next investigated whether TSPAN7A were associated with low C-peptide levels in individuals with LADA in the cross-sectional study. Compared with individuals who had high C-peptide levels, those with poorer beta cell function had a lower BMI (p < 0.001), a higher GADA titre (p < 0.01) and a higher frequency of IA-2A (p < 0.05) and positivity for TSPAN7A (p < 0.01) (Table 2). For further analysis, sex, age at onset, duration, BMI, GADA titre, IA-2A, ZnT8A and TSPAN7A were evaluated in the logistic regression. Analysis of logistic regression demonstrated that duration of disease (OR 1.83, p = 0.019), GADA titre (OR 2.67, p = 0.009) and positivity for TSPAN7A (OR 2.87, p = 0.034) were risk factors for beta cell function, while BMI was a protective factor (OR 0.34, p = 0.001) (Table 3). We further compared C-peptide levels between single-antibody individuals who were positive and negative for TSPAN7A. Both FCP and PCP levels were decreased in individuals with LADA who were positive for TSPAN7A (p < 0.05; Fig. 2).

To further confirm that TSPAN7A were risk factors for beta cell function, we followed up 41 TSPAN7A-negative and 12 TSPAN7A-positive individuals with LADA at baseline, 1, 2 and 3 years. The clinical characteristics, including sex, age, duration, BMI, HbA_1c, GADA titre, FPG, PPG, FCP or PCP were comparable at baseline (ESM Table 1). However, after a 3-year follow-up in individuals with LADA, both FCP and PCP decreased significantly in TSPAN7A-positive, compared with TSPAN7A-negative individuals with LADA (Fig. 3). The median annual decrease in rates of FCP and PCP in individuals with LADA who were positive for TSPAN7A was much higher than in individuals who were negative for TSPAN7A (34.6% vs 7.9%, p = 0.043; 33.2% vs 11%, p = 0.041, respectively).