Low-grade Inflammatory Marker Profile May Help to Differentiate Patients With LADA, Classic Adult-Onset Type 1 Diabetes, and Type 2 Diabetes

In adults, autoimmune diabetes may manifest as the classic acute-onset type 1 diabetes phenotype or, alternatively, may initially produce a clinical form of slowly progressive insulin-dependent diabetes, widely known as latent autoimmune diabetes in adults (LADA) (1). Patients with LADA share some genetic, metabolic, clinical, and immunologic characteristics with both classic adult-onset type 1 and also type 2 diabetes (2,3). Since its clinical features significantly overlap with type 2 diabetes, LADA is often misdiagnosed when only based on the clinical phenotype (2,3). This, together with a lack of specific randomized clinical trials conducted in patients with LADA, translates into an often inappropriately treated condition (3).

As for immunological features, it is estimated that between 4% and 14% of patients initially categorized as having type 2 diabetes have circulating diabetes-associated autoantibodies (2), a hallmark of the adaptive immune response leading to pancreatic β-cell destruction. On the one hand, classic type 1 diabetes and LADA are indistinguishable regarding innate immunity (e.g., levels of immune mediators such as cytokines, chemokines, or adhesion molecules), which is involved in the acceleration of β-cell destruction (4–6). On the other hand, patients with type 2 diabetes have higher levels of proinflammatory cytokines and adhesion molecules (which are known to contribute to insulin resistance) than patients with LADA or classic type 1 diabetes (4–6). Therefore, systemic inflammation seems to be a common feature of all of these diabetes types, albeit to different extents, and a shared pathogenic mechanism has been suggested (6).

The innate immunity profile of patients with type 2 diabetes is associated mainly with the metabolic correlates of adiposity (7,8). Namely, obesity is associated with chronic low-grade inflammation as a result of chemotactic and stress signals sent by the altered adipocyte levels and increased levels of free fatty acids, which increase the numbers of cytokines such as tumor necrosis factor-α (TNF-α), interleukin (IL)-6, IL-1β, and interferon-γ (7,8). When prolonged, this obesity-induced chronic inflammation leads to the development of components of metabolic syndrome, insulin resistance, and diabetes (7,8).

Subjects with LADA span the spectrum from lean to obese, and the role of obesity and the degree of insulin resistance in these subjects is a matter of debate (9). In the Spanish population, patients with LADA have distinctive adiposity and metabolic profile characteristics compared with patients with classic adult-onset type 1 diabetes and type 2 diabetes (10). Therefore, we hypothesized that patients with LADA might also show a differential profile regarding low-grade inflammatory mediators according to specific adiposity- and metabolic-related traits. For this, we selected and measured biomarkers observed in the context of chronic inflammation, obesity, and also insulin resistance, as follows: IL-6 and CRP plasma levels, which are increased in metabolic syndrome (11); adiponectin, which has been demonstrated to be involved in the pathophysiology of type 2 diabetes, with prospective studies showing that higher levels are associated with a lower risk of type 2 diabetes (12); soluble TNF-α receptor 2 (sTNFRII) levels, which are correlated with insulin sensitivity and are proportional to the percentage of body fat, BMI, and triglyceride (TG) levels in healthy subjects; and the number of total leukocytes, which is increased in obesity-induced chronic inflammation (8). In addition, sTNFRII has been related recently to activated CD4 T regulatory cells (aTregs). Tregs are involved in the maintenance of immunological unresponsiveness to self-antigens and in the suppression of excessive immune responses deleterious to the host (13,14).

This study aimed to assess serum immune mediators in a cohort from a previously recruited sample that evaluated patients with LADA from our region regarding autoimmunity, insulin secretion, and clinical characteristics compared with patients with adult-onset type 1 diabetes and type 2 diabetes (10). Briefly, the original study included a total of 78 patients with type 1 diabetes and 480 with type 2 diabetes classified according to the American Diabetes Association standard criteria that were current at the time of the study (15). Moreover, LADA was diagnosed in 82 patients based on age (30–70 years of age at diagnosis), no insulin requirement during the initial 6 months of the disease course, and the presence of GAD autoantibodies (GAD65) or IA-2 antigen (IA-2A) autoantibodies. The original study collected clinical and demographic data including age, age at diagnosis, sex, diabetes duration, height, weight, BMI, waist-to-hip circumference ratio (WHR), systolic blood pressure (SBP), and diastolic blood pressure (DBP) and biochemical determinations including glycosylated hemoglobin (HbA_1c), fasting plasma glucose levels, lipid profile, creatinine levels, and absolute leukocyte count.

From the original study population, frozen fasting serum was available for the determination of inflammatory biomarkers from a total of 597 samples: 73 patients with type 1 diabetes, 442 patients with type 2 diabetes, and 82 patients with LADA. Peripheral blood samples were collected according to standard procedures after overnight fasting. After a maximum of 60 min at room temperature, the blood samples were centrifuged for 10 min at 1,500g to separate the serum, and the supernatant aliquots were stored at −80°C and were only thawed just prior to performing the assay. Autoantibodies against GAD65 and IA-2A were determined by ELISA (DRG Diagnostic, Marburg, Germany), as previously described (10). Optimal cutoff values were set at 5 and 15 units/mL for GAD65 and IA-2A, respectively (16,17). In the Diabetes Antibody Standardization Program (DASP) Workshop in 2009, the sensitivity and specificity scores for GAD65 were 82% and 95%, and for IA-2A were 60% and 100%, respectively.

Serum hs-CRP levels were measured with an immunoturbidimetric assay (Roche Diagnostics, Mannheim, Germany), and participants with levels >10 mg/mL were excluded because this usually indicates acute inflammation. Serum levels of adiponectin, sTNFRII, and IL-6 were measured using available sandwich ELISA kits following the manufacturer instructions. The adiponectin (Mediagnost, Reutlingen, Germany) assay range was 0.88–39.32 µg/mL with an intra-assay and interassay coefficient of variation of <10%. The sTNFRII (BioSource, Nivelles, Belgium) assay range was 0–20 ng/mL with an intra-assay and interassay coefficient of variation of <10%, and the IL-6 (Bender MedSystems, Vienna, Austria) assay range was 0–8.7 pg/mL with an intra-assay and interassay coefficient of variation of 3.4% and 5.2%, respectively.

The median and interquartile range values were computed for quantitative variables; for qualitative variables, absolute and relative frequencies were used. Differences between groups were tested by the Student t test, ANOVA, or the nonparametric Mann-Whitney or Kruskal-Wallis tests, where appropriate. Differences between groups in qualitative variables were assessed by the χ² test or Fisher exact test. To account for multiple testing, we used the Tukey correction. Correlations were assessed by the Spearman rank correlation coefficient.

Crude and adjusted odds ratios (ORs) with their 95% CIs were estimated by logistic regression. Multivariable logistic regression analyses were performed by the Enter method using covariables that were clinically or statistically associated. The variables assessed in the models with the type of diabetes included hs-CRP, cytokines (adiponectin, sTNFRII, and IL-6), and leukocyte count. Models were adjusted for sex, age, WHR, BMI, TGs, SBP, DBP, HDL cholesterol (HDL-c), LDL cholesterol (LDL-c), HbA_1c, and diabetes duration. Goodness-of-fit logistic model assumptions were evaluated using the Hosmer-Lemeshow test and Shapiro-Wilk test for normality. Logistic regression analyses were conducted to evaluate the frequency of LADA versus classic adult-onset type 1 diabetes and type 2 diabetes and of autoimmune diabetes versus type 2 diabetes using the explanatory variables as predictors. The diagnostic accuracy of the logistic regression models was checked by receiver-operating characteristic (ROC) curves and the area under the ROC curve (AUC_ROC). Statistical significance was established at a P value of <0.05. Data management and analyses were performed with the free software environment R, version 3.3.1 (18), and SPSS Statistics software (version 20; SPSS, Chicago, IL).

The clinical characteristics of the three study groups (75 patients with LADA, 67 patients with classic adult-onset type 1 diabetes, and 390 patients with type 2 diabetes) are shown in Supplementary Table 1. There were significant differences among the three types of diabetes for all studied variables, except for sex distribution and total cholesterol, and this is in agreement with the results from the original study (10).

There were significant differences in the median serum concentrations of inflammatory markers among all three diabetes types except for IL-6. The leukocyte count and sTNFRII values gradually increased from classic adult-onset type 1 diabetes to LADA and from LADA to type 2 diabetes (i.e., 6,230 leukocytes in type 1 diabetes, 6,310 leukocytes in LADA, and 6,715 leukocytes in type 2 diabetes; P = 0.005), and for sTNFRII (2.06 ng/mL in type 1 diabetes, 2.73 ng/mL in LADA, and 3.05 ng/mL in type 2 diabetes; P < 0.001) (Fig. 1 and Supplementary Table 1). The levels of adiponectin followed exactly the opposite pattern and were highest in patients with classic adult-onset type 1 diabetes (8.30 vs. 6.22 µg/mL in LADA and 4.82 µg/mL in type 2 diabetes; P < 0.001) (Fig. 1 and Supplementary Table 1). Between groups, we observed that patients with LADA differed from patients with type 1 diabetes only in that they had significantly higher levels of sTNFRII (P = 0.001). In addition, they differed from patients with type 2 diabetes in that they had higher levels of adiponectin (P = 0.03) but lower levels of hs-CRP (1.28 vs. 2.03 mg/L; P = 0.022) and total leukocytes. The comparisons between patients with autoimmune diabetes (LADA plus type 1 diabetes) and patients with type 2 diabetes are shown in Supplementary Table 2.

When considering all three diabetes types together, the adiposity markers BMI, WHR, and TG were negatively correlated with adiponectin (P < 0.01), but were positively correlated with hs-CRP, sTNFRII, and total leukocytes (P values between <0.05 and <0.001) (Fig. 2). The opposite pattern was observed for HDL-c, which was positively correlated with adiponectin (P < 0.001) but was negatively correlated with hs-CRP, sTNFRII, and total leukocytes (P values between <0.05 and <0.001). Moreover, adiponectin and sTNFRII were positively correlated with age (P < 0.001), and IL-6 and hs-CRP were positively correlated with HbA_1c levels (P < 0.05). Finally, SBP was positively correlated only with hs-CRP (P < 0.01). Correlations in the autoimmune subgroup (classic adult-onset type 1 diabetes and LADA) were similar to those of the three diabetes types together (Supplementary Fig. 1).

The multiple logistic model of LADA versus classic adult-onset type 1 diabetes showed that increased odds of LADA were associated with older age (OR 1.05; P = 0.018) and increasing sTNFRII levels (OR 1.9; P = 0.047), and decreased odds were associated with lower HDL-c levels (OR 0.14; P = 0.001) and hs-CRP levels (OR 0.78; P = 0.019) (Fig. 3A). Conversely, older age, high BMI, and a higher number of total leukocytes lowered the risk of LADA compared with type 2 diabetes (OR 0.98, P = 0.036; OR 0.89, P = 0.001; and OR 0.98, P = 0.036, respectively), whereas high HDL-c levels increased the risk (OR 2.41, P = 0.003, respectively) (Fig. 3B). The model comparing autoimmune diabetes with type 2 diabetes is shown in Supplementary Fig. 2.

Diagnostic models were constructed with the identified candidate parameters to explain the different types of diabetes. The LADA versus classic adult-onset type 1 diabetes model explained 35% of the variation in the dependent variable and had a good discriminatory ability (AUC_ROC 0.83; P < 0.001) (Fig. 3C). The discriminative power for single explanatory biomarkers was 0.62 (P = 0.06) and 0.71 (P = 0.001) for hs-CRP and sTNFRII, respectively. The LADA versus type 2 diabetes model explained 15% of the variation in the dependent variable, with an AUC_ROC of 0.73 (P < 0.001) (Fig. 3D), and the AUC_ROC for the single biomarker leukocyte count was 0.60 (P = 0.007). The model of autoimmune diabetes is shown in Supplementary Fig. 2.

Our study included a panel of markers that, altogether, have been specifically implicated in the obesity-associated chronic inflammatory state. Moreover, we performed analyses to include three types of diabetes and metabolic variables not studied at large, and all of it within a single European population, which differs from Asian populations regarding the prevalence of LADA and metabolic syndrome. Our observations showed that patients with LADA have clinical features at the crossroads among diabetes types, in accordance with previous reports from the European Action LADA cohorts (19–21) and also from studies conducted in Italy (22), Finland (23), and China (24) and in our setting (10).

A previous report from the European Action LADA cohort showed that concentrations of IL-6 and TNF-α were positively correlated with BMI and waist circumference in patients with type 1 and type 2 diabetes and also in patients with LADA (5). We corroborate these results and add to the literature that, for all studied diabetes types together, adiposity-related markers (BMI, WHR, and TG) were positively correlated with levels of hs-CRP and sTNFRII and leukocyte count but were negatively correlated with adiponectin. Conversely, HDL-c followed the opposite pattern and was positively correlated with adiponectin but negatively correlated with all studied markers including IL-6.

Multiple logistic models indicated that, even after stepwise adjustment for multiple variables, increased sTNFRII levels persisted as a predictor for a higher risk of LADA than for classic adult-onset type 1 diabetes. This is of particular interest because TNFRII is the primary receptor for CD4 Tregs and the most potent for triggering the differentiation of functionally resting Tregs to aTregs (13,14). aTregs are indispensable in maintaining immunological unresponsiveness to self-antigens and in suppressing excessive immune responses deleterious to the host (25,26). A defect in Treg activation in type 1 diabetes has been demonstrated (27), and it was reported that lower numbers of aTregs were associated with a trend for less residual C-peptide secretion and poorer glycemic control (28). It is tempting to speculate that this activation defect (reduced number of aTregs) could be poorer in patients with classic type 1 diabetes than in patients with LADA and therefore linked to a more severe clinical course requiring an earlier start of insulin treatment.

There were significant differences in the median serum systemic concentrations of all inflammatory markers among the three diabetes types except for IL-6. This is in contradiction with other studies showing that median IL-6 serum concentrations were significantly higher in patients with type 2 diabetes than in patients with type 1 diabetes, patients with LADA, and healthy participants (5,29). This could be attributed to this cytokine circulating at very low concentrations, with a significant diurnal variability and a short half-life of between 2 and 4 h (30). C-reactive protein has a longer half-life (19–20 h) than IL-6 or TNF-α, with no diurnal variations and greater stability over prolonged periods (30), and that is why hs-CRP is considered a more convenient measure of systemic inflammation compared with assays of other circulating cytokines (31).

In our study, the leukocyte count and sTNFRII values gradually increased from classic adult-onset type 1 diabetes to LADA and from LADA to type 2 diabetes, whereas adiponectin followed the opposite direction. These patterns have been previously described in Action LADA populations for TNF-α (5) and in a Chinese population for adiponectin (29). However, the between-group comparisons in these studies differ from our results: in the Action LADA study (5), TNF-α levels were significantly lower in patients with LADA than in those with type 2 diabetes but were similar to those in patients with type 1 diabetes. These differences may be attributed to the fact that the European study measured serum TNF-α levels, whereas we measured its soluble receptor type 2, which is a more stable protein that modulates the activation of TNF-α in adipocytes, which is associated with obesity and is a predictor of insulin resistance (32). Regarding adiponectin, a recent Chinese report (29) found that levels were increased in patients with classic type 1 diabetes and LADA, but not in patients with type 2 diabetes. These differences may be attributed to a leaner body shape in patients with LADA in China compared with other ethnic groups and to the fact that the frequency of metabolic syndrome in Chinese patients with LADA is similar to that in patients with type 2 diabetes (29), whereas in our region the frequency of metabolic syndrome was higher in patients with LADA than in those with classic type 1 diabetes and was lower than in patients with type 2 diabetes (10). Moreover, the levels of hs-CRP among patients with LADA in our study were lower than in patients with classic adult-onset type 1 diabetes and much lower than in patients with type 2 diabetes. These results are in agreement with those of two other reports, one conducted in China and the other in Bulgaria (33,34) regarding in type 2 diabetes, but in contradiction with the observation for classic type 1 diabetes in the Chinese patients (33). Further studies with larger samples of patients with LADA are warranted in order to elucidate the role of this inflammatory marker in the pathogenesis of the disease.

Finally, the results of the current study suggest that low-grade inflammatory markers independently contributing to the risk of a specific diabetes type were not optimal as single predictors for discriminating patients with or without LADA. This could be explained by the small sample size or, alternatively, be because these biomarkers might be involved in more than one specific pathophysiological pathway. However, they provided added value when combined with other markers and established conventional risk factors (mainly age, HDL-c level, and BMI). In particular, sTNFRII was useful to distinguish LADA from classic adult-onset type 1 diabetes. Taken together, our data show that sTNFRII could be a novel, clinically relevant biomarker for potential phenotyping. Additionally, the multiple biomarker models that we applied had good discriminatory ability (i.e., >0.8), hence potentially representing objective and useful tools to identify LADA.

Although the current study has yielded some preliminary findings, it has several limitations. The first limitation concerns the cross-sectional design, which prevents establishing the direction of causality (e.g., metabolic syndrome and altered levels of immune mediators as a cause of or response to low-grade chronic inflammation). In addition, the cross-sectional design only reflected the current status of diabetes-associated autoantibodies, and persistence and confirmation could not be determined at follow-up. Although the specificity for both anti-GAD65 and anti-IA2 ELISA kits was high (95% and 100%, respectively), we cannot rule out the presence of cases of false-positive results within the subgroup with type 2 diabetes, which could have resulted in shared inflammatory status between diabetes groups. In addition, the number of patients with autoimmune diabetes was much lower than the number of patients with type 2 diabetes, which might have lowered the statistical power, either overestimating the effect size or reducing the chance of detecting a true effect. Moreover, we did not include a control group, and we cannot reject the hypothesis that some of the changes that we observed in the concentrations of the immune mediator could also be present in other populations, such as lean nondiabetic offspring of subjects with type 2 diabetes or young obese subjects with normal and impaired glucose tolerance. Finally, we did not include factors known to alter the risk for some components of metabolic syndrome that could alter circulating levels of inflammatory biomarkers, such as dietary intake, modifiable lifestyle factors (e.g., physical activity), and the presence of diabetes-related complications or comorbid conditions (e.g., cardiovascular disease, nonalcoholic fatty liver, or polycystic ovary syndrome). Therefore, replication of results in an independent data set is warranted to support both the internal validity and the generalizability of the present findings.

In summary, our study provided evidence of significant differences in median serum concentrations of circulating adipose-related inflammatory biomarkers across the main diabetes types. Most importantly, we provided preliminary evidence that low-grade inflammatory biomarkers may be useful predictors in the context of multiple marker models, including conventional risk factors, and provided good discriminative values in the diagnosis of the different diabetes types. If the performance of these models is externally validated in independent populations, they may have potential clinical utility to identify individuals with higher odds of having LADA than classic adult-onset type 1 diabetes or type 2 diabetes, along with islet autoantibody positivity.

Acknowledgments. The authors thank the patients and IRBLleida (B.0000682) Biobank, integrated in the Spanish National Biobanks Network of the Carlos III National Institute of Health PT13/0010/0014, for their collaboration. The authors also thank Amanda Prowse for editorial support (Lochside Medical Communications Limited, Glasgow, U.K.).

This work has been carried out within the framework of the Doctorate in Medicine of the Autonomous University of Barcelona.

Funding. This research was supported by grants from the Spanish Ministry of Health, the Carlos III National Institute of Health (ISCIII) (grants PI12/0183 and PI15/0625), and European Regional Development Fund. CIBERDEM and the Centre for Biomedical Research on Physiopathology of Obesity and Nutrition are initiatives of the ISCIII, Spain.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. E.C. contributed to the study design, conduct of the study, data analysis, and writing of the manuscript. M.H. and D.M. contributed to the study design and coordination, conduct of the study, data analysis, and writing of the manuscript. A.C. and J.R. contributed to the statistical analysis, data interpretation, and writing of the manuscript. M.G., J.F.-N., and J.-M.F.-R. contributed to data interpretation and the writing of the manuscript. A.E., A.M., and A.R.-M. contributed to data collection and the conduct of the study. All authors critically reviewed the manuscript and approved the final version for publication. D.M. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in oral form at the XXVIII Congress of the Spanish Society of Diabetes 2017, Barcelona, Spain, 5–7 April 2017, and as a poster presentation at the 53rd Annual Meeting of the European Association for the Study of Diabetes, Lisbon, Portugal, 11–15 September 2017.