figure a

Introduction

The major causes of morbidity and mortality in individuals with type 2 diabetes are cardiovascular (CV) complications [1]. Type 2 diabetes is associated with a twofold increased risk of death and a threefold increased risk of coronary artery disease (CAD) [2]. CV disease risk stratification in those with type 2 diabetes is typically based on the evaluation of classical CV risk factors and on diabetes-related variables [3, 4]. Moreover, single biomarkers associated with atherosclerosis (e.g. osteopontin and galactin-3 [5, 6]), inflammation (e.g. TNF receptor 1 [TNFR1] and angiopoietin-like 2 [7, 8]) and congestive heart failure (e.g. N-terminal pro-B-type natriuretic peptide [NT-proBNP] [9, 10]), or a combination of biomarkers [11,12,13] have been proposed to improve risk stratification in individuals with type 2 diabetes. Although promising data have been obtained using these biomarkers, up until now, none of them have been incorporated into international guidelines [1]. Thus, the research for new predictive biomarkers of CV disease in type 2 diabetes remains highly relevant.

Tenascin-C (TN-C) is a large hexameric extracellular matrix (ECM) glycoprotein expressed in most normal adult tissues [14, 15]. It is implicated in the regulation of cell behaviour and tissue remodelling in many developmental and pathological processes [14, 16]. At sites of tissue injury and inflammation, TN-C is markedly upregulated [17, 18]: it is transiently upregulated during acute inflammation, such as after acute myocardial infarction (MI), during which serum TN-C levels rise within hours and peak after 5 days, and may be used to predict long-term CV events [19, 20]. In chronic inflammation and during tissue repair, TN-C levels are continuously upregulated [16]. Beyond acute coronary events, TN-C is an emerging candidate biomarker for diagnosis and/or risk stratification in various CV diseases, such as aortic aneurysm and dissection [21], dilated cardiomyopathy [22] and atherosclerosis [23]. TN-C is consequently considered as a marker of inflammation in CV diseases.

In individuals with diabetes, a significant association between increased TN-C levels and the severity of proliferative diabetic retinopathy has been reported [24,25,26]. There are no clear data in the literature examining the prognostic usefulness of serum TN-C levels for CV outcomes specifically involving type 2 diabetes. Recently, a big-data approach in >8400 individuals with dysglycaemia identified TN-C, among other biomarkers, as an independent determinant of CV outcomes or death [11, 12]. In individuals with chronic kidney disease (CKD; ~40% with type 2 diabetes), higher serum TN-C levels were significantly associated with CV-associated morbidity and/or death [27, 28]. The association between TN-C and CV events in a population of purely diabetic individuals remains to be demonstrated.

In the present study, we examined the association of TN-C with all-cause death and CV-associated morbidity in a prospective cohort of individuals with type 2 diabetes [7, 8, 29]. We also assessed the incremental prognostic value of serum TN-C concentration in addition to established CV risk markers for long-term risk assessment of individuals with type 2 diabetes.

Methods

Participants from the SURDIAGENE (SUivi Rénal, DIAbète de type 2 et GENEtique) cohort were recruited at the University Hospital of Poitiers, France, from 2002 to 2012 [29]. The local ethics committee (CPP Ouest III) approved the study design and written informed consent was obtained from all participants. Participants were prospectively followed-up until death, or until 31 December 2015, whichever came first.

Participants with a baseline eGFR < 30 ml min−1 [1.73 m]−2 and/or a history of prior renal replacement therapy and/or a follow-up duration of <3 months were excluded from the present analysis. Demographic and clinical data, morphometric measurements, and ECG and biological variables were obtained at baseline. A self-reported history of MI or symptomatic peripheral artery disease was noted.

Assays

Blood samples and second morning urine samples were obtained in individuals after an overnight fast. HbA1c and serum creatinine concentrations were centrally determined using a chromatography method (ADAMS A1c HA-8160 analyser; Menarini, Florence, Italy) and a colorimetric method on an automated analyser (KONE Optima; Thermo Clinical Labsystems, Vantaa, Finland), respectively. eGFR was calculated using the Chronic Kidney Disease Epidemiology Collaboration formula [30]. Urinary creatinine was measured on a Hitachi 911 automatic analyser (Roche Diagnostics, Meylan, France), and urinary albumin by nephelometry on a Modular System P (Roche Diagnostics).

Samples were processed under standardised conditions and stored at −80°C in the Poitiers Biological Resource Center (BRC BB-0033-00068, Poitiers, France), undergoing only one freeze-thaw cycle prior to assay. Plasma C-reactive protein (CRP) was measured at baseline using an immunoturbidimetric assay (Roche/Hitachi cobas c systems; Roche Diagnostics). Serum concentrations of TNF receptor 1 (soluble TNFR1) and angiopoietin-like 2 (ANGPTL2), two additional markers of inflammation, were measured using a human soluble TNFR1 ELISA kit (EKF Diagnostics, Dublin, Ireland) and a human ANGPTL2 ELISA kit (Cloud-Clone, Houston, TX, USA), according to the manufacturer’s instructions [7, 8]. Serum TN-C concentrations were measured using a commercially available enzyme immunoassay ELISA kit for large TN-C (catalogue no. JP27751; IBL International, Hamburg, Germany), according to the manufacturer’s instructions. The limit of detection was 0.38 ng/ml, the intra-assay coefficient of variability was 3.8–6.6%, and the inter-assay CV was 3.9–7.0%. Serum was diluted 1:8 in all samples.

Study outcomes

The primary endpoint was all-cause mortality and the secondary endpoint was a composite of CV death, non-fatal MI and non-fatal stroke (major adverse CV events [MACE]) [31]. Vital status and CV endpoints were determined from participants’ hospital records, interviews with their general practitioners and inquiry to the French National Death Registry. Specific cause of each death event was identified by the ICD code on the death certificate, which was issued by authorised clinical practitioners (ICD-9-CM before 1 January 2012, www.icd9data.com/2007/Volume1; ICD-10-CM thereafter, http://apps.who.int/classifications/icd10/browse/2016/en). The present analysis takes the most recent available data (December 2015) into account. Each endpoint was reviewed by an independent adjudication committee according to the international definitions of clinical outcomes. Deaths were categorised as being due to CV disease, malignant tumour, infection, renal causes and other causes. The adjudication committee was blinded with regard to TN-C concentrations. We defined history of CV disease by a history of MI and/or stroke.

Statistical analysis

Qualitative variables are reported as absolute values and percentages, while quantitative variables are described by mean ± SD or median (interquartile range; 25th to 75th percentile), as appropriate. Qualitative variables were compared between tertile groups using the χ2 test, whereas quantitative variables were compared using the Student’s t test or Mann–Whitney U test, as appropriate. Correlation between quantitative variables was tested using a Spearman’s test. The time to event was plotted by the Kaplan–Meier method curves according to tertiles of TN-C, and compared with the logrank test. Risk-prediction models established by Cox proportional hazard models were used to analyse the effect on study outcomes. The HR and 95% CI are presented.

We tested each model for log-linearity and proportionality assumptions and covariates were log-transformed if appropriate.

We adjusted the risk of all-cause death and MACE for predetermined covariates (age, sex, active smoking, statin treatment, hypertension, insulin treatment, eGFR and history of CV disease) as they represent established key risk factors for our outcomes. The relative integrated discrimination improvement (rIDI) index was calculated to assess the improvement in 5 year risk prediction of TN-C in addition to these predetermined risk factors. The 5 year risk was selected since it represents clinically relevant information. Harrell’s C-index was also determined to assess discrimination of the multivariate models. Improvement of increment of TN-C was assessed by bootstrap procedure.

We performed three sensitivity analyses to account for biomarkers of inflammation (CRP, ANGPTL2 and TNFR1), a marker of heart failure (NT-proBNP), additional CV risk factors, such as BMI, HbA1c, diabetes duration, LDL levels, systolic BP (SBP), and the presence of diabetic retinopathy.

All hypotheses were tested at the 5% level of significance. Statistical analyses were carried out using the SAS version 9.4 software package (SAS, Cary, NC, USA).

Results

Out of 1468 participants, the 1321 who met the selection criteria, had available serum samples and had complete covariate data were analysed. The demographic, clinical and biological characteristics of the study population are summarised in Table 1, according to TN-C tertiles.

Table 1 Baseline participant characteristics according to tertiles of serum TN-C concentration
Full size table

Higher TN-C concentrations were significantly associated with older age, longer duration of diabetes, history of CV disease and severe diabetic retinopathy (Table 1). Higher TN-C concentrations were also significantly associated with lower BMI, higher resting heart rate and higher serum concentrations of TNFR1 and ANGPTL2 (Table 1).

We found a weak negative correlation between baseline eGFR and TN-C (ρ = − 0.16, p < 0.0001) and weak positive correlations between baseline albumin/creatinine ratio (ACR) and TN-C (ρ = 0.10, p = 0.0003) and between baseline CRP and TN-C (ρ = 0.19, p < 0.001).

Compared with women, men had significantly higher mean TN-C concentrations (78.3 ± 45.6 ng/ml vs 72.3 ± 40.1 ng/ml; p = 0.0109). However, since there was no significant statistical interaction of sex on the association of TN-C with outcomes, we presented results pooled by sex.

All-cause death and MACE

Median follow-up duration was 89 months (57–130), corresponding to 9965 person-years. The overall mortality rate was 4.4% of total person-years (95% CI 4.0, 4.8; n = 442 participants) and the MACE rate was 5.2% of total person-years (95% CI 4.8, 5.7; n = 497). The major causes of death were CV diseases (n = 240; 54%), cancer (n = 74; 17%), infections (n = 39; 9%), renal causes (n = 4; 1%) and other causes (n = 85; 19%). Non-white ethnicity had no significant impact neither on death (p = 0.438) nor MACE (p = 0.699).

The Kaplan–Meier survival curve for all-cause death across TN-C tertiles is shown in Fig. 1a and for MACE in Fig. 1b. The comparable survival rates in the first two TN-C tertiles and the significantly lower survival rates in the third tertile are explained by the narrow distribution of serum TN-C levels in the tertiles 1 and 2 and a wide range of TN-C concentrations in the tertile 3. Baseline serum TN-C concentrations were higher in the group of individuals who died compared with survivors (89.3 ± 49.8 ng/ml vs 69.0 ± 38.1 ng/ml; p < 0.0001) and higher in individuals with incidental MACE compared with those without (87.2 ± 48.4 ng/ml vs 68.9 ± 38.6 ng/ml; p < 0.0001).

Fig. 1
figure 1

Kaplan–Meier survival curves according to TN-C tertiles for (a) overall risk of death and (b) MACE. p values were calculated using the logrank test

Full size image

After adjustment for predetermined risk factors, higher TN-C concentrations were significantly associated with risk of all-cause death (adjusted HR per 1 SD: 1.27 [95% CI 1.17, 1.38]; p < 0.0001) and MACE (adjusted HR per 1 SD: 1.23 [95% CI 1.13, 1.34]; p < 0.0001) (Table 2).

Table 2 Multivariate Cox analysis for the risk of all-cause death and MACE
Full size table

TN-C and discrimination

Adding TN-C concentrations to these established prognostic factors (age, sex, active smoking, statin treatment, hypertension, insulin treatment, eGFR and history of CV disease) in type 2 diabetes modestly, but significantly improved the accuracy of the risk modelling for all-cause death (change in C-Statistics: 0.0111 [95% CI 0.0107, 0.0114]; rIDI: 8.2%, p = 0.0006) and for MACE (change in C-Statistics: 0.0095 [95% CI 0.0092, 0.0099]; rIDI: 6.7%, p = 0.0014).

Sensitivity analyses

After accounting for predetermined covariates, CRP plus TNFR1 and ANGPTL2 (adjusted HR per 1 SD of TN-C concentration: 1.15 [95% CI 1.05, 1.27], p = 0.0025 and 1.13 [95% CI 1.03, 1.23], p = 0.0085, for all-cause death and MACE, respectively) or after accounting for predetermined covariates and NT-proBNP (adjusted HR per 1 SD: 1.16 [95% CI 1.06–1.27], p = 0.0008 and 1.13 [95% CI 1.04, 1.23], p = 0.0041, for all-cause death and MACE, respectively) in the survival model containing predetermined risk factors, the association findings were roughly unchanged for each outcome.

Additional adjustments for BMI, HbA1c, diabetes duration, LDL levels, SBP and the presence of diabetic retinopathy showed that the associations between serum TN-C concentrations and outcomes were robust and unchanged for each outcome (HR for 1 SD of TN-C: 1.25 [95% CI 1.14, 1.36], p < 0.0001 and 1.31 [95% CI 1.19, 1.43], p < 0.0001, for MACE and all-cause death, respectively).

Discussion

In this study we have shown that higher serum TN-C concentrations were significantly associated with an increased risk of all-cause death and MACE in individuals with type 2 diabetes after adjusting for predetermined risk factors. TN-C remained significantly associated with MACE and all-cause death, even after adjustment for inflammatory markers such as ANGPTL2 and TNFR1, which have been shown previously to be associated with these outcomes in type 2 diabetes [7, 8, 32,33,34,35,36]. This may suggest that the proinflammatory factor TN-C is independently associated with the development of diabetes-related CV disease.

TN-C has been shown to be overexpressed in retinal endothelial cells cultured from diabetic human donors [25], as well as in retinal vascular basement membranes during long-term diabetes in an autopsy series [26]. Accordingly, we found that higher TN-C concentrations were associated cross-sectionally with more severe stages of diabetic retinopathy. Our results are also in accordance with a prior study, which shows that high serum concentrations of TN-C was weakly correlated with GFR [27]. In a population of 92 individuals with CKD (including 42% with type 2 diabetes), Liabeuf et al. reported that TN-C levels increased with the severity of CKD stage and could be used to predict MACE, CV death and all-cause death [27]. Of note, in this study the proportion of individuals with type 2 diabetes was not significantly different between groups with TN-C levels above/below the median [27]. In addition, serum TN-C concentrations have been reported to be significantly higher in individuals with diabetic nephropathy (n = 11) compared with healthy control participants (n = 176), even if no significant correlation between serum TN-C and creatinine clearance was found [37]. In our study, higher TN-C concentrations were not associated with diabetic nephropathy. In high-risk haemodialysed patients (38% with type 2 diabetes), serum TN-C concentrations were found to predict all-cause- and CV-mortality [28]. In the latter study, high TN-C concentrations were associated with high blood glucose concentrations, and the authors attributed this relationship to the microvascular complications related to increased expression of TN-C (among other ECM components) that has been reported in the retina of individuals with diabetes [26]. Indeed, chronic inflammation related to diabetes seems to promote ECM remodelling and to induce the expression of ‘diabetic-specific ECM proteins’, including TN-C, which are not normally present in healthy retinal basement membranes [26].

In a prospective study performed in individuals after acute MI (38% with type 2 diabetes), serum TN-C predicted MACE beyond the cardiac biomarker NT-proBNP [20]. Similar to our findings, in a cross-sectional study of individuals with or without CAD (41% with type 2 diabetes), elevated TN-C was associated with CAD history [23]. In a prospective study of individuals with dilated cardiomyopathy, myocardial expression of TN-C was associated with type 2 diabetes [22]. In the latter study, high TN-C cardiac expression was associated with adverse left ventricular (LV) remodelling and overall mortality [22]. The authors proposed that diabetes, through TGF-β signalling and inflammation, may promote cardiac ECM remodelling including TN-C expression [22], possibly contributing to the decline in LV function [22]. Interestingly, we showed that the prognostic value of TN-C persisted after adjustment for baseline history of MI/stroke and even after considering baseline NT-proBNP as a proxy of LV dysfunction (data not shown). Globally, a link between elevated TN-C concentrations and diabetes has been established in the kidney, retina and heart.

TN-C is considered as a damage-associated molecule pattern (DAMP) protein, which are associated with various chronic inflammatory diseases, including diabetes [38, 39]. TN-C is an ECM protein and ECM remodelling occurs in both human and rodent models of diabetes and insulin resistance [40]. Chronic inflammation induced by obesity or hypoxia in the skeletal muscle, the liver and the adipose tissue is known to promote ECM remodelling. In these organs, increased ECM protein synthesis may contribute to insulin resistance through three hypothetical patways: (1) ECM may act as a physical barrier and prevent glucose and insulin diffusion; (2) ECM may impair vascular function and decrease vascular insulin delivery; (3) ECM may directly modulate insulin action [40]. In this model, inflammation-induced ECM remodelling promotes diabetes. Since TN-C overexpression is a key feature of inflammation-induced ECM remodelling, it seems plausible to assume that, in the SURDIAGENE cohort, TN-C could be involved, not only in the progression of CV-associated diabetes complications, but also in the pathogenesis of diabetes itself. The potential role of TN-C as an aetiological factor in diabetes deserves further study.

In individuals with dysglycaemia, TN-C was found to be predictive of death only, but not MACE [11, 12]. Interestingly, TN-C showed a significant predictive value in our cohort not only for all-cause death, but also for MACE, suggesting that TN-C overexpression in individuals with type 2 diabetes might be linked to CV disease development and progression.

Our study has a number of potential limitations; for example, the French participants from the SURDIAGENE study were recruited in a prospective hospital-based monocentric cohort and this design may limit extrapolation of our findings.

The strengths of this study include an extended follow-up period and independent adjudication for outcomes. Samples were collected under standardised conditions and stored at −80°C in a certified biobank, undergoing only one prior freeze–thaw cycle. To our knowledge, this is the first prospective study in type 2 diabetes reporting the association of serum TN-C with risk of death or severe CV complications.

On the basis of the data collected in this prospective type 2 diabetes cohort, we conclude that increased serum TN-C concentration is independently associated with risk of all-cause death and MACE, a finding facilitating improvement of risk prediction on top of traditional CV risk factors. Thus, TN-C could be a promising biomarker for improved risk stratification in individuals with type 2 diabetes. Further research will decipher whether TN-C is a potential therapeutic target in individuals with type 2 diabetes.