Higher Collagen VI Formation Is Associated With All-Cause Mortality in Patients With Type 2 Diabetes and Microalbuminuria

Type 2 diabetes is the most common etiology in patients with chronic kidney disease (CKD). Patients with diabetes are prone to develop microalbuminuria, which is associated with an accelerated deterioration of kidney function (1). Furthermore, albuminuria is as a prominent risk factor for renal (2) and cardiovascular (3,4) disease. The hazardous effect of albuminuria is partially explained by promoting a proinflammatory environment that causes further damage to the kidney and thus exacerbates the condition (5). The management of blood pressure (e.g., renin-angiotensin-aldosterone system (RAAS) blockers and diuretics) and lifestyle has substantially delayed progression of renal and cardiovascular complications in patients with type 2 diabetes (6,7). However, despite the standard of care, patients with type 2 diabetes are still at a higher risk of developing CKD, cardiovascular events, and mortality than the general population (8,9).

Currently, no antifibrotic drugs are available, so the treatment of patients with advanced CKD relies on dialysis and ultimately transplantation. Biomarkers, which add diagnostic and prognostic information to the currently available risk factors, are required to promote the development of such compounds.

Despite the underlying etiology, CKD is assumed to be driven by renal fibrosis caused by an abnormal shift in the turnover of components of the renal extracellular matrix (10,11). Collagen is an essential part of the fibrotic structure, acting as a scaffold for a range of interaction partners and cell adhesion (12–14). Because collagen formation and development of fibrosis are closely linked, assessment of collagen formation may identify patients with active fibrosis at higher risk for cardiovascular events, mortality, and rapid deterioration of kidney function.

In this study, we focused on PRO-C6, a biomarker in serum and urine that detects the COOH-terminal propeptide of collagen VI (COL VI), which is released during deposition in the extracellular matrix (15). COL VI is expressed in the renal extracellular matrix during organ development but is expressed only at low levels in the healthy adult kidney (16). It is positioned at the interface between the interstitial matrix and the glomerular basement membrane, where it forms an intricate meshwork of microfilaments with a physiological role in maintaining structure and function, controlling the organization of the matrix and cell orientation (17). Increased deposition has been reported in situ in patients with diabetes in whom an increased COL VI deposition is seen throughout nodular lesions (18) and seemingly replaces COL IV in the glomerular basement membrane (18).

Seminal studies have shown that collagens are capable of interacting both locally and with distant sites through the release of fragments termed matrikines that stimulate signaling activity (11,19). During production of COL VI, the COOH-terminal propeptide of the α-3 chain is released (15), which contains the matrikine endotrophin (20). Endotrophin has a range of potentially deleterious effects, such as attracting macrophages, increasing transforming growth factor-β expression, and promoting epithelial-mesenchymal transition, adipose tissue fibrosis, and metabolic dysfunction (20). Because of sequence overlap in the epitope, the fragment recognized by PRO-C6 also contains endotrophin. These findings support a potential role of PRO-C6 as a marker that may reflect both COL VI formation and a fragment that amplifies tissue injury in the local microenvironment.

The role of COL VI formation and endotrophin in the pathophysiology of diabetic kidney disease has not been elucidated. We therefore hypothesized that PRO-C6 represents pathophysiological processes associated with COL VI formation and thus predicts cardiovascular events, mortality, and disease progression in patients with type 2 diabetes and microalbuminuria. To address this hypothesis, we evaluated PRO-C6 as a prognostic marker for these outcomes in a prospective, observational study of patients with type 2 diabetes and microalbuminuria without symptoms of coronary artery disease.

Two hundred participants were recruited between January 2007 and February 2008 from the outpatient clinic at Steno Diabetes Center Copenhagen as previously described (21). Inclusion criteria were 1) type 2 diabetes according to World Health Organization criteria, 2) no history of coronary artery disease or symptoms suggestive of cardiac disease (assessed from patient interviews and records), and 3) persistent urinary albumin excretion ratio (UAER) >30 mg/24 h (in two of three consecutive measurements). Exclusion criteria were normal UAER or nonpersistent elevated UAER, symptoms of heart failure, angina pectoris, ischemic heart disease or diagnosed myocardial infarction, ischemic heart failure, malignant arrhythmia, and atrial fibrillation. The study complied with the Declaration of Helsinki, the research protocol was approved by the local ethics committee (De Videnskabsetiske Komitéer for Region Hovedstaden, Hillerød, Denmark), and all participants gave written informed consent. Median (range) follow-up time was 6.5 (0.3–7.2) years.

Serum samples from 198 patients and urine from 190 patients were available for the current study. PRO-C6 was measured in serum (S-PRO-C6) and urine (U-PRO-C6) by using a competitive ELISA developed by Nordic Bioscience (Herlev, Denmark). The monoclonal antibody used in the ELISA specifically detects the last 10 amino acids of the α-3 chain of COL VI (3168′KPGVISVMGT′3177) (22). Intra- and interassay variations of the assay were 3.2 and 7.9%, respectively. To normalize for urine output, U-PRO-C6 levels were divided by urinary creatinine levels.

The assay was carried out as previously described (22). Briefly, a streptavidin-coated 96-well ELISA plate (cat. 11940279; Roche) was coated with a biotinylated peptide. The plate was washed five times in washing buffer followed by incubation with standard peptide or sample together with horseradish peroxidase–conjugated monoclonal antibody. The plate was washed five times followed by incubation with 3,3′,5,5′-tetramethylbenzidine (Kem-En-Tec, Taastrup, Denmark) in the dark. To end the reaction, a 1% sulfuric acid solution was added, and the plate was analyzed on the ELISA reader at 450 nm, with 650 nm as the reference. All incubation steps were carried out with constant shaking at 300 rpm. Concentrations were corrected for dilution factor of the samples (twofold dilution for serum and undiluted for urine).

UAER was measured in 24-h urine samples by enzyme immunoassay (VITROS; Ortho Clinical Diagnostics, Raritan, NJ). Current smoking was defined as one or more cigarettes, cigars, or pipes per day. Brachial blood pressure was measured twice after a 10-min rest in seated patients with an oscillometric device using an appropriate cuff size and then averaged.

Patients with samples available were traced through the Danish National Death Register and the Danish National Health Register on 1 January 2014. Definitions of the three predefined end points have been previously described (23,24). The combined cardiovascular events included cardiovascular mortality, stroke, ischemic cardiovascular disease, and heart failure. For patients with multiple events, only the first was included. Moreover, 175 of the 198 patients (88.4%) had yearly measurements for all visits. Plasma creatinine from these visits was used to calculate the estimated glomerular filtration rate (eGFR) by the Chronic Kidney Disease Epidemiology Collaboration equation (25). As proposed by Coresh et al. (26), the renal end point was defined as a decline in eGFR >30% (evaluated as change from baseline to the last available measurement). In the current study, this end point was termed disease progression. Because some patients did not attend all visits, data on disease progression were available only for the aforementioned 175 (88.4%) patients.

All variables, except eGFR and systolic and diastolic blood pressure, had a non-Gaussian distribution (nonnormally distributed), as assessed by the D’Agostino and Pearson normality test, and are summarized as the median with interquartile range (IQR). Summary statistics for eGFR and systolic and diastolic blood pressure are presented as mean ± SD. Categorical variables are summarized as total numbers with corresponding percentages. S-PRO-C6 and U-PRO-C6 were either categorized into tertiles or per doubling (log2 transformed).

To test for differences in potential confounders, we stratified the study population according to S-PRO-C6 tertiles. Differences among tertiles were assessed with a one-way ANOVA for normally distributed variables, Kruskal-Wallis test for nonnormally distributed variables, and χ² test for categorical variables.

In the study, levels of S-PRO-C6 and U-PRO-C6 were tested against three outcomes: 1) the cardiovascular end point, 2) all-cause mortality, and 3) disease progression. Differences in S-PRO-C6 and U-PRO-C6 levels in patients with and without outcomes were assessed with the nonparametric Mann-Whitney test. We applied the Kaplan-Meier function to compare the risks of the three end points according to the tertiles of S-PRO-C6 and U-PRO-C6.

Cox proportional hazards regression analysis was used to generate unadjusted hazard ratios (HRs) with 95% CIs per doubling of S-PRO-C6 and U-PRO-C6 for the three end points. HRs of S-PRO-C6 and U-PRO-C6 were also calculated after adjustment for conventional cardiovascular risk factors: sex, age, BMI, systolic blood pressure, LDL cholesterol, smoking, HbA_1c, plasma creatinine, and UAER. Because of missing data for the conventional risk factors, seven patients were excluded from the analysis for cardiovascular events and all-cause mortality (final n = 191), and five patients were excluded for disease progression (final n = 170).

To quantify the added predictive value of S-PRO-C6 and U-PRO-C6, we calculated receiver operating characteristic (ROC) curves. We applied C statistics to compare the area under the curve (AUC) for the model, including conventional cardiovascular risk factors (base model), and the AUC for the model that included conventional cardiovascular risk factors plus S-PRO-C6 (base model + S-PRO-C6) and U-PRO-C6 (base model + U-PRO-C6).

Because C statistics often do not significantly improve even after the addition of powerful predictors (27) and do not consider time to event, we also calculated relative integrated discrimination improvement (rIDI) (27). rIDI is a measure suggested by Pencina et al. (27) as a relevant method to demonstrate usefulness of a biomarker when added to a model of conventional risk factors. rIDI is defined as the increase in discrimination slope when adding S-PRO-C6 or U-PRO-C6 to conventional risk factors (model 2) divided by the discrimination slope of the model only consisting of the conventional risk factors (model 1) minus 1 (27). The rIDI was calculated as follows:

All two-tailed P < 0.05 were considered significant. Statistical analyses were made using MedCalc (Ostend, Belgium), SAS 9.4 (SAS Institute, Cary, NC), and GraphPad Prism 6 (GraphPad Software, San Diego, CA) statistical software.

The population of 198 patients was followed for a median (range) of 6.5 (0.3–7.2) years and had a median (IQR) age of 60 (54–65) years and diabetes duration of 12 (7–18) years; 149 (75%) were men. The patients had a mean ± SD eGFR of 90 ± 17 mL/min/1.73 m², systolic blood pressure of 130 ± 17 mmHg, diastolic blood pressure of 75 ± 11 mmHg, and median (IQR) HbA_1c of 58 (52–73) mmol/mol. The patients were treated with oral antidiabetic medications (85%), insulin (62%), and cardiovascular medications, including antihypertensive drugs (99%), RAAS blockers (98%), statins (94%), aspirin (92%), and β-blockers (14%).

The clinical characteristics of the 198 patients enrolled in the study are presented in Table 1. Patients were stratified by tertiles of S-PRO-C6 levels. Age and BMI were significantly higher in the third tertile than in the first and second tertiles (P = 0.004 and 0.01, respectively). Plasma creatinine and known duration of diabetes increased significantly (both P < 0.0001), and eGFR decreased (P < 0.001) with tertiles. No differences among tertiles were observed for any of the other variables.

We first investigated whether levels of S-PRO-C6 and U-PRO-C6 had a Gaussian distribution. For both S-PRO-C6 (K2 = 86.1; P < 0.0001) (Supplementary Fig. 1A) and U-PRO-C6 (K2 = 251.0; P < 0.0001) (Supplementary Fig. 1B) a non-Gaussian distribution was seen. Log transformation led to a borderline Gaussian distribution for S-PRO-C6 (K2 = 6.0; P = 0.05) but not for U-PRO-C6 (K2 = 70.9; P < 0.0001). No significant correlation between S-PRO-C6 and U-PRO-C6 was demonstrated (Spearman r = 0.03; P = 0.70) (Supplementary Fig. 1C). Because U-PRO-C6 was corrected for creatinine, thereby assuming similar renal excretion pathways as creatinine, we also analyzed S-PRO-C6 and U-PRO-C6 not normalized for urinary creatinine. No association between unadjusted U-PRO-C6 and S-PRO-C6 were observed. A significant correlation between S-PRO-C6 and plasma creatinine at baseline was observed (Spearman r = 0.48; P < 0.0001) (Supplementary Fig. 1D) but not with UAER (Spearman r = 0.00; P = 0.96) (Supplementary Fig. 1E). No correlation was observed between U-PRO-C6 and plasma creatinine or UAER.

During follow-up, 38 (19.2%) patients experienced a cardiovascular event, 26 (13.1%) died, and 42 (21.2%) had >30% deterioration of kidney function, which was defined as renal disease progression. Patients experiencing a cardiovascular event had significantly higher S-PRO-C6 (median [IQR] 9.0 [6.6–11.5] vs. 6.9 [5.7–8.8] ng/mL; P = 0.003) and U-PRO-C6 (0.24 [0.19–0.33] vs. 0.22 [0.17–0.27] ng/mL; P = 0.03) levels than unaffected patients. Deceased patients had a significantly higher S-PRO-C6 level than survivors (9.7 [7.4–11.5] vs. 6.8 [5.8–8.8] ng/mL; P = 0.0006). S-PRO-C6 levels were not different in patients who died as a result of cardiovascular disease versus other causes (10.7 [8.7–11.8] vs. 8.7 [6.6–10.3] ng/mL; P = 0.20). No difference in U-PRO-C6 levels between groups was seen. Patients with renal disease progression had significantly higher S-PRO-C6 levels than those without renal disease progression (8.4 [6.6–10.1] vs. 6.7 [5.5–8.7]; P = 0.0004). No difference in U-PRO-C6 levels was seen between groups.

To determine whether PRO-C6 was associated with the investigated outcomes, we applied the Kaplan-Meier failure function to compare the risks of fatal and nonfatal cardiovascular events, all-cause mortality, and disease progression according to the tertiles of S-PRO-C6 and U-PRO-C6 (Fig. 1). A significantly increased risk of experiencing cardiovascular events (χ² = 10.8; P = 0.005) (Fig. 1A), all-cause mortality (χ² = 21.2; P < 0.0001) (Fig. 1B), and disease progression (χ² = 12.6; P = 0.002) (Fig. 1C) was seen with increasing tertiles of S-PRO-C6. No association between U-PRO-C6 tertiles and outcomes was demonstrated for cardiovascular events (χ² = 5.43; P = 0.07) (Fig. 1D), all-cause mortality (χ² = 3.05; P = 0.22) (Fig. 1E), or disease progression (χ² = 0.97; P = 0.62) (Fig. 1F).

We next investigated whether the levels of S-PRO-C6 and U-PRO-C6 were independently associated with these outcomes in continuous analyses. We applied Cox proportional hazards regression analysis for the biomarkers unadjusted and adjusted for conventional risk factors (age, sex, BMI, systolic blood pressure, LDL cholesterol, smoking, HbA_1c, plasma creatinine, and UAER). Higher S-PRO-C6 was associated with cardiovascular events, all-cause mortality, and disease progression (all P ≤ 0.001), and remained associated with all outcomes after adjustment (all P ≤ 0.01) (Table 2). In unadjusted analyses, higher U-PRO-C6 was associated with cardiovascular events (HR per doubling 1.54 [95% CI 1.03–2.29]; P = 0.04) and borderline associated with all-cause mortality (1.63 [1.00–2.65]; P = 0.05). After adjustment, both associations were lost. The remaining associations were not significant (P ≥ 0.11) (Table 2). No association between U-PRO-C6 unadjusted for urinary creatinine and the various outcomes was observed. The variables associated with the outcomes in the adjusted models were higher S-PRO-C6 and age for cardiovascular events; higher S-PRO-C6, smoking, and lower plasma creatinine for all-cause mortality; and higher S-PRO-C6 and UAER and lower BMI for disease progression (Supplementary Table 1).

We first assessed whether the addition of S-PRO-C6 and U-PRO-C6 to a conventional model of risk factors led to a significant increase in C statistics by using ROC curve analyses (Fig. 2). The model containing conventional risk factors yielded an AUC of 0.76 (95% CI 0.70–0.82; P < 0.0001) for cardiovascular events (Fig. 2A), 0.74 (0.67–0.80; P < 0.0001) for all-cause mortality (Fig. 2B), and 0.73 (0.65–0.79; P < 0.0001) for disease progression (Fig. 2C). The addition of S-PRO-C6 increased the C statistics for all outcomes, but not significantly (P > 0.17) (Fig. 2). The addition of U-PRO-C6 increased the C statistics for cardiovascular events and all-cause mortality, but not significantly (P > 0.22) (Fig. 2).

To further assess whether S-PRO-C6 or U-PRO-C6 added incremental predictive value, we used rIDI as suggested by Pencina et al. (27). rIDI quantifies the increase in separation of events and nonevents on a relative scale and takes time to event into consideration. The addition of S-PRO-C6 to the model containing conventional risk factors improved rIDI by 22.5% (P = 0.02) for cardiovascular events, 76.8% (P = 0.002) for all-cause mortality, and 53.3% (P = 0.004) for disease progression, indicating that S-PRO-C6 adds incremental predictive value. The addition of U-PRO-C6 to the models did not lead to a significant improvement of rIDI for any outcome.

In a study of 198 patients with type 2 diabetes and microalbuminuria without a history or symptoms of coronary artery disease, we demonstrate that higher levels of a fragment reflecting COL VI formation, S-PRO-C6, is an independent predictor of cardiovascular events, all-cause mortality, and loss of renal function. The addition of either S-PRO-C6 or U-PRO-C6 to a model containing conventional risk factors increased C statistics, but not significantly. Because C statistics often do not change considerably, even for potent risk markers, and do not consider time to event, rIDI has been suggested as a relevant addition to the conventional methods of determining incremental added value of a marker to risk prediction (27). By using rIDI, we found that the addition of S-PRO-C6 to the model increased the prediction of all outcomes beyond conventional risk markers.

Even when adjusted for conventional risk factors (among others age, sex, and plasma creatinine found in the eGFR equations), S-PRO-C6 remained significantly associated with renal disease progression as defined by a deterioration of kidney function of >30% from baseline. Plasma creatinine in the current cohort was not associated with future loss of kidney function, but S-PRO-C6, BMI, and UAER were. This finding could indicate that S-PRO-C6 is an early marker of future loss of kidney function in patients with type 2 diabetes and microalbuminuria and, on average, normal kidney function.

The involvement of COL VI in kidney disease has mainly been investigated by studies that used immunohistochemical analysis (18). Increased COL VI has been shown to be prognostic for disease progression in patients with idiopathic membranous nephropathy (28) and diagnostic for kidney disease (29). To our knowledge, only studies performed by our group have since attempted to elucidate the association of COL VI formation with diabetes and kidney disease. We have shown previously that higher PRO-C6 identifies a subpopulation of patients with type 2 diabetes who respond to insulin sensitizing treatment (30) and is independently associated with mortality in patients with CKD of mixed etiology (31). The current findings are in line with previous studies; therefore, PRO-C6 likely reflects a pathophysiological alteration of the extracellular matrix associated with disease activity and progression.

CKD is characterized by a progressive reduction of GFR as a result of loss of functional filtration surface area, a process believed to be driven by fibrosis (10). Biomarkers reflecting pathological alterations to the extracellular matrix, therefore, may be superior to biomarkers reflecting functional changes (13). The most well-investigated biomarkers reflecting changes to the extracellular matrix in patients with CKD are those of proteomic studies, such as the urinary CKD273 classifier (32). These studies mostly contain fragments of COL I and III, not COL VI (32). Because various collagens have diverse functions, relating the current findings for PRO-C6, and thus COL VI, to those of the proteomic studies in patients with CKD is difficult.

S-PRO-C6 and BMI were positively associated in this study. Obesity, among other risk factors, is a leading contributor to morbidity and mortality worldwide because it promotes cardiometabolic disease and fibrosis (33). We show that PRO-C6 is independently associated with outcomes after adjustment for BMI and adds incremental predictive value to a model of conventional cardiovascular risk factors, including BMI. This finding indicates that PRO-C6 is not merely a result of obesity but partially a result of dysfunctional adipose tissue characterized by chronic low-grade inflammation that spreads to several tissues and eventually causes cardiovascular disease (33).

One of the strengths of the study is that we were able to adjust for important conventional risk factors. The comprehensive adjustments for the conventional risk factors in conjunction with the relatively small sample size and number of events may, however, lead to unstable model estimates as a result of overfitting. Another strength is that the patients had microalbuminuria, whereas on average, their kidney function was preserved. The cohort, therefore, comprised patients in early to moderate stages of kidney disease (CKD stage 1 to early 3B), which allowed us to test whether COL VI formation acts as an early marker for poor prognosis in patients with relatively nonadvanced CKD. This is crucial as therapeutic intervention is more likely to be effective during the early period of kidney disease because the reversibility of fibrosis during advanced stages becomes questionable as a result of various processes, such as cross-linking of the extracellular matrix (34). The finding that both S-PRO-C6 and albuminuria were significantly associated with disease progression in the model with all conventional risk factors supports the involvement of COL VI in the pathophysiology leading to poor outcome in patients with type 2 diabetes. Together with the improvement in rIDI of the model containing conventional risk factors, including albuminuria, this finding supports the notion that PRO-C6 adds incremental predictive value to a marker reflecting microvascular injury.

As a limitation, we were unable to show a mechanistic link between higher S-PRO-C6 and poor prognosis; therefore, we could not provide causal evidence for the increased risk of cardiovascular events, mortality, and disease progression in patients with higher S-PRO-C6. Moreover, because the main findings were shown in serum and not urine, S-PRO-C6 is likely to be a marker of widespread systemic expansion of the extracellular matrix in diabetes. Because COL VI is upregulated in kidney disease (18), the myocardium (35), atherosclerotic lesions (36), processes leading to heart failure (37), and cirrhotic liver disease (38), higher PRO-C6, and thus COL VI, may reflect an increased fibrogenic burden at a systemic level where more than one organ contribute to the pool of PRO-C6 in serum. A mechanistic link between COL VI formation and poor prognosis was shown in a study by Naugle et al. (39) where COL VI promoted differentiation of cardiac fibroblasts in vitro and increased COL VI colocalized with myofibroblasts in the myocardium of rats subjected to myocardial infarction. Furthermore, because deficiency of the COL VI α-1 chain reduced the damage after myocardial infarction (40), the upregulation of COL VI exacerbates injury. Combined, the notion that COL VI is not only a consequence but also a potential cause of poor prognosis is supported. Another limitation is that information on measured GFR/creatinine clearance was not available. On the basis of these and previously published findings for PRO-C6, we believe that S-PRO-C6 merits further investigation as a risk marker in both patients with kidney disease and patients with diabetes. A determination of whether changes in PRO-C6 predict response to kidney and/or cardioprotective therapies is important.

In conclusion, higher S-PRO-C6 is independently associated with cardiovascular events, all-cause mortality, and disease progression. S-PRO-C6 adds significant incremental predictive value to conventional risk factors in patients with type 2 diabetes and microalbuminuria.

Funding. This work was supported by the Danish Research Foundation and the Danish Agency for Science, Technology and Innovation (ID 4135-00023).

Duality of Interest. D.G.K.R., S.H.N., M.A.K., and F.G. are full-time employees of Nordic Bioscience. Nordic Bioscience is a privately owned, small- to medium-sized enterprise partly focused on the development of biomarkers. None of the authors received fees, bonuses, or other benefits for the work described in the article. M.A.K. holds stocks in Nordic Bioscience. The patent for the ELISA used to measure PRO-C6 levels is owned by Nordic Bioscience. The funder provided support in the form of salaries for D.G.K.R., S.H.N., M.A.K., and F.G. but did not play any additional role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. Outside this work, B.J.v.S. is employed by Novo Nordisk, and P.R. has received research grants from Novo Nordisk and AstraZeneca; has acted as consultant for Novo Nordisk, Bayer, Astellas, Boehringer Ingelheim, AbbVie, and AstraZeneca (honoraria to institution); and has shares in Novo Nordisk. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. D.G.K.R. wrote the manuscript. D.G.K.R., T.W.H., and S.H.N. acquired and researched data. D.G.K.R., M.T., F.G., and P.R. contributed to the discussion and reviewed/edited the manuscript. T.W.H., B.J.v.S., H.R., H.-H.P., M.A.K., and P.K.J. reviewed/edited the manuscript. D.G.K.R. 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 abstract form at the annual meeting of the American Society of Nephrology, New Orleans, LA, 31 October–5 November 2017.