Biomarkers That Predict Diabetic Nephropathy: The Long Road From Finding Targets to Clinical Use

Diabetic kidney disease is the leading cause of renal failure in the U.S. and Western industrialized countries, affecting 20–30% of type 1 and type 2 diabetic patients. With the expansion of the obesity and type 2 diabetes epidemic worldwide and particularly in Asia, we can only anticipate an increase in the worldwide incidence of diabetic renal disease (1). It is difficult to overstate the physical, social, economic, and personal burdens created by diabetic nephropathy. And yet, caregivers are in many ways adrift in their efforts to predict which diabetic patients will develop progressive renal dysfunction. Those who wish to develop new pharmaceuticals to prevent or reverse diabetic nephropathy are also at a loss to accurately identify the subpopulation of patients with diabetes that will develop nephropathy and potentially benefit from new treatments. The judgment of clinicians as to who will develop diabetic nephropathy has come unmoored as a result of the loss of microalbuminuria as a biomarker of progressive renal dysfunction in patients with diabetes. In type 1 diabetic patients, confidence in the predictive value of microalbuminuria has declined as a result of observations that a large proportion of patients with microalbuminuria revert to normoalbuminuria and, in a significant percentage of patients, renal function has declined before the onset of microalbuminuria (2,3). In type 2 diabetes, approximately 20% of patients may develop stage 3 chronic kidney disease and remain normoalbuminuric (4). Thus, while microalbuminuria may be an indicator of renal damage, considerable doubt has emerged that it is a predictor of progressive chronic renal disease in patients with diabetes. This growing doubt as to the relevance of microalbuminuria as a diabetic nephropathy predictor has promoted a vigorous search for new biomarkers.

In this issue of Diabetes, Zürbig et al. (5) report the results of a proteomic analysis of urine from type 1 and type 2 diabetic patients that focused on the identification of biomarkers that predict renal dysfunction. The study samples were a longitudinal cohort, and capillary-electrophoresis coupled to mass spectrometry was used to profile the low–molecular weight proteome in urine. This low–molecular weight protein fraction (<2,000 Daltons) is frequently referred to as the “peptidome” and is composed of intact low–molecular weight proteins, as well as peptide fragments. The peptidome has been studied using mass spectrometry to identify biomarker candidates for a variety of disease states, including malignancy and renal disease (6–8). Zürbig et al. used a previously developed urine peptide pattern to determine if it was predictive of macroalbuminuria. The previously described peptide panel was composed of 273 different urine peptides and was developed to identify the risk of progressive chronic kidney disease regardless of the cause (9). In subsequent work, the authors used tandem mass spectrometry to identify all of the 273 marker peptides and found that 74% were collagen fragments. In the current study, the 273 peptide classifier was applied to mass spectra obtained from urine of patients 1–5 years before the onset of macroalbuminuria. The classifier system was able with good sensitivity to identify those patients that 5 years later developed macroalbuminuria and was superior in this regard to microalbuminuria as a predictor. Interestingly, specific peptides had lower urine expression in patients that subsequently developed macroalbuminuria. These peptides included fragments of multiple collagens, clusterin, and CD99. Finally, the authors compared the specificity and sensitivity of urine markers to predict macroalbuminuria relative to “early” and “late” specimens obtained 3–5 years and 0–2 years, respectively, prior to the onset of macroalbuminuria. The performance of the 273 peptide classifier did not vary substantially between early and late specimens and performed better in early specimens than microalbuminuria.

The strengths of this study are its use of longitudinal specimens and the analysis of the biomarker as a predictor over time. Biomarker discovery studies that rely on cross-sectional specimens appear less likely to identify markers that remain robust in subsequent analyses. The current study also provides additional evidence that renal collagen homeostasis is perturbed early in diabetic nephropathy, as reported previously by the authors and other workers in the field (10,11). Most importantly, this study adds to a growing number of candidate biomarkers that may predict in advance the 20–40% of patients with diabetes that are fated to develop progressive renal disease. Such a biomarker would create new opportunities to focus resources and current therapies on those at risk, as well as provide a valuable tool for therapeutic development.

However, the road from discovery of a biomarker candidate to implementation of a clinical assay is a long one. The work by Zürbig et al. represents a mixture of discovery and validation as illustrated in Fig. 1. The initial discovery of the 273-marker panel required the identification and quantification of several thousand peptides in a very limited number of specimens. Although the current study uses a panel of only 273 peptides to predict the development of diabetic nephropathy, that number remains unwieldy when constructing a clinically applicable biomarker test that must be validated in several thousand specimens. The authors correctly point out that the transition from a mass spectrometry–based assay to a clinical assay will be challenging. The vast majority of our current clinical assays are quantitative antibody–based tests. The development of antibodies to the specifically modified collagen fragments would be quite difficult and construction of an ELISA to such small peptides has its own unique set of challenges. The alternative would be to perform high-throughput quantitative mass spectrometry in the clinical setting (12).

• Download figure
• Open in new tab
• Download powerpoint

FIG. 1.

Schema of biomarker discovery, validation, and implementation.

Despite these challenges, Zürbig et al. have made a notable contribution to the effort to identify new biomarkers that predict diabetic nephropathy and to a better understanding of its pathophysiology. Their work provides additional evidence that changes in urine extracellular matrix proteins occur quite early in the course of diabetic nephropathy. This furthers the concept that urine collagen expression has significant potential as a biomarker to predict diabetic nephropathy. They correctly suggest that these changes in collagen metabolism should be studied further in animal models. Their work indicates that the study of collagen homeostasis in early diabetic nephropathy may give us important mechanistic insights into this important clinical problem.

• © 2012 by the American Diabetes Association.