Urinary Proteomics as a Tool to Identify Kidney Responders to Dipeptidyl Peptidase-4 Inhibition: A Hypothesis-Generating Analysis from the MARLINA-T2D Trial

2.1 Patients and Sample Collection

Samples were derived from the completed MARLINA-T2D study (NCT NCT01792518) and the study design, methodology and key results have been reported.11, 25 In brief, 360 T2D patients with prevalent albuminuria (30–3000 mg/g) despite stable background RAS blockade that met the eligibility criteria after a 2-week placebo run-in period, were randomized 1:1 to receive double-blind, once-daily oral treatment with linagliptin 5 mg or placebo (on top of recommend standard of care for DKD) for 24 weeks. Urine samples available for proteome analyses were collected after the placebo run-in period (i.e., at baseline) and after the 24 weeks of treatment (end of study) according the standardized study procedures as reported.26 Both urine samples were available for 89.7% (n = 323) of the total study population. Estimated glomerular filtration rate (eGFR) and UACR levels were analyzed every 6 weeks during study conduct. The study was designed and conducted in accordance with the standards of good clinical practice and principles of the Helsinki Declaration. Written informed consent was obtained from all participants. MARLINA-T2D was approved by independent ethics committees institutional review boards at each participating centre.

2.2 Urinary DPP-4 Activity

The DPP-4 assay was carried out using 10 µL human urine. Briefly, 40 µL of buffer containing 100 mm Tris and 100 mm NaCl, pH 7.8 was transferred to 10 µL urine. Subsequently, 50 µL Ala-Pro-AFC (H-Ala-Pro-7-amido-4-trifluoro-methylcoumarin) with a final concentration of 100 µ µm was added and incubated for 60 min at room temperature. The AFC release was measured using SpectraMax M5 fluorescence microplate reader (Molecular Devices) at 405 nm excitation/535 nm emission respectively. Urinary DPP-4 activity was normalized to urinary creatinine.

2.3 Urinary Proteome Analysis

Urine samples were prepared and analyzed using a P/ACE MDQ capillary electrophoresis system (Beckman Coulter, Fullerton, USA) on line coupled to a MicroTOF MS (Bruker, Bremen, Germany) described previously in detail.27 The MosaiquesVisu software was used for data evaluation.28 Accuracy, precision, selectivity, sensitivity, reproducibility, and stability of the CE-MS method have been previously described.29 Internal standard peptides were used for calibration, as previously described.30 All detected peptides were deposited, matched, and annotated in a MicrosoftSQL database, allowing for further analysis and comparison between groups.

To assess the CKD273 score a previously generated classifier was used.17 CKD273 is a support vector machine-based classification model that allows the classification of samples in a high-dimensional parameter space using MosaCluster software (version 1.7.0).31, 32 MosaCluster calculated classification scores are based on amplitudes of identified biomarkers within CKD273. Technicians performing proteomics analyses were blinded in terms of treatment allocation and study endpoints. To retrieve amino acid sequences of identified peptides, the Human Urinary Proteome Database, version 2.0.33 was utilized and respective peptides from this database were sequenced, as previously described.33, 34

2.4 Statistical Analysis

CKD273 scores, eGFR, and UACR values were not normally distributed, and therefore a nonparametric Spearman's rank correlation coefficient was used for estimating respective correlation coefficients by MedCalc (version 12.1.0.0, http://www.medcalc.be; MedCalc Software, Mariakerke, Belgium).35 GFR slopes calculation based on within-patient regression analysis using all available eGFR values estimated between the baseline and follow-up visit. Renal function loss was defined based on calculated eGFR % slopes per year. These eGFR slopes served as a dependent variable in multiple regression analyses and the additional predictive value of the CKD273 was estimated on top of relevant baseline covariates (UACR and eGFR, independent variables). To exclude potential bias introduced by linagliptin treatment, in this analysis, only data of 155 patients treated with placebo were used. The eGFR values between the linaglitin and placebo-treated groups were compared using the nonparametric Mann–Whitney test. A paired Wilcoxon test was used to analyze the eGFR changes before and after linaglitin or placebo treatment (MedCalc35).

To identify peptides significantly altered by linagliptin treatment, UPP from baseline and follow-up samples were compared. Statistical analysis was performed using non-parametric Wilcoxon rank sum test. Only peptides that were present at a frequency of 30% or higher in either baseline or follow-up samples were investigated, ensuring that altered peptides were not biased by preanalytics and/or potential population-specific genetic variability. A false discovery rate adjustment of Benjamini–Hochberg36 was employed to correct for multiple testing. In addition, a correlation analysis between baseline peptide amplitude (signal intensity) and follow-up DPP-4 peptidase activity (µg/g urine creatinin) was performed also by MedCalc35 using Spearman's rank correlation analysis. Overall, a p-value < 0.05 was considered to be statistically significant.

3.1 Urinary CKD273 Classification Score and Parameters of Kidney Disease at Baseline

As expected in our cross-sectional analyses, the MARLINA-T2D cohort confirmed previous results from other cohorts reporting a significant correlation between eGFR and UACR values at baseline (rho = −0.30; Figure 1A). Moreover, we confirmed previous reports of a significant, inverse correlation between the CKD273 and eGFR (rho = −0.41; Figure 1B) and a positive correlation of the CKD273 with UACR (rho = 0.54; Figure 1C).

3.2 Urinary CKD273 Classification Score and Prediction of CKD Progression after 24 Weeks

In the next step, we aimed to explore whether CKD273 could predict CKD progression in MARLINA-T2D, as previously described in several cohorts of patients with CKD stages 1–4.20, 22, 23, 37 To exclude potential bias introduced by linagliptin treatment, we utilized the available baseline and follow-up data of 155 patients treated with placebo and calculated the percent decrease in eGFR expressed as annual slopes as described previously.20 In order to assess if the CKD273 could independently predict CKD progression (i.e., annual eGFR decrease), this parameter was added alongside baseline UACR and eGFR as an independent variable into a multivariate regression analysis. In result, the CKD273 (p = 0.027) as well as baseline eGFR values (p = 0.015) were independently associated with progressive GFR loss.

3.3 Urinary CKD273 Classification Score and Kidney Treatment Response to Linagliptin

Among the total study population, the renal function, expressed as eGFR (Table S1, Supporting Information), was not significantly different between the placebo and linagliptin group at baseline (median GFR placebo: 92 mL/min /1.73 m², interquartile range [IR]: 63–114 vs median eGFR linagliptin: 95 mL/min/1.73 m², IR: 71–130, p = 0.184) and there were no clinically meaningful changes in either group after 24 weeks of treatment (median eGFR placebo: 89 mL/min/1.73 m², IR: 59–116 vs median eGFR linagliptin: 96 mL/min/1.73 m², IR: 70–121, p = 0.301) (see Figure 2A). To explore our first main hypothesis, we stratified patients according to their respective baseline CKD273 classification score into a high or low CKD progression risk group, respectively (cut-off 0.343, as established previously17). There were no significant differences in baseline eGFR among the high CKD-risk individuals receiving either placebo (n = 64) or linagliptin (n = 70, median eGFR placebo: 66.5 mL/min/1.73 m² [IR: 56–95] vs median eGFR linagliptin: 82.5 mL/min/1.73 m² [IR: 61–114], p = 0.165). Notably, after 24 weeks, we observed a trend toward preservation of eGFR for the linagliptin treated high CKD-risk group (median GFR: 90 mL/min/1.73 m², IR: 56–109) as compared to the high CKD-risk placebo group (median GFR: 65 mL/min/1.73 m², IR: 51–97; p = 0.0563; Figure 2B). A paired analysis of the eGFR values before and after placebo treatment resulted in a significant (p = 0.035) decrease of the renal function at study end. The observed increase of the eGFR values after linagliptin treatment at study end was not significant from matching baseline the GFR values. There again, eGFR changes did not differ between linagliptin and placebo arms for the low CKD-risk group before (median eGFR placebo: 101 mL/min/1.73 m², IR: 82–121 vs median eGFR linagliptin: 106 mL/min/1.73 m², IR: 82–140, p = 0.436) and after active treatment (median eGFR placebo: 99 mL/min/1.73 m², IR: 77–129 vs median eGFR linagliptin: 99 mL/min/1.73 m², IR: 78–126, p = 0.7915), respectively (Figure 2C). In the case of albuminuria, no differences between the linagliptin and placebo-treated group were observed before and after the treatment in the whole, low and even high CKD-risk group. Not even a trend toward significance (see Figure S1, Supporting Information) was detected.

Since it was expected that linagliptin would alter the CKD273 score, we further assessed the correlation between the CKD273 score at week 24 with eGFR values at week 24 (i.e., end of study correlation). Consistent with results at baseline, we found a significant correlation between CKD273 and eGFR for the linagliptin and placebo groups (rho = −0.360, p < 0.0001; and rho = −0.259, p = 0.0012, respectively).

3.4 Linagliptin Treatment and Changes in Urinary Proteome Profiles

To test our second main hypothesis, we compared the UPP of n = 163 and n = 153 individuals treated with either linagliptin or placebo, respectively after 24 weeks. Comprehensive UPP of the two treatment groups at baseline and after 24 weeks, respectively, are shown in Figure 3A.

As expected, no significant changes in UPP before and after treatment were observed for the placebo group. In contrast, the analysis of the UPP of the linagliptin group resulted in the identification of 993 peptides significantly changed after 24 weeks of treatment. We were able to identify the amino acid sequence for 314 of these peptides (Figure 3B). For more in-depth investigations, we selected the top 30% of sequenced altered peptides (n = 95 peptides, defined by lowest p-values) as shown in Table S2, Supporting Information, and further depicted in Figure 3C. The majority of these peptides were fragments of different collagens (n = 77) and 86% were found to be increased after 24 weeks of linagliptin treatment. It is expected that inhibition of an abundant protease in proximal tubular cells, such as DPP-4, responsible for cleavage of various peptides, significantly alters UPP. DPP-4 cleaves X-proline dipeptides from the N-terminus of polypeptides and we thus investigated if the altered UPP after linagliptin would be reminiscent of this proteolytic pattern; 53 of the 95 peptides had a DPP-4 cleavage site (i.e., proline in the penultimate position), and are thus expected to increase after linagliptin treatment. Moreover, peptide levels with an N-terminal X-proline-containing sequence were also elevated, whereas peptides that may have resulted from the cleavage of those N-terminal amino acid sequences were decreased. As an example for the above described and proline-dependent alterations in peptides after linagliptin treatment, collagen alpha-1 (I) fragments as shown in Figure 4.

3.5 Correlation of Urinary DPP-4 Peptidase Activity with Urinary Peptide Abundance

To test for our third hypothesis we correlated urinary DPP-4 peptidase activity after linagliptin treatment with the abundance (signal intensity) of all 314 linagliptin-affected sequenced urinary peptides at baseline (i.e., before drug exposure). We identified 20 peptides showing significant correlations (p < 0.05) with urinary DPP-4 activity, thereby indicative of dynamic changes after linagliptin treatment. These peptides are listed in Table S3, Supporting Information. Two specific examples of peptides depicting either an increase or decrease after linagliptin treatment are further shown in Figure 5. The majority of the above identified peptides showed a negative correlation with urinary DPP-4 activity (n = 17, 85%).