Urine peptidome analysis in cardiorenal syndrome reflects molecular processes

The cardiorenal syndrome (CRS) is defined as the confluence of heart-kidney dysfunction. This study investigates the molecular differences at the level of the urinary peptidome between CRS patients and controls and their association to disease pathophysiology. The urinary peptidome of CRS patients (n = 353) was matched for age and sex with controls (n = 356) at a 1:1 ratio. Changes in the CRS peptidome versus controls were identified after applying the Mann–Whitney test, followed by correction for multiple testing. Proteasix tool was applied to investigate predicted proteases involved in CRS-associated peptide generation. Overall, 559 differentially excreted urinary peptides were associated with CRS patients. Of these, 193 peptides were specifically found in CRS when comparing with heart failure and chronic kidney disease urinary peptide profiles. Proteasix predicted 18 proteases involved in > 1% of proteolytic cleavage events including multiple forms of MMPs, proprotein convertases, cathepsins and kallikrein 4. Forty-four percent of the cleavage events were produced by 3 proteases including MMP13, MMP9 and MMP2. Pathway enrichment analysis supported that ECM-related pathways, fibrosis and inflammation were represented. Collectively, our study describes the changes in urinary peptides of CRS patients and potential proteases involved in their generation, laying the basis for further validation.

The cardiorenal syndrome (CRS) is a complex pathological disorder, which reflects the interplay between heart and chronic kidney diseases 1 . Epidemiologic data reveal that about 45-63% of chronic heart failure (HF) patients developed chronic kidney disease (CKD) 2 . CRS is classified into five subtypes, referring to primary organ dysfunction, each with different underlying pathological mechanisms 3 . The term "CRS" has been applied to the relation of these two diseases, but the definition and classification have not been clearly established on a molecular pathophysiologic level until now 4 . However, multi-factorial mechanisms including hemodynamic changes, fibrosis, vascular calcification, neurohormonal activity, immunologic imbalance, inflammation, apoptosis, endothelial injury, thrombosis, and oxidative stress have been proposed to explain the general pathophysiology of CRS 5 .
Urinary peptidomics focuses on the analysis of naturally occurring peptides and small proteins in urine. Application of the approach enabled identification of biomarkers associated with kidney and heart diseases including, among others, CKD 15 , kidney fibrosis 16 , diabetic kidney disease (DKD) 17 , acute kidney injury (AKI) 18 as well as HF 19 , asymptomatic LV diastolic dysfunction (LVDD) 20 , and ischemic and dilated cardiomyopathy 21 ; with a biomarker panel for CKD also having reached the level of clinical implementation for patient stratification in large clinical trials 22 .
Due to the lack of specific CRS biomarkers on a molecular level, we designed the current study to identify molecular differences at the level of the urinary peptidome between CRS and matched controls, prompted by the Urine peptidome analysis. A total of 3184 peptides were detected and considered when comparing CRS patients with controls. After applying the Benjamini-Hochberg (BH) adjustment, 1480 differentially excreted peptides (BH, P < 0.05) were identified when comparing cases with controls. To shortlist the more representative features, a 30% frequency threshold, frequently used in such analyses 23,24 was applied, which resulted in a final list of 559 differentially excreted peptides (see Fig. 1a and Supplementary Table S1). These peptides derived from 110 unique protein precursors, which based on functional annotation, reflected largely extracellular matrix (ECM) changes, cell-ECM interactions, collagen formation/degradation, metabolic and inflammatory processes. The top 15 most frequently observed protein precursors, associated with at least 400 differentially excreted peptides are presented in Fig. 1b. Of these 559 differentially excreted peptides, 313 (55.9%) originated from fibril-forming collagens and were fragments of collagen type I alpha 1 chain (COL1A1), collagen type I alpha 2 chain (COL1A2), and collagen type III alpha 1 chain (COL3A1). Together the collagens accounted for 72% of the quantified peptidome. The 20 most significant differentially excreted peptides included fragments of additional collagen types (COL4A1, COL4A3, COL9A3, COL5A1, COL5A2, COL5A3 and COL19A1) and other peptides originating from plasma proteins (such as apolipoprotein A1 (APOA1) and B2M) (see Table 2). Among the peptides showing the most prominent increase in abundance in patients versus controls were fragments of B2M, albumin, and alpha-1-antitrypsin (A1AT), whereas fragments of collagen types (IV, V and VI), clusterin (CLU) and ubiquitin-associated protein 1-like (UBAP1L) were the peptides with the most decreased abundance levels in CRS (see Fig. 1c). Urinary peptide differences in CRS, CKD and HF. We further investigated if common peptides were identified between patients with HF and CRS as well as between CKD and CRS. We compared the 559 peptides NT-proBNP: N-terminal pro b-type natriuretic peptide; BP: blood pressure; HF: heart failure; HFrEF: heart failure with reduced ejection fraction; HFpEF: heart failure with preserved ejection fraction; HFmrEF: heart failure with mid-range ejection fraction; N/A: not applicable. www.nature.com/scientificreports/ of our study with the 218 and 577 differentially excreted peptides which were associated with CKD and HF, respectively, and described in previous studies 15,19 . Of these 218 CKD-associated peptides, 31 (14.2%) were commonly identified in our study as well. These common 31 peptides were sequenced fragments of albumin, B2M, A1AT, APOA1, alpha-1B-glycoprotein (A1BG), alpha-2-HS-glycoprotein (AHSG), sodium/potassium-transporting ATPase subunit γ (FXYD2), osteopontin (SPP1), collagen types I and III and other proteins (see Supplementary Table S2). In addition, 341 out of 577 (59%) of the HF-specific peptides were commonly identified in CRS patients (see Supplementary Table S3). These shared peptides showed the same directionality of difference and originated mostly from collagen types (I, II, III, IV, V, IX, XI), B2M, A1AT and uromodulin (UMOD). Overall, 193 of 559 (34.5%) CRS-associated peptides were not among those identified in the HF and CKD studies (see Supplementary Table S4). Most of these 193 peptides, which were found in the CRS cohort only, were collagen fragments (n = 134, 69.4%), with collagen types I, II, III, IV and V represented by the largest number of collagen fragments (n = 117, 60.6%). The top most represented protein precursors were collagen types I (40.9%) and III (9.3%), UMOD (4.1%) fibrinogen alpha chain (FGA, 2.6%), COL5A2 (2.1%), CD99 (1.6%), CLU (1.6%), FXYD2 (1.6%) and polymeric immunoglobulin receptor (PIGR, 1.6%).
On a protein level, we further investigated if any of the protein precursors represented by 193 CRS-associated peptides was not identified in the CKD and HF urinary profiles. Totally, 30 protein precursors were detected only in CRS cohort but not in HF and CKD. These protein precursors included several collagen types (COL4A2, COL4A4, COL6A5, COL7A1, COL8A1 and COL13A1), PIGR, secreted and transmembrane protein 1 (SECTM1), ankyrin repeat domain-containing protein 17 (ANR17), ubiquitin-like protein ATG12 (ATG12),   Pathway enrichment analysis of the shortlisted proteases along with the protein precursors of the differentially excreted peptides was performed by Metascape. The analysis revealed that ECM-related pathways (i.e. ECM organization and degradation) were significantly affected. Together, cathepsins K and S along with various metalloproteinases (i.e. MMP1, MMP2, MMP3, MMP8, MMP9, MMP12, MMP13, MMP14, MMP20 and MMP26) and KLK4 were mostly involved in the structure, organization and degradation of ECM. Additionally, pathway analysis highlighted that the predicted proteases along with the protein precursors were also inolved in collagen formation, protein processing, regulation of inflammatory response and neutrophil mediated immunity (see Fig. 3).

Discussion
In the present case-control matching study, a large-scale urine peptidome analysis of patients with CRS and matched controls was performed, targeting to identify the disease-specific urine peptidome profile and its potential links to pathophysiology. Importantly, our study identifies a high number of differentially excreted urine peptides between CRS patients and matched controls.
The results of the current study suggest that the urine peptidome of CRS patients integrates changes of 559 peptides originated from various protein precursors. These peptides originate to a large extent from proteins involved in ECM, collagen formation/degradation, inflammation, metabolism but also transcriptional regulation  25 . A large number of overlapping urine peptides between CRS and HF as well as CRS and CKD patients were expected, as the examined patients of our study have combined both pathologies. These overlapping peptides derived, among others, from plasma proteins which are well-known renal and heart failure biomarkers including albumin, B2M, SPP1, AHSG, but also uromodulin and collagens 26 .
In addition, to the above overlapping protein precursors, two proteins including, APOA1 (a protein which is associated with renal dysfunction in HF patients 27 ) and FXYD2 (a protein which mediates the function of the Na, K-ATPase in mammalian kidney epithelial cells 28 ) are commonly identified in CRS and CKD, similarly reflecting the common underlying molecular profiles of these diseases. The highly abundant COL1A1, COL1A2, and COL3A1 proteins were represented by multiple urinary fragments, reflecting most likely fibrosis and changes in the ECM in both organs. In a healthy heart, the ECM is more than 95% composed of the fibrillar profibrotic collagens, including collagen types I and III 29 , important for maintaining the tensile strength, elasticity and extensibility of the myocardium 29 . Alterations of collagen type I and III expressions have been associated with myocardial fibrosis in HF patients 30 and vascular calcification in the heart 31 in animal studies. Similarly, collagen types I and III are the most abundant collagen types in the renal ECM 32 linked to ECM dysregulation and CKD progression 33 . In addition, the activity of proteases may also be linked with the high number of the detected collagen-related peptides as it is well reported that metalloproteinases regulate cardiac and renal remodeling as well as fibrosis by facilitating ECM turnover and inflammatory cells 34,35 .
Further expected changes based on current knowledge were consistently detected. Among the most pronounced was the selective enrichment of increased albumin peptides in CRS, as expected for renal dysfunction and consequently albuminuria. Similarly, we observed an enrichment of peptides derived from COL9A3 and COL5A3, which are selectively highly expressed at the protein level in the heart and not detected in the kidneys www.nature.com/scientificreports/ based on proteomics databases (ProteomicsDB). Along the same lines, the abundance of a matrix GLA peptide (MGP) was strongly increased in CRS patients. MGP was already described as a calcification inhibitor protein with a strong association to HF indices and mortality 36 .
Similarly, in line with the literature, among protein precursors represented by decreased fragments in CRS patients were COL4A1 and COL4A3, collagens which are associated with nephropathy 37 . Furthermore, circulating proteins portrayed by increased fragments in CRS included AHSG (a protein which is associated with vascular calcification, cardiovascular mortality and kidney dysfunction 38 ), APOA1 (a protein which is associated with HF and kidney dysfunction 27 ), B2M (a protein which is associated with cardio-renal remodeling 11 and inflammation 39 ), COL18A1 (endostatin, generated from COL18A1, is associated with the development of cardiovascular events in CKD patients 40 ), MGP and thymosin beta-4 (TMSB4X, a protein which is associated with renal fibrosis 41 ).
Protein precursors detected only in CRS cohort but not in HF and CKD were also investigated. These included several collagen types, including COL4A2, COL4A4, COL6A5. Although few studies have aimed to investigate the role of collagen type IV and VI in cardio and renal failure, recently, it was suggested that COL4A2 is associated with cardiac fibrosis phenotype 42 , as well as with glomerular basement membrane alterations 43 , suggesting that COL4A2 may play an important role in CRS. In addition, COL4A4 and COL6A5 are associated with CKD 44,45 but no evidence was found to link COL4A4 and COL6A5 with HF. Interestingly, two additional proteins, ROBO1 and HUWE1 are linked with cardiac and renal fibrosis 46,47 whereas ANR17 protein may play a key role in the formation and maintenance of the blood vessels 48 . Moreover, several proteins associated with the immune system were uniquely found in the CRS cohort. These proteins were CD14 (a protein which is associated with heart and renal dysfunction 49 ), CD99 and IRF6, suggesting the importance of the immune response in CRS. In addition, urinary PIGR peptides have been previously associated with cardiorenal dysfunction 50 , nevertheless, the exact role of the protein in the disease pathology remains unknown.
Our protease prediction analysis suggested the activity of 18 proteases responsible for more than 1% of the cleavage events. These predictions can be mostly supported by the existing bibliography on cardiovascular and renal pathologies. Notably, 9 out of the 11 predicted metalloproteases (MMP13, MMP9, MMP2, MMP20, MMP3, MMP12, MMP14, MMP8 and MMP1) were previously positively correlated with HF and were increased after myocardial infarction 34,51 . Interestingly, the transcriptional activation of MMP13 induces the vascular smooth muscle cell (VSMC) apoptosis and ECM breakdown via the FOXO3a activation 52 . These observations can be reflected by the contribution of these metalloproteases to vascular and kidney damage. Specifically, MMP2, MMP3, MMP8, MMP9 and MMP12 are involved in the ECM deposition in the glomeruli; MMP2, MMP3, MMP9, MMP13 and MMP14 induce epithelial-to-mesenchymal transition that leads to kidney fibrosis; and MMP2, MMP3 and MMP9 are correlated with vascular calcification, arterial stiffening and atherogenesis 53 . As reflected also by the cleavage events in our results, MMP2 and MMP9 have a prominent role in these processes   54,55 . The role of MEP1B and MMP26 predicted by our analysis to mediate a large number of cleavage events, yet not previously linked to renal or heart dysfunction, apparently merits further investigation in the context of CRS. From the 4 predicted proprotein convertases, three PCSK6 PSCK5 and PCSK7, have been previously associated with heart failure, to our knowledge. PCSK6 converts procorin to corin which in turn can activate natriuretic peptides, regulating cardiovascular and renal function 56 . Interestingly, PCSK6 is a key regulator of smooth muscle cell function (SMCs) in vascular remodeling and a novel player in cardiac remodeling after myocardial infraction 57,58 . Additionally, the inactivation of PCSK6 along with PCSK5 in endothelial cells leads to decreased collagen deposition and cardiovascular hypotrophy via IGF-1/Akt/mTOR signaling 59 . Moreover, PCSK7 is associated with both cardiovascular disease (CVD) 60 and end-stage kidney disease 61 . However, the role of PCSK4 and KLK4 has not yet been linked to HF or CKD and would merit further investigation in the context of CRS based on our results.
Regarding cathepsins, elevated levels of CTSK were correlated with the presence of chronic HF 62 and with major adverse cardiac and cerebrovascular events in CKD patients 63 , while circulating CTSS levels increase with CKD progression and GFR decline 64 . Levels of the mRNA, protein and activity of CTSS were found increased in the left ventricular myocardium of humans and rats with HF compared with controls, suggesting its participation in cardiac remodeling 65 . CTSS was also shown to affect epithelial-to-mesenchymal transition and ECM deposition in mouse models of mild and severe hydronephrosis, indicating its role in the regulation of renal fibrosis 66 .
Collectively, the pathology of CRS is complex and a number of pathways including ECM-related, fibrosis and inflammation are involved. Urinary peptidomics analysis reflects such CRS-associated changes, occasionally overlapping, as expected, with changes earlier observed in HF and CKD; but also alterations such as collagen type IV (COL4A2 and COL4A4), type VI (COL6A5), HUWE1, CD14, ANR17, PIGR and ROBO1 as well as a number of predicted proteases including MEP1B, MMP26, PCSK4 and KLK4, meriting further investigation in the context of CRS.
Among the strengths of our study was the large sample size allowing for meaningful patient matching. Based on this large sample size as well as on the case-control matching, we are confident in data validity and reliability. Additionally, all urinary peptides were analyzed using the same analytical platform and protocols (CE-MS). In contrast to any MS/MS approach, (CE-)MS does not provide sequence information. However, also as a result of the excellent reproducibility, the dataspace can be well defined, which results in the identification of over 4000 urine peptides and assignment of sequence based on mass and migration time with very high confidence 67  The study has some limitations: these include that it is retrospective, based on already available published data. However, the sample size, the multicenter design and the very high consistency and significance of the observed changes reduce the risk of this bias. Further, we limited our study to a subgroup of CRS: to subjects with CKD and HF. This is owed to the fact that we aimed towards depicting molecular changes. Such an approach requires homogeneity in molecular pathophysiology. In addition, a number of known kidney-specific and heart-specific protein biomarkers (such as KIM-1, L-FABP, NGAL, and NAG) 14 could not be detected in our study; this is in fact expected as the applied technique (CE-MS) resolves the peptidome (< 10 kDa molecular mass peptides). Finally, the analysis is descriptive which, however, still opens multiple research avenues towards understanding the functional impact from the generation of the presented significant fragments.
In conclusion, this study reports the detection of a high number of urinary CRS-associated peptides when compared with controls. The underlying molecular mechanisms for the CRS pathology, as reflected by these peptides, represent fibrosis, ECM-related pathways, collagens formation/degradation and inflammation, in line with the existent knowledge. However, a number of peptides/protein precursors, not highlighted previously in association to HF or CKD, such as peptides of PIGR, CD14, ANR17, COL4A2 COL4A4, COL6A5, ROBO1 and HUWE1 may be important players in the mechanisms of CRS. As the present work is the first attempt to explore the urine peptidome profile of CRS patients, the findings of our study require further validation.

Methods
Study design: patient data and selection. Urinary peptidomics datasets from patients with CRS (in this study, CRS patients simultaneously combine both pathologies; HF and CKD), as well as individuals with no signs of diseases (controls) at urine sampling were used. These datasets corresponded to urine samples from cohorts described in several published studies investigating mainly renal failure (including Syskid 15 , FSGS-Aachen 24 ), or heart and cardiovascular failure (including NTCVD-Urin 69 , FROG-ICU 70 , PCHF-Urin 71 and HOMAGE 72 ). The data were investigated for the availability of information on both kidney (i.e. CKD) and heart (i.e. HF) disease. The kidney function was assessed via the eGFR assessed based on ´Chronic Kidney Disease Epidemiology Collaboration´ (CKD-EPI). The ´European Society of Cardiology´ guidelines were used for the HF diagnosis and subtyping 73 . Clinical, pathophysiological, and molecular variables such as ejection fraction, systolic blood pressure, diastolic blood pressure, serum creatinine, hypertension, NT-proBNP levels, left ventricular ejection fraction (LVEF) as well as information on the age and sex, were retrieved (see Table 1). These datasets (corresponding to 3463 individuals) were then separated into the CRS cohort consisting of patients with both pathologies; HF and CKD with an eGFR < 60 ml/min/1.73 m 2 . The control group with no signs of heart disease and preserved kidney function (eGFR > 60 ml/min/1.73 m 2 ) was selected (see Table 1). The study had received ethics approval (ΕΚ163/19 Ethik-Commission of the medical faculty of the RWTH Aachen), fulfilling all the requirements on the protection of the individuals participating in medical research and in accord- www.nature.com/scientificreports/ ance with the principles of the Declaration of Helsinki 74 . All data sets received were anonymized. All experiments were performed in accordance with relevant named guidelines and regulations. Each patient has written informed consent to use part of the tissue for scientific research.
Case-control matching. CRS patients were matched on age and sex with controls at a 1:1 ratio. Following the case-control matching procedure, the final groups consisted of 353 patients with CRS compared with 356 individuals with no signs of either disease as listed in Table 1.

CE-MS.
The urine samples have been prepared and measured by CE-MS as stated before 75 . The P/ACE MDQ capillary electrophoresis system (Beckman Coulter, USA) connected to a micro-TOF-MS (Bruker Daltonic, Germany) was used for the CE-MS analysis. The probabilistic clustering algorithm along with isotopic distribution and conjugated masses for charge have been used for RAW MS data evaluation as described previously 67 .
Totally, twenty-nine fragments of collagens that were not affected by disease were used for the normalization of the CE-MS data. Proteases analysis. The Proteasix (http:// www. prote asix. org), an open-source tool was used for the protease prediction analysis 81 . As such, the potentially involved proteases were linked with the generation of the identified CRS-associated peptides. In brief, Proteasix uses information about the protease/cleavage site associations from a number of protease databases including the MEROPS, the CutDB, the UniProt Knowledgebase and the literature. The generated list of proteases is divided into two types; (a) "predicted" and (b) "observed". For the "observed" proteases the protease/cleavage site association is collected from the literature, whereas for the "predicted" proteases the predicted proteolysis is determined by the MEROPS database. To improve the reliability of the proteolytic data, we decided to focus only on the "observed" proteases.
Statistical analysis. The Kolmogorov-Smirnov normality test was used to determine the distribution of the urine peptidome data. Statistical analysis of the abundance of urinary peptides was performed using the nonparametric Mann-Whitney U test, followed by correction for multiple testing using the Benjamini-Hochberg (BH) method. Statistical analysis was performed using SPSS software version 20.0 (SPSS, Inc., Chicago, Illinois). A BH-adjusted P-value < 0.05 was considered to be statistically significant. The abundance of urinary peptides was analyzed and plotted using GraphPad Prism 7 (GraphPad Software, La Jolla, California, USA). Data are presented as mean ± standard deviation (s.d.) (*P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).

Data availability
All data generated or analysed during this study are included in this published article (and its Supplementary Information files).

Funding
Open Access funding enabled and organized by Projekt DEAL.