Urinary peptidomics in a rodent model of diabetic nephropathy highlights epidermal

growth factor as a biomarker for renal deterioration in patients with type 2 diabetes

1,2Boris B. Betz, 3Sara J. Jenks, 4Andrew D Cronshaw, 5Douglas J. Lamont, 6Carolynn Cairns, 7Jonathan

R. Manning, 8Jane Goddard, 9David Webb, 10John J. Mullins, 1,8Jeremy Hughes, 3Stella McLachlan,

10Mark W.J. Strachan, 3Jackie F. Price, 6,8Bryan R. Conway

1Centre for Inflammation Research, University of Edinburgh

2Institute of Clinical Chemistry and Laboratory Medicine, Jena University Hospital

3Centre for Population Health Sciences, University of Edinburgh

4Proteomics Facility, University of Edinburgh

5Proteomics Facility, College of Life Sciences, University of Dundee

6Centre for Cardiovascular Science, University of Edinburgh

7Centre for Regenerative Medicine, University of Edinburgh

8Department of Renal Medicine, Royal Infirmary, Edinburgh

9Clinical Pharmacology Unit, University of Edinburgh

10Metabolic Unit, Western General Hospital, Edinburgh

Corresponding author

Bryan Conway

Room W3.06,

Centre for Cardiovascular Science,

Queen’s Medical Research Institute,

University of Edinburgh

EH16 4TJ

E-mail: bryan.conway@ed.ac.uk

Tel: +44(0)131 2426691; FAX: +44(0)131 2426779

Word count: 3860 Abstract: 1385 characters

Abstract

Many diabetic patients suffer from declining renal function without developing albuminuria. To

identify alternative biomarkers for diabetic nephropathy (DN) we performed urinary peptidomic

analysis in a rodent model in which hyperglycaemia and hypertension synergise to promote renal

pathological changes consistent with human DN. We identified 297 increased and 15 decreased

peptides in the urine of DN rats compared with controls, including peptides derived from proteins

associated with DN and novel candidate biomarkers. We confirmed by ELISA that one of the parent

proteins, urinary epidermal growth factor (uEGF), was more than x2-fold reduced in DN rats in

comparison with controls. To assess the clinical utility of urinary EGF we examined renal outcomes in

642 participants from the Edinburgh Type 2 Diabetes Study (ET2DS) who were normoalbuminuric

and had preserved renal function at baseline. After adjustment for established renal risk factors, a

lower uEGF:creatinine ratio was associated with new-onset eGFR <60 ml/min/1.73m2 (OR 0.48;

95%CI [0.26-0.90]), rapid (>5% per annum) decline in renal function (OR 0.44; 95%CI [0.27-0.72]) or

the composite of both outcomes (OR 0.38; 95%CI [0.24-0.62]). The utility of low uEGF:creatinine

ratio as a biomarker of progressive decline in renal function in normoalbuminuric patients should be

assessed in additional populations.

Keywords: peptidomics; diabetic nephropathy; epidermal growth factor

Introduction

Diabetic Nephropathy (DN) remains the single-most common cause of end-stage renal disease

(ESRD) in the western world 1. Furthermore, the development of nephropathy in patients with

diabetes confers an increased risk of mortality 2. Albuminuria is an established biomarker in DN,

indicating increased risk for progressive renal failure and mortality 3, 4, however not all patients with

albuminuria develop renal impairment. Conversely, not all patients with impaired renal function

have concomitant albuminuria, particularly those with type 2 diabetes. For example, in the

Edinburgh Type 2 Diabetes Study (ET2DS), a representative cohort of patients with type 2 diabetes

aged 60-75 years in the Lothian region of Scotland, of the 24% of patients with estimated glomerular

filtration rate (eGFR) <60ml/min/1.73m2, 74% were normoalbuminuric 5. Considering these

shortcomings and the prevalence of DN world-wide, there is an urgent need for reliable markers that

can predict the risk of renal failure in patients with diabetes, thereby enabling individualisation of

therapy.

Rodent models can be a useful source of biomarkers for human disease. For example, kidney injury

molecule-1 (KIM-1) was first identified by non-biased screening of a rodent model of ischemia-

reperfusion injury and subsequently confirmed in human disease 6. However, classical rodent models

of DN display only the earliest stages of human disease 7and consequently have limited translational

potential for the detection of biomarkers 8. To overcome this limitation, our lab recently developed a

rodent model in which induction of hyperglycaemia and renin-dependent hypertension in

Cyp1a1mRen2 rats synergise to promote transcriptomic and pathological changes characteristic of

moderately advanced human DN 5.

In the current study, we firstly assessed whether profiling of the urinary peptidome in the

Cyp1a1mRen2 model could identify candidate biomarkers of DN. Secondly, we wished to assess

whether one of the candidates identified by this approach, urinary epidermal growth factor (uEGF)

that has been recently demonstrated to be a predictive marker for progression of CKD 9, could be

used to identify patients with type 2 diabetes and normoalbuminuria who were at risk of rapid loss

of renal function and incident stage III chronic kidney disease (CKD).

Results

Urinary peptidomics identifies potential markers of DN

To identify potential biomarkers of DN in a rodent model, we performed peptidomic analysis on

urine from control rats and from rats following induction of hyperglycaemia alone (DM),

hypertension alone (HTN) or a combination of both hyperglycaemia and hypertension (DM+HTN) for

28 weeks. The pathophysiological features of the model have been published previously 5. In brief,

renin-dependent hypertension and hyperglycaemia synergized to produce massive albuminuria

(Table 1) and marked glomerulosclerosis and tubulointerstitial fibrosis (Figure 1a and 1b)5.

Virtual principal components (PC) were calculated by the program Progensis QI from all detected

peptides to assess quality control and identify outliers. PC analysis of the urinary peptidomics

demonstrated clustering of samples within each group with urine from DM+HTN animals segregating

clearly from normal controls and from animals that were hypertensive or diabetic alone (Figure 2).

This was consistent with our previous study where a combination of hyperglycaemia and

hypertension was required to produce robust changes in renal gene expression that were

representative of those observed in the kidneys of patients with DN 5. Hence for further analysis, we

focused on the differences in the peptidome in the DM+HTN animals compared with controls.

Altogether 1165 urinary peptide sequences were identified with 1142 peptides detected in both the

control and DM+HTN animals. Volcano plot analysis demonstrated that the majority of peptides

exhibited increased abundance in the HTN+DM group compared with controls, of which 297

peptides were significantly increased (p<0.01, Figure 3). Only 15 peptides were of significantly lower

abundance in comparison to the control group. Table 2 lists key peptide sequences that were

increased (Table 2a) or decreased (Table 2b) including the corresponding protein of origin. The

complete list of peptide sequences detected can be found in the supplementary table 1. The list of

peptide sequences increased in the urine of DM+HTN animals included many serum proteins

(Apolipoprotein, albumin, Plasminogen, alpha-2-HS-glycoprotein, haptoglobin), indicating that many

of the peptides are likely to be derived from increased passage across a damaged glomerular

filtration barrier. However, other molecules detected in the peptidomic analysis were also found to

be similarly up-regulated (complement C3, osteopontin, collagen type 1 alpha-1 chain, and clusterin)

or down-regulated (epidermal growth factor) at a gene expression level by microarray analysis of the

kidneys of DM+HTN rats 5 or in patients with DN 10 11 (www.nephromine.org) compared with

respective controls (annotated in Table 2).

Reduced renal EGF expression and urinary EGF concentration in DM+HTN animals

We elected to focus on urinary epidermal growth factor (uEGF) for further investigation as it was

uniquely the only down-regulated peptide in the peptidomic analysis that exhibited corresponding

reductions in renal gene expression on microarray analysis in both the rodent model5 (confirmed by

real-time PCR, Figure 4A) and in human disease10. ELISA analysis determined that a significant

reduction in urinary EGF excretion was apparent by eight weeks after onset of diabetes and

hypertension (Figure 4B). Immunohistochemistry demonstrated a focal reduction in EGF protein

expression specifically in injured (kidney injury molecule-1+) and atrophic tubules in DM+HTN rats

(Figure 5).

uEGF correlates with baseline renal function

To determine whether urinary EGF could act as a biomarker in patients with type 2 diabetes who

have no clinical evidence of renal dysfunction, we used urine specimens from 642 (out of 1076, 60%)

participants from the Edinburgh Type 2 Diabetes Study who had preserved renal function (eGFR > 60

ml/min per 1.73m2) and were normoalbuminuric (albumin:creatinine ratio (ACR)<2.5 mg/mmol in

males and 3.5 mg/mmol in females) at baseline. The baseline characteristics of these subjects

stratified by uEGF tertiles are supplied in table 3. Renal function correlated significantly, but weakly

with uEGF:creatinine ratio (uEGF/Cr, r=0.11, p<0.01) while a greater proportion of patients in the

highest tertile of uEGF/Cr had ACR >0.5 mg/mmol. To assess whether these associations were also

observed in other renal diseases, we quantified uEGF/Cr in a separate cohort of non-diabetic

patients (n=95) who had a wide range of eGFR and albuminuria (patient characteristics summarised

in Suppl Table 2)12 and determined that uEGF/Cr ratio was positively correlated with eGFR (r=0.71,

p<0.001) and inversely correlated with albuminuria (r=-0.51, p<0.001, Suppl Figure 2).

uEGF is associated with incident eGFR<60ml/min/1.73m2 or rapid decline of renal function

During the follow-up period in the ET2DS, 133 patients (21%) exhibited a rapid decline in eGFR (>5%

per annum) and 91 of the 642 patients (14%) developed eGFR<60ml/min/1.73m2. Altogether 161

(25%) participants developed either outcome. When we stratified patients according to whether or

not they developed the composite endpoint, there were significant differences in the baseline

characteristics between the groups including renal function (GFR), age, sex, systolic blood pressure

duration of diabetes, HbA1c, ACR and uEGF. (Table 4). Results were similar when stratification was

performed for either endpoint alone (Supplement Tables 3 and 4).

Kaplan-Meier analysis demonstrated that patients in the tertile with lowest uEGF concentration at

baseline had a significantly higher risk (p=0.003) of developing an eGFR<60ml/min/1.73m 2 during

the follow-period while an ACR was not a useful prognostic indicator (Figure 6).

Logistic regression was used to investigate the association between Log(uEGF/Cr) ratio and

progression of renal disease. The unadjusted odds ratios of Log(uEGF/Cr) for incident

eGFR<60ml/min/1.73m2, rapid decline in renal function or a combination of either endpoint were

0.48 (95% confidence interval [CI], 0.30–0.80), 0.54 (95% CI, 0.35–0.84) and 0.45 (95% CI, 0.30–0.68)

respectively (Table 5). These associations remained significant following adjustment for baseline

characteristics (eGFR, age, gender, ACR(categorized), HbA1c, duration of diabetes, systolic and

diastolic blood pressure).

We analysed the area under the receiver operating curve (AUC-ROC) for the individual and combined

renal outcomes using a panel of established risk factors (reference panel) including: age, gender,

eGFR, ACR, duration of diabetes, HbA1c, systolic and diastolic blood pressure (Table 6). The addition

of uEGF/Cr ratio resulted in a small increase in AUC-ROC compared to the reference panel (Table 6),

which did not reach significance (p=0.07 for combined end-point).

Discussion

In the current study, we characterised the urinary peptidome in an innovative rodent model of

moderately advanced DN to identify candidate disease biomarkers. We determined that whilst the

urinary excretion of many peptides was greater in diabetic and hypertensive animals compared with

controls, amongst the few peptides at lower concentration were those derived from EGF. In a large

cohort of patients with type 2 diabetes but without albuminuria or a reduction in renal function, low

urinary EGF concentration was associated with increased risk for rapid loss of renal function or

incident stage 3 CKD independent from classical risk factors.

To the authors’ knowledge there is only one other study addressing urinary low-molecular weight

proteomics/peptidomics in an animal model of DN. 8 Although this study employed Zucker diabetic

rats, there was a significant overlap in the identity of most parent proteins between this study and

our own, including collagen alpha-1 and alpha-2 chains, osteopontin, apolipoproteins, complement

component C3 and albumin. Explanations for the incomplete overlap of peptides/proteins detected

may include: differences in sample processing, mode of detection (LC-MS/MS vs capillary

electrophoresis coupled MS) and biological variation between the DN models.

A major advantage of using the Cyp1a1mRen2 rat model is that we have access to archived

microarray analysis of the gene expression profile in the kidneys of these rats. Hence we can

compare the urinary excretion of peptides of interest with the corresponding renal gene expression.

When altered urinary excretion of these peptides is concordant with altered renal expression of the

corresponding genes, it may, at least in part, be that this reflects altered production from the injured

kidney per se and may be informative regarding the activity of intrinsic renal pathological pathways.

We focused on EGF for further investigation for a number of reasons. Firstly, we hypothesised that

biomarkers that are down-regulated may be more likely to reflect intrinsic renal pathophysiology, as

up-regulated peptides are more likely to be due to increased passage across a damaged glomerular

filtration barrier, and hence may not add significant prognostic value to albuminuria. In support of

this hypothesis, we observed a concordant reduction in EGF gene expression in the kidneys of

diabetic and hypertensive Cyp1a1mRen2 rats compared with controls, as has previously been

reported for diabetic mRen-2 rats 13. This is consistent with the focal reduction in EGF expression in

atrophic and injured (KIM-1+) tubules that we observed in the diabetic and hypertensive

Cyp1a1mRen2 rats. This suggests that uEGF may be a biomarker of healthy tubular mass and indeed

EGF may be important in maintaining tubular cell health10. Secondly, consistent with the results from

Cyp1a1mRen2 animal model and other DN models8 or rodent models of reduced renal mass14, a

reduction in renal gene and protein EGF expression has also been observed in the renal

tubulointerstitium in patients with DN 10 11 9

We found a weakly positive correlation of uEGF/Cr ratio and baseline eGFR in patients with

preserved renal function in the ET2DS, and a much stronger correlation in a separate cohort of

patients with non-diabetic CKD across a broader range of renal function, and these findings are in

keeping with the results of previous studies 9, 15-20. uEGF/Cr ratio has recently been demonstrated to

predict rapid decline of renal function and ESRD in three independent cohorts of patients with

moderately advanced CKD9, however these cohorts included only a minority of patients with DN. We

elected to study more than six hundred patients with type 2 diabetes from the EdT2DS who had

preserved renal function (eGFR>60 mL/min/1.73 m2) and were normoalbuminuric at baseline as it is

particularly difficult to predict renal outcomes in this subgroup. Indeed, it is known that

albuminuria and decline in renal function can be uncoupled phenotypes in T2D21, therefore

additional prognostic biomarkers are required.

We performed Kaplan Meier analysis and determined that those in the lowest tertile of uEGF/Cr

ratio were at higher risk of incident eGFR<60ml/min/1.73m2.Furthermore, in binary logistic

regression analysis we determined that the uEGF/Cr ratio was associated with the combined

outcome of incident CKD or rapid decline in renal function. However, the baseline characteristics of

those who progressed were significantly different to non-progressors. Indeed, the lower baseline

eGFR in progressors raises the possibility that the association between uEGF/Cr ratio and renal

outcome may be explained by the weak, but significant correlation between uEGF/Cr and baseline

eGFR. Therefore we performed multivariate analysis and determined that the association between

uEGF/Cr ratio and renal outcome remained significant even after adjustment for age, sex, HbA1c,

systolic/diastolic blood pressure, albuminuria and eGFR. The addition of uEGF/Cr ratio to a panel of

established risk factors resulted in a very modest increase in the ROC-AUC, however it has been

proposed that this method can be insensitive when adding a new predictor to a model, even if such

a predictor has an independent and statistically significant contribution to a prediction model22.

This study has limitations. Biopsies are rarely performed in presumed DN therefore the aetiology of

the decline in renal function and incident CKD cannot be confirmed histologically. The ET2DS was

designed to be representative of patients with type 2 diabetes aged 60-75 in South-East Scotland;

the cohort comprises almost exclusively Caucasians and the results may not be generalizable to

other diabetic populations. Additionally, only a single measurement of uEGF was assessed. We

determined renal function based on eGFR values calculated using the EPI-CKD creatinine equation,

which is not as accurate as measurements using inulin clearance especially in patients with

preserved renal function 23. For the peptidomic study we used tryptic digestion and this may reduce

the comparability of our results with other studies, but we believed that this would enable us to

detect as many peptides as possible.

In conclusion, the study demonstrates that urinary peptidomics in an innovative rodent model of

combined hypertensive and diabetic renal damage provides an attractive source for biomarker

discovery. A low uEGF:creatinine ratio is associated with rapid decline in renal function and incident

CKD independent of standard risk factors. The validity of using uEGF levels to risk stratify patients

with diabetes who have no evidence of kidney disease should be assessed in other populations

including those with type 1 diabetes.

Materials and methods

Experimental setting

The Cyp1a1mRen2 model has been described in detail recently5. In summary, hyperglycaemia and

hypertension were induced in isolation or combination in adult male Cyp1a1ren2 rats by intravenous

administration of 20mg/kg streptozotocin and/or activation of the murine renin2 transgene by

dietary supplementation with 0.125% indole-3-Carbinol (I-3-C) to bind to the Cyp1a1 promoter. After

28 weeks, 24hr urine was collected in metabolic cages, spun at 500g for 10min to remove debris and

the supernatant was frozen at -80C until analysis. All procedures were performed under a UK Home

Office license and approved by the Ethics Committee at the University of Edinburgh.

Urine Sample Preparation and Mass Spectrometry Analysis

5% of the total 24hr urinary volume was placed in Millipore Amicon Ultra-4 spin concentrators

(Millipore, Billerica, MA) with a 10kD cut-off cellulose membrane and spun at 3220g for 15 minutes

at 4C. Ultrafiltrate containing peptides and low-molecular weight (LMW) proteins was acidified with

5% trifluoroacetic acid (TFA), loaded onto a C8 StageTip 24 and centrifuged at low speed (500-2000g)

for 1-5min, until no residual sample remained upstream of the tip. Acetonitrile (80% in H2O) was

used to elute the peptides by centrifugation. The effluent was dried to remove acetonitrile,

resuspended in 50mM ammonium bicarbonate and reduced in 5mM dithiothreitol at 60°C for 15

min. After cooling to room temperature samples were alkylated with iodoacetamide (15 mM) for 15

minutes in the dark. Proteolytic digestion was performed by incubating with trypsin (2.5ng/l)

overnight at 32°C. The rationale of using trypsin was our goal to detect the broadest spectrum of

possible biomarker candidates. We were concerned that in the absence of trypsin high molecular

weight peptides and small proteins at approximately 10kD might not fly and fragment in the

instrument and be difficult to detect by LC-MS/MS25. Directly prior to MS-Analysis, digested material

was loaded on a C18 StageTip and the final volume eluted in Acetonitrile and reduced by vacuum

evaporation 26.

A volume of 5ul of the sample was loaded onto an Ultimate 3000 Series HPLC (Dionex) with a PicoTip

Emitter (FS 360-100-8-N-20-C12, New Objective) in series with an LTQorbitrap mass spectrometer

(ThermoScientific). The PicoTip Emitter was packed with Reprosil-Pur C18-AQ 3um (Dr Maisch

GmbH) to a length of 6.5-7.5cm. The PicoTip column was equilibrated with solvent A (0.5% acetic

acid in 5% acetonitrile) and eluted with a linear gradient, from 0%B for 9min; from 0 to 20%B over 9

to 40min; from 20 to 80%B over 40 to 48min; solvent B (0.5% acetic acid in 99.5% acetonitrile), over

65min at a flow rate of 0.7ul/min for the first 9min and 0.3ul/min thereafter. Data dependent

acquisition was controlled by Xcalibur software (ThermoScientific).

Data analysis

Data analysis was performed using Progenesis QI software (version 4.0, Nonlinear Dynamics Limited,

Newcastle upon Tyne, UK) and the MASCOT search engine (Matrix Science, London, UK) 27. The raw

LC-MS/MS data were imported into Progenesis software and visually checked for defects. One of the

DM+HTN samples was selected as a reference and all other ion intensity maps were aligned to it

using the ‘Automatic Alignment’ function in the software with manual corrections applied where

necessary. After all the runs were aligned with the reference, the peptide ions with charge states 2+

to 5+ were selected. The sample files were transferred to the MASCOT search engine

(www.matrixscience.com – in-house installation on the server of the Fingerprints Proteomics

Facility, University of Dundee) for peptide identification. The search was performed against the

Rattus norvegicus Uniprot database, version 2013, which contains 37202 sequences.

The search was not restricted by tryptic cleavage in order not to lose natural peptides. Further

search criteria were: peptide tolerance of 6ppm, and the allowed modifications were

Carbamidomethyl, Dioxidation, Gln>pyro-Glu and Oxidation. Data were also searched against a

decoy database. The false discovery rate for matches above the identity threshold was 0.84% and for

matches above the identity or the homology threshold was 4.12%. MS/MS spectra of the top five

‘ranks’ for each feature were exported to the MASCOT search engine. Using the peptide score

distribution from the MASCOT search results, peptides below an ion score threshold of 50

(corresponding to p-value 0.05) were discarded. The filtered MASCOT search results were re-

imported into the Progenesis QI software. Only the peptides with at least three hits were retained

for further analysis. The differences in peptide intensity between sample groups were calculated by

the ANOVA test in Progenesis QI software. The threshold for the p-value was set at 0.01 and q-value

was calculated to control for false positive p-values. We used a q-value threshold of <0.05 implying

that 5% of significant tests will result in false positives. (see Suppl Fig 1 for flow diagram of analytic

approach).

Immunohistochemistry

Staining was performed on 4M paraformaldehyde fixed, paraffin-embedded rat kidney sections.

Staining for periodic acid-Schiff and picrosirius red was performed according to standard protocols.

After antigen retrieval (boiling for 15min in sodium citrate buffer), goat-anti EGF (1:25, Santa Cruz,

Texas, USA) or goat anti-KIM-1 (1:50, R+D Systems, Minneapolis, MN) primary antibodies were

incubated overnight at 4C followed by an anti-goat, biotinylated secondary antibody (1:300, Vector

Labs, Peterborough, UK) and developed using the Vector ABC kit and DAB.

Gene expression

RNA was extracted from the renal cortex of each animal using Nucleospin RNA II kit (Macherey-

Nagel, Duren, Germany), reverse transcribed (Applied Biosystems High Capacity cDNA Reverse

Transcription Kit, Life Technologies, Paisley, UK) and relative expression of EGF determined by real-

time PCR using inventoried Taqman gene expression assay kits (Life Technologies) for EGF

(Rn00563336) and the housekeeping gene (TATA-binding protein: Rn01455648_m1).

ELISA

For rodent and human specimens, we used enzyme-linked immunosorbent EGF assay kits (DY3689

and DY1756, respectively, R&D systems). In rats, the urinary EGF was expressed as amount excreted

in 24h hours, while in humans the uEGF concentration was corrected by reference to the urinary

creatinine concentration.

Edinburgh Type 2 Diabetes Cohort

We used specimen from 642 patients from the EdT2DS who had no evidence of kidney disease at

baseline (eGFR > 60 ml/min/1.73m2 and ACR<2.5 mg/mmol in males and <3.5 mg/mmol in females).

The detailed study protocol has been published previously 28. Briefly, at the baseline clinic visit all

participants underwent venepuncture and provided an early morning urine sample. From the urine

samples albumin and creatinine concentrations were determined and from the venous blood

sample, HbA1c and isotope dilution mass spectrometry–traceable serum creatinine levels were

measured according to standard protocols in the Department of Biochemistry, Western General

Hospital, Edinburgh, UK. The CKD-EPI equation was employed for estimating GFR. Urine was frozen

at -20C. Subjects were invited to re-attend the research clinic 4 years after their initial visit. For

persons who did not attend the follow-up appointment, the Lothian laboratory database was

interrogated to estimate GFR by the CKD-EPI equation, based on the isotope dilution mass

spectrometry–traceable serum creatinine concentration at their last outpatient clinic visit. In this

way, we excluded rises in creatinine occurring as a result of an acute illness. The rate of change in

renal function was determined by subtracting the Year 4 eGFR (or last outpatient eGFR for those not

attending the Year 4 visit) from the baseline eGFR, and dividing by the time between

measurements5. To determine incident eGFR <60 ml/min/1.73m2 and new-onset albuminuria we

obtained all outpatient results during follow-up using the Lothian laboratory database. Only those

who had eGFR <60 ml/min/1.73m2 or ACR: >2.5mmol/l (>3.5mmol/l for females) on 2 out of 3

consecutive samples were recorded as meeting end-points. Primary outcomes were new onset

eGFR<60ml/min/1.73m2, rapid decline in renal function (>5% per year) or a combination of both

end-points.

All the study participants gave their informed consent and ethics permission was obtained from the

Lothian Medical Research Ethics Committee.

Non-diabetic CKD cohort

In addition, we determined uEGF/Cr ratios in archived urine samples from a cohort of patients

(n=95) with non-diabetic CKD from a study which has been described previously12. In brief, this study

was designed to assess factors influencing vascular dysfunction across a wide range of renal function

and the baseline patient characteristics are given in Suppl table 2.

Statistical analysis

Data are means (±standard deviation) and medians (interquartile range) for parametric and non-

parametric data, respectively. Differences between groups were assessed by 2-test for categorical

variables and for parametric data by t-test or ANOVA and non-parametric data by Mann-Whitney or

Kruskal-Wallis. For regression analysis uEGF:creatinine was log transformed. As most (73%)

measurements of ACR were <0.5mg/mmol we categorised ACR values as: <0.5mg/mmol and 0.5-

1mg/mmol, and 1-2.5mg/mmol (males) or 1-3.5mg/mmol (females). For logistic regression analysis

ACR was transformed in log(ACR)+1. Associations between baseline characteristics including

log(urinary EGF:creatinine) and outcomes were determined by univariable and multivariable binary

logistic regression models. The baseline characteristics for the multivariable model were age,

gender, systolic & diastolic blood pressure, HbA1c, duration of diabetes, log(ACR)+1 and eGFR.

Unless explicitly stated otherwise in the text, two-sided P-values < 0.05 were considered significant

for statistical procedures. AUC-ROC curves were calculated using Medcalc version 15.8 (MedCalc

Software, Mariakerke, Belgium). All other statistical analyses were performed with Progenesis QI for

proteomics Software, the SPSS package (SPSS Inc., Chicago, IL, USA) and GraphPad Prism software

(GraphPad Prism Software Inc., San Diego, CA, USA).

Acknowledgements

This research was supported by an Innovation Grant from Kidney Research UK, the Edinburgh and

Lothians Diabetes Research Foundation and the Chief Scientist Office of the Scottish Government.

The ET2DS is funded by the Medical Research Council UK.

Disclosure

None of the authors have any competing interests to disclose in relation to the manuscript

Legend

Tables

Table 1. End-point pathophysiological parameters in control animals (n=10) and in animals subjected

to 28 weeks of hyperglycaemia (n=6), hypertension (n=7) or a combination of diabetes and

hypertension (n=10). *p<0.05, **p<0.01, ***p<0.001 vs control, ###p<0.001 vs diabetes alone, $p<0.05

vs Hypertension alone.

Table 2. Peptides of interest that were either significantly (p<0.01) increased (2A) or decreased (2B)

in the DM+HTN animal group in comparison with controls. *,#Peptides for which the corresponding

genes exhibit similar expression changes in the kidney in the Cyp1a1mRen2 rodent model (*) or in

human renal biopsies (#) compared with respective controls.

Table 3. Baseline characteristics of the study population stratified by urinary EGF:creatinine tertiles.

Comparison by2 , Kruskal–Wallis test or ANOVA

Abbreviations: EGF, epidermal growth factor; eGFR, estimated glomerular filtration rate;

albumin:creatinine ratio (ACR).

Table 4. Baseline characteristics of the study population stratified by the combination of the

outcomes incident eGFR<60ml/min/1.73m2 or rapid decline of renal function. Comparison by2 , t-

test or Mann-Whitney test.

Table 5. Unadjusted and adjusted odds ratio (OR, 95% CI) from logistic regression models per 1SD

increase in the log urinary EGF:creatinine ratio in relation to the outcomes incident

eGFR<60ml/min/1.73m2, rapid decline of renal function or the combination of both outcomes. The

following baseline variables were entered into the adjusted model: age, gender, eGFR, logACR+1,

duration of diabetes, HbA1c, systolic and diastolic blood pressure, Abbreviations: OR, odds ratio; CI,

confidence interval; ACR albumin:creatinine ratio; eGFR, estimated glomerular filtration rate.

Table 6. Receiver operating characteristic area under the curves (ROC-AUC) for log(uEGF/Cr), a

reference model (REF) that includes age, gender, eGFR, logACR+1, duration of diabetes, HbA1c,

systolic and diastolic blood pressure. The third model includes in addition to the reference model

log(uEGF/Cr).

Figures

Figure 1. Diabetes and hypertension together promote glomerulosclerosis and tubulointerstitial

fibrosis. (A) Representative images of periodic acid-Schiff staining in the glomeruli from kidneys from

each group. Bars represent 20 M. (B) Representative images of picrosirius red staining of the renal

cortex from each group. Bars represent 100 M.

Figure 2. To visualise similarities and differences in the variation in the global urinary peptide profiles

between animals in each group we performed principal component analysis (PCA). In the plot the

combination of peptides that explains the most variation is denoted as principal component 1 (PC1)

on the x-axis with a second principal component (PC2) on the y-axis. It is important to emphasise

that the principal components are not actual physical variables and PCA is used here as a quality

control to visualise the degree of clustering and identify outliers rather than to analyse the degree of

changes between the groups. Urines from each animal are represented in blue (Control), violet (DM

alone), orange (HTN alone) or green (DM+HTN). It is apparent that urine samples from animals

within the same group cluster together, with samples from the DM+HTN animals clearly segregating

from those from animals with either DM or HTN alone

Figure 3. Volcano plot demonstrates the fold change (log2) in abundance (x-axis) plotted against the

P-value (–log100, y-axis) for urinary peptides from the DM+HTN animal group compared with control

untreated rats. Altogether 1165 peptides were compared. The dotted line represents p=0.01.

Figure 4. (A) Renal cortical EGF mRNA expression was significantly reduced in the DM+HTN animals

compared with control animals ***p<0.001. (B) Serial measurements of the mean (s.e.) excretion

(g/24hrs) of urinary epidermal growth factor (EGF) in DM+HTN rats. *p<0.01 v baseline.

Figure 5. EGF and kidney injury molecule-1 (KIM-1) expression in sequential sections of renal cortex

from control and DM+HTN animals (A) There was focal reduction in EGF protein expression in

tubules in DM+HTN animals (*) (B) The reduced EGF expression was observed principally in injured

(KIM-1+) tubules. Bars represent 100 M.

Figure 6. Kaplan Meier survival curves for incident eGFR<60ml/min/1.73m2 stratified according to (A)

tertiles of uEGF/Cr ratio or (B) baseline ACR categories: <0.5. 0.5-0.1, 1-2.5 (male) or 1-3.5mg/mmol

(female).

Supplementary data

Suppl table 1. List of all peptides identified by Mascott-Search, grouped according to

increase/decrease of intensity in DM+HTN group compared to control.

Suppl table 2. Baseline characteristics of a separate cohort of 95 patients with non-diabetic CKD.

Values are means (±SD) or median (IQR). APKD: adult polycystic kidney disease, FSGS: focal

segmental glomerulosclerosis; ACR: albumin:creatinine ratio; uEGF/Cr: urinary epidermal growth

factor:creatinine ratio. Other causes of CKD include: mesangial proliferative glomerulonephritis

(n=2), cysteinuria (n=2), medullary sponge kidney (n=2), congenitally absent kidney (n=2), Alport’s

syndrome (n=1); thin basement membrane disease (n=1). Unknown include those with microscopic

haematuria (n=12) or low grade proteinuria (n=3) who were not biopsied.

Suppl table 3. Baseline characteristics of the study population stratified by the outcome of rapid

decline of renal function. Comparison by2 , t-test or Mann-Whitney test. $ Number or patients (%)

in categories of ACR <0.5mg/mmol , 0.5-1mg/mmol , 1- 2.5mg/mmol (males) or 3.5mg/mmol

(females).

Suppl table 4. Baseline characteristics of the study population stratified by the by the outcome of

incident eGFR<60ml/min/1.73m2. Comparison by2 , t-test or Mann-Whitney test. $ Number or

patients (%) in categories of ACR <0.5mg/mmol , 0.5-1mg/mmol , 1- 2.5mg/mmol (males) or

3.5mg/mmol (females).

Suppl Figure 1. Flow-Chart of the analytical procedure to exclude non-relevant features/peptides

using Progenesis/Maxquant software including the numbers of features/peptides at each step.

Suppl Figure 2. Correlation between uEGF:creatinine ratio and (A) eGFR and (B)

albuminuria:creatinine ratio.

References

1. Eggers PW. Has the incidence of end-stage renal disease in the USA and other countries stabilized? Curr Opin Nephrol Hypertens 2011; 20: 241-245.

2. Lind M, Svensson AM, Kosiborod M, et al. Glycemic control and excess mortality in type 1 diabetes. N Engl J Med 2014; 371: 1972-1982.

3. Keane WF, Brenner BM, de Zeeuw D, et al. The risk of developing end-stage renal disease in patients with type 2 diabetes and nephropathy: the RENAAL study. Kidney Int 2003; 63: 1499-1507.

4. Gerstein HC. Diabetes and the HOPE study: implications for macrovascular and microvascular disease. International journal of clinical practice Supplement 2001: 8-12.

5. Conway BR, Rennie J, Bailey MA, et al. Hyperglycemia and renin-dependent hypertension synergize to model diabetic nephropathy. Journal of the American Society of Nephrology : JASN 2012; 23: 405-411.

6. Ichimura T, Bonventre JV, Bailly V, et al. Kidney injury molecule-1 (KIM-1), a putative epithelial cell adhesion molecule containing a novel immunoglobulin domain, is up-regulated in renal cells after injury. The Journal of biological chemistry 1998; 273: 4135-4142.

7. Brosius FC, 3rd, Alpers CE, Bottinger EP, et al. Mouse models of diabetic nephropathy. Journal of the American Society of Nephrology : JASN 2009; 20: 2503-2512.

8. Siwy J, Zoja C, Klein J, et al. Evaluation of the Zucker diabetic fatty (ZDF) rat as a model for human disease based on urinary peptidomic profiles. PLoS One 2012; 7: e51334.

9. Ju W, Nair V, Smith S, et al. Tissue transcriptome-driven identification of epidermal growth factor as a chronic kidney disease biomarker. Science translational medicine 2015; 7: 316ra193.

10. Lindenmeyer MT, Kretzler M, Boucherot A, et al. Interstitial vascular rarefaction and reduced VEGF-A expression in human diabetic nephropathy. J Am Soc Nephrol 2007; 18: 1765-1776.

11. Woroniecka KI, Park AS, Mohtat D, et al. Transcriptome analysis of human diabetic kidney disease. Diabetes 2011; 60: 2354-2369.

12. Lilitkarntakul P, Dhaun N, Melville V, et al. Blood pressure and not uraemia is the major determinant of arterial stiffness and endothelial dysfunction in patients with chronic kidney disease and minimal co-morbidity. Atherosclerosis 2011; 216: 217-225.

13. Kelly DJ, Cox AJ, Tolcos M, et al. Attenuation of tubular apoptosis by blockade of the renin-angiotensin system in diabetic Ren-2 rats. Kidney Int 2002; 61: 31-39.

14. Thulesen J, Jorgensen PE, Torffvit O, et al. Urinary excretion of epidermal growth factor and Tamm-Horsfall protein in three rat models with increased renal excretion of urine. Regul Pept 1997; 72: 179-186.

15. Torffvit O, Jorgensen PE, Kamper AL, et al. Urinary excretion of Tamm-Horsfall protein and epidermal growth factor in chronic nephropathy. Nephron 1998; 79: 167-172.

16. Lev-Ran A, Hwang DL, Miller JD, et al. Excretion of epidermal growth factor (EGF) in diabetes. Clin Chim Acta 1990; 192: 201-206.

17. ter Meulen CG, Bilo HJ, van Kamp GJ, et al. Urinary epidermal growth factor excretion is correlated to renal function loss per se and not to the degree of diabetic renal failure. Neth J Med 1994; 44: 12-17.

18. Kawaguchi M, Kamiya Y, Ito J, et al. Excretion of urinary epidermal growth factor in non-insulin dependent diabetes mellitus. Life Sci 1993; 52: 1181-1186.

19. Dagogo-Jack S, Marshall SM, Kendall-Taylor P, et al. Urinary excretion of human epidermal growth factor in the various stages of diabetic nephropathy. Clin Endocrinol (Oxf) 1989; 31: 167-173.

20. Mathiesen ER, Nexo E, Hommel E, et al. Reduced urinary excretion of epidermal growth factor in incipient and overt diabetic nephropathy. Diabet Med 1989; 6: 121-126.

21. Retnakaran R, Cull CA, Thorne KI, et al. Risk factors for renal dysfunction in type 2 diabetes: U.K. Prospective Diabetes Study 74. Diabetes 2006; 55: 1832-1839.

22. Cook NR. Statistical evaluation of prognostic versus diagnostic models: beyond the ROC curve. Clin Chem 2008; 54: 17-23.

23. Inker LA, Schmid CH, Tighiouart H, et al. Estimating glomerular filtration rate from serum creatinine and cystatin C. The New England journal of medicine 2012; 367: 20-29.

24. Rappsilber J, Ishihama Y, Mann M. Stop and go extraction tips for matrix-assisted laser desorption/ionization, nanoelectrospray, and LC/MS sample pretreatment in proteomics. Anal Chem 2003; 75: 663-670.

25. Fliser D, Novak J, Thongboonkerd V, et al. Advances in urinary proteome analysis and biomarker discovery. J Am Soc Nephrol 2007; 18: 1057-1071.

26. Ohta S, Bukowski-Wills JC, Sanchez-Pulido L, et al. The protein composition of mitotic chromosomes determined using multiclassifier combinatorial proteomics. Cell 2010; 142: 810-821.

27. Atrih A, Mudaliar MA, Zakikhani P, et al. Quantitative proteomics in resected renal cancer tissue for biomarker discovery and profiling. Br J Cancer 2014; 110: 1622-1633.

28. Price JF, Reynolds RM, Mitchell RJ, et al. The Edinburgh Type 2 Diabetes Study: study protocol. BMC Endocr Disord 2008; 8: 18.