While engaged in this project, Dr. Christos Argyropoulos was a salaried employee of the University of Pittsburgh. Since the completion of the work presented herein Dr. Argyropoulos has served as a consultant to the Medical Department of the Greek Affiliate of Abbott Laboratories, a global healthcare company that manufactures diabetes care products and develops pharmaceuticals for the treatment of diabetic nephropathy. The views and opinions in this research project are solely those of the contributing authors and do not necessarily reflect those of Abbott Laboratories.
Conceived and designed the experiments: JJ DG TO KW CA JB DE. Performed the experiments: KW SM DH. Analyzed the data: CA KW. Contributed reagents/materials/analysis tools: KW DG TO DE SM DH CA. Wrote the paper: CA JJ KW DG JB TO DE.
Patients with Type 1 Diabetes (T1D) are particularly vulnerable to development of Diabetic nephropathy (DN) leading to End Stage Renal Disease. Hence a better understanding of the factors affecting kidney disease progression in T1D is urgently needed. In recent years microRNAs have emerged as important post-transcriptional regulators of gene expression in many different health conditions. We hypothesized that urinary microRNA profile of patients will differ in the different stages of diabetic renal disease.
We studied urine microRNA profiles with qPCR in 40 T1D with >20 year follow up 10 who never developed renal disease (N) matched against 10 patients who went on to develop overt nephropathy (DN), 10 patients with intermittent microalbuminuria (IMA) matched against 10 patients with persistent (PMA) microalbuminuria. A Bayesian procedure was used to normalize and convert raw signals to expression ratios. We applied formal statistical techniques to translate fold changes to profiles of microRNA targets which were then used to make inferences about biological pathways in the Gene Ontology and REACTOME structured vocabularies. A total of 27 microRNAs were found to be present at significantly different levels in different stages of untreated nephropathy. These microRNAs mapped to overlapping pathways pertaining to growth factor signaling and renal fibrosis known to be targeted in diabetic kidney disease.
Urinary microRNA profiles differ across the different stages of diabetic nephropathy. Previous work using experimental, clinical chemistry or biopsy samples has demonstrated differential expression of many of these microRNAs in a variety of chronic renal conditions and diabetes. Combining expression ratios of microRNAs with formal inferences about their predicted mRNA targets and associated biological pathways may yield useful markers for early diagnosis and risk stratification of DN in T1D by inferring the alteration of renal molecular processes.
Diabetic nephropathy (DN) is the leading cause of End Stage Renal Disease (ESRD) in the Western world, accounting for more than 40% of cases. Patients with either type 1 (T1D) or 2 (T2D) diabetes are at risk of DN, but the disease burden is higher in the former group
In recent years microRNAs (miRNAs), a family of short (average of 22nt long), naturally occurring, small antisense non-coding RNAs have emerged as important post-transcriptional regulators of gene expression (see review
Urine samples from participants in the Pittsburgh Epidemiology of Diabetes Complications (EDC) study were examined. The EDC study is a historical prospective cohort which recruited patients from Children’s University Hospital of Pittsburgh Registry of all cases of T1D, diagnosed or seen within a year of diagnosis between January 1st 1950 and May 31st 1980. Participants were followed thereafter with repeat exams biennially for 10 years and again at 18 years. Follow up of all participants in the EDC was censored for this analysis on December 31st 2000.
In the EDC, diabetic renal disease was characterized in terms of its progression from a normoalbuminuric urine examination to progressively higher amounts of albumin in the urine (microalbuminuria) to overt nephropathy. Microalbuminuria was defined as 20–200 µg/min in at least two of three timed urines (24hr, overnight, and 4 hr clinic visit) and was further classified as
The RNA from urine was isolated using the miRNeasy kit (Qiagen, Germantown, MD). In brief, 700 µl of QIAzol reagent was added to 200 µl of urine sample. The sample was mixed in a tube followed by adding 140 µl of chloroform. After mixing vigorously for 15 seconds, the sample was then centrifuged at 12,000×g for 15 minutes at 4°C. The upper aqueous phase was carefully transferred to a new collection tube, and 1.5 volume of ethanol containing binding buffer from the kit was added and mixed. The sample was then applied directly to a silica membrane containing column and the RNA was retained and cleaned by using buffers provided in the kit. The immobilized cleaned RNA was then eluted from the membrane into a collection tube with a low salt elution buffer or water. The quality and quantity of the RNA was evaluated by 260/280 ratio and Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara CA).
In brief, the cDNA was generated from 20 µl of RNA using buffer and enzyme provide in the Qiagen kit. After incubating the cDNA synthesis reaction at 42°C for 60 minutes, the cDNA was diluted to 8 ml with SYBR containing PCR reagents from Exiqon and water. The plates were then loaded onto ABI 7900HT real-time PCR system and the threshold cycle (Ct) was measured with standard methods. Exiqon miRNA qPCR panels 1 and 2 (Version 1) were used, that included probes for 748 unique miRNA. Each miRNA species was assayed once per panel with the exception of miR-423-5p, miR-103, miR-191 and the three non-coding RNA species U6, SNORD38B and SNORD39A for which duplicate reactions was set up as per panel manufacturer instructions. Although suggested as reference genes (biological controls) by the panel manufacturer the 6 microRNAs/small nuclear RNAs were not used as referents during normalization. Nevertheless their presence in multiple technical replicates in any given panel, allowed us to derive panel specific normalization factors which were applied to the raw expression levels of all microRNAs. A single inter-plate calibrator spiked in control (UniSP3) was run 6 times per plate and was used to normalize the expression levels of all miRNAs included in each of the qPCR panels. A second spiked in control (UniSP6) was included in some but not all urine reactions as a dual positive – negative control and was thus not considered in subsequent analyses (including normalization). We also included a no-template negative control in all assays (nine replicates per assay) as per manufacture guidelines. To resolve discrepancies in the nomenclature of miRNA species, we mapped names of miRNAs present in the Exiqon plates to the most current ones in miRBase (version 18, November 2011) and the associated MIMAT accession numbers (
In order to classify individual patient samples and visualize the resemblance in the corresponding profiles we applied Principal Component Analysis (PCA) to the corrected Cq values obtained from the raw Cq measurements after subtracting the quantification cycle number of the spiked-in control. To handle missing data in the expression of miRNAs across samples we applied a specific variety of PCA, i.e. Probabilistic PCA (PPCA)
To analyze the difference in miRNA expression within patient groups, we quantified the relative expression level of each miRNA, its normalized threshold cycle difference (ΔCq) i.e. the difference between the quantification cycle in the experimental (E) and the reference (R) state : ΔCq = Cq(E) – Cq(R), with positive ΔCq values indicating lower concentrations. To ensure a sufficient amount of data for downstream analyses, only those miRNAs that were detected in at least 2/3 of patient samples in each comparison were analyzed.
A mixed effects model was used simultaneously accounting for matching patients within pairs while normalizing ΔCq values for PCR related factors. Normalization of quantification cycle signals occurred in two steps (
Replicate qPCR reactions were analyzed with a hierarchical linear mixed model in order to estimate panel specific correction factors that were subtracted from the raw Cq signals of unreplicated reactions (first step), while simultaneously estimating the difference (ΔCq) between an experimental and referent state. In the second step, the ΔCq of the spiked in control was subtracted from the non-control ΔCq values to calibrate the relative fold changes according to the Delta-Delta method. Both steps of the normalization procedure acknowledged the uncertainty implicit in estimating the ΔCq of both control and non-control signals (shown as a density plot at the bottom part of the figure), by performing this subtraction probabilistically i.e. by Monte Carlo methods.
Bayesian models were programmed in the BUGS language (
To infer putative targets of differentially expressed miRNAs we utilized three different algorithms: miRanda (release August 2010)
A change in the ligand concentration between an experimental state (
To the extent that miRNAs function as negative regulators of mRNA translation a positive log-odds ratio (larger bound fraction) would imply a propensity for the target mRNA expression to be reduced in the experimental state.
To synthesize the evidence from multiple
Baseline characteristics of patients included in this study are shown in
Group A | Group B | |||
Clinical Classification | Clinical Classification | |||
Normal | Overt Nephropathy | Intermittent Microalbuminuria | Persistent Microalbuminuria | |
|
10 | 10 | 10 | 10 |
|
10 | 10 | 20 | 20 |
|
10 | 10 | 19 | 14 |
|
42.8±5.1 | 41.4±6 | 29.4±6.3 | 27.5±5.3 |
|
5 | 5 | 5 | 5 |
|
34.1±5.8 | 34.4±6.4 | 20.7±5.4 | 21.3±5.8 |
|
3 | 1 | 0 | 1 |
|
0 | 0 | 0 | 0 |
|
2 | 2 | 0 | 1 |
|
5 | 5 | 0 | 0 |
|
4 | 1 | 1 | 5 |
|
1 | 6 | 0 | 0 |
|
8.2±1.1 | 8.2±1.0 | 9.9±1.9 | 10.2±2.4 |
|
103.5±20.2 | 106.3±44.2 | 100.3±20.6 | 115.9±57.5 |
|
1 | 8 | 0 | 0 |
|
1 | 5 | 0 | 0 |
Due to the insufficient amount of RNA, we did not obtain good quality mRNA measurement in 6 samples from 3 patient pairs in the PMA sub-group and one urine sample from the IMA group. Nevertheless, reproducibility of un-normalized Cq signals from urine in the rest of the samples was high (
A global view of the changes in urinary miRNA profiles according to the clinical classification was performed with PCA and the results for the first five principal components (PC) are shown in
To present the results of the five dimensional PCA, we utilized bivariate projections in which each component is plotted against all e.g. the second plot in the first row plots the first principal component (PC1) against the second (PC2). Each individual urine sample is color and symbol coded according to the disease classification at the time it was collected. N: patients without nephropathy, DN: patients with overt nephropathy, IMA(B): normoalbuminuric samples from patients who had intermittent microalbuminuria, PMA(B): last normoalbuminuric samples from patients who had persistent albuminuria, IMA: micro-albuminuric samples from patients who had intermittent micro-albuminuria, PMA: micro-albuminuric samples from patients who had persistent microalbuminuria.
To explore whether patients who had been matched into pairs had similar microRNA profiles we plotted the results of the PCA according to the pair identifier. These results which are shown in
This figure utilizes the same bivariate projection setup as
219 miRNAs yielded measurable signals in >75% of the 33 urine samples profiled for this comparison. We observed only a few differences in the baseline samples. Relative to the IMA group, patients with PMA demonstrate decreased
miRNA | Fold Change | 95% Credible Interval | P |
Under-expressed | |||
hsa-miR-323b-5p |
0.07 | 0.01–0.42 | 0.0030 |
hsa-miR-221-3p |
0.15 | 0.03–0.80 | 0.0280 |
hsa-miR-524-5p | 0.19 | 0.04–0.88 | 0.0350 |
hsa-miR-188-3p | 0.28 | 0.08–0.98 | 0.0454 |
Over-expressed | |||
hsa-miR-214-3p |
8.71 | 1.97–38.05 | 0.0050 |
hsa-miR-92b-5p |
8.65 | 1.11–67.46 | 0.0394 |
hsa-miR-765 | 7.22 | 1.78–30.98 | 0.0046 |
hsa-miR-429 | 5.92 | 1.42–23.94 | 0.0136 |
hsa-miR-373-5p |
4.50 | 1.19–17.27 | 0.0296 |
hsa-miR-1913 | 4.37 | 1.30–15.47 | 0.0156 |
hsa-miR-638 | 3.71 | 1.02–13.81 | 0.0464 |
For miRNAs whose name changed after the introduction of the 18th version of MiRBase, we provide both the previous (in
miRNA | Fold Change | 95% Credible Interval | P |
Under-expressed | |||
hsa-miR-589-5p |
0.05 | 0.00–0.98 | 0.048 |
hsa-miR-373-5p |
0.07 | 0.01–0.45 | 0.007 |
hsa-mir-520h | 0.12 | 0.02–0.80 | 0.026 |
hsa-miR-92a-3p |
0.14 | 0.02–0.98 | 0.048 |
Over-expressed | |||
hsa-miR-323b-5p |
31.51 | 2.91–368.48 | 0.0044 |
hsa-miR-433 | 16.24 | 1.39–196.0 | 0.028 |
hsa-miR-17-5p |
14.82 | 1.09–214.0 | 0.044 |
hsa-miR-222-3p |
11.22 | 1.13–102.0 | 0.036 |
hsa-let-628-5p | 7.59 | 1.07–52.2 | 0.044 |
For miRNAs whose name changed after the introduction of the 18th version of MiRBase, we provide both the previous (in italics) and the recent (regular font) name.
283 miRNAs yielded measurable signals in >75% of the 20 urine samples from these 10 patient pairs. In
miRNA | Fold Change | 95% Credible Interval | P |
Under-expressed | |||
hsa-miR-221-3p |
0.25 | 0.07–0.86 | 0.0330 |
Over-expressed | |||
hsa-miR-619 | 6.98 | 1.86–27.80 | 0.0030 |
hsa-miR-486-3p | 6.43 | 1.36–26.66 | 0.0290 |
hsa-miR-335-5p |
5.81 | 1.70–20.71 | 0.0050 |
hsa-miR-552 | 5.47 | 1.19–27.50 | 0.0310 |
hsa-miR-1912 | 4.72 | 1.01–23.92 | 0.0490 |
hsa-miR-1224-3p | 4.45 | 1.10–17.48 | 0.0430 |
hsa-miR-424-5p |
4.38 | 1.35–15.13 | 0.0130 |
hsa-miR-141-3p |
3.81 | 1.29–11.17 | 0.0140 |
hsa-miR-29b-1-5p |
3.03 | 1.09–8.61 | 0.0370 |
For miRNAs whose name changed after the introduction of the 18th version of MiRBase, we provide both the previous (in italics) and the recent (regular font) name.
Albuminuric vs Normoalbuminuric in the MA group | Overt vs Normal | |||
Pathway | P-value | Fraction | P-value | Fraction |
|
||||
Signaling by SCF-KIT | 0.006 | 18/76 | 0.001 | 41/76 |
Signaling by Insulin receptor | 0.009 | 23/109 | <0.001 | 65/109 |
Signaling by NGF | 0.016 | 38/212 | <0.001 | 119/212 |
Signaling by Rho GTPases | 0.024 | 24/125 | <0.001 | 71/125 |
Signaling by ERBB4 | 0.027 | 16/76 | <0.001 | 45/76 |
Signaling by ERBB2 | 0.035 | 19/97 | <0.001 | 59/97 |
Signaling by PDGF | 0.040 | 22/118 | <0.001 | 67/118 |
Signaling by VEGF | 0.041 | 4/11 | ||
Signaling by EGFR | 0.044 | 20/106 | <0.001 | 64/106 |
Dowstream signaling of activated FGFR | 0.038 | 19/98 | <0.001 | 61/98 |
Signaling by BMP | 0.001 | 16/23 | ||
Signaling by TGFβ | 0.004 | 11/15 | ||
DAG and IP3 signaling | 0.010 | 20/31 | ||
PIP3 activates AKT signaling | 0.020 | 15/26 | ||
RAF/MAP kinase cascade | 0.031 | 7/10 | ||
Signaling by Notch | 0.036 | 13/23 | ||
Interaction of integrin α5β3 with fibrillin | 0.044 | 2/3 | ||
Interaction of integrin α5β3 with von Willbrand factor | 0.044 | 2/3 | ||
Integrin cell surface interactions | 0.024 | 40/85 | ||
|
0.009 | 57/122 | ||
|
||||
G0 and early G1 | 0.040 | 12/21 | ||
|
||||
Metabolism of lipids and lipoproteins | 0.022 | 51/305 | 0.005 | 132/205 |
Cysteine formation from homocysteine | 0.016 | 2/2 | ||
Integration of energy metabolism | 0.009 | 45/93 | ||
|
||||
Post-translational protein modification | 0.045 | 30/173 | 0.019 | 76/173 |
|
0.007 | 67/396 | <0.001 | 189/396 |
|
0.032 | 40/84 | ||
|
||||
Caspase-8 is formed from procaspase-8 | 0.019 | 4/9 | ||
|
||||
RNA Polymerase II Transcription | 0.050 | 19/101 | ||
Capping complex formation | 0.039 | 7/26 | ||
Nuclear Receptor Transcription | 0.005 | 28/51 | ||
|
||||
Vitamin D (calciferol) metabolism | 0.048 | 3/7 | ||
|
0.008 | 12/18 | ||
|
||||
|
0.047 | 3/3 | ||
Transmission across Chemical Synapses | <0.001 | 66/108 | ||
|
||||
Interleukin-2 signaling | 0.029 | 10/41 | ||
14-3-3 zeta binding allows recruitment of PI3K | 0.033 | 5/15 | ||
Signaling by interleukins | 0.002 | 54/105 | ||
|
<0.001 | 206/426 | ||
Platelet homeostasis | 0.008 | 14/56 | ||
Platelet activation, signaling and aggregation | 0.011 | 35/187 |
P-value: the p-value of the hypergeometric test unadjusted for multiple comparisons, Fraction: number of proteins in the pathway that are targets of differentially expressed miRNAs over the total number of proteins in each pathway.
In this paper we report the changes of urinary miRNA spectrum in T1D patients with different stages of albuminuria and nephropathy. We found concentration changes on specific miRNAs that may involve in specific pathways known to be altered in various forms of renal diseases. Since the kidney is the most likely source of these urinary miRNAs, we suggest that these miRNAs may be of biological and clinical significance in T1D.
A global Principal Component Analysis viewpoint of the microRNA profiles analyzed in this report suggests that there are some differences in the expression of urinary microRNA which appear to follow the clinical classification of patients and urinary samples with respect to albumin excretion. The apparent clustering of profiles from patients who had been matched into pairs, suggests that there are other factors affecting urinary microRNA besides the clinical classification of disease. Such factors are likely related to the variables we used in patient matching e.g. age, sex, and duration of disease and level of glycemic control. This observation justifies post-hoc our decision to explore specific microRNA signatures across the spectrum of clinical classification of patients and samples using a matched case control design.
Our matched case-control Bayesian analyses highlight a set of 27 differentially regulated miRNAs across different clinical stages of diabetic renal disease. Previous work using experimental, clinical chemistry or biopsy samples has demonstrated differential expression of many of these miRNAs in a variety of renal conditions:
Based on miRNA target prediction databases, miRNAs showing concentration changes in diabetic urine may regulate genes that play key roles in renal physiology and pathophysiology:
In addition, intriguing connections in heterologous systems have been reported for other miRNAs highlighted in this report: miR-221-3p/222-3p (neovascularization and vascular neointimal hyperplasia
With the samples used in this study, we could not verify the association of miR-192 with DN. Higher miR-192 levels have been previously linked to renal damage in the streptozocin (T1D) and the db/db (T2D) mouse nephropathy models
Most of the identified miRNAs exhibited changes in one disease state rather than showing a quantitative trend of increasing or decreasing expression paralleling the severity of albuminuria. To understand this pattern we examined the predicted targets of these miRNAs and the corresponding pathways using structured vocabularies for biological annotation. Despite the disparate identity of the miRNAs, the mRNAs that are predicted to be targeted by them map to pathways that have been previously shown to be pathophysiologically relevant to DN: TGF (the prototypical “renal-fibrosis” culprit
Our analyses suggest the involvement of NGF (Nerve Growth Factor, a prototypical Central Nervous System trophic molecule) in diabetic nephropathy. This may lead to a new direction toward the development of T1D associated nephropathy since so far the renal expression of NGF has been thought to reflect the level of glycemic control
Growth Factor as well as other pathways (e.g. cell-cell and cell-matrix) are targeted from the microalbuminuric stage, while the number of targeted genes in these pathways increased at the overt nephropathy stage. Hence an “exposure-response” relation appears at the target (mRNA) rather than the regulator (miRNA) level. This relation stems from the overlapping, combinatorial, binding specificities of miRNAs to their mRNA targets so that the same pathways may be targeted by rather different sets of miRNAs depending on the prevailing cellular context.
An interesting aspect of the targets associated with the miRNAs identified in this study is the lack of an overwhelming association between growth factor transduction pathways and the
The findings of our study should be interpreted in light of a number of limitations. First, we analyzed urine samples from an era in which current therapies for diabetic nephropathy (angiotensin converting enzyme inhibitors and angiotensin receptor blockers) were not widely used early in the disease process. Hence most of the patients with MA were not on ACEi/ARB inhibition even though evidence from randomized trials suggest that these agents delay the appearance of microalbuminuria
In summary, a set of 27 differentially miRNAs were identified in matched urine samples from T1D patients with different stages of diabetic nephropathy, whose renal outcomes had been ascertained after prolonged follow up. These miRNAs map to pathways of known relevance to the development of diabetic renal disease, strongly suggesting the renal source of the miRNAs. Our results suggest that a number of miRNAs in urine may serve not only as molecular signatures of distinct clinical phenotypes in diabetic nephropathy but also as early indicators of alterations in specific biological processes in the kidney which can be of importance in individualizing emergent therapies for diabetic kidney disease. Further studies are needed to extend these observations in the setting of T2D and clarify the potential utility of these miRNAs in early diagnosis, risk stratification for progression and treatment selection or monitoring.
(TIF)
(PNG)
(XLS)
(PDF)
(ZIP)
(ZIP)
(ZIP)
(ZIP)
(ZIP)
We thank Ms. Yue Yuan for RNA isolation.