Signatures of miR-181 a on the Renal Transcriptome and Blood Pressure

Subjects Human kidney samples were collected during the TRANScriptome of renaL humAn TissuE (TRANSLATE) Study (1). In brief, samples were obtained after surgery in 200 patients who underwent elective unilateral nephrectomy because of non-invasive renal cancer in one of three nephrology-urology centres [Silesian Renal Tissue Bank (2, 3), TRANSLATE P (recruitment conducted in Western Poland) and TRANSLATE Z (recruitment conducted in Southern Poland) (1)]. Each subject underwent standardized clinical phenotyping, including personal history (through anonymous coded questionnaires), anthropometry (i.e. weight, height) and triplicate measurements of blood pressure (BP) using a mercury sphygmomanometer (in a subset of samples – Silesian Renal Tissue Bank) (2, 3) or automatic digital BP monitoring (in the rest of the patients). Hypertension was diagnosed based on the criteria used previously (4, 5). All individuals were of white-European ancestry. Serum samples from 200 subjects recruited in the Genetic Regulation of Arterial Pressure of Humans In the Community (GRAPHIC) Study were obtained for the replication experiment. Details of recruitment and phenotyping of GRAPHIC are described elsewhere (6, 7). Briefly, 520 nuclear families of white-European ancestry, with both parents aged 40–60 years and two adult offspring aged ≥ 18 years were identified through general practice records in Leicestershire, UK. All subjects (n = 2,037) were extensively phenotyped including both clinic BP (calculated as a mean of the second and third of three readings) and 24-h ambulatory BP (7). The subset of 199 subjects obtained for the present replication cohort were selected from the parental generation of the GRAPHIC cohort (n = 1,032) after the following initial exclusions: poor quality phenotype or genotype data (n = 9); on antihypertensive or lipidlowering medication (n = 150); serum sample missing (n = 30) or low volume (n = 4). Following these exclusions, from amongst eligible subjects, 200 were selected on the basis of sex and extremes of mean 24-h ambulatory systolic BP. Thus, 100 men with the 50 highest and lowest values for mean 24-h ambulatory systolic BP were included in the replication sample, along with 100 women using the same criteria. One subject was further excluded due to a failure of RNA extraction leaving 199 included in the statistical analyses. No patients were receiving heparin and just six were receiving antiplatelet therapy. All subjects gave written informed consent and the studies were approved by ethics committees at each institution. Renal Tissue, Serum Processing and RNA Extraction The renal tissue samples in the TRANSLATE Study were collected immediately after nephrectomy from a healthy (unaffected by cancer) pole of the kidney and immersed in RNAlater (Life Technologies) before storage at –80°C, as previously described (1-3, 8). RNA was extracted from all kidney tissues using the miRNeasy kit (Qiagen) and stored at –80°C. RNA was also extracted from 100 μl (TRANSLATE) or 200 μl (GRAPHIC) of serum (secured from blood collected during recruitment and stored at –80°C) using the miRNeasy Serum/Plasma kit (Qiagen) from matching subjects. No pooling of samples was performed.

nuclear families of white-European ancestry, with both parents aged 40-60 years and two adult offspring aged ≥ 18 years were identified through general practice records in Leicestershire, UK. All subjects (n = 2,037) were extensively phenotyped including both clinic BP (calculated as a mean of the second and third of three readings) and 24-h ambulatory BP (7). The subset of 199 subjects obtained for the present replication cohort were selected from the parental generation of the GRAPHIC cohort (n = 1,032) after the following initial exclusions: poor quality phenotype or genotype data (n = 9); on antihypertensive or lipidlowering medication (n = 150); serum sample missing (n = 30) or low volume (n = 4). Following these exclusions, from amongst eligible subjects, 200 were selected on the basis of sex and extremes of mean 24-h ambulatory systolic BP. Thus, 100 men with the 50 highest and lowest values for mean 24-h ambulatory systolic BP were included in the replication sample, along with 100 women using the same criteria. One subject was further excluded due to a failure of RNA extraction leaving 199 included in the statistical analyses. No patients were receiving heparin and just six were receiving antiplatelet therapy.
All subjects gave written informed consent and the studies were approved by ethics committees at each institution.

Renal Tissue, Serum Processing and RNA Extraction
The renal tissue samples in the TRANSLATE Study were collected immediately after nephrectomy from a healthy (unaffected by cancer) pole of the kidney and immersed in RNAlater (Life Technologies) before storage at -80°C, as previously described (1)(2)(3)8). RNA was extracted from all kidney tissues using the miRNeasy kit (Qiagen) and stored at -80°C. RNA was also extracted from 100 μl (TRANSLATE) or 200 μl (GRAPHIC) of serum (secured from blood collected during recruitment and stored at -80°C) using the miRNeasy Serum/Plasma kit (Qiagen) from matching subjects. No pooling of samples was performed.

Quantitative Real-time PCR (qPCR)
The first-strand complementary synthesis reaction (cDNA) was performed using the High Capacity Reverse Transcriptase cDNA Synthesis kit (Life Technologies) for mRNA and the TaqMan ® MicroRNA Reverse Transcription kit (Life Technologies) for miRNAs. Amplification reactions used the TaqMan ® Fast Advanced Master Mix (Life Technologies) in a Viia7 TM qPCR system (Life Technologies). TaqMan ® assays were used to measure mRNA and miRNA expression (Supplemental Table S1) following the cycling conditions recommended by the supplier. β-actin (ACTB) mRNA was used as an internal reference transcript for renal mRNA expression and RNA U6 small nuclear 2 (RNU6B) for miRNA expression in the kidney; both were chosen based on previously published data. (8) In serum, the spike-in cel-miR-39 was used to normalize experimental variability in miRNA levels, in agreement with previously published studies (9)(10)(11)(12). All expression experiments were run in duplicate (TRANS-LATE) or triplicate (GRAPHIC). The delta cycle threshold (dCt), derived from normalization of mRNA and miRNAs to reference mRNAs, was used as qPCR measure of renal expression (13). dCt is an inverse measure of mRNA expression -the higher the value, the lower the abundance of the target examined in the tissue.

Pathway Analysis of Human Kidney Transcriptomes Characterized by Next-Generation RNA-Sequencing
Next-generation RNA-sequencing and small RNA-sequencing raw data (level 1) from kidney tissue unaffected by cancer (labeled as "normal-matched") collected from 69 individuals with clear cell renal carcinoma were downloaded from The Cancer Genome Atlas (TCGA) database (http://tcga-data.nci.nih.gov) in April 2014 (application 26966-1 approved by TCGA Data Access Committee). Raw 50 base pair reads were first aligned to the GRCh37 reference genome (as used by Ensembl release 70) using the TopHat programme (15). On average, 68 million paired-end reads per sample were aligned successfully. Expression values were calculated for all Ensembl transcripts and genes using Cufflinks and Cuffdiff software (16). The final expression values were computed in log 2 TPM+1 (TPM: transcripts per million) units as described in previous studies (17,18). Technical variation between the samples was accounted for by normalization based on probabilistic estimation of expression residuals (PEER) (19,20). Analysis of association between miR-181a and mRNAs was conducted using Limma (21), with each mRNA expression value as the response variable and miR-181a expression, age and sex as independent parameters included in the regression model.
To identify renal pathways associated with expression of miR-181a, we used gene-set enrichment analysis (22) (in preranked mode) with Kyoto Encyclopedia of Genes and Genomes (KEGG), BioCarta and Reactome as pathway repositories. The t-statistics generated were averaged by gene. The pathway overlap, and therefore edge width, was calculated using the overlap coefficient. The resultant pathway network was visualized using Gephi (23) and annotated manually to highlight clusters of pathways with similar functions. Correction for multiple testing in both individual mRNA-and pathway-based analyses was conducted by calculation of false discovery rate q-values and the cor- tive staining (renin) was brown, while the counterstain was haematoxylin (blue staining). For ISH, the positive staining was blue while the counterstain was nuclear fast red (red staining). Both ISH and immunohistochemistry were conducted by Bioneer A/S (Denmark). Images were acquired directly from the slides by digital photography using a Nikon Eclipse E800 microscope with 100x, 200x and 400x magnification.

Biochemical Analysis
Serum renin levels were measured by immunoradiometric assay (Renin III, Cisbio, Gif-sur-Yvette) in 192 TRANSLATE subjects from whom both serum sample and kidney tissue samples were available.

Statistical Analysis
In the TRANSLATE Study, distributions of all mRNA expression measurements and biochemical data were examined using the Skewness and Kurtosis test prior to further analysis. Serum renin was non-normally distributed, and was therefore logarithmically transformed to achieve a normal distribution. Unadjusted association analysis between two quantitative phenotypes was explored first by Pearson's linear correlation. Further analyses were conducted using step-wise multiple linear regression models with clinical (age, sex, body mass index, recruitment centre) and experimental (cDNA or qPCR plate) variables as independent parameters (crite-ria of F-entry probability: 0.15, removal: 0.20). Additional biochemical (serum levels of renin) or molecular (renal expression of renin mRNA) parameters were included in the models where appropriate. Three types of sensitivity analysis were used after analysis of association between gene expression or biochemical phenotypes and BP values. In the first type of sensitivity analysis, BP from patients receiving antihypertensive medication were corrected for the BP lowering effect of medications by adding a constant of 15 mmHg and 10 mmHg to the systolic BP (SBP) and diastolic BP (DBP) measurements, respectively, in line with the algorithms used before (14). The second sensitivity analysis was based on exclusion of all subjects on antihypertensive therapy. The third sensitivity analysis included only individuals with clinic SBP ≤ 140 mmHg and DBP ≤ 90 mmHg, who were not taking antihypertensive medication.
In the GRAPHIC Study, data were first checked to ensure that assumptions for multiple regression analyses were met. Multiple linear regression analyses were performed using clinic SBP or clinic DBP as the dependent variable and miR-181a dCt, age, body mass index and sex as independent variables.
Results obtained at the P < 0.05 level were considered statistically significant. Statistical package SPSS for Windows (Release 21.0 for TRANSLATE and 20.0 for GRAPHIC) was used for the statistical analyses. rected threshold for statistical significance was set at q < 0.01.

Hierarchical Clustering of Multidimensional Scaling of Renal Transcriptomes Characterized by Next-Generation RNA-Sequencing
To exclude the potential effect of coexistent cancer on the transcriptome of the tumour-unaffected renal tissue from where the specimen was sampled, we compared global expression profiles of healthy tissue from nephrectomies due to cancer with those from non-cancer kidneys and clear cell renal carcinoma. The RNA-sequencing data were collected from three sources: TCGA (69 normalmatched and 10 randomly chosen renal cancer samples), Genotype-Tissue Expression (GTEx) Project (24) (8 non-cancer kidneys harvested from donors identified through low post-mortem interval) and TRANSLATE P Study (where 32 kidneys underwent next-generation RNAsequencing).
RNA-sequencing sample preparation and quality control -TRANSLATE Study. Approximately 30 milligrams of tissue was mechanically homogenized (PowerGen™ Model 125 Homogenizer) in Qiagen RTL buffer. RNA extraction was conducted using RNeasy Mini kit (Qiagen) according to the supplier's guidelines. Eluted RNA samples were assessed for concentration using a Nanodrop 8sample spectrophotometer ND-800 prior to further processing. RNA sequencing libraries were prepared using a non-strandspecific Illumina TruSeq v2 kit and 1 μg of poly-A selected input RNA at the Australian Genome Research Facility (AGRF). Libraries were sequenced on an Illumina HiSeq 2000 using 100 base pair pairedend reads, producing an average of ~52 million reads per sample. Raw reads were checked for quality using FastQC (http://www.bioinformatics.babraham.ac .uk/projects/fastqc/). Post-alignment, the squared coefficient of variation and expression dispersion estimates were assessed using CummeRBund (25).
RNA-sequencing data processing -GTEx Project. GTEx samples were pre-Supplementary Figure S2. Hierarchical clustering of renal transcriptomes characterized by next-generation RNA-sequencing. The expression values for 12551 commonly expressed mRNAs were represented in three dimensions via classical multidimensional scaling (MDS) and then clustered into a hierarchical structure from a pairwise Euclidean distance matrix. The horizontal axis represents Euclidean distance between samples in the three MDS dimensions. Blue, the Cancer Genome Atlas (TCGA) "normal-matched" samples; green, TRANSLATE samples from tissue unaffected by cancer; orange, Genotype-Tissue Expression (GTEx) non-cancer renal samples; red, TCGA renal cancer samples.