Study populations
For biomarker discovery using small RNA-sequencing, participants were recruited from kidney and cardiovascular clinics at the Belfast City and Royal Victoria Hospitals, Belfast. The presence of CKD was determined on the basis of ≥ 2 eGFR measurements < 60 ml/min/1.73 m2 > 3 months apart or persistent albuminuria expressed as albumin-to-creatinine ratio (ACR) ≥ 3 mg/mmol in line with the clinical definition of CKD (6). Type 2 diabetes mellitus (T2DM) was diagnosed according to the American Diabetes Association 2010 criteria (30). Patients with systemic metabolic disease other than T2DM, or those who had experienced infection within the previous month, were excluded from the study. In the discovery cohort (n = 33) patients were categorised into 3 groups: type 2 diabetes with kidney disease (n = 9) (T2DKD - T2DM and eGFR < 60 ml/min/1.73m2, ACR in 3-30mg/mmol range), type 2 diabetes and normal kidney function (n = 13) (T2DNRF - T2DM and eGFR > 60 ml/min/1.73m2, ACR ≤3mg/mmol) and no diabetes and normal kidney function (n = 11)(NDNRF).
To assess the utility of miR-190-5p as a prognostic biomarker in patients with unselected CKD patients, we employed 549 patients from a prospective cohort study – ‘Non-invasive biomarkerS Of Kidney disease’ (seNSOR), which recruited patients with unselected aetiology of CKD from outpatient clinics at the Royal Infirmary Edinburgh, Scotland between March 2017 to March 2019. Participants with acute kidney injury (AKI) were excluded which was determined using the KDIGO creatinine-based criteria (31).
In both centres, participant information collected upon enrolment included age, ethnicity, diabetes status, blood pressure, aetiology of CKD and current prescribed medications. Baseline laboratory data obtained included: serum creatinine, albuminuria (as urinary albumin to creatinine ratio, uACR) and glycated haemoglobin A1c (HbA1c). Glomerular filtration rate was estimated from serum creatinine using the 2009 Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation including/excluding race (31). For the seNSOR cohort, kidney outcome data were captured using the NHS Lothian patient record systems. The primary outcome of CKD progression was defined as reaching ESKD (starting kidney replacement therapy (KRT) or maintaining an eGFR < 15ml/min/1.73m2 for > 90 days) or > 30% reduction in kidney function from eGFR at baseline (maintained for > 90 days). Reaching ESKD alone was also used as a secondary outcome. Ethical approval was obtained from the respective Offices for Research Ethics Committees (Northern Ireland REC/14/NI/1132; Scotland REC/15/ES/0094).
Small RNA-sequencing (sRNA-Seq)
Whole blood was collected in an ethylenediaminetetraacetic acid (EDTA) collection tubes, centrifuged at 3,000 g for 10 minutes at 4°C and the supernatant aliquoted in RNase/DNase-free tubes and frozen at -80°C. RNA was extracted from 600 µl of plasma using the NextPrep™ MagnaZol™ cfRNA Isolation Kit (Bio Scientific Corp, Texas, USA) and quantified using the Qubit miRNA Assay kit (ThermoFisher Scientific, San Jose, USA).
microRNA libraries were prepared using QIAseq miRNA Library Kit (Qiagen, Hilden, Germany). Library concentrations were quantified using the Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific) and quality control was performed using a High Sensitivity NGS Fragment Analysis Kit (Agilent Technologies, Santa Clara, CA, USA). Libraries were pooled at 4nM molarity and sequenced using the NextSeq 500 High-Output Kit (Illumina, San Diego, CA, USA). Raw read quality control was performed with FastQC, adapters trimmed with Trim Galore, reads collapsed with seqcluster and aligned with Bowtie1 against miRBase mature miRNA and miRBase hairpin (miRbase Ver 22, available at http://mirbase.org). Post-alignment processing of miRBase hairpin was performed using SAMtools and EdgeR (3.36.0) for miRBase analysis. The FASTQ files and expression matrix are available in Gene Expression Omnibus (GEO). This provided a count table of miRNA reads ready for downstream analysis in Rstudio (Ver 4.1.1). As previously performed (32), exploratory principal component analysis was first performed using DESeq2 (Version 1.34.0). After count normalisation and removal of lowly expressed reads, differential expression analysis was performed using EdgeR (Version 3.36.0). A fold change > 1.5 and false discovery rate (FDR, Benjamini-Hochberg method) < 0.05 was considered significant.
Gene targets of selected miRNAs were identified using the R package multiMiR (1.6.0). multiMiR allows the user to identify validated miRNAs from multiple databases. For predicted miR-gene interactions we only selected interactions which were conserved and present in more than one database. We performed pairwise correlation of proximal tubular cell expression changes of predicted miR-190-5p gene targets. Genes anti-correlated to miR-190-5p (correlation coefficient of < -0.6) were considered potential gene targets. Finally, the pathways enriched for miR-190a-5p target genes were explored using the package ClusterProfiler (3.12.0).
RNA-Sequencing of bulk cortical tissue and fluorescence-activated cell sorting of specific kidney cell types.
As previously described (33), male 8- to 12-week-old C57BL/6 mice (Envigo) (n = 5/group) underwent laparotomy and unilateral ureteric obstruction (UUO) followed by ureteric reimplantation to reverse the obstruction for the R-UUO group. Mice were culled at day 2 after UUO (UUO-2), at day 7 after UUO (UUO-7), and in the R-UUO group at 14 days after reversal by exsanguination under terminal isoflurane anaesthesia. Immediately after culling, mice were perfused with 5 mL PBS. The kidneys were harvested, with the kidney capsule removed and placed in ice-cold PBS and processed as previously described (32). Cells were sorted for 4 positive cell type populations (PT [LTL], Mac [F4/80], endothelial [CD31], and fibroblast [PDGFR-β]. Cells were sorted into RLT buffer (Qiagen) at room temperature. RNA was extracted using RNeasy Plus Micro Kit (Qiagen) for transcriptomic analysis and for sRNA-Seq with a modification for miRNA extraction. For the amended protocol for miRs, cells were homogenized by vortexing; the lysate was pipetted onto a gDNA Eliminator spin column and centrifuged at 10,000g for 30 seconds. The gDNA column was discarded and 1.5 times volume molecular grade pure ethanol was added to the lysate and placed on a RNeasy MinElute column. The column was spun at 10,000g for a further 30 seconds, 700 µL RWT buffer was added, and the extraction continued as per manufacturer’s protocol. RNA integrity was checked using Agilent Technologies picochips. All samples utilized for sRNA-Seq had a minimum RNA integrity score of 7. For the bulk and single-population sRNA-Seq, RNA underwent sRNA-Seq by Genewiz, using the Illumina HiSeq platform, generating paired-end reads of 50 bp (n = 4 per group). For gene expression analysis of sorted cell populations, RNA integrity was 9 for all samples, cDNA was amplified using the NuGen Ovation kit and sequenced by Genewiz on the llumina HiSeq platform with 2x150bp configuration (n = 4 per group). Data available at GEO.
Analysis of miR-190a-5p in seNSOR samples
Whole blood was collected into a BD Vacutainer® SST™ Tube, centrifuged at 3,000 g for 10 minutes at 4°C, and the supernatant was aliquoted in RNase/DNase-free tubes and frozen at -80°C. C.elegans miR-39 (Qiagen) was added as an exogenous reference miRNA prior to RNA extraction from 200 µl of serum using the miRNeasy serum/plasma kit. Reverse transcription was performed using the microRNA reverse transcription kit (ThermoFisher Scientific) with specific primers for hsa-miR-190-5p (ID 000489) and cel-miR-39 (ID 000200). qPCR was performed using specific primer:probes (ThermoFisher Scientific) with Universal Master Mix II (ThermoFisher Scientific) on the Quantstudio 5 (Applied Biosystems). Cycle threshold was obtained for miR-190-5p and normalised to cel-miR-39 and relative miRNA expression calculated using the 2-Δ CT method.
Cell culture of kidney proximal tubular cells (RPTECs)
Human kidney proximal tubular epithelial cells (RPTEC/TERT1) (ATCC® CRL-4031™ LGC Limited, UK) were maintained in complete growth media consisting of DMEM/F-12 (2.5 mM L-glutamine, 15 mM HEPES, 0.5 mM sodium pyruvate, and 1200 mg L− 1 sodium bicarbonate) (Gibco, New York, USA), geneticin 2 µL mL− 1 (Gibco, New York, USA), hydrocortisone 25 ng mL− 1 (Sigma Aldrich, Gillingham, UK), ascorbic acid 3 ug mL− 1 (Sigma-Aldrich, Gillingham, UK), sodium selenite 6.7 ng mL− 1 transferrin 5.5 ug mL− 1, insulin 10 ug mL− 1 (Gibco, New York, USA), triiodo-L-thyronine 6 pM (Sigma-Aldrich, UK), prostaglandin E1 25 ng mL− 1 (Sigma-Aldrich, Gillingham, UK), rhEGF 10 ng mL− 1 (Promega, Southampton, UK). Cells were grown at 37°C, 5% CO2, in T75 cm2 flasks to 100% confluence. For passaging, cells were trypsinised, spun and resuspended in fresh media.
Manipulation of miR-190-5p levels in RPTECs.
Dharmacon miR-190a-5p miRNA mimic (C-300639-03-0002) and inhibitor (IH-300639-05-0002), along with negative control mimic (CN-001000-01-05) and miRNA hairpin Inhibitor negative control (IN-001005-01-05) were purchased from Horizon Discovery. RPTECs were transfected with 100 nM of miR-190-5p inhibitor or 50 nM miR-190-5p mimic or controls using siPORT™ NeoFX™ Transfection agent in Opti-MEM for 6 hrs before adding complete media. Media was replaced after 24 hrs and cells left for a further 48 hrs prior to being collected in Qiazol.
RNA was extracted from cells using the miRNeasy mini kit (Qiagen) following manufacturer’s instructions. cDNA was prepared using the reverse transcription kit with random hexamers (ThermoFisher Scientific). qPCR was performed using specific primer:probes for ADAM10 (Hs00153853_m1) and PPIA was used as the housekeeper. Reverse-transcription was performed using the microRNA reverse transcription kit (ThermoFisher Scientific) using specific primers for has-miR-190-5p (ID 000489) and snoRNA RNU48 (001006) as an endogenous housekeeper and qPCR performed using specific primer:probes (ThermoFisher Scientific). All qPCR was performed with Universal Master Mix II (ThermoFisher Scientific) on the Quantstudio 5 (Thermo Fisher Scientific). Relative expression was calculated as change in cycle threshold (ΔCT) for each sample to the housekeeper and plotted as 2−ΔCT and relative standard error of the mean plotted.
Single-cell analyses
For correlation analysis of ADAM10 and TLN2 single nucleus multiome (paired gene expression and chromatin accessibility) data was downloaded from GEO in Seurat (v4.4.0) format (GSE254185). Due to sparsity and technical dropouts in single cell transcriptomics data, we opted to assess the average expression levels in partitions of the nearest neighbour graph instead of individual cells for gene-to-gene correlations.
For this purpose, using the original cell annotations, the data was subset to proximal tubule (PT) cells and subclusters of the weighted shared nearest neighbour (wsnn) graph were constructed using the FindClusters function with resolution = 8 (median: 131 cells/cluster). Clusters were subsequently pruned by discarding clusters with less than 20 cells. The average log-scale expression level of ADAM10 and TLN2 was derived using the AverageExpression function and cluster-level average expression was used for visualisation as scatterplot. Expression values were fitted with a linear regression slope with respective Spearman rho and p-value. Violin plots were generated using Seurat’s VlnPlot function.
Statistical analyses
The normality of distribution for all key variables was assessed by Shapiro-Wilk test. Clinical characteristics for continuous data are expressed as mean ± standard deviation for normally distributed data and median (interquartile range) when not normally distributed. Categorical variables were expressed as counts (percentages). For continuous values with normal distribution, t-tests or ANOVA were used to compare 2 groups or > 2 groups, respectively. For continuous values without normal distribution, Mann-Whitney test was used to compare 2 groups and a Kruskal-Wallis test was employed if > 2 groups. Categorical values were assessed using a Chi-square test. For determination of linear correlation between variables, correlation coefficients (r) were estimated using Spearman’s rank tests (the data was not normally distributed). Kaplan-Meier survival curves were constructed for the primary kidney endpoint comparing above and below median miR-190a-5p separately for each uACR stage (< 3 mg/mmol, 3-300 mg/mmol, > 300 mg/mmol), with the log-rank test used to compare curves. Cox univariate and multivariate proportional hazards survival models were performed to assess factors that predicted the endpoints. MiR-190a-5p was log transformed before entering the model due to non-normal distribution. For miR-190 mimic and inhibitor experiments biological repeats were used (n = 3), with 3 technical replicates per sample and results analysed by Student’s unpaired T-test. All tests were performed using SPSS version 22.0 (IBM, Armonk, NY), R version 4.1.2 (R foundation for Statistical computing, Vienna, Austria) or Graphpad prism version 8.2.1 (Graphpad software, Boston, Massachuetts. A p value < 0.05 was considered significant.