microRNA signatures of perioperative myocardial injury after elective non-cardiac surgery: prospective observational cohort study

Inclusion/exclusion criteria

Inclusion criteria: Perioperative myocardial injury was defined as troponin I (TnI) >99^th centile (>0.04 ng.ml^-1; Troponin-I Centaur CP assay, Siemens, UK). Patients with TnI>99^th centile after surgery were matched with patients in whom TnI remained ≤99^th centile after surgery. Samples were matched by age, sex, Revised Cardiac Risk Index Score and by day after surgery. All patients with confirmed PMI had an electrocardiogram reviewed by an independent adjudication committee We excluded patients with clinically confirmed myocardial infarction and/or pulmonary embolus, those receiving renal replacement therapy and participants with elevated troponin-I (TnI) >99^th centile before surgery (Figure 1).

• Download figure
• Open in new tab

Figure 1: METS study 1 flow diagram, illustrating patient/sample selection.

METS study sample preparation

Samples were obtained once before surgery and then again on each day after surgery, until postoperative day 3. 10ml whole blood was collected in a plain tube, allowed to clot at room temperature before centrifugation (3000 rpm for 10 minutes). Serum was then aliquoted into two 2ml RNA-free tubes. The two aliquoted samples were then frozen at −70°C and subsequently used separately for troponin quantification and microRNA analysis.

METS study microRNA Isolation

Detailed methods and references are provided in Supplementary data. In brief, 200μl serum samples were screened for any obvious red discolouration (haemolysis) before isolation using the miReasy Serum/Plasma Advanced Kit (Qiagen, Denmark) in accord with manufacturer instructions. For quality control, 1μL of spike-in mixture containing UniSp2, UniSp4 and UniSp5 synthetic microRNAs (RNA Spike-in kit, Qiagen) was added to each sample (Supplementary data). 20uL RNA was isolated using RNeasy UCP MiniElute columns and stored at −20°C for subsequent batched analysis.

Circulating microRNAs

We selected microRNA-1, 21, 146, 133, 208 and 499 as established reported microRNA signatures for ACS (Supplementary table). Meta-analysis of microRNA discovery data has highlighted these microRNAs as potential biomarkers of myocardial infarction¹⁴. The isoforms selected were chosen from the clinical studies that had isolated microRNA from plasma/serum. The nomenclature was crosschecked using the miRBase database¹⁵ (Supplementary Material).

Quantitive RT-PCR

Detailed methods are provided in Supplementary Material. 2uL of RNA was reversed transcribed in a 10uL final reaction volume using the miRcury LNA RT kit. In accord with manufacturer recommendation, serum microRNA amounts are standardised by starting sample volume rather than RNA concentration. The reactions were performed in a Primus 96 plus thermal cycler (MWG Biotech, Ebersberg, Germany) for 60 minutes at 42°C, followed by incubation (inactivation of reaction) at 95°C for 5 min and then allowed to cool at 4°C. cDNA was then stored at −20°C. For qualitative PCR cDNA was diluted (1:40) and added to mastermix before plating on miRNA PCR custom made plates containing pre-designed miRCURY LNA miRNA PCR Assays in triplicate (Supplementary Data). Amplification was performed using the Applied Biosystem StepOneplus RT-PCR system (Applied Biosystems, California, USA). Cycle quantification (Cq) values were determined using the second derivative method (Applied Biosystem StepOneplus software, California, USA).

Quality control

Cq values were calibrated using an interplate calibrator control assay, UniSp3 pre-aliquoted on to each plate. The calibrated Cq data generated from the spike controls was used to compare each sample and identify outlier samples to be excluded from analysis. If the Cq values of the technical control produce similar values for each sample, isolation efficiency across all samples can be considered consistent. For this study, UniSp5 was used as the isolation technical control. If the Cq value was ≥37, then the sample and patient was flagged as an outlier and excluded from analysis. The synthetic RNA, UniSp6 was added to the RT reaction prior to cDNA synthesis. The outlier Cq values for UniSp6 were used to evaluate RT efficiency. Both hsa-miR-23a and hsa-miR-451were used to detect haemolysis, which interferes with miR detection ¹⁶. Samples that breached published limits of detection for haemolysis were excluded. Amplification efficiency was calculated (LinRegPCR software, Version 2017.1; Heart Failure Research Centre, Amsterdam, Netherlands)¹⁷ and individual melt curves for each PCR reaction were inspected (Supplementary data).

Cq normalisation and microRNA expression

The Qiagen miRcury LNA miRNA Serum/Plasma Focus Panel was used to determine two reference gene(s) for this study as recommended by consensus guidelines, since there are no universal reference microRNA defined for microRNA expression studies. ¹⁸ The focus panel has 179 microRNA assays commonly found in the serum and plasma. The NormFinder algorithm ¹⁹ was used to determine the most suitable reference microRNA (Supplementary material). MicroRNA miR-152-3p and hsa-miR-361-5p had the lowest combined stability value (0.077) and selected as reference microRNA. Expression levels were compared using the relative Cq method, where microRNA Cq values were normalised to the geometric mean of the reference microRNA Cq values to give ΔCq values. The ΔCq values were converted to a fold change (FC) by applying the formula of 2^{−ΔCq. 20}

Pathway analysis

We used DNA Intelligent Analysis (DIANA)-miRPath v.3.0 pathway analysis software²¹ to assess in an unbiased manner which transcriptomic processes may be altered by microRNAs (microT-CDS v5.0).

Extracellular vesicles samples

To examine microRNA content in extracellular vesicles, a separate cohort of patients was recruited to prepare larger volumes of plasma which required specific different preparation to the samples obtained in METS. Plasma samples were prepared from 11 patients who underwent elective non cardiac surgery.¹³ Perioperative myocardial injury was defined as plasma high sensitivity Troponin T assay ≥99^th centile (≥15 ng.L^-1) within 2 days of surgery (Cobas assay. Roche Diagnostics, Mannheim, Germany). 48h after surgery whole blood was collected into EDTA anticoagulant bottles, centrifuged at 3000 RPM for 10 minutes (Cepr-AL6 Flexicool Rota, Capricorn Laboratory Equipment, UK). The plasma layer was sterile filtered (0.8μm membrane filter, Starlabs, Hamburg, Germany) and transferred to Eppendorf tubes. The samples were immediately stored at −80°C until batch processing. At the same time blood samples were collected, an ECG was reviewed independently to ensure no patient with myocardial infarction was included.

microRNA isolation from extracellular vesicles

Plasma samples were shipped to Exiqon (Vedbaek, Denmark) to isolate microRNA in extracellular vesicles and exosomes for next generation sequencing, masked to troponin values and clinical details. Quality control checks ensured high isolation efficiency and lack of haemolysis.¹⁶ (Supplementary Material) Library preparation was done using the NEBNext® Small RNA Library preparation kit (New England Biolabs, Massachusetts, USA). The library pool was sequenced on a NextSeq500 sequencing instrument (Illumina, San Diego, USA). Raw data was de-multiplexed and FASTQ files for each sample was generated using the bcl2fastq software (IIllumina, San Diego, USA). The MicroRNA sequence reads (FASTQ) files were then mapped to Homo sapien genome (GRCh37) and the microRNA sequence database, MiRBase V20, to generate MicroRNA sequence counts using Bowtie2 (Version 2.2.4.) software.²² MicroRNA counts were normalised for library size using Trimmed mean of M (TMM) method and differential expression of TMM analysis performed using EdgeR statistical software package.²³ Differentially expressed microRNA were used to perform Gene ontology enrichment analysis using topGO.²⁴

Statistical Analysis

Clinical characteristics of the patients were stratified according to presence, or absence, of troponin elevation >99^th centile values. Categorical data are summarised as absolute values (percentage). Continuous data are presented as mean (SD), or median (IQR), unless stated otherwise. ΔCq values were checked for normality using the Shapiro–Wilk test. For microRNA assays that were expressed in >95% across all samples, ΔCq values were analysed using a multi-level (mixed error-component) model. The levels examined were; presence/absence of perioperative myocardial injury (>99^th centile) and sampling time (before, or after, surgery).

Individual comparisons between groups were calculated using post-hoc Bonferroni tests. For microRNA expressed in <95% samples, McNemar Chi-square analysis was performed. P<0.05 was considered statistically significant. All statistical analyses were undertaken using NCSS 12 (Kaysville, UT, USA). Data have been deposited in NCBI’s Gene Expression Omnibus²⁵ and are accessible through GEO Series accession number GSE144777.

Sample Size Calculation

Receiver operating character curve analysis has demonstrated that miR-499, miR-1-3 and miR-21 can be used to identify patients with acute coronary syndrome, area under the curve (AUC) = 0.89. ⁹ If these microRNA were present and/or elevated in patients at the time of detecting perioperative myocardial injury, >11 patients would be required to detect a difference in relative expression serum microRNA after surgery (α=0.01; 1-β=0.9). The sample size calculation was performed using the pROC package (R software, v3.4.3).²⁶

Patient characteristics

Patients recruited from METS who developed myocardial injury (median TnI: 0.08ng.ml^-1 (interquartile range: 0.05-0.13ng.ml^-1)) were clinically similar to patients in whom troponin remained within normal limits (Table 1).

Plasma circulating microRNA expression

Circulating microRNA expression levels increased for hsa-miR-1-3p (mean fold-change:3.99 (95%CI: 1.95-8.19); F_[1,47.5] =24.16; p<0.001) and hsa-miR-133-3p (fold-change:5.67 (95%CI: 2.94-10.91); F_[1,42] = 31.13; p<0.001) after surgery, but were independent of elevated troponin levels after surgery (mixed model level accounting for presence/absence of PMI: p>0.05). Expression levels of hsa-miR146a-5p decreased after surgery only in patients with elevated troponin after surgery (F_[1,34.5]=6.43, p=0.03). There was no change in expression for hsa-miR-21-5p after surgery (Figure 2). In patients with troponin elevation after surgery, 91% (20/22) of the samples detected hsa-miR-208b-3p, compared with 50% (12/24) with troponin levels below the 99^th centile after surgery (odd ratio (OR):10.00 (95%CI:1.90-52.55); p<0.01)). hsa-miR-499a-5p was also detected in a higher proportion of samples from patients with elevated troponin levels after surgery (19/22 vs 13/24; OR:5.36 (95% CI 1.32-24.17); p<0.05) surgery. (Supplementary Material; Figure 1)

• Download figure
• Open in new tab

Figure 2: Heatmaps showing relative expression levels (Fold change) of detectable serum microRNA before and after non-cardiac surgery, in relation to the development of myocardial injury.

A-D: Heatmap showing relative serum expression levels for each matched sample using the relative Cq method. MicroRNA Cq values were normalised to the geometric mean of the reference microRNA Cq values to give ΔCq values. The ΔCq values were converted to a fold change (FC) by applying the formula of 2^−ΔCq. Bars below each microRNA heatmap indicate range of fold-change expression (log₂ scaled). Black bars denote sample that failed strict quality control threshold. E-H: Relative change in MicroRNA expression levels after surgery. Bars are mean fold change with 95% confidence intervals. P-values refer to comparison; before surgery vs after surgery (analysed by mixed-model repeated measures and Bonferroni test); * p<0.05, ** p <0.01 **p<0.001. PMI: perioperative myocardial injury (defined as troponin I (TnI) >99th centile (>0.04 ng.ml¹).

Pathway analysis

Nine signalling pathways were identified by DIANA in-silica pathway analysis (false-discovery rate <0.01) for microRNA that changed after surgery (Figure 3A). This agnostic analysis approach highlighted the involvement of adrenergic signalling in cardiomyocytes (hsa04261; p=0.005) and arrythmogenic right ventricular cardiomyopathy (hsa05412; p= 0.006). All microRNAs that were elevated in serum after surgery (n=4) regulated expression of SLC8A1, a gene that controls calcium overload through the NCX1 antiporter membrane protein by removing calcium from cells (Figure 3B). Subunits of protein phosphatase (which regulates cardiac excitability through dephosphorylation of cardiac ion channels), angiotensin receptors and the ATP sarcoplasmic/endoplasmic reticulum calcium pump were also identified as targets of microRNAs that changed after surgery. (Figure 3B)

• Download figure
• Open in new tab

Figure 3: Bioinformatic analysis identifying pathways and genes regulated by serum microRNAs that change after non-cardiac surgery.

A. Signalling pathways (false-discovery rate threshold p value <0.01), as derived from DIANA mir-Path v3.0, which bioinformatically links microRNA with their regulatory roles in biological and signalling pathways. Cardiac-specific pathways are highlighted in red. B. Shared genes regulated by serum microRNA identified as changing after non-cardiac surgery. The four microRNAs identified as changing after surgery regulate mRNA expression of SLC8A1 (NCX1 sodium-calcium exchanger) and subunits of the protein phosphatase enzyme.

Patient characteristics

Characteristics of the 11 patients from whom samples were obtained 48 hours after surgery are presented in Table 2. The median (IQR) troponin-T values were 19 (17-23) ng.L ^-1 in patients with troponin elevation, compared with patients in whom plasma troponin remained lower than the 99th centile value (median: 6 (5-9) ng.L^-1; p=0.006).

Next generation sequencing

A mean of 9.8 million read were obtained for each sample. Mean mapping rate to the genome and microRNA annotation was 62.9%. Mapping to miRbase, with normalisation for differences in the sequences depth within each sample, revealed 345 unique microRNA within all samples where Tags per million (TPM) ≥1 and 164 unique microRNAs where TPM ≥10. Thirty-two microRNAs were differentially expressed in patients with troponin elevation after surgery (Figure 4A). After correcting for false discovery rate <0.05, three microRNAs remained differentially expressed (Figure 4B). The top two biological processes identified through Gene Ontology were associated with cardiac muscle physiology. (Figure 4C; Supplementary material).

• Download figure
• Open in new tab

Figure 4: Next generation sequencing of microRNAs detected in extracellular vesicles.

A. Heat map showing all differentially expressed microRNA (P<0.05), comparing patients with PMI (hsTnT >14ngL^-1) versus no troponin elevation. Values are Trimmed mean of M values. (Z score transformation; red indicates higher expression levels).

B. Volcano plot of microRNAs identified in all samples. Green dots denote differentially expressed microRNA with adjusted p-value (FDR; false discovery rate) <0.05. Orange dots denote differentially expressed microRNA with p-value <0.05 and red dots denote microRNA that were not differentially expressed.

C. GO terms that were significantly enriched among the differentially expressed microRNA, comparing patients with PMI (hsTnT >14ngL^-1) versus no troponin elevation.

D. Refined hypothesis of mechanisms contributing to perioperative myocardial injury, summarising chief microRNA findings and their potential roles in cardioprotection. Blue coloured microRNA species have been associated with cardioprotection, while red coloured microRNAs promote cardiac injury in the setting of ischaemia/reperfusion injury. Trop I-troponin I; Trop C-troponin C; Trop T-troponin T. Adrenergic stress is a common feature of the perioperative period, leading to the release of microRNA from cardiac cells, which may be cardiomyocytes, fibroblasts or other cell types. In individuals at highest risk of elevated troponin after surgery (e.g. elderly, patients with cardiac failure, diabetes mellitus, renal disease), pre-exisiting (preoperative), or acquired, loss of cardioprotective signalling mechanisms promotes the likelihood of calcium overload from adrenergic stress, resulting in mitochondrial dysfunction, bioenergetic compromise and myocardial injury manifest clinically by the measurement (at low levels) of troponin T in plasma/serum.