Droplet digital PCR as a novel detection method for quantifying microRNAs in acute myocardial infarction

International Journal of Cardiology 257 (2018) 247–254

Contents lists available at ScienceDirect

International Journal of Cardiology journal homepage: www.elsevier.com/locate/ijcard

Droplet digital PCR as a novel detection method for quantifying microRNAs in acute myocardial infarction S. Robinson a,b,1, M. Follo c,1, D. Haenel a,1, M. Mauler a,1, D. Stallmann a,1, M. Tewari d,e,f,g,h,1, D. Duerschmied a,1, K. Peter b,h,1, C. Bode a,1, I. Ahrens a,i,1,2, M. Hortmann a,⁎,1,2 a

Department of Cardiology and Angiology I, Heart Center Freiburg University, Faculty of Medicine, University of Freiburg, Germany Department of Medicine, Monash University, Melbourne, Australia Department of Medicine I, Lighthouse Core Facility, Medical Center - University of Freiburg, Faculty of Medicine, University of Freiburg, Germany d Department of Internal Medicine, Hematology/Oncology Division, University of Michigan Medical School, Ann Arbor, MI, USA e Department of Biomedical Engineering, University of Michigan, Ann Arbor, MI, USA f Biointerfaces Institute, University of Michigan, Ann Arbor, MI, USA g Center for Computational Medicine and Bioinformatics, University of Michigan, Ann Arbor, MI, USA h Baker IDI Heart and Diabetes Institute, Melbourne, Australia i Department of Cardiology, Augustinerinnen Hospital, Academic Teaching Hospital, University of Cologne, 50678 Cologne, Germany b c

a r t i c l e

i n f o

Article history: Received 21 July 2017 Received in revised form 22 October 2017 Accepted 31 October 2017 Keywords: ST-segment elevation myocardial infarction Droplet digital PCR Quantitative real time PCR Micro-RNAs Ischemia-reperfusion injury

a b s t r a c t Background: micro-RNAs have shown promise as potential biomarkers for acute myocardial infarction and ischemia-reperfusion injury (I/R). Most recently droplet digital polymerase chain reaction (ddPCR) has been introduced as a more reliable and reproducible method for detecting micro-RNAs. Aims: We aimed to demonstrate the improved technical performance and diagnostic potential of ddPCR by measuring micro-RNAs in ST-elevation myocardial infarction (STEMI). Methods: A dilution series was performed in duplicate on synthetic Caenorrhabditis elegans-miR-39, comparing quantitative real-time PCR (qRT-PCR) and ddPCR. We used ddPCR and qRT-PCR to quantify the serum levels of miR-21, miR-208a and miR-499 between STEMI patients (n = 24) and stable coronary artery disease (CAD) patients (n = 20). In STEMI, I/R injury was assessed via measurement of ST-segment resolution. Results: In the dilution series, ddPCR demonstrated superior coefficient of variation (12.1%vs.32.9%) and limit of detection (0.9325 vs.2.425copies/μl). In the patient cohort, ddPCR demonstrated greater differences in miR-21 levels (2190.5 vs. 484.7 copies/μl; p = 0.0004 for ddPCR and 136.4 vs. 122.8 copies/μl; p = 0.2273 for qRT-PCR) and in miR-208a (0 vs. 24.1 copies/μl, p = 0.0013 for ddPCR and 0 vs. 0 copies/μl, p = 0.0032 for qRT-PCR), with similar differences observed in miR-499 levels (9.4 vs. 81.5 copies/μl, p b 0.0001 for ddPCR and 0 vs. 19.41 copies/μl, p b 0.0001 for qRT-PCR). ddPCR also more accurately defined STEMI for all miRNAs (area under the curve (AUC) of 0.8021/0.7740/0.9063 for miR-21/208a/499 with ddPCR vs. AUC of 0.6083/0.6917/0.8417 with qRTPCR). However, there was no association between miR-21/208a/499 levels and ischemia-reperfusion injury. Conclusion: ddPCR demonstrates superiority in both technical performance and diagnostic potential compared to qRT-PCR. Ultimately, this supports its use as a diagnostic method for quantifying micro-RNAs, particularly in large multi-center trials. © 2017 Elsevier B.V. All rights reserved.

1. Introduction Ischemic heart disease is the leading cause of morbidity and mortality worldwide [1]. Early diagnosis and intervention is integral in ⁎ Corresponding author at: Department of Cardiology and Angiology I, Heart Center Freiburg University, Faculty of Medicine, University of Freiburg, Hugstetter Street 55, 79106 Freiburg, Germany. E-mail address: [email protected] (M. Hortmann). 1 This author takes responsibility for all aspects of the reliability and freedom from bias of the data presented and their discussed interpretation. 2 Drs Hortmann and Ahrens contributed equally to this article.

https://doi.org/10.1016/j.ijcard.2017.10.111 0167-5273/© 2017 Elsevier B.V. All rights reserved.

reducing injury in acute myocardial infarction (AMI). The current gold standard for diagnosis uses high-sensitivity cardiac troponins (hs-cTn), however these markers are not without their limitations. hs-cTn have compromised specificity, with an increased detection of no-ischemic damage [2–4]. Moreover, raised troponins are not entirely AMI specific, and they can be chronically raised in patients with congestive cardiac failure or chronic kidney disease [2–5]. The establishment of novel biomarkers for AMI has therefore become an important focus of medical research. Furthermore, there is a relative paucity of established biomarkers for reperfusion injury. In recent years miRNAs have been identified as promising markers for a number of diseases, including

248

S. Robinson et al. / International Journal of Cardiology 257 (2018) 247–254

AMI [6]. miRNAs are short (~22 nucleotide) non-coding single-stranded RNA molecules involved in the regulation of messenger RNA (mRNA) via inhibitory effects on translation and/or stability, ultimately influencing gene expression [7]. Under physiological or pathological conditions, miRNAs can be released from their cells of origin into the circulation and are thought to act upon cells and tissues at distant sites [8]. miRNAs are attractive candidates for disease biomarkers. They are highly stable in the circulation and measureable in a variety of bodily fluids [7,8]. Moreover, their levels can change significantly in pathological states, and some miRNAs show high tissue and disease specificity [7]. A number of miRNAs have been investigated as valid biomarkers in AMI. Of particular interest for this study are miRNA-21, miRNA-208a and miRNA-499. These miRNAs have been shown to be upregulated in AMI, and miRNA499 has even shown potential as a biomarker for reperfusion injury [9–15]. Although results are encouraging, a variety of issues have arisen which have limited their use [16]. Foremost amongst these limitations is the current method used to quantify miRNAs, qRT-PCR [17]. ddPCR is a relatively novel method of PCR, which partitions a 20 μl sample into ~20,000 individual droplets [18]. The ddPCR system uses a Poisson statistical analysis of fluorescent signals from positive and negative droplets to allow for absolute quantification [18]. ddPCR has demonstrated a number of advantages over conventional qRT-PCR, which may aid in mitigating the current limitations on using circulating miRNAs as biomarkers. In experiments to date, ddPCR has demonstrated a high degree of linearity and quantitative correlation in measuring miRNAs within its dynamic range, and has shown greater reproducibility and less inter- and intra-assay variability compared to qRT-PCR [18,19]. Through these advantages ddPCR could offer greater day-today comparability and reliability and hence greater utility as both a diagnostic method as well as in validating miRNAs in large multi-center clinical trials. The aim of this study was to investigate ddPCR as a novel method of quantifying serum miRNAs and to determine whether its use leads to an improvement in the diagnostic potential of validated miRNAs for AMI and I/R injury. 2. Methods 2.1. Synthetic oligonucleotide dilution series of C. elegans-miR-39 A lyophilised 5′-phosphorylated synthetic oligonucleotide for Caenorhabditis elegans-miR-39 (C. elegans-miR-39) (Integrated DNA Technologies) (Table A.1), with a known starting mass of 7.6 nmol was centrifuged according to the company's protocol and diluted in molecular grade Tris-EDTA (TE) buffer (Thermo Fisher Scientific) to a final concentration of 10 pmol/μl. A 12-step dilution series using nuclease-free water (Applichem Panreac) from 2500 copies/μl to 0 copies/μl was performed for C.elegans-miR-39 (Table A.2 and Fig. A.1). Each sample was briefly centrifuged for 10 s at 8000g before being diluted into the next sample in the dilution series. Each sample then underwent reverse transcription (RT) using 4.16 μl/well nuclease-free water, 1.50 μl/well RT buffer, 0.15 μl/well 100 nM dNTP, 0.19 μl/well RNAse inhibitor, 1.00 μl/well multi-scribe reverse transcriptase and 3 μl/well specific RT primer (Applied Biosystems, Inc. ID: 000200). For RT reaction 10 μl mastermix and 5 μl sample were combined and spun on a microplate centrifuge at 4 °C for 2 min at 2000g (Thermo Scientific). Samples underwent a 15 μl thermal cycling protocol using the C1000 Touch™ Thermal Cycler (BioRad) at 16 °C for 30 min, 42 °C for 30 min, and 85 °C for 5 min and held at 4 °C. The RT product was then further used for qRT-PCR and ddPCR in equal volume. For qRT-PCR, 7.67 μl/well nuclease-free water, 10 μl/well universal PCR master mix and 1 μl/well of a specific 20 × hydrolysis primer/probe (Applied Biosystems, Inc. ID: 000200) were combined, briefly centrifuged and partitioned into 18.67 μl/well with 1.33 μl/well RT product. Each sample was prepared in duplicate. Samples were spun on a microplate centrifuge at 4 °C for 2 min at 2000g (Thermo Scientific). Samples underwent a 20 μl qRT-PCR protocol at 95 °C for 10 min, 95 °C for 15 s and 60 °C for 60 s, before being repeated for 39 more cycles and held at 4 °C. All qRT-PCR data was analyzed using CFX Manager™ (BioRad). For ddPCR, 7.67 μl/well nuclease-free water, 10 μl/well ddPCR™ supermix for probes (no dUTP) (BioRad) and 1 μl/well 20× specific hydrolysis primer/probe were combined, briefly centrifuged and partitioned into 18.67 μl/well with 1.33 μl/well RT product. 20 μl of the sample was pipetted into each well in an 8-well cassette with 70 μl droplet generation oil for probes (BioRad) and placed in the QX200™ Droplet Generator (BioRad). 40 μl of the droplet-formed sample was pipetted into separate wells. Each sample was prepared in duplicate. Samples were transferred to the C1000™ Thermal Cycler (BioRad) and underwent thermal cycling set for 40 μl with 2.5 °C/s ramp rate at 95 °C for 10 min, 94 °C

for 30 s, and 60 °C for 60 s, before being repeated for 39 more cycles and thereafter taken to 98 °C for 10 more minutes and held at 12 °C. The technical performance of ddPCR versus qRT-PCR was compared statistically by a number of methods. Coefficient of variation was calculated by the equation: CV% ¼ ðStandard deviation=MeanÞ  100% Before calculating the CV% for each diluent in qRT-PCR, the Ct values were converted to absolute copies using Pfaffl analysis [20]: R ¼ Etarget ΔCPtarget ðcontrol−sampleÞ where Etarget represents the efficiency of each individual qRT-PCR run according to the slope of the line of best fit: Etarget ¼ −1 þ 10ð−1=slopeÞ The control for the Pfaffl equation was calculated by calculating the mean of the highest concentration (i.e. 2500 copies/μl), and comparing that to each sample. Each sample could then be converted to absolute copies by the following equation: Absolute copies ¼ R  2500 The qRT-PCR data could then be analyzed for CV% and compared with ddPCR. Reduction in CV% was calculated by the equation:

Difference in CV% ¼ CV%qRT−PCR −CV%ddPCR = CV%qRTPCR The lower limit of linear range (LLLR) was determined by runs-testing [21], removing successive dilution points until the p-value was N0.05 (Prism Version 5.0c software). The limit of detection (LOD) was calculated by the following equation [22]: LOD ¼ bxNblank þ 1:645σ blank þ 1:645σ low where bx N blank = mean of negative controls σblank = standard deviation of negative controls σlow = standard deviation of lowest concentration measurement The limit of quantification (LOQ) was defined as the lowest concentration tested which still remained above or equal to both the LLLR and the LOD. Linearity of ddPCR and qRT-PCR for each dilution series was assessed using the r-value for each line of best fit (GraphPad Prism v5.0). Values of quantification are presented as log transformed for both qRT-PCR and ddPCR measurements in order to display every dilution point. 2.2. Patient samples 2.2.1. Patient population In this single-center prospective study, peripheral venous blood was obtained from 24 STEMI patients and 20 patients with stable CAD between October 2013 and August 2015. Patients with STEMI as evidenced by ST-elevation N0.1 mV in at least two contiguous leads were included. Major exclusion criteria included LVEF b30% and cardiogenic shock. The final study population consisted of 44 patients. The protocol of this study conforms to the ethical guidelines of the 1975 Declaration of Helsinki and was approved by the institutional ethical committee of the University Medical Center of Freiburg (CF16/15462016000808). Written informed consent was obtained from all patients. 2.2.2. Percutaneous coronary intervention All STEMI patients received aspirin (minimum of 250 mg) and an ADP receptor blocker (prasugrel 60 mg, ticagrelor 180 mg or clopidogrel 600 mg). Unfractionated heparin (5000 U) was administered prior to angiography. Eptifibatide was given at the discretion of the cardiologist. Peripheral blood was obtained 5 ± SD 2.7 h after PCI. Blood was drawn post-PCI in order to also assess for an association between miRNA levels and ST-resolution post-PCI, indicating clinically relevant I/R-injury. 2.2.3. miRNA extraction Serum samples underwent miRNA extraction according to the manufacturer's instructions (Qiagen miRNeasy Serum/Plasma Kit, ID: 217184). After the initial Qiazol denaturation step, all samples were spiked with 5 μl of 9.635 × 105 copies/μl C. elegans-miR-39 to act as an exogenous control. 2.2.4. Reverse transcription, qRT-PCR and ddPCR Samples underwent reverse transcription, qRT-PCR and ddPCR according to the above-mentioned protocols, using specific hydrolysis primer/probes for each miRNA analyzed (ThermoFisher, ID: 000397, 000511, 001352). Every sample also underwent RT and qRT-PCR/ddPCR for C. elegans-miR-39, performed in duplicate in separate wells to act as an exogenous control. Final ddPCR concentrations for each miRNA were equal to the results

S. Robinson et al. / International Journal of Cardiology 257 (2018) 247–254

249

as calculated by the software multiplied by the dilution factor of the template for the reverse transcription and ddPCR reactions. 2.2.5. Ischemia-reperfusion injury ECGs were analyzed within the STEMI group before and 2.25 h (1.5–3.5) after PCI. A decrease in ST-elevation by a minimum of 70% was set as the cut-off for a patient to be considered adequately reperfused [23]. Four of the STEMI patients had to be excluded from analysis, due to a lack of required ECGs in the hospital's database. 2.2.6. Statistical analysis Continuous variables are presented as mean ± standard deviation (SD) if found to follow a Gaussian distribution according to a D'Agostino–Pearson omnibus normality test, or as a median and interquartile range if non-Gaussian. Normally distributed variables were compared using Student's t-test and non-normally distributed variables using the Mann– Whitney U test. Categorical variables are presented as percentages. A Chi-square analysis compared categorical variables. A ROC analysis compared the sensitivities and specificities of qRT-PCR and ddPCR, with the endpoint analysis being the AUC. miR-499 and miR-21 levels were further analyzed as a ratio of miR-499/miR-21 for each patient, and a Mann– Whitney U test as well as a ROC analysis was performed for both qRT-PCR and ddPCR. miR-21/208a/499 levels between STEMI patients with I/R injury were compared to those with adequate reperfusion using a Mann–Whitney U test, and ROC analyses were performed for each miRNA. Analyses for correlations between clinical characteristics and miR-21/miR-208a/miR499 levels were performed by Pearson analysis for parametric data or a Spearman test for non-parametric data. A p-value of ≤0.05 was considered statistically significant for all analyses except for the analysis of correlation. This analysis required correction by Bonferroni's method. This set the new statistically significant p-value at 0.0029. All analyses were performed using Graph Pad Prism v5.0 (GraphPad Software, Inc., La Jolla, CA).

3. Results 3.1. Dilution series of synthetic C. elegans-miR-39 ddPCR consistently exhibited a superior coefficient of variation (CV%) between most dilution points across all four runs (Figs. 1a/b and A.2). qRT-PCR had an average CV% of 32.9%, whilst ddPCR's average CV% was 12.1%. ddPCR demonstrated an average decrease in CV% of 50.3%. ddPCR also demonstrated superior linearity across most runs (Fig. A.3), with an average r2 of 0.983 compared to 0.945 for qRT-PCR. Finally, ddPCR demonstrated superior or equal LOD, LOQ and LLLR across all runs (Fig. 1c). It must be noted that with absolute quantification, absolute numbers of copies detected were consistently lower than absolute copies predicted across every run, with an average detection efficiency of 40.1% (ranging between 17.2 and 73.4%) (Table A.3). 3.2. Patient characteristics The patient population consisted of 73% males and the average age was 65.7 ± SD 13.4 years. Table A.4 presents the baseline characteristics of STEMI and stable CAD groups. A higher level of the cardiac biomarkers creatine kinase (CK) and cTnT at peak were demonstrated in STEMI patients. STEMI patients were more likely to receive prasugrel, ticagrelor and eptifabitide. Stable CAD patients were more likely to be overweight. 3.3. Comparison of ddPCR and qRT-PCR for the diagnosis of AMI ddPCR altogether demonstrated superior sensitivity, specificity, variability and significance of miR-21/208a/499 levels comparing patients with stable CAD and STEMI patients (Figs. 2–4). ddPCR generally exhibited lower variation between duplicate measurements, with an average CV% of 6.77% for miR-21, 66.4% for miR-208a and 41.1% for miR-499 compared to 22.1%, 81.5% and 36.4% for qRT-PCR respectively. For miR-21, qRT-PCR demonstrated minimal differences in median levels between stable CAD and STEMI groups (136.4 copies/μl vs. 122.8 copies/μl p = 0.2273), whereas ddPCR reached statistical significance (2190.5 copies/μl vs. 484.7 copies/μl; p = 0.0004) (Fig. 2). For miR-208a, qRT-PCR demonstrated inferior significance (0 copies/μl vs. 0 copies/μl; p = 0.0032 for qRT-PCR, and 0 copies/μl vs. 24.1 copies/μl; p = 0.0013 for ddPCR) (Fig. 3). For miR-499, qRT-PCR and ddPCR

Fig. 1. Comparison of qRT-PCR and ddPCR, including coefficient of variation (CV%), limit of detection (LOD), limit of quantification (LOQ) and lower limit of linear range (LLLR) for a dilution series of C.elegans-mir-39 performed in duplicate across four separate runs. (a) Trendline comparing coefficient of variation (CV%) between qRT-PCR (blue trendline) and ddPCR (red trendline) at different concentrations for a dilution series of C.elegans-mir-39 performed in duplicate across four separate runs. (b) Average CV% for qRT-PCR and ddPCR and difference in CV% for every run. (c) Average LOD, LOQ and LLLR for qRT-PCR and ddPCR for every run. (d) Radar chart displaying the main characteristics of qRT-PCR versus ddPCR.

demonstrated equal significance (0 copies/μl vs. 19.41 copies/μl; p b 0.0001 for qRT-PCR, and 9.4 copies/μl vs. 81.5 copies/μl; p b 0.0001 for ddPCR) (Fig. 4). ddPCR also exhibited decreased variability in patient measurements, with an overall decrease in CV% of 10.2%. The measurements of absolute copies for each patient were correlated between ddPCR and qRT-PCR for both miR-21 (Spearman r = 0.6421) and miR-499 (Spearman r = 0.8393), but not for miR-208a (Spearman r = 0.02528) (Fig. A.4). A ROC analysis demonstrated that ddPCR more accurately classified STEMI as reflected by an increased AUC for miR-21 (AUC of 0.8021 for ddPCR vs. AUC of 0.6083 for qRT-PCR), miR-208a (AUC of 0.7740 for ddPCR vs. AUC of 0.6917 for qRT-PCR) and miR499 (AUC of 0.9063 for ddPCR vs. AUC of 0.8417 for qRT-PCR) (Figs. 2–4). No single miRNA analyzed was superior to cTnT in diagnostic utility (AUC of 0.9432). miRNA-499/miRNA-21 as expressed as a ratio demonstrated greater significance and AUC for both qRT-PCR and ddPCR than either miRNA alone (Fig. A.5) (0 vs. 0.2097; p b 0.0001; AUC of 0.8688 for qRT-PCR and 0.007 vs. 0.16; p b 0.0001; AUC of

250

S. Robinson et al. / International Journal of Cardiology 257 (2018) 247–254

0.9792 for ddPCR). ddPCR demonstrated an AUC that outperformed cTnT in diagnostic utility (AUC of 0.9432 for cTnT vs. AUC of 0.9792 for miR499/miR-21 as measured by ddPCR). 3.4. Comparison of qRT-PCR and ddPCR for the diagnosis of ischemia-reperfusion injury Neither qRT-PCR nor ddPCR were able to demonstrate significant differences in miR-21, miR-208a or miR-499 levels between patients who were successfully reperfused and those who suffered I/R injury (Figs. A.6 and A.7). 3.5. Correlation analysis Tables A.5/6/7/8 display correlation analyses for main clinical data for miR-21, miR-208a, miR-499 and miR-499/miR-21 respectively. After Bonferroni's correction, the only clinical correlations to reach statistical significance included a negative correlation between miR-21 and peak CK levels within 24 h of presentation as measured by ddPCR, a positive correlation between miR-499 and peak CK levels as measured by ddPCR and qRT-PCR. We further demonstrated a positive correlation

between miR-499/miR-21 ratio for peak CK-MB as measured by qRTPCR, cTnT as measured by ddPCR and peak CK as measured by both systems. 4. Discussion Droplet digital PCR (ddPCR) is a relatively novel method of PCR that allows absolute quantification of nucleic acids. With the exception of Hindson et al., at current there has been limited experimentation comparing ddPCR to qRT-PCR for quantifying miRNAs [19]. Furthermore, to the best of our knowledge, there has been negligible experimentation of ddPCR and its utility for diagnosing AMI or reperfusion injury with miRNAs. With this report, we aim to demonstrate these proposed advantages of ddPCR over qRT-PCR using synthetic oligonucleotides and validated endogenous miRNA markers for AMI and reperfusion injury in patient serum. 4.1. ddPCR demonstrates superior utility for the quantification of synthetic C. elegans-miR-39 in a dilution series ddPCR demonstrated superior technical qualities, including greater reproducibility and superior linearity across every run. Furthermore,

Fig. 2. Quantification of circulating miRNA-21-5p (a) Duplicate measurements of miR-21 for each patient sample represented as median and interquartile range (IQR) between stable CAD and STEMI groups for both qRT-PCR (blue) and ddPCR (red) with corresponding p-values. (b) ROC analysis from data in a, representing the sensitivity and specificity for cTnT (black), qRTPCR (blue) and ddPCR (red) with corresponding area under the curve values. (c) Median and IQR of all measurements by qRT-PCR and ddPCR for miRNA-21 for STEMI patients and patients with stable CAD given in copies/μl.

S. Robinson et al. / International Journal of Cardiology 257 (2018) 247–254

251

Fig. 3. Quantification of circulating miRNA-208a-3p (a) Duplicate measurements of miR-208a for each patient sample represented as median and interquartile range (IQR) between stable CAD and STEMI groups for both qRT-PCR (blue) and ddPCR (red) with corresponding p-values. (b) ROC analysis from data in a, representing the sensitivity and specificity for cTnT (black), qRT-PCR (blue) and ddPCR (red) with corresponding area under the curve values. (c) Median and IQR of all measurements by qRT-PCR and ddPCR for miRNA-208a for STEMI patients and patients with stable CAD given in copies/μl.

ddPCR demonstrated superior sensitivity, as evidenced by a superior LOD, LLLR and LOQ. In our experiments, we noted absolute measurements significantly lower than the theoretically input copies. Hindson et al. also reported absolute measurements between 49 and 114% of the theoretical input across all oligonucleotides investigated [19]. A number of factors could account for this, of which the most likely is suboptimal reverse transcription efficiency [24]. Despite this, both qRT-PCR and ddPCR underwent the same cycling conditions with the same PCR products. Hence, the results presented here allow us to compare the two methods in depth. In summary, ddPCR demonstrated superior technical qualities for the analysis of synthetic C. elegans-miR-39. Therefore, the issues of excess variability, sensitivity and specificity for detecting miRNAs are somewhat reduced with ddPCR. 4.2. miRNA-21, miRNA-208a and miRNA-499 levels are significantly dysregulated in STEMI patients compared to patients with stable CAD All miRNAs were significantly dysregulated when quantified by ddPCR. miR-21 levels decreased in STEMI patients. These findings conflict with current studies, which demonstrate an increase in miR-21 levels in patients with AMI [9,10]. In Wang's and Oliveri's studies,

miR-21 levels were measured at time of admission, whereas for this project, blood was taken 5 ± SD 2.7 h after PCI. It is plausible that miR-21 levels increase in the acute phase of AMI and decrease according to its kinetics, or as a direct result of PCI, possibly demonstrating therapeutic success. Further human time-course studies investigating levels before and after therapeutic intervention are required. Moreover, discrepancies between our study and other studies may be partially attributable to different normalisation methods undertaken. In both Wang's and Oliveri's studies, different normalisation protocols were used. This variation reflects the current lack of consensus in gold-standard reference controls for miRNA analysis. Regardless, in our project a dysregulation in miR-21 reached statistical significance for ddPCR but not for qRTPCR, giving clinically relevant evidence to the superiority of ddPCR over qRT-PCR for quantifying differences in miRNA expression in serum. ddPCR also demonstrated improved significance for miR-208a levels. miR-208a levels were increased significantly for both qRT-PCR and ddPCR, which is consistent with the current literature [11,25]. However, miR-208a has shown variable data concerning its detection [11, 25–27]. This is also consistent with our findings, where the levels of miR-208a were detectable in only 5% of CAD patients and 42% of STEMI patients by qRT-PCR and 40% and 75% respectively by ddPCR. Given that miR-208a is only expressed within the heart, if its extraction

252

S. Robinson et al. / International Journal of Cardiology 257 (2018) 247–254

Fig. 4. Quantification of circulating miRNA-499-5p (a) Duplicate measurements of miR-499 for each patient sample represented as median and interquartile range (IQR) between stable CAD and STEMI groups for both qRT-PCR (blue) and ddPCR (red) with corresponding p-values. (b) ROC analysis from data in a, representing the sensitivity and specificity for cTnT (black), qRT-PCR (blue) and ddPCR (red) with corresponding area under the curve values. (c) Median and IQR of all measurements by qRT-PCR and ddPCR for miRNA-499 for STEMI patients and patients with stable CAD given in copies/μl.

and RT were optimized, then miR-208a could become a highly specific and sensitive biomarker [11,28,29]. Finally, ddPCR and qRT-PCR demonstrated equal significance for miR-499 levels, whereby both methods showed an increase in STEMI patients. These findings are also consistent with the current literature [11,12,30]. In past studies, miR-499 has shown differences from 80 to 3 × 105-fold between AMI and control groups [12,30]. In this study, ddPCR only demonstrated an approximately 9-fold difference between median values. The reason for this may once again be due to the time the blood was drawn, this being after PCI, whereas the abovementioned studies analyzed blood taken at the time of admission. In Chen's study, there was a 58% decrease in miR-499 levels 24 h after PCI compared to admission, and 48% lower levels of miR-499 in those who received PCI compared to those who did not [13]. Therefore, the lower levels found in our study may be a marker of therapeutic reperfusion. In summary, ddPCR demonstrated improved significance and sensitivity for the quantification of miR-21/208a/499 in STEMI and stable CAD patients.

4.3. ddPCR more accurately classifies STEMI cases from stable CAD patients using miRNA-21, miRNA-208a and miRNA-499 ROC analyses demonstrated that ddPCR more accurately differentiated STEMI patients from stable CAD patients for all miRNAs analyzed. Furthermore, ddPCR demonstrated diminished variability, with lower CV% between duplicate measurements for miR-21 and miR-208a, and lower CV% between mean values for every miRNA. We therefore propose that ddPCR is more resistant to sample variability and the presence of endogenous inhibitors, which have been problematic in interfering with PCR efficiency. This is in keeping with prior studies, which have demonstrated ddPCR's increased tolerance to common PCR inhibitors such as EDTA and heparin [31]. The improvement in the AUC for miR-21 is encouraging, whilst to date miR-21 has shown variable results in AMI diagnosis [9,10]. ddPCR may quantify this miRNA more accurately, and therefore be able to better differentiate STEMI populations from those with stable CAD. The same can be said for miR-208a and miR-499. Both of these miRNAs demonstrate desirable qualities as AMI biomarkers. For example, miR-

S. Robinson et al. / International Journal of Cardiology 257 (2018) 247–254

208a is known to be cardiac specific and miR-499 is unaffected by renal function [11,32,33]. In this study, we were unable to demonstrate AUCs superior to cTnT for any single miRNAs. However, miR-499/miR-21 combined and expressed as a ratio demonstrated a superior AUC and hence diagnostic utility to cTnT as measured by ddPCR. This is in keeping with prior studies which demonstrate that miRNAs exhibit superior diagnostic potential when combined in a panel, and when utilized appropriately, may even out-compete cTn in their diagnostic value [34,35]. From our study, we propose that ddPCR is the superior method for quantifying a panel of miRNAs for diagnostic purposes. 4.4. Neither qRT-PCR nor ddPCR demonstrated a significant association between miR-21/208a/499 and reperfusion injury In our study, miR-21/208a/499 did not show a significant association with reperfusion injury after PCI. miR-499, as measured by ddPCR, was the only miRNA to show a trend. This might show promise, because miR-499 has been shown to play a physiological role in cardiac I/R injury [14,15]. 4.5. miR-21 and miR-499 demonstrated significant correlation with myocardial necrosis markers We found a significant correlation between peak CK levels and miR21 and miR-499 measured by ddPCR, and a significant correlation between peak CK and cTnT levels and miR-499/miR-21 as expressed as a ratio for ddPCR. All of these cardiac enzymes, and in particular, their levels at peak, have been associated with both size of infarct and prognosis [36,37,38,39,40,41,42,43]. Therefore, the above-mentioned miRNAs may have similar utility, giving further interest into these miRNAs as biomarkers for AMI. It is also interesting that miR-499/ miR-21 as expressed as a ratio demonstrated greater correlation with CK and cTnT for ddPCR, giving further credence to the improved utility of miRNAs when used as a panel. 4.6. Cost of ddPCR compared to qRT-PCR We estimated the direct cost, starting from cDNA to completion of each analysis between qRT-PCR and ddPCR, including all necessary disposables and reagents. In these experiments, ddPCR cost almost double the price of qRT-PCR per well. Despite this, it must be kept in mind that ddPCR is an absolute quantification method that does not require standard curves and can accurately quantify single samples, whereas qRTPCR optimally requires both standard curves and samples performed in triplicate to produce reliable results. Therefore, overall cost difference between both systems would be negligible depending on the experiments being performed. 5. Limitations The results of this study support our hypotheses. However, our study has some limitations. The clinical characteristics were obtained retrospectively and hence we were not able to collect certain data and our results may be subject to bias and confounders. Furthermore, the clinical data is limited by lack of randomization and blinding. This study was also limited by its small patient size as well as the small number of miRNAs analyzed. A potential strength is that blood for miRNA analysis was drawn at similar time points between patient groups. However, a time-course study was not performed. A further limitation is that we only used a single exogenous spike-in-control in order to normalise our data, and we did not compare its utility to an endogenous control or a combination method. Moreover, we only analyzed a single surrogate sign of I/R injury. Finally, this study lacked a healthy control group. Ultimately, the aim of this study was more technical rather than clinical and we did not aim to demonstrate a superiority of miRNAs over cTnT for the diagnosis of AMI. Our findings of ddPCR's superiority

253

for the quantification of miRNAs needs to be verified in studies with larger patient cohorts, analyzing a larger panel of miRNAs over different time points. 6. Conclusions In conclusion, we have shown that ddPCR consistently demonstrated superiority in both technical proficiency and diagnostic potential compared to qRT-PCR. However, further investigation is required in larger patient cohorts. Ultimately, our findings support the use of ddPCR over qRT-PCR for more accurate and reproducible quantification of miRNAs in biological expressions in cardiovascular biology, particularly for the use in large multi-center clinical trials. Supplementary data to this article can be found online at https://doi. org/10.1016/j.ijcard.2017.10.111. Funding sources/disclosures The authors have no relevant financial information or potential conflicts of interest to disclose. References [1] WHO, Projections of mortality and causes of death, 2015 and 2030[Data file]. Available from: http://www.who.int/healthinfo/global_burden_disease/projections/ en/ 2016 (9 May 2016). [2] K. Thygesen, J.S. Alpert, A.S. Jaffe, M.L. Simoons, B.R. Chaitman, H.D. White, Third universal definition of myocardial infarction, Circulation 126 (2012) 2020–2035. [3] M. Rubini Gimenez, R. Twerenbold, T. Reichlin, et al., Direct comparison of highsensitivity-cardiac troponin I vs. T for the early diagnosis of acute myocardial infarction, Eur. Heart J. 35 (2014) 2303–2311. [4] A. Al-Saleh, A. Alazzoni, S. Al Shalash, et al., Performance of the high-sensitivity troponin assay in diagnosing acute myocardial infarction: systematic review and meta-analysis, CMAJ Open 2 (3) (July 2014) 199–207. [5] M. Roffi, C. Patrono, J.P. Collet, et al., 2015 ESC guidelines for the management of acute coronary syndromes in patients presenting without persistent ST-segment elevation, Eur. Heart J. 3 (37) (2016) 267-31. [6] T. Sun, Y.H. Dong, W. Du, C.Y. Shi, K. Wang, M.A. Tariq, et al., The role of MicroRNAs in myocardial infarction: from molecular mechanism to clinical application, Int. J. Mol. Sci. 18 (4) (March 31, 2017) https://doi.org/10.3390/ijms18040745 (PMID: 28362341, pii: E745). [7] A.M. Mohr, J.L. Mott, Overview of microRNA biology, Semin. Liver Dis. 35 (1) (February 2015) 3–11. [8] P. Diehl, A. Fricke, L. Sander, et al., Microparticles: major transport vehicles for distinct microRNAs in circulation, Cardiovasc. Res. 93 (4) (2012) 633–644. [9] F. Wang, G. Long, C. Zhao, et al., Atherosclerosis-related circulating miRNAs as novel and sensitive predictors for acute myocardial infarction, PLoS One 9 (2014) 9. [10] F. Olivieri, R. Antonicelli, M. Lorenzi, et al., Diagnostic potential of circulating miR-499-5p in elderly patients with acute non ST-elevation myocardial infarction, Int. J. Cardiol. 167 (2) (July 31, 2013) 531–536. [11] G.K. Wang, J.Q. Zhu, J.T. Zhang, et al., Circulating microRNA: a novel potential biomarker for early diagnosis of acute myocardial infarction in humans, Eur. Heart J. 31 (6) (March 2010) 659–666. [12] Y. Devaux, M. Vausort, E. Goretti, et al., Use of circulating microRNAs to diagnose acute myocardial infarction, Clin. Chem. 58 (2012) 559–567. [13] X. Chen, L. Zhang, T. Su, et al., Kinetics of plasma microRNA-499 expression in acute myocardial infarction, J. Thorac. Dis. 7 (5) (May 2015) 890–896. [14] J.X. Wang, J.Q. Jiao, Q. Li, et al., miR-499 regulates mitochondrial dynamics by targeting calcineurin and dynamin-related protein-1, Nat. Med. 17 (2011) 71–78. [15] J. Wang, Z. Jia, C. Zhang, et al., miR-499 protects cardiomyocytes from H2O2 induced apoptosis via its effects on Pdcd4 and Pacs2, RNA Biol. 11 (4) (2014) 339–350. [16] C. Schulte, T. Zeller, microRNA-based diagnostics and therapy in cardiovascular disease - summing up the facts, Cardiovasc. Diagn. Ther. 5 (1) (February 2015) 17–36. [17] J. Koshiol, E. Wang, Y. Zhao, F. Marincola, M.T. Landi, Strengths and limitations of laboratory procedures for microRNA detection, Cancer Epidemiol. Biomark. Prev. 19 (4) (April 2010) 907–911. [18] B.J. Hindson, K.D. Ness, D.A. Masquelier, et al., High-throughput droplet digital PCR system for absolute quantitation of DNA copy number, Anal. Chem. 83 (22) (November 15, 2011) 8604–8610. [19] C.M. Hindson, J.R. Chevillet, H.A. Briggs, et al., Absolute quantification by droplet digital PCR versus analog real-time PCR, Nat. Methods 10 (10) (October 2013) 1003–1005. [20] M.W. Pfaffl, Quantification Strategies in Real-Time PCR, Bustin SA, Editor. A-Z of Quantitative PCR [Internet], International University Line (IUL), La Jolla, CA, 2004 (Chapter 3. Available from: http://gene-quantification.org/chapter-3-pfaffl. pdf. (4 August 2016)). [21] D.K. Massart, Data Handling in Science and Technology, in: D.L. Massart, B.G.M. Vandeginste, L.M.C. Buydens, S. De Jong, P.J. Lewi, J. Smereys-Verbeke (Eds.),

254

[22] [23]

[24]

[25]

[26] [27] [28] [29] [30]

[31]

[32] [33]

[34]

S. Robinson et al. / International Journal of Cardiology 257 (2018) 247–254 Handbook of Chemometrics and Qualimetrics, Elsevier Science B.V., Amsterdam, 1998 (Part A). A. Shrivastava, V.B. Gupta, Methods for the determination of limit of detection and limit of quantification of the analytical methods, Chron. Young Sci. 2 (2011) 21–25. L. De Roeck, S. Vandamme, B.R. Everaert, et al., Adiponectin and ischemiareperfusion injury in ST segment elevation myocardial infarction, Eur. Heart J. Acute Cardiovasc. Care 5 (2016) 71–76. J.S. McDonald, D. Milosevic, H.V. Reddi, S.K. Grebe, A. Algeciras-Schimnich, Analysis of circulating microRNA: preanalytical and analytical challenges, Clin. Chem. 57 (6) (2011) 833–840. S. De Rosa, S. Fichtlscherer, R. Lehmann, B. Assmus, S. Dimmeler, A.M. Zeiher, Transcoronary concentration gradients of circulating microRNAs, Circulation 124 (2011) 1936–1944. Y. D'Alessandra, P. Devanna, F. Limana, et al., Circulating microRNAs are new and sensitive biomarkers of myocardial infarction, Eur. Heart J. 31 (2010) 2765–2773. T. Adachi, M. Nakanishi, Y. Otsuka, et al., Plasma microRNA 499 as a biomarker of acute myocardial infarction, Clin. Chem. 56 (2010) 1183–1185. T.E. Callis, K. Pandya, H.Y. Seok, et al., MicroRNA-208a is a regulator of cardiac hypertrophy and conduction in mice, J. Clin. Invest. 119 (2009) 2772-278. X. Ji, R. Takahashi, Y. Hiura, G. Hirokawa, Y. Fukushima, N. Iwai, Plasma miR-208 as a biomarker of myocardial injury, Clin. Chem. 55 (2009) 1944–1949. F. Olivieri, R. Antonicelli, M. Lorenzi, et al., Diagnostic potential of circulating miR499-5p in elderly patients with acute non ST-elevation myocardial infarction, Int. J. Cardiol. 167 (2013) 531–536. T.C. Dingle, R.H. Sedlak, L. Cook, K.R. Jerome, Tolerance of droplet-digital PCR versus real-time quantitative PCR to inhibitory substances, Clin. Chem. 59 (11) (2013) 1670–1672. M. Kemp, J. Donovan, H. Higham, J. Hopper, Biochemical markers of myocardial injury, Br. J. Anaesth. 93 (1) (April 2004) 63–73. M.F. Corsten, R. Dennert, S. Jochems, et al., Circulating microRNA-208b and microRNA-499 reflect myocardial damage in cardiovascular disease, Circ. Cardiovasc. Genet. 3 (2010) 499–506. R. Zhang, C. Lan, H. Pei, G. Duan, L. Huang, L. Li, Expression of circulating miR-486 and miR-150 in patients with acute myocardial infarction, BMC Cardiovasc. Disord. 15 (June 16, 2015).

[35] C. Li, Z. Fang, T. Jiang, et al., Serum microRNAs profile from genome-wide serves as a fingerprint for diagnosis of acute myocardial infarction and angina pectoris, BMC Med. Genet. 6 (2013) 16. [36] A. Di Chiara, M. Plewka, M. Werren, L.P. Badano, C. Fresco, P.M. Fioretti, Estimation of infarct size by single measurements of creatine kinase levels in patients with a first myocardial infarction, J. Cardiovasc. Med. 7 (5) (2006) 340–346. [37] S. Savonitto, C.B. Granger, D. Ardissino, et al., The prognostic value of creatine kinase elevations extends across the whole spectrum of acute coronary syndromes, J. Am. Coll. Cardiol. 39 (1) (2002) 22–29. [38] S. Chia, F. Senatore, C.O. Raffel, H. Lee, F.J.T. Wackers, I. Jang, Utility of cardiac biomarkers in predicting infarct size, left ventricular function, and clinical outcome after primary percutaneous coronary intervention for ST-segment elevation myocardial infarction, JACC Cardiovasc. Interv. 1 (4) (2008) 415–423. [39] A. Halkin, G.W. Stone, C.L. Grines, et al., Prognostic implications of creatine kinase elevation after primary percutaneous coronary intervention for acute myocardial infarction, J. Am. Coll. Cardiol. 47 (5) (2006) 951–961. [40] Y. Gerber, A.S. Jaffe, S.A. Weston, R. Jiang, V.L. Roger, Prognostic value of cardiac troponin T after myocardial infarction: a contemporary community experience, Mayo Clin. Proc. 87 (3) (2012) 247–254. [41] T. Dohi, A. Maehara, S.J. Brener, et al., Utility of peak creatine kinase-MB measurements in predicting myocardial infarct size, left ventricular dysfunction, and outcome after first anterior wall acute myocardial infarction (from the INFUSE-AMI trial), Am. J. Cardiol. 115 (5) (2015) 563–570. [42] A.K. Hassan, S.C. Bergheanu, H. Hasan-Ali, et al., Usefulness of peak troponin-T to predict infarct size and long-term outcome in patients with first acute myocardial infarction after primary percutaneous coronary intervention, Am. J. Cardiol. 103 (6) (2009) 779–784. [43] E. Hedström, K. Aström-Olsson, H. Ohlin, et al., Peak CKMB and cTnT accurately estimates myocardial infarct size after reperfusion, Scand. Cardiovasc. J. 41 (1) (2007) 44–50.