Circulating microRNAs: Potential Markers of Cardiotoxicity in Children and Young Adults Treated With Anthracycline Chemotherapy

Background Biomarkers for early detection of anthracycline (AC)‐induced cardiotoxicity may allow cardioprotective intervention before irreversible damage. Circulating microRNAs (miRNAs) are promising biomarkers of cardiovascular disease, however, have not been studied in the setting of AC‐induced cardiotoxicity. This study aimed to identify AC‐induced alterations in plasma miRNA expression in children and correlate expression with markers of cardiac injury. Methods and Results Candidate plasma profiling of 24 miRNAs was performed in 33 children before and after a cycle of AC (n=24) or noncardiotoxic chemotherapy (n=9). Relative miRNA changes between the pre‐ and postcycle time points (6, 12, and 24 hours) were determined within each treatment group and compared across groups. Plasma miRNA expression patterns were further explored with respect to AC dose and high‐sensitivity troponin T. Greater chemotherapy‐induced dysregulation was observed in this panel of candidate, cardiac‐related plasma miRNAs in patients receiving anthracyclines compared with those receiving noncardiotoxic chemotherapy (24‐hour MANOVA; P=0.024). Specifically, plasma miRs‐29b and ‐499 were upregulated 6 to 24 hours post‐AC, and their postchemotherapy expression significantly correlated with AC dose. Patients with acute cardiomyocyte injury (high‐sensitivity troponin T increase ≥5 ng/L from baseline) demonstrated higher expression of miR‐29b and miR‐499 post‐AC compared with those without. Conclusions In this pilot study, cardiac‐related plasma miRNAs are dysregulated following ACs. Plasma miR‐29b and ‐499 are acutely elevated post‐AC, with dose response relationships observed with anthracycline dose and markers of cardiac injury. Further evaluation of miRNAs may provide mechanistic insight into AC‐induced cardiotoxicity and yield biomarkers to facilitate earlier intervention to mitigate cardiotoxicity.

A nthracycline (AC)-related cardiotoxicity is among the most significant long-term threats to cancer survivors, with cardiac events comprising the most common nonmalignant cause of death in this population. 1,2 The estimated prevalence of cardiomyopathy in 50-year-old survivors of childhood cancer exposed to cardiotoxic chemotherapy or radiation is 21%. 3 Once AC-induced cardiomyopathy develops, it is commonly progressive, with no definitive treatment other than implantation of a left ventricular assist device and/or heart transplantation. [4][5][6] Currently, methods to detect subclinical cardiac injury early in the course of progressive cardiotoxicity are lacking. Early identification of patients at high risk for cardiac morbidity could allow implementation of cardioprotective measures during a stage when these therapies are likely to impact cardiac outcome. 7,8 Pediatric studies exploring anthracycline (AC)-induced trends in cardiac troponin demonstrate a correlation between early troponin elevations and late echocardiographic abnormalities. 9,10 However, studies have yet to reveal quantitative trends in troponin levels that specifically predict late cardiac outcomes, and thus troponins are insufficient to guide cardioprotective intervention.
MicroRNAs (miRNAs) are short, noncoding RNAs that regulate fundamental cellular processes and play a critical role in cardiovascular development and disease. 11,12 Circulating miRNAs are promising biomarkers of cardiovascular disease attributed to their surprisingly high stability in plasma, tissue-or cell-specific distribution, and ease of quantitative measurement. 13,14 The demonstrated sensitivity and specificity of miRNAs in various cardiovascular disease states indicate a potential role for miRNAs in early detection of cardiotoxicity. 15 Although the role of miRNAs in modulating anthracyclineinduced cardiac injury has been demonstrated in animal models, [16][17][18] little is known about AC-induced alterations in plasma miRNAs. The aim of this study was to explore acute AC-induced alterations in plasma miRNA expression in children and correlate miRNA expression with known markers or predictors of cardiac injury. We hypothesized that children receiving AC chemotherapy would demonstrate a unique plasma miRNA profile compared with controls receiving noncardiotoxic chemotherapy, and that this plasma miRNA signature could identify children with AC-induced cardiac injury.

Study Design
This was a prospective cohort study of patients aged <18 years receiving AC or noncardiotoxic chemotherapy (controls) at Children's Medical Center Dallas (Figure 1), approved by the University of Texas Southwestern Institutional Review Board (Dallas, TX). Patients with a history of congenital heart disease, Down syndrome, or those who received radiation to the heart were ineligible. Patients receiving AC, high-dose cyclophosphamide (>2000 mg/m 2 per cycle or cancer were enrolled before a cycle of anthracycline (AC) or noncardiotoxic (Control) chemotherapy. Plasma was collected before and at several points following the chemotherapy cycle. Plasma miRNA expression was evaluated according to chemotherapy group (AC vs noncardiotoxic) and troponin elevation. AC indicates anthracycline; Acute cTn(+), patients with hs-cTnT ≥14 ng/L whose post-AC cTnT was ≥5 ng/L above the precycle-AC baseline; Chronic cTnT(+), patients with hs-cTnT ≥14 ng/L whose pre-and postcycle-AC troponin levels remained stable (<5 ng/L rise from baseline); cTn(À), patients with plasma cTnT level <14 ng/L at all time points; cTn(+), patients with plasma hs-cTnT level ≥14 ng/L at any time point (0, 6, 12, or 24 hours); CTRL, control group; hs-cTnT, high-sensitivity cardiac troponin T assay.
>1200 mg/m 2 per dose), or tyrosine kinase inhibitors or those with any evidence of cardiac injury (high-sensitivity cardiac troponin T [hs-cTnT] >14 ng/L) were not eligible for the control group. Potential subjects were identified and referred for recruitment by their primary oncology provider. Those meeting eligibility criteria were approached for enrollment and informed consent was obtained. In all cases, AC chemotherapy was given by bolus infusion over 15 to 30 minutes. In most (n=21 of 25) cases, the AC group underwent sample collection during their final cycle of AC; controls underwent collection during a late treatment cycle of a noncardiotoxic chemotherapy regimen. In the AC group, a "pre-AC" cycle baseline sample was drawn before the first cycle dose of AC, with post-AC sample collection 6, 12, and 24 hours following the final AC dose. The 12-hour sample was only drawn from patients who were hospitalized (collected in 24 of 33 patients).
Because this study was designed to assess miRNA alteration following the maximal cumulative AC exposure, all but 1 patient in the AC group had anthracycline exposure before to their "day 0" precycle baseline. Controls underwent the same schedule of plasma collection timed around a dose of noncardiotoxic chemotherapy.

Sample Processing, RNA Isolation, and MicroRNA Profiling
Plasma samples were collected either by peripheral venipuncture or from a central line (see Data S1 for details). Briefly, RNA was isolated using a miRNeasy serum/plasma microRNA isolation kit (Qiagen, Hilden, Germany) and reverse transcribed using the Taqman miRNA reverse transcription kit. Custom TaqMan Array MicroRNA Cards (Product# 4346799, Format 32; Applied Biosystems, Foster City, CA) were designed to assay 24 candidate miRNAs and 7 miRNAs for quality assessment and normalization. Candidate miRNAs were selected based on their demonstrated roles as biomarkers or mediators of cardiovascular disease (Table S1). The final endogenous miRNA controls used for miRNA normalization were chosen based on stability score (Table S2).

High-Sensitivity Troponin Assay
hs-cTnT was measured using an automated immunoassay (Troponin T hs STAT, Elecsys-2010; Roche Diagnostics, Indianapolis, IN). 19 Five nanograms per liter represents the limit of detection for this assay. An absolute threshold of ≥14 ng/L is considered abnormal for this assay. 20,21 Assessment of change in troponin over time is recommended to discriminate patients with and without acute cardiomyocyte injury. 22,23 Therefore, we considered patients to have a significant chemotherapy-induced troponin rise if hs-cTnT 6 to 24 hours after chemotherapy was at least 5 ng/L greater than the precycle baseline. Patients were categorized into 3 hs-cTnT categories, those with: acute hs-cTnT elevations (hs-cTnT ≥14 ng/L at any time point, plus ≥5 ng/L rise from baseline); chronic elevation (hs-cTnT >14 ng/L at any time point, but <5 ng/L rise from baseline); or normal (<14 ng/L at all time points).

Statistical Analysis
Comparisons across groups were made using Fisher's exact tests for nominal characteristics and rank-sum tests for continuous characteristics. Individual patient troponin values across 0-, 6-, 12-, and 24-hour time points were plotted by type of chemotherapy received ( Figure S1). Differences in miRNA regulation from baseline were compared between AC and control groups at each time point (6,12, and 24 hours) using linear mixed-effects regression models for each of the 24 target miRNAs. The models included fixed effects of group (anthracycline versus noncardiotoxic chemotherapy or cTnT rise ≥5 ng/L versus <5 ng/L), time (0, 6, 12, and 24 hours), condition (target miRNA versus control miRs-140 and -484), and all 2-and 3-factor cross-terms. The models included random intercept, time, condition, and time9condition by patient effects to account for correlated cycle threshold (Ct) levels, changes over time, differences between conditions, and differences between changes over time by condition within patients, respectively. Additionally, MANOVA was applied to miRNA levels averaged by patient over the three Ct replicates at the 6-and 24-hour time points and the 2 control conditions to test the hypothesis of no overall group effect on the entire panel of candidate miRNAs. The number of samples collected at 12 hours was insufficient to perform this test. Twenty-four miRNAs were measured in the cohort and 3 were selected for further analyses based on their consistent differential regulation between treatment groups with between group P<0.05 (miRs-1 and -29b) or between group P<0.1 with miRNA with fold-change >2 at all time points in the AC group with no change in the treatment group (miR-499). These miRNAs were subsequently compared across troponin groups (hs-cTnT rise ≥5 ng/L versus <5 ng/L) using similar regression modeling to that used to analyze treatment groups. Two miRNAs with differential expression between troponin groups were further analyzed to explore their relationship with relevant covariates including interim and cumulative AC dose, wherein normalized miRNA cycle threshold values at each time point were regressed linearly and separately on each covariate using data averaged over the 3 Ct replicates by patient and time. Finally, logistic regression models were fit in the subgroup of patients receiving ACs to explore the sensitivity and specificity of normalized miR-29b and -499 expression for predicting patients with a significant rise in hs-cTnT post-AC (≥5 ng/L above baseline). These models used average Ct data by patient, time, and miRNA to construct receiver operator characteristic curves.
A priori, we targeted recruitment of 34 patients (24 ACexposed; 10 controls), which would provide 80% power to detect a 2-fold mean group change in miRNA expression from baseline between groups, with an alpha of 0.05. Reported P values were not adjusted for multiple comparisons. All analyses were programmed in SAS/STAT software (version 9.4; SAS Institute Inc., Cary, NC). Additional details on both mRNA and troponin assays as well as the statistical modeling can be found in Data S1.

Clinical Data and High-Sensitivity Troponin T Assay
A total of 37 patients were enrolled (n=25 AC and n=12 control), 33 of whom were included in the miRNA analysis (n=24 AC and n=9 control). Three patients receiving intermediate-dose cyclophosphamide (1200 mg/m 2 ) were excluded from the control group because of low-level elevations in hs-cTnT, in order to maintain a control group free of any evidence of cardiac injury. Additionally, 1 patient receiving doxorubicin had uniformly low Ct values (global mean <2 SD below that of the entire group) in 3 of the 4 sample time points, thus was excluded because of technical problems with the sample/assay. The AC group contained patients with various diagnoses (Table) who received doxorubicin (n=15), mitoxantrone (n=7), daunorubicin (n=1), or idarubicin (n=1) fractionated over 1 to 5 days. The mean cumulative anthracycline dose before baseline plasma collection and the individual cycle dose around which the plasma samples were collected were 321AE30 and 102AE13 mg/m 2 in doxorubicin equivalents, respectively. Eight patients received dexrazoxane in addition to their cycle dose of anthracycline, either per institutional protocol (n=7) or for left ventricular dysfunction following a previous anthracycline  cycle (n=1). The control group contained patients with a variety of tumor types who received non-AC containing chemotherapeutic regimens, including cisplatin, carboplatin, cyclophosphamide, ifosfamide, etoposide, bleomycin, clofarabine, 5-fluorouracil, vincristine, or dactinomycin. The AC and control groups did not differ by age (P=0.83) or sex (P=0.81) at study entry. Of the 24 AC patients analyzed, 15 had hs-cTnT concentrations ≥14 ng/L following the cycle dose(s) of AC (cTn(+) AC). Of these, 8 patients had troponin ≥14 ng/L at baseline, preceding this cycle of AC (mean cTnT: 23AESD 4.1 ng/L; Figure 1). Seven of the 15 hs-cTnT-positive patients had an acute hs-cTnT rise of ≥5 ng/L from baseline. Nine patients had no hs-cTnT elevations at any time point. Within the AC group, the patients with acute elevations were similar in age, sex, and cumulative AC dose compared to those with chronically elevated or normal hs-cTnT (Table). However, the patients with acute elevations received a higher interim dose of AC than those with normal or chronically elevated hs-cTnT (P=0.004).

Evaluation of Plasma MicroRNA Dysregulation Over Time by Chemotherapy Group
Greater overall chemotherapy-induced dysregulation was observed in the 24 candidate plasma miRNAs in patients receiving AC chemotherapy compared with those receiving noncardiotoxic chemotherapy (MANOVA at 24 hours, P=0.024). Change in plasma expression from baseline of 2 specific miRNAs differed significantly between AC and controls (P value for anthracycline versus control at 6 hours, <0.05; Table S3). Plasma miR-1 was significantly upregulated 6, 12, and 24 hours post-AC, compared with controls in whom it was unchanged at all time points (Figure 2; Table S3). Plasma miR-29b demonstrated 2.7-fold upregulation 6 hours post-AC, while unchanged in controls ( Figure 2). Additionally, miR-499 was significantly upregulated by >2-fold 6, 12, and 24 hours post-AC compared with controls in whom it was unchanged and showed a trend toward differential regulation between treatment groups at 24 hours (P=0.09; Figure 2; Table S3).

Evaluation of Plasma MicroRNA Expression by Troponin Concentration
Further analysis of chemotherapy-induced plasma miR-1, -29b, and -499 regulation revealed significant differences between patients with and without evidence of acute cardiac injury following chemotherapy (hs-cTnT increase ≥5 ng/L from baseline). Patients with acute AC-induced hs-cTnT elevations had significant upregulation in miR-29b at 6 hours (6.6-fold; P<0.01) and miR-499 at 24 hours (5.2-fold; P=0.03) compared with AC or control patients with chronically elevated or normal hs-cTnT ( Figure 3; between group P=0.01 for miR-29b and P=0.07 for miR-499). Plasma miR-1 was similarly upregulated over time regardless of hs-cTnT concentrations. Similar results were observed when normalized plasma miR-29b and miR-499 expression (DCt) was compared across groups at each pre-and postchemotherapy time point. Patients with acute cardiac injury demonstrated significantly higher plasma expression of miR-29b at 6 hours and miR-499 at 6 and 24 hours compared with those with chronically elevated or normal hs-cTnT ( Figure 4).

Discussion
This study is the first to analyze circulating miRNAs in the setting of AC-induced cardiotoxicity. Although these findings are preliminary given the exploratory nature and small sample size, there were several key findings that support further exploration of miRNAs as clinically relevant markers of ACinduced cardiotoxicity. First, overall dysregulation of a panel of candidate plasma miRNAs with cardiac relevance was greater following AC compared with noncardiotoxic chemotherapy. Second, plasma miR-29b and miR-499 were upregulated in children with acute AC-induced cardiac injury, as evidenced by rising troponin concentrations post-AC. Third, post-AC miR-29b and miR-499 expression significantly correlated with cumulative AC dose, an established predictor of cardiotoxicity risk. Finally, plasma miR-29b and miR-499 expression levels 6 to 24 hours following AC provided moderate discrimination of patients with and without acute, cardiac injury. Together, these findings indicate that miRNAs may have a role in detecting subclinical cardiotoxicity. miRNAs are critical modulators of cardiovascular development and disease, wherein miRNA-guided expression of downstream targets governs the process of regeneration, repair, or pathological remodeling. 11 Specific profiles of miRNA expression have been identified in numerous cardiovascular disease states, wherein circulating expression profiles may provide insight into underlying injury or remodeling occurring in the cardiac tissue. 24,25 These circulating miRNA profiles are not only windows into the molecular landscape of heart disease, but offer the potential for novel therapeutics aimed at the modulation of miRNA and downstream target expression. 15 Additionally, accumulating evidence supports the value of miRNAs as robust biomarkers of cardiovascular disease. 13,15 Relevant to our findings, miR-499 is a muscle-specific, cardiomyocyte-enriched miRNA that has been shown to have high sensitivity and specificity for detecting myocardial ischemia in previous reports. 15,26,27 Plasma miR-499 upregulation postinfarction has been observed even earlier than hs-cTnT and thus may provide superior early sensitivity than troponins. 28 Additionally, higher plasma miR-499 expression following acute myocardial infarction has been correlated with increased risk for heart failure or mortality, indicating that miRNA profiles in the setting of acute injury may have important prognostic implications. 27 The protective role of miR-499 against oxidative-stress-induced apoptosis following acute myocardial infarction 29 has important relevance to AC-induced cardiotoxicity given that the most accepted mechanism of AC-induced cardiotoxicity is through the generation of reactive oxygen species and mitochondrial dysfunction within the cardiomyocyte. 30 Interestingly, in our small cohort, the protective role of dexrazoxane against AC-induced oxidative stress abrogated the relationship between cumulative anthracycline dose and miR-499 elevations, perhaps indicating that dexrazoxane was protective against oxidative-stress-related miR-499 elevations, even at high anthracycline doses.
The miR-29 family members are potent inhibitors of cardiac fibrosis and play a key role in cardiac remodeling following cardiomyocyte injury. 31 Plasma miR-29a upregulation has been reported following myocardial injury, where the degree of mR-29a upregulation was associated with the extent of late remodeling post-acute myocardial infarction. 24 miR-29b, reported herein, targets many genes involved in the extracellular matrix (ECM), such as fibronectin, collagen, and matrix metalloproteinases. Given the critical role of early and late ECM remodeling in response to AC-induced cardiotoxicity, 32,33 the upregulation of plasma miR-29b in this study has intriguing mechanistic relevance and may reflect early remodeling in response to AC-induced cardiac injury. Furthermore, modulation of miR-29 in various settings has successfully reduced maladaptive remodeling postinjury and allowed a more-efficient ECM remodeling response. 31,34 With ongoing progress toward miRNA-directed therapy, the elucidation of the role of miR-29b in AC-induced cardiotoxicity could reveal an important therapeutic target that might allow prevention of AC-induced remodeling leading to preserved cardiac function.
This study has several notable limitations. First, the miRNA expression patterns identified in this study are based on a small, heterogeneous group of patients who received varying anthracycline regimens for a variety of malignancies. Additionally, the disproportionate sample sizes between the control group and AC group poses a significant statistical limitation. However, despite this heterogeneity and small sample size, miRNA expression correlated with multiple indicators/predictors of cardiotoxicity, including AC dose and hs-cTnT. Although troponin elevation has not consistently been shown to correlate with late cardiac outcomes in ACexposed childhood cancer survivors, it is clearly a reflection of cardiomyocyte injury. 35 Thus, use of high-sensitivity troponin changes as a cardiotoxicity correlate for this preliminary miRNA analysis is reasonable. Unfortunately, changes in echocardiogram parameters are typically a late finding in pediatric cardiotoxicity occurring many years from AC exposure. Thus, more-definitive imaging correlates on which to compare miRNA expression were not feasible. The candidate approach to miRNA assessment in the setting of AC-induced cardiotoxicity may have limited our ability to identify miRNAs with important diagnostic or mechanistic implications. However, the 2 miRNAs found to have the most significant plasma dysregulation in patients with AC-induced cardiac injury are promising, particularly given their intriguing mechanistic relevance to that implicated in cardiotoxicity.
A final limitation of this and other circulating miRNA profiling studies is that the etiology and mechanistic significance of plasma miRNA dysregulation is unknown. Although it is clear that many miRNAs (eg, miRs-499, -208, -133, and -1) are released into the plasma following cellular injury (as demonstrated after myocardial infarction), miRNAs are also exported from intact cells by being bound to stabilizing proteins or packaged within exosomes or microvesicles. Given evidence of troponin elevations following AC exposure, plasma miRNA expression described herein may, in part, reflect miRNA release following myocardial injury. Given the small sample size of this study, it cannot be definitively concluded that the other candidate miRNAs previously associated with cardiomyocyte injury (eg, in the setting of acute myocardial infarction, such as miR-133a) 15 are not associated with AC-induced cardiomyocyte injury. However, the lack of relationship observed in this study between these cardiac miRNAs and AC might reflect the unique molecular functions of each individual miRNA, which might not be significantly involved in the heart's response to AC. Further work is warranted to determine the source and mechanistic implications of miR-29b and 499 elevations in the plasma.
In conclusion, we report that plasma miR-29b and -499 elevations occur in the acute setting following AC exposure and may be useful in detecting cardiomyocyte injury. Further work is warranted to evaluate the mechanistic role of these miRNAs in AC-induced cardiac injury and determine whether these miRNAs are useful for identifying patients at high risk of developing AC-induced cardiomyopathy.

Disclosures
The following authors have relationships with industry to disclose: Dr DeLemos: Amgen (advisory role, research funding provided), Roche Diagnostics (advisory role, research funding provided), and Abbott Diagnostics (research funding provided); Dr Mammen: CareDx Inc (advisory role, speaker's bureau, and research funding/honoraria/travel accommodations provided), HeartWare (advisory role, honoraria/travel accommodations provided), PhaseBio (advisory role, honoraria/travel accommodations provided), and California Institute for Regenerative Medicine (advisory role); Nielson: United Therapeutics (advisory role, speaker's bureau, and travel accommodations provided). Additional disclaimer: Roche Diagnostics provided hs-cTnT measurements, however had no role in the study.

RNA Quantification & Processing of miRNA Expression Data
Custom TaqMan® Array MicroRNA Cards (Applied Biosystems; Product# 4346799, Format 32) were designed to assay 24 candidate miRNAs and 7 miRNAs for quality assessment and normalization.
Candidate miRNAs were selected based on their demonstrated roles as biomarkers or mediators of cardiovascular disease (Table S1) and include let-7g, miRs-1, 133a, 133b, 145, 146a, 195, 208a, 208b,   21, 210, 214, 215, 216b, 29a, 29b, 29c, 30a-5p, 34c, 361, 367, 423-3p, 423-5p, and 499-5p. A group of candidate endogenous control miRNAs, including miRs-103, 140, 191-5p, 484, U6, and synthetic cel-miR-39, were also selected based on reports of successful normalization achieved using these miRNAs. [1][2][3][4] Additionally miR-451 and miR-23a were assayed as a quality control measure to assess sample hemolysis. [5] Reverse transcription utilizing the Taqman custom RT primer pool was performed using the Taqman miRNA reverse transcription kit. Reaction volumes were modified based on the Taqman technical note entitled "Optimized protocols for human or rodent microRNA profiling with precious samples" (Product Cycle threshold data for all samples were imported into OmicsOffice Statminer software program's real time qPCR analysis feature. Undetermined cycle threshold values were set to 36, the detection limit. Candidate endogenous control miRNAs were evaluated using the normfinder algorithm aimed at identifying the miRNA(s) with the lowest variation within groups (AC vs. control). The aggregate of miRs-140 and -484 was selected as the endogenous control for this study based on its lowest stability value (Table S2).

Plasma microRNA Normalization
Currently there is a lack of consensus regarding circulating miRNA normalization. Although many groups are normalizing serum and plasma miRNA expression to a spiked in synthetic miRNA, this does not account for the effects of plasma processing and other pre-analytic variables on circulating miRNA measurement. [6] Given the high content of many intracellular miRNAs relative to that of plasma, hemolysis and platelet contamination of plasma can introduce significant miRNA variability from one sample to the next. [6,7] We employed several strategies to overcome these technical challenges, including an additional high-speed centrifugation step to minimize platelet contamination and assessing the degree of RBC-derived miRNA contamination due to hemolysis. [7,8] Fortunately, tissue specific miRNAs with little or no expression in circulating blood cells (including miR-499 from this study) are relatively unaffected by these processing variables. However, miR-29b and other miRNAs examined herein are highly expressed in blood cells, thus may be sensitive to processing conditions and inadequately normalized by the synthetic spike-in strategy. [9] For this reason, endogenous miRNA controls, chosen based on stability score (Table S2), were used for candidate miRNA normalization in this study.