Genome-Wide Association Study Identifies ROBO2 as a Novel Susceptibility Gene for Anthracycline-Related Cardiomyopathy in Childhood Cancer Survivors

PURPOSE

Interindividual variability in the dose-dependent association between anthracyclines and cardiomyopathy suggests a modifying role of genetic susceptibility. Few previous studies have examined gene-anthracycline interactions. We addressed this gap using the Childhood Cancer Survivor Study (discovery) and the Children's Oncology Group (COG) study COG-ALTE03N1 (replication).

METHODS

A genome-wide association study (Illumina HumanOmni5Exome Array) in 1,866 anthracycline-exposed Childhood Cancer Survivor Study participants (126 with heart failure) was used to identify single-nucleotide polymorphisms (SNPs) with either main or gene-environment interaction effect on anthracycline-related cardiomyopathy that surpassed a prespecified genome-wide threshold for statistical significance. We attempted replication in a matched case-control set of anthracycline-exposed childhood cancer survivors with (n = 105) and without (n = 160) cardiomyopathy from COG-ALTE03N1.

RESULTS

Two SNPs (rs17736312 [ROBO2]) and rs113230990 (near a CCCTC-binding factor insulator [< 750 base pair]) passed the significance cutoff for gene-anthracycline dose interaction in discovery. SNP rs17736312 was successfully replicated. Compared with the GG/AG genotypes on rs17736312 and anthracyclines ≤ 250 mg/m², the AA genotype and anthracyclines > 250 mg/m² conferred a 2.2-fold (95% CI, 1.2 to 4.0) higher risk of heart failure in discovery and an 8.2-fold (95% CI, 2.0 to 34.4) higher risk in replication. ROBO2 encodes transmembrane Robo receptors that bind Slit ligands (SLIT). Slit-Robo signaling pathway promotes cardiac fibrosis by interfering with the transforming growth factor-β1/small mothers against decapentaplegic (Smad) pathway, resulting in disordered remodeling of the extracellular matrix and potentiating heart failure. We found significant gene-level associations with heart failure: main effect (TGF-β1, P = .007); gene*anthracycline interaction (ROBO2*anthracycline, P = .0003); and gene*gene*anthracycline interaction (SLIT2*TGF-β1*anthracycline, P = .009).

CONCLUSION

These findings suggest that high-dose anthracyclines combined with genetic variants involved in the profibrotic Slit-Robo signaling pathway promote cardiac fibrosis via the transforming growth factor-β1/Smad pathway, providing credence to the biologic plausibility of the association between SNP rs17736312 (ROBO2) and anthracycline-related cardiomyopathy.

INTRODUCTION

Anthracyclines play a critical role in the treatment of multiple childhood cancer types, such that 60% of childhood cancer survivors carry a history of anthracycline exposure.^1–4 However, there is a strong dose-dependent association between anthracycline exposure and cardiomyopathy,^5,6 potentially limiting the full therapeutic potential of the drug. Several other factors, such as younger age at exposure, female sex, chest radiation, and presence of cardiovascular risk factors (CVRFs; eg, diabetes, hypertension, and dyslipidemia), modify the risk of anthracycline-related cardiomyopathy.^2,7 However, the interindividual variability in the dose-dependent association between anthracycline exposure and risk of cardiomyopathy suggests a possible role of genetics. Previous efforts led by our team and other investigators have identified more than 100 genetic variants associated with anthracycline-related cardiomyopathy.⁸ With few exceptions, most studies lack robust patient numbers, have not examined gene-environment (gene-anthracycline dose) interactions, or have not attempted replication in independent populations.

CONTEXT

• Key Objective

• Do genetic variants interact with anthracycline dose to increase the risk of cardiomyopathy in children with cancer?

• Knowledge Generated

• Using a genome-wide association study approach, we identified single-nucleotide polymorphisms rs17736312 [ROBO2] exceeding the significance cutoff for gene-anthracycline dose interaction in discovery and replicating successfully in an independent childhood cancer survivor population. ROBO2 encodes transmembrane Robo receptors that bind Slit ligands. Slit-Robo signaling pathway promotes cardiac fibrosis by interfering with the transforming growth factor-β1/Smad pathway, resulting in disordered remodeling of the extracellular matrix and potentiating heart failure. We found significant gene-level associations with heart failure for TGF-β1, ROBO2, and SLIT2.

• Relevance (J.W. Friedberg)

• These results, including external validation, provide strong evidence for the association of a common single-nucleotide polymorphisms with the development of anthracycline-mediated cardiomyopathy, and may lead to precision approaches to prevent this devastating side effect of cancer therapy.*

  *Relevance section written by JCO Editor-in-Chief Jonathan W. Friedberg, MD.

We addressed these limitations by leveraging the resources offered by two large nonoverlapping populations: the Childhood Cancer Survivor Study (CCSS)⁹ cohort to identify single nucleotide polymorphisms (SNPs) that potentially modify the dose-dependent risk of anthracycline-related cardiomyopathy in discovery, and the Children's Oncology Group (COG) study COG-ALTE03N1^5,10,11 to replicate SNPs that surpassed a prespecified threshold for statistical significance in the discovery stage.

METHODS

Study Design and Population

Discovery set.

CCSS is a multi-institutional cohort study of long-term (≥ 5 years) survivors of childhood cancer in North America. Survivors were diagnosed between 1970 and 1986 at age 21 years or younger.⁹ They were followed prospectively through longitudinal surveys querying health conditions (including congestive heart failure [CHF]), health-related behaviors, and health care use.^12–14 Responses to questionnaire items related to cardiomyopathy, CHF, heart transplantation, and medications were classified and graded using the Common Terminology Criteria for Adverse Events, using previously described methodology to define CHF.^12,15 Only those outcomes graded as severe (grade 3; self-reported cardiomyopathy or CHF, plus medications), life-threatening (grade 4; requiring heart transplantation), or fatal (grade 5) were included. A composite binary variable for CVRFs (yes [presence of any of the following: diabetes, hypertension, and dyslipidemia]; no [absence of all CVRFs]) was ascertained through self-report. The protocol was approved by the human subjects committee at participating institutions. Participants provided informed consent.

Replication set.

COG-ALTE03N1 uses a matched case-control design to understand the pathogenesis of cardiomyopathy in childhood cancer survivors. COG member institutions enrolled patients after obtaining approval from local institutional review boards. Written informed consent or assent was obtained from patients, parents, or legal guardians. Cases and controls were identified from individuals diagnosed with cancer at age 21 years or younger. Cases consisted of childhood cancer survivors who developed cardiomyopathy. For each case, one to four controls with no signs or symptoms of cardiomyopathy were randomly selected from the same COG childhood cancer survivor cohort, matched on primary cancer diagnosis, year of diagnosis (± 5 years), and race/ethnicity. The selected controls also needed to have a longer duration of cardiomyopathy-free follow-up compared with time from cancer diagnosis to cardiomyopathy for the corresponding case. Cases fulfilled American Heart Association criteria for cardiac compromise by presenting with signs and/or symptoms (dyspnea, orthopnea, fatigue, edema, hepatomegaly, and/or rales); or, in the absence of signs and/or symptoms, had echocardiographic features of left ventricular dysfunction (ejection fraction ≤ 40% and/or fractional shortening ≤ 28%). All participants provided blood or saliva for germline DNA. Participants also provided blood in PAXgene blood RNA tubes for RNA.

Therapeutic Exposures

Participating sites (CCSS and COG-ALTE03N1) abstracted anthracycline chemotherapy and chest radiation information from medical records. Anthracycline exposure included exposure to doxorubicin, daunorubicin, idarubicin, epirubicin, or mitoxantrone (an anthraquinone). Lifetime anthracycline exposure was calculated by multiplying the cumulative dose (milligram per squared millimeter) of individual anthracyclines received by a factor that reflects the drug's cardiotoxic potential,¹⁶ and then summing the results. Radiation to the chest with heart in the field was captured as a yes/no variable.

Genotyping and Quality Control

Discovery set.

Genotyping of study samples and quality control (QC) replicates was conducted at the Cancer Genomics Research Laboratory of the National Cancer Institute. Germline DNA from 5,739 CCSS participants was genotyped on HumanOmni5Exome arrays (Illumina Inc, San Diego, CA). Details of the DNA sample extraction, QC, mapping, variant identification, and annotation are as described previously.¹⁷ Imputation on the basis of the 1000 Genomes Project release v.3 reference haplotypes using IMPUTE v.2.3.0 yielded 26,135,904 high-quality SNPs and small insertions or deletions. Multidimensional scaling¹⁸ was used to cluster individuals into non-Hispanic White (n = 5,589) and others (n = 150). Both the overall genomic control inflation factor (λ = 1.07) and the quantile-quantile plot (Data Supplement, online only) for genome-wide marginal effect tests did not suggest any large-scale systematic bias because of population stratification. From the 26,135,904 autosomal SNPs, we removed 19,151,628 SNPs with minor allele frequency < 0.05 and 852,536 SNPs that failed the Hardy-Weinberg equilibrium test (P < .000005), yielding 6,131,740 common autosomal SNPs in the final data set consisting of 1,866 non-Hispanic White childhood cancer survivors who carried a history of anthracycline exposure.

Replication set.

Germline DNA was isolated from peripheral blood (QIAamp/Qiagen kits) or saliva (Oragene kits). Genotyping was performed on the Juno system (Fluidigm, San Francisco, CA) according to manufacturer's instructions. End point fluorescence values were measured on the BioMark HD system, and the Fluidigm SNP Genotyping Analysis software program was used to generate genotyping calls for each sample.

RNA Isolation, Library Construction, and Sequencing

RNA was isolated using the PAXgene whole blood RNA kit (Qiagen Inc, Valencia, CA). RNA concentration was measured using Nanodrop ND-1000 Spectrophotometers (Thermo Fisher Scientific Inc, Waltham, MA). RNA quality was checked on Bioanalyzer Nanochip (Agilent Technologies, Palo Alto, CA) and samples with RNA integrity number > 7 were submitted to the Genomic Services Laboratory at HudsonAlpha Institute for Biotechnology, Huntsville, AL. Poly-adenylated RNAs were isolated using NEBNext Magnetic Oligo d(T)25 beads. Libraries were prepared using the TruSeq RNA Sample Preparation Kit (Illumina Inc). Each library was pair-end sequenced (100 bp) by using the TruSeq SBS Kit v4-HS (Illumina Inc), on a HiSeq2500 platform. Raw reads were demultiplexed using bcl2fastq Conversion Software (Illumina Inc) with default settings.

Differential Gene Expression Analysis

TrimGalore!¹⁹ was used to trim off primer adapter sequences found in the raw FASTQ files. STAR was used to align the trimmed RNA-Seq fastq reads to the human reference genome from Gencode (GRCh38 p7 Release 25).²⁰ HTSeq-count was used to count the number of reads mapping to each gene from the STAR alignments.²¹ Normalization and differential expression were applied to the count files using DESeq2.²² The q-values (adjusted P values) from DESeq2 are adjusted for multiple testing with the Benjamini-Hochberg procedure, which controls false-discovery rate.

Statistical Analyses

Analyses were conducted using R,²³ SAS (SAS Institute Inc, Cary, NC) software, and PLINK.¹⁸

Discovery stage.

Main effect.

Cox regression model (Data Supplement, online only) was used to investigate the main effect of each of the 6,131,740 SNPs (assuming an additive genetic model) that passed QC, minor allele frequency, and Hardy-Weinberg equilibrium filters on the risk of anthracycline-related cardiomyopathy; time to cardiomyopathy started from primary cancer diagnosis. In addition to the SNPs and the top 10 genotype-based principal components, we included age at diagnosis of primary cancer (continuous variable), sex, chest radiation (yes/no), anthracycline dose, and CVRFs (yes/no), on the basis of previous evidence regarding association of these variables with cardiomyopathy.^2,5,10–12

Gene-environment interaction.

We used Cox regression model (Data Supplement) to detect a role for gene-environment (anthracycline dose) interaction on the risk of anthracycline-related cardiomyopathy. In addition to the SNPs, SNP-anthracycline interactions, and the top 10 genotype-based principal components, we included age at diagnosis of primary cancer (continuous variable), sex, chest radiation (yes/no), anthracycline dose, and CVRFs (yes/no) in the analysis.

SNPs with P ≤ 5 × 10⁻⁸ in the main effect or gene-environment interaction analysis were retained from the discovery analysis, and verified in the replication data set.

Replication stage.

Main effect.

Conditional logistic regression (Data Supplement) was used to investigate the main effect of each SNP (assuming an additive genetic model) identified as significantly associated with cardiomyopathy in the discovery stage. Age at diagnosis of primary cancer (continuous variable), sex, chest radiation (yes/no), anthracycline dose, CVRFs (yes/no), and the SNPs of interest were included in the analysis.

Gene-environment interaction.

We used conditional logistic regression (Data Supplement) to replicate the gene-environment (anthracycline dose) interaction effect of a SNP identified as significantly associated with anthracycline-related cardiomyopathy in the discovery stage. Age at diagnosis of primary cancer (continuous variable), sex, and chest radiation (yes/no), anthracycline dose, CVRFs (yes/no), SNP, and SNP-anthracycline interaction were included in the analysis.

RESULTS

Demographic and Clinical Characteristics

Discovery cohort.

The discovery cohort included 126 childhood cancer survivors with heart failure and 1,740 without. As shown in Table 1, the median age at primary cancer diagnosis for cases and controls was 7 years (range, 0-20 years) and 11 years (range, 0-20 years), respectively. Cases received a higher cumulative anthracycline exposure (median dose, 362.1 mg/m² [range, 50-917.8] v 277.9 mg/m² [10-1,120], P < .0001), and were more likely to carry a history of CVRFs (79.4% v 33.2%, P < .0001). Median time between cancer diagnosis and heart failure was 22 years (range, 0-39 years); survivors without heart failure were followed for a significantly longer period (median, 31 years [14-45], P < .0001).

TABLE 1.

Demographic and Clinical Characteristics of Anthracycline-Exposed Individuals

Replication case-control set.

The replication set included 105 cases and 160 matched controls. As shown in Table 1, the median age at primary cancer diagnosis for cases and controls was 6.6 years (range, 0-20.6 years) and 8.6 years (range, 0.3-21.7 years), respectively. Cases received a higher cumulative anthracycline exposure (median dose, 348 mg/m² [60-760] v 250 mg/m² [9-600], P = .003), and were more likely to carry a history of CVRFs (39.1% v 6.9%, P < .0001). Median time between cancer diagnosis and cardiomyopathy was 7.1 years (range, 0.1-27.7 years); controls were followed for a significantly longer period (median, 10.7 years [1.2-33], P <.0001).

Main Effect Analysis

No SNPs crossed the prespecified threshold for genome-wide significance (Table 2) in the discovery cohort. The smallest P value was observed for SNP rs9418663 on gene DOCK1 (P = 5.83 × 10⁻⁷).

TABLE 2.

Genome-Wide Significant Associations Identified in the CCSS Cohort and Replicated in the COG Case-Control Set

Gene-Environment Interaction Analysis

As shown in Table 2 and Figure 1, two SNPs in the discovery cohort demonstrated significant gene-anthracycline interaction effect (rs113230990, near a CCCTC-binding factor [CTCF] insulator [< 750 base pairs], and rs17736312 on gene ROBO2 [also known as SAX3]). After adjustment for age at primary cancer diagnosis, sex, chest radiation, anthracycline dose, and CVRFs, the per-allele and per-mg/m² anthracycline exposure hazard ratio was estimated to be 1.009 (95% CI, 1.006 to 1.012; P = 3.58 × 10⁻⁹) for rs113230990 and 1.004 (95% CI, 1.003 to 1.005; P = 5.17 × 10⁻⁸) for rs17736312. The gene-anthracycline interaction effect of rs17736312 (ROBO2) was successfully replicated in the COG case-control set (hazard ratio, 1.005; 95% CI, 1.001 to 1.009; P = .032). SNP rs17736312 resides on gene ROBO2, which is a relatively high recombination region (recombination rate > 40 cM/Mb; Fig 2). In humans, the genome-wide average recombination rate ranges from 0.029 cM/Mb to 4.26 cM/Mb for a 5-Mb window.²⁴

FIG 1.

Manhattan plot of a genome-wide association study on SNP by anthracycline interaction analysis for anthracycline-related cardiomyopathy in the discovery cohort of Childhood Cancer Survivor Study of European ancestry. SNP*anthracycline association with anthracycline-related cardiomyopathy are expressed as –log10 (P) on y-axis. Chromosomes 1-22 are labeled on the x-axis. SNPs, single-nucleotide polymorphisms.

FIG 2.

LocusZoom plot of the index SNP rs17736312 (chr3:76484466) on gene ROBO2. SNP by anthracycline association with anthracycline-related cardiomyopathy are expressed as –log10 (P) on the y-axis obtained from analysis of common variants in the discovery cohort of Childhood Cancer Survivor Study of European ancestry. The results within ± 250 kb of the index SNP rs17736312 are shown. SNPs are shown as circles and the index SNP has the largest –log10(P). All SNPs are color-coded according to the strength of linkage disequilibrium with the index SNP (as measured by r² in the European 1000 Genomes project data). SNPs, single-nucleotide polymorphisms.

As shown in Table 3, individuals exposed to high-dose anthracyclines (> 250 mg/m²) and carrying AA genotypes for rs17736312 had a higher risk of cardiomyopathy compared with individuals carrying GG/AG genotype with low-dose (≤ 250 mg/m²) anthracycline exposure (discovery cohort: odds ratio [OR], 2.2; 95% CI, 1.2 to 4.0; P = .009; replication set: OR, 8.2; 95% CI, 2.0 to 34.4; P = .004). Individuals exposed to low-dose anthracyclines (≤ 250 mg/m²) and carrying AA genotypes had a higher risk of cardiomyopathy compared with individuals carrying GG/AG genotype (discovery cohort OR, 1.8,; 95% CI, 1.1 to 2.9; P = .02; replication set = 2.96; 95% CI, 1.1 to 7.7; P = .03).

TABLE 3.

Main and Modifying Effect of ROBO2 rs17736312 Genotypes on Dose-Dependent Risk of Anthracycline-Related Cardiomyopathy in the Discovery Set and the Replication Set

The ROBO family of genes encode transmembrane receptors (Robo) that bind Slit guidance ligands (SLIT)^25,26 with functions linked to cell adhesion, migration, growth and survival.²⁷ A recent observation suggests that the Slit-Robo signaling pathway promotes cardiac fibrosis via the transforming growth factor (TGF)-β1/small mothers against decapentaplegic (Smad) pathway.²⁸ We examined the main effect, gene*anthracycline, gene*gene, and gene*gene*anthracycline interaction effects for genes ROBO2, SLIT1, SLIT2, and TGF-β1 using the CCSS cohort (Data Supplement). As shown in the Data Supplement, the following significant gene-level associations with anthracycline-related cardiomyopathy were identified: main effect (TGF-β1, P = .007); gene*anthracycline interaction effect (ROBO2*anthracycline, P = .0003); and gene*gene*anthracycline interaction (SLIT2* TGF-β1*anthracycline, P = .009). In addition, we demonstrated an association between TGF-β1 gene expression and ROBO2 and SLIT2 genotypes (Data Supplement).

DISCUSSION

We performed a genome-wide association study to identify genetic variants associated with anthracycline-related cardiomyopathy. In the discovery cohort, we identified two SNPs demonstrating significant gene-anthracycline interaction effect (rs113230990, near a CTCF insulator binding site and rs17736312 on gene ROBO2). SNP rs17736312 on gene ROBO2 was successfully replicated. Thus, individuals exposed to high-dose anthracyclines and carrying the AA genotype were at a two-fold to eight-fold higher risk of developing cardiomyopathy/heart failure in the discovery cohort and replication set respectively, when compared with patients who carried the GG/AG genotype and had received low-dose anthracyclines. To our knowledge, this is the first study reporting the modifying effect of a polymorphic variant in ROBO2 in the dose-dependent association between anthracyclines and cardiomyopathy.

SNP rs17736312 resides on gene ROBO2, which is a member of the roundabout (ROBO) family of proteins that encodes highly conserved transmembrane receptors. There are four human Robo homologs, ROBO1, ROBO2, ROBO3, and ROBO4, with some redundancy in the function of Robo receptors. Robo receptors are members of the immunoglobulin superfamily of cell adhesion molecules. These transmembrane receptors bind Slit guidance ligands (SLIT)^25,26 with functions linked to cell adhesion, migration, growth, and survival.²⁷ Slit-Robo signaling is involved in many aspects of heart development, including migration and alignment of the cardiac cell, formation of the lumen, the heart chamber, the ventricular septum, semilunar and atrioventricular valves, caval veins, and pericardium in Drosophila, zebrafish, and murine models.^29–35 Slit-Robo signaling is implicated in ventricular septal defects, bicuspid aortic valves, and tetralogy of Fallot in humans.^36,37

Extracellular matrix remodeling plays an important role in anthracycline-related cardiomyopathy. Cardiac fibroblast is the keystone of fibrogenesis, such that activation of cardiac fibroblasts into myofibroblasts enhances the synthesis of collagen fibers, fibronectin, and profibrotic mediators, thus disrupting normal extracellular matrix.^38–41 The collagen fibers embed in the myocardium, altering its architecture and potentiating heart failure.⁴² In anthracycline-related cardiomyopathy, TGF-β1–induced fibroblast activation results in adverse extracellular matrix remodeling,^43–51 involving genes related to TGF-β and collagen turnover.^46,52–57 Given the observation that the Slit–Robo signaling pathway promotes cardiac fibrosis via the TGF-β1/Smad pathway,²⁸ we examined the role of ROBO2, SLIT, and TGF-β1 in the development of anthracycline-related cardiomyopathy. We successfully demonstrated a significant gene-level association between anthracycline-related cardiomyopathy and TGF-β1. We were able to also demonstrate gene*gene*anthracycline interaction effect between SLIT2, TGF-β1, and anthracycline exposure. Finally, we were able to demonstrate an association between SNPs in ROBO2 and SLIT2 and TGF-β1 gene expression. These studies provide credence to the biologic plausibility of our identified association between SNPs on ROBO2 and anthracycline-related cardiomyopathy, suggesting that high doses of anthracyclines combined with genetic variants involved in the profibrotic Slit-Robo signaling pathway promote cardiac fibrosis via the TGF-β1/Smad pathway, thus increasing the risk of anthracycline-related cardiomyopathy. Furthermore, these findings suggest that suppressing Robo signaling and/or the TGF-β1/Smad signaling can possibly lead to suppression of cardiac fibrosis.

This study should be considered in the context of its limitations. The diagnosis of cardiomyopathy is based on self-report in the CCSS. However, we used a stringent definition of cardiomyopathy, such that only those outcomes graded as severe (grade 3; self-reported cardiomyopathy or CHF, plus medications), life-threatening (grade 4; requiring heart transplantation), or fatal (grade 5) were included. The COG replication set differed from the CCSS discovery set in study design (matched case-control), the methodology used for ascertaining cardiomyopathy (source validation with physical examination and echocardiographic indices), and the population characteristics (shorter length of follow-up, lower prevalence of CVRFs, and a diverse racial/ethnic background). However, despite these differences in the discovery cohort and replication case-control set, we were successful in replicating our findings generated in the discovery cohort.

These limitations notwithstanding, the study has considerable strengths. We used CCSS for discovery—a large childhood cancer survivor population with detailed therapeutic exposures and phenotypic data. The COG replication set is one of the largest populations of clinically validated cardiomyopathy occurrences among childhood cancer survivors. The successful replication of a common SNP (allele frequency for A allele is 57%) on ROBO2 and the associated biological plausibility underscores the robustness of the findings of the connection between the identified SNP rs17736312 on gene ROBO2 and anthracycline-related cardiomyopathy.

AUTHOR CONTRIBUTIONS

Conception and design: Xuexia Wang, Purnima Singh, Smita Bhatia

Financial support: Gregory T. Armstrong, Smita Bhatia

Administrative support: Gregory T. Armstrong, Smita Bhatia

Provision of study materials or patients: Melissa M. Hudson, Saro H. Armenian, Joseph P. Neglia, A. Kim Ritchey, Jill P. Ginsberg, Leslie L. Robison, Gregory T. Armstrong, Smita Bhatia

Collection and assembly of data: Purnima Singh, Lindsey Hageman, Melissa M. Hudson, Saro H. Armenian, Joseph P. Neglia, A. Kim Ritchey, Jill P. Ginsberg, Leslie L. Robison, Gregory T. Armstrong, Smita Bhatia

Data analysis and interpretation: Xuexia Wang, Purnima Singh, Liting Zhou, Noha Sharafeldin, Wendy Landier, Paul Burridge, Yutaka Yasui, Yadav Sapkota, Javier G. Blanco, Kevin C. Oeffinger, Melissa M. Hudson, Eric J. Chow, Joseph P. Neglia, Douglas S. Hawkins, Jill P. Ginsberg, Smita Bhatia

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Genome-Wide Association Study Identifies ROBO2 as a Novel Susceptibility Gene for Anthracycline-Related Cardiomyopathy in Childhood Cancer Survivors

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/jco/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).