Genetic Susceptibility to Anthracycline-Induced Cardiotoxicity in Australian Pediatric Cancer Patients

Graphical abstract

What is the clinical question being addressed?

Which germ line variants are associated with anthracycline-induced cardiotoxicity in Australian pediatric cancer survivors?

What is the main finding?

TTN rs2303838 showed a nominally protective association, whereas SLC28A3 rs7853758 was associated with increased risk; exploratory analyses highlighted 5 novel candidate genes with potential relevance to cardiotoxicity.

Survival rates for childhood cancers have improved over the past decades, due in part to the use of anthracyclines, which are highly effective chemotherapy agents and used in more than half of pediatric chemotherapy regimens. Despite their antitumor efficacy, their use is limited by the risk of anthracycline-induced cardiotoxicity (ACT), which commonly presents as symptomatic cardiomyopathy or asymptomatic left ventricular dysfunction. Risk is influenced by cumulative exposure and clinical factors, yet these do not fully explain interindividual susceptibility. Germ line variants in genes related to drug transport, metabolism, oxidative stress, myocardial structure, DNA damage response, and intracellular signaling have been associated with ACT, although replication has been inconsistent.¹

This study investigated genetic susceptibility to ACT in Australian pediatric cancer survivors by assessing previously ACT-associated variants and conducting an exploratory analysis.

Ethics approval was obtained from the Royal Children’s Hospital HREC (HREC38130), with written consent from participants selected from 2 pediatric cohorts (ACTive, ACOR). Eligible patients had a cancer diagnosis, anthracycline exposure, baseline echocardiography, no pre-existing cardiac abnormality, and a sample available in the biobank. ACT cases were identified using Cancer Therapy-Related Cardiac Dysfunction criteria, defined as ejection fraction <53% on echocardiography ≥21 days after anthracycline administration. Controls had no clinical evidence of cardiotoxicity. Cases and controls were matched as close as possible by treatment era (within 10 years), cancer diagnosis, age group, and cumulative anthracycline dose, with residual variation accounted for in adjusted analyses. Demographic, clinical, and treatment data were extracted from medical records/REDCap. Cumulative anthracycline dose was calculated as doxorubicin equivalents.

Genomic DNA was extracted from peripheral blood-derived biospecimens, with whole exome sequencing available for 18 patients and whole genome sequencing performed for 34 additional patients. Reads were aligned to GRCh38, and variant calling was restricted to the intersecting exome-capture regions. Genotypes were callable if they met thresholds for read depth (≥10), genotype quality (≥20), and allele balance (20%-80% alternate reads for heterozygotes, ≥85% for homozygous alternate, and <5% for homozygous reference); variants with >10% missing genotype calls were excluded and not imputed. Genetic relatedness was assessed using PCA with SNPRelate v1.46.0, with no case-control separation observed.

Previously reported ACT-associated variants were curated from 2 systematic reviews^1,2 and tested when present in the final callset. Carrier status was compared between cases and controls using binomial logistic regression adjusted for age at diagnosis, sex, and cumulative anthracycline dose, with uncorrected P < 0.05 used for nominal associations. Fisher exact test was used for post hoc power calculation. SLC28A3 rs7853758 meta-analysis used a random-effects model to combine previously published study-level ORs and 95% CIs³ with the present cohort estimate using the “metafor” and “forestplot” R package.

For exploratory analysis, functional (missense, frameshift, stop-gain, start-lost, splice-site, protein-altering, in-frame indel) rare or low-frequency variants (minor allele frequency [MAF]<5%) were retained, and excluding genes listed in Catalogue of Somatic Mutations in Cancer. Gene-based association was tested using the adjusted sequence kernel association test R-package, with unweighted and Combined Annotation-Dependent Depletion (CADD)-weighted models adjusted for age, sex, and cumulative anthracycline dose. Multiple testing was addressed using the Benjamini-Hochberg false discovery rate. Leave-one-variant-out jackknife analysis assessed whether gene-level signals were driven by individual variants. Variant impact was interpreted using CADD and a composite multi-tool score (MTS) from AlphaMissense, MutationAssessor, PROVEAN, PolyPhen-2, and VEST, using thresholds for pharmacogenomics⁴ with higher values indicating greater support for a deleterious effect.

The cohort included 52 patients in 26 matched case-control pairs. The mean age at cancer diagnosis was 7.4 years (SD: 5.6; range 0-17), 55.8% were female, and common diagnoses were acute myeloid leukemia (40.4%) and acute lymphoblastic leukemia (23.1%). The mean cumulative anthracycline dose was 314.8 mg/m² (SD: 255.7; range: 20.0-961.0).

Among 16 previously reported variants present in the final callset, 2 showed nominal associations. TTN rs2303838 CT or TT genotypes were carried by 5/26 cases and 14/26 controls and were associated with reduced odds of ACT (OR: 0.20; 95% CI: 0.06-0.71, P = 0.0125). SLC28A3 rs7853758 GA or AA were carried by 10/26 cases and 3/26 controls and were associated with increased odds of ACT (OR: 4.85; 95% CI: 1.15-20.56, P = 0.032). Post hoc power estimates were 67.4% and 50.2%, respectively. Updated SLC28A3 rs7853758 meta-analysis (Figure 1) incorporating the present cohort resulted in a pooled OR of 1.21 (95% CI: 0.42-3.49; P = 0.721; 95% prediction interval: 0.08-18.29) with high heterogeneity (tau² 1.63, I² 92.7%, Q(6) 35.5; P = 0.00001).

Figure 1.

SLC28A3 rs7853758 Meta-Analysis

This forest plot shows a meta-analysis of the association between SLC28A3 rs7853758 and anthracycline-induced cardiotoxicity after incorporation of the present study results. The pooled estimate was not statistically significant but showed high and significant between-study heterogeneity.

Exploratory gene-based analysis included 32,530 rare or low-frequency functional variants across 11,525 genes. While no gene survived multiple-testing correction (false discovery rate >0.65), the genes with nominal P < 0.01 in either unweighted or CADD-weighted analysis were prioritized as potential candidate genes (TRIM34, SLC12A4, HTR3D, HPS5, SERPIND1) for variant annotation and biological plausibility assessment. Jackknife analysis showed each signal was largely driven by one specific variant: TRIM34 rs3740997, a missense variant present only in 6 cases (MAF 0.0338; CADD 22.0; MTS 0.8); SLC12A4 rs770789419, a 12 bp deletion in 5′ untranslated region in 7 cases (MAF 0.0268; CADD 11.72); HTR3D rs77099580, a splice acceptor variant in 6 cases (MAF 0.0310; CADD 23.10); HPS5 rs61884288, a missense variant in 6 cases (MAF 0.0236; CADD 19.68; MTS 0.4); and SERPIND1 rs5903, a missense variant in 5 cases (MAF 0.0067; CADD and MTS 0).

This study provides the first analysis of genetic susceptibility to ACT in an Australian pediatric cancer survivor cohort. Among previously reported variants, 2 showed nominal associations using uncorrected P values: TTN rs2303838 showed a protective association, supporting a finding previously reported in only one study, whereas SLC28A3 rs7853758 was enriched in cases. However, given the number of candidate variants tested, wide CIs, and low-to-moderate post hoc power estimates, the observed effect sizes may be overestimated and should be interpreted as exploratory findings. SLC28A3 rs7853758 was therefore further assessed in the context of previous studies through meta-analysis. Although inclusion of the present study shifted the pooled estimate toward increased risk, the overall association remained nonsignificant and substantial heterogeneity persisted, likely reflecting differences across studies in cardiotoxicity definitions, study populations, ancestry composition, and duration of follow-up. The wide prediction interval further indicates uncertainty in the expected effect across populations, cautioning against utilizing this variant as either protective or risk-conferring for ACT.

The exploratory analysis identified 5 novel but preliminary candidate genes, which should be interpreted as exploratory findings only, as none survived multiple-testing correction and each signal was largely due to a single variant. Among these, TRIM34, SLC12A4, and HTR3D had the clearest biological plausibility. Specifically, TRIM34 may modify cardiomyocyte susceptibility to ferroptosis under anthracycline-induced oxidative and iron stress, given the gene’s role in ferroptosis resistance and preclinical evidence implicating other TRIM-family proteins (TRIM65, TRIM52, and TRIM25) in doxorubicin cardiotoxicity. SLC12A4 encodes the K⁺/Cl^– cotransporter (KCC1), which is expressed in cardiac tissue and involved in ion homeostasis and cell volume regulation. HTR3D encodes the serotonin receptor subunit identified in cardiac mitochondria, where it regulates mitochondrial activity and cellular function. HPS5 and SERPIND1 relate to platelet dense granule biology, coagulation, and thrombin-mediated cardiac injury pathways, but their association with ACT remains less direct and more speculative. Clinical translation of these findings would require replication in larger, independent, and ancestrally diverse cohorts, consistent direction and ideally functional validation before a variant could be used for screening or cardioprotective intervention.

Although limited by exonic analysis and modest sample size, which reduced power for discovery, this study identified preliminary biologically plausible candidates requiring replication and functional follow-up and provided additional evidence for 2 previously ACT-associated genetic variants.

Funding support and author disclosures

This work was supported by a National Heart Foundation Cardio Oncology Strategic Grant (APP105525) and a Kids Cancer Project Grant. Dr Conyers was supported by a VESKI FAIR Fellowship, Genomics Health Futures Mission Medical Research Future Fund grants (2044744, 2024900), and a Murdoch Children’s Research Institute (MCRI) Clinician Scientist Fellowship. Dr Mateos was supported by the Kids’ Cancer Project Col Reynolds Clinical Fellowship. Dr Elliott is a principal investigator of the Novo Nordisk Foundation Center for Stem Cell Medicine, reNEW (NNF21CC0073729), and is supported by the Medical Research Future Fund (MRF2024440). Drs Elliott and Rossello receive support from the Novo Nordisk Foundation Center for Stem Cell Medicine: reNEW (NNF21CC0073729). The MCRI is supported by the Victorian Government’s Operational Infrastructure Support Program. Dr Rossello has received institutional and salary support as a coinvestigator and subcontractor with the Peter MacCallum Cancer Centre for an investigator-initiated trial which receives funding support from Regeneron Pharmaceuticals, and as a coinvestigator on a translational research project funded by a Regeneron Pharmaceuticals grant; and has also received travel expenses from MGI Australia and New Zealand. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.