Whole-Genome Sequencing Explores Why Low-Risk Patients Develop Bleomycin Lung Injury

Bleomycin remains one of the most valuable drugs in modern oncology, forming the backbone of curative-intent chemotherapy for germ cell tumours and Hodgkin lymphoma. Yet its use is shadowed by bleomycin-induced pneumonitis, a lung toxicity that can range from a silent decline in lung function to progressive interstitial fibrosis and death. A new multicentre study has now applied whole-genome sequencing to a national cohort of patients who developed this complication despite carrying few or none of the recognised clinical risk factors, in an effort to uncover whether inherited genetic variation might explain their susceptibility.

The retrospective observational study, conducted across nine treating institutions in Australia between June 2020 and June 2022, enrolled 21 patients through the Exceptional Responders program. Thirteen had germ cell tumours and eight had Hodgkin lymphoma. Eligible cases required prior bleomycin exposure and clinically significant pneumonitis, defined either as new respiratory symptoms with radiographic changes and pulmonary function test decline, or as an asymptomatic drop in diffusing capacity for carbon monoxide exceeding 25 percent during therapy that could not be attributed to infection or tumour involvement. Crucially, the investigators excluded patients with high baseline risk, specifically those with an estimated glomerular filtration rate of 70 mL/min or below, a cumulative bleomycin dose of 400,000 IU or more, or age 70 or above, leaving a cohort considered predominantly clinically low risk.

The clinical picture that emerged underscored how unpredictable this toxicity remains. The median age at diagnosis was 39 years for patients with germ cell tumours and 49 years for those with Hodgkin lymphoma, and 67 percent of the cohort were lifetime non-smokers. Most pneumonitis cases, 76 percent, were graded below CTCAE grade 3, yet nearly a quarter reached grade 3 or higher. Corticosteroids were given to 52 percent of patients and two received additional immunosuppressive agents. Two deaths occurred among bleomycin-exposed patients: one patient with Hodgkin lymphoma treated with ABVD died in intensive care from pneumonitis complicated by acute infection, the single BIP-related death in the cohort, while a second patient, treated for germ cell tumour at age 33, later died at 54 from hypoxic respiratory failure due to COVID-19 in the context of bleomycin-related pulmonary fibrosis. Before pneumonitis onset, patients had received a median of nine bleomycin doses in the germ cell tumour group and four in the lymphoma group, corresponding to median cumulative doses of 270,000 IU and 69,200 IU respectively.

To search for genetic contributors, the team obtained blood samples for whole-genome sequencing from 18 of the 21 patients. Sequencing was performed at the Garvan Institute of Medical Research on an Illumina NovaSeq 6000 platform to a minimum depth of 30x, with reads aligned to the GRCh38 reference genome and germline variants called using GATK best practices. Rather than scanning the entire genome unguided, the analysis centred on a curated panel of 29 candidate genes compiled from a literature review of genes and alleles previously implicated in bleomycin toxicity. This panel spanned genes involved in bleomycin detoxification such as BLMH, inflammatory signalling including IL6 and IL6ST, surfactant biology including SFTPA2 and ABCA3, mucociliary clearance through MUC5B, telomere maintenance through TERT and RTEL1, antioxidant defence via NFE2L2, and innate immune signalling through TLR4.

The variant search followed several filtering strategies. Five candidate variants with predicted high impact, meaning likely to severely disrupt protein expression, were initially found in four genes, but none survived manual curation. Two predicted frameshift variants in MUC5AC sat close together in a haplotype expected to preserve the reading frame and therefore be non-damaging; variants in NFE2L2 and SFTPA2 were dismissed because their alternative alleles are in fact the common major alleles in population datasets, with the SFTPA2 variant classified as benign in ClinVar; and a splice donor variant in PARN affected only transcripts that are not protein-coding. Structural variants within panel genes were likewise discarded, either because they were unlikely to be damaging or because their predicted breakpoints fell in repeat-rich regions where sequence alignment was unreliable.

Attention then turned to moderate-impact variants, such as missense changes that may alter protein function without abolishing it. After applying a population allele frequency filter of 0.01 or below and inspecting sequence alignments, 14 rare missense variants remained across nine genes, distributed across 11 samples, each present in a single patient. Among these, two ABCA3 variants, rs141621969 (Ala501Glu) and rs45592239 (Pro766Ser), each observed in separate patients, carry general population allele frequencies of 0.003 and 0.002. Although neither has previously been linked to bleomycin-induced pneumonitis, both alleles have been described in association with interstitial lung disease, providing a rationale for further investigation of ABCA3, a key regulator of surfactant metabolism in type II alveolar cells. The mucin genes MUC5AC and MUC5B showed the greatest sequence variability, and in one patient a distinctive region of high heterozygosity of roughly 20 kilobases extended across the large central exon 31 of MUC5B, a finding the authors consider unlikely to be artefactual and one that may warrant long-read sequencing validation.

The analysis also flagged a depletion signal in IL6ST, the gene encoding the gp130 signalling subunit of the interleukin-6 receptor complex. Two linked variants, rs2228044 and rs2228043, forming the Gly148Arg-Leu397Val haplotype, have population allele frequencies of 0.21 and 0.17 respectively, yet only one of 36 alleles in the cohort carried the haplotype, an observed frequency of 0.028 with an unadjusted p value below 0.01. This depletion did not withstand correction for multiple testing, but its biological plausibility is notable: phosphorylated STAT3 is elevated in bleomycin-induced pneumonitis, and experimental STAT3 inhibition mitigates bleomycin-induced lung injury in models. Separately, the BLMH variant rs1050565 (Ile443Val), annotated in the ClinPGx pharmacogenetic database with associations between the C/C genotype and bleomycin toxicity phenotypes, showed a possible but non-significant enrichment in the cohort, with a variant allele frequency of 0.41 against an expected 0.29 and a p value of 0.09.

Polygenic risk score analysis offered a different lens on common genetic variation. Twelve patients of European ancestry were compared against a reference group of 667 European individuals from the 1000 Genomes Project using 38 polygenic risk scores drawn from the Polygenic Score Catalog, covering lung function, idiopathic pulmonary fibrosis, interstitial lung disease and lung cancer, together with scores for autoimmune and autoinflammatory diseases. After Bonferroni correction for multiple comparisons, with significance set at p below 0.0013, there was no significant difference in the distribution of score percentiles between the pneumonitis cases and the reference population. The authors caution that this negative result should not be read as evidence against a genetic contribution to susceptibility, since the available scores were developed for common pulmonary and immune traits rather than for this rare toxicity, and the small cohort limited statistical power.

The study’s limitations are substantial and openly acknowledged. The cohort was small and retrospectively assembled, a consequence of the rarity of clinically significant pneumonitis in low-risk patients, and the sample size was determined by eligible patient numbers rather than a formal power calculation. Without a control cohort of bleomycin-exposed patients who did not develop pneumonitis, distinguishing true predisposing variants from random genomic variation is difficult. Polygenic analyses were restricted to individuals of European ancestry and may not capture the genetic architecture across diverse populations, a point reinforced by the authors’ observation that the minor allele frequency of the IL6ST Gly148Arg variant ranges from 3.52 percent in East Asian to 44.25 percent in African American gnomAD populations.

What the study delivers, ultimately, is a hypothesis-generating map rather than an answer. No single high-penetrance variant or unifying genomic signal emerged to explain why predominantly low-risk patients develop this toxicity, underscoring its likely multifactorial nature. Instead, the work establishes a clinically annotated germline whole-genome sequencing resource, with data deposited in the European Genome-phenome Archive under accession EGAD50000002221 to enable global collaboration, and nominates candidate genes and pathways, particularly in surfactant biology and inflammatory signalling, for validation in larger cohorts. The authors argue that large-scale genomic comparisons of bleomycin-exposed patients with and without pneumonitis, integrating functional analysis and detailed phenotyping, will be needed to define the biological determinants of susceptibility, with particular focus on variants in genes such as BLMH, ABCA3 and IL6ST. For a drug that remains integral to curative regimens for many young patients, refining the ability to anticipate who will be harmed could preserve bleomycin’s therapeutic benefit while minimising avoidable long-term harm.

Subject of Research: Germline genetic contributors to bleomycin-induced pneumonitis in clinically low-risk patients with germ cell tumours or Hodgkin lymphoma

Article Title: A multicenter whole-genome sequencing cohort of bleomycin-induced pneumonitis in predominantly clinically low-risk patients

Article References: Hitchen, N., Handel, E. E., Leslie, C., Di Ciaccio, P. R., Joshua, A., Ardolino, L., Grimison, P., Honeyball, F., Tran, B., McCarthy, L., Milliken, S., Viiala, N., Gangatharan, S., Bilmon, I., Zhang, A. Y., Lewin, J., Marx, G., Murphy, C., Trinh, A., … Barnet, M. B. (2026). A multicenter whole-genome sequencing cohort of bleomycin-induced pneumonitis in predominantly clinically low-risk patients. Molecular Genetics and Genomics, 301(1), 208. https://doi.org/10.1007/s00438-026-02534-9

Image Credits: AI Generated

DOI: 10.1007/s00438-026-02534-9

Keywords: bleomycin-induced pneumonitis, whole-genome sequencing, germ cell tumours, Hodgkin lymphoma, polygenic risk scores, ABCA3, IL6ST, BLMH, MUC5B, lung toxicity, germline variants, chemotherapy