Childhood Cancer Treatment and Clonal Hematopoiesis

Childhood cancer survivors experience a heavy burden of long-term complications, including earlier onset of aging-associated cardiometabolic conditions and significant risk of secondary malignancies (1, 2). Predicting and mitigating these late effects of treatment have emerged as pressing challenges in an era when most children with cancer can be cured of their primary malignancy. Cytotoxic treatments, encompassing chemotherapeutics and ionizing radiation, are the fulcrum of effective childhood cancer therapy and are also accountable for the bulk of late effects. Cytotoxic treatments may damage organs through direct cell killing or by inflicting sublethal mutational damage that interferes with cellular function. Moreover, mutational damage may impart a selective advantage on some cells that leads to the distortion of the clonal composition of damaged tissue. This clonal diversity constitutes the substrate for natural selection at the cellular level, which can culminate in secondary cancers.

Due to ease of representative sampling, the hematopoietic system offers an insightful vantage point for studying the impact of age and cytotoxic exposures on tissue clonal dynamics over time. Clonal hematopoiesis (CH) refers to the disproportionate expansion of one somatically mutated hematopoietic stem cell (HSC) clone relative to others. CH becomes increasingly common with age in the general population and is associated with increased overall mortality attributable to both cardiometabolic and malignant conditions (3). Numerous studies have demonstrated that CH is particularly prevalent in adult cancer patients (3). Moreover, CH has emerged as a potentially promising biomarker for the risks of second malignancies and other adverse outcomes of cancer treatment in adults (3). However, the long-term prevalence, genetic landscape, natural history, and clinical significance of CH in childhood cancer survivors are poorly understood. It is in this context that Hagiwara and colleagues contribute a landmark study (4) of CH in survivors of childhood cancer drawn from the St. Jude Lifetime Cohort Study (SJLIFE).

Established in 1962, SJLIFE constitutes one of the largest and most meticulously conducted long-term follow-up studies of childhood cancer survivors (2). Hagiwara and colleagues examined the blood of 2,860 SJLIFE participants and 324 controls for the presence of CH by deep (>1,700-fold coverage) targeted sequencing of 39 relevant cancer genes. Overall, the findings provide robust evidence that childhood cancer survivors experience precocious CH. CH was detected in 15% of survivors, nearly double the prevalence of 8.6% in controls. Analysis of telomere attrition rates provides convincing evidence that the early onset of CH is attributable to mutations acquired in childhood, rather than a higher HSC division rate. Collectively, these results, when considered in the context of previous studies of the general population, suggest that childhood cancer survivors exhibit rates of CH expected of individuals two to three decades older (Fig. 1). Through a series of careful statistical analyses, the investigators then teased out associations between CH and exposure to specific cytotoxic therapies (alkylating agents, bleomycin, estimated ionizing radiation dose to active bone marrow), directly linking CH to childhood cancer treatment.

A further extraordinary finding is the marked enrichment of STAT3-driven CH in Hodgkin lymphoma survivors. The authors leverage single-cell combined genotyping and phenotyping analysis to demonstrate that STAT3-mutated clones are biased toward T-cell differentiation, suggesting a cellular origin distinct from the original Hodgkin lymphoma (a B-cell malignancy). By analyzing whole-genome sequencing data from STAT3-mutated clones, Hagiwara and colleagues were able to investigate the mutational processes responsible for the STAT3 mutations. In all cases studied, the specific DNA sequence context of the STAT3 mutations conferred a very high (98.5%–100%) probability that these variants were caused by a mutational signature (SBS25) attributed to procarbazine, a key cytotoxic agent used in Hodgkin lymphoma treatment. Interestingly, dacarbazine, a closely related alkylating agent, does not leave this mutational imprint and has been adopted in some treatment protocols, as it lacks the substantial germ cell toxicity of procarbazine (with similar anticancer efficacy; refs. 5, 6). If STAT3-driven CH is associated with specific adverse outcomes, Hagiwara's observation would lend further support for substituting procarbazine with dacarbazine.

It is worth reflecting on what Hagiwara and colleagues did not find. Most prominently, they did not observe an association between CH and exposure to platinum chemotherapy. Several studies, including a large survey of CH predominantly in adult cancer patients, identified exposure to platinum agents as a key risk factor for CH (7, 8). Furthermore, detailed investigations of small numbers of children who developed therapy-related leukemia have identified CH laden with platinum-induced mutations (9, 10). It is conceivable that the lack of platinum mutational signatures in the SJLIFE cohort is related to the long interval between end of treatment and blood draw in this study. Platinum-induced clones may have evanesced in the absence of ongoing genotoxic stress. Much more likely, however—the absence of a platinum mutational footprint in the SJLIFE cohort is due to the inability of targeted gene sequencing to detect “driverless” CH—that is, clonal expansions that do not harbor mutations in the 39 genes interrogated. Clones without apparent driver mutations have accounted for the majority of CH in whole-genome sequencing studies of the general population. Targeted sequencing of candidate genes is a sensitive and specific tool for detecting very low variant allele frequency mutations in particular genes but cannot assess mutations elsewhere in the genome. Accordingly, the CH discovered by targeted sequencing in this study may be only the tip of the iceberg. Indeed, studies that have deployed whole-genome sequencing for the detection of CH in children with cancer have readily identified “driverless” CH that would not have been detected by targeted sequencing of cancer genes (8, 9). Therefore, whilst the fundamental discovery of accelerated CH in childhood cancer survivors is likely to stand the test of time, some of the negative signals in this study may have to be revisited in the future as unbiased sequencing methods become more tractable for large studies of CH.

The findings of this study justify further large-scale, longitudinal efforts to understand the impact of childhood cancer treatments on the clonal architecture of blood. Most importantly, we need to understand the significance of CH for our patients. Is there a relationship between CH and premature physiologic aging in long-term childhood cancer survivors? To what extent is CH associated with late effects such as treatment-related leukemia or early onset of aging-related chronic health conditions? Furthermore, a precise, quantitative understanding of the effects of specific cytotoxic agents on the clonal composition of blood may inform chemotherapy choices in common treatment protocols. In an ideal world, pediatric oncology clinical trials would be able to incorporate biomarkers predictive of the late-effects burden of novel treatment approaches. It is currently very challenging to predict whether improvements in 5-year overall or event-free survival may be outweighed by late effects 10 or 15 years down the road. It is conceivable that CH could be helpful in this regard. These are questions that can be answered only through arduous and expensive long-term efforts, such as SJLIFE, which are challenging to deliver in an era of short-termism in research funding and scientific endeavors.