Cancer therapy leaves a genomic record long after treatment ends
Cancer therapy is not biologically silent once the infusion stops or the radiation field closes. Genotoxic treatment can leave behind DNA changes that are still readable years later, turning a tumor genome into an archive of prior exposure, clonal survival, and, in some settings, resistance.
In a JCI viewpoint, Xiaolu Wang and Nilay Sethi examine how therapy-associated mutational signatures are being used to interpret the long genomic afterlife of cancer treatment. Their focus is not simply on whether chemotherapy and radiotherapy damage DNA — that part is already established — but on what those scars actually mean when they show up in relapsed tumors, surviving normal lineages, or late secondary lesions in long-term survivors.
A mutation signature can indicate exposure without proving mechanism. A clone can carry the mark of treatment because it survived the assault, expanded afterward, and remained available for sequencing years later.
Relapsed childhood tumors carry treatment scars, but not every scar explains relapse
The centerpiece of the discussion is a broad genomic analysis by Layeghifard and colleagues involving 611 whole-genome-sequenced tumors from 544 children across three precision-medicine programs. Those tumors were paired with detailed treatment histories spanning 86 therapeutic agents, making it possible to ask which therapies leave detectable genomic imprints, how quickly those imprints emerge, and whether they track with later tumor behavior.
Tumors that progressed after treatment carried higher variant burdens than treatment-naive tumors across several classes of genomic variation, although the pattern depended on disease stage and the time elapsed after exposure. Treated primary tumors sampled fewer than four months after therapy did not show the same increase seen in treated advanced tumors, including relapsed and metastatic lesions, which were sampled nearly two years later. In this setting, time matters because it determines whether a therapy-marked lineage has had enough opportunity to expand to a detectable level.
The strongest exogenous signals came from chemotherapy and radiotherapy. Platinum exposure stood out most clearly. Established platinum-associated signatures — SBS31, SBS35, and DBS5 — were enriched in treated tumors, and additional platinum-linked genomic features also appeared. Platinum-associated signatures could become detectable within months, emerging as early as 91 days after therapy initiation in tumors with sufficient mutation burden. Their prevalence rose with time, crossing more than 33% of platinum-treated tumors within one year and reaching roughly 50% by nearly three years.
A signature is a record of survival, not a verdict on causation
A detectable signature does not automatically mean therapy caused the relapse. It means a cell lineage encountered a mutagen and persisted long enough to be sampled.
That distinction is the backbone of the argument. A signature-positive relapse may reflect therapy-driven mutagenesis, but it may also simply reflect clonal survival after a genotoxic bottleneck. Detection depends on dose, elapsed time, mutation burden, clonal expansion, and the sequencing depth of the sample. A negative result does not exclude prior exposure, because the relevant clone may sit below the limit of detection or may never have acquired a strong enough genomic record to register.
This is especially important when therapy-associated variants are interpreted in recurrent disease. The genomic scar is real. The biological conclusion is narrower. In relapsed tumors, the scar marks the survivor more reliably than it identifies the exact mechanism that produced resistance.
Platinum resistance appears linked to gene expression, not recurrent driver mutations
The discussion also ties these genomic scars to treatment response. Platinum signatures were associated with poorer response to later platinum treatment in independent cohorts, echoing prior clinical observations. Yet the resistance phenotype was linked to higher expression of resistance-associated genes rather than to recurrent therapy-induced driver mutations.
That pattern fits the biology of a resistant clone better than the image of a single new oncogenic lesion. Under selective pressure, surviving cells can reorganize transcriptional programs, alter DNA repair pathways, or shift cell state without needing to acquire a brand-new driver mutation every time. The signature shows where the clone has been. The resistance phenotype tells clinicians what the clone can now tolerate.
Normal stem cell compartments and tumors may be shaped by therapy in different ways
Wang and Sethi use the viewpoint to reconcile two literatures that can appear to point in opposite directions. In relapsed tumors, chemotherapy looks like a mutagenic driver of clonal evolution. In normal blood and other stem cell compartments, therapy often behaves more like a selective filter that enriches for rare preexisting clones over longer timescales.
Both processes can operate at once. In a tumor, therapy can create new variation among surviving cells. In normal tissue, most newly induced mutations disappear unless the cell carrying them persists or gains a selective advantage. Over decades, that difference becomes critical. A normal stem cell still has to acquire the same malignant capabilities that a cancer cell has already assembled.
Therapy-associated polyposis shows how old exposure can reappear as new disease
The viewpoint extends that logic to therapy-associated polyposis (TAP), a rare condition in which survivors of childhood or young adult cancer develop numerous premalignant gastrointestinal polyps decades after treatment, without a recognized hereditary polyposis syndrome.
In a cohort of 29 survivors, the original cancer had been diagnosed at a median age of 16, and polyposis appeared nearly three decades later. Most evaluable patients had received alkylating chemotherapy. TAP adenomas carried SBS25, a signature associated with methyl-hydrazine alkylators such as procarbazine and dacarbazine, along with more structural genomic disruption than adenomas from hereditary conditions.
The long latency is biologically revealing. Colonic epithelium turns over continuously, so a damaged cell is usually lost. Thirty years later, what remains is not preserved injury but a preserved lineage — one that survived treatment, persisted through tissue renewal, and eventually entered neoplastic growth. In TAP, the signature is a molecular trace of exposure and lineage survival, not proof that the signature itself caused the polyposis.
Survivorship care may need a wider genomic frame
These findings raise a direct question for survivorship medicine: how should a therapy signature be used in clinical practice?
Exposure reconstruction is the easiest use case. When treatment records are incomplete or decades old, a sufficiently specific mutational signature may provide molecular evidence that a tissue lineage experienced a particular class of therapy. Response prediction is harder but beginning to look plausible in contexts such as platinum rechallenge, where clinical decision-making already uses the platinum-free interval as a proxy for ongoing sensitivity.
Risk prediction is the most difficult application. A genomic scar is not the same as a cancer-risk calculator. Some therapies increase later malignancy risk without leaving a clean base-substitution signature. Anthracyclines are a key example: they leave no detectable base-substitution signature in normal cells or in clonal whole-genome assays, yet they show a linear dose-response relationship with later breast cancer risk.
The practical implication is blunt. More signature does not automatically mean more future cancer.
Surveillance strategies for childhood cancer survivors have traditionally emphasized radiation field and dose. The TAP data argue that alkylating chemotherapy may also deserve attention when late gastrointestinal risk is being assessed. The unresolved issue is not whether therapy leaves a genomic imprint — it does — but how to separate a record of past exposure from a forecast of future disease, especially when the same scar can mark both a surviving clone and a changing tissue compartment.