TRACERx first 100: tracking how lung cancers evolve, and how chromosomal chaos predicts relapse
Multi-region sequencing of 327 regions from the first 100 TRACERx lung cancers showed that most driver mutations are early and shared while copy-number chaos continues to evolve, and that tumours with high copy-number heterogeneity were nearly five times more likely to relapse or kill the patient.
TRACERx (TRAcking Cancer Evolution through therapy) is a Cancer Research UK prospective study following patients with resected stage I-IIIA non-small-cell lung cancer from surgery to relapse or death. This first report analysed 327 tumour regions from 100 patients with whole-exome sequencing.
Intratumour heterogeneity was pervasive: a median of 30% of mutations were subclonal, and 48% of tumours had subclonal driver alterations. Driver mutations in EGFR, MET, BRAF and TP53 were almost always clonal (early), whereas alterations in PIK3CA, NF1 and chromatin modifiers were often late. Ongoing chromosomal instability (subclonal copy-number alterations) rather than mutational heterogeneity predicted recurrence-free survival: patients whose tumours had elevated copy-number heterogeneity had a hazard ratio of 4.9 for recurrence or death.
A companion paper (Abbosh, Nature 2017) showed phylogenetic ctDNA tracking could detect relapse a median 70 days before imaging.
- 327 regions from 100 tumours; median 30% of mutations subclonal, 48% of tumours with subclonal drivers
- Elevated copy-number intratumour heterogeneity associated with recurrence or death: HR 4.9 (95% CI 1.8-13.1)
- EGFR, MET, BRAF and TP53 mutations almost always clonal; PIK3CA, NF1 and chromatin-modifier mutations often subclonal
- Whole-genome doubling occurred in most tumours and preceded much of the copy-number diversification
- Companion ctDNA paper: relapse detected a median 70 days before CT in tracked patients
Lung cancers keep evolving after they form, and it is ongoing chromosomal instability rather than the number of mutations that best predicts who will relapse. This gives a rationale for targeting the earliest (clonal) drivers and neoantigens and for tracking evolution in blood after surgery.
- Early-stage, surgically resected tumours only; the interim cohort of 100 was later expanded to 421
- Exome sequencing does not capture non-coding or structural events fully
- The prognostic value of copy-number heterogeneity needed validation in the full cohort and other cancers
- Clinical utility of clonal-neoantigen targeting remained hypothetical at the time
Relapse after surgery is driven by particular subclones that can be identified in the primary tumour and tracked in blood, which argues for evolution-aware adjuvant strategies. The pollution finding reframes carcinogenesis: some agents promote already-mutant cells rather than causing mutations.
A single biopsy is an incomplete picture of a patient's cancer. Truncal mutations shared by all cells (in kidney cancer, VHL) are the most reliable drug targets, whereas mutations in only some branches predict resistance. This is why liquid biopsy and multi-region sampling matter.