Summary The survival of residual tumor cells following therapy and their eventual recurrence constitutes one of the biggest obstacles to obtaining cures in breast cancer, but it remains unclear how the clonal composition of tumors changes during tumor relapse. We used cellular barcoding to directly monitor clonal dynamics during tumor recurrence in a genetically engineered mouse model. We found that the clonal diversity of tumors progressively decreased during tumor regression, residual disease, and recurrence. Only a fraction of subclones survived oncogene withdrawal and persisted in residual tumors. The minimal residual disease phase itself was accompanied by a continued attrition of clones, suggesting an ongoing process of selection during dormancy. The reactivation of dormant residual cells into recurrent tumors followed several distinct evolutionary routes. Approximately half of the recurrent tumors exhibited a striking clonal dominance in which one or two subclones comprised the vast majority of the tumor. The majority of these clonal recurrent tumors exhibited evidence of de novo acquisition of Met amplification, and were sensitive to small-molecule Met inhibitors. A second group of recurrent tumors exhibited marked polyclonality, with thousands of subclones and a clonal architecture very similar to primary tumors. These polyclonal recurrent tumors were not sensitive to Met inhibitors, but were instead dependent upon an autocrine IL-6 – Stat3 pathway. These results suggest that the survival and reactivation of dormant tumors can proceed via multiple independent routes, producing recurrent tumors with distinct clonal composition, genetic alterations, and drug sensitivities.

Adaptation and Selection Shape Clonal Evolution During Residual Disease and Recurrence