Background: Primary lung cancer is the leading cause of cancer-related mortality, with metastatic disease being responsible for the majority of deaths. To gain insight into the lethal process of metastasis, we report on the longitudinal evolutionary analysis of the TRACERx 421 paired primary-metastasis cohort.

Methods: 712 tumor samples, of which 485 were primary tumor and 227 were metastatic samples, from 129 metastatic non-small cell lung cancer (NSCLC) patients with detailed clinical annotation were collected from 18 UK hospital sites and whole exome sequenced. Mutations and copy number events were integrated to resolve the evolutionary history of each tumor.

Results: We observe that metastases generally diverge relatively late in molecular time, after the majority of mutations in the primary tumor have accumulated, with a substantial minority (33%) diverging prior to the last clonal sweep in the primary tumor. For this minority of cases, divergence is estimated to have occurred at tumor volumes below the limit of computed tomography (CT) detection. Our extensively sampled cohort reveals that sampling bias may result in erroneous inference of metastatic trajectories. 79% of metastases diverging after the last clonal sweep in the primary tumor would be misclassified as diverging prior to the last clonal sweep if only a single region of the primary tumor is considered.

Patterns of dissemination range from monoclonal, involving a single metastatic clone (68% - where metastatic potential is likely acquired once in the life-history of the tumor), to polyclonal and polyphyletic (17%), where metastatic potential may have arisen multiple times during lung cancer development, or at a single time point early in the development of the tumor. We find that thoracic lymph node disease resected at surgery was responsible for less than 20% of subsequent disease dissemination, suggesting that lymph nodes likely represent a hallmark of metastatic potential rather than a gateway to further metastases.

Furthermore, we observe that clones which seed the metastases are generally dominant within the primary, reflecting positive selection and acquisition of subclonal mutations in specific cancer genes (e.g. RB1, PIK3CA). In squamous cell carcinomas, we find that non-metastatic primary tumors show no significant evidence of positive subclonal selection. We find that 35% of metastases harbor driver mutations not identified in the primary tumor and identify somatic copy number alterations that are enriched in metastases (e.g. CCND1 gains in lung adenocarcinoma).

Conclusions: These data highlight the potential to apply evolutionary measures to primary tumors to predict metastatic risk, the limitations to current screening approaches particularly for early tumor divergence, and the importance of future precision adjuvant therapies to target disseminated micro-metastatic clones.

Citation Format: Ariana Huebner, Maise Al Bakir, Carlos Martinez Ruiz, Kristiana Grigoriadis, Thomas B. Watkins, Oriol Pich, David A. Moore, Selvaraju Veeriah, Sophia Ward, Joanne Laycock, Diana Johnson, Andrew Rowan, Maryam Razaq, Mita Akther, Cristina Naceur-Lombardelli, Sonya Hessey, Michelle Dietzen, Emma Colliver, Alexander M. Frankell, Emilia Lim, Takahiro Karasaki, Christopher Abbosh, Crispin T. Hiley, Mark S. Hill, Daniel Cook, Gareth Wilson, TRACERx consortium, Allan Hackshaw, Nicolai J. Birkbak, Simone Zaccaria, Mariam Jamal-Hanjani, Charles Swanton, Nicholas McGranahan. TRACERx: Mapping the evolution of metastases in non-small cell lung cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 3792.