NEW YORK — Using sequencing data, researchers characterized intra-tumor heterogeneity of more than three dozen cancer types, finding subclonal populations among most of the cancer types analyzed.
Tumor cells accumulate somatic mutations, some of which provide advantages and lead those cells to expand. Characterizing intra-tumor heterogeneity, though, has been challenging due to the need to disentangle signals from various cancer cells and even nearby non-cancer cells in bulk tumor samples in combination with the sheer number of tumor samples that need to be characterized.
Researchers from the Pan-Cancer Analysis of Whole Genomes (PCAWG) initiative developed a new strategy to call copy-number and other mutations from tumor sequencing data to gauge intra-tumor heterogeneity, which can hinder targeted therapies aimed at specific genetic alterations. After applying this approach to more than 2,600 cancer samples, the researchers found evidence of intra-tumor heterogeneity in most tumors and noted particular patterns common among certain cancer types, as they reported on Wednesday in the journal Cell.
"This study was the first of its kind and allowed us to look at intra-tumor heterogeneity along the whole genome of many cancer types in unprecedented detail," senior author Peter Van Loo and postdoc Maxime Tarabichi, both at the Francis Crick Institute in the UK, wrote in an email.
The researchers developed a consensus-based approach to reconstruct intra-tumor heterogeneity. After sequencing the bulk cell population, they used four different methods to call structural variants and five methods to call SNVs and then reconstructed the subclonal populations in silico and, again, used multiple methods to gauge copy number and subclonal structure. They applied this approach to 2,658 tumor samples from 38 different cancer types from the PCAWG initiative.
Nearly all — 95.1 percent of the 1,705 samples with enough resolution for analysis — contained one or more distinct subclones, the researchers reported.
Subclones from the same tumors mostly exhibited sibling relationships marked by the development of branching lineages, the researchers found. They estimated that two subclones are about 3.11-fold more likely to have a sibling relationship than a parent-child relationship.
There was also no correlation between mutational burden and the proportion of linear or branching subclones, which suggested to the researchers that branching evolution is common across tumor types, including among those with low mutation rates.
Some subclonal mutations cropped up repeatedly within tumors of the same type. For instance, SETD2 was frequently mutated at subclonal levels in clear cell renal cell carcinoma, while driver mutations in MEN1 were present in a portion of pancreatic neuroendocrine tumors and CDKN2A in some pancreatic adenocarcinomas. Similarly, structural variants appear to be driver mutations in about a third of ovarian adenocarcinomas and about 40 percent of leiomyosarcomas.
Intra-tumor heterogeneity, Van Loo and Tarabichi noted, is a major clinical challenge. For targeted treatments to have their intended effect of killing all cancer cells, the mutations being targeted have to be present in all cells. "However, our study shows that a significant fraction of clinical targets are in fact only present in a subset of the cancer cells, and that fraction is likely an underestimate," they said.
Being able to better predict which mutations are more likely to be present in all cells could help identify more suitable treatment strategies and targets, they added.
Going forward, Van Loo and Tarabichi said they and the wider community are now turning to new approaches and improved tools to examine intra-tumor heterogeneity with increased granularity. In addition, they are folding in samples from other cancer types not included in this dataset.