Cancer Evolution Is Associated with Pervasive Positive Selection on Globally Expressed Genes

Abstract

Cancer is an evolutionary process in which cells acquire new transformative, proliferative and metastatic capabilities. A full understanding of cancer requires learning the dynamics of the cancer evolutionary process. We present here a large-scale analysis of the dynamics of this evolutionary process within tumors, with a focus on breast cancer. We show that the cancer evolutionary process differs greatly from organismal (germline) evolution. Organismal evolution is dominated by purifying selection (that removes mutations that are harmful to fitness). In contrast, in the cancer evolutionary process the dominance of purifying selection is much reduced, allowing for a much easier detection of the signals of positive selection (adaptation). We further show that, as a group, genes that are globally expressed across human tissues show a very strong signal of positive selection within tumors. Indeed, known cancer genes are enriched for global expression patterns. Yet, positive selection is prevalent even on globally expressed genes that have not yet been associated with cancer, suggesting that globally expressed genes are enriched for yet undiscovered cancer related functions. We find that the increased positive selection on globally expressed genes within tumors is not due to their expression in the tissue relevant to the cancer. Rather, such increased adaptation is likely due to globally expressed genes being enriched in important housekeeping and essential functions. Thus, our results suggest that tumor adaptation is most often mediated through somatic changes to those genes that are important for the most basic cellular functions. Together, our analysis reveals the uniqueness of the cancer evolutionary process and the particular importance of globally expressed genes in driving cancer initiation and progression.

Extremely high genetic diversity in a single tumor points to prevalence of non-Darwinian cell evolution

“…The non-Darwinian prediction is consistent with the high K a /K s ratios (nonsynonymous/synonymous SNVs per site) observed in 400 cancer genomes (29) and in The Cancer Genome Atlas (TCGA) data (35). The ratio is statistically indistinguishable from 1 in most studies (36), thus indicating ineffective selection against protein sequence changes in tumors. Cases of K a /K s ∼ 1 are rarely seen in nature; for example, K a /K s < 0.3 between humans and other primates (1).…”

The reverse evolution from multicellularity to unicellularity during carcinogenesis

“…4b). Using this new method, we analysed data from The Cancer Genomic Atlas (TCGA) and successfully identified as many as 134 novel putative tumour suppressors under a false discovery rate of 0.1 (Supplementary Table 3), a finding reminiscent of a recent study 23 and challenging the view that the growth of the cancer gene list has reached a plateau 17 . It is conceivable that most of the newly identified tumour suppressors are minor cancer drivers, conferring small fitness advantages in primary tumours.…”

Mutation-Profile-Based Methods for Understanding Selection Forces in Cancer Somatic Mutations: A Comparative Analysis

“…Analyses of large-scale cancer somatic mutation data have revealed that the effects of positive selection are much stronger on cancer cells than on germline cells [16,17]. Given that many of the positively selected genes in tumor development act as the driving force behind tumor initiation and development and are thus considered "driver genes", it is understandable that almost all previous studies have focused on positively selected genes in cancer genomes [3,[18][19][20][21].…”

“…Many previous studies have used the ratio of nonsynonymous to synonymous substitution rates to identify genes that might be under strong positive selection both in organismal evolution and carcinogenesis [11,16,17,[23][24][25][26]. However, most of these studies applied conventional methods, which are usually based on simple nucleotide mutation/substitution models, e.g., the simplest equal-rate model assuming that every mutation or substitution pattern has the same probability [27].…”

“…However, most of these studies applied conventional methods, which are usually based on simple nucleotide mutation/substitution models, e.g., the simplest equal-rate model assuming that every mutation or substitution pattern has the same probability [27]. Unfortunately, this may not be a realistic biological model because many recent cancer genomics studies have shown that mutation profiles vary greatly between different cancer samples [17,28]. In addition, context-dependent mutation bias (i.e., base-substitution profiles that are influenced by the flanking 5′ and 3′ bases of each mutated base) should also be considered [28,29].…”