Cross-Cancer Analysis Reveals Novel Pleiotropic Associations—Response

Genomic research based on mouse models has led to major advances in our understanding of disease biology in humans, and it is one of the most widely used mammalian model organisms, due in part to their genetic tractability and the high number of human orthologous genes partitioned into regions of conserved synteny (1, 2). We are pleased to see that our cross-cancer genome-wide analysis based on data of five cancers in human (3) has inspired further investigation using existing experimental data based on mouse models (4). Using recombinant congenic strains and microsatellite markers, Quan and colleagues investigated lung and colon cancer susceptibility loci in the mouse genome. They identified significant colocalization of lung and colon cancer susceptibility loci in 27 gene clusters, defined as loci mapped within 10 centiMorgan (cM), while our study did not identify pleiotropic loci for lung and colorectal cancers. We found the work of Quan and colleagues very valuable, and the fact that some of the lung–colon gene clusters identified in Quan and colleagues fall in human–mouse syntenic regions is intriguing, as it may suggest that there are more pleiotropy regions related to carcinogenesis in mammals to be identified. The differences between our results may be due to several fundamental differences in our respective approaches, and the organismal differences between humans and mice such as tissue-specific gene regulatory networks, which we outline in the following paragraphs.Our previous cross-cancer analysis was based on a two-stage genome-wide association approach, analyzing approximately 9.9 million common germline sequence variants across human genome for the risk of lung, ovary, breast, prostate, and colorectal cancer based on 61,851 cancer patients and 61,820 controls in the discovery set. Those that showed pleiotropic signals were then further replicated in independent studies of 55,789 patients and 330,490 controls (3). In this analysis, we aimed to detect biological pleiotropic effects; therefore, we were primarily focused on signals from the same genomic regions (within the same gene, or the same region of high linkage disequilibrium in intergenic regions; ref. 5). This is different from the definition of clusters used by Quan and colleagues, which considered the overlapping regions of lung and colon susceptibility loci within 10 centiMorgan (10 cM). Although there is no fixed ratio between cM and basepairs, 1 cM on average would correspond to crudely 1 to 1.2 million basepairs (Mb) in human genome, with the ratio varying by sex, chromosome position, and other factors (6, 7). The majority of the susceptibility loci that were investigated in our analysis were less than 3 Mb in sizes, with the exception of the MHC region. Therefore, colocalization at 10 Mb would not be detected in our analysis.Moreover, even though mouse and human models shared similarities in the control networks of gene activities, the two species in different mammalian orders have significantly diverged at the sequence level, and only approximately 40% of the human nucleotides can be mapped to the mouse genome (1, 2). In addition, the expression profiles of many mouse genes are different from their human orthologs, and similar genes may be engaged in different biological pathways in two different species (2). Transcription factor binding between human and mouse is rapidly evolving (8) and that transcriptional remodeling of regulatory regions themselves is pervasive (9), which in principle could alter the tissue-specific expression of susceptibility loci. These biological differences between mouse and human may explain at least part of differences in our findings.Finally, given the large amount of statistical testing one needed to perform for the common sequence variation, our interpretation of the statistical significance was penalized by the multiple comparisons. It is possible that there are colon–lung cancer pleiotropy loci that were not identified in our analysis given the stringent statistical threshold applied. For example, we did observe several loci with nominal associations with both lung and colorectal cancers in the GAME-ON/GECCO discovery set, such as variants in 6p21.33 and 13q13.1, but not in the replication stage, which could be due to reduced power of the colorectal cancer study in the latter. This was acknowledged in our discussion, and the full list of loci with potential pleiotropic effects was shown in Supplementary Table S1 to facilitate further investigations.Overall, the mouse model and the genome-wide association approaches are highly complementary, and each has its own strengths and limitations. Mouse models are essential for studying human cancers, but it is clear from the considerable efforts being taken to humanize mice, the development of patient-derived xenografts, as well as the need for different types of animal models in cancer research, that not all aspects of human cancer susceptibility and biology can be recapitulated in mice (10, 11). Combining both approaches could potentially provide greater insights into cancer etiology and tumorigenesis mechanism.See the original Letter to the Editor, p. 6042No potential conflicts of interest were disclosed.We thank Prof. James R. Woodgett for the insightful comments and suggestions.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.