This is a tale of serendipity and stubbornness. It all began in 2014, a time when we were younger, had more hair, and did not panic whenever we heard someone cough in the supermarket line. Our laboratory at the Evolutionary Biology Institute recently joined the Pan-Cancer Analysis of Whole Genomes consortium. I had just started my PhD, and this was the very first time our lab worked with cancer data. So, honoring the old adage of “when all you have is a hammer, everything looks like a nail”, we decided that my PhD project would focus on analyzing human cancer from the point of view of evolution.  And, like all things in the big data era, we sought to do so by crunching large amounts of data against each other: cancer samples, and human and non-human great apes (NHGA) population data.

We planned to analyze this from three points of view: presence of cancer driver mutations in chimpanzee and gorilla samples, high-impact nonsynonymous mutations in cancer predisposition genes, and inter-species hotspots of mutations. Long story short, there are almost no cancer driver mutations in great apes (and the few there are, are probably false positives (1)), and the only two (BRCA1&2) cancer predisposition genes accumulating large numbers of nonsynonymous mutations in chimpanzee and gorilla were already described (2). Up until this point, all was disappointment, as the potentially most sexy results were not so sexy after all and, moreover, they were already taken.

So, we moved on to analyze the presence of mutation hotspots in the different datasets. Having no prior experience with cancer whole-genome data, we naively expected to find patterns of mutation mostly similar between human and non-human great apes, with a few mutation hotspots in cancer here and there. We also hoped that some of the cancer hotspots (different from human) could be explained by the reactivation of ancestral mutation hotspots present in chimpanzee and gorilla. But, oh boy, we were in for a treat!

What we found is that the distribution of mutations across the genome was only moderately correlated between humans and non-human great apes, and much weaker than the correlation between chimpanzee and gorilla populations. When looking at cancer, the human germline mutation distribution was very poorly correlated with cancer (we didn’t know that at the time, but this had already been described before (3),(4)), but, surprisingly, the correlation between cancer and the germlines of chimpanzee and gorilla was much, much higher, almost as strong as the correlation between human and non-human great ape germlines! 

So, we did what any honest, hard-headed scientist would do: we did not believe our results. And then, we spent a whole year relentlessly redoing all the analyses, using new datasets, and testing all possible errors we could imagine. Whatever we tried, our observations stood there, unmoved. We finally had to surrender our efforts and accept the unbearable truth: the results were real; the mutation distribution of tumors looked more like chimpanzees than humans.

 What is the cause behind this? Both tumors and non-human great apes accumulate more mutations in closed chromatin regions, while the human germline accumulates mutations in GC-rich open chromatin regions. We also observed that the species’ diversity played a role in this cancer-chimpanzee similitude. However, CpG>T transitions and mutations at non-CpG sites presented different dynamics: Population bottlenecks cause a loss of diversity in all sites of the genome. However, CpG sites, having a 10-fold higher mutation rate, recover their apparent diversity much faster when the population grows again. This causes a decoupling of the mutation load in CpG vs non-CpG sites, affecting more GC-rich regions (which are also enriched in open chromatin, exons, promoters, enhancers, … functional things, yay!). These results suggest that (ancient) human demography played a big role in the poor human-tumor correlation. 

We also wondered if differences in the mutation mechanisms acting across the genome of the species could contribute to these human-chimpanzee discrepancies. However, we could detect none. This suggests that the tumor-chimpanzee similitude is driven, not by specific mutation mechanisms, but by the accumulation of mutations dynamics present in normal human cells.

So, for once, it seems that cancer is not the bad guy in this story (ok, it still is). Instead, humans are the odd ones this time. Cancers, like chimpanzees and gorillas, just show the full landscape of mutation of a normal human cell. It is us, humans, with our turbulent distant past, the ones showing a skewed landscape. Our study highlights the relevance of the study of non-human great apes and their conservation for human health, representing a mirror in which to reflect the findings detected in human.

References:

1- Predicting the clinical impact of human mutation with deep neural networks. Sundaram, Nature Genetics 2018. https://doi.org/10.1038/s41588...

2- Rapid evolution of BRCA1 and BRCA2 in humans and other primates. Lou, BMC Evolutionary Biology 2014. https://doi.org/10.1186/1471-2...

3- The large-scale distribution of somatic mutations in cancer genomes. Hodgkinson, Human Mutation 2012. https://doi.org/10.1002/humu.2...

4- Chromatin organization is a major influence on regional mutation rates in human cancer cells. Schuster-Böckler, Nature 2012. https://doi.org/10.1038/nature...