In this issue of Cancer Discovery, Fowler and colleagues conduct a thorough characterization of the dynamics of mutant clones in phenotypically normal human skin. Their results extend previous studies by showing that human skin is composed in large part of clones harboring mutations frequently observed in human cancer, while at the same time they uncover a previously unappreciated biological heterogeneity among nearby clones and across different body sites.
See related article by Fowler et al., p. 340.
In recent years, thanks to improvements in sequencing methods, a number of studies have increased our understanding of the pattern of somatic mutations in a variety of human tissues (1–5). Invariably, these studies have shown that normal tissues are composed of a patchwork of mutant clones, sometimes representing the majority of the cells and often containing mutations thought to be oncogenic, but without evidence of malignant transformation or compromised tissue function.
From a cancer-initiation perspective, these results represent a conundrum in the fact that the pervasive presence of these mutations in most individuals is hard to reconcile with the fact that cancer is a relatively rare occurrence, mostly arising in old age and from a single rogue clone out of approximately 40 trillion cells in an adult human.
In the skin, previous studies have shown that mutant clones are widespread and characterized by the presence of similar mutations to those observed in basal cell carcinoma (BCC) and squamous cell carcinoma (SCC; refs. 1, 2), the two types of cancer originating from keratinocytes, the most abundant cells in normal skin.
In this issue, Fowler and colleagues explore the mutational patterns of normal skin obtained from 35 Caucasian donors, from a variety of body regions (6). They sequenced small biopsies each containing ∼105 cells targeting 74 genes frequently mutated in keratinocyte cancers at an average coverage of 690×, allowing detection of mutation in 1% or less of cells. It is important to stress that observation of a mutation requires sufficient representation of that mutation in a tissue to be detected, which occurs through clonal expansion. So detection of a mutation reflects some combination of mutation occurrence with clonal expansion driven by either selection, drift, or hitchhiking.
The ratio of nonsynonymous to synonymous mutations (dN/dS) is an unbiased metric to define the contribution of a particular gene toward fitness (whether organismal or somatic). An excess of nonsynonymous variants implies positive selection for protein-altering mutations in a gene. By analyzing the dN/dS ratios, they found that 11 genes were positively selected. This list includes genes previously identified using the same method (1, 2). In addition, five genes were found to be under significant negative/purifying selection, including PIK3CA, which undergoes frequent activating mutations in various cancers. A more detailed examination showed that mutations in NOTCH1 and TP53 were more frequently found in domains with critical protein functions, whereas in PIK3CA, activating mutations observed in cancer were significantly more frequent than expected, further confirming their negative and positive roles in somatic fitness, respectively. A CRISPR screen validated that three of the five genes under negative selection (CUL3, PIK3CA, and DICER1) are essential in a keratinocyte cell line.
They further observed that half or more of the human skin is composed of mutant clones, with NOTCH1 mutations being present at frequencies similar to those observed in BCC and SCC, suggesting that although these mutations have a clear positive fitness effect in normal skin, they might not be further positively selected during tumor formation. These observations raise the question of whether NOTCH1 loss-of-function contributes to BCC or SCC pathology. In contrast, other mutations such as in TP53, NOTCH2, and KMT2D are enriched in the cancers relative to normal skin, indicating further selection during cancer evolution.
Different areas of the human skin have different relative risks of developing SCC and BCC; for example, the relative risk of BCC and SCC is higher in the head compared with the trunk or legs. To understand whether the pattern of mutations can explain these differences, the authors analyzed the frequencies of each mutation and the mutational signatures in tissues obtained from different body sites. They found that TP53 mutations are overrepresented and FAT1 mutations are underrepresented in the head, whereas NOTCH1 and FAT1 mutations are overrepresented in the legs. Moreover, the UV-associated G[T>C]T substitutions are overrepresented in the head and underrepresented in the trunk. Although these associations are suggestive, it remains unclear if there is a causal relationship with cancer risk and what the underlying mechanism is. Differences in cancer risk across sites may relate more to microenvironmental changes in the skin (themselves affected by exposures such as to sunlight) that alter adaptive landscapes, either favoring or disfavoring somatic evolution to SCC or BCC (Fig. 1). Further mechanistic studies will be needed to address this quandary.
Perhaps not surprisingly, most mutations observed in donor skin biopsies are associated with UV signatures, which is in agreement with a large body of literature showing that UV exposure is an important risk factor for skin cancer (7, 8). Although this association was observed in specific cases within the same sample, when comparing more sun-exposed to less sun-exposed regions of the body, this relationship was not universal and some samples showed no correlation. Interestingly, in one patient, long-term immunosuppressive treatment appears to have induced more mutations than those associated with UV damage, exemplifying the importance of other environmental factors.
To gain further spatial and molecular resolution, the authors subdivided six skin samples into smaller specimens composed of about 2,400 cells and subjected them to targeted sequencing for 324 cancer-associated genes, with additional samples subjected to whole-genome sequencing. These analyses resulted in several interesting findings: first, there is a striking heterogeneity in the number of mutations and size of clones across different donors; second, within the same sample, different clones show a remarkable variation in terms of clone size, telomere length, and mutational signature; third, although this analysis showed two more genes under positive selection (PPM1D and ASXL1), it provided evidence that PATCH1 mutations, which are frequently observed in BCC and are known to contribute to cancer maintenance (9), may represent a passenger event due to proximity to the NOTCH1 gene, facilitating subsequent contributions to the cancer phenotype.
Finally, the authors analyzed the spectrum of mutations within hair follicles, which have been proposed to be the site of origin for BCC and SCC. They subdivided the follicle into three sections, the upper, the middle (which contain the stem cells in the bulge region), and the bulb section, and separately analyzed them along with tissue from the nearby skin. This analysis showed that the number of mutations and their variant allele frequencies decrease in a top-to-bottom manner, likely due to reduced UV penetration in the lower layers of the skin. The pattern of mutations was not different from that observed in the nearby skin, while at the same time the number of mutations and the variant allele frequencies were lower in the follicle. Interestingly, the hair follicle stem cell pool appears to be polyclonal throughout life, in contrast to other niche structures such as the colonic crypt that undergo waves of clonal fixation.
This study provides several new findings that advance our understanding of cancer-initiation dynamics in the skin. At the same time, it raises new intriguing questions. In particular, although there are differences in terms of mutation burden, clone size, and mutational signature in areas with different relative cancer risk, the connection between the molecular and epidemiologic findings remains elusive. Similarly, whether the mutational/clonal burden can directly inform on the subjective risk of an individual toward cancer development remains an open question, particularly when considering that relative differences in the distribution of mutations between skin regions, with different relative cancer risks, are small, suggesting that other factors likely play a role (Fig. 1). In the hematopoietic system, it has been reported that certain mutations and their allele frequency can inform on the person's risk of developing leukemia (10)—similar relationships are not readily evident in the skin, although future analyses of mutational patterns in the skin may reveal such connections. Another curious observation is the fact that both the number of mutations and their variant allele frequencies in the skin show only a weak association with age, whereas epidemiologic data clearly show that age is an important factor for skin cancer development (11), again pointing to other unappreciated biological mechanisms that contribute to cancer risk. Additionally, analogous studies of individuals representing other races will be critical for understanding how evolved differences in skin pigmentation shape mutational dynamics in this critical barrier tissue.
One fundamental question is the extent that observed oncogenic mutation-driven clonal expansions in our skin matter. Given that much of this somatic evolution is happening in younger individuals, where natural selection has long favored our survival so as to maximize reproductive success, the impact on human fitness is doubtfully very negative and may even be positive (by maintaining somatic cell function and/or by limiting somatic evolution toward more malignant phenotypes). Given the surprising quantity of oncogenic mutation–driven clones in human skin and other tissues, we can tip our hats to the power of natural selection to frame our somatic tissues to nonetheless impede cancer evolution, such that although ∼40% of us will develop cancers, these cancers will mostly arise in our post-reproductive years and do so from a single rogue clone. It will be fundamentally important to understand not only how such a rogue clone progresses from an oncogenically initiated cell to a full-blown cancer, but also the evolved tumor-suppressive mechanisms that prevent such cancer progression for millions of initiated clones.
By uncovering fascinating new insight into mutational patterns in normal skin, Fowler and colleagues have provided a foundation upon which we can build a better understanding for how environmental factors shape mutational and microenvironmental landscapes in the skin to dictate the odds of cancer development.
No disclosures were reported.
J. DeGregori is supported by VA Merit award 1-I01-BX004495.