Genes, environment, and “bad luck”

We note with interest the Perspective by M. Nowak and B. Waclaw on "Gene, environment, and 'bad luck'" (24 March, p.1266-7), and we also appreciate their support of our conclusion that 'one cannot use the relationship (between stem cell division and cancer risk) to differentiate between contributions from replication, hereditary, and environmental factors to cancer risk' [1]. This is a fundamental conclusion [2] that unfortunately was still partly ignored in the recent study by Tomasetti et al [3]. Moreover, and contrary to the assertion by Nowak and Waclaw that Tomasetti et al. "never claimed such a possibility" in their 2015 publication [4], these authors clearly articulated that "our (their) analysis shows that stochastic effects associated with DNA replication [intrinsic factors] contribute in a substantial way to human cancer incidence" [4].

Nowak and Waclaw fault our analysis of the contribution of intrinsic vs extrinsic factors, and they cite two specific problems:

1) That we assumed that if two cancers have the same D (stem cell divisions) value, then the variation in R (an unfortunate use of terms with R here indicating tissue-specific life time risk) is only caused by extrinsic risk factors (we will avoid this use of R here to obviate confusion with the R of Tomasetti et al).
We believe that the source of confusion is the imprecise definition and use of key terminology, especially that for intrinsic risk. This intrinsic factor effect, defined as "R" by Tomasseti et al [3] and also clearly defined (by us [2] and by Tomasetti et al [4]) as the basal errors in DNA mutations that occur during normal cell division, universally estimated at around three mutations/cell division. The use of this specific definition is critical for two reasons. First, it is well quantitated. Second, it is the predominant effect in dictating unmodifiable cancer risk. As such, all other effects, including the ones mentioned by Nowak and Waclaw as affecting R (such as immune surveillance or mutations), are themselves non-intrinsic factors (non-R), and therefore potentially modifiable.
Notably, Nowak and Waclaw reach the conclusion that "very different factors lead to variation in tissue lifetime risk of cancer given the same D". This was precisely our conclusion. However, Nowak and Waclaw then claim that we restricted these factors to only environmental or hereditary. This is simply not correct. We allowed for all non-R factors including immune, ROS, viral infection.
Importantly, our estimation of the intrinsic risk from the correlation analysis is based on one biologic fact: namely, the intrinsic mutation rate (R) is a measurable entity (approximately 3 mutations/cell division), and this appears to be constant in various tissues. Thus, tissues with the same R and with similar level of D (small intestine, liver, pancreas, and colon) should have similar effect of R on cancer risk; however, this contradicts real data showing that small intestine with a rapidly dividing epithelium shows very little cancer risk (a range of 100-1000 fold less cancer risk for small intestine) in spite of having a similar R as other gastrointestinal tissues. Clearly, this indicates the operation of other (i.e. non-R) factors in driving hepatic (e.g. cirrhosis or HVC), pancreatic, and colorectal cancers.
Additionally, Nowak and Waclaw disregard the fact that extrinsic risks estimated with our approach of "intrinsic risk line" are consistent with those obtained from many large-scale epidemiological studies. This consistency validates our approach.

2). That our model ignores clonal expansion.
This is not accurate. We realized the issue of clonal expansion and addressed it in the second part of our model that assumes every tissue cell to be a stem cell, which is equivalent to clonal expansion to the tissue size at the very early stage of tissue development. This is an extreme case of clonal expansion, and even with that we found that the theoretical intrinsic risks are still quite low. It should also be noted that we did not consider other important factors that may actually reduce cancer incidence: namely, cell death, oncogene-induced senescence, and cell turnover. We 'allowed' cells with one mutation to 'live forever' waiting for subsequent mutations. Including this effect may actually further lower the theoretical intrinsic cancer risk. Also, Nowak and Waclaw disregard the fact that we did incorporate the variation in number of genes required to initiate cancer and we did incorporate whether these effects were in oncogenes or tumor suppressor genes.
In contrast, Nowak and Waclaw's model is limited at two levels. First, their analysis is restricted to one- and two-hit models, whereas most investigators, including Tomasetti et al argue for at least 3 for the majority of cancers. Second, cancer risk from their model prediction would shoot above one for large D or D^2/N values in their figure, although this section of the figure is not shown. This indicates their model parameters may be wrongly set up, e.g. an unrealistically high mutation rate μ that inflates the estimation of intrinsic risks in their model.

In conclusion, while Nowak and Waclaw have raised some interesting hypotheses, some definitions are being used loosely, and that may account for a lot of the existing 'debate' and confusion. For example, Nowak and Waclaw attribute to Tomasetti et al the conclusion that 'as much as 66% of driver mutations are due to replication". This is not what Tomasetti el al concluded. They concluded that as much as 66% of mutations are due only to R factor effects. Many non-R factors contribute to mutagenesis in addition to the intrinsic R.

Reference:
1. Nowak, M.A. and B. Waclaw, Genes, environment, and "bad luck". Science, 2017. 355(6331): p. 1266-1267.
2. Wu, S., et al., Substantial contribution of extrinsic risk factors to cancer development. Nature, 2016. 529(7584): p. 43-7.
3. Tomasetti, C., L. Li, and B. Vogelstein, Stem cell divisions, somatic mutations, cancer etiology, and cancer prevention. Science, 2017. 355(6331): p. 1330-1334.
4. Tomasetti, C. and B. Vogelstein, Variation in cancer risk among tissues can be explained by the number of stem cell divisions. Science, 2015. 347(6217): p. 78-81.