Substantial contribution of extrinsic risk factors to cancer development

Journal name:
Nature
Volume:
529,
Pages:
43–47
Date published:
DOI:
doi:10.1038/nature16166
Received
Accepted
Published online

Abstract

Recent research has highlighted a strong correlation between tissue-specific cancer risk and the lifetime number of tissue-specific stem-cell divisions. Whether such correlation implies a high unavoidable intrinsic cancer risk has become a key public health debate with the dissemination of the ‘bad luck’ hypothesis. Here we provide evidence that intrinsic risk factors contribute only modestly (less than ~10–30% of lifetime risk) to cancer development. First, we demonstrate that the correlation between stem-cell division and cancer risk does not distinguish between the effects of intrinsic and extrinsic factors. We then show that intrinsic risk is better estimated by the lower bound risk controlling for total stem-cell divisions. Finally, we show that the rates of endogenous mutation accumulation by intrinsic processes are not sufficient to account for the observed cancer risks. Collectively, we conclude that cancer risk is heavily influenced by extrinsic factors. These results are important for strategizing cancer prevention, research and public health.

At a glance

Figures

  1. Schematic showing how intrinsic processes and extrinsic factors relate to cancer risks through stem-cell division.
    Figure 1: Schematic showing how intrinsic processes and extrinsic factors relate to cancer risks through stem-cell division.

    This hypothesis maintains the strong role of stem-cell division in imparting cancer risk, but it also illustrates the potential contributions of both intrinsic and extrinsic factors operating through stem-cell division. Other effects, for example, through division of non-stem cells, are considered later in this analysis.

  2. Correlation analysis of stem-cell division and cancer risk does not distinguish contribution of extrinsic versus intrinsic factors to cancer risk.
    Figure 2: Correlation analysis of stem-cell division and cancer risk does not distinguish contribution of extrinsic versus intrinsic factors to cancer risk.

    The black dots are data from figure 1(also shown in supplementary table 1) of Tomasetti & Vogelstein5, and the black line shows their original regression line. The blue diamonds represent the hypothesized quadrupled cancer risks due to hypothetical exposure to an extrinsic factor such as radiation. The blue regression line for the hypothetical risk data maintains the same correlation as the original black line, albeit reflecting a much higher contribution of extrinsic factors to cancer risk.

  3. Estimation of the proportion of lifetime cancer risk that is not due entirely to ‘bad luck’.
    Figure 3: Estimation of the proportion of lifetime cancer risk that is not due entirely to ‘bad luck’.

    a, b, Estimations based on total tissue stem-cell divisions originally reported in Tomasetti & Vogelstein5 (a) and total tissue cell divisions (b). Red dots are cancers used to compute the ‘intrinsic’ risk linear regression lines (red dashed lines). Blue dots are cancers known to have substantial extrinsic risks from epidemiology studies. The numbers in parentheses are the estimated percentages of cancer risks that are due to factors other than intrinsic risks.

  4. Theoretical lifetime intrinsic risks (tLIR) for cancers based on different number of hits (k) required for cancer onset.
    Figure 4: Theoretical lifetime intrinsic risks (tLIR) for cancers based on different number of hits (k) required for cancer onset.

    a, b, The green (a) and blue (b) dashed lines are the ‘intrinsic’ risk lines estimated on the basis of total reported stem-cell numbers and total homeostatic tissue cells, respectively. The intrinsic stem-cell mutation rate (r) is assumed to be 1 × 10−8 per cell division. The red dashed lines are the ‘intrinsic’ risk lines estimated on the basis of the observed data using the same mechanism as Fig. 3a. Adjusted (adj.) basal and adjusted melanoma represent cancer risks after adjusting for the effect of sun exposure and UV radiation. AML, acute myeloid leukaemia.

  5. Examples of increased cancer incidence trends from 1973–2012 in SEER data.
    Extended Data Fig. 1: Examples of increased cancer incidence trends from 1973–2012 in SEER data.

    The cancer types include melanoma, thyroid cancer, kidney cancer, liver cancer, small intestine cancer, testicular cancer, non-Hodgkin lymphoma (NHL), anal and anorectal cancer and thymus cancer. The horizontal dashed lines indicate the historical minimal incidence. The vertical solid lines indicate the most recent year. The numbers represent the minimal percentage of extrinsic risk. The cervix uteri cancer, gallbladder cancer and oesophageal cancer are examples with declining or consistent incidence trend. The incidence rate is per 100,000 people.

  6. Sensitivity analysis of different mutation rates on tLIR when the number of hits (k) required is 3.
    Extended Data Fig. 2: Sensitivity analysis of different mutation rates on tLIR when the number of hits (k) required is 3.

    a, b, Theoretical intrinsic lifetime risks (tLIR) for cancers have been calculated based on five different mutation rates: r = 1 × 10−10, 1 × 10−9, 1 × 10−8, 1 × 10−7, 1 × 10−6. The red dashed lines are the ‘intrinsic’ risk lines based on the observed data following the same estimation mechanism as the intrinsic risk line in Fig. 3a. The green (a) and blue (b) dashed lines are the ‘intrinsic’ risk lines estimated based on total reported stem-cell numbers and total homeostatic tissue cells, respectively.

  7. Sensitivity analysis of different mutation rates on tLIR when the number of hits (k) required is 4.
    Extended Data Fig. 3: Sensitivity analysis of different mutation rates on tLIR when the number of hits (k) required is 4.

    a, b, Theoretical intrinsic lifetime risks (tLIR) for cancers have been calculated based on five different mutation rates: r = 1 × 10−10, 1 × 10−9, 1 × 10−8, 1 × 10−7, 1 × 10−6. The red dashed lines are the ‘intrinsic’ risk lines based on the observed data following the same estimation mechanism as the intrinsic risk line in Fig. 3a. The green (a) and blue (b) dashed lines are the ‘intrinsic’ risk lines estimated based on total reported stem-cell numbers and total homeostatic tissue cells, respectively.

  8. Intrinsic cancer risk modelling.
    Extended Data Fig. 4: Intrinsic cancer risk modelling.

    Part 1 of 2: propagation diagram of driver gene mutation states between generations in one stem cell, from which the stem-cell mutation transition probabilities from one generation to the next are computed.

  9. Intrinsic cancer risk modelling.
    Extended Data Fig. 5: Intrinsic cancer risk modelling.

    Part 2 of 2: schema of stem-cell divisions and driver gene mutations, from which the theoretical lifetime intrinsic risks (tLIR) for cancer due to k driver gene mutations are computed. Each coloured circle represents the mutation of a new driver gene in the given stem cell (yellow, first mutation; green, second mutation; red, third mutation). If the mutation of 3 designated driver genes would induce a cancerous stem cell (k = 3), then this diagram shows a cancer occurrence as the second stem cell in the last generation (generation n) that has accumulated all 3 driver gene mutations.

Tables

  1. Robustness analysis on total stem-cell divisions and cell divisions estimates in Fig. 3
    Extended Data Table 1: Robustness analysis on total stem-cell divisions and cell divisions estimates in Fig. 3
  2. Epidemiological studies on the extrinsic risks of various cancers
    Extended Data Table 2: Epidemiological studies on the extrinsic risks of various cancers
  3. Percentages of intrinsic versus extrinsic MS with known and unknown causes in different cancer types
    Extended Data Table 3: Percentages of intrinsic versus extrinsic MS with known and unknown causes in different cancer types
  4. Percentages of extrinsic risks based on the reported stem-cell estimates and total homeostatic tissue cells, as shown in Fig. 4
    Extended Data Table 4: Percentages of extrinsic risks based on the reported stem-cell estimates and total homeostatic tissue cells, as shown in Fig. 4

References

  1. Sell, S. Stem cell origin of cancer and differentiation therapy. Crit. Rev. Oncol. Hematol. 51, 128 (2004)
  2. Reya, T., Morrison, S. J., Clarke, M. F. & Weissman, I. L. Stem cells, cancer, and cancer stem cells. Nature 414, 105111 (2001)
  3. Cairns, J. Mutation selection and the natural history of cancer. Nature 255, 197200 (1975)
  4. Visvader, J. E. Cells of origin in cancer. Nature 469, 314322 (2011)
  5. Tomasetti, C. & Vogelstein, B. Variation in cancer risk among tissues can be explained by the number of stem cell divisions. Science 347, 7881 (2015)
  6. Ashford, N. A. et al. Cancer risk: role of environment. Science 347, 727 (2015)
  7. Wild, C. et al. Cancer risk: role of chance overstated. Science 347, 728 (2015)
  8. Potter, J. D. & Prentice, R. L. Cancer risk: tumors excluded. Science 347, 727 (2015)
  9. Gotay, C., Dummer, T. & Spinelli, J. Cancer risk: prevention is crucial. Science 347, 728 (2015)
  10. Song, M. & Giovannucci, E. L. Cancer risk: many factors contribute. Science 347, 728729 (2015)
  11. O’Callaghan, M. Cancer risk: accuracy of literature. Science 347, 729 (2015)
  12. Tomasetti, C. & Vogelstein, B. Cancer risk: accuracy of literatureresponse. Science 347, 729731 (2015)
  13. Altenberg, L. Statistical problems in a paper on variation in cancer risk among tissues, and new discoveries. Preprint at http://arxiv.org/abs/1501.04605 (2015)
  14. Torre, L. A. et al. Global cancer statistics, 2012. CA Cancer J. Clin. 65, 87108 (2015)
  15. Gray, J., Evans, N., Taylor, B., Rizzo, J. & Walker, M. State of the evidence: the connection between breast cancer and the environment. Int. J. Occup. Environ. Health 15, 4378 (2009)
  16. Shimizu, H. et al. Cancers of the prostate and breast among Japanese and white immigrants in Los Angeles County. Br. J. Cancer 63, 963966 (1991)
  17. Haggar, F. A. & Boushey, R. P. Colorectal cancer epidemiology: incidence, mortality, survival, and risk factors. Clin. Colon Rectal Surg. 22, 191197 (2009)
  18. Johnson, I. T. & Lund, E. K. Review article: nutrition, obesity and colorectal cancer. Aliment. Pharmacol. Ther. 26, 161181 (2007)
  19. Parkin, D. M., Mesher, D. & Sasieni, P. 13. Cancers attributable to solar (ultraviolet) radiation exposure in the UK in 2010. Br. J. Cancer 105 (Suppl 2), S66S69 (2011)
  20. Koh, H. K., Geller, A. C., Miller, D. R., Grossbart, T. A. & Lew, R. A. Prevention and early detection strategies for melanoma and skin cancer. Current status. Arch. Dermatol. 132, 436443 (1996)
  21. Blot, W. J. et al. Smoking and drinking in relation to oral and pharyngeal cancer. Cancer Res. 48, 32823287 (1988)
  22. Kamangar, F., Chow, W.-H., Abnet, C. & Dawsey, S. Environmental causes of esophageal cancer. Gastroenterology Clin. North Am. 38, 2757 (2009)
  23. Bosch, F. X. et al. Prevalence of human papillomavirus in cervical cancer: a worldwide perspective. J. Natl. Cancer Inst. 87, 796802 (1995)
  24. Frisch, M. et al. Sexually transmitted infection as a cause of anal cancer. N. Engl. J. Med. 337, 13501358 (1997)
  25. Chaturvedi, A. K. et al. Human papillomavirus and rising oropharyngeal cancer incidence in the United States. J. Clin. Oncol. 29, 42944301 (2011)
  26. El-Serag, H. B. Epidemiology of viral hepatitis and hepatocellular carcinoma. Gastroenterology 142, 12641273 (2012)
  27. Helicobacter and Cancer Collaborative Group. Gastric cancer and Helicobacter pylori: a combined analysis of 12 case control studies nested within prospective cohorts. Gut 49, 347353 (2001)
  28. Surveillance, Epidemiology, and End Results (SEER) Program. SEER*Stat Database: Incidence—SEER 9 Regs Research Data, Nov 2014 Sub (1973–2012) (National Cancer Institute, DCCPS, Surveillance Research Program, Surveillance Systems Branch, 2015)
  29. Dela Cruz, C. S., Tanoue, L. T. & Matthay, R. A. Lung cancer: epidemiology, etiology, and prevention. Clin. Chest Med. 32, 605644 (2011)
  30. Alexandrov, L. B. & Stratton, M. R. Mutational signatures: the patterns of somatic mutations hidden in cancer genomes. Curr. Opin. Genet. Dev. 24, 5260 (2014)
  31. Alexandrov, L. B. et al. Signatures of mutational processes in human cancer. Nature 500, 415421 (2013)
  32. Jones, S. et al. Comparative lesion sequencing provides insights into tumor evolution. Proc. Natl Acad. Sci. USA 105, 42834288 (2008)
  33. Tomasetti, C., Vogelstein, B. & Parmigiani, G. Half or more of the somatic mutations in cancers of self-renewing tissues originate prior to tumor initiation. Proc. Natl Acad. Sci. USA 110, 19992004 (2013)
  34. Bozic, I. & Nowak, M. A. Unwanted evolution. Science 342, 938939 (2013)
  35. Tomasetti, C., Marchionni, L., Nowak, M. A., Parmigiani, G. & Vogelstein, B. Only three driver gene mutations are required for the development of lung and colorectal cancers. Proc. Natl Acad. Sci. USA 112, 118123 (2015)

Download references

Author information

Affiliations

  1. Department of Applied Mathematics and Statistics, Stony Brook University, Stony Brook, New York 11794, USA

    • Song Wu,
    • Scott Powers &
    • Wei Zhu
  2. Stony Brook Cancer Center, Stony Brook University, Health Sciences Center, Stony Brook, New York 11794, USA

    • Song Wu,
    • Scott Powers,
    • Wei Zhu &
    • Yusuf A. Hannun
  3. Department of Pathology, Stony Brook University, Health Sciences Center, Stony Brook, New York 11794, USA

    • Scott Powers &
    • Yusuf A. Hannun
  4. Department of Medicine, Stony Brook University, Health Sciences Center, Stony Brook, New York 11794, USA

    • Yusuf A. Hannun
  5. Department of Biochemistry, Stony Brook University, Health Sciences Center, Stony Brook, New York 11794, USA

    • Yusuf A. Hannun

Contributions

Y.A.H. formulated the hypothesis. S.W. and Y.A.H. designed the research. S.W. and W.Z. performed mathematical and statistical analysis. S.W., S.P., W.Z. and Y.A.H. performed research. S.W., S.P., W.Z. and Y.A.H. wrote the paper.

Competing financial interests

The authors declare no competing financial interests.

Corresponding authors

Correspondence to:

Author details

Extended data figures and tables

Extended Data Figures

  1. Extended Data Figure 1: Examples of increased cancer incidence trends from 1973–2012 in SEER data. (224 KB)

    The cancer types include melanoma, thyroid cancer, kidney cancer, liver cancer, small intestine cancer, testicular cancer, non-Hodgkin lymphoma (NHL), anal and anorectal cancer and thymus cancer. The horizontal dashed lines indicate the historical minimal incidence. The vertical solid lines indicate the most recent year. The numbers represent the minimal percentage of extrinsic risk. The cervix uteri cancer, gallbladder cancer and oesophageal cancer are examples with declining or consistent incidence trend. The incidence rate is per 100,000 people.

  2. Extended Data Figure 2: Sensitivity analysis of different mutation rates on tLIR when the number of hits (k) required is 3. (201 KB)

    a, b, Theoretical intrinsic lifetime risks (tLIR) for cancers have been calculated based on five different mutation rates: r = 1 × 10−10, 1 × 10−9, 1 × 10−8, 1 × 10−7, 1 × 10−6. The red dashed lines are the ‘intrinsic’ risk lines based on the observed data following the same estimation mechanism as the intrinsic risk line in Fig. 3a. The green (a) and blue (b) dashed lines are the ‘intrinsic’ risk lines estimated based on total reported stem-cell numbers and total homeostatic tissue cells, respectively.

  3. Extended Data Figure 3: Sensitivity analysis of different mutation rates on tLIR when the number of hits (k) required is 4. (207 KB)

    a, b, Theoretical intrinsic lifetime risks (tLIR) for cancers have been calculated based on five different mutation rates: r = 1 × 10−10, 1 × 10−9, 1 × 10−8, 1 × 10−7, 1 × 10−6. The red dashed lines are the ‘intrinsic’ risk lines based on the observed data following the same estimation mechanism as the intrinsic risk line in Fig. 3a. The green (a) and blue (b) dashed lines are the ‘intrinsic’ risk lines estimated based on total reported stem-cell numbers and total homeostatic tissue cells, respectively.

  4. Extended Data Figure 4: Intrinsic cancer risk modelling. (166 KB)

    Part 1 of 2: propagation diagram of driver gene mutation states between generations in one stem cell, from which the stem-cell mutation transition probabilities from one generation to the next are computed.

  5. Extended Data Figure 5: Intrinsic cancer risk modelling. (115 KB)

    Part 2 of 2: schema of stem-cell divisions and driver gene mutations, from which the theoretical lifetime intrinsic risks (tLIR) for cancer due to k driver gene mutations are computed. Each coloured circle represents the mutation of a new driver gene in the given stem cell (yellow, first mutation; green, second mutation; red, third mutation). If the mutation of 3 designated driver genes would induce a cancerous stem cell (k = 3), then this diagram shows a cancer occurrence as the second stem cell in the last generation (generation n) that has accumulated all 3 driver gene mutations.

Extended Data Tables

  1. Extended Data Table 1: Robustness analysis on total stem-cell divisions and cell divisions estimates in Fig. 3 (578 KB)
  2. Extended Data Table 2: Epidemiological studies on the extrinsic risks of various cancers (260 KB)
  3. Extended Data Table 3: Percentages of intrinsic versus extrinsic MS with known and unknown causes in different cancer types (565 KB)
  4. Extended Data Table 4: Percentages of extrinsic risks based on the reported stem-cell estimates and total homeostatic tissue cells, as shown in Fig. 4 (567 KB)

Supplementary information

PDF files

  1. Supplementary Information (198 KB)

    This file contains Supplementary Text and Data and additional references.

Comments

  1. Report this comment #67437

    James DeGregori said:

    The article by Wu et al argues that extrinsic risk factors contribute far more to cancer risk than calculated by Tomasetti and Vogelstein 1. While I share their critique of the deficiencies in the assumptions and conclusions made by Tomasetti and Vogelstein 2, I would argue that they make a major error in solely regarding extrinsic factors as mutagens; i.e. they calculate the added risk of cancer caused by these factors (such as smoking) as solely coming from increases in mutation frequency. In fact, their modeling (see Fig 4) requires very large numbers of lifetime stem cell divisions, in part because they do not consider any selective impact of mutations. The accumulation of multiple mutations in a single cell lineage would be extremely improbable without clonal expansion (to increase the target size) following each mutation. Since Nowell (1976) 3, the cancer field has largely considered these mutational events to be inherently advantageous. Thus, most cancer models have been primarily focused on the occurrence of mutations, assuming that each oncogenic mutation immediately and inevitably leads to clonal proliferation and is thus rate-limiting for cancer progression. But this understanding of the fitness effect of mutations is discrepant with evolutionary theory, whereby the fitness value of a mutation is entirely dependent on context (genetic, environmental, etc.).

    Extrinsic risk factors like smoking, as well as intrinsic risk factors like aging, will do much more than affect mutation load ? they will drastically alter tissue landscapes and thus influence the selective value of mutations 4. The major driver of organismal evolution is environmental change, largely by impacting selection and drift. To give just one example, the hominid lineage leading to modern humans has undergone drastic phenotypic change in the last 5+ million years, and yet I doubt that any evolutionary biologist would argue that this was due primarily to mutation accumulation. Instead, changing environments and selective pressures drove human evolution. Our ape and chimp cousins took a different path, due to different environmental pressures, not due to differences in mutation rates. At the organismal level, it is environmental perturbations that lead to evolutionary change as organisms adapt to new environments. So why do cancer biologists so often ignore the role of altered selection driven by environmental (i.e. tissue microenvironment) changes when considering links between cancer incidence and factors such as aging, smoking, obesity, etc.?

    When the dynamic evolutionary concept of fitness is incorporated into our understanding of cancer, then cancer progression, as a type of somatic evolution, can primarily be understood as a microenvironment-dependent process. While the natural selection driven maintenance of tissues through periods of likely reproduction promote stabilizing selection in stem and progenitor cell pools (limiting somatic evolution), alterations in tissue landscapes (whether from aging, smoking or other insults) will change adaptive landscapes, promoting selection for mutations that are adaptive to this new microenvironment. Some of these mutations can be oncogenic, and thus contexts that promote tissue change like aging promote selection for adaptive oncogenic mutations. Of course, mutations are still necessary (and thus cell divisions are necessary), but mutations without the other evolutionary forces of selection and drift would be insufficient to account for increased rates of cancer in old age, in smokers, and for other cancer-promoting contexts. Hopefully, the impact of etiologic factors on cancer risk will be more frequently considered in terms of how they impact tissue microenvironments and selection, in addition to how they impact mutation frequency.

    James DeGregori
    Department of Biochemistry and Molecular Genetics
    University of Colorado School of Medicine
    james.degregori@ucdenver.edu

    1 Tomasetti, C. & Vogelstein, B. Cancer etiology. Variation in cancer risk among tissues can be explained by the number of stem cell divisions. Science 347, 78-81, doi:10.1126/science.1260825 (2015).
    2 Rozhok, A. I., Wahl, G. M. & DeGregori, J. A Critical Examination of the ?Bad Luck? Explanation of Cancer Risk. Cancer Prevention Research 8, 762-764, doi:10.1158/1940-6207.capr-15-0229 (2015).
    3 Nowell, P. C. The clonal evolution of tumor cell populations. Science 194, 23-28 (1976).
    4 Rozhok, A. I. & DeGregori, J. Toward an evolutionary model of cancer: Considering the mechanisms that govern the fate of somatic mutations. Proceedings of the National Academy of Sciences of the United States of America 112, 8914-8921, doi:10.1073/pnas.1501713112 (2015).

  2. Report this comment #67457

    Song Wu said:

    We appreciate the thoughtful comment, which raises a very interesting point that tumor microenviroment may also affect tissue-specific cancer incidences. We agree with that. However, it is important to note that we adopted a very specific definition of intrinsic cancer risk factors as defined in this article, as well as in the previous study in Science. In this formulation intrinsic risk refers only to the internal mutation rate in those dividing cells (stem or otherwise). As such, this factor is most susceptible to randomness. In our further analyses including all our four distinct approaches, we remained agnostic as to the nature of extrinsic factors. These would include not only environmental factors but also factors in the organism that are extrinsic to the tumor including inflammatory mediators, immune responses, hormones, and tissue microenviroment. These are all potentially modifiable conditions, and should belong to the domain of extrinsic factors.

    We also agree that external factors may act through avenues other than stem cell, which is the reason that we specifically did not say that external factors (or even internal factors) act exclusively through stem cells. Additionally, in some components of our analyses, such as evidence from epidemiological data and mutational signatures, the results are independent of whether external factors act through stem cells or not. In this regards, our initial approaches were primarily directed at the question of whether the strong correlation between stem cell division and cancer risk can distinguish the effects of intrinsic from extrinsic factors, and our results show that it does not.

    Overall, our main message is to promote further research into the causes of cancer and how they could be prevented.

    -The Authors

  3. Report this comment #67545

    Francis Newman said:

    The work by Wu et al. is predicated on the assumption that external factors increase cancer risk. While this is a reasonable assumption, it is by no means logically inevitable.

Subscribe to comments

Additional data