Intratumor heterogeneity reflects clinical disease course

Intratumoral genetic heterogeneity of driver somatic mutations is present in a variety of tumor types, yet the extent of heterogeneity is variable. We propose that this variation is a reflection of the inherent biology of a given tumor type, representing the pace of metastatic dissemination and hence clinical disease course.

Cancer genome instability is the result of somatic mutations, copy-number alterations, dynamic changes in chromosome number and structure and epigenomic instability. These changes lead to cell-to-cell variation and intratumor heterogeneity (ITH) and thereby provide the molecular events on which selection acts1. ITH is associated with cancer progression and contributes to cancer drug resistance2. As a consequence, many new therapies fail to deliver meaningful survival benefits, which increases the health economic cost of therapeutic development1,3. The development of a broader understanding of tumor subclonal population structures (Fig. 1a) and the impact of neutral evolution and selection during disease progression may improve risk predictions regarding disease relapse and the metastatic cascade4 (Box 1 provides a glossary of relevant evolutionary terms). Elucidation of how minor subclones may sustain or restrain tumor growth, combined with delineation of microenvironmental influences on tumor subclonal progression, may provide new opportunities for targeted intervention and tumor control1,3,5.

Fig. 1: Evolutionary trajectories in solid tumors and relationship to clinical course.

a, Conceptual framework of truncal (clonal) mutations, which are present in every cancer cell, and branched (subclonal) mutations, which are present in only a subset of cancer cells. Mutations in these groups are typically distinguished through inference of the MRCA clone; i.e., the clone containing events common to all cancer cells analyzed. Mutations in the MRCA are considered truncal or clonal (i.e., mutation in gene A), and other mutations are considered branched or subclonal (i.e., mutations in genes B and C). b, A highly branched Darwinian growth pattern. c, A monoclonal punctuated growth pattern. d, The spectrum of ITH across cancer types, based on multi-region sequencing data, plotted as the proportion of driver mutations found to be subclonal (y-axis) versus tumor types (x-axis). In parentheses (below x-axis), incidence rate of each tumor type, according to data from the UK Office for National Statistics 2017 cancer registry. Below the plot are illustrative phylogenetic trees that demonstrate the spectrum of intratumor heterogeneity, from high (left) to low (right): green dots represent tumor subclones, and lightning bolts indicate subclonal driver mutation. e, Relationships among ITH, surgical resection rates and patient survival, plotted as ITH (measured as the proportion of drivers that are subclonal; x-axis) versus the rate of major surgical resection (y-axis) versus 5-year overall survival (data from UK Office for National Statistics; z-axis). ITH data are from refs. 7,14. Surgical resection rates are from data from the UK National Cancer Registration and Analysis Service, 2004–2006. Overall survival data are from the UK Office for National Statistics, with data averaged (mean) across sexes (for non–sex-specific cancers).

Work over the past decade has revealed wide variation in cancer subclonal structures across tumor types, with key insights emerging as to how ITH may manifest. Broadly speaking, these evolutionary phenomena illustrate that tumors at one end of the spectrum follow a clear Darwinian growth pattern, displaying extensive ITH characterized by branched heterogeneous somatic mutations and selection of subclonal driver mutations and DNA copy-number events6 (Fig. 1b). At the other end of the spectrum are tumors characterized by a monoclonal growth pattern that display limited ITH for somatic mutations and have multiple clonal (truncal) driver events, and alterations in both somatic coding regions and copy number. In such tumors, these characteristics are all present in the most recent common ancestor (MRCA), and subsequent evolutionary trajectories seem to be genetically restrained7 (Fig. 1c). How and why these distinct tumor evolutionary growth patterns emerge is not clear. Cell of origin, disease latency, strength and contingencies of the originating clonal driver(s), and the native and tumor microenvironment, together with underlying germline constraints, mutational processes and environmental carcinogens, may all play a role8. These have yet to be systematically investigated in the appropriate models.

Two additional variables that have not been fully considered in interpretation of the extent and prevalence of ITH within and between tumor types are the clinical course of each disease and the standards of care for managing them. Different tumor types have very different evidence-based approaches to their management that reflect the clinical course of each cancer type or subtype; hence, the opportune time to sample them for evolutionary analyses varies. For example, clear-cell renal-cell carcinoma (ccRCC) is a tumor type for which surgical management plays a major role. The presence of limited metastatic disease is not mutually exclusive with surgery, and such patients commonly undergo cytoreductive nephrectomy. In selected patients, particularly in those with single-site lung metastases or adrenal metastases, metastasectomy with complete resection can result in impressive 5-year overall survival (40% or 60%, respectively)9. It is not uncommon for tumor thrombi to extend intraluminally within the vena cava, a serious and dramatic presentation of the disease, yet surgery remains a viable option for management of this scenario. This clinical paradigm has supported evolutionary studies of cancer because the primary tumor and matched synchronous metastases can be sampled during surgery. So far, samples of primary and paired metastases in ccRCC have been obtained largely during primary surgical resection and through metastasectomies later in the disease course.

By contrast, surgical management plays a more limited role in pancreatic ductal adenocarcinoma (PDA). Although both the kidneys and pancreas are retroperitoneal organs, the latter lies deep within the abdomen adjacent to vital structures. Moreover, PDA is asymptomatic until very advanced stages; thus, the majority of patients have unresectable disease because it is locally advanced at diagnosis and/or has metastasized to distant organs with rapid disease progression10. Given these clinical presentations, no more than 15% of patients are eligible for potentially curative surgical resection and the majority will succumb to their disease within 5 years11. Thus, samples of advanced-stage PDA used for multiregion sequencing and evolutionary analyses have been obtained predominantly through post-mortem collection.

The distinct biology of each tumor type is probably at the root of the divergent clinical behaviors noted above. At one end of this extreme are certain ccRCC cases, with metastases confined to a few sites (oligometastatic disease) where adaption and selection occur at the level of the genome, which leads to ITH, parallel evolution and convergence on a common phenotype6,12. In advanced-stage tumors, copy-number variation takes on a preeminent role and leads in part to loss of CDKN2A (a driver gene) on chromosome 9p, selected in the metastatic subclone12. At the other extreme are patients who present with extremely aggressive metastatic PDA in which almost all alterations in major driver genes (KRAS, CDKN2A, TP53 and SMAD4) are present in the MRCA. In this tumor type, adaption and selection occur at the level of the epigenome, and this leads to phenotypic and metabolic plasticity13. Thus, the biology of each cancer type, partially reflected by their distinct evolutionary patterns, has influenced the time at which samples have been obtained for study. What remains unknown is if the clinical time during which samples are taken reflects the full spectrum of evolutionary trajectories that may occur within each disease, or whether there are underappreciated biases from tissue sampling related to the extent and prevalence of ITH both within tumor types and between tumor types.

In part by the choice of tumor type to study, the very different clinical behaviors and the paradigms used for clinical management of advanced-stage ccRCC versus that of PDA, our respective laboratory efforts have led to us to study what we believe are extremes of a spectrum of evolutionary trajectories that may occur across tumor types (Fig. 1d). This notion is supported by multi-region sequencing efforts from multiple groups that have demonstrated that endometrial, colorectal, prostate, brain, esophageal, breast and lung tumors fall between these two extremes14. Of clinical note are high-incidence cancer types such as prostate cancer (173.1 cases per 100,000 men; Fig. 1d), in which high levels of ITH have been documented15.

Nevertheless, several technical caveats should be also considered in assessment of ITH via multi-region sequencing, notably tumor size and the purity and number of biopsy samples taken, as well as differences in sequencing methodology. As these differences have not been controlled for in a standardized pan-cancer multi-region sequencing study, the extent to which estimates of ITH will change as studies become more comprehensive remains to be seen. In the cases thus far in which the relationship between tumor size and ITH has been assessed, larger tumors (>10 cm maximal dimension) were found to have greater ITH than that of smaller tumors (0–4 cm)6.

Adding to the complexity, outliers within a tumor type can also be observed; for example, in studies of advanced-stage ccRCC in the Renal TRACERx cohort, a minority of patients suffered rapid progression and dissemination across multiple metastatic sites and succumbed to disease early12, presenting with aggressive disease biology more akin to PDA. Genomic profiling of these tumors has revealed the presence of multiple truncal (clonal) driver-gene alterations, relatively monoclonal structures, high chromosomal complexity and loss of chromosome 9p as an early clonal event in tumor evolution. We suggest that such tumor evolutionary growth patterns occur in patients with inoperable or rapidly progressing disseminated metastatic disease, which renders post-mortem analyses the main approach for further defining disease biology in this high-risk group. Similarly, a minority of patients with PDA have tumors with a more indolent growth pattern, oligometastatic disease and relatively long-term survival, as may be seen in ccRCC. These tumors may have fewer clonal driver-gene alterations. Deep genomic profiling of one exceptional patient who survived 44 months with a conservatively treated locally advanced PDA revealed a CTNNA2 deletion as the sole truncal driver-gene event16.

Treatment history can also profoundly influence the subclonal driver landscape and the degree of ITH. For example, multiple parallel evolutionary resistance mechanisms have been described in distinct metastatic lesions obtained post-mortem from patients treated with inhibitors of the receptor EGFR (epidermal growth factor receptor) or the kinase PI(3)Kα (phosphatidylinositol-4,5-bisphosphate 3-kinase, catalytic subunit-α) in EGFR-activating-mutant non–small-cell lung cancer17 or PI(3)Kα-activating-mutant breast cancer18, respectively. It is not known how sequential lines of therapy alter the somatic landscape across multiple disease sites both within tumor types and between tumor types, or how such interventions may affect conclusions regarding the degree of ITH in cancers. As longitudinal and post-mortem studies increase in breadth across tumor types, it will be intriguing to assess how the extent of ITH within and between tumor types correlates with the clinical disease course, treatment histories and patterns and tempo of metastatic dissemination. Emerging findings indicate that ccRCC, endometrial tumors and central nervous system tumors lie at one end of the spectrum, and PDA and non–small-cell lung cancer lie at the other. For instance, in the first 100 patients to be enrolled in the non–small-cell lung cancer TRACERx cohort, the proportion of subclonal somatic driver mutations was 33%, compared with 55% in the first 100 patients in the ccRCC TRACERx cohort19.

The data outlined above raise two points for future investigation. First, all studies of ITH should strive to take into account clinical context, treatment histories, pace of disease progression and extent of dissemination to metastatic sites. For example, both post-mortem analyses and longitudinal analyses will be required for full appreciation of the extent of ITH within a single tumor type at diagnosis versus in advanced-stage disease. We expect that when these factors are taken into account and such studies are performed for many tumor types, this will more accurately reflect the full spectrum of evolutionary trajectories, together with how specific evolutionary trajectories are related to specific truncal driver genes, chromosomal instability, the dynamic microenvironment and the predatory immune system.

Second, it will be intriguing to elucidate the level of influence that the tissue or cell of origin has on the degree of ITH within each cancer type. For example, ccRCC and papillary renal-cell carcinoma (pRCC), despite originating from the same tissue and probably the same cell type (convoluted proximal tubular cells), have considerable distinctions in their evolutionary patterns20. pRCC seems to be a disease driven predominantly by copy number (with changes in chromosome 7 being common clonal events), followed by subclonal driver mutations in genes encoding components of the transcription factor NRF2 (NFE2L2/NRF2 nuclear factor, erythroid 2–like 2) pathway. In contrast to ccRCC, pRCC is characterized by an absence of clonal mutations in VHL (which encodes the tumor suppressor VHL (Von Hippel–Lindau tumor suppressor)), clonal or sublconal mutations in PBRM1 (which encodes the tumor suppressor PBRM1 (polybromo 1)) and fewer mutations in genes encoding components of the PI(3)Kα signaling pathway.

Finally, a pertinent question for the field is to what extent ITH, surgical sampling and survival might be interlinked. Research output thus far (Fig. 1e) has revealed various tumor types with low ITH, low surgery rates and low survival (pancreatic, lung, brain and esophageal tumors), whereas other tumor types have elevated ITH, surgery and survival rates (kidney, colorectal, breast, prostate and endometrial tumors). An interesting area for future studies is to what extent ITH and patterns of Darwinian and punctuated evolution might influence survival outcomes. Ultimately, a deep understanding of evolutionary trajectories within and across tumor types, before and after treatment, may distinguish patients with more indolent disease biology with oligometastatic progression from those with more rapid dissemination within and across multiple organs. Such insights may in turn facilitate clinical trial stratification and clinical management, which would enable prioritization of patients for metastasectomy or for stereotactic or focal ablative approaches to metastatic lesions in the palliative setting.


  1. 1.

    Greaves, M. Cancer Discov. 5, 806–820 (2015).

    CAS  Article  Google Scholar 

  2. 2.

    Greaves, M. Nat. Rev. Cancer 7, 213–221 (2007).

    CAS  Article  Google Scholar 

  3. 3.

    Gillies, R. J., Verduzco, D. & Gatenby, R. A. Nat. Rev. Cancer 12, 487–493 (2012).

    CAS  Article  Google Scholar 

  4. 4.

    Williams, M. J. et al. Nat. Genet. 50, 895–903 (2018).

    CAS  Article  Google Scholar 

  5. 5.

    Greaves, M. & Maley, C. C. Nature 481, 306–313 (2012).

    CAS  Article  Google Scholar 

  6. 6.

    Turajlic, S. et al. Cell 173, 595–610 (2018).

    CAS  Article  Google Scholar 

  7. 7.

    Makohon-Moore, A. P. et al. Nat. Genet. 49, 358–366 (2017).

    CAS  Article  Google Scholar 

  8. 8.

    Maley, C. C. et al. Nat. Rev. Cancer 17, 605–619 (2017).

    CAS  Article  Google Scholar 

  9. 9.

    Krabbe, L.-M., Bagrodia, A., Margulis, V. & Wood, C. Semin. Intervent. Radiol. 31, 27–32 (2014).

    Article  Google Scholar 

  10. 10.

    Kleeff, J. et al. Nat. Rev. Dis. Primers 2, 16022 (2016).

    Article  Google Scholar 

  11. 11.

    Winter, J. M. et al. Ann. Surg. Oncol. 19, 169–175 (2012).

    Article  Google Scholar 

  12. 12.

    Turajlic, S. et al. Cell 173, 581–594 (2018).

    CAS  Article  Google Scholar 

  13. 13.

    McDonald, O. G. O. G. et al. Nat. Genet. 49, 367–376 (2017).

    CAS  Article  Google Scholar 

  14. 14.

    Gomez, K. et al. BMC Cancer 18, 85 (2018).

    Article  Google Scholar 

  15. 15.

    Gundem, G. et al. Nature 520, 353–357 (2015).

    CAS  Article  Google Scholar 

  16. 16.

    Kohutek, Z. A. et al. Cold Spring Harb. Mol. Case Stud. 3, a001701 (2017).

    Article  Google Scholar 

  17. 17.

    Blakely, C. M. et al. Nat. Genet. 49, 1693–1704 (2017).

    CAS  Article  Google Scholar 

  18. 18.

    Juric, D. et al. Nature 518, 240–244 (2015).

    CAS  Article  Google Scholar 

  19. 19.

    Jamal-Hanjani, M. et al. N. Engl. J. Med. 376, 2109–2121 (2017).

    CAS  Article  Google Scholar 

  20. 20.

    Kovac, M. et al. Nat. Commun. 6, 6336 (2015).

    CAS  Article  Google Scholar 

Download references


Research support by CA179991 and CA220508 to C.A.I.D., who also receives research support from Bristol-Myers Squibb. K.L. is supported by a UK Medical Research Council Skills Development Fellowship Award (grant reference number MR/P014712/1). C.S. is supported by the Francis Crick Institute that receives its core funding from Cancer Research UK (FC001169,FC001202), the UK Medical Research Council (FC001169, FC001202), and the Wellcome Trust (FC001169, FC001202) and from Cancer Research UK (TRACERx, PEACE and CRUK Cancer Immunotherapy Catalyst Network), the CRUK Lung Cancer Centre of Excellence, Stand Up 2 Cancer (SU2C), the Rosetrees Trust, NovoNordisk Foundation (ID16584) and the Breast Cancer Research Foundation (BCRF), European Research Council under the European Union’s Seventh Framework Programme (FP7/2007-2013)/ ERC grant agreement n°FP7 – 617844 (PROTEUS) and Marie Curie Network PloidyNet.

Author information



Corresponding authors

Correspondence to Christine A. Iacobuzio-Donahue or Charles Swanton.

Ethics declarations

Competing interests

The authors declare no competing interests.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Iacobuzio-Donahue, C.A., Litchfield, K. & Swanton, C. Intratumor heterogeneity reflects clinical disease course. Nat Cancer 1, 3–6 (2020).

Download citation

Further reading


Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing