Division hierarchy leads to cell heterogeneity


Cellular diversity can hamper cancer treatment. Analysis of tumour cell-division patterns now reveals how such heterogeneity can arise by a hierarchical pattern of stem-cell divisions yielding a mosaic of different cells. See Article p.227

The heterogeneity of the diverse and evolving mosaic of cells that form a tumour can be a major clinical problem because it means that some cells might dodge the effects of treatment. It has been proposed1,2,3 that cancer stem cells (CSCs, also known as cancer-initiating cells), which can be genetically identical to other cells in the tumour, are distinguished by key functional properties, including their capacity to self-renew and to initiate tumours. CSCs exhibit stem-cell-like characteristics and are considered to be responsible for the ability of tumours to spread to other parts of the body and develop resistance to chemotherapy and radiotherapy1,2,3. On page 227, Lan et al.4 provide insight into the role of CSCs in brain tumours, monitoring the division patterns of individual cells to investigate their fate and how cellular heterogeneity can arise.

The selection pressures that a cancer is under, including possible exposure to anti-cancer therapeutics, can drive the evolution of tumour-cell heterogeneity following the Darwinian principle of the survival of the fittest. Moreover, neutral evolution — random genetic changes in a population — can be a source of heterogeneity in the absence of a selective pressure5. Cellular diversity is crucial for cancer because it provides a tumour with the versatility to adapt and survive in a selective environment6.

The origin of tumour-cell heterogeneity resides either in genetic differences between cells or in non-genetic differences, such as epigenetic alterations (which can include several types of change, including those that affect DNA or the proteins that bind DNA) that are related to the maturation state of a cell, known as its differentiation state7. Cells in tumours can accumulate different mutations as the tumour grows, which increases genomic diversity5,8. Also, the differentiation state can lead to a heterogeneous cell population through a process known as hierarchical differentiation, which can occur when a subpopulation of CSCs gives rise to more-specialized descendants that have diverse capacities to proliferate, self-renew and differentiate1,2,3.

Lan and colleagues decided to study intratumour heterogeneity by labelling cells from patient-derived brain tumours called glioblastoma using a technique known as DNA barcoding. In this approach, a small, unique DNA sequence is introduced into a cell that enables that cell and its descendants — a clonal lineage of cells — to be tracked (Fig. 1). The authors used this approach to monitor the clonal evolution of individual human tumour cells that were transplanted into the brains of immunodeficient mice that do not reject human tissue. The authors used this system in a series of tumour transplantation experiments. By measuring the number and the size of clones after rounds of transplantation, the authors found that the number of cells that are able to initiate glioblastoma tumours was greater than expected.

Figure 1: Tracking cellular heterogeneity in cancer.

Lan et al.4 investigated whether the patterns of tumour-cell divisions could give rise to cellular heterogeneity. To do this, they inserted unique DNA barcode sequences into individual cells from a type of human brain tumour called glioblastoma to enable them to monitor individual cell lineages. The barcoded cells were used for rounds of transplants into the brains of immunodeficient mice that do not reject human tissue. The cellular patterns the authors observed were consistent with a model of a hierarchical pattern of cell divisions for the descendants of a given barcoded cell. This pattern consists of slowly dividing stem cells (yellow) dividing to give rise to slowly dividing stem cells and rapidly dividing progenitor cells (green). The rapidly dividing progenitor cells give rise to more rapidly dividing progenitor cells and to short-lived differentiated cells (purple).

The researchers showed that clone size was distributed in a broad pattern following a negative binomial mathematical distribution. This indicated that the clonal dynamics were the result of a neutral process within an equipotent population of tumour-initiating cells, rather than dependent on the intrinsic fitness of specific cellular clones. This finding has important implications because it suggests that the genomic diversity within a tumour does not affect the tumour-initiating capacity of cells. The growth of each cancer clone depended more on the CSC hierarchy of cell divisions than on genomic heterogeneity. More studies will be needed to assess whether this phenomenon is a feature of all cancers or whether it depends on the specific genomic heterogeneity of the tumour. For example, tumours with different numbers and complexities of genomic subclones should be tested.

The negative binomial dependence of clone size observed by Lan and colleagues was consistent with a model in which clones generated by a slow-cycling stem-cell-like population of cells give rise to a rapidly dividing progenitor population that subsequently produces short-lived, non-dividing differentiated cellular descendants. Similar results have been reported9 in mice using a cell-lineage tracing approach, and the findings also recapitulate the pattern observed during the differentiation process of radial glial cells during the normal development of the nervous system10. Lan et al. observed a hierarchical differentiation pattern in glioblastoma that mimicked the differentiation pattern observed in healthy tissues. Other groups have reported1 that the CSC hierarchy can be bidirectional, such that differentiated cells can 'revert back' to stem-cell-like cells. Although this was not observed by Lan et al., it would be interesting to assess whether this could happen in their model under certain specific conditions — for example, if the CSC compartment was depleted after drug treatment.

The authors assessed how temozolomide (TMZ), a DNA-damaging agent used as standard treatment for glioblastoma, affected clone initiation and evolution. From this analysis, the authors identified two distinct clonal groups: those that were sensitive to TMZ; and some that became enriched in the cell population on TMZ treatment. By contrast with the results previously obtained using cell-lineage tracing in mice11, the authors found that the size of TMZ-sensitive clones maintained a negative binomial distribution, indicating that TMZ does not affect their proliferative hierarchy. However, the fact that some clones became enriched suggests that some CSCs might be responsible for clonal evolution in response to TMZ. These results indicate that there might be two functionally distinct CSCs present in the tumour that have different responses to TMZ.

The authors also investigated whether CSCs can be targeted therapeutically. Because the CSC-induced diversity was a process of differentiation and hence of epigenetic regulation, they tested whether the resistant clones could be tackled with compounds that target epigenetic modifications. Using an in vitro drug-screening approach, the authors found that TMZ-sensitive clones could be targeted by a molecule called MI-2-2. MI-2-2 interrupts the interaction between the proteins menin and MLL, and thereby inhibits a complex that can regulate the epigenetic state of the cell. The TMZ-resistant clones could be targeted by an inhibitor of the protein EZH2, which also has a role in epigenetic regulation. Treatment with both compounds eradicated the in vitro self-renewal capacity of the CSCs.

Although evaluation of this reported phenomenon in other model systems and in different tumour types is certainly warranted, Lan and colleagues' work has demonstrated that a proliferative hierarchy of CSCs can be a source of intratumour heterogeneity. However, this model of human tumour cells transplanted into an immunodeficient mouse has limitations. For example, the model lacks the normal immune-system response, which could affect subclonal genomic evolution. Furthermore, the time span of the experiment differs from the reality for patients, and the process in the mouse model might be too rapid for genomic clonal evolution to work. Moreover, the transfer of tumour cells into the mouse brain might prevent the formation of the correct tumour-cell niche and microenvironment, which could also have an effect on clonal dynamics. Nevertheless, this study constitutes a breakthrough in our understanding of tumour heterogeneity and puts the spotlight on CSCs as possible therapeutic targets for glioblastoma, and perhaps even for other tumour types with a similarly poor prognosis.

Footnote 1


  1. 1.

    See all news & views


  1. 1

    Kreso, A. & Dick, J. E. Cell Stem Cell 14, 275–291 (2014).

    CAS  Article  Google Scholar 

  2. 2

    Singh, S. K. et al. Nature 432, 396–401 (2004).

    ADS  CAS  Article  Google Scholar 

  3. 3

    Bao, S. et al. Nature 444, 756–760 (2006).

    ADS  CAS  Article  Google Scholar 

  4. 4

    Lan, X. et al. Nature 549, 227–232 (2017).

    ADS  CAS  Article  Google Scholar 

  5. 5

    Sottoriva, A. et al. Proc. Natl Acad. Sci. USA 110, 4009–4014 (2013).

    ADS  CAS  Article  Google Scholar 

  6. 6

    Yap, T. A., Gerlinger, M., Futreal, P. A., Pusztai, L. & Swanton, C. Sci. Transl. Med. 4, 127ps10 (2012).

    Article  Google Scholar 

  7. 7

    Marusyk, A., Almendro, V. & Polyak, K. Nature Rev. Cancer 12, 323–334 (2012).

    CAS  Article  Google Scholar 

  8. 8

    Johnson, B. E. et al. Science 343, 189–193 (2014).

    ADS  CAS  Article  Google Scholar 

  9. 9

    Vanner, R. J. et al. Cancer Cell 26, 33–47 (2014).

    CAS  Article  Google Scholar 

  10. 10

    Pollen, A. A. et al. Cell 163, 55–67 (2015).

    CAS  Article  Google Scholar 

  11. 11

    Chen, J. et al. Nature 488, 522–526 (2012).

    ADS  CAS  Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Joan Seoane.

Related links

Related links

Related links in Nature Research

Cancer: Tumour stem-cell surprises

Cancer: Tumours build their niche

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Seoane, J. Division hierarchy leads to cell heterogeneity. Nature 549, 164–166 (2017).

Download citation

Further reading


By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.