A persistent look at how tumours evade therapy – Nature.com


You have full access to this article via your institution.

Cancer can recur when a subset of tumour cells, called persister cells, survive chemotherapy. Most of these persisters are non-dividing (quiescent) in the presence of the therapeutic drug, but a rare subpopulation can re-enter the cell cycle during treatment, which enables them to proliferate. Much research has focused on the genetic mechanisms underlying such resistance to treatment. However, emerging data suggest that non-genetic mechanisms (such as changes to the complex of DNA and protein called chromatin) might also have a role in the development of a persistent state. Writing in Nature, Oren et al.1 examine the cellular lineages and gene-expression profiles of persister cells by using a method called DNA barcoding to trace tumour cells and their descendants. Their findings illuminate the role of non-genetic, reversible mechanisms in resistance to chemotherapy for a range of tumours from different tissues.

The authors analysed cell divisions in human lung cancer cells grown in vitro that have a mutation in the gene encoding the epidermal growth factor receptor (EGFR). The cells were treated with osimertinib, an inhibitor of this receptor. Oren and colleagues tracked the outcomes for cellular lineages of the tumour cell line and found that 8% of the lineages gave rise to persister cells after 14 days, and 13% of the persisters resumed the cell cycle and proliferated to form cell colonies. These results show that these cycling and non-cycling persisters arise early during the course of treatment, and that they evolve from separate cell lineages.

To characterize the molecular mechanisms associated with cycling and non-cycling persister cells, the authors developed a system that they call Watermelon, to simultaneously trace each cell’s lineage, proliferation status and transcriptional state (Fig. 1). To determine whether the persister state was due to a genetic, irreversible property of the persister cells, the authors re-exposed the persister cell population to osimertinib after a pause in treatment. They found that cells from both cycling and non-cycling populations reacquired drug sensitivity, suggesting that a non-genetic, reversible mechanism underlies persistence.

figure 1

Figure 1 | A method to analyse persister cells in tumours. Oren et al.1 present the Watermelon technique for analysing persisters — tumour cells that evade destruction by chemotherapeutic drugs such as osimertinib. The authors introduced an engineered DNA sequence into human tumour cells grown in vitro. The sequence contains a unique DNA ‘barcode’, which identifies the cells in that lineage. Also included are genes encoding a green fluorescent protein (mNeon), and a red fluorescent protein, mCherry, that requires the molecule doxycycline for its expression. The authors treated the cells with osimertinib and doxycycline, and analysed the surviving persister cells using single-cell analysis. The non-cycling (non-dividing) persister cells had a higher level of mCherry compared with the cycling (dividing) persisters. The persister cells contain reactive oxygen species (ROS), which cause oxidative damage. Cycling persister cells had lower levels of ROS and higher levels of oxidized fatty acids compared with the non-cycling persister cells. Cycling persister cells display hallmarks of antioxidant defences, including the expression of antioxidant molecules and the transcription factor NRF2.

The authors assessed gene expression using the method of single-cell RNA sequencing at different time points during a two-week treatment, and compared these signatures in cycling and non-cycling persisters. The cycling persistent state was uniquely characterized by the upregulation of defence programs that produce antioxidant molecules — including expression signatures characteristic of the metabolism of the antioxidant glutathione, as well as production of the protein NRF2, which is a transcription factor induced in response to oxidative stress. Moreover, the expression of several genes that are NRF2 targets correlated with lineages that had a large number of descendant persister cells, and the genetic engineering of cells to deplete a negative regulator of NRF2 resulted in an increase in the fraction of persisters that were cycling.

Osimertinib treatment induced the formation of reactive oxygen species (ROS), which can cause oxidative stress. At the end of treatment, cycling persisters had significantly lower levels of ROS compared with the non-cycling persisters. When the authors decreased ROS levels in cells through the addition of ROS scavenger molecules, the fraction of persister cells that were cycling increased. These analyses therefore suggest that the redox state of cells has a role in the regulation of cycling persisters.

Recognizing that redox balance is linked to metabolism, the authors profiled the products of metabolism in the cycling and non-cycling persisters, and identified 56 products that differed in their abundance between these two cell populations. The authors found a greater abundance of fatty acids linked to the molecule carnitine (a result of a preliminary step in the oxidation of fatty acids) in cycling states than in non-cycling states. The authors also noted an increase in the oxidation of fatty acids as a consequence of osimertinib treatment. Modulation of the pathway affecting fatty-acid oxidation revealed that increasing or decreasing fatty-acid oxidation leads to an increase or decrease in the fraction of cycling persisters, respectively. These results support the idea that a metabolic shift in fatty-acid oxidation affects the proliferative capacity of persisters.

To test whether their observations extended beyond the model system of lung cancer, Oren et al. generated Watermelon models of further types of human cancer, using melanoma, lung, breast and colorectal tumours. They treated the cells with suitable inhibitors, characteristic of chemotherapies, depending on the genetics underlying the particular cancer. In most of these models, the cycling persisters showed elevated fatty-acid metabolism, antioxidant responses and NRF2 signatures compared with the non-cycling persisters, showing that the authors’ findings extend to cancer types other than lung cancer.

These in vitro findings were validated using an engineered mouse model in which the animals had an inducible version of a mutant EGFR in lung tumours. After osimertinib treatment, the persister cells had a higher level of ROS and gene-expression signatures characteristic of fatty-acid metabolism compared with the cells in mice that had not received treatment. The authors also assessed gene-expression changes before and after chemotherapy in samples of cells from people with EGFR-driven lung adenocarcinoma, with melanoma driven by a mutant version of the enzyme BRAF (treated with inhibitors of BRAF and the enzyme MEK), and with breast cancer driven by a mutant version of the HER2 protein (treated with lapatinib). In all these scenarios, signatures of ROS production and fatty-acid metabolism were increased in the persister cells after treatment compared with samples of untreated tumour cells, and were higher in cycling than in non-cycling persisters.

Oren and colleagues’ study fits into the wider context of current work highlighting the importance of non-genetic mechanisms in persister-cell survival and proliferation24. One major problem when studying persisters is that they are a small fraction of the initial population of tumour cells, making it difficult to characterize them by sequencing cells in bulk. The value of the authors’ Watermelon method is that it enables the detailed characterization of persisters at the resolution of single cells. One future direction might be to apply similar single-cell approaches to study non-genetic mechanisms of resistance in other types of cancer, such as pancreatic5 or prostate6 tumours, which are fields where such research is emerging.

Understanding the dynamics of persister cells is crucial to the development of more-effective chemotherapies for cancer treatment. Previous studies found that the response pathway to the hormone oestrogen7, which has a role in breast cancer, and the pathway related to the cell-death process termed ferroptosis8,9 are associated with the persister state. Oren et al. found that, although inhibiting these pathways did decrease the amount of persister cells, there was an increase in the fraction of persisters that were cycling, suggesting that these would not be optimal chemotherapy targets.

By contrast, the authors report that inhibiting the pathway for fatty-acid oxidation using the inhibitor drug etomoxir resulted in a decrease in both the fraction of cells that were persisters and the fraction of the persisters that were cycling. This promising result indicates that modulation of this pathway, and genes that have functions related to this pathway, might be worth considering in the development of new treatment strategies.

doi: https://doi.org/10.1038/d41586-021-02117-1


  1. 1.

    Oren, Y. et al. Nature http://doi.org/10.1038/s41586-021-03796-6 (2021).

    Article  Google Scholar 

  2. 2.

    Salgia, R. & Kulkarni, P. Trends Cancer 4, 110–118 (2018).

    PubMed  Article  Google Scholar 

  3. 3.

    Sharma, S. V. et al. Cell 141, 69–80 (2010).

    PubMed  Article  Google Scholar 

  4. 4.

    Hata, A. N. et al. Nature Med. 22, 262–269 (2016).

    PubMed  Article  Google Scholar 

  5. 5.

    Chen, P.-Y. et al. Cancer Res. 78, 985–1002 (2018).

    PubMed  Article  Google Scholar 

  6. 6.

    Jolly, M. K., Kulkarni, P., Weninger, K., Orban, J. & Levine, H. Front. Oncol. 8, 50 (2018).

    PubMed  Article  Google Scholar 

  7. 7.

    Garon, E. B. et al. J. Thorac. Oncol. 8, 270–278 (2013).

    PubMed  Article  Google Scholar 

  8. 8.

    Angeli, J. P. F., Shah, R., Pratt, D. A. & Conrad, M. Trends Pharmacol. Sci. 38, 489–498 (2017).

    PubMed  Article  Google Scholar 

  9. 9.

    Hangauer, M. J. et al. Nature 551, 247–250 (2017).

    PubMed  Article  Google Scholar 

Download references

Competing Interests

The authors declare no competing interests.


Latest on:

Nature Briefing

An essential round-up of science news, opinion and analysis, delivered to your inbox every weekday.

Leave a Reply

Your email address will not be published. Required fields are marked *