Cancer treatment

Bacterial snack attack deactivates a drug

Tumour cells can develop intrinsic adaptations that make them less susceptible to chemotherapy. It emerges that extrinsic bacterial action can also enable tumour cells to escape the effects of drug treatment.

From birth, the surfaces and cavities of the human body are populated by microbes that, in tight partnership with the host, maintain a complex ecosystem that underlies many essential physiological processes1. One key feature of our resident microbes is their tremendous metabolic capacity. Our bacterial population contains millions of genes2 encoding enzymes that can process substances that have been derived from nutrients or the environment, or that have been administered as drugs. Such metabolism generates other compounds that can affect host homeostasis3. However, microbial metabolism is not always beneficial for the host. Writing in Science, Geller et al.4 report that bacteria within a tumour can metabolize an anticancer drug into an inactive form and thereby render it ineffective.

It was previously observed5 that the in vitro culture of two types of human tumour cell together with non-cancerous cells called fibroblasts resulted in unexpected tumour-cell survival after treatment with the chemotherapy drug gemcitabine. Geller and colleagues carried out some detective work to investigate this phenomenon. By using DNA-sequence analysis, they found that the fibroblast sample was contaminated with a bacterium called Mycoplasma hyorhinis. Other examples of bacteria being linked to cancer treatment outcomes have been reported. For instance, people who have colorectal cancer often harbour high numbers of tumour-associated Fusobacterium nucleatum bacteria, and this association is linked to poor clinical prognosis6. This bacterium can trigger the activation of a pathway in colorectal-cancer cells that results in the tumour developing resistance to the drugs fluorouracil and oxaliplatin7.

The cytidine deaminase enzyme of M. hyorhinis might be responsible for converting gemcitabine into an inactive form through a metabolic process called deamination8. Geller and colleagues searched for cytidine deaminase genes in a database of bacterial genomes. The gene exists as long and short versions, and the authors found that the long version was present in around 11% of the species in the database. Of these species, more than 98% are in the bacterial class called Gammaproteobacteria, to which the model organism Escherichia coli belongs.

The authors used an E. coli strain that can detect and invade tumours to study the effect of cytidine deaminase on gemcitabine treatment. A technique called high-performance liquid chromatography–tandem mass spectrometry enabled Geller and colleagues to monitor the form of gemcitabine present in the culture medium of bacterial cells grown in vitro. When the authors deleted the long version of cytidine deaminase from E. coli, it abolished the microbe's capacity to metabolize gemcitabine into its inactive form. However, if a copy of the gene was reintroduced to the bacterium, this restored its ability to metabolize the drug.

The researchers investigated this phenomenon in vivo by analysing the effect of bacteria on gemcitabine treatment of mouse colon-cancer cells that were transplanted into mice below the surface of their skin. If the animals were inoculated with E. coli, tumour growth was unaffected by the drug. However, tumours responded to gemcitabine treatment if the animals were treated with antibiotics or if the E. coli strain used lacked the long form of the cytidine deaminase gene. To test whether bacterial-mediated drug metabolism occurs in the tumour microenvironment or elsewhere in the body, the authors used a miniature device to deliver gemcitabine directly into tumours. The results were consistent with bacterial metabolic activity in the tumour microenvironment.

To determine whether this phenomenon has clinical relevance, the authors studied a human cancer called pancreatic ductal adenocarcinoma (PDAC), which is often treated with gemcitabine. In 86 out of 113 tissue samples isolated from PDAC tumours, microscopy techniques, molecular-detection approaches and sequencing analysis uncovered signs of bacterial presence. Conducting the same tests using healthy pancreatic samples revealed only rare cases (3 out of 20 samples) of bacterial tissue infiltration.

DNA sequence analysis revealed that the bacteria in PDAC samples were mainly Gammaproteobacteria from the Enterobacteriaceae family, which harbours the long form of the cytidine deaminase gene and to which E. coli belongs. The authors isolated and cultured bacteria from 15 PDAC samples. For 14 of these 15 samples, if the bacteria were cultured in vitro with human colon-cancer cells, the tumour cells were unaffected by gemcitabine administration. Together, the authors' findings are consistent with a model in which tumours are shielded from the effect of drug treatments if bacteria in the tumour microenvironment metabolize and deactivate the therapeutic agent (Fig. 1).

Figure 1: Bacteria can deactivate a drug treatment that targets cancer.

Geller et al.4 investigated how bacteria can affect cancer treatment and observed that bacterial presence was associated with a lack of tumour-cell response to a chemotherapeutic drug called gemcitabine. a, The authors analysed samples isolated from human pancreatic cancer and found that the tumours contained bacteria from the Enterobacteriaceae family. Such bacteria contain a long version of the cytidine deaminase enzyme (CDDL) that can deactivate gemcitabine. b, The authors used a mouse model system to study the in vivo relationship between the presence of Enterobacteriaceae bacteria and the effectiveness of gemcitabine in targeting tumour cells from the colon. If antibiotic treatment was used to kill the bacteria, gemcitabine treatment destroyed the colonic tumour cells.

Recognition of the therapeutic potential of our resident bacteria is growing. Interactions between bacteria and immune cells are essential for the efficacy of chemotherapeutic agents including cyclophosphamide, oxaliplatin and cisplatin9,10,11, and for immunotherapy approaches that boost the cancer-targeting activity of the immune system12,13. Geller and colleagues' work now adds another perspective to the effects of microbes on therapy, by instead demonstrating that bacteria might prevent a treatment from working.

Their findings also reveal that co-administration of antibiotics during chemotherapy might have complex consequences. For example, could antibiotics prevent cancer-drug deactivation by depleting the microbes responsible, or might they destroy beneficial organisms necessary for optimal therapeutic response? Untangling the nature of microbial networks that promote or hinder therapeutic regimens will undoubtedly open more treatment avenues for people who have cancer.

As exciting as Geller and colleagues' work is, further studies are needed before these insights result in new clinical strategies. Their model system, in which cells from cancer cell lines were transplanted into mice and high levels of E. coli were injected into the animals' bloodstream, does not reflect the natural process of cancer microenvironment formation, and might not capture how tumours are usually infiltrated by microbes. Whether microbes colonize tumours in animal pancreatic-cancer models should be explored, as well as the extent to which gemcitabine treatment is affected by microbes in such a system. Moreover, the amount of microbial biomass necessary for gemcitabine deactivation in the pancreas is unknown, so whether such a quantity could be reached in the human pancreas is unclear.

Geller and colleagues demonstrated that healthy pancreatic tissue is not readily colonized by bacteria, therefore a bacterial community is presumably not normally needed in this organ and such a presence probably reflects a pathological situation. The tumour microenvironment can generate metabolic products that are detected by sensor proteins on bacteria, activating bacterial migration into tumours14. Dissecting the circumstances and mechanisms leading to microbial colonization of tumours should illuminate15 how bacteria affect tumour progression and treatment response.

Episodes of pancreatic inflammation known as pancreatitis represent a major risk factor for pancreatic cancer. It would therefore be worth determining whether bacteria are present in the pancreas during pancreatitis and, if so, whether these bacteria contribute to cancer development and possible escape from the effects of subsequent cancer drug treatment.

Changes in response to cancer treatment can be driven by several factors, including the genetic background of the tumour cells, or bacterial-induced changes to tumour-cell signalling. Geller and colleagues' work adds microbial metabolism of anticancer drugs as another potential contributor to poor prognosis when tumours are treated. Although more data will be needed to firmly establish whether microbial metabolism within tumours normally affects the success of tumour treatment, the findings reveal a potential new therapeutic strategy for anticancer treatment.Footnote 1


  1. 1.

    See all news & views


  1. 1

    Kau, A. L., Ahern, P. P., Griffin, N. W., Goodman, A. L. & Gordon, J. I. Nature 474, 327–336 (2011).

    CAS  Article  Google Scholar 

  2. 2

    Ursell, L. K., Metcalf, J. L., Parfrey, L. W. & Knight, R. Nutr. Rev. 70, S38–S44 (2012).

    Article  Google Scholar 

  3. 3

    Koppel, N., Maini Rekdal, V. & Balskus, E. P. Science 356, eaag2770 (2017).

    Article  Google Scholar 

  4. 4

    Geller, L. T. et al. Science 357, 1156–1160 (2017).

    ADS  CAS  Article  Google Scholar 

  5. 5

    Straussman, R. et al. Nature 487, 500–504 (2012).

    ADS  CAS  Article  Google Scholar 

  6. 6

    Mima, K. et al. Gut 65, 1973–1980 (2016).

    CAS  Article  Google Scholar 

  7. 7

    Yu, T. et al. Cell 170, 548–552 (2017).

    CAS  Article  Google Scholar 

  8. 8

    Vande Voorde, J. et al. J. Biol. Chem. 289, 13054–13065 (2014).

    CAS  Article  Google Scholar 

  9. 9

    Viaud, S. et al. Science 342, 971–976 (2013).

    ADS  CAS  Article  Google Scholar 

  10. 10

    Iida, N. et al. Science 342, 967–970 (2013).

    ADS  CAS  Article  Google Scholar 

  11. 11

    Daillère, R. et al. Immunity 45, 931–943 (2016).

    Article  Google Scholar 

  12. 12

    Vétizou, M. et al. Science 350, 1079–1084 (2015).

    ADS  Article  Google Scholar 

  13. 13

    Sivan, A. et al. Science 350, 1084–1089 (2015).

    ADS  CAS  Article  Google Scholar 

  14. 14

    Kasinskas, R. W. & Forbes, N. S. Cancer Res. 67, 3201–3209 (2007).

    CAS  Article  Google Scholar 

  15. 15

    Tsilimigras, M. C. B., Fodor, A. & Jobin, C. Nature Microbiol. 2, 17008 (2017).

    CAS  Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Christian Jobin.

Related links

Related links

Related links in Nature Research

Cancer: Tumours addicted to drugs are vulnerable

Cancer: Double trouble for tumours

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Jobin, C. Bacterial snack attack deactivates a drug. Nature 550, 337–339 (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.