Keeping it real to kill glioblastoma

The results of in vitro and in vivo screens to identify genes that are essential for the survival of a type of brain cancer show almost no overlap, underlining the need for caution when interpreting in vitro studies. See Letter p355.

Glioblastoma is the most deadly form of brain cancer in adults, and is responsible for more than 12,000 brain-tumour diagnoses in the United States each year1. Extensive characterization of its molecular basis, largely made possible through massive genomic studies2,3, has provided a comprehensive understanding of the cancer-causing genes and aberrant pathways that underlie glioblastoma development, relapse and resistance. But putative therapies that target these genes and pathways have shown little efficacy in clinical trials4, motivating investigation of alternative targets. On page 355, Miller et al.5 report results from an elegant, systematic screen, which aimed to identify therapeutic targets for glioblastoma that regulate the epigenome — a collection of heritable chemical modifications that control gene activity without altering the underlying DNA sequence.

Epigenetic modifications include those on DNA (for instance, the presence of methyl groups on cytosine residues), and those on histone proteins (including methylation, acetylation and phosphorylation), around which DNA is packaged. Collectively, these modifications dictate whether genes are expressed or silent, typically in a cell-type-specific manner. Enzymes called epigenetic modifiers are responsible for establishing, maintaining, reading and removing these modifications in a highly orchestrated, multifaceted manner.

Cancer biologists have taken a keen interest in the role of epigenetic modifiers in tumour development, largely because mutations in the genes that encode these enzymes are prevalent in a wide range of cancers6. Various classes of epigenome-targeting drug have created considerable enthusiasm of late, and clinical trials of all phases are under way for such drugs7.

Miller et al. sought to investigate the cellular requirements for epigenetic modifiers in glioblastoma (Fig. 1). The authors used cancerous cells derived from people with glioblastoma, which they either grew in culture or transplanted into the brains of mice. They used these models to perform parallel in vitro and in vivo screens to look for those genes on which cancer cells depend by introducing RNA molecules that inhibit the expression of particular genes (one RNA in each glioblastoma cell). The authors grew the cells for two to three weeks, and looked for genes whose inhibition caused cell depletion, compared with controls in which the inhibitory RNAs were inactive. Finally, they compared inventories of genes depleted under the two experimental conditions.

Figure 1: Parallel screens for essential genes.

Miller et al.5 took cells from patients with the cancer glioblastoma, and introduced different inhibitory RNA molecules into different cells (colour coded). Each RNA molecule inhibited expression of a gene involved in epigenetic modification — regulation of chemical modifications on DNA and associated proteins that alter gene activity without changing the underlying DNA sequence. The authors grew the cells in culture or in the brains of mice for two to three weeks, then analysed cell depletion, to see which genes must be properly expressed for normal glioblastoma-cell growth. The in vivo and in vitro results showed almost no overlap. Genes vital for growth from the in vitro screen related mostly to cell metabolism and macromolecule production, whereas those from the in vivo screen related to a regulatory process in gene transcription called pause release. (Adapted from Fig. 1 of the paper5.)

Surprisingly, the researchers found that the candidate gene classes identified differed vastly between experiments, with only three genes picked out under both conditions. Cells grown in culture, where the nutrients required for tumour-cell proliferation are abundant, depended on genes encoding epigenetic modifiers that regulate cell metabolism and macromolecule production. By contrast, cells propagated in mice depended on a class of gene that encodes epigenetic modifiers involved in the regulation of transcriptional pause release — a mechanism by which gene transcription is restarted after a pause. Epigenetic modification of transcriptional pause release allows the timing, rate and magnitude of transcriptional responses to be highly coordinated under certain physiological conditions8. The targets of this class of modifier include stress-response genes, and the authors suggest that activation of these genes enables tumour cells in vivo to adapt to the harsh conditions of their microenvironment.

Among the top hits emerging from the in vivo dependency screen was the gene Jumonji C-domain-containing protein 6 (JMJD6), which encodes a bifunctional enzyme that can modify molecules by removing specific methyl groups or adding hydroxyl groups. Miller et al. demonstrated that the JMJD6 protein binds to active regulatory DNA sequences that control the expression of stress-response genes in their glioblastoma models. Epigenetic modification by JMJD6 involves9 coordination of transcriptional pause release, promoting gene transcription; presumably, activation of stress-response genes by JMJD6 promotes tumour-cell survival. Targeted inhibition of JMJD6 or related factors reproducibly prolonged the survival of mice with glioblastomas, but had no detectable effect on glioblastoma cells in culture.

The authors' results have implications that extend far beyond the specifics of the current report. The overwhelming majority of similar dependency screens reported in the literature were conducted in vitro. Most use cancer-cell lines that were established decades ago and propagated in standard tissue-culture conditions, which involve nutrient-containing culture medium mixed with growth-factor-containing serum. In the current study, the in vitro screen used patient-derived cells grown under serum-free conditions — conditions that are arguably more faithful, physiologically, to the real-life conditions in which tumours grow — and still showed little overlap with the in vivo results.

This lack of congruence raises serious concerns about the validity of results obtained using standard in vitro approaches, even if such results have subsequently been reproduced in other cell lines. Similar context-dependent readouts have been observed in other in vivo dependency studies10,11, underlining the need for careful consideration of the experimental system used when conducting such screens.

Miller and co-workers' findings add to a growing list of potential epigenetic therapies for cancer. Drugs that inhibit the activity of proteins of the BET family of epigenetic modifiers have yielded promising results across a range of cancers and are now being evaluated in early clinical trials12,13. Among those causing the biggest stir are drugs that target BRD4, which (in addition to other roles) functions in concert with JMJD6 to regulate transcriptional pause release. BRD4 inhibitors have shown preliminary efficacy in glioblastoma cells and mouse models14,15, consistent with the effects of JMJD6 inhibition reported in the current study.

Although Miller and colleagues' work suggests that inhibiting transcriptional pause release might be an attractive option for treating glioblastoma, translation of these early results to the clinic is a long way off. There are currently no drugs approved by the US Food and Drug Administration that specifically modulate transcriptional pause release. However, so far no targeted approaches to treating this universally fatal form of brain cancer have provided any appreciable survival advantage to patients. Strategies such as those implicated in the current study are most certainly worth a closer look.Footnote 1


  1. 1.

    See all news & views


  1. 1

    Ostrom, Q. T. et al. Neuro Oncol. 18, v1–v75 (2016).

  2. 2

    Brennan, C. W. et al. Cell 155, 462–477 (2013).

  3. 3

    Cancer Genome Atlas Research Network. Nature 455, 1061–1068 (2008).

  4. 4

    Bastien, J. I. L., McNeill, K. A. & Fine, H. A. Cancer 121, 502–516 (2015).

  5. 5

    Miller, T. E. et al. Nature 547, 355–359 (2017).

  6. 6

    Feinberg, A. P., Koldobskiy, M. A. & Göndör, A. Nature Rev. Genet. 17, 284–299 (2016).

  7. 7

    Jones, P. A., Issa, J.-P. J. & Baylin, S. Nature Rev. Genet. 17, 630–641 (2016).

  8. 8

    Liu, X., Kraus, W. L. & Bai, X. Trends Biochem. Sci. 40, 516–525 (2015).

  9. 9

    Liu, W. et al. Cell 155, 1581–1595 (2013).

  10. 10

    Possik, P. A. et al. Cell Rep. 9, 1375–1386 (2014).

  11. 11

    Meacham, C. E. et al. Genes Dev. 29, 483–488 (2015).

  12. 12

    Wadhwa, E. & Nicolaides, T. Cureus 8, e620 (2016).

  13. 13

    Chaidos, A., Caputo, V. & Karadimitris, A. Ther. Adv. Hematol. 6, 128–141 (2015).

  14. 14

    Cheng, Z. et al. Clin. Cancer Res. 19, 1748–1759 (2013).

  15. 15

    Berenguer-Daizé, C. et al. Int. J. Cancer 139, 2047–2055 (2016).

Download references

Author information

Correspondence to Paul A. Northcott.

Related links

Related links

Related links in Nature Research

Reproducibility: Experimental mismatch in neural circuits

Cancer: Exploiting collateral damage

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Northcott, P. Keeping it real to kill glioblastoma. Nature 547, 291–292 (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.