Cell biology

Unconventional translation in cancer

Translation of RNA into proteins is a fundamental process for all cells. Analysis of a mouse model of skin cancer uncovers an atypical RNA-translation program that has a vital role in tumour formation. See Article p.494

Cancer remains a prevailing problem worldwide, yet many aspects of the process by which cellular identity is changed during tumour development remain unclear. Much attention has been directed towards understanding how altered gene transcription shapes cancer growth. On page 494, Sendoel et al.1 explore another level of gene regulation, and identify a switch in the translation machinery that has an impact on tumour formation.

Squamous-cell carcinoma is a common form of skin cancer that often involves aberrant activation of the RAS–MAPK signalling pathway and subsequent expression of the gene SOX2 (ref. 2). To study this disease, Sendoel and colleagues used a mouse model that is prone to tumour development owing to overexpression of Sox2 in the outer layer of skin — the epidermis. The authors focused on skin cells at pre-neoplastic stages, before tumour formation.

To take a snapshot of translation in the mutant epidermis, Sendoel et al. profiled the complete collection of messenger RNA sequences bound by the cells' protein-synthesis machinery (the ribosomes)3. They complemented this profiling with an analysis of the overall mRNA levels in the cells, thereby providing insight into the cells' transcriptional state.

The researchers found that differences in transcription and translation between normal and pre-neoplastic skin were not always correlated. Overall, translation seemed to be reduced in pre-neoplastic cells. Nevertheless, the profiling data suggested that some mRNAs were more efficiently translated (had more ribosomes bound per transcript) in pre-neoplastic than in normal cells. Notably, many of these mRNAs encoded proteins associated with tumour progression.

Further analyses revealed a pre-neoplastic increase in ribosome occupancy of 5′ untranslated regions (UTRs), regulatory regions that precede the normal protein-coding sequence of mRNAs. About 50% of 5′ UTRs contain sequences that can themselves code for protein, such as upstream open reading frames (uORFs)4. Translation of uORFs can result in synthesis of short proteins called peptides, which may have varying stability. Alternatively, initiation of translation within 5′ UTRs can lead to generation of extended proteins (Fig. 1).

Figure 1: A translation switch during tumour development.

a, Messenger RNAs contain regulatory 5′ untranslated regions (5′ UTRs) upstream of their main protein-coding sequence (red). Many of these 5′ UTRs (including those found in certain mRNAs associated with cancer) also contain sequences with coding potential (yellow). Under normal conditions, the eIF2 protein complex and associated factors, in combination with a subunit of the cell's translational machinery (the ribosome), control translation initiation from the beginning of the main coding sequence. Ribosomes then direct protein synthesis. b, During cancer development, eIF2 can be inactivated by phosphorylation (P), leading to an overall decline in translation. Sendoel et al.1 report that, in a mouse model of squamous-cell carcinoma, the alternative initiation factor eIF2A then promotes translation of a subset of cancer-associated mRNAs from initiation sites in 5′ UTRs. This can lead to the synthesis of small proteins (peptides) or extended proteins in addition to the main protein product. This translational switch is an early and essential event during tumour formation in this mouse model of skin cancer.

Currently, it is not clear how many 5′-UTR-encoded products have physiologically relevant functions. However, ribosome engagement with uORFs is known to affect translation of downstream coding sequences. Under conditions of stress, uORFs promote translation of some mRNAs; in non-stressful conditions, they are generally associated with impaired protein synthesis4.

The pattern of ribosome occupancies observed by Sendoel et al. suggests that many uORFs are translated in the pre-neoplastic epidermis, often from initiation sites that have an unconventional genetic code. Notably, many cancer-related mRNAs show a skewed distribution of ribosomes, with a higher proportion binding to the 5′ UTR in the pre-neoplastic cells than in normal cells. This enhanced use of upstream translation-initiation sites is frequently observed for mRNAs that have high translation efficiency, implying that the two phenomena are mechanistically linked.

In search of the mechanism underlying these changes in translation, Sendoel and colleagues inhibited the expression of genes that encode translational regulators in the skin of mouse embryos and identified those required for cell growth. They found that depletion of the eIF2 protein complex — a central factor for translation initiation in normal cells — had less effect on proliferation in the pre-neoplastic epidermis than in normal tissue. Growth in pre-neoplastic tissue seems to rely instead on the alternative initiation factor eIF2A (Fig. 1). This protein has been linked to uORF translation and use of unconventional start sites5,6,7.

To confirm the relevance of eIF2A to tumour development, the authors turned to transplant experiments. They deleted the Eif2a gene in skin cells that harboured cancer-causing mutations in the genes encoding the H-Ras and Tgfbr2 proteins, and engrafted these cells into mice. This revealed that eIF2A is required for tumour growth in vivo.

Next, the researchers demonstrated that loss of Eif2a reduces ribosome binding to 5′ UTRs in skin-cancer cells. Finally, they provided evidence that eIF2A-directed use of 5′ UTR initiation sites is linked to the efficient translation of a subset of cancer-associated mRNAs under conditions of both stress in vitro and tumour formation in vivo.

Several stress pathways are known to lead to phosphorylation and inactivation of the eIF2 complex4,8, and eIF2 inactivation has been linked to cancer development8. After incorporating Sendoel and co-workers' data, a model emerges in which, when eIF2 is inactivated, eIF2A redirects translation through 5′ UTRs to cancer-relevant mRNAs to fuel tumour growth. Strikingly, the authors reported that mutating the uORF of the gene Ctnnb1, which is required for skin-cancer development, also impaired tumour development. Their results therefore provide strong evidence that a switch to eIF2A-dependent translation and use of upstream initiation sites can be early and important events in tumour formation.

The current study extends the body of evidence8,9 that supports translational rewiring as a central player in cancer development. The atypical translation mechanism has the potential to shape the cancer proteome (the complete collection of proteins in a cancer cell) and hence affect tumour development in many ways.

Protein levels cannot be inferred solely through ribosome profiling, which allows predictions about protein synthesis but not about post-translational protein regulation10. To address this, Sendoel et al. confirm some of their findings by mass spectrometry. In the future, however, it will be valuable to analyse in more detail how the translational switch they have discovered affects both protein abundance and peptide diversity.

A central question raised by the study is whether the dependency of pre-neoplastic cells on eIF2A is solely due to altered translation of typical protein-coding sequences or whether the 5′ UTR translation events generate products that are involved in tumour formation. Interestingly, these translation products might be involved in immune responses7 and could potentially be targets for cancer immunotherapy strategies.

Many mechanistic aspects of eIF2A function await further exploration. Particularly, it will be of interest to delineate the basis on which specific RNA sequences are selected for eIF2A-mediated translation. It also remains to be established whether eIF2A dependency and translation initiation in 5′ UTRs are features of tumour development across multiple tissues, or whether they are dictated by specific mutations or cellular contexts.

Sendoel et al. describe recurrent amplifications of EIF2A in human squamous-cell carcinomas, and find a correlation between high EIF2A mRNA levels and poor prognosis. It was recently reported11 that mice survive without eIF2A, making EIF2A-dependent processes attractive targets for drug development. Identification of the key mRNAs targeted by EIF2A could likewise reveal new avenues for therapeutic intervention. It is intriguing to think that exploration of EIF2A dependency could reveal an Achilles heel in cancer. Footnote 1


  1. 1.

    See all news & views


  1. 1

    Sendoel, A. et al. Nature 541, 494–499 (2017).

    ADS  CAS  Article  Google Scholar 

  2. 2

    Boumahdi, S. et al. Nature 511, 246–250 (2014).

    ADS  CAS  Article  Google Scholar 

  3. 3

    Ingolia, N. T., Ghaemmaghami, S., Newman, J. R. & Weissman, J. S. Science 324, 218–223 (2009).

    ADS  CAS  Article  Google Scholar 

  4. 4

    Barbosa, C., Peixeiro, I. & Romão, L. PLoS Genet. 9, e1003529 (2013).

    CAS  Article  Google Scholar 

  5. 5

    Liang, H. et al. Cell Metab. 19, 836–848 (2014).

    CAS  Article  Google Scholar 

  6. 6

    Starck, S. R. et al. Science 336, 1719–1723 (2012).

    ADS  CAS  Article  Google Scholar 

  7. 7

    Starck, S. R. et al. Science 351, aad3867 (2016).

    Article  Google Scholar 

  8. 8

    Koromilas, A. E. Biochim. Biophys. Acta 1849, 871–880 (2015).

    CAS  Article  Google Scholar 

  9. 9

    Bhat, M. et al. Nature Rev. Drug Discov. 14, 261–278 (2015).

    CAS  Article  Google Scholar 

  10. 10

    Brar, G. A. & Weissman, J. S. Nature Rev. Mol. Cell Biol. 16, 651–664 (2015).

    CAS  Article  Google Scholar 

  11. 11

    Golovko, A. et al. Cell Cycle 15, 3115–3120 (2016).

    CAS  Article  Google Scholar 

Download references

Author information



Corresponding author

Correspondence to Kim B. Jensen.

Related links

Related links

Related links in Nature Research

Molecular biology: The ends justify the means

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Pedersen, M., Jensen, K. Unconventional translation in cancer. Nature 541, 471–472 (2017). https://doi.org/10.1038/nature21115

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.


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