Cancer genomics

Spot the difference

Article metrics


  • A Correction to this article was published on 02 February 2017

A molecular analysis of human oesophageal cancers reveals abnormalities that might be targetable by existing drugs, and indicates that the current stratification of these tumours into subtypes is incomplete. See Article p.169

The classification of cancers of the oesophagus and stomach into different subtypes for different treatments is a subject of debate in the cancer community. Conventionally, these cancers have been classified according to the positions at which they arise in the body. On page 169, an international collaboration, the Cancer Genome Atlas Research Network, presents a comprehensive genomic analysis of human oesophageal squamous-cell carcinomas (ESCCs) and oesophageal adenocarcinomas (EACs). These are the two major subtypes of oesophageal cancer recognized according to microanatomical categorization, and together they cause more than 400,000 deaths worldwide each year2. The study reveals previously unknown genomic aberrations — a discovery that may lead to more-rational clinical trials for oesophageal cancer treatments. Furthermore, it provides evidence that oesophageal and gastric cancers should be stratified by molecular features, rather than by anatomical site.

The Cancer Genome Atlas Research Network ( produces open-access genomic data about human cancers. Together with other similar projects, it has revealed the genomic landscape of many types of human cancer across nearly all organ sites, confirming some previously understood similarities and differences between cancers and uncovering unexpected molecular profiles that might lead to new avenues for cancer prevention and treatment3.

In the current study, the consortium first demonstrated that ESCCs and EACs have distinct molecular profiles through DNA sequencing, analyses of gene and protein expression and analyses of methylation of DNA (which can lead to alterations in gene expression). The authors outline several biological pathways that are genetically altered in ESCC, EAC or both, and which confirm the need for different therapeutic strategies in these two diseases.

Analysis of genetic copy-number alterations (in which large portions of a chromosome are deleted or replicated, altering gene-expression levels) revealed amplification of several different genes in ESCC and EAC. For instance, amplification of the genes ERBB2 and VEGFA was found in samples from EAC but not ESCC. An inhibitor of ERBB2 called trastuzumab has been approved for use by the US Food and Drug Administration, and an inhibitor of the VEGF signalling pathway called ramucirumab is frequently used in a subset of EACs — these therapies can now be quickly deployed in patients who harbour these genetic mutations. By contrast, mutations that lead to dysregulation of the cell cycle occurred in both ESCC and EAC. As the authors suggest, available inhibitors of cell-cycle proteins such as the drug palbociclib might effectively combat both diseases.

The division of ESCCs and EACs into further subgroups could lead to more-effective, precision treatments. The consortium identified subgroups within ESCC that were characterized by specific genetic alterations. The first subgroup was characterized mainly by mutations that activated the NRF2 signalling pathway, including activating mutations in the NFE2L2 gene, which is associated with resistance to chemoradiotherapy and poor patient outcome4. NFE2L2 encodes a transcription factor, whose activity is inhibited by the protein KEAP1 — this pathway could therefore potentially be targeted for treatment if drugs could be developed to inhibit mutant, activated NRF2. Intriguingly, molecular features of this ESCC class bear a striking resemblance to classical subtypes of lung and head and neck cancers (Fig. 1), suggesting that these diseases may respond similarly to targeted therapies.

Figure 1: A molecular classification of cancers.

Cancers are conventionally stratified by the site at which they arise, for instance as head and neck cancers, oesophageal squamous-cell carcinomas (ESCCs), oesophageal adenocarcinomas (EACs) and gastric cancers. But cancers at the same site can be further subdivided on the basis of molecular profiles (represented as differently coloured portions of each pie chart. Greyed out segments are not molecularly related to each other). This has already been done for head and neck and gastric cancers9. Genomic analyses of ESCCs and EACs by the Cancer Genome Atlas Research Network1 reveal that EACs are molecularly the same disease as a type of gastric cancer (purple), that ESCCs can be subdivided into three subtypes, and that one ESCC subtype (dark green) resembles some head and neck cancers.

The second ESCC subgroup is characterized in part by a decrease in DNA methylation compared with the other subgroups and by overexpression of the gene BST2, which encodes a protein that modulates immune activity and that is associated with poor prognosis in ESCC (ref. 5). Here, too, there is therapeutic potential, because an anti-BST2 antibody has shown antitumour activity in other cancers6,7,8. The third subtype is characterized by mutation in the gene SMARCA4 and activation of the PI3K signalling pathway. PI3K inhibitors that are currently in clinical trials may therefore be effective in this subgroup.

There has been some debate about how to classify EACs that arise near to or at the entrance to the stomach, because it can be unclear on which side of the gastroesophageal junction the tumour originated. So the authors next sought to compare molecular features of EAC with those of gastric cancers. The findings indicate a striking similarity between EAC and one subtype of gastric cancer, denoted as having chromosomal instability (CIN). Recurrent gene amplifications are shared between EAC and gastric CIN tumours, but not between EAC and ESCC. Although there are subtle differences in the frequencies at which some genetic alterations arise, the authors' evidence indicates that EAC and gastric CIN are the same disease at the molecular level (Fig. 1).

This study is an excellent example of the potential for molecular technologies and bioinformatic approaches to contribute to our understanding of human disease. Conventionally, cancers are treated according to where they arise in the body and their appearance under a microscope. We now understand that cancers that develop in the same location can vary greatly at the molecular level. Treatment decisions made without this new wealth of genomic data can lead to patients receiving a drug that is unlikely to be beneficial, or not receiving one predicted to be effective. Many cancers continue to be treated on the basis of organ site alone, but clinicians are beginning to embrace molecular-classification systems — a trend that is likely to accelerate as similar studies are conducted across cancer types.

Clinical trials mainly enrol patients who have cancers at the same organ site. But the current study demonstrates that EAC and ESCC should be separated in the clinical setting so that each can be targeted according to its specific genomic features. In addition, basket clinical trials (in which multiple organ sites are combined on the basis of common molecular features) can now pool ESCC with head and neck cancer, and EAC with gastric CIN cancer, to maximize the benefit to patients and accelerate the approval of effective therapies for each disease. Large-scale genomic studies such as this one may thereby hold the key that unlocks the promise of precision medicine.Footnote 1


  1. 1.

    See all news & views


  1. 1

    The Cancer Genome Atlas Research Network. Nature 541, 169–175 (2017).

  2. 2

    Torre, L. A. et al. CA Cancer J. Clin. 65, 87–108 (2015).

  3. 3

    Weinstein, J. N. et al. Nature Genet. 45, 1113–1120 (2013).

  4. 4

    Shibata, T. et al. Neoplasia 13, 864–873 (2011).

  5. 5

    Mukai, S. et al. Ann. Surg. Oncol. (2016).

  6. 6

    Wang, W. et al. Lung Cancer 63, 23–31 (2009).

  7. 7

    Staudinger, M. et al. Blood Cancer J. 4, e219 (2014).

  8. 8

    Hiramatsu, K. et al. Cancer Sci. 106, 1474–1478 (2015).

  9. 9

    The Cancer Genome Atlas Network. 517, 576–582 (2015).

  10. 10

    The Cancer Genome Atlas Research Network. Nature 513, 202–209 (2014).

Download references

Author information

Correspondence to Noah D. Peyser or Jennifer R. Grandis.

Related links

Related links

Related links in Nature Research

Genomics: Comparisons across cancers

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Peyser, N., Grandis, J. Spot the difference. Nature 541, 162–163 (2017) doi:10.1038/nature21112

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.