Many mutations in one clinical-trial basket

When abnormality in a gene is linked to cancer and a drug targets the encoded protein, how can the patients who will respond to the drug be identified if the gene is mutated in many different ways in many different cancers?

Cancer usually arises from genomic abnormalities. However, the number and complexity of genetic alterations in tumours can make it difficult to predict whether, and in which tissues, a particular mutation in a specific cancer-linked gene will drive tumour growth. This poses a challenge when trying to identify effective treatments. For example, if a drug that targets a specific protein can treat a person with breast cancer who has a mutation in the gene encoding the protein, could the drug treat another patient who has a different mutation in that gene? And could it treat a person with a mutation in the same gene, but in a tumour that has developed in a different tissue? In a paper in Nature, Hyman et al.1 report the outcome of a clinical trial testing the ability of the drug neratinib, which inhibits HER2 and HER3 tyrosine kinase enzymes, to reduce or eliminate tumours. The drug was tested on 21 types of cancer in 141 people who had a total of 42 different mutations affecting one of the enzymes.

Studies in the 1970s revealed that certain chromosomal DNA aberrations can be linked to the development of specific cancer types, and that an amplification in the number of copies of particular genes can have a tumour-promoting effect2. For example, a highly lethal type of breast cancer is linked3 to amplification of the gene ERBB2 and an increase in the level of the HER2 protein that it encodes. HER2 amplification occurs in several other cancers4, including colorectal adenocarcinoma and bladder cancer. This understanding led to efforts to develop treatments to stop the action of such overexpressed proteins, resulting in several HER2-targeted therapies that are used in the clinic5 to prolong survival in people whose cancers have amplification of ERBB2. Other links between ERBB2 abnormalities and cancer have been identified; for example, single-nucleotide mutations in ERBB2 have been found in breast cancers6 that do not have amplified ERBB2, and in lung adenocarcinomas7.

The rapid development of therapeutics targeting specific cancer-associated proteins has coincided with the rise in DNA sequencing of tumours. In the past decade, the genomic alterations in tens of thousands of cancers have been characterized at single-nucleotide resolution. This has revealed that cancer-associated genes can be altered in myriad ways and that such alterations can be found in primary tumours that arise in many different tissues. However, such variability makes it hard to predict whether a specific drug will have an effect on a patient’s cancer; this, in turn, complicates the decision of who to enrol in a clinical trial. One approach to this problem involves introducing the mutated genes in question into preclinical model systems such as genetically engineered mice or cell-line models, but these models are not practical for large-scale investigations of many different gene alterations in different tissue types.

The design of clinical trials testing targeted therapeutics has changed substantially in the era of cancer genomics. Early-phase trials, in particular, now often include people who have an altered target gene, regardless of the tissue in which the tumour is present. These ‘basket’ trials seek to identify the combination of mutations and tissues that respond to treatment, offering the opportunity, if a trial progresses to a later stage, to focus on tumours in those tissues that are most likely to respond.

The ability of neratinib to target tumours with ERBB2 mutations had been demonstrated6 in human-tumour samples transplanted into mice. Hyman et al. used a basket-trial approach to test the effects of the drug on many patients with known tumour-driving ERBB2 mutations; they also examined its effects on a small number of patients who had either rare ERBB2 mutations or mutations in ERBB3, the gene that encodes HER3 and that has also been linked to tumour growth8. An interesting feature of the trial design is that it included people with mutations that had not previously been tested for a response to the drug. Some tumour types studied by Hyman and colleagues were not represented in sufficient numbers for the team to assess whether treatment had had a statistically significant effect, and enrolment in the trial is continuing for specific tissues.

The authors found that the effect of neratinib therapy varied in different mutational and tissue contexts. For example, some people who had breast, small-cell lung, cervical, biliary or salivary cancers, and who had certain ERBB2 mutations, responded to the treatment; the greatest effect was observed for breast cancers containing amino-acid alterations in the extracellular or kinase domains of HER2 (Fig. 1). Several patients with previously uncharacterized ERBB2 variants responded to neratinib, supporting the role of these mutations as tumour drivers. Neratinib had no effect on tumours with ERBB3 mutations, nor did it affect colorectal or bladder cancers that had ERBB2 mutations. The bladder-cancer result is consistent with previous studies9,10 in which HER2 targeting did not affect this type of cancer. Lack of response to neratinib provides circumstantial evidence that rare alterations in ERBB2 are unlikely to be tumour drivers.

Figure 1 | Results of a cancer clinical trial. Hyman et al.1 report the outcome of a study testing how effectively the drug neratinib can treat tumours. The tyrosine kinase enzymes HER2 and HER3 have been linked to tumour growth and can be inhibited by neratinib. The 141 patients tested had a range of mutations that altered HER2 or HER3, and, between them, had many different tumour types. The protein structures are shown, and arrows indicate the domains or interdomain locations at which protein alterations due to mutations were found. For the HER2 data shown, the cancers were grouped into ten cancer-type categories: biliary, bladder, breast, cervical, colorectal, endometrial, gastro-oesophageal, lung, ovarian or other (for all other cancer types). Responding patient numbers indicate those whose best overall response to the drug was a partial or complete response — a decrease or absence, respectively, of detectable cancer at the end of the trial.

Hyman and colleagues’ results indicate that preclinical model studies, such as those suggesting that ERBB3 can drive tumour growth8, can sometimes be misleading when trying to infer what happens in a human tumour. This might be because of how the overall genomic context influences the effect of a mutation. A tumour that has an altered target gene could also have alterations in other cancer-promoting genes. Another source of inconsistency between human and mouse studies might be the particular tissue context.

Finally, the genomic heterogeneity of tumour cells (the presence of groups of cells in the tumour that contain different genetic alterations) might be important in determining treatment response. Sequencing analysis conducted by Hyman and colleagues for certain ERBB2 mutations demonstrated that most patients whose ERBB2 mutations were present in all the tumour cells responded to neratinib, whereas those with ERBB2 mutations in only a subset of the tumour cells did not respond.

The authors noted that response to treatment could be affected by the particular genetic mutation, the location of the tumour and the specific pattern of other mutated cancer-associated genes present. This will probably hold true for most, if not all, future basket trials of targeted inhibitor therapies and is quite instructive for such studies. More-complete genomic characterization of tumours, beyond the gene(s) being targeted, will be needed to determine the genomic context linked to response or resistance to treatment. The genomic profiles and therapeutic-response data from basket trials such as this one should be made publicly available as a way of improving the design of clinical trials of other agents. Such data sets might contribute to the development of diagnostics that enable the precise identification of those patients who are most likely to benefit from targeted treatment. The data could also help to streamline the design of clinical trials and thereby hasten cancer therapeutics towards regulatory approval.

Nature 554, 173-175 (2018)


  1. 1.

    Hyman, D. M. et al. Nature 554, 189–194 (2018).

  2. 2.

    Semba, K. et al. Proc. Natl Acad. Sci. USA 82, 6497–6501 (1985).

  3. 3.

    Slamon, D. J. Science 235, 177–182 (1987).

  4. 4.

    Scholl, S., Beuzeboc, P. & Pouillart P. Ann. Oncol. 12, S81–S87 (2001).

  5. 5.

    Slamon, D. J. et al. N. Engl. J. Med. 344, 783–792 (2001).

  6. 6.

    Bose, R. et al. Cancer Discov. 3, 224–237 (2013).

  7. 7.

    Stephens, P. et al. Nature 431, 525–526 (2004).

  8. 8.

    Jaiswal, B. S. et al. Cancer Cell 23, 603–617 (2013).

  9. 9.

    Oudard, S. et al. Eur J. Cancer 51, 45–54 (2015).

  10. 10.

    Powles, T. et al. J. Clin. Oncol. 35, 48–55 (2017).

Download references

Nature Briefing

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


Sign up to Nature Briefing

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

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