News & Views | Published:

Medical research

Personalized test tracks cancer relapse

Nature volume 545, pages 417418 (25 May 2017) | Download Citation

  • A Correction to this article was published on 26 July 2017

This article has been updated

Genomic analysis of lung-tumour evolution has been used to create personalized blood tests that enable successful clinical monitoring for early signs of cancer relapse — a promising step on the road to precision medicine. See Article p.446

The treatment of tumours that have spread, through a process called metastasis, to occupy locations beyond their primary site is usually decided on by the visual inspection of cells using an approach called histology, or from analysis of DNA sequences in a tumour fragment. Such tumour fragments are obtained either from what is thought to be the tumour's primary site or from its most accessible site. However, this approach does not accurately capture the intrinsic genetic heterogeneity of these malignancies, nor the ability of cancer genomes to evolve dynamically over time. This can mean that people with cancer can receive multiple types of therapy, often lasting years, if not decades, on the basis of what might be an outdated molecular analysis of their disease. Two articles, one on page 446 by Abbosh et al.1 and the other in The New England Journal of Medicine by Jamal-Hanjani et al.2, describe a clinical study that aimed to create a detailed genetic profile tracking how lung tumours evolve over time in single individuals. The results enabled Abbosh and colleagues to use patient-specific information to create a test that can identify whether relapse has occurred after surgical tumour removal.

Investigations of how tumours change over time are being revolutionized by a procedure called liquid biopsy3, which involves sequencing tumour DNA fragments, known as circulating tumour DNA (ctDNA), obtained from a blood sample. Cell-free fragments of DNA are shed into the bloodstream during cell-death processes called apoptosis and necrosis, or can enter the bloodstream by other mechanisms4. In their experiments, Abbosh and colleagues observed a relationship between their ability to detect ctDNA and the presence of cell necrosis, which is frequently found in some forms of non-small-cell lung cancer. Current advances in the sensitivity and accuracy of liquid biopsy are leading to clinical applications4,5. Abbosh et al. now offer a further advance in ctDNA analysis by testing whether liquid biopsies can be used to monitor the evolution of a cancer at an early stage of disease, and to identify relapse.

The clinical study reported in the current papers is called TRACERx, and is designed to study how non-small-cell lung cancers change over time. The authors began with a cohort of 100 individuals who underwent surgery to remove their tumours. However, efforts to remove all tumour cells are not always successful. Moreover, residual malignant cells might persist after surgery if a few tumour cells have already migrated to other locations in the body. If the patient relapses, additional treatment options such as chemotherapy can be considered.

The authors set out to track the evolution of individual tumours by testing ctDNA from blood samples obtained at different time points. However, the sensitivity of liquid biopsies has been a major challenge because tumour DNA can represent less than 1% of the DNA present in a blood-plasma sample3. To devise highly sensitive tests that can identify such a genetic 'needle in a haystack', Abbosh et al. used information obtained in Jamal-Hanjani and colleagues' analysis of surgically removed lung tumours. Jamal-Hanjani et al. sequenced the entire protein-coding region of the genome in several parts of these excised tumours and sequenced healthy tissue for comparison, to identify DNA changes, known as single nucleotide variants (SNVs), present specifically in the tumour.

The SNV profiles associated with different regions of a given tumour show both similarities and differences. Genetic mutations that occur early in tumour development are usually present in all tumour cells (clonal alterations), whereas mutations arising later occur in only a subset of cells (subclonal alterations). Analysis of the mutations in different tumour regions enabled Jamal-Hanjani and colleagues to infer the pattern of tumour genetic evolution and to construct phylogenetic trees showing the relationships between mutations (Fig. 1). From these trees, Abbosh et al. could identify an individual's clonal and subclonal SNVs and use this information to design liquid biopsies to check for cancer relapse. To improve the detection sensitivity of this technique, SNVs were analysed using a sequencing approach involving a technique called multiplex polymerase chain reaction, in which more than one SNV can be detected during sample analysis by liquid biopsy. The authors tested 10–22 NVs per person. The tumour SNVs detected mainly corresponded to early clonal SNVs, and the majority of the patients whose liquid biopsies contained clonal SNVs also had subclonal SNVs.

Figure 1: Monitoring lung-cancer evolution to detect early relapse after surgery.
Figure 1

In the TRACERx clinical study1,2, non-small-cell lung tumours were surgically removed from individuals, and samples were isolated from various regions of the tumour (indicated by the numbers 1–4). Using these samples, Jamal-Hanjani et al.2 sequenced protein-coding genomic regions, and identified DNA changes (shown as X symbols) that arose in these areas of the tumour. By analysing the patterns of DNA changes, the authors could determine the relationship between mutations. For example, early mutations are likely to be present in all tumour cells (clonal alterations), whereas later mutations occur in only some cells (subclonal alterations). The authors created a phylogenetic tree showing how an individual's tumour evolved. Abbosh et al.1 used the genetic profiles obtained by Jamal-Hanjani et al. to design a liquid-biopsy blood test for circulating tumour DNA (ctDNA) that could be used to spot the specific genetic hallmarks of cancer relapse appearing in an individual. This personalized testing approach identified signs of tumour recurrence a median of 70 days before relapse was identified through imaging scans. Such recurrence is often seen at different locations from the primary site of tumour formation.

This work provides an impressive level of insight into lung-cancer evolution, and is pioneering on several levels. Of the 100 individuals enrolled in the TRACERx study, Abbosh and colleagues followed 24 after their surgery, using both regular imaging surveillance and ctDNA testing of 12–30 SNVs. The median interval between identification of tumour-associated SNVs in post-operative liquid-biopsy analysis and the subsequent confirmation of relapse by computed tomography (CT) scans was 70 days. Although ctDNA analysis has previously enabled relapse to be predicted in breast and colorectal cancers6,7,8, Abbosh and colleagues' work was achieved through blinded, prospective analysis in a highly challenging setting. This context of early-stage disease, in which ctDNA levels are known to be extremely low, makes their findings particularly remarkable.

Results from the phylogenetic trees used for ctDNA analysis, or from additional analysis of tissue samples from individual patients, provided information that could potentially be acted on. For example, samples obtained from one patient who had relapsed were dominated by a subclone carrying an amplification of the ERBB2 gene, a cancer-promoting alteration that can be targeted with existing drugs. Furthermore, the authors found that when patients received chemotherapy, liquid-biopsy testing for tumour SNVs in ctDNA provided an indication of whether the treatment was successful. The potential to predict relapse at an early stage or to gain insights into which alterations are increasing in frequency suggest that ctDNA analysis can potentially anticipate cancer's next move, in a similar way to how an experienced chess player can often predict their opponent's next gambit.

This approach also helps in understanding how cancer heterogeneity can drive disease progression. For example, Abbosh and colleagues report the post-mortem analysis of one individual. A detailed profile of the patient's tumour recurrences and the SNVs associated with their relapse provided information on which subclones gave rise to metastatic tumours, and offered useful spatial and temporal insights into how the cancer evolved during treatment.

Another important finding by Abbosh and colleagues was that a higher frequency of mutant SNVs present in ctDNA (known as the variant allele frequency, or VAF) was associated with a higher tumour volume as determined from CT scans. A VAF value of 0.1% was correlated with a tumour volume that the authors calculated contains approximately 300 million cells. Applying this parameter might help investigators performing similar extrapolations to assess tumour size in future liquid-biopsy studies of lung or other tumour types.

Although this study demonstrates the power of liquid biopsies to track cancer over time, the approach is technically challenging, cumbersome and requires costly patient-specific assays. It is therefore not feasible for use in routine clinical practice at present. However, as sequencing costs continue to decrease, this could change, and such precision-medical treatment might become an everyday reality. The wealth of clinically relevant information about cancer that can be garnered today from genetic analysis of blood samples would have been considered science fiction only ten years ago.

The multidisciplinary team assembled for this project included, in addition to oncologists and radiologists, various non-medical specialists such as geneticists, statisticians, mathematicians and evolutionary biologists. These studies thus provide compelling evidence that 'cross-pollination' of different realms of expertise can enable major progress in the delivery of precision cancer care.

Notes

Change history

  • 20 July 2017

    An earlier version of this article spelled the ERBB2 gene name incorrectly.

References

  1. 1.

    et al. Nature 545, 446–451 (2017).

  2. 2.

    et al. N. Engl. J. Med. (2017).

  3. 3.

    & J. Clin. Oncol. 32, 579–586 (2014).

  4. 4.

    , , & Nature Rev. Clin. Oncol. (2017).

  5. 5.

    et al. Nature Rev. Cancer 17, 223–238 (2017).

  6. 6.

    et al. Clin. Cancer Res. 19, 2643–2650 (2014).

  7. 7.

    et al. Sci. Transl. Med. 7, 302ra133 (2015).

  8. 8.

    et al. Sci. Transl. Med. 8, 346ra392 (2016).

Download references

Author information

Affiliations

  1. Alberto Bardelli is in the Department of Oncology, University of Torino, 10060 Candiolo, Italy, and at the Candiolo Cancer Institute-FPO, IRCCS.

    • Alberto Bardelli

Authors

  1. Search for Alberto Bardelli in:

Corresponding author

Correspondence to Alberto Bardelli.

About this article

Publication history

Published

DOI

https://doi.org/10.1038/545417a

Further reading

Comments

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.

Newsletter Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing