Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.


Genetic clues can be used to predict whether early-stage cancer will form an invasive tumour

A type of non-invasive cancer called carcinoma in situ (CIS) can occur in the human lung. Some cases of CIS will progress to form an invasive type of cancer known as lung squamous cell carcinoma (LUSC), but until now no method had been developed that could reliably identify which CIS growths would progress. Writing in Nature Medicine, Teixeira et al.1 report their analyses of CIS samples from human lung tissue and the identification of a set of genomic alterations that can be used to predict whether CIS is likely to progress to form an invasive tumour.

Biopsy sampling of CIS growths is possible during bronchoscopy surveillance of patients’ lungs. Teixeira and colleagues used such biopsies to study the development of LUSC by monitoring CIS growths over time using imaging and by taking cellular samples of a given CIS at different time points. In the people they studied, a subset of the CIS growths either progressed to form LUSC or regressed and regained a normal appearance (Fig. 1). The authors focused on 129 CIS biopsy samples that had been obtained from 85 people before visible signs of progression or regression had been detected. They performed a range of genomic analyses on different subsets of these samples, including whole-genome DNA sequencing, analysis of RNA expression and profiling of a DNA modification called methylation that can influence gene expression.

Figure 1 | Hallmarks of invasive lung tumours. a, A type of abnormal cellular growth called carcinoma in situ (CIS) can occur in the human lung. b, c, Lung CIS can regress or it can progress to form a malignant invasive cancer known as lung squamous cell carcinoma (LUSC). It is hard to predict which of these paths a CIS will follow, and this has implications for decisions on clinical treatment. Teixeira et al.1 report their analysis of CIS samples taken from patients over time. They report that patterns of gene expression, patterns of a type of DNA modification called methylation (Me) and alterations in chromosomal structure, such as deletions of DNA sequences, could be used to predict whether a CIS would progress to form an invasive cancer. The authors report that a CIS can regress despite the presence of mutations in cancer-associated genes, and that a mutation in the gene TP53 is a common hallmark of CIS growths that progress.

In the whole-genome DNA-sequencing analysis of 29 samples from individuals whose CIS progressed to LUSC and of 10 regressive CIS samples, the authors found that, overall, the progressive samples had significantly more mutations and more alterations in the number of copies of some genes than the regressive samples had. The most striking finding was that, unlike the regressive samples, almost all of the progressive samples had mutations in the gene TP53 — a tumour-suppressor gene that helps to prevent the development of cancer. In addition, the progressive samples had a distinct pattern of chromosomal amplifications and deletions of sequences that are commonly found in squamous-cell carcinomas2 (tumours that originate from cells in tissues that line internal body cavities). The CIS samples that regressed generally lacked notable chromosomal aberrations. Remarkably, of four CIS growths that had TP53 mutations, many copy-number alterations and that were initially classified as regressive by visual monitoring of the lungs, three subsequently progressed.

Using a statistical method of data analysis called principal component analysis, Teixeira and colleagues demonstrated that the DNA-methylation patterns in most of the regressive CIS samples were more similar to those of normal lung cells than to the patterns in progressive tumours, except for the cases in which tumours that were initially classified as regressive later progressed. The authors’ method could distinguish between regressive and progressive CIS samples using either data for DNA methylation or gene copy number, and the results were consistent regardless of which of these two types of data were used. This indicates that the methylation patterns characteristic of progressive CIS might be influenced by underlying mutations and changes in gene copy number that are associated with progressive CIS.

The authors developed a profile of gene-expression signatures, sets of specific methylated DNA sequences and copy-number alterations for specific genes that could be used to accurately determine the probability of CIS progression. They confirmed the predictive value of their approach by analysing CIS biopsies that had previously been set aside to validate their results. The authors also tested their predictive system by analysing DNA-sequence data from human lung samples obtained from The Cancer Genome Atlas (TCGA) project. They found that their approach could distinguish most LUSC tumours (those likely to be similar to progressive CIS) from the control samples of normal lung tissue (which would be more similar to regressive CIS). Many of the genes for which there were differences in expression and DNA-methylation patterns between progressive and regressive CIS are involved in lung cancer. Teixeira et al. also observed differences in expression and methylation of genes associated with chromosomal instability, which is characterized by structural alterations such as large insertions or deletions of DNA.

An important unanswered question concerns how lung CIS growths progress to become LUSC. The authors can identify which CIS growths will progress, but do not know which changes are essential for an invasive cancer to form. The mutation frequencies of key cancer-driver genes in the people who had progressive CIS in Teixeira and colleagues’ study are largely similar to the mutation frequencies of these genes in the LUSC tumours in the TCGA database. However, there are some notable differences. For example, a tumour-suppressor gene called KMT2D is reported3 to have a significant association with LUSC, but Teixeira and colleagues found fewer cases of inactivating mutations in this gene associated with progressive CIS than were associated with LUSC tumours in the TCGA database. This discrepancy raises the question of whether KMT2D inactivation is a crucial step in the development of LUSC.

In addition to mutations of individual genes, another type of DNA alteration that can be a driver of LUSC and of other cancers3 are small regional (focal) alterations to chromosomes that increase the number of cancer-promoting genes or delete tumour-suppressor genes. Teixeira et al. report broad patterns of gene copy-number changes, but don’t report on focal alterations, and so whether this type of alteration is involved in the acquisition of invasive characteristics by CIS is unknown. An issue for future study is which molecular mechanisms drive the regression of lung CIS, because some genetic alterations characteristic of CIS were still present in the growths that regressed.

The evolution of CIS to form an invasive carcinoma has also been studied in breast cancer. In general, CIS growths in the ductal region of the breast have many of the same types of alteration found in invasive breast tumours4. Intriguingly, mutations known to promote the development of a skin cancer called melanoma and of oesophageal cancer have been identified57 in samples of normal skin and oesophagus, respectively. Together with Teixeira and colleagues’ results, these findings confirm that the accumulation of all the genomic alterations required for invasive cancer to develop takes time, and they suggest that we might all be carrying the burden of cancer-associated mutations in our seemingly healthy tissues.

Teixeira and colleagues’ study sample size was too small for rare genomic alterations to be detected. It would therefore be useful to follow it up with a larger study, which might enable the identification of other genomic alterations involved in the progression of lung CIS to LUSC.

As well as providing the first reported whole-genome-sequencing analysis of lung CIS, Teixeira and colleagues’ work offers a glimpse of a future in which precancerous growths that look similar under the microscope could be evaluated on a molecular level to accurately estimate the likelihood that they will develop into invasive cancer. This would enable physicians to tailor therapeutic interventions to the probability of disease progression.

Nature 566, 336-337 (2019)



  1. 1.

    Teixeira, V. H. et al. Nature Med. (2019).

    Article  Google Scholar 

  2. 2.

    Hoadley, K. A. et al. Cell 158, 929–944 (2014).

    PubMed  Article  Google Scholar 

  3. 3.

    Campbell, J. D. et al. Nature Genet. 48, 607–616 (2016).

    PubMed  Article  Google Scholar 

  4. 4.

    Casasent, A. K., Edgerton, M. & Navin, N. E. J. Pathol. 241, 208–218 (2017).

    PubMed  Article  Google Scholar 

  5. 5.

    Martincorena, I. et al. Science 348, 880–886 (2015).

    PubMed  Article  Google Scholar 

  6. 6.

    Martincorena, I. et al. Science 362, 911–917 (2018).

    PubMed  Article  Google Scholar 

  7. 7.

    Yokoyama, A. et al. Nature 565, 312–317 (2019).

    PubMed  Article  Google Scholar 

Download references


Nature Careers


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


Quick links