TOP-OCC and TOP-OCD

https://1000wgs.twbiobank.org.tw/search_result

23    Ongen, H., C. L. Andersen, J. B. Bramsen, B. Oster, M. H. Rasmussen, P. G. Ferreira, J. Sandoval, E. Vidal, N. Whiffin, A. Planchon, I. Padioleau, D. Bielser, L. Romano, I. Tomlinson, R. S. Houlston, M. Esteller, T. F. Orntoft, and E. T. Dermitzakis. Putative cis-regulatory drivers in colorectal cancer. (2014). Nature. pdf 4129.    The cis-regulatory effects responsible for cancer development have not been as extensively studied as the perturbations of the protein coding genome in tumorigenesis. To better characterize colorectal cancer (CRC) development we conducted an RNA-sequencing experiment of 103 matched tumour and normal colon mucosa samples from Danish CRC patients, 90 of which were germline-genotyped. By investigating allele-specific expression (ASE) we show that the germline genotypes remain important determinants of allelic gene expression in tumours. Using the changes in ASE in matched pairs of samples we discover 71 genes with excess of somatic cis-regulatory effects in CRC, suggesting a cancer driver role. We correlate genotypes and gene expression to identify expression quantitative trait loci (eQTLs) and find 1,693 and 948 eQTLs in normal samples and tumours, respectively. We estimate that 36% of the tumour eQTLs are exclusive to CRC and show that this specificity is partially driven by increased expression of specific transcription factors and changes in methylation patterns. We show that tumour-specific eQTLs are more enriched for low CRC genome-wide association study (GWAS) P values than shared eQTLs, which suggests that some of the GWAS variants are tumour specific regulatory variants. Importantly, tumour-specific eQTL genes also accumulate more somatic mutations when compared to the shared eQTL genes, raising the possibility that they constitute germline-derived cancer regulatory drivers. Collectively the integration of genome and the transcriptome reveals a substantial number of putative somatic and germline cis-regulatory cancer changes that may have a role in tumorigenesis.
—————–
A comparison of colorectal cancer and normal cells from 103 patients identifies dozens of genes that are differently expressed in tumour cells as a result of altered regulation of transcription.

How important for cancer incidence and progression is genetic variation that affects gene expression? This fundamental question has received remarkably little attention in recent studies of cancer genomes1, perhaps because of a prevailing view that the cancer-causing mutations that can be targeted by drugs are those that disrupt protein structure2. In a paper published on Nature’s website today, Ongen et al.3 demonstrate how simultaneous gene-expression profiling and whole-genome genotyping can be used to dissect the regulation of gene transcription in colorectal cancer3. The findings provide two thought-provoking insights: that cancer-driving changes may be identifiable among an excess of regulatory mutations, and that ‘cryptic’ regulatory genetic variation has the potential to modify cancer progression.

It is well established that gene expression is altered in cancer. Despite their independent derivation, tumours of the same type tend to converge on a common, new gene-expression profile. Various studies, primarily from The Cancer Genome Atlas project1, have noted differential transcription of tumour-driving and tumour-suppressing genes in advanced tumours, but so many gene transcripts are altered in these tumours that it is difficult to know which ones ‘drive’ the altered behaviour and which ones are ‘passengers’, just going along for the ride. Furthermore, epigenetic alterations — those that modify gene expression without involving sequence mutations — have been implicated in cancer, including colorectal cancer4, 5. Broad surveys of transcriptional and epigenetic changes in tumours have been conducted6, 7, but not on the scale and resolution achieved by Ongen and colleagues. They used a method known as RNA-Seq, in which the transcriptome of a cell (its whole complement of RNA molecules) is sequenced.

Over the past couple of years, sequencing of the exome of cancer cells (in essence, just the protein-coding regions) has suggested that around 250 genes are mutated in cancer cells significantly more often than expected by chance8, 9, 10. Many of these are pan-cancer genes, and some are tumour-type specific. It is less straightforward to perform similar analyses for regulatory DNA, for two reasons: we are just beginning to learn how to identify regulatory functions in the hundreds of kilobases that surround genes, and altered gene expression is often due to changes in genes located elsewhere in the genome. Ongen et al. overcome these limitations by focusing on changes in the ratio of expression of heterozygous alleles (sites at which the DNA sequence differs between the two copies of the sequence in a cell) between tumour and matched normal cells, as had also been done in another recent analysis of colorectal cancer11. They call the hundreds of instances of this phenomenon that they find per sample ‘genes with allelic dysregulation’ (GADs).

Although allele-specific expression can also be attributed to changes at other genes, it is highly likely that in many cases it is due to a locally acting regulatory mutation. Ongen et al. observe a significant correlation between the somatic (non-germline) mutation rate and altered allele-specific expression and, in each of the 103 matched normal–tumour pairs they analysed, approximately 200 transcripts showed a cancer-specific deviation in the allelic expression ratio at heterozygous sites. Their interpretation is that one allele is transcribed more than the other owing to the action of regulatory-sequence variation.

The authors show that some of this deviation can be attributed to familiar cancer-associated mutation types, including loss of heterozygosity and copy-number alteration, and that some is due to inferred (yet to be defined) regulatory mutations. Tallying these instances over all of the samples, and taking two approaches to controlling for statistical biases, they arrive at a list of 71 GADs that occur more frequently in tumours than in normal cells, 9 of which are shared with an existing list of pan-cancer driver mutations8. These observations provide a smoking gun for the idea that regulatory mutations can drive cancer. Perhaps there is no need to distinguish them with their own name, but the term ‘GPS mutations’ comes to mind, because they are instructing driver mutations on what to do, but it is not altogether clear that the cancer cells would not still attain a tumorous state without their help — much like a satellite-based navigation system instructing a driver on how to get somewhere.

A related term, ‘back-seat driver’, has been invoked to describe another class of mutation that probably has a role in mediating cancer progression or metastatic spread, and that is conditional on the status of other driver mutations12. Ongen and colleagues’ second major contribution is to suggest that, in addition to GPS mutations in GADs, another important source of cancer regulation is cryptic genetic variation (Fig. 1). These are genetic variants that are not relevant under normal circumstances, but become so only in a perturbed state13. They may play a key part in modifying the expression of cancer-driving genes. Specifically, when the authors looked for common variants that associate with differences in gene expression among individuals, they found that at least one-third of the expression-regulating variants (eSNPs) are tumour specific.
Figure 1: Cryptic regulation in tumour cells.
Cryptic regulation in tumour cells.

a, In normal cells, the two copies (alleles) of a gene in which the alleles are heterozygous, meaning that they contain a different nucleotide at a specific site (here, A and C), are both expressed at the same, low level. In this example, the alleles in the transcript are associated with a normally irrelevant, or ‘cryptic’, variant allele in a regulatory region of the gene (A with T and C with G). b, In a tumour cell that is characterized by enhanced expression of a transcription factor, this cryptic variant becomes relevant, because the transcription factor can bind to the T-containing allele but not to one containing G. This leads to enhanced transcription of the gene containing the A allele relative to that containing C, and thus an altered allelic-expression ratio.

    Full size image (104 KB)

Furthermore, these polymorphisms are enriched for binding sites for six known cancer-related transcription factors, all of which are upregulated in colorectal cancer. The idea is that when one of these factors (IRX3, E2F4, NFIL3, TFAP2A, CUX1 or LEF1) is in excess, polymorphisms that in normal cells do not influence the expression of the adjacent gene become relevant. Whether or not these are key back-seat modifiers of cancer progression is unclear, because there is only a mild enrichment for these polymorphisms in a genome-wide association study (GWAS) of colorectal cancers14. Perhaps they would be more enriched in a GWAS that assessed tumour progression.

Two words of caution about this study are in order. The first is that there has been no attempt to demonstrate functionality of any of the candidate mutations — the findings are all based on statistical association. The second is that anyone using RNA-Seq quickly realizes that there are many points at which the analysis can provide divergent results. Given recent interest in the repeatability of findings, it could be argued that it would be a good idea for journals to require independent parallel analyses from a different group, conducted blind, to corroborate such results before publication. This suggestion is not made to denigrate the careful and insightful analyses reported by Ongen et al., but is rather a generic comment on the inherent complexity of RNA-Seq, GWASs and enrichment analyses. Different analysts are likely to find quite different details. Yet the prospect that acquired variants drive cancer by controlling gene expression against a background of cryptic regulatory modifiers opens up a new perspective on cancer research. Similar analyses can now be performed on other data sets, and also on diseases other than cancer in which regulation of gene expression is altered15, 16. The next challenge is to establish the clinical utility of the identified regulatory variation.
References

    References•
    Author information

    The Cancer Genome Atlas Research Network. Nature Genet. 45, 1113–1120 (2013).
        Article
        Show context
    Hopkins, A. L. & Groom, C. R. Nature Rev. Drug Discov. 1, 727–730 (2002).
        Article
        Show context
    Ongen, H. et al. Nature http://dx.doi.org/10.1038/nature13602 (2014).
        Show context
    Bogaert, J. & Prenen, H. Ann. Gastroenterol. 27, 9–14 (2014). .