Skip to main content
Premium Trial:

Request an Annual Quote

Researchers Report on Analysis of Cancer Cell Line Transcriptomes

NEW YORK (GenomeWeb) – A group of researchers from Genentech have analyzed the transcriptomes of more than 675 commonly used human cancer cell lines, potentially paving the way for the development of new drugs.

As it reported in Nature Biotechnology today, the team led by Zemin Zhang amassed RNA sequencing and SNP array data from 675 cancer cell lines. From this, they were able to put together a picture of overall gene expression in these cell lines to identify novel gene fusion pairs, uncover previously unknown ways in which known oncogenes are regulated, and predict how certain cell lines may respond to treatment with particular inhibitors.

This, the researchers noted, will help enable the development of targeted therapeutics for cancer.

"The availability of this transcriptome and SNP array data can greatly enhance our understanding of drug response in clinically relevant models and thereby expedite the development of effective personalized medicine," Zhang and his colleagues wrote in their paper.

The researchers performed RNA sequencing using the Illumina HiSeq 2000 platform and SNP array analysis using the Illumina HumanOmni2.5_4v1 arrays on 675 commonly used cancer cell lines. They discarded some lines from further analysis, as they were genetically quite similar to other lines, despite their distinct names, and generated a median 61 million reads per sample, which they mapped and aligned to the human reference genome.

The Genentech team also compared its results to two previous studies of cancer cell lines, one from the Cancer Cell Line Encyclopedia and the other from the Sanger Institute. Some 276 lines were common to all three studies, though the study by Zhang's group included nearly 150 additional lines not present in either previous work. Gene expression between the overlapping lines was consistent, the researchers reported.

Zhang and his colleagues then began to sift through their data, both to confirm previous findings and search for new patterns.

For instance, they found recurrent copy number variations in cancer genes like MYC and ERBB2, and the loss of CDKN2A.

Additionally, they examined how the expression of various cancer genes correlated with one another. The expression of EGFR, EPHA2, AATGA3, and CAV2 are correlated with the expression of MET, the researchers reported. Expression of MET and EGFR, they added, was correlated in most tissue types, with the exception of pancreatic cancer cell lines and head and neck cancer cell lines.

To determine whether co-expression indicated co-regulation, the researchers set up a series of perturbation studies in which they targeted MET, EGFR, and their downstream effectors.

Inhibition of MET by short hairpin RNA or of EGFR by pharmacological inhibitors led to lower levels of not only MET and EGFR, but also of EPHA2 and ITGA.  Likewise, upregulation of MET and EGFR led to increased levels of those genes in cells with a PTEN deletion, but not in wild-type cells. This, the researchers said, indicates that PTEN typically repressed the regulation of these four genes.

To see whether this effect was linked to either the PI3K/AKT/mTOR or MAPK/ERK signaling pathways — which are both downstream of active MET and EGFR — the researchers treated cells harboring a PTEN deletion with either a PI3K inhibitor or a MEK inhibitor. Either treatment, they reported, led to reduced expression of the four genes, indicating that both pathways regulate those genes.

"[O]ur results argue that MET and EGFR signaling activates a previously undescribed positive-feedback pathway downstream of the PI3K and MEK signaling pathways, increasing their own expression as well as that of ITGA3 and EPHA2," Zhang and his colleagues said. They added that further mining of expression patterns in cancer cells could uncover additional patterns of regulation.

RNA sequencing of the cell lines also enabled the researchers to identify novel gene fusions.

The researchers found more than 2,000 unique pairs of gene fusions within these cell lines, 168 of which had previously been found. While 21 of the cell lines were models for canonical oncogenic fusions, the researchers also identified 11 cell lines that hadn't been previously been known to carry gene fusion. The RMPI 2650 cell line, for example, has a BRD4-C15orf55 fusion, which the researchers noted is found in aggressive midline carcinomas, a rare and hard-to-treat cancer.

This finding, they added, suggests that this cell line, and others like it that have newly been identified as harboring fusions, could be new systems in which to study such fusions.

They also identified some 17 known oncogenic fusion genes, including ALK, FGFR2, and RAF1, with previously unknown fusion partners. All the ALK fusions, both known and unknown, were sensitive to treatment with Xalkor (crizotinib), the researchers noted.

In addition, they uncovered a fusion in a glioma cell line that brought together EML4 and NTRK3. EML4 is a common fusion partner of ALK, and the researchers tested whether this fusion was functional by examining how glioma lines responded to crizotinib treatment. One line — G111, the one with the EML4-NTRK3 fusion — was sensitive, suggesting to Zhang and his colleagues that these fusion may represent a therapeutic target for glioma.

Zhang and his colleagues also combined the transcriptome data they generated on these cell lines with genomic information from them to try to bolster predictions of how cell lines will respond to certain therapeutics.

They integrated all the data on high-confidence mutations, copy number changes, and gene fusions for a set of pathways known to be altered in cancer. For each cell line, they scored the pathways as being either normal or aberrant.  They then examined how well these data-driven pathway predictions matched with how the cell lines responded to five targeted drugs candidates, two MEK inhibitors, two PI3K inhibitors, and an FGFR inhibitor.

From this, they noted that PI3K pathway aberrations were linked with response to PI3K inhibitor-treatment and MAPK pathway aberrations were similarly associated with MEK inhibitor treatment. A de-regulated PI3K pathway, the researchers noted, was linked to a worse response to a MEK inhibitor.

"With the rapidly expanding discovery and development of 'rationally targeted' therapeutics, it becomes critical to identify biomarkers predictive of clinical benefit," Zhang and his colleagues said. "We show how integration of various dimensions of genomic and transcriptomic data can improve the prediction of sensitivity to targeted therapeutics. Going forward, incorporation of the vast body of knowledge of pathway aberration into patient assessment will lead to more effective cancer treatment."