| dc.description.abstract | Background: The five year overall survival rate for colorectal cancer (CRC) patients
varies between 38.8% and 59.9%. Selecting patients who are likely to respond to
therapy remains a clinical and pathological challenge, hence the need for predictive
and prognostic biomarkers. The objectives of this study were: 1) to establish which
genes were differentially expressed with respect to sensitivity to treatment, 2) to
integrate the list of differentially expressed genes with copy number to systematically
identify predictive biomarkers, and 3) to establish which genes are commonly gained
in the panel of CRC cell lines. As proof of concept of the approach the copy number
variations of the identified genes were assessed in a cohort of Dukes’ A and B
cancers, in order to analyse the likelihood of these genes acting as useful biomarkers.
Methods: Cell viability assays were carried out on a panel 15 CRC cell lines. IC50s
were measured for 5-fluoruracil (5-FU), oxaliplatin (L-OHP), and BEZ-235, a
PI3K/mTOR inhibitor. We carried out a systematic array-based survey of gene
expression and copy number variation in CRC cell lines, and compared these to
responses to different treatments. Cell lines were profiled using array comparative
genomic hybridisation (aCGH; NimbleGen 135k), Illumina gene expression analysis,
reverse phase protein arrays (RPPA), and targeted sequencing of KRAS hotspot
mutations. The associations between the biological variables and drug sensitivity
were assessed using correlation coefficients, chi-square analysis, and the Mann
Whitney-U test. Tissue microarrays (TMA) were constructed for a cohort of CRC
patients (n=118) and TRIB1 and MYC amplifications were measured using
fluorescence in situ hybridisation (FISH). The protein expression for trib1 and 14
associated biomarkers were investigated using Automated Quantitative Analysis
(AQUA) and analysed using the Pearson’s correlation coefficient.
Results: Twenty-three regions were frequently gained, and fourteen regions were
lost across the cell line panel. Gains were observed at 2p, 3q, 5p, 7p, 7q, 8q, 12p,
13q, 14q, and 17q, and losses at 2q, 3p, 5q, 8p, 9p, 9q, 14q, 18q, and 20p. Frequently
gained regions contained EGFR, PIK3CA, MYC, SMO, TRIB1, FZD1, and BRCA2,
while frequently lost regions contained FHIT and MACROD2. Gene enrichment
analysis showed that differentially expressed genes with respect to treatment
response were involved in Wnt signalling, EGF receptor signalling, apoptosis, cell
cycle, and angiogenesis. Stepwise integration of copy number and gene expression
data yielded 47 candidate genes that were significantly correlated (corrected p-value
≤0.05). Differentially expressed genes common to all three treatment responses
included AEBP2, DDX56, MRPL32, MRPS17, MYC, NSMCE2, and TBRG4. TRIB1
(n=76) and MYC (n=81) were amplified (FISH score ≥1.8) in 14.5% and 7.4% of the
CRC cohort, respectively. TRIB1 and MYC amplifications were significantly
correlated (corrected p-value ≤ 0.0001). Trib1 protein expression in the patient
cohort was significantly correlated (corrected p-value ≤ 0.01) with protein expression
of pErk, Akt, and Caspase 3.
Conclusions: The CRC in-vitro model was used effectively in this study for
discovery of both predictive and prognostic biomarkers. A set of candidate
predictive biomarkers for 5-FU, L-OHP, and BEZ235 have been described, worthy
of further study. Amplification of the putative oncogene TRIB1 has been assessed for
the first time in a cohort of CRC patients. Inhibition of TRIB1 may be a synthetic
lethal approach when MYC amplifications are present, which requires further clinical
and experimental validation. | en |
