Comparison of multiple gene expression platforms for measuring a bladder cancer hypoxia signature.

Tumour hypoxia status provides prognostic information and predicts response to hypoxia­modifying treatments. A previous study by our group derived a 24­gene signature to assess hypoxia in bladder cancer. The objectives of the present study were to compare platforms for generating signature scores, identify cut­off values for prospective studies, assess intra­tumour heterogeneity and confirm hypoxia relevance. Briefly, RNA was extracted from prospectively collected diagnostic biopsies of muscle invasive bladder cancer (51 patients), and gene expression was measured using customised Taqman Low Density Array (TLDA) cards, NanoString and Clariom S arrays. Cross­platform transferability of the gene signature was assessed using regression and concordance analysis. The cut­off values were the cohort median expression values. Intra­ and inter­tumour variability were determined in a retrospective patient cohort (n=51) with multiple blocks (2­18) from the same tumour. To demonstrate relevance, bladder cancer cell lines were exposed to hypoxia (0.1% oxygen, 24 h), and extracted RNA was run on custom TLDA cards. Hypoxia scores (HS) values showed good agreement between platforms: Clariom S vs. TLDA (r=0.72, P<0.0001; concordance 73%); Clariom S vs. NanoString (r=0.84, P<0.0001; 78%); TLDA vs. NanoString (r=0.80, P<0.0001; 78%). Cut­off values were 0.047 (TLDA), 7.328 (NanoString) and 6.667 (Clariom S). Intra­tumour heterogeneity in gene expression and HS (coefficient of variation 3.9%) was less than inter­tumour (7.9%) variability. HS values were higher in bladder cancer cells exposed to hypoxia compared with normoxia (P<0.02). In conclusion, the present study revealed that application of the 24­gene bladder cancer hypoxia signature was platform agnostic, cut­off values determined prospectively can be used in a clinical trial, intra­tumour heterogeneity was low and the signature was sensitive to changes in oxygen levels in vitro.

Selection of endogenous control genes for normalising gene expression data derived from formalin-fixed paraffin-embedded tumour tissue.

Quantitative real time polymerase chain reaction (qPCR) data are normalised using endogenous control genes. We aimed to: (1) demonstrate a pathway to identify endogenous control genes for qPCR analysis of formalin-fixed paraffin-embedded (FFPE) tissue using bladder cancer as an exemplar; and (2) examine the influence of probe length and sample age on PCR amplification and co-expression of candidate genes on apparent expression stability. RNA was extracted from prospective and retrospective samples and subject to qPCR using TaqMan human endogenous control arrays or single tube assays. Gene stability ranking was assessed using coefficient of variation (CoV), GeNorm and NormFinder. Co-expressed genes were identified from The Cancer Genome Atlas (TCGA) using the on-line gene regression analysis tool GRACE. Cycle threshold (Ct) values were lower for prospective (19.49 ± 2.53) vs retrospective (23.8 ± 3.32) tissues (p < 0.001) and shorter vs longer probes. Co-expressed genes ranked as the most stable genes in the TCGA cohort by GeNorm when analysed together but ranked lower when analysed individually omitting co-expressed genes indicating bias. Stability values were < 1.5 for the 20 candidate genes in the prospective cohort. As they consistently ranked in the top ten by CoV, GeNorm and Normfinder, UBC, RPLP0, HMBS, GUSB, and TBP are the most suitable endogenous control genes for bladder cancer qPCR.