Molecular profiling from FFPE tissue: Stratification of the Stage II colorectal cancer population

C138 The majority of clinical tissue specimens in current clinical practice are processed and captured as formalin-fixed paraffin-embedded (FFPE) tissues. We have developed a range of microarray tools (Cancer DSATM ), based on the transcriptome of an individual disease, which enables the capture of gene expression data from FFPE specimens. We have used this tool to develop prognostic signatures in colorectal cancer and there are a number of cancer therapeutics for which gene signatures as markers of response are under investigation.
 Designed to maximise the expression data obtained from the chosen disease setting, these arrays enable us to generate significantly more information relevant to the disease of interest than that obtained with generic arrays. The unique content and design of these high-density microarrays, based on the Affymetrix Genechip system, means the array can be employed for novel molecular target discovery. We have analysed the ability of the DSA tools to retrieve data from FFPE tumour samples using 5 replicates of matched FFPE and RNA-later stored samples. >70% of probesets that were consistently detected in RNA-later samples were also detected from the matched FFPE tissue. 96% of probesets consistently present in FFPE are also consistently present in RNA-later samples demonstrating the reproducibility of the technology.
 One clinical application of this technology will also be presented: using this tool we have developed an accurate prognostic signature using FFPE tissues capable of predicting disease recurrence in patients with stage II colorectal cancer. Primary tumours from 143 patients with stage II CRC were collected from 7 different hospitals, in 3 countries (USA, UK, Ireland). Within the study group we defined 2 prognostic cohorts: a “poor prognostic cohort” that relapsed within 5 years post surgery and a “good prognostic cohort” that remained disease free. Patient samples were randomly divided to generate a training set of 93 and an independent test set of 50 samples. RNA expression data was analysed using PLS to identify transcripts that best discriminate between patients at high risk of relapse versus patients likely to remain disease free, post surgery.
 A 91-transcript signature was identified that was capable of stratifying the population into good- and poor-prognosis. The ability of the signature to predict recurrence was then evaluated in the independent test set of 50 samples. This gene signature stratifies this Stage II CRC patient population with a high accuracy and a NPV of 92%.
 Using our DSA technology and FFPE tumour samples we have derived and validated a prognostic signature that accurately stratifies stage II CRC patients according to their risk of recurrence post surgery. This demonstrates how the potential of clinical tumour banks of FFPE samples can be unlocked and used to stratify disease populations to better determine the utility of new cancer therapeutics.