Abstract

Microarray-based expression profiling has identified prognostic gene signatures for many cancers and validation is required in clinical samples. However, most clinical material is in the form of formalin fixed and paraffin embedded tissue (FFPET) in which gene expression analysis is problematic. We have developed a generic quantum dot (QD) based multiplexed in-situ hybridization (ISH) method enabling quantitative localization of multiple mRNA targets in FFPET. We expand on our previous work introducing a method for standardization of ISH signal, enabling comparative measurement of gene expression across multiple samples. This was applied to tissue microarrays (TMAs) using archived trephine biopsies from patients with acute myeloid leukaemia (AML) to identify prognostic genes. A total of 15 TMAs were prepared using FFPET samples from 240 patients with AML diagnosed and treated between 1994 and 2005 at Manchester Royal Infirmary (Manchester, UK). For the analysis, 192 patients were included as the remainder either died before, during or immediately after one course of chemotherapy or there was incomplete data collection. The median age was 52 years (range 17–77) and all patients received intensive chemotherapy according to standard UK MRC AML protocols. Three cores were taken from each sample for TMA preparation. A standard was prepared using a cell pellet obtained from whole blood white cells which was embedded, in triplicate, in each TMA. QD-ISH was performed for nine genes recognized to be of prognostic value in AML. Triplex QD-ISH using QD labeled anti-sense cDNA oligonucleotides was performed for the following targets: Bcl2, survivin and XIAP; DNMT1, DNMT3A and DNMT3B; HOXA4, HOXA9 and Meis1. Signal intensity for each gene was measured using spectral imaging. Scrambled sense cDNA oligonucleotides were used to measure the level of background staining for each gene in each core. Background noise was corrected for by dividing expression levels of anti-sense probes by that of the scrambled probe, for both samples and standards. This enabled direct comparison between TMAs as gene expression values of samples were normalized against the standard. The mean expression of each gene was calculated for each patient, divided into quartiles and correlated with clinical outcome data. Statistical analysis was performed using contingency tables, the chi-square test and Mann Whitney-U. Overall survival (OS) and disease free survival (DFS) were displayed using the Kaplan-Meier method and Cox regression was performed for univariate and multivariate analysis. The OS in this cohort of patients was 43% at 5 years with 80% achieving complete remission (CR) after induction chemotherapy. Patient age (<60 years), WCC (<100×109/l) at diagnosis, cytogenetics (good and intermediate risk) and low HOXA4 expression (median 577 [95%CI 325–828]) were all associated with improved OS (p<0.0001; p=0.02; p<0.0001; p=0.013) and DFS (p<0.0001; p=0.013; p<0.0001; p=0.025) on univariate and multivariate analysis. High expression of HOXA9 (median 0.843 [0.145–7.479]; p<0.0001) and DNMT3A (median 1.305 [0.073–5.477]; p=0.04) were associated with failure to achieve CR. High Meis1 expression was found to be of borderline significance for poor response to chemotherapy (median 0.716 [0.051–7.840]; p=0.05). Expression levels of the remaining 5 genes did not show any correlation with CR, DFS or OS. These findings are consistent with recently published data regarding the prognostic significance of various new markers. In line with others we have demonstrated low expression of HOXA4 is an independent good prognostic marker in adult AML. Although high expression of HOXA9, DNMT3A and Meis1 was associated with inferior CR rates in our study, OS and DFS were not adversely affected. This may be related to improvements and more aggressive clinical practice (eg stem cell transplantation) over recent years which can overcome potential deleterious gene effects. These results demonstrate that the application of a standardized, quantitative multiplex QD-ISH can be used for identification of prognostic markers in FFPET samples. The advantages of this method include its application to TMAs which allows high sample throughput, use of archived materials and its transferability across a spectrum of malignancies.

Disclosures: No relevant conflicts of interest to declare.

Author notes

Corresponding author