Testing and Validating High Throughput Genomic Analysis of FFPE Samples: Systematic Gene Expression Differences between Matched Frozen and FFPE Samples Due to Rapid mRNA Degradation Signals.
Rafael Cabal, Raya Khanin, Agnes Viale, Nicholas Socci, Robert Lucito, Adam Olshen, Jack Zhang, William Gerald, Scott Powers, Marc Ladanyi. Memorial Sloan Kettering Cancer Center, New York, NY; Cold Spring Harbor Laboratory, Cold Spring Harbor, NY; University of California, San Francisco
Background: To assess the performance of current platforms in comprehensive genomic profiling of formalin-fixed paraffin-embedded (FFPE) samples, we used duplicate frozen and matched FFPE tumor samples of 5 lung adenocarcinomas (total=20 samples) to evaluate array-based and next generation sequencing-based DNA and RNA profiling approaches.
Design: For genomic copy number profiling, we compared Affymetrix SNP 6.0 arrays, Affymetrix MIP arrays, Agilent 1M arrays, and ABI SOLiD low coverage sequencing. For transcriptional profiling, we evaluated Illumina whole genome DASL arrays, Affymetrix Exon arrays, and ABI SOLiD transcriptome sequencing, using Affymetrix U133Plus 2.0 array data from the frozen tumors as a reference.
Results: Agilent 1M arrays and ABI SOLiD low coverage sequencing performed best by different measures including noise, replicate concordance, and matched frozen-FFPE concordance. ABI SOLiD sequencing showed better reproducibility of frozen vs FFPE replicates than any of the array platforms. ABI SOLiD sequencing (even low coverage) outperformed various array platforms in both transcriptional and genomic copy number profiling on FFPE samples. Correlations between expression data from frozen samples and FFPE samples were worse than any cross-platform correlations using the same sample type, suggesting that there are systematic gene expression differences between matched frozen and FFPE samples. We performed an unbiased search of 3'UTRs for 7-9 bp motifs that found an AU-rich motif (AAAUAUU) to be significantly associated (r=0.73) with gene expression differences between FFPE and frozen. Similarly, the average number of the canonical AUUUA mRNA degradation signals showed a significant linear relationship (r=0.78) to fold-changes between FFPE and frozen.
Conclusions: ABI SOLiD sequencing outperforms various array platforms in both transcriptional and copy number profiling on FFPE samples. AU-rich motifs are significantly over-represented in mRNAs that are downregulated in FFPE vs frozen samples, suggesting a possible relationship to AU-rich rapid mRNA degradation signals. This result suggests a biological, rather than technical, basis for the discordance between expression profiles of FFPE vs frozen samples.
Category: Special Category - Pan-genomic/Pan-proteomic approaches to Cancer
Tuesday, March 1, 2011 8:00 AM
Platform Session: Section H 1, Tuesday Morning