Mining the Archives: A Cross-Platform Analysis of Gene Expression Profiles in Archival Formalin-Fixed Paraffin-Embedded Tissues

Abstract

Formalin-fixed paraffin-embedded (FFPE) tissue samples represent a potentially invaluable resource for transcriptomic research. However, use of FFPE samples in genomic studies has been limited by technical challenges resulting from nucleic acid degradation. Here we evaluated gene expression profiles derived from fresh-frozen (FRO) and FFPE mouse liver tissues preserved in formalin for different amounts of time using 2 DNA microarray protocols and 2 whole-transcriptome sequencing (RNA-seq) library preparation methodologies. The ribo-depletion protocol outperformed the other methods by having the highest correlations of differentially expressed genes (DEGs), and best overlap of pathways, between FRO and FFPE groups. The effect of sample time in formalin (18 h or 3 weeks) on gene expression profiles indicated that test article treatment, not preservation method, was the main driver of gene expression profiles. Meta- and pathway analyses indicated that biological responses were generally consistent for 18 h and 3 week FFPE samples compared with FRO samples. However, clear erosion of signal intensity with time in formalin was evident, and DEG numbers differed by platform and preservation method. Lastly, we investigated the effect of time in paraffin on genomic profiles. Ribo-depletion RNA-seq analysis of 8-, 19-, and 26-year-old control blocks resulted in comparable quality metrics, including expected distributions of mapped reads to exonic, untranslated region, intronic, and ribosomal fractions of the transcriptome. Overall, our results indicate that FFPE samples are appropriate for use in genomic studies in which frozen samples are not available, and that ribo-depletion RNA-seq is the preferred method for this type of analysis in archival and long-aged FFPE samples.

Keywords: FFPE; RNA-seq; archival RNA; biorepositories; microarray; toxicogenomics.

FIG. 1.

Differential gene expression. Expression profiles following exposure to furan (8 mg/kg/day) in FRO, 18 h in formalin (FFPE-18 h), or 3 weeks in formalin (FFPE-3 weeks) liver samples evaluated using different DNA microarray and RNA-seq protocols. (A) Correlation analysis of DEG fold changes for FRO versus FFPE-18 h (top row) and FRO versus FFPE-3 weeks (bottom row) (P < 0.0001 for all linear regressions). Genes were significantly changed in at least one list (FDR P < 0.05, fold change > ± 1.5 compared with control). (B) Hierarchical clustering of all DEGs (FDR P < 0.05, fold change > ± 2).

FIG. 2.

Pathway analysis for (A) ribo-depletion RNA-seq, (B) poly-A-enrichment RNA-seq, (C) 1-color microarrays, and (D) 2-color microarrays. The top 15 pathways for the FRO groups (blue) are listed vertically. The level of enrichment of the corresponding FFPE groups (white-FFPE-18 h; grey-FFPE-3 weeks) is plotted. Pathway significance was calculated in IPA using a Fisher Exact test; −log(P-value) = 1.3 corresponds to a P = 0.05 (indicated in red). Venn diagrams indicate the overlap of each full list. For all analyses, only mapped DEGs with FDR P < 0.05 and fold change > ± 1.5 were used and only pathways with at least 4 DEGs were considered. Nrf2 Oxidative Stress Response was the only pathway that was enriched in all 12 groups (blue chevrons); p53 was enriched in 11/12 groups (grey chevrons). Full color version available online.

FIG. 3.

Examination of RNA-seq data quality metrics. (A–C) Mapping distributions and GC content distributions for the furan-treated sampled. (A) The percentage of reads that covers each nucleotide position of all genes scaled to 100 bins, from 5′ UTR to 3′ UTR for FRO and FFPE samples. The ribo-depleted RNA samples consistently showed more uniform gene coverage than did their poly-A–selected counterparts. (B) GC content distributions for FRO and FFPE samples. (C) The percentage of FRO and FFPE reads that map to various genic regions. Overall, we detected more intronic reads from ribo-depleted RNA samples than from poly-A–enriched libraries; this trend increased as a function of time in formalin. (D,E) Mapping distributions and GC content distributions, expressed as a percentage, for 8-, 19-, and 26-year-old control FFPE tissues. (F) Inter-group Kendell's tau comparison.