Direct formalin fixation induces widespread transcriptomic effects in archival tissue samples

Abstract

Sequencing technologies now provide unprecedented access to genomic information in archival formalin-fixed paraffin-embedded (FFPE) tissue samples. However, little is known about artifacts induced during formalin fixation, which could bias results. Here we evaluated global changes in RNA-sequencing profiles between matched frozen and FFPE samples. RNA-sequencing was performed on liver samples collected from mice treated with a reference chemical (phenobarbital) or vehicle control for 7 days. Each sample was divided into four parts: (1) fresh-frozen, (2) direct-fixed in formalin for 18 h, (3) frozen then formalin-fixed, and (4) frozen then ethanol-fixed and paraffin-embedded (n = 6/group/condition). Direct fixation resulted in 2,946 differentially expressed genes (DEGs) vs. fresh-frozen, 98% of which were down-regulated. Freezing prior to formalin fixation had ≥ 95% fewer DEGs vs. direct fixation, indicating that most formalin-derived transcriptional effects in the liver occurred during fixation. This finding was supported by retrospective studies of paired frozen and FFPE samples, which identified consistent enrichment in oxidative stress, mitochondrial dysfunction, and transcription initiation pathways with direct fixation. Notably, direct formalin fixation in the parent study did not significantly impact response profiles resulting from chemical exposure. These results advance our understanding of FFPE samples as a resource for genomic research.

Conflict of interest statement

The authors declare no competing interests. The authors confirm that all methods were carried out in accordance with relevant guidelines and regulations.

Figures

Figure 1

Design of experiments to prospectively (Study 1) and retrospectively (Study 2) evaluate formalin effects. PB phenobarbital, DEHP di(2-ethylhexyl) phthalate, NBF neutral buffered formalin, FR frozen, FR > OH frozen first then fixed in 70% ethanol prior to paraffin-embedding, FR > FFPE frozen first then fixed in 10% NBF for 18–24 h prior to paraffin-embedding, FFPE directly fixed in 10% NBF for 18–24 h or 3 weeks prior to paraffin-embedding.

Figure 2

Formalin effects on gene expression by RNA-seq. (a) Principal component analysis on scaled, low-expression filtered, counts per million-normalized, log2-transformed gene counts altered by different preservation procedures. ● indicates vehicle control (Con) samples. ▲ indicates phenobarbital (PB)-treated samples. Different colors designate the preservations: frozen control (FR), 70% ethanol-fixed (FR > OH), frozen first then 18-h formalin-fixed (FR > FFPE), and 18-h formalin-fixed (FFPE) liver tissue samples. FR > OH, FR > FFPE, and FFPE samples were all paraffin-embedded. (b) Overlap in significant differentially expressed genes between different preservation groups relative to the paired frozen controls. Significant differentially expressed genes identified by FDR-adjusted p-value < 0.05 and fold-change cutoff ± 2 using Partek Flow Gene Specific Analysis.

Figure 3

Effects of formalin fixation on detection of chemical treatment responses. (a) Venn diagram showing overlap in significant phenobarbital (PB)-induced differentially expressed genes (DEGs, FDR-adjusted p-value < 0.05, ± twofold-change) across preservation groups: frozen control (FR), 70% ethanol-fixed (FR > OH), frozen first then 18-h formalin-fixed (FR > FFPE), and 18-h formalin-fixed (FFPE) liver tissue samples. FR > OH, FR > FFPE, and FFPE samples were all paraffin-embedded. (b) Expression pattern of top 10 PB-induced differentially expressed genes. Across preservations, FR > OH, FR > FFPE, and FFPE resulted in significant reductions of counts per million (CPM) gene counts for all preservations relative to frozen control (FR). Error bars represent standard error. Significant differences between preservation group compared to frozen were determined by two-tailed Wilcoxon signed-rank test with Holm correction. **p-value = 0.0029, *p-value = 0.01 (c) Significant upstream regulators (Right-Tailed Fisher’s Exact Test, p-value < 0.05) predicted from PB-induced DEG patterns across preservation groups. « indicates CAR, PXR, and PB. (d) Z-scores of upstream regulators matched to significant p-values with positive z-scores predicting activation and negative scores predicting repression of regulator. « indicates CAR, PXR, and PB.