Gene expression profiles from formalin fixed paraffin embedded breast cancer tissue are largely comparable to fresh frozen matched tissue

Abstract

Background and methods: Formalin Fixed Paraffin Embedded (FFPE) samples represent a valuable resource for cancer research. However, the discovery and development of new cancer biomarkers often requires fresh frozen (FF) samples. Recently, the Whole Genome (WG) DASL (cDNA-mediated Annealing, Selection, extension and Ligation) assay was specifically developed to profile FFPE tissue. However, a thorough comparison of data generated from FFPE RNA and Fresh Frozen (FF) RNA using this platform is lacking. To this end we profiled, in duplicate, 20 FFPE tissues and 20 matched FF tissues and evaluated the concordance of the DASL results from FFPE and matched FF material.

Methodology and principal findings: We show that after proper normalization, all FFPE and FF pairs exhibit a high level of similarity (Pearson correlation >0.7), significantly larger than the similarity between non-paired samples. Interestingly, the probes showing the highest correlation had a higher percentage G/C content and were enriched for cell cycle genes. Predictions of gene expression signatures developed on frozen material (Intrinsic subtype, Genomic Grade Index, 70 gene signature) showed a high level of concordance between FFPE and FF matched pairs. Interestingly, predictions based on a 60 gene DASL list (best match with the 70 gene signature) showed very high concordance with the MammaPrint® results.

Conclusions and significance: We demonstrate that data generated from FFPE material with the DASL assay, if properly processed, are comparable to data extracted from the FF counterpart. Specifically, gene expression profiles for a known set of prognostic genes for a specific disease are highly comparable between two conditions. This opens up the possibility of using both FFPE and FF material in gene expressions analyses, leading to a vast increase in the potential resources available for cancer research.

Figure 1. Unsupervised hierarchical clustering with the most informative 5444 probes of all FFPE samples (n = 45) (A) and all FF samples (n = 44) (B).

In both cases (A and B) the samples cluster together with their biological and technical (if present) replicates. We employed simple scaling normalization without background subtraction and log2 transformed the data. The replicates are color coded. On the left side of the dendogram is reported the Pearson correlation (from 0.6 to 1).The samples are indicated with the FFPE_ID or FF_ID and the Pair_ID combined. FFPE = Formalin Fixed Paraffin Embedded, FF = Fresh Frozen.

Figure 2. Heat maps of the distance matrix representing the pairwise distances between all FFPE and FF samples (n = 89) using all probes that passed the p-value filter (n = 21178).

Prior to the computation of the distance matrix (where distance = 1-Pearson correlation), data were normalized with a simple scaling normalization without background correction and then median centered per probe, separately for FFPE and FF samples. The paired samples are indicated with a number from 1 to 20 and are color coded. The tissue type is indicated with the vertical bar (black = FFPE, white = FF). (A) Heat map using median centered log2 normalized data. (B) Heat map using non-median centered log2 normalized data.

Figure 3. Box plots of the Pearson correlation coefficient between FF replicates (n = 44), FFPE replicates (n = 45) and FFPE/FF pairs (n = 44).

The box plots were generated using all probes that passed the p-value filtering (all) or using the informative probes (highvar). The data were normalized across the arrays with a simple scaling normalization and median centered per probe. The outlier samples are indicated in the graph with the array_ID (see Table 1). FF_replicates =  FF replicates, FF_random =  FF random replicates, FFPE_replicates =  FFPE replicates, FFPE_random =  FFPE random replicates, FFPEvsFF_pairs =  FFPE-FF paired samples, FFPE vs FF_random_pairs =  random FFPE-FF pairs.

Figure 4. Percentage overlap between the two-, three- and four-fold regulated probes in FFPE and FF paired samples for the up-regulated probes (A) and down-regulated probes (B).

Figure 5. Comparison of the 60 gene index as derivative of the 70 gene prognosis signature of the FFPE samples to the 60 gene index of the paired FF samples.

Scatterplot of the Pearson correlation of the 60 genes from each sample to the average expression profile of the good outcome patients as described in , . This correlation is defined as the 60 gene index. The circles are colored based on the diagnostic results for these patients of the customized MammaPrint 8-pack test : red = high risk outcome, black = low risk outcome.

Figure 6. Distribution of the probewise correlation between FFPE and FF matching pairs.

The x-axis reports the probewise correlation between FFPE and FF matching pairs, the y-axis the corresponding number of probes. The 1000 and 500 most and least correlating probes are highlighted in white and blue respectively.

Figure 7. Density distribution of the GC content (%) in the top 500 most highly correlating and bottom 500 least correlating probes.

Below the x axis the statistical significance based on the T-test and Wilcoxon signed-rank test is shown. Similar density functions were observed for the top 1000 and bottom 1000 probes.