Targeted high throughput sequencing in clinical cancer settings: formaldehyde fixed-paraffin embedded (FFPE) tumor tissues, input amount and tumor heterogeneity

Abstract

Background: Massively parallel sequencing technologies have brought an enormous increase in sequencing throughput. However, these technologies need to be further improved with regard to reproducibility and applicability to clinical samples and settings.

Methods: Using identification of genetic variations in prostate cancer as an example we address three crucial challenges in the field of targeted re-sequencing: Small nucleotide variation (SNV) detection in samples of formalin-fixed paraffin embedded (FFPE) tissue material, minimal amount of input sample and sampling in view of tissue heterogeneity.

Results: We show that FFPE tissue material can supplement for fresh frozen tissues for the detection of SNVs and that solution-based enrichment experiments can be accomplished with small amounts of DNA with only minimal effects on enrichment uniformity and data variance.Finally, we address the question whether the heterogeneity of a tumor is reflected by different genetic alterations, e.g. different foci of a tumor display different genomic patterns. We show that the tumor heterogeneity plays an important role for the detection of copy number variations.

Conclusions: The application of high throughput sequencing technologies in cancer genomics opens up a new dimension for the identification of disease mechanisms. In particular the ability to use small amounts of FFPE samples available from surgical tumor resections and histopathological examinations facilitates the collection of precious tissue materials. However, care needs to be taken in regard to the locations of the biopsies, which can have an influence on the prediction of copy number variations. Bearing these technological challenges in mind will significantly improve many large-scale sequencing studies and will - in the long term - result in a more reliable prediction of individual cancer therapies.

Figure 1

Comparison of FFPE and snap frozen tissue material for whole exome re-sequencing approaches. (A) Exonwise coverage comparison of snap frozen and FFPE DNA preparations for a benign tissue sample. Coefficients of variation are calculated exonwise and plotted by the smallest coverage of each exon-exon comparison. As reference, coefficients of variation were calculated for two sequencing replicates of the snap frozen preparation (B) Mean coverage by GC content for snap frozen and FFPE DNA preparations. All exons were split into 800 bins by GC content and the average exon coverage was averaged within each bin. (C) Comparison of SNVs and InDels detected in snap frozen and FFPE DNA preparations for a benign tissue sample. False negative SNVs/InDels are detected in the snap frozen preparation but not in the FFPE preparation. False positive SNVs/InDels are detected in the FFPE preparation but not in the snap frozen preparation. (D) Comparison of SNVs detected in snap frozen and FFPE DNA preparations with Affymetrix SNP array 6.0 plotted by minimal coverage.

Figure 2

Different DNA amounts for targeted re-sequencing approaches. (A) Exonwise coverage comparisons obtained with different amounts of input DNA. Coefficient of variations were calculated for each comparison and plotted by the smallest coverage of each exon-exon comparison. (B) Variant/Reference ratio distribution for different amounts of input DNA. Depicted is the density curve for each preparation and distribution. (C) Comparison of SNVs detected with different amounts of input DNA. The Y-axis depicts the percentage of foci in concordance for different preparations at different coverage levels. (D) Comparison of InDels detected with different amounts of input DNA. The y-axis depicts the percentage of foci in concordance for different preparations at different coverage levels.

Figure 3

Comparison of different tumor biopsies for targeted re-sequencing approaches. (A) Normalized coverage-distribution plot for two foci of each of three tumor tissues. The mean coverage per exon was divided by the overall mean coverage of all exons and plotted as normalized coverage (x-axis). The fraction of bait-covered exons in the genome achieving coverages equal or lower than the overall mean coverage is indicated on the y-axis. (B) Exonwise coverage comparison of two foci of each of three tumor tissues. A coefficient of variation is calculated for each comparison and plotted by the smallest coverage of each exon-exon comparison. (C) Comparison of SNVs detected in two foci of each of three tumors. Discordant SNVs are those detected in focus A but not focus B and vice versa. (D) Comparison of InDels detected in two foci of each of three tumors. Discordant InDels are those detected in focus A but not focus B and vice versa.

Figure 4

Comparison of copy number profiles of different tumor biopsies. DNA read frequencies and subsequent normalized log ratios for tumor versus normal were determined for chromosomal intervals (bins) of 55-190 Kb. Copy number changes were calculated as running median of the log ratios of 20 bins. Differences in copy number between the two foci of one tumor are depicted as the difference of the two running median vectors. Differences greater or equal 0.2 were highlighted in magenta.