Genotyping concordance in DNA extracted from formalin-fixed paraffin embedded (FFPE) breast tumor and whole blood for pharmacogenetic analyses

Abstract

Background: Cancer pharmacogenetic studies use archival tumor samples as a DNA source when germline DNA is unavailable. Genotyping DNA from formalin-fixed paraffin embedded tumors (FFPE-T) may be inaccurate due to FFPE storage, genetic aberrations, and/or insufficient DNA extraction. Our objective was to assess the extent and source of genotyping inaccuracy from FFPE-T DNA and demonstrate analytical validity of FFPE-T genotyping of candidate single nucleotide polymorphisms (SNPs) for pharmacogenetic analyses.

Methods: Cancer pharmacogenetics SNPs were genotyped by Sequenom MassARRAYs in DNA harvested from matched FFPE-T, FFPE lymph node (FFPE-LN), and whole blood leukocyte samples obtained from breast cancer patients. No- and discordant-call rates were calculated for each tissue type and SNP. Analytical validity was defined as any SNP with <5% discordance between FFPE-T and blood and <10% discordance plus no-calls.

Results: Matched samples from 114 patients were genotyped for 247 SNPs. No-call rate in FFPE-T was greater than FFPE-LN and blood (4.3% vs. 3.0% vs. 0.5%, p < 0.001). Discordant-call rate between FFPE-T and blood was very low, but greater than that between FFPE-LN and blood (1.1% vs. 0.3%, p < 0.001). Samples with heterozygous genotypes were more likely to be no- or discordantly-called in either tissue (p < 0.001). Analytical validity of FFPE-T genotyping was demonstrated for 218 (88%) SNPs.

Conclusions: No- and discordant-call rates were below concerning thresholds, confirming that most SNPs can be accurately genotyped from FFPE-T on our Sequenom platform. FFPE-T is a viable DNA source for prospective-retrospective pharmacogenetic analyses of clinical trial cohorts.

Keywords: Cancer; FFPE; Formalin-fixed paraffin embedded; Germline genome; Pharmacogenetics; Somatic genome.

Figure 1

Consort Diagram for Patients and SNPs. Quality control filtering of patients whose samples failed genotyping in blood or tumor (left) and SNPs that failed genotyping in blood (right). After quality control 114 patients with matched blood, FFPE‐T and FFPE‐LN samples genotyped for 247 independent SNPs were included in the analysis. Footnote: Abbreviations: MAF: Minor allele frequency, HWE: Hardye—Weinberg Equilibrium.

Figure 2

Genotyping for FFPE‐T. Concordance between FFPE‐T and blood for all patients (n = 114) and SNPs (n = 247) that passed quality control. (A) Discordant calls (blue) and no‐calls in either the tumor (light blue) or blood (white) are represented by colors. (B) The number of possible discordant calls (strict discordant calls plus no‐calls in tumor or blood) for all SNPs with stacked bars colored similarly to Figure 1A. The horizontal lines indicate cutoffs of 5%, 10%, and 20%. Overall, the tumor genotyping was highly concordant with few SNPs that failed FFPE‐T genotyping.

Figure 3

Strict discordance rates for FFPE‐T and FFPE‐LN. Histogram of the strict discordance rates in FFPE‐T and FFPE‐LN, each compared with blood, for all SNPs included in the analysis. Overall the discordance rates were quite low, with very few SNPs having more than 1% discordant calls between either tissue of interest and blood.