Functional DNA quantification guides accurate next-generation sequencing mutation detection in formalin-fixed, paraffin-embedded tumor biopsies

Abstract

The formalin-fixed, paraffin-embedded (FFPE) biopsy is a challenging sample for molecular assays such as targeted next-generation sequencing (NGS). We compared three methods for FFPE DNA quantification, including a novel PCR assay ('QFI-PCR') that measures the absolute copy number of amplifiable DNA, across 165 residual clinical specimens. The results reveal the limitations of commonly used approaches, and demonstrate the value of an integrated workflow using QFI-PCR to improve the accuracy of NGS mutation detection and guide changes in input that can rescue low quality FFPE DNA. These findings address a growing need for improved quality measures in NGS-based patient testing.

Figure 1

Study design. The study design coupled pre-analytical FFPE DNA characterization across three methods (spectrophotometry, fluorescence dye-based quantification, and QFI-PCR) with variant calling results from targeted NGS and confirmation assays to assess the impact of template quality and set thresholds for minimum ‘functional’ DNA inputs. dsDNA, double-stranded DNA; HS, high sensitivity.

Figure 2

FFPE DNA characterization by QFI-PCR and fluorescence-based assays from 165 tumor DNA samples. (A) Distribution of FFPE DNA quantification using QFI-PCR and the fluorescence-based Qubit dsDNA HS assay from 5 ng bulk DNA input as determined by NanoDrop spectrophotometry. A total of 27 samples were undetected by fluorescence assay (<0.1 ng/μl, equivalent to 2% ‘functional’ DNA templates; shown as open circles); these samples produced between 0.03 and 2.5% QFI using qPCR. Five samples were undetected by QFI-PCR and produced between 2.4 to 4.0% of the spectrophotometrically determined DNA concentration using Qubit. Five samples were undetected using both methods (not shown in figure). (B) A scatterplot of QFI-PCR and Qubit relative DNA quantification revealed a linear trend for templates with at least 3% QFI (R = 0.66; N = 86). No correlation, however, was observed for the poorest quality samples (R = 0.07; N = 42; shown as open circles).

Figure 3

Effect of amplifiable FFPE DNA copy number on the detection of mutational load by targeted NGS. The amplifiable DNA copy number (Cp#) for two clinical FFPE samples was calculated to be 370 and 289 copies based on the QFI (9% for both samples) and a well-characterized mutation frequency of 30.0 ± 3.4% for BRAF V600E or 38.4 ± 4.6% for PIK3CA H1047R, respectively. BRAF and PIK3CA loci were enriched using PCR [14] and sequenced on the MiSeq. The graph shows a dilution series of the DNA, from 24 to 3,030 amplifiable DNA copies and each point (blue, BRAF; red, PIK3CA) represents the fraction of reads with the target mutation.

Figure 4

Effect of FFPE DNA copy number on NGS variant calling using a commercial targeted enrichment cancer panel. A) Correlation between QFI and number of variants detected on the PGM following enrichment using the AmpliSeq Cancer Panel. (B) Samples with the lowest QFI produced a significantly larger number of variants from AmpliSeq NGS compared to those with >6% QFI (P-value 0.006).