Steiert, T. A., Fuß, J., Franke, A., & Forster, M. (2023). A critical spotlight on the paradigms of FFPE-DNA sequencing. Nucleic Acids Research, 51(14), 7143–7158.

Steiert et al. (2023) — A Critical Spotlight on the Paradigms of FFPE-DNA Sequencing

Bibliographic Reference

Steiert, T. A., Fuß, J., Franke, A., & Forster, M. (2023). A critical spotlight on the paradigms of FFPE-DNA sequencing. Nucleic Acids Research, 51(14), 7143–7158. DOI: 10.1093/nar/gkad528.

Core Argument

Most FFPE-DNA literature has focused on a single damage type — cytosine deamination (C→T) — but the complete spectrum of FFPE-induced DNA damage is broader and includes oxidation products (C→A), other base substitutions (T→A, T→C), and three mechanisms of information loss (duplication inflation, insert size collapse, coverage non-uniformity). Using case-matched 13-year-old FFPE and FF liver specimens with extensive replication, the authors demonstrate that C→T/G→A artifacts show a 7-fold increase over FF, but other artifact types are equally prevalent and cannot be addressed by UDG treatment alone. They synthesize recommendations across four critical parameters — (I) pre-analytical QC, (II) DNA repair, (III) sample preparation, (IV) bioinformatic analysis — and provide the “ERROR-FFPE-DNA” checklist for minimal reporting standards in FFPE-based sequencing publications.

Methods

• Case-matched comparison: 13-year-old FFPE and FF liver specimens from the same individual, analyzed with target-enriched sequencing across many replicates
• Four-parameter framework: (I) Pre-analytical sample specifications (tumor content, DIN, DNA quantity), (II) DNA repair treatments (UDG + FA-pyrimidine-DNA glycosylase), (III) Analytical sample preparation (library type, target enrichment vs. amplicon, PCR cycles), (IV) Bioinformatic analysis (duplicate removal, VAF threshold, strand-specific filtering)
• Empirical demonstration: Comparison of multiple DNA repair strategies and library preparation methods on the 13-year-old specimens
• ERROR-FFPE-DNA checklist: Minimal reporting items for FFPE-NGS publications

Key Findings

• The FFPE artifact spectrum extends beyond C→T. While C→T/G→A showed the highest fold-increase over FF (7×), other substitution types — C→A/G→T (oxidation), T→A/A→T, T→C/A→G — were equally prevalent in absolute terms and contributed to the total artifact repertoire. UDG treatment alone, which addresses only uracil-derived C→T artifacts, leaves the majority of non-deamination artifact types unaffected.

• Three mechanisms of information loss compound artifact risk. (1) Duplication ratio is 2× higher in FFPE vs. FF libraries — half the unique coverage per sequencing dollar. (2) Insert sizes are ~½ of FF libraries due to fragmentation — shorter inserts mean paired-end reads overlap, reducing unique bases sequenced, and make unique mapping more difficult. (3) Coverage uniformity is decreased, with systematic dropouts in AT-rich regions. These three mechanisms jointly create low-coverage regions where artifact-bearing fragments are overrepresented, producing artifact allele frequencies (AAF) that can exceed 10%.

• Artifact allele frequencies regularly exceed the standard 5% filter threshold. In the 13-year-old FFPE specimen, some AAFs exceeded 10%, particularly in regions of low sequencing coverage. The highest AAF was not a C→T deamination artifact but a C→A/G→T change. This means that the common practice of filtering variants with VAF <5% — intended to remove sequencing errors — may fail to remove a substantial fraction of FFPE artifacts, which can masquerade as genuine subclonal mutations.

• Pre-analytical QC is the most critical parameter. DNA integrity number (DIN), tumor cell content estimation by a pathologist, and PCR-based quantification of amplifiable templates (not spectrophotometry) are essential. The authors show that even low-quality FFPE-DNA (DIN 2.0, where only 5% of fragments amplify at ≥129 bp) can yield usable data when all four parameters are optimized — but only if the limitations are documented and the analysis is fit-for-purpose.

• The ERROR-FFPE-DNA checklist. The authors propose minimal reporting standards for FFPE-NGS publications: specimen age, fixation protocol (formalin concentration, pH, duration), storage conditions, DNA extraction method, DIN or equivalent fragmentation metric, amplifiable template quantity (qPCR), library preparation method (amplicon vs. capture, UDG treatment, polymerase type), sequencing depth and coverage uniformity, bioinformatic filtering strategy, and VAF threshold rationale.

Concepts Introduced or Used

FFPE-DNA damage spectrum, artifact allele frequency (AAF), DNA integrity number (DIN), duplication ratio, insert size collapse, coverage uniformity, AT-rich dropout, ERROR-FFPE-DNA checklist, pre-analytical QC, DNA repair treatments, target-enriched vs. amplicon sequencing

Entities Referenced

• EASI-Genomics consortium — European advanced sequencing infrastructure
• 100,000 Genomes Project — reported systematic FFPE dropouts in AT-rich regions
• DIN (DNA integrity number) — electrophoresis-based fragmentation metric
• UDG, FA-pyrimidine-DNA glycosylase — DNA repair enzymes

Limitations (as stated by authors)

• The experimental demonstration used a single 13-year-old case-matched FFPE/FF pair — results may vary with specimen age, tissue type, and fixation protocol
• DIN 2.0 represents severely degraded DNA; results obtained under these conditions may not generalize to higher-quality FFPE specimens
• The ERROR-FFPE-DNA checklist is a proposal; adoption depends on journal and community acceptance
• Some damage types (e.g., interstrand crosslinks) cannot be fully reversed and cause permanent information loss

Relevance to Clonal Evolution

This paper is the modern operational manual for FFPE-DNA sequencing, extending the mechanistic framework of Do & Dobrovic (2015) with quantitative empirical data and practical recommendations. Its relevance to clonal evolution is threefold:

1. The 5% VAF filter is not a safety net for FFPE artifacts. Standard bioinformatic practice removes variants with VAF <5% as putative sequencing errors. Steiert et al. demonstrate that FFPE artifacts can exceed 10% VAF in low-coverage regions, meaning that artifacts penetrate the standard filtering threshold. For subclonal reconstruction, this is critical: subclonal mutations are expected at VAFs of 5–25% — precisely the range where FFPE artifacts also appear. The joint distribution of true subclonal SNVs and FFPE artifacts in FFPE-derived data is unidentifiable without orthogonal validation (variant-allele-fraction).

2. The information-loss triad — duplication, insert collapse, coverage dropout — is a confounder for CNA inference. Decreased coverage uniformity, with systematic dropout in AT-rich regions, directly affects logR-based copy number calls. The 2× duplication inflation reduces effective coverage for a given sequencing depth, increasing the CNA detection floor. The shortened insert sizes reduce the number of unique mappable bases, further degrading coverage in repetitive or GC/AT-extreme regions. These effects are region-specific — they produce apparent copy number changes that are preservation artifacts, not biological CNA events (copy-number-alteration).

3. The ERROR-FFPE-DNA checklist provides a framework for assessing FFPE-derived clonal evolution studies. When evaluating published clonal evolution results from FFPE cohorts, the checklist items — specimen age, fixation protocol, DIN, UDG treatment, library type, coverage uniformity — are the minimum information needed to assess whether the reported subclonal mutations are trustworthy. Studies that report SMF values from FFPE data without documenting these parameters should be interpreted with caution, because the SMF may reflect an unknown mixture of true subclonal mutations, CNA-induced miscalibration, and FFPE-induced artifacts spanning multiple base substitution types — only one of which (C→T) is partially addressed by UDG pretreatment (subclonal-reconstruction).