Skip to main content

Fixation Effects on Variant Calling in a Clinical Resequencing Panel.

Parker JDK, Yap SQ, Starks E, Slind J, Swanson L, Docking TR, Fuller M, Zhou C, Walker B, Filipenko D, Xiong W, Karimuddin AA, Phang PT, Raval M, Brown CJ, Karsan A

Publication: J Mol Diagn (2019) • Vol 21 • Page 705-717PMID: 31055024View on PubMed

Purpose

The purpose of this paper was to compare histology and next-generation sequencing (NGS) quality metrics and results in matched tumor and normal colorectal carcinoma specimens fixed in formalin, Eco-FixMB, Glyo-Fixx, or UMFIX. The effects of storage of UMFIX and formalin-fixed blocks on NGS results was also investigated.

Conclusion

Both pathologists preferred formalin, Eco-FixMB, or Glyo-Fixx specimens to those fixed in UMFIX, but differences in staining were minor compared to batch differences. The authors report that libraries constructed from specimens fixed with Glyo-Fixx failed quality control and were not sequenced. Sequencing metrics were largely comparable when UMFIX or OCT was used but use of formalin or Eco-FixMB produced fewer on-target reads (Exom capture or WGS), lower percentage of reads of 150-199 MB, more fluctuation in per-base contents between positions, and identification of more variants (mostly C>T/G>A, transitions). The median coverage per amplicon was negatively correlated with amplicon length and GC content in formalin- and EcoFixMB-fixed specimens but not for OCT-embedded or UMFIX-specimens. The effects of fixative on variant identification could be attenuated through changing the variant allele frequency (VAF) cut-off.

Block storage had an adverse effect on the percentage of on-target reads, number of excess variants, and per base variation in formalin specimens compared to the prior analysis. In contrast, the percentage on-target reads for UMFIX specimens was not affected and the increase in the number of variants detected was smaller.

 

Studies (1)

Purpose

The purpose of this study was to compare histology and NGS quality metrics and results in matched tumor and normal colorectal carcinoma specimens fixed in formalin, Eco-FixMB, Glyo-Fixx, or UMFIX. The effects of storage of UMFIX and formalin-fixed blocks on NGS results was also investigated. Excess tumor and normal adjacent tissues were collected from 10 patients during colectomy or hemicolectomy and transported to the laboratory in RPMI. Specimens were washed in PBS and excess fat, muscle, and grossly necrotic areas were removed. The remaining specimen was trimmed to 50-100 mg and matched specimens were fixed within 2 h of resection in formalin, UMFIX, Eco-FixMB, or Glyo-Fixx overnight at 4°C or placed in sucrose. The UMFIX specimens were then processed using a Tissue-Tek Xpress Rapid Tissue Processor while the other fixed specimens were processed and embedded using a Shandon Tissue Processor. The sucrose specimens were immersed in OCT for 1 h and frozen at -80°C. H&E staining was assessed by two pathologists blinded to the fixative type. One pathologist evaluated the normal and tumor tissues in at least three fixatives from 10 cases, while the other evaluated 6 cases. DNA was extracted from fixed specimens using the AllPrep FFPE DNA Extraction Kit and from OCT specimens using the AllPrep DNA Extraction Kit and quantified by fluorometry. RainDance amplicon libraries were constructed using equal inputs and sequenced using a MiSeq. Variants were identified by Samtools software version 2.3.6. Exome libraries were constructed using SureSelectXT Human All Exon V5þUTRs 96 capture probes and sequenced on a HiSeq2500. Variants were called using VarScan2.

Summary of Findings

Both pathologists preferred formalin, Eco-FixMB, or Glyo-Fixx specimens to those fixed in UMFIX, but differences in staining were minor compared to batch differences. The authors report that libraries constructed from specimens fixed with Glyo-Fixx failed quality control and were not sequenced. Specimens fixed with formalin or Eco-FixMB had lower counts of contaminant reads (P<0.05 and P<0.01, respectively), but also fewer on-target reads (P<0.05 and P<0.01, respectively) and more off-target reads than from OCT libraries (P=0.01 and P<0.001, respectively). In contrast, UMFIX libraries had fewer contaminant and unaligned reads and more on-target reads than libraries from OCT specimens (P=0.0077, P=0.000041, and P=0.0021; respectively). Compared to OCT libraries, the per-base contents fluctuated between positions more in libraries from formalin-fixed and EcoFixMB-fixed specimens (P=0.00072 and P=0.00023, respectively) and less in libraries from UMFIX specimens (P=0.0039). UMFIX and OCT specimens had a similar percentage of reads of 150-199 bp to the prediction of 55% (49 and 56%, respectively), but the proportion was significantly lower in formalin (40%, P=0.007) and EcoFixMB (40%, P=0.0081) specimens which had increased representation of inserts of 0-49 bp and >250 bp compared to OCT controls and the expectation. Library coverage histograms showed lower coverage for specimens fixed with formalin and EcoFixMB compared to OCT specimens, but UMFIX specimens were of similar or higher coverage than OCT specimens. The median coverage per amplicon in formalin- and EcoFixMB- fixed specimens was negatively correlated with amplicon length (R2=0.07, P=4 e-8 and R2=0.17, P=7e-19; respectively) and GC content (R2=0.18, P=6e-21 and R2=0.19, P=4e-19; respectively) but median coverage was not correlated with amplicon length or GC content for OCT-embedded or UMFIX-specimens. Importantly, more variants (mostly C>T/G>A transitions) were identified in formalin and EcoFixMB specimens than in OCT specimens (P=0.0001), particularly in low coverage areas, but this was not found for UMFIX specimens. Increasing the VAF from 0.01 to 0.1 allowed the libraries from formalin and EcoFixMB specimens to cluster with OCT specimens. Only 2 germline variants were not found in all replicate libraries and both of these were found in EcoFixMB-fixed specimens. The interquartile range for heterozygous variants was also larger for formalin (0.11) and EcoFixMB (0.12) specimens rather than OCT (0.07) or UMFIX (0.05) specimens.

When DNA was extracted from the same blocks 14 months later, the percentage of on-target reads was reduced and per base variation increased non-significantly in formalin specimens compared to the prior analysis, but there was no change for UMFIX specimens. There were also more variants identified in the libraries from formalin and UMFIX blocks stored for 14 months than when extraction was done earlier (P=8.5e-6 and P=0.047, respectively). When the number of excess variants was plotted versus block storage duration, a clear increase with storage was found for formalin and EcoFixMB-fixed specimens but the increase was hardly noticeable in UMFIX specimens.

  • Biospecimens

    TissueColorectal
  • Preservation Types

    FormalinOCTOther Preservative
  • Diagnoses

    NeoplasticCarcinoma
  • Platforms

    MorphologyH-and-E microscopy
    DNANext generation sequencing
  • Pre-analytical Factors

    ClassificationPre-Analytical FactorValue(s)
    Biospecimen PreservationType of fixation/preservation
    EcoFixMB
    Formalin (buffered)
    GlycoFixx
    OCT
    UMFIX
    StorageStorage duration
    Unspecified (short)
    Up to 14 months