Skip to main content

The impact of RNA extraction method on accurate RNA sequencing from formalin-fixed paraffin-embedded tissues.

Marczyk M, Fu C, Lau R, Du L, Trevarton AJ, Sinn BV, Gould RE, Pusztai L, Hatzis C, Symmans WF

Publication: BMC Cancer (2019) • Vol 19 • Page 1189PMID: 31805884View on PubMed

Purpose

This paper compared RNA integrity metrics as well as whole transcriptome and targeted RNA sequencing results obtained from matched RNAlater-preserved and formalin-fixed paraffin-embedded (FFPE) breast tumors. The effects of using three different FFPE extraction methods were also investigated.

Conclusion

Median RNA integrity numbers (RINs) and percentage of fragments >200 bp (DV200) were lower for FFPE than RNAlater-preserved specimens but median transcript integrity number (TIN) were higher for FFPE than RNAlater specimens. RIN and TIN values were comparable among RNA extracted from the FFPE specimens using the three kits but DV200 was higher for RNA extracted from FFPE specimens using the HighPure Kit than the Norgen or AllPrep kits.

Compared to sequencing libraries made with RNA extracted from FFPE specimens, those constructed from RNA extracted from RNAlater-preserved specimens had more duplicate reads and a higher percentage of GC content and, after alignment to the reference genome, fewer unique reads, higher expression of protein-coding genes, and more reads for miscellaneous RNA. RNA from FFPE specimens had more reads mapping to intronic regions and long non-coding RNAs. PCA of expression of the protein coding genes clustered the FFPE specimens together, regardless of extraction method, and separately from the matched RNAlater specimen. The concordance correlation coefficient (CCC) of gene expression between matched FFPE and RNAlater preserved specimens using whole transcriptome or targeted RNAseq was high, particularly for genes with higher expression, but was even higher among the matched FFPE specimens subjected to different extraction methods. Although over 50% of genes displayed significant differences in expression between RNAlater and FFPE specimens, only approximately 1000 showed differences >2-fold, 78.53% of which were the same regardless of extraction method. Although the numbers of genes with higher and lower expression in FFPE specimens than RNAlater were similar, more of those with the highest fold changes had higher expression in FFPE than RNAlater. The breast cancer signatures calculated using EndoPredict and SETER/PR were highly concordant between FFPE and RNAlater specimens, but the recurrence score was biased towards higher scores in RNAlater specimens, regardless of FFPE extraction kit. Linear mixed effects modeling showed a small positive bias in expression between FFPE and RNAlater preserved specimens and higher variance in genes with low expression.

Studies (1)

Purpose

This study compared RNA integrity metrics as well as whole transcriptome and targeted RNA sequencing results obtained from matched RNAlater-preserved and FFPE breast tumors. The effects of using three different FFPE extraction methods were also investigated. Resected breast tumors from 12 patients with invasive breast cancer were diced and mixed. Half of each specimen was placed in RNAlater, stored at 4°C for 6-72 h, and then frozen at -80°C while the other half was fixed in 10% buffered formalin for 8-72 h at room temperature followed by embedding in paraffin. RNA was extracted from RNAlater-preserved specimens using Qiagen RNeasy Kit and from FFPE blocks using the Norgen FFPE RNA Purification Kit, Qiagen AllPrep DNA/RNA FFPE Kit, and Roche High Pure FFPE RNA Micro Kit. RNA was DNAse treated, quantified by a Nanodrop spectrometer, and evaluated for integrity using a bioanalyzer. Sequencing libraries were prepared using the RNA HyperPrep Kit with RiboErase and sequenced on an Illumina HiSeq 4000. Targeted RNA sequencing libraries were constructed using RainDance Source System and sequenced on an Illumina MiSeq.

Summary of Findings

As expected, median RINs and DV200 were lower for FFPE than RNAlater-preserved specimens (2.5 versus 7.2 and 77 versus 88, respectively). RIN values were comparable between RNA extracted from FFPE specimens with the Norgen (median = 2.4, range = 2–7.1), AllPrep (median = 2.5, range = 1.9–4.6), and HighPure (median = 2.5, range = 1.9–7) kits. However, DV200 values were higher for RNA extracted from FFPE specimens using the HighPure Kit rather than the Norgen or AllPrep kits (median = 83, range = 70–92 versus median 79.5, range = 57–90, P=0.0211 and median = 73, range = 63–87, P<0.001, respectively). In contrast, the median transcript integrity number (TIN) was higher for FFPE than RNAlater specimens (81.66 versus 75.84, P<0.0001), regardless of kit, but did not differ among the extraction kits.

Compared to sequencing libraries made with RNA extracted from FFPE specimens, those constructed from RNA extracted from RNAlater-preserved specimens had more duplicate reads (1.65-fold; P< 0.001), higher percentage of GC content (1.15-fold; P<0.001), and more Illumina adapter sequences (7.29-fold; P< 0.001). After alignment to the reference genome, RNAlater preserved specimens had fewer unique reads, higher expression of protein-coding genes (1.69-fold; P<0.01), and more reads for miscellaneous RNA, but RNA from FFPE specimens had more reads mapping to intronic regions and long non-coding RNAs. PCA of expression of the protein coding genes clustered the FFPE specimens together, regardless of extraction method, and separately from the matched RNAlater specimen. However, the concordance correlation coefficient of gene expression between matched FFPE and RNAlater-preserved specimens was high (>0.5) with higher CCC (approximately 0.15) observed for genes with higher expression (P<0.0001) or using targeted RNAseq (>0.90). The CCC was even higher among the matched FFPE specimens subjected to different extraction methods (>0.93). Although over 50% of genes displayed significant differences in expression between RNAlater and FFPE specimens, only approximately 1000 showed differences >2-fold, 78.53% of which were the same regardless of extraction method. The most differentially expressed genes were found when extraction was with the Norgen Kit (1125 versus 757 with AllPrep and 958 with HighPure). Although the numbers of genes with higher and lower expression in FFPE specimens than RNAlater were similar, more of those with the highest fold changes had higher expression in FFPE than RNAlater. Although the breast cancer signatures calculated using EndoPredict and SETER/PR were highly concordant between FFPE and RNAlater specimens (CCC >0.9), the recurrence score was biased towards higher scores in RNAlater specimens (CCC 0.49-0.56). Importantly, FFPE extraction kit did not affect the results. Linear mixed effects modeling showed a small positive bias in expression between FFPE and RNAlater-preserved specimens and a higher variance in genes with low expression and found that the bias and variance between replicates were lowest when extraction was with HighPure and Norgen kits, respectively

  • Biospecimens

    TissueBreast
  • Preservation Types

    FormalinRNAlater
  • Diagnoses

    NeoplasticCarcinoma
  • Platforms

    RNAAutomated electrophoresis/Bioanalyzer
    RNANext generation sequencing
  • Pre-analytical Factors

    ClassificationPre-Analytical FactorValue(s)
    Biospecimen PreservationType of fixation/preservation
    Formalin (buffered)
    RNAlater
    Analyte Extraction and PurificationAnalyte isolation method
    Norgen FFPE RNA purification Kit
    Qiagen AllPrep DNA/RNA FFPE kit
    Roche High Pure FFPE RNA Micro Kit
    Next generation sequencing SpecificData handling
    EndoPredict
    Reoccurrence Score
    SET ER/PR
    Next generation sequencing SpecificTechnology platform
    Targeted RNAseq
    Whole transcriptome
    Automated electrophoresis/Bioanalyzer SpecificTechnology platform
    DV200
    RIN
    TIN