Application of the 3' mRNA-Seq using unique molecular identifiers in highly degraded RNA derived from formalin-fixed, paraffin-embedded tissue

Abstract

Background: Archival formalin-fixed, paraffin-embedded (FFPE) tissue samples with clinical and histological data are a singularly valuable resource for developing new molecular biomarkers. However, transcriptome analysis remains challenging with standard mRNA-seq methods as FFPE derived-RNA samples are often highly modified and fragmented. The recently developed 3' mRNA-seq method sequences the 3' region of mRNA using unique molecular identifiers (UMI), thus generating gene expression data with minimal PCR bias. In this study, we evaluated the performance of 3' mRNA-Seq using Lexogen QuantSeq 3' mRNA-Seq Library Prep Kit FWD with UMI, comparing with TruSeq Stranded mRNA-Seq and RNA Exome Capture kit. The fresh-frozen (FF) and FFPE tissues yielded nucleotide sizes range from 13 to > 70% of DV200 values; input amounts ranged from 1 ng to 100 ng for validation.

Results: The total mapped reads of QuantSeq 3' mRNA-Seq to the reference genome ranged from 99 to 74% across all samples. After PCR bias correction, 3 to 56% of total sequenced reads were retained. QuantSeq 3' mRNA-Seq data showed highly reproducible data across replicates in Universal Human Reference RNA (UHR, R > 0.94) at input amounts from 1 ng to 100 ng, and FF and FFPE paired samples (R = 0.92) at 10 ng. Severely degraded FFPE RNA with ≤30% of DV200 value showed good concordance (R > 0.87) with 100 ng input. A moderate correlation was observed when directly comparing QuantSeq 3' mRNA-Seq data with TruSeq Stranded mRNA-Seq (R = 0.78) and RNA Exome Capture data (R > 0.67).

Conclusion: In this study, QuantSeq 3' mRNA-Seq with PCR bias correction using UMI is shown to be a suitable method for gene quantification in both FF and FFPE RNAs. 3' mRNA-Seq with UMI may be applied to severely degraded RNA from FFPE tissues generating high-quality sequencing data.

Keywords: 3′ mRNA-Seq; FFPE; Gene expression; PCR amplification bias; UMI.

Fig. 1

The overall experimental design

Fig. 2

The PCR bias-corrected QuantSeq 3′ mRNA-Seq data in the UHR. A; Percentage of mapped reads out of total reads between standard input and low input/FFPE protocols by different input amounts. B & C; Total number of detected genes in the different inputs between protocols. D; Similarity matrix between the input amounts and protocols. Data were normalized by log₂ (TPM + 1)

Fig. 3

Data comparison between the QuantSeq 3′ mRNA-Seq and TruSeq Stranded mRNA-Seq kit. A; Correlation plot. Data were normalized by log₂ (TPM + 1). Each dot constitutes a gene. B; Distribution of mapped reads. The incompatible paired-end reads (15%) were not reflected in the TruSeq Stranded mRNA-Seq data. C; Number of detected genes between two platforms. D; Percentage of mapped reads distribution by RNA biotypes

Fig. 4

The PCR bias-corrected QuantSeq 3′ mRNA-Seq data in the degraded RNA (DV200 > 30%). A; Percentage of mapped reads out of total reads. Blue, non-PCR bias-corrected reads; Orange, PCR bias-corrected reads. B; Total number of detected genes. C; Similarity matrix in the paired FF and FFPE samples at 1 ng and 10 ng input. D; Correlation plot at 10 ng input between FF and FFPE samples. Samples 6-FF-70 and 8B-FFPE-70 are paired samples. Data were normalized by log₂ (TPM + 1). 6-FF-70, 70% of DV200; 8B-FFPE-70, 70% of DV200;; 2-FFPE-50, 50% of DV200;; 3-FFPE-40, 40% of DV200;; 2-FF-68, 68% of DV200. Each dot constitutes a gene

Fig. 5

The QuantSeq 3′ mRNA-Seq data comparison using highly degraded RNA (DV200 ≤ 30%). A; Percentage of mapped reads out of total reads by different input amounts and average fragment size of RNA. Blue, non-PCR bias-corrected reads; Orange, PCR bias-corrected reads. B; Total number of detected genes in the different inputs and average fragment size of FFPE RNA. C & D; Similarity matrix at the 10 ng and 100 ng input amounts of EF1-FFPE-30 and GT1-FFPE-30. GT1-FFPE-13, 13% of DV200; GT1-FFPE-30, 30% of DV200; EF1-FFPE-20, 20% of DV200; EF1-FFPE-30, 30% of DV200; JB1-FFPE-19, 19% of DV200; 1-FF-20, 20% of DV200

Fig. 6

Data comparison between the QuantSeq 3′ mRNA-Seq and RNA Exome Capture. A; Correlation analysis. Data were normalized by log₂ (TPM + 1). Each dot constitutes a gene. B; Distribution of mapped reads. Data are means of EF1-FFPE-30 and GT1-FFPE-30 samples from each kit ± SD.***, p < 0.001; **, p < 0.01. The incompatible paired-end reads (11%) were not reflected in the RNA Exome Capture data. C; Number of detected protein-coding genes between two platforms. EF1-FFPE-30, 30% of DV200; GT1-FFPE-30, 30% of DV200