Transposase mediated construction of RNA-seq libraries

Abstract

RNA-seq has been widely adopted as a gene-expression measurement tool due to the detail, resolution, and sensitivity of transcript characterization that the technique provides. Here we present two transposon-based methods that efficiently construct high-quality RNA-seq libraries. We first describe a method that creates RNA-seq libraries for Illumina sequencing from double-stranded cDNA with only two enzymatic reactions. We generated high-quality RNA-seq libraries from as little as 10 pg of mRNA (∼1 ng of total RNA) with this approach. We also present a strand-specific RNA-seq library construction protocol that combines transposon-based library construction with uracil DNA glycosylase and endonuclease VIII to specifically degrade the second strand constructed during cDNA synthesis. The directional RNA-seq libraries maintain the same quality as the nondirectional libraries, while showing a high degree of strand specificity, such that 99.5% of reads map to the expected genomic strand. Each transposon-based library construction method performed well when compared with standard RNA-seq library construction methods with regard to complexity of the libraries, correlation between biological replicates, and the percentage of reads that align to the genome as well as exons. Our results show that high-quality RNA-seq libraries can be constructed efficiently and in an automatable fashion using transposition technology.

Figure 1.

RNA-seq methods overview. (A) In the standard adapter ligation RNA-seq library construction protocol, double-stranded cDNA made from fragmented mRNA is subjected to end repair, dATP addition, adapter ligation, size selection, and PCR. (B) For the Tn–RNA-seq method described here, double-stranded cDNA is incubated with transposome (transposase complexed with transposon) and then undergoes PCR. (C) In the directional RNA ligation approach (standard directional), poly(A)-selected mRNA is fragmented with heat and end repaired. 3′ and then 5′ adapters are ligated onto single-stranded RNA before reverse transcription, followed by PCR. (D) The directional Tn–RNA-seq library construction described here starts with double-stranded cDNA, in which the second strand synthesized contains uracils instead of thymines. cDNA synthesis is followed by transposition of sequencing adapters, gap fill-in/ligation, USER digestion, and PCR. P5 and P7 correspond to Illumina cluster generation primers. R1 identifies the sequencing primer and cR1 indicates custom sequencing primer. (′) Reverse complement.

Figure 2.

The directional Tn–RNA-seq method exhibits complete strand specificity. Aligned reads for the standard, Tn–RNA-seq, and directional Tn–RNA-seq method are displayed for three genomic loci. Reads mapping to the positive strand are shown in orange and reads mapping to the negative strand are shown in blue. Arrows indicate the direction of transcription for each RefSeq gene.

Figure 3.

Transposon-based libraries show expected depletion of coverage at 5′ ends of transcripts. The percentage of coverage (y-axis), averaged across all transcripts is plotted as a function of distance across the transcripts (x-axis). 0% corresponds to the 5′ ends and 100% corresponds to the 3′ ends of transcripts.

Figure 4.

Expression values are consistent between standard RNA-seq library construction and transposon-based RNA-seq library construction in ECC-1. (A) Scatterplot showing expression values for standard RNA-seq library construction (y-axis) and the Tn–RNA-seq library construction (x-axis). The Pearson correlation between the standard and Tn–RNA-seq protocols is 0.959, and the Spearman rank correlation is 0.979. (B) Scatterplot displaying expression values for standard RNA-seq library construction (y-axis) and the directional Tn–RNA-seq library construction (x-axis). The Pearson correlation between the standard and directional Tn–RNA-seq protocols is 0.934, and the Spearman rank correlation is 0.959. (C) Scatterplot displaying expression values for Tn–RNA-seq library construction (x-axis) and the directional Tn–RNA-seq library construction (y-axis). The Pearson correlation between the Tn–RNA-seq and directional Tn–RNA-seq protocols is 0.971, and the Spearman rank correlation is 0.953.

Figure 5.

Tn–RNA-seq libraries constructed from as low as 10 pg of mRNA are high quality and show highly correlated expression levels. (A) Library complexity, calculated as the number of different alignment positions in a random set of 1 million aligned reads divided by 1 million, is shown for libraries made with between 50 ng and 1 pg of mRNA. (B) Rank correlations of expression measurements between the library constructed with 50 ng of mRNA and every other Tn–RNA-seq library are displayed.