Efficient and robust RNA-seq process for cultured bacteria and complex community transcriptomes

Abstract

We have developed a process for transcriptome analysis of bacterial communities that accommodates both intact and fragmented starting RNA and combines efficient rRNA removal with strand-specific RNA-seq. We applied this approach to an RNA mixture derived from three diverse cultured bacterial species and to RNA isolated from clinical stool samples. The resulting expression profiles were highly reproducible, enriched up to 40-fold for non-rRNA transcripts, and correlated well with profiles representing undepleted total RNA.

Figure 1

Performance evaluation of five rRNA depletion methods. (a) Shown is the distribution of RNA-seq reads aligning to protein-coding sequences (CDS; blue), rRNA (red), and other regions (tRNA, non-coding RNA, small RNA, and intergenic regions; gray) for undepleted total RNA (top) and five rRNA depletion protocols. (b) The lengths of the black bars represent the coefficient of determination (R²) for RPKM values before and after rRNA depletion using different rRNA-depletion methods. Ribo-Zero, normalization using duplex-specific nuclease (DSN) and Ovation were tested on a 1:1:1 pool (by mass) of total RNA prepared from P. marinus, E. coli, and R. sphaeroides. MICROBExpress and mRNA-ONLY were performed on individual RNA preparations without pooling.

Figure 2

Depletion of rRNA in a mixture of total RNAs from E. coli, R. sphaeroides and P. marinus with Ribo-Zero is reproducible and works well with fragmented total RNA. (a) The pie charts represent the mapped read distributions of protein-coding genes (CDS; blue), rRNA (red), and other reads (tRNA, non-coding RNA, small RNA and intergenic regions; gray) for undepleted total RNA, two technical replicates of Ribo-Zero treatment of intact total RNA and for Ribo-Zero treatment of fragmented total RNA. (b,c) Double-log scatter plots of RPKM values and the coefficient of determination (R²) for the technical Ribo-Zero replicates (b) and for Ribo-Zero treatment of fragmented versus intact total RNA (c). Points on the axes represent CDSs with zero coverage in one of the two samples. The number of data points in the diagonal cloud and on the axes is indicated. The total number of annotated CDSs in the three bacterial genomes is 10,278.

Figure 3

Strand specificity of RNA-seq reads. Shown is a 17-kb window of the E. coli genome viewed with the Artemis browser [28]. The mapped reads aligning to the top strand (green) or bottom stand (purple) consistent with the direction of the annotated genes as represented by the blue boxes with arrows and corresponding gene ID numbers and operons below (for example, genes b3196 through b3206).

Figure 4

Bacterial rRNA depletion and CDS enrichment in RNA extracted from two human stool samples. Shown is the distribution of sequencing reads aligning to annotated protein coding genes (CDS; blue), rRNA (red) and other features (tRNA, intergenic regions and contigs with no annotations; gray) of 19 reference genomes representing the most abundant bacterial species in stools A and B.

Figure 5

Ribo-Zero depleted gene-expression profiles from intact and partially degraded human stool samples: technical reproducibility and correlation with total RNA-seq data. (a-d) Shown are double-log scatter plots and the coefficient of determination (R²) comparing RPKM values for 64,752 annotated CDSs (points in black) in 19 bacterial genomes between technical Ribo-Zero replicates (a,b) as well as with and without Ribo-Zero depletion (c,d). RNA extracted from stool A was largely intact (RIN = 9). RNA from stool B was partially degraded (RIN = 7). Input amount was 5 µg total RNA. Points on the axes represent CDSs without coverage in one of the two samples. The number of data points on the diagonal and on the axes is indicated. Points in green represent the unannotated rRNAs.