Comprehensive analysis of the Corynebacterium glutamicum transcriptome using an improved RNAseq technique

Abstract

Background: The use of RNAseq to resolve the transcriptional organization of an organism was established in recent years and also showed the complexity and dynamics of bacterial transcriptomes. The aim of this study was to comprehensively investigate the transcriptome of the industrially relevant amino acid producer and model organism Corynebacterium glutamicum by RNAseq in order to improve its genome annotation and to describe important features for transcription and translation.

Results: RNAseq data sets were obtained by two methods, one that focuses on 5'-ends of primary transcripts and another that provides the overall transcriptome with an improved resolution of 3'-ends of transcripts. Subsequent data analysis led to the identification of more than 2,000 transcription start sites (TSSs), the definition of 5'-UTRs (untranslated regions) for annotated protein-coding genes, operon structures and many novel transcripts located between or in antisense orientation to protein-coding regions. Interestingly, a high number of mRNAs (33%) is transcribed as leaderless transcripts. From the data, consensus promoter and ribosome binding site (RBS) motifs were identified and it was shown that the majority of genes in C. glutamicum are transcribed monocistronically, but operons containing up to 16 genes are also present.

Conclusions: The comprehensive transcriptome map of C. glutamicum established in this study represents a major step forward towards a complete definition of genetic elements (e.g. promoter regions, gene starts and stops, 5'-UTRs, RBSs, transcript starts and ends) and provides the ideal basis for further analyses on transcriptional regulatory networks in this organism. The methods developed are easily applicable for other bacteria and have the potential to be used also for quantification of transcriptomes, replacing microarrays in the near future.

Figure 1

Experimental workflow for the preparation of a whole transcriptome library (a) and of a library enriched for primary 5′-transcript ends (b). Both protocols start with isolated total RNA. Stable RNA is then depleted using the Ribo-Zero rRNA removal kit and the obtained RNA is fragmented my metal hydrolysis to a size of 200 - 500 nt. For the whole transcriptome library (a) the 5′-triphosphate ends are processed to 5′-monophosphate ends by a RNA 5′-polyphosphatase, unphosphorylated 5′-ends are phosphorylated, and phosphorylated 3′-ends are then dephosphorylated using T4 polynucleotide kinase. For the native 5′-end protocol (b), all fragments containing a 5′-monophosphate are degraded by treatment with a 5′-phosphate dependent exonuclease and the 5′-triphosphate ends of native transcripts are then processed to 5′-monophosphate ends by a RNA 5′-polyphosphatase. Next, for both libraries RNA adapters are ligated to the 5′-ends carrying a 5′-monophosphate group. The tagging of the 3′-end of the RNA with flanking sequences necessary for reverse transcription is performed in a ligation-free approach with a loop DNA adapter containing seven unpaired wobble bases at its 3′-end. After reverse transcription of the RNA fragments into cDNA fragments, the cDNA fragments are amplified, tagged with sequencing linkers at their ends by PCR and finally sequenced. Stable RNA species (rRNA, tRNA) are depicted in red, other RNAs are given in green, and DNA in blue.

Figure 2

Classification of TSSs obtained with RNAseq. (a) Illustration of categories for TSS classification based on genomic context. The first TSS classification level is divided into two categories: TSSs that belong to annotated genes (gray shaded arrows) and TSSs that belong to new transcripts (black shaded arrows). TSSs belonging to annotated genes were classified into single TSSs or multiple TSSs. TSSs belonging to new transcripts were arranged into antisense, intragenic or intergenic TSSs. (b) Identification, filtering, and classification of TSSs. From the automatically detected TSSs those TSSs were removed that belong to rRNA or tRNA, false-positive, or alternative TSSs.

Figure 3

Distribution of nucleotides within the -10 and -35 regions of C. glutamicum σ^A promoters. Relative occurrence of a nucleotide at a particular position is represented by the size of the nucleotide. The representation is based on 2,522 -10 and 704 -35 regions identified with Improbizer[43]. The core -10 and -35 regions are underlined. The sequence logo was created with WebLogo[45].

Figure 4

Distribution of 5′-UTR length of mRNAs belonging to annotated protein-coding genes in C. glutamicum. The distribution is based on 2,147 TSSs assigned to mRNAs. The bar labeled leaderless represents an UTR length of zero. The other bars represent UTR length in increments of five (1 - 5, 6 - 10, 11 - 15, etc.).

Figure 5

Analysis of ribosome binding sites in C. glutamicum. (a) Frequencies of purines (G or A) compared to frequencies of pyrimidines (T or C) within the first 20 bases (relative to the start codon) of 922 different 5′-UTRs. (b) Analysis of the spacing between the RBS and the start codon, based on all identified RBS motifs by Improbizer. (c) Information content within the identified RBS motif of C. glutamicum including three leading and lagging bases, a spacer of 1 - 7 nt, as well as the translational initiation codon. The y-axis shows the information content (measured in bits). The analysis is based on 673 RBS motifs identified with Improbizer within a spacing of 7 ± 3 nt only. The sequence logo was created with WebLogo[45].

Figure 6

Classification and identification of operon structures in C. glutamicum shown for an example region. Black color denotes cumulated reads derived from primary transcripts (upper part) or from the whole transcriptome (bottom part) that are both used to detect operon structures. The y- and x-axis represents coverage and genomic position. Primary operons that were found by combined read pairs covering neighboring genes: cg2332-cg2331-cg2330, and cg2338-cg2337; monocistronic operons that were indicated by reads covering only one gene: cg2333, cg2334, and cg2336; sub-operons that were identified by TSSs of the primary transcript ends library (stacks in the upper part) within primary operons: cg2331-cg2330, and cg2337.

Figure 7

Analysis of the gene numbers in monocistronic transcripts, primary operons, and sub-operons in C. glutamicum. Operons differing in the number of genes are shown in different colors.

Figure 8

Examples of transcript ends determined by RNAseq and predicted rho-independent terminators. The genomic regions, predicted terminators, and accumulated reads were shown. Protein-coding regions are indicated by arrows. The picture includes two examples of unidirectional terminators (a and b) and one example of a bidirectional terminator structure (c). Rho-independent terminators were predicted by TransTermHP[54].