Genome-wide mutational biases fuel transcriptional diversity in the Mycobacterium tuberculosis complex

Abstract

The Mycobacterium tuberculosis complex (MTBC) members display different host-specificities and virulence phenotypes. Here, we have performed a comprehensive RNAseq and methylome analysis of the main clades of the MTBC and discovered unique transcriptional profiles. The majority of genes differentially expressed between the clades encode proteins involved in host interaction and metabolic functions. A significant fraction of changes in gene expression can be explained by positive selection on single mutations that either create or disrupt transcriptional start sites (TSS). Furthermore, we show that clinical strains have different methyltransferases inactivated and thus different methylation patterns. Under the tested conditions, differential methylation has a minor direct role on transcriptomic differences between strains. However, disruption of a methyltransferase in one clinical strain revealed important expression differences suggesting indirect mechanisms of expression regulation. Our study demonstrates that variation in transcriptional profiles are mainly due to TSS mutations and have likely evolved due to differences in host characteristics.

Fig. 1

Global transcriptomic profiles of the samples. a The PCA plot shows that samples belonging to the same phylogenetic clade tend to group closely, except for two cases. b A cluster analysis reinforces the trend derived from the PCA, with almost all the samples belonging to the same lineage clustering together. The heatmap colour scale reflects the Euclidean distance between each sample, calculated from the complete transcriptomic signatures

Fig. 2

Gene expression changes across the MTBC phylogeny. a Number of genes differentially expressed (red up, blue down) in each of the main branches of the MTBC phylogeny. The phylogeny was constructed using Illumina sequencing data, the Maximum-Likelihood algorithm and a bootstrapping of 1000 replicates. Sample N1177 is included to shown the complete phylogenetic picture, but it was not included for further analyses. b Number of PDEG genes in each of the main MTBC branches plotted against the genetic distances

Fig. 3

Non-random processes impact the emergence and disruption of Pribnow boxes. a Distribution of new (green) and disrupted (purple) Pribnow boxes in 1000 random simulations. Red arrows mark the observed value for each type of event in our dataset. b Mutation bias towards new A and T alleles inferred from 235,212 substitution obtained from 4595 clinical samples of the MTBC and normalised by GC content as in ref.

Fig. 4

Impact of natural mutations in the appearance and disruption of Pribnow boxes. a Effect of the new/disrupted Pribnow boxes over the expression of downstream genes. New boxes tend to upregulate gene expression while disrupted boxes tend to downregulate transcription (wilcoxon test, p-value = 5.37-E09). Blue circles represent those changes in expression detected in the PDEG analysis (adj-pval < 0.05, log2 fold-change in expression > 1.5). Red circles represent subtle changes in gene expression, thus not identified by the PDEG analyses. b New Pribnow boxes can increase sense and/or antisense expression, depending on the genomic context in which the mutation appears. c The G2726105A mutation, common to all L3 strains, creates two new Pribnow boxes in the intergenic region of oxyR and ahpC. These new boxes are the potential explanation for the observed upregulation of oxyR, ahpC and ahpD in the L3 strains

Fig. 5

Methyltransferase activity of the MTBC. a Distribution of the characterised mutations that potentially impair methyltransferase function on a global dataset (n = 4595 samples). b Gene expression differences between SigA recognition motifs differentially methylated by each of the three methyltransferases. Red line marks a 0 fold-change in gene expression (no differences). The expression of each gene was tested in both situations, methylated and non-methylated strains, in independent lineages (when possible). c Different overlapping patterns found between SigA recognition motifs and the methylated motifs. The red adenines in the motif are the methylated ones

Fig. 6

Gene expression differences due to an hsdM deletion. a Overall transcriptomic profiles of the wild-type versus the ΔhsdM strains. b Volcano-plot of the gene expression differences of the wild-type versus de mutant strains. A small numbers of genes showed differential expression (3 downregulated and 10 upregulated)