Frequently arising ESX-1-associated phase variants influence Mycobacterium tuberculosis fitness in the presence of host and antibiotic pressures

Abstract

Mycobacterium tuberculosis (Mtb) exhibits an impressive ability to adapt to rapidly changing environments, despite its genome's apparent stability. Recently, phase variation through indel formation in homopolymeric tracts (HT) has emerged as a potentially important mechanism promoting adaptation in Mtb. This study examines the impact of common phase variants associated with the ESX-1 type VII secretion system, focusing on a highly variable HT upstream of the ESX-1 regulatory factor, espR. By engineering this frequently observed indel into an isogenic background, we demonstrate that a single nucleotide insertion in the espR 5'UTR causes post-transcriptional upregulation of EspR protein abundance and corresponding alterations in the EspR regulon. Consequently, this mutation increases the expression of ESX-1 components in the espACD operon and enhances ESX-1 substrate secretion. We find that this indel specifically increases isoniazid resistance without impacting the effectiveness of other drugs tested. Furthermore, we show that two distinct observed HT indels that regulate either espR translation or espACD transcription increase bacterial fitness in a mouse infection model. The presence of multiple ESX-1-associated HTs provides a mechanism to combinatorially tune protein secretion, drug sensitivity, and host-pathogen interactions. More broadly, these findings support emerging data that Mtb utilizes HT-mediated phase variation to direct genetic variation to certain sites across the genome in order to adapt to changing pressures.

Importance: Mycobacterium tuberculosis (Mtb) is responsible for more deaths worldwide than any other single infectious agent. Understanding how this pathogen adapts to the varied environmental pressures imposed by host immunity and antibiotics has important implications for the design of more effective therapies. In this work, we show that the genome of Mtb contains multiple contingency loci that control the activity of the ESX-1 secretion system, which is critical for interactions with the host. These loci consist of homopolymeric DNA tracts in gene regulatory regions that are subject to high-frequency reversible variation and act to tune the activity of ESX-1. We find that variation at these sites increases the fitness of Mtb in the presence of antibiotic and/or during infection. These findings indicate that Mtb has the ability to diversify its genome in specific sites to create subpopulations of cells that are preadapted to new conditions.

Keywords: bacterial genetics; phase variation; tuberculosis.

Fig 1

espR homopolymer indels frequently occur across diverse clinical isolates. (A–D) tSNE plots generated from 26,903 clinical Mtb isolates. (A) All isolates colored according to lineage. L1 in blue, L2 in orange, L3 in green, L4A in red, L4B in purple, and L4C in brown. (B and C) All isolates colored by espR homopolymer genotype with (B) homopolymer +1 insertions marked in blue, or (C) homopolymer −1 deletions marked in red, (D) L1 isolates colored by both RD236a genotype and espR homopolymer genotype. Isolates marked blue have an intact RD236a locus and are WT at the espR homopolymer locus. Isolates marked in orange have an intact RD236a locus with an indel at the espR homopolymer locus. Isolates marked in green have a deleted RD236a locus, with a WT espR locus. Isolates marked in red have an RD236a deletion and an espR homopolymer indel. (E) A core SNP-based phylogenetic tree of lineage 1 strains (933 publicly available L1 strains predominantly from Vietnam and India). Outer ring represents whether RD236a is present (gray) or deleted (red). Tree tips are colored according to genotype at the espR homopolymer locus. Tree is rooted using Mycobacterium canettii as outgroup.

Fig 2

espR upstream insertion results in upregulation of EspR regulon. (A) Schematic showing the variable 7G homopolymer upstream of espR. The homopolymer begins 144 bp upstream of the espR translational start site. The predicted transcriptional start site rests within the homopolymer, with the wild-type allele predicted to include a 5G stretch at the 5′ end of the UTR. The canonical translational start site and accompanying ribosomal binding site are labeled as START SITE 1, with a second putative start codon and associated ribosomal binding site labeled START SITE 2, with start codons underlined. (B and C) Volcano plots showing RNAseq results from (B) espR ins variant and (C) ∆espR. Genes found in the espACD operon are highlighted in green, select lip genes in blue, and espR in red. (D) Venn diagram showing the overlap of all significantly differentially expressed genes (P-adj ≤ 0.05) in both mutants according to DeSeq2, with the significance of the overlap calculated using a hypergeometric test. (E) Correlation plot showing all genes significantly dysregulated in both mutants with espACD operon highlighted in green. (F) qPCR showing changes in relative espA expression, normalized to sigA levels and compared using one-way ANOVA followed by a Holm-Sidak test. Points indicate individual biological replicates, and the median of these values is shown. P-value: *, <0.05; **, <0.01.

Fig 3

espR-ins mutants result in changes at the protein level. (A) mRNA levels observed in RNAseq for espA, espR, and esxB (CFP-10) in the espR homopolymer mutant strain. Error bars denote standard error. (B–D) Mass spectrometry was performed to evaluate EspR, EspA, and CFP-10 protein levels, respectively. Individual points represent biological replicates, with median value marked as a line. Values were normalized to SigA levels for each sample. Individual points represent four biological replicates with median indicated by line. Samples were compared using ANOVA followed by Dunnett’s multiple comparison test. P-value: *, <0.05; **, <0.01.

Fig 4

espR-ins results in a differentially folded 5′UTR and increased translational efficiency. (A) Folding prediction of WT and +1 espR untranslated region with color denoting the probability of base pairing at each base. The poly-G motif is enclosed with a box at the 5′ end of the transcript, and two putative Shine-Dalgarno sequences are circled. (B) Schematic of espR translational reporter. We fused synthetic promoter p16 to the predicted espR 5′UTR from TSS to start codon, driving a nanoluc reporter. (C) Luminescence of constructs containing WT, +1, and no UTR were measured and compared by ANOVA with Dunnett’s multiple corrections test (D) mRNA levels of nanoluc were compared by qPCR and normalized to sigA as an endogenous control and then compared by ANOVA with Dunnett’s multiple corrections test. Points indicate individual biological replicates, and the median of these values is shown. P-value: *, <0.05; **, <0.01; ***, <0.001; ****, <0.0001.

Fig 5

espR overexpression leads to an increase in isoniazid MIC. (A–F) Minimum inhibitory concentration curves showing percent inhibition of mutant strains and revertants in the presence of (A) isoniazid, (B) rifampicin, (C) ethambutol, (D) moxifloxacin, (E) bedaquiline, or (F) linezolid. Curves were generated using nonlinear regression comparing IC50 values while constraining all other parameters using GraphPad Prism 8. Individual IC50 values were then compared using ANOVA followed by Dunnett’s test when applicable. An additional Bonferroni correction was then performed to account for the number of antibiotics tested. (G) MIC curves comparing WT to an espR overexpression strain in the presence of isoniazid. After generating curves, IC50 values were compared by t-test. (H) MIC curves comparing WT, espR KO, and complemented espR strains. After generating curves from three biological replicates, IC50 values were compared using one-way ANOVA followed by Tukey’s post-test: P-value: *, <0.05; **, <0.01; ***, <0.001; ****, <0.0001.

Fig 6

ESX-1 associated homopolymeric indels do not detectably alter macrophage cytokine profiles or cell death. For all macrophage experiments, BMDMs were infected with Mtb at a theoretical MOI of 9. (A) Four hours post-infection cells were lysed and plated for CFU. Three biological replicates are shown with mean and standard deviation indicated. (B) Twenty-four hours post-infection, macrophage viability was assessed using cell titer glo. Mean and standard deviation of nine replicates pooled across two independent experiments are shown. (C) Four hours post-infection, macrophage RNA was harvested. qPCR was performed on IFN-β, using actB as an endogenous control. Mean and standard deviation of 6 biological replicates pooled from two independent experiments shown. (D, E, and F) Cell supernatant was harvested 24 h post-infection and ELISA was performed to quantify IL-1β, IL-6, and TNF-α respectively. Mean and standard deviation of 9 replicates pooled across two independent experiments are shown. All statistics were performed using ANOVA followed by Dunnett’s multiple comparison testing. P-value: *, <0.05; **, <0.01; ***, <0.001; ****, <0.0001.

Fig 7

ESX-1 mutants show fitness advantage during mouse infection in a T-cell independent manner. (A) Schematic showing experimental design of pooled infection of barcoded Mtb strains (total CFU ~ 400). (B) Lung CFU at day 0 and day 30 from WT or ∆TCR-α mice aerosol infected with minipool. Mean and standard deviation from each timepoint plotted. (C and D) Normalized sequencing counts of barcodes from mouse lung homogenate at day 0 and day 30 in (C) WT B6 and (D) ∆TCR-α mice. Two independently derived barcoded WT, espA, and espR strains and one ∆eccB1 were pooled for analysis. Different point shapes represent individual barcodes for each strain from four different mice with median plotted. Two-way ANOVA followed by Dunnett’s test for multiple comparisons was performed on paired barcode counts at D0 and D30 for each genotype. P-value: *, <0.05; **, <0.01; ***, <0.001; ****, <0.0001.