Bacterial strains and growth conditions

M. tuberculosis H37Rv and its derivatives were grown in Middlebrook 7H9 broth supplemented with 10% OADC (0.5 g/L oleic acid, 50 g/L bovine serum albumin fraction V, 20 g/L dextrose, 8.5 g/L sodium chloride, and 40 mg/L catalase), 0.2% glycerol and 0.05% Tween 80. Liquid cultures were grown in 50 mL conical polypropylene tubes at 37°C with a shaker speed of 200 rpm except when indicated otherwise. For growth on solid media, Middlebrook 7H10 supplemented with 0.5% glycerol and OADC was used. For M. tuberculosis auxotrophic strain mc²6230 (ΔpanCD, ΔRD1, [31]) pantothenate was added to 7H9 or 7H10 to a final concentration of 24 μg/mL. When required for resistant bacteria selection or plasmid maintenance, the following concentrations of antibiotics were used: 25 μg/mL kanamycin (KAN), 50 μg/mL hygromycin (HYG) or 25 μg/mL zeocin (ZEO). Knock-out strains were constructed using the recombineering system described by Murphy and collaborators [32]. For genetic complementation, an L5-site integrating plasmid or Giles-site integrating plasmid was used, and for gene overexpression the episomal plasmid pMV762 was used [33]. A description of all strains used in this study is provided in S1 Table.

For growth of Mtb mc²6230 in minimal media, log phase cultures grown in minimal media were sub-cultured to an OD_600nm = 0.01 in the same media. The minimal media had the following composition: 0.5 g/L asparagine, 1 g/L KH₂PO₄, 2.5 g/L Na₂HPO₄, 50 mg/L ferric ammonium citrate, 0.5 g/L MgSO₄·7H₂O, 0.5 mg/L CaCl₂, and 0.1 mg/L ZnSO₄. Minimal media was supplemented with 0.05% Tween 80, 24 μg/mL pantothenate, and 0.2% glycerol.

Antibiotic susceptibility testing

To determine the minimum inhibitory concentration (MIC) for Mtb mc²6230 strains the agar proportion method was used [34]. Briefly, antibiotics were added to 7H10 plates to obtain the following concentrations: 1, 0.5, 0.25, 0.125, 0.06125, 0.0306, 0.01531 and 0.0077 μg/mL of rifampicin (RIF) or isoniazid (INH). Direct 5 μL aliquots and serial dilutions of 7H9 mid-log cultures of each strain were plated on antibiotic-containing and antibiotic-free plates. The MIC for each strain was determined as the drug concentration that reduced CFU by 90% compared to the control.

Drug killing experiments

For Mtb mc²6230 and H37Rv strains, log phase cultures grown in 7H9 were diluted to an initial OD_600nm = 0.1 in triplicate in absence of antibiotics. After 24 hours the following antibiotics were added to the indicated final concentrations: 0.6 μg/mL of RIF, 2.4 μg/mL of INH, 2.5 μg/mL of clarithromycin (CLA), 1 μg/mL of ofloxacin (OFX), 2 μg/mL ethambutol (EMB), or 500 μg/mL erythromycin (ERY). Triplicate cultures of each strain were incubated in absence of drug as a control. Aliquots were periodically taken, and serial dilutions plated on 7H10 agar plates without drug. CFUs were counted after 20–35 days.

Determination of the fraction of survival in INH at different growth phases

Log phase cultures of Mtb mc²6230 WT or Δrnj were sub-cultured to an OD_600nm = 0.02. A total of 24 tubes were used per strain (6 replicates per timepoint). INH was added after 24, 48, 72, or 96 hours to a final concentration of 2.4 μg/mL. For each timepoint, CFUs were measured before adding INH (time 0 for each timepoint) and after 2 days of incubation with INH. The fraction of survival in INH was determined as the ratio between CFUs at day 2 over CFUs at time 0.

RNA purification and quantitative PCR

For RNA purification from Mtb mc²6230, frozen cultures stored at -80°C were thawed on ice and centrifuged at 4,000 rpm for 5 min at 4°C. For Mtb H37Rv, cultures were pelleted and processed immediately. The pellets were resuspended in 1 mL Trizol (Life Technologies) and placed in tubes containing Lysing Matrix B (MP Bio). Cells were lysed by bead-beating (2 cycles of 9 m/sec for 40 s, with 2 min on ice in between) in a FastPrep 5G instrument (MP Bio). 300 μL chloroform was added and samples were centrifuged for 15 min at 4,000 rpm at 4°C. The aqueous phase was collected, and RNA was purified using Direct-Zol RNA miniprep kit (Zymo) according to the manufacturer’s instructions. For Mtb mc²6230 samples, the optional on-column DNase treatment step was used. For H37Rv samples, in-tube DNase treatment was done using DNase Turbo (Ambion) followed by purification with a Zymo Clean & Concentrator kit.

For cDNA synthesis, 600 ng of RNA were mixed with 0.83 μL 100 mM Tris, pH 7.5, and 0.17 μL of random primers (3 mg/mL NEB) in a total volume of 5.25 μL. The mix was denatured at 70°C for 10 min and placed on ice for 5 min. For reverse transcription, the following reagents were added to achieve the specified amounts or concentrations in 10 μL reactions: 100 U ProtoScript II reverse transcriptase (NEB), 10 U RNase inhibitor (murine; NEB), 0.5 mM each deoxynucleoside triphosphate (dNTP), and 5 mM dithiothreitol (DTT). Samples were incubated for 10 min at 25°C and reverse transcription was performed overnight at 42°C. RNA was degraded by addition of 10 μL containing 250 mM EDTA and 0.5 N NaOH and heating at 65°C for 15 min, followed by addition of 12.5 μL of 1 M Tris-HCl, pH 7.5. cDNA was purified using the MinElute PCR purification kit (Qiagen) according to the manufacturer’s instructions.

RNA abundance was determined by quantitative PCR (qPCR) using iTaq SYBR green (Bio-Rad) with 200 pg of cDNA and 0.25 μM each primer in 10 μL reaction mixtures, with 40 cycles of 15 s at 95°C and 1 min at 61°C (Applied Biosystems 7500). Primers used in this study are listed in S2 Table.

Measurement of RNA half-life

For Mtb mc²6230, 5 mL of biological triplicates of log phase cultures were treated with RIF (final concentration of 50 μg/mL) to inhibit transcription, as reported by [35]. After RIF addition, tubes were placed into liquid nitrogen after 0, 2, 5, 10, 20, 40, and 80 min. Cultures were frozen at -80°C until RNA purification. RNA was purified using Direct-zol kit (Zymo) as described above. Half-lives were calculated as previously described [36,37]. Timepoints 40 and 80 min were excluded from half-life calculations in cases where there were missing values due to low abundance or cases where they did not follow the initial exponential decay trend (S8 Fig).

Construction and analysis of RNA expression and 5’ end-directed libraries

For RNAseq libraries, 5 mL of log phase cultures of Mtb H37Rv WT, Δrnj, and Δrnj::rnj_OE strains grown in 7H9 were placed in liquid nitrogen and stored at -80°C. RNA purification and construction of both RNA expression and 5’ end-directed (non-5’ pyrophosphohydrolase-converted) libraries were performed as previously reported [38,39]. For both types of libraries, Illumina HiSeq 2000 paired-end sequencing producing 50 nt reads was used. Sequencing was performed at the UMass Medical School Deep Sequencing Core Facility. Raw and processed data are available in GEO, accession number GSE196357.

Bioinformatic tools and analyses

Reads from Mtb datasets were aligned to the NC_000962 reference genome using Burrows-Wheeler Aligner [40]. The FeatureCounts tool was used to assign mapped reads to genomic features, and DESeq2 was used to assess changes in gene expression in RNA expression libraries [41,42]. To differentiate fully upregulated genes from those having accumulation of reads in specific parts of the transcripts, we designed a pipeline (S1 Fig). Each gene was divided into non-overlapping 10-nucleotide (nt) windows and the mean read depth (coverage) in each window was calculated for Δrnj and WT replicates, using the Bedtools Genomecov function with the -pc option to computationally fill in coverage between reads [43], and a custom Python script to compute the coverage for each 10 nt window. The log₂ ratio of coverage in Δrnj/WT was computed for each window in each gene. Then, within each gene, the window with the median log₂ coverage was identified. DESeq2 also reports log₂ ratios for each gene, and the genome-wide mean of these log₂ ratios was close to zero as expected given the assumption that the majority of genes do not differ between strains. However, the genome-wide mean of the median 10 nt window log₂ ratios was -0.07. We therefore normalized the log₂ ratios of the median 10 nt windows for all genes by adding 0.07. Next, we determined the absolute difference between the DESeq2 log₂ fold change and the normalized median 10 nt window log₂ ratio for each gene. The standard deviation (SD) of these differences was then calculated. Among the genes reported as differentially expressed by DESeq2, we set a cutoff of two SDs, such that genes with differences of ≥2 SDs were classified as partially up or down-regulated, and genes with differences of <2 SD were classified as fully up or down-regulated. For genes that were partially up-regulated, the minimum free energy (MFE) and thermodynamic ensemble predictions of the overexpressed regions and adjacent upstream and downstream regions of the same length were determined using the RNAfold web server [44].

To specifically evaluate the characteristics of the 5’ regions of RNA fragments that accumulated in the Δrnj strains, the normalized read depths for 5’ ends in the 5’ end-directed non-pyrophosphohydrolase-treated libraries were compared. 5’ ends were considered if they had a minimum read depth of five reads and the highest read depth in a 5-nt window (S6 Table). Those 5’ ends with read depth ratios ≥10 in Δrnj/WT were classified as enriched in the Δrnj strain. A total of 381 5’ ends were classified as enriched. For a comparison, we selected 1,000 5’ ends that were equally represented in the WT and Δrnj strains. The first 50 nt downstream of each 5’ end were analyzed. The minimum free energy (MFE) of each 50-nt sequence was determined using the RNAfold web server [44], and the %GC was determined.

Analysis of the sequence and structural contexts of mutations

Published crystal structures were used to visualize the structural contexts of RNase J point mutations using Mol* Viewer [45] on RCSB PDB. The structures used were Deinococcus radiodurans 4XWW and 4XWT [46], Thermus thermophilus 3T3O [47] and 3BK1 and 3BK2 [48], and Streptomyces coelicolor 5A0T [49]. To assess the sequence context similarity between Mtb RNase J and the orthologs used for structural analysis, RNase J orthologs were aligned by Clustal Omega 1.2.4 [50].

Mutations in RNase J are associated with drug resistance in clinical Mtb strains

We recently identified mutations in non-canonical drug resistance genes that are strongly associated with drug resistance in clinical Mtb strains isolated from China [16]. As mutations in these genes do not confer high-level drug resistance, we hypothesized that they may arise as intermediate steps in the acquisition of high-level drug resistance. One of the genes associated with drug resistance was Rv2752c, hereafter referred to as rnj, which encodes the bifunctional exo/endoribonuclease RNase J [23]. Genome-wide association studies (GWAS) revealed that mutations in rnj were associated with INH resistance in Mtb [16]. Our results were consistent with two previous GWAS studies that also identified rnj mutations in association with multidrug resistance in Mtb [17,18].

In our data set, rnj was highly polymorphic in INH-resistant strains compared to drug sensitive strains and mutations were distributed throughout the CDS (Fig 1A, Table 1 and S3). In INH-resistant strains nonsynonymous mutations were the most prevalent (∼80% of the total mutations), while frameshift and nonsense mutations were also found throughout the gene, suggesting selection for loss of function variants. We used crystal structures and mutational analysis data from three other bacterial RNase J orthologs to analyze the structural contexts of clinical Mtb rnj mutations (Figs 1B and S2) [46–49]. Several mutations caused non-conservative changes to residues involved in Zn²⁺ coordination at the catalytic site (S3 Table) [46–49]. Several others introduced prolines predicted to disrupt α-helices or β-sheets in the C-terminal domain, which is required for dimerization and efficient cleavage in Bacillus subtilis [48]. Some mutations affected residues involved in RNA binding [46,47,49].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. RNase J is mutated in many clinical Mtb strains.

A. Vertical lines indicate mutations identified in clinical isolates in Hicks et al., 2018. Frameshift (red), nonsynonymous (blue), synonymous (orange), and nonsense (green) mutations are highlighted. Positions in the Mtb H37Rv genome are indicated in parentheses. Numbers in grey indicate mutations that evolved twice independently. B. Point mutations in INH-resistant clinical Mtb strains modeled on a Thermus thermophilus crystal structure (PDB 3T3O, Dorleans et al., 2011). Gold and gray indicate the two monomers, with Mtb mutated residues indicated in red on the gray monomer. A 4-mer RNA is indicated in blue. A catalytic zinc is indicated by “Zn.” The second catalytic zinc is not present in the structure due to an active-site mutation needed to capture a stable RNA co-complex (Dorleans et al., 2011). Structures were annotated on RCSB PDB by Mol* Viewer (Sehnal et al., 2021). The view on the left is oriented to best visualize mutations in the C-terminal domain while the view on the right is oriented to best visualize the mutations in the catalytic region.

https://doi.org/10.1371/journal.ppat.1010705.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Summary of rnj mutations identified in clinical strains of M. tuberculosis and their predicted consequences.

https://doi.org/10.1371/journal.ppat.1010705.t001

Bellerose and collaborators [51,52] recently found, using a TnSeq screening approach, that disruption of rnj increased survival of Mtb in mice treated with the clinical first line drug regimen. Taking this finding together with the observation of potentially deleterious mutations described above, we hypothesized that loss of RNase J function might confer a survival advantage to the bacterium in the presence of drug treatment, and that mutations in clinical strains arise as a part of an evolutionary pathway to the acquisition of high-level drug resistance.

Loss of RNase J increases tolerance to several drugs

To investigate the link between RNase J and drug sensitivity in Mtb, we first constructed RNase J deletion (Δrnj₆₂₃₀) and complemented strains (Δrnj₆₂₃₀::rnj) in the Mtb mc²6230 background (ΔpanCD, ΔRD1) [31]. To determine the effects of RNase J on drug sensitivity, we assessed both MICs and bactericidal activities of clinically relevant drugs. The MICs for RIF and INH were the same for the Δrnj₆₂₃₀ strain and its WT parent (0.125 μg/mL for both drugs), indicating that loss of RNase J does not confer resistance to either drug. However, the Δrnj₆₂₃₀ strain displayed increased survival in the presence of lethal concentrations of RIF and INH compared to the WT and the complemented strains (Fig 2A), suggesting that the absence of RNase J leads to increased drug tolerance. We observed a similar behavior when exposing the strains to lethal concentrations of EMB, OFX, CLA, and ERY (Fig 2A).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Loss of RNase J increases tolerance to several drugs.

A. Time-kill curves in presence of the indicated drugs. The concentrations used were: 0.6 μg/mL RIF, 2.4 μg/mL INH, 2.5 μg/mL CLA, 1 μg/mL OFX, 2 μg/mL EMB, or 500 μg/mL ERY. B. Time-kill curves in presence of lethal concentrations of both INH (2.4 μg/mL) and OFX (2 μg/mL). Both experiments were performed using Mtb mc²6230 strains. *p<0.05, **p<0.01, ***p<0.001 two-way ANOVA for comparisons of Δrnj to WT. FDR 0.05 (Benjamini and Hochberg). Curves are representative of at least two independent experiments.

https://doi.org/10.1371/journal.ppat.1010705.g002

We sought to verify these results in virulent H37Rv, constructing RNase J deletion (Δrnj_H37Rv) and complemented strains (Δrnj_H37Rv::rnj). We tested killing by RIF, INH, and OFX, and found that in each case, the Δrnj_H37Rv strain had a survival advantage compared to the WT and complemented strains (S3 Fig). Furthermore, complementation of Δrnj_H37Rv with rnj^H86A, which is predicted to be catalytically dead [48], did not restore the WT sensitivity phenotypes for RIF, INH, or OFX. While we cannot exclude the possibility that Mtb RNase J with the catalytic site mutation is unstable, these data are consistent with the idea that the greater survival of the Δrnj_H37Rv strain is due to loss of RNase J catalytic activity (S3 Fig).

To further characterize the survival advantage conferred by loss of rnj, we exposed cultures to a combination of drugs (INH and OFX) at lethal concentrations to prevent outgrowth of resistant mutants. We observed that both strains reached a plateau at a similar number of CFUs (Fig 2B), suggesting that loss of RNase J does not affect persistence but rather increases drug tolerance by reducing the killing rate. However, we cannot exclude the possibility that RNase J affects persister levels, since detection and quantification of persister cells are affected by the choice of assay and culture conditions (many works including [7,53–56]).

We noted that Δrnj₆₂₃₀ colonies were smaller than those of the WT and complemented strains. We therefore measured the growth characteristics of the three strains (Fig 3A). The growth rate of Δrnj₆₂₃₀ in mid-log phase was statistically indistinguishable from the growth rates of the WT and complemented strains. However, Δrnj₆₂₃₀ had a slightly longer lag phase than WT cells. The lag phase delay was condition-dependent, as we did not observe it when growth curves were started at a higher OD (S4A Fig), or when the strains were grown in minimal media (S4B Fig). To test the possibility that the better survival of Δrnj was due to the subtle growth defect observed at low ODs, we sub-cultured Δrnj₆₂₃₀ and WT cultures to a low OD, added INH at a lethal concentration after one, two, three, or four days, and compared the fraction of survival after two days of incubation with drug (Fig 3B). We observed that the higher fraction of survival for Δrnj₆₂₃₀ was consistent regardless of the growth phase at which drug was added (Fig 3C; growth phases spanned lag phase through late log phase). In addition, we performed RIF and INH time-killing curves in minimal media, where Δrnj₆₂₃₀ does not show a longer lag phase, and found that Δrnj₆₂₃₀ had better survival than WT in this condition as well (S4C Fig). While we cannot exclude the possibility that a longer lag phase contributes to the tolerance of Δrnj₆₂₃₀ to some drugs, these data suggest that unlike what was previously reported in E. coli in response to intermittent treatment with ampicillin [57], lag phase differences do not explain the increased tolerance of Δrnj₆₂₃₀ to continuous exposure to lethal concentrations of INH or RIF.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Drug tolerance in Δrnj Mtb is not due to its lag phase growth defect.

A. Growth kinetics of Mtb mc²6230 WT, Δrnj, and Δrnj::rnj in 7H9 media. Slopes were statistically equivalent for all three strains in mid-log phase (days 3–6, linear regression). ODs were significantly different for Δrnj vs WT for all time-points from day 3 onward (t-tests with FDR 1% correction) B. Schematic of experiment to determine the effect of growth phase on drug survival. Created with Bio.Render.com. C. Data from the experiment shown in B. Bars indicate the proportion of cells that survived after two days of incubation with INH (2.4 μg/mL) starting at the indicated days. This proportion is the value in the lower yellow bar in B divided by the value in the upper yellow bar in B. *p<0.05, **p<0.01, ***p<0.001, t-tests with Benjamini and Hochberg FDR 0.05.

https://doi.org/10.1371/journal.ppat.1010705.g003

RNase J affects rRNA processing and expression of a subset of genes and mRNA fragments

To investigate how RNase J mediates drug tolerance in Mtb, we assessed its role in RNA metabolism by performing RNAseq expression profiling as well as RNA 5’ end mapping. These studies were done with WT_H37Rv and Δrnj_H37Rv strains transformed with the empty vector pJEB402 as well as the Δrnj_H37Rv strain complemented with rnj under a strong constitutive promoter, Δrnj_H37Rv::rnj_OE [58]. As a quality control, we evaluated the 23S rRNA as previous work has shown that RNase J is necessary for 23S rRNA maturation in M. smegmatis [23]. Consistent with these data, we found that the 23S rRNA transcript was 15 nt longer in Δrnj_H37Rv compared to WT_H37Rv (S5 Fig), indicating that RNase J also plays a role in 23S rRNA processing in Mtb.

Standard analysis of the RNAseq expression data using DESeq2 indicated that 57 and 16 genes had increased or decreased transcript abundance, respectively, in Δrnj_H37Rv compared to WT_H37Rv (fold change ≥1.5, adj p value ≤0.01) (Fig 4A, S4 Table). Comparison of changes in transcript abundance in the Δrnj_H37Rv and Δrnj_H37Rv::rnj_OE strains showed a significant negative correlation (S6 Fig), reflecting the opposite effects of rnj overexpression and deletion. However, visual inspection of RNAseq expression library coverage revealed that several differentially abundant transcripts did not have increased abundance across the entire gene in Δrnj_H37Rv, but rather displayed increased read coverage for only short segments of the genes in question (Fig 4B). We hypothesized that these short RNA fragments might have arisen from incomplete degradation of mRNAs in the absence of RNase J. Thus, standard DESeq2 analysis is not sufficient to discriminate between genes for which the abundance of the whole transcript changes and genes marked by the accumulation of RNA fragments. To address this issue, we developed a bioinformatics pipeline (see Materials and Methods and S1 Fig) to identify genes for which the abundance of the whole transcript was altered. Of the 57 genes that were reported as significantly overexpressed in Δrnj_H37Rv by DESeq2, 31 reflected increases in the whole transcript (Fig 4A, red dots and S4 Table) while the remaining genes had increased abundance of only subsections of their transcripts (Fig 4A, light pink dots, and S4 Table). All the genes with reduced abundance in Δrnj_H37Rv except for one (Fig 4A, light blue dot) showed changes in the abundance of the entire transcript (Fig 4A, blue dots, and S4 Table).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. RNase J affects expression of genes and causes highly structured mRNA fragments to accumulate in Mtb.

A. Volcano plot showing the genes affected by RNase J in Mtb H37Rv strains. Partially and fully over/under expressed genes are distinguished with different colors. B. Schematics of the read depth of two genes presenting full overexpression (upper panel) or partial overexpression (lower panel) in Δrnj. C. Read depth of expression libraries in Mtb H37Rv strains for four genes that displayed partial overexpression in Δrnj. Grey lines below the arrows denote the sequences targeted by qPCR for regions of the genes displaying accumulation of short fragments in Δrnj (R2) and regions with similar read coverage in all strains (R1). For all H37Rv experiments, the WT and Δrnj strains contained the empty vector pJEB402. D. Determination of half-life for the gene regions shown in C using Mtb mc²6230 strains. *p<0.05, **p<0.01, ***p<0.001. E. The minimum free energy of folding was predicted for overexpressed regions in A and for regions of equal lengths immediately upstream and downstream of each overexpressed region. Upstream and downstream MFEs were only calculated when the region fell within the coding sequence of the gene. Blue indicates overexpressed regions for which only the upstream adjacent region was available, purple indicates regions for which only the downstream adjacent region was available, and red indicates regions for which both upstream and downstream regions were available. **p<0.01, Wilcoxon matched-pairs signed rank test. F and G. 5’ end-mapping libraries were used to identify transcripts overexpressed in Δrnj. The 50 nt downstream of both overexpressed and unchanged 5’ ends were analyzed to predict the minimum free energy of secondary structure formation and determine G+C content. ****p<0.0001, Mann Whitney test.

https://doi.org/10.1371/journal.ppat.1010705.g004

RNase J has a specialized role in degrading mRNA fragments with strong secondary structure and high G+C content

We hypothesized that the transcript fragments that accumulated in the Δrnj_H37Rv strain could be RNase J targets that were inefficiently degraded in the mutant strain. To test this, we chose four genes with partial-transcript abundance increases in the Δrnj_H37Rv strain and measured the half-lives of both the overrepresented regions of each transcript and regions for which read coverage was similar in both strains (Fig 4C). The overrepresented regions had longer half-lives in the absence of rnj, while half-lives of equally abundant regions did not differ between the strains (Fig 4D).

To understand why RNase J has an apparently rate-limiting role in degradation of certain mRNA fragments, we calculated the predicted minimum free energy of folding (ΔG) for each overrepresented region (S5 Table) and for adjacent upstream and downstream regions of equal size, when those regions were present within the same coding sequence. The ΔGs of the overrepresented regions were in most cases lower than the ΔGs of the upstream adjacent regions (Fig 4E, p = 0.0021), indicating that formation of secondary structure was more energetically favorable for the overrepresented regions. There was an apparent trend toward the ΔGs of the overrepresented regions also being lower than those of the downstream adjacent regions, but this was not statistically significant. We repeated the analysis using the free energies of thermodynamic ensemble predictions [44] and the results were nearly identical (p = 0.0017 for accumulated regions vs upstream regions). Inspection of the minimum free energy structures did not reveal any trends regarding the types and locations of predicted secondary structure elements. The library construction method did not permit identification of the exact 5’ and 3’ ends of each overrepresented region, which precluded analysis of the basepairing probabilities of those ends.

Since the overrepresented regions appeared to be more structured than adjacent upstream regions and RNase J is known to have 5’ to 3’ exonuclease activity, we hypothesized that RNase J may contribute to degradation of transcripts that have high levels of secondary structure near their 5’ ends. We therefore used a separate 5’-end-directed RNAseq dataset to analyze the properties of 5’-end-adjacent sequences in WT_H37Rv and Δrnj_H37Rv transcriptomes. We assessed the G+C content and predicted secondary structure of the 50 nt sequences adjacent to RNA 5’ ends that had increased abundance in the Δrnj_H37Rv strain or were equally abundant in the Δrnj_H37Rv and WT_H37Rv strains (S6 Table). We found that the 5’-end-adjacent sequences with increased abundance in Δrnj_H37Rv had significantly more negative predicted minimum free energies of folding (ΔG) than those that were equally represented in Δrnj_H37Rv and WT_H37Rv (Fig 4F). In addition, the median G+C content of the sequences overrepresented in Δrnj_H37Rv was significantly higher than that of the equally represented sequences (Fig 4G). These results are consistent with the idea that RNase J targets RNAs with relatively strong secondary structure, particularly near their 5’ ends.

Having found that loss of RNase J increased stability of short RNA fragments, we wondered if the increased abundance of the fully overexpressed genes in Δrnj_H37Rv was a direct consequence of slower degradation rates or increased transcription. We measured the half-lives of six fully overexpressed genes and found that transcript stability was not altered (S7A and S8 Figs). However, these genes did not have high G+C contents, and technical difficulties prevented us from obtaining high-confidence half-lives for those fully overexpressed genes that were GC-rich (S7B Fig). There may therefore be genes that were fully overexpressed due to reduced degradation rates, and this should be addressed in future work. We also considered the possibility that the absence of RNase J could lead to accumulation of antisense transcripts that could affect mRNA degradation or translation. However, there were no changes in median antisense coverage for genes differentially expressed in Δrnj_H37Rv or for all expressed genes in this strain. Together, these data suggest that RNase J affects the transcript abundance of some genes through altered transcription rather than by altering their mRNA stability.

Genes differentially expressed in the absence of RNase J are enriched for sRNAs, PE/PPE family genes, SigM targets, and genes with roles in hypoxia response and carbon source switching

Examination of the genes that were fully overexpressed or underexpressed revealed several themes. First, the differentially expressed genes were enriched for sRNAs and genes of the PE/PPE family, including nine overexpressed PE_PGRS genes (S9 Fig). Second, six of the underexpressed genes (ppe50, ppe51, fadD26, ppsA, Rv0885, and Rv3137) were reported to be negatively regulated by the stress-responsive alternative sigma factor SigM, while one of the overexpressed genes (Rv3093c) was reported to be positively regulated by SigM [59]. This suggests that there may be increased SigM activity in Δrnj_H37Rv, although the sigM gene itself was not increased at the transcript level. Finally, the differentially expressed genes included several associated with hypoxia responses (the sRNAs MTS2823 and F6, and the protein-coding genes fdxA, ppe31, clgR, and Rv3740c) and several associated with utilization of various carbon sources (ppe50, ppe51, mcm1C, prpC, Rv1066) [60–68], suggesting that the metabolic status of the Δrnj_H37Rv may differ from that of WT_H37Rv.

Overexpression of the sRNAs Mts2823 and Mcr11 is necessary but not sufficient for INH tolerance in Δrnj Mtb

The sRNAs Mts2823 and Mcr11 were two of the most overexpressed genes in the Δrnj_H37Rv strain (S4 Table). Since sRNAs have been implicated in adaptation to different stresses, we sought to investigate if their increased expression in Δrnj contributed to drug tolerance. We therefore deleted each of the two sRNAs in both the WT₆₂₃₀ and Δrnj₆₂₃₀ strains and performed time-killing curves in presence of RIF or INH. We found that deletion of either of these sRNAs in the Δrnj₆₂₃₀ background decreased INH tolerance to levels near the WT₆₂₃₀ (Fig 5A, 5B, 5D and 5E). In contrast, deletion of these sRNAs had no effect on INH tolerance in the WT₆₂₃₀ background (Fig 5C and 5F). Thus, Mts2823 and Mcr11 are necessary for the INH tolerance conferred by loss of RNase J in the Δrnj₆₂₃₀ strain. To determine whether either sRNA was sufficient for INH tolerance, we then constructed strains overexpressing either Mts2823 (Mts2823_OE) or Mcr11 (Mcr11_OE) in the WT₆₂₃₀ background. However, we found that both showed drug sensitivity levels comparable to that of WT₆₂₃₀ (Fig 5C and 5F), suggesting that overexpression of either Mts2823 or Mcr11 alone is not sufficient to increase drug tolerance. No consistent effects were observed for strains with deletions or overexpression of these sRNAs in RIF, suggesting that effectors acting downstream of RNase J deletion may act in part via drug specific mechanisms (S10 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Overexpression of the sRNAs Mts2823 and Mcr11 is necessary but not sufficient for INH tolerance in Δrnj Mtb.

Time-killing curves in presence of INH (2.4 μg/mL) for strains with deletion or overexpression of Mts2823 (A-C) or Mcr11 (D-F) strains are shown. *p<0.05, **p<0.001 two-way ANOVA. Pink stars: comparison of Δrnj ΔMts2823 to Δrnj. Red stars: comparison of WT to Δrnj. Tan stars: comparison of Δrnj ΔMcr11 to Δrnj.

https://doi.org/10.1371/journal.ppat.1010705.g005

Downregulation of ppe50-ppe51 is necessary for the INH and RIF tolerance phenotypes of Δrnj Mtb, and deletion of ppe50-ppe51 confers drug tolerance to a WT strain

Two of the most strongly downregulated genes in the Δrnj_H37RV strain were ppe50 and ppe51, which are expressed in an operon and have previously been implicated in drug sensitivity [51,52,69]. To test the hypothesis that downregulation of ppe50 and ppe51 contributes to the drug tolerance of the Δrnj strains, we ectopically overexpressed them in the Δrnj₆₂₃₀ background and performed drug killing experiments. Overexpression of the ppe50-ppe51 operon in the Δrnj₆₂₃₀ strain impaired survival of Mtb in the face of both INH and RIF, producing drug sensitivity levels similar to those observed in the WT strain (Fig 6A). Thus, the dysregulated expression of ppe50-ppe51 is also necessary for the altered drug susceptibility associated with loss of rnj. To test if loss of ppe50-ppe51 is sufficient to confer drug tolerance, we deleted the ppe50-ppe51 operon from WT₆₂₃₀ and measured killing by INH and RIF (Fig 6B). Δppe50-ppe51₆₂₃₀ displayed increased tolerance to both drugs compared to WT₆₂₃₀, despite rnj being intact. Normal drug sensitivity was restored when we complemented Δppe50-ppe51₆₂₃₀ with ectopically expressed ppe50-ppe51. Loss of ppe50-ppe51 in a WT background therefore appears to be sufficient to confer tolerance to both RIF and INH.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Downregulation of ppe50-ppe51 is required for the drug tolerance phenotype of Δrnj Mtb, and deletion of ppe50-ppe51 is sufficient to induce drug tolerance in a WT background.

Time-kill curves in the presence of RIF (0.6 μg/mL) or INH (2.4 μg/mL) in Mtb mc²6230 strains. A. ppe50-ppe51 was ectopically expressed from a strong constitutive promoter in the Δrnj strain. *p<0.05, **p<0.01, two-way ANOVA. Blue stars: comparison of WT::Empty_pJEB402 to Δrnj::Empty_pJEB402. Magenta stars: comparison of Δrnj::ppe50/ppe51 to Δrnj::Empty_pJEB402. B. ppe50-ppe51 was deleted from the WT strain and then ectopically expressed from a strong constitutive promoter. *p<0.05, **p<0.01 ***p<0.001, two-way ANOVA. Blue stars: comparison of WT to WT Δppe50-ppe51. Lavender stars: comparison of WT Δppe50/ppe51::ppe50/ppe51 to WT Δppe50-ppe51. FDR 0.05 (Benjamini and Hochberg) for all comparisons.

https://doi.org/10.1371/journal.ppat.1010705.g006