Loss of RNase J leads to multi-drug tolerance and accumulation of highly structured mRNA fragments in Mycobacterium tuberculosis

Abstract

Despite the existence of well-characterized, canonical mutations that confer high-level drug resistance to Mycobacterium tuberculosis (Mtb), there is evidence that drug resistance mechanisms are more complex than simple acquisition of such mutations. Recent studies have shown that Mtb can acquire non-canonical resistance-associated mutations that confer survival advantages in the presence of certain drugs, likely acting as stepping-stones for acquisition of high-level resistance. Rv2752c/rnj, encoding RNase J, is disproportionately mutated in drug-resistant clinical Mtb isolates. Here we show that deletion of rnj confers increased tolerance to lethal concentrations of several drugs. RNAseq revealed that RNase J affects expression of a subset of genes enriched for PE/PPE genes and stable RNAs and is key for proper 23S rRNA maturation. Gene expression differences implicated two sRNAs and ppe50-ppe51 as important contributors to the drug tolerance phenotype. In addition, we found that in the absence of RNase J, many short RNA fragments accumulate because they are degraded at slower rates. We show that the accumulated transcript fragments are targets of RNase J and are characterized by strong secondary structure and high G+C content, indicating that RNase J has a rate-limiting role in degradation of highly structured RNAs. Taken together, our results demonstrate that RNase J indirectly affects drug tolerance, as well as reveal the endogenous roles of RNase J in mycobacterial RNA metabolism.

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.

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.

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.

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.

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.

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.