Defining the role of RNase E in the mycobacterial degradosome-like network

Abstract

mRNA degradation is a fundamentally important process that is regulated in response to stress in the globally important pathogen Mycobacterium tuberculosis. Several mycobacterial ribonucleases (RNases) are hypothesized to function together to coordinate mRNA degradation, but the interactions among them are mostly undefined. One of the rate-limiting enzymes, RNase E, contains intrinsically disordered regions (IDRs). Here, we aimed to define the interactions between major mycobacterial mRNA degradation enzymes and identify the function(s) of the two IDRs of RNase E in the nonpathogenic model Mycolicibacterium smegmatis. We found that the two IDRs differentially impact mRNA degradation rates in vivo but are largely functionally redundant in their impacts on steady-steady transcript abundance. In vitro, the IDRs are uninvolved in catalysis but play major roles in RNA binding and interactions with other mRNA degradation enzymes, namely PNPase, RNase J, and RhlE1. In vivo, these enzymes localize with RNase E, but its IDRs play only a minor role, suggesting substantial redundancy in subcellular localization mechanisms. Collectively, we propose a degradosome-like network model in mycobacteria, held together by dynamic, transient interactions among RNA degradation enzymes and RNA that can be disrupted during physiologically relevant stress to allow for adaptability.

GRAPHICAL ABSTRACT

INTRODUCTION

Mycobacteria are a clinically relevant group of bacteria that includes Mycobacterium tuberculosis, the causative agent of tuberculosis, which causes millions of reported cases and deaths each year [1]. M. tuberculosis treatment is challenging and requires a 4–6 months course of multiple antibiotics [1]. During M. tuberculosis infections, the bacteria face several stressors from the human host such as hypoxia, low pH, and reactive oxygen and nitrogen species (reviewed in [2]), yet the bacteria can still survive and often tolerate antibiotics. One way that mycobacteria respond to energy stress in particular is global stabilization of their transcriptomes by decreasing mRNA degradation rates [3, 4], as it is energetically expensive to synthesize and degrade RNA. However, this stabilization cannot be explained by changes in the abundance of RNA processing and degradation enzymes [3, 5]. The mechanisms by which mycobacteria post-translationally regulate their RNA processing and degradation enzymes are unknown.

To understand the regulation of mRNA degradation, it is necessary to define the components of the molecular machinery that carries out this process and the protein-protein interactions that organize this machinery. Endoribonucleases, exoribonucleases, and RNA helicases are the principal enzymes that carry out mRNA degradation. In mycobacteria, the endoribonuclease RNase E, exonuclease PNPase, and dual endo/exoribonuclease RNase J have established roles in this process [6–9].

The essential endoribonuclease RNase E plays a rate-limiting step in the degradation of at least 80% of mRNAs in the nonpathogenic model organism Mycolicibacterium smegmatis [10], and its essentiality and large transcriptional impact in M. tuberculosis suggest that it has a similarly import role [11]. It also plays an important role in rRNA maturation [6, 7]. Structurally, RNase E is known to bind magnesium for efficient catalysis of RNA [6] and produces a 5’ monophosphate and 3’ hydroxyl upon RNA cleavage [12]. In Escherichia coli, zinc binding is required for multimerization of RNase E [12, 13] to form dimers, which are needed to form the catalytic interface, as well as tetramers (dimers of dimers) [6]. RNase E cleaves long but not short RNAs efficiently when multimerized as tetramers or larger species suggesting a post-translational regulatory mechanism [14].

In the Proteobacteria, RNase E has a long C-terminal intrinsically disordered region (IDR) that mediates interactions with other proteins [15, 16]. However, the IDR of RNase E in Caulobacter crescentus is required for an additional noncatalytic function: formation of liquid-liquid phase-separated biomolecular condensates [17]. Condensates represent a mechanism of cytoplasmic organization by the formation of membraneless organelles in both bacteria and eukaryotes (reviewed in [18, 19]). In C. crescentus, RNase E condensates allow for rapid assembly and disassembly of RNA degradation hubs for efficient degradation when needed and adaptability during stress [17, 20–22]. The condensates formed by RNase E can also recruit other enzymes to promote their activity in C. crescentus [20]. Strikingly, mycobacterial RNase E has two long IDRs flanking a central catalytic domain [15], suggesting potential additional functionalities in condensate formation and/or specific protein-protein interactions compared to the better-studied Proteobacterial RNase Es which have single IDRs.

Other relevant RNA processing and degradation enzymes in mycobacteria include RNase J, PNPase, and RhlE. RNase J is a dimer with dual endo- and 5’ to 3’ exoribonuclease activities [7] and has gained attention because its deletion in M. tuberculosis leads to multidrug tolerance [23, 24]. It has a non-essential role in rRNA maturation and a specialized role in degradation of a subset of highly structured mRNAs [7, 24]. PNPase is trimer with 3’ to 5’ exonuclease activity [8, 9] that can also act as a polymerase in the reverse direction to add 3’ poly-A tails, which in bacteria generally facilitate mRNA degradation (reviewed in [25]). It is essential in M. tuberculosis and M. smegmatis, suggesting that its roles in mRNA degradation and/or stable RNA maturation are required for growth [11]. Notably, mycobacteria lack two 3’ to 5’ exonucleases (RNase R and RNase II) that have redundant roles with PNPase in E. coli, where PNPase is not essential [26, 27]. RhlE is a DEAD Box RNA helicase that is ATP-dependent and proposed to participate in mRNA degradation in mycobacteria [11].

In some bacteria, multiprotein complexes of RNA degradation proteins termed RNA degradosomes have been described. These are mediated by specific interactions and thus are distinct from condensates of RNA degradation proteins, which are formed by many weak, low-specificity interactions. However, it is likely that degradosomes and condensates can coexist, with proteins such as RNase E participating in both types of interactions simultaneously [22]. Degradosomes are thought to be sites of accelerated RNA degradation due to the close proximity of related enzymes as well as sites of allosteric activation mediated by direct protein-protein interactions in some cases [16, 28, 29]. The core E. coli degradosome has been well-defined, with RNase E serving as a scaffold that binds RhlB (an RNA helicase), PNPase, and enolase with defined stoichiometries [30–35]. These stoichiometries can be impacted by factors such as the presence of RNA, and other degradation proteins associate with the core degradosome components at sub-stoichiometric levels that may vary depending upon condition [35]. For example, the composition of the degradosome can change in response to stress such as cold-shock [36]. RNase E binds its partner proteins via a long C-terminal IDR [15, 16, 37], and is membrane-bound in E. coli [38].

Variations of the degradosome have been described in other bacteria including C. crescentus, Bacillus subtilis, and Helicobacter pylori. In C. crescentus, RNase E binds aconitase and PNPase using its C-terminal IDR and interacts with either of two DEAD-box RNA helicases [39]: RhlB during log phase growth and RhlE during cold shock [40]. B. subtilis lacks RNase E but encodes the endoribonuclease RNase Y, which is considered a functional equivalent of RNase E. RNase Y facilitates direct and indirect interactions among PNPase, RNase J1, RNase J2, CshA (an RNA helicase), enolase, and phosphofructokinase [41] using its IDR [42]. This association of proteins appears less stable than the E. coli degradosome; for example, RNase Y can only pull down its interacting proteins from cell lysates if crosslinking is first performed [42]. A complex of these components with defined stoichiometry has not been defined. The B. subtilis RNA degradation proteins have therefore been described as forming a degradosome-like network involving transient protein-protein interactions [41–45]. In Helicobacter pylori, an RNase J-based degradosome has been proposed in which it interacts with RhpA (an RNA helicase) and associates with the membrane independent of the RNase Y that this organism encodes [46].

There is clearly great diversity in the composition of RNA degradosomes across different species, as well as a diversity of modes of interaction among degradosome proteins. The mycobacterial RNA degradosome has yet to be directly defined. Immunoprecipitation and mass spectrometry experiments have implicated RNase E, PNPase, RNase J, and RhlE as being potential degradosome members in M. tuberculosis [11], but the design of these experiments did not distinguish direct interactions from indirect protein-protein interactions or indirect interactions mediated by RNA. The only direct interaction tested and confirmed was between PNPase and RNase J [11]. Beyond that single interaction, no direct evidence has previously shown the arrangement and composition of a mycobacterial degradosome.

Here, we use the nonpathogenic model organism M. smegmatis to investigate the roles of the IDRs in mycobacterial RNase E as well as define the protein-protein interactions that organize the major mycobacterial RNA degradation proteins. We found that the IDRs of RNase E have major and partially redundant roles in RNA binding, which is reflected in the phenotypic impacts of their deletion in cells. The IDRs of RNase E also have roles in condensate formation and mediate interactions with three other RNA degradation proteins in vitro, but remarkably, these functions only have minor impacts on subcellular localization of the RNA degradation machinery in vivo. Other components of the RNA degradation machinery also have direct physical interactions, which may underlie the robustness of their localization patterns in cells. Localization of most of the major mRNA degradation proteins became more diffuse in response to energy stress, consistent with the idea that subcellular localization may contribute to global regulation of mRNA degradation rates. Together, our findings suggest that mycobacterial RNA degradation proteins are organized into a degradosome-like network that is tunable during clinically relevant stress conditions.

MATERIALS AND METHODS

Bacterial Strains and Culture Conditions:

Middlebrook 7H9 liquid medium supplemented with 5 g/L bovine serum albumin fraction V, 2 g/L glucose, 0.85 g/L NaCl, and 4 mg/L catalase, 0.2% glycerol, 0.05% Tween 80 was used to grow M. smegmatis Mc²155 and its derivatives. Middlebrook 7H10 solid medium supplemented with 5 g/L bovine serum albumin fraction V, 2 g/L glucose, 0.85 g/L NaCl, and 4 mg/L catalase, 0.5% glycerol, was used to grow M. smegmatis Mc²155 and its derivatives. Liquid cultures were grown at 37°C with 200 rpm shaking. Carbon starvation media for microscopy was composed of 34.2 μM EDTA, 6.8 μM CaCl₂ × 2 H₂O, 491.9 μM MgCl₂ × 6 H₂O, 0.83 μM Na₂MoO4 × 2 H₂O, 1.8 μM CoCl₂ × 6 H₂O, 0.80 μM CuSO₄ × 5 H₂O, 5.1 μM MnCl₂ × 4 H₂O, 7.0 μM ZnSO₄ × 7 H₂O, 180 μM FeSO₄ × 7 H₂O, 8.9 mM K₂HPO₄, 70.8 mM NaH₂PO₄, 15 mM (NH₄)₂SO₄, and 0.05% tyloxapol in water. Antibiotic concentrations for M. smegmatis were 25 μg/mL kanamycin, 150 μg/mL hygromycin B, 40 μg/mL nourseothricin, and 1 μg/mL apramycin.

All M. smegmatis strains used in this study are listed in Supplemental Table 1, and all M. smegmatis plasmids are listed in Supplemental Table 2. For experiments shown in Figures 2, 4A–F, 6, and 7 and Supplemental Figure 1, a single copy of the indicated version of the rne gene was present at an L5-integrated plasmid expressed from its native promoter and 5’ UTR (436 nt upstream of the native start codon). These were constructed by first integrating into strain Mc²155 a copy of rne with its native promoter/UTR and an N-terminal 6xhis-3xFLAG tag into the L5 site on a kan-marked plasmid, then deleting all but the last 150 nt of the native rne coding sequence by two-step homologous recombination. Finally, the resulting strain was transformed with NAT-marked L5-integrating plasmids encoding the desired RNase E variants (N-terminal his-FLAG tags and N-terminal mCherry fusion with full-length RNase E or various domain deletions), and colonies in which the NAT-marked plasmid replaced the kan-marked plasmid were identified.

Figure 2. Deletion of the two IDRs of RNase E in M. smegmatis has differential impacts on cell size, growth rate, mRNA degradation, and gene expression.

A) Cell length of RNase E IDR mutants. Cells were imaged on a spinning disk microscope, and cell length was measured on ImageJ with the FIJI plugin. The circle colors represent biological replicates, and the squares represent the median of that biological replicate. Black line represents the mean of the medians. Significance stars represent a one-way ANOVA with Kruskal-Wallis multiple comparisons test comparing each IDR mutant to full length. B) Growth curve of the IDR deletion strains. Points represent the average of three biological replicate strains. Linear regression was used for significance testing. C) RNA half-lives of selected genes in the IDR mutants. Longer half-lives indicate less efficient RNA degradation. Transcript half-lives for the indicated genes were measured by blocking transcription with 150 μg/mL rifampicin and measuring RNA abundance at several timepoints by quantitative PCR. Error bars denote 95% confidence interval. Half-lives were compared using linear regression analysis. D-E) RNAseq was used to compare the steady-state transcriptomes of strains expressing the indicated mutants or full-length RNase E. Genes were classified as differentially expressed if their log₂ foldchange was <-1 (downregulated) or >1 (upregulated) in the truncations compared to full-length, with adjusted p <0.05. In A-C, ns p > 0.05, *** p ≤ 0.001, **** p ≤ 0.0001

Figure 4. RNase E IDRs play a minor role in condensate formation in vivo.

A) Representative cells from spinning disk microscopy of mCherry::RNase E with or without 100 μg/mL rifampicin. B) In ImageJ with the FIJI plugin, a line was drawn through each cell and gray value versus distance was plotted. Cell regions where the signal intensity was greater than the mean were counted as foci, and the number of foci was normalized by cell length. A Mann-Whitney test was used for significance testing. C) A schematic of mCherry tagged RNase E constructs imaged by spinning disk microscopy with representative cells shown in D). Cells were quantified in E) using the same method for foci per micron calculations as in B) and in F) using coefficient of variation of signal intensity for a line drawn through the length of each cell. A one-way ANOVA with the Kruskal-Wallis multiple comparisons test was performed. G) A schematic of constructs used to assess the ability of RNase E IDRs to induce foci formation by mScarlet. Full-length RNase E was used as a positive control. Images obtained by spinning disk microscopy with representative cells are shown in H) and quantified in I) and J) in the same way as panels B) and F), respectively. For panels B, D-E, and I-J, the circle colors represent biological replicates, and the squares represent the median of that biological replicate. Black line represents the mean of the medians. ns p > 0.05, * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001

Figure 6. Localization of degradosome-like network proteins in vivo.

Representative cells from spinning disk microscopy of RNase E::mCherry with IDR mutants and A) PNPase::eGFP, C) RNase J::dendra2, and E) RhlE1::dendra2. Quantification of localization in B), D), and F), respectively, was done by drawing a line through a cell in ImageJ with the FIJI plugin, plotting green versus red channels, and computing the Pearson’s R value. Values at +1 indicate complete correlation, 0 indicates no correlation, and −1 indicates anticorrelation. The circle colors represent biological replicates, and the squares represent the median of that biological replicate. Black line represents the mean of the medians. A one-way ANOVA with the Kruskal-Wallis multiple comparisons test was performed. ns p > 0.05, **** p ≤ 0.0001

Figure 7. Localization of degradosome-like network proteins in vivo during carbon starvation stress.

A) Single-tagged strains of mCherry::RNase E, PNPase::eGFP, RNase J::dendra2, and RhlE1::dendra2 were imaged by spinning disk microscopy during both carbon rich and carbon starvation conditions. Foci per micron was calculated as in Figure 4B. Representative cells from spinning disk microscopy of RNase E::mCherry with IDR mutants and B) PNPase::eGFP, D) RNase J::dendra2, and F) RhlE1::dendra2 are shown during carbon rich and carbon starvation conditions. Quantification of localization in C), E), and G), respectively, was done as in Figure 6 B, D, and F. The circle colors represent biological replicates, and the squares represent the median of that biological replicate. Black line represents the mean of the medians. A one-way ANOVA with the Kruskal-Wallis multiple comparisons test was performed. ns p > 0.05, **** p ≤ 0.0001

For experiments shown in Figure 4G–J, full-length RNase E or the indicated RNase E IDRs were expressed as fusions with mCherry from the native rne promoter and 5’ UTR from Giles-integrating plasmids in strains where the native copy of rne was intact.

For experiments shown in Figures 6–7, eGFP-tagged PNPase or Dendra2-tagged RhlE1 or RNase J were expressed from the UV15 promoter and 5’ UTR [47] from Giles-integrating plasmids.

Lennox Luria-Bertani broth and solid media were used to grow E. coli. Antibiotic concentrations used for E. coli were 50 μg/mL kanamycin, 250 μg/mL hygromycin B, 60 μg/mL nourseothricin, and 20 μg/mL chloramphenicol. E. coli NEB-5-alpha (New England Biolabs) was used for cloning. HiFi Assembly (New England Biolabs) was used to make all plasmids for this study.

RNase E Variants Growth Assay:

RNase E variants were grown to mid-log phase (OD₆₀₀ of ~0.5) then diluted to an OD₆₀₀ of 0.05 in 5 mL. Every 3 hours for 9 hours, a 1 mL aliquot was taken from the shaker, with each tube only being removed for approximately 10 seconds. Cultures were read by OD₆₀₀, diluted in duplicate, and 5 μL spots were plated in duplicate (four spots total) on drug-free plates. Colonies were counted on day 2 and additional colonies were counted on day 3. The four spots were averaged, and CFU/mL was normalized to time 0 and reported. Significance testing was done by comparing the slopes of the line for each variant to full-length in GraphPad Prism.

RNA Half-lives:

RNA extraction, cDNA synthesis, and quantitative PCR to determine mRNA half-lives were done as described in [3].

RNASeq:

Strains were grown in triplicate to an OD₆₀₀ of ~0.42, and RNA was extracted as described in [3]. rRNA was depleted, and Illumina-compatible RNAseq libraries were constructed as described in [48, 49]. Libraries were sequenced at the UMass Medical School Deep Sequencing core facility on an Illumina HiSeq instrument to obtain 100 nt paired end reads. Reads were aligned to the NC_008596 reference genome using Burrows-Wheeler Aligner [50]. The FeatureCounts tool was used to assign mapped reads to genomic features. DESeq2 [51] was used to assess changes in gene expression. Genes with an adjusted p-value less than 0.05 and log2 fold changes >2 or <-2 were regarded as differentially expressed. Venn Diagrams were made with BioVenn [52]. Raw and processed data are available on GEO, accession number GSE313996.

Microscopy:

Nikon CSU W spinning disk confocal microscope with a CFI 60 plan apochromat λ D 100X oil immersion objective was used to obtain microscopy images. 1 mL of exponential phase M. smegmatis cultures were washed once and resuspended in less than 200 μL of 7H9 without dextrose, catalase, or BSA. 10 μL of cells were spotted onto 1.2% agarose pads (w/v agarose in water) and imaged in a chamber at 37°C. Brightfield channel used a 300 ms exposure time. mCherry and mScarlet were imaged with a 561 nm excitation laser and a ET605/52m emission filter. mCherry used 80% laser power and 6 s exposure time, and mScarlet used 80% laser power and 3 s exposure time. Dendra2 and eGFP were imaged using a 488 nm excitation laser and a ET525/36m emission filter. For Dendra2, 80% laser power and 200 ms exposure time were used, and for eGFP, 80% laser power and 100 ms exposure time were used. For rifampicin treatment, exponential phase cultures were treated for 30 minutes with 100 μg/mL rifampicin at 37°C shaking at 200 rpm. Images were acquired, denoised, and deconvoluted using Nikon Elements software (5.42.06).

Image analysis was performed using ImageJ with the FIJI plugin. For cell length measurements, straight lines were drawn through cells, and the measure tool was used. For assessment of the number of foci and coefficient of variation, segmented lines were drawn through cells in the brightfield channel, and the plot profile feature was used in the fluorescent channel. Foci were defined as instances where the signal intensity plotted along the length of the cell rose above and then dipped below the mean intensity. The number of foci in each cell was divided by the cell length to calculate peaks/micron. CV values (standard deviation/mean) were calculated for each cell from the signal intensity values along the length of the line. For assessment of colocalization, segmented lines were drawn through the cell in the brightfield channel, and the plot profile was plotted in both fluorescent channels. The intensity in the eGFP or dendra2 channel was plotted versus the mCherry channel, and the Pearson’s R correlation value was calculated for each cell in GraphPad Prism. For all microscopy analyses, 50 cells were counted in each biological replicate, and there were three biological replicates per condition.

CRISPRi Growth Curve for PNPase Functionality:

Exponential phase cultures were diluted to OD 0.005 and treated with either water (negative control) or 50 ng/mL anhydrous tetracycline (ATc) to induce the CRISPRi system in a 96 well plate. The plate was read by an EPOCH2 microplate reader (BioTek^®) using Gen5 version 3.04 BioTek^® software at 37°C with continuous orbital shaking at 1807 cpm 1mm at 600 nM every 10 minutes for 2 days. Biological triplicates were analyzed.

Rifampicin Killing Curve for RNase J Functionality:

Biological triplicate strains were diluted to an OD₆₀₀ of 0.1 in 5 mL of media with or without a final concentration of 12 μg/mL rifampicin. 5 μL spots were diluted and plated in duplicate on drug-free plates every day for 5 days. Colonies were counted on day 2 and additional colonies were counted on day 3. The average of the spots was reported, and CFU/mL was normalized to time 0 and reported.

CRISPRi Growth Assay for Lysates:

M. smegmatis cultures were grown to mid-log phase (OD₆₀₀ of ~0.5) then diluted to an OD₆₀₀ of 0.05 in 10 mL. 50 ng/mL anhydrous tetracycline to induce the CRISPRi system or water as a negative control was added to each culture. Every 3 hours for 12 hours, a 1 mL aliquot was taken from the shaker, with each tube only being removed for approximately 10 seconds. Cultures were read by OD₆₀₀. Biological triplicates were analyzed.

Cell Lysates Cleavage Assays:

10 mL M. smegmatis cultures were grown to exponential phase and pelleted by centrifugation at 10 minutes at 3,900 rpm and 4°C. Cells were washed with 10 mL of 7H9 containing no BSA, dextrose, or catalase 3 times then resuspended in 1 mL of 20 mM tris-HCl pH 7.9, 100 mM NaCl, 5% glycerol, 0.01% IGEPAL, 0.1 mM DTT, 10 mM MgCl2 and 1X Halt Protease Inhibitor Cocktail, EDTA-Free (ThermoFisher). Cells were lysed with a FastPrep-24 bead-beater (MP Biomedicals) using 4 cycles of 30 sec at 6.5 m/s, 1 minute on ice between cycles in tubes containing 100 μm zirconium beads (OPS Diagnostics, PFMB 100-100-12). Lysates were centrifuged for 10 min 13,000 rpm at 4°C and the supernatants were collected. Pierce BCA Protein Assay Kit (ThermoFisher) was used to determine total protein concentration in the lysate.

For CRISPRi and RNase E repression lysate cleavage assays, a final concentration of 200 nM of 29 nt FAM-labeled RNA purchased from Integrated DNA Technologies (Supplemental Figure 8D–E, Supplemental Table 3) and 400 ng of total protein from the lysate was added to cleavage assay reactions with 20 mM tris-HCl pH 7.9, 100 mM NaCl, 5% glycerol, 0.01% IGEPAL, 0.1 mM DTT, and 10 mM MgCl₂. Reactions were incubated at 37°C in a thermocycler for 1 hour (PNPase and RNase J) or 2 hours (RNase E), stopped using a final concentration of 12.5 mM EDTA pH 8.0 in formamide, and heated at 70°C for 10 minutes. Samples were run on a 20% urea-PAGE gel then imaged on an Azure 600 imager using the Cy3 channel (524 nm excitation). Two (PNPase and RNase J) or three (RNase E) biological replicates of each strain were grown and lysed. Substrate and product band intensities were calculated using ImageJ with the FIJI plugin and used to calculate the percent of substrate remaining. Ordinary one-way ANOVA with Sidak’s multiple comparisons test was used for significance testing.

For hypoxia lysate cleavage assays, a 22 nt FAM-labeled RNA was used (Supplemental Figure 8G, Supplemental Table 3) at a final concentration of 50 nM with cells grown in hypoxia bottles as described in [3] for 3 days prior to harvesting. The CRISPRi protocol described above was modified using 3.8 μg total protein, then the reactions were stopped after the indicated amount of time. Duplicate strains were grown and lysed. Substrate and product band intensities were calculated using ImageJ with the FIJI plugin and used to calculate the percent of substrate remaining. Simple linear regression was used for significance testing.

For carbon starvation lysate cleavage assays, the 22 nt FAM-labeled RNA was used (Supplemental Figure 8I, Supplemental Table 3) at a final concentration of 50 nM with cells that were grown in carbon starvation media for 24 hours prior to harvesting using the protocol described in [3]. The hypoxia protocol was modified using 4.8 μg total protein.

Western Blotting of Cell Lysates:

15 mL M. smegmatis cultures were grown to exponential phase and pelleted by centrifugation at 10 minutes at 3,900 rpm and 4°C. Cells were washed with an equal volume of 7H9 containing no BSA, dextrose, or catalase three times then resuspended in 300 μL of 20 mM tris-HCl pH 7.9, 100 mM NaCl, 5% glycerol, 0.1 mM DTT, 0.01% IGEPAL, and 1X Halt Protease Inhibitor Cocktail, EDTA-Free (ThermoFisher). Cells were lysed with a FastPrep-24 bead-beater (MP Biomedicals) using 4 cycles of 30 sec at 6.5 m/s, 1 minute on ice between cycles in tubes containing 100 μm zirconium beads (OPS Diagnostics, PFMB 100-100-12). Lysates were centrifuged for 10 min 13,000 rpm at 4°C and the supernatants were collected. Pierce BCA Protein Assay Kit (ThermoFisher) was used to determine total protein concentration in the lysate, and 3 μg of total protein was loaded onto an SDS-PAGE gel then transferred to a PVDF membrane. Membranes were stained with LiCor Revert 700 Total Protein Stain (LiCor, 926–11021) following the manufacturer’s instructions and imaged to verify equal loading. The membrane was blocked with TBS (20 mM tris-HCl pH 7.6, 137 mM NaCl) with 3% nonfat dry milk (w/v) for 30 min at room temperature, washed with TBS three times, and incubated with monoclonal anti-FLAG M2 mouse antibody (1 μg/mL) (Sigma-Aldrich, F1804) in TBS 3% nonfat dry milk (w/v) overnight at 4°C. Membranes were washed with TBS three times then incubated with goat anti-mouse HRP antibody for 45 minutes at room temperature. The membrane was washed 6 times with TBS with 0.05% Tween-20 (v/v) then once with TBS (no Tween). Membranes were developed with Azure Radiance ECL (AC2204) in an Azure 600 imager.

Overexpression and Purification of Recombinant RNase E Variants:

RNase E variants were recombinantly expressed and purified using E. coli BL21(DE3) pLysS transformed with pET42-derived plasmids (Supplemental Table 4). 2 L cultures were grown at 37°C and were shaken at 200 rpm to an OD₆₀₀ of 0.5–0.8. Cultures were induced with 1 mM IPTG at 28°C and were shaken at 200 rpm for 4 hours prior to harvest by centrifugation. Cell pellets were resuspended in 10 mL of buffer (20 mM tris-HCl, 150 mM NaCl, 5% glycerol, 0.01% IGEPAL, 10 mM imidazole) with 1x Halt Protease Inhibitor Cocktail, EDTA-Free (ThermoFisher), 40 mg of lysozyme, and 16 U Turbo DNase (Invitrogen). Cells were lysed with a BioSpec Tissue-Tearor (6 cycles of speed 6 then 4 cycles of speed 9 that were 30 s each with 30 s on ice between cycles). Lysates were cleared by centrifugation with 8 mL His-Pur nickel–nitrilotriacetic acid resin 50% slurry (ThermoScientific) and the NaCl concentration in the lysate was increased to 1 M before incubation for 60 minutes on at room temperature with end-to-end rotation. The resin was washed three times with 10 mL of 20 mM tris-HCl, 1 M NaCl, 5% glycerol, 0.01% IGEPAL, and 20 mM imidazole then eluted with 4 mL of 20 mM tris-HCl, 150 mM NaCl, 5% glycerol, 0.01% IGEPAL, 150 mM imidazole with 1x Halt Protease Inhibitor Cocktail, EDTA-Free (ThermoFisher) three times, on end-to-end rocker for 10 minutes each. Eluates were concentrated with Microcon PL-100 (full-length) or PL-30 (other variants) (100,000 or 30,000 NMWL) protein concentrators (MilliporeSigma) to a volume of about 300 μL. Samples were loaded onto 1 cm diameter, 38 ml Sephacryl S-400 (full-length, N-deletion, and C-deletion) or S-200 (N- and C-deletion) High Resolution resin (GE Healthcare) size exclusion chromatography columns run with a BioLogic LP chromatography system (BioRad). A flow rate of 0.25 mL/minute was used, and 500 μL fractions were collected using SEC buffer (20 mM tris-HCl, 150 mM NaCl, 5% glycerol, 0.01% IGEPAL, 1 mM DTT, 1 mM EDTA). Fractions were combined, concentrated, and buffer exchanged using the same Microcon protein concentrators with 20 mM tris-HCL, 100 mM NaCl, 5% glycerol, 0.01% IGEPAL, and 0.1 mM DTT.

Overexpression and Purification of Recombinant Other Degradodosome proteins:

PNPase, RhlE1, and RNase J variants were expressed and purified as described above for RNase E, scaled down to 1 L growth (Supplemental Table 4 with corresponding M. smegmatis gene IDS in Supplemental Table 5). Microcon PL-30 (30,000 NMWL) protein concentrators (MilliporeSigma) were used for all proteins, and Sephacryl S-400 (PNPase) or S-200 (RNase J and RhlE1) High Resolution resin (GE Healthcare) were used for size exclusion chromatography.

Cleavage Assays:

Hairpin RNA oligo purchased from Integrated DNA Technologies (Supplementary Figure 6C, Supplementary Table 3) at a final concentration of 316 nM was incubated with 100 nM of RNase E and 100 nM of PNPase in 20 mM tris-HCl pH 7.9, 100 mM NaCl, 10 mM MgCl₂, 10 μM ZnCl₂, 0.5% glycerol, 0.01% IGEPAL, and 1 mM DTT for 1 hour at 37°C in a 10 μL reaction. Reactions were stopped by adding 1X of RNA loading dye (ThermoScientific) then heated for 10 minutes at 70°C. Samples were run on a 15% TBE-urea PAGE gel and stained with 40 mL of 1X TBE with 1X Sybr^™ Gold (Invitrogen) on a rocker for 10 minutes then imaged with an Azure 600 imager at 302 nm excitation. ImageJ with the FIJI plugin was used to quantify band intensities, which were normalized to a mock reaction with no enzymes.

This protocol was modified using a final concentration of 949 nM of the same RNA hairpin (Supplementary Figure 6D, Supplementary Table 3) and 138 nM RhlE1 (N-terminal tags) and 95 nM PNPase in 20 mM tris-HCl pH 6.5, 100 mM NaCl, 5 mM sodium phosphate monobasic, 100 mM MgCl₂, 0.5% glycerol, 0.01% IGEPAL, 1 mM DTT, and 1 mM ATP in a final volume of 10 μL. Samples were incubated at 37°C for 75 minutes then stopped and imaged as described above.

A 29 nt FAM-labeled RNA purchased from Integrated DNA Technologies (Supplementary Figure 8A, Supplementary Table 3) was used in a cleavage assays with RNase E, RNase J, and PNPase. RNase E and RNase J used the same buffer as Supplementary Figure 6C, and PNPase used the same buffer as Supplementary 6D described above. Reactions contained 80 ng of enzymes (final concentrations 101 nM RNase E, 95 nM PNPase, 127 nM RNase J, and 138 nM RhlE1 (N-terminal tags)) and 200 nM RNA were incubated at 37°C degrees for 1 hour in a 10 μL reaction. Reactions were stopped by adding a 10 μL with a final concentration of 12.5 mM EDTA pH 8.0 in formamide. Reactions were then heated for 10 minutes at 70°C. Samples were run on a 20% urea-PAGE gel then imaged on an Azure 600 imager using the Cy3 channel (524 nm excitation).

RNase E Kinetics:

The same 29 nt FAM labeled RNA containing exactly one RNase E cleavage site as in Supplemental Figure 8A (at final concentrations of 150, 300, 445, 590, 715, 1410, 3020 nM; Figure 3A–B, Supplementary Table 3) was incubated with 101 nM of each of the RNase E variants at 37°C in a thermocycler. The same buffer described above for Supplementary Figure 6C was used. Reactions were stopped every 30 seconds for 4 minutes (except for reactions with 3020 nM RNA which were stopped every 2 minutes for 16 minutes) by addition 10 μL of EDTA pH 8.0 to a final concentration of 12.5 mM in formamide then heated at 70°C for 10 minutes. Samples were run on a 20% urea-PAGE gel then imaged on an Azure 600 imager using the Cy3 channel (524 nm excitation).

Figure 3. The IDRs of RNase E are dispensable for catalytic activity but have partially redundant roles in RNA binding.

A) Kinetic analysis of IDR deletion RNase E mutants in vitro using a 29 nt FAM labeled substrate, quantified in B). Initial rates were calculated by simple linear regression of the disappearance of substrate, quantified on ImageJ with the FIJI plugin. Points represent the average of three replicates. Nonlinear regression with a K_cat model was used for determining kinetic constants, and each mutant was compared to full-length using the extra sum of squares F-test. C) EMSA binding assays with a 721 nt fluorocein-UTP labeled RNA and catalytically dead RNase E, fit with the one-site specific binding equation quantified in D). The fraction bound was quantified on ImageJ with the FIJI plugin, and each point represents the average of three replicates. Comparisons were made using the extra sum of squares F-test.

Band intensities were quantified using ImageJ with the FIJI plugin and normalized to a mock reaction containing no enzyme run on the same gel. Initial rates were calculated by using the slope of the line of the amount of substrate remaining versus time and were averaged for three replicates. Initial rates were plotted versus substrate concentration and were fit using the nonlinear regression K_cat model in GraphPad Prism. Because the substrate concentrations used were not all dramatically higher than the enzyme concentration, we report the results as apparent kinetic parameters.

Electromobility Shift Assays:

HiScribe^® T7 High Yield RNA Synthesis Kit (New England Biolabs) kit was used with a 20 μL reaction containing 7.5 μL of NTP + Buffer mix, 1.5 μL of T7 RNA Polymerase, 250 ng of the mCherry DNA template containing a T7 promoter, and 3 μL of fluorescein-12-UTP (Sigma) and incubated overnight at 37°C. 2 μL of DNase I (New England Biolabs) and 30 μL of water were added to the reactions and incubated at 37°C for 15 minutes. RNA Clean and Concentrator-5 (Zymo) was used according to the manufacturer’s instructions with an elution volume of 11 μL.

A final concentration of 200 nM RNA (Figure 3C–D, Supplemental Table 3) was mixed with the indicated concentration of catalytically dead RNase E in 20 mM tris-HCl pH 7.9, 100 mM NaCl, 10 mM MgCl₂, 10 μM ZnCl₂, 0.5% glycerol, 0.01% IGEPAL, and 1 mM DTT at 37°C for 15 minutes in a 10 μL reaction. Reactions were stopped with final concentrations of 0.45% SDS, 9.1 mM EDTA, 22.5% glycerol. Reactions were run on a 1% agarose gel at 4°C and imaged on an Azure 600 imager Cy3 channel (524 nm excitation). Band intensities were quantified using ImageJ with the FIJI plugin to calculate the fraction bound, which was then plotted versus the RNase E concentration and fit with a one site - specific binding nonlinear model in GraphPad Prism.

Coimmunoprecipitations with RNase E:

43 nM catalytically dead RNase E was mixed with 86 nM catalytically active degradosome enzymes (PNPase, RhlE1 (C-terminal tag), or RNase J) in a final volume of 1 mL with 20 mM tris-HCl pH 7.9, 100 mM NaCl, 5% glycerol, 0.1 mM DTT, 0.01% IGEPAL, 10 mM MgCl₂, and 10 μM ZnCl₂ and preincubated with end-over-end rotation at 37°C for 1 hour. 20 μL of anti-FLAG magnetic beads (Sigma) were added and incubated at 37°C for an additional 1 hour. Supernatant was removed and beads were washed 3 times with TBS then eluted with 75 μL of 0.1 M glycine-HCl pH 2.8 and immediately run on SDS-PAGE, loading an equal fraction of mock reaction and elution. Western blots were run as described above, using the same protocol for anti-FLAG antibody and 0.33 μg/mL anti-HA antibody (Biolegend, 901533). ImageJ with the FIJI plugin was used to quantify band intensities.

Coimmunoprecipitation without RNase E:

100 ng/μL catalytically active degradosome enzymes (PNPase, RhlE1 (N-terminal tags), or RNase J) in a final volume of 70 μL were combined in 20 mM tris-HCl pH 7.9, 100 mM NaCl, 5% glycerol, 0.1 mM DTT, 0.01% IGEPAL, 10 mM MgCl₂, and 10 μM ZnCl₂ and preincubated with end-over-end rotation at 37°C for 1 hour. 12 μL of anti-HA magnetic beads (Sigma) were added and incubated at 37°C for an additional 1 hour. The supernatant was removed and the beads were washed 3 times with TBS then eluted with 30 μL of 0.1 M glycine-HCl pH 2.8 and immediately run on SDS-PAGE, loading an equal fraction of mock reaction and elution. SDS-PAGE gels were stained with Bio-Safe Coomassie G-250 Stain (BioRad) and imaged in an Azure 600 imager. Three-way coimmunopreciptations were done in the same way with 2 μg/μL each protein in a final volume of 70 μL.

In vitro Condensate Formation Microscopy:

10 μM RNase E was incubated with 20 mM Tris pH 7.4, 100 mM NaCl, 1 mM DTT buffer in a total reaction volume of 10 μL for 30 min at room temperature. When indicated, incubations included 40 ng/μL E. coli total RNA or 1 mM magnesium chloride. The entire 10 μL was spotted on a slide and covered with a coverslip before imaging with a Thermo Scientific EVOS microscope with 100X magnification..

The number and diameter of condensates were quantified using the Analyze Particles function in ImageJ. Briefly, three images per protein construct were processed by applying a particle size filter of 0.2 μm to infinity and a circularity range of 0.01 – 1.0. The resulting condensate counts and diameters were then plotted using GraphPad Prism and ordinary one-way ANOVA with Sidak’s multiple comparisons test was used. For constructs that did not show detectable condensates, both condensate number and diameter were assigned a value of zero for the purpose of data visualization.

Phosphate Assay:

EnzChek Phosphate Assay Kit (Invitrogen) was scaled down to a final volume of 200 μL in a 96 well plate and read in a Victor Nivo^® (Perkin Elmer) multimode plate reader using VICTOR Nivo Control Software^® 4.5.0. Reactions containing 50 mM tris-HCl (pH 7.5), 1 mM MgCl₂, 0.1 mM sodium azide, 200 mM MESG, 0.2 U purine nucleoside phosphorylase, 0.1 μM enzymes (catalytically active RhlE1 (C-terminal tags), catalytically dead PNPase, and/or catalytically dead RNase E), 1 μM hairpin RNA (Supplemental Figure 6A–B, Supplemental Table 3) were preincubated for 10 minutes at room temperature then ATP was added to a final concentration of 2 mM and immediately read on the plate reader every 20 seconds for 15 minutes. Absorbance values were normalized to time 0 then converted to phosphate concentration using a standard curve made with KH₂PO₄.

SEC-HPLC:

A Nexera Lite Inert High Performance Liquid Chromatography System (Shimadzu) was used with a TOSOH Bioscience G3000SWXL column (7.8 mm diameter, 30 cm length). A program with an isocratic flow of 1 mL/minute with a buffer containing 20 mM Tris-HCl pH 7.9, 100 mM NaCl, 5% glycerol, 0.1 mM DTT, 10 mM MgCl₂, 10 μM ZnCl₂, and 0.01% IGEPAL was utilized. Molecular weight size standard (Sigma 69385) was run, and 20 μL of 2.6 μM of each catalytically dead RNase E variant was injected. UV detection was monitored at 280 nm, and LabSolutions Version 6.117 was used for data collection and analysis.

RESULTS

Mycobacterial RNase E has two intrinsically disordered regions

In the organisms where RNase E has been best studied, such as E. coli [15, 16] and C. crescentus [17], this enzyme contains an intrinsically disordered region (IDR) that mediates noncatalytic functions including interaction with other proteins and subcellular localization. We sought to know if RNase E in mycobacteria also contains an IDR. We used IUPred2 to predict regions of disorder in M. smegmatis RNase E (Figure 1A) and found two IDRs: one on the N-terminus and one on the C-terminus, flanking the catalytic domain. AlphaFold3 showed a similar prediction (Figure 1B). Multiple sequence alignments in ClustalOmega suggest that the presence of two IDRs flanking the catalytic domain is conserved in mycobacteria and more broadly in the Actinobacteria, in contrast to the Proteobacteria (Figure 1A) [15]. These IDRs are quite large, at 330 and 214 amino acids long in M. smegmatis and similar sizes in M. tuberculosis. While the IDR sequences are not well conserved across mycobacteria, they consistently include acidic and arginine rich regions. To our knowledge, the Actinobacteria is the only class whose RNase E contains both N- and C-terminal IDRs, so we sought to learn the physiological importance of these regions. To determine the essentiality of the IDRs, we generated an M. smegmatis strain expressing full-length RNase E from a plasmid integrated at the L5 phage integration site and deleted rne (MSMEG_4626) from its native locus. We then exchanged the L5 plasmid for variants encoding truncated RNase E and marked with a different antibiotic resistance gene. All variants had N-terminal FLAG-tags. An N-terminal IDR deletion (retaining AA 330–1037), a C-terminal IDR deletion (retaining AA 2–823), and a N- and C-terminal IDR double deletion (retaining AA 330–823) all supported bacterial survival. In contrast, we were unable to obtain a strain expressing RNase E with a deletion of the first 375 amino acids as its sole copy, suggesting that the essential catalytic domain [7] was compromised (Figure 1C).

Figure 1. The structure of RNase E contains two intrinsically disordered regions.

A) AphaFold3 structure of RNase E from M. smegmatis. Blue indicates areas of high confidence and red/yellow indicate areas of low confidence. B) IUPred2 disorder prediction of RNase E. Multiple sequence alignments in ClustalOmega show location and relative size and position of RNase E IDRs in a few species of bacteria. C) Schematic of IDR deletion mutants of RNase E constructed and used in this study.

The N-terminal IDR of RNase E impacts growth and mRNA degradation rates

To begin to understand the physiological roles of the RNase E IDRs, we investigated the phenotypic consequences of deleting them. First, we performed western blotting to assess the relative amounts of RNase E in the IDR deletion strains and confirmed that each truncated version had abundance similar to, or higher than, the full-length (Supplementary Figure 1A–C). Therefore, any phenotypes we observe with the mutant strains are not due to protein instability upon IDR deletion. The higher RNase E protein levels in the IDR-deletion strains could be due to loss of autoregulation of mRNA stability by cleavage of its 5’ untranslated region (UTR), as has been shown in E. coli and C. crescentus [17, 53, 54]; the M. smegmatis rne 5’ UTR has an RNA cleavage site [24], consistent with a UTR-cleavage-based autoregulatory system. When assessing cell length, we found that deleting the N-terminal IDR resulted in significantly shorter cells compared to strains with full-length RNase E and the C-terminal IDR deletion (Figure 2A).

Additionally, the cell size distributions of the N-terminal IDR deletion and N- and C-terminal IDR double deletion strains were broader than those of the strains with full-length or C-terminal deleted versions. This suggests that deleting the N-terminal IDR of RNase E causes a cell growth or division defect. Additionally, the strains with RNase E N-terminal IDR deletions had growth defects observed by optical density (Figure 2B) and colony forming units (CFUs) (Supplementary Figure 1D).

To determine if the IDR deletions affected mRNA degradation, we measured the half-lives of the transcripts of four arbitrarily selected genes (atpB, rnj, rraA, and sigA). Deleting the N-terminal IDR caused an increase in RNA half-life for all four transcripts, indicating that mRNA was degraded less efficiently. This effect was exacerbated when the C-terminal IDR was additionally deleted (Figure 2C), but deletion of the C-terminal IDR alone did not have a detectable impact. This suggests that the two IDRs have partially redundant functions that are required for normal mRNA degradation rates, while the N-terminal IDR also has a unique function.

To globally assess the impact of the RNase E IDRs on the steady-state transcriptome, we performed RNAseq to identify genes with differentially abundant transcripts. It is important to note that these changes in steady-state transcript abundance likely reflect the combination of two types of effects: (1) differences in degradation rate (direct impacts of perturbed RNase E function) and (2) differences in transcription rate resulting from cellular adaptation to the perturbed RNase E function. Relatively small numbers of genes were differentially expressed in the N-terminal and C-terminal individual IDR deletion strains (68 and 20 respectively), but the double deletion strain showed a large set of differentially expressed genes (311 genes; Figure 2D–E; Supplementary Table 6). This implies that there is functional redundancy between the two IDRs. The proportions of up- and down-regulated genes were similar for all three strains, suggesting that many of the expression changes were due to regulatory responses rather than passive effects of slower mRNA degradation. There were no annotated GO or KEGG pathways enriched in the differentially expressed genes for any of the three strains. However, some pathways were identifiable among differentially expressed genes. Several mycofactocin biosynthesis genes were differentially expressed in the double-IDR deletion strain and similar patterns were present in the N-terminal IDR single deletion strain. This pathway may be involved in redox balance [55] Eleven genes in the ESX-1 type VII secretion system locus were upregulated in the double-IDR deletion strain, of which four were upregulated in the N-terminal IDR single deletion strain. The ESX-1 secretion system is involved in virulence in M. tuberculosis and conjugation in M. smegmatis [56–58]. There were some expression differences in genes associated with DNA replication; MSMEG_6892, encoding the replicative DNA helicase, was upregulated in both the N-terminal IDR single deletion strain and the double IDR deletion strain, and MSMEG_2416, encoding a septation-associated protein [56, 58], was upregulated in the double IDR deletion strain. Additionally, four ribonucleoside reductase genes were upregulated in the double-IDR deletion strain (of which one was also upregulated in the N-terminal IDR single deletion strain) (Supplemental Table 7). The only possible RNase differentially expressed was an annotated endoribonuclease L-psp family protein, which was downregulated in the double-IDR deletion strain. A gene annotated as encoding a ribonuclease inhibitor, MSMEG_0209, was also downregulated. The functions of these proteins have not been studied, and the annotations should therefore be treated with caution. None of the four genes shown to have altered degradation kinetics in Figure 2C were differentially expressed, indicating that their slower degradation in the IDR deletion strains was likely compensated by correspondingly slower transcription, as we have previously reported [10].

The RNase E IDRs contribute to RNA binding but are dispensable for multimerization and catalysis in vitro

Since we uncovered several phenotypes associated with deleting the IDRs of RNase E, we wanted to understand the biochemical mechanism responsible. We overexpressed N-terminal 6x-His and FLAG tagged M. smegmatis RNase E and the IDR deletion mutants recombinantly and purified it by immobilized metal affinity followed by size exclusion chromatography. We purified catalytically inactive versions of each RNase E variant (D694R and D738R [10, 59] in the full-length protein, and the equivalent residues of each variant) and tested their cleavage activities in parallel to optimize purification procedures to prevent co-purification of E. coli RNases.

We tested the impact of the RNase E IDRs on catalysis directly by measuring the kinetics of the cleavage of a short 5’ FAM-labeled 29 nt RNA. This substrate was derived from the atpB gene and was previously shown to have exactly one RNase E cleavage site [10]. The apparent V_max and apparent K_m for each IDR deletion protein were compared to full-length, and none were statistically significantly different (Figure 3A–B). This indicates that the IDRs do not impact catalysis, which is consistent with the apparent distance of the IDRs from the active site as well as reports that the IDRs of RNase E in other organisms were shown to be involved in noncatalytic activities [15–17, 39, 42].

Since there are regions within the IDRs that consist of several consecutive arginine residues (Figure 1C), we hypothesized that these positively charged regions might contribute to RNA binding. To test this, we performed electromobility shift assays (EMSAs) with a 721 nt RNA. We reasoned that an RNA of this length would allow for simultaneous binding of multiple sites on RNase E (e.g. the catalytic domain and both IDRs). Full-length RNase E bound RNA better than both the N-terminal and the C-terminal IDR deletions (Figure 3C–D and Supplementary Figure 2A–C), and the double IDR deletion bound RNA so weakly that it could not be detected by EMSA (Supplementary Figure 2D). The N-terminal and C-terminal IDR deletions appeared to bind RNA with similar affinity (Figure 3C–D), implying that there is functional redundancy between the two IDRs. This functional redundance is consistent with the RNAseq results (Figure 2D).

In both E. coli [12, 13] and M. tuberculosis [6], RNase E exists as monomers, dimers, and tetramers. To determine if the IDRs impact the multimerization state of M. smegmatis RNase E, we performed analytical size-exclusion chromatography. We found monomers, dimers, and tetramers in all IDR deletions, suggesting that the IDRs are not necessary for multimerization (Supplementary Figure 3A–D). This is unsurprising, as it has been reported in E. coli that zinc binding is necessary for multimerization [12–14], and the zinc binding residues are within the catalytic domain of M. smegmatis. Interestingly, we observed that the full-length, N-terminal IDR deletion, and double IDR deletion were predominantly in the tetramer state, whereas the C-terminal IDR deletion was primarily in the monomer and dimer states. This suggests that there may be some regulation of multimerization state by the C-terminal IDR, but this does not seem to be important in the growth conditions we used, as deletion of the C-terminal IDR had little impact on the cellular phenotypes we investigated (Figure 2A–C).

The RNase E IDRs play a minor role in condensate formation

Broadly, bacteria are thought to form biomolecular condensates to quickly bring enzymes and substrates together, allow reactions to happen, then disassemble to allow for rapid responses to environmental changes (reviewed in [18, 19]). In C. crescentus, an established function of the RNase E IDR is to form biomolecular condensates [17, 20–22]. To determine if RNase E from M. smegmatis forms condensates, we first fused RNase E to mCherry to observe its subcellular localization. The RNase E-mCherry fusion was confirmed to be functional because cells were able to grow with this as the only copy of RNase E, an essential protein [7], when the native copy had been deleted. We observed RNase E-mCherry foci and showed that these are dynamic bodies by treating the cells with rifampicin. We saw a decrease in the number of foci per micron (Figure 4A–B) following rifampicin treatment, consistent with our previous report with an RNase E-mScarlet fusion [60]. Since rifampicin blocks transcription, this result suggests that as RNA levels globally decrease following rifampicin treatment, RNase E condensates are disassembled.

To determine if the IDRs of RNase E are required for foci formation, the sole copy of RNase E with each of the different domain deletions was fused to mCherry (Figure 4C–D). When we looked at the metric of foci per micron, we did not observe any differences between the full-length and deletion strains (Figure 4E). Another metric for assessing the heterogeneity of protein distribution is to measure the coefficient of variation (CV) in fluorescence intensity along a line drawn through the long-axis of the cell. Punctate signal arising from dense and dilute phases will produce a higher CV than homogenous signal. Upon measuring CV, we saw some variations among the strains with the N-terminal and the C-terminal single IDR deletions causing similar decreases in CV and the double IDR deletion causing a greater decrease (Figure 4F).

However, all RNase E fusions had higher CVs than mCherry alone. The subtle changes caused by deletion of the IDRs may be due to differences in RNA binding (Figure 3C–D) since RNA is presumably a component of these condensates.

To determine if the IDRs alone were sufficient for condensate formation, we fused the IDRs to mScarlet individually and in combination without the catalytic domain of RNase E (Figure 4G–H). No differences in foci per micron were observed between mScarlet alone and the mScarlet-IDR fusions (Figure 4I). However, fusion of both RNase E IDRs to mScarlet led to a significant increase in CV, suggesting that in combination they may be capable inducing condensate formation (Figure 4J). While this effect was not seen for the individual IDR fusions, it is possible that this was due to lower expression levels. Together, these data suggest that while the IDRs of RNase E may contribute to condensate formation in the context of the full-length protein, they are not the major determinants of RNase E localization within M. smegmatis cells, nor are theyalone sufficient to mediate condensate formation to the extent seen for the full-length protein. This contrasts with previous reports in C. crescentus ([17, 20–22]; see discussion).

We also looked in vitro at the ability of purified RNase E to form condensates big enough to be seen by phase-contrast microscopy. In contrast to the in vivo foci which may occur due to interactions with its interaction partners, in vitro condensation can determine directly if RNase E has the intrinsic capacity to drive phase separation. We observed that while the full-length, N-terminal IDR deletion, and C-terminal IDR deletion variants of RNase E formed condensates, the double IDR deletion did not under any conditions (Supplementary Figure 4A). Surprisingly, the single IDR deletion variants had a greater propensity for condensate formation than the full-length protein; both the N- and C-terminal deletion variants formed condensates with and without magnesium and RNA, while the full-length version only formed visible condensates in the presence of magnesium. The N-terminal IDR deletion protein formed notably more condensates than any of the other variants (Supplementary Figure 4B), including the full-length protein, and the C-terminal deletion formed irregularly shaped condensates, which may suggest that it reaches the equilibrium state more slowly [61]. We found that the addition of magnesium, a metal bound by RNase E for catalysis, had variable effects on condensate size, which could possibly be explained by copurification of magnesium with RNase E (Supplementary Figure 4C). In contrast to previous reports in C. crescentus RNase E [17, 20–22], we did not find that RNA stimulated the formation of condensates (Supplementary Figure 4A) by the full-length protein, but RNA did increase the size of the condensates formed by the single IDR deletion mutants (Supplementary Figure 4C). Collectively, these data suggest that the IDRs of mycobacterial RNase E can both stimulate and inhibit condensate formation, depending on local conditions. The discrepancies between the in vitro and in vivo data may be explained by the other protein components that interact with RNase E and could play roles in condensate formation; for example, RhlE1 has a long IDR similar to those of helicases shown to phase separate in other organisms [62]. The apparent condensates we observe in vivo may be more similar to P-bodies found in yeast, where removing one protein involved is not sufficient to dissolve condensates [21, 63]. The in vitro assay is likely an oversimplification of what is happening in cells, given that we assayed RNase E in the absence of other proteins.

Defining the protein-protein interactions of the M. smegmatis degradosome-like network in vitro

RNase E is known to bind specific proteins involved in mRNA degradation using its IDR to form a multiprotein complex called the RNA degradosome in E. coli [16, 30–34, 37] and C. crescentus [39]. A similar situation occurs in Bacillus subtilis where RNase Y binds RNA degradation and metabolic enzymes [41] via its IDR [42]. We sought to know if the RNase E IDRs in M. smegmatis also mediate protein-protein interactions. Previous work in the literature predicted that PNPase, RNase J, and the helicase RhlE may interact with RNase E to form the RNA degradosome in mycobacteria, based on RNase E pull-downs from cell lysates followed by mass spectrometry [11]. However, these experiments could not distinguish direct from indirect or RNA-mediated interactions and had yet to be verified biochemically. M. smegmatis has two homologs of RhlE (RhlE1 and RhlE2) but M. tuberculosis only has one. We chose to test RhlE1 as its sequence aligns more closely to that of the M. tuberculosis protein.

To probe the direct protein-protein interactions of the degradosome-like network directly, we performed coimmunoprecipitation experiments with purified proteins. RNase E was pulled down with its FLAG tag, and the protein partner was western blotted for its hemagglutinin (HA) tag. RNase E was also western blotted for FLAG in each experiment as a control. Western blotting was necessary to obtain clear data in these experiments because RNase E was somewhat unstable during the immunoprecipitation process, resulting in the presence of degradation products of similar sizes as the other protein partners (Supplementary Figure 5A). We identified direct interactions between RNase E and PNPase (Figure 5A–B), between RNase E and RNase J (Figure 5C–D), and between RNase E and RhlE1 (Figure 5E–F). To determine if these interactions required one or both IDRs of RNase E, we assessed the impact of IDR deletions. Deletion of the C-terminal IDR of RNase E largely abolished its ability to pull down PNPase, while deletion of the N-terminal IDR did not have any detectable impact by this method (Figure 5A–B). This suggests that PNPase interacts with RNase E via its C-terminal IDR. A similar result was seen for the interaction with RNase E and RNase J, which was abolished by deletion of the C-terminal IDR of RNase E (Figure 5C–D). In the case of RhlE1, deletion of either the N- or the C-terminal IDR of RNase E abolished the coimmunoprecipitation (Figure 5E–F). This suggests that both IDRs participate in the interaction between RNase E and RhlE1.

Figure 5. Defining the protein-protein interactions of the degradosome-like network in vitro.

Representative images of pull downs of Flag-tagged RNase E and coimmunoprecipitation of HA-tagged A) PNPase, C) RNase J, and E) PNPase, detected by western blot. In all cases “IP” indicates immunoprecipitation, and samples incubated in buffer without beads were run in parallel for comparison. Pull downs were quantified in B), D), and F), respectively, using ImageJ with the FIJI plugin. Pull down by IDR deletions were expressed as percentages of the full-length pull down. Dots represent replicates, and bars represent averages. A one-way ANOVA was performed with Sidak’s multiple comparison test. Representative images of Coomassie SDS-PAGE gels of HA coimmunoprecipitations of G) HA-RNase J and untagged PNPase, H) HA-RhlE1and untagged PNPase, and I) HA-tagged RhlE1 and untagged RNase J. J) Summary of the verified 2-way protein-protein interactions and a model of the possible degradosome-like network in mycobacteria. ns p > 0.05, * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001

A direct interaction between M. tuberculosis PNPase and RNase J was previously reported [11]. We wanted to confirm that this interaction also occurred in M. smegmatis and to probe additional possible interactions among PNPase, RNase J, and RhlE1. We performed reciprocal coimmunoprecipitations, pulling down HA-tagged versions of each protein and assessing if the other untagged protein coimmunoprecipitated by SDS-PAGE and Coomassie staining. We identified direct interactions between PNPase and RNase J (Figure 5G and Supplementary Figure 5B) as well as between PNPase and RhlE1 (Figure 5H and Supplementary Figure 5C). In both cases, these interactions were observed when either of the partner proteins was pulled down. However, we did not detect an interaction between RNase J and RhlE1 when either of the two proteins was pulled down (Figure 5I and Supplementary Figure 5D). These results suggest that PNPase may play a hub-like role, with at least three possible direct interaction partners.

We propose a model in which these pairs of interactions could occur individually, as we have demonstrated by coimmunoprecipitation in vitro (Figure 5J). It is possible that multiple interactions could occur simultaneously as was shown in E. coli [30, 34], but we have no direct evidence for that idea in mycobacteria. To probe multiple simultaneous interactions, we performed a three-component coimmunoprecipitation experiment pulling down HA-tagged RhlE1 in the presence of both PNPase (its direct interaction partner) and RNase J (which could only be pulled down via its interaction with PNPase). While RhlE1 pulled down PNPase as expected, RNase J was not pulled down, indicating that a stable ternary complex was not formed (Supplementary Figure 5E). This suggests that either the interactions of PNPase with RhlE1 and RNase J are mutually exclusive, or the simultaneous interactions are too weak to be identified by pulldown. Our methods were not suitable to test the ability of RNase E to simultaneously bind to multiple protein partners in mycobacteria. In E. coli, the degradosome was defined using techniques including copurification, size exclusion chromatography, yeast-two hybrid assays, and coimmunoprecipitation from lysates without crosslinking [15, 16, 30, 32, 33] leading to the conclusion that a stable, stoichiometrically defined degradosome exists in cells. We had to employ much more sensitive techniques to probe the degradosome-like network in M. smegmatis, suggesting that a less rigid system exists (see discussion).

Having defined direct protein-protein interactions between the RNA degradation and processing proteins in mycobacteria, we wanted to know if this binding causes allosteric changes that enhance activity. There is a precedent for this idea in E. coli where the interaction of the RNA helicase RhlB with RNase E causes the helicase to unwind RNA at a faster rate [16, 28, 29]. To assess if RNase E has the same impact on RhlE1 in mycobacteria, we utilized a phosphate assay with a structured hairpin RNA. RhlE1 is an ATP-dependent RNA helicase that produces inorganic phosphate when it unwinds RNA, and we saw an increase in phosphate with catalytically active RhlE1 but not a catalytically dead mutant (K55A mutation [64]). With the addition of catalytically dead RNase E (D694R and D738R mutations [10, 59]) and RNA at levels expected to saturate RNase E, we observed no changes in the rate of ATP hydrolysis by RhlE1 (Supplementary Figure 6A) suggesting that RhlE1 binding to RNase E does not cause allosteric changes that enhance activity. Because PNPase also interacts with RhlE1, we utilized the same assay to test if RhlE1 binding to catalytically dead PNPase (R431A mutation [8]) enhances its ATPase activity (Supplementary Figure 6B), but we observed no changes. We used RNA cleavage assays to assess the cleavage activities of protein combinations using the same structured hairpin RNA. We observed no differences in the cleavage activity of RNase E in the presence of catalytically dead PNPase (Supplementary Figure 6C) or the cleavage activity of PNPase in the presence of RhlE1 (Supplementary Figure 6D). In the classical model of E. coli mRNA degradation, cleavage by RNase E is the initiation step, while the helicase can unwind secondary structure so that the exonucleases can finish the job by processive cleavage [30, 65]. Our data collectively suggest that the RNA degradation proteins in mycobacteria are unlikely to be allosterically enhancing the activity of the others upon binding. It is more likely that the close proximity of the enzymes within the degradosome-like network, caused by both condensate formation and protein-protein interactions, allows for the products of one enzyme to be readily passed to the next enzyme to degrade mRNA more efficiently.

Localization of degradosome-like network proteins in vivo

To determine if the degradosome-like network protein interactions identified in vitro are reflected in the subcellular localization patterns of these proteins in vivo, we performed microscopy and assessed the localization of mCherry-tagged RNase E and partner proteins tagged with green fluorescent protein (GFP). PNPase was tagged with eGFP, and the functionality of the fusion was tested using a growth curve with PNPase knockdown using the CRISPRi system [66]. PNPase is essential in mycobacteria and causes a growth defect when knocked down, but complementation with a CRISPRi-resistant PNPase::eGFP fusion restored growth back to wildtype levels (Supplementary Figure 7A). RNase J was tagged with Dendra2, and its functionality was tested by a rifampicin killing curve because rifampicin causes an increase in killing in an RNase J knockout in M. smegmatis (which is notably different from M. tuberculosis where knockout of RNase J increases rifampicin tolerance [24]). Complementation of an RNase J deletion strain with the RNase J::Dendra2 fusion restored killing back to wildtype levels (Supplementary Figure 7B). No known phenotypes of RhlE1 deletion strains in M. smegmatis have been reported to our knowledge, so we were unable to test its functionality in vivo. RhlE1 has a C-terminal IDR, and we fused Dendra2 to the C-terminus of the protein to minimize potential impact on the catalytic domain. In these strains, the sole copy of RNase E was fused to mCherry (with the IDR mutants as in Figure 4C–F), and they were merodiploid for the other degradosome proteins that was fused to a GFP.

To quantify the degree to which the proteins were localized in the same space in the cell, the fluorescence intensity in the green channel (PNPase, RNase J, or RhlE1) was plotted versus the signal in the red channel (RNase E), and the Pearson’s correlation coefficient between both channels was reported. All three partner proteins (PNPase, RNase J, and RhlE1) had localization patterns that were significantly correlated with that of full-length RNase E (Figure 6). Similar to the in vitro immunoprecipitation results, PNPase showed reduced localization with RNase E in vivo when the C-terminal IDR of RNase E was deleted (Figure 6A–B), suggesting that its localization in cells is closely tied to its interaction with the C-terminal IDR of RNase E. Unexpectedly, deletion of the N-terminal IDR of RNase E led to a small but statistically significant increase in localization with PNPase.

RNase J also showed a modest increase in localization with RNase E upon deleting the N-terminal IDR of RNase E, but it localized with the C-terminal IDR deletion and double IDR deletion of RNase E similarly to full-length RNase E (Figure 6C–D). This suggests that the interaction observed in vitro between RNase J and the C-terminal IDR of RNase E is not a major contributor to the localization of RNase J in cells; perhaps interactions with RNA or another unidentified protein are responsible for the correlated in vivo localization of RNase J and RNase E.

The impact of the RNase E IDRs on subcellular localization patterns of RhlE1 was very similar to that of RNase J. RhlE1 had a modest increase in localization with RNase E when the N-terminal IDR of RNase E was deleted, but the C-terminal IDR deletion and double IDR deletion of RNase E correlated with RhlE1 to the same extent as full-length RNase E (Figure 6E–F). This suggests that the localization of RhlE1 is primarily controlled by other factors besides its direct interaction with the IDRs of RNase E in cells. RhlE1 itself contains an IDR on its C-terminus, so it is possible that this region plays a dominant role in its localization in cells as condensate formation was shown to be conserved among RNA helicases in bacteria [21].

In vivo localization of degradosome-like network proteins changes during carbon starvation stress

When M. tuberculosis is inside the human host during infection, the bacteria face many stressors including hypoxia within the granuloma [67, 68]. Hypoxia has been shown to dramatically reduce rates of mRNA degradation in mycobacterial cells [3, 4] which has been explained as an energy conservation mechanism during periods of stress for the bacteria. Carbon starvation has also been shown to cause stabilization of mRNA [3], and like hypoxia represents a type of energy stress. We hypothesized that weak, transient interactions and localization patterns within the cell were primarily responsible for efficient cleavage rates during log phase growth, rather than fixed interactions or post-translational modifications such as phosphorylation. To test this, we first confirmed that the degradosome enzymes RNase E, PNPase, and RNase J cleaved the 29 nt fluorescently labeled oligo from Figure 3A–B (Supplementary Figure 8A). To identify which enzyme was primarily responsible for cleavage of this short RNA in a cell lysate, we utilized the CRISPRi system to knock down PNPase in wildtype and RNase J knockout backgrounds (Supplementary Figure 8B–C). An RNase E knockdown strain was made by replacing the native promoter with an inducible one in the chromosome [10]. We then lysed these strains, expecting that lysis would disrupt weak interactions and subcellular localization patterns but not strong interactions or post-translational modifications. We assayed cleavage the fluorescently labeled RNA by lysates. We found that PNPase was primarily responsible for the cleavage of this short oligo in a lysate (Supplementary Figure 8D and F) and that RNase E knockdown had no impact (Supplementary Figure 8E–F). This suggests that RNase E is unlikely to be cleaving short RNAs in cells, consistent with its suggested role in initiation of mRNA degradation. To assess the impact of stress on the ability of PNPase to cleave short RNAs in lysates, we used a 22 nt fluorescently labeled RNA and performed cleavage assays using lysates from cells that had been grown in hypoxia or carbon starvation stress since these conditions are known to stabilize mRNA globally. We found that there were no differences in cleavage rates in lysates from either stress condition compared to log phase (Supplementary Figures 8G–J), consistent with the idea that transient interactions and localization patterns control mRNA cleavage rates in mycobacteria.

Furthermore, we looked at the impact of stress on RNA degradation protein localization in living M. smegmatis cells by assessing protein localization with live-cell microscopy using the previously generated fluorescently labeled protein strains from Figure 6. However, imaging of hypoxic cells was not technically feasible with our imaging system. Carbon starvation was therefore utilized to assess in vivo localization of degradosome proteins in M. smegmatis in a condition causing stress-induced mRNA stabilization. First, the localization patterns of each protein were individually assessed by counting foci per micron during log phase growth (carbon rich) and following 24 hours of carbon starvation (Figure 7A). In all cases, protein localization was impacted by carbon starvation stress. RNase E, PNPase, and RNase J showed a decrease in the number of foci per micron, suggesting that the degradosome-like network dissolves or becomes more diffuse during energy stress. RhlE1 showed less foci than the other three proteins during log phase growth but had an increase in foci formation during carbon starvation.

The correlations of PNPase, RNase J, and RhlE1 localization with RNase E were also assessed in carbon starvation in a similar way as in Figure 6. PNPase (Figure 7B–C) and RNase J (Figure 7D–E) showed small but statistically significant decreases in Pearson’s R values during carbon starvation stress. This suggests that these protein interactions with RNase E are disrupted when RNA is stabilized, furthering the idea of a dissolved degradosome-like network during stress. The changes were likely small due to the proteins spreading out to similar extents leading to more uniform signal across both channels in the cell during carbon starvation. RhlE1 did not show a statistically significant change in its localization pattern with RNase E (Figure 7F–G), but we had previously suggested that the localization pattern of RhlE1 may not be fully dependent upon RNase E (Figure 6E–F). The changes in localization patterns of the mRNA processing and degradation proteins during a clinically relevant stress condition are consistent with a model in which mycobacteria can reorganize their degradosome-like network to slow down their mRNA degradation rates to conserve energy and survive inside the host during an infection. The roles of the RNase E IDRs in the formation of the degradosome-like network and the changes in protein localization during carbon starvation are summarized in Figure 8.

Figure 8. Model of the degradosome-like network in M. smegmatis during log phase with RNase E IDR deletion mutants and carbon starvation.

The circles in the cells show the foci identified in vivo with teal representing RNase E, blue representing RNase J, red representing PNPase, and yellow representing RhlE1. Fuzzy circles represent diffuse foci. Black dashed lines represent transient interactions with RNase E identified in vitro.

DISCUSSION

RNase E is known to play major roles in RNA processing and degradation in many species of bacteria. In E. coli, RNase E is known to control global mRNA decay rates [69–71], and this activity is influenced by the degradosome that is scaffolded by its one IDR [15, 16, 37]. While RNase E was known to have a major role in RNA degradation in mycobacteria [10], its interactions with other putative degradosome components were previously unknown. Here, we found that RNase E in mycobacteria, and the Actinobacteria more broadly, has two IDRs. In M. smegmatis, the two IDRs are nonessential and do not affect catalytic activity of a short oligo substrate. Deleting either IDR results in a relatively small number of differentially expressed genes, but the double deletion shows many upregulated and downregulated genes indicating substantial functionally redundancy. Consistent with this, the IDRs both contribute to RNA binding in a partially redundant way. The IDRs also contribute to condensate formation and interactions with other RNA degradation proteins. Furthermore, we have defined many of the direct protein-protein interactions that mediate formation of degradosomes or, more likely, degradosome-like networks in mycobacterial cells.

RNA degradosomes previously described in other bacteria have not been shown to include both RNase E and RNase J. For the first time, we have identified a direct protein-protein interaction between these two proteins, as well as interactions between each protein and other degradation proteins. RNase E directly interacts with the other three RNA processing and degradation enzymes that we tested, consistent with expectations for an RNase E-scaffolded degradosome. Previous work in M. tuberculosis suggested, but did not directly test, the presence of PNPasescaffolded degradosomes as well [11], which has been defined in eukaryotic mitochondria by association with the RNA helicase Suv3 (reviewed in [72]). Our in vitro results were also consistent with this idea; PNPase directly binds each of the three other enzymes that we tested. RNase J and RhlE1 each bind to two other enzymes. A great number of multi-protein complex configurations are therefore possible. However, any complex of more than two different degradosome components in mycobacteria remains hypothetical. We did not detect the one possible three-way interaction that we were able to assay by coimmunoprecipitation in this study (Figure S4E). In our hands, the interactions between degradosome components were too weak to analyze by size-exclusion chromatography. We were also unable to immunoprecipitate degradosomes from M. smegmatis or M. tuberculosis cell lysates; in the absence of crosslinking, RNase E did not pull down any other proteins in appreciable quantities, and the addition of crosslinking agents caused RNase E to partition into very large, insoluble complexes. Collectively with our in vivo localization data (discussed later), these observations, and our in vitro data are consistent with the idea of a degradome-like network as has been previously described in B. subtilis [41], with many relatively weak protein-protein and protein-RNA interactions producing a dynamic RNA degradation network.

We found that the IDRs of RNase E contribute to protein-protein interactions, but the importance of these roles in cells differed from what our in vitro data suggested. The C-terminal IDR was required for RNase E to interact with PNPase, RhlE1, and RNase J in vitro, and the N-terminal IDR was additionally required for the RNase E-RhlE1 interaction. In the case of PNPase, deletion of the C-terminal IDR of RNase E did impact the colocalization of the two proteins within cells, as predicted. As the C-terminal IDR deletion did not affect degradation of the transcripts we tested, this suggests that PNPase and RNase E do not need to be localized in close proximity in cells for normal mRNA degradation rates. This is similar to what was reported in E. coli with respect to the interaction between PNPase and RNase E, which does not appear to be important for normal in vivo mRNA degradation rates [37]. However, it contrasts with findings that deletion of the C-terminal RNase E IDR reduces degradation efficiency of mRNA in E. coli [37] and 5’ UTR cleavage in C. crescentus [17]. It is possible that the interaction of the N-terminal IDR of mycobacterial RNase E with RNA mitigates the impact of loss of the C-terminal IDR (discussion further below).

We observed an increase in localization between all three degradosome proteins when the N-terminal IDR of RNase E was deleted. However, deleting the N-terminus caused less efficient degradation of the mRNAs that we tested, which suggests that the efficiency of the degradation is likely controlled by factors outside of the degradosome protein interactions. It is possible that other mechanisms with different interaction partners we have not yet identified, such as RNA binding proteins or RNA decapping enzymes, are responsible for fine-tuning the efficiency of mRNA degradation rates and interact with the N-terminus of RNase E. Surprisingly, deletion of the C-terminal IDR of RNase E did not significantly impact its colocalization with RNase J and RhlE1 in cells, despite being required for direct protein-protein interaction in vitro. RhlE1 contains its own IDR, which could possibly contribute to its localization pattern in cells. This suggests that the localization of these proteins within cells is a product of multiple protein-protein and protein-RNA interactions, and is therefore robust to perturbation of direct interactions with RNase E. This is further evidence that mycobacteria do not have a stable degradosome complex with fixed stoichiometric ratios like that of E. coli[30–34], but rather dynamic, transient interactions between the RNA processing and degradation enzymes dominate in mycobacterial cells allowing for rapid adaptation to environmental changes.

Additionally, while we only tested enzymes previously predicted to interact with RNase E here [11], it is likely that RNase E has other binding partners that have yet to be identified and may contribute to the proposed degradosome-like network.

In addition to organization mediated by specific protein-protein interactions, cytoplasmic organization of RNA processing and degradation by biomolecular condensates has been found in both prokaryotes and eukaryotes. In eukaryotes, stress granules have been described as membraneless organelles that can sequester intronic RNA damaged from UV light [73] (and reviewed further in [74]). In yeast and other eukaryotes, P-bodies form and organize mRNA decapping and decay (reviewed in [75]). This idea has been shown in prokaryotes such C. crescentus where RNase E condensates have been shown to stimulate mRNA degradation with the decay process being initiated by an RNase E cleavage event [22]. The RNase E IDR was necessary for condensate formation in C. crescentus [17], and RNase E can recruit PNPase to the condensate to stimulate PNPase activity on cleaved RNAs [20]. In mycobacteria, we observed that the IDRs of RNase E were partially redundant and necessary for condensate formation in vitro but not in vivo. One possibility is that the double deletion formed small condensates that were not large enough to be visualized by diffraction limited microscopy. Another possibility for this observation is that the in vitro condensates are fundamentally different than the in vivo ones, possibly because they lack additional components found in cells such as other proteins. In vivo, these condensates appear to be more like P-bodies in yeast, where disrupting one component does not disrupt the entire condensate [63], and multiple proteins from the body can drive phase-separation [76], furthering the idea of a degradosome-like network held together by a combination of RNA binding, condensates, and transient protein-protein interactions.

It has been well-established in the literature that IDRs across all life can play roles in substrate recognition and nucleic acid binding (reviewed in [77]). We established that the IDRs of RNase E are necessary for robust RNA binding in M. smegmatis, which is consistent with literature on E. coli where the sole IDR of RNase E contains a known RNA binding regions [78, 79] that recognizes specific RNA substrates [79]. It has been shown that one IDR of a human RNA helicase, DDX3X, is able to recognize a specific type of secondary structure called G-quadraplexes in RNA [80]. It has also been shown that M. tuberculosis has more RNA G-quadraplexes than M. smegmatis and that chemical stabilization of these RNAs causes slowed growth and caused translational inhibition [81]. RNase E has been shown to preferentially cleave before a cytidine in M. smegmatis [10], but no specific sequence motif for cleavage has been identified. It is possible that one or both IDRs in RNase E (and perhaps the IDR of RhlE1) contributes to recognition of RNA secondary structures such as G-quadraplexes; this is an avenue for further investigation.

We hypothesized that the protein-protein interactions that do occur in cells may cause allosteric changes that promote activity of the enzymes. Evidence for this idea was found in E. coli, where binding the of the helicase RhlB to RNase E causes allosteric changes in the helicase that allow for more efficient RNA unwinding [15, 28, 29]. To our surprise, our findings do not support the idea of allosteric changes within the enzymes upon formation of the degradosome-like network in mycobacteria. Instead, we propose that the close proximity of the enzymes within the network allows for complete mRNA degradation so that the products of one enzyme are readily available to enter as the substrate for the next enzyme in the pathway. Transient interactions are unlikely to be ideal for allosteric changes within an enzyme, which may explain these results in the context of a degradosome-like network. In E. coli, disruption of the degradosome impacts the degradation of mRNA but not rRNA processing [37]. While we only assessed the impact on mRNA here, it is likely that this is the case in mycobacteria which could be assessed in the future.

Finally, we demonstrated that M. smegmatis disrupts its degradosome-like network in response to carbon starvation stress. RNA stabilization has been proposed as an energy conservation strategy in mycobacteria [3], so these results align with the post-translational regulation of these enzymes to prevent rapid mRNA decay during stress. RNA stabilization is a clinically relevant stress response in M. tuberculosis, as it has been shown to occur during hypoxia [4, 64] which occurs inside granulomas [67, 68]. As a result, potential therapeutic strategies to investigate further for treating pathogenic mycobacteria include an RNase E activator or forcing formation of the degradosome-like network during stress when the cells do not have the energy to replace the degraded transcripts.

The results presented here support the IDRs of RNase E playing multiple noncatalytic roles in the cell including RNA binding, condensate formation, and protein-protein interactions. We propose a model in which RNase E forms a degradosome-like network composed of dynamic, transient interactions with other proteins in cells that can be disrupted upon RNA stabilization with a physiologically relevant stressor.

Supplementary Material

Supplement 1

Supplementary Table 6. Differentially expressed genes with the RNase E IDR deletions

Supplement 2

Supplementary Table 1. M. smegmatis strains

Supplementary Table 2. Plasmids integrated into M. smegmatis

Supplementary Table 3. RNA oligos

Supplementary Table 4. Plasmids used for protein overexpression in BL21 E. coli

Supplementary Table 5. M. smegmatis genes and proteins

Supplementary Table 7. Ribonucleoside reductase genes upregulated in the RNase E double IDR deletion

1

Supplementary Figures 1–8

Funding Statement

This work was supported in part by NIH-NIAID award 5TP01AI143575-02 to SSS, by NIH-NIAID award R21 AI156415-01A1 to SSS, by NSF-CAREER award 1652756 from the Directorate of Biological Sciences to SSS, an award from the Potts Foundation to SSS, NIH NIGMS award R35GM124733 to JMS, DoEd GAANN training grant P200A240115 to ARR, and a Dr. Armand P. Ferro and Mary H. Ferro Summer Fellowship to ARR. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.