The Mycobacterium tuberculosis sRNA F6 regulates expression of groEL/S

Almost 140 years after the identification of Mycobacterium tuberculosis as the etiological agent of tuberculosis, important aspects of its biology remain poorly described. Little is known about the role of post-transcriptional control of gene expression and RNA biology, including the role of most of the small RNAs (sRNAs) identified to date. We have carried out a detailed investigation of the M. tuberculosis sRNA, F6, and show it to be dependent on SigF for expression and significantly induced during in vitro starvation and in a mouse model of infection. However, we found no evidence of attenuation of a ΔF6 strain within the first 20 weeks of infection. A further exploration of F6 using in vitro models of infection suggests a role for F6 as a highly specific regulator of the heat shock repressor, HrcA. Our results point towards a role for F6 during periods of low metabolic activity similar to cold shock and associated with nutrient starvation such as that found in human granulomas in later stages of infection.


INTRODUCTION
In spite of claiming more than 3000 lives on a daily basis, the impact of Mycobacterium tuberculosis as a major human pathogen is currently being overshadowed by the ongoing COVID-19 pandemic. Notoriously difficult to work with and very different from both Gram-positive and Gram-negative model organisms, we still know preciously little about many aspects of its basic biology including RNA biology and post-transcriptional control of gene expression. However, the study of regulatory RNA in M. tuberculosis is slowly gaining momentum, aided by nextgeneration sequencing (NGS) applications, which have provided significant insights into the abundance and dynamics of non-coding RNA over a range of growth conditions, the location of transcription start sites on a global scale and translated versus untranslated transcripts e.g. (1)(2)(3)(4)(5). Still, much remains to be uncovered about the post-transcriptional control exerted by regulatory RNA in M. tuberculosis in particular what role these molecules play in the pathogenesis, latency and persistence of M. tuberculosis. A multitude of M. tuberculosis small regulatory RNAs (sRNAs) have been identified in the last decade, but only few have been investigated and even less, potentially linked to latency and persistence e.g. (1,3,4,(6)(7)(8)(9)(10)(11)(12)(13).
F6 (ncRv10243) was one of the first M. tuberculosis sRNAs to be identified, not by RNA-seq but by cDNA cloning (6). F6 is conserved in a range of pathogenic mycobacteria and in Mycobacterium smegmatis with the 5' end showing the highest degree of conservation (6,7). The location of F6, i.e. between the convergent fadA2 (encoding an acetyl-CoA transferase) and fadE5 (encoding an acyl-CoA hydrogenase), is also highly conserved suggesting a regulatory role for F6 in lipid metabolism, which is critical for intracellular survival (14). Initial analysis, combining northern blotting with 5' and 3' RACE, revealed that F6 is expressed as a 102-nucleotide transcript, which is 3' processed to the dominant transcript of 58 nucleotides, comprising the more conserved part of the transcript (6). F6 is upregulated in stationary phase, during oxidative stress, and in low pH, while overexpression from a multicopy-number plasmid leads to a slow-growth phenotype on solid media (6). The F6 promoter contains a typical SigF promoter motif and exhibits the highest SigF occupancy according to ChIP-chip analysis, while overexpression of SigF leads to slow growth akin to overexpression of F6 (6,15). SigF is a non-essential sigma factor, conserved in most mycobacteria; its expression is induced by several stresses including anaerobiosis, nutrient starvation, oxidative stress and cold shock, while heat shock downregulates its expression, and deletion leads to attenuation in mice (16)(17)(18). (19).
In the current study, we have investigated M. tuberculosis F6 expression under different growth conditions and found that F6 is significantly upregulated in an in vitro model of persistence (nutrient starvation) and in mouse lungs. We have generated an F6 deletion strain (DF6) and using microarrays we have analysed differential gene expression which, in combination with in silico prediction of sRNA-mRNA interactions, has led to the identification of a potential regulatory target of F6. Finally, we have assessed the fitness of the DF6 strain in vitro and in vivo. Our results indicate that although deletion of F6 does not lead to attenuation in macrophages or our specific mouse model, it may still be critical for certain aspects of infection. F6 is highly upregulated in mice, it does control hrcA expression and it does lead to impaired recovery from a Wayne model of hypoxia, suggesting a role in later stages of infection.

Bacterial strains and plasmids
Escherichia coli DH5α was used for plasmid construction and grown in LB Agar or Broth using kanamycin at 50 μg/ml and X-Gal where necessary at 200 μg/ml.
Mycobacterium tuberculosis was grown on 7H11 agar plus 10% OADC (Becton Dickinson). Liquid cultures were grown in standard Middlebrook 7H9 medium supplemented with 0.5% glycerol, 10% Middlebrook ADC (Becton Dickinson) and 0.05% Tween-80 at 37°C in a roller bottle (Nalgene) rolling at 2 rpm or 50 ml falcon tubes (Corning) in an SB3 Tube Rotator (Stewart) at 28 rpm. Kanamycin was added where required at 25 μg/ml and X-Gal where required at 50 μg/ml. All plasmids and oligos used in this study are listed in Tables S1 and S2.

Wayne Model
Cultures were grown to exponential phase (OD600=0.6-0.8) and subsequently diluted to OD600=0.005 in 7H9 in triplicate in Wayne tubes, each containing a sterile stirring bar. Cultures were incubated at 37°C on a stirring platform and OD monitored over time until cultures reached NRP-2. Cultures were then either diluted into 7H9 OD600=0.05 and OD monitored, or serially diluted onto 7H11 agar and CFU enumerated.

Construction of an F6 deletion strain and complemented strain
Allelic replacement techniques were used to generate an M. tuberculosis knockout mutant as per published protocol (20). Briefly, approximately 1.5kb of the flanking region from either side of F6 (5' flank coordinates 292520-293605 and 3' flank 293709-294600) were cloned into the suicide vector pBackbone. The sRNA was removed by site directed mutagenesis to produce the targeting plasmid pJHP04.
Electrocompetent H37Rv was transformed with pJHP04 and subsequent single cross overs (SCOs) and double cross overs (DCOs) selected.
To complement the ΔF6 mutant, electrocompetent cells were prepared and transformed with an integrating plasmid (pKP186) containing the cloned region 293428-293876 from H37Rv.

Preparation of starved cultures
M. tuberculosis H37Rv, ΔF6 and complemented strain were grown in Middlebrook 7H9 supplemented with 0.4% glycerol, 0.085% NaCl, 0.5% BSA and 0.05% Tyloxapol in roller bottle culture (2 rpm at 37°C). Exponentially growing bacteria were pelleted and washed 3 times with PBS supplemented with 0.025% Tyloxapol and finally resuspended in PBS with 0.025% Tyloxapol to an equivalent starting volume and incubated statically for 24 or 96 hours.

RNA isolation
RNA isolation from in vitro cultures was done as described previously (6). Briefly, cultures were harvested with rapid cooling by adding ice directly to the culture and subsequent centrifugation at 10,000 rpm for 10 minutes. RNA was isolated from the pellet using the FastRNA Pro Blue Kit from MP Biomedicals following manufacturer's instructions.
To isolate RNA from bacteria grown in mice, lung homogenates were spun at 13,000 rpm for 5 minutes to collect the bacteria. These were resuspended in 1 ml Trizol (Invitrogen) with 150 micron glass beads and the samples disrupted in a fast Prep (MPBio) at a setting of 6.0 for 40 secs. The RNA was extracted according to manufacturer's guidelines. RNA concentration was measured by nanodrop and RNA integrity measured by the 2100 Bioanalyzer using a Nano chip.

Quantitative RT-PCR
Total RNA was treated with Turbo DNase (Ambion) until DNA free. cDNA was synthesized using Superscript III (Invitrogen) and random hexamers. Primers were designed using the Applied Biosystems software Primer Express, and sequences are listed in Table S1. Each 20 µl qRT-PCR reaction, contained 16SYBRgreen (Applied Biosystems), 900 nm each primer and 5 µl of template cDNA. Absolute quantitation was performed and all genes were normalised to 16S expression.

Mapping genome data and variant calling
WGS was performed as described elsewhere using the Illumina HiSeq platform (21).
Sequencing reads for the ΔF6 strain were mapped against the M. tuberculosis H37Rv reference genome (AL123456) using BWA (22), and variants called with SAMtools (23). Variant filtering was performed by inclusion of only those variants with a minimum mapping quality of 10 and maximum read depth of 400. Finally, heterozygous calls or those found in repetitive or mobile elements (genes annotated as PE/PPE/insertions/phages) were removed.

Macrophage Infection
Bone marrow derived macrophages (BMDMs) were generated from 6-8 week old Plates were incubated for 3-4 weeks to obtain CFU counts.

Expression of F6 is up-regulated by starvation and during infection
SigF is the suspected master regulator of F6 expression, and this was validated by northern blotting, which demonstrated a complete absence of F6 in a DsigF strain (Fig.   S1). As SigF is highly expressed during nutrient starvation and infection, we speculated that upregulation of F6 under these conditions was highly likely (16). A widely used in vitro model of M. tuberculosis latency and persistence is static incubation in PBS, effectively starving the bacteria for nutrients and to some extent oxygen, which leads to a significant up-regulation of SigF (16). We subjected cultures of M. tuberculosis H37Rv to starvation by washing and resuspending log-phase cultures in PBS and incubating them in standing cultures for 96 hours before isolating total RNA. F6 expression was measured by qRT-PCR (normalised to 16S rRNA) and compared to log-phase levels, and the results, shown in Fig. 1, demonstrate that F6 expression increased by almost two logs (>75-fold) in response to starvation, suggesting a role for this sRNA under limiting nutrient availability that may be seen during infection. Next, we wanted to investigate F6 expression in a mouse model of infection. BALB/c mice were infected with ~100 CFU of H37Rv via aerosol route and left for three weeks before culling and isolation of M. tuberculosis total RNA from lung tissue. The RNA was analysed as described above demonstrating that F6 expression was up-regulated approximately 6-fold in mouse lung tissue compared to in vitro log-phase levels, but clearly not as dramatically as during in vitro starvation (Fig. 1). Together our results demonstrate that the expression of F6 is highly dynamic and that it may play a role during infection, more specifically in nutrient-poor environments.

Deletion of F6 in M. tuberculosis H37Rv
To identify a potential regulatory role of F6 with minimal impact on the two flanking fad genes, we generated an unmarked F6 deletion strain (ΔF6, see materials and methods for details). Candidates were screened by PCR amplification and sequencing of the PCR product, which confirmed the deletion of the sRNA from the M. tuberculosis genome (Fig. S2).
To further verify the deletion and to determine if there were any secondary mutations that might affect subsequent phenotypic analysis, ΔF6 strain was sequenced using high-throughput sequencing. Alignment of the genomic sequence of ΔF6 with H37Rv identified two single-nucleotide polymorphisms (SNPs) in addition to those generated by the deletion scar; one was C2864730T, resulting in an R102W substitution in Rv2541 while the other was C4178146T in the 5' UTR of Rv3729.
To ensure that these SNPs did not influence any potential phenotypes of the F6 deletion strain, we constructed a complemented strain, in which expression of F6 was driven by its native promoter from a single-copy plasmid integrated on the chromosome. F6 expression in wildtype, H37Rv, DF6 and the complemented strain was compared using qRT-PCR, and the results demonstrate that there was no statistical significance (unpaired t-test, p<0.05) between wildtype and the complemented strain, while there was no detectable expression in DF6 ( Figure S3).
The DF6 strain was assessed for an in vitro growth phenotype by comparing its growth to that of wildtype and the complement strain in 7H9 roller bottle cultures over two weeks. Under these conditions, there was no significant difference between the three strains (Fig. 2).

Recovery from Wayne Model hypoxia is impaired in DF6
In addition to the starvation model of latency, another widely used in vitro-model is the Wayne model, in which the amount of available oxygen is gradually limited in sealed cultures of M. tuberculosis (24). When oxygen concentrations decrease to the microaerobic level (1% oxygen saturation), the cells enter a state of non-replicating persistence (NRP-1) followed by NRP-2 when oxygen saturation reaches 0.06%.
SigF is highly induced during anaerobiosis such as that observed the Wayne model, suggesting that its regulon, including F6, may play a role in low-oxygen conditions (18).
We therefore decided to evaluate the fitness of the DF6 strain using the Wayne model.
To measure potential differences in respiration rate, the depletion of oxygen for each strain was monitored by a methylene blue indicator tube set up for each strain at the start of incubation. After 14 days incubation, cultures were plated onto 7H11 agar for determination of colony forming units (CFU). In addition, NRP-2 cells were assessed for their ability to be resuscitated by dilution into fresh 7H9 media and monitoring of growth (OD600). Enumeration was performed to ensure that an equal number of bacteria was inoculated.
According to the CFU counts, there was no significant difference in survival between wildtype H37Rv, ΔF6 and the complemented strain after incubation in the Wayne model indicating equal numbers of live bacteria per unit of OD (Fig. 3A). This also demonstrated that an equal number of bacteria were used in the inoculation of fresh media for regrowth. Conversely, the growth curve of the NRP-2 cells transferred to fresh media indicated that the ΔF6 strain was impaired for recovery/resuscitation following incubation in the Wayne model, while the complemented strain displayed an intermediate phenotype (Fig. 3B). These results suggest that F6 may play a role in resuscitation of nonreplicating M. tuberculosis, associated with the reactivation of latent infection.

Deletion of F6 induces a heat shock response during starvation.
To establish a potential role for F6 during nutrient starvation, we performed microarray analysis on RNA isolated from PBS-starved cultures of wildtype H37Rv and DF6. To strike a balance between starvation and RNA quality, which showed significant degradation after 96 hours in PBS, we decided to reduce the PBS incubation to 24 hours. Experiments were performed in biological triplicates and technical (dye-swap) duplicates using Agilent gene expression microarrays (see materials and methods).
To our surprise, only four genes showed ≥ 2-fold change (between 2.5 and 4 fold) and these were all upregulated in the mutant (Table 1). Strikingly, all of these are under the control of the transcriptional repressor HrcA (Rv2374c), although encoded in different parts of the genome (25). Two of these are Rv0440 (groEL2) and Rv3418c (groES), i.e. heat-shock inducible chaperonins, the other two genes are conserved hypotheticals, Rv0990c and Rv0991c, which form an operon (25).  The results confirmed that groES, groEL2, and Rv0991c were significantly upregulated in DF6 during starvation, and that the phenotype could be largely complemented by providing a copy of F6 in trans (Fig. 4). The changes for Rv0990c

Gene number Gene name Fold change P-value
(co-transcribed with Rv0991c) were not significant according to qRT-PCR.

In silico prediction of F6-hrcA RNA interaction
The HrcA-regulon comprises very few genes in M. tuberculosis, i.e. the four in table 1 as well as groEL1, which is co-transcribed with groES (25). Deletion of hrcA leads to upregulation of those specific genes, and the most likely explanation for the observed coordinated upregulation in the DF6 strain, is a down-regulation of HrcA expression leading to de-repression of its regulon. In other words, F6 activates HrcA expression, directly or indirectly. hrcA was not among the significantly differentially expressed genes in the microarray analysis, but a direct interaction between mRNA and sRNA does not necessarily affect mRNA levels. Importantly, a lack of significant change in mRNA levels does not exclude regulation by a given sRNA.
The 5' untranslated region (UTR) of the hrcA mRNA is relatively short (34 nucleotides), but nevertheless has the potential to form a stable stem-loop that sequesters the start codon and the region immediately upstream, excluding the potential Shine-Dalgarno (Fig. 5A). As our results were largely consistent with a role for HrcA in the coordinated upregulation of the two operons, we decided to investigate a potential interaction between F6 and the hrcA mRNA in silico, using the targetRNA2 webserver (26). We changed the default of 80 nucleotides upstream to 34 nucleotides in line with the mapped transcription start site for hrcA (2) and used hrcA as single target with a hybridisation seed of six nucleotides. The result shown in Fig. 5B suggests that F6 can interact with the first 12 nucleotides of the hrcA mRNA, i.e. the exact region that sequesters part of the translation initiation region. This interaction was confirmed by RNAhybrid ( Fig. 5) (27). Further studies are required to validate this prediction, but is seems plausible that F6 is involved in activating hrcA translation by opening an otherwise autoinhibitory structure similar to what has been described for the interaction between the E. coli rpoS mRNA and several sRNAs (28).

Fitness of the ΔF6 strain in macrophages and mouse models of infection
M. tuberculosis is an intracellular pathogen that is subjected to a variety of stresses within the host, including but not limited to oxidative stress and nutrient deprivation. F6 is induced under oxidative stress, starvation and infection, and expression is under control of SigF, which is critical for intracellular survival in mice and guinea pigs (17,29). Moreover, the DF6 strain displayed impaired recovery from the Wayne model. CFUs. For both naïve and activated macrophages we observed no significant differences between the three strains, suggesting that under these conditions, F6 does not result in attenuation in the mouse macrophage model (Fig. 6). H37Rv and ΔF6 at any of the time points (Fig. 6).

DISCUSSION
M. tuberculosis F6 is a SigF-controlled sRNA, which is well-expressed in exponential phase, but up-regulated by a variety of stresses (6).
In this study we have shown that F6 is also highly up-regulated after three weeks of murine infection and to an even higher extent in the starvation model of 96 hours starvation in PBS, suggesting that this sRNA plays a role in the early stages of pathogenesis but possibly a more significant one for survival in nutrient deficient environments such as those encountered in human granulomas in later stages of infection. M. tuberculosis has the ability to resuscitate after long periods of low metabolic activity/dormancy (24,30), and using the Wayne model of dormancy, we found that recovery from NRP-2 was impaired for DF6 versus wildtype/complement, although initial survival appeared to be unchanged, according to CFU counts.
Our study indicates that starvation for 24 hours leads to an upregulation of three loci (encoding groES, groEL2, and Rv0991c) in the ΔF6 strain compared to the wildtype and complemented strains. Together, these loci make up the HrcA regulon suggesting that this heat shock repressor is likely coordinating and responsible for the observed differences (25). We observed no significant change in the levels of hrcA mRNA, but we did identify a potential autoinhibitory structure within the 5' UTR of the transcript, which could be sequestered by interaction with F6 to expose the entire translation initiation region. SigF has previously been shown to be downregulated during heat shock (18), offering the possibility that F6 represents a critical regulatory link between heat/cold shock and the expression of essential chaperonins by directly targeting the hrcA mRNA and thereby GroEL/S (Fig. 7).  The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.