A Riboswitch-Based Inducible Gene Expression System for Mycobacteria

Research on the human pathogen Mycobacterium tuberculosis (Mtb) would benefit from novel tools for regulated gene expression. Here we describe the characterization and application of a synthetic riboswitch-based system, which comprises a mycobacterial promoter for transcriptional control and a riboswitch for translational control. The system was used to induce and repress heterologous protein overexpression reversibly, to create a conditional gene knockdown, and to control gene expression in a macrophage infection model. Unlike existing systems for controlling gene expression in Mtb, the riboswitch does not require the co-expression of any accessory proteins: all of the regulatory machinery is encoded by a short DNA segment directly upstream of the target gene. The inducible riboswitch platform has the potential to be a powerful general strategy for creating customized gene regulation systems in Mtb.


Introduction
Tools for manipulating gene expression are fundamental to genetic studies. Inducible systems, usually under the control of a small molecule, are particularly useful because they permit exquisite experimental control over both the dose and timing of gene expression. Inducible promoters are widely used to silence genes via direct transcriptional control or antisense methodologies and to overexpress proteins for biochemical and structural studies [1]. In bacteria, inducible systems have been used to elucidate gene function, determine gene essentiality, and validate drug targets [2,3,4].
Although several inducible expression systems exist for Gramnegative bacteria, adaptation to distantly related bacteria has proven difficult. Species that lack diverse regulated expression tools include the mycobacteria, among them Mycobacterium tuberculosis (Mtb), which causes tuberculosis in humans, and species that are commonly used as models for Mtb such as the fish and amphibian pathogen M. marinum, the non-pathogenic M. smegmatis (Msmeg), and the vaccine strain M. bovis BCG [5]. Unique challenges inherent to the biology of these medically relevant organisms, such as their pathogenesis, slow growth rate, and inefficient DNA uptake, have significantly hindered molecular genetics studies [6].
The earliest described mycobacterial inducible system, and the only one derived from endogenous mycobacterial machinery, is the acetamide-inducible Msmeg acetamidase promoter [7]. Although the system has proven useful for conditional mutant construction and protein overexpression, the acetamidase promoter exhibits a high level of basal activity and is prone to recombination in Mtb [7,8,9,10]. Alternatives have been derived from transposons or the regulons of other Gram-positive bacteria [5,11,12,13,14,15], including several tetracycline repressor-based (TetR) systems [16,17,18,19].
For all of the above regulons, response to the inducer is mediated by one or more accessory proteins that must be imported into mycobacteria. Expression of these exogenous regulators can require optimization to achieve desired levels of induction [19]. Moreover, further adapting such systems can entail protein engineering and extensive characterization to verify function [20]. While these tools have proven useful in some contexts, alternative regulatory strategies, especially those that circumvent these limitations, would be valuable additions to the mycobacterial genetic toolkit.
We have recently reported a series of theophylline-responsive riboswitches that can control gene expression in a range of Gramnegative and Gram-positive bacteria, including Msmeg, a fastgrowing, non-pathogenic species that is a widely used model system for Mtb [21]. Significantly, no exogenous regulator proteins are involved in the induction mechanism, making the system easy both to modify and to implement in different strains or species. The machinery is encoded by a single ,300-bp DNA segment comprising a mycobacterial promoter (a variant of Phsp60, a widely used constitutive promoter from BCG [22]) and a synthetic RNA aptamer that binds to theophylline ( Figure 1) [21]. Here we characterize this promoter-riboswitch combination and show that it can be used to induce and repress gene expression reversibly; to control a conditional gene knockdown; and to regulate expression in a macrophage model of infection.

Characterization of riboswitch-controlled gene expression in mycobacteria
To assess the generality of riboswitch response across different target genes, we created two constructs, ribo-gfp and ribo-lacZ, which are designed to express GFP and b-galactosidase under control of the riboswitch, and assayed their fluorescence or enzyme activity. With both constructs, we observed dosedependent induction by theophylline with maximum reporter expression at ,2 mM theophylline (Figure 2A), whereas significant growth attenuation was only observed at $10 mM theophylline in Msmeg ( Figure S1). Also, the activation ratio, defined as the ratio of the reporter gene readout at 2 mM vs. 0 mM theophylline, was similar for the two reporter genes (89612 for b-galactosidase and 6568 for GFP). Msmeg ribo-gfp cells were further characterized by flow cytometry after incubation in 0-4 mM theophylline ( Figure 2B). Each sample comprised a single distribution, and the mean GFP fluorescence intensity increased with theophylline concentration (Table S1).
In Mtb containing ribo-gfp, a similar dose response was observed as in Msmeg (Figure 2A). Growth rates of Mtb were slightly more sensitive to theophylline, with attenuation in Mtb observed at $5 mM ( Figure S1). For the negative controls in the absence of theophylline, a higher fluorescence signal was observed from Mtb than Msmeg, perhaps due to higher scattering from Mtb. While this resulted in a lower calculated activation ratio in Mtb (8.260.84), the overall dose response and maximum fluorescence levels were similar between the two species ( Figure 2A). To confirm that GFP expression in Mtb is fully repressed in the absence of theophylline, anti-GFP immunoblots were performed on whole-cell lysates ( Figure 2C). In the absence of theophylline, no signal above background was detected in cells containing ribogfp or vector only.
The kinetics of both gene expression and repression were measured in GFP fluorescence and immunoblot assays. In both Mtb and Msmeg, maximal GFP expression was observed after two doubling times ( Figure 3A; as in Figure 2A, higher background fluorescence signal was detected for the Mtb vs. Msmeg negative controls). To demonstrate gene repression upon theophylline removal, Msmeg strains were induced for ,1.3 doubling times (4 h) and then exchanged into fresh media with or without 2 mM theophylline. After theophylline removal, GFP fluorescence levels in Msmeg ribo-gfp cells were significantly reduced within five doubling times, whereas cells with theophylline maintained high GFP expression over the same period ( Figure 3B; note that the data in Figures 3A and 2B were acquired on a different instruments, so RFU values cannot be directly compared between the two). These data, particularly the reversibility of theophylline induction, were further confirmed by flow cytometry. After the 4-h initial induction period, two identically induced ribo-gfp cell populations exhibited the same level of GFP expression. Five doubling times after the media exchange, ribo-gfp cells without theophylline were indistinguishable from vector control, whereas continuously induced cells maintained GFP expression ( Figure 3C, Table S1). In Mtb ribo-gfp cells, the effect of theophylline removal was analyzed by immunoblot. Whereas as GFP expression was maintained over 4 days (,4 doubling times) for cells incubated in theophylline, no GFP was detected one and two days after theophylline removal ( Figure 3D).

Theophylline-dependent knockdown of katG in Msmeg
To assess the ability of the riboswitch-promoter combination to control theophylline-dependent knockdown mutants, we targeted Msmeg katG (MSMEG_6384), a homologue of Mtb katG (Rv1908c). The katG gene encodes a catalase-peroxidase that converts the   prodrug isoniazid into its active form [23]. We generated a homologous recombinant strain, RiboS-katG, in which katG is under riboswitch regulation, and confirmed the single crossover event by PCR ( Figure 4A).
For both the wild-type and RiboS-katG strains, the halfmaximum effective concentration of isoniazid (EC 50 ) was determined in a growth assay in 0-5 mM theophylline. RiboS-katG did not express KatG at levels detectable by immunoblot, indicating efficient repression in the absence of theophylline, and therefore exhibited isoniazid resistance compared to the wild type (EC 50 ,10 mg/mL) (Figure 4). In response to theophylline, dosedependent KatG expression and increasing isoniazid sensitivity were observed. Addition of 2-5 mM theophylline induced KatG at levels sufficient to restore the wild-type isoniazid EC 50 .

Theophylline-dependent expression in Mtb in a macrophage infection model
To test the utility of the riboswitch regulatory system in the context of infection, we infected the murine macrophage-like RAW 264.7 cell line with Mtb harboring the ribo-gfp construct. Macrophages were then induced with 0 or 0.5 mM theophylline for one day, fixed, and imaged by fluorescence microscopy. GFP fluorescence was observed from intracellular Mtb containing ribogfp only in the presence of theophylline ( Figure 5).

Discussion
Both the activation ratio and time response of the riboswitchbased system compare favorably with the ,100-fold activation ratios and 2-day induction times reported for the nitrile-inducible and Tn10-derived Tet systems [14,19]. In a direct comparison between the riboswitch and Tet systems in Msmeg, the activation ratio for GFP was equivalent (6963 for the riboswitch vs. 7265 for Tet; data not shown). We further determined that the riboswitch functions as a titratable system, like the Tn10 TetR repressor, rather than as a bistable switch, like the nitrile-inducible system [14]. Also, the effect of theophylline is reversible in both Msmeg and Mtb upon removal of the inducer.
Gene expression in an intracellular pathogen such as Mtb is often regulated in response to changes in the host environment, such as internalization by macrophages [24]. The ability to modulate expression levels during infection is critical to determining how specific genes affect bacterial survival and disease progression in the host. Chromosomal gene knockouts are commonly used to examine gene function or determine gene essentiality, and conditional gene knockdowns afford the additional power of inducing expression or repression at a defined phase of growth or infection. The results from the conditional katG knockdown and macrophage infection experiments show that the riboswitch affords control of mycobacterial gene expression both in vitro and within host macrophages.
We have confirmed that the riboswitch can regulate gene expression in both the model organism Msmeg and the pathogen Mtb, suggesting that the mechanism of riboswitch induction is species-independent and that factors affecting intracellular theophylline concentration-such as membrane penetration, metabolism, and efflux-operate similarly in both species. These data also demonstrate the consistency of riboswitch response to theophylline across a variety of in vitro and cell-based applications that are relevant to the study of Mtb and other mycobacteria. Finally, the similarity in responses shows that Msmeg can serve as a host for screening further iterations of riboswitch-based mycobacterial gene regulation. We anticipate that the construction of promoter-riboswitch libraries and the ability to screen by fluorescence, as demonstrated here, will facilitate the engineering of enhancements such as increased dynamic range, decreased basal expression, and inducible repression. Such screens could be performed not only under standard culture conditions, but also in a macrophage infection model or various in vitro culture models, such as hypoxia [25] and carbon starvation [26], that mimic tuberculosis disease states.
Application of the theophylline-responsive riboswitch system to animal models of infection would be facilitated by the fact that theophylline is an FDA-approved drug and well tolerated in mice and guinea pigs. However, because theophylline is a bronchodilator, its use may complicate Mtb infection studies. Given that the TetR system is the only mycobacterial inducible system currently available for use in animals, the application of theophylline riboswitch in animal models of tuberculosis nevertheless warrants further investigation.

Reporter gene assays for riboswitch-regulated constructs
See Tables S1, S2, and S3 for summaries of all strains, constructs and primers as well as Text S1 for detailed methods used in this study. Riboswitch-reporter plasmids were constructed by assembly PCR methods as previously described [21]. Wholecell GFP fluorescence assays were performed as reported for Msmeg [21]. Briefly, for dose response curves, cultures were grown from early to late log phase (optical density at 600 nm [OD 600 ] of 0.2 or 0.3 to ,1) over two doubling times (6 h) in media containing 0-5 mM theophylline. Emission from whole-cell suspensions was measured at 510 nm with excitation at 450 nm and a 495 nm high-pass cutoff filter in a Gemini XPS fluorescence microplate reader (Molecular Devices Corporation). For Mtb, cells were resuspended and incubated at room temperature for 1 h in 200 mL phosphate-buffered 10% formalin prior to fluorescence measurement.
To measure the kinetics of GFP repression upon theophylline removal, the OD 600 and GFP fluorescence of Msmeg cultures were monitored continuously in a 96-well plate. Briefly, bacteria were inoculated at OD 600 of 0.1 in 300 mL in black, clear-bottom 96well plates in media containing 0 or 2 mM theophylline and incubated without shaking at 37uC. After 4 h, cells were pelleted and exchanged into fresh media with or without theophylline and monitored for an additional 15-17 h. For these assays, fluorescence was measured with a FLUOstar Optima plate reader (BMG Labtech) with 485-nm excitation and 520-nm emission filters (30 flashes per well, constant gain of 1000). All GFP data are reported as relative fluorescence (RFU) normalized by the OD 600 for each sample. b-Galactosidase activity in whole-cell lysates was measured as previously described [27].

Flow cytometry
For each sample, 1-3610 8 Msmeg cells (based on OD 600 of 1 = 3610 8 cells/mL) were washed twice with 1 mL of PBS and resuspended in 1 mL 10% formalin. After sonication in an ice water bath for 2 minutes, cell clumps were pelleted by centrifugation for 10 min at 2006 g, and 900 mL supernatant was removed for analysis with a Becton Dickinson FACScan flow cytometer (Clinical Flow Cytometry Laboratory, Stony Brook University). For each experiment, the vector control sample was used to set a gate based on forward and side scatter channels to select against debris and any remaining cell clumps. Histograms were calculated from approximately 2610 4 cells per sample using Cyflogic software (CyFlo Ltd., Finland).

Immunoblot analysis of Mtb lysates
Mtb strains were inoculated at OD 600 of 0.05 in 30-mL cultures and incubated with shaking at 37uC in medium containing 2 mM theophylline. On day two, a subset of cultures was pelleted and resuspended in medium without theophylline, and all cultures were grown for an additional two days. On each day of the experiment, samples were removed for OD 600 measurements and to obtain lysates for immunoblot analysis. Cells were killed by boiling and lysed by bead-beating (265k rpm for 30 s each). After removing cell debris by centrifugation, supernatants were stored at 280uC until further analysis.

Construction and characterization of Msmeg with katG under riboswitch control
A homologous recombinant Msmeg strain was generated as described using pRiboS-katG, a suicide vector containing a segment of the Msmeg katG gene MSMEG_6384 [28]. Homologous recombination at the katG locus was verified by PCR on genomic DNA isolated from a single clone. The half-maximum effective concentration of isoniazid (EC 50 ) was determined in a growth assay at each theophylline concentration. Expression of KatG in Msmeg wild-type and RiboS-katG strains was confirmed in whole-

Macrophage infection and microscopy
RAW 264.7 cells (ATCC TIB-71) were incubated in an Mtb suspension at an MOI of 5 for 4 h and washed with PBS. Infected macrophages were allowed to recover in medium until replacement at 24 h post infection with fresh medium containing 0 or 0.5 mM theophylline. After an additional day, macrophages were washed with PBS and fixed in phosphate-buffered 10% formalin for 1 h. Coverslips were mounted on glass slides with VectaShield mounting medium plus DAPI stain (Vector Laboratories). Image stacks were acquired on a Zeiss Axiovert 200 M using a 10061.3 numerical aperture lens. Following digital deconvolution using the nearest-neighbors algorithm in Slidebook (Intelligent Imaging Solutions), final images were generated by z-projection of 30-40 frames at 0.34 mm separation. Text S1 Detailed protocols for DNA constructs and assays.

Supporting Information
(DOC)