Unravelling the Secrets of Mycobacterial Cidality through the Lens of Antisense

One of the major impediments in anti-tubercular drug discovery is the lack of a robust grammar that governs the in-vitro to the in-vivo translation of efficacy. Mycobacterium tuberculosis (Mtb) is capable of growing both extracellular as well as intracellular; encountering various hostile conditions like acidic milieu, free radicals, starvation, oxygen deprivation, and immune effector mechanisms. Unique survival strategies of Mtb have prompted researchers to develop in-vitro equivalents to simulate in-vivo physiologies and exploited to find efficacious inhibitors against various phenotypes. Conventionally, the inhibitors are screened on Mtb under the conditions that are unrelated to the in-vivo disease environments. The present study was aimed to (1). Investigate cidality of Mtb targets using a non-chemical inhibitor antisense-RNA (AS-RNA) under in-vivo simulated in-vitro conditions.(2). Confirm the cidality of the targets under in-vivo in experimental tuberculosis. (3). Correlate in-vitro vs. in-vivo cidality data to identify the in-vitro condition that best predicts in-vivo cidality potential of the targets. Using cidality as a metric for efficacy, and AS-RNA as a target-specific inhibitor, we delineated the cidality potential of five target genes under six different physiological conditions (replicating, hypoxia, low pH, nutrient starvation, nitrogen depletion, and nitric oxide).In-vitro cidality confirmed in experimental tuberculosis in BALB/c mice using the AS-RNA allowed us to identify cidal targets in the rank order of rpoB>aroK>ppk>rpoC>ilvB. RpoB was used as the cidality control. In-vitro and in-vivo studies feature aroK (encoding shikimate kinase) as an in-vivo mycobactericidal target suitable for anti-TB drug discovery. In-vitro to in-vivo cidality correlations suggested the low pH (R = 0.9856) in-vitro model as best predictor of in-vivo cidality; however, similar correlation studies in pathologically relevant (Kramnik) mice are warranted. In the acute infection phase for the high fidelity translation, the compound efficacy may also be evaluated in the low pH, in addition to the standard replication condition.

Introduction smegmatis mc 2 155, and Mycobacterium tuberculosis H37Rv ATCC 27294 were used for this study. We rationally selected five target genes for this study from the list of Sassetti's classification on essential targets [11] i.e. rpoB, rpoC, aroK, ilvB, and ppk. The recombinants of these target genes for AS-RNA were generated by cloning full-length genes (sequences were taken from KEGG and Tuberculist databases) of these in the reverse/ antisense orientation into the vector pAZI9018b by replacing the lacZ gene [8].Three different independent transformations were performed for the target AS constructs. The AS recombinants of Mtb were selected from 7H10 agar plates supplemented with 50μg/ml Hygromycin (Hyg50) and were grown in 7H9 broth containing Hyg50. The O.D.600 nm was adjusted to 0.1, and the cells were induced at 10 μM IPTG. Transformants showed a slower growth rate, so wherever required; the suspensions were concentrated to match the required O.D. (O.D.600nm to 0.1, approximately 10 7 cells/ml).

Target selection for in-vivo validation
Though there are about 600 in-vitro essential genes in Mtb [11], we made a particular choice by selecting a few key target genes from different pathways: 1). Transcription: RpoB (Rv0667) and rpoC(Rv0668), encoding β and β'-subunits of RNA polymerase. RpoB is a well known clinically validated target, hence was also used as a bactericidal target control for this study. Its partner rpoC which is involved in the same process of transcription was selected to compare if it has an equal partnership and cidality potential. 2). Fundamental amino acid (aa) biosynthesis pathways: We chose two targets, aroK, and ilvB, one each from two amino acid biosynthesis pathways; aromatic (shikimate) and branched chain aa pathways. Both the targets aroK and ilvB, being in-vitro essential and have no human homologs, present a great opportunity towards the development of non-toxic drugs [15][16][17][18]. AroK (Rv2539c), from aromatic (shikimate) aa pathway, encodes the 5 th enzyme shikimate kinase, phosphorylating shikimic acid to shikimate-3-phosphate during chorismate biosynthesis. IlvB (ilvB1, Rv3003c)from branchedchain aa pathway, encodes Acetohydroxyacid synthase, (AHAS), the 1 st and most important enzyme [19] out of the several genes (ilvB1, ilvB2, ilvG, and ilvX). Among all the genes in shikimate pathway, aroK is the only gene which is exclusive and in-vitro essential in Mtb. The essential amino acids can conditionally modulate bacterial growth [8,15,16], hence may act as regulatory tools to delineate their essentiality. Mtb exists in multiple milieu in-vivo [9], so, the targets of amino acid biosynthesis pathway were chosen as suitable tools for concept validation under various physiologies encountered in-vivo. In our earlier report [8], we have demonstrated in-vitro cidality of ilvB under replicating growth condition [8,12]. It is regulatable with the addition of physiological concentrations of ILV and P (Isoleucine, Leucine, Valine and Pantothenate) showing auxotrophy [20,21]. 3) The central energy metabolism: Ppk (Rv2984)or polyphosphate kinase plays an important regulatory role in the transition of the bacteria to the persistence phase under the growth-limiting conditions (phosphate depletion, amino acid starvation, or osmotic stress). It tries to accumulate polyP intracellularly, modulating several bacterial processes (protein synthesis, nucleotide balance, lipid metabolism, energy utility, and susceptibility to antibiotics) [12]. Ppk was chosen as an unvalidated target for this study because this target gene has not been characterised as per Sassetti's list [11]. It is neither reported as essential (in-vitro or in-vivo) nor a non-essential target [11]. Earlier, we had established a partial validation of ppk as essential and cidal target under in-vitro replication condition only [12]. Hence, we chose ppk for this study as an un-validated target gene for further characterisation of its cidality under in-vivo condition.
Other than the clinically validated target from transcription subunit rpoB, rest of the targets ppk, aroK, rpoC, and ilvB are unvalidated targets since their in-vivo essentiality in Mtb is not reported yet. Therefore, in order to delineate in-vivo bactericidal potential of these targets in the present study, we used IPTG (Isopropyl β-D-1-thiogalactopyranoside) inducible AS-RNA silencing [8] as a gene-specific inhibitor for the chosen targets. It was used under multiple physiologies of granuloma simulated in-vitro (replicating conditions, and, under the stressful condition of nitrosative free radicals, acidic, hypoxic, or under starvation of C and N2) as well as under in-vivo conditions (experimental TB of mice).
In-vitro cidality of Mtb targets under replicating conditions using survival kinetics after AS-RNA inhibition The antisense effect was estimated in triplicate from 3 independent transformants for AS recombinants of rpoB, rpoC, and aroK in Mtb. The conditional AS-gene-silencing was induced by 10 or 100μM [8] IPTG under an in-vitro replicating growth condition. The survivors (from AS-silencing) were enumerated by plating on different generation times (Day 0, 1, 7, 14, 21, up to 63 days with a gap of 1 week), and the data was analysed using Prism software (Graph Pad Software, Inc., San Diego, Calif.). Appropriate controls were used. Two earlier established cidal targets (ppk, ilvB) for AS-repression under only the in-vitro 'replicating growth conditions' [8,12]; were used as positive controls in this entire replication condition AS-silencing study. The clinically validated target, rpoB, was another cidal positive control throughout for all the validations. The negative controls for antisense repression were the WT = wild type Mtb; and the V = empty vector in Mtb (also the control for the recombinants). The genes aroK, ilvB, and rpoC were selected as un-validated target genes along with another uncharacterised target gene ppk.

In-vitro cidality of Mtb targets under different physiologies
A total of seven different Mtb strains, i.e. five AS-recombinants (rpoB, rpoC, aroK, ilvB, ppk) and two control strains (WT and vector in Mtb) were taken for this study. We had to profile the survival kinetics of all the selected seven Mtb strains/ AS-recombinants, in triplicate, under six different assay conditions, (Platings/day = 7strainsX 3plicateX 6assaysX 5dil.sX 9time points) by plating 5 dilutions for cfu, up-to 35 days, under the constraints of bio-safety level-3 containment facility. A throughput method was ideal to conduct and compare all the experiments with replicates in parallel. Hence, a validated throughput SPOT-MBC assay [22] was used for this study because of its efficient way of testing a large number of samples/replicates from various in-vitro models, to be explored in parallel under BSL3 containment. Moreover, this relatively straightforward throughput method has been validated for different phenotypes of Mtb (AS-recombinants, vector controls in Mtb as well as WT, sensitive or resistant Mtb) as well; and had yielded cfu counts similar to conventional plating method [22]. The standard conventional plating method would have given an identical output; hence, it was prudent to use this efficient SPOT-MBC assay system. The ppk target was validated earlier [22] with SPOT-MBC and hence was taken as a positive control for this validation. The rest of the selected uncharacterised targets (rpoB, rpoC, aroK, ilvB) were validated in this study by conventional vs. SPOT-MBC methods in parallel under in-vitro replicating growth conditions, before using it for various physiological screens.
We further investigated the vulnerability of selected genes by checking cidality under all the in-vivo simulated multiple physiological conditions (the equivalent of macrophages and the lung granuloma) by performing survival kinetics in-vitro. All the strains were tested under six different physiological in-vitro assay conditions for investigations in parallel, in triplicate: replicating, hypoxia [7], nutrient starvation [5], low pH [2], nitric oxide stress [3], and nitrogen starvation [6]. The survivors were monitored by enumerating up to 35 days in a kinetic manner from all the screens in parallel using SPOT-MBC assay [22]. Since our previously reported ppk AS-repression (SPOT vs. conventional assay) experiments were performed only under proliferating assay conditions [12,22], and no other alternate physiological environments; we, therefore, investigated ppk AS-repression under all other the physiological conditions as well.
In-vivo cidality of Mtb targets after AS-RNA repression All the seven Mtb strains, including the controls and AS recombinants, were investigated for their growth or survival kinetics under in-vivo conditions in BALB/c mice, for confirmation of in-vitro cidality. This important in-vivo evaluation provided us with the proof of concept on cidality of targets under experimental TB in mice, the actual disease condition.
Ethics statement. The in-vivo study was carried out in strict accordance with the recommendations of the Institutional Animal Ethics Committee (IAEC), registered with the Committee for the Purpose of Control and Supervision (CPCSEA), Government of India (registration no. CPCSEA1999/5). All the protocols for animal experimentation and animal usage were reviewed and approved in advance by the IAEC. Carbon dioxide (CO 2 ) was used for euthanasia.
Animals. Male BALB/c mice were purchased from RCC Hyderabad, India. Mice (6-8weeks) with an average body weight of 20-25 grams were randomly assigned to groups of three per cage and were allowed two weeks of acclimatisation before experimentation. The animals were housed under standard conditions in the facility with a 12 hr. day-night cycle. The infected mice were maintained in individually ventilated cages (Allentown Technologies, USA) in bio-safety level 3 (BSL-3) facilities. All the procedures including dosing of the infected mice were performed under strict BSL-3 bio-containment guidelines. Feed (the sterile commercial diet) and water were given ad libitum. We provided IPTG to mice in their drinking water at a 5mM concentration, as per the earlier established methods [23,24], for in-vivo target induction/ repression. The animals were euthanized with CO 2 at the respective time points, and the lungs were sterically removed, homogenised and were plated for cfu enumeration of survivor Mtb bacilli.
Course of Infection (COI) studies in mouse-tuberculosis infection model. BALB/c mice were infected in the bio-safety level 3 (BSL3) facility, via inhalation in an aerosol infection chamber [25] with the suitable modification that delivered~10 4 bacilli/mouse lung, a higher dose of infection [26]. This method was standardised and validated in our laboratory [25,26]. The groups of mice (n = 3) were infected with different Mtb strains (WT and AS-recombinants of Mtb). High dose bacterial infection to lungs was delivered by increasing the strength of the bacterial inoculum (10 9 cfu/ml) used for inhalation [26] that delivered 10 4 cfu/lung, to allow evaluation of cfu reduction by capturing enough window following AS-repression. Instillation of 10 4 cfu to lung/mouse was enumerated by harvesting and plating the lungs at day 3 post infection. The course of infection (COI) of all the strains was monitored on 3, 7, 14, 28, 42, and 56 days post infection by cfu enumeration.
Rifampicin was used at 30mg/kg bw as a reference drug control (for the WT Mtb) formulated in CMC (Carboxymethyl Cellulose) suspension [25]. Drug treatment with rifampicin started after 3 days following the establishment of acute infection. The drug was administered by oral gavage. AS-repression was induced with IPTG, given as 5mM in drinking water ad libitum beginning on day 3, based on the previously validated and established method [23,24]. The water bottles containing IPTG were replaced every 48 hrs. At every time point, the mice were euthanized with CO2, and the lungs were harvested and homogenized [25] in PBS containing 0.1% bovine gelatine and 0.1% triton-X100 using tissue grinders (W012576; Wheaton). Each suspension was serially diluted in 10-fold steps and plated on Middlebrook 7H11 agar supplemented with 10% albumin-dextrose-catalase. Plates were incubated at 37°C with 5% CO2 for 3 weeks, and colony forming unit (CFU) counts were enumerated on 3, 7, 14, 28, 42, and 56 days.
Transcriptome analysis of in-vivo antisense repression of target genes. Lung homogenates of AS-recombinants of Mtb along with the WT and vector control cultures were collected from different time points (day 3,7,14,28,42,56), and pelleted. One ml of TrizolH was added to the cell pellets (from 2 ml homogenate/s) to stabilize and arrest the mRNA. These samples were flash frozen in dry ice and stored at -70°C until further processing. Wild-type (WT) M. tuberculosis culture control as well as the rifampicin-treated homogenates were also included. While processing, pellets were thawed on ice, cells were disrupted by bead beating using 0.1 mm diameter zirconium beads (Biospec), followed by a 5 min centrifugation at 14,000 g. Total RNA was isolated, and qRTPCR was performed using SYBR Green chemistry (Brilliant II SYBR1 Green QPCR Master Mix) in the Mx3005P Stratagene system as reported [8,12] using respective forward and reverse primers (Table A in S1 File). The transcript levels were measured against SigA (Rv2703) as a validated house-keeping control gene, from our previous study of AS-repression [12], as well as taking clues from other reports including a detailed analysis by Manganelli et al. [27,28]. The fold difference was calculated against vector control using Delta-delta Ct method [29].

Statistical Analysis
The data from in-vitro as well as in-vivo experimentation was analysed using Prism software (Graph Pad Software, Inc., San Diego, Calif.) for all the calculations for data. In-vitro data included survival plots as well as the correlation of cfu/ml from the SPOT-MBC vs. conventional plating method. The in-vivo data included the COI of various Mtb strains tested after AS-repression in-vivo. Two-way analysis of variance (ANOVA) with Bonferroni post hoc test correction was used to compare the net bacterial loads at each time point using Prism software (Graph Pad Software, Inc., San Diego, Calif.), to check the significance of AS repression for invivo cidality. The correlation between the outcomes from various in-vitro screens vs. in-vivo studies was analysed to narrow down to the best in-vitro screen, the predictor of in-vivo cidality.

Throughput cfu enumeration assay for AS-silencing in-vitro
Validation data of all the AS-recombinants by the throughput "SPOT-MBC assay" vs. conventional [22] plating for cfu enumeration (S1 Fig) demonstrated an overlapping data with a strong positive correlation (Pearson's): rpoB (R 2 = 0.5857), rpoC (R 2 = 0.7770), and aroK (R 2 = 0.9863) vs. vector controls. However, in the case of ilvB, we found a variation in the Emax (SPOT = 3.3 log 10 and conventional = 5.2 log 10 reduction), probably, the smaller surface area of the agar media in SPOT-assay, allowed the colonies to grow in proximity. It enabled socializing through the exchange of the required amino-acids between the healthy and the sick or nearly dead bacilli, helping them to rejuvenate. Overall, the SPOT-MBC assay was aptly and significantly suitable (EEE) [22] for kinetic profiling of Mtb AS-recombinants and controls under multiple tests. Henceforth, spot assay was used for cfu enumeration for all the in-vitro screens.

AS-silencing under physiological stress conditions in-vitro
AS-driven survival kinetics, under various alternate physiological in-vitro conditions (Fig 2), was a yardstick to understand the overall importance (cidality) of respective targets. As expected, the WT and V controls exhibited condition-specific, normal growth under all the physiologies. Since ppk was reported cidal under proliferating assay conditions [13,22]; in this study, under different physiological conditions, ppk exhibited the maximum total AS-repression among the tested. Whereas, under the replication condition alone, ilvB had demonstrated Mtb enters the host via inhalation. The first encounter in-vivo is with the immune cells: the macrophages which phagocytose and attack Mtb with their ammunition like low pH, hydrolases, free-radicals, etc. This first encounter is unanimously responded by all the hosts irrespective of their immune status. Beyond this point, the countering of the pathogen is host population specific. The outcome as TB or no-TB is a delicate balance and is outcome result of the battles between the pathogen and the immuno-competence of the host to outsmart. Further, it progresses into a larger immune structure: the granuloma. During the process, the other stresses in the granuloma, especially in the caseating/ necrotizing lesions are the gradually decreasing levels of O 2 , N, C, etc. representing hypoxia, the arrest of various biosynthesis processes and poor nutrition nearly close to starvation. If the host wins, the pathogen is contained in the solid granuloma, which may gradually heal with time. But in case the pathogen Mtb overpowers the immune pressure, the granuloma progresses as a caseating and necrotising granuloma. In this case Mtb bacilli multiply to finally break open from granuloma, and disseminate to other organs of the body or come out into the bronchi and get coughed out to infect other hosts. We attempted to simulate some of these stress conditions in the form of various in-vitro screens to identify the ideal in-vitro screen/ condition. maximum cidality. The other targets were also significantly (P<0.001) silenced with different cidality magnitudes: ppk(7.5)>ilvB(7.3)>rpoB(7.1)>rpoC(5.2)>aroK(3.2) (Fig 3). The stringent conditions where most of the targets were cidal, showed a rank-order for the starvation of N>C>O>pH. Surprisingly, the genes rpoB, rpoC, aroK, and ppk were significantly repressed upon the exposure to acid. However, starvation data revealed that Nitrogen depletion (Msx) accelerated "silencing of most of the targets in-vitro" followed by nitric oxide (NO) and low pH (LpH) pressures as compared to the replicating (REP) growth conditions.

Cidality SCORE
TB is a complex disease where Mtb can efficiently demonstrate multiple dimensions of pathogenesis under paradoxical milieus of the host. We asked the question if the total compounded sum of all these stringent pressures on a target would show a potential link between in-vivo persistence; as it would behave under multiple milieus in a diseased situation? Therefore, we calculated a 'compounded potential' of the chosen targets. The SCOREs were assigned based on the extent of AS-repression of targets, representing the net cidality of a target from all the replicating/ non-replicating screens (Table B in S1 File, Fig 4) whether negative (growth) or positive (kill). The cidality SCORE graphs were plotted (Fig 4), demonstrating the total potential (vulnerability) in the following order of target cidality (log 10 cfu reduction): ppk(7.5)>ilvB (7.3)>rpoB(7.1)>rpoC(5.2)>aroK(3.2). The resulting numeral indicated the "cidality SCORE" and a measure of cidality potential of the respective targets. The entire data was statistically significant (Dunnett's multiple comparisons) vs. the controls. However, in the case of ilvB, an attenuation of the target was observed due to the revival of additional colonies upon extended incubation (up to 45 days).
The AS-repression under regular and stringent growth demonstrated that the in-vitro cidal targets showed maximum SCOREs up to 7.5 (log 10 cfu reduction), compared to the controls  (Fig 4). This outcome of the in-vitro alternate physiological conditions established specifically few targets (rpoB, aroK, ilvB) cidal under all the conditions (Fig 4). Our next objective step was to validate cidality of these targets under the in-vivo experimental model of tuberculosis.

In-vivo AS-silencing of in-vitro cidal targets in mice
Validation of the SCORE based in-vitro cidality was confirmed in-vivo by infecting BALB/c mice with the respective Mtb strains (AS-recombinants and WT and vector controls) via an aerosol route [25,26]. In-vivo studies were approved by IAEC (CPCSEA), as mentioned in the ethics statement. The in-vivo AS-silencing was achieved by providing IPTG (5mM) to mice in drinking water [23,24], a non-invasive technique of IPTG delivery in drinking water; thereby minimising the handling of infected mice in the bio-containment set-up.
In-vivo cidality of the targets on 28 th day The first visible and quantitative in-vivo trends of cidality appeared in the lungs after four weeks post-infection. The control Mtb strains grew as expected, with fully developed visible granulomas in the lungs (Fig 5), whereas the AS-recombinants showed reduced growth on the 28 th day, due to target specific AS-repression and cidality. RpoB showed the maximum cidality in-vivo (3.9 log 10 reduction), despite only 2.5 log 10 cfu reduction in the replicating in-vitro model. The culture plates were incubated up to 45 days, and the colony characteristics were observed. None of the AS-recombinants, except ilvB, showed any signs of re-growth or delayed appearance of additional colonies on the media plates. The colony counts remained constant upon further incubation; suggesting cidality due to AS-repression in ppk, rpoB, rpoC, and aroK. On the contrary, in the case of ilvB, the appearance of additional colonies on extended incubation point towards attenuation rather than cidality. IlvB showed a negligible cidality (0.36 log 10 cfu reduction) under the in-vivo condition on day 28 and exhibited granuloma formation similar to the control strains. Targets like ppk, rpoC and aroK were repressed by their respective AS-counterparts and showed significant (two-way ANOVA, ÃÃ P = 0.0086 to Ã P = 0.0402), but varying levels of log 10 cfu reduction: rpoB(2.5)> aroK(1.9)> ppk(1.2)> rpoC (1.5). The in-vivo model was validated by including a positive treatment control; using the rifampicin treatment (30mg/kg) of WT Mtb that demonstrated~3.8 log 10 cfu reduction. This drug treatment control was a non-AS-RNA based inhibition control.
In-vivo cidality of targets on 56 th day The bacterial reduction in the lungs clearly delineated cidal targets from the static ones. The rank order of log 10 cfu reduction magnitude was: rpoB(3.9)> aroK(2.4)> ppk(1.6)> rpoC(1.59); significantly different ( ÃÃÃ P = <0.0001 to 0.0004) when compared to the WT. The vector control Mtb demonstrated no difference from the WT strain (ns, P = 0.6283) (Fig 5). In contrast, ilvB ( ÃÃ P = 0.0012, two-way ANOVA) behaved as a static-to no-effect target under the in-vivo conditions. It demonstrated a marginal cfu reduction (0.36 log 10 cfu reduction) vs. the control Mtb. The treatment (IPTG and Rifampicin) was initiated 3-day post-infection (10 4 cfu/Lung), to substantiate and compare the killing kinetics of AS induction vs. the drug treatment. It was "more reflective of treating individuals who are recently infected" and not the chronic TB infection.
Pharmacokinetics (PK) of AS-RNA. We evaluated the levels of AS-RNA generated by the Mtb-recombinants (during the span of infection in mice), indirectly, by measuring the net fold target repression, representing bio-availability of AS-RNA. It was estimated by RTPCR (Table C in S1 File, Fig 6) of transcriptome from Mtb strains recovered from the infected lung homogenates. Each target was repressed, though with different magnitudes (13-to 103-fold), proving that Mtb AS-RNA had no PK issues. Although the order of cidality was rpoB>aroK>ppk>rpoC>ilvB; the transcript down-regulation rank order was: ppk>rpoC>aroK>ilvB>rpoB (Fig 6). The Rifampicin treated WT Mtb transcript showed a mere~3-fold down-regulation of the target gene (rpoB, maximum 2.6 fold) in-vivo, deficient enough to show cidality of 3.8 log 10 cfu reduction. Maximum cidality demonstrated by rpoB had the minimum fold transcript down-regulation, hence vulnerable. The maximum transcript was down-regulated in ppk that showed the least cidality. The safety and bioavailability of IPTG are reported earlier [23].

Correlation of in-vitro score vs. the in-vivo cidality
The in-vitro replicating/non-replicating data had a varied cidality/gene repression response under in-vivo conditions with different magnitudes. Most of the 'in-vitro cidal' targets (4 out of 5; 80%) demonstrated cidality under in-vivo condition. Cidality data from all the in-vitro assay conditions vs. the in-vivo efficacy data, showed a significant correlation (Spearman R = 0.9856) with the low pH in-vitro assay condition (Fig 7), establishing that the "low pH in-vitro screen" optimally simulated the in-vivo disease condition to select in-vivo cidal inhibitors.

MIC of anti-TB drugs under low pH
Most of the frontline drugs had MIC on pH adapted Mtb cultures under low pH (6.2) condition (Table D in S1 File) except isoniazid.

Discussion
The slow-growing intracellular Mtb remains focused on capturing the fortress of lungs. Mtb strategically evolves new sets of artillery (combination of genes) during the phases of replication and persistence. It implies that the conditionally expressed targets required in the in-vivo phase during the disease might be the relevant ones under all the host-physiologies to achieve  cidality. It prompted us to identify the targets that are bactericidal under all in-vivo physiologies, like rpoB, the target of rifampicin.
In this study, we investigated various in-vivo simulated physiologies of the granuloma-like environment, under in-vitro conditions. Using AS-RNA-silencing as a cidality tool (genetic inhibitor), we examined five rationally selected targets cidal under all the in-vitro screens and correlated the outcome from in-vitro versus the in-vivo data. Our data demonstrated rpoB and aroK as the best mycobactericidal targets under all the physiologically relevant conditions. RpoB was used as a clinically proven cidal target control.
AroK(encoding shikimate kinase) emerged as an in-vivo validated Mtb-cidal and vulnerable target under all the physiologically relevant conditions, suggesting its potential for TB-drug discovery. Shikimate pathway has been previously reported as in-vivo essential only for aroA and aroC in the ESKAPE pathogen, Acinetobacter baumannii [30]. Few pathogens like E. coli, Salmonella typhimurium, and Yersinia pestis have one extra shikimate kinase gene isoform (aroL), which is catalytically superior, better regulated and plays a pivotal role [30][31][32]. On the contrary, no clarity on the physiological function of aroK suggests its non-essentiality in these pathogens [33]. On the contrary, in M. tuberculosis shikimate kinase is unique and exclusive enzyme essential for its survival; as is the finding reported in another intracellular pathogen  Table B in S1 File. Helicobacter pylori [15,16,34]. It presents an excellent opportunity for exploring aroK target for TB drug discovery. The in-vivo essentiality of aroK in Mtb has not been reported yet. However, in the case of Acinetobacter baumannii, a very recent report has confirmed aroK as an in-vivo essential enzyme [30,32].
In the present study, Mtb AroK performed consistently well across all the simulated physiologies in-vitro (cidality SCORE = 3.2) as well asunder in-vivo condition (2.4 log 10 cfu reduction;39.7-fold transcript repression in-vivo, Figs 1 and 3-5) despite a comparatively lower invitro cidality SCORE, thus, delineating its cidality. Blocking aroK target kills Mtb under invivo, clearly indicating nearly zero availability of amino acids (tryptophan, tyrosine, and phenylalanine) in the in-vivo milieu. However, whether it will translate into a therapeutically valid target for clinical usage is worth further investigations.
Our studies confirmed rpoB as the most vulnerable target with a mere 13-fold transcript repression in-vivo (Table C in S1 File) translating into an excellent in-vivo cidality (3.9 log 10 cfu reduction). Emerging from a good in-vitro cidality SCORE (7.1, Fig 4) with cidality under low pH, it established that inhibiting rpoB with AS-therapeutics (unique MOA of blocking mRNA specifically) is equipotent to rifampicin treatment, validating our experimental approach. However, it's another RNAP partner, the rpoC, demonstrated a lower "cidality SCORE" (and a negligible cidality in low pH) compared to rpoB (5.2vs. 7.1, Fig 4), though marginally better cidality under replicating condition than rpoB (Fig 1). Despite a high (43-fold) in-vivo transcript repression for rpoC, the cidality SCORE could, in fact, delineate rpoC as just a marginally in-vivo bactericidal target (1.6 log 10 cfu reduction) as compared to rpoB (Table C in S1 File, Fig 5).
Targets, ilvB, and ppk were "highly in-vitro bactericidal" (Figs 1 and 4), but comparatively less cidal targets under in-vivo condition ( Fig 5); despite their best repressed in-vivo transcript levels (ilvB = 31.1-fold, ppk = 103.6-fold; Table C in S1 File, Fig 6). One of the best in-vitrocidal targets, ilvB, demonstrated negligible in-vivo cidality (0.36 log 10 cfu reduction, Fig 5). This outcome is attributable to its auxotrophic nature [8,35] and suggests compensation of the effect by the availability of trace amino acids (isoleucine, leucine, valine, and pantothenate) under in-vivo milieu. It is hard to delineate attenuation and cidality, but probably ilvB flaunted target attenuation under the in-vitro condition as well, because of the revival of additional colonies upon extended incubation. The presence of trace amino acids (ILVP) nullified its cidality potential [8] leading to attenuation. Other targets studied (ppk, rpoB, rpoC, aroK), did not demonstrate this phenomenon, suggesting their bactericidal nature.
Next best in-vitro cidal target ppk (Figs 1 and 4), was marginally in-vivo-cidal (1.6 log 10 cfu reduction, Fig 5) despite its maximum in-vivo transcript repression (103.6-fold; Table C in S1 File, Fig 6). Since ppk is proven to be a stationary phase specific target [12], a long-term infection model in mice may be required to demonstrate its cidality. The in-vivo model used for the present study was a hybrid model encompassing from acute to chronic infection states. We intentionally infected mice with high dose (10 4 cfu/lung) to facilitate a measurable window for cfu enumeration, whether the bacterial numbers increase (up to 10 8 cfu/lung) or reduce. To begin with, this mouse infection model is an acute model harbouring replicating bacilli. But with time (beyond four weeks) bacterial growth of WT control and the recombinant strains slows down, it acquires more of a nonreplicating state in-vivo while the AS effect is still on. Hence, the later part of the course of infection represents the AS-effect on a chronic disease. Since AS-repression based cidality was observed beyond 4 weeks as well (Fig 5), it suggested a possibility of killing Mtb even during the chronic phase. However, AS-based cidality studies in a chronic mouse infection model are warranted. Upon a continuous expression of AS-RNA under IPTG induction, the respective targets were being silenced during the course of infection in mice from acute to chronic; hence, demonstrating a stage-specific cidality effect, if the target was essential under the respective conditions.
Interestingly during 4-8 weeks period (chronic stage of infection) in particular, a further growth of ppk recombinant was significantly inhibited (ns, P = 0.1080, one-way ANOVA, Dunnett's multiple comparisons) by the antisense. Whether the bacterial count will reduce further beyond 8-weeks, needs to be determined. However, we need to delayer it separately, with careful investigations in a long term course-of-infection model beyond 8-weeks.
The cidality SCORE correlated well with the in-vivo data outcome in 4 out of 5 targets (80%) confirming in-vivo-cidality. AroK target, which showed a moderate in-vitro cidality under replicating condition, would have been missed out if had not been checked under alternate physiological conditions as well. It outlines the importance of cidality-SCORE. A consistent performance of aroK under all the physiological conditions tested demonstrated a cidality SCORE of 3.2, and revealed its cidality potential; and hence, selected for in-vivo validation. It emerged as a bactericidal target under in-vivo condition also, probably because it was cidal under all the physiologies studied. RpoB was far more superior to rpoC. The cidality-SCORE approach was a worthwhile investigation, implying that a target is preferred if it is cidal under all the physiologies encountered in the host. It was apparently confirmed by rpoB and aroK, but not by rpoC. The cidality SCORE could intricately delineate the in-vivo bactericidal potential of rpoB and rpoC targets. The only target which did not show in-vitro to in-vivo correlation was ilvB, because of other issues like auxotrophy and attenuation observed in this study. Overall, the in-vitro cidality SCORE could predict the in-vivo cidality of the targets, irrespective of their magnitudes. Since frontloading of all alternate physiological screens in parallel may not be feasible; we questioned which screen out of the six conditions tested is the best in-vivo cidality predictor?
Antimycobacterials are tested under the replicating growth conditions in 7H9 media [8,36,37], and confer limited translation of in-vitro cidality into the in-vivo efficacy, thus resulting in high attrition rate in drug discovery. This physiology of 'in-vitro replication' growth condition does not represent the 'actual microenvironment encountered under host immune pressures' to a large extent, thus overlooking the selection of potential in-vivo cidal compounds. As a consequence, inappropriate compounds get selected. During the paradoxical intra-macrophagic phase, the acidification renders Mtb almost inactive, even before it understands the in-vivo milieu to adapt in [38]. The activation-specific pH of macrophage (~6.2), further drops upon phagosome-lysosomal (P-L) fusion to~4.5 within 15 to 60 minutes under various immune pressures [39]. Despite this acidified external milieu of macrophage, the internal pH of Mtb (~7) remains neutral [40]. In order to refrain from the P-L fusion (Fig 2), and to retain itself within the macrophage to progress further, Mtb invariably fools the system by several mechanisms of interference by inducing virulence proteins and efficiently expressing macrophage-specific survival genes [39,41]. The stage-specific transcriptome levels in macrophage [42] fluctuate with the physiological transitions of replication and death rate (Table C in S1 File) in the host. The macrophage attack on Mtb is the 1st battle to be won by the host (Fig 2) representing the most significant universal host-pathogen interaction. It highlights the potential link of low pH to the intraphagosomal survival of Mtb [42][43][44][45], urging researchers to simulate this 'in-host' stage under a single in-vitro screen to select bactericidal agents [43,44]. We conducted comprehensive studies on in-vivo simulated microenvironments under in-vitro conditions and their correlation versus in-vivo cidality data. This correlation from a limited set of five genes investigated in the acute infection in BALB/c mouse point to low pH as probably the most unfavourable microenvironment (R = 0.9856) encountered by Mtb, among the various immune pressures (low pH, hypoxia, nutrient starvation, free radicals, etc.). However, similar studies using a larger set of genes under all the in-vitro physiologies as well as more relevant mouse models with human-like pathology may unleash these cidality correlations better.
Target aroK, like rpoB, the best in-vivo-cidal target, demonstrated a predominant cidality in the low pH screen (Table B in S1 File). It was interesting to discover that a seldom used, low pH screen, correlating best with in-vivo confirmation of cidal targets (rpoB and aroK); can statistically (R = 0.9856) predict in-vivo cidality (Fig 7). Does it mean that the replicating condition (7H9 media) and the rest of the in-vitro conditions are obsolete? May be/ maybe not. These correlations need to be verified in other mice strains like Kramnik that endows human like lung lesions and chronic disease.
Under the starvation of essential nutrients (Carbon, Nitrogen) in the granuloma, the nonreplicating Mtb undergoes a global metabolic shift for energy conservation, shutting down some of its pathways, re-routing Carbon flow from central fatty acid metabolism to lipid and glutamate biosynthesis. Nitrogen is required in both replicating and non-replicating Mtb for biosynthesis of amino acids, nucleotides, organic cofactors to control critical molecular events (asparagine hydrolysis, ammonia release, pH buffering, growth) in the acidic environment, with α-ketoglutarate/glutamate as the key 'nodal-point of pathways' [46][47][48][49][50]. In this study, most of the targets were cidal under in-vitro Nitrogen depletion condition (R = 0.4928), that enhances acid stress in granuloma; and restores the activity of lysosomal hydrolases to kill Mtb [46][47][48][49][50]. Thus, a "chicken or egg" conundrum, what triggers first? Are this Nitrogen depletion and low pH conditions, part of a vicious cycle; or, operate in tandem metabolically, along with other hostile conditions? Whatsoever, it requires a careful investigation and needs to be teased apart meticulously. However, difficult to crack, but it will be exciting and path-breaking.
The low pH, is probably subsidiary to another useful host defence mechanism, nitric oxide (NO, R = 0.5508), attacking Mtb with free radicals from innate immune cells; as the iNOS-/-(inducible NO synthase-deficient) mice, are reported to succumb to Mtb infection [4]. However, the cidality data from hypoxia or replicating growth conditions failed to correlate with invivo cidality outcome from BALB/c mouse model studies (Fig 7). Alternatively, it justifies the use of an appropriate in-vivo model that exhibits hypoxic lesions (not formed in BALB/c mouse) for better correlations.
Mtb infection in the BALB/c mouse leads to solid-granulomas harbouring primarily intracellular bacteria, but these lesions are not hypoxic and do not best represent lung pathology of human-TB [51]. Whereas, in the case of the human host, where mixes of both intracellular, as well as extracellular Mtb populations, exist, other conditions like hypoxia, and nutrient starvation may also play important roles. Though, the BALB/c mouse model has its limitations but is still better than investigating the in-vitro conditions alone.
In the recent years, development of several improvised gene knock-out mouse models has lead to better understanding of TB pathophysiology. These models display a human-like disease pathology in the lungs with hypoxia or other microenvironments of granuloma; harbouring mixed populations of Mtb: 1) Kramnik mice strains (C3HeB/FeJ [52]; 2). NOD-SCID/γc null of NSG model [53]; 3). iNOS-/-for NO synthase [4], etc. Each of these models has their pros and cons [54].
BALB/c mice have a strong precedence of use for testing drug efficacy including the investigational new drugs in the current pipeline, like TMC 207, PA824, SQ109, ADZ-5847, and Benzothiazinones, [55][56][57][58][59][60] and for target essentiality [61,62]. Hence, we used BALB/c mice strain in these studies. Although, BALB/c mice models are well validated for tuberculosis and are the standard route for translation in drug discovery, but, for cidality studies under different physiologies, it may not best represent the micro-environments and pathophysiology of human-TB. The solid granulomas harbouring primarily intracellular bacteria do not particularly develop hypoxia in BALB/c mice.
The low pH in-vitro screen emerging as the best-correlated predictor of in-vivo cidality, from BALB/c mouse model, may be a partial conclusion because of the two reasons: 1). Despite an effort to establish a hybrid of acute-chronic infection model in BALB/c, the high dose infection established represented mostly an acute-like infection with replicating bacteria analogous to early human disease. Hence, in the acute phase of infection for the high fidelity translation, the compound efficacy may also be evaluated in the low pH, in addition to the standard replication condition.2). The BALB/c mice used in the present study primarily form solid granuloma (only a single pathophysiological lesion type) harbouring largely the intracellular Mtb, unlike those observed (intra/extra-cellular Mtb with hypoxic microenvironment) in the cavitary lesions of human TB patients. Therefore, it is pertinent to perform such correlation studies using AS-recombinants of a larger set of target genes in humanised chronic infection models (Kramnik-C3HeB/FeJ) that may unleash these cidality correlations better. Kramnik model represents both extra/intra-cellular pathogen populations, along with most of the cavitary lesion types, as well as multiple stringent μ-environments of human-like granuloma that influence the pathogen to modulate its survival-specific genes for adaptation [51,54].
Our treatment therapy is effective on the acute phase of infection; it may not reflect what happens in the chronic phase. The low pH may not be important in general, as we earlier envisaged, but may be only relevant to the acute phase. The importance of the low pH or the appropriate screen in the chronic phase may be separately investigated.
Treatment of TB is complex; eradication of the dormant Mtb takes longer to treat due to hide-n-seek being played by this pathogen [63]. It cannot be attributed only to poor bio-availability of drugs [64,65], but may primarily be dictated by either the physiologic heterogeneity of bacteria in the tissues [66] or penetration of drugs in caseating foci/granuloma [66]. Rifampicin and PZA, the best sterilizing drugs responsible for shortening treatment duration in humans, penetrate better in to granulomas vs. moxifloxacin that largely concentrates in the periphery [66,67]. Interestingly all of these best tuberculocidals exhibit good MIC in acidic pH (Table D in S1 File), except isoniazid, which is inactive under acidic environment, hence, demonstrates a reduced activity on intracellular Mtb. The alarming statistics [1] on drug resistance in TB demand novel PZA-like sterilizing drugs that work best in low/acidic pH, kill non-replicating populations of Mtb and shorten the treatment duration [38,68]. In a separate study, we have demonstrated the importance of low pH screens by selecting PZA-hybrid molecules invitro [69]. Recently, PZA was also reported to enhance the cidality of various combinations of existing drugs [70]and is an integral component of emerging novel combinations in the clinic like PAMZ (PA-824, Moxi, PZA), in STAND (Shortening Treatment by Advancing New Drugs, www.clinicaltrial.com) clinical trial. Our findings unequivocally suggest that anti-TB activity in the low pH environment may be predictive of in-vivo tuberculocidality, especially for the acute phase of infection.
The low pH condition in-vivo appears to be a cumulative sum of various triggers and several secondary mechanisms. These triggers need to be explored in intricate details, and their appropriateness to identify tuberculocidal therapeutics may be strengthened further using large compound libraries and potential targets. The targets of stationary phase like ppk, need to be explored in the long-term infection models of experimental TB.
Our studies on in-vivo AS-repression of cidal targets lead to the killing of Mtb in-vitro and in-vivo, hence, demonstrating the bactericidal effect. It suggests that in the long run, AS-therapeutics can be explored in patients suffering from drug-sensitive and drug-resistant TB, for which currently there are only a few effective drugs available. It is a futuristic goal, but AS-therapeutics is an emerging radical approach to treating various diseases like anti-viral infections [71] (Fomivirsen or Vitravene) anti-cholesterol [72] (mipomersen), or even anti-cancer [73] (AP 12009) without any safety issues. However, as antibacterial, there is a need to develop improved AS-delivery systems for its successful applications. A lot of investigations are underway to overcome significant obstacles towards the development of efficient delivery systems like CPP (cell penetrating peptides), nanoparticles, or nanotubes, etc. [74,75] in the long process. Some companies are developing the AS delivery systems for bacterial therapeutic AS, ISIS Pharmaceuticals [ISIS Pharmaceuticals Inc.; now called Ionis pharmaceuticals http://www. ionispharma.com] is the leader in the field of AS-therapeutics.
An astute observation through the lens of antisense into a kaleidoscope of physiology-specific gene expression under in-vivo pressures in mouse; or, simulated in-vitro screens; unravelled a few secrets while a lot remain unrevealed.
Our studies with in-vivo AS-RNA-silencing of mycobacterial targets unravelled following features on target cidality: 1). AroK emerged as the in-vivo cidal target under diverse physiologies. These findings suggest aroK a potential target for developing mycobactericidal agents. 2). Use of rpoB target validated the entire concept on in-vitro to in-vivo translation. Like rpoB, aroK demonstrated that a druggable target needs to be cidal under all the physiological states in the host. 3). Cidality-SCORE is a rational approach to rank order cidality potential of the targets. Cidality-SCORE positively correlated in-vitro to the in-vivo translation of 4 out of 5 targets (rpoB, aroK, rpoC, and ppK) expressed under multiple physiologies. However, ilvB failed to translate to in-vivo cidality due to its auxotrophic nature and hence attenuation. 4). Low pH appeared to be an in-vivo predictor of cidality in acute infection. Hence, in the acute phase of infection for the high fidelity translation, the compound efficacy may also be evaluated in the low pH, in addition to the standard replication condition. Further, in-vitro to in-vivo translation correlations need to be studied in the chronic humanised models to narrow-down to a single in-vitro screen that alone can predict in-vivo (acute-chronic) cidality potential of the targets/ inhibitors.
We used a small subset of only five, potentially cidal target genes. Studies with a larger set of mixes of validated and unvalidated genes and their investigation in physiologically more relevant humanised mouse disease models like Kramnik and others are required to unravel other unrevealed shades. (TIF) S1 File. " Table A in S1 File", " Table B in S1 File", "Table C in S1 File", and "Table D in S1 File". (DOCX)

Supporting Information
Obtained regulatory permissions for in-vitro rDNA experiments and in-vivo experiments using mice: PK RKS NK SN.