Characterization of Phosphofructokinase Activity in Mycobacterium tuberculosis Reveals That a Functional Glycolytic Carbon Flow Is Necessary to Limit the Accumulation of Toxic Metabolic Intermediates under Hypoxia

Metabolic versatility has been increasingly recognized as a major virulence mechanism that enables Mycobacterium tuberculosis to persist in many microenvironments encountered in its host. Glucose is one of the most abundant carbon sources that is exploited by many pathogenic bacteria in the human host. M. tuberculosis has an intact glycolytic pathway that is highly conserved in all clinical isolates sequenced to date suggesting that glucose may represent a non-negligible source of carbon and energy for this pathogen in vivo. Fructose-6-phosphate phosphorylation represents the key-committing step in glycolysis and is catalyzed by a phosphofructokinase (PFK) activity. Two genes, pfkA and pfkB have been annotated to encode putative PFK in M. tuberculosis. Here, we show that PFKA is the sole PFK enzyme in M. tuberculosis with no functional redundancy with PFKB. PFKA is required for growth on glucose as sole carbon source. In co-metabolism experiments, we report that disruption of the glycolytic pathway at the PFK step results in intracellular accumulation of sugar-phosphates that correlated with significant impairment of the cell viability. Concomitantly, we found that the presence of glucose is highly toxic for the long-term survival of hypoxic non-replicating mycobacteria, suggesting that accumulation of glucose-derived toxic metabolites does occur in the absence of sustained aerobic respiration. The culture medium traditionally used to study the physiology of hypoxic mycobacteria is supplemented with glucose. In this medium, M. tuberculosis can survive for only 7–10 days in a true non-replicating state before death is observed. By omitting glucose in the medium this period could be extended for up to at least 40 days without significant viability loss. Therefore, our study suggests that glycolysis leads to accumulation of glucose-derived toxic metabolites that limits long-term survival of hypoxic mycobacteria. Such toxic effect is exacerbated when the glycolytic pathway is disrupted at the PKF step.


Introduction
Despite the availability of effective anti-tubercular drugs, tuberculosis (TB) remains a scourge of public health with 8.8 million people infected with active TB in 2010 [1]. The metabolic versatility of M. tuberculosis, the etiological agent of TB, represents one of the key virulence strategies developed by the pathogen to persist in many microenvironments within its host [2]. Earlier studies on carbon metabolism have shown that M. tuberculosis can utilize a variety of carbon substrates [3].
During infection, several studies have shown that the gluconeogenic pathway is required for infection and persistence, suggesting that fatty acids constitute one of the main carbon and energy source utilized by M. tuberculosis [4][5][6][7]. In addition, the sequencing of M. tuberculosis genome revealed that fatty acid b-oxidation genes are extensively duplicated [8] and are up-regulated during infection in macrophages [9] and in mice [10]. Beside fatty acids, host cholesterol is another possible carbon source used by M. tuberculosis during infection [11,12].
However, M. tuberculosis displays a unique versatility for carbon metabolism that we are only starting to appreciate and understand. A recent study has shown that M. tuberculosis is not subjected to catabolic repression and is therefore capable of co-catabolizing several carbon sources simultaneously for optimal growth [13]. In particular, the simultaneous catabolism of glucose and lipids was found to potentiate bacterial growth, at least in vitro [13]. Therefore, even though b-oxidation is required for virulence, it is conceivable that other carbon and energy sources may be utilized by M. tuberculosis for optimal infection and persistence in the various microenvironments within its host. Glucose represents one of the most abundant sources of carbon and energy, and it is thus not surprising that the glycolytic pathway is highly conserved in almost all living organisms. Indication that glucose metabolism might be important for M. tuberculosis during infection arises from a study where putative carbohydrate transporters and a hexose kinase were found essential for infection in mice [10]. Studies in Salmonella enterica serovar Typhimurium, an intracellular pathogen, have also shown that glycolysis is required for infection [14,15].
The M. tuberculosis genome reveals an intact glycolytic and pentose phosphate pathway but no Entner-Doudoroff pathway [8]. Early in vitro studies suggested that glucose is predominantly oxidized through glycolysis while a small fraction enters the pentose phosphate pathway [16]. The key-committing step of glycolysis is catalyzed by a phosphofructokinase (PFK) activity, which irreversibly catalyzes the phosphorylation of fructose-6phosphate to fructose-1,6-bisphosphate ( Fig. 1). Two putative PFKs (PFKA and PFKB) encoding genes have been annotated in M. tuberculosis genome, namely Rv3010c and Rv2029c, respectively. Although both enzymes are proposed to catalyze the same enzymatic reaction, they belong to different subfamilies; PFKA belongs to the PFK protein family whereas PFKB belongs to the pfkB subfamily of ribokinase superfamily. Furthermore, not only their amino acid sequence greatly differs, but their gene expressions are different whereby pfkB, is a member of the DOS regulon and is up-regulated during hypoxia [17,18] and in activated macrophages [9]. TraSH-based mutagenesis screen indicated that both pfkA and pfkB are not essential for the survival of M. tuberculosis in vitro and in vivo growth [10,19].
In this work, we investigated the role of PFKA and PFKB in M. tuberculosis glucose metabolism. We demonstrated that pfkA encodes a functional PFK that is essential for growth on glucose as sole carbon source, and is responsible for the total PFK activity in M. tuberculosis. No functional redundancy between pfkA and pfkB could be established. Our data indicate that a functional glycolytic pathway is required to limit the intracellular accumulation of glucose-derived toxic metabolic intermediates during co-metabolism. We also report a strong detrimental effect of glucose metabolism for the long-term survival of hypoxic non-replicating mycobacteria.

Ethics statement
All the animal experiments were approved by and carried out under the guidelines of the Institutional Animal Care and Use Committee (IACUC) of Novartis Institute for Tropical Diseases, Singapore. Non-terminal procedures were performed under anesthesia, and all efforts were made to minimize suffering.

Mycobacterial strains and growth conditions
M. tuberculosis H37Rv and derivative mutant strains were grown in Middlebrook 7H9 liquid medium supplemented with 0.2% glycerol, 0.5% bovine serum albumin fraction V, 0.2% glucose, 0.085% sodium chloride and 0.05% Tween-80 or on Middlebrook 7H11 agar supplemented with 10% oleic acid-dextrose-albumincatalase enrichment and 0.5% glycerol. When required, culture media were supplemented with hygromycin at final concentration of 80 mg/ml or with kanamycin at a final concentration of 25 mg/ ml.
For growth kinetics studies in defined culture broth media, mycobacteria were first cultured in 7H9 medium until mid-log phase, washed once with defined culture broth medium (no carbon source) and then inoculated at an initial optical density at 600 nm (OD 600 ) of 0.05. The defined culture broth media contained 2.5 g Na 2 HPO 4 , 1 g KH 2 PO 4 , 0.424 g glutamic acid, 1 mg pyridoxine, 0.5 mg biotin, 15 mg ferric ammonium citrate, 40 g MgSO 4 , 0.5 mg CaCl 2 , 0.6 mg ZnSO 4 , 0.6 mg CuSO 4 , 0.8 g NaCl, 0.5 g Tyloxapol and 0.1% fatty-acid free bovine serum albumin (Sigma A8806) per litre of medium. Glucose, acetate or glycerol was added at a final concentration of 0.2% as carbon source. Bacterial growth was monitored by measuring the optical density at 600 nm over time.
For growth kinetics studies in Dubos medium, mycobacteria were first culture in Dubos liquid medium (Difco) supplemented with 0.5% BSA fraction V, 0.085% NaCl and 0.03% Tween-80 until mid-log phase. The cells were inoculated at an initial OD 600 of 0.05 in either Dubos liquid medium described previously or in complete Dubos liquid medium (further supplemented with 0.75% glucose). Bacterial growth was monitored at OD 600 over time.

Construction of M. tuberculosis knockout and complemented strains
The knockout mutants were obtained by double homologous recombination using plasmid pYUB854 as described before [20]. Briefly, fragments of ,1 kb flanking pfkA or pfkB opening reading frames (ORFs) were PCR amplified from H37Rv genomic DNA using primer pairs: PFKA5F-PFKA5R, PFKA3F-PFKA3R, PFKB5F-PFKB5R and PFKB3F-PFKB3R (Table 1). The 59 and 39 flanking fragments were then cloned into pYUB854 flanking the hygromycin-resistance gene. The sacB-lacZ cassette excised from pGOAL17 [21] was finally cloned into the unique PacI site of pYUB854. The final plasmids were UV-irradiated prior to electroporation into H37Rv strain. Positive clones of knockout mutants were selected as white colonies on 7H11 agar supplemented with hygromycin and X-Gal. Deletion of the target gene was verified by PCR (primers sequence in Table 1) and confirmed by Southern blot. Unmarked knockout mutant of pfkA was obtained upon removal of the hygromycin cassette using the resolvase gene-containing plasmid pWM19 as described previously [22]. Colonies were first selected on 7H11 agar containing gentamycin at 31uC and followed by selection on 7H11 agar supplemented with 2% sucrose at 39uC.
For complementation of DpfkA mutant, the pfkA ORF and its putative promoter region were PCR amplified with primers PFKAcF and PFKAcR (Table 1) from H37Rv genomic DNA and cloned into the promoterless integrative vector pMV306 [23]. For complementation of DpfkB mutant, the pfkB ORF was PCR amplified with primers PFKBcF and PFKBcR (Table 1) and cloned into replicative vector pMV262 under the constitutive hsp60 promoter [23]. The XbaI-HindIII fragment from recombinant pMV262-hsp60pfkB plasmid was then cloned into pMV306, giving pMV306-hsp60pfkB. All the complemented strains were verified by PCR.

Southern blot analysis of knockout mutants
Genomic DNA of the parental and mutant strains were digested with restriction enzyme BamHI for confirmation of pfkA deletion and EcoRI for confirmation of pfkB deletion. Digested DNA was separated on a 0.8% agarose gel, transferred onto nylon membrane and probed for modification of the loci. Southern blot analysis was performed using DIG Nonradioactive Nucleic Acid Labelling and Detection System (Roche), following the manufacturer's instruction. DIG-labelled probeA and probeB (Fig. 2) were PCR-amplified using primer pairs of PFKAusF-PFKAsR and PFKBusF-PFKBsR (Table 1) respectively. Bands were visualized using chemiluminescent detection.

Complementation of an E.coli PFK knockout mutant
Mycobacterial pfkA and pfkB ORFs were PCR amplified using primer pairs PFKApQE60F-PFKApQE60R and PFKBcF-PFKBcR respectively (Table 1). pfkA was cloned into expression vector pQE-60 and and pfkB was cloned into pQE-30 (Qiagen). These vectors allow an IPTG-inducible T5 promoter-driven expression in E. coli. The recombinant plasmids were electroporated into DpfkADpfkB E. coli mutant RL257 (CGSC, Yale) [24]. The transformed RL257 strains and control RL257 strain were grown in Luria Bertani (Miller) broth medium (Difco) until midlog phase and washed once with M9 solution (Difco). OD 600 was then adjusted to 0.01 before streaking onto minimal medium agar containing M9 minimal salts (Difco) supplemented with 2 mM MgSO 4 and 0.2% glucose or glycerol as carbon source. Expression of mycobacterial pfkA and pfkB genes was induced by 0.01 mM IPTG in the agar plates. Bacteria were incubated at 37uC overnight.

Cloning, expression and purification of PFKA and PFKB
The pfkA and pfkB ORFs were PCR amplified from H37Rv genomic DNA with primer pairs PFKApET29F-PFKApET29R and PFKBpET15F-PFKBpET15R respectively (Table 1). pfkA was cloned into expression vectors pET-29a(+) and pfkB was cloned into pET15b (Novagen). pfkA was expressed as C-terminal 6xHistag recombinant protein while pfkB was expressed as N-terminal 6xHis-tag recombinant protein in E. coli BL21(DE3) (Stratagene). Bacteria cultures were grown at 37uC in LB broth until mid-log phase and then transferred to 16uC. Induction of recombinant protein expression started with the addition of 0.1 mM IPTG and bacteria were cultured at 16uC for 20 hrs. Bacterial cells were harvested and disrupted by sonication. Cell debris were removed by centrifugation and the His-tagged proteins were purified under native conditions on Ni-NTA agarose column (Qiagen) followed by size exclusion chromatography on Superdex 200 10/300 GL column (GE Healthcare). The proteins were stored at 280uC in buffer containing 50 mM Tris-HCl pH7.5 and 5 mM MgCl 2 .

Preparation of cell-free crude extracts for enzyme and metabolite assays
Cell-free crude extracts of M. tuberculosis strains were prepared by harvesting mid-log phase culture grown in 7H9 medium. For metabolites measurement, bacterial strains were culture in complete Dubos liquid medium or Dubos liquid medium without glucose. Mycobacterial cells were washed twice with PBS/0.05% Tween-80 and resuspended in lysis buffer [50 mM Tris-HCl pH 7.5, 5 mM MgCl 2 , 1 mM dithiothreitol and complete protease inhibitor (Roche)]. Mycobacterial cells were disrupted mechanically by 0.1 mm glass beads in FastPrep FP220A beadbeater (Qbiogen). Lysates were clarified by centrifugation and then filtered through 0.22 mm filter. Total protein concentration was measured with BCA protein assay reagent kit (Pierce). Cell-free crude extracts to be used for metabolite assays were boiled for 10 min and centrifuged at 13,000 rpm for 10 min at 4uC.

Phosphofructokinase activity assay
Phosphofructokinase activity was measured in an enzymecoupled assay in which fructose-1,6-bisphophate formation is coupled to the oxidation of NADH [25]. The standard assay mixture (0.1 ml) contained 5 mM fructose-6-phosphate, 1 mM ATP, 0.3 mM NADH (Roche), 1 unit each of aldolase, triosephosphate isomerase and glycerol-3-phosphate dehydrogenase in 50 mM Tris-HCl pH7.5 and 5 mM MgCl 2 . For enzymatic assay with purified recombinant proteins, 1 mM fructose-6-phosphate and 0.1 mM ATP were used instead. The enzyme activities were measured by monitoring the decrease in absorbance at 340 nm using a SpectraMax spectophotometer (Molecular Devices) at room temperature. Quantification was done with a NADH standard curve. All enzymes and reagents used in enzyme-coupled assay were purchased from Sigma-Aldrich, unless otherwise stated.

Immunoblotting
Samples of cell-free crude extract were separated on a NuPAGE 4-12% polyacrylamide get (Invitrogen) and transferred to PVDF membrane. PFKB was detected with rabbit polyclonal anti-PFKB antibody, raised against the recombinant PFKB, and visualized with SuperSignal West Pico Chemiluminescent Substrate kit (Pierce). Ponseus-S (Sigma-Aldrich) staining of the membrane was done to check for equal loading of the cell lysates.

Measurement of intracellular metabolites
Intracellular metabolites concentration were enzymatically determined as described by Hasan et al [26] with slight modifications. Glucose-6-phosphate concentration was determined by measuring the increase in absorbance at 340 nm in an enzymatic assay reaction (0.2 ml) containing 0.3 mM NADP + (Roche), 0.1 unit of glucose-6-phosphate dehydrogenase in 50 mM Tris-HCl pH7.5 and 5 mM MgCl 2 . To determine the concentration of fructose-6-phosphate, 0.1 unit of phosphoglucose isomerase was added to the reaction mixture after glucose-6phosphate reaction was completed and the change in absorbance at 340 nm was recorded. Quantification was done with a NADPH standard curve. All enzymes and reagents used in metabolites measurement were purchased from Sigma-Aldrich, unless otherwise stated.

Mouse infection
Animals were housed in specific pathogen-free conditions in individual ventilated cages in an ABSL3 facility. Female BALB/c mice of 6-8 weeks old were nasally infected with 10 3 CFU of H37Rv parental and DpfkA strains. Four animals per group were sacrificed at the indicated time points. The lungs and spleen were aseptically harvested and homogenized in PBS/0.05% Triton X-100. The bacterial load were quantified by plating serial dilutions of the organ homogenates on 7H11 agar supplemented with cycloheximide and ampicillin each at 10 mM. The number of CFU was recorded after 16 days incubation at 37uC.

Wayne model of hypoxia
M. tuberculosis H37Rv and mutant strains were subjected to slow withdrawal of oxygen as described before [27]. Mycobacteria were first cultured in Dubos liquid medium (without glucose). The cells were then diluted in either Dubos liquid medium (without glucose) or complete Dubos liquid medium to a final OD 600 of 0.002. 17 ml of the diluted culture was aliquoted into screw-cap test tubes (20 mm by 125 mm) to maintain a head-to-space ration of 0.5 and the test tubes were tightly capped. The cultures were then stirred gently at 170 rpm on magnetic stirring platform at 37uC. Methlylene blue (1.5 mg/ml) was added to two representative tubes to monitor oxygen depletion. Growth and survival of mycobacteria were determined by enumeration of CFU after 2 to 3 weeks of incubation at 37uC on 7H11 agar plated out at various time-points..

Results
PfkA is responsible for the total PFK activity in M. tuberculosis Two genes pfkA (Rv3010c) and pfkB (Rv2029c) have been annotated to encode a PFK in M. tuberculosis. To investigate the relative contribution of each gene product to the overall M. tuberculosis PFK activity, M. tuberculosis mutants deleted for either pfkA or pfkB were constructed by homologous recombination in M. tuberculosis H37Rv. Since pfkA is part of an operon, an unmarked pfkA KO mutant was constructed to avoid any polar effect on downstream open reading frame (ORF) gatB (Fig. 2). Deletion at the correct genetic locus was confirmed by Southern blot (Fig. 2). Complementation was then performed whereby an intact copy of the pfkA ORF and its promoter region was re-introduced into the DpfkA bacterial chromosome using the promoterless integrative plasmid pMV306. A PFK enzymatic assay was developed using cell-free extracts from the parental (WT), KO and complemented strains. Results showed that PFK activity could not be detected over background level in the DpfkA mutant ( Table 2). The PFK activity in DpfkB mutant was comparable to that measured in the WT strain, suggesting that pfkB does not encode for a functional PFK. The PFK activity could be restored to parental level in the DpfkA mutant upon complementation with a wild-type copy of pfkA ( Table 2). These data thus suggested that pfkA is responsible for the total PFK activity in M. tuberculosis, at least under aerobic conditions.
Since pfkB was previously reported to be upregulated under hypoxic condition and in activated macrophages [9], we hypothesized that PFKB may contribute to M. tuberculosis PFK activity under hypoxia but not during aerobic growth. To test whether pfkB encodes for a functional PFK, the pfkB ORF was cloned in a replicative plasmid (pMV262) under the control of the constitutive hsp60 promoter, and expressed in the DpfkA mutant. PFKB over-expression in the DpfkA mutant was confirmed by Western blot (Fig. 3), but did not lead to detectable PFK activity levels in the cell free extracts, further supporting that PFKB does not contribute to the mycobacterial PFK activity (Table 2). Consistently, when M. tuberculosis PFKA and PFKB were expressed in a pfkA/pfkB double KO strain of E. coli (RL257) [24], only mycobacterial pfkA but not pfkB allowed the growth of E. coli RL257 on minimal medium with glucose as the sole carbon source (Fig. 4), strongly suggesting that pfkB does not encode for a PFK enzyme.
Finally, His-tagged PFKA and PFKB proteins were overexpressed in E. coli, purified and tested in enzyme-coupled assay for their PFK activity. PFKA was able to catalyze the phosphorylation of fructose-6-phosphate to fructose-1,6-bisphosphate efficiently, whereas no significant activity was detected from PFKB ( Table 2).
Taken together, these data strongly support that pfkA is responsible for the overall PFK activity in M. tuberculosis H37Rv, and that pfkB does not catalyze fructose-6-phosphate in vivo.
pfkA is necessary and sufficient for M. tuberculosis growth on glucose as sole carbon source To further study the role of pfkA and pfkB in mycobacterial glucose metabolism, we tested the ability of both the DpfkA and DpfkB mutants to grow in the presence of glucose as sole carbon source. The results showed that DpfkA mutant was unable to grow efficiently on glucose as sole carbon source whereas it displayed a parental growth kinetic on acetate or glycerol (Fig. 5A-C). The growth defect on glucose observed with mpfkA mutant was restored upon complementation with a wild-type copy of pfkA (Fig. 5A). Consistent with the PFK activity data (Table 2), the growth defect of DpfkA mutant could not be reversed upon constitutive expression of pfkB (Fig. 5A). In contrast, DpfkB mutant showed no growth defect on glucose (Fig. 5D) which is in agreement with the PFK activity measured in this mutant strain (Table 2). These data thus further support that PFKA is indispensable for the glycolytic pathway in M. tuberculosis during aerobic growth with no functional redundancy with PFKB.

PFKA is not required for virulence and survival in the mouse model of tuberculosis infection
Genes encoding putative disaccharide transporters were predicted to be required for the first week of mouse infection [10]. Studies in Salmonella enterica serovar Typhimurium have shown that PFK is important during mouse infection [14,15]. To study the role of PFKA in M. tuberculosis virulence, BALB/c mice were nasally infected with the parental or DpfkA strains, and their infection profiles in the lung and spleen were monitored. The results indicated that the bacterial loads in both organs recovered from both mouse groups were comparable (Fig. 6). This result thus suggested that PFKA, and therefore glycolysis, is not crucial for M. tuberculosis survival and persistence in the mouse lungs and spleen. This result does not rule out the possibility that PFKA may be required for survival in other animal models where pathology and physiology of the bacterium are closer to those observed during human infection.

PFKA is required for survival of hypoxic non-replicating M. tuberculosis
During the course of in vitro aerobic growth we noticed that DpfkA mutant displayed impaired fitness upon reaching stationary phase in Dubos medium with no visible signs of clumping (Fig. 7A). Dubos medium contains large amount of glucose, amino acids and lipids as main sources of carbon and energy. When glucose was depleted from the Dubos medium, DpfkA mutant survived as well as the parental strain during the stationary phase (Fig. 7B). We thus hypothesized that accumulation of toxic glucose-derived sugar-phosphates such as glucose-6-phospate and fructose-6phosphate in the DpfkA mutant may account for the growth defect observed during co-metabolism when oxygen tension becomes limiting. Sugar-phosphates have indeed been shown to be highly toxic in many bacteria, including M. tuberculosis [28]. Consistently, the pool of glucose-6-phospate and fructose-6-phosphate measured during the exponential growth of the DpfkA mutant was 50% higher compared to the parental strain (Table 3). It is interesting to note that while accumulation of sugar-phosphates occurs during the growth exponential phase, the toxic phenotype instead was only observed during the stationary phase, linking the detrimental effect of sugar-phosphate accumulation with oxygen depletion.
This observation prompted us to extend our study to the survival of DpfkA mutant under anaerobic conditions using the well-established in vitro Wayne model of hypoxia in which gradual depletion of oxygen triggers the bacterium to enter a nonreplicating state [27]. In this model, Dubos medium supplemented with glucose is classically employed by the vast majority of research groups to study the physiology of hypoxic non-replicating mycobacteria [27]. The DpfkA mutant multiplied efficiently before oxygen depletion. However after day 6, which coincided with decolourization of the oxygen probe methylene blue, DpfkA bacteria displayed a significant viability loss compared to the parental and complemented strains (Fig. 7C). To test whether the attenuated phenotype was linked to the accumulation of toxic glucose-derived metabolites, the same experiment was performed in culture medium in which addition of glucose was omitted. In these culture conditions DpfkA mutant survived as well as the parental strain, demonstrating that in the presence of exogenous glucose, absence of PFK activity leads to the accumulation of toxic metabolic intermediates in hypoxic non-replicating mycobacteria (Fig. 7D).

Glucose is detrimental for long term M. tuberculosis survival in the Wayne model
The toxic effect of glucose observed in the Wayne model when the glycolytic pathway is disrupted prompted us to take a closer look at the limited viability traditionally observed with M. tuberculosis whereby non-replicating mycobacteria do not survive longer than 25 days after which they start to die at an accelerated rate with less than 0.1% of the initial inoculum of non-replicating  cells still viable at day 40 [27]. While the reason for this limited long term viability has never been investigated, we hypothesized that this phenomenon might be explained by the accumulation of glucose-derived toxic metabolites over time. Consistently, when grown in medium without glucose, M. tuberculosis viability was maintained over 60 days, whereas a steep decrease in viability was observed in the presence of glucose (Fig. 8) as reported before. Furthermore, analysis of the intracellular metabolites pool showed significantly higher level of glucose-6-phosphate in hypoxic cells compared to non-hypoxic bacteria (Table 4). More importantly the level of glucose-6-phosphate in cells cultured in the glucosesupplemented medium was 2-fold higher than that in cells cultured in absence of glucose. Interestingly, the rate of methylene blue decolourization in culture without glucose was significantly slower than that observed in culture with glucose, suggesting that the rate of respiration in M. tuberculosis is slower in the absence of active glycolysis. Altogether, these observations thus indicate that longterm survival of hypoxic mycobacteria in the presence of exogenous glucose is limited by accumulation of toxic glucosederived metabolic intermediates.

Discussion
M. tuberculosis is believed to encounter a range of different microenvironments in its host as it progresses from the initial infection of alveolar macrophages to the development of granulomas and extra-pulmonary dissemination. One way that M. tuberculosis copes with the challenge of changing and hostile environments is by metabolic adaptation. M. tuberculosis is able to utilize a variety of carbon sources at least in vitro and genomic analysis has revealed the presence of a number of carbon metabolic pathways, including the highly conserved glycolytic pathway. Previous study has shown the presence of a functional  glucose kinase which phosphorylates glucose to glucose-6-phosphate (Fig. 1) [29]. Here we demonstrated the presence in M. tuberculosis H37Rv of a phosphofructokinase (PFK) activity, the key regulatory enzyme of glycolysis. Similar to E. coli, two mycobacterial genes were annotated as PFK encoding genes, namely pfkA and pfkB. In E. coli PFKB is a minor isoenzyme and accounts for about 10% of the bacterial PFK activity [30]. Furthermore, a pfkAdeleted E. coli mutant was shown to be able to grow on glucose provided that PFKB was present and functional [24,31]. Here, we have generated strong experimental evidence supporting that PFKA accounts for the total M. tuberculosis PFK activity without functional redundancy with PFKB. No PFK activity was detected in crude extract from DpfkA M. tuberculosis mutant; the DpfkA M. tuberculosis mutant could not be complemented with pfkB expressed under a constitutive strong promoter; a PFK-deficient E. coli mutant could be complemented when expressing M. tuberculosis pfkA but not M. tuberculosis pfkB; purified recombinant M. tuberculosis PFKA displayed a PFK activity in vitro while PFKB showed minimal activity. Although purified recombinant PFKB catalyzes fructose-6-phosphate in vitro, albeit at very low efficiency, it is not able to complement the loss of PFKA in vivo. This suggests that fructose-6-phosphate might not be the true substrate of M. tuberculosis PFKB. Predictive three-dimensional protein structure generated by Phyre2 server [32] showed that M. tuberculosis PFKB shares 40% identity with E. coli PFKB (data not shown). Based on the presence of the conserved catalytic motif GXGD in its amino acid sequence, M. tuberculosis PFKB has been classified as a member of the ribokinase superfamily, PFKB subfamily. Analysis of T. gondii adenosine kinase's crystal structure suggested that enzymes from the ribokinase family are able to adapt easily to a variety of sugar-based substrates [33]. Members of the PFKB subfamily which share high degree of structural conservation have been shown to phosphorylate a variety of substrates beside fructose-6-phosphate; examples are fructose-1-phosphate in E. coli   [34,35] and tagatose-6-phosphate in S. aureus [36] and E. coli, although with a lower efficacy than fructose-6-phosphate [37]. Thus, it is possible that M. tuberculosis PFKB is able to phosphorylate sugar-based substrates other than fructose-6-phosphate. So far none of the studies on the kinases from the PFKB subfamily have identified amino acid residues involved in substrate specificity. As such the nature of the M. tuberculosis PFKB substrate cannot be deduced from its amino acid sequence and has yet to be elucidated.
Co-metabolism experiments showed a defect in cell viability with DpfkA mutant upon entry into the stationary phase. The attenuation phenotype was correlated with accumulation of toxic metabolic intermediates in the glycolytic pathway upstream of PFKA. Consistently, removal of glucose from the culture medium restored viability of the DpfkA mutant. Significantly higher intracellular pools of glucose-6-phosphate and fructose-6-phosphate in the DpfkA mutant, compared to the parental strain, further supports the hypothesis that these sugar-phosphates may be toxic to the bacterial cell and accumulate in a PFK-deficient mutant. This finding is consistent with a previous study where we showed that excessive metabolic intermediates such as glycerol phosphate, dihydroxyacetone phosphate and methylglyoxal are toxic to M. tuberculosis [28]. Accumulation of sugar-phosphate may have various physiological consequences including mRNA destabilization [38], stimulation of gene expression [39], and activation of pyruvate kinase [40], all of which may contribute to impair the cell viability. The toxic effect of sugar-phosphate in M. tuberculosis was previously reported whereby accumulation of maltose-1phosphate leads to bacterial death in vitro and in mice [41]. It is of interest to note that maltose-1-phosphate is a product of trehalose catabolism with glucose-6-phosphate being a precursor of trehalose, thereby linking glycolysis and trehalose pathway. In addition to the accumulation of toxic metabolic intermediates, the impact of glycolysis disruption on other metabolic pathways may also play a role in reduced cell viability. Also, it must be noticed that the DpfkA mutant did not exhibit significant growth defect in standard 7H9 medium including during the stationary phase (data not shown). This discrepancy between Dubos and 7H9 media may be attributed to the higher concentration of glucose in Dubos medium compared to 7H9 (0.75% versus 0.2%), with the idea that a threshold glucose concentration may be necessary before a toxic effect can be observed. However, other differences in the composition between both media may also explain the difference in the phenotype observed.
Interestingly, a detrimental effect of glucose was observed in wild-type M. tuberculosis grown in vitro under hypoxia in the wellestablished Wayne model. We showed that M. tuberculosis could persist in a non-replicating state for much longer when glucose was omitted in the culture medium. There has been much speculation on the possible reasons of the limited persistence of M. tuberculosis in the NRP2 phase of the Wayne model, such as nutrients exhaustion or low levels of ATP. Here we show that the limited mycobacterial persistence is linked to the presence of glucose in the medium and is likely due to the accumulation of glucose-derived toxic metabolic intermediates. We believe that this finding is of great importance and advocates for revisiting the mechanisms employed by M. tuberculosis for long-term persistence in the absence of growth.
The metabolomic profile of M. tuberculosis infected murine tissues was recently analyzed and revealed that the level of glucose and glycogen in those tissues decreased along with the increase in phospholipids level [42]. This may suggest that M. tuberculosis switches from the carbohydrate to lipid metabolism in order to adapt to its microenvironment. However, we did not observe any significant difference between the DpfkA and parental strains in their ability to colonize and persist in the mouse lungs and spleen. This suggests that the glycolytic pathway is dispensable during mice infection and also indicates that the toxic effect observed in vitro in Dubos medium is not observed in vivo. It is therefore possible that M. tuberculosis replicates and persists in an environment where access to glucose is limited. Alternatively, since attenuation of the DpfkA mutant was seen mostly under hypoxia in vitro, absence of hypoxic granuloma or lesions in mice may not allow recapitulating such attenuation [43]. It would be interesting to determine the fitness of DpfkA mutant in animal models where hypoxic granulomatous lesions are formed. Regardless, the Table 3. Concentration of intracellular glucose-6-phosphate and fructose-6-phosphate in aerobic M. tuberculosis strains.  maintenance of an intact glycolytic pathway in M. tuberculosis throughout evolution indicates that glycolysis could play an important role in mycobacterial infection and persistence in microenvironments not recapitulated in mice.
In conclusion, we provide here the experimental evidence that PFKA is responsible for the overall PFK activity in M. tuberculosis and that there is no functional redundancy with PFKB. Furthermore, our work demonstrates that in the presence of exogenous glucose, hypoxic mycobacteria tend to accumulate toxic glucose-derived metabolic intermediates that impair the bacilli long-term survival. Disruption of the glycolytic pathway further accentuates accumulation of these toxic intermediates.