Pyruvate Dehydrogenase Kinase Inhibitor Dichloroacetate Improves Host Control of Salmonella enterica Serovar Typhimurium Infection in Human Macrophages

Global increases in the prevalence of antimicrobial resistance highlight the urgent need for novel strategies to combat infectious diseases. Recent studies suggest that host metabolic pathways play a key role in host control of intracellular bacterial pathogens. In this study we explored the potential of targeting host metabolic pathways for innovative host-directed therapy (HDT) against intracellular bacterial infections. Through gene expression profiling in human macrophages, pyruvate metabolism was identified as potential key pathway involved in Salmonella enterica serovar Typhimurium (Stm) infections. Next, the effect of targeting pyruvate dehydrogenase kinases (PDKs) – which are regulators of the metabolic checkpoint pyruvate dehydrogenase complex (PDC) – on macrophage function and bacterial control was studied. Chemical inhibition of PDKs by dichloroacetate (DCA) induced PDC activation and was accompanied with metabolic rewiring in classically activated macrophages (M1) but not in alternatively activated macrophages (M2), suggesting cell-type specific effects of dichloroacetate on host metabolism. Furthermore, DCA treatment had minor impact on cytokine and chemokine secretion on top of infection, but induced significant ROS production by M1 and M2. DCA markedly and rapidly reduced intracellular survival of Stm, but interestingly not Mycobacterium tuberculosis, in human macrophages in a host-directed manner. In conclusion, DCA represents a promising novel HDT compound targeting pyruvate metabolism for the treatment of Stm infections.


INTRODUCTION
The rising prevalence of antimicrobial drug-resistance poses a serious threat to the control of bacterial infectious diseases. Drug-resistance among common bacterial infections continues to cause an estimated mortality rate of 700,000 people annually (1). Two major examples of antibiotic-resistant pathogens are Mycobacterium tuberculosis (Mtb) and Salmonella enterica. The World Health Organization estimated that around 10 million people developed active tuberculosis (TB) in 2017, and that 3.5% of new TB cases and 18% of previously treated TB cases were infected by rifampicin-resistant Mtb strains or multidrugresistant Mtb-strains (2). S. enterica is a common food-borne pathogen that can cause gastroenteritis and typhoid fever, the latter being responsible for an estimated 200,000 human deaths annually (3). Non-typhoidal S. enterica caused invasive disease in 535,000 people in 2017 and two highly invasive drug-resistant S. enterica serovar Typhimurium (Stm) strains have been described (4)(5)(6). Drug-resistance in intracellular pathogens such as Mtb and Stm is a major concern to global health and poses a critical need for developing new and more effective treatments against such bacteria.
Host-directed therapy (HDT) aims at augmenting host immune responses, mostly as adjunct therapy to currently used antibiotics. HDT may not only provide more effective treatment regimens but may also help shortening current lengthy treatment regimens and thus reducing risk of de-novo resistance mutations as well as reducing antibiotics-induced adverse effects (7). The development of novel HDT, however, requires a better understanding of the biology of intracellular infections and bacterial survival tactics. Recent studies have connected regulation of host defense against bacteria with host metabolic pathways (8). Rapid upregulation of glycolysis in macrophages is a hallmark of Mtb infection and LPS stimulation, and resembles the Warburg effect seen in cancer cells, but the mechanisms underlying this metabolic remodeling in bacterially-infected host cells remain poorly understood. Although multiple studies have suggested that rapid upregulation of glycolysis favors host immunity by meeting urgent energy demands required to induce protective immune responses (9)(10)(11)(12)(13)(14)(15)(16), more recent studies have shown that tricarboxylic acid (TCA) cycle intermediates, including itaconate, citrate and succinate, are associated with host immune responses against intracellular pathogens as well (17)(18)(19)(20)(21)(22)(23)(24). Host metabolic pathways and immune control of intracellular pathogens are thus interconnected in a highly complex manner. The specific immunometabolic pathways underlying macrophage-mediated inflammation and bacterial control remain largely unresolved. This particularly applies to infection-induced metabolic remodeling of human macrophages, which is significantly underrepresented in the literature. In the current study, we therefore modelled host-pathogen interactions in two different, polar human macrophage subsets which are known to respond differently to external stimuli. Classically activated macrophages (M1) display a pro-inflammatory phenotype and function in response to pathogen-associated molecular patterns (PAMPS) whereas alternatively activated macrophages (M2) display an anti-inflammatory phenotypes and function in response to PAMPS (25,26). Although the metabolic divergence between M1 and M2 is well described for murine macrophages, surprisingly little is known about human monocyte-derived macrophages (27)(28)(29).
Here, we identify the metabolic impact of infections with intracellular bacteria Stm and Mtb on primary human macrophages and report the prominent effect of the metabolic checkpoint pyruvate dehydrogenase complex (PDC) on Stm but not Mtb infections. PDC catalyzes the conversion of pyruvate into acetyl-CoA to supply the TCA cycle and is a gate-keeping enzyme regulating the balance between glycolysis and oxidative phosphorylation (OXPHOS). PDC consists of three subunits: pyruvate dehydrogenase (PDH, subunit E1), dihydrolipoamide acetyltransferase (subunit E2) and dihydrolipoamide dehydrogenase (subunit E3). PDH activity is the rate-limiting step during the conversion of pyruvate into acetyl-CoA (30). PDH activity is blocked upon phosphorylation by pyruvate dehydrogenase kinases (PDKs) and restored upon dephosphorylation by pyruvate dehydrogenase phosphatases (PDPs) (31). PDK inhibitors have been developed for treating cancer and type II diabetes (32)(33)(34).
In the current study, we report that PDKs are significantly modulated during Stm infection in primary human macrophages at the transcriptional level. Dichloroacetate (DCA), a chemical inhibitor of PDKs, activated PDC in M1 and M2 and induced a metabolic shift from glycolysis to OXPHOS in M1. DCA markedly increased production of reactive oxygen species (ROS) by M1 and M2. DCA induced a significant reduction of intracellular Stm (but not Mtb) in human macrophages, but did not exert a direct antibiotic effect on Stm extracellular bacteria, indicating that DCA induces Stm control via modulation of host metabolic pathways. In conclusion, DCA represents a promising novel HDT compound targeting pyruvate metabolism for the treatment of difficult to treat Stm infections.

Infection-Induced Glycolysis in Primary Human Macrophages
Although LPS stimulation and bacterial infections have been shown to shift cellular metabolism towards a more glycolytic phenotype in various cell types (9)(10)(11)(12)(13)(14)(15)(16), a number of studies have reported that human macrophages do not increase glycolysis in response to bacterial stimuli (35,36). To assess the potential of targeting host metabolic pathways to treat intracellular bacterial infections through HDT, we evaluated the effect of Stm and Mtb on cellular metabolism in primary human macrophages. Extracellular acidification rates (ECAR) and oxygen consumption rates (OCR) were measured as indicators of glycolytic and OXPHOS energy phenotypes, respectively, using a Seahorse XFe96 Extracellular Flux Analyzer(experimental outline in Figure 1A). Macrophages were pretreated with bacterial lysates, which we used as a proxy for live infections,  Figure 1D).
To assess modulation of metabolism by live bacteria, extracellular lactate levels (as an indicator of glycolysis) were determined in macrophages infected with live Stm and Mtb ( Figure 1E). M1 and M2 secreted significantly higher levels of lactate upon Stm infection compared to uninfected cells, supporting infection-induced glycolysis. Live Mtb infection significantly increased lactate production in M2 compared to uninfected cells, but surprisingly not in M1. The absence of lactate production by Mtb-infected M1 corroborates recent findings in mouse bone marrow-derived macrophages (BMDMs) and human monocyte-derived macrophages that Mtb is able to actively suppress induction of glycolysis (37,38).
To confirm induction of a metabolic shift at the transcriptional level, we studied mRNA expression levels of the glycolytic key gene hexokinase 2 (HK2) and the TCA cycle key genes succinate dehydrogenase complex flavoprotein subunit A (SDHA) and isocitrate dehydrogenase 2 (IDH2) in M1 and M2. HK2 mRNA levels were significantly increased 4 hours post Stm infection in M2 ( Figure 1F), but were not altered in Stm-or Mtbinfected M1( Figure 1B). SDHA and IDH2 mRNA levels were downregulated in Stm-and Mtb-infected M1 and M2, albeit to different extents, suggesting decreased activity of the TCA cycle pathway. Taken together, our data suggest that Stm infections are able to shift human M1 and M2 metabolism towards a more glycolytic phenotype.
To study whether the increased glycolysis contributes to host defense, we first treated Stm-and Mtb-infected M1 and M2 with glycolysis inhibitor 2-deoxy-D-glucose (2-DG), since earlier reports showed that 2-DG increased the burden of Mtb in murine BMDMs and in mouse lungs (39,40). PKA inhibitor H-89 was used as positive control and in agreement with previous findings (41,42) significantly lowered outgrowth of Stm, while being less effective in reducing Mtb bacterial load (Figures 2A, B). Surprisingly, treatment of human M1 with 2-DG significantly decreased outgrowth of intracellular Stm, but only decreased intracellular Mtb in M2 at the highest concentration (5 mM). Importantly, 2-DG did not show cytotoxicity in both macrophage subsets ( Figure 2C). The finding that 2-DG treatment did not (M1) or only moderately (M2) affect Mtb bacterial load in primary human macrophages contrasts with earlier reports, which concluded that 2-DG treatment increased the burden of Mtb in murine BMDMs and in mouse lungs (39,40). However, 2-DG has recently been described to decrease Legionella pneumophila survival independently of inhibition of glycolysis in mouse BMDMs (43) and also, high concentrations of 2-DG have been shown to inhibit not only glycolysis but also OXPHOS in mouse BMDMs (44). Therefore, in order to be able to evaluate the effect of glycolysis independently from the use of 2-DG, we determined bacterial survival after substitution of D-glucose by D-galactose in the cell culture medium. Substitution of D-glucose by D-galactose reduces glycolytic flux and thus could independently validate our findings (44). Indeed, the absence of glucose significantly decreased Stm outgrowth in M1, but not M2, and tended to increase intracellular Mtb levels ( Figures 2D,  E). These data suggest that inhibition of glycolysis impacts Stm and Mtb in opposite directions; it can inhibit Stm outgrowth, while it can promote Mtb survival as described (39,40). Also of importance, these data suggest that pathways other than inhibition of glycolysis are responsible for the antibacterial effect of 2-DG.

Gene Expression Analysis Identifies Pyruvate Dehydrogenase Kinases as Potential Host Targets for HDT During Stm and Mtb Infections
Next, we aimed to identify metabolic key genes that are modulated during Stm and Mtb macrophage infections. We performed computational analysis of published RNAseq data by Blischak et al. (45) to study the expression of metabolic genes (i.e. all genes in Reactome pathways 'Glucose metabolism' and 'TCA cycle and respiratory electron transport') in human M2 macrophages infected for 18 hours with Stm or Mtb (MOI of 5) compared to their time-matched controls (mock infection). Stm infection had a more prominent effect on the regulation of glycolytic genes and genes involved in the TCA cycle and respiratory electron transport compared to Mtb (Supplemental Figure 2). Two key glycolytic genes, enolase (ENO) 2 and HK2, were significantly upregulated 18 hours post Stm or Mtb infection. In contrast, most genes involved in the TCA cycle and respiratory electron transport were downregulated upon Stm and Mtb infection.
We hypothesized that metabolic genes that are modulated during Stm and Mtb infections may provide novel targets for HDT. Surprisingly, among the most up-and downregulated metabolic genes during Stm and Mtb infection, two PDK isoforms were regulated in opposite directions: PDK4 was strongly downregulated upon Stm infection, whereas PDK1 was strongly upregulated upon Mtb infection (Supplemental Figure 3). The differential expression of PDK isoforms is interesting, because PDKs are regulators of the metabolic checkpoint PDC, linking glycolysis with TCA cycle activity and OXPHOS ( Figure 3A). More detailed kinetic analysis of the  Figure 3C). A similar trend could be observed in Mtb-infected macrophage subsets, albeit not significant ( Figure 3D). Taken together, these data indicate that pyruvate metabolism is markedly modulated during Stm infections and to a lesser extent during Mtb infections in human M1 and M2, as shown by a downregulation of PDK2, PDK4 and PDP2 at the transcriptional level.

Dichloroacetate Can Be Used as a Tool to Study the Effect of Pyruvate Dehydrogenase Kinases on Intracellular Stm and Mtb Control in Human Macrophages
To study the effect of pyruvate metabolism during Stm and Mtb infections in more detail, we aimed to modulate this pathway through chemical inhibition of PDK using dichloroacetate (DCA). To validate inhibition of PDKs as the mechanistic target of DCA, we first examined whether DCA treatment in Stm and Mtb-infected macrophage subsets resulted in dephosphorylation of PDH at serine residues 293 and 300, which corresponds with increased PDC activation. Pyruvate, a natural PDK inhibitor, was added as a positive control for PDH dephosphorylation. Quantification of the western blot protein band intensities revealed that both DCA treatment and addition of pyruvate resulted in PDH dephosphorylation at serine residues 293 and 300 in Stm and Mtb-infected M1 and M2 compared to DMSO ( Figure 4A). In summary, these data  Since DCA has been reported to shift metabolism from glycolysis to OXPHOS in various cell lines (32,46,(48)(49)(50), we next evaluated whether DCA could reverse the Stm-and Mtb-induced metabolic shift towards glycolysis that was predominantly observed in human M1 macrophages ( Figure 1C). First, we studied gene expression levels of metabolic key genes (HK2, PDK2, PDK4, IDH2, SDHA) and downstream targets of succinate accumulation (HIF1A and VEGF) (19,22,51), as an indication of TCA cycle activity,  in human macrophages ( Figure 4B). DCA did not significantly modify metabolic gene expression at the transcriptional level, but VEGF was significantly increased by DCA in Stm-infected M1, in accordance with increased TCA cycle activity ( Figure 4C). Next, we studied the effect of DCA on OCR and ECAR levels in macrophages that were prestimulated with Stm or Mtb lysate ( Figure 4D). DCA treatment resulted in increased OCR levels and decreased ECAR levels in M1, consistent with a classical  DCA-induced shift from glycolysis to OXPHOS ( Figure 4E). Strikingly, a shift from glycolysis to OXPHOS could not be detected in M2, indicating that PDC activation does not necessarily lead to glycolytic reprogramming and that this is cell-type specific.
In conclusion, DCA can be used as a chemical tool to study the effect of PDC activation on Stm and Mtb infections in human macrophages, which is accompanied by metabolic rewiring from glycolysis to OXPHOS in a M1 cell type specific manner.

DCA Treatment Has Minor Impact on Cytokine and Chemokine Production, but Induces ROS Production by Human Macrophages
To study whether PDC activation can modulate effector functions of human macrophages, we assessed the production of 41 chemokines/cytokines by multiplex beads assay. Clearly, Stm-infection markedly induced the production of proinflammatory cytokines, anti-inflammatory cytokines and chemokines, which contrasted with the lower magnitude changes seen for Mtb-infected macrophages ( Figure 5A and Supplemental Table 1). DCA treatment did not have any major impact on overall chemokine/cytokine secretion profiles of both M1 and M2, although several individual chemokines and cytokines were significantly modified by DCA in M1 and M2, dependent on the cell-type and the infection type (Stm versus Mtb) ( Figure 5B and Supplemental Table 1).
Since PDC activation by DCA can promote OXPHOS, which is associated with mitochondrial ROS production (46, 52), we next assessed the effect of DCA on ROS production by M1 and M2. Interestingly, DCA significantly induced ROS production in both M1 and M2 and even was a more potent ROS inducer compared to positive control phorbol 12-myristate 13-acetate (PMA) (Figures 5C, D). Surprisingly, ROS production was increased immediately after administration of DCA, even before the first measurement could be taken. ROS production was still observed 1h post DCA stimulation in M1, but reached plateau levels in M2, suggesting that DCA-induced ROS production was halted. DCA-induced ROS production could be partially inhibited by N-acetylcysteine (NAC), which is a scavenger of ROS.

DCA Treatment Induces Intracellular Stm Control in Human Macrophages
Since we showed that PDC is a target of interest for HDT and DCA activated PDC, resulting in altered metabolism and increased ROS production, we investigated whether DCA could impact bacterial control by human macrophages. Macrophages were treated with DCA or with H-89 as positive control. As expected, H-89 markedly decreased Stm and tended to decrease Mtb in M1 and M2 (41). Exposure of Stm-and Mtbinfected M1 and M2 to 5-20 mM DCA resulted in a profound decrease in Stm but not Mtb outgrowth ( Figure 6A). Importantly, cell viability was not affected ( Figure 6B). Interestingly, Stm outgrowth was significantly more inhibited by DCA treatment in M1 compared to M2, which is in agreement with the more profound effect of DCA on M1 metabolism ( Figure 4E). Furthermore, DCA decreased the percentage of Stm-infected M1 and M2 cells compared to DMSO as shown by flow cytometry using DsRed-expressing Stm (Supplemental Figure 4A). Interestingly, the effect of DCA on inhibiting intracellular Stm outgrowth was already significant after 2 hours of DCA treatment in M2 and after 4 hours in both macrophage subsets, suggesting that DCA induces a host protective response during the early phase of infection (Supplemental Figure 4B).
Not unexpectedly, DCA had little effect on intracellular Mtb in M1 and M2 ( Figure 6A). Even a prolonged exposure (up to 72h) of Mtb-infected macrophages to DCA did not result in reduction of Mtb CFU levels (Supplemental Figure 4C), suggesting that Mtb modulates host-signaling pathways in a manner different from Stm, that cannot be redressed by DCA. Potential direct antibiotic properties of DCA were excluded by adding DCA to Stm and Mtb cultures in the absence of human cells, showing that DCA acts via host-directed pathways in human macrophages to control intracellular Stm ( Figure 6C).
Taken together, our data suggest that DCA inhibits intracellular Stm infection, but not Mtb infection, in a hostdirected manner. DCA reduced Stm levels rapidly after administration, suggesting that DCA is a potent molecule for treatment against Stm.

DISCUSSION
Physiological metabolic modulation of phagocytes is essential to effective host control of intracellular bacterial pathogens (15). Here, we show that human M1 and M2 adapt their metabolic profile upon Stm and Mtb lysate stimulation and Stm infection by increasing glycolysis. We identified metabolic key genes, PDKs, as novel targets for HDT against Stm and report DCA as a functional chemical tool to study PDK inhibition in the context of bacterial infections. Furthermore, we report cellspecific effects of DCA on host cell metabolism and chemokine and cytokine production and we show major and rapid effects of DCA on Stm outgrowth in infected human macrophages. DCA has been described to induce bacterial clearance in the peritoneum of septic mice (53), and to our knowledge, this is the first report of a host-directed anti-bacterial effect of DCA on Stm.
Rapid upregulation of glycolysis is a hallmark of intracellular bacterial infections and is thought to rapidly provide energy to sustain biosynthesis of inflammatory molecules during the initial infection stage (9)(10)(11)(12)(13)(14)(15)(16). However, most studies have been performed in mouse models, and the number of studies that assessed induction of glycolysis by bacterial infections in human cellular models is limited. Only few studies have reported that human macrophages do not increase glycolysis in response to LPS and intracellular trypanosomatid parasites (Leishmania donovani, L. amazonensis and T. cruzi) (35,36). In the current study, M1 and M2 subsets increased their glycolytic metabolism upon infection with Stm, as demonstrated by increased  (27)(28)(29)54). In fact, in support of our data, LPS stimulation enhanced the glycolytic pathway already after 3 hours in human M1 and M2 macrophages that were differentiated from monocytes by using M-CSF and IFNg or M-CSF, respectively (55). Moreover,    intracellular Mtb (39). Moreover, mice treated in vivo with 2-DG had higher Mtb bacterial burdens in their lungs after 14 days (40). Although 2-DG is classically reported to inhibit glycolysis, it also inhibits OXPHOS when used in high concentrations that are similar to the ones in our study (up to 10 mM) (44). Furthermore, the inhibitory effect of 2-DG on Stm and Mtb infection could be dependent on other pathways than its impairment of the glycolytic pathway or OXPHOS, as has already been suggested for Legionella pneumophila (43). In agreement with others, we thus propose caution in the interpretation of data generated using 2-DG in the context of glycolysis (44,57,58). Alternatively, substitution of glucose for galactose can provide a more reliable method to inhibit glycolysis (44). In our work substitution of D-glucose for D-galactose tended to increase intracellular Mtb, in agreement with reported data in mouse BMDM (39), and in line with the hypothesis that glycolysis is required for protective host responses against Mtb. Importantly, substitution of D-glucose for D-galactose significantly decreased intracellular levels of Stm, suggesting that the classical "host-protective" role of glycolysis during bacterial infections does not apply to Stm in human macrophages. PDKs are considered gate-keeping kinases between glycolysis and OXPHOS and chemical inhibition of PDKs by DCA resulted in a significant reduction of intracellular Stm outgrowth in human macrophages. DCA is a potent small molecule in clinical use for the treatment of lactic acidosis and metabolic disorders (59) and has gained serious interest as anti-cancer therapeutic in the past decade (60). DCA has passed phase I/II toxicity testing in humans, which can accelerate the translation towards to clinic application for other diseases. Administration of DCA displayed a broad spectrum regarding toxicity and safety, ranging from well-tolerance (61) and common gastrointestinal side-effects (62), to severe side-effects like peripheral neuropathy (63,64). Despite reported toxicities, DCA has not lost interest as potential therapeutic, and severe side-effects may be suppressed with adjuvant therapy (65). Tolerance of DCA in patients infected with Salmonella remains to be determined and is crucial for its applicability as HDT compound for the treatment of Salmonella infections, which has a mild disease course in most patients, but can be persistent and even lethal in others (66,67). Building upon this potential, we have identified PDKs as potential novel host targets for HDT to treat Stm infections. Of note in this context, several novel, potentially safer, DCA derivatives have been synthesized recently (59).
DCA classically induces a metabolic shift from glycolysis to OXPHOS, which was reported in cancer cell lines (32,46,(48)(49)(50) and mouse embryonic stem cells (68). Similarly, DCA significantly decreased ECAR levels in human macrophages differentiated with M-CSF or M-CSF with IFN-y (55). Here, we report a similar result using DCA in M1 stimulated with Stm or Mtb lysate, but not in M2. Differences in ECAR levels between M-CSF-derived macrophages published by Chiba et al. and our M-CSF-derived M2 may be explained by differences in DCA concentrations tested (20 mM versus 10 mM in our study), or by the use of pre-stimulation with Stm or Mtb lysate in our study. In addition to decreased ECAR levels, we observed increased OCR levels immediately after DCA treatment in M1, indicative of a metabolic shift from glycolysis to OXPHOS. Interestingly, we observed increased levels of VEGF mRNA levels in Stm-infected M1 and not in M2. Since succinate has been associated with HIF-1a stabilization, we hypothesize that the increased VEGF mRNA levels are a consequence of increased flux into the TCA cycle leading to succinate accumulation. Thus, importantly, we show that DCA induces cell-type specific responses, but not general metabolic rewiring in all cell types. Since DCA-induced metabolic effects were only observed in M1 and not in M2, we propose that improved control of Stm was not caused by a metabolic shift from glycolysis to OXPHOS. While not investigated here, DCA treatment may have increased or decreased accumulation of specific TCA cycle metabolites that modulate innate immune responses of macrophages to Stm, such as succinate (69), or may have deprived the cells from host carbon sources that can be exploited by Stm, for instance lactate (70).
In previous studies, DCA induced the functional polarization of murine macrophages towards an anti-inflammatory phenotype based on cytokine production profiles (71,72). Similarly, in human cells, DCA decreased IL-1b and IL-10 levels in LPS-stimulated human THP-1 cells (73) and IL-6 levels in LPS-stimulated human macrophages (55). In contrast with these studies, we did not observe significant DCA-inhibiting effects on IL-1b, IL-10 and IL-6 secretion by Stm-or Mtbinfected human macrophages, possibly because infection alone already strongly skewed the cytokine response ( Figure 5A). Generally, we observed that the effect of DCA on chemokine/ cytokine secretion was dependent on both cell-type (M1 versus M2) and infection (Stm versus Mtb). We therefore conclude that, based on their cytokine secretion profile, DCA was not able to (re)polarize Stm-or Mtb-infected human M1 and M2 towards an anti-inflammatory phenotype.
Potential mechanism by which DCA induces Stm control in human macrophages is through increasing ROS production, as DCA clearly induced ROS in both M1 and M2. In agreement with our data, DCA increased intracellular ROS production by tumor cells, head and neck cancer cells and multiple myeloma cells (46,52,74). Although ROS production appears a host strategy to cope with Stm (75-77), it can also promote Stm survival, and thus ROS acts as a double-edged sword in the perspective of the host and Stm (78). While DCA-induced ROS production has been suggested to originate from increased OXPHOS activity in the mitochondria (46,74), we cannot not exclude other sources of ROS production based on our data. Increased ROS production in M2, despite no detectable changes in OXPHOS in these cells, in fact might suggest that ROS may have originated from other sources, for instance NADPH oxidase-dependent ROS production of phagosomal origin (79). Given that the type III secretory system of Salmonella interferes with host NADPH oxidase trafficking to the phagosome [as reviewed in (80)] to evade ROS-induced killing by phagocytes, we speculate that NADPH localization may be restored by DCA treatment. This effect of DCA on NADPH oxidase localization and the effect of ROS production on intracellular Stm control requires future investigation.
In the current study, we showed that PDK expression is modulated during Stm infections and can be targeted by DCA to improve host control against Stm infections. Furthermore, we evaluated effects of DCA on Stm from a host perspective. Whether the effect of DCA on Stm control is caused by limited pyruvate access as carbon source remains to be determined, but may explain why DCA did not decrease intracellular Mtb levels. Together, our data highlight differences in survival of two pathogenic bacteria that can use the same host niche, and show the clinical potential of DCA as novel HDT treatment against Stm.

Seahorse Real-Time Extracellular Flux Analysis
Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were determined as a measure of mitochondrial respiration and glycolysis, respectively, by using the Agilent Seahorse XFe96 Extracellular Flux Analyzer (Seahorse Bioscience) according to manufacturer's instructions and as described before (82). Briefly, cells were plated in a Seahorse XF cell culture microplate at a density of 30,000 cells/well and incubated overnight at 37°C/5% CO 2 in RPMI 1640 medium (10% FBS, 2 mM L-Alanyl-L-Glutamine). Cells were stimulated with 1 or 10 µg/ml Stm or Mtb lysate and incubated for 4 hours at 37°C/5% CO 2 . Plates were washed with glucose-free RPMI 1640 medium (Sigma-Aldrich) containing 5% FBS and incubated for

Macrophage Infection
Ongoing

Lactate Assay
Cell supernatants were harvested and kept at -20°C until further use. 10

CFU Assay
Intracellular bacterial burden was determined by colony forming unit (CFU) assays as described previously (81). Cells were washed once with PBS and lysed in 0.05% sodium dodecyl sulfate (SDS) solution (Thermo Fisher Scientific). Serially diluted cell lysates were plated on LB agar (Stm) or on 7H10 agar containing 10% OADC and 5% Glycerol (Mtb). Plates were incubated at 37°C and CFUs were counted from triplicate wells.

Cellular Toxicity Assay
The number and percentage of dead cells based on plasma membrane integrity of the adherent cell population was quantified by analysis of microscopy images. Cells in 96-well flat bottom plates (30,000 cells/well) were stained with 2 µg/ml propidium iodide (PI, Sigma-Aldrich) and 2 µg/ml Hoechst 33342 (H3570, Sigma-Aldrich) in 40 µl/well phenol red-free RPMI 1640 (Sigma-Aldrich) supplemented with 10% FBS and 2 mM L-Alanyl-L-Glutamine and incubated for 5 minutes at room temperature (RT). Cells were imaged using a Leica AF6000 LC fluorescence microscope (Leica Microsystems, Wetzlar, Germany) combined with a 10x dry objective. Total and dead cell numbers were quantified by respectively counting the nuclei and the number of PI-positive cells using Fiji/ImageJ software (83).

Cell-Free Bacterial Growth Assay
Compounds were diluted in Difco ™ LB broth or in Difco ™ Middlebrook 7H9 broth and added (100 µl/well) to 100 µl/well bacterial culture (OD 600 of 0.1) in flat bottom 96-well plates. Gentamycin (50 µg/ml) and rifampicin (20 µg/ml) were added as positive controls to inhibit Stm and Mtb growth, respectively. Absorbance was measured directly after plating and after 16

Reactive Oxygen Species (ROS) Assay
Procedure was performed as described previously (81). Cells (40,000 cells/well) were seeded in black ultra-thin clear flat bottom 96-well plates (Corning) in RPMI 1640 (10% FBS, 2 mM L-Alanyl-L-Glutamine) and incubated overnight at 37°C/ 5% CO 2 . Cells were washed with 150 µl PBS/well and RPMI 1640 without phenol red was added and background fluorescence was measured using a SpectraMax i3x plate reader (l ex 493 nm and l em 522 nm) at 37°C. Cells were washed with 150 µl PBS/well. 50 µl/well of 5 µM CM-H2DCFDA, a general ROS probe, in RPMI 1640 without phenol red was added for 30 minutes at 37°C/5% CO 2 . Cells were washed with 100 µl PBS/well and 100 µl/well RPMI 1640 without phenol red was added containing DMSO (0.1% v/v), DCA (10 mM) or PMA (300 nM) as positive control. ROS production was measured kinetically for 60 minutes at 37°C by measuring fluorescence at 522 nm using the SpectraMax i3x.

Cytokine and Chemokine Secretion by Enzyme-Linked Immuno Sorbent Assay (ELISA) and Multiplex Beads Assay
Cell supernatants were collected and sterilized by centrifugation using a 96-well filter plate containing a 0.2 µm PVDF membrane (Corning). IL-12p40/p70 production was quantified using a human IL-12p40/p70 Enzyme-Linked Immuno Sorbent Assay

Statistical Analysis
Data were tested with a paired sample t-test when comparing two groups and RM one-way ANOVA followed by Dunnett's multiple comparison test, unless mentioned otherwise, when comparing three or more groups. Statistical tests were considered significant when p < 0.05 at 95% confidence interval. All statistical analyses were carried out using GraphPad Prism 8 software (Graphpad Software, San Diego, CA, USA).

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Institutional Review Board of the Leiden University Medical Center, The Netherlands. The patients/participants provided their written informed consent to participate in this study.

AUTHOR CONTRIBUTIONS
TO and CD devised the project and the main conceptual ideas. MH and FV contributed to the design and implementation of the research. CD, GS, JE, SV, and MH performed the experiments and analyzed the experimental data. KW contributed to preparation of materials. CD and TO wrote the manuscript. TO supervised the project. All authors contributed to the article and approved the submitted version.

FUNDING
The study received support from the Netherlands Organization for Scientific Research (NWO-TOP Grant Agreement No. 91214038); and the NWO-TTW NACTAR grant (Grant Agreement No. 16444). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.