Mycobacterium tuberculosis (Mtb) H37Rv was the mycobacterial strain used in the study (Borah et al., 2019a,b). The THP-1 monocytic cell line was used as the human macrophage model used in this work (Borah et al., 2019a,b).

Mtb H37Rv was cultured using Middlebrook 7H11 agar (Merck) and Middlebrook 7H9 broth (Merck) containing 10% (vol/vol) oleic acid-albumin-dextrose-catalase enrichment medium supplement (OADC) (Becton Dickenson) and 0.5% (vol/vol) glycerol (Merck). Mtb was streaked onto brain heart infusion agar (Merck) plates to check the purity of cultures prior to infection assays.

THP-1 monocytic cell line was grown in RPMI 1640 growth media (Merck) supplemented with 10% heat inactivated fetal bovine serum (FBS) (Merck) at 37°C, 5% CO₂ and 95% humidity. For ¹³C labelling experiments, modified RPMI growth media (Gibco™ 11,879,020) without glucose was prepared by adding 100% [U-¹³C₆]glucose (w/v, CK Isotopes).

Isotopic labelling experiments were conducted by growing THP-1 macrophages in RPMI media containing 100% [U-¹³C₆]glucose. Briefly, monocytes were grown in unlabelled RPMI media followed by differentiation into macrophages for three days using 50 nM Phorbol 12-myristate 13-acetate (PMA) in 175 cm² tissue culture flasks and in 6 well plates. Around 3 × 10⁷ and 1 × 10⁶ macrophages were cultivated in 175cm² flasks and 6 well plates, respectively.

THP-1 monocytes were differentiated into macrophages using PMA in RPMI media containing 100% [U-¹³C₆]glucose as described in the previous section. On the day of infection, macrophages were cultivated either in 175 cm² tissue culture flasks or 6 well plates and washed with phosphate buffered saline (PBS) supplemented with 0.49 mM Mg2+ and 0.68 mM Ca2+ (PBS⁺) three times prior to infection (Borah et al., 2019a). Mtb broth cultures were grown in 7H9 media with 10% OADC and 0.5% glycerol. Exponentially growing Mtb cultures were washed three times with PBS and added to macrophages at a multiplicity of infection (MOI) of 5. Macrophages infected with Mtb were incubated at 37°C, 5% CO₂ and 95% humidity for 3–4 h, followed by washing the macrophages with PBS⁺ three times. Fresh media with 100% [U-¹³C₆]glucose was added to infected macrophages and left for 48 h prior to cell harvest for metabolomics and isotopomer analysis. Uninfected macrophages were washed and incubated for 48 h in 100% [U-¹³C₆]glucose containing RPMI as controls.

After 48 h of infection, uninfected and infected macrophages cultivated in 6-well plates were washed twice with PBS and cells were scraped. Cells were spun at 300 g for five minutes. 6 M hydrochloric acid (HCl) was added to the cells and boiled for 10 min (Borah et al., 2019a). Amino acid hydrolysates were prepared from cell lysates by incubating the lysates in 6 M HCl at 100°C for 24 h. Gas-chromatography mass spectrometry (GC–MS) metabolomics was performed to measure amino acid pool sizes and for mass isotopomer analysis. Amino acid hydrolysates were dried and derivatized using pyridine and tert-butyldimethyl silyl chloride (TBDMSCl) (Merck) (Borah et al., 2019a). Amino acids were analyzed using a VF-5 ms inert 5% phenyl-methyl column (Agilent Technologies) on GC–MS. Mass spectra were extracted using chemstation GC–MS software. Mass spectrometry datasets were corrected for natural isotope effects using the Isotopomer network compartmental analysis (INCA) platform (Young, 2014; Borah et al., 2021a). Average ¹³C in an amino acid was calculated from the fractional abundance of the mass isotopomers in the amino acid fragments. The corrected mass isotopomer distributions (MIDs) for non-essential amino acids and organic acids were included into the macrophage metabolic model for flux estimation using ¹³C-MFA.

Biomass measurements were calculated for uninfected and infected macrophages. 3 × 10⁷ macrophages cultivated in 175 cm2 tissue culture flasks were used for biomass assays. For calculating dry cell weight, macrophage samples were distributed into eppendorfs, freeze dried and weighed. Dry cell pellets were used to measure protein using BCA (bicinchoninic acid) protein determination kit (Merck), lipids using lipid Assay Kit (Abcam ab242305), total DNA using Qubit dsDNA BR Assay (Fisher 10,146,592), total RNA using QUANT-IT RNA BR assay kit (Fisher 10,266,793) and carbohydrates using carbohydrate kit (Merck, MAK104). Amino acids were quantified using GC–MS method as described in the previous section.

Metabolic flux estimations were performed using INCA, version 2.2 (Young, 2014). The MIDs included in the model for flux estimation are provided in Supplementary File S2. Uninfected and infected macrophages were at pseudo-steady state (metabolic and isotopic) after 48 h of ¹³C-labelling and Mtb-infection (Beste et al., 2013; Borah et al., 2019a). Fluxes were estimated using a non-linear weighted least squares fitting approach to calculate net and exchange flux distributions that were the best-fit for generation of the isotopic labelling data and biomass constraints (Young, 2014; Borah et al., 2021a). A random initial guess and multistart optimisation approach with 200 restarts was used for flux estimations. Flux profiles with the minimum statistically acceptable SSR were considered the best-fit. The goodness of fit of the flux maps was assessed by comparing the simulated and experimental measurements. The uncertainties in the flux estimations were checked using the Monte Carlo analysis function built into INCA. The confidence intervals and frequency distribution plots were calculated using Monte Carlo analysis, with significance level α = 0.05, 100 trials per iteration and a relative error tolerance of 0.005.

Statistical analysis including chi-square test and Monte Carlo analysis was performed using INCA and t-tests were calculated using Graphpad version 8.2.1.

Metabolic modelling of a macrophage was conducted using INCA, version 2.2 (Young, 2014). The isotopomer model consists of a central carbon metabolic network with 57 net flux reactions and 25 exchange fluxes (Supplementary File S1). Metabolic reactions include glycolysis, gluconeogenesis, the PPP, the TCA cycle, anaplerosis, biomass reaction and amino acid synthesis (glutamate, glutamine, proline, arginine, aspartate, asparagine, alanine, serine, glycine, cysteine, and tyrosine). The biomass reaction used to constrain the model was constructed from the experimentally measured biomass proportions of uninfected and infected macrophages (Figure 1). The network also consists of uptake reactions for glucose and amino acids: glutamine, alanine, glycine, serine, tyrosine, aspartate, asparagine, cysteine, arginine, glutamate, and proline. The glucose uptake flux was fixed at 100 and fluxes were estimated relative to the glucose uptake rate.

This section outlines the methodology for conducting ¹³C-MFA in THP-1 macrophages (uninfected or infected with Mtb H37Rv) (Figure 1). [U-¹³C₆]glucose was chosen as the isotopic substrate for the labelling experiment. THP-1 monocytes were differentiated into macrophages with PMA in RPMI growth media containing 100% [U-¹³C₆]glucose for 72 h. We did not find any significant increase in ¹³C incorporation into the biomass of macrophages when monocytes were cultivated in [U-¹³C₆]glucose for 48 h prior to differentiation into macrophages. Hence, we chose to grow macrophages in ¹³C labelling media during differentiation and infection. Macrophages were uninfected (control) or infected with Mtb H37Rv for 48 h in ¹³C labelling media. The infection period of 48 h was chosen as the pseudo steady-state labelling period in infected macrophages, as established from our previously published work (Beste et al., 2013; Borah et al., 2019a). Uninfected or infected macrophages were harvested after 48 h for metabolomics and isotopic labelling analysis. In parallel, biomass composition analyses were conducted for uninfected and Mtb-infected macrophages grown in unlabelled RPMI media. The biomass composition was used to construct the biomass equation that constrained the isotopomer metabolic network model which was constructed to include the central carbon metabolic reactions for glycolysis/ gluconeogenesis, pentose phosphate pathway (PPP) and tricarboxylic acid (TCA) cycle. The model was constructed based on published literature and databases including HumanCyc and KEGG (Amaral et al., 2011; Duckwall et al., 2013; Antoniewicz, 2018; Lagziel et al., 2019). The isotopomer data from the labelling experiments were incorporated into the model and computationally fitted using the MATLAB-based isotopomer network compartmental analysis (INCA) platform to compute the best-fit flux maps.

Biomass measurements were used to constrain the metabolic model for flux estimations. We measured macromolecular biomass composition: nucleic acids, protein, lipids, and carbohydrates of uninfected and Mtb-infected macrophages (Figures 2A–E). Protein accounted for the highest proportion of cell dry weight (>50%), followed by lipids and carbohydrates (Figures 2F,G). There were no significant changes in total cellular lipids, protein, RNA and DNA upon Mtb infection. Total cellular carbohydrates in infected macrophages were significantly higher when compared to uninfected macrophages demonstrating the shift in carbohydrate synthesis upon Mtb infection (Figure 2E).

We quantified cellular amino acid pools using GC–MS metabolomics to identify changes induced in infected macrophages (Figure 2H). There were no significant differences in the pool sizes of essential amino acids (valine, leucine, isoleucine, methionine, threonine, lysine, histidine, tryptophan, phenylalanine) between the two groups as these are not synthesised in human cells but taken up from the growth media. Amongst the non-essential amino acids, proline and cysteine levels were significantly lower in infected macrophages. Cysteine is synthesised directly from serine; the pool size of cellular serine was over two-fold lower in infected macrophages. Aspartate and glutamate, the two primary precursors for amino acid biosynthesis did not change significantly upon Mtb-infection.

In this section, we present the ¹³C mass isotopomer distribution profiles for macrophage amino acids. The carbon backbone in amino acids is synthesised from metabolic intermediates generated from central carbon metabolism. The carbon backbone of alanine (Ala) and serine (Ser) are produced from glycolytic intermediates pyruvate and 3-phosphoglyceric acid, respectively. The ¹³C distribution profile for Ser is strikingly different for infected macrophages; the proportional distribution of ¹³C is highest in isotopomer family M3 for uninfected macrophages but M1 for infected macrophages (Figure 3). Quantitative analysis in Figure 4 shows that the fractional abundance or proportion of ¹³C in M2 were significantly higher for infected macrophages. Glycine (Gly) and cysteine (Cys) which were synthesised from Ser also showed higher proportions of M2 and M3 in infected macrophages. This was expected as these three amino acids are synthesised from the same carbon precursor derived through glycolysis. The proportion of ¹³C in Ala M3 is higher in uninfected macrophages (Figure 4). The differences in these amino acids indicate likely changes in carbon fluxes that generated respective metabolic precursors through glycolysis in infected macrophages. Tyrosine is derived from the PPP and glycolytic intermediates; the ¹³C distribution profile is significantly different between uninfected and infected groups (Figures 3, 4) indicating changes in fluxes through these central pathways. The TCA cycle derived intermediates fumarate (FUM) and malate (MAL) and amino acids aspartate (Asp) and glutamate (Glu) exhibited quantitatively distinct proportional ¹³C in infected macrophages (Figure 5). Although the total cellular pool sizes of Asp., Glu and Tyr showed no significant differences between uninfected and infected macrophages, the ¹³C isotopomer analysis was distinct between the two groups; this confirms that the measurement of pool sizes alone may not be an accurate indicator of changes in carbon fluxes through a pathway and ¹³C isotopomer analyses could provide more detailed information. The mass isotopomer distribution profiles obtained from the labelling experiments were incorporated into the metabolic model to estimate carbon fluxes through the central metabolic network, which are presented in the next section.

Metabolic fluxes for central carbon metabolism (CCM) were quantified using ¹³C-MFA (see Figure 1 for the overview of the steps for MFA). The details of reactions included in the isotopomer network model of a macrophage are provided in Supplementary File S1. Fluxes were resolved for the pathways involved in CCM: glycolysis/gluconeogenesis, PPP, and the TCA cycle (Supplementary File S3). The flux maps for uninfected and infected macrophages are shown in Figure 5. Fluxes were quantified relative to the glucose uptake rate set arbitrarily at 100. The uptake of glucose in macrophages has been reported to be mediated by GLUT1, a member of the SLC2 transporter family (Nonnenmacher and Hiller, 2018). In the cytosol, glucose is converted to glucose 6 phosphate (G6P) by hexokinase (Nonnenmacher and Hiller, 2018). Flux from G6P was partitioned to glycolysis via fructose 6 phosphate (F6P) and to PPP via 6-phosphogluconic acid (PG6). Reactions R4 to R10 of glycolysis carried the highest carbon fluxes relative to the PPP and the TCA cycle. The uninfected and infected macrophages profile exhibited oxidative TCA cycle fluxes. The TCA cycle fluxes were relatively lower than that in glycolysis and the PPP. Anaplerotic carbon metabolism was modelled between phosphoenolpyruvate (PEP) and oxaloacetate (OAA) and between pyruvate (PYR) and malate (MAL). The cytosolic and mitochondrial pools of OAA and MAL were lumped together to improve flux resolution in these reactions. Anaplerotic reactions carried the lowest carbon flux relative to the rest of the central pathways. We compared the best-fit carbon flux estimates between uninfected and infected macrophages and described the significantly altered pathways in the next section.

The statistical significance of the differences between the fluxes of Mtb-infected and uninfected macrophages was assessed using Monte Carlo analysis. The frequency distribution plots for glycolytic fluxes are shown in Figure 6 and Supplementary Figure S1. Glycolytic fluxes R4, R5, R6, R7, R8, R9 and R10 were significantly higher in infected macrophages. Glycolysis is a faster route for ATP production and generation of biosynthetic precursors, which are required to mount host defence responses against Mtb (Shi et al., 2016). We have selectively illustrated the statistical comparisons for fluxes R4, R6 and R10 in Figure 6. Flux R4 from G6P to F6P, the first step of glycolysis was significantly higher in infected macrophages. The plots for R4 are significantly different between uninfected and infected macrophages. For uninfected macrophages, the maximum probable and best-fit net flux value of R4 is 66.9, but for infected macrophages, this value is 94.7. Similarly, the plots for R6 (from fructose 1,6-bisphosphate (FBP) to glyceraldehyde 3-phosphate (GAP) and dihydroxyacetone phosphate (DHAP)) show the maximum probable best-fit flux of 98.2 for R6 in infected macrophages vs. 88.9 in uninfected macrophages. The comparisons for rest of the glycolytic fluxes R5, R7, R8 and R9 are shown in Supplementary Figure S1 which also shows significant changes in uninfected vs. infected macrophages. The plots for R10, the last step rate-limiting step of glycolysis from PEP to PYR catalysed by the pyruvate kinase enzyme show a maximum flux of 197.2 for infected macrophages vs. 188.4 in uninfected macrophages. Carbon flux from PYR is withdrawn for the synthesis of amino acids including ALA and lactate. The statistical resolution for these output fluxes was not refined, however the total flux withdrawn from PYR to synthesise ALA and lactate was higher in infected macrophages (195.19 in infected vs. 188.20 in uninfected macrophages).

The fluxes for both oxidative and non-oxidative branches of the PPP were significantly reduced in Mtb-infected macrophages. The plots for flux R11 (the first step of the PPP: G6P to PG6) and R12 (PG6 to ribulose 5-phosphate (RU5P)) show the best-fit flux value of 33 in uninfected macrophages vs. 5.2 in infected macrophages (Figure 7; Supplementary Figure S2). These oxidative PPP fluxes generate NADPH coenzymes that fuel cellular lipid synthesis. The fluxes through the non-oxidative branches of the PPP (R13, R14, R15, R16, R17, R18, R19) were also reduced in infected macrophages (Figure 7; Supplementary Figure S2). The plots for R13 and R14 in Figure 7 shows over a five-fold reduction of fluxes in infected macrophages and no overlap of the maximum value in the frequency distribution plots with that of uninfected macrophages.

We have demonstrated that infected macrophages have significantly higher glycolytic fluxes and reduced PPP fluxes. The flux partitioning between glycolysis and PPP was altered upon infection, channelling higher carbon catabolism through glycolysis. Glycolytic glucose catabolism generates adenosine triphosphate (ATP) and nicotinamide adenine dinucleotide (NADH) which are important coenzymes/cofactors for biological functions and energy production in cells (Figure 6). The increase in glycolytic fluxes resulting in higher production of ATP and NADH might be required to fuel the immune responses against Mtb. Targeting the reprogrammed fluxes in infection could help improve TB treatment. For instance, if we could boost glycolytic fluxes in infected cells through therapies this could enhance the elimination of Mtb.

Fluxes through the TCA cycle were significantly lower compared to glycolysis. The TCA cycle generates co-enzymes that participate in the mitochondrial electron transport chain for ATP production. Macrophages are non-diving cells, so the relatively low TCA cycle fluxes indicate a low demand for carbon oxidation through this cycle and subsequent ATP production. Instead, the cellular energy and biosynthetic demand is met by glycolysis and the PPP fluxes. Thus, the TCA cycle fluxes are not the primary route for carbon oxidation in macrophages. Flux profiles show complete use of TCA cycle fluxes R21, R22, R24, R25, R26, R27, R28 and R29 in the oxidative phase (Figure 5). The frequency distribution plots in Supplementary Figure S3 shows no statistically significant differences in the TCA cycle and anaplerotic (malic enzyme and PEP carboxykinase) fluxes between uninfected and infected macrophages although the net fluxes were over 10 times higher in infected macrophages. These fluxes were statistically less well-resolved as compared to glycolysis and the PPP fluxes probably because of the choice of the ¹³C-substrate. Here we chose glucose as the substrate, which is primarily utilised through glycolysis and the PPP, and therefore, these fluxes are statistically well-resolved.

We also quantified the carbon fluxes for non-essential amino acid biosynthesis (Figure 8). The TCA cycle and glycolysis are the primary sources of metabolic precursors required for the biosynthesis of amino acids. The TCA cycle-derived aspartate, arginine, glutamate, and glutamine and PYR-derived alanine biosynthetic fluxes were higher in infected macrophages. Aspartate, glutamate, and glutamine are acquired by the pathogenic Mtb as carbon and nitrogen nutrient sources during intracellular growth (Beste et al., 2013; Borah et al., 2019a). The precise role of these amino acids in macrophage metabolism and immunity still needs further investigation. The increased flux to arginine synthesis can be explained by the involvement of this amino acid in macrophage metabolism and immune responses against Mtb (McKell et al., 2021). Proline and asparagine fluxes did not change in infected macrophages and tyrosine fluxes (not shown in Figure 8) were higher in infected macrophages. The fluxes to glycolytic intermediate PG3-derived serine, cysteine, and glycine synthesis were reduced in infected macrophages (Figure 8). We have previously demonstrated that serine from host macrophages is not available to intracellular Mtb (Borah et al., 2019a). Mtb with deletion in de novo serine biosynthetic pathway failed to survive in human THP-1 macrophages (Borah et al., 2019a). The reduced serine biosynthetic flux in host macrophages is a metabolic phenotype associated with proinflammatory M1 macrophages that restricts bacterial growth.