TCA cycle remodeling is associated with IL-1Î²-mediated proinflammatory eicosanoid signaling in humans with pulmonary tuberculosis

The metabolic signaling pathways that drive pathologic tissue inflammation and damage in humans with pulmonary tuberculosis (TB) are not well understood. Using combined methods in plasma high-resolution metabolomics, lipidomics and cytokine profiling from a multicohort study of humans with pulmonary TB disease, we discovered that IL-1{beta}-mediated inflammatory signaling was closely associated with TCA cycle remodeling, characterized by accumulation of TCA cycle intermediates such as succinate and decreased itaconate. This inflammatory metabolic network was particularly active in persons with multidrug-resistant (MDR)-TB after receiving 2 months of ineffective treatment and was only reversed after 1 year of appropriate anti-TB chemotherapy. Both succinate and IL-1{beta} were closely associated with increases in proinflammatory lipid signaling, including increases in the products of phospholipase A2, increased arachidonic acid formation, and metabolism of arachidonic acid to proinflammatory eicosanoids. Together, these results indicate that decreased itaconate and accumulation of succinate and other TCA cycle intermediates are important drivers of IL-1{beta}-mediated proinflammatory eicosanoid signaling in humans with pulmonary TB disease. Host-directed therapies that mitigate such metabolic reprograming may have potential to limit excessive pulmonary inflammation and tissue damage.


Introduction
Tuberculosis (TB) remains the leading global cause of infectious disease mortality, accounting for ~1.4 million deaths each year (1). Treatment outcomes are particularly poor for persons with multidrug resistant (MDR)-TB, who achieve treatment success in an estimated 57% of cases (1). Even those successfully completing MDR-TB treatment demonstrate high rates of post-TB obstructive lung disease (2,3). While poor treatment outcomes among persons with MDR-TB are multifactorial, propagation of pathologic inflammatory responses during delays in initiation of appropriate anti-TB chemotherapy may play an important role. Even when available, sputum culture with drug susceptibility testing can take up to 3 months to perform, often resulting in treatment with ineffective antibiotics for several months. Prompt diagnosis of MDR-TB using newer molecular diagnostic tests leads to more rapid sputum culture conversion (4) and higher rates of cure (4,5), suggesting treatment delays meaningfully contribute to long-term adverse outcomes.
Elucidating host response pathways that promote pathologic inflammation and tissue damage in persons with MDR-TB will be critical to identify those at greatest risk for adverse pulmonary outcomes as well as targets for host-directed therapeutics.
Animal models of pulmonary TB disease show increases in proinflammatory eicosanoid signaling lead to greater neutrophilic infiltration, bacterial burden and tissue damage (6,7). While early IL-1 production following infection with Mycobacterium tuberculosis (Mtb) is critical for control of bacterial replication (7), ongoing IL-1 signaling can lead to upregulation of proinflammatory eicosanoids and an influx of neutrophils, which promote further bacterial replication and tissue destruction (6). IL-1 blockade in macaques with pulmonary TB disease limits tissue inflammation and damage, further supporting a pathologic role for dysregulated IL-1 signaling in pulmonary TB (8).
Recent studies in macrophage biology show tricarboxylic acid (TCA) cycle remodeling may play an important role in regulating IL-1 signaling (9). Inflammatory macrophage activation leads to accumulation of TCA cycle intermediates such as succinate, which, in turn, results in upregulation of the proinflammatory IL-1b-HIF-1a axis (9). This metabolic remodeling is regulated by the metabolite itaconate, which inhibits succinate dehydrogenase thereby limiting this proinflammatory cascade.
However, the contribution of this host metabolic response pathway to IL-1-mediated inflammation in human pulmonary TB has not been previously described.
Using combined approaches in targeted and untargeted high-resolution metabolomics (HRM) and lipidomics, we sought to determine whether host metabolic phenotypes in persons with MDR-TB receiving ineffective therapy differed from persons with drug susceptible (DS)-TB at the time of diagnosis. We further sought to determine whether differences in host metabolism contribute to pathologic inflammatory responses and poor treatment outcomes in MDR-TB. We found that after 2-3 months of ineffective antibiotic therapy, persons with MDR-TB exhibit significant increases in arachidonic acid (AA) metabolism characterized by increased phospholipase A2 activity and increased conversion of AA to proinflammatory eicosanoids. We show that such changes are strongly associated with increased plasma concentrations of IL-1b, as well as TCA cycle intermediates succinate, fumarate and malate and negatively correlated with plasma concentrations of itaconate. Finally, we show TCA cycle remodeling is reversed only after prolonged treatment with effective anti-MDR-TB chemotherapy. These findings  The y-axis shows significantly regulated metabolic pathways and the x-axis shows thelog p-value for pathway enrichment. All comparisons were adjusted for age, sex and HIV status. The green dashed line indicates statistical significance at p<0.05.
The first step in AA metabolism occurs when phospholipase A2 hydrolyzes the sn-2 acyl bond of phospholipids, releasing AA and lysophosphatidic acid (13).
Therefore, we reasoned that increases in AA metabolism in the MDR-TB cohort would be accompanied by increases in phospholipase A2 activity. To more fully characterize lipid metabolism, we performed untargeted lipidomics on plasma samples from a subset of the above population including HIV-positive persons with MDR-TB (n=29), HIV- All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted January 26, 2021. ; https://doi.org/10.1101/2021.01.23.21250380 doi: medRxiv preprint negative persons with MDR-TB (n=21), persons with DS-TB (n=30) and controls with and without LTBI (n=40). Of plasma lipids with confirmed chemical identities, persons with MDR-TB with or without HIV demonstrated significant upregulation of multiple species of lyso-phospholipids and significant decreases in phospholipids with two acyl chains compared to persons with LTBI (Figure 2A) and persons with DS-TB ( Figure   2B), consistent with significant increases in phospholipase A2 activity.  All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted January 26, 2021. ; Following formation from phospholipase A2, AA is further metabolized through one of three metabolic pathways to regulate inflammation: cyclooxygenases (COX) to form anti-inflammatory prostaglandins, lipoxygenases (LOX) to form proinflammatory eicosanoids and CYP450 enzymes to form the less biologically active dihydroxyeicosatrienoic acids (DHETs) (14). Prior studies have shown that proinflammatory eicosanoids formed through the LOX pathway are significantly upregulated in persons with pulmonary TB disease and associated with more severe clinical disease (7,15) and cavity formation (16). We therefore hypothesized that increased phospholipase A2 activity and AA metabolism in MDR-TB patients were driving increased production of proinflammatory eicosanoids. To further evaluate which AA metabolites were upregulated in South Africans with MDR-TB versus Georgians with DS-TB and asymptomatic controls, we performed a targeted oxylipin assay to identify and quantify relevant eicosanoids in plasma samples (17). (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted January 26, 2021. ; concentrations of the arachidonic acid (AA) metabolites (A) 11,12-DHET and (B) 14,15-DHET as well as (C) the linoleic acid (LA) metabolite 9,10-DIHOME metabolized by the Prior studies have shown that IL-1 plays a critical role in regulating the balance of proinflammatory eicosanoids and anti-inflammatory prostaglandins following infection with Mtb (7). While IL-1 is necessary for initial control of Mtb replication (7), dysregulated IL-1b signaling in later stages of TB disease is associated with eicosanoidmediated inflammation and tissue damage (6), which is ameliorated with IL-1 blockade (8). Thus, we posited that increased arachidonic acid metabolism and proinflammatory eicosanoid signaling in MDR-TB was due to increased IL-1b signaling. Indeed, we found that South Africans with MDR-TB had significantly higher plasma concentrations IL-1b versus Georgians with DS-TB and controls without TB disease ( Figure 4A). Furthermore, we found plasma concentrations of IL-1b were significantly and positively correlated with plasma concentrations of pro-inflammatory eicosanoids 5-HETE ( Figure   4B) and 12-HETE ( Figure 4C). All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted January 26, 2021. ; Accumulation of TCA cycle intermediates has been described as a potent driver of IL-1b-mediated inflammatory responses in multiple disease states (9,18,19). We therefore hypothesized that increased IL-1b-mediated proinflammatory eicosanoid signaling in MDR-TB may be driven by TCA cycle remodeling and accumulation of TCA cycle intermediates such as succinate. To test this hypothesis, we analyzed plasma HRM data run in parallel using a C18 negative LC system to better identify and quantify organic acids. All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted January 26, 2021. ; https://doi.org/10.1101/2021.01.23.21250380 doi: medRxiv preprint We detected 9,787 m/z features in C18 negative ionization mode. Unbiased pathway analysis (12) revealed the TCA cycle was one of the most significantly regulated metabolic pathway in persons with MDR-TB versus controls without TB disease ( Figure 5A) and versus persons with DS-TB ( Figure 5B). In the subset of MDR-TB patients where plasma samples were available over the course of the 2-year treatment period (n=17), we also found the TCA cycle was also the most significantly regulated metabolic pathway during MDR-TB treatment ( Figure 5C).  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted January 26, 2021. ; https://doi.org/10.1101/2021.01.23.21250380 doi: medRxiv preprint TCA cycle metabolites that significantly differed between groups were confirmed and quantified by accurate mass, MS/MS and retention time relative to authentic standards (20). TCA cycle remodeling in MDR-TB was characterized by significant increases in plasma concentrations of succinate, fumarate and malate versus persons with DS-TB and asymptomatic controls (Figure 6 A-C), consistent with accumulation of TCA cycle intermediates. Conversely, plasma concentrations of itaconate, which has been shown to negatively regulate TCA cycle flux by inhibiting succinate dehydrogenase (9), were significantly decreased in persons with MDR-TB compared to persons with DS-TB and controls ( Figure 6D). Plasma concentrations of itaconate were significantly and negatively correlated with plasma concentrations of succinate ( Figure   6E). Similarly, plasma itaconate concentrations were significantly and negatively correlated with plasma concentrations of fumarate and malate (r= -0.47 and -0.48 respectively, p<0.001 for both), consistent with a role in negatively regulating these metabolites. In persons with DS-TB from Georgia, itaconate concentrations were significantly lower in persons with a persistently positive AFB sputum smear at enrollment versus those who had converted to a negative AFB sputum smear ( Figure   6F). All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
We then sought to examine the relationship between plasma concentrations of TCA cycle metabolites and plasma concentrations of IL-1b and proinflammatory eicosanoids.
We found plasma concentrations of succinate were significantly and positively correlated with concentrations of 5-HETE, 12-HETE and IL-1b, confirming a close association between TCA cycle remodeling and IL-1b-mediated proinflammatory eicosanoids signaling (Figure 7 A-C). Conversely, we found significant and negative correlations between plasma concentrations of itaconate and those of 5-HETE, 12-HETE and IL-1b (Figure 7 D-F). (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted January 26, 2021. ; In this multicohort study, we used an unbiased approach to plasma metabolic pathway analysis combined with targeted and untargeted lipidomics to discover that TCA cycle remodeling is strongly associated with IL-1b-mediated proinflammatory eicosanoid signaling in persons with pulmonary TB disease. We found this inflammatory cascade, characterized by increased plasma concentrations of TCA cycle intermediates succinate, fumarate and malate and decreased concentrations of itaconate, was particularly upregulated in persons with MDR-TB after 2-3 months of ineffective anti-TB chemotherapy and was reversed only after 1 year of efficacious treatment. Collectively, these results indicate that TCA cycle remodeling is an important driver of IL-1bmediated proinflammatory eicosanoid signaling in pulmonary TB. Furthermore, our findings suggest prolonged delays in effective treatment allow this proinflammatory cascade to perpetuate, potentially contributing to poor treatment outcomes in MDR-TB.
These findings also indicate that administration of host-directed therapies that limit TCA cycle flux may be an effective way to reduce proinflammatory eicosanoid signaling and limit tissue damage (6)(7)(8)(9).
Accumulation of TCA cycle intermediates such as succinate are known to drive pathologic inflammation in macrophages through activation of the HIF-1a-IL-1 axis that is an important mediator of tissue damage in ischemic-reperfusion injury (18,19). The strong association between TCA cycle remodeling and IL-1b-mediated proinflammatory eicosanoid signaling in our study suggests TCA cycle intermediates also play a critical role in regulating the human inflammatory response to pulmonary TB. One potential source of accumulating TCA cycle intermediates in persons with MDR-TB is the host microbiome (21). The microbiome produces a large proportion of succinate circulating in All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted January 26, 2021. ; https://doi.org/10.1101/2021.01.23.21250380 doi: medRxiv preprint plasma (21), thus providing a potential avenue through which microbial composition could regulate the inflammatory response in TB disease. Indeed, microbiome differences in rhesus macaques have been associated with TB disease severity and bacterial burden (22). Another potential explanation that is not mutually exclusive is the ability of Mtb to induce aerobic glycolysis (23,24). This process, also termed the "Warburg Effect", involves a shift in cellular energy metabolism to increased glycolysis and reduced oxidative phosphorylation, thereby causing TCA cycle intermediates to accumulate (25). Indeed, in vitro studies of human pulmonary macrophages show the ability of Mtb to induce aerobic glycolysis is necessary to stimulate production of IL-1b (23,26).
IL-1b is known to play an important role in regulating the products of AA metabolism, which evolves over the course of infection with Mtb (6-9). Initially, IL-1 provides a counterbalance to type-1-IFN signaling (7,27) by promoting metabolism of AA to anti-inflammatory prostaglandins rather than proinflammatory eicosanoids (7). This signaling pathway promotes early control of Mtb replication, with IL-1 receptor blockade leading to greater abundance of proinflammatory eicosanoids and increased bacterial burden while IL-1 receptor agonism attenuates disease severity (7,27).
Similarly, early IL-1 signaling by pulmonary macrophages triggered by induction of aerobic glycolysis is necessary to control Mtb growth (23). However, as TB disease progresses, IL-1b becomes a driver of proinflammatory eicosanoids signaling leading to an influx of neutrophilic inflammation that is permissive to bacterial growth (6). In later stages of TB disease in mice, IL-1 blockade is therapeutic and limits tissue damage (8).
In macaques with TB disease, plasma concentrations of IL-1b are highly correlated with All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted January 26, 2021. ; https://doi.org/10.1101/2021.01.23.21250380 doi: medRxiv preprint levels of pulmonary inflammation as measured by PET-CT and IL-1R blockade limits inflammation and tissue damage (8). Thus, accumulating TCA cycle intermediates during later stages of TB disease may be another example of Mtb exploiting a host metabolic adaption aimed at controlling Mtb replication during the initial stages of infection (28,29). While TCA cycle remodeling initially promotes control of bacterial replication, these results suggest that in later stages of human TB disease it increases AA metabolism and conversion to proinflammatory eicosanoids, potentially worsening tissue damage and increasing disease severity.
Clinical studies also indicate IL-1b primarily acts as a driver of inflammation and tissue damage in human TB disease. Independent studies consistently show elevated plasma concentrations of IL-1b in persons with pulmonary TB versus asymptomatic controls (30,31) that correlate with markers of inflammation such as ESR and CRP (30). Further, elevated plasma concentrations of IL-1b in pulmonary TB are associated with greater extent of disease and cavitation on chest radiograph (31,32). Our findings suggest such signaling may be caused by TCA cycle remodeling and that tissue damage associated with elevated IL-1b is mediated by increased AA metabolism and production of proinflammatory eicosanoids.
The observation that dysregulated IL-1b signaling occurs disproportionately in persons with MDR-TB has been reported previously (30). We hypothesize this is caused by delays in adequate treatment in persons with MDR-TB, which allows for propagation of a proinflammatory cascade. However, rifampin resistance mutations in Mtb isolates have been associated with differential metabolic responses and secretion of IL-1b in human macrophages (33). Thus, it is possible the different host responses All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
This study is subject to several limitations. Though we demonstrate a strong association between plasma concentrations of TCA cycle intermediates and proinflammatory eicosanoids, the observational nature of the study precludes us from definitively establishing a causal relationship. All persons with MDR-TB were from a single study site, so it remains possible that metabolic differences in this group are driven by host genetic differences and differences in microbial composition rather than MDR-TB status. In future studies it will be important to evaluate the relative contribution of pulmonary macrophages and the host microbiome to plasma concentrations of TCA cycle intermediates in persons with pulmonary TB.
In summary, we demonstrate that IL-1b-mediated proinflammatory eicosanoid signaling is strongly associated with TCA cycle remodeling in humans with pulmonary TB. This remodeling is characterized by significant increases in plasma succinate concentrations and significant decreases in concentrations of itaconate. These findings provide evidence that pathologic eicosanoid signaling in pulmonary TB is driven by accumulation of TCA cycle intermediates. The TCA cycle may therefore represent a promising target for host-directed therapies aimed at limiting pulmonary inflammation and tissue damage.

Sample collection
For all cohorts, blood was collected in ethylenediaminetetraacetic acid (EDTA)containing tubes and centrifuged; isolated plasma was immediately frozen and stored at All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  (37).

Plasma metabolomics analysis
De-identified samples were randomized by a computer-generated list into blocks of 40 samples prior to transfer to the analytical laboratory where personnel were blinded to clinical and demographic data. Thawed plasma (65 μL) was treated with 130 μl acetonitrile (2:1, v/v) containing an internal isotopic standard mixture (3.5 μL/sample), All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted January 26, 2021. ; https://doi.org/10.1101/2021.01.23.21250380 doi: medRxiv preprint as previously described (38). The internal standard mix for quality control consisted of 14 stable isotopic chemicals covering a broad range of small molecules (38) were performed for each plasma sample and median summarized (42).

Metabolite identification and reference standardization
Identities of metabolites of interest were confirmed using ion dissociation methods (tandem MS/MS). Fragmentation spectra were generated using a Q Exactive HF Hybrid Quadrupole-Orbitrap Mass Spectrometer with parallel reaction monitoring mode using a targeted inclusion list. TCA cycle metabolites were confirmed and quantified by accurate mass, MS/MS and retention time relative to authentic standards (20).

Untargeted lipidomics
Lipids were extracted from each plasma sample using a high throughput methyl t butyl ether (MtBE) extraction procedure with an automated and robust liquid handling All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
Extracted samples were dried under nitrogen and reconstituted in 200 µl 1:1 chloroform:methanol prior to injection into the LC/MS system and lipids were resolved using a Thermo Acclaim C18 reverse phase column on a Thermo Vanquish UPLC coupled to a Thermo Fusion IDX mass spectrometer (Thermo, Waltham, MA) (44). Data were acquired at a resolution of 240,000 FWHM and deep MS/MS data was collected on pooled samples using an iterative data dependent strategy at multiple collision energies. Data was processed using LipidSearch (Thermo Fisher, San Jose, CA). Lipids that contained a signal to noise ratio of greater than 10 and had high confidence (MS/MS) identifications with CVs less than 30% across pooled QCs were used for downstream analysis.

Targeted measurement of eicosanoids
We selectively targeted oxylipins and endocannabinoids (OXYs), which are highly regulated, resulting in low abundance in human plasma. To enrich these lipids from bulk membrane lipids, we performed solid phase extraction methods using the Biotage Extrahera liquid handling system as previously described (17,45,46). The resulting extracts were analyzed using a multiple reaction monitoring (MRM)-based LC/MS protocol on a QTrap 5500 (Sciex, Waltham, MA), whereby detected oxylipins and endocannabinoids were fragmented and quantified against external standard curves. Oxylipins and endocannabinoids detected in less than 60% of patient samples were excluded from this analysis. Missing values were imputed using half of the minimum detected value for each lipid.

Plasma cytokine detection
All rights reserved. No reuse allowed without permission.
(which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
The copyright holder for this preprint this version posted January 26, 2021. ; https://doi.org/10.1101/2021.01.23.21250380 doi: medRxiv preprint The U-PLEX assay (Meso Scale MULTI-ARRAY Technology) commercially available by Meso Scale Discovery (MSD) was used for plasma cytokine detection. This technology allows the evaluation of multiplexed biomarkers by using custom made U-PLEX sandwich antibodies with a SULFO-TAG™ conjugated antibody and next generation of electrochemiluminescence (ECL) detection. The assay was performed according to the manufacturer's instructions (https://www.mesoscale.com/en/technical_resources/technical_literature/techncal_notes _search). In summary, 25μL of plasma from each participant was combined with the biotinylated antibody plus the assigned linker and the SULFO-TAG™ conjugated detection antibody; in parallel a multi-analyte calibrator standard was prepared by doing 4-fold serial dilutions. Both samples and calibrators were mixed with the Read buffer and loaded in a 10-spot U-PLEX plate, which was read by the MESO QuickPlex SQ 120. The plasma cytokines values (pg/mL) were extrapolated from the standard curve of each specific analyte.

Statistical analysis
Statistical comparisons of metabolite and lipid intensity values (abundance) and concentrations were performed in R version 3.5.0. For untargeted metabolomics and lipidomics analyses, metabolite intensity values were log2 transformed and compared between groups using linear regression, controlling for age, sex and HIV status (47).
Metabolic pathway enrichment analysis was performed using mummichog, a Pythonbased informatics tool that leverages the organization of metabolic networks to predict functional changes in metabolic pathway activity (12,28,48). Following quantification of selected metabolites and lipids, cross-sectional comparison of plasma concentrations