Spatially heterogeneous lipid dysregulation in tuberculous meningitis

Abstract

Tuberculous (TB) meningitis is the deadliest form of extrapulmonary TB which disproportionately affects children and immunocompromised individuals. Studies in pulmonary TB have shown that Mycobacterium tuberculosis can alter host lipid metabolism to evade the immune system. Cholesterol lowering drugs (i.e., statins) reduce the risk of infection, making them a promising host-directed therapy in pulmonary TB. However, the effect of M. tuberculosis infection on the young or adult brain lipidome has not been studied. The brain is the second-most lipid-rich organ, after adipose tissue, with a temporally and spatially heterogeneous lipidome that changes from infancy to adulthood. The young, developing brain in children may be uniquely vulnerable to alterations in lipid composition and homeostasis, as perturbations in cholesterol metabolism can cause developmental disorders leading to intellectual disabilities. To begin to understand the alterations to the brain lipidome in pediatric TB meningitis, we utilized our previously published young rabbit model of TB meningitis and applied mass spectrometry (MS) techniques to elucidate spatial differences. We used matrix assisted laser desorption/ionization-MS imaging (MALDI-MSI) and complemented it with region-specific liquid chromatography (LC)-MS/MS developed to identify and quantify sterols and oxysterols difficult to identify by MALDI-MSI. MALDI-MSI revealed several sphingolipids, glycerolipids and glycerophospholipids that were downregulated in brain lesions. LC-MS/MS revealed the downregulation of cholesterol, several sterol intermediates along the cholesterol biosynthesis pathway and enzymatically produced oxysterols as a direct result of M. tuberculosis infection. However, oxysterols produced by oxidative stress were increased in brain lesions. Together, these results demonstrate significant spatially regulated brain lipidome dysregulation in pediatric TB meningitis.

1. Introduction

Tuberculosis (TB), caused by the pathogen Mycobacterium tuberculosis, is responsible for at least 1.2 million deaths globally each year and remains a major global health challenge (Chakaya et al., 2021; WHO, 2020). Despite accounting for 5–10 % of extrapulmonary TB cases, TB meningitis constitutes the most lethal form, disproportionately impacting children under five years of age and immunocompromised individuals (Jain et al., 2018; Seddon et al., 2019). Neuroinflammation is integral to its pathogenesis and is believed to underlie its high mortality and morbidity, but the mechanisms remain poorly understood (Jain et al., 2018; Seddon et al., 2019; Tucker et al., 2016). Corticosteroids are currently the standard of care host-directed therapy used to address neuroinflammation. However, they do not prevent neurological sequelae and do not improve the survival of several patient populations (Donovan et al., 2023; Prasad et al., 2016; Thwaites et al., 2004). Indeed, genetic polymorphism of fatty acid metabolism by leukotriene A4 hydrolase (LTA4H), an enzyme that regulates the production of proinflammatory mediators from fatty acids (i.e., leukotriene B4), influences corticosteroid efficacy, with the hyperinflammatory genotype (TT) conferring a survival benefit in patients treated with corticosteroids (Thuong et al., 2017; Tobin et al., 2012). There is an urgent need for more effective host-directed therapy. Although the brain is a lipid-rich organ, there remains a critical knowledge gap of the effect that M. tuberculosis infection has on lipid metabolism and lipidome composition in the brain and bridging this gap could lead to the development of novel host-directed therapy.

The brain trails behind only adipose tissue as the second-most lipid-rich organ, primarily composed of cholesterol, sphingolipids, and glycerophospholipids. Lipids levels within the brain are tightly regulated by in situ, de novo biosynthesis and restricted transport across the blood-brain barrier (Yoon et al., 2022). The brain lipidome is temporally and spatially dynamic, changing from infancy to adulthood (Oliveira et al., 2022). In the young, developing brain, perturbations to cholesterol metabolism (i.e., Smith-Lemli-Opitz syndrome and 7-dehydrocholesterol [7-DHC] reductase) result in microcephaly and intellectual disabilities, highlighting the importance of lipid species to normal brain development (Porter, 2008). In addition to serving as key structural components of the central nervous system (CNS), these lipid species are active participants in the neuroinflammatory cascade. Under pathological conditions, enzymatic cleavage of glycerophospholipids releases arachidonic acid that leads to the production of eicosanoids, including leukotrienes, prostaglandins and thromboxanes (Farooqui et al., 2007). Cholesterol biosynthesis pathway intermediates (i.e., sterols) and their oxidized products (i.e., oxysterols) regulate differentiation and proliferation of immune cells (Spann and Glass, 2013). Although some oxysterols are formed by enzymatic reactions, others are formed by auto-oxidation of cholesterol with reactive oxygen species (ROS) secondary to oxidative stress, which contributes to lipid peroxidation. Regardless of the mechanism, oxysterols can be cytotoxic and induce neuroinflammation (Anderson et al., 2020; Son et al., 2023). Similarly, sphingolipids are active modulators of innate and adaptive immunity (Lee et al., 2023). For instance, sphingolipid-deficient macrophages exhibit deficient phagocytosis of M. tuberculosis due to disrupted phagolysosome formation (Niekamp et al., 2021).

Granulomas, a hallmark of TB pathology, represent complex tissue microenvironments carrying signatures of both the host’s innate and adaptive immunological response to infection. From the initial mobilization of macrophages to peripheral lymphocytic cuffing as well as dynamic remodeling leading to caseation and cavitation, the intricate host-pathogen interactions in TB granulomatous lesion formation have been scrutinized heavily in the context of pulmonary TB (Marakalala et al., 2016; Mattila et al., 2013; Ramakrishnan, 2012). At the core of the tissue destruction seen in these granulomas lies the dysregulated host lipid metabolism (Kim et al., 2010). One such manifestation is the development of lipid-laden foamy macrophages, which serve as the primary carbon source for the survival and proliferation of M. tuberculosis, bypassing the nutrient-depriving sequestration by host immune cells (Babunovic et al., 2022; Brandenburg et al., 2021; Guerrini et al., 2018; Pandey and Sassetti, 2008). It is believed that reorganization of host lipid metabolism is one way M. tuberculosis evades the immune system and obtains nutrients (Pawełczyk et al., 2021; Russell et al., 2009; Shim et al., 2020). Indeed, the mycobacterium genome exhibits upregulated of genes of lipid and fatty acid metabolism and increased levels of major cellular lipid components (Cole et al., 1998; Lee et al., 2013). Statins, known to lower serum cholesterol levels by inhibiting 3-hydroxy-3-methylglutaryl-CoA reductase (HMGCR), promote intracellular killing of M. tuberculosis and reduce the risk of pulmonary TB in retrospective clinical studies (Dutta and Karakousis, 2021). Indeed, alterations in the lipidome within M. tuberculosis-infected tissues have critical implications for not only pathogen survival but also for drug delivery and penetration into infected tissue for effective sterilization (Greenwood et al., 2019; Kokesch-Himmelreich et al., 2022).

Although granulomas have been studied extensively in the lungs, TB can lead to granulomatous inflammation in the CNS, often referred to as tuberculomas or brain lesions (Sholeye et al., 2022; Zaharie et al., 2020). Postmortem studies in humans and animal models of TB meningitis show that necrotizing and non-necrotizing brain lesions as well as foamy macrophages can develop in the CNS (Ruiz-Bedoya et al., 2022; Zaharie et al., 2020). Alterations in host lipid metabolism has also been described in cerebrospinal fluid (CSF) of TB meningitis patients, with a proteomic analysis revealing elevated apolipoprotein (i.e., ApoB) (Mu et al., 2015). Modest enrichment of glycerophospholipid, arachidonic acid and sphingolipid metabolism have also been observed by metabolomic pathway analysis of patient CSF (van Laarhoven et al., 2018). Changes to the brain lipidome may have important implications for young children who are not only at highest risk for TB meningitis but also uniquely vulnerable due to ongoing myelination within the CNS (Buyanova and Arsalidou, 2021; Seddon, 2019). Despite mounting evidence highlighting the role of lipid dysregulation in the pathogenesis of pulmonary TB, there is a dearth of literature characterizing this process in TB meningitis. These knowledge gaps present opportunities to enhance our understanding and expand new targets for host immunomodulation.

To begin to understand the alternations to the brain lipidome in TB meningitis, we utilized matrix assisted laser desorption/ionization-mass spectrometry imaging (MALDI-MSI) to identify the spatial distribution of sphingolipids, glycerophospholipids, and glycerolipids. We complemented this imaging with region-specific liquid chromatography tandem mass spectrometry (LC-MS/MS) developed to identify sterols and oxysterols difficult to identify and quantify by MALDI-MSI (Tallman et al., 2021; Yutuc et al., 2020). MALDI-MSI has been performed in pulmonary TB granulomas and elucidated an abundance of arachidonic acid with spatially organized inflammatory signaling, showing proinflammatory eicosanoids in the necrotic center and anti-inflammatory ones on the periphery (Marakalala et al., 2016). In this study, we utilized our previously described model of TB meningitis in young rabbits (~5 days old at the time of infection) to characterize lipid dysregulation in the young, developing brain secondary to M. tuberculosis. To account for the fact that the brain lipidome is in flux during development, we compared results to age-matched uninfected healthy rabbits (Oliveira et al., 2022). We performed MALDI-MSI of whole brains with regional comparisons done by drawing regions of interest (ROI) and then performed complementary region-specific LC-MS/MS using microdissections of several areas of the brain in another cohort of rabbits of the same age and duration of infection. Fig. 1 shows our experimental schematic.

Fig. 1.

Experimental design. (A) Experimental workflow. After three weeks post-injection, two cohorts of rabbits were utilized for matrix assisted laser desorption/ionization-mass spectrometry imaging (MALDI-MSI; upper panel) or liquid chromatography (LC)-MS/MS (lower panel). (B) Similar regions of interest (ROI) were used for comparison in MALDI-MSI (left) and LC-MS/MS (right). Other abbreviations: DHB, 2,5-dihydroxybenzoic acid; GM, grey matter; PBS, phosphate buffered saline; WM, white matter.

2. Materials and methods

2.1. General

All protocols were approved by the Johns Hopkins University Animal Care and Use (RB19M417) and Biosafety committees. All chemicals and solvents were purchased from commercial suppliers and were high-performance liquid chromatography (HPLC) or LC-MS grade.

2.2. Rabbit model of TB meningitis

We utilized our previously published young rabbit model of TB meningitis (Ruiz-Bedoya et al., 2022; Tucker et al., 2018; Tucker et al., 2016; Tucker et al., 2024). In brief, young (5–7 day old) male and female New Zealand White rabbits (Robinson Services Inc.) had topical lidocaine applied, were sedated with isoflurane/dexmedetomidine, and then injected intraventricularly with 20 μL of phosphate buffered saline (PBS, healthy) or M. tuberculosis H37Rv inoculum (i.e., M. tuberculosis-infected) via the bregma. The infection was allowed to incubate for three weeks prior to tissue harvest and processing after rabbits were euthanized with pentobarbital overdose.

2.3. MALDI-MSI

2.3.1. Sample preparation

When euthanized, rabbits were perfused with formalin. The brain was harvested and stored in formalin at 4 °C for 24 h. Brains were removed from formalin, blotted dry, and frozen at −80 °C for a minimum of 24 h prior to sectioning. Brains were cryosectioned at 10 μm thickness in a − 20 °C Leica CM1860 UV cryostat (Leica Biosystems, Nussloch, Germany) onto indium tin oxide coated slides (Delta Technologies, Loveland, CO). Prior to sectioning, slides were checked for conductivity using a conductivity meter and washed for 10 min each with hexanes and ethanol while sonicating. Slides were allowed to dry overnight before sectioning. M. tuberculosis-infected brains (n = 3) and uninfected healthy brains (n = 3) were cryosectioned and placed on slides in pairs (i.e., one M. tuberculosis-infected brain and one healthy brain per slide). Brain sections were chosen in M. tuberculosis-infected samples based on lesion location and key anatomical landmarks were matched in uninfected healthy samples. Sectioned slides were kept in −80 °C freezer until MALDI-MSI.

2.3.2. Imaging

Tissues were sprayed with 40 mg/mL 2,5-dihydroxybenzoic acid (DHB) in 50 % HPLC grade acetonitrile (CH₃CN) using a HTX M5 Sprayer (HTX, Chapel Hill, NC) with the following parameters: 75 °C nozzle temperature, 8 passes, 0.1 mL/min flow rate, 1200 mm/min velocity, 3 mm track spacing, criss cross pattern, 10 pounds per square inch (PSI) pressure, 3 L/min gas flow rate, and 10 s drying time between passes. MALDI-MSI was performed in a rapifleX^® MALDI TOF/TOF Instrument (Bruker Daltonics, Billerica, MA) in the reflector positive mode at 100 μm spatial resolution. Data were imported into SCiLS Lab (Bruker Daltonics, Billerica, MA). The ROIs were manually drawn for analysis in each sample for the following regions: individual brain lesion (i.e., lesion) in the M. tuberculosis-infected brain and the uninfected healthy control brain ROI was drawn as the whole brain (i.e., healthy). A Receiver Operating Characteristic (ROC) curve was used to determine significant differences in lipid levels between the brain lesion from M. tuberculosis-infected brain and healthy brain.

2.3.3. Lipid identification and verification

Identification of lipids was performed using LIPID MAPS publication database (lipidmaps.org). Confirmation of lipid identity was performed on a 7 T Bruker scimaX by high mass resolution. Lipids were then categorized based on the nomenclature used by LIPID MAPS and described by Fahy et al. (Fahy et al., 2009).

2.4. Sterol and oxysterol LC-MS/MS

2.4.1. Sample preparation

Similarly, rabbits were euthanized and perfused with PBS. M. tuberculosis-infected brains (n = 8 rabbits) were harvested and underwent microdissection to sample core tuberculous brain lesions (n = 1–3 lesions/brain), tissue surrounding the brain lesion (i.e., paralesional; n = 1–3 paralesional/brain), and white matter and grey matter areas (n = 6 other regions/brain) of the hemisphere contralateral to the lesion (“unaffected”). Out of 15 focal tuberculous brain lesions observed, 9 were periventricular and 6 deep parenchymal. Supplemental Fig. 1 shows representative brain lesions and paralesional areas on gross pathology. The white matter and grey matter microdissections were repeated for uninfected healthy brains (n = 13 rabbits; n = 8 samples/brain). Tissues were submerged in methanol (MeOH; 10 μL/mg tissue) and homogenized with a microbead homogenizer. An antioxidant cocktail (10 μL of 2.5 mg triphenylphosphine [TPP] and 1.0 mg butylated hydroxytoluene [BHT] per mL MeOH) was added to prevent ex vivo isomerization or oxidation before storage at −80 °C.

2.4.2. Sterol and oxysterol extraction

50 μL (5 mg) of tissue homogenate was mixed with 10 μL internal standards (Sterols: 30 μM d₇-cholesterol, d₆-lanosterol, d₇-dHLan [dihydrolanosterol]; 3 μM d₇–7-DHC, d₆-desmosterol, d₆–7-DHD [7-dehydrodesmosterol], d₇-lathosterol, d₆-DHL [dehydrolathosterol]; Oxysterols: 1 μM of d₇-α-epChol [5,6-α epoxycholesterol], d₇–7-KetoChol [7-ketocholesterol], d₇-DHCAO [3,5-dihydroxycholestanone], d₇–7-OHChol [7-hydroxycholesterol], d₇–24-OHChol [24-hydroxycholesterol], d₆–25-OHChol [25-hydroxycholesterol], d₅–27-OHChol [27-hydroxycholesterol], and 0.28 μM d₇-β-epChol [5,6-β epoxycholesterol]). The sterols and oxysterols were extracted by adding 600 μL Folch solution (2:1 = chloroform [CHCl₃]: MeOH) and 400 μL 0.9 % NaCl. The CHCl₃ layer was collected, dried, and derivatized by addition of 100 μL of a fresh made N,N-dimethylglycine (DMG) solution (per 1 mL of reagent: 20 mg 2-methyl-6-nitrobenzoic acid anhydride, 14 mg DMG, 6 mg dimethylaminopyridine [DMAP], 0.1 mL anhydrous triethylamine [Et₃N] in 0.9 mL anhydrous CHCl₃). After 30 min agitation at room temperature, the resulting solution was dried and reconstituted in 100 μL MeOH. Samples were analyzed on an Acquity UPLC system (Waters, Milford, MA) by injecting 10 μL onto the column (Agilent Poroshell EC-C18, 10 cm × 2.1 cm, 1.9 μm) with CH₃CN: MeOH:H₂O 70:25:5 (0.01 % formic acid (v), 1 mM ammonium acetate [NH₄OAc]; Agilent, Santa Clara, CA) mobile phase at a column temperature 40 °C. The flow rate was 0.4 mL for 12 min then ramped to 0.6 mL with a total run for 17 min. A TSQ Quantum Ultra mass spectrometer (ThermoFisher, Waltham, MA) was used for MS detections in positive electrospray ionization (ESI) mode, and data were acquired and analyzed with Finnigan Xcalibur software package (ThermoFisher, Waltham, MA). A full list of selected reaction monitoring (SRM) of sterols and oxysterols were published previously (Tallman et al., 2021). Final sterol and oxysterol numbers were reported as nmol/mg tissue.

2.5. Histology

Sections (10 μm) were collected sequentially to the ones used in MALDI-MSI. These were subjected to hematoxylin and eosin (H&E) staining to better observe anatomical structures. Each step of the staining took place in its own container. The H&E staining procedure was as follows: 24 h in 100 % ethanol for fixation, 3 min in 100 % ethanol, 3 min in 96 % ethanol, 3 additional min in 96 % ethanol, 3 min in 70 % ethanol, 3 min in 70 % ethanol, 3 min in tap water, 3 min in hematoxylin, 3 min in distilled water, 30 s in eosin, tap water rinse, 1 min in xylene, 30 s in 100 % ethanol.

2.6. Statistical analysis

Data were analyzed using Prism 10, version 10.2.3 (GraphPad Prism, La Jolla, CA). For LC-MS/MS data, comparisons were made using Kruskal-Wallis test for nonparametric data and P value ≤0.05 were considered statistically significant. For MALDI-MSI data, ROC area under the curve (AUC) of <0.25 or > 0.75 was used to determine significant differences in lipid levels.

3. Results

3.1. Glycerophospholipids, sphingolipids, and glycerolipids are decreased in M. tuberculosis-infected brain lesions

Evidence of native lipid species alteration was observed as a direct result of M. tuberculosis infection in the brain. We were able to identify the brain lesion by visual inspection of the tissue, however, hematoxylin and eosin (H&E) staining was performed to confirm (Fig. 2A). A ROI was drawn encompassing the brain lesion within the M. tuberculosis-infected brain. This region was compared to the uninfected healthy brain using ROC analysis. MALDI-MSI analysis revealed that several lipid species were downregulated in brain lesions of the M. tuberculosis-infected brain compared to the uninfected healthy brain (Fig. 2B–L). These data correlated with the H&E staining which demonstrated significant damage to the tissue (Fig. 2A). The recorded lipid masses and the matched lipid identities can be found in Table 1. Of the eleven downregulated lipids: six were glycerophospholipids, three were sphingolipids, and two were glycerolipids. No lipids within our m/z detection range were found to be upregulated in the brain lesions.

Fig. 2.

MALDI-MSI of identified lipids. (A) H&E staining of healthy control (top) and M. tuberculosis-infected brain (bottom). (B-G) MALDI images of identified glycerophospholipids, (H-J) identified sphingolipids, and (K-L) identified glycerolipids in healthy control brain (top) and M. tuberculosis-infected brain (bottom). White dotted outline in H&E and red dotted outline in MALDI image correspond to brain lesion. Abbreviations: DG, diradylglycerolipid; HexCer, hexosylceramide; PA, glycerophosphate; PC, glycerophosphocholine; PE, glycerophosphoethanolamine; PG, glycerophosphoglycerol; TG, triadylglycerolipid. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 1.

Lipid masses and the matched lipid identities from MALDI-MSI. Lipid mass to charge ratios (m/z) identified by MALDI-MSI, with the associated identities from LIPID MAPS. Area under the curve (AUCs) closer to 1 indicate higher concentrations in healthy samples and AUCs closer to 0 indicate higher concentrations in M. tuberculosis-infected samples. DG, diradylglycerolipids; HexCer, hexosylceramide; PC, glycerophosphocholine; PE, glycerophosphoethanolamine; PG, glycerophosphoglycerol; PA, glycerophosphate; TG, triradylglycerolipid.

Detected m/z (Daltons)	Matched m/z (Daltons)	Delta (Daltons)	Formula	Ion	Possible Identities	Lipid Class	Lipid Subclass	Respective AUC’s (N1,N2,N3)
734.563	734.5694	0.0064	C40H80NO8P	[M + H]+	PC 32:0 / PE 35:0	Glycerophospholipid	Diacylglycerophosphocholine / Diacylglycerophosphoethanolamine	0.987, 0.921, 0.777
735.559	735.5534	0.0056	C40H79O9P	[M + H]+	PG O-34:1	Glycerophospholipid	1-akyl, 2-acylglycerophosphoglycerol	0.989, 0.844, 0.843
753.654	753.6367	0.0173	C47H86O5	[M + Na]+	DG 44:3	Glycerolipid	Diacylglycerol	0.871, 0.989, 0.801
754.678	754.6967	0.0187	C47H83O6	[M + H]+	TG(14:0/16:1(9Z)/14:0)(d5)	Glycerolipid	Triacylglycerol	0.961, 0.966, 0.934
762.591	762.5854	0.0056	C43H81NO8	[M + Na]+	HexCer 37:2; O2	Sphingolipid	Simple Glucose Series	0.993, 0.922, 0.903
773.603	773.6055	0.0025	C44H85O8P	[M + H]+	PA 41:1	Glycerophospholipid	Diacylglycerophosphate	0.889, 0.963, 0.800
774.516	774.5068	0.0092	C44H72NO8P	[M + H]+	PC 36:8 / PE 39:8	Glycerophospholipid	Diacylgycerophosphocholine / Diacylglycerophosphoethanolamine	0.891, 0.756, 0.868
782.65	782.6504	0.0004	C46H87NO8	[M + H]+	HexCer 40:2; O2	Sphingolipid	Simple Glucose Series	0.994, 0.994, 0.867
798.643	798.6454	0.0024	C46H87NO9	[M + H]+	HexCer 40:2; O3	Sphingolipid	Simple Glucose Series	0.884, 0.947, 0.852
799.528	799.525	0.003	C42H81O9P	[M + K]+	PG O-36:2	Glycerophospholipid	1-akyl, 2-acylglycerophosphoglycerol	0.926, 0.871, 0.804
800.635	800.6528	0.0178	C46H90NO7P	[M + H]+	PC O-38:2	Glycerophospholipid	1-akyl, 2-acylglycerophosphoglycerol	0.925, 0.966, 0.875

3.1.1. Glycerophospholipids are decreased in brain lesions

Six glycerophosopholipids were downregulated in the lesion of the M. tuberculosis-infected brain (Fig. 2B–G, Table 1). Glycerophosphocholines and glycerophosphoethanolamines, also known as phosphatidylcholines (PC) and phosphatidylethanolamines (PE) respectively, are the most abundant phospholipids in all mammalian cell membranes (van der Veen et al., 2017) and were among the decreased glycerophospholipids within brain lesions of the M. tuberculosis-infected brain. Specifically, we found PC 32:0/PE 35:0, PC O-38:2, and PC 36:8/PE 39:8 were downregulated in the brain lesion (Fig. 2B–D). These data are consistent with the H&E stain, as there is substantial disruption of tissue architecture within the brain lesion. Glycerophosphoglycerol, also known as phosphatidylglycerol (PG), is a phospholipid found in the mitochondria cristae and is necessary for cardiolipin synthesis which regulates several mitochondrial proteins, including the electron transport complex (Sun et al., 2023). PG O-34:1 and PG O-36:2 were both found to be downregulated in the brain lesion as compared to the healthy brain (Fig. 2E–F). Glycerophosphates, also known as phosphatidic acids (PA), mainly function as lipid mediators (Yao et al., 2023). We identified PA 41:1 to be downregulated in the brain lesion (Fig. 2G). These results indicate a substantial remodeling of tissue architecture within the brain lesion compared to the surrounding tissue and the healthy brain.

3.1.2. Sphingolipids are decreased in brain lesions

Sphingolipids are among the most common lipids in the brain and three of the downregulated lipids in the brain lesion were sphingolipids, specifically glycosphingolipids (Fig. 2H–J). Glycosphingolipids like hexosylceramide (HexCer) are involved in cell proliferation and inflammation (Düsing et al., 2023). We found HexCer 37:2;O₂, HexCer 40:2;O₂, and HexCer 40:2;O₃ to be notably absent from the brain lesion of the M. tuberculosis-infected brain (Fig. 2H–J). These results taken together indicate disruption in cell proliferation and inflammatory response in and around the brain lesion during M. tuberculosis infection.

3.1.3. Glycerolipids are decreased in brain lesions

Glycerolipids have an important role in cell signaling and are primary components of the cell membrane. Diacylglycerols (DG) are secondary messengers needed for intracellular signaling, components of cell membranes, and precursors to glycerophospholipids (Eichmann and Lass, 2015). We found DG 44:3 to be down in the brain lesion (Fig. 2K). Triacylglycerols (TG) mainly function to store excess fatty acid (Li et al., 2018). We found TG 44:1 to be downregulated in the brain lesion as compared to the uninfected healthy brain (Fig. 2L). These results indicate a breakdown in cell membranes and fatty acids within the brain lesion during M. tuberculosis infection.

3.2. Sterols are decreased in M. tuberculosis-infected brains

The cyclization of squalene-2,3-oxide into lanosterol constitutes the first committed step of the enzymatically regulated, multi-step sterol biosynthesis pathway culminating in cholesterol synthesis (Fig. 3A). The post-squalene pathway is traditionally divided into two branches, the Bloch and Kandutsch-Russell (KR) pathways. While Bloch intermediates at each step can theoretically be converted into KR metabolites by the action of sterol-Δ²⁴-reductase (24-dehydrocholesterol reductase [DHCR24]), previous studies have described conflicting preferences for one branch over the other in the brain (Genaro-Mattos et al., 2019; Mitsche et al., 2015).

Fig. 3.

Sterols are decreased in M. tuberculosis-infected brain. (A) Post-squalene cholesterol biosynthesis pathway demonstrating intermediates measured in LC-MS/MS analysis. (B–C) After three weeks post-injection, the brains of healthy (injected with PBS; grey) and M. tuberculosis-infected (injected with M. tuberculosis H37Rv; blue colors) rabbits were microdissected. Darkest blue (lesion) corresponds to the brain lesion and lighter blue denotes areas farther from brain lesion center (with lightest blue in contralateral, unaffected hemisphere of M. tuberculosis-infected brain). (B) Bloch pathway sterol intermediates. (C) Kandutsch-Russell pathway sterol intermediates. * P < 0.05, ** P < 0.01, *** P < 0.001. Data are represented as median ± interquartile range. Statistical comparisons were made using a Kruskal-Wallis test. Other abbreviations: DHCR7, 7-dehydrocholesterol reductase; DHCR24, 24-dehydrocholesterol reductase or sterol-Δ²⁴-reductase; EBP, emopamil binding protein; LSS, lanosterol synthase; SC5D, lathosterol 5-desaturase. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

To investigate if any major steps of the pathway were differentially impacted by infection with M. tuberculosis, we quantified the level of lanosterol, cholesterol, and 6 major intermediates using LC-MS/MS (Fig. 3B–C). While desmosterol, the Bloch pathway-defining intermediate, was the most abundant species after cholesterol, we found that tissues were also enriched with lathosterol, a KR pathway intermediate, suggesting that both branches may be dominant at specific steps along the overall pathway. Compared to those in uninfected healthy brains, cholesterol and its post-squalene sterol intermediates were significantly decreased in the brain lesion and surrounding tissue (i.e., paralesional) of the M. tuberculosis-infected brain. Notably, levels of DHL, lathosterol, 7-DHD, and desmosterol were suppressed in the hemispheres contralateral to the visible brain lesion of the M. tuberculosis-infected brain (i.e., unaffected) compared to healthy control, resembling a dose-response-like decrease of sterol intermediates with increasing proximity to the brain lesion. Compared to grey matter, white matter in both M. tuberculosis-infected brain and uninfected healthy brains had higher levels of cholesterol and most of the sterol intermediates (Fig. S2).

3.3. Oxysterol are dysregulated in M. tuberculosis-infected brains

3.3.1. Oxysterol products of enzymatic oxidation are decreased in M. tuberculosis-infected brain

Oxysterols encompass numerous compounds which are derived from the oxidation of not only cholesterol, but also various sterols and associated metabolites such as Vitamin D. These bioactive sterol derivatives can be classified into two families based on their genesis: enzymatic and non-enzymatic oxysterol products. Among several possible functional groups (hydroxyl, carbonyl, epoxide) which modify the molecular polarity and biological function, several side-chain hydroxylated oxysterols (24-OHChol, 25-OHChol, 27-OHChol) are known products of enzymatic oxidation (Russell, 2000). 24-OHChol, also known as cerebrosterol due to its generation by neuron-specific enzyme CYP46A1 and ability to cross the blood-brain barrier, has been previously suggested to be a marker of CNS cholesterol turnover (Ramirez et al., 2008). Cholesterol 25-hydroxylase, the enzyme that catalyzes the production of 25-OHChol, is upregulated by Toll-like receptor activation in dendritic cells and macrophages (Bauman et al., 2009; Park and Scott, 2010). Additionally, 25-OHChol has been found to be immunomodulatory through several mechanisms, such as the inhibition of lymphocyte proliferation and IgA class-switching (Bauman et al., 2009; Kato et al., 1998). 27-OHChol, another blood-brain barrier traversing oxysterol, regulates cholesterol metabolism by downregulating sterol-regulated element binding protein-1 (SREBP-1a) and HMGCR, the latter of which is the mechanistic target of statins (An et al., 2017; Kim et al., 2022a). Several oxysterols are also produced nonenzymatically, often driven by free radical species (or ROS) under inflammatory conditions through cholesterol auto-oxidation (Iuliano, 2011). This schema of oxysterol classification is delineated in Fig. 4A.

Fig. 4.

Oxysterols are dysregulated in M. tuberculosis-infected brain. (A) Schematic representation of oxysterols produced by enzymatic reactions (B) and reactive oxygen species (ROS, C). (B–C) After three weeks post-injection, the brains of healthy (injected with PBS; grey) and M. tuberculosis-infected (injected with M. tuberculosis H37Rv; blue colors) rabbits were microdissected. Darkest blue (lesion) corresponds to the brain lesion and lighter blue denotes areas farther from brain lesion center (with lightest blue in contralateral, unaffected hemisphere of M. tuberculosis-infected brain). (B) Oxysterols produced by enzymatic reactions are decreased in M. tuberculosis-infected brains. (C) Oxysterols produced by ROS are increased in M. tuberculosis-infected brains. * P < 0.05, ** P < 0.01, *** P < 0.001. Data are represented as median ± interquartile range. Statistical comparisons were made using a Kruskal-Wallis test. Other abbreviations: CH25H, cholesterol-25-hydroxylase; CYP46, cholesterol-24-hydroxylase; CYP27, human sterol-hydroxylase; ROS, reactive oxygen species. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Consistent with the marked reduction in the parent molecule cholesterol under M. tuberculosis infectious conditions, we found similar dose-response suppression of 24-OHChol and 25-OHChol in regions adjacent to or within the tuberculous brain lesions (Fig. 4B). 27-OHChol, the least abundant enzymatic oxysterol product of the three, was decreased in the unaffected, paralesional, and brain lesion areas although the magnitude of reduction did not correlate with proximity to lesion core (Fig. 4B). Unlike cholesterol and most of the sterol intermediates, the differences between grey matter and white matter in both M. tuberculosis-infected and uninfected healthy brains were not statistically significant, with the more significant differences most saliently observed between M. tuberculosis-infected and uninfected healthy (Fig. S3).

3.3.2. Oxysterols produced by oxidative stress are increased in M. tuberculosis-infected brain

We further quantified levels of non-enzymatically produced oxysterols arising from the free radical oxidation of cholesterol, a significant pathophysiological process observed in bacterial meningitis (Scheld et al., 2002; Yin et al., 2011). Despite the suppression of the endogenous sterol biosynthetic pathway and reduction in enzymatic oxysterol products, we observed elevated levels of lipid peroxidation or oxidative stress markers α-epChol, β-epChol, 7-KetoChol, and DHCAO most pronounced in the brain lesion (Fig. 4C). Our method does not distinguish between 7-αOHChol, a product of cholesterol 7 alpha-hydroxylase (CYP7A1), and 7-βOHChol, a non-enzymatic auto-oxidation product. However, given that CYP7A1 catalyzes the rate-limiting step of the classic bile acid synthesis pathway and is exclusively expressed in hepatocytes, the observed increase in 7-OHChol quantified in the brain lesion can be predominantly attributable to elevated 7-βOHChol in the setting of increased ROS (Chiang and Ferrell, 2020) (Fig. 4C).

4. Discussion

TB meningitis is the most severe form of TB and causes severe neurological sequelae or death in the majority of patients. The brain is a complex, lipid-rich organ with a dynamic lipidome that changes both spatially and temporally during development (Oliveira et al., 2022). These lipids are not only important for structure but also brain function, with roles in myelin stability, neurogenesis, neurotransmitter regulation, and cell signaling to name a few (Yoon et al., 2022). During neuroinflammatory and neurodegenerative conditions, the brain lipidome can significantly change. For example, alterations in sphingolipid and sterol metabolism are found in Alzheimer’s disease patients (Bandaru et al., 2009) and decreased TGs are described in the visual cortex of patients with Parkinson’s disease (Cheng et al., 2011). Using our young rabbit model of TB meningitis, we demonstrated there was significant lipid dysregulation in the M. tuberculosis-infected brain. Through complementary methods, we established that major lipids in the brain - cholesterol, sphingolipids, and glycerophospholipids – were decreased. A total of 25 lipids, across two different MS methods, were found to have differing concentrations in the brain due to TB meningitis.

Using MALDI-MSI, we were able to identify a total of 11 lipids across multiple classes which were significantly different between the brain lesion and the healthy brain. This is the first ever study of brain lipidome in TB meningitis using MALDI-MSI. MALDI-MSI has been used extensively to study drug distributions in pulmonary TB infection (Kokesch-Himmelreich et al., 2022; Prideaux et al., 2011; Prideaux et al., 2015; Treu et al., 2020), as well as to examine pro-inflammatory pathways (Marakalala et al., 2016). Many of the studies that exist focus specifically on lipids associated with the mycobacterium itself and not the host-lipid profile. We focused primarily on the host response to M. tuberculosis infection in the brain lesion and the impact the lesion has on the host lipids, although the mycobacterium is often concentrated at the brain lesion. Of these 11 lipids, six of them were identified as glycerophospholipids which are key structural lipids. Interestingly, in lipidomic studies of plasma from pulmonary TB patients, a number of glycerophospholipids were downregulated. These downregulated lipids did not include the same PCs, PGs, and PAs we found in the brain which may be due to differences in the lung and brain microenvironments (Wood et al., 2018). A similar lipidomic study in pulmonary TB also showed downregulation of certain HexCer sphingolipids similar to those we observed in our data (Anh et al., 2023). Many literature sources show an increase in glycerolipids, specifically TGs, associated with granulomatous lesions in pulmonary TB (Dawa et al., 2021). We observed two glycerolipids, DG 44:3 and TG 44:1, which were decreased in the brain lesion. We are unable to determine if this is a global trend in glycerolipids because MALDI-MSI is not optimized for glycerolipids. It is, however, possible that the trend may be different in pediatric brain than in pulmonary TB infection. Further studies are needed to evaluate the role of glycerolipids in the young, developing brain. The downregulated lipids identified by MALDI-MSI taken together indicate substantial remodeling, breakdown in cell membranes and fatty acids, and disruption in cell proliferation and inflammatory response within the brain lesion compared to the surrounding tissue and the healthy brain.

Using microdissection, we saw a region-specific decrease in cholesterol and its major sterol intermediates in close proximity to M. tuberculosis infection, with the lowest levels in the brain lesion where we have previously shown the bacilli are concentrated (Tucker et al., 2016). M. tuberculosis is known to utilize cholesterol as a carbon source, especially when sequestered in granulomas during pulmonary TB infection (Pandey and Sassetti, 2008). Therefore, the decreased cholesterol may be a result of mycobacterial consumption. However, there was also a decrease in cholesterol and enzymatically produced oxysterols where M. tuberculosis is not found histologically (i.e., in the unaffected hemisphere of the M. tuberculosis-infected brain), implying that these lipid perturbations could be related to widespread neuroinflammation. 7-DHC, a sterol intermediate decreased in M. tuberculosis-infected brain, has recently been shown to suppress ferroptosis, an iron-dependent form of cell death triggered by excessive lipid peroxidation (Freitas et al., 2024). Additionally, M. tuberculosis triggers ferroptosis by secreting protein tyrosine phosphatase A which leads to inhibition of glutathione peroxidase 4, thus promoting mycobacterial dissemination (Qiang et al., 2023). The reduction of 7-DHC in the setting of M. tuberculosis infection may create an extremely pro-ferroptotic environment, making ferroptosis an intriguing pathway that warrants further study in TB meningitis.

The young, developing brain in children may be uniquely vulnerable to alterations in lipid composition and homeostasis, as perturbations in cholesterol metabolism can cause developmental disorders with microcephaly and intellectual disabilities (i.e., Smith-Lemli-Opitz syndrome and 7-DHC reductase) (Porter, 2008). This has important implications for TB meningitis where children <5 years of age are most commonly infected (Seddon et al., 2019). At this age, myelination is still occurring and microglia, the resident brain immune cells, play an important role in neurodevelopment such as axonal guidance and synaptogenesis (Buyanova and Arsalidou, 2021; Kim et al., 2022b; Schafer et al., 2012). The reduction of lipids has a significant impact on brain structure and function, affecting myelination and cell signaling pathways, and may contribute to poor neurodevelopmental outcomes in children after TB meningitis (Rohlwink et al., 2016; Schoeman et al., 2002).

In TB meningitis, microglia are infected by M. tuberculosis and become activated and proinflammatory (Rock et al., 2005; Tucker et al., 2016). Although direct infection is an important cause of microglial activation and neuroinflammation in TB meningitis, alternations in the lipid homeostasis may be another mechanism of microglial activation. A study looking at aging brains found that lipid-droplet accumulation in microglia led to dysfunctional microglia with increased proinflammatory cytokines and oxidative stress and decreased phagocytosis (Marschallinger et al., 2020). Studies have shown that TB meningitis causes oxidative stress, with an in vitro study showing that ROS are upregulated in microglia secondary to M. tuberculosis infection (Yang et al., 2007) and transcriptomic analysis of ventricular CSF in children demonstrating an upregulation of genes associated with nitric oxide (Rohlwink et al., 2019). Our data show significant oxidative stress in brain lesions with increased concentration of only oxysterols produced by ROS (e.g., 7-KetoChol, 7-OHCHol). 7-KetoChol is one of the most common and toxic non-enzymatically produced oxysterols and is well described in atherosclerotic plaques where foamy macrophages, also described in TB pulmonary and brain lesions, are pathologically characteristic (Anderson et al., 2020). Although our current study did not focus on microglia, our previous studies demonstrated increased microglia density and activated morphology colocalizing with brain lesions where mycobacterium bacilli reside (Ruiz-Bedoya et al., 2022; Tucker et al., 2016), suggesting that activated microglia may have an important role in oxidative stress, especially lipid peroxidation.

Although the role of lipids and their metabolism in TB meningitis is understudied compared to pulmonary TB, several studies have investigated the role of fatty acid metabolism through the genetic polymorphisms of the enzyme, LTA4H, that regulates the conversion of arachnoid acid to eicosanoids, many of which are proinflammatory mediators (i.e., leukotriene B4). First studied in zebrafish, LTA4H influences the balance of proinflammatory and anti-inflammatory eicosanoids and has been found to affect survival outcomes in TB meningitis in conjunction with corticosteroid treatment in many populations (Thuong et al., 2017; Tobin et al., 2012; Tobin et al., 2010), although not all populations (van Laarhoven et al., 2017). Interestingly, 7-KetoChol, a non-enzymatically produced oxysterol we found increased in brain lesions, is known to induce apoptosis and increase transcription of phospholipase A2 genes that releases arachidonic acid from plasma membranes (Anderson et al., 2020; Cheng et al., 2021). Metabolomic pathway analysis of the CSF of TB meningitis patients also showed a modest enrichment of glycerophospholipid, arachidonic acid and sphingolipid metabolism (van Laarhoven et al., 2018). Two studies recently investigated the effect of aspirin, a host-directed therapy tested in several TB meningitis trials, on several lipid mediators, including those produced by the arachidonic acid pathway (i.e., thromboxane and prostaglandin) and those termed specialized pro-resolving mediators (SPMs) (Colas et al., 2019; Mai et al., 2018). SPMs are host protective lipid mediators produced from essential fatty acids and can be produced by several cell types, including microglia (Chen et al., 2018; Serhan et al., 2002). Colas et al. found that SPMs were lower in those with more severe TB meningitis at presentation and predicted mortality, and treatment with aspirin increased SPMs in both studies (Colas et al., 2019; Mai et al., 2018). Although the current study provides insights into the effect M. tuberculosis has on the brain lipidome, we did not investigate the concentration of these lipid mediators, such as SPMs or eicosanoids, which may have important implications for pathogenesis and response to treatment.

Our study has some limitations. We performed postmortem analysis at a single time point which may simplify the temporal changes that occur to the brain lipidome during M. tuberculosis infection. However, we accounted for intrinsic lipidomic variability seen during neurodevelopment by comparing M. tuberculosis-infected brains to age-matched healthy brains. Additionally, the time point was chosen based on gross pathological lesions and symptomatology in previous cohorts. There could be sampling bias from the microdissections for the LC-MS/MS studies, however a similar decrease in lipid species was shown with MALDI-MSI which retains spatial integrity. Future studies utilizing noninvasive imaging techniques (e.g., positron emission tomography [PET]) could complement the spatial resolution of MALDI-MSI and allow for longitudinal monitoring in live animals. Although we did not test the effect of treatment, antimicrobial or host-directed therapy may induce changes to the lipidome that need to be further elucidated.

5. Conclusions

Our study is the first spatial analysis of lipid dysregulation in brain parenchyma during TB meningitis infection in the young, developing brain. Through complementary methods, we demonstrated that major lipids in the brain – cholesterol, sphingolipids, and glycerophospholipids – were decreased in brain lesions, with many showing a region-specific decrease as proximity to brain lesions (with M. tuberculosis bacilli) increased. Conversely, we show significant, proximity-dependent lipid peroxidation in brain lesions with increased concentration of oxysterols produced by ROS in the lesion center. Our data provide insights into possible mechanisms, such as increased lipid peroxidation and reduced lipid concentration, that may contribute to poor neurodevelopmental outcomes in children after TB meningitis. Future studies to further investigate the role of microglia, ferroptosis and SPMs in TB meningitis pathogenesis are needed.

Supplementary Material

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.nbd.2024.106721.

Abbreviations

AUC

area under the curve

BHT

butylated hydroxytoluene

CHCl₃

chloroform

CH₃CN

acetonitrile

CH25H

cholesterol-25-hydroxylase

Chol

cholesterol

CNS

central nervous system

CSF

cerebrospinal fluid

CYP7A1

cholesterol 7 alpha-hydroxylase

CYP46

cholesterol-24-hydroxylase

CYP27

human sterol-hydroxylase

Des

desmosterol

DG

diradylglycerolipid

DHB

2,5-dihydroxybenzoic acid

7-DHC

7-dehydrocholesterol

DHCAO

3,5-dihydroxycholestanone

DHCR7

7-dehydrocholesterol reductase

DHCR24

sterol-Δ²⁴-reductase

7-DHD

7-dehydrodesmosterol

DHL

24-dehydrolathosterol

dHLan

dihydrolanosterol

DMAP

dimethylaminopyridine

DMG

N,N-dimethylglycine

EBP

emopamil binding protein

α-epChol

5,6-α epoxycholesterol

β-epChol

5,6-β epoxycholesterol

ESI

electrospray ionization

Et₃N

trimethylamine

HexCer

hexosylceramide

HPLC

high-performance liquid chromatography

7-KetoChol

7-ketocholesterol

KR

Kandutsch-Russell

Lan

lanosterol

Lath

lathosterol

LC

liquid chromatography

LSS

lanosterol synthase

LTA4H

leukotriene A4 hydrolase

MALDI

matrix assisted laser desorption/ionization

MeOH

methanol

MS

mass spectrometry

NH₄OAc

ammonium acetate

7-OHChol

7-hydroxycholesterol

24-OHChol

24-hydroxycholesterol

25-OHChol

25-hydroxycholesterol

27-OHChol

27-hydroxycholesterol

PA

glycerophosphate

PBS

phosphate buffered saline

PC

glycerophosphocholine

PE

glycerophosphoethanolamine

PG

glycerophosphoglycerol

PSI

pounds per square inch

ROC

receiver operating characteristic

ROI

region of interest

ROS

reactive oxygen species

SC5D

lathosterol 5-desaturase

SPM

specialized pro-resolving mediator

SRM

selected reaction monitoring

TB

tuberculosis

TG

triadylglycerolipid

TPP

triphenylphosphine

Data availability

Data will be made available on request.