Infection with Mycobacterium tuberculosis depletes host fatty acids in humans and non-human primates

Summary

Background

The ways in which Mycobacterium tuberculosis (Mtb) infection affects host metabolism and its implications for disease pathogenesis remain poorly understood. Understanding how Mtb shapes the metabolic state of the host could lead to nutritional and host-directed therapies to improve TB outcomes.

Methods

We performed high-resolution metabolic and lipid profiling on male rhesus macaques infected with Mtb (n = 17). Serial plasma samples were analysed beginning at the time of infection and up to 15–16 weeks post infection. Observed host metabolic changes were examined in a cohort of Mtb-infected (TBI, positive QuantiFERON Gold [QFT]; n = 92) and Mtb-uninfected (TBU, negative QFT; n = 102) participants as well as those with pulmonary TB (PTB; n = 68).

Findings

Using principal component analysis, we found the greatest metabolic changes in NHPs occurred 15–16 weeks after experimental infection. Metabolic changes were characterised by declines in saturated, monounsaturated, and polyunsaturated long chain fatty acids (LCFAs) including linoleic acid, linolenic acid, docosahexaenoic acid, and palmitoleic acid. Humans with TBI also experienced significant declines in plasma concentrations of nearly all species of LCFAs 6 months after Mtb exposure. In participants with PTB, LCFAs were further depleted relative to those with TBI but gradually normalised six months post-TB treatment.

Interpretation

Our study shows the host metabolic response to infection with Mtb is characterised by the systemic depletion of host lipids, which is exacerbated in persons with PTB. Interventions that supplement host lipids should be tested to determine their impact on TB outcomes.

Funding

This work was supported by grants from the U.S. National Institute of Allergy and Infectious Diseases (NIAID) [R01 AI182244, R21 AI178324, K23 AI144040, P30 AI168386, P30 AI050409, K24 AI114444, U19 AI111211]; and the National Center for Advancing Translational Sciences [UL1 TR002378], Bethesda, MD, USA. The study sponsors had no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the manuscript; or in the decision to submit the manuscript for publication.

Research in context.

Evidence before this study

It has long been evident that Mycobacterium tuberculosis (Mtb) has a bidirectional relationship with host metabolism. Before Koch's discovery of the tubercle bacillus, tuberculosis (TB) disease was often referred to as “consumption” due to the wasting syndrome induced in the host. At the same time, a low BMI at the time of Mtb infection greatly increases the risk of progression to TB disease. This effect is most pronounced in people with low BMI, but individuals with a BMI in the normal or overweight range are also at greater risk of TB disease versus those with a BMI in the obese range. While these observations suggest there is an evolving, bidirectional relationship between TB and host metabolism, there is little available data from either human or non-human primate (NHP) studies to know how host metabolism changes after infection with Mtb. High-resolution metabolomics is a technology that has emerged over the last decade to provide an increasingly granular assessment of host metabolism. To date, most TB metabolomics studies have focused on metabolic changes observed at the time of TB diagnosis, when clinical disease is already evident. This makes it difficult to know which observed metabolic changes are the result of Mtb infection versus clinical sequelae of TB disease. A deeper understanding of the molecular metabolic response to Mtb infection is needed to help inform nutritional interventions with the potential to improve TB outcomes.

Added value of this study

Here, we use high-resolution metabolomics performed on plasma samples from experimentally infected NHPs, as well as rigorously conducted human cohorts, to describe a common host metabolic response to Mtb. We found that NHPs experience depletion of long chain fatty acids 15–16 weeks after infection. This metabolic response was reproduced in a cohort of human household contacts exposed to persons with pulmonary TB for over 5 h per day for at least 5 days per week. In household contacts not progressing to TB disease, fatty acid depletion was transient and resolved without intervention, whereas participants with TB disease had more profound depletion, particularly of polyunsaturated fatty acids. This metabolic response resolved with TB treatment, but plasma concentrations of long chain fatty acids did not normalise until six months post-TB treatment completion.

Implications of all the available evidence

These data provide evidence for a common host metabolic response to Mtb across humans and NHPs, characterised by fatty acid depletion. They also offer crucial context for multiple clinical and epidemiologic observations including the association between low BMI and risk for TB disease, as well as the lower risk for TB progression in those receiving nutritional supplementation. In future studies, it will be essential to better understand the role of increased host fatty acid utilisation after Mtb infection. While mouse models indicate dietary supplementation of long chain fatty acids may reduce Mtb burden and limit immunopathology, further investigation is warranted to determine whether supplementation of mono- and/or polyunsaturated long chain fatty acids either after Mtb infection or at pulmonary TB diagnosis could improve clinical outcomes in humans.

Introduction

An estimated 1.23 million people die of tuberculosis (TB) disease each year, making it the leading global cause of infectious disease death.¹ For centuries, TB disease was colloquially referred to as “consumption” due to the severe wasting it induces in the human host.² Yet the relationship between TB disease and host metabolism remains poorly understood. This is due, in part, to the bidirectional nature of the relationship. While low BMI is highly prevalent among persons with TB disease³ and associated with poor treatment outcomes,4, 5, 6, 7, 8 it is also a TB risk factor.^9,10 WHO estimates that nearly one million people develop TB disease each year due to undernutrition (BMI <18.5 kg/m²); greater than HIV, smoking, or diabetes.¹ In a cluster randomised trial from 2023, household contacts of people with pulmonary TB receiving a nutritional support intervention had a 40% reduction in TB incidence compared to those not receiving the intervention.¹¹ Together, these observations indicate that host metabolic abnormalities can be both a cause and effect of TB disease.⁹

However, despite the long-standing clinical observation that people with TB disease frequently experience weight loss and wasting, little is known about how Mtb infection itself impacts host metabolism. Most descriptions of the metabolic sequelae of TB disease in humans come from cross-sectional evaluations of participants at the time of TB diagnosis.^12,13 It is often difficult to determine which observed metabolic abnormalities were present prior to progression to TB disease, which were caused by TB symptoms such as anorexia, and which might be caused by Mtb. This may explain why nutritional markers such as vitamin A and vitamin D, which are depleted at the time of TB diagnosis,14, 15, 16 have had little to no impact on TB outcomes when provided as supplemention.17, 18, 19 Determining how Mtb infection and disease impact host metabolism is critical to understand the ways in which Mtb and metabolic abnormalities intersect to drive progression to TB disease and worsen TB outcomes.

With recent advances in high-resolution metabolomics (HRM), the prospect of detailed metabolic phenotyping across human and animal TB studies has become attainable. This technology measures thousands of metabolic features in biologic samples at relatively low cost, allowing for simultaneous measurement of numerous metabolic parameters in large numbers of biologic specimens. Using a combination of untargeted and targeted HRM methods, we performed detailed metabolic phenotyping of plasma samples from multiple cohorts of non-human primates (NHPs) experimentally infected with Mtb.^20,21 We then examined our findings in a human cohort of household contacts (HHCs) with Mtb infection (TBI) and a separate human cohort with pulmonary TB (PTB). By performing detailed metabolomic analysis of serial plasma samples from experimental infection studies and well-characterised human cohorts, we sought to describe the host metabolic changes that result from Mtb infection. Our secondary objective was to determine whether observed metabolic responses to Mtb infection were exacerbated in those with TB disease, and whether such changes were reversed with TB treatment.

Methods

Rhesus macaque infections

Full data from the animals in this study were previous reported.^20,21 We analysed plasma samples from three cohorts of male Indian-origin rhesus macaques infected with Mtb H37Rv (GenBank Accession AL123456) instilled in 2 mL of saline via a bronchoscope: 1) animals infected with 30–50 CFU of Mtb (n = 6), 2) animals infected with 30–50 CFU of Mtb and treated with anti-PD-1 blocking mAb every 2 weeks (n = 6; anti-PD-1 humanised clone EH12-2132/2133), and 3) animals infected with 120–150 CFU of Mtb (n = 5). The anti-PD-1 antibody was developed by Gordon Freeman. The antibody was shown to block interaction between PD-1 and its ligands in vitro, with specificity for PD-1 in Indian-origin rhesus macaques confirmed previously.²² The first available plasma sample for each animal was analysed as the baseline sample. For most animals (n = 10) this was taken before Mtb infection (week 0) and for the remainder (n = 7) it was taken shortly after Mtb infection (week 2). Serial plasma samples were then analysed at week 3–5, week 10–13, and week 15–16 after infection (Fig. 1). One animal developed respiratory distress and was euthanised at 11 weeks. All animals were humanely euthanised with intravenous Beuthanasia-D 1 mL/10 lbs animal weight administered by trained personnel according to the guidelines of the American Veterinary Medical Association.²³ No animals were excluded. A summary of Mtb burden and PET-CT findings at the time of necropsy is provided in Supplemental Table S1).

Fig. 1.

Rhesus macaque study design and metabolomics workflow: (A) Design of the rhesus macaque experiments for which plasma samples were analysed. (B) High-resolution metabolomics workflow. Created in BioRender. Collins, J. (2026) https://BioRender.com/odi0ile.

Household contact cohort

We prospectively enrolled HHCs exposed for ≥5 h/day and ≥5 days/week to a person with PTB and a positive sputum acid-fast bacilli (AFB) smear or medium/high Xpert MTB result in Addis Ababa, Ethiopia from December 2017 to July 2021 (Fig. 2A). HIV seronegative participants ≥15 years were eligible for enrolment and were evaluated for TB infection with a QuantiFERON® Gold Test (QFT). Male and female participants were enrolled, and sex was obtained by self-report (see Supplemental Table S2 for full inclusion and exclusion criteria). Participants with a positive QFT were considered to have TBI²⁴ and were evaluated every six months for a total of 24 months to monitor for disease progression with plasma samples collected at each visit. Participants did not receive TB preventive treatment as is standard of care for HIV-negative adults in Ethiopia. Study visits included a clinical evaluation and symptom questionnaire. If there was clinical concern for progression to TB disease, a chest x-ray and sputum testing with Xpert MTB/RIF and Mtb culture were performed. Participants that developed TB symptoms between visits also underwent an unscheduled evaluation.

Fig. 2.

Flow chart of human study participants: (A) Flow diagram of Ethiopian study participants enrolled in the household contact cohort. This included participants that had a Mtb infection (TBI) as evidenced by a positive Quantiferon (n = 92) as well as those without Mtb infection (TBU) as indicated by a negative Quantiferon (n = 102). (B) Study flow diagram of participants enrolled in the pulmonary TB treatment cohort (n = 68). Created in BioRender. Collins, J. (2026) https://BioRender.com/9wamplp.

TB disease cohort

Persons diagnosed with PTB²⁴ were enrolled in Addis Ababa, Ethiopia from December 2019 to July 2022 (Fig. 2B). Participants who: were ≥18 years old; did not have a prior history of TB disease; were HIV seronegative; had a positive sputum Xpert MTB/RIF and Mtb culture, were eligible for enrolment. Patients received a standardised treatment regimen that included two months of rifampin, isoniazid, ethambutol and pyrazinamide followed by four additional months of rifampin and isoniazid administered via directly observed therapy. Plasma samples were collected within 5 days (median 3 days) and at 2 and 6 months after TB treatment initiation, as well as 6 months post-TB treatment completion (recovery period).

Ethics

Written informed consent was obtained from all study participants, and study approval was obtained from the institutional review boards (IRBs) of Armauer Hansen Research Institute (Addis Ababa, Ethiopia; Protocol P005/18), the Ethiopian National Research Ethics Review Committee (Addis Ababa, Ethiopia; Approval 02/246/994/22), and Emory University (Atlanta, Georgia, USA; Study IRB00100966). Rhesus macaques were housed in biocontainment racks and maintained in accordance with the Animal Welfare Act, the Guide for the Care and Use of Laboratory Animals, and all applicable regulations, standards, and policies of a fully AAALAC International accredited Animal Biosafety Level 3 vivarium. All procedures were performed utilising appropriate anaesthetics as listed in a NIAID DIR Animal Care and Use Committee (ACUC) approved animal study proposal LPD-25E. Animals received enrichment and husbandry and technical staff monitored behaviour daily, with investigators notified of any change in condition.^20,21

Untargeted metabolomics analysis

All samples analysed for this study were collected using EDTA-containing tubes, with plasma isolated and stored at −80 °C prior to analysis. De-identified plasma samples were randomised 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 (50 μl) was treated with 130 μl acetonitrile (2:1, v/v) containing an internal isotopic standard mixture (3.5 μl/sample), as previously described.²⁵ Samples were mixed and placed on ice for 30 min prior to centrifugation to remove protein. The resulting supernatant was transferred to low-volume autosampler vials maintained at 4 °C. We performed dual hydrophilic interaction liquid chromatography (HILIC) and C18 liquid chromatography with positive and negative electrospray ionisation (ESI) respectively, using an injection volume of 10 μl. Analyte separation for HILIC was performed with a Waters XBridge BEH Amide XP HILIC column (2.1 × 50 mm², 2.6 μm particle size) and gradient elution with mobile phases A: LCMS grade water, B: LCMS grade acetonitrile, C: 2% formic acid. The initial 1.5 min period consisted of 22.5% A, 75% B, and 2.5% C, followed by a linear increase to 75% A, 22.5% B, and 2.5% C at 4 min and a final hold of 1 min. C18 chromatography was performed on an end-capped C18 column (Higgins Targa C18 2.1 × 50 mm², 3 μm particle size) with mobile phases A: water, B: acetonitrile, C: 10 mM ammonium acetate. The initial 1 min period consisted of 60% A, 35% B, and 5% C followed by a linear increase to 0% A, 95% B, and 5% C at 3 min and held for the remaining 2 min. The mobile phase flow rate was 0.35 mL/min for the first min, and increased to 0.4 mL/min for the final 4 min. High-resolution mass spectrometry was performed using an Orbitrap Q Exactive Mass Spectrometer (Thermo Scientific, San Jose, CA, USA) at 120k resolution operated over a scan range of 85–1275 mass/charge (m/z). Tune parameters for sheath gas were 45 for ESI+ and 30 for ESI−. Auxiliary gas was set at 25 for ESI+ and 5 for ESI−. Spray voltage was set at 3.5 kV for ESI+ and −3.0 kV for ESI−. Data were stored as. Raw files²⁶ and extraction and alignment were performed using apLCMS (version 6.6.8)27 and xMSanalyzer (version 2.0.7.999)²⁸ with each feature defined by specific m/z value, retention time and integrated ion intensity.²⁶ Three technical replicates were performed for each plasma sample and intensity values were median summarised.²⁹ Identities of select targeted metabolites were confirmed and quantified by accurate mass, MS/MS and retention time relative to authentic standards.³⁰

Targeted measurement of fatty acids

Non esterified fatty acids (NEFA) were analysed from plasma samples using a LC/MS based method.³¹ Briefly, 100 μl of plasma was vortexed with 1 mL 50:50 Acetonitrile:Methanol. Samples were then centrifuged (4000 rpm 10 min) to pellet any solids; the supernatant was retained for the remainder of the analysis. Fatty acids in solution were derivatised using 20 μl of 200 mM 3-Nitrophenylhydrazine (3-NPH), 20 μl 20 mL of 120 mM N-(3-dimethylaminopropyl)-N′-ethylcarodiimide (EDC), and 20 mL of 6% Pyridine at 40 °C for 30 min. External calibration curves were prepared for fatty acids in the linear range of 0.005 μM–10 μM. Samples were analysed by liquid chromatography (LC) using a Sciex ExionLC AC system on a 20-min linear gradient consisting of 0.1% formic acid in water (Solvent A) and 0.1% formic acid in acetonitrile (Solvent B). The LC system was coupled to a triple quadrupole mass spectrometer (QTRAP5500, ABSciex) using a multiple reaction monitoring (MRM)-based method with negative polarity and spray voltage of 4500 V, ion source gas 1 of 60 psi, ion source gas 2 of 50 psi, curtain gas of 20 psi, and CAD gas of 6, with a source temperature of 650 °C. See Supplemental Tables S3 and S4 for full LC-MS running conditions and MRM transitions respectively.

Targeted measurement of bile acids

Bile acids were extracted from 100 μl of plasma using 500 μl of a 1:4 v/v water:methanol solution. Samples were centrifuged at 4000 rpm for 5 min at 4 °C and supernatants were removed to a new vial. Methanol was added to precipitate the sample, and the extract was dried under nitrogen. The samples were then reconstituted with 200 μl of methanol. Samples were analysed using a 1290 Infinity II Agilent Zorbax Eclipse Plus C18 (2.1 × 100 mm, 1.8 μm) liquid chromatography column at flow rate of 0.5 mL/min at 65 °C during a 21-min gradient coupled to an Agilent 6495C triple quadrupole mass spectrometer. The mobile phase of UPLC grade solvents consisted of solvent A: 0.1% formic acid in water and Solvent B: 0.1% formic acid in acetonitrile. The sample injection volume was 5 μl, and analysis was completed in negative mode. Bile acids were identified and quantified by MRM relative to authentic standard curves in the linear range of 0.01 nM–10 μM using external standards. See Supplemental Tables S5 and S6 for full running conditions and MRM transitions respectively.

Statistics

All comparisons were performed in R version 4.4.1. Metabolite intensity values and concentrations were log₂ transformed for statistical comparison. For untargeted analyses, metabolites with >50% missing values were excluded from the analysis. For remaining missing values, half the minimum value for a feature intensity or metabolite concentration was used. Changes in metabolite concentrations over time in humans and NHPs were assessed using a repeated measures ANOVA to control for interindividual variability and visualised using principal component analysis (PCA). Metabolic pathway enrichment analysis was performed using mummichog (version 2.7), a Python-based informatics tool that leverages the organisation of metabolic networks to predict functional changes in metabolic pathway activity.³² Cross-sectional comparisons of metabolite concentrations between the TBI and TBU groups, as well as persons with TBI and those with PTB, were performed using linear regression, controlling for age, sex, and BMI to minimise any confounding of these variables on the metabolome. We used a Benjamini-Hochberg false discovery rate (FDR) to account for multiple comparisons.³³ In the initial discovery analysis, a less stringent q-value (FDR-adjusted p-value) of 0.2 was used to minimise type II error. For targeted analyses, a q-value ≤0.05 was considered statistically significant. Effect sizes between groups and time points were standardised across metabolites by calculating the log2 fold change: a logarithmic representation of the concentration ratio between two conditions.

Role of funders

The study sponsors had no role in the study design; in the collection, analysis, and interpretation of data; in the writing of the manuscript; or in the decision to submit the manuscript for publication.

Results

Metabolic changes after Mtb infection of NHPs

We first sought to understand how and when the plasma metabolome changed in NHPs following experimental Mtb infection in the absence of antibiotic therapy. In three cohorts of NHPs (n = 17 total), we performed untargeted HRM on plasma obtained either prior to (n = 10) or 2 weeks after (n = 7) Mtb infection, then serially at 3–5 weeks, 10–13 weeks and 15–16 weeks post-Mtb infection. We then assessed which metabolites were differentially abundant at any follow up time point relative to baseline (2-way repeat measures ANOVA; q-value<0.2), controlling for either receipt of a PD1 inhibitor (received by six animals) or infection with a higher dose of Mtb (five animals received 120–150 CFU rather than 30–50 CFU. For positive and negative ionisation mode in both cases, we found the largest number of metabolite features were significantly associated with time post-Mtb infection, with relatively few metabolites changing over time based on anti-PD1 treatment or infectious Mtb dose (Supplemental Fig. S1A–H). PCA analysis revealed that the 15–16-week time point had clear separation from the other three time points, indicating the time of greatest metabolic change (Fig. 3A and B). Comparing metabolite concentrations at the 15–16-week time point versus baseline, we identified 944 differentially abundant metabolic features in negative ionisation mode and 231 in positive ionisation mode (Fig. 3C and D). Using metabolic pathway analysis, we found metabolites that declined after Mtb infection mapped primarily to overlapping pathways in lipid and bile acid metabolism while those metabolites that increased mapped to tyrosine metabolism and the valine, leucine, and isoleucine degradation pathways (Fig. 3E).

Fig. 3.

Untargeted and pathway analysis of the plasma metabolome following Mtb infection of rhesus macaques: (A) Principal component analysis of metabolic features that significantly change over time in plasma following Mtb infection of rhesus macaques (n = 17) measured in positive ionisation mode using HILIC chromatography and (B) negative ionisation mode using c18 chromatography. (C) Volcano plots of plasma metabolic features significantly increased 15–16 weeks after Mtb infection versus baseline (red; repeat measures ANOVA q-value <0.2) and those significantly decreased (blue) using positive ionisation mode with HILIC chromatography and (D) negative ionisation mode with c18 chromatography. (E) Metabolic pathways of plasma metabolic features that significantly increased 15–16 weeks after Mtb infection (orange) and those that significantly decreased (blue; P < 0.05) based on the mummichog pathway analysis program.³²

Using reference standardisation,³⁰ we verified and quantified a targeted set of metabolites mapping to statistically significant metabolic pathways. Quantified metabolites that significantly differed for at least one post-infection time point versus baseline (repeat measures ANOVA q ≤ 0.05) are shown in Fig. 4A. We found there were significant declines in saturated and unsaturated LCFAs, acyl carnitines, and bile acids following Mtb infection relative to baseline. These classes of molecules followed a similar trajectory with a smaller magnitude decrease in concentration 3–5 weeks after infection, followed by recovery to baseline concentrations at the 10–13-week time point and then a larger magnitude drop at 15–16 weeks (Fig. 4B). At the 15–16-week time point, the concentration of plasma LCFAs was significantly correlated with concentrations of valerobetaine (Pearson r > 0.5 and P < 0.05; Fig. 4C), a microbiome-derived metabolite known to inhibit fatty acid oxidation.³⁴

Fig. 4.

Targeted analysis of plasma metabolomic changes in rhesus macaques after Mtb infection: In rhesus macaques (n = 17) (A) shows a balloon plot with annotated and quantified metabolites that differed significantly from baseline for at least one time point following Mtb infection (repeat measures ANOVA q < 0.05). The colour scale shows the log2 fold change in metabolite concentration at each time point versus baseline while the size of each balloon is determined by the -log10 p-value of the association. (B) Line plot with dots showing the mean and error bars showing the standard error of plasma concentrations of palmitoleic acid, linolenic acid, and docosahexaenoic acid following M. tuberculosis infection of rhesus macaques (repeat measures ANOVA versus baseline; ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001). (C) Correlation plot showing a network of highly correlated metabolites (Pearson r > 0.5 and p < 0.05) that included multiple species of long chain fatty acids and valerobetaine, a microbiome-derived inhibitor of beta-oxidation. (D) Targeted analysis of fatty acids and bile acids showing log2 fold change increase (orange) or decrease (blue) in plasma 15–16 weeks after M. tuberculosis infection versus baseline (repeat measures ANOVA; ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001).

To gain a more detailed understanding of the changes in fatty acid and bile acid metabolism, we performed a targeted analysis of these pathways to quantify additional metabolites not well detected by the metabolomics method. Of the short chain fatty acids, we found formic acid was significantly increased 15–16 weeks after infection versus baseline (0.39 versus 0.34 μM; repeat measures ANOVA P = 0.02), while concentrations of propionic acid (2.29 versus 2.29 μM; P = 0.99) and valeric acid (0.33 versus 0.38 μM; P = 0.27) were not significantly changed (Fig. 4D). However, the microbiome-derived butyric acid significantly declined after Mtb infection (1.04 versus 1.95 μM; P = 0.001). Similar to the findings from our metabolomics analysis, we found medium chain fatty acids, as well as saturated and unsaturated LCFAs declined. In the bile acid pathway, both conjugated bile acids and microbiome-derived secondary bile acids significantly declined post-Mtb infection. Together, these results show experimental Mtb infection of rhesus macaques profoundly reshapes host metabolism at a systemic level. Metabolic changes in NHPs after Mtb infection are characterised by significantly reduced plasma concentrations of bile acids, as well as saturated and unsaturated LCFAs. Further, these changes are strongly associated with declines in microbially-derived metabolites including butyrate, valerobetaine, and secondary bile acids, suggesting an important component of cross talk with the functional microbiome.

Plasma LCFAs decline in humans infected with Mtb

To evaluate if a similar metabolic response to Mtb infection was evident in human cohorts, we analysed plasma samples from 194 Ethiopian HHCs of whom 92 had evidence of Mtb infection (TBI) with a positive QFT and 102 that were Mtb uninfected (TBU) as evidenced by a negative QFT at cohort enrolment (Table 1, Supplemental Table S7, and Fig. 5A). The TBU group was 35% male and had a median age of 26 years (IQR 21–35), while the TBI group was 57% male with a median age of 34 years (IQR 26–39). The TBI group was followed for 2 years with plasma collected every six months during which time none progressed to TB disease. At the time of enrolment (<3 months after exposure), there were no significant differences between the TBI and TBU groups (Fig. 5B). However, at the 6-month follow-up visit, plasma concentrations of numerous saturated and unsaturated LCFAs had significantly decreased in TBI participants (Fig. 5C) compared to baseline. Similar to the trajectory of palmitoleic acid in Fig. 5D, saturated and unsaturated LCFAs declined 6 months after Mtb exposure, with concentrations increasing again at the 12-month follow up visit. These results suggest humans and NHPs have a shared metabolic response to Mtb infection characterised by depletion of both saturated and unsaturated LCFAs.

Table 1.

Demographic data of Ethiopian study participants.

Variables	Pulmonary TB (PTB; n = 68)	Mtb infected household contacts (TBI; n = 92)	Mtb uninfected household contacts (n = 102)	TB progressors (n = 8)	P-valuea
Sex, n (%)
Male	44 (65)	52 (57)	36 (35)	4 (50)	<0.001
Female	24 (35)	40 (43)	66 (65)	4 (50)
Age, median (IQR)	26 (21, 35)	34 (26, 39)	26 (23, 36)	22 (20,26)	<0.001
Never Smoker, n (%)	59 (87)	82 (89)	98 (96)	8 (100)	0.10
No alcohol consumption, n (%)	48 (71)	43 (47)	53 (52)	5 (62.5)	0.02
BMI, n (%)a
<18.5	31 (46)	15 (16)	24 (24)	4 (50)	<0.001
18.5–25	37 (54)	53 (58)	65 (64)	4 (50)
25.1–30	0 (0)	18 (20)	9 (9)	0 (0)
>30	0 (0)	5 (5)	3 (3)	0 (0)
History of diabetes, n (%)	3 (4)	4 (4)	4 (4)	0 (0)	0.94
AFB sputum smear, n (%)
Negative/Scanty	15 (22)	–	–	–
1–2+	36 (53)
3–4+	16 (24)
Positive Xpert MTB/RIF, n (%)	100 (100)	–	–	–
Positive Mtb culture, n (%)	100 (100)	–	–	–
Mtb culture positive after 2 months, n (%)b	3 (6)	–	–	–
Cavity on chest x-ray, n (%)	29 (43)	–	–	–

QFT, Quantiferon; Mtb, Mycobacterium tuberculosis.

^aDerived using chi-square test for categorical variables and ANOVA for continuous variables.

^b17 participants were missing a culture result at 2 months.

Fig. 5.

Plasma metabolic changes following Mtb exposure in humans: (A) Design of the longitudinal household contact cohort. Created in BioRender. Collins, J. (2026) https://BioRender.com/vznyr73 (B) Bubble plot showing targeted metabolites that significantly changed in Mtb infected contacts (TBI; n = 92) after high-level exposure to a person with pulmonary TB (PTB; n = 54). (C) Volcano plot of targeted metabolites that significantly increased (red) and decreased (blue) in the plasma of persons with TBI 6 months after exposure to Mtb versus baseline (n = 80). (D) Plasma palmitoleic acid concentrations in TBI (n = 92) versus Mtb uninfected (TBU; n = 102) household contacts, as well as the change over time in persons with TBI (n = 80). The box plots depict the minimum and maximum values (whiskers), the upper and lower quartiles, and the median. The length of the box represents the interquartile range. A repeat measures ANOVA test was used to compare metabolite changes over time and a linear regression model controlling for age, sex, and body mass index was used for cross-sectional comparisons between groups (∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001).

Plasma LCFAs decline further in TB disease

Given asymptomatic persons with TBI were found to have a decrease in plasma concentrations of multiple LCFAs six months after Mtb exposure that partially recovered 12 months after exposure, we sought to determine whether greater reductions in plasma concentrations of LCFAs were evident in persons with PTB. We analysed plasma samples from a cohort of 68 Ethiopian adults with PTB, of whom 44 (65%) were male with a median age of 26 years (IQR 21–35). All persons with PTB had both a positive sputum Xpert MTB test result and positive sputum culture for Mtb. Of these, 61% also had a ≥1+ AFB positive sputum smear. We initially compared plasma metabolite concentrations in a subset of persons with PTB (n = 54) to persons with TBI six months after exposure to Mtb. Participants with PTB exhibited significantly lower concentrations of nearly all plasma LCFAs. Decreases in polyunsaturated LCFAs were more pronounced in PTB participants (Fig. 6A–F), though plasma concentrations of saturated and monounsaturated LCFAs were also significantly lower (Supplemental Fig. S2A–F).

Fig. 6.

Plasma concentrations of polyunsaturated long chain fatty acid in participants with Mtb infection and pulmonary TB: Plasma concentrations of (A) homolinolenic acid, (B) docosapentaenoic acid, (C) linoleic acid, (D) eicosadienoic acid, (E) linolenic acid, and (F) docosahexaenoic acid in household contacts with Mtb infection (TBI) at enrolment (n = 92) and after 6 months (n = 80) as well as participants with pulmonary TB (PTB) at diagnosis (n = 54). Participants with TB disease were compared with household contacts using a linear regression model controlling for age, sex, and body mass index while change in metabolite concentration over time in contacts was compared using a repeat measures ANOVA (∗P ≤ 0.05, ∗∗P < 0.01, and ∗∗∗P < 0.001). The box plots depict the minimum and maximum values (whiskers), the upper and lower quartiles, and the median. The length of the box represents the interquartile range.

To understand the time course of the decline in plasma LCFA concentrations prior to PTB diagnosis, we analysed baseline plasma samples from the HHCs with TBI that progressed to TB disease during the study follow up period (n = 8). Six (75%) were diagnosed with PTB and two were diagnosed with extrapulmonary TB. We found that at the time of study enrolment (2–13 months prior to TB disease diagnosis), plasma LCFA concentrations did not significantly differ in TB progressors versus those with TBI who did not progress during the two-year follow up period (Supplemental Fig. S3A–L). While this sample size is limited, it indicates that low plasma LCFA concretions did not precede TB disease progression.

We then analysed plasma samples from the full cohort of PTB participants (Fig. 7A), most of whom also had plasma samples collected two months after treatment start (n = 64; 94%), six months after treatment start (n = 58; 85%), and six months post-TB treatment (n = 33; 49%). Following TB treatment initiation, we found plasma concentrations of all saturated and unsaturated LCFAs significantly increased (Fig. 7B). However, the increase was gradual, with concentrations not significantly rising until the end of TB treatment and six months post-TB treatment. The unsaturated LCFAs docosahexaenoic acid, linolenic acid, and palmitoleic acid had representative trajectories as depicted in Fig. 7C. For nearly all LCFAs including docosapentaenoic acid, this increase continued until six months post-TB treatment (Fig. 7D).

Fig. 7.

Plasma metabolic changes after treatment initiation in persons with pulmonary TB: (A) Study design of the longitudinal TB disease cohort. Created in BioRender. Collins, J. (2026) https://BioRender.com/qwlkip8 (B) Balloon plot depicting metabolic differences in plasma concentrations between persons with Mtb infection (TBI; n = 92) those with pulmonary TB (PTB; n = 68), as well as change over time in persons with PTB during and after treatment. (C) Line plot with dots showing the mean and error bars showing the standard error of plasma concentrations of palmitoleic acid, linolenic acid, and docosahexaenoic acid after treatment initiation in persons with TB disease (n = 68; repeat measures ANOVA versus baseline; ∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001). (D) Plasma docosapentaenoic acid concentrations in persons with TB disease during and after treatment (n = 68) as well as controls without TB disease (n = 104). The box plots depict the minimum and maximum values (whiskers), the upper and lower quartiles, and the median. The length of the box represents the interquartile range. A repeat measures ANOVA was used to compare metabolite changes over time and a linear regression model controlling for age, sex, and body mass index was used for cross-sectional comparisons between groups (∗P < 0.05, ∗∗P < 0.01, ∗∗∗P < 0.001).

Discussion

We show that Mtb infection of both humans and NHPs is associated with dramatic changes in host fatty acid metabolism. Mtb infection led to depletion of acyl carnitines and both saturated and unsaturated LCFAs in NHPs, even in those infected with a relatively low Mtb inoculum. Such changes were strongly associated with decreased concentrations of multiple microbiome-derived metabolites that play a role in fatty acid metabolism including valerobetaine, butyrate, and secondary bile acids.³⁴ The decline in plasma saturated and unsaturated LCFAs was also evident in asymptomatic HHCs with TBI 6 months after high-level exposure to Mtb via a person with PTB. For participants with PTB, all measured polyunsaturated LCFAs were further depleted compared to those with TBI. After TB treatment initiation, plasma LCFAs of PTB participants increased to a similar concentration as those seen at baseline in TBI and TBU participants, though this increase did not occur until six months post-TB treatment. Together, these results demonstrate a characteristic metabolic response to infection with Mtb in humans and NHPs that is defined by a systemic increase in the utilisation of essential and non-essential fatty acids. This shows that Mtb infection alone reduces plasma LCFA concentrations and that larger magnitude decreases are associated with PTB.

The impact of TB disease on metabolic health has long been evident.^3,7,13,35, 36, 37 Weight loss is one of four cardinal symptoms used to screen for TB disease³⁸ and prior to the identification of the tubercle bacillus, TB was known as “consumption” based on the cachectic appearance many patients exhibit.² Yet most studies characterising the metabolic abnormalities associated with TB disease have been cross-sectional at the time of diagnosis, limiting our understanding of whether metabolic changes observed could be accounted for by anorexia and immune activation.¹³ Our results among experimentally infected NHPs and highly exposed HHCs provide evidence that Mtb infection itself remodels host metabolism even in the absence of TB symptoms or clinical illness. Declines in plasma concentrations of LCFAs were observed almost immediately after Mtb infection in NHPs. Further, most of the animals studied (12/17) received a relatively low inoculum of Mtb and many had only a small number of bacilli present in pulmonary granulomas on necropsy.^20,21 This indicates such responses take place even in the absence of a high bacterial burden. The same metabolic changes were also observed in humans with TBI six months after exposure to Mtb, a population without clinical symptoms that did not progress to TB disease. Thus, we show that Mtb infection alone is associated with systemic host metabolic changes in pathways related to fatty acid metabolism.

Decades of epidemiologic evidence linking TB disease to undernutrition,^9,10,39,40 paired with more recent experimental data,⁴¹ suggest appropriate regulation of host fatty acid metabolism may be important for controlling Mtb replication. One hypothesised aetiology of the association between TB outcomes and undernutrition has been protein energy malnutrition, which is known to have a detrimental impact on cell mediated immune responses.^13,42 Yet this hypothesis is not consistent with the observation that people with BMIs in the overweight or obese ranges have a significantly lower risk of TB disease relative to those with BMIs in the healthy or underweight range.^43,44 In our study, changes in lipid metabolism predominated in NHPs and humans after Mtb infection. Alterations in amino acid metabolism have primarily been observed in individuals at the time of TB disease diagnosis^35,45, 46, 47 or shortly before,⁴⁸ suggesting protein malnutrition may not explain the link between low BMI and TB progression. Our results instead implicate changes in fatty acid utilisation as the initial metabolic change associated with Mtb infection, which is exacerbated in those with TB disease. We observed that those with TBI not progressing to TB disease had their plasma LCFA concentrations recover without intervention. Persons with PTB had more significant declines at the time of diagnosis, especially for polyunsaturated LCFAs. Interestingly, the eight HHCs who went on to progress to TB disease during follow up did not have lower plasma LCFA concentrations at baseline versus non-progressors with TBI. These data suggest that low plasma LCFA concentrations did not precede disease progression. The lower concentrations at PTB diagnosis may therefore indicate higher LCFA utilisation is a necessary response to Mtb that increases with increased bacterial burden.

Experimental evidence from independent studies suggests increased LCFA utilisation may be necessary to limit immunopathology. Mice infected with Mtb that were given a supplement of the unsaturated LCFA palmitoleic acid had a lower burden of Mtb in the lungs and lower levels of pulmonary inflammation versus those receiving a supplement of the short chain fatty acid butyric acid.⁴¹ Taken together, these data indicate that supplementing LCFAs could have a therapeutic effect by replacing those LCFAs utilised as part of the host response to Mtb. This may explain why people with a higher BMI (and therefore higher fat stores) have a lower risk of TB disease progression compared to individuals with lower BMIs.

One possible mechanism linking systemic changes in fatty acid metabolism to control of Mtb infection is the innate immune response. It is well established that the metabolic state of macrophages impacts their ability to control Mtb replication both in vitro⁴⁹ and in vivo.50, 51, 52 Multiple studies demonstrated that a phenotype of increased glycolysis and reduced oxidative phosphorylation (also termed aerobic glycolysis) results in control of Mtb replication.^49,50,53 However, in vitro macrophage models generally capture early events after Mtb infection, while experimental murine models often result in rapid progression of TB disease over weeks. Their generalisability to humans, who control Mtb replication over months to years, is uncertain. Our observation that Mtb infection in humans and NHPs is marked by substantial increases in fatty acid utilisation (resulting in LCFA depletion) suggests there may be greater complexity to this traditional immunometabolic paradigm. Our findings are in line with other evidence from multicohort human studies, which have shown monocyte transcriptional and epigenetic changes related to lipid metabolism distinguish people who are highly exposed to Mtb but show no evidence of infection (i.e. resistors).^54,55 While it is unknown whether the metabolic changes observed in our study are directly linked to monocyte immune function, they highlight the need to better understand the intersection between metabolism and innate immunity in the context of Mtb control and progression.

It is important to note that the present study captures early metabolic changes after Mtb infection and includes participants with TB disease that were enrolled early in their clinical course. Thus, depletion of LCFAs represents an early response Mtb infection that is initially exacerbated in persons with PTB. Our prior work has shown if TB disease goes untreated and increases in severity, fatty acid oxidation eventually shuts down.⁴⁵ After months of inadequate TB treatment, LCFAs and TCA cycle intermediates accumulate, consistent with severely impaired fatty acid oxidation. This, in turn, is strongly associated with generation of proinflammatory eicosanoids and cytokines that exacerbate lung destruction.⁴⁵ In experimental models, this metabolic phenotype has been shown to cause increased mitochondrial dysfunction, generation of reactive oxygen species, and tissue damage.⁵⁶ Taken together, we believe this supports a model whereby LCFA utilisation protects against immunopathology and that impaired fatty acid breakdown is associated with greater disease severity.

The strength of this study is that we demonstrate a common host metabolic response to Mtb in NHPs and in human cohorts with TBI and PTB. However, the study is also subject to limitations. Due to the observational nature of the study design, we are unable to conclude whether this response is protective to the host or whether it enhances TB pathogenesis. The included NHP cohorts were exclusively made up of rhesus macaques, so observations in these animals may not be generalisable to all species of NHPs. In human participants with TB disease, we only measured LCFAs prior to TB progression in eight participants, so the time course of depletion in relation to disease progression remains uncertain. In future studies, it will be important to evaluate the impact of interventions aimed at mitigating depletion of LCFAs through dietary supplementation in experimental conditions and human populations. While no participants in the longitudinal TBI cohort were known to progress to TB disease, it is impossible to completely rule out the possibility that subclinical disease occurred at some point during the follow up period.⁵⁷ Additionally, Mtb infection was assessed by QFT, which is unable to distinguish recent from remote infection.

In summary, we show that both Mtb infection and TB disease result in a systemic metabolic response in humans and NHPs characterised by depletion of LCFAs. In humans not progressing to TB disease, depletion of LCFAs is transient, while those who develop TB disease experience further depletion that is not reversed until six months post-TB treatment completion. These results have the potential to create a new paradigm for the link between TB and host metabolism and suggest nutritional interventions aimed at supplementing mono- and/or polyunsaturated long chain fatty acids could have potential to improve TB outcomes.

Contributors

Conceptualisation: JMC, LW, JDE, DLB, HMB. Data curation: JMC, LW, NRG, MN, AN, KMS, DPJ. Formal analysis: JMC, MSYL, MN. Funding acquisition: JMC, LW, NRG, CLD, JR, JDE, HMB. Investigation: JMC, LW, NRG, RRK, MN, AN, KMS, MJM, DPJ. Methodology: JMC, KHL, MN, KMS, EAO, DPJ. Project administration: JMC, LW, KB, NRG, CLD, JR, SS, KDK, JDE, DLB, HMB. Software: JMC, MN, KMS, EAO, DPJ. Resources: JMC, EAO, DPJ. Supervision: JMC, LW, KB, NRG, EAO, DPH, JDE, DLB, HMB. Validation: JMC, KHL, MSYL, MN, KMS. Visualisation: JMC, KHL, MN, MJM, DPJ, HMB. Writing—original draft: JMC. Writing—review & editing: all authors. JMC, MN, KMS, and DPJ accessed and verified the underlying data. All authors reviewed and approved the final manuscript. The TBRU-ASTRa consortium led the observational human cohort studies, including participant recruitment, study visits, sample collection, and outcome adjudication.

Data sharing statement

De-identified data may be accessed in the Supplemental Data Files 1–9.

Declaration of interests

The TBRU-ASTRa consortium was funded by the award U19 AI111211 from the U.S. National Institute of Allergy and Infectious Diseases. The authors JMC, NRG, RRK, MSYL, CLD, JR, AN, DPJ, MJM, EAO, JDE, and HMB receive support from additional grants from the U.S. National Institutes of Health (NIH) outside of those that supported this work. Grants from the U.S. NIH also supported research-related travel JMC, NRG, RRK, CLD, JR, DPJ, MJM, EAO, and HMB. HMB is supported by a contract from the U.S. State of Georgia Department of Public Health. HMB is a voting member of the NIH Workforce Development Enterprise Committee. CLD is an Executive Committee voting member of the AIDS Clinical Trial Group.