Latent tuberculosis infection is associated with an enrichment of short chain fatty acid producing bacteria in the stool of women living with HIV

Abstract

Background:

Latent tuberculosis infection (LTBI) is common in people living with HIV (PLHIV) in high TB burden settings. Active TB is associated with specific stool taxa; however, little is known about the stool microbiota and LTBI, including in PLHIV.

Method:

Within a parent study that recruited adult females with HIV from Cape Town, South Africa into predefined age categories (18–25, 35–60 years), we characterised the stool microbiota of those with [interferon-γ release assay (IGRA)- and tuberculin skin test (TST)-positive] or without (IGRA- and TST- negative) LTBI (n=25 per group). 16S rRNA DNA sequences were analysed using QIIME2, Dirichlet Multinomial Mixtures, DESeq2 and PICRUSt2.

Results:

No α- or β-diversity differences occurred by LTBI status; however, LTBI-positives were Faecalibacterium-, Blautia-, Gemmiger-, Bacteroides-enriched and Moryella-, Atopobium-, Corynebacterium-, Streptococcus-depleted. Inferred metagenome data showed LTBI-negative-enriched pathways included several involved in methylglyoxal degradation, L-arginine, putrescine, 4-aminobutanoate degradation and L-arginine and ornithine degradation. Stool from LTBI-positives demonstrated differential taxa abundance based on a quantitative response to antigen stimulation (Acidaminococcus-enrichment and Megamonas-, Alistipes-, and Paraprevotella-depletion associated with higher IGRA or TST responses, respectively). In LTBI-positives, older people had different β-diversities than younger people whereas, in LTBI-negatives, no differences occurred across age groups.

Conclusion:

Amongst female PLHIV, those with LTBI had, vs. those without LTBI, Faecalibacterium, Blautia, Gemmiger, Bacteriodes-enriched, which are producers of short chain fatty acids. Taxonomic differences amongst people with LTBI occurred according to quantitative response to antigen stimulation and age. These data enhance our understanding of the microbiome’s potential role in LTBI.

Introduction

Tuberculosis (TB) is major cause of death, with 167 000 deaths among persons living with HIV (PLHIV) in 2022¹. One strategy to prevent active TB is control latent tuberculosis infection (LTBI). LTBI is inferred from a positive tuberculin skin test (TST) or interferon-gamma release assay (IGRA). TB preventive treatment (TPT) strategies play a key role in TB prevention especially in vulnerable populations such as PLHIV.

PLHIV are at greater risk for Mycobacterium tuberculosis (MTB) infection and progression to active TB². However, it is poorly understood why some individuals in TB-endemic countries are never infected despite high exposure and why a large proportion of infected individuals never progress. We need more information on correlates of infection and progression, which may have prognostic value.

The microbiome has important immunomodulating effects and the microbiome’s role, including people with LTBI, is an emerging area of interest. For example, in the lung, Lactobacillus is enriched in people with LTBI compared to active pulmonary TB group and LTBI-negatives³. In the nasopharynx of LTBI-positives, Staphylococcus and Corynebacterium dominate the microbiome compared to healthy control and active TB cases⁴, and the nasopharyngeal microbiota of LTBI-positives has lower alpha-diversity than LTBI-negatives⁵.

The stool microbiota has potentially an important immunomodulatory role in respiratory disease, including active TB⁶. However, it is comparatively understudied in latent TB. One study among individuals with poorly-controlled diabetes, showed LTBI-positives people to be Bacteroides-, Alistipes-, and Blautia-enriched compared to LTBI-negatives⁷.

During LTBI infection, comparisons of TB cases, HIV-negative LTBI-positive individuals, LTBI-negative and active TB gut microbiomes showed trends of changes in Bacteriodes and Firmicutes, however no significant difference in composition of the stool microbiota⁸. In HIV-negative LTBI-positive individuals showed positive correlation between relative abundances of Coriobacteriaceae and IFN-gamma against MTB antigens more likely associated with of CD4 + T cell⁹. Those studies, however, did not include PLHIV where changes in the gut microbiome are a well-characterised complication of HIV¹⁰, commonly characterized by reductions in diversity in the stool microbiota¹¹. If microbial dysbiosis is detected early in natural history of TB disease (after infection), it may be indicative of early microbial and immune dysregulation associated with subsequent risk of active TB. Tests of progression to active TB are a major public health priority, as is understanding the biological drivers of LTBI. To continue to strengthen our understanding of LTBI and the microbiome, we evaluated the stool microbiota of PLHIV with and without LTBI, nested within a larger study studying resistance to infection.

Methods

Recruitment

Participants (18–60 years) were recruited from community health care clinics in Cape Town, South Africa as part of a published parent study (ResisTB)¹². This cohort was predominantly female, and age was a surrogate for TB exposure, resulting in two groups of 18–25 years and 35–60 years. All people had to be TB symptom screen negative, HIV-positive and stable on ART for ≥ 1 year. Study procedures were approved by the Stellenbosch University Human Research Ethics Committee (N16/03/033A) and each participant provided written informed consent.

Definitions

IGRA and TST positive (LTBI-positive) people was defined by two QuantiFERON-TB Gold Plus-positive and a positive TST (> 0 mm). IGRA and TST negative (LTBI-negative) was defined by two negative QuantiFERON-TB Gold Plus-negative and a TST (0 mm).

Microbiota specimen collection and processing

At TST administration, participants were provided with a home stool sampling kit containing an EasySampler (ALPCO, Salem, USA) and a receptacle containing DNA stabilization buffer (Stratec Biomedical, Birkenfeld, Germany). Generally, buffered stools were collected the night before TST reading and returned at TST reading. Upon receipt at the laboratory, buffered stool was frozen at −20°C until batched DNA extraction carried out using the PSP Spin Stool DNA Plus Kit (Stratec Biomedical, Birkenfeld, Germany).

16S rRNA gene sequencing and microbiota analysis

V4 region sequencing of the bacterial 16S rRNA gene (150 bp read length, paired-end) was done using Illumina Miseq as described⁶. Sequences was analysed with Quantitative Insights into Microbial Ecology (QIIME2, version 2.0.8). Cluster analysis was carried out using Dirichlet-Multinomial Mixtures (DMM)¹³. Alpha diversity was calculated by Shannon’s diversity with Mann-Whitney testing using GraphPad Prism (v8; GraphPad Software, USA). Beta diversity was calculated using Bray-Curtis with permutational multivariate ANOVA (PERMANOVA) using R (v4.2.2; R Core Team). The functional metagenome was inferred from sequencing data using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt2; v2.0.0)¹⁴. Differentially abundant taxa and metabolic pathways were identified using DESeq2 (v1.22.2)¹⁵ and Benjamini–Hochberg correction adjustment for multiple comparisons (significance level 0.20)^6,16. For comparisons groups, taxa at higher relative abundance in one group were described as enriched (those at lower relative abundance were described as depleted). For the whole cohort, age and field site correction was applied for DESeq2 analyses. Linear discriminant analysis (LDA) effect size (LEfSe)¹⁷ was used to compare the clusters to each other. The proportions test was done using STATA (v18; StataCorp, USA) to determine whether a specific variable was more frequent in different groups. Data and code are available upon publication.

Results

Population

We collected stool from female PLHIV stable on ART and whom did not have previous TB, with or without LTBI (n = 25 per group; Fig. 1). Demographic data are in Table 1. People with LTBI were younger and more likely to be from Khayelitsha (Site B) Community Health Clinic (Table 1).

Figure 1. Study profile.

We collected and analysed stool from 25 LTBI-positive and 25 LTBI-negative people enrolled in a parent study (ResisTB). HIV: Human immunodeficiency virus; ART: Antiretroviral therapy; IGRA: Interferon gamma release assay; TST: Tuberculin skin test; LTBI: latent TB infection; TB: Tuberculosis.

Table 1. Demographic and clinical characteristics.

People with LTBI were younger and more likely to from Khayelitsha (Site B) CHC. Abbreviations: LTBI: Latent TB infection; BMI: Body mass index; INH: Isoniazid; CHC: Community Health Centre; CDC: Community Day Centre. Data are median (IQR) or n (%).

Characteristic*	Overall (n = 50)	LTBI-positive (n = 25)	LTBI-negative (n = 25)	p-value
Age, years,	24 (23–39)	24 (22–25)	38 (24–41)	0.005
18–25	33/50 (66)	21/25 (84)	12/25 (48)
35–60	17/50 (34)	4/25 (16)	13/25 (52)	0.007
CD4 (cells/mm3)	468 (368–683)	471 (346–675)	434 (375–705)	0.881
BMI (kg/m)	29 (25–33)	30 (28–34)	26 (24–32)	0.137
Current Tobacco smoker	1/49 (2)	0/24 (0)	1/25 (4)	0.312
Alcohol	31/50 (62)	17/25 (68)	14/25 (56)	0.382
INH prophylaxis	12/50 (24)	6/25 (24)	6/25 (24)	>0.999
Co-trimoxazole prophylaxis	7/50 (14)	2/25 (8)	5/25 (20)	0.221
Field site
Khayelitsha (Site B) CHC	10/50 (20)	8/25 (32)	2/25 (8)	0.034
Khayelitsha (Site B) Youth	7/50 (14)	4/25 (16)	3/25 (12)	0.684
Kraaifontein CHC	11/50 (22)	1/25 (4)	10/25 (40)	0.002
Site C Youth	19/50 (38)	12/25 (48)	7/25 (28)	0.145
Du Noon CDC	3/50 (12)	0/25 (0)	3/25 (12)	0.074

^*Missing data: Current Tobacco smoker (n = 1)

Stool from people with LTBI is Moryella-, Atopobium-, Corynebacterium-, Streptococcus-depleted and Faecalibacterium, Blautia, Gemmiger and Bacteriodes-enriched

Overall, no differences were seen by LTBI status for α- (p = 0.168, Fig. 1A) and β-diversity (PERMANOVA, p = 0.841, Fig. 1B). However, Moryella-, Atopobium-, Corynebacterium-, Streptococcus were depleted and Faecalibacterium, Blautia, Gemmiger and Bacteriodes-enriched in LTBI-positives compared to LTBI-negatives (Fig. 1C). People on TB preventative treatment (INH prophylaxis) were, compared to those not on treatment, Blautia-enriched and Moraxella-, Megamonas- and Actinobacillus-depleted (Supplementary Fig. 1).

Comparisons according to age groups (18–25 vs. 35–60 years)

When we compared age groups within people of the same LTBI status, within LTBI-positive individuals, no α-diversity differences occurred (p = 0.789, Fig. 2A), however, β-diversity differed (p = 0.003, Fig. 2B). Older LTBI-positives were Ochorobactrum-, Neisseria- and Mycoplasma-enriched and Catenibacterium-, Alistipes-, and Methanobrevibacter-depleted (Fig. 2C). LTBI-negative individuals did not differ in α-diversity (p = 0.205, Fig. 2E), β-diversity (p = 0.442, Fig. 2F) older LTBI-negatives were also Methanobrevibacter-depleted and Actinobacillus-enriched. (Fig. 3G). We calculated Bray-Curtis distances for older vs younger LTBI-positives and LTBI-negatives. Both age groups demonstrated similar distances (p = 0.952) across LTBI-statuses (Supplementary Fig. 2).

Figure 2. Stool from LTBI-positives is Moryella-, Atopobium-, Corynebacterium-Streptococcus-depleted and Faecalibacterium, Blautia, Gemmiger and Bacteriodes-enriched.

(A) Comparison of Shannon’s diversity index of LTBI-positive and LTBI-negative groups. (B)Principal coordinate analysis of Bray-Curtis distances between groups. (C) Volcano plot depicting differentially abundant taxa. More discriminatory taxa (bolded) appear closer to the left or right, and higher above the threshold (red dotted line, FDR=0.20). Relative abundance of taxa is indicated by circle size. LTBI: Latent TB infection.

Figure 3. Distinct stool microbiotas by age in LTBI-positive.

Shannon’s diversity index, Principal coordinate analysis of Bray-Curtis distances, and volcano plot depicting differentially abundant taxa enriched in 35–60 and 18–25 age groups are for LTBI-positive only (A, B and C) and for LTBI-negative only (E, F and G). More discriminatory taxa (bolded) appear closer to the left or right, and higher above the threshold (red dotted line, FDR=0.20). LTBI: Latent TB infection.

Distinct metabolic pathway associations with LTBI-positives and -negatives

LTBI-negatives were enriched in methylglyoxal, L-arginine, putrescine, 4-aminobutanoate, and L-ornithine degradation pathways (Fig. 3). LTBI-positives had no differential enrichment.

Microbial cluster identification and their characteristics

Taxonomic analyses

Three clusters (C1, C2 C3) were identified (Supplementary Fig. 3) with C1 vs C2 (p = 0.361) and C2 vs C3 (p = 0.299) having similar α-diversity whilst C1 vs C3 differed (p = 0.040) (Fig. 5A). β-diversity differed between clusters (PERMANOVA p = 0.001) (Fig. 5B). C1, C2, and C3 were characterised by high Bacteriodes, Streptococcus, and Prevotella abundances, respectively (Fig. 5C). When cluster pairs were compared (Fig. 5D–F), C1 was, compared to C2, Methanobrevibacter-, Catenibacterium- and Megamonas-enriched and Actinobacillus-, Moraxella- and Ochrobactrum-depleted and, compared to C3, Megamonas-, Akkermansia- and Anaerostipes-enriched and Succinivibrio-, Prevotella-, Dialister-depleted. C3 was, compared to C2, Roseburia-, Anaerovibrio-, and CF231-enriched and Actinobacillus-, Moraxella- and Ochrobactrum-depleted. When all three clusters were compared together (Supplementary Fig. 4), C1 was, relative to the others, the most enriched in Bacteroides, Oscillospirai and Parabacteroides., C2 most enriched in Streptococcus, Veillonella, and Actinomyces, and C3 most enriched in Prevotella and Catenibacterium.

Figure 5. Three distinct microbial clusters, with a high abundance of Bacteroides, Streptococcus and Prevotella, were identified.

Comparison of (A) alpha- and (B) diversity- by cluster. (C) Heatmap shows the composition of each cluster. (D, E, F)Volcano plots depicting differentially abundant taxa compared across cluster pairs. More discriminatory taxa (bolded) appear closer to the left or right, and higher above the threshold (red dotted line, FDR=0.20). LTBI: Latent TB infection.

Clinical and demographic characteristics across clusters

β-diversity differed by field site (Supplementary table 1). There were no differences in the proportion of LTBI-positives per cluster (Table 2). C1 and C3 were more likely to be on INH prophylaxis than C2 and more likely to be from sites other than Khayelitsha (Site B) Youth. C2s were, compared to C3s, more likely to be from sites other than Khayelitsha (Site B) CHC Youth and Du Noon CDC.

Table 2. Demographic and clinical characteristics of the three clusters found in cohort.

C1 was more likely than C2 to be on current INH prophylaxis and more likely to be recruited from Khayelitsha (Site B) CHC, Kraaifontein CHC, Site C Youth or Du Noon CDC. C2 was more likely than C3 to be recruited from Khayelitsha (Site B) CHC, Kraaifontein CHC or Site C Youth. Abbreviations: LTBI: Latent TB infection; BMI: Body mass index; INH: Isoniazid; CHC: Community Health Centre; CDC: Community Day Centre. Data are median (IQR) or n (%).

Characteristics*	C1 (n = 23)	C2 (n = 16)	C3 (n = 11)	p-value C1 vs C2	p-value C2 vs C3	p-value C1 vs C3
Age, years	24 (22–25)	37 (24–40)	25 (24–33)	0.114	0.723	0.146
LTBI-positive	11/23 (48)	9/16 (56)	5/11 (46)	0.601	0.581	0.900
CD4 (cells/mm3)	575 (375–700)	395 (372–516)	487 (317–642)	0.242	0.914	0.445
BMI (kg/m)	29 (25–32)	30 (28–35)	27 (23–34)	0.248	0.277	0.612
Current Tobacco smoker	0/23 (0)	0/15(0)	1/11 (9)	-	0.234	0.142
Alcohol	15/23 (65)	9/16 (56)	7/11 (64)	0.574	0.704	0.927
INH prophylaxis	8/23 (35)	1/16 (6)	3/11 (27)	0.037	0.130	0.662
Co-trimoxazole prophylaxis	3/23 (13)	4/16 (25)	0/11 (0)	0.337	0.072	0.211
Field site
Khayelitsha (Site B) CHC	0/23 (0)	10/16 (63)	0/11 (0)	< 0.001	< 0.001	-
Khayelitsha (Site B) Youth	1/23 (4)	3/16 (19)	3/11 (27)	0.1440	0.597	0.051
Kraaifontein CHC	7/23 (30)	0/16 (0)	4/11 (36)	0.014	0.009	0.726
Site C Youth	15/23 (65)	0/16 (0)	4/11 (36)	< 0.001	0.009	0.114
Du Noon CDC	0/23 (0)	3/16 (19)	0/11 (0)	0.030	0.127	-

^*Missing data: Current Tobacco smoker (n = 1)

Within LTBI-positives, taxa are differentially enriched based on the magnitude of the response to antigen stimulation

Taxonomic abundances between quantitative responses to antigen stimulation were compared overall and in LTBI-positives. No differential abundances were identified between the overall cohort above or vs the median average IGRA or TST quantitative response (Supplementary Fig. 5A and B). LTBI-positives with an IGRA response above the median were Acidaminococcus-enriched and Granulicatella-depleted (Fig. 6A) and LTBI-positives below the median TST response were Megamonas-, Alistipes-, and Paraprevotella-enriched (Fig. 6B).

Figure 6. LTBI-positives had, based on the magnitude of the quantitative responses, differentially enriched taxa (those with IGRA responses above the median were Acidaminococcus-enriched and Granulicatella-depleted, those with TST responses below the median were Megamonas-, Alistipes-, Paraprevotella-enriched).

(A) Volcano plot depicting differentially abundant taxa based on the median IGRA quantitative response to antigen (median value 5.56). More discriminatory taxa appear closer to the left or right, and higher above the threshold (red dotted line, FDR=0.20). (B)Volcano plot depicting differentially abundant taxa based on the median TST response (Median value 18). LTBI: Latent TB infection.

Discussion

We compared, in PLHIV, the stool microbiota of LTBI-positive vs. LTBI-negative women. Our key findings are: 1) stool from LTBI-positives differed from LTBI-negatives in terms of known short chain fatty acid (SCFA)-producing taxa and these were associated with a depletion of metabolite degradation pathways, 2) three taxonomic clusters occurred, characterised by high abundances of Bacteroides, Streptococcus and Prevotella, and clustering associated with INH prophylaxis and health facility location, 3) LTBI-positives with greater quantitative response to mycobacterial antigen stimulation were, vs. LTBI-positives with lesser responses, Acidaminococcus-enriched (IGRA readouts) and Megamonas-, Alistipes-, and Paraprevotella-depleted (TST), and 4) β-diversity differed by age group only in LTBI-positives but not - negatives. Our findings help lay a foundation for understanding the microbiome’s role in LTBI.

Stool from people with LTBI was Moryella-, Atopobium-, Corynebacterium-, Streptococcus-depleted and Faecalibacterium, Blautia, Gemmiger and Bacteriodes-enriched. Faecalibacterium and Gemmiger is a known producer of the butyrate¹⁸, which is a SCFA that increases incident TB risk¹⁹. Bacteroides produces SCFAs like acetate and propionate²⁰. Blautia is enriched in people with active TB and independently predicts upregulation of pro-inflammatory pathways⁶. Although the role of Atopobium is unclear, Streptococcus, which we found to be depleted in LTBI-positives, produces acetate, which mitigates host inflammation²¹.

Three taxonomic clusters occurred [Bacteroides (C1), Streptococcus (C2), Prevotella-enriched (C3)], however, these were not associated with LTBI status (other studies have documented specific clusters associated with active TB¹⁶). People within each cluster were more likely to be recruited from different facilities, suggesting potential geographic associations to be considered in future studies. Additionally, C1 was more likely than C2 to be receiving INH prophylaxis. INH prophylaxis itself was associated with Blautia-enrichment and Moraxella-, Megamonas- and Actinobacillus-depletion. Other studies have shown Clostridiales, Coprococcus, Lachnospiraceae, and Ruminococcaceae-enriched and Clostridium_XIVa, Romboutsia, and Roseburia-depleted stool to occur during rifamycin-based tuberculosis preventive therapy²². To our knowledge, our study is the first to show Blautia-enriched and Moraxella-, Megamonas- and Actinobacillus-depleted stool in humans on isoniazid TB preventive therapy. This association is interesting as isoniazid itself is a drug thought to have an extremely narrow antimicrobial spectrum (Mycobacteria only)²³.

Additionally, LTBI-positives who had a larger quantitative response to antigen stimulation were, when IGRA readouts were used, Acidaminococcus-enriched and Granulicatella-depleted, and, when TST readouts was used, Megamonas-, Alistipes-, Paraprevotella-depleted. Acidaminococcus produces acetate and butyrate²⁴ and is primarily influenced by diet²⁵; its enrichment likely reflects lifestyle differences within LTBI-positives. Paraprevotella (like Alistipes) is a SCFA-producer and generally considered beneficial²⁶. This finding is notable because higher quantitative responses are associated with greater risk of incident TB²⁷, suggesting such taxa may contribute to this risk, however, this requires prospective confirmation in controlled studies.

Younger people were more likely to be LTBI-positive than older people so, to adjust for age as a potential confounder, we dichotomised people (18–25, 35–60 years). Within LTBI-positives, older vs. younger individuals were enriched in Ochrobactrum (an opportunistic pathogen²⁸), Neisseria (role unclear), Mycoplasma (induces a proinflammatory cytokines²⁹) and depleted in Catenibacterium (enriched in PLHIV³⁰ and active TB³¹), Alistipes (SCFA producer with potentially anti-inflammatory effects³²), and Methanobrevibacter (a methane producer³³). Older LTBI-negatives were also Methanobrevibacter-depleted and enriched in Actinobacillus (inversely associated with amino acid production³⁴) but did not show β-diversities differences. This could suggest LTBI results in greater age-related microbiome differences but requires further investigation.

Our study has strengths and limitations. This is a cross-sectional study that, to enhance feasibility, leveraged (but was constrained) the parent ResisTB study. We only evaluated women with HIV and other populations may result in different findings, however, PLHIV do have elevated risk of incident TB. Although people were measured once, our study generates useful data to inform hypothesis-driven interventions to potentially modulate the microbiome.

In conclusion, amongst women living with HIV, those with LTBI were, vs. those without LTBI, primarily differentially abundant in SCFA-producing anaerobic bacteria. Taxonomic differences also occurred amongst people with LTBI by age group, suggesting the age-related microbiome perturbations are more pronounced in LTBI-positives. Longitudinal studies are needed to further delineate the microbiome’s role in LTBI, which this work helps provide a justification for.

Figure 4. Distinct microbial metabolic pathways are associated with LTBI.

LTBI-positives had a depletion of degradation-associated pathways. LTBI: Latent TB infection.

Funding Statement

This work and authors were supported by the Deutscher Akademischer Austauschdienst (ST32-PKZ-91770486), the European & Developing Countries Clinical Trials Partnership (EDCTP; project numbers SF1041, TMA2017CDF-1914-MOSAIC and TMA2019CDF-2738-ESKAPE-TB), National Research Foundation (NRF), the South African Medical Research Council (SAMRC), the Harry Crossley Foundation, Stellenbosch University Faculty of Health Sciences, and the National Institute of Allergy and Infectious Diseases of the National Institutes of Health under award numbers: R01AI136894; U01AI152087; U54EB027049; D43TW010350; K43TW012302. The content is the solely the responsibility of the authors and does not necessarily represent the official views of the funders.

Availability of data and materials

Data is available on reasonable request. Study protocol and datasets generated in this study maybe requested from the corresponding author.