Mycobacterium tuberculosis infection disrupts gut and respiratory microbial communities and networks with incomplete restoration after two months of treatment Authors: Sohyoung Won, Juyong Hong, Heebal Kim, Junho Lee, Young Chun Ko, Bumhee Yang, Sun-Hyung Kim, Seok Shin Koh, Jeeyeon Park, Gee Oh Guak, Gyeong In Lee, Youngeun Choi, Seung Heon Lee, Gisu Kang, Seoae Cho, Hyejin Kim Published: 2026 ABSTRACT Pulmonary tuberculosis (TB), caused by Mycobacterium tuberculosis (MTB), remains a major global health concern. Respiratory infections can influence the gut microbiome, and vice versa, through communication between the gut and respiratory microbiome—the gut–lung axis. Yet how TB infection and treatment relate to this connection remains unclear. To investigate the associations between TB and gut–respiratory microbiome interactions, we analyzed paired gut and respiratory samples from patients with TB at baseline (TB-0) and two months after treatment (TB-trt), along with matched healthy controls (HC). Microbial communities were profiled at the species level, and integrated gut–respiratory microbial networks were constructed and compared. MTB infection altered both microbial composition and network structure. Patients with TB showed lower levels of short-chain fatty acid (SCFA)–producing species and higher levels of inflammation-associated taxa. Consistent trends emerged in the gut–respiratory networks: SCFA-producing species were central in the networks of HC and persisted in TB-trt, whereas inflammation-associated species occupied influential positions in TB-0 and remained, though less prominently, after treatment. The size and complexity of the networks also differed across groups. TB infection was accompanied by broad disruptions of the gut–respiratory microbiome, and treatment resulted in only partial restoration. Species associated with these shifts, including Bifidobacterium spp. and Sutterella wadsworthensis, have the potential to serve as TB biomarkers or as adjuncts to improve treatment outcomes, with further experimental validation required.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13099-026-00834-4. Introduction Pulmonary tuberculosis (TB), caused by Mycobacterium tuberculosis (MTB), remains a serious global health concern, and the World Health Organization estimated approximately 1.23 million deaths in 2024 [1]. TB primarily affects the lungs; however, MTB can spread to other organs, resulting in complications that extend beyond the respiratory system [2]. The human microbiome plays a crucial role in regulating immunity, metabolism, and susceptibility to infection [3–5]. Previous studies have also revealed connections between the gut microbiome and respiratory health, suggesting that disturbances in the gut may influence the lungs, and vice versa, in both healthy states and diseases such as TB [6]. The emerging concept of the gut–lung axis suggests their bidirectional interactions, in which respiratory and intestinal microbes can influence one another through immune and metabolic pathways [7, 8]. Recent studies have revealed multiple ways in which TB is associated with alterations of the gut microbiome [9]. Several bacterial taxa differ in abundance between patients with TB and healthy controls [4, 10–13]. For instance, members of Actinomycetota, Bacillota, and Pseudomonadota are often enriched in patients with TB, whereas Bacteroidota, Blautia, and Bifidobacterium are consistently depleted [14–16]. In addition to compositional shifts, many studies have reported a reduction in alpha diversity among patients with TB, along with a decrease in short-chain fatty acid (SCFA) production [15–17]. TB treatment can alter the gut microbiome imbalance caused by MTB infection [18]. Longitudinal studies comparing the gut microbiome before and after treatment report a decline in microbial diversity, accompanied by an enrichment of Escherichia and Shigella with a reduction in Akkermansia [19] and highlighted the value of following these temporal changes to better guide microbiome-targeted interventions during TB treatment [20]. Although many studies have explored the relationship between TB and either the gut or the respiratory microbiome [21], much less is known about the interaction between the two communities. To address this key knowledge gap, we compared paired gut and respiratory samples from the same participants and constructed integrated microbial networks to examine how TB shapes the gut–respiratory microbiome. By combining case–control comparisons with longitudinal sampling during treatment, we investigated the effects of both infection and treatment on microbial communities. Methods Recruitment This study, nested within a prospective longitudinal study, was conducted from November 2022 to December 2023 in South Korea, including Kwangju Christian Hospital, Chungbuk National University Hospital, three Double-barred Cross TB clinics, and the Korean Institute of Tuberculosis (KIT). Eligible participants were adults (≥ 19 years). Participants were excluded if they were pregnant, had used antibiotics within the previous month, or had comorbid conditions affecting the lung or gut microbiota. All patients with TB were recruited prior to TB treatment initiation, and no TB treatment had been administered at the time of enrolment. TB diagnosis was defined as a positive result on at least one of the following: smear microscopy, culture, Xpert MTB/RIF assay. Radiological findings were used as supportive evidence only. Healthy controls were recruited from the community and included only if they had no history of TB, no TB-related symptoms, and tested negative for latent TB infection using QuantiFERON-TB Gold Plus. Specimen collection Respiratory and stool samples were collected from all participants. For patients with TB, samples were collected prior to treatment initiation and at a follow-up visit two months later. Healthy controls provided samples at enrolment only. Respiratory samples from patients with TB were collected as expectorated sputum. In one patient who was unable to produce sputum, bronchoalveolar lavage (BAL) fluid was collected due to the inability to expectorate sputum; however, this sample was excluded from downstream analyses due to the absence of follow-up data. Accordingly, no BAL sample was included in the final analysis. HCs provided deep-cough respiratory samples and stool samples at enrollment (Figure S1). For HCs who had difficulty producing sputum, samples were obtained by collecting a mixture of saliva from the deep pharynx and respiratory secretions following a deep cough in the morning. Details on specimen quality assessment and BAL collection are provided in the Supplementary Materials. Stool samples were collected using a T-swab Clinical Virus Transport Medium Advance Kit (Noble Biosciences, Korea). Respiratory samples (sputum and BAL) and stool samples were transported at 4 °C to the central laboratory at KIT, where DNA was extracted. Extracted DNA from all samples (sputum, BAL, and stool) was stored at − 80 °C until further analysis. DNA extraction Total DNA was extracted from stool samples using the E-Z 96 DNA Kit (Omega Bio-tek, USA) and from respiratory samples using the High Pure PCR Template Preparation Kit (Roche, Switzerland), according to the manufacturers’ instructions. DNA quality control was performed as described in the Supplementary Materials. Amplicon sequencing and preprocessing PCR was performed with an input of over 1 ng of DNA using BX-Taq DNA polymerase (NICSROgene™, Seoul, Korea) and a barcode primer set specifically designed for the study, as previously described [22]. The resulting 4.2 kb amplicons are verified through gel electrophoresis and HM SPRI beads. Amplicons were quantified using the PicoGreen dsDNA assay and pooled to achieve a final concentration of 1,000 ng. Library preparation end–repair, adapter ligation, and cleanup were performed following the Oxford Nanopore Technologies protocol (Ligation sequencing amplicons V14, SQK-LSK114), and sequencing was conducted on a SpotON flow cell. Base-calling was performed using Guppy (version 6.5.7) [23], with demultiplexing and adapter trimming conducted via Porechop (version 0.2.3_seqan2.1.1) [24]. Quality assessment and statistical analysis of the sequencing reads were executed using Nanoplot (version 1.39.0) [25]. Reads were filtered by length (3,500 to 5,000 bp) using NanoFilt [25], and Chimeric reads were identified and removed using minimap2 (version 2.26-r1175) [26] and yacrd (version 0.6.2) [27]. Microbiome profiling and data processing Microbiome classification and profiling were performed using the MIrROR pipeline (version 1.0.0) with default settings, as described previously [16]. Sample matching and filtering We used a matched case–control design to compare HCs and patients with TB, and a matched-pair design to compare the same patients at baseline and 2 months after treatment (TB-0 and TB-trt, respectively). For the case–control comparison, participants were matched by sex, age, body mass index (BMI), smoking, alcohol consumption, and diabetes status. Differences between HCs and patients with TB were assessed using chi-squared tests for categorical variables and the Mann–Whitney U test for continuous variables. To minimize potential contamination from oral and nasal microbes in the respiratory microbiome analysis, we excluded respiratory samples with a high abundance of oral and nasal commensals from the Human Oral Microbiome Database [28]. To preserve the study design, we retained only matched sets in which the respiratory samples from the HC and the corresponding patient with TB, both at baseline and after treatment, showed < 70% cumulative relative abundance of oral and nasal commensals. We analyzed the respiratory microbiome of 12 participants per group in the downstream analysis. Power estimation based on sample size We quantified the adequacy of the sample size using a bootstrap-based observed-effect power analysis, following the simulation-based PERMANOVA power framework described by Kelly et al. (2015) [29]. Statistical analysis of microbial communities Core gut and respiratory microbiome were identified as taxa with ≥ 90% prevalence within each group (HC, TB-0, and TB-trt). Alpha-diversity was measured as the Shannon index and beta diversity was assessed via Principal coordinates analysis (PCoA). Non-parametric test were used for testing (Mann-Whitney and PREMANOVA). Differential abundance analysis was conducted for two comparisons: between HCs and TB-0, and between TB-0 and TB-trt. Significant taxa had FDR < 0.05 from a normalized linear model and an absolute LDA score > 2 from linear discriminant analysis. We performed taxon set enrichment analysis and compared pathway abundance, and results with FDR < 0.05 were considered statistically significant in both. Gut and respiratory microbiome co-occurrence network analysis We constructed co-occurrence networks combining gut and respiratory taxa for each group following the methods described in Mac Aogain et al. [30]. We filtered the data to include only species with an abundance of at least 0.1% in 3 or more samples and retained significant edges (FDR < 10⁻⁸). Nodes ranking in the top 5% were labelled. Results Study population or cohort characteristics We enrolled n = 36 people with active TB pre-treatment and n = 36 healthy controls with matching demographic and clinical characteristics to people with TB (Table 1). There were no differences in sex, age, BMI, comorbidities, smoking, or alcohol consumption between patients with TB and HCs (p > 0.05). The effect of the recruitment center was also not significant in either the gut or respiratory microbiota (PERMANOVA, p > 0.05). For the respiratory microbiome, we analyzed the data of 12 participants per group with low levels of potential contamination. The estimated power in gut microbiome composition was 0.84, whereas the corresponding power for lung microbiome composition was 0.71, which are within the range of values often interpreted as adequate for exploratory detection of moderate community differences when sample sizes are inherently limited [31]. Table 1The demographic and clinical characteristics between healthy controls and patients with TBCharacteristicsHealthy controlsPatients with TBp valueN = 36 (%)N = 36 (%)Sex0.62 Male16 (44.4)22 (61.1) Female20 (55.6)14 (38.9)Age (years)0.35 Median (Interquartile range)63.0 (44.8–73.0)67.5 (51.8–74.0) 20–2925.0 (25.0–27.0)27.0 (24.0–27.0) 30–3931.0 (31.0–35.0)0 40–4946.5 (44.8–47.2)44.0 (42.0–47.0) 50–5954.0 (52.5–54.5)55.0 (51.8–58.2) 60–6963.0 (61.5–65.2)65.5 (62.5–68.8) 70–7973.0 (73.0-73.8)73.0 (70.0–74.0) ≥ 8085.0 (82.0–86.0)86.0 (83.5–87.0)BMI (kg/m2)0.25 Median (Interquartile range)22.7 (20.3–25.7)21.2 (19.8–23.1) Underweight (BMI < 18.5)17.7 (17.1–18.0)17.8 (17.8–18.0) Normal weight (BMI 18.5–24.9)21.6 (20.1–22.7)21.0 (20.0-21.9) Obesity (BMI ≥ 25.0)26.8 (25.9–28.4)28.0 (26.5–29.9)Comorbidity> 0.99 No19 (52.8)16 (44.4) Yes17 (47.2)20 (55.6)Type of comorbidity Hypertension10 (27.8)14 (38.8)0.87 Diabetes mellitus5 (13.9)10 (27.8)0.66 Dyslipidemia9 (25.0)5 (13.9)0.83 Othera8 (22.2)12 (33.3)> 0.99Smoking status0.57 Never smoked23 (63.9)14 (38.9) Smoke currently9 (25.0)11 (30.6) Formerly smoked4 (11.1)11 (30.5)Drinking0.48 No16 (44.4)17 (47.2) Yes20 (55.6)19 (52.8)Previous history of TB0.20 No35 (97.2)0 New030 (83.3) Relapseb1 (2.8)6 (16.7)The p-value indicates the statistical significance of differences between healthy controls and patients with TBa Other diseases included cardiovascular diseases, neurologic diseases, renal dysfunction, chronic liver diseases, and osteoporosisb Relapse was defined as TB occurring within 2 years of treatment completion Table 1The demographic and clinical characteristics between healthy controls and patients with TBCharacteristicsHealthy controlsPatients with TBp valueN = 36 (%)N = 36 (%)Sex0.62 Male16 (44.4)22 (61.1) Female20 (55.6)14 (38.9)Age (years)0.35 Median (Interquartile range)63.0 (44.8–73.0)67.5 (51.8–74.0) 20–2925.0 (25.0–27.0)27.0 (24.0–27.0) 30–3931.0 (31.0–35.0)0 40–4946.5 (44.8–47.2)44.0 (42.0–47.0) 50–5954.0 (52.5–54.5)55.0 (51.8–58.2) 60–6963.0 (61.5–65.2)65.5 (62.5–68.8) 70–7973.0 (73.0-73.8)73.0 (70.0–74.0) ≥ 8085.0 (82.0–86.0)86.0 (83.5–87.0)BMI (kg/m2)0.25 Median (Interquartile range)22.7 (20.3–25.7)21.2 (19.8–23.1) Underweight (BMI < 18.5)17.7 (17.1–18.0)17.8 (17.8–18.0) Normal weight (BMI 18.5–24.9)21.6 (20.1–22.7)21.0 (20.0-21.9) Obesity (BMI ≥ 25.0)26.8 (25.9–28.4)28.0 (26.5–29.9)Comorbidity> 0.99 No19 (52.8)16 (44.4) Yes17 (47.2)20 (55.6)Type of comorbidity Hypertension10 (27.8)14 (38.8)0.87 Diabetes mellitus5 (13.9)10 (27.8)0.66 Dyslipidemia9 (25.0)5 (13.9)0.83 Othera8 (22.2)12 (33.3)> 0.99Smoking status0.57 Never smoked23 (63.9)14 (38.9) Smoke currently9 (25.0)11 (30.6) Formerly smoked4 (11.1)11 (30.5)Drinking0.48 No16 (44.4)17 (47.2) Yes20 (55.6)19 (52.8)Previous history of TB0.20 No35 (97.2)0 New030 (83.3) Relapseb1 (2.8)6 (16.7)The p-value indicates the statistical significance of differences between healthy controls and patients with TBa Other diseases included cardiovascular diseases, neurologic diseases, renal dysfunction, chronic liver diseases, and osteoporosisb Relapse was defined as TB occurring within 2 years of treatment completion The demographic and clinical characteristics between healthy controls and patients with TB The p-value indicates the statistical significance of differences between healthy controls and patients with TB a Other diseases included cardiovascular diseases, neurologic diseases, renal dysfunction, chronic liver diseases, and osteoporosis b Relapse was defined as TB occurring within 2 years of treatment completion Effects of MTB infection and treatment on gut and respiratory microbiome compositions After removing singletons, 778 bacterial species were identified in the gut of HCs, whereas patients with TB harbored 801 species at baseline and 836 after treatment. In the respiratory microbiome, 450 species were detected in HCs, whereas patients with TB-0 and TB-trt had 575 and 376 species, respectively. Overall, the gut microbiome exhibited higher species richness than the respiratory microbiome, with only modest differences among groups (HC, TB-0, and TB-trt) in the gut, whereas species richness in the respiratory microbiome peaked in TB-0 and declined following treatment (TB-trt). Core species analysis showed significant differences among groups in the gut microbiome (Fig. 1C; Table S1). HCs had the largest number of core species, whereas MTB infection was associated with a substantial loss of core species: 19 species (47.7%), including Bifidobacterium pseudocatenulatum. Only four new core species, including Prevotella copri, emerged in patients with TB. After TB treatment, the core microbiome contracted further, leaving only one core species, Erysipelatoclostridium ramosum. The most abundant species in the gut also differed according to TB infection and treatment status. P. copri and Bacteroides vulgatus were prevalent across all groups, whereas Faecalibacterium prausnitzii was more abundant in HCs but decreased in TB-0 and TB-trt. In contrast, Sutterella wadsworthensis was enriched in TB-0 (Fig. 1A). Alpha diversity did not differ between patients with TB and HCs (p = 0.981) but decreased significantly after treatment (p = 0.004; Fig. 2A). Beta diversity differed between HCs and TB-0 (p = 0.039) as well as between TB-0 and TB-trt (p = 0.001). In the respiratory microbiome, HCs also had the greatest number of core species, followed by TB-0, as the pattern observed in the gut (Fig. 1D; Table S2). A total of 54 species (61.3%) were identified as core species only in HCs, including species within the genera Neisseria and Leptotrichia. Four core species were specific to TB-0, including species within the genus Porphyromonas. The 15 core species identified after treatment all overlapped with those in HCs and included Streptococcus species. Common lung microbiome taxa such as Streptococcus, Neisseria, and Prevotella species [32] showed high abundances in the respiratory microbiome (Fig. 1B). In contrast, Pseudomonas aeruginosa was abundant only in TB-0, whereas Serratia marcescens became abundant after TB treatment. Alpha diversity differed only between TB-0 and TB-trt, whereas beta diversity differed between HCs and TB-0. Fig. 1Relative abundance and core microbiome analysis results for the gut and respiratory tracts of HC, TB-0, and TB-trt. The top-12 species in the gut (A) and respiratory (B) microbiome are shown based on percentage relative abundance. The numbers of core species identified in the gut (C) and respiratory (D) microbiome for HC, TB-0, and TB-trt are shown as a Venn diagram Relative abundance and core microbiome analysis results for the gut and respiratory tracts of HC, TB-0, and TB-trt. The top-12 species in the gut (A) and respiratory (B) microbiome are shown based on percentage relative abundance. The numbers of core species identified in the gut (C) and respiratory (D) microbiome for HC, TB-0, and TB-trt are shown as a Venn diagram Fig. 2Diversity analysis of the gut and respiratory microbiomes in HC, pTB-0, and TB-trt. Alpha-diversity was measured via the Shannon index, with p-values obtained from the Wilcoxon rank-sum test for the gut (A) and respiratory (B) microbiome. PCoA based on Bray-Curtis distance was used to assess beta-diversity in the gut (C) and respiratory (D) microbiome, and PERMANOVA was applied to test for significant differences between the groups Diversity analysis of the gut and respiratory microbiomes in HC, pTB-0, and TB-trt. Alpha-diversity was measured via the Shannon index, with p-values obtained from the Wilcoxon rank-sum test for the gut (A) and respiratory (B) microbiome. PCoA based on Bray-Curtis distance was used to assess beta-diversity in the gut (C) and respiratory (D) microbiome, and PERMANOVA was applied to test for significant differences between the groups Alterations in gut–respiratory microbial network properties and key taxa following MTB infection and treatment Gut–respiratory microbiome networks differed in topology across HC, TB-0, and TB-trt (Fig. 3; Tables S3–S5). TB-0 had the largest network, with the greatest numbers of nodes and edges, whereas TB-trt had the smallest (Table 2). In contrast, network complexity measured by density, clustering coefficient, and modularity increased progressively from HC to TB-0 and reached its highest level in TB-trt (Table 2). We used the Mantel test to assess whether the differences between networks were significant and found that the interactions differed between HC and TB-0, and between TB-0 and TB-trt (p < 0.01). Fig. 3Cooccurrence networks of the gut and respiratory microbiome in HC (A), TB-0 (B), and TB-trt (C). Lines within the network represent statistically significant edges at FDR < 10⁻⁸. Colors indicate clusters within the networks, and shapes distinguish gut and respiratory samples. Node size corresponds to betweenness centrality, and only nodes within the top 5% of betweenness are labeled with species names Cooccurrence networks of the gut and respiratory microbiome in HC (A), TB-0 (B), and TB-trt (C). Lines within the network represent statistically significant edges at FDR < 10⁻⁸. Colors indicate clusters within the networks, and shapes distinguish gut and respiratory samples. Node size corresponds to betweenness centrality, and only nodes within the top 5% of betweenness are labeled with species names Table 2Network properties of gut and respiratory microbiome co-occurrence networks in healthy controls and patients with TB before and after treatmentNodesHealthy controlsPatients with TBbefore treatmentPatients with TBafter treatment171186108Edges21,43725,7459,634Density0.7370.7480.834Clustering coefficient0.7960.8090.873Average degree250.7276.8178.4Modularity0.0840.1300.136Nodes represent the number of bacterial taxa included in the network, while edges indicate the number of connections between taxa. Density refers to the proportion of possible connections that are present in the network. Clustering coefficient measures the tendency of nodes to form local clusters. Average degree represents the mean number of connections per node. Modularity quantifies the extent to which the network is divided into distinct modules, with higher values indicating greater modularity Table 2Network properties of gut and respiratory microbiome co-occurrence networks in healthy controls and patients with TB before and after treatmentNodesHealthy controlsPatients with TBbefore treatmentPatients with TBafter treatment171186108Edges21,43725,7459,634Density0.7370.7480.834Clustering coefficient0.7960.8090.873Average degree250.7276.8178.4Modularity0.0840.1300.136Nodes represent the number of bacterial taxa included in the network, while edges indicate the number of connections between taxa. Density refers to the proportion of possible connections that are present in the network. Clustering coefficient measures the tendency of nodes to form local clusters. Average degree represents the mean number of connections per node. Modularity quantifies the extent to which the network is divided into distinct modules, with higher values indicating greater modularity Network properties of gut and respiratory microbiome co-occurrence networks in healthy controls and patients with TB before and after treatment Nodes represent the number of bacterial taxa included in the network, while edges indicate the number of connections between taxa. Density refers to the proportion of possible connections that are present in the network. Clustering coefficient measures the tendency of nodes to form local clusters. Average degree represents the mean number of connections per node. Modularity quantifies the extent to which the network is divided into distinct modules, with higher values indicating greater modularity Nodes with high betweenness, representing species that mediate microbial interactions, varied across groups (Tables S6–S8). In the HC network, gut Bifidobacterium and Bacteroides species, and respiratory Streptococcus and Gemella species showed top 5% betweenness (labeled in Fig. 3). Diverse respiratory species, including Haemophilus influenzae, Actinomyces viscosus, and Peptostreptococcus anaerobius, emerged as bridge taxa in the TB-0 network. After TB treatment, Bifidobacterium and Bacteroides species distinct from those in HCs, along with gut Absiella innocuum, exhibited the highest betweenness, whereas some species, such as gut Sutterella wadsworthensis and respiratory Gemella sanguinis, remained central in both HC and TB-0. Many of the edges were unique to each group, in which HC had 11,927 unique edges (56% of total edges), and TB-0 had 16,235 unique edges (63%) when comparing the two networks. Comparing TB-0 and TB-trt, 21,336 edges were unique to TB-0 (83%), whereas 5,225 edges were unique to TB-trt (54%). Gut and respiratory microbial species and pathways significantly affected by MTB infection and its treatment Comparison between HCs and TB-0 identified four gut bacterial species, including Bacteroides dorei, Bifidobacterium adolescentis, Bifidobacterium pseudocatenulatum, that were significantly different in both differential abundance analysis and LEfSe (Fig. 4A; Table S9). Among these, S. wadsworthensis increased with TB infection and declined following treatment, whereas the other three species were consistently depleted in patients with TB. When comparing TB-0 and TB-trt, we detected significant changes in 32 gut species, including Absiella innocuum and Bacteroides ovatus, which became more abundant after treatment (Fig. 4B; Table S10). Fig. 4Differential abundance analysis results. Statistically significant results (FDR < 0.01) are shown with their log fold change (logFC) values and LDA scores. (A) compare the gut microbiome between HC and TB-0, while (B) compare between TB-0 and TB-trt. The bar size represents the log fold changes and LDA scores, with blue, red, and green bars indicating species more abundant in healthy controls, patients with TB before treatment, and patients with TB after treatment, respectively. The numeric values are FDR values Differential abundance analysis results. Statistically significant results (FDR < 0.01) are shown with their log fold change (logFC) values and LDA scores. (A) compare the gut microbiome between HC and TB-0, while (B) compare between TB-0 and TB-trt. The bar size represents the log fold changes and LDA scores, with blue, red, and green bars indicating species more abundant in healthy controls, patients with TB before treatment, and patients with TB after treatment, respectively. The numeric values are FDR values No respiratory species passed the significance threshold; however, when accounting for structural zeros (species present in one group but absent in another), several taxa were differentially represented between groups. For instance, M. tuberculosis was detected only in TB-0 but not in HCs and was also significant in LEfSe. Also, Lactobacillus amylovorus and Serratia marcescens were found exclusively in TB-trt, and the enrichment was significant compared to TB-0. We performed taxon set enrichment analysis to look at metabolites related to changes in the microbial community. Genes responsible for the synthesis of metabolites such as 4-hydroxyphenylacetic acid and phenylalanine were significantly lower in TB-0 than in HC (FDR < 0.05; Table 3). No significant differences were detected between TB-0 and TB-trt. Species associated with the production of each metabolite are listed in Supplementary Table S11. Table 3Significantly enriched microbial gene sets identified through taxon set enrichment analysis based on differential abundance analysis between healthy controls and patients with TB before treatmentTaxon SetMedian rankFDRProducers of 4-hydroxyphenylacetic acid-0.6380.008producers of phenylalanine-0.8230.008Producers of kaempferol-0.7670.020Producers of indole-3-acetic acid-1.000.023Taxon set name refers to microbial gene sets associated with the production of specific metabolites. Median rank of set members represents the relative enrichment of the microbial gene set, with negative values indicating higher enrichment in healthy controls. FDR is the adjusted p-value accounting for multiple comparisons, with values below 0.05 considered statistically significant Table 3Significantly enriched microbial gene sets identified through taxon set enrichment analysis based on differential abundance analysis between healthy controls and patients with TB before treatmentTaxon SetMedian rankFDRProducers of 4-hydroxyphenylacetic acid-0.6380.008producers of phenylalanine-0.8230.008Producers of kaempferol-0.7670.020Producers of indole-3-acetic acid-1.000.023Taxon set name refers to microbial gene sets associated with the production of specific metabolites. Median rank of set members represents the relative enrichment of the microbial gene set, with negative values indicating higher enrichment in healthy controls. FDR is the adjusted p-value accounting for multiple comparisons, with values below 0.05 considered statistically significant Significantly enriched microbial gene sets identified through taxon set enrichment analysis based on differential abundance analysis between healthy controls and patients with TB before treatment Taxon set name refers to microbial gene sets associated with the production of specific metabolites. Median rank of set members represents the relative enrichment of the microbial gene set, with negative values indicating higher enrichment in healthy controls. FDR is the adjusted p-value accounting for multiple comparisons, with values below 0.05 considered statistically significant Pathway abundance analysis revealed 29 functional pathways that were substantially more prevalent in TB-0 compared to HCs (Table 4). Seven of these decreased even more after TB treatment, which were related to the biosynthesis of coenzyme A, the non-oxidative pentose phosphate pathway, mycolic acid biosynthesis, methyl-branched fatty acid biosynthesis, the glyoxylate cycle, cholesterol degradation, and the methylcitrate cycle. Table 4Differentially abundant pathways between healthy controls and patients with TB before treatment based on PICRUSt2 analysis and LinDA (Linear models for differential abundance)Pathway IDPathway nameHC vs. TB-0TB-0 vs. TB-trtlogFCFDRlogFCFDRARO-PWYChorismate biosynthesis I-2.380.02-2.050.08COA-PWY*Coenzyme A biosynthesis I-2.420.02-2.940.03COA-PWY-1Coenzyme A biosynthesis II-2.450.02-1.910.09NONOXIPENT-PWY*Pentose phosphate pathway (non-oxidative branch) I-3.140.01-2.590.02PEPTIDOGLYCANSYN-PWYPeptidoglycan biosynthesis I-2.370.02-1.840.09PWY-1042*Glycolysis IV-3.350.01-3.080.02PWY-5384Sucrose degradation IV-2.530.02-2.880.06PWY-5667*Cytidine diphosphate (CDP)-diacylglycerol biosynthesis I-3.200.01-2.710.02PWY-5686Uridine monophosphate (UMP) biosynthesis I-2.360.02-1.840.09PWY-61215-aminoimidazole ribonucleotide biosynthesis I-2.280.02-1.840.09PWY-61225-aminoimidazole ribonucleotide biosynthesis II-2.190.04-1.850.09PWY-6123Inosine-5-phosphate biosynthesis I-2.370.02-1.840.09PWY-6124Inosine-5-phosphate biosynthesis II-2.370.02-1.820.09PWY-6163Chorismate biosynthesis from 3-dehydroquinate-2.350.02-2.050.08PWY-6277Superpathway of 5-aminoimidazole ribonucleotide biosynthesis-2.190.04-1.850.09PWY-6385Peptidoglycan biosynthesis III (mycobacteria)-2.450.02-1.850.09PWY-6386Uridine diphosphate (UDP)-N-acetylmuramoyl-pentapeptide biosynthesis II-2.530.02-1.950.09PWY-6387UDP-N-acetylmuramoyl-pentapeptide biosynthesis I-2.380.02-1.860.09PWY-6609*Adenine and adenosine salvage III-4.270.01-4.320.01PWY-6628Superpathway of L-phenylalanine biosynthesis-2.350.02-2.010.08PWY-6630Superpathway of L-tyrosine biosynthesis-2.350.02-2.010.08PWY-7220Adenosine deoxyribonucleotides de novo biosynthesis II-2.410.02-2.400.08PWY-7221Guanosine ribonucleotides de novo biosynthesis-2.410.02-1.880.09PWY-7222Guanosine deoxyribonucleotides de novo biosynthesis II-2.410.02-2.400.08PWY-7790UMP biosynthesis II-2.390.02-1.860.09PWY-7791UMP biosynthesis III-2.370.02-2.310.12PWY-8178*Pentose phosphate pathway(non-oxidative branch) II-3.150.01-2.580.02PWY0-1319*CDP-diacylglycerol biosynthesis II-3.200.01-2.710.02THRESYN-PWYSuperpathway of L-threonine biosynthesis-3.010.02-2.180.09Pathway identifier (ID) corresponds to the MetaCyc pathway identifier, and Pathway Name is the name of the pathway. logFC represents the log₂ fold change in pathway abundance, where positive values indicate higher abundance in patients with TB before treatment compared with healthy controls for the “HC vs. TB-0” comparison, and higher abundance after treatment compared with before treatment for the “TB-0 vs. TB-trt” comparison. FDR is the adjusted p-value accounting for multiple comparisons, with values below 0.05 considered statistically significant. Pathways marked with an asterisk (*) were also significantly decreased after treatment in patients with TB (FDR < 0.05) Table 4Differentially abundant pathways between healthy controls and patients with TB before treatment based on PICRUSt2 analysis and LinDA (Linear models for differential abundance)Pathway IDPathway nameHC vs. TB-0TB-0 vs. TB-trtlogFCFDRlogFCFDRARO-PWYChorismate biosynthesis I-2.380.02-2.050.08COA-PWY*Coenzyme A biosynthesis I-2.420.02-2.940.03COA-PWY-1Coenzyme A biosynthesis II-2.450.02-1.910.09NONOXIPENT-PWY*Pentose phosphate pathway (non-oxidative branch) I-3.140.01-2.590.02PEPTIDOGLYCANSYN-PWYPeptidoglycan biosynthesis I-2.370.02-1.840.09PWY-1042*Glycolysis IV-3.350.01-3.080.02PWY-5384Sucrose degradation IV-2.530.02-2.880.06PWY-5667*Cytidine diphosphate (CDP)-diacylglycerol biosynthesis I-3.200.01-2.710.02PWY-5686Uridine monophosphate (UMP) biosynthesis I-2.360.02-1.840.09PWY-61215-aminoimidazole ribonucleotide biosynthesis I-2.280.02-1.840.09PWY-61225-aminoimidazole ribonucleotide biosynthesis II-2.190.04-1.850.09PWY-6123Inosine-5-phosphate biosynthesis I-2.370.02-1.840.09PWY-6124Inosine-5-phosphate biosynthesis II-2.370.02-1.820.09PWY-6163Chorismate biosynthesis from 3-dehydroquinate-2.350.02-2.050.08PWY-6277Superpathway of 5-aminoimidazole ribonucleotide biosynthesis-2.190.04-1.850.09PWY-6385Peptidoglycan biosynthesis III (mycobacteria)-2.450.02-1.850.09PWY-6386Uridine diphosphate (UDP)-N-acetylmuramoyl-pentapeptide biosynthesis II-2.530.02-1.950.09PWY-6387UDP-N-acetylmuramoyl-pentapeptide biosynthesis I-2.380.02-1.860.09PWY-6609*Adenine and adenosine salvage III-4.270.01-4.320.01PWY-6628Superpathway of L-phenylalanine biosynthesis-2.350.02-2.010.08PWY-6630Superpathway of L-tyrosine biosynthesis-2.350.02-2.010.08PWY-7220Adenosine deoxyribonucleotides de novo biosynthesis II-2.410.02-2.400.08PWY-7221Guanosine ribonucleotides de novo biosynthesis-2.410.02-1.880.09PWY-7222Guanosine deoxyribonucleotides de novo biosynthesis II-2.410.02-2.400.08PWY-7790UMP biosynthesis II-2.390.02-1.860.09PWY-7791UMP biosynthesis III-2.370.02-2.310.12PWY-8178*Pentose phosphate pathway(non-oxidative branch) II-3.150.01-2.580.02PWY0-1319*CDP-diacylglycerol biosynthesis II-3.200.01-2.710.02THRESYN-PWYSuperpathway of L-threonine biosynthesis-3.010.02-2.180.09Pathway identifier (ID) corresponds to the MetaCyc pathway identifier, and Pathway Name is the name of the pathway. logFC represents the log₂ fold change in pathway abundance, where positive values indicate higher abundance in patients with TB before treatment compared with healthy controls for the “HC vs. TB-0” comparison, and higher abundance after treatment compared with before treatment for the “TB-0 vs. TB-trt” comparison. FDR is the adjusted p-value accounting for multiple comparisons, with values below 0.05 considered statistically significant. Pathways marked with an asterisk (*) were also significantly decreased after treatment in patients with TB (FDR < 0.05) Differentially abundant pathways between healthy controls and patients with TB before treatment based on PICRUSt2 analysis and LinDA (Linear models for differential abundance) Pentose phosphate pathway (non-oxidative branch) II Pathway identifier (ID) corresponds to the MetaCyc pathway identifier, and Pathway Name is the name of the pathway. logFC represents the log₂ fold change in pathway abundance, where positive values indicate higher abundance in patients with TB before treatment compared with healthy controls for the “HC vs. TB-0” comparison, and higher abundance after treatment compared with before treatment for the “TB-0 vs. TB-trt” comparison. FDR is the adjusted p-value accounting for multiple comparisons, with values below 0.05 considered statistically significant. Pathways marked with an asterisk (*) were also significantly decreased after treatment in patients with TB (FDR < 0.05) Discussion This study investigates the impact of TB infection and treatment on the gut and respiratory microbiome at the species level, focusing on community composition and network structure. Using both gut and respiratory samples from the same participants, we sought to demonstrate the relationship between TB and the gut–respiratory axis. Matching between HCs and TB-0 was performed to minimize inter-individual variability, and longitudinal comparisons were made using samples collected at diagnosis (TB-0) and two months after treatment with anti-TB drugs (TB-trt). Together, our data shows (1) identified species associated with TB infection and treatment in terms of abundance and importance within the microbial network, (2) how microbiome diversity and global network metrics correlated with TB. Core microbiome depletion, which may lead to a loss of community stability [33], and decreased abundance of several Bifidobacterium species were observed in TB-0 compared to HC. Particularly, B. pseudocatenulatum was a core species only in HC, and its network centrality was reduced in TB-0. Bifidobacterium species produce short-chain fatty acids (SCFAs), which contribute to gut health and immune regulation, and decreases in these species in patients with TB have also been reported in previous studies [15, 34, 35]. In addition to SCFAs, TB-0 showed a decrease in metabolites such as 4-hydroxyphenylacetic acid and phenylalanine, which are involved in maintaining the intestinal barrier and reducing inflammation [36, 37]. Further, kaempferol acts as an antioxidant, and indole-3-acetic acid promotes mucin sulfation [38, 39]. Pathways that are essential for MTB survival and pathogenesis, including Coenzyme A and lipid biosynthesis, glycolysis, nucleotide recycling, and the pentose phosphate pathway, were enriched in TB-0 but decreased in TB-trt [40–44]. These findings suggest that MTB infection may compromise gut homeostasis and reshape metabolic pathways to support its persistence, and this is partially restored by treatment. Gut and respiratory microbiome networks further showed that SCFA–producing and gut homeostasis-associated species—Bifidobacterium spp., Agathobacter faecis (also known as Roseburia faecis), Alistipes shahii, Bacteroides uniformis, and Eubacterium ventriosum—served as bridge taxa in HC but not in TB-0 [45]. In contrast, respiratory pathogens or opportunistic pathogens, including H. influenzae, A. viscosus, and P. anaerobius, showed high betweenness in the networks of TB-0 [46, 47]. TB infection altered not only the composition of the gut and respiratory microbial community but also their interactions, shifting the networks toward dysbiosis. SCFA-producing species, such as (A) innocuum, Anaerostipes hadrus, and (B) ovatus, reemerged as bridge taxa in TB-trt [48, 49], and (A) innocuum and (B) ovatus were also significantly enriched after treatment. On the other hand, Streptococcus pseudopneumoniae, a pulmonary pathogen, also became central [50]. This reflects the restoration of the network after treatment, albeit incompletely. Disruptions in the gut microbiome caused by TB treatment have also been reported in other studies and appear to persist over time [51]. S. wadsworthensis in the gut had high betweenness in both HC and TB-0, with an enrichment in TB-0, which decreased after treatment, suggesting its potential as a TB-associated bacterial marker. Its rise during infection may reflect immune or inflammatory responses, whereas its decline after treatment could indicate partial microbiome recovery, with possible involvement in inflammation and gut dysbiosis [52]. The construction of combined networks of the gut and respiratory microbiome from the same participant revealed numerous significant connections between species at the two sites, for instance, between gut Bifidobacterium species and respiratory Gemella species, supporting the existence of the gut–respiratory axis, in which alterations in one microbiome can influence the other [7]. Our analysis showed that MTB infection and treatment contribute to changes in the microbial networks. The large number of edges unique to each group suggests that TB infection and treatment alter gut–respiratory bacterial interactions by eliminating existing connections and forming new ones. Network comparisons indicated that MTB infection expands overall network size, whereas treatment reduces it, likely as a result of microbial diversity loss [53]. Network complexity was higher in TB-0 than in HC and increased further in TB-trt, indicating tighter associations among surviving microbes and reflecting microbial community restructuring. This study has certain limitations regarding the collection of respiratory tract samples, particularly from HCs. Although sputum samples are commonly used as a proxy for the lower airway microbiome, they may include contributions from oral and nasal communities [54]. We attempted to approximate lower airway microbiomes by excluding sputum samples with a high cumulative relative abundance of oral and nasal commensals. Still, the results should be interpreted with caution. As this study is observational, there remains the possibility of reverse causation, residual confounding, and shared etiology, even though we attempted to control for these factors in the study design. In addition, sampling was performed only two weeks after treatment initiation and did not extend to six months after treatment completion, due to the project duration and participant dropout. Future studies that include sampling after full treatment could provide further insights into the effects of TB treatment on the gut and respiratory microbiome. Lastly, the functional predictions have inherent limitations because they are based on reference databases rather than experimental validation. Further experimental verification, such as animal colonization experiments or clinical intervention trials, would help to validate the findings of this study, including the proposed microbial markers. MTB infection and treatment exert significant effects on the gut–respiratory axis. TB infection caused a loss of SCFA-producing and intestinal barrier–associated species, along with an increase in inflammation-related species and microbial pathways essential for MTB survival. Similar patterns were observed in the gut–respiratory microbial networks, which changed significantly with TB infection and treatment. SCFA-producing species acted as bridge taxa in the network of HC and reemerged in TB-trt, whereas pathogenic species became central in TB-0 and remained partly present in TB-trt. The microbial community disruptions caused by TB were restored after treatment, but incompletely, leading to the need for strategies to rebalance the microbiome and improve treatment outcomes [51]. The microbial taxa identified as significant in this study are not only associated with TB but also have broader clinical relevance across respiratory diseases. Decreases in core microbiome taxa and depletion of Bifidobacterium have been reported in chronic obstructive pulmonary disease (COPD), and Bifidobacterium has also been linked to other respiratory diseases, including COVID-19 and influenza [55–57]. Beyond TB, short-chain fatty acid (SCFA)–producing bacteria are known to suppress lung inflammation and have been associated with multiple respiratory infections, such as those caused by SARS-CoV-2 [58]. Consistent with these observations, Bifidobacterium and other SCFA-producing species have been reported to have therapeutic potential in respiratory diseases. Further experimental validation of their effects in TB could therefore provide important insights into their roles in MTB infection and treatment. Through analyses of community composition, diversity, and network structure, we provide a comprehensive view of how TB and its treatment are associated with the gut and respiratory microbiome. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2. Supplementary Material 3. Supplementary Material 3.