Metatranscriptomic analysis of upper and lower respiratory tract microbiomes in patients with tuberculosis and community-acquired pneumonia Authors: Liangyu Li, Meng Xu, Mengling Liu, Jieyu Mao, Pei Xiong, Ruiyun Li, Jianjun Chen, Xiaojun Wu Published: 2026 ABSTRACT AbstractBackgroundAlthough changes in the lung microbiome have been observed in many respiratory diseases, the lung microbiome of patients with tuberculosis (TB) remains largely undefined. The aim of this study was to determine and compare the composition of upper and lower respiratory microbial communities, changes in host gene expression, and functional pathway activation in patients with TB and community-acquired pneumonia using a Metatranscriptomic approach.MethodsFrom November 2020 to November 2021, 42 bronchoalveolar lavage fluid samples, 10 oropharyngeal swabs, and 10 nasopharyngeal swabs were collected from patients hospitalized with TB or community-acquired pneumonia for RNA-sequencing, within 72 h of admission. Data from 28 healthy controls were downloaded from the National Center for Biotechnology Information database.ResultsThe most common microorganisms in the samples from patients with TB were Prevotella, Escherichia, Mycobacterium, and Verrucoccus. Notable differences in microorganism diversity were observed between the TB and community-acquired pneumonia groups as well as between the upper and lower respiratory tracts compared to that in healthy controls. Altered microbial interactions were also observed. Rothia and Mycobacterium were identified as marker microorganisms in the TB group, which exhibited increased expression of SNAP25, MGAM2, CNTN4, MMP12, and GPR174, in parallel to IL36A, RHCG, and CYP2B6 downregulation relative to the community-acquired pneumonia group. The pathways enriched for differential genes were similar among all patient groups, particularly involving neuroactive ligand receptor interaction and Ca2+ signaling. Patients with TB differed from those with community-acquired pneumonia, particularly with regard to the cytokine–cytokine receptor interaction pathway. The most commonly detected antibiotic resistance genes conferred resistance against β-lactams and macrolides. Antibiotic resistance genes were more abundant in oropharyngeal than in nasopharyngeal samples.ConclusionsThe composition of microbial communities and host expression landscapes differs between TB and community-acquired pneumonia as well as between different respiratory tract sites. These changes in lung microbiota may impact lung disease development and prognosis.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12879-026-13614-x.Keywords: Pulmonary alveolar lavage fluid, Respiratory tract infection, Pulmonary tuberculosis, Metatranscriptomic Background Lower respiratory tract infections (LRTIs) are a long-standing global public health problem. Owing to advances in sequencing technology, an increasing number of studies have found that the lung microbiome plays an important role in the pathophysiological processes driving respiratory disease [1]. The lungs are not sterile, with the main bacterial genera found in the lower respiratory tract of healthy individuals, including Prevotella, Veillonella, and Streptococcus [2]. Acute and chronic lung diseases alter lung microbial communities, influencing the migration, elimination, and regional growth conditions of microbial species [3]. Enhanced comprehension regarding the influence of the lung microbiome on health and disease can provide insights for the development of novel diagnostic and therapeutic approaches. Tuberculosis (TB) is caused by the Mycobacterium tuberculosis (Mtb) complex and is one of the leading causes of death worldwide. According to the World Health Organization, 1.23 million people will die from TB in 2024 alone [4]. Changes in the lung microbiome have been observed in many respiratory diseases, including chronic obstructive pulmonary disease, asthma, and cystic fibrosis [5–10]. However, the lung microbiome of patients with TB remains largely undefined. Studies investigating the airway microbiome of patients with TB using next-generation sequencing techniques are scarce [11], with inconsistent and sometimes contradictory results [9]. To our knowledge, there has been no comparative analysis of the lung microbiome between patients with TB and those with other lung infections, such as community-acquired pneumonia (CAP). Therefore, we performed a metatranscriptomic analysis of the lung microbiome in bronchioalveolar lavage fluid (BALF) from patients with TB and CAP, with the goal to determine whether different infections are associated with a distinct lung microbiome. Methods Study design and study population This prospective observational study was conducted at the East Campus of Renmin Hospital of Wuhan University, Wuhan, China, between November 2020 and November 2021. The study protocol was approved by the Institutional Review Board of Renmin Hospital of Wuhan University (approval number: WDRY2024-K190), and all procedures were performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants prior to any study-related procedures. Patients hospitalized with suspected lower respiratory tract infection were screened for enrollment. Participants with a final definitive clinical diagnosis of community-acquired pneumonia (CAP) or tuberculosis (TB) were included in the final analysis. The diagnosis of CAP was established in accordance with the clinical practice guidelines issued by the Chinese Thoracic Society [12]. The diagnosis of active pulmonary TB was based on the Chinese Health Industry Standard Diagnosis for Pulmonary Tuberculosis (WS 288–2017) [13]. Exclusion criteria for all participants were as follows: (1) immunocompromised status; (2) pre-existing structural lung disease; (3) history of pulmonary tuberculosis within 5 years prior to enrollment; (4) intolerance to fiberoptic bronchoscopy; (5) receipt of systemic anti-tuberculosis treatment (TB group) or systemic antibiotics (CAP group) prior to specimen collection. All fiberoptic bronchoscopy procedures and bronchoalveolar lavage fluid (BALF) collection were performed for clinically indicated diagnostic purposes as part of routine patient care; no additional invasive procedures were performed solely for the purposes of this study. Specimen collection and processing For all enrolled participants, 5 mL of bronchoalveolar lavage fluid (BALF) was collected via bronchoscopy performed within the first 12 h of study enrollment. Paired BALF, nasopharyngeal (NP) swabs, and oropharyngeal (OP) swabs were collected simultaneously from a subset of CAP patients. All swab specimens were immediately preserved in sterile commercial sample preservation medium (Mandelab, Shenzhen, China) after collection. All specimens were immediately stored at − 80 °C following collection until downstream processing. Publicly available BALF metatranscriptomic data from healthy individuals were retrieved from the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA) database [14]. All raw sequencing data from control datasets were processed using the identical bioinformatic pipeline as our in-house specimens, with consistent sequencing platforms and analytical parameters applied to minimize technical heterogeneity between groups. Demographic characteristics, clinical symptoms, laboratory test results, and treatment outcomes were collected for all enrolled participants using standardized case report forms. RNA extraction and sequencing Total RNA was extracted from 200 µL of BALF specimens and 2 ml of swab preservation buffer from OP and NP specimens using RNeasy Plus Universal Mini Kit (QIAGEN, USA), strictly following the manufacturer’s instructions. Strand-specific RNA sequencing libraries were constructed with the Revelo RNA-Seq Kit (TECAN, USA) per the manufacturer’s protocol. Quantified by Qubit. All qualified libraries were sent to GENEWIZ (Suzhou, China) for high-throughput sequencing on the Illumina NovaSeq 6000 platform, with a paired-end 150 bp sequencing strategy and a raw data output of 10 Gb per sample. The human reference genome GRCh38 used for read alignment was downloaded from the National Genomics Data Center (NGDC) [15, 16]. Data processing and analysis Quality assessment of the raw data was performed using FastQC (v0.11.9) [17]. Quality filtering and splice trimming were performed using Fastp (v0.21.0) [18]. Clean reads were mapped to the human reference genome (GRCh38/hg38) using Hisat2 (v2.2.1) [19]. Host and microbial reads were extracted using SAMtools (v1.12) [20] and BEDTools (2.30.0) [21], respectively. Microbial reads were assembled using SPAdes software (v3.15.5) [22]. Resistance genes were annotated on the assembled contigs using ABRicate (https://github.com/tseemann/abricate). Microbial reads were compared to the National Center for Biotechnology Information non-redundant nucleic acid database for species annotation using Kraken2 (v2.1.2) [23]. Diversity analysis was performed using the vegan package. The linear discriminant analysis (LDA) effect size (LEfSe) method was used to identify species with significant differences in relative abundance. Microbial interaction networks were constructed using the SParse InversE Covariance estimation for the Ecological ASsociation Inference (SPIEC-EASI) package (https://github.com/zdk123/SpiecEasi). Gephi (v0.10.1) software was used for visualization editing. The host reads were quantified by applying featureCounts (v2.0.3) [24], and expression matrices were generated. The DESeq2 package was applied to the expression matrix to obtain differential expression files. The enrichment plot package was applied to the differential expression files for the Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis. Statistical analysis All statistical analyses were performed using R software (version 4.2.1, R Foundation for Statistical Computing, Vienna, Austria). For measurement data, all variables were presented as mean (interquartile range, M [P25, P75]). For comparisons between two independent groups, the independent samples t-test was used for normally distributed data, and the Mann-Whitney U test (Wilcoxon rank-sum test) for non-normally distributed data. For comparisons among three groups, one-way analysis of variance (ANOVA) was applied for normally distributed data, followed by the Bonferroni method for post-hoc pairwise comparisons if the overall ANOVA result was statistically significant; the Kruskal-Wallis H test was used for non-normally distributed data, with Bonferroni correction for pairwise comparisons. For categorical data, results were presented as number. Between-group comparisons were performed using the Pearson χ² test, or Fisher’s exact test when the expected frequency was < 5 in more than 20% of the cells. All tests were two-sided, and a P-value < 0.05 was considered statistically significant. Results Study population A total of 120 subjects were included in the final analysis of this study. The study cohort was divided into the disease group and healthy control group, with further subgrouping based on clinical diagnosis and specimen type. Among the prospectively enrolled subjects, 26 patients were definitively diagnosed with tuberculosis (TB) and 36 patients with community-acquired pneumonia (CAP) according to the pre-specified diagnostic criteria. Bronchoalveolar lavage fluid (BALF) specimens were collected from all 62 patients during clinically indicated bronchoscopy. Paired nasopharyngeal (NP) swabs, oropharyngeal (OP) swabs and BALF specimens were simultaneously collected from 10 patients in the CAP group. Four BALF specimens were excluded from downstream analysis due to library construction failure (post-construction quality control failed to meet sequencing standards). For the control groups, metatranscriptomic data of 28 microbiologically healthy individuals (SRP305669) and 20 healthy controls (HRA000143) for host transcriptome analysis were retrieved from the National Center for Biotechnology Information (NCBI) database. The study flow and cohort grouping are shown in Fig. 1. Quality control was performed on the downloaded raw sequencing data using the identical pipeline as in-house specimens. The detailed data in Supplementary Table 1. Number of cases in different subgroups and flowchart for data analysis The demographic and clinical characteristics of the enrolled TB and CAP patients are summarized in Table 1. Compared with the CAP group, the TB group had a significantly shorter interval from admission to sampling, longer symptom duration before admission, shorter total hospital stay, lower rate of antibiotic exposure before sampling, and higher peripheral blood basophil count (all P < 0.05). The majority of patients in the TB group were transferred to other medical institutions. No significant between-group differences were observed in gender, age, baseline vital signs, or clinical symptom profiles (all P > 0.05). Table 1Demographic and clinical characteristics of the study participantsVariableCAP (n = 36)TB (n = 26)P-value General condition Age, years51.94(39,62)54.57(37.5,67)0.6474Sampling day6.53(2,9)3.36(2,4)0.0129Hospital length of stay11.2(9,13)6.88(5,8)0.0000Sick day48.05(6,16)66.06(15,60)0.0043Body temperature (℃)36.7(36.3,36.5)36.43(36.3,36.5)0.0968Pulse (per min)86.38(78,93)87.69(77.25,103.25)0.7391Breath (per min)19.41(19,20)19.88(20,20)0.2009Systolic pressure (mmHg)133.93(117,146)123.54(110.75,126.75)0.0646Diastolic pressure (mmHg)78.38(72.25,85)75.08(69.25,78.75)0.2838 Laboratory values WBC 109cells/L7.67(4.11,10.73)6.23(5.38,7.01)0.1291Neutrophils %65.06(52.25,80.17)58.06(53.25,65.65)0.1220Lymphocyte %22.79(10.02,34.97)26.28(20.1,33)0.3133Monocyte %9.9(7.9,11.5)8.86(6.7,11.05)0.3148Eosinophil %6(1.35,5.45)1.88(0.95,2.3)0.1496Basophil %0.66(0.5,0.8)0.83(0.55,0.9)0.2027Neutrophils 109cells/L4.82(2.17,6.79)6.45(3.42,4.82)0.5323Lymphocyte 109cells/L5.1(1.04,1.97)3.01(1.15,2.19)0.5922Monocyte 109cells/L0.61(0.43,0.7)0.62(0.46,0.68)0.8916Eosinophil 109cells/L0.18(0.01,0.2)0.14(0.1,0.18)0.5805Basophil 109cells/L0.03(0.02,0.04)0.06(0.04,0.08)0.0086CRP mg/L31.75(0.5,48.4)18.69(0.5,32.2)0.3416PCT ng/mL0.81(0.04,0.87)0.72(0.04,0.09)0.9011 Department 0.7800Respiratory medicine3423Thoracic surgery11Others12 Gender 0.5600Male2012Female1613 Outcome 2.40E-10Improved355Transferred121 Antibiotic 7.32E-06Used3310Unclear316 Clinical feature Cough18171Sputum13140.6072Fever1020.0225Pharynx discomfort300.2379Runny nose or nasal congestion211Chest pain500.054Chest tightness730.3098Dyspnea400.1134Nodular shadow270.0735Wheezing560.7341Hemoptysis221Lung auscultation960.5357Smoking (past or current)530.7061Drinking (past or current)211Hypertension541Diabetes211Hyperlipidemia211Chronic heart disease211Chronic lung disease410.3713Chronic liver disease250.2389Chronic kidney disease001Digestive Disease101Notes: Data are presented as means (Interquartile Range) or number. Abbreviations: CAP, community-acquired pneumonia; TB, tuberculosis Demographic and clinical characteristics of the study participants Notes: Data are presented as means (Interquartile Range) or number. Abbreviations: CAP, community-acquired pneumonia; TB, tuberculosis Composition of the microbiota The dominant phyla were the same between patients with different lung diseases and healthy controls. The healthy group showed an absolute predominance of the Proteobacteria phylum (60.22%), whereas patients with CAP and those with TB showed a relative decrease in Proteobacteria detection (26.11% and 25.48%, respectively). The predominant phyla detected at different sites were consistent, but the proportions varied considerably, with dominated by Bacteroidetes, followed by Proteobacteria (Sfig. 1). At the genus level, the most commonly detected pathogens were Prevotella, Neisseria, Escherichia, Mariprofundus, and Veillonella. In the TB group, these included Prevotella, Escherichia, Mycobacterium, Pedobacter, and Veillonella. In the CAP group, pathogens detected were Prevotella, Neisseria, Pedobacter, Escherichia, and Veillonella. Most common in the Health group were Mariprofundus, Prevotella, Ralstonia, Escherichia, and Neisseria. In the lower respiratory tract (BALF) group, most were Prevotella, Neisseria, Pedobacter, Escherichia, and Veillonella versus Prevotella, Capnocytophaga, Neisseria, Gemmatirosa, and Campylobacter in the OP group and Malassezia, Klebsiella, Candidatus, Thalassarchaeum, and Staphylococcus in the NP group (Fig. 2). Top 50 genera of microorganisms by disease group and site based on mean abundance Diversity of microbiota Alpha diversity analysis revealed lower Observed indices in patients with CAP and TB compared with those in the healthy controls. Observed indices were decreased in the BALF samples compared to those in the NP and OP (Fig. 3A and C). PCoA based on Bray-Curtis distance scaling analysis showed that the centroids of the CAP and TB groups were significantly separated from those of the healthy group, indicating that the occurrence of different diseases led to significant changes in the mean community composition of the patients. Moreover, statistically significant differences in β-diversity were observed between BALF, OP, and NP, suggesting distinct microbial compositions at different sites (Fig. 3B and D). Diversity analysis at different disease groups and sites. (A) Alpha diversity in different disease groups. (B) Beta diversity in different disease groups. (C) Alpha diversity at different sites. (D) Beta diversity at different sites Differences in microbiota Linear discriminant analysis Effect Size (LEfSe) was performed to identify discriminative microbial marker taxa between groups, with a significance threshold set at linear discriminant analysis (LDA) score ≥ 4 and P < 0.05. The CAP group for Escherichia, Pedobacter, Klebsiella, Enterovirus, and Streptococcus, and TB group for Mycobacterium, Rothia, and Capnocytophaga. Moreover, significant differences were found for Prevotella, Capnocytophaga, Neisseria, Veillonella, Campylobacter, Gemmatirosa, Leptotrichia, and Abyssicoccus in the OP group; Malassezia, Klebsiella, Ndidatus Thalassarchaeum, Corynebacterium, and Staphylococcus in the NP group; and Streptococcus, Enterovirus, and Porphyromonas in the BALF group (Fig. 4A and B). We further conducted exploratory correlation analysis to investigate the potential interactions between these differential marker taxa and host differential genes, with detailed results presented in Supplementary Table 2. We further analyzed the signature microbes and differential host gene expression. However, limited by the low microbial biomass in BALF samples and the small sample size, the results did not survive strict multiple testing correction. Only weak yet biologically plausible host-microbe interaction signals were identified. Streptococcus was the most prominent core marker taxon for the CAP group, and Rothia was the key signature taxon for the TB group (Fig. 4C and D). Interactions across different sampling sites are presented in Sfig 2. LEfSe analysis at different disease groups and sites. (A) LEfSe analysis at different disease groups. (B) LEfSe analysis at different sites. (C) Interaction networks between significantly differential microbes and host genes in the CAP group. (D) Interaction networks between significantly differential microbes and host genes in the TB group. Network description: Red nodes denote significantly differential microbes in the corresponding disease group; blue nodes denote differentially expressed genes in the corresponding disease group; gray edges represent potential interactions between microbes and host genes. LefSe, linear discriminant analysis effect size. CAP, community-acquired pneumonia; TB, tuberculosis; OP, oropharyngeal; NP, nasopharyngeal Antibiotic resistance Annotation of the antibiotic resistance genes in all samples revealed that the most common type of resistance was against beta-lactams, followed by macrolides, aminoglycosides, chloramphenicol, lincosamides, and tetracyclines. There were no significant differences in drug resistance among the different disease groups nor between disease groups and the healthy group. However, in the analysis of different sites, antibiotic resistance genes were more common in OP samples and less common in the NP samples, and the resistance genes detected in BALF samples were distinct from those in the OP and NP samples (Fig. 5). Resistance genes detected in different disease groups and sites Microbial interaction networks Microbial co-occurrence networks were constructed for different disease groups and respiratory tract sites using the SPIEC-EASI algorithm. Significant differences in network scale were observed among disease groups (Fig. 6A): the healthy control group had the largest network (3867 nodes, 155840 edges), followed by the CAP group (2972 nodes, 87593 edges) and the TB group (2596 nodes, 68608 edges) (Fig. 6B, Supplementary Table 3). Network density was comparable across all groups (0.0198–0.0208). The healthy group had the lowest global clustering coefficient (0.3107) and modularity (0.4318), which were significantly elevated in the CAP (0.4143/0.5334) and TB (0.4880) groups (Fig. 6C), indicating enhanced local microbial clustering under infectious conditions and more prominent functional differentiation of the microbial community in CAP patients. Hub species analysis revealed unique core hubs in the healthy group. The CAP and TB groups shared some hubs but with significantly lower node degrees, and disease-specific hubs were also identified (Supplementary Table 4). Overall, infection was associated with reduced network scale, altered topological structure, and reshaped hub species composition, indicating niche specialization and functional reorganization of the respiratory microbial community. SPIEC-EASI microbial networks of different disease groups. (A) Microbial networks at different disease groups . (B) Global Network Metrics by different disease groups . (C) Local Network Metrics Distribution in different disease groups . Network: dots represent genera of microorganisms, and connections between dots indicate the presence of interactions; different sizes of dots represent the number of connections. CAP, community-acquired pneumonia; TB, tuberculosis Marked differences in network topological features were observed across BALF, OP, and NP specimens from CAP patients (Sfig. 3 A): the BALF network had the smallest scale, longest network diameter and average path length, as well as the highest clustering coefficient (0.387) and modularity (0.682); the OP network had the highest density and shortest average path length; the NP network had the highest species richness but the sparsest interspecies interactions (Sfig. 3B and 3 C, Supplementary Table 3). Cross-regional common hubs and site-specific hubs were identified across different sites (Supplementary Table 4), which exerted universal regulatory roles and site-specific ecological functions, respectively. Host gene expression Compared with the healthy group, we found significant changes in host gene expression in the different disease groups. Host gene expression also varied among sampling sites. The CAP group exhibited an upregulation of 2,118 genes, such as OR10P1, TNNC2, and C19orf81, in addition to 921 downregulated genes, including SPRR3, SPRR2A, and STATH. In the TB group, 2,027 genes were upregulated, including STATH and C12orf75, in addition to 1,121 downregulated genes, such as TNNC2, NTRK1, CHRND, RTBDN, and TAF6L(Supplementary Table 5, Sfig. 4 A). Compared with BALF, there were 1,689 upregulated genes in OP, including FAM25G, KRT80 KRT16, and CRYBG2; in addition to 1,257 downregulated genes, including DYNLRB2, C20orf85, TFF1, SCGB3A1, and FAM216B. In NP, we detected 8,508 upregulated genes, including TXNDC8, DEFB124, AADACL2, and SLC9C1, versus 135 downregulated genes, including DEFA3 FABP4, HTN3, and CCL18 (Supplementary Table 6, Sfig. 4B). KEGG enrichment analysis revealed that the pathways enriched for differentially expressed genes were similar in patients with CAP and TB. The TB group differed mainly in the cytokine–cytokine receptor interaction pathway. Pathways involved in endocytosis and protein processing in the endoplasmic reticulum were enriched in the NP samples, whereas pathways related to coronavirus 2019 and ribosomes were enriched in the OP samples, compared to those in BALF samples (Fig. 7). Pathway enrichment analysis of host differentially expressed genes in different disease groups and sites Additional files Discussion In this study, we performed a comprehensive metatranscriptomic analysis of the upper and lower respiratory tract microbiomes in TB and CAP patients, and compared them with healthy controls. Our results revealed significant differences in microbial community composition, diversity, interaction patterns, host gene expression profiles, and functional pathway enrichment between TB and CAP patients, as well as between different anatomical sites of the respiratory tract. To our knowledge, this is one of the first studies to simultaneously characterize the active respiratory microbiome and host transcriptome in TB patients using a metatranscriptomic approach, providing new insights into the host-microbe interactions in TB. Our study found that the dominant phyla of the respiratory microbiome were consistent between TB, CAP patients and healthy controls, but with significant differences in relative abundance. The healthy control group was dominated by Proteobacteria, while the relative abundance of Proteobacteria was significantly decreased in the TB and CAP groups, which is consistent with previous studies on the respiratory microbiome in pulmonary infectious diseases. At the genus level, Prevotella, Escherichia, Mycobacterium, and Verrucoccus were the most abundant genera in the TB group [25]. We found significant differences in the microbiota at different sampling sites, and the Prevotella, Veillonella, and Neisseria genera were more likely to be detected in the OP samples than in the lower respiratory tract. These results are in line with the findings of Leung et al. [26] regarding the microbiology of flu-infected OP fluid in which the oral commensals, Prevotella, Veillonella, and Neisseria, were found. In contrast, Klebsiella spp. were more easily detected in the NP samples. The notion that Klebsiella pneumoniae colonizes local NP samples was also suggested in a study by Yasir et al., who explored changes in the NP microbiota of patients with obesity from Saudi Arabia infected with severe acute respiratory syndrome coronavirus 2 [27]. LEfSe analysis identified Mycobacterium and Rothia as marker microorganisms for the TB group. Rothia is an opportunistic pathogen most commonly found in immunocompromised patients and patients with TB. Cases of endocarditis, bacteremia, central nervous system infections, peritonitis, osteomyelitis, and LRTI have been reported to result from Rothia mucilaginosa infection [28]. Rothia mucilaginosa has also been reported as the single causative agent of pneumonia, and Zeidler et al. [29] found that it is closely associated with reactive arthritis. Hong et al. [1] found that R. mucilaginosa was an anchoring species in the network of bacteria co-infecting with Mtb. Conversely, another study found that R. mucilaginosa was decreased during active inflammation in patients with TB [30]. Further studies are therefore necessary to clarify the relationship between Rothia and TB as well as determine whether the initial infection of patients with Rothia promotes TB infection, and vice versa. This will provide insights into therapeutic strategies for the management of complex co-infections in patients with TB. Owing to the inherently low microbial abundance and restricted sample size of bronchoalveolar lavage fluid samples, no statistically significant associations between microbial profiles and host transcriptional signatures remained after strict correction for multiple comparisons. Nevertheless, we observed marginal host-microbe interaction signals with potential biological relevance, which require confirmation in larger, well-powered cohorts. For antibiotic resistance gene profiling, we found that the most common antibiotics resistance genes (ARGs) were those conferring resistance to β-lactams and macrolides, which are the most widely used antibiotics for respiratory tract infections in clinical practice [27]. There were no significant differences in ARG abundance between the TB, CAP and healthy groups, which may be related to the fact that none of the enrolled patients received antibiotic or anti-tuberculosis treatment before specimen collection, avoiding the selective pressure of antimicrobial drugs on ARGs. We also found that ARGs were more abundant in OP specimens than in NP specimens, which is consistent with the fact that the oropharynx is a major reservoir of ARGs in the respiratory tract [31]. We constructed microbial co-occurrence networks to explore the interactions between microorganisms under different disease conditions. Our results showed that compared with the healthy group, the TB and CAP groups had smaller network size (fewer nodes and edges), but higher clustering coefficient and modularity. These findings suggest that under disease conditions, the respiratory microbial community contracts into a tighter and more modular structure [32], which may reflect niche specialization and functional reorganization of the microbial community in response to infection and host immune pressure. The hub species in the healthy group were completely different from those in the disease groups, and the node degrees of the common hub species in the disease groups were significantly lower than those in the healthy group, indicating that the core “hub” function of the healthy microbial network was weakened in disease states, which may be associated with the disruption of microbial homeostasis in pulmonary infection. Our study revealed significantly altered gene expression profiles in patients with TB. SPRRs are a group of proteins produced by small proline-rich protein genes. SPRR2A is an antimicrobial protein that shapes intestinal microbiota composition and protects the gut from infections [33]. Studies have revealed that SPRR3 plays a regulatory role in allergic airway inflammation [34]. OR10P1 is associated with beta-blocker-induced changes in heart rate [35]. STATH is expressed in nasal secretions and saliva, but not in other body fluids. TNNC2 is a gene that encodes skeletal muscle fast troponin C, which is associated with muscle contraction. TNNC2 has also been found to be associated with intestinal barrier repair, facilitating improvement in functional recovery and prognosis after intestinal inflammation or injury [36]. Neurotrophic receptor tyrosine kinase 1 (NTRK1) regulates neuronal injury, and the NTRK1 gene encodes a high-affinity receptor for nerve growth factors, which regulate neural development and differentiation [37]. Chromosome 12 open reading frame 75 (C12orf75) has been associated with neuronal injury. Further, differential expression of C12orf75 is strongly associated with cancer progression [38]. The epigenetic regulator, TAF6L, can coordinate gene expression to control self-renewal and thus maintain mouse embryonic cell stemness [39]. Pathway enrichment analysis showed that the differential genes of all patient groups were commonly enriched in neuroactive ligand-receptor interaction and calcium signaling pathways, which are involved in the regulation of host mucosal immunity and inflammatory response in respiratory infections [40]. Notably, the TB group showed specific enrichment in the cytokine-cytokine receptor interaction pathway compared with the CAP group, which reflects the distinct host immune response patterns between TB and CAP infection. TB is a chronic granulomatous infection, while CAP is an acute bacterial infection, leading to different cytokine response profiles, which is consistent with previous immunological studies on TB [41]. Further research is needed to elucidate the precise mechanisms through which differentially expressed genes and pathways act in the process of pathogenic infection. This study has several limitations that need to be acknowledged. First, the healthy control data were downloaded from public databases, and the grouping was completely parallel to the sequencing batch, resulting in complete confounding of batch effects and group effects, which cannot be corrected by bioinformatic algorithms. We tried to minimize technical heterogeneity by using the same sequencing platform, data processing pipeline, and consistent sample processing methods, but the potential confounding effect of batch differences cannot be completely excluded. Second, the sampling design was asymmetric: TB patients only had BALF specimens, while CAP patients had paired upper and lower respiratory tract specimens, which limited the direct comparison of upper respiratory tract microbiome between TB and CAP patients. This was because our primary research aim was to compare the lower respiratory tract microbiome and host response between TB and CAP patients, and upper respiratory tract specimens from CAP patients were collected as a control to assess the potential contamination of BALF by upper respiratory tract microorganisms. Third, we did not perform a detailed analysis of Mycobacterium tuberculosis transcriptional activity, which will be the focus of our follow-up research. Fourth, we did not perform longitudinal follow-up of the enrolled patients, because TB patients need to be transferred to specialized tuberculosis hospitals for treatment after diagnosis, making it impossible to collect longitudinal specimens. Finally, the sample size of this study was relatively small, and the results need to be validated in larger cohorts. Conclusions In conclusion, this metatranscriptomic study revealed significant differences in respiratory microbial community composition, diversity, interaction patterns, and host gene expression landscapes between TB and CAP patients, as well as between different anatomical sites of the respiratory tract. We identified Mycobacterium and Rothia as marker microorganisms for TB, and a set of differential genes and immune-related pathways associated with TB infection. These findings provide new insights into the pathogenesis of TB and the host-microbe interactions in respiratory infections, and may contribute to the development of new strategies for the diagnosis and treatment of TB. Further studies with larger sample sizes and functional experiments are needed to validate our findings and elucidate the causal relationship between respiratory microbiota alterations and TB progression. Additional File 1.docx. Supplementary Figures. Supplementary Table 1. Sample Information. Supplementary Table 2. Target microbe correlations. Supplementary Table 3. Global network metrics. Supplementary Table 4. Local network metrics. Supplementary Table 5. different gene TB and CAP. Supplementary Table 6. different gene OP and NP. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Supplementary Material 5 Supplementary Material 6 Supplementary Material 7 Acknowledgements The authors thank Sanling Fan, Haiyue Zhang, Hailing Liu, Jing Yin, Min Shang, Yuchuan Luo, and others for their help with this work. Abbreviations Lower respiratory tract infection Tuberculosis Mycobacterium tuberculosis Community-acquired pneumonia Bronchioalveolar lavage fluid Linear discriminant analysis Linear discriminant analysis effect size Nasopharynx Oropharynx Kyoto Encyclopedia of Genes and Genomes Author contributions Material preparation, data collection, and analysis were performed by LL, MX, ML, JM, PX, RL. Funded and guided by JC, XW. The first draft of the manuscript was written by LL, the revision was by MX, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. Funding This work was supported by The National Natural Science Foundation of China (82370008, 31970174), National Key R&D Program of China (2023YFC260550). Data availability The datasets generated and/or analyzed during the current study are not publicly available due to the necessity of data for further experimental analysis but are available from the corresponding author on reasonable request. Declarations Ethical approval and consent to participate This study was approved by the Ethics Committee of Wuhan University Renmin Hospital (WDRY2024-K190). The procedures were conducted in accordance with the principles of the Declaration of Helsinki (1964, amended in 2008). Informed consent was obtained from all the participants. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Liangyu Li and Meng Xu contributed equally to this work. Contributor Information Jianjun Chen, Email: chenjj@wh.iov.cn. Xiaojun Wu, Email: wuxiaojunrmyy@126.com. References Associated Data Supplementary Materials Supplementary Material 1 Supplementary Material 2 Supplementary Material 3 Supplementary Material 4 Supplementary Material 5 Supplementary Material 6 Supplementary Material 7 Data Availability Statement The datasets generated and/or analyzed during the current study are not publicly available due to the necessity of data for further experimental analysis but are available from the corresponding author on reasonable request.