Ifnar gene variants influence gut microbial production of palmitoleic acid and host immune responses to tuberculosis

Both host genetics and the gut microbiome have important effects on human health, yet how host genetics regulates gut bacteria and further determines disease susceptibility remains unclear. Here, we find that the gut microbiome pattern of participants with active tuberculosis is characterized by a reduction of core species found across healthy individuals, particularly Akkermansia muciniphila. Oral treatment of A. muciniphila or A. muciniphila-mediated palmitoleic acid strongly inhibits tuberculosis infection through epigenetic inhibition of tumour necrosis factor in mice infected with Mycobacterium tuberculosis. We use three independent cohorts comprising 6,512 individuals and identify that the single-nucleotide polymorphism rs2257167 ‘G’ allele of type I interferon receptor 1 (encoded by IFNAR1 in humans) contributes to stronger type I interferon signalling, impaired colonization and abundance of A. muciniphila, reduced palmitoleic acid production, higher levels of tumour necrosis factor, and more severe tuberculosis disease in humans and transgenic mice. Thus, host genetics are critical in modulating the structure and functions of gut microbiome and gut microbial metabolites, which further determine disease susceptibility. Chen et al. show how host genetics influence the abundance of commensal Akkermansiamuciniphila in the gut, which in turn produce palmitoleic acid, thus contributing to shaping host immune responses to Mycobacterium tuberculosis infection.

represent key rationales leading to severe and recurrent TB infections, the excessive inflammatory response (for example, tumour necrosis factor (TNF)) may be a cause of exacerbated TB during M. tuberculosis infection [22][23][24][25][26] . Therefore, regulating immune homeostasis of the gut-lung axis may serve as a key approach for successful TB control. Although it was shown that gut microbiota aberration might correlate with TB pathogenesis [27][28][29] , it is still unclear whether and how host genetics influence the structure and function of the gut microbiome to maintain immune homeostasis of the gut-lung axis and therefore determine the variable TB pathogenicity and susceptibility in M. tuberculosis-infected individuals.
In this study, we focused on investigating the relationship among interindividual variations in host genetic factors, gut microbial abundance and inflammatory responses in influencing susceptibility and severity of TB. We identified a single-nucleotide polymorphism (SNP) in the type I interferon (IFN-I) receptor (IFNAR1) gene from 3,125 participants with active TB and 3,387 healthy controls (HCs) in three independent cohorts. Importantly, we found that the IFNAR1 rs2257167 G allele was linked to enhanced IFN-I signalling, impaired intestinal colonization of A. muciniphila, reduced production of the A. muciniphila-mediated metabolite palmitoleic Ifnar gene variants influence gut microbial production of palmitoleic acid and host immune responses to tuberculosis 1 and 2 helper T (T H 1 and T H 2) cell cytokines (TNF, IFN-γ, interleukin (IL)-10, IL-2, IL-4 and IL-6). These cytokines were selected for analyses based on their well-established importance in mediating TB susceptibility and severity [33][34][35][36][37][38] . Ex vivo production of TNF/IFN-γ was higher in samples from participants with TB than in those of HCs, and such increased production of TNF, IFN-γ and IL-10 was at least partially specific to M. tuberculosis as ex vivo restimulation using lysates of M. tuberculosis further enhanced their productions (Fig. 1h,i). Interestingly, the abundance of A. muciniphila was negatively correlated with TNF production (Fig. 1j). However, there was an erratic association between A. muciniphila abundance and IFN-γ (Fig. 1k) or IL-10 response (Fig. 1l), and also no significant correlation between B. vulgatus abundance and TNF, IFN-γ and IL-10 production (Extended Data Fig. 3). Collectively, these results indicated that a higher TNF response was linked to the decreased abundance of A. muciniphila but not B. vulgatus in M. tuberculosis infection, highlighting that A. muciniphila might function to regulate TB susceptibility or severity via regulation of TNF response.
We then investigated potential in vivo effects of A. muciniphila and B. vulgatus in M. tuberculosis-infected mice. An aerosol M. tuberculosis infection model 39 was developed, and antibiotic-treated mice exhibited more severe pathological impairment and much higher Bacillus burdens in lungs (Extended Data Fig. 4a-e), confirming that gut microbiota were necessary for mediating anti-TB protection. Moreover, M. tuberculosis-infected mice were orally treated with A. muciniphila or B. vulgatus, and it was mice gavaged with A. muciniphila but not B. vulgatus that showed much less haemorrhage and pathological impairment in lungs ( Fig. 2a-f) and exhibited lower pulmonary Bacillus burdens (Fig. 2g,h). Additionally, it showed that M. tuberculosis-infected mice with A. muciniphila but not B. vulgatus treatments exhibited significantly lower levels of TNF, but not IFN-γ, IL-2, IL-4, IL-6 and IL-10 ( Fig. 2i and Supplementary Fig. 1a). These therapeutic effects in M. tuberculosis-infected mice with A. muciniphila treatments were correlated with higher abundance of A. muciniphila ( Supplementary  Fig. 2). However, such anti-TB protection effects of A. muciniphila in M. tuberculosis-infected mice were partially compromised with TNF signalling blockade (Extended Data Fig. 5).
Furthermore, as it has been shown that excessive TNF could mediate the TB pathogenesis 22,25 , we next investigated whether and how A. muciniphila directly mediates TNF production during M. tuberculosis infection. We previously found that histone modification states at Tnf loci determined TNF production by T cells 40 . Therefore, we examined whether A. muciniphila had any effect on M. tuberculosis-specific TNF expression by mediating histone modification of the Tnf promoter. Immunoblot and chromatin immunoprecipitation followed by quantitative PCR (ChIP-qPCR) analysis showed A. muciniphila indeed induced less enrichment of monomethylated histone H3 Lys9 and trimethylated histone H3 Lys4 (denoted H3K9Me1 and H3K4Me3, respectively, hereafter) at Tnf promoter regions (Extended Data Fig. 6). Both H3K9Me1 and H3K4Me3 are histone modifications capable of promoting gene transcription 41 . Thus, these data collectively suggested that A. muciniphila may confer anti-TB protection via epigenetically modulating TNF signalling, and such protection effects require intact host TNF signalling.

Palmitoleic acid mediates the functions of A. muciniphila.
We then investigated what bacterial components of A. muciniphila mediated anti-TB effects. Metabolite signatures of bacterial cell components and culture supernatants of A. muciniphila were subjected to gas chromatography combined with time-of-flight mass spectrometry (GC-TOF-MS) and random forest analyses. They showed that A. muciniphila-derived 2-hydroxybutanoic acid, 3-hydroxybutyric acid and palmitoleic acid were potential dominant metabolites involved in mediating anti-TB immunity (Fig. 3a TB  HC  TB  HC  TB  HC  TB   HC  TB  HC  TB  HC  TB  HC  TB   HC  TB  HC  TB  HC  TB  HC    and Extended Data Fig. 7a). However, only palmitoleic acid but not 2-hydroxybutanoic acid and 3-hydroxybutyric acid could significantly reduce M. tuberculosis-specific in vitro TNF production in culture supernatants of mouse T cells ( Fig. 3c and Supplementary  Fig. 1b-d). While both palmitoleic acid (C16:1n-7) and palmitic acid (C16:0) are long-chain fatty acids, which share a similar structure, palmitoleic acid but not palmitic acid showed significant anti-inflammatory effects ( Supplementary Fig. 1e,f). Consistently, Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis revealed that A. muciniphila was less likely to express long-chain fatty acid CoA ligase than B. vulgatus (Extended Data Fig. 7b-f) and the expression levels of the fatty acyl-ACP thioesterase (Fat/FatA) gene were significantly increased in faeces of mice with A. muciniphila treatment (Extended Data Fig. 7g), which might together allow more production and accumulation of palmitoleic acid in A. muciniphila.   . Data are representative of three experiments with three independent biological replicates. c, Pooled bar graphic data showed the in vitro expression of IL-2, IL-4, IL-6, IFN-γ, TNF and IL-10 in culture supernatants of CD3 + T cells of wild-type mice. T cells isolated from spleens from wild-type mice were co-cultured with M. tuberculosis lysates (10 μg ml −1 ) only or with 2-hydroxybutanoic acid, 3-hydroxybutyric acid or palmitoleic acid for 3 d. N = 6 mice per group. Box-and-whisker plots show the centre line as the median and the box limits as the first and third quartiles. The whiskers represent the minimum and maximum of all points. d, Palmitoleic acid concentrations in plasma derived from mice with oral administration of A. muciniphila were higher than those in plasma of mice treated with B. vulgatus, saline or drinking water. e,f, The quantitative analyses of A. muciniphila abundance in stool samples (e) and palmitoleic acid concentrations in plasma and faecal samples (f) derived from mice with drinking water only, mice with drinking water plus oral administration of A. muciniphila and antibiotic-treated mice with oral administration of inactive A. muciniphila or A. muciniphila at the third day after bacteria gavage. g, Palmitoleic acid showed much higher concentrations in faecal samples from HCs than the TB group (HC = 28, TB = 26; Shenzhen cohort). h, Concentrations of palmitoleic acid in plasma of HCs were higher than those in plasma of the TB group (HC = 28, TB = 26; Shenzhen cohort). N = 6 mice per group. Data are representative at least two biological replicates. Data are presented as the mean ± s.d. P values were calculated by Mann-Whitney test (two-tailed; a) or Student's two-tailed unpaired t-test (c-h).
Furthermore, oral administrations of A. muciniphila, but not B. vulgatus, significantly increased palmitoleic acid concentrations in plasma of mice (Fig. 3d). Oral administrations of live but not inactivated A. muciniphila significantly increased palmitoleic acid concentrations in both plasma and faeces of mice (Fig. 3e,f). Indeed, we found that M. tuberculosis-infected mice gavaged with live but not inactivated A. muciniphila showed much less haemorrhage and pathological impairment in lungs, and exhibited lower pulmonary Bacillus burdens (Extended Data Fig. 8). Moreover, KEGG analysis showed that other gut bacteria, which displayed significantly differential abundance between the HC and TB samples, could not produce or accumulate palmitoleic acid as like A. muciniphila (Extended Data Fig. 7h-j). These data collectively indicate that intestine-colonized A. muciniphila are an important contributor to higher palmitoleic acid levels.
Importantly, the concentrations of palmitoleic acid were significantly higher in both faeces and plasma from HCs than those from individuals with TB in our human cohorts (Fig. 3g,h), indicating that decreased palmitoleic acid concentration correlated with active TB in humans. Indeed, M. tuberculosis-infected mice with dietary treatments of palmitoleic acid but not butyrate showed much less pulmonary unresolved haemorrhage and pathological impairment ( Fig. 4a-d) and exhibited lower pulmonary Bacillus burdens (Fig. 4e,f). It should be noted that such therapeutic effects in palmitoleic acid-treated mice were directly linked to higher levels of palmitoleic acid after oral treatments of palmitoleic acid ( Supplementary Fig. 3). Consistently, there was no significant difference in butyrate concentrations in faeces between HCs and individuals with TB ( Fig. 3g), suggesting that palmitoleic acid might play a more important role than butyrate in mediating TB protection or susceptibility.
Additionally, M. tuberculosis-infected mice treated with dietary palmitoleic acid showed significantly lower expression of TNF but not IFN-γ, IL-2, IL-4, IL-6 and IL-10 in serum compared to those receiving butyrate or water treatment ( Fig. 4g and Supplementary  Fig. 1g). Importantly, palmitoleic acid also significantly decreased M. tuberculosis-specific TNF production in culture supernatants of mouse T cells, and reduced in vivo expression of intracellular TNF in CD4 + /CD8 + T cells ( Supplementary Fig. 4a-e), and epigenetically reduced enrichment of H3K4Me3 at the Tnf promoter during M. tuberculosis infection (Supplementary Fig. 4f-h). These findings collectively indicate that A. muciniphila used the metabolite palmitoleic acid to induce less euchromatin at the Tnf promoter and inhibit M. tuberculosis-specific TNF expression. These results indicate that such a palmitoleic acid-mediated inhibition effect against pro-inflammatory TNF may be one of key mechanisms for A. muciniphila-conferred anti-TB protective immunity.
A. muciniphila colonization is influenced by type I interferon signalling. IFN-I signalling plays critical roles in determining TB pathogenicity, intestinal epithelium development and homeostasis of the gut microbial ecosystem [42][43][44][45][46][47] . Thus, we hypothesized that IFN-I signalling affected the abundance of A. muciniphila and, in turn, resulted in varied TB susceptibility and severity. We fed wild-type and Ifnar1 −/− mice with HC and TB faecal samples, respectively (Extended Data Fig. 9a). Importantly, we found that Ifnar1 −/− mice harboured approximately 2-6 times more abundance of anaerobic bacteria in the jejunum or the caecum compared to wild-type mice (Extended Data Fig. 9b,c). Interestingly, regardless of whether faeces came from HCs or from TB samples, A. muciniphila exhibited much higher levels of colonization and abundance in the jejunum, ileum and caecum of Ifnar1 −/− mice compared to wild-type controls (Extended Data Fig. 9d,e). Together, these results suggest that deletion or reduction of IFN-I signalling favoured intestinal colonization and abundance of A. muciniphila. Ifnar1 rs2257167 G allele dictates stronger type I interferon signalling. We further postulated that there might be host genetic variants present in Ifnar1 that altered the strength of IFN-I signalling and resulted in interindividual variations in the abundance and immunoregulatory function of A. muciniphila and A. muciniphila-mediated metabolites during M. tuberculosis infection. Targeted sequencing was performed to analyse the human IFNAR1 gene in a discovery cohort in Guangzhou, China (Supplementary  Table 2). Nine SNPs in the intron and only one SNP, rs2257167 (which would result in a valine-to-leucine substitution at codon 168 in human IFNAR1 (termed as hIFNAR1-p.Val168VLeu; Supplementary Fig. 5), in the exon, were identified (Supplementary Table 3). Because Ifnar1 rs2257167 GG and GC as well as rs1041868 GG and GA genotypes showed the most significant trend towards higher risk of developing active TB in the Guangzhou cohort (Supplementary Table 3), two larger independent cohorts, including the Shenzhen cohort (HC = 1,445; TB = 1,533) 48 and Foshan cohort (HC = 1,679; TB = 1,328) in China were recruited to further investigate whether rs2257167 G or rs1041868 G was indeed a risk allele of developing active TB or more severe TB (Supplementary Table 2). Indeed, further sequencing analysis 49 validated that individuals carrying the rs2257167 G allele exhibited a significantly increased risk of developing active TB in both Shenzhen and Foshan cohorts (Supplementary Table 4).
Notably, we also discovered that individuals carrying rs2257167 genotypes GG and GC displayed higher expression levels of phosphorylated (p)-STAT1, IFNB1 and ISG15 than those carrying genotype CC from our human cohorts ( Fig. 5a and Supplementary  Fig. 6a-c), suggesting that the G allele enhanced IFN-I signalling.
To confirm this, we developed two transgenic mouse lines, denoted as IFNAR1-p.Val168Val (harbouring the G allele) and IFNAR1-p. Val168Leu (harbouring the C allele), respectively ( Supplementary  Fig. 7a). There was no significant difference in gene copies or expression levels of hIfnar1 between IFNAR1-p.Val168Val and IFNAR1-p.Val168Leu mice ( Supplementary Fig. 7b,c). However, the expression of p-STAT1, IFNB1 and ISG15 in bone marrow-derived macrophages (BMDMs) were higher in IFNAR1-p.Val168Val mice than in IFNAR1-p.Val168Leu mice ( Fig. 5b and Supplementary  Fig. 6d,e). Furthermore, the levels of p-STAT1 in intestinal epitheliums as well as IFN-β protein in serum were higher in IFNAR1-p. Val168Val mice than in IFNAR1-p.Val168Leu mice, but these differences weren't observed in the lungs during M. tuberculosis infection ( Supplementary Fig. 6f,g). Additionally, IFNAR1-p.Val168Val mice showed higher expression of IFNB1 and ISG15 in BMDMs and intestinal epitheliums, but not in the lungs, compared to IFNAR1-p. Val168Leu mice ( Supplementary Fig. 6h-l). Collectively, these results suggest that the IFNAR1/Ifnar1 rs2257167 G allele enhanced IFN-I signalling in humans and mouse models.

Ifnar1 rs2257167 G allele impairs A. muciniphila colonization.
To explore whether Ifnar1 rs2257167 G allele affected the intestinal colonization and abundance of A. muciniphila as well as production of A. muciniphila-mediated palmitoleic acid, IFNAR1-p.Val168Leu and IFNAR1-p.Val168Val mice with antibiotics pretreatment were orally given with the same amounts of A. muciniphila (Fig. 5c). IFNAR1-p.Val168Val mice indeed harboured less A. muciniphila in intestinal epithelia, compared to their IFNAR1-p.Val168Leu counterparts (Fig. 5d-g and Supplementary Fig. 8). Importantly, IFNAR1-p.Val168Val mice displayed lower concentration levels of palmitoleic acid in plasma than IFNAR1-p.Val168Leu mice did after A. muciniphila treatment (Fig. 5h). These findings collectively reveal that Ifnar1 rs2257167 G allele dictating stronger IFN-I signalling could impede the intestinal abundance of A. muciniphila and reduce palmitoleic acid levels.
Furthermore, we analysed whether increased A. muciniphila abundance could ameliorate the Ifnar1 rs2257167 G allele-induced TB severity. Interestingly, A. muciniphila treatment showed protective effects against M. tuberculosis infection (Fig. 6a-e and Supplementary Fig. 10a) and significantly lower levels of TNF in IFNAR1-p.Val168Val mice ( Fig. 6f and Supplementary Fig. 10b). Such anti-TB protection in IFNAR1-p.Val168Val mice with A. muciniphila treatments was linked to higher A. muciniphila abundance ( Supplementary Fig. 10c,d). Also, dietary palmitoleic acid treatments increased palmitoleic acid levels in both plasma and faecal samples ( Supplementary Fig. 10e,f), alleviated pulmonary TB pathology and reduced Bacillus burdens and TNF production in M. tuberculosis-infected IFNAR1-V168V mice (Fig. 6g-k and Supplementary Fig. 10g). These results suggest that oral administration of A. muciniphila or palmitoleic acid could ameliorate Ifnar1 rs2257167 G allele-induced TB severity and confer anti-TB protection in a genotype-specific manner.   -test (f, h, j, m, n and p).

Effects of IFNAR1 gene variants are observed in participants with tuberculosis.
We then investigated whether IFNAR1 rs2257167 G allele could affect A. muciniphila abundance in active human TB. Participants with TB carrying genotypes GC and GG possessed a significantly lower abundance of A. muciniphila than participants carrying genotype CC (Fig. 7a,b). Importantly, the relative concentrations of palmitoleic acid were significantly lower in both faeces and plasma from participants with TB carrying Ifnar1 rs2257167GC and GG than carriers of the CC genotype (Fig. 7c,d).
Additionally, compared to participants with TB carrying Ifnar1 rs2257167 CC, carriers of genotypes GC and GG showed much higher (approximately five-to six-fold) ex vivo expression levels of M. tuberculosis-specific TNF (Fig. 7e,f and Extended Data Fig. 10). Furthermore, high-resolution computed tomography (HRCT) images were assessed 24,50,51 to analyse the association of Ifnar1 SNP rs2257167 with clinical features of active TB. Participants with TB carrying the Ifnar1 rs2257167 G allele showed significantly higher HRCT scores (suggesting much more severe disease; Fig. 7g) and less favourable disease outcomes following a 2HRZE/4HR treatment regimen recommended by the World Health Organization ( Fig. 7h) than participants carrying genotype CC. Collectively, these data indicate that IFNAR1 rs2257167 G allele is associated with reduced abundance of A. muciniphila and hyper-reactive TNF, which lead to an increased risk of developing active TB and more severe TB disease in humans.

Discussion
The exact contributions and underlying mechanisms of host genetics, environmental factors and diet to the structure and function of gut microbiota are still unclear despite recent research advan ces [3][4][5][6][7]11,13,15 . The precise mechanisms of TB pathogenesis also remain elusive as TB continues to pose as a leading global public health threat 16 . While this study and other groups have shown that the disruption of the microbiota contributes to increased susceptibility to TB 29,52 , faecal transplantation 29 or oral administration of gut bacteria such as Lactobacillus plantarum 53 has shown the possibility of restoring anti-TB immunity or preventing extrapulmonary dissemination of pulmonary TB. However, how human genetics modulates compositional shifts of the gut microbiome and gut microbial metabolites and elicits varying degrees of TB pathogenicity and severity in different individuals is still unclear. This work has demonstrated that host genetic factors modulated IFN-I signalling for better colonization and composition of protective gut bacteria, increased levels of gut microbial metabolites, and a more successful control of M. tuberculosis infection and TB pathology. Even though long-term or short-term previous medications 54 or stool consistency 55 might also directly or indirectly contribute to altered gut microbiota structures and host immune homeostasis, this work implicated that host genetics might exert an important impact on the colonization and abundance of gut microbiota, which further affect the intestinal or systemic levels of gut microbial metabolites, in some biological settings. Thus, further studies may be useful to uncover how medications including antibiotics may impact gut microbiota and how gut microbiota may facilitate anti-TB treatments in a genotype-specific manner.
While IFN-I signalling is correlated with highly heterogeneous outcomes of TB in different models, with either a protective or a pathogenic role [43][44][45][56][57][58][59][60] , the exact mechanisms of how IFNAR1 (that is, IFN-I) signalling modulates the intestinal colonization of A. muciniphila require further studies. As IFN-I signalling in the gut is the primary innate defence mechanism to directly or indirectly facilitate the control of gut pathogens 46 , IFN-I signalling driven by a protective Ifnar1 allele may help establish a favourable gene network or ecosystem for colonization of selected gut bacteria, which may lead to increased growth, colonization and abundance of A. muciniphila. Moreover, because gut bacteria might utilize soluble IgA antibody for colonization in the intestinal mucosa 10 , the gut IFN-I signalling could affect other immune mediators like IgA antibody and further influence the intestinal colonization of A. muciniphila. It is possible for these participants carrying the germline IFNAR1 rs2257167 G allele to be clinically distinguished by a form of more severe primary progressive TB, and also presumably much easier to develop secondary/recurrent TB or poorer prognosis. It will also be interesting for further studies to investigate whether there are more TB patients carrying particular germline genotypes in areas with high TB prevalence.
TNF is essential to activate the innate immunity programme for killing intracellular M. tuberculosis and is required for TB control 61,62 . Indeed, anti-TNF therapies are associated with an increased risk of developing active TB 61,63 . However, hyper-reactive TNF inflammation may also be detrimental 22,24 and the delicate modulation of moderate TNF inflammation levels was, therefore, one of the central ideas for avoiding development of active TB disease. While highly excessive or inappropriate TNF production was induced by host genetics-dictated lower abundance of A. muciniphila in the intestinal compartment in severe TB disease, colonization dominance of A. muciniphila might help prevent aberration of the microbial ecosystem and avoid impairment of homeostasis of immune networks.
In addition, this work demonstrates that host genetics can determine the structure and function of gut bacteria and impact the concentration levels and immunoregulatory functions of gut microbial metabolites. We have shown that A. muciniphila used the metabolite palmitoleic acid to regulate systemic M. tuberculosis-specific TNF expression. Moreover, it has been shown that palmitoleic acid could reduce inflammatory response in other disease models, such as nonalcoholic fatty liver disease [64][65][66] . It would also be interesting to explore whether and what metabolites derived from gut bacteria may migrate into the pulmonary compartment and regulate local inflammation (for example, local TNF/IL-1β response in lung granulomas 24 ) during M. tuberculosis infection. It is worth exploring whether manipulating other host genetic factors and/or other gut bacteria or metabolic pathways capable of producing palmitoleic acid could also confer similar or better anti-TB protection effects. Notably, although butyrate appeared to inhibit TNF expression in some disease models or biological settings 67-69 , we did not observe a notable inhibitory effect on TNF expression and anti-TB protective effects of butyrate in this work.
In summary, this study not only demonstrates the previously unrecognized importance of a host genetics-gut microbiota axis in the regulation of TB susceptibility and severity but also provides a new paradigm for deciphering the role of gut microbiota in human health and disease as well as for developing genotype-dictated precision therapeutics of TB, particularly multidrug-resistant TB, targeting gut microbiota and gut microbial metabolites.

Study participants. All experiments were approved by the internal review and the ethics boards of Zhongshan School of Medicine of Sun Yat-sen University.
This project aimed to investigate the relationship among interindividual variations in host genetic factors, gut microbial abundance and inflammatory responses in influencing susceptibility and severity of TB. This study used two independent Chinese Han population cohorts for 16S rDNA sequencing analysis: the Shenzhen cohort, consisting of 28 HCs (age range, 23.46-54.65 years) and 26 individuals with TB (age range, 23.80-47.96 years); the Foshan cohort, consisting of 17 HCs (age range, 23.36-43.84 years) and 19 individuals with TB (age range, 21.85-48.78 years). On entry to the study, participants were screened using a standardized questionnaire by a general practitioner and demographic information was recorded. Individuals were excluded if they had constipation or diarrhoea, or received probiotic treatment 3 months before their recruitment into the study. Early-morning stool samples were obtained from participants with newly diagnosed pulmonary TB as soon as possible before initiation of anti-TB therapies. Fresh stool samples (morning samples) were obtained from HCs on the day of their hospital examination. Detailed information on the distributions of age, sex, height, weight and body mass index for each cohort sample is provided in Supplementary Table 1 and Supplementary Fig. 11a.
For SNP analysis, we used three independent Chinese Han population cohorts: the Guangzhou cohort as a discovery cohort and the Shenzhen cohort and Foshan cohort as replication cohorts. The Guangzhou cohort comprised 263 HCs (age range, 20.78-45.18 years) and 264 individuals with TB (age range, 19.91-55.51 years). The replication cohorts comprised 2,978 individuals from the Shenzhen cohort (1,445 HCs, age range 17.00-56.75 years; 1,533 individuals with TB, age range 21.38-52.69 years) and 3,007 individuals from the Foshan cohort (1,679 HCs, age range 19.90-46.70 years; 1,328 individuals with TB, age range 21.21-49.08 years). These individuals are not related, neither from the same village or the same family. Detailed information on the distributions of age, sex, height, weight and body mass index for all cohort samples is provided in Supplementary Table 2 and Supplementary Fig. 11b.
All the participants were prospectively recruited on the basis of TB symptomatic features. The diagnosis of TB was based on clinical signs and symptoms, chest radiography, acid-fast bacilli identification (sputum smear, M. tuberculosis culture or nucleic acid amplification techniques assay) and response to anti-TB treatment. Participants with TB who had cancer, diabetes, hypertension or HIV infection were excluded. HCs were defined as individuals with no clinical history of TB, normal chest radiograph findings, a negative interferon gamma release assay test result to confirm no latent TB infection, no other infectious diseases and major conditions like cancers, diabetes and hypertension. The study was designed to recruit the HCs in the same regions at the same time. No participant in the cohorts dropped out during the study. All participants provided written informed consent.
Isolation of peripheral blood mononuclear cells. PBMCs were isolated from freshly collected heparin lithium blood by Ficoll-Paque Plus density gradient centrifugation. Briefly, Ficoll-Paque Plus was loaded into the blood and centrifuged at 1,000g for 20 min at 20 °C. The isolated monocytes were washed with pH 7.4 PBS (Gibco), and finally resuspended in 10% fetal bovine serum/RPMI-1640 (Gibco) medium and cultured for further research.
Cytometric bead array analysis. Culture supernatants of PBMCs derived from HCs and participants with TB, or serum isolated from the M. tuberculosis-infected mice were analysed for the following cytokines: IL-2, IL-4, IL-6, IL-10, TNF, IFN-γ or IL-17A using the BD CBA T H 1/T H 2 Cytokine Kit or T H 1/T H 2/T H 17 Cytokine Kit (purchased from BD) as previously described 40 . Data were collected on the Beckman CytoFLEX S and analysed as previously described 40 .

DNA extraction, 16S rDNA gene amplification and pyrosequencing (human faecal samples).
Fresh faecal samples were collected from participants with TB and HCs. DNA was isolated with the DNA Stools Kit (Qiagen) by following the manufacturer's instructions. Agarose gel electrophoresis detection was performed for the purity and concentration of DNA in faecal samples from the Shenzhen cohort. The corresponding primers were used to amplify DNA: 341F and 806R for V3 + V4 regions, 515F and 907R for V4 + V5 regions, 515F and 806R for V4 regions. Amplified DNA fragments were sequenced by paired-end, 250-bp reads using the Illumina NovaSeq platform. For DNA of faecal samples derived from the Foshan cohort, the corresponding primers were used to amplify DNA: 341F-805R for V3 + V4 regions, 338F-533R for the V3 Regions, and 967F-1046R for the V6 regions. The index sequences were added and enriched after the extraction was complete. The Qubit 2.0, Agilent 2100 and Bio-Rad CFX 96 were used to quantify the concentrations and purity of the library to ensure DNA quality. Afterward, the library was sequenced on an Illumina HiSeq 2500 using the 250-bp paired-end protocol.
Antibiotic treatment. A mixture of ampicillin (1 mg ml −1 ), streptomycin (5 mg ml −1 ) and colistin (1 mg ml −1 ; Sigma-Aldrich) was added to sterile drinking water for mice. Solutions and bottles were changed three times per week. Antibiotic activity was confirmed by cultivating faecal pellets resuspended in brain heart infusion (BHI) + 15% glycerol at 0.1 g ml −1 and cultured on blood agar plates for 48 h at 37 °C in aerobic and anaerobic conditions weekly.
Gut colonization with commensal species. A. muciniphila (ATCC BAA-835) and B. vulgatus (ATCC-8482) were purchased from the American Culture Type Collection (ATCC) and grown on Columbia Agar plates with 5% Sheep Blood in an anaerobic atmosphere at 37 °C for at least 72 h and 48 h, respectively. Bacteria were collected from the agar plates and suspended in sterile saline with 10% glycerol to obtain suspensions of 10 9 CFUs per ml at an optical density of 600 nm. Our pilot studies of dosage ascendance suggested that oral gavage with higher amounts (within the ranges of optimized dosage of A. muciniphila) of A. muciniphila conferred better anti-TB protection effects in wild-type mice. Herein, gut colonization of antibiotic-treated mice was performed by oral gavage with 200 µl suspension containing 2 × 10 8 bacteria three times per week.
Dietary treatment of metabolites. Palmitoleic acid (Sigma) or butyrate (Sigma) was added to drinking water containing sodium hydroxide at a final concentration of 36 mM 70 . Water with same concentration of sodium hydroxide was used as control, and 100 mg of palmitoleic acid was dissolved in a 0.1 M sodium hydroxide aqueous solution.
Mice and infection. All animal experimental procedures were approved by the ethics boards of Zhongshan School of Medicine of Sun Yat-sen University (protocol no. 2016-078) according to the national legislation for animal care. Specific pathogen-free C57BL/6J mice were purchased from Sun Yat-sen University Laboratory Animal Center. Transgenic mice were constructed using microinjection technology for DNA fragments consisting of the human IFNAR1 gene G or C allele to randomly integrate into the mouse's intestinal epithelium genome in the C57BL/6J background (Supplementary Fig. 7a). Female and male C57BL/6J mice aged approximately 6 weeks were used for the study. The mice were housed under specific pathogen-free conditions and reared in line with standardized methods at 22 ± 1 °C on a 12-h light/dark cycle with free access to food and water.
For M. tuberculosis infection, each mouse was infected through aerosols with approximately 150 CFUs M. tuberculosis (H37Rv) for 5 weeks at Biosafety Level-3 Laboratory of Sun Yat-sen University. No mouse was excluded from the analyses.

Bacterial and histopathological analysis of M. tuberculosis-infected mice.
To determine M. tuberculosis burden, lungs of mice were homogenized carefully for M. tuberculosis CFU counting analysis as previously described 40 . For tissue histopathological analysis, lung tissues of mice were fixed in 10% zinc formalin and embedded in paraffin. Sections (5-μm thick) were stained with H&E and images were obtained using the Digital Slide Scanning System AxioScan Z1. Images of acid-fast staining were obtained under a microscope (Olympus BX51). An overall histology score was assigned to the lungs of mice based on the degree of granulomatous inflammation as follows: 0, no lesion; 1, minimal lesion (1-10% area of tissue in the section involved); 2, mild lesion (11-30% area involved); 3, moderate lesion (30-50% area involved); 4, marked lesion (50-80% area involved); 5, severe lesion (>80% area involved).
High-resolution computed tomography examination and radiological scoring. HRCT was performed at 10-mm-section intervals (120 kV, 50-450 mAs, 1-mm-slice thickness, 1.5-s scanning time) with a window level between 2,550 and 40 Hounsfield Units and window width between 300 and 1,600 Hounsfield Units using the Toshiba Aquilion 64 CT Scanner (Toshiba). HRCT scans were analysed by two independent chest radiologists and conclusions on the findings were reached by consensus. Radio-pathological changes were quantified using a scoring system developed by Ors et al. The arbitrary scores were based on the percentage of lung parenchyma abnormality as we and others have previously described 50,51 .
DNA extraction and single-nucleotide polymorphism genotyping. Fresh blood samples were collected from participants with TB and HCs. Genomic DNA was isolated using the DNA Blood Kit (Qiagen) following the manufacturer's instructions. In the TB discovery study, we used targeted sequencing of the 11 exons in human IFNAR1 and found the SNP rs2257167 in the fourth exon in HCs and participants with TB. For validation cohorts, rs2257167 was genotyped in HCs and in participants with TB using TaqMan assays (ABI).
Western blots and chemical reagents. Cells were isolated from indicated individuals or mice and were treated differently based on specific experimental settings. Cells were co-cultured with Sendai virus in some experiments. Total proteins were extracted from cells using RIPA lysis buffer and quantified using BCA Protein Assay Kit (Thermo Fisher Scientific). Extracted proteins mixed with loading buffer were electrophoresed through 10% SDS-PAGE and then transferred to PVDF membranes (Bio-Rad), which were then blocked in 5% fat-free milk for 1 h. The membranes were incubated with primary antibodies at 4 °C overnight, and further incubated with a secondary antibody (Thermo Fisher) for 1 h at room temperature. Signals were detected using Immobilon Western Chemiluminescent HRP Substrate (Millipore). Images were captured by the Bio-Rad ChemiDoc Touch machine and analysed by ImageJ 1.43 software.

Detection of anaerobic bacteria and A. muciniphila.
Samples from the jejunum, caecum, ileum and faeces of each mouse were homogenized in PBS. Tenfold serial dilutions were made using PBS and plated on Columbia Agar plates with 5% Sheep Blood, and incubated at 37 °C in anaerobic conditions for 48 h or at least 72 h based on experimental settings. Genomic DNA was isolated from tissues or faecal samples using the QIAamp DNA Tissue or Stools Mini Kit (Qiagen) following the manufacturer's instructions. Targeted qPCR systems were applied using either TaqMan or SYBR Green technology. The primers and probes were used as previously described 71 . Bacterial cell ingredients, bacterial culture supernatant and plasma metabolites extraction. Samples (100 μl) were added into 2-ml microcentrifuge tubes. Then, Faecal metabolite extraction. Samples (100 ± 1 mg) were added into 2-ml microcentrifuge tubes. Then, 0.48 ml of extraction liquid (methanol: chloroform in a 3:1 ratio) containing 20 μl of l-2-chlorophenylalanine (1 mg ml −1 stock in dH 2 O) was added to each sample as internal standard followed by vortex mixing for 30 s. Samples were homogenized in a ball mill for 4 min at 45 Hz, then treated with ultrasound for 5 min (incubated in ice water) five times and centrifuged for 15 min at 4 °C and 10,000g. The supernatant (0.4 ml) was transferred into a fresh 2-ml GC/MS glass vial and then speed-vacuumed to dry. Then, 40 μl methoxyamine hydrochloride (20 mg ml −1 in pyridine) was added to each sample and incubated for 30 min at 80 °C. Next, 60 μl of the BSTFA reagent (1% TMCS, vol/vol) was added to the sample aliquots and incubated for 1.5 h at 70 °C. All samples were analysed using a GS system coupled with a Pegasus high-throughput TOF-MS.

GC-TOF-MS analysis.
All GC-TOF-MS analyses were performed using an Agilent 7890 GS system coupled with a Pegasus high-throughput TOF-MS. The system utilized a DB-5MS capillary column coated with 5% diphenyl cross-linked with 95% dimethylpolysiloxane (30 m × 250 μm inner diameter, 0.25-μm film thickness; J & W Scientific). A 1-μl aliquot of the analyte was injected in splitless mode. Helium was used as the carrier gas; the front inlet purge flow was set at 3 ml min −1 , and the gas flow rate through the column was 1 ml min −1 . The initial temperature was kept at 50 °C for 1 min, and then raised to 310 °C at a rate of 10 °C per min, and kept at 310 °C for 8 min. The injection, transfer line and ion source temperatures were 280 °C, 280 °C and 250°C, respectively. The energy was −70 eV in electron impact mode. The mass spectrometry data were acquired in full-scan mode with a m/z range of 50-500 at a rate of 20 spectra per second after a solvent delay of 6.27 min.

Analysis of GC-TOF-MS.
Data were collected using the Chroma TOF 4.3X software (LECO Corporation) and the LECO-Fiehn Rtx5 database was used for raw peak exacting, data baseline filtering and calibration of the baseline, peak alignment, deconvolution analysis, peak identification and integration of the peak area 72 . Both mass spectrum match and retention index match were considered in metabolite identification.

RNA extraction and real-time PCR.
Total RNA was extracted from tissues or PBMCs using TRIzol reagent (Takara) and total RNA concentration was determined by a Nanodrop spectrometer. First-strand cDNA was synthesized using the PrimeScript One Step Strand cDNA Synthesis Kit (Takara) following the manufacturer's instructions; qPCR was performed in technical triplicates using SYBR Green to determine the expression levels of IFNB1 and ISG15. GADPH was used as an endogenous control to normalize gene expression (Supplementary Table 5). Relative mRNA expression levels were presented as means ± s.d. Statistical differences were analysed by Student's t-test.
A. muciniphila colonization study using transmission electron microscopy. The jejunum, cecum and ileum from mono-colonized mice were trimmed into 2-to 3-mm cube sizes and immediately fixed with an ice-cold solution of 3% glutaraldehyde in 0.1 M sodium cacodylate buffer at 4 °C overnight. Samples were then post-fixed in 2% osmium tetroxide buffered solution and were embedded in epoxy resin. Subsequently, samples were processed as previously described 10 . Micrographs were produced using an FEI Tecnai G2 Spirit BioTwin 634 transmission electron microscope.
Fluorescence in situ hybridization and immunofluorescence of A. muciniphila in caecum. Paraformaldehyde-fixed, paraffin-embedded colon tissue sections (5 µm) were deparaffinized. A specific fluorescein-labelled oligonucleotide probe 71 targeting one region of the 16S rDNA gene of A. muciniphila was used to detect bacterial colonization. Monitoring of non-specific hybridizations was performed using the probe Non-EUB 73 as a negative control. These probes were hybridized to the tissue overnight at 50 °C. Then, slides mounted with ProLong Gold with DAPI (Invitrogen) were sealed with coverslips, left to dry at 4 °C in the dark overnight, and imaged using a confocal microscope (LSM 880 with Airyscan).
Statistical analysis. Statistical analysis was performed using GraphPad Prism software. For SNPs and TB susceptibility studies: (1) the Hardy-Weinberg Equilibrium analysis for IFNAR1 SNP distribution was used in participants with TB and HCs; (2) the Pearson χ 2 test was used to compare the genotypic and allelic frequencies of SNPs between participants with TB and controls; (3) unconditional logistic regression adjusted by gender and age was used to calculate the odd ratios, 95% confidence intervals and corresponding P values using four alternative models (multiplicative, additive, dominant and recessive). For statistical analysis of other experiments, normal distributions of data were determined with the D' Agostino-Pearson omnibus test, and statistical significance was determined using one-way ANOVA with Tukey's multiple-comparison test, Student's two-tailed unpaired t-test, Mann-Whitney U test or PERMANOVA test. P < 0.05 was considered as statistically significant. Odds ratios and P values for SNP analyses were determined with SPSS.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Data of 16S rDNA sequencing are available in a public repository at https:// dataview.ncbi.nlm.nih.gov/. The accession number of 16S rDNA sequencing data is PRJNA609532. Figure 1, Extended Data Figs. 1 and 2 have associated raw data. Source data are provided with this paper. All other data supporting the findings of this study are available from the corresponding authors on reasonable request.

Code availability
Code is available at https://github.com/ZhenhuangGe/Akkermansia-muciniphila/. . (c) A. muciniphila was belonged to top 10 bacterial species in fecal microbiota of HC. Species with enriched relative abundance in HC are adjusted P < 0.05 and log2 (HC/TB) > 0, species with enriched relative abundance in TB are adjusted P < 0.05 and log2 (HC/TB) < 0. The red-boxed area marks the A. muciniphila. (d) The absolute abundance of A. muciniphila in the fecal microbiota from HC (n = 17) and TB (n = 19). (e)Predictive power of top 10 species enriched in HC (that is the top 10 most reduced species in TB) assessed by random forest analysis. Blue boxplots correspond to minimal, average, and maximum Z-score of shadow species, which were shuffled version of real species introduced to random forest classifier and act as benchmarks to detect truly predictive species. Red boxplots represent rejected species, yellow boxplots represent suggestive species, and green boxplots represent confirmed species. The red arrowhead marks the A. muciniphila. HC = 17, TB = 19. (f) Histogram of the Linear discriminant analysis (LDA) coupled with effect size measurements (LEfSe) identified the species with different abundance in HC and TB. Higher abundant species in TB are shaded in green, higher abundant species in HC are shaded in red. Red arrowhead pointed A. muciniphila, black arrowhead pointed B. vulgatus. (g) Circle charts showed the relative abundance of six bacteria with differentiated relative abundance between HC and TB in Foshan cohort, and these six bacteria were also observed at both Shenzhen and Foshan cohorts. Histogram showed the fold change of relative abundance of six bacteria, including B. vulgatus, B. uniformis, A. muciniphila, B. caccae, P. merdae and E. ramosun between HC and TB (calculated as HC/TB) in Foshan cohort. Bacteria are identified by color bars above the chart. Data are presented as mean + /-SD. Pvalue was calculated by PERMANOVA test (a), Mann-Whitney test [(c)and (d)].