Akkermansia muciniphila protects the host from SFTSV infection by attenuating systemic inflammation
Faecal samples were taken from 260 patients hospitalized with SFTS (surviving patients were designated as the SF-S group, and deceased patients were designated as the SF-D group; Tables S1 and S2), 176 non-SFTSV febrile patients (non-SF) and 19 healthy controls (HC) via 16S ribosomal RNA-sequencing (RNA-seq). We observed a significantly greater abundance of Akkermansia, Lactobacillus, Enterococcus and Parabacteroides in the SF-S group compared with the HC group (Fig. 1a), whereas only Akkermansia was markedly reduced in the SF-D group compared with the SF-S group (Fig. 1b and Extended Data Fig. 1a,b). Moreover, the abundance of Akkermansia increased with disease course in the SF-S group but remained constant in the non-SF group, implying that this phenotype was specific to SFTSV infection (Fig. 1c). Systemic SFTSV burden was higher in the SF-D group than in the SF-S group but was not correlated with Akkermansia abundance in SFTS patients (Fig. 1d). By contrast, compared with SF-D patients, SF-S patients showed markedly lower expression of the proinflammatory cytokines interleukin-1β (IL-1β), interleukin-6 (IL-6) and tumour necrosis factor-α (TNF-α) in serum, the concentrations of which were inversely correlated with the relative abundance of Akkermansia (Fig. 1e,f and Extended Data Fig. 1c).
Subsequently, we prepared faecal samples pooled from four recovered SFTSV-infected patients (with a high abundance of Akkermansia) or from three patients who died of SFTSV infection (with a low abundance of Akkermansia) and performed faecal microbiota transplantation (FMT) in microbiota-depleted mice that were pretreated orally with a cocktail of four broad-spectrum antibiotics (hereafter referred to as the Abx mice)20. As expected, FMT from SFTSV-infected patients into Abx-treated mice significantly restored the total faecal 16S rRNA copy number at day 0; however, it did not restore copy number to the same level as the PBS control (Extended Data Fig. 2a), thus implying less efficient colonization of human microbiota in the mouse gut. Moreover, the FMT experiments were not influenced by SFTSV faecal shedding because no viral genomic copy number was detectable in the stool of infected patients.
A lethal infection mouse model pretreated with an anti-interferon alpha receptor 1 (IFNAR1)-blocking antibody, was used to evaluate the pathogenesis of SFTSV infection throughout the entire study21. Remarkably, FMT from recovered patients (FMT-S group), but not from succumbed donors (FMT-D group), administered to Abx mice significantly increased the survival rate from 23% to 54% by the end of the observation period (Fig. 1g). Moreover, FMT-S mice exhibited greatly diminished proinflammatory cytokine expression in the spleens, and significantly improved pathological changes in the lung, liver and spleen, with similar splenic viral burden compared with those of non-FMT or FMT-D recipients (Fig. 1h,i).
Because SFTSV infection caused over 60% mortality in Abx mice, but only approximately 20% lethality in PBS-treated mice, it implies that intestinal microbes confer protection against lethal SFTSV infection in our mouse model (Fig. 2a). Subsequently, we collected faecal samples from PBS-treated mice at 3 days post-infection (d.p.i.) with SFTSV and performed FMT experiments in Abx mice. Indeed, Abx mice that received FMT from surviving mice had a significantly enhanced survival rate of SFTSV infection, whereas FMT from donor mice that succumbed to infection exhibited no protective effect (Fig. 2a). Moreover, FMT-reconstituted Abx mice displayed significantly lower Il1b and Il6 expression and ameliorated tissue damage in the lung, liver and spleen, but exhibited similar splenic viral burden to Abx mice without FMT (Extended Data Fig. 2b,c). Further faecal 16S rRNA analysis of PBS-treated mice at 3 d.p.i. revealed a significantly increased relative and absolute abundance of Akkermansia in the mice that survived, but not in those mice that succumbed to SFTSV systemic infection (Fig. 2b and Extended Data Fig. 2d).
We further selected 20 representative faecal samples from recovered patients with high Akkermansia abundance for deep sequencing and identified two operational taxonomic units (OTUs) belonging to A. muciniphila, with OTU1 detected in 15 samples and OTU6 in 5 samples (Fig. 2c). To determine the role of A. muciniphila in the protection against SFTSV infection, we used live and pasteurized A. muciniphila to gavage Abx mice and then inoculated them with SFTSV. Because the relative abundance of Lactobacillus was also increased, whereas the relative abundance of Enterococcus was decreased in the surviving SFTSV-infected patients (Fig. 1a), a well-studied human symbiotic isolate Lactobacillus reuteri (which belongs to the Lactobacillus genus) and Enterococcus faecalis (which is a major cause of nosocomial infection) were used as controls of unrelated commensal bacteria. Strikingly, A. muciniphila colonization significantly protected Abx mice from lethal SFTSV infection, whereas L. reuteri gavage did not alter the lethality rate, and E. faecalis even exacerbated the mortality rate (Fig. 2d), despite effective colonization (Extended Data Fig. 2e).
To further confirm the role of A. muciniphila in modulating host inflammatory responses against SFTSV infection, we performed bacteria monocolonization in a gnotobiotic mouse model. At 3 d.p.i., significantly greater titres of SFTSV were detected in the spleen of non-colonized germ-free (GF) mice compared with in A. muciniphila-colonized GF mice (Fig. 2e). Moreover, non-colonized GF mice exhibited more robust proinflammatory cytokine expression, as well as more severe tissue inflammatory lesions, compared with GF mice colonized by A. muciniphila (Fig. 2e–g). Similarly, in Abx mice, A. muciniphila reconstitution reversed SFTSV titre differences in the spleen, liver and lung at 3 d.p.i. (Extended Data Fig. 2f) and decreased the expression of Il1b, Il6 and Tnfa at the transcriptional and protein levels (Fig. 2h and Extended Data Fig. 2g). At 5 d.p.i., SFTSV titres were higher in the spleen, but not in the liver or lung of Abx compared with PBS-treated and A. muciniphila-colonized Abx mice (Extended Data Fig. 2f). Even without an obvious reduction in viral titres, A. muciniphila colonization significantly diminished proinflammatory cytokine expression (Fig. 2h) and alleviated tissue damage and inflammatory infiltration in tissues at 5 d.p.i. (Extended Data Fig. 2h).
To define the protective component of A. muciniphila, wild-type (WT) B6 mice were gavaged with filtered A. muciniphila supernatant or pasteurized A. muciniphila, which was initiated in parallel with Abx administration until the end of the experiments (Extended Data Fig. 1d), to maintain the level of active metabolites or pasteurized A. muciniphila cells. Interestingly, the administration of Abx-treated mice with A. muciniphila supernatant significantly alleviated SFTSV-induced mortality, whereas gavage with pasteurized A. muciniphila did not alter mortality (Fig. 2d), even though the pasteurized cells were precisely detected in the faeces of treated animals before SFSTV infection (Extended Data Fig. 2e), suggesting that the protective effect was very likely dependent on the microbial metabolites rather than the bacterial components. Together, these concordant results in both GF and Abx animals indicate a role for A. muciniphila-driven metabolites in alleviating inflammatory responses in both the peripheral and distal organs after SFTSV systemic infection.
A. muciniphila-driven conjugated primary bile acid protects the host from SFTSV infection by dampening systemic inflammatory responses
We next performed untargeted metabolomics analysis on a total of 405 serum samples collected from SF-S (n = 222), SF-D (n = 21) and non-SF (n = 132) patients, and detected 69 metabolites that were differentially regulated between the SF-S and non-SF groups and identified 153 differential metabolites between the SF-S and SF-D groups (Fig. 3a). Among these differential metabolites, a series of bile acids (BAs), including chenodeoxycholic acid (CDCA), GCDCA, TCDCA and taurodeoxycholic acid (TDCA), were found to be significantly increased in the SF-S group compared with the other two groups (Fig. 3a,b). Furthermore, a high GCDCA serum concentration was positively correlated with an increased relative abundance of A. muciniphila and negatively correlated with IL-1β and IL-6 expression in serum (Fig. 3c,d).
CDCA is predominantly conjugated to glycine (to form GCDCA) and rarely conjugated to taurine (which would form TCDCA) in humans22. Peripheral blood mononuclear cells (PBMCs) from three SFTS patients in the acute phase treated with GCDCA exhibited significantly reduced expression of IL1B (but not IL6), compared with the untreated control (dashed line), whereas CDCA increased IL1B expression by 2.5-fold (Fig. 3e and Extended Data Fig. 3a). In addition, we pretreated PBMC extracts from five healthy donors with CDCA or GCDCA and discovered a remarkable decrease in IL-1β and IL-6 expression and comparable viral replication in the GCDCA-pretreated PBMCs compared with the untreated control post SFTSV infection (Fig. 3f,g and Extended Data Fig. 3b). TCDCA pretreatment also resulted in significantly decreased IL1B and IL6 transcripts independent of viral replication, whereas TDCA pretreatment greatly increased proinflammatory cytokine expression (Fig. 3h).
Next, we infected PBS-, Abx- or A. muciniphila-colonized Abx mice with SFTSV and subjected the serum samples to untargeted metabolomics analyses. Partial least-squares discrimination analysis revealed significantly different metabolomics profiles of infected mice compared with mock mice (Extended Data Fig. 4a). KEGG analyses indicated that the differential metabolites in PBS- or A. muciniphila-colonized mice compared with Abx mice were indeed enriched in bile secretion and cholesterol metabolism (Fig. 4a and Extended Data Fig. 4b). Furthermore, the relative abundances of GCDCA and a series of taurine-conjugated BAs, including TCDCA, taurocholate acid (TCA), TDCA and taurocholate-α-muricholic acid (T-α-MCA), were markedly elevated in the serum of mice colonized with A. muciniphila (Fig. 4b,c). Among them, GCDCA and TCDCA significantly downregulated Il6, Il1b and Tnfa expression in SFTSV-infected mouse PBMCs, whereas TCA, TDCA and T-α-MCA had no or even an augmenting effect on proinflammatory cytokine expression compared with unprimed cells independent of viral replication (Fig. 4d and Extended Data Fig. 5a,b). In addition, we confirmed that TCDCA downregulated proinflammatory cytokine expression in a dose-dependent, viral replication-independent manner (Extended Data Fig. 5c,d).
Because the vast majority of CDCA is conjugated to taurine instead of glycine in rodents22, we administered Abx mice with TCDCA in drinking water for 4 weeks to test whether TCDCA is the proximate BA metabolite that could likewise ameliorate inflammatory responses post SFTSV infection. Interestingly, TCDCA treatment significantly increased its serum concentration (Extended Data Fig. 5e) and reduced the mortality rate of infected Abx mice by 40% (Fig. 4e). Moreover, we observed significantly reduced proinflammatory cytokine expression and alleviated tissue damage, but no altered viral replication in TCDCA-treated mice compared with Abx control mice (Fig. 4f–h and Extended Data Fig. 5f). These findings suggest a role for A. muciniphila in the induction of inflammation-suppressing GCDCA and TCDCA, which can dampen host inflammatory damage resulting from SFTSV infection in vivo.
The A. muciniphila metabolite HAL suppresses systemic inflammatory responses resulting from SFTSV infection by upregulating BAAT expression in hepatocytes
Hepatocytes synthesize primary BAs via hydroxylation of cholesterol to generate cholic acid and CDCA, and subsequently conjugate to either glycine or taurine by bile acyl-CoA synthetase (BACS) and BAAT23. First, we excluded the possibility that A. muciniphila colonization alone produces more cholesterol to generate more conjugated primary BAs by showing comparable hepatic concentrations of cholesterol among PBS-, Abx- and A. muciniphila-colonized mice (Extended Data Fig. 6a). We also discovered an equivalent upregulation of the apical sodium-dependent BA transporter in the ileum of the Abx- or A. muciniphila-colonized mice compared with PBS controls (Extended Data Fig. 6b), suggesting that the reabsorption of conjugated primary BAs into the circulation does not account for the A. muciniphila-associated protection. Regarding the impact of A. muciniphila colonization on TCDCA/GCDCA biosynthesis, we determined that CYP7A1, CYP8B1, CYP27A1, CYP7B1 and BACS expression was not significantly altered, whereas the expression of BAAT was markedly augmented (Fig. 5a and Extended Data Fig. 6c,d). Because L. reuteri colonization did not protect Abx mice from SFTSV infection (Fig. 2d), we used L. reuteri as an unrelated control and demonstrated that no upregulated levels of BAAT were detected in the livers of L. reuteri-colonized mice in comparison with PBS or Abx mice (Fig. 5a and Extended Data Fig. 6d). Analogous to the Abx model, GF mice colonized with A. muciniphila also had markedly higher BAAT levels in hepatic tissues compared with vehicle-treated GF mice (Fig. 5b and Extended Data Fig. 6e,f). As expected, A. muciniphila-colonized Abx mice or GF mice had remarkably augmented serum levels of TCDCA, as well as other taurine-conjugated BAs, including T-α-MCA and TCA, compared with uncolonized Abx or GF mice (Fig. 5c and Extended Data Fig. 6g).
Because primary BAs are produced by hepatocytes, and modified and bio-transformed by commensal microbiota24, we subsequently hypothesized that A. muciniphila generates certain bioactive small metabolites to enhance BAAT-driven TCDCA production and to ultimately yield protection from SFTSV infection. Therefore, we fractionated the A. muciniphila cultured supernatants to generate a series of fractions based on molecular mass25. Intriguingly, the 10 kDa or less filtrate substantially augmented BAAT expression in a dose-dependent manner in the Huh-7 cell line (similar to what occurred in the unfractionated supernatant control), whereas this phenotype was not observed for the other filtrates (Extended Data Fig. 6h,i). Treatment with proteinase K did not interfere with the effect of the 10 kDa filtrate, thus indicating that the effective factors responsible for the induction of BAAT expression are very likely small metabolic molecules or small microbial proteins that are not affected by proteinase K (Extended Data Fig. 6i).
To investigate the bacterial metabolites mediating BAAT augmentation, we examined the supernatant of A. muciniphila cultured in brain heart infusion medium via untargeted metabolome analyses. Major metabolites that A. muciniphila produces (such as acetate and propionate) were identified using gas chromatography mass spectrometry (Extended Data Fig. 6j). In addition, liquid chromatography–tandem mass spectrometry (LC–MS/MS) analysis yielded five A. muciniphila-associated metabolites with relatively higher concentrations in both positive and negative ion-exchange chromatography (Extended Data Fig. 7a). The concentration of HAL was positively correlated with A. muciniphila cell number (Extended Data Fig. 7b). We pretreated Huh-7 cells with the five metabolites, and one of them, HAL, appeared to induce markedly higher levels of BAAT in both the Huh-7 cell line and mouse primary hepatocytes (MPHs) in a dose-dependent manner (Fig. 5d,e and Extended Data Fig. 7c,d). Because A. muciniphila releases short-chain fatty acids in vivo and in vitro1, we included acetate in the study as an unrelated metabolite control and observed no effect on BAAT expression (Fig. 5e). Furthermore, we administered HAL together with CDCA to MPHs and found that the concentration of CDCA decreased significantly, whereas the concentration of TCDCA increased markedly. This phenotype was lost when small interfering RNAs targeting BAAT were transfected into MPHs, implying that the taurine-conjugating metabolic activities were driven by HAL in a BAAT-dependent manner (Fig. 5f and Extended Data Fig. 7e). Consistently, HAL markedly increased expression of BAAT but not CYP7A1, CYP8B1, CYP27A1, CYP7B1 or BACS in hepatic tissues of Abx animals, in comparison with expression in untreated Abx mice (Fig. 5g and Extended Data Fig. 7f,g), whereas acetate induced significantly lower levels of BAAT compared with HAL (Fig. 5g). As expected, HAL-treated Abx mice or GF mice had markedly augmented serum TCDCA levels compared with uncolonized Abx or GF mice (Extended Data Fig. 8a,b). In addition, Abx mice had no HAL in their serum samples, whereas Abx mice colonized with A. muciniphila had picomolar quantities of HAL in serum (Extended Data Fig. 8c). Notably, a higher concentration of HAL was determined in serum from recovered SFTSV-infected patients compared with recovered febrile patients without SFTSV infection (Extended Data Fig. 8d). These concordant results demonstrated that the A. muciniphila-derived metabolite HAL promotes BAAT expression both in vitro and in vivo.
A further in vivo protective experiment showed that HAL-treated Abx mice exhibited an increased survival rate compared with non-HAL-treated, acetate-treated or propionate-treated Abx mice, albeit to a lesser extent than PBS-treated controls (Fig. 5h), thus suggesting an anti-SFTSV effect exerted by HAL in the context of microbiota deficiency. Systemic inflammatory cytokine levels as well as SFTSV-associated histopathological changes in various tissues were apparently ameliorated in HAL-treated GF mice in comparison with vehicle-treated controls (Fig. 5i, Extended Data Fig. 8e,f). We believe the inflammation-suppressing effect was not directly exerted by HAL because mouse PBMCs pretreated with HAL displayed equivalent proinflammatory cytokine expression compared with the vehicle control post SFTSV infection (Extended Data Fig. 8g). Moreover, HAL treatment failed to provide efficient protection for Abx WT B6 mice transiently transfected with siBAAT (Fig. 5j and Extended Data Fig. 8h), suggesting that HAL confers protection in a BAAT-dependent manner.
Conjugated primary BA GCDCA suppresses NF-κB-mediated inflammatory responses in a manner dependent on the BA receptor TGR5
To investigate the molecular mechanisms by which GCDCA suppresses SFTSV-induced inflammation, we used a human cellular model, THP-1 cells, because monocytes are recognized as the main target cells for SFTSV infection in human PBMCs21,26,27. Indeed, GCDCA drastically downregulated IL-1β and IL-6 in SFTSV-infected THP-1 cells in a dose-dependent and viral replication-independent manner (Fig. 6a–c and Extended Data Fig. 9a,b), whereas CDCA significantly upregulated IL1B and IL6 (Extended Data Fig. 9c,d).
Furthermore, we employed transcriptome analysis of SFTSV-infected THP-1 cells with or without GCDCA pretreatment and found that a large pool of inflammatory response-related genes was specifically upregulated in SFTSV-infected versus mock THP-1 cells, according to Gene Ontology (GO) analysis, whereas significant downregulation of TLR8 and downstream signalling pathways was observed in GCDCA-pretreated infected cells (Fig. 6d,e). However, the knockdown of TLR8 and MyD88, GCDCA still inhibited the expression of IL-1β and IL-6 under SFTSV infection (Extended Data Fig. 10a,b).
In total, several genes in the NF-κB signalling pathway were significantly downregulated in GCDCA-pretreated, SFTSV-infected THP-1 cells, including RELA and NFKB1 (Fig. 6e). We further demonstrated that GCDCA pretreatment markedly decreased the level of the P50 subunit and the phosphorylated form of the P65 subunit (p-P65) in both the cytosolic and nuclear fractions, as well as specifically reducing the P65 level in the nucleus in a MyD88 signalling-independent manner (Fig. 6f and Extended Data Fig. 10c,d).
BAs regulate innate immune responses by activating different receptors, particularly TGR5 and farnesoid X receptor (FXR). TGR5, as a BA-activated membrane receptor, exhibits anti-inflammatory function in macrophages. The strongest activators of TGR5 are lithocholic acid and deoxycholic acid. CDCA has the greatest FXR-activating potential, followed by deoxycholic acid, lithocholic acid and finally cholic acid28,29,30. To identify the receptor(s) that GCDCA utilizes to confer anti-inflammatory responses in the context of SFTSV infection, we knocked down TGR5 and FXR using siRNA followed by GCDCA treatment and SFTSV infection (Fig. 6h and Extended Data Fig. 10h). Intriguingly, the suppressive effect of GCDCA pretreatment on SFTSV-induced IL-1β and IL-6 upregulation was significantly impaired under TGR5 (but not FXR) knockdown (Fig. 6g and Extended Data Fig. 10g,h). Furthermore, following TGR5 depletion, the differences in P50 and p-P65 protein levels were also lost in GCDCA-treated cells compared with the non-treated controls (Fig. 6h and Extended Data Fig. 10e,f). Correspondingly, the significant difference in the survival rate between the TCDCA-treated and untreated Abx WT mice (Fig. 4e) was completely lost in the TCDCA-treated and untreated Abx TGR5−/− mice (Fig. 6i and Extended Data Fig. 10i). Consistent with these results, the protective effect of A. muciniphila colonization (Extended Data Fig. 10j) also required TGR-5 signalling, because the fatality rate was not restored in TGR5−/− mice after A. muciniphila reconstitution (Fig. 6j). Together, these data indicate that A. muciniphila-driven TCDCA confers protection against SFTSV systemic infection via TGR5 signalling in vivo.