We first investigated the effect of the reference F. duncaniae strain A2-165, which exhibits strong anti-inflammatory activities (43, 55). C57BL/6 mice were supplemented daily via the intragastric administration of A2-165 or vehicle for five days before the administration of a sub-lethal dose of IAV ( Figure 1A ). The treatment was maintained until sacrifice on day 7 post-infection (7 dpi), the peak of the acute phase. Mock-infected mice served as controls. On average, the IAV-infected mice had lost 12% of their initial body weight on 7 dpi ( Figure 1B ). Supplementation with A2-165 attenuated the body weight loss resulting from infection, although the difference was not statistically significant. We next sought to determine whether A2-165 influenced viral replication in lungs. Determination of the lung infectious titers in a tissue culture infectious dose (TCID₅₀) assay indicated that A2-165 treatment was associated with a significantly lower viral load at 7 dpi ( Figure 1C ).

The pathogenesis of viral infection is associated with an excessive inflammatory response. We therefore sought to determine the consequences of F. duncaniae A2-165 treatment on lung inflammation, as judged from gene expression assays and a pathology assessment. We first measured the mRNA expression of IFN-stimulated genes (ISGs) and inflammatory genes in the lungs, using quantitative RT-PCR assays. Although it did not reach significance, the level of gene expression of the inducible protein Mx1 - the expression of which is proportional to the virus load at this time – tended to be lower in A2-165-treated animals than in nontreated controls ( Figure 1D ). Accordingly, the mRNA expression levels of type I IFN (IFN-β) and type II IFN (IFN-γ) - both of which were associated with uncontrolled pro-inflammatory processes and the impairment of lung epithelial regeneration at 7 dpi (56) - were lower in A2-165-treated animals. This was also the case for inflammatory genes, such as Tnfa, Il1b, Il6, Ccl2 and Cxcl2 ( Figure 1D ). Influenza results in an impairment of the epithelial barrier function (57). Accordingly, the expression of transcripts encoding the transcription factor forkhead box protein J1 (Foxj1, involved in cilium formation) and the Clara cell secretory protein (Cc10, involved in the repair and maintenance of airway integrity) was lower in IAV-infected animals ( Figure 1E ). Interestingly, A2-165 administration significantly attenuated the relative fall in Foxj1 and Cc10 (not significant) gene expression - suggesting that administration of the bacterium was associated with reinforcement of the lung barrier. Examination of lung sections from IAV-infected mice revealed significant pathological lesions at 7 dpi ( Figure 1F , left panel). These changes were alleviated in F. duncaniae-treated animals, as evidenced by less alveolar wall thickening and less inflammatory cell infiltration of the alveoli. The lung histopathology scores were significantly lower in mice treated with A2-165 ( Figure 1F , right panel). Taken as a whole, these results show that F. duncaniae A2-165 not only hampered viral replication in the lungs but also diminished lung inflammation.

In order to confirm that F. duncanie’s protective effect was not strain-specific, we evaluated the impact of a second strain (I-4574). Supplementation with I-4574 was associated with significantly less body weight loss during influenza, which indicated a better disease outcome ( Figure 2A ). Accordingly, the viral load (as determined by the TCID50) was lower in mice that received I-4574 ( Figure 2B , left panel). Quantification of the M1 protein transcripts in RT-qPCR assays confirmed this finding ( Figure 2B , right panel). To better explore the mechanisms involved in reduced viral load, we evaluated the impact of I-4574 supplementation on the antiviral immune response at early time point (2 dpi). To this end, we measured levels of various IFNs in bronchoalveolar lavage fluids; IFNs are known to be involved in the control of viral replication. Mock-infected mice treated with vehicle or I-4574 were used as controls. While I-4574 had no effect on the production of the type I (IFN-α and IFN-β) and type II (IFN-γ) IFN, it augmented that of the type III IFN IFN-Λ2,3 ( Figure 2C and not shown). At this early time point, we failed to detect IFN production in mice infected with influenza (without treatment). It is noteworthy that in the absence of IAV, I-4574 also enhanced IFN-Λ2,3 production, albeit not in a significant manner. Quantification of inflammatory gene expression in the lungs at 7 dpi confirmed the efficacy of I-4574 in this model of infection (7 dpi, Figure 2D ). Accordingly, the BALF concentrations of TNF-α and CCL-2 proteins were lower (albeit not significantly) in I-4574-treated mice than in control mice ( Figure 2E ). Lastly, I-4574 attenuated the lower mRNA expression of markers involved in lung barrier functions, such as Foxj, Cc10, and the tight junction protein Zonula occludens 1 (Zo1) ( Figure 2F ). We conclude that the protective effect of F. duncaniae is not strain-specific and is associated with enhanced IFN-Λ production soon after infection.

Influenza infection in mice associates with gut dysbiosis at 7 dpi (21, 28, 58–60). To investigate the potential effects of F. duncaniae supplementation on gut microbiota’s composition during infection, 16S rRNA gene amplicon sequencing was performed on fecal samples. As expected, IAV induced changes in the composition of the gut microbiota at 7 dpi ( Figure 3 ). Weighted Unifrac-based principal component analysis (PCA) plot revealed that the two IAV-infected groups clustered away from the mock-infected group (P = 0.004 for the vehicle group and P = 0.001 for the F. duncaniae group as assessed by permutational multivariate analysis of variance, PERMANOVA) ( Figure 3A and not shown). The infected group that received F. duncaniae separated from the vehicle control group, although not significantly (P = 0.136, PERMANOVA). We next analyzed the composition of the gut microbiota at different taxonomic levels. At the phylum level, IAV led to a significant increased relative abundance of Bacteroidota (previously known as Bacteroidetes), Verrucomicrobiota, Pseudomonadota (Proteobacteria) and Cyanobacteriota ( Figure 3B , left panel and Supplementary Figure 1B ). On the other hand, IAV associated with a drop of Bacillota (Firmicutes). Supplementation with F. duncaniae significantly attenuated changes in Verrucomicrobiota and Cyanobacteriota relative proportions ( Figure 3B ). A significant impact of F. duncaniae was also observed at lower taxonomic levels ( Supplementary Figure 1C for the class). Linear discriminant analysis effect size (LEfSe) analyses revealed that, at genus levels, F. duncaniae supplementation augmented the relative frequency of the SCFA producing bacteria Dubosiella (Bacilllota) and Muribaculaceae UBA3263 (Bacteroidota) ( Figures 3C, D ). On the other hand, it significantly reduced the IAV-induced augmentation of Akkermansia (Verrucomicrobiota), Odorobacter (Bacteroides) and crytpobacteroides (Bacteroides). In line with the fact that it does not colonize in the mouse gut (61, 62), we failed to detect F. duncaniae by quantitative PCR. Collectively, F. duncaniae administration impacted on gut dysbiosis during influenza.

The production of SCFAs depends strongly on the gut microbiota’s functional activity. As observed previously (28, 29), cecal levels of acetate, propionate and butyrate fell massively upon IAV infection ( Figure 3D ). In contrast, levels of isobutyrate, isovalerate and valerate did not change significantly ( Supplementary Figure 1D ). Interestingly, F. duncaniae treatment was associated with attenuation of the drop in acetate, propionate and butyrate (not significant) levels at 7 dpi. Given that acetate and butyrate are important during influenza (24, 25, 28, 29, 32), this partial restoration might contribute to the better disease outcomes associated with F. duncaniae administration.

To investigate how metabolites (e.g. butyrate) might contribute to the F. duncaniae’s protective influence, we metabolically inactivated bacteria by pasteurization. Treatment with the pasteurized bacteria was associated with an attenuation (not significant) of body weight loss in IAV-infected mice ( Figure 4A ). Although the pasteurized form was less efficacious than the live form, it was associated with a significantly lower viral load in the lungs ( Figure 4B ). Accordingly, the expression level of gene encoding Mx1, Isg15 and IFN-β was lower in mice supplemented with pasteurized F. duncaniae than in control (vehicle) mice ( Figure 4C ). Here again, the effects were less strong than with live F. duncaniae, particularly for IFN-β. Pasteurized F. duncaniae failed to influence the expression of transcripts encoding inflammatory cytokines such as CXCL2 ( Figure 4C and data not shown). Influenza infection is known to be associated with gut injury, poor mucosal barrier function, and greater levels of gut inflammation (29, 58, 63). Given that F. duncaniae has been linked to lower levels of gut inflammation, we compared the effects of live vs. pasteurized F. duncaniae in IAV-infected mice. The colon length - a marker of intestinal inflammation - was lower in IAV-infected mice ( Figure 4D ). Interestingly, live F. duncaniae (but not pasteurized F. duncaniae) abrogated the IAV-induced reduction in colon length. Influenza led to greater expression of ISGs and inflammatory genes in the colon ( Figure 4E ). Live F. duncaniae and, to a lesser extent (for Ifng) pasteurized F. duncaniae, abrogated cytokine production, as assessed in quantitative RT-PCRs ( Figure 4E ). We conclude that the live form of F. duncaniae alleviates influenza outcomes more potently than the pasteurized form does.

Individuals with severe influenza are prone to develop secondary bacterial infections. This enhanced susceptibility is due - at least in part - to impaired antibacterial defenses and altered lung integrity and barrier functions (64, 65). We showed recently that the drop in acetate production during IAV causes bacterial superinfection (28, 29). In view of F. duncaniae’s effect on SCFA levels, we next investigated the bacterium’s potential effect on secondary bacterial infections after influenza. To this end, F. duncaniae-supplemented, IAV-infected mice were infected secondarily with Streptococcus pneumoniae ( Figure 5A ). The number of viable bacteria in the lung was counted 30 hours after the inoculation of S. pneumoniae. As shown in Figure 5B , supplementation of F. duncaniae failed to significantly lower the number of viable bacteria in the lung. In the model studied here, S. pneumoniae disseminates in the blood to reach the spleen. Of interest, supplementation of F. duncaniae was associated with significantly less systemic bacterial translocation in doubly infected mice ( Figure 5C ). This reduced systemic invasion is superinfected mice probably relies on improved lung integrity. We conclude that preventive supplementation with F. duncaniae in mice lowers the severity of both influenza and bacterial superinfection.

C57BL/6J male mice (8-week-old) were purchased from Janvier-Labs (Le Genest-Saint-Isle, France). Mice were housed under specific pathogen-free condition in the biosafety level 2 animal facility of the Institut Pasteur de Lille. Animals were maintained with a strict 12-hour dark/light cycle and were given ad libitum access to regular chow and water. Mice were fed a standard rodent chow (SAFE A04) (SAFE, Augy, France) and water ad libitium. This diet contains11.8% fiber including 10% water-insoluble fiber (3.6% cellulose) and 1.8% water-soluble fiber. After a one-week acclimation period, mice were treated with F. duncaniae (see below) and then infected as follows. Mice were anesthetized by intraperitoneal injection of 1.25 mg of ketamine plus 0.25 mg of xylazine in 200 µl of phosphate buffered saline (PBS), and then intranasally (i.n.) infected with 50 µl of PBS containing (or not, for mock group) 75 plaque forming units (p.f.u.) of the H3N2 IAV strain A/Scotland/20/1974. For the superinfection, mice were intranasally infected with S. pneumoniae (serotype 1, clinical isolate E1586, 1x10³ viable bacteria) 7 days post-influenza (28). Mice were euthanized by cervical dislocation. The number of live bacteria in lungs and spleen was determined 30 hours after S. pneumoniae challenge as described (28). All experiments involving IAV and S. pneumoniae were performed within the Biosafety Level 2 facility of the Institut Pasteur de Lille. The protocols were validated by the local committee for the evaluation of the biological risks and complied with current national and institutional regulations and ethical guidelines (Institut Pasteur de Lille/B59- 350009). The experimental protocols using animals were approved by the institutional ethical committee, Comite d’Ethique en Experimentation Animale (CEEA) 75, Nord Pas-de-Calais. The animal study was authorized by the Education, Research and Innovation Ministry under registration number APAFIS#13743-2018022211144403v2 and APAFIS#33193-2021092318476492.

F. duncaniae A2–165 (DSM 17677, DSMZ collection, Braunschweig, Germany) and I-4574 (CNCM collection, Paris, France) strains, both isolated from human fecal stool, were grown at 37°C in Brain-heart infusion medium supplemented with 0.5% yeast extract (Difco, Detroit, USA), 2 mM acetate, 2 mg/ml fructose and 0.5 mg/ml cysteine HCl (Sigma-Aldrich), in an anaerobic chamber (90% N2, 5% CO,2 and 5% H2). Bacteria were washed and pellets were resuspended in PBS containing 16% glycerol at 5 x 10⁹ CFU/ml aliquoted and frozen at -80°C until use. Mice were daily treated by intragastric administration (200 µl) of PBS containing 16% glycerol (mock-infected and IAV-infected mice) or of a suspension of F. duncaniae at 5 x 10⁹ CFU/ml (1 x 10⁹ CFU/mouse/day), 5 days before infection and until the end of the procedure. Pasteurization was obtained after 30 min at 70°C. Integrity of the bacteria was checked by microscopy analysis.

Body weight was monitored daily after IAV infection. Mice were sacrificed at 2 dpi or 7 dpi. Bronchoalveolar lavage fluids were collected using 0.5 ml PBS containing 0.2% BSA and centrifuged at 1100 x g for 2 min at 4°C. Supernatants were stored at -20°C until ELISA or Multiplex analysis. After perfusion with PBS, the left lobe of the lung was stored in 4% PBS-buffered formaldehyde for 2 days, rinsed in PBS, transferred into 70% ethanol solution and processed into paraffin-embedded tissue blocks for histology assessment. Part of the right lobe was stored in RNAlater^® (Ambion, Thermo Fisher Scientific, Waltham, MA) and frozen at -80°C until qRT-PCR analysis. The remaining part of the lung was immediately frozen in liquid nitrogen and stored at -80°C until TCID₅₀ evaluation. Colon length was measured as described in (29). Fecal samples were collected from each individual mouse, immediately frozen in liquid nitrogen and stored at -80°C until SCFA quantification and 16S rRNA gene amplicon sequencing.

Virus load in lung homogenates was assessed by titration of 50% tissue culture infectious dose (TCID50) using Madin-Darby canine kidney (MDCK) cells. Quantification of viral load and of gene expression was as follows. After homogenization of lung and colon tissue in RA1 buffer (Macherey-Nagel, Hoerdt, Germany), total RNA was purified using the NucleoSpin RNA isolation kit (Macherey-Nagel). The quantity and quality of RNA were checked using a NanoDrop spectrophotometer (Thermo Fisher Scientific). RNA was reverse transcribed using the High-Capacity RNA-to-cDNA™ Kit (Applied Biosystems, Hammonton, NJ). Quantification of the viral load was determined by measuring the copy number of RNA encoding the viral M1 gene. Briefly, after treatment with RNAse OUT (Invitrogen), RNA was reverse-transcribed using the SuperScript^® II Reverse Transcriptase (Invitrogen, Life Technologies, Carlsbad) and primer specific for M1 (5′-TCT AAC CGA GGT CGA AAC GTA-3′). Amplification was performed using TaqMan Universal PCR Master Mix (Applied Biosystems), the specific primers for M1 (forward: 5′-AAG ACC AAT CCT GTC ACC TCT GA-3′; reverse: 5′-CAA AGC GTC TAC GCT GCA GTC C-3′) and M1 specific TaqMan probe (FAM) 5′-TTT GTG TTC ACG CTC ACC GTG CC-3′ (TAMRA), on the QuantStudio™ 5 Flex Real-Time PCR System (Applied Biosystems). The standard curve was performed using a plasmid constructed with a synthetic gene containing the IAV M1 sequence (segment 7).

After homogenization and RNA extraction of lung and intestinal (colon) tissues, RNA was reverse transcribed using the High-Capacity RNA-to-cDNA™ Kit (Applied Biosystems). qRT-PCR was performed using the Power SYBR Green PCR Master Mix on the QuantStudio™ 5 Flex Real-Time PCR System (Applied Biosystems) ( Table 1 for the sequences of the oligonucleotides). The relative gene expression (2^-▵▵CT) was normalized according to the PCR cycle thresholds (Ct) for the gene of interest and the housekeeping gene coding glyceraldehyde 3-phosphate dehydrogenase (Gapdh) (^▵Ct) and to the ^▵Ct values between non-infected (mock) and infected animals.

Bronchoalveolar lavage was performed by instillation of 500 μL of PBS containing 2% bovine serum albumin via the exposed trachea into the lungs, followed by aspiration after 30 s, using a 18G IV cannula (B. Braun, Melsungen, Germany). Supernantants were centrifuged at 10,000 g for 2 minutes before cytokine dosage. The concentrations of IFN Λ2−3 was determined by ELISA (R&D systems DuoSet, Minneapolis, MN). Concentrations of IFN-γ, IFN-β, IFN-α, TNF-α, and CCL2 were determined using multiplex analysis following the manufacturers’ instructions (LEGENDplex, Biolegend, San Diego, CA).

Lung tissue samples were fixed in 10% PBS-buffered formaldehyde for 48h, rinsed in PBS, transferred in 70% ethanol and then paraffin-embedded under standardized conditions. Tissue sections (3 µm thick) were stained with hematoxylin and eosin. Lung sections were scanned with a Nanozoomer (Hamamatsu Photonics, city, Hamamatsu, Japan), and morphological changes were assessed using a semi-quantitative dual histopathology score adapted from (93, 94), which evaluates each following score: cellular death/necrosis, alveolar and/or perivascular edema, hyaline membrane or fibrin, inflammation, thrombi, congestion, hemorrhage, type II hyperplasia, and syncytia from 0 to 4 (0 = absent, 1 = 1-10% of the lung section affected, 2 = 11-25% of the lung section affected, 3 = 26-50% of the lung section affected, and 4=>50% of the lung section affected). The degree of peribronchial and perivascular inflammation was scored on a subjective scale of 0 (no) to 5 (severe) in a blind manner. The percentage of the altered/inflamed zone of the section (lung lesion) was also evaluated (from score 0 to score 10). The score was then defined as the sum of the peribronchial and perivascular scores, and the percentage of altered zone. Representative photomicrographs were taken to illustrate the major distinguishing morphological features among the experimental groups.

Genomic DNA extraction from feces was conducted following an adaptation from González-Cabrera et al. (95). Briefly, around 50 mg of feces were crushed in 300 µl of SLX-Mlus Buffer (Omega Bio-tek, Norcross, GA) using a plastic homogenizer. A volume of 50 µl was then sampled and incubated at 99°C for 10 min with 16.7 µl of 1 M NaOH. The solution was neutralized adding 66.7 µL of 0.1248 M HCl, 0.125 M Tris-HCl and 0.5% Triton X-100 before a last incubation at 99°C for 10 min. Genomic DNA suspensions were stored at -20°C. Microbial diversity and composition were determined for each sample by targeting a portion of the ribosomal genes. A set of 12 primers (forward) and 3 primers (reverse) were used to amplify the V3 and V4 hypervariable region of the 16S rRNA gene fragment using an optimized 16S-amplicon-library preparation protocol (Biomnigene, Besançon France). Briefly, 16S rRNA gene PCR was performed using 5 µl of 1/40 diluted genomic DNA using GoTaq^® Rapid PCR Master Mix (Promega, Madison, WI) using 15 barcoded primers at final concentrations of 0.2 μM and an annealing temperature of 55°C for 38 cycles. The PCR products were multiplexed at equal concentrations and purified using a PippinHT system (Sage Science, Beverly, MA). Sequencing was performed using a 250-bp paired-end sequencing protocol on an Illumina MiSeq platform (Illumina, San Diego, CA).

A step of removal of low-quality reads from the raw paired-end reads were performed using Fastp. Remaining sequences were assigned to samples based on barcode matches using cutadapt (version 4.4). Data were then imported in Qiime2 (2023.7) and forward and reverse primer sequences were removed using the cutadapt plugin. The sequences were denoized using the DADA2 method, and reads were classified using the GTDB database (release 214). A total of 666,810 paired-end reads were analyzed, with an average of 22,227 per sample (range: 10,492 to 47,244). Beta diversity was computed using Qiime2. Principal Coordinate analyses of the Weighted Unifrac distance were performed to assess beta diversity. Differences between groups were tested using PERMANOVA analysis. Raw sequence data are accessible in the National Center for Biotechnology Information (project number PRJNA1046012), biosample accession numbers SAMN38468615 to SAMN38468644 (https://dataview.ncbi.nlm.nih.gov/object/PRJNA1046012?reviewer=7da92vicemsbkcq6oqsf79pbph). Differential analysis was performed using the linear discriminant analysis effect size (LEfSe) pipeline.

Fecal concentrations of SCFAs were quantified as previously described (28). After extraction of fecal contents with water (wt g/vol) and centrifugation at 12,000 × g for 15 min, the proteins were precipitated from the suspensions using a phosphotungstic acid saturated solution. Supernatants (100 µl) were analyzed using a gas chromatograph (GC7890B; Agilent Technologie, Les Ulis, France). All samples were analyzed in duplicate. Data were collected and peaks integrated using the Open Lab Chemstation software (Agilent Technology, Santa Clara, CA). Results were expressed as μmol/gramm of fecal content.

For infectious markers, results are expressed as mean ± standard error of the mean (SEM). Statistical analyses were performed using GraphPad Prism v8.0.2 and R v4.0.2 softwares. The Mann-Whitney U test was used to compare the two groups, unless otherwise stated. Comparisons of more than two groups were analyzed using the one-way ANOVA Kruskal-Wallis test (nonparametric), followed by Dunn’s post-test. *P < 0.05; ** P < 0.01, *** P < 0.001.