A Pulmonary Lactobacillus murinus Strain Induces Th17 and RORγt⁺ Regulatory T Cells and Reduces Lung Inflammation in Tuberculosis

The lung has been considered as a sterile organ until recently (1, 2). Technological improvements (e.g., high-throughput 16S rRNA sequencing) provided direct evidence for the presence of transient microorganism niches along the respiratory tract; these range from the oral cavity to the lung, and they are limited by environmental parameters and immune defense mechanisms (1, 3–6). Despite the low biomass, there is a core of lung-resident bacteria among humans, dominated by Streptococcus, Prevotella, and Veillonella genera, and they can be altered by environmental factors and respiratory diseases often leading to an increase in Proteobacteria (7–11). The airway bacterial community, distinct from those of the gut, seems to play an essential role in respiratory health, as the acquisition of specific bacterial classes after birth is key for lung architecture development and immune tolerance against respiratory inflammatory pathologies later in life (12–15). It is well established that the gut microbiota shape lung immunity via the gut–lung axis (16, 17, 18) and that oral administration of gut probiotics improves immune responses during respiratory infections (19–21). Similarly, approaches using aerosolized antibiotic administration or intranasal (i.n.) delivery of bacterial strains indicate that the lung microbiota can also modulate lung immunity (15, 22–24). In line with this notion, recent studies show that administration of bacteria, either directly isolated from the airways or closely related to resident bacteria but from other sources, confers a potent probiotic effect to prevent and treat respiratory diseases (2, 25, 26). Therefore, the identification of pulmonary bacterial strains capable of modulating lung immunity and their use as probiotics holds promise but still remains in its infancy.

Multiple studies established that microbiota at mucosal sites, such as the gastrointestinal tract, are associated with the abundance of Th17 and regulatory T cells (Tregs) (27, 28). Not unexpectedly, germ-free mice exhibit a diminished number of Th17 and Tregs along with an altered intestinal barrier and defects in the establishment of peripheral tolerance (29–32). Intestinal recolonization of germ-free mice with segmented filamentous bacteria or Clostridia bacteria (clusters IV and XIVa) induces local induction of Th17 or Tregs, respectively (29–32). In general, Th17 cells express the transcription factor retinoic acid receptor–related orphan nuclear receptor γ (RORγt), and contribute to type 3 immunity (33). Most of what we know about Th17 cells is within the context of inflammation. Although they often confer protection against extracellular pathogens, particularly through neutrophil recruitment to mucosal tissue, Th17 cells are also implicated in autoimmune and inflammatory disorders when uncontrolled (33). However, seminal studies point out toward a dichotomous nature of Th17 cells, as they also engage in homeostatic functions (34, 35). For example, the presence of these nonpathogenic resident Th17 cells, depending on defined components of the gut microbiota (e.g., segmented filamentous bacteria), improves intestinal epithelial damage repair via the production of IL-22 in a model of Citrobacter rodentium infection (29, 36). In the case of Tregs, the role of microbiota is much clearer. By controlling effector T cells, Tregs are key actors of immune tolerance that ensure the protection of the host against tissue immunopathology. When dysregulated, however, Tregs are involved in pathogen susceptibility (37). These cells are found at mucosal sites, express the transcription factor forkhead box P3 (Foxp3) and originate either from the thymus or are induced at peripheral sites (38, 39). Thymic Tregs, or conventional Tregs (cTregs), express the transcription factor Helios, and they are specific for self-antigens and migrate to peripheral tissues to reduce autoimmunity (39). Induced Tregs at mucosal peripheral sites display a degree of plasticity to adapt to a given environment. In particular, one subset known for its coexpression of Foxp3 and RORγt, referred hereafter as to “RORγt⁺ Tregs,” is essential to limit immune activation toward resident microorganisms. Indeed, the presence of RORγt⁺ Tregs at mucosal sites highly depends on microbiota (31, 32). Analyses of their TCR repertoire reveal their specificity for commensal bacteria (28, 38), and multiple studies suggest that these cells have protective functions in intestinal homeostasis (41).

Both Th17 and RORγt⁺ Tregs are present in the lung in low numbers at steady-state conditions, but the impact of local microbiota on these cells is unknown (41). In the case of lung Th17 cells, most studies have been performed in the context of inflammation (42). Although airway Th17 cells are protective against both extracellular (e.g., Candida albicans) and intracellular (e.g., Mycoplasma pneumonia) pathogens, they can also contribute, if not properly controlled, to inflammatory lung diseases, such as autoimmunity, fibrosis, asthma, and chronic obstructive pulmonary disease (42). It remains unclear how these different contexts of action for Th17 and lung microbiota are interconnected beyond the correlation between their abundance and physiopathological situations. For instance, the abundance of Proteobacteria in the lower airways correlates with Th17-mediated asthma (43) as well as the enrichment of the lung microbiome with oral taxa and Th17-driven inflammation (44), suggesting that certain airway microbiota communities regulate these lymphocytes. With regards to RORγt⁺ Tregs, the only report that describes their presence in the lung was done by Lochner et al. (45) in 2008. Yet the authors did not investigate how these cells are influenced by the lung microbiota. Other than this, the induction of RORγt⁺ Tregs along with that of Th17 cells was observed during dysbiosis of oral mucosa microbiota upon Candida infections (46). In mice, antibiotic-mediated depletion of oropharyngeal microbiota not only reduces the abundance of both resident Th17 and RORγt⁺ Tregs, but also increases the incidence of tissue damage and high fungal burden in oropharyngeal mucosa, arguing that the interrelationship between resident bacteria and these leukocytes is key against periodontal inflammation generated during infection (46–48). The primordial evidence for a role of pulmonary microbiota in the induction of lung Tregs comes from studies in neonate mice. Although specific pathogen–free (SPF) neonates are as sensitive to allergic asthma as germ-free mice, postbirth microbiota acquisition in the lung transiently modulates the dendritic cell phenotype, which foments the differentiation of Tregs to establish long-term tolerance, resulting in adult SPF mice resistance against asthma (2, 13, 14). In line with these studies, Le Noci et al. (22) recently demonstrated that the use of aerosolized antibiotics induces dysbiosis specifically of the lung microbiota, resulting in a decrease of local Tregs and better immune control of lung cancer (22, 23). Therefore, although direct evidence is lacking about the capacity of specific pulmonary bacterial strains to modulate the abundance and phenotype of Th17 and RORγt⁺ Tregs, these studies make it plausible that airway microbiota may exert host protective functions via the regulation of these leukocytes.

In this study, we investigate the capacity of a recently isolated pulmonary bacterial strain, Lactobacillus murinus CNCM I-5314, to regulate the abundance and phenotype of lung Th17 and RORγt⁺ Tregs, and whether this modulation correlates with any potential beneficial effect in the context of a respiratory chronic infection.

The microbiota bacterial strain used in this study was previously isolated from neonatal mouse lung homogenates (2). This strain was deposited at the French National Collection of Microorganism Cultures (CNCM) under the name CNCM I-5314 (associated with patent file PCT N° WO2020201145A1). Based on its 16S and genomic sequence, CNCM I-5314 was identified as Lactobacillus murinus. CNCM I-5314 was cultivated in Man, Rogosa, Sharpe liquid medium (BD Difco; Thermo Fisher Scientific) without shaking at 30°C (for overnight preculture) or 37°C (for shorter cultures) or grown in 15% agar containing Man, Rogosa, Sharpe medium at 37°C for CFU assays.

Mycobacterium tuberculosis (H37Rv) was grown in 7H9 liquid medium (BD Difco) supplemented with 10% albumin–dextrose–catalase (BD Difco), 0.5% glycerol (Promega), and 0.05% Tween 80 (EUROMEDEX) or in 7H10 Agar medium (BD Difco) supplemented with 0.1% peptone (Thermo Fisher Scientific), 10% oleic acid–albumin–dextrose–catalase (BD Difco) and 0.5% glycerol (49).

Fresh cultures in exponential growth (3–4 h for CNCM I-5314 and 24 h for M. tuberculosis) were used for all inoculum preparation. Bacterial concentration was determined based on OD at 600 nm and previously determined correspondence between OD and bacterial concentration using CFU assays; for CNCM I-5314, in exponential growth, 1uDO is equivalent to 2 × 10⁸ CFU/ml). Pellets were harvested by centrifugation at 3,000 × g at 4°C, washed twice in PBS (MgCl₂, CaCl₂ free; Life Technologies) and resuspended in PBS at 5 × 10⁸ CFU/ml (for i.n. administration) or 5 × 10⁹ CFU/ml (for intragastric administration [gavage] [i.g.]) for CNCM I-5314 or at 5 × 10⁴ CFU/ml for M. tuberculosis.

In some experiments, heat-inactivated CNCM I-5314 was obtained by incubation of a part of the inoculum for 30 min at 70°C, a pasteurization process conserving bacterial component integrity (50). Other experiments were performed with labeled bacteria. Accordingly, CFSE-stained bacteria were obtained after incubation of the inoculum in 100 µg/ml CFSE (CellTrace CFSE Cell Proliferation Kit; Invitrogen) for 10 min at 37°C, three washes in PBS 10% FCS (PAN-Biotech), and resuspension in PBS for administration (51).

Animal care and experimentations were consistent with the French guidelines, and were approved by the Ministry of Higher Education and Research (agreement APAFIS 5704). Six- to eight-week-old female SPF C57BL/6 mice were purchased from Charles River Laboratories. Five to thirteen mice per group were used per experiment. To perform M. tuberculosis infection, mice were i.n. inoculated with H37Rv (1 × 10³ CFU per mouse in 20 µl PBS) under 4% isoflurane anesthesia (Vetflurane; Virbac). CNCM I-5314 (1 × 10⁷ CFU in 20 µl of PBS) or mock (PBS) were administered i.n. three times a week during a 2-wk period before sacrifice (naive mice) or infection under 4% isoflurane anesthesia. During M. tuberculosis infection, CNCM I-5314 administration was continued twice per week until sacrifice. When indicated, CNCM I-5314 was administered i.g. (1 × 10⁹ CFU in 200 µl of PBS) every day during a 10-d period before sacrifice.

Mice (three to six per group) were sacrificed at 42 d postinfection by i.p. administration of pentobarbital (Doléthal; Vétoquinol) at lethal dose. The lungs were inflated and fixed with 10% neutral buffered formalin (Sigma-Aldrich, Merck), and embedded in paraffin. Lung sections (5 µm) were stained with H&E or immunostaining by immunohistochemistry (IHC). For the IHC characterization of mouse leukocytes in tissues, rabbit polyclonal anti–inducible NO synthase (INOS) (clone 54, 1/100; BD Transduction Laboratories) or rabbit polyclonal anti-mouse myeloperoxidase (MPO) (RB-373, 1/100; Thermo Fisher Scientific) were incubated overnight at 4°C after heat-induced epitope retrieval with 10 mM Tris–1 mM EDTA (pH 9) or 10 mM citrate (pH 6), respectively. The sections were incubated at room temperature with goat anti-rabbit HRP (P0448, 1/400; Dako) for 2 h, streptavidin– HRP (GT85912, 1/200; GeneTex) for 1 h, revealed with DAB (C09-12; GBI Labs), and counterstained with hematoxylin (HEMML-OT-100; BioGnost). All sections were digitized using a Pannoramic 250 scanner (3DHISTECH). Histopathological scoring was determined by dividing the area of infiltrated lung tissue by the total area of lung tissue from the H&E staining or IHC immunostaining using the CaseViewer software (3DHISTECH).

Mice (five to eight per group) were sacrificed by cervical dislocation under 4% isoflurane anesthesia 21 or 42 d postinfection or 15 d after first CNCM I-5314 administration (naive mice). In some experiments presented in Supplemental Fig. 1A, bronchoalveolar lavages were performed by gentle intratracheal injection of 1 ml of PBS. Alternatively, lung and spleen were prepared individually into cell suspensions (49). To discriminate by flow cytometry lung-resident leukocytes from those in blood, when indicated, mice were injected i.v. with 2 µg of anti-CD45.2 Ab (BD Biosciences) 5 min prior to sacrifice (52). Lung and spleen were collected in C Tubes (Miltenyi Biotec), homogenized with a gentleMACS Dissociator (Miltenyi Biotec), incubated with 2 mg/ml collagenase D (Roche, Merck), and 0.1 mg/ml DNase I (Roche) for 30 min at 37°C under 5% CO₂. For M. tuberculosis–infected mice, a part of the lung and spleen cell suspension was serially diluted in PBS and spread on Agar medium for CFU scoring to assess M. tuberculosis bacterial load. The remaining homogenates were filtered on 70-µm cell strainers (ClearLine; Dominique Dutscher). Alternatively, spleen collected from noninfected mice and used specifically for flow cytometry were directly crushed on 70-µm cell strainers. After centrifugation of lung and spleen cell suspensions at 330 × g for 5 min, supernatants were 0.22-µm filtered (Millex-GP Sterile Syringe Filters with PES, Millipore Express Plus polyethersulfone; Merck) to exclude M. tuberculosis and stored at −80°C. Cytokine concentration in supernatants was later assessed by ELISA using BD OptEIA Sets (BD Biosciences) following the manufacturer’s instructions. Cell pellets were resuspended in 1 ml of PBS. RBCs were lysed in 150 mM NH₄Cl, 10 mM KHCO₃, and 0.1 mM EDTA (pH 7.2) buffer for 5 min; RBC lysis was stopped by the addition of 10% FCS-containing medium (RPMI 1640, GlutaMAX Supplement, HEPES; Thermo Fisher Scientific) (53). After centrifugation, cell pellets were 40-µm filtered (ClearLine). A portion of the cell pellet was conserved in TRIzol reagent (Invitrogen) at −80°C for RNA quantification and the rest was used for flow cytometry.

For a fraction of the lung cell suspension, cytokine production was stimulated in RPMI supplemented with 10% FCS, 50 ng/ml PMA (Sigma-Aldrich) and 500 ng/ml ionomycin (Sigma-Aldrich) and blocked with brefeldin A (GolgiPlug, 1/1000; BD Biosciences) and monensin (GolgiStop, 1/2000; BD Biosciences) for 4 h at 37°C with 5% CO₂. The remaining cell suspension was stored in Cell Staining Buffer (CSB; BioLegend) at 4°C for the assessment of extracellular marker and transcription factor staining. Cell staining was performed in V-bottom, 96-well plates, and centrifugations were performed at 600 × g (or 700 × g after cell fixation) for 2 min. Cell suspensions were first incubated 20–30 min at 4°C with a mixture containing blocking anti–CD16/32 Ab (TruStain FcX; BioLegend), a viability marker (LIVE/DEAD Fixable Blue Dead Cell Stain Kit; Invitrogen), and the indicated extracellular Abs were diluted in CSB. The following Abs (clone) were used: anti-CD45.2 (104), anti–Siglec F (E50-2440), anti–Ly-6G (1A8), anti-CD4 (SK3), anti–PD-1 (J43) (from BD Biosciences); anti-CD86 (GL1) from eBioscience; anti-CD11b (M1/70), anti-CD11c (N418), anti-MerTK (2B10C42), anti-CD64 (X54-5/7.1), anti-MHCII (M5/114.15.2), anti-CD24 (M1/69), anti-CD103 (2E7), anti–Ly-6C (HK1.4), anti–TCR-β (H57-597), anti-CD3 (17A2), anti-CD8a (53-6.7), anti-CD4 (RM4-5 and GK1.5), anti-ICOS (C3998.4A), anti-CCR6 (29-2L17), anti-CD80 (16-10A1), anti-CD206 (C068C2), and anti-CD16/32 (93) from BioLegend. Cells were washed in CSB, fixed 30 min at room temperature, and permeabilized for 15 min at room temperature using reagents from the Foxp3/Transcription Factor Staining Buffer Set (eBioscience and Thermo Fisher Scientific). Cells were stained with an Ab set to detect intracellular factors diluted in the permeabilization buffer for 45 min at room temperature. Abs (clone) used included anti-RORγt (Q31 378), anti-LAP (TW7-16B4), and anti-TNF (MP6-XT22) from BD Biosciences; anti–T-bet (4B10), anti-FOXP3 (FJK-16s), anti-Ki67 (SolA15), anti-Helios (22F6), and anti–IL-17 (17B7) from eBioscience; and anti–CTLA-4 (UC10-4B9) and anti–IFN-γ (XMG1.2) from BioLegend. In the context of M. tuberculosis–infected mice, the stained cell preparation was fixed for 2 h with 4% paraformaldehyde (Pierce, Thermo Fisher Scientific) at room temperature to ensure M. tuberculosis killing. Just before acquisition, 10–20 µl of counting beads (CountBright; Molecular Probes and Thermo Fisher Scientific) were added to cell suspensions to determine absolute cell numbers. Staining was acquired with an LSR II or LSRFortessa flow cytometer (BD Biosciences) and analyzed with the FlowJo V10 software (Tree Star). Of note, cells were first gated on singlets (forward scatter height versus forward scatter width and side scatter height versus side scatter width) and live cells before further analysis, as shown in (Supplemental Fig. 2).

RNeasy spin columns (RNeasy Mini Kit; QIAGEN) were used to extract RNA from frozen TRIzol samples, according to the manufacturer’s instructions. RNA was then reverse-transcribed into cDNA using M-MLV Reverse Transcriptase (Invitrogen). Quantitative RT-PCR (RT-qPCR) was performed using a 7500 Real-Time PCR System, and data were analyzed using the 7500 Software version v2.3 (Applied Biosystems and Thermo Fisher Scientific). Values were normalized using the housekeeping β-actin gene and expressed as a fold change between experimental (CNCM I-5314–treated mice) relative to control samples. The gene-targeted primer sets used in this study were as follows: β-actin, forward primer (Fw) 5′-GCTGTGCTGTCCCTGTATGCCTCT-3′ and reverse primer (Rv) 5′-CCTCTCAGCTGTGGTGGTGAAGC-3′; Tnf, Fw 5′-CAAAATTCGAGTGACAAGCCTGT-3′ and Rv 5′- CCACTTGGTGGTTTGCTACGA-3′; Ifng, Fw 5′-CAGCAACAGCAAGGCGAA-3′ and Rv 5′-GGACCTGTGGGTTGTTGACCT-3′; Il-6, Fw 5′-ATGTTCTCTGGGAAATCGTGGA-3′ and Rv 5′-AGAATTGCCATTGCACAACTCTT-3′; Il-1b, Fw 5′-GCCCATCCTCTGTGACTCAT-3′and Rv 5′-AGGCCACAGGTATTTTGTCG-3′; Mpo, Fw 5′-GCACAATATGGCACGCCCAA-3′ and Rv 5′-CCAAGGCCTGTCTCTGCTGT-3′; Inos, Fw 5′-TCCTCACGCTTGGGTCTTGTTC-3′ and Rv 5′-TCCAACGTTCTCCGTTCTCTTGC-3′; and Il-10, Mm_Il10_1_SG (GT00106169, QuantiTect Primer Assay; QIAGEN).

Statistical analyses were performed using the software Prism 7 (GraphPad Software, San Diego, CA). A minimum of four mice per group was used in each experiment. Pooled results obtained from two to five experiments are presented in most figures as indicated in figure legends. For M. tuberculosis–infected mice, M. tuberculosis infection was verified for all mice, and uninfected mice were not included in the analyses. When samples followed a normal distribution, the mean value for each group was represented, and a Student t test (comparison between two groups) or one-way ANOVA followed by a Holm-Sidak posttest (to compare more than two groups) were applied to establish significance relative to controls. In the absence of a normal distribution, the bar in the graphs represents the median of each group, and a Mann–Whitney U test or a Kruskal–Wallis followed by Dunn posttest was applied to compare each experimental group with controls. When indicated, a two-way ANOVA followed by a Sidak multiple comparisons test was performed to compare PBS- to CNCM I-5314–treated mice among different cell types. A significant statistical difference was represented by *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001; NS results were indicated when the p value was greater than 0.05.

We previously isolated 20 bacterial strains from lung homogenates of neonatal SPF mice (2). Importantly, i.n. administration of one of these strains (i.e., Enterococcus faecalis, CNCM I-4969) in neonates dramatically reduced type 2 immunity and the incidence of allergic asthma in mouse lungs, suggesting that these pulmonary bacteria have the potential to modulate lung immunity (2). To expedite the search for prospective probiotics, our main approach was to first test those strains closely related to generally recognized as safe bacteria, which are mainly composed of Lactobacillus species (21, 54). Among our collection of pulmonary bacteria, we identified L. murinus (CNCM I-5314; Lagilactobacillus murinus under new taxonomic reclassification) (2, 55), a common resident of the murine gastrointestinal tract associated with the regulation of gut Th17 and Tregs (56, 57). Different studies reported a decrease in Lactobacillus species, particularly L. murinus, in different pathological conditions in the gastrointestinal tract (56–60). In some of these models, administration of L. murinus recapitulated the protective effect by transplantation of gut microbiota, and this was associated with a strong induction of Tregs via the local production of TGF-β and IL-10 (56). More recently, overcolonization of the gastrointestinal tract by L. murinus was shown to result in the particular expansion of Tregs in the lungs, leading to protection against allergic airway inflammation (61). Therefore, the CNCM I-5314 strain seemed like an obvious candidate to be tested as a prospective probiotic. To this end, we first verified that the i.n. delivery of CFSE-labeled CNCM I-5314 reached the lungs and bronchoalveolar space as free bacteria as well as internalized by resident leukocytes, as illustrated in supplementary Supplemental Fig. 1. Having confirmed this, we then assessed the capacity of CNCM I-5314 to induce lung Th17 and RORγt⁺ Tregs by flow cytometry after i.n. or i.g. delivery in naive mice to probe the local versus systemic effects, respectively (Fig. 1A, Supplemental Fig. 2A). The i.n. delivery of CNCM I-5314 resulted in an increase in the proportion and cell number of lung Th17 and RORγt⁺ Tregs but not lung cTregs when compared with those levels observed in mock-treated mice (Fig. 1B–D). By contrast, i.g. administration of CNCM I-5314 resulted in a slight increase in the proportion and cell number of RORγt⁺ Tregs as well as in the total number of cTregs but had no effect on Th17 cells in the lungs (Fig. 1B–D). Importantly, with the exception of a slight (but statistically significant) increase in splenic RORγt⁺ Tregs, the proportion of Th17 and cTregs in the spleen remained intact after i.n. administration of CNCM I-5314 as compared with that in mock-treated mice, suggesting that this bacterial strain mostly exerts a local effect in the lung rather than a systemic one reflected in secondary lymphoid organs (Supplemental Fig. 2B). Our data also indicated that CNCM I-5314 modulates Th17 and RORγt⁺ Tregs in the lung tissue and airway because the exclusion of circulating leukocytes in the lung vasculature yielded the same proportion and cell number (Supplemental Fig. 2C, 2D). Interestingly, the proportion of lung Th17 cells relative to RORγt⁺ Tregs at steady state decreased upon i.n. delivery of CNCM I-5314, indicating that this pulmonary bacterial strain favors RORγt⁺ Tregs over Th17 cells (Fig. 1E). In fact, among total lung Foxp3⁺ CD4⁺ T cells, the proportion of RORγt⁺ Tregs increased from 15% up to 45% after delivery of CNCM I-5314 (Fig. 1E). Furthermore, i.n. administration of heat-inactivated (HK) CNCM I-5314 induced a similar increase in the proportion and cell number of Th17 and RORγt⁺ Tregs as live bacteria when compared with mock-treated mice (Fig. 1B–E), suggesting that a structural component rather than a secreted metabolite, for example, is responsible for this effect. Notably, the different variations of CNCM I-5314 delivery had no impact on the total number of lung CD4⁺ T cells (Fig. 1D), a selective effect which was previously observed for probiotic strains with beneficial effects in disease context (62). Altogether, these findings indicate an important and local rather than systemic (i.e., through the gut–lung axis) effect of CNCM I-5314 in the regulation of lung Th17 and RORγt⁺ Tregs.

Although RORγt⁺ Tregs were previously identified in the lung (45), their characteristics and modulation by the microbiota have not yet been assessed. Because RORγt⁺ Tregs and Th17 cells induced by the microbiota may play a key role in respiratory health and disease (31, 32, 36, 63), we performed a formal characterization of these lung T cell populations at steady state and after i.n. delivery of CNCM I-5314. As described for RORγt⁺ Tregs of gut origin (31, 32, 63), lung RORγt⁺ Tregs expressed Foxp3 and RORγt at a similar level to cTregs and Th17, respectively (Fig. 2A). In addition, a smaller proportion of lung RORγt⁺ Tregs (12%) expressed Helios when compared with cTregs (50%), suggesting a peripheral rather than thymic origin (Fig. 2A, 2B) as previously proposed for these leukocytes in the gut (32, 31, 64). In terms of their activation status, we noticed that, like Th17 cells, there was a high proportion of lung RORγt⁺ Tregs expressing the ICOS receptor (68%), which was further increased across all CD4⁺ T cells by CNCM I-5314 in comparison with the PBS control group (Fig. 2A, 2B). Likewise, 96% of lung RORγt⁺ Tregs expressed the immune-suppressive receptor CTLA-4, which was higher than in cTregs (71%). In addition, 47% of lung Th17 cells were positive for CTLA-4, which was further increased (67%) in CNCM I-5314–treated mice when compared with PBS-treated mice (Fig. 2A, 2B). Moreover, 52% of lung RORγt⁺ Tregs displayed the suppressive receptor PD-1, which was higher than in cTregs (19%) and Th17 (29%) cells; CNCM I-5314 only affected the PD-1 expression in Th17 cells (50%) relative to that found in PBS-treated mice (Fig. 2A, 2B). As previously described (30), we confirmed that lung Th17 (12%) and RORγt⁺ Tregs (23%) expressed the CCR6 chemokine receptor; CNCM I-5314 enhanced CCR6 expression only in Th17 cells (23%) when compared with those from PBS-treated mice (Fig. 2A, 2B). Moreover, when we assessed the proliferation status of the lung CD4⁺ T cells by measuring the Ki67 nuclear marker, we noticed that lung Th17 and RORγt⁺ Tregs had a higher proliferation index compared with cTregs, which was further enhanced in both leukocytes by CNCM I-5314 treatment (Fig. 2A, 2B). In terms of cytokine production, lung Th17 (44%), cTregs (33%), and RORγt⁺ Tregs (15%) exhibited a lower proportion of TNF expression compared with other CD4⁺ T cells (61%), and administration of CNCM I-5314 decreased TNF production across all lung CD4⁺ T cells when compared with the PBS treatment (Fig. 2C). By contrast, the percentage of leukocytes producing IL-17A was uniquely higher in Th17 cells (31%) and increased by CNCM I-5314 treatment specifically in Th17 (48%) and RORγt⁺ Tregs (20%) (Fig. 2C). In the case of immunosuppressive cytokines, we failed to detect lung CD4⁺ T cells expressing IL-10 by our intracellular cell staining approach, independent of the CNCM I-5314 administration. However, the proportion of leukocytes expressing the proform of TGF-β1 was higher in both Th17 (72%) and RORγt⁺ Tregs (82%) compared with cTregs (28%); the CNCM I-5314 treatment only had an increasing effect in cTregs (45%) when compared with PBS control group (Fig. 2C).

Collectively, this characterization indicates that lung RORγt⁺ Tregs share phenotypic characteristics reminiscent of immunosuppressive leukocytes at steady state, which is maintained after administration of CNCM I-5314. This is mirrored by the expansion of Th17 cells with an anti-inflammatory phenotype that may indicate potential lung homeostatic functions rather than proinflammatory responses.

To better understand the immunomodulatory effect enacted by CNCM I-5314 in the context of pulmonary disease, we assessed its potential prophylactic effect in the context of one of the most historical and deadly respiratory pathogens to afflict humans, M. tuberculosis, the etiological agent of tuberculosis (TB). In fact, both Th17 and cTregs play protective and pathogenic roles in TB, depending on the stage and severity of the disease (65–71). As illustrated in (Fig. 3A, mice were inoculated i.n. with CNCM I-5314, or vehicle control (PBS alone) three times per week during a 2-wk period prior to i.n. infection with M. tuberculosis, or mock infection (PBS alone). Thereafter, mice were inoculated twice per week postinfection with CNCM I-5314 (or PBS alone) until the indicated sacrifice time points postinfection. We assessed lung T cells by flow cytometry at 42 d postinfection, which is a standard reference time point to observe the dedicated adaptive immune response against M. tuberculosis (Fig. 3B, Supplemental Fig. 2A). Unlike Th1 cells, which are characterized by the lineage T-box expressed in T cells (T-bet) transcription factor, and are the crucial effector T cells in the control of M. tuberculosis, the administration of CNCM I-5314 increased the proportion and number of lung Th17 cells compared with those obtained in mock-treated mice (Fig. 3B–D). In addition, whereas the proportion of cTregs was slightly diminished, CNCM I-5314 administration strongly increased the proportion and cell number of RORγt⁺ Tregs (Fig. 3B–D). CNCM I-5314 also modulated Th17 and RORγt⁺ Tregs locally in the lung tissue and airway because the exclusion of circulating leukocytes in the lung vasculature yielded the same proportion and cell number (Fig. 3C, 3D). Unlike in steady-state conditions, in which the proportion of Th17 cells relative to RORγt⁺ Tregs decreases upon delivery of CNCM I-5314 (Fig. 1E), this ratio remained constant during M. tuberculosis infection, indicating that the higher abundance of Th17 cells over RORγt⁺ Tregs might be favored in an inflammatory context (Fig. 3E). When considering all lung Foxp3⁺ CD4⁺ T cells in infected mice, however, the proportion of the RORγt⁺ Tregs among total Tregs increased from 14% up to 50% after CNCM I-5314 delivery (Fig. 3E). Notably, compared with the minute proportion of these cells obtained in the PBS-treated M. tuberculosis–infected animals, the CNCM I-5314 appeared responsible for the higher proportion of Th17 and RORγt⁺ Tregs (11% and 8%, respectively) in the lung CD4⁺ T cell compartment in M. tuberculosis–infected mice (Fig. 3E). Because the probiotic use of Lactobacillus strains modulates the levels and phenotype of CD4⁺ T cells (62), we also examined their cytokine production in the presence of CNCM I-5314 at day 42 postinfection with M. tuberculosis (Fig. 4A, 4B). Although most cytokines (TNF, IFN-γ, pro–TGF-β) were not modified by CNCM I-5314 administration, there was an increase in IL-17 production by lung CD4⁺ T cells in CNCM I-5314–treated M. tuberculosis–infected mice (Fig. 4A, 4B) that correlated with the Il17a mRNA expression detected in whole lung tissue from these animals (Fig. 4C). Interestingly, whereas the CNCM I-5314 treatment did not alter the cytokine mRNA gene expression (Tnf, Ifng, Il6, Il10, or Il1b) in lung tissue (Fig. 4C), it decreased significantly the TNF and IFN-γ (but not IL-6 or IL-10) protein production found in lung exudate from M. tuberculosis–infected mice (Fig. 4D).

Altogether, these findings establish that the immunomodulatory effect of exogenous administration of CNCM I-5314 on lung CD4⁺ T cells, distinguished by an increase in Th17 and RORγt⁺ Tregs, overcomes the well-known capacity of M. tuberculosis to modulate lung immunity (72).

To evaluate whether the immunomodulatory effect exerted by CNCM I-5314 influences TB infection outcome, we first verified that the CNCM I-5314 treatment did not result in unwanted inflammation in the lung. As measured by histology at day 42 after mock infection, we confirmed that CNCM I-5314 delivery during mock infection (PBS only) did not result in leukocyte infiltration indicative of lung lesions during M. tuberculosis infection (Fig. 5A). Next, in a 2 × 2 experimental design, in which we compared PBS-treated mock-infected mice with PBS-treated M. tuberculosis–infected mice and CNCM I-5314–treated mock-infected mice with CNCM I-5314–treated M. tuberculosis–infected mice (Supplemental Fig. 3A), we noticed that CNCM I-5314 treatment did not increase the mRNA profiles of proinflammatory genes, such Il1b, Mpo, and Inos in lung tissue at day 42 postinfection with M. tuberculosis (or mock) (Supplemental Fig. 3B). Of note, unlike PBS-treated M. tuberculosis–infected mice, in which there is an expected induction in Inos mRNA expression in lung tissue compared with that obtained in PBS-treated mock-infected mice, we observed a decreasing trend in the expression of this gene in lung tissue from both CNCM I-5314–treated mock-infected mice and CNCM I-5314–treated M. tuberculosis–infected mice (Supplemental Fig. 3B). At the protein level, the CNCM I-5314 treatment did not provoke the release of proinflammatory (TNF, IFN-γ, and IL-6) or anti-inflammatory (IL-10) cytokines in the lung exudate as expected for day 42 postinfection with M. tuberculosis (Supplemental Fig. 3C). Moreover, flow cytometry analyses of intracellular staining for hallmark Th1 markers, such as T-bet and IFN-γ, revealed that the rise of Th1 cells is specific to M. tuberculosis infection, as they become abundant in the lungs of PBS-treated M. tuberculosis–infected mice compared with PBS-treated mock-infected mice (Supplemental Fig. 3D, 3E). Similar to PBS-treated mock-infected mice, Th1 cells are nonexistent in CNCM I-5314–treated mock-infected mice (Supplemental Fig. 3D, 3E). By contrast, we confirmed that the rise of Tregs and IL-17A–producing cells is specific to the CNCM I-5314 treatment, as these cells are only significantly augmented in the lungs of CNCM I-5314–treated mock-infected mice in comparison with PBS-treated mock-infected mice (Supplemental Fig. 3D, 3E). Collectively, these data clearly eliminate the possibility that treatment with CNCM I-5314 causes unwanted inflammation in the lungs of mock- and M. tuberculosis–infected animals.

Next, we measured the M. tuberculosis burden and lung inflammation in infected mock- and CNCM I-5314–treated mice (Fig. 3A). Histological analyses at 42 d postinfection revealed that administration of CNCM I-5314 reduces lung leukocyte infiltration associated with M. tuberculosis infection (Fig. 5A, 5B). However, administration of CNCM I-5314 did not modify the pathogen burden in the lungs nor dissemination to the spleen at 21 or 42 d postinfection (Fig. 5C). IHC staining for lung inflammation markers associated with M. tuberculosis infection, such as MPO and induced INOS, revealed that the CNCM I-5314 treatment decreased significantly the abundance of lung leukocytes positive for these markers at 42 d postinfection (Fig. 5D, 5E). In the case of MPO, quantitative RT-PCR analyses indicated the total mRNA level of the Mpo gene in lung tissue was not affected by CNCM I-5314 treatment at 42 d postinfection (Fig. 5D). Moreover, flow cytometry analyses revealed that neutrophils (Ly-6G^hiCD11b^hi), one of the main leukocytes known to express MPO (73), were not altered by CNCM I-5314 delivery in M. tuberculosis–infected mice (Supplemental Fig. 4A). With regards to INOS, whereas the total mRNA level of the Inos gene was not affected by CNCM I-5314 (Fig. 5E), we identified Ly-6C⁺ inflammatory monocytes as the only leukocyte population being diminished in the lung, as measured by flow cytometry at 42 d postinfection (Supplemental Fig. 4A, 4B). Inflammatory monocytes are known to express both MPO and INOS proteins (73–75). Interestingly, in contrast to interstitial macrophages, dendritic cells, and eosinophils, the total number of alveolar macrophages was augmented in the lungs of CNCM I-5314–treated mice (Supplemental Fig. 4A, 4B). Further analyses of these macrophages seem to indicate that CNCM I-5314 administration established a wound-healing/tissue-repair phenotype characterized by upregulation of the mannose receptor (CD206) and of the FcγRIIIa/FcγRIIIb receptors (CD16/CD32) (76) that was mirrored by the downregulation of proinflammatory markers, such as CD80, ultimately correlating with the overall drop of MPO⁺ and INOS⁺ cells upon CNCM I-5314 treatment (Supplemental Fig. 4B). Indeed, both MPO and INOS are considered markers for proinflammatory (M1) macrophages (76, 77). These data argue that the CNCM I-5314 administration modulates the number of inflammatory monocytes/macrophages infiltrating the lung and the abundance and phenotype of resident macrophages during M. tuberculosis infection.

Altogether, these findings indicate that CNCM I-5314 reduces TB-associated lung inflammation without affecting M. tuberculosis lung colonization or extrapulmonary dissemination or generating unwanted pulmonary inflammation.

Respiratory diseases, such as chronic obstructive pulmonary disease, asthma, acute lower respiratory tract infections, and TB, are among the most common causes of death and severe illness (78). Dysbiosis of the lung microbiota is one of the shared features among these diseases, often associated with deleterious consequences to the host, suggesting that resident microorganisms contribute to respiratory health and disease (8, 9, 79, 80). There is an emerging interest in deciphering whether pulmonary microbiota modulate local immunity. This knowledge could shed light on mechanisms operating in lung homeostasis and response to respiratory pathogens. In this study, we addressed these issues and propose the following contributions to the field.

Pulmonary CNCM I-5314 can regulate the levels of local Th17 cells distinguished by a nonpathogenic phenotype. Recent studies in the intestine have shed light on the dichotomous nature of Th17 cells, which allows them to support the intestinal barrier integrity (34). At steady state, Th17 cells are mainly found in the intestine, but they are also located in small quantities in other barrier organs, such as the skin, the oral cavity, and the lungs (34). These cells are characterized by the core expression of RORγt, CCR6, ICOS, IL-17A and IL-17F, which is a common feature of all RORγt⁺ leukocytes (42). We observe that resident Th17 cells are present, albeit scarcely, in the lung at steady state, confirming previous studies (34, 42, 45). They display a phenotype distinguished by a CCR6^loICOS^intCTLA-4^intPD-1^intIL-17A^intpro-TGF-β1^hiTNF^intKi67^hi profile. Interestingly, a high proportion of lung Th17 cells is positive for immunosuppressive markers, such as CTLA-4, PD-1, and, particularly, pro–TGF-β1, suggesting these cells are nonpathogenic. Moreover, the high percentage of Ki67⁺ Th17 cells indicates a proliferation capacity in situ. Upon CNCM I-5314 i.n. administration, the proportion and number of Th17 cells are increased by a factor of four in the lungs, and their immunosuppressive-oriented phenotype is enhanced, as reflected by the higher expression of CTLA-4 and PD-1 for instance. This effect seems independent of the gut–lung axis because an oral delivery failed to reproduce the induction of lung Th17 cells. These are important findings, given the emerging interest about the role of airway microbiota in Th17-mediated respiratory immunity and inflammatory disorders (44).

Pulmonary CNCM I-5314 can control the homeostasis of lung RORγt⁺ Tregs. As reported in the literature (45), we confirm the low presence of these lymphocytes in the lung, which constitutes ∼15% of the total Treg population at steady-state conditions. Like Th17 cells, RORγt⁺ Tregs are specifically increased in the lungs by i.n. administration of CNCM I-5314, comprising 50% of all Tregs. In fact, this effect is reproduced by the HK version of CNCM I-5314, suggesting that soluble molecules produced by live bacteria are not responsible. Unlike Th17 cells, i.g. administration of CNCM I-5314 also increases (albeit to a lower extent than i.n. delivery) the proportion and numbers of lung cTregs and RORγt⁺ Tregs, alluding to a partial contribution by the gut–lung axis. This is in line with a recent study reporting that overcolonization of the gastrointestinal tract by L. murinus results in an expansion of lung Tregs, leading to protection against allergic airway inflammation (61). Likewise, Sefik et al. (32), who used monoassociation of germ-free mice with 22 bacterial species from the human gastrointestinal tract, including those from the Firmicutes phylum (e.g., Lactobacillus species), demonstrated that multiple but not all strains induce intestinal RORγt⁺ Tregs. We also provide the characterization of lung RORγt⁺ Tregs as Helios^loCD25^loICOS^intCTLA-4^hiPD-1^intCCR6^intIL-17^lopro-TGF-β1^hiTNF^loKi67^hi, which resembles an immunosuppressive phenotype that is slightly enhanced after CNCM I-5314 administration. As in the gut, lung RORγt⁺ Tregs expressed ICOS, which is considered a mucosal Treg marker that is essential for their development and production of IL-10 (81). Lung RORγt⁺ Tregs highly express CTLA-4 and PD-1 compared with cTregs, indicating a potential immunosuppressive capacity (31). This is supported by a study demonstrating that the setup of a tolerogenic lung environment during colonization by the microbiota was dependent on the generation of PD-1⁺Helios⁻ Tregs (13). Like Th17 cells, there is a high proportion of lung RORγt⁺ Tregs positive for Ki67, which is further augmented by CNCM I-5314 administration, arguing for the capacity of these cells to proliferate in situ. Regarding the cytokine profile, lung RORγt⁺ Tregs produce IL-17A, accompanied by contrasting levels of pro–TGF-β1 (81%) and TNF (12%), which becomes accentuated upon administration of CNCM I-5314. IL-17 production by these cells was described to be pathogenic in other contexts and linked with aggravated inflammation (82–86). However, it is worth noting that a TGF-β–rich environment supports an anti-inflammatory role for cells producing IL-17 in the regulation of an immune response (87). Thus, the CNCM I-5314–induced RORγt⁺ Treg-driven production of IL-17A, in combination with a high ratio of pro–TGF-β1:TNF, may play a homeostatic role that contributes to tissue integrity.

CNCM I-5314 increases Th17 and RORγt⁺ Tregs during M. tuberculosis infection. In the TB context, Th17 cells mainly play a proinflammatory role, resulting in either protection or pathology as typically reflected by the recruitment of neutrophils into the lungs (88, 89). In our mouse model of M. tuberculosis infection, the presence of the Th1 cells dwarfs that of Th17 cells in the lung, which is expected because of the crucial importance of Th1 cells in TB. In CNCM I-5314–treated mice infected with M. tuberculosis, lung Th17 cells augment considerably without affecting the abundance of Th1 cells, but this is not accompanied by neutrophil recruitment in the lungs, suggesting that these cells do not provoke an inflammatory effect during infection. A recent study demonstrated that intestinal Th17 cells (characterized by IL-22 and IL-17 production) induced by local commensals are not involved in models of systemic pathological inflammation or in C. rodentium clearance (36). Instead, they decrease the barrier permeability and favor epithelium damage repair during C. rodentium infection. By contrast, pathogen-elicited Th17 cells exhibit high plasticity toward proinflammatory cytokine (e.g., IFN-γ) production and a better engagement in C. rodentium clearance but also contributed to systematic pathological inflammation (36). Concerning RORγt⁺ Tregs, we observed similar trends as Th17 cells in our infection model, despite the well-known capacity of M. tuberculosis to shape the lung immune response. Whereas this Treg subset represents a minor cell population compared with cTregs in mock-treated animals, it augments considerably in infected animals treated with CNCM I-5314. These observations are interesting, considering that the role of Tregs in TB is also controversial (71). On the one hand, early presence of Tregs in the lung is thought to allow bacterial escape from immune control. On the other hand, Tregs are critical to avoid overinflammation, leading to necrosis and lung tissue damage at later stages of M. tuberculosis infection (71). However, whether these cells belong to the RORγt⁺ Treg subtype and whether the RORγt⁺ Tregs induced by lung microbiota could alter the specific response to M. tuberculosis is unknown. Like Th17 cells, the physiological function for RORγt⁺ Tregs is complex and highly dependent on the microenvironment (41). Most of the reports allude a protective role for intestinal RORγt⁺ Tregs against inflammatory disorders such as allergy and colitis (12, 31, 32, 41). Alternatively, there are reports about RORγt⁺ Tregs acquiring Th17-like pathogenic features to promote inflammation in, for example, autoimmunity. Likewise, their presence can also be dysregulated to promote immunosuppression to aggravate cancer and chronic inflammation or to play a deleterious role in the context of helminth infection (82–86). All things considered, we propose that part of the biological role for the pulmonary CNCM I-5314 strain is to regulate the dichotomy of RORγt⁺ Tregs toward a possible protective role during the inflammation generated by respiratory infection such as TB.

CNCM I-5314 decreases lung inflammation associated with M. tuberculosis infection without altering pathogen load. We and others have recently reported an increased susceptibility to M. tuberculosis infection in mice in which the microbiota composition was altered by administration of broad-spectrum antibiotics (21, 49, 80, 90). However, the contribution of lung microbiota to shape the immune response against M. tuberculosis is still poorly understood. In the current study, we demonstrate that CNCM I-5314 reduces lung leukocyte infiltration and IFN-γ/TNF production when administered i.n. to M. tuberculosis–infected mice. Not only do we show a lower abundance of lung MPO- and INOS-positive leukocytes, but also, we describe a lower abundance of inflammatory monocytes mirrored by a significant increase of alveolar macrophages with an inflammatory phenotype. Future studies will assess the direct impact of CNCM I-5314 in the activation of inflammatory monocytes/macrophages and their capacity to generate inflammation and engage in lung homeostatic functions such as wound healing and tissue repair. Notably, our data indicate that CNCM I-5314 has anti-inflammatory properties without affecting M. tuberculosis load in the lung and spleen, nor does it interfere with the generation of Th1 cells and the CD4⁺ T cell–mediated production of IFN-γ and TNF during infection. Although leukocyte infiltration following M. tuberculosis infection is associated with immunopathology leading to lung tissue damage (91), it is difficult to conclude whether the reduction of leukocyte infiltration and inflammatory signals observed in the lungs of CNCM I-5314–treated mice ultimately yield a net health benefit to the host. One factor influencing these results is the choice of SPF mice with an intact microbiota in the lungs. We predict that CNCM I-5314 administration will have different effects in lung immunity in mouse models lacking microbiota (e.g., germ free) or in which dysbiosis is induced specifically in the lungs (e.g., aerosolized antibiotics).

Collectively, our findings provide evidence for a potential role of the pulmonary microbiota in respiratory health and disease. In a model of chronic lung inflammation, our data highlight the importance of Th17 and RORγt⁺ Tregs as probable mediators for the immunomodulatory effect enacted by a lung L. murinus strain. Whether additional pulmonary bacterial strains replicate the same effect remains to be investigated. Yet, as we predict that this may be a general feature of resident lung bacteria, we propose that a better characterization of the mechanisms leading to the regulation of these cell populations and their role in lung immunity will deeply improve our understanding of the microbe–host interactions at this mucosal site. Likewise, the identification of new beneficial pulmonary bacterial strains, especially from the human lung, will represent an essential microbial source of therapeutic molecules for multiple respiratory diseases.

We greatly acknowledge F. Capilla, A. Alloy, and T. Al Saati (US006/CREFRE) for histological analyses; P. Constant, F. Levillain, F. Moreau, C. Berrone, and B. Raynaud-Messina (Institut de Pharmacologie et de Biologie Structurale [IPBS] and Genotoul Anexplo-IPBS platform), for accessing the BSL3 facilities; A. Tridon and G. Marsal (IPBS) for access to zootechnics facilities; and E. Näser and the Genotoul TRI-IPBS facilities for imaging and flow cytometry. We thank A. Charton (IPBS) for her key assistance with in vivo experiments, M. Dupont (IPBS) for RT-qPCR analyses, L. Bermudez (Micalis Institute) for advice on handling of commensal bacterial strains, and Y.-M. Boudehen (IPBS) for technical expertise provided in molecular biology. We are grateful to C. A. Spinner, B. Raymond, C. Gutierrez (IPBS), and V. Saint-Criq (Micalis Institute) for critical reading of the manuscript and helpful comments.

The authors have no financial conflicts of interest.