Macrophage cytokine responses to commensal Gram-positive Lactobacillus salivarius strains are TLR2-independent and Myd88-dependent

The mechanisms through which cells of the host innate immune system distinguish commensal bacteria from pathogens are currently unclear. Toll-like receptors (TLRs) are a class of pattern recognition receptors (PRRs) expressed by host cells which recognize microbe-associated molecular patterns (MAMPs) common to both commensal and pathogenic bacteria. Of the different TLRs, TLR2/6 recognize bacterial lipopeptides and trigger cytokines responses, especially to Gram-positive and Gram-negative pathogens. We report here that TLR2 is dispensable for triggering macrophage cytokine responses to different strains of the Gram-positive commensal bacterial species Lactobacillus salivarius. The L. salivarius UCC118 strain strongly upregulated expression of the PRRs, Mincle (Clec4e), TLR1 and TLR2 in macrophages while downregulating other TLR pathways. Cytokine responses triggered by L. salivarius UCC118 were predominantly TLR2-independent but MyD88-dependent. However, macrophage cytokine responses triggered by another Gram-positive commensal bacteria, Bifidobacterium breve UCC2003 were predominantly TLR2-dependent. Thus, we report a differential requirement for TLR2-dependency in triggering macrophage cytokine responses to different commensal Gram-positive bacteria. Furthermore, TNF-α responses to the TLR2 ligand FSL-1 and L. salivarius UCC118 were partially Mincle-dependent suggesting that PRR pathways such as Mincle contribute to the recognition of MAMPs on distinct Gram-positive commensal bacteria. Ultimately, integration of signals from these different PRR pathways and other MyD88-dependent pathways may determine immune responses to commensal bacteria at the host-microbe interface.


Results
Lactobacillus salivarius strains induce cytokine responses that are TLR2 independent but MyD88 dependent in macrophages. We investigated the role of TLR2 and MyD88 in the recognition and response by murine macrophages to the Gram-positive bacterium L. salivarius. Live cells of thirty-three L. salivarius strains from different environmental sources (Table 1), control probiotic and pathogenic bacteria and individual TLR ligands were screened using MSD multi-plex cytokine assays for their ability to stimulate WT, TLR2 −/− and MyD88 −/− BMDMs to produce a panel of cytokines (TNF-α, IL-6, IL-10, IL-12p70, IL-1β and IFN-γ) and the chemokine KC/GRO. TLR2 agonists (Pam3csk4 and FSL-1) did not trigger TNF-α responses in TLR2 −/− BMDMs, but TLR4 agonist (LPS) triggered cytokine response at equal magnitude in both WT and TLR2 −/− BMDMs ( Supplementary Fig. S1), thus confirming the phenotype of the TLR2 −/− macrophages used in this screen.
Irrespective of the environmental source of the L. salivarius strains, cytokine responses triggered by these bacterial strains were not reduced in TLR2 −/− BMDMs in comparison to WT BMDMs (Fig. 1). Indeed, we observed an increased cytokine response from TLR2 −/− BMDMs in these initial screens. The TLR2 independent cytokine responses triggered by L. salivarius did not correlate with the host species or tissue of origin of the strains ( Table 1). The human commensal strain, L. salivarius UCC118 induced highest levels of cytokines, whereas L. salivarius CCUG44481 and the probiotic L. rhamnosus GG (LGG) induced the lowest, in WT and TLR2 −/− BMDMs (Fig. 1).
However, MyD88 −/− macrophages did not produce cytokines or chemokines in response to stimulation with L. salivarius strains, while exposure to two intracellular pathogens (Gram-negative Salmonella typhimurium SJW1103 and Gram-positive Listeria monocytogenes EDGe) resulted in production of TNF-α, IL-10 and KC/ GRO, at fourfold lower levels compared to WT macrophages (Fig. 1). In addition, L. salivarius UCC118 was unable to induce NF-κB driven reporter activity (Fig. 2a) and TNF-α secretion (Fig. 2b) in MyD88 −/− THP-1 human monocyte-like cells thus confirming the requirement of MyD88 signaling for L. salivarius induced cytokine responses in macrophages. TriDAP (a NOD1 agonist), activated NF-κB driven reporter activity at comparable levels in WT and MyD88 −/− THP-1 cells thus confirming the capacity for MyD88-independent activation of NF-κB in these cell lines (Fig. 2a). Since TLR2 was dispensable for inducing cytokine responses by L. salivarius UCC118, we investigated if the ability to trigger TLR2-independent cytokine responses was a general feature of Gram-positive commensal bacteria by measuring cytokine responses triggered by Bifidobacterium breve UCC2003 (B. breve) in WT and TLR2 −/− BMDMs. B. breve is another sub-dominant member of the adult human gut microbiota and one of the first colonizers of the human gastrointestinal tract 35 www.nature.com/scientificreports/ shown to have immunomodulatory and other probiotic properties owing to the presence of an exopolysaccharide (EPS) on its cell surface 37 .
L. salivarius UCC118 cytokine responses are TLR4 independent in murine BMDMs. TLR4 is a PRR required for recognition of the Gram-negative bacterial cell-envelope component, LPS. However, recent studies suggest that TLR4 also contributes to the recognition of Gram-positive bacteria such as Streptococcus pneumoniae 39 . It is also required for triggering protective host responses in response to exopolysaccharide (EPS) from Gram-positive Lactobacillus spp. and Bacillus spp. and from Gram-negative bacteria like Bordetella spp. [40][41][42] .
Since EPS production is an important characteristic of Lactobacillus spp. 43 The heat map represents TNF-α, IL-10, IL-12p70, IL-6, IFN-ɣ, IL-1β and mKC responses in WT, MyD88 −/− and TLR2 −/− BMDMs treated with Listeria monocytogenes EDGe, Lactobacillus rhamnosus GG, Salmonella typhimurium SJW1103 and L. salivarius strains isolated from human ileal caecal region (hc), saliva (hs), blood (hb), gall bladder (hg) animals (a), food (f) or unknown sources (un). BMDMs were treated with the respective ligands or bacteria at MOI of 10 for 20 h followed by cytokine expression analysis using MSD 7-plex assay. The gradient of the heat maps generated represent log of median of cytokine concentration (pg/ml) and goes from blue to red (on a range of − 5 to + 15). Data shown are the average of 3 independent experiments (n = 3). www.nature.com/scientificreports/ after co-culture with L. salivarius UCC118 (Fig. 4a-f). Blocking TLR4 receptor activation in WT BMDMs using the TLR4 selective inhibitor Tak242 38 prevented induction of TNF-α responses by LPS but not by L. salivarius UCC118 (Fig. 5). This suggested that L. salivarius UCC118 induced cytokine responses in BMDMs were TLR2 and TLR4 independent. Because of the surprising TLR2-independent nature of the cytokine responses to L. salivarius UCC118 from BMDMs derived from TLR2 −/− mice purchased from Jackson labs 39 (Fig. 3a-f), we also compared cytokine responses to L. salivarius UCC118 from BMDMs generated from different TLR2 −/− mice purchased from Oriental Bioservice 33 (Fig. 4a-f) in order to exclude confounding variables such as effects from different genetic backgrounds, vendors or methods used to generate the targeted KOs (knockout). Cytokine responses to LPS and E. coli (TLR4 agonists) were TLR2-independent, but TLR4-dependent ( Fig. 4a-f). Cytokine responses to Pam3csk4, FSL-1 were TLR2-dependent as observed in mice from Jackson labs ( Fig. 3) but cytokine responses to HKLM (TLR2 agonist) were TLR2-dependent in BMDMs from mice purchased from Oriental Bioservice ( Fig. 4a-f) as opposed to BMDMs from mice sourced from Jackson labs ( Fig. 3a-f). Interestingly, in BMDMs generated from mice sourced from Oriental Bioservice, increased production of KC in response to L. salivarius UCC118 was abrogated in TLR2 −/− and TLR2/4 −/− compared to WT BMDMs (Fig. 4d) while this effect was not observed in the BMDMs generated from TLR2 −/− mice purchased from Jackson labs (Fig. 3d). Another major difference was the increased Il12-p70 response from TLR2 −/− BMDMs generated from mice purchased from Jackson labs (Fig. 3e) and the increased TNF-α response (Fig. 4c) in mice purchased from Oriental Bioservice. All these BMDMs expressed classic macrophage markers, CD11b and F4/80 at identical levels as assessed by FACS, while as expected all TLR2 −/− BMDMs did not express TLR2 (Fig. 6). Thus, the TLR2-independent nature of the BMDM cytokine response to L. salivarius UCC118 was confirmed in BMDMs from two different strains of TLR2 −/− mice. However, we did observe different TLR2-independent responses which is suggestive of possible genetic differences between the different KO mice strains on the C57BL/6 background. L. salivarius UCC118 induced TNF-α response is partially Mincle receptor dependent. Since neither TLR2 nor TLR4 were required for L. salivarius UCC118 induced cytokine responses, we analysed the expression of 75 PRR pathway-associated genes in WT BMDMs co-cultured with L. salivarius UCC118 for 20 h, in an attempt to identify PRRs involved in the recognition of L. salivarius UCC118. Most of the cytokine and chemokine genes (Il6, Il1a, Il1b, Tnfa, Cxcl10, Csf3) were significantly upregulated in BMDMs co-cultured with L. salivarius UCC118 in comparison to NT (non-treated) BMDMs (Fig. 7a). Among the PRR genes whose expression was upregulated, Clec4e coding for Mincle showed the greatest increased fold change, followed in rank order by the PRRs-Tlr1 and Tlr2 in L. salivarius UCC118 co-cultured with BMDMs. Expression of all the other TLRs (Tlr3, Tlr4, Tlr7, Tlr8 and Tlr9) were downregulated while Tlr6 expression did not vary between the groups in these BMDMs (Fig. 7a). Upregulation of Tlr1, Tlr2, Clec4e, Il6, Tnfa and Il1b was confirmed separately at 4, 8, 12 and 20 h by separate RT-qPCR experiments (Fig. 7b). Because Clec4e was the most upregulated PRR in BMDMs co-cultured with L. salivarius UCC118 we then investigated L. salivarius UCC118 induced cytokine responses in BMDMs from WT and Clec4e −/− mice. Absence of Clec4e expression in Clec4e −/− BMDMs was www.nature.com/scientificreports/  www.nature.com/scientificreports/ confirmed by RT-qPCR (Fig. 8a). Heat-killed Mycobacterium tuberculosis (HKMT), a reported Mincle and TLR2 agonist 45,46 , was used a positive control for TNF-α response triggered by Clec4e/Mincle in BMDMs. TNF-α response to HKMT was reduced in Clec4e −/− macrophages as expected (Fig. 8b). TNF-α responses triggered by L. salivarius UCC118 were also partially reduced in Clec4e −/− BMDMs in comparison to BMDMs from co-housed WT littermates (Fig. 8b). LPS (TLR4 agonist) and B. breve showed no difference in TNF-α responses between WT and Clec4e −/− BMDMs (Fig. 8b). Interestingly, FSL1 (TLR2 agonist) also had reduced TNF-α responses in Clec4e −/− BMDMs (Fig. 8b), suggesting a possible role for Mincle in the regulation of TLR2 induced cytokine Statistical analysis was performed with 2 tailed student t test in GraphPad Prism, p < 0.05 (denoted by *) was considered statistically significant. ns nonsignificant.  Fig. S2) supporting our observation of upregulated expression of Tlr2 and Clec4e in L. salivarius UCC118 co-cultured BMDMs (Fig. 7). However, there was no evidence for a direct functional association between Mincle and TLR2 in either humans or mice, but putative homologs of Mincle (like CLEC18A, CLEC4M, CLEC7A) were found to interact with TLR2 in humans ( Supplementary Fig. S2). Overall, these results suggested that Clec4e or its downstream sigalling components may be potentially important contributors to TLR2 mediated induction of cytokine responses by TLR2 agonist ligands (like FSL-1 and HKLM) and to responses to L. salivarius UCC118.

Discussion
In this study we investigated the mechanisms underpinning the recognition of Gram-positive commensal bacterial species Lactobacillus salivarius by mouse macrophages and a human monocyte-like cell line. We found that macrophage cytokine responses to these bacteria were TLR2 independent yet completely MyD88 dependent and associated with the upregulation of Tlr1, Tlr2 and Clec4e PRR genes in these cells. Since MyD88 signalling was required for triggering of macrophage cytokine responses to L. salivarius, it is possible that other MyD88-dependent pathways such as TLR7 and TLR9 and/or the IL-1R signalling pathway might be important contributors to L. salivarius recognition and response by these cells 48 . In fact, it is known that other commensal Lactobacillus species such as L. reuteri can activate IFN signalling through TLR7 in plasmacytoid dendritic cells 49 and L. plantarum can induce elafin secretion via TLR9 in CaCO2 cells 50 . The involvement of MyD88 independent PRRs such as TLR3 51 or NLRs (NOD like receptors) 52 as the predominant players in transducing cytokine responses by these bacteria can be excluded given that L. salivarius trigger macrophage cytokine responses in a MyD88 dependent manner. In support of these observations, we found that while the stimulation of BMDMs with L. salivarius UCC118 downregulated the expression of Tlr3, Tlr4, Tlr7, Tlr8 and Tlr9, it upregulated Tlr1, Tlr2 and Ilr1 which signal in a MyD88 dependent manner. Interleukin-1 receptor (IL-1R) mediated signaling might be another contributing factor involved in L. salivarius induced cytokine responses. L. salivarius induced the expression of Il1a, Il1b and Ilr1 genes (Fig. 7) and secretion of IL1β cytokine when they were co-cultured with bone-marrow derived macrophages. MyD88 acts as an adaptor component in the IL-1R signaling pathway 53-55 . Thus, it is possible that IL-1α or/and IL1β secreted upon initial recognition and internalization of L. salivarius by macrophages might participate in the further induction of cytokine responses in an autocrine and paracrine manner by signalling through IL-1R in a Myd88 dependent manner. This would explain the reduced cytokine response observed in Myd88 −/− macrophages. In addition, stimulation of BMDMs with L. salivarius UCC118 was also able to induce an unexpected hyper-inflammatory cytokine response in TLR2 −/− BMDMs. This suggests a possible immunoregulatory role of TLR2 in modulating macrophage responses to L. salivarius. It is possible that in addition to its importance in host cells for the initiation of a protective inflammatory response against invading pathogens TLR2 might also have evolved an alternative immunoregulatory role or other protective function in the context of host-commensal interactions 56 . Indeed, bacterial components of the microbiota are able to www.nature.com/scientificreports/ directly downregulate some downstream MyD88 effector genes in zebra fish at steady-state through TLR2 57 . An immunoregulatory role for TLR2 has also been previously suggested in disease contexts such as in mouse models of arthritis 58 and colitis 59,60 . Although addressing the mechanistic basis and significance of this observed phenotype in TLR2 −/− BMDMs is beyond the scope of this study, this aspect should be further explored in the future. The purity of BMDMs and TLR2 status were validated by flow cytometry. The TLR2 function of BMDMs used in this study were validated using selective TLR2 agonists-Pam3csk4, FSL1 and HKLM. We observed some notable differences in cytokine responses to stimulation with whole bacteria versus individual MAMPs between BMDMs generated from TLR2 KO mice sourced from two different suppliers used in this study. These included, a diminished cytokine response against HKLM and abrogated KC responses against all treatments in TLR2 −/− BMDMs derived from the mice purchased from Oriental Bioservices, which could not be observed in BMDMs from the mice purchased from Jackson Labs. Nonetheless, the TLR2-dependent nature of the responses to individual TLR2-selective MAMPs, and to the commensal bacteria B. breve and the TLR2-independent nature of cytokine responses to L. salivarius UCC118 was consistently observed in BMDMs generated from both TLR2 −/− mouse strains sourced from these suppliers. This highlights the importance of carrying out similar in vitro and in vivo experiments in mice of different genetic backgrounds, breeding facilities, suppliers and/or vendors before making major conclusions of a particular phenotype as the observed effects may be affected by these confounding factors 61,62 .
The host has evolved multiple primary and secondary protective barrier-associated mechanisms to contain commensal bacteria within the gut lumen. Despite this, exposure of host innate immune cells to members of the microbiota through various combinations of MAMP-PRR interactions might still occur in the gut lamina propria or in systemic compartments in the context of pathologies such as inflammatory bowel disease (IBD). In this context, certain members of the gut microbiota may take advantage of the compromised gut barrier to translocate from the intestinal tract into the systemic circulation 63 . Defective TLR1 signaling by intestinal epithelial cells has been associated with disruption of intestinal homeostasis and increased inflammatory responses against microbiota 64 and defective MyD88 signaling is linked to reduced antibacterial responses and enhanced adherence and translocation of gut-associated bacteria from lumen to epithelial cells 29,65 . In this study we chose a widely studied gut-associated commensal bacteria: L. salivarius in order to understand the relative contribution of TLR2 to their recognition and subsequent triggering of NF-κB dependent cytokine responses in macrophages. The expression of Clec4e, Tlr1 and Tlr2 were significantly upregulated in BMDMs co-cultured with L. salivarius UCC118. Clec4e (also known as Mincle) is a C-type lectin receptor that is broadly known to recognize mannosyl fatty acids ligands associated with Mycobacterium tuberculosis bacteria and Candida albicans fungus and triggers intracellular signaling through its adaptor Syk (spleen tyrosine kinase) 34,66,67 . Clec4e PPI network retrieved from STRING database suggested co-expression of Clec4e and TLR2 as well as functional association between Clec4e homologs and TLR2 in both mice and humans. However, a recent study has reported direct interaction of Clec4e itself with TLR2 68 . Lipomannan and Lipoarabinomannan from Corynebacterium have been shown to upregulate cell surface Clec4e expression in macrophages through a TLR2-MyD88 pathway and Clec4e consequentially www.nature.com/scientificreports/ binds to Corynebacterium glycolipids to induce NF-кB activation and induction of inflammatory responses including nitrite and granulocyte colony stimulating factor (G-CSF) secretion thus supporting a co-operation between TLR2 and Clec4e to sense bacterial glycolipids 68 . Glycolipids derived from L. plantarum are also able to trigger transcription factor nuclear factor of activated T-cells (NFAT) in reporter cell lines through human and mouse mincle 69 . A recent study also shows that microbial interaction with Clec4e is important in triggering cytokine production by DCs (dendritic cells), for the promotion of intestinal barrier integrity and prevention of translocation of gut bacteria to systemic tissues at steady-state 70 . In addition, Lactobacillus sps. were recognised as having the highest binding affinity to Clec4e amongst other commensal bacteria reported in this study 70 . Our study also supports the possibility of Lactobacillus derived ligands binding to Clec4e since macrophage TNF-α responses to L. salivarius UCC118 were partially dependent on Clec4e. Clec4e has been recently reported to contribute to phagocytic function and to control intracellular growth of Mycobacterium tuberculosis by triggering autophagy 71 . Clec4e expression increased steadily from 4 to 24 h in BMDMs co-cultured with L. salivarius. Thus, it is also possible that Clec4e may act as phagocytic receptor to trigger internalization of L. salivarius by macrophages. In the future, it would be of interest to identify if there are glycolipids in L. salivarius similar to that in L. plantarum which could mediate cytokine responses directly through Clec4e and to investigate if Clec4e might have a role in the internalization of L. salivarius in addition to its importance in the induction of cytokine responses.
Our study also provides evidence of the differential requirements for individual PRR pathways in macrophage cytokine responses to Gram-positive commensal bacteria since B. breve required TLR2 to induce cytokine responses in macrophages but L. salivarius UCC118 did not even though both of these bacteria were Gram-positive. The difference in the induction of macrophage cytokine responses by these two bacteria maybe attributed to differences in L. salivarius cell surface architecture 72,73 . Collectively, our data supports a model whereby the response of host macrophages to L. salivarius is TLR2 and TLR4 independent but MyD88-dependent. Our data also suggests Mincle (Clec4e) as an important contributor to TLR2-dependent responses against TLR2 ligands as well as to TLR2-independent responses to bacteria like L. salivarius. These observations support a model whereby the integration of signals from different PRR pathways and MyD88-dependent pathways may ultimately determine immune responses to commensal bacteria at the host-microbe interface.

Materials and methods
Mice. All mice used in this study were 8-12 week old male mice from a C57BL/6 background. TLR2 −/− , TLR4 −/− , TLR2/4 −/− mice and WT (wildtype) controls were obtained from Oriental BioService (Kyoto, Japan) and TLR2 −/− , MyD88 −/− and WT controls were purchased from Jackson Laboratories (Bar Harbor, USA). All these animals were housed in the Biological Service Unit animal housing facility at University College Cork (UCC) under specific pathogen-free (SPF) conditions using individually ventilated cages (IVC). Standard housing and environmental conditions were maintained (temperature 21 °C, 12 h light and 12 h darkness with 50% humidity). Animals were fed regular chow food (purchased from Envigo (Cambridgeshire, UK). Animals were also given water ad libitum. WT and Clec4e −/− mice were littermates born from heterozygous parents and housed in the animal housing facility at Centro Nacional de Investigaciones Cardiovasculares Carlos III (CNIC) until taken for bone-marrow isolation. Mice at CNIC were housed under SPF conditions using IVC. Standard housing and environmental conditions were maintained (temperature 21 °C, 12 h light and 12 h darkness with 50% humidity). Animals were fed with chow diet (LASQC diet, Altromin international, Lage, Germany). Animals were also given water ad libitum. For in vitro analysis of cytokine responses bone-marrow derived macrophages (BMDMs) were generated from WT, TLR2 −/− , TLR4 −/− , TLR2/4 −/− and Clec4e −/− mice (n = 3, age: 8-12 weeks, all males), and experiments were repeated a total of three times (n = 3) in technical triplicates per experiment. For gene expression analysis, RNA was harvested from WT or Clec4e −/− (n = 3) BMDMs that were either nontreated (NT) or L. salivarius UCC118 co-cultured and experiments were repeated a total of three times (n = 3) with technical duplicates per experiment. All animal work in CNIC was approved by the local animal ethics committee. Bacterial strains and culture conditions. Lactobacillus salivarius and other bacterial strains used in this study and their sources are listed in Table 1. Lactobacillus strains were routinely cultured at 37 °C under micro-aerobic conditions (5% CO2) in de Man Rogosa-Sharpe medium (MRS, Difco). Bifidobacterium breve UCC2003 was cultured in reinforced clostridium medium (RCM, Sigma-Aldrich) for 10 h, and subcultured in MRS supplemented with 1% cysteine overnight at 37 °C in anaerobic conditions. E. coli EC101 was grown aerobically in Luria-Bertani broth (LB, Sigma-Aldrich) at 37 °C and 200 rpm. Multiplicity of infection (MOI) values of these bacteria for co-culture experiments were calculated by measuring their cell numbers corresponding to their optical density values at 600 nm (OD 600 ) after overnight (16 h)  Flow cytometry analysis. BMDM cells were washed twice in PBS supplemented with 1% bovine serum albumin (BSA) and 0.1% sodium azide. Nonspecific binding of antibodies (Abs) to Fc receptors was blocked by pre-incubation of cells with monoclonal Abs (mAb) 2.4G2 directed against the FcgRIII/II CD16/CD32 (0.5 ng mAb per 10 6 cells). 1 × 10 6 cells were incubated with 0.5 ng of the relevant mAb for 20 min at 4 °C, and washed again twice. The following mAbs from ebiosicence were used: APC-conjugated mAb binding CD45, PECy-7-conjugated mAb binding CD11b, FITC-conjugated mAb binding F4/80 and PE-conjugated mAb binding TLR2. Cells were analysed using BD Accuri C6 or BD FACSCalibur. Data were analysed using FCS Express V5 Flow Cytometry software (Copyright De Novo Software 2017).
Heat map generation of cytokine responses against bacterial strains. The log of the median of cytokine concentration (pg/ml) from three independent experiments for different bacterial strains was visualized in the form of heatmap (Fig. 1). The heatmap was generated using ComplexHeatMap R package 75 . The gradient of the heat maps generated goes from blue to red (on a range of − 5 to + 15) to depict the low, intermediate, www.nature.com/scientificreports/