Lacticaseibacillus casei AMBR2 modulates the epithelial barrier function and immune response in a donor-derived nasal microbiota manner

Live biotherapeutic products (LBP) are emerging as alternative treatment strategies for chronic rhinosinusitis. The selection of interesting candidate LBPs often involves model systems that do not include the polymicrobial background (i.e. the host microbiota) in which they will be introduced. Here, we performed a screening in a simplified model system of upper respiratory epithelium to assess the effect of nasal microbiota composition on the ability to attach and grow of a potential LBP, Lacticaseibacillus casei AMBR2, in this polymicrobial background. After selecting the most permissive and least permissive donor, L. casei AMBR2 colonisation in their respective polymicrobial backgrounds was assessed in more physiologically relevant model systems. We examined cytotoxicity, epithelial barrier function, and cytokine secretion, as well as bacterial cell density and phenotypic diversity in differentiated airway epithelium based models, with or without macrophage-like cells. L. casei AMBR2 could colonize in the presence of both selected donor microbiota and increased epithelial barrier resistance in presence of donor-derived nasal bacteria, as well as anti-inflammatory cytokine secretion in the presence of macrophage-like cells. This study highlights the potential of L. casei AMBR2 as LBP and the necessity to employ physiologically relevant model systems to investigate host–microbe interaction in LBP research.

and patients are therefore still often treated with antibiotics, despite the lack of high-quality scientific evidence supporting the use of systemic or topical antibiotics in CRS 17,21,22 and the fact that antibiotic treatment should only be considered in specific cases of CRS 20 . This inappropriate use of antibiotics may increase the spread of antibiotic resistance genes and can result in adverse side effects in patients, such as gastrointestinal discomfort and skin rashes (reviewed in 21 ). In recent years, various studies on the association between the sinonasal microbiome and CRS etiopathology and persistence have revealed new insights in the role of bacteria in CRS 11,[23][24][25][26][27][28][29] . A number of pathogenic bacterial species has been associated with CRS exacerbations and persistence, notably Staphylococcus aureus, Pseudomonas aeruginosa, Streptococcus pneumoniae, and Haemophilus influenzae [29][30][31][32] . It is not straightforward to summarise the findings of the different studies, however, most studies point towards an overall dysfunctional interaction between the sinonasal epithelium and its resident microbiota. Restoring this dysfunctional interaction is thus a major objective to restore host health 27,33,34 . The use of probiotics or live biotherapeutic products (LBPs) has been proposed as a possible strategy to restore the balanced interaction between host and microbiota in the sinonasal cavities by reinforcing beneficial bacteria of the microbiota, strengthen the epithelial barrier, and reduce inflammation 34 .
LBPs are biological products that: (1) contain live organisms, such as bacteria; (2) are applicable to the prevention, treatment, or cure of a disease or condition of human beings; and (3) are not vaccines 35 . They form a special class of probiotics, which are defined as live microorganisms that, when applied in adequate amounts, confer a health benefit to the host. They are an emerging alternative strategy to treat upper respiratory tract diseases 33,34,[36][37][38][39][40] . Their mechanisms of action can be divided in the exclusion or inhibition of pathogens, improvement of epithelial barrier function, and modulation of local or systemic host immune responses 34 . De Boeck et al. 41 have demonstrated that certain lactobacilli are reduced in the nasopharynx of CRS patients compared to healthy controls. One particular Lacticaseibacillus casei strain, the species formerly known as Lactobacillus casei 42 , strain AMBR2, was isolated from the URT of a healthy volunteer. This strain was shown to express fimbriae enabling strong adherence to URT epithelium and inhibit the growth and virulence of several URT pathogens. These adaptive traits that enable survival and colonisation in the human URT could potentially increase the efficacy of L. casei AMBR2 as live biotherapeutic product. Indeed, an explorative first in-human study showed that the strain could persist in the nose of healthy volunteers up to 2 weeks after the last administration, thus longer than the nasal clearance time of 15 min 41 , but a detailed study on the colonisation dynamics of this strain has not yet been performed.
The screening and efficacy testing of potential LBPs for the respiratory tract generally starts in cell culture models as pure strains 41 and 43 , and/or in simple competition experiments with pathogenic species 40,41,44,45 . However, one of the first biological barriers for the successful engraftment of an LBP is formed by the resident microbiota 46 . When nasally applied as a (prophylactic) treatment, these strains will be introduced in an environment colonised by an existing microbial community, with interindividual differences in composition 11,[47][48][49][50] , microbial load 23,51 , and, especially in case of a pathogen-dominated dysbiosis, potentially permissiveness to engraftment [52][53][54] . In the gut, the mucosal colonisation permissiveness of probiotics has been shown to be highly person-specific, with both the microbiota and the immune responses determining successful colonisation 55,56 . In order to better predict the efficacy and responders of LBPs, it is important to assess the potency of novel LBPs in a microbial community that is more representative for the one that will be encountered upon administration, including interindividual variability. Few in vitro host-microbe interaction models of the sinonasal cavities include natural nasal microbiota 57,58 .
We have previously developed dual and triple co-culture in vitro model systems to mimic and study host-microbe interactions in the human URT microenvironment for 72 h using models containing respectively differentiated Calu-3 respiratory epithelial cells apically inoculated with donor-derived microbiota (dual co-culture), and the former combination with the addition of THP-1 derived macrophage-like cells (triple coculture) 58 . The differentiation of a Calu-3 cell layer at the air-liquid interface (ALI) results in an epithelial layer with functioning epithelial barrier, ciliated cells and mucus-producing goblet cells [59][60][61][62] . The inclusion of THP-1 derived macrophage-like cells allows to study crosstalk between epithelial cells, bacteria, and macrophages 63 .
In this study, we examined the colonisation potential and host-microbe interaction of L. casei AMBR2 in donor-derived polymicrobial background, as presented in Fig. 1. In an initial screening, we assessed growth and attachment of L. casei AMBR2 on submerged respiratory epithelial monolayers in the presence of nasal microbial communities obtained from 10 different healthy donors. Two donors were selected from this group based on the ability of AMBR2 to grow in their nasal communities. A sample of their nasal microbiota was used to inoculate dual and triple co-culture models, which were then treated with L. casei AMBR2. We evaluated the colonisation behaviour of L. casei AMBR2 and the host responses in these complex model systems, and the manner in which the inclusion of immune cells impacted bacterial and epithelial behaviour.

Results
Donor screening. L. casei AMBR2 survival and growth in polymicrobial and host cell background. 24 h after inoculation, we separately sampled bacteria that were adhered or non-adhered to Calu-3 respiratory epithelial cell monolayers, as adherence to the epithelial cell layer is an important trait to avoid rapid mucociliary clearance of LBP in the URT. Adhered and non-adhered bacteria were plated on MRS agar to determine the amount of cultivable lactic acid bacteria in the samples. No colonies were found on plates with samples from nasal microbiota without probiotic, or on the sterile control group. Growth occurred in all samples from wells inoculated with L. casei AMBR2, as displayed in Fig. 2a. We found no significant differences in growth in polymicrobial backgrounds from the different donors. The number of adhered lactic acid bacteria ranged from 10 4 CFU mL −1 to 5.10 × 10 6 CFU mL −1 with an average of 1.89 × 10 6 CFU mL −1 ± 1.39 × 10 6 CFU mL −1 (standard deviation), 11.06% of the total (adhered + non-adhered) colony forming community on MRS agar. Non-adhered Scientific Reports | (2020) 10:16939 | https://doi.org/10.1038/s41598-020-73857-9 www.nature.com/scientificreports/ lactic acid bacteria ranged from 2.80 × 10 6 CFU mL −1 to 2.73 × 10 7 CFU mL −1 with an average of 1.52 × 10 7 CFU mL −1 ± 6.20 × 10 6 CFU mL −1 , 88.94% of the total colony forming community. In samples with L. casei AMBR2 without polymicrobial background, lactic acid bacteria grew to an average of 1.60 × 10 6 CFU mL −1 ± 1.80 × 10 6 C FU mL −1 adhered cells (12.5% of the total colony forming community) and 1.12 × 10 7 CFU mL −1 ± 5.94 × 10 7 CF U mL −1 non-adhered cells (87.5% of the total colony forming community) in the medium. Donor 3 and donor 7 were selected for further testing based on having respectively the lowest and the highest average attached colony counts on MRS agar. We used Random Forest classifiers to further estimate the number of intact L. casei AMBR2 cells in the adhered and non-adhered polymicrobial donor-derived communities (Figs. 2b, 3), which corresponded to the lactic acid bacteria count we observed using selective plating. This indicated that L. casei AMBR2 could survive or grow in all donor backgrounds, except for donor 3, where a decrease in non-adhered intact cell counts compared to the inoculum was observed. Incubation with donor 3 derived microbiota resulted in significantly less estimated non-adhered intact AMBR2 cells (K-W, p = 0.03; Wilcoxon tests, p = 0.05) than with other donor backgrounds. This trend was observed, yet non-significant, in non-adhered microbial communities as well.
Bacterial growth and phenotypic diversity. To explore the microbial community factors that could account for differences in facilitating the growth or survival of L. casei AMBR2, we evaluated bacterial cell density and phenotypic diversity in the nasal swab communities used to inoculate the epithelial cell monolayers during the donor screening. The bacterial cell density in the swab samples obtained from the donor's nasal cavities, is displayed in Fig. 4a. While average intact cell density in the nasal samples was 2.69 × 10 5 intact cells mL −1 ± 1.98 × 10 5 cells mL −1 , donor 3 displayed a bacterial density of 3.98 × 10 3 intact cells mL −1 ± 1.19 × 10 3 cells mL −1 , which was lower (p = 0.05, Wilcoxon test) than all other donors, except for donor 9. Donor derived inocula were brought to the same concentration prior to inoculation to evaluate the effect of community composition rather than cell density. Phenotypic diversity was calculated based on intact bacterial cell flow cytometric data (10 3 events/sam- www.nature.com/scientificreports/ ple), and is displayed in Fig. 4b. Intact bacterial cell densities in samples from donor 3 were too low to calculate phenotypic diversity. The density of intact bacteria adhered to the Calu-3 monolayers after 24 h of co-culture with or without L. casei AMBR2 was measured using flow cytometry (Fig. 5). On average, bacterial cell densities with AMBR2 were 2.46 × 10 6 intact cells mL −1 ± 2.54 × 10 6 cells mL −1 . Adhered intact cell densities without L. casei AMBR2 were on average 9.75 × 10 5 intact cells mL −1 ± 1.29 × 10 6 intact cells mL −1 , and showed large variation between donors.
Phenotypic β-diversity analysis on adhered and non-adhered microbial communities demonstrated that except for donor 3, microbial communities inoculated with L. casei AMBR2 clustered together (PCoA, Fig. 6a,b), in line with the Lacticaseibacillus-dominated communities observed after partial 16S rRNA gene sequencing (Supplementary Figure 1). Communities derived from donor 3 with or without L. casei AMBR2 clustered separately from all other communities.  www.nature.com/scientificreports/ Phylogenetic microbial community analysis. The relative abundances of the eight most abundant genera present in the nasal bacterial communities of the donors are displayed in Fig. 7a. Nasal swab bacterial community richness was estimated by the Chao1 index ( Fig. 7b) and the community alpha-diversity was calculated using the Inverse Simpson index (Fig. 7c). An overview of richness and diversity in the different sample types and treatment groups is given in Table 1. Richness was significantly increased in cultured samples, with and without AMBR2 (Wilcoxon test, p < 0.01), whereas diversity was decreased after 24 h of culture (Wilcoxon test, p < 0.001).
In the adhered bacterial communities, introduction of AMBR2 resulted in an increase in richness (p = 0.017, Wilcoxon test), but not diversity. In the non-adhered communities, introduction of AMBR2 did not result in significant differences in richness or diversity. Taxa from the Lactobacillus genus complex, likely L. casei AMBR2, were able to attach and survive in all donor communities; after 24 h Lactobacillus sp. had a relative abundance of 76.7% ± 34.3% (average ± standard deviation) in the non-adhered communities, and 70.7% ± 33.7% in the adhered communities in which it was introduced (Supplementary Figure 1). The consensus sequence of the most abundant Lactobacillus OTU, OTU2, was compared to the expanded Human Oral Microbiome Database (eHOMD 54 ), which indicated that it was Lactobacillus paracasei, reclassified as Lacticaseibacillus paracasei 42 (100% identity ), or Lactobacillus casei, reclassified as Lacticaseibacillus casei 42 (99.8% identity). In 7 out of 10 communities where AMBR2 was not  www.nature.com/scientificreports/ introduced, an overgrowth of Staphylococcus sp. occurred (OTU1). This genus (represented by OTU1) was also the second most abundant (after Lactobacillus sp.) in 8 out of 10 communities where AMBR2 was introduced.
Because Staphylococcus taxa can be important pathobionts in CRS 64,65 , we assessed whether there was a correlation between the initial relative abundance of Staphylococcus sp. and the relative abundance of Lactobacillus sp. after 24 h. Both Spearman (r S = − 0.68, p = 0.029, n = 10) and Pearson (r P = − 0.62, p = 0.054, n = 10) correlation indicated a negative relationship between the abundances of both genera. Comparison of the consensus sequence of the partial 16S rRNA gene with eHOMD 54 indicated OTU1 was Staphylococcus epidermidis (identity = 100%). A notable exception was the bacterial community of donor 3, where Lactobacillus sp. made up only 1.37% (non-adhered) and 2.06% (adhered) of the relative abundance 24 h after introduction of AMBR2, whereas unclassified Enterobacteriaceae (OTU4) accounted for 95.5% (non-adhered) and 94.4% (adhered) of the relative abundance in these communities. At the start of the experiment, donor 3 had 0.53% unclassified Enterobacteriaceae in their nasal community. Two other donors, 8 and 10, had respectively 0.04% and 0.29% unclassified Enterobacteriaceae (OTU4) in their nasal community. However, after 24 h, their communities were dominated by Staphylococcus sp. (without AMBR2) or Lactobacillus sp. (with AMBR2). We compared the consensus sequence of the partial 16S rRNA gene with eHOMD 54 , which indicated OTU4 was Klebsiella aerogenes (identity = 100%).

Colonisation of L. casei AMBR2 in dual and triple upper respiratory tract host-microbe interaction set-ups with selected donor microbial backgrounds. Cytotoxicity. Cytotoxic stress of Calu-3
or Calu-3 and THP-1 cells exposed to bacterial communities was assessed through lactate dehydrogenase (LDH) leakage to the medium in the basal compartment (Supplementary Figure 2). Cytotoxicity remained below 5% for all co-cultures and control groups throughout the experiment. The donor background had no significant effect on cytotoxicity, nor did treatment with L. casei AMBR2 alter LDH release. The presence of macrophages resulted in a small yet significant reduction in percentage cytotoxicity (ANOVA followed by Tukey HSD, p < 0.0001).
Transepithelial electrical resistance. The epithelial barrier function was measured throughout the experiment as transepithelial electrical resistance (TEER), as displayed in Fig. 8a. At the start of the experiment, the TEER of co-cultures without THP-1 derived macrophages was significantly higher than co-cultures with THP-1 derived macrophages, while after 48 h, the opposite occurred (Wilcoxon Rank Sum test). There were no statistically significant differences in TEER between dual and triple co-cultures 24 h and 72 h after the start of the incubation. 72 h after the start of the experiment, cell layers treated with L. casei AMBR2 had a higher epithelial resistance than cell layers without the probiotic (Wilcoxon Rank Sum test). Looking at the different donor backgrounds, we observed that dual and triple co-culture systems inoculated with donor microbiota from donor 7 treated with L. casei AMBR2 had higher epithelial resistances than their counterparts without AMBR2 (K-W). For donor 3, this effect of AMBR2 only occurred in dual co-culture (K-W). Without donor background, no such effect of AMBR2 treatment was observed (K-W).
Cytokine secretion. We evaluated release of pro-and anti-inflammatory cytokines in response to the nasal microbiota and L. casei AMBR2 to investigate the effect of L. casei AMBR2 on inflammation or immune surveillance. Incubation with L. casei AMBR2 had an impact on the IL-8 release in both model systems and for both donor backgrounds (K-W, p = 0.001335, Fig. 8b). In dual co-culture, IL-8 release in AMBR2 treated groups showed to be slightly but significantly lower after 24 h than in non-treated groups (KW, p = 0.00919). From 48 h after inoculation on, AMBR2 treated cell cultures released significantly more IL-8 than non-treated cell cultures (K-W, p = 0.00035 for both model systems at 48 h and 72 h). While this increase in IL-8 occurred in both model systems, the presence of macrophages was a determining factor of the IL-8 release profile after AMBR2 treatment. In dual co-culture systems, the concentration of IL-8 was significantly higher (K-W) than in triple co-culture systems, p = 0.00035) at 48 h, while this effect disappeared at 72 h (K-W). The donor-derived polymicrobial background was a less determining factor and only resulted in a significant difference in dual co-cultures treated with L. casei AMBR2 72 h after the start of the experiment (K-W, post hoc non-significant).
IL-10 concentrations in the basolateral medium were below the quantification limit, except in co-cultures containing macrophages after 72 h of co-culture (Fig. 8c). In co-cultures including donor-derived microbial communities, treatment with L. casei AMBR2 increased IL-10 release in a noticeable, yet non-significant manner, whereas there was no difference between the sterile control group and cell layers treated with AMBR2 alone. www.nature.com/scientificreports/ TNF-α release in the basolateral compartment was below the quantification limit for all groups on time points 0-48 h; therefore, only the 72 h time point is shown (Fig. 8d). After 72 h of co-culture, set-ups containing macrophage-like cells had higher concentrations of TNF-α than set-ups without macrophage-like cells (p = 0.014, K-W). Presence of L. casei AMBR2 elicited an increase in TNF-α release in triple co-culture (p = 0.038, K-W), independent of the donor background. In dual co-culture, this was observed as well (p = 0.0003, K-W), however, further inspection of the data showed that TNF-α release was only above quantification limit in cell layers treated with AMBR2 alone.
Bacterial growth and phenotypic diversity. Intact bacterial cell density in the apical washing fluid 72 h after the start of the experiment is displayed in Fig. 9a. L. casei AMBR2 displayed presence and growth on all cell layers on which it was inoculated (5.0 × 10 7 ± 1.9 × 10 7 intact cells mL −1 in dual co-culture, 6.2 × 10 7 ± 6.7 × 10 7 intact cells mL− −1 in triple co-culture, average ± standard deviation). Nasal bacteria derived from donor 3 showed limited growth in dual (1.7 × 10 5 ± 1.7 × 10 5 intact cells mL −1 ) and triple co-culture set-ups (2.1 × 10 5 ± 1.0 × 10 5 intact cells mL −1 ), contrarily to those derived from donor 7, which were able to grow in both model systems (4.0 × 10 7 ± 3.0 × 10 7 intact cells mL −1 in dual co-culture, 2.3 × 10 7 ± 1.8 × 10 7 intact cells mL −1 in triple co-culture) to similar cell densities as L. casei AMBR2 alone. In communities where AMBR2 was inoculated in a background of donor 7, cell densities were higher than the donor or AMBR2 alone, indicating growth of both inocula or a synergistic effect between the two (1.7 × 10 8 ± 6.1 × 10 7 intact cells mL −1 in dual co-culture, 2.1 × 10 8 ± 3.7 × 10 7 www.nature.com/scientificreports/ intact cells mL −1 in triple co-culture). The presence of macrophages did not significantly affect the intact cell density, except for donor 3 with AMBR2, where the intact cell count was lower in triple co-culture (Wilcoxon test, p = 0.022). The phenotypic alpha diversity (D2) is presented in Fig. 9b. For donor 3, phenotypic alpha diversity in both model systems was decreased in presence of L. casei AMBR2 (K-W, p < 0.001, followed by Dunn's test, p < 0.001), whereas the type of model system did not significantly affect phenotypic diversity with or without AMBR2. Phenotypic diversity decreased over time in donor 3 communities with AMBR2 (K-W, p < 0.001, Dunn's post hoc, p < 0.05), as well as donor 3 communities without AMBR2 (K-W, p = 0.012, Dunn's post hoc, p = 0.01 between 48 and 72 h). For donor 7, large variation in phenotypic diversity was observed between time points and samples in communities in which AMBR2 was not introduced and there was no significant decrease in phenotypic diversity over time. In both dual and triple co-culture, a decrease upon introduction of AMBR2 in these communities occurred from 24 to 48 h, whereas after 72 h phenotypic diversity had increased again to the level it was at 24 h (K-W, p < 0.001, Dunn's post hoc, p < 0.001).
We therefore assessed whether loss of phenotypic evenness was a driver for phenotypic diversity loss upon introduction of AMBR2 (Fig. 9c). Without AMBR2, donor 3 had a more even community than donor 7 in dual (Wilcoxon test, p < 0.01) and triple co-culture systems (Wilcoxon test, p < 0.01), whereas with AMBR2, donor 7 had a more even community than donor 3 with AMBR2 (Wilcoxon test, p = 0.02) in co-cultures without macrophage-like cells, as well as in co-cultures with macrophage-like cells (Wilcoxon test, p = 0.01). Donor communities in which AMBR2 was introduced had a significantly lower evenness than communities consisting of donor microbiota alone (K-W, p < 0.001), regardless of donor or model system.

Discussion
We examined the ability of the potential live biotherapeutic L. casei AMBR2 to grow on and attach to respiratory epithelial cell layers in the presence of donor-derived nasal microbiota from 10 different healthy donors. After the selection of two of these donors, the host-microbe interaction of AMBR2 was tested in more physiologically relevant host-microbe interaction models in the polymicrobial background of these donors.
During the donor screening, we observed that AMBR2 could adhere and survive or grow in all donor backgrounds, albeit with interindividual differences. Phylogenetic community analysis showed that all of our donors www.nature.com/scientificreports/ belonged to the intermixed Staphylococcus-Corynebacterium-Dolosigranulum diverse profile 66 , which represented 91% of the participants (n = 90) in a Belgian-Dutch cohort. They also belonged to the most prevalent nasal community state types found by Liu et al. 51 in a Danish cohort and by Lehtinen et al. 67 in a U.S. American cohort. Incubation of these nasal bacterial communities on submerged Calu-3 respiratory cell layers resulted in outgrowth of either Staphylococcus sp. (without AMBR2) or Lactobacillus sp. (with AMBR2), with the exception of donor 3, which had a community dominated by unclassified Enterobacteriaceae, regardless of presence of AMBR2, after 24 h. In our, and other similar, more complex model systems 57,58 , a consistent increase of Staphylococcus sp. in co-cultured communities compared to the original inoculum has been observed. The growth of Staphylococcus community members was favoured as well here in this simplified submerged monolayer set-up when no additional lactobacilli were introduced. We observed a negative correlation between the relative abundance of Staphylococcus spp. in the inoculum and the relative abundance of Lactobacillus spp. 24 h after introduction of L. casei AMBR2 (partial 16S rRNA gene sequencing). However, in 8 out of 9 of these communities Lactobacillus sp. relatively represented > 75% of the community as determined by partial 16S rRNA gene sequencing and flow cytometry. Furthermore, through selective plating and flow cytometry, we detected increased lactic acid bacteria and predicted AMBR2 cell counts after 24 h compared to the inoculum, indicating growth in 8 out of 10 donor backgrounds. This colonisation success can partly be explained by the high inoculation density of AMBR2 relative to the nasal microbiota, as it has been observed that absolute abundance is a driving factor in nasal bacterial interactions 51 .
Most striking was the bacterial community of donor 3, who had relative abundances of > 90% of an OTU from the genus Enterobacteriaceae sp., presumably Klebsiella aerogenes (eHOMD 54 ), with or without introduction of AMBR2. This OTU was present in other nasal communities, yet no overgrowth was observed in other donors and no conclusions could be made on whether this was the same species or strain. Klebsiella aerogenes, previously known as Enterobacter aerogenes, is an opportunistic pathogen, which had a prevalence of 2.6% in a healthy German study population (n = 1878) 68 . Interestingly, they showed that significant differences in the co-colonisation pattern of S. aureus and enterobacteria existed, with individuals being colonised with S. aureus having less frequent colonisation of Enterobacteriaceae. Donor 3 was characterised by lower intact bacterial cell counts in their nasal swab than the other donors (10 4 cells mL −1 vs. between 10 5 and 10 6 cells mL −1 for all other donors); the fact that the inocula of other donors required more dilution at the start of the experiment, could have stochastically decreased the relative abundance of Enterobacteriaceae and thus have reduced the potential for Enterobacteriaceae sp. overgrowth in the other donors in which it occurred.
The nasal communities included in our screening assay represented the most commonly observed community types in the nose 51,66,67 . Future studies would benefit from including individuals with other community types, such as those dominated by e.g. Moraxella or Enterobacteriaceae, and a broader age range and background to better reflect the general population and elucidate bacterial community patterns associated with degree of colonisation permissiveness.
When applied in dual and triple co-culture systems, we observed that nor the donor communities, nor AMBR2, induced a cytotoxic response during the course of the experiment. In presence of a polymicrobial background, L. casei AMBR2 increased the epithelial resistance of differentiated epithelial cell layers, whereas this did not occur without donor-derived microbiota. The presence of donor-derived bacteria without L. casei AMBR2 resulted in a slight decrease in epithelial resistance, which was restored through the introduction of AMBR2. TLR-2 (Toll-like Receptor-2) ligands peptidoglycan and the synthetic bacterial lipopeptide Pam3CSK4 were previously found to increase epithelial barrier functionality in a Calu-3 cell-based setup through upregulation of tight junction proteins claudin-1 and zonula occludens 1 69 . Martens and colleagues (submitted, 2020) have demonstrated that L. casei AMBR2 could increase epithelial barrier function in primary cell layers derived from CRSwNP (CRS with nasal polyps) patients via upregulation of the same tight junction proteins, presumably through a TLR-2 mediated mechanism. This mechanism remains however to be substantiated, as well as the responsible microbe associated molecular patterns associated with TLR2.
Throughout the experiment, pro-inflammatory IL-8 release remained in the range of sterile controls in similar set-ups 70,71 , or on Calu-3 monolayers treated with AMBR2 41 , and was six-to tenfold lower than a similar set-up treated with 10 7 Staphylococcus aureus cells mL −171 . The introduction of AMBR2 in the dual and triple co-culture systems elicited an increase in IL-8 release. We observed a difference in IL-8 release profile dependent on the presence or absence of THP-1 derived macrophages. Without THP-1 derived macrophages, a peak in IL-8 release was observed after 48 h, whereas this did not occur in the triple co-culture systems containing THP-1-derived macrophages. Several probiotic lactobacilli, living or heat-killed, can also induce IL-10 release in THP-1 derived macrophages in direct contact 72,73 . We observed an increase in IL-10 release in triple co-culture systems, albeit in 100-fold lower concentrations than observed by Rocha-Ramirez et al. 72 , possibly due to the lack of direct contact between macrophages and bacterial cell wall compounds, and their use of primary human monocyte derived macrophages instead of the THP-1 cell line. In mice, remote induction of IL-10 by Lactobacillus plantarum, reclassified as Lactiplantibacillus plantarum 42 in the gut could reduce the induction of IL-8 after nasal infection with Klebsiella pneumoniae 74 , a mechanism which might possibly explain the reduced IL-8 release in our triple co-culture model system. This underlines the importance of incorporating immune cells as well as epithelial cells to study host-microbe interactions in the URT.
L. casei AMBR2 could grow in model systems based on differentiated airway epithelial structures, regardless of the presence of natural nasal microbiota or macrophages. Regrowth occurred after daily washes to mimic the nasal fluid flow, indicating adherence of AMBR2 to the respiratory cell layer. While AMBR2 colonisation in the microbial background of donor 3 was reduced compared to AMBR2 alone or in presence of donor 7 in the simplified screening set-up, this did not occur in the more complex dual and triple co-culture set-ups, despite a relative abundance of 0.29% unclassified Enterobacteriaceae in the inoculum of donor 3. De Boeck et al. 41 showed that AMBR2 has superior adaptive properties for survival in the URT, such as fimbriae for attachment, and the www.nature.com/scientificreports/ ability to cope with oxidative stress through the presence of catalase genes. The dual and triple co-culture setups mimic the microenvironment of sinonasal cavities more closely than a submerged monolayer in terms of nutrient availability 75 , host response 76 , and growth surface 61,62 . Liu et al. 6 demonstrated that S. epidermidis could exclude opportunistic pathogens through host-mediated competition by the induction of antimicrobial peptides. Furthermore, L. casei AMBR2 has antimicrobial activity against staphylococci and Enterobacteriaceae 41 . These factors may explain the lack of Enterobacteriaceae overgrowth in dual and triple co-culture systems inoculated with nasal bacteria from donor 3 compared to the submerged monolayer co-culture, as well as a similar growth of AMBR2 in donor backgrounds in these more relevant model systems.
Major limitations of this study are the use of cancer cell lines, the limited number of cell types tested, and the low number of individuals used for nasal microbiota sampling. A first major disadvantage of our study is the small number of individuals used for nasal microbiota sampling, their narrow age range, the homogeneity of their environment, and their high interindividual variability. We did not fully cover the variability observed in the population, therefore further validation studies require a larger and more heterogeneous sample group. Next to this, the incorporation of the microbiota of a larger and more heterogeneous sample cohort in our model system can result in improved understanding of the mechanisms driving bacterial community evolution upon introduction in the in vitro environment of the model system and elucidate possible differences between community types. Due to the existence of bacterial community state types, a preselection of this sample cohort would be appropriate in order to cover sufficient variability.
Our screening and colonisation assays were performed using nasal microbiota obtained from clinically healthy individuals. In order to assess the factors driving colonisation potential and probiotic potency in a polymicrobial background more representative of the CRS environment, CRS patient-derived sinonasal communities should be included, stratified according to CRS sub-or endotypes 10,26,77 . Furthermore, the respiratory epithelial dual and triple co-culture systems are based on non-inflamed epithelial cell layers with intact tight junctions. A dysfunctional epithelial barrier is characteristic for CRSwNP sinonasal epithelium [78][79][80] , not only resulting in increased potential of microorganisms crossing the epithelial barrier, but also leakage of nutrients into the sinonasal cavity 75 . Furthermore, mucus production and oxygen concentrations can be affected in CRS, potentially steering the sinonasal microbiota towards dysbiosis [81][82][83] . Mechanistic studies of probiotic potency should therefore not only include CRS-patient derived microbiota, but also an epithelial microenvironment mimicking that found in CRS.
Our goal was to develop a co-culture model system for the human upper respiratory tract with relevant readouts for responses to microbial colonisation and inflammation. We have chosen cancer cell lines for their reproducible behaviour, as a generic model system where the interindividual variability only originates from the bacterial community, rather than from both host and bacterial model constituents. We have chosen the Calu-3 cell line for its potential to be differentiated at air-liquid interface, mucin production, and establishment of tight junctions. Next to this, the behaviour of this cell line is well characterised [59][60][61][62]84 . Other (nasal) respiratory cell lines have limitations for use in this type of model system due to forming leaky epithelium and/or not differentiating at ALI, for example NuLi-1 cells 85 , which is not representative for healthy nasal epithelium. BEAS-2B cells, a human bronchial epithelial cell line, are frequently used in URT research, fail to establish tight junctions and do not form pseudostratified epithelium upon differentiation, both of which are essential in the sinonasal setting 60 . 16HBE cells, another human bronchial epithelial cell line, form tight junctions, yet have lower mucin production than their Calu-3 counterpart, rendering these cells suboptimal to establish a sinonasal model system 86 . It should however be noted that Calu-3 cells are limited in reflecting the responses to inflammatory stimulation observed in primary cells 70 . Recently, advances have been made in the immortalisation and differentiation of primary sinonasal epithelial cells 57 and in the generation of upper respiratory tract cells from pluripotent stem cells 87 . These cell types can be more representative than the cancer cell line used in this model system.

Conclusion
L. casei AMBR2, a potential live biotherapeutic for the human upper respiratory tract, is able to attach to and survive on a respiratory epithelial monolayer in the presence of donor-derived nasal microbiota, representative for the most prevalent community type in a healthy population. While attachment and survival of AMBR2 occurred in all donor-derived communities, its growth was severely inhibited in one case of Enterobacteriaceae overgrowth. This type of overgrowth did not occur in differentiated respiratory epithelial cell layers at the air-liquid interface, emphasising the importance of employing physiologically representative model systems in host-microbe interaction research. Introduction of AMBR2 in dual and triple co-culture systems did not induce cytotoxicity and could inhibit the reduction in epithelial barrier function caused by the nasal community. Inclusion of macrophage-like cells showed that introduction of AMBR2 could restore the release of the anti-inflammatory cytokine IL-10 in a polymicrobial background. This underlines the potential potency of L. casei AMBR2 as a live biotherapeutic product in the polymicrobial background in which it will be introduced.

Materials and methods
Human cell culture. Calu www.nature.com/scientificreports/ Aldrich) to assess viability and counted using a Neubauer counting chamber. Cells were seeded at density of 3.95 × 10 4 cells cm −2 , and maintained until confluent. 7 days prior to the bacterial inoculation the medium was switched to FBSi supplemented medium without antibiotic/antimycotic solution.
In this assay, dual and triple co-culture models were used 58 . In the dual co-culture model, Calu-3 cells were combined with bacteria, while in the triple co-culture model, Calu-3 cells were combined with bacteria and macrophage-like cells. Calu-3 cells (passage 9) were seeded at a density of 10 5 cells cm −2 in 6-well Transwell inserts (4.67 cm 2 culture surface, Corning, Sigma-Aldrich), and maintained under submerged conditions until fully confluent. When confluent, the cell layers were brought to air-liquid interface (ALI) for differentiation for 21 days. 7 days prior to the bacterial inoculation the medium was switched to FBSi supplemented medium without antibiotic/antimycotic solution. THP-1 cells (ATCC TIB-202, passage [43][44][45][46][47][48][49][50] were seeded in 6-well plates at a density of 10 5 cells/well in RPMI 1640 medium with Glutamax, supplemented with 10% FBSi. Monocytes were differentiated into macrophages using phorbol-12-myristate-12-acetate (PMA, Sigma-Aldrich) at a concentration of 100 nM for 72 h. THP-1 derived macrophages and Calu-3 cell layers were combined prior to bacterial inoculation by placing the inserts with differentiated Calu-3 cell layers in the 6-well plate containing THP-1 derived macrophages.
Donor recruitment and sampling. Nasal swabs were collected from 10 different healthy donors, aged between 24 and 32 (27 ± 2.5 years) and 80% female/20% male). Donor characteristics are included in Supplementary Table 1. Inclusion criteria were: clinical health, never-smoker, aged between 18 and 65, no asthma or atopy, no antibiotic use in the past 4 weeks, no use of topical corticosteroids in the past 4 weeks and no history of chronic or recurrent respiratory illnesses such as (chronic) rhinosinusitis, chronic obstructive pulmonary disease or aspirin exacerbated respiratory disease. Research incubation work with nasal microbiota from human origin was approved by the ethical committee of the Ghent University hospital under registration number B670201214538 and the research was performed in accordance with the relevant guidelines and regulations. Informed consent was obtained from all participants. Two nasal swabs (Flocked Swab, sterile, Copan, VWR, Oud-Heverlee, Belgium) were taken from the posterior nose and care was taken to avoid contamination from the nostrils. One swab was stored at − 20 °C until further analysis. The other swab tip was placed in an Eppendorf tube in sterile phosphate buffered saline (PBS) and placed on a Vortex shaker during 60 s. The tip was then removed from the tube and dilutions were prepared for flow cytometry and inoculation of the respiratory tract cell layers.
For dual and triple co-culture, two donors were selected from the donor pool of the screening assay, based on the ability of L. casei AMBR2 to grow in their microbial background. Donor 3 and donor 7 were used, since they displayed respectively the lowest and the highest growth of L. casei AMBR2 in the polymicrobial background of their nasal swab. Sampling was performed as described earlier.
Bacterial culture. L. casei AMBR2 41 was retrieved from a − 80 °C glycerol stock and transferred to MRS agar (De Man, Rogosa, and Sharpe agar, Carl Roth, Karlsruhe, Germany), 37 °C. A colony was picked and transferred to MRS broth (Carl Roth), 37 °C for overnight incubation. Prior to determination of the concentration using flow cytometry, and inoculation on the cell layers, the bacterial culture was washed twice (5 min at 5000×g) using sterile PBS, and afterwards suspended in sterile PBS.
Colonisation of L. casei AMBR2 in upper respiratory tract host-microbe interaction set-ups. Donor screening assay. For each donor (n = 10), Calu-3 layers were inoculated with donor microbiota only (n = 3), or with donor microbiota and L. casei AMBR2 (biological replicates, n = 3). Sterile controls (n = 3), and L. casei AMBR2 only (n = 3) wells were included. First, a suspension of 10 3 intact nasal bacterial cells in PBS was added to the Calu-3 cell layer. Two hours after the inoculation of nasal bacteria, 100 µL of 10 8 intact cells mL −1 PBS of L. casei AMBR2 was introduced in the nasal community. In sterile wells, or wells with donor communities, but without L. casei AMBR2, an equal amount of PBS was added. The cell layers with microbial communities were incubated for 24 h at 37 °C, 10% CO 2 , and 90% relative humidity. Bacteria that were not strongly adhered to the Calu-3 cell layer were sampled by pipetting up and down and removing the medium from the cell layers, and gently pipetting sterile PBS to remove the remaining non-adhered bacteria in an additional washing step. The cell layers were then incubated with 0.2% v/v Triton X-100 in PBS to detach the bacteria that were bound to the cell layers (adhered bacteria). Samples were taken for flow cytometry, DNA extraction, and plating on MRS agar. Microbial suspensions that were plated after sampling from the epithelial cell layers, were first diluted in sterile PBS, then plated on MRS agar and incubated at 37 °C for 48 h.
Colonisation of L. casei AMBR2 in dual and triple co-culture set-ups. In this assay, a comparison was made between the colonisation of L. casei AMBR2 in dual (Calu-3 + microbiota) and triple co-culture (Calu-3 + THP-1 + microbiota) set-ups inoculated with nasal bacteria from identical donors as described in De Rudder et al. 58 . Nasal microbiota samples were obtained from two healthy donors selected from the donor screening as earlier described, and for each donor we inoculated Calu-3 cell layers in co-culture with THP-1 cells (n = 6) and Calu-3 cell layers without THP-1 cells (n = 6) with 100 µL of 10 4 intact cells mL −1 . Two hours after the inoculation of nasal bacteria, 100 µL of 10 9 intact cells mL −1 PBS of L. casei AMBR2 was introduced in half of the nasal communities in dual (n = 3) or triple (n = 3) co-culture models. Sham inoculated cell layers (n = 3), and L. casei AMBR2 only inoculated cell layers (n = 3) were included for dual and triple co-culture set-ups.
Cytotoxicity. Cytotoxicity of the epithelial layers was assessed by quantifying the lactate dehydrogenase (LDH) leaked from the cell layer in the basolateral compartment using the Pierce LDH Cytotoxicity Assay Kit (Thermo Intact/damaged staining and flow cytometry. Intact bacterial cell densities were measured using flow cytometry, through SYBR Green/propidium iodide staining (SGPI), as described in Van Nevel et al. 88 . Samples were incubated for 13 min at 37 °C for SGPI staining. Flow cytometric analysis of the microbial cells present in the suspension was performed using a C6 + Accuri flow cytometer (BD Biosciences, Erembodegem, Belgium) equipped with a 488 nm laser, following previously described methods 89 . Fluorescence events were monitored using the fluorescence detector (FL)1 533/30 nm and FL3 > 670 nm optical detectors. Forward and sidewaysscattered light was collected as well. A threshold value of 1000 was applied on the FL1 channel. Data for the donor screening experiment are available on Flow Repository (FR-FCM-Z2JS).

Flow cytometric fingerprinting and classification of populations. Flow cytometric fingerprints
were calculated using the R package Phenoflow 90 . Fingerprints were constructed based on forward scatter, sideward scatter, SG fluorescence and PI fluorescence of the sample using 128 bins per channel (nbin = 128) and a bandwidth of 0.01 (bw = 0.01) for the kernel density estimator calculated at each bin. Fingerprint-based ecological parameters (alpha-and beta-diversity, Pareto-evenness) were calculated using the Diversity rf() function, the beta div fcm() function and the Evenness() function from the R package Phenoflow 90 . Phenotypic beta-diversity was calculated using Bray-Curtis distances. Principal Coordinates Analysis (PCoA) and non-metric multidimensional scaling (NMDS) was used to create ordination plots of the beta-diversity.
To estimate the intact cell density of donor-derived bacteria and L. casei AMBR2 in wells containing a combination of both, Random Forest classifiers were created using RandomF FCS(), with tenfold cross-validation (repeated 3 times). These classifiers were then applied to predict the intact cell counts of AMBR2 and donor bacteria in a mixed sample using RandomF predict(), both from the R package Phenoflow 90 . Models were trained and tested with flow cytometry data of L. casei AMBR2 alone and the respective donor community alone on 2000 events for the non-adhered communities (75% training, 25% testing). In samples with low bacterial cell densities, less events were used (Table S1-part 1). Accuracy and Cohen's Kappa statistic were evaluated for each individual model, and classifiers with accuracy ≥ 90% and a Kappa value ≥ 85% were used to estimate the events originating from either the donor microbiota or L. casei AMBR2. Accuracy and Kappa values for each classifier are presented in Table S2-part 1. Adhered communities had lower cell densities than non-adhered communities, therefore between 20 and 200 events could be used to train and test the classifiers (75% training, 25% testing). Accuracy and Cohen's Kappa statistic were evaluated for each individual model, and classifiers with accuracy ≥ 80% and a Kappa value ≥ 60% were used to estimate the events originating from either the donor microbiota or L. casei AMBR2 (Table S2-part 2). DNA extraction, partial 16S rRNA gene sequencing and data analysis. DNA was extracted from microbial communities as described in Vilchez-Vargas et al. 91 . Sequencing of the V3-V4 region of the 16S rRNA gene with forward primer (341F) and reverse primer (785R) was done on an Illumina MiSeq in PE 2X300 bp using V3 chemistry (Illumina) by Biofidal (Vaulx-en-Velin, France-https ://www.biofi dal-lab.com), as described in Supplementary Method (6).
The amplicon data were processed as described in 92 using mothur (v.1.39.5), largely based on the protocol by the Schloss lab 93,94 and described in more detail in Supplementary Method (6). Richness was estimated using the Chao1 index 95 and diversity was estimated the Inverse Simpson index. Samples from the dual and triple coculture set-ups yielded insufficient biomass for sequencing, therefore only the swab samples (t = 0 h, inoculum) are displayed. Data are available under the BioProject accession number PRJNA627613.
Statistical analysis. All statistical analyses were performed and graphs were created using R version 3.6.1 (The R Foundation for Statistical Computing). The normality of the data was assessed using the Shapiro-Wilk test, and homoscedasticity was assessed with the Bartlett test for equal variances. In case normality and homoscedasticity were rejected, the Kruskal-Wallis test (K-W) was used for multiple comparisons, followed by Scientific Reports | (2020) 10:16939 | https://doi.org/10.1038/s41598-020-73857-9 www.nature.com/scientificreports/ Dunn's post hoc test with Holm's correction or Wilcoxon Ranked Sum tests. When normality and homoscedasticity were not rejected, two-way ANOVA was performed, followed by Tukey's HSD post hoc test. Significance was considered at the 5% level (α = 0.05).