Rothia from the Human Nose Inhibit Moraxella catarrhalis Colonization with a Secreted Peptidoglycan Endopeptidase

ABSTRACT Moraxella catarrhalis is found almost exclusively within the human respiratory tract. This pathobiont is associated with ear infections and the development of respiratory illnesses, including allergies and asthma. Given the limited ecological distribution of M. catarrhalis, we hypothesized that we could leverage the nasal microbiomes of healthy children without M. catarrhalis to identify bacteria that may represent potential sources of therapeutics. Rothia was more abundant in the noses of healthy children compared to children with cold symptoms and M. catarrhalis. We cultured Rothia from nasal samples and determined that most isolates of Rothia dentocariosa and “Rothia similmucilaginosa” were able to fully inhibit the growth of M. catarrhalis in vitro, whereas isolates of Rothia aeria varied in their ability to inhibit M. catarrhalis. Using comparative genomics and proteomics, we identified a putative peptidoglycan hydrolase called secreted antigen A (SagA). This protein was present at higher relative abundance in the secreted proteomes of R. dentocariosa and R. similmucilaginosa than in those from non-inhibitory R. aeria, suggesting that it may be involved in M. catarrhalis inhibition. We produced SagA from R. similmucilaginosa in Escherichia coli and confirmed its ability to degrade M. catarrhalis peptidoglycan and inhibit its growth. We then demonstrated that R. aeria and R. similmucilaginosa reduced M. catarrhalis levels in an air-liquid interface culture model of the respiratory epithelium. Together, our results suggest that Rothia restricts M. catarrhalis colonization of the human respiratory tract in vivo.

In this study, we characterized the nasal bacterial communities of children with and without M. catarrhalis colonization. Among the bacterial taxa that were more abundant in the noses of healthy children without M. catarrhalis colonization than in children with cold symptoms and M. catarrhalis, we identified Rothia spp. (phylum Actinobacteria, family Micrococcaceae). We isolated Rothia from nasal lavage samples and determined that Rothia dentocariosa and a potentially novel species similar to Rothia mucilaginosa, which we provisionally named "Rothia similmucilaginosa," were able to fully inhibit the growth of M. catarrhalis in in vitro coculture assays. In contrast, isolates of Rothia aeria were more variable in their ability to inhibit M. catarrhalis in vitro. Using a combination of comparative genomics and proteomics, we identified a putative peptidoglycan hydrolase called secreted antigen A (SagA). The SagA protein was present at higher relative abundance in the secreted proteomes of R. dentocariosa and R. similmucilaginosa than in those from R. aeria. To determine whether SagA can degrade M. catarrhalis peptidoglycan, we heterologously produced the protein and demonstrated its activity using zymography and against M. catarrhalis cultures. Finally, we showed that R. aeria and R. similmucilaginosa can protect cultures of differentiated airway epithelial cells from M. catarrhalis infection. Together, these data suggest that Rothia in the human nasal cavity may protect the host against M. catarrhalis colonization using the peptidoglycan endopeptidase SagA.

RESULTS
The nasal bacterial communities of healthy and children with cold symptoms and Moraxella colonization are distinct. We were interested in identifying commensal nasal bacteria that can inhibit the growth of M. catarrhalis. As a first step to identify these candidates, we compared the communities of healthy children and children with cold symptoms and PCR-confirmed M. catarrhalis colonization from a subset of nasal lavage specimens collected as part of the Genetic Influences on Rhinovirus-Influenced Asthma (RhinoGen) study (25,43) (see Table S1 in the supplemental material). We extracted total DNA from specimens collected from 10 children with varying degrees of cold symptoms (median age: 7.95 years, interquartile range [IQR]: 6.6 to 8.2 years; 100% male; median cold severity score: 7, IQR: 6 to 8) and 10 children with no symptoms (median age: 8.05 years, IQR: 8 to 8.2 years; 60% male; cold severity score: 0) and used 16S rRNA gene V4 region amplicon sequencing to characterize their nasal bacterial communities. In total, we identified 235 operational taxonomic units (OTUs) at $97% sequence identity from the 20 nasal communities (Table S2). There was a slight but significant difference in the Shannon diversity index between the nasal communities isolated from healthy children (mean = 1.88, standard deviation [SD] = 0.79) and those from children exhibiting any degree of cold symptoms (mean = 1.19, SD = 0.47; Wilcoxon rank-sum test: P = 0.029).
To determine whether nasal bacterial community composition was associated with donor health status, we investigated beta diversity using Bray-Curtis dissimilarity and nonmetric multidimensional scaling (NMDS). In general, we found that the bacterial communities from healthy children were distinct from those from children with cold symptoms (analysis of similarity [ANOSIM] R = 0.29, P = 0.0039) (Fig. 1). However, three nasal bacterial communities from healthy children grouped with nine of the communities from children with cold symptoms. In addition, one of the bacterial communities from a child with minor symptoms (cold severity score = 4) did not group with any other community in this analysis (Fig. 1). We identified 45 bacterial OTUs with significant trends (a = 0.05) across the ordination space (Table S2). Otu0001 (R 2 = 0.82, P = 0.00001) and Otu0222 (R 2 = 0.51, P = 0.0023), both corresponding to Moraxella spp., were strongly associated with communities isolated from children with cold symptoms. In contrast, we identified 24 OTUs that were associated with healthy communities. These OTUs included Otu0034 (R 2 = 0.40, P = 0.024) and Otu0039 (R 2 = 0.34, P = 0.047), which correspond to Rothia mucilaginosa (phylum Actinobacteria, family Micrococcaceae) and an unclassified Rothia sp., respectively.
Rothia specifically inhibit Moraxella species. Given the negative correlation of Rothia OTUs with Moraxella OTUs (Fig. 1), we hypothesized that these bacteria may be able to inhibit the growth of M. catarrhalis. To test this hypothesis, we first isolated Rothia from frozen nasal lavage specimens donated as part of the RhinoGen study. In addition, we supplemented these isolates with two strains that we previously isolated from specimens donated as part of the Childhood Origins of Asthma (COAST) study (38,44,45).
We assessed the ability of a subset of the cultured Rothia isolates to inhibit the growth of M. catarrhalis in vitro using pairwise inhibition assays. We cultured 14 different Rothia isolates on brain heart infusion (BHI) agar plates. After the colonies were established, we inoculated six different M. catarrhalis strains next to the Rothia colony, then scored them for inhibition by visual assessment of the coculture after overnight incubation. Consistent with our hypothesis, we found that some Rothia isolates were able to inhibit the growth of multiple M. catarrhalis strains. In general, we did not observe differences in inhibition among different M. catarrhalis strains (Fig. S1). Across technical replicates of our assays, we generally observed concordance in M. catarrhalis inhibition. However, between different experimental replicates, Rothia isolates that generally inhibited M. catarrhalis or had no effect on its growth sometimes showed weak inhibition in another replication. Therefore, we assigned each Rothia isolate into one of three categories based on its average ability to inhibit M. catarrhalis: strong inhibition (fully inhibited most M. catarrhalis strains in nearly all replicates, n = 7), weak inhibition (partially inhibited most M. catarrhalis strains in nearly all replicates or had variable inhibition across replicates, n = 3), and no inhibition (did not inhibit most M. catarrhalis strains in nearly all replicates, n = 4) ( Fig. 2A).
We then wanted to determine whether this inhibition was specific to M. catarrhalis or whether other Gammaproteobacteria were susceptible to inhibition by the Rothia isolates. We repeated the inhibition assays and included strains of Acinetobacter baumannii, Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa, and Serratia marcescens. In addition, we tested whether these Rothia isolates inhibited the growth of Moraxella bovoculi and Moraxella ovis, which were isolated from a cow and sheep, respectively, with conjunctivitis (Fig. 2B). In general, we observed that all three Moraxella species behaved identically in the inhibition assays (i.e., if a Rothia isolate inhibited M. catarrhalis, then it also inhibited M. bovoculi and M. ovis). We never observed inhibition of a non-Moraxella species by any of the 14 Rothia isolates that we tested (Fig. 2B). These 14 Rothia isolates were also tested for activity against a panel containing three fungi (Aspergillus flavus, Candida albicans, and Trichosporon asahii), three Gram-positive bacteria (Bacillus subtilis, Enterococcus faecalis, and S. aureus), and one additional Gram-negative bacterium (Enterobacter cloacae). No inhibitory activity was observed against any fungi or bacteria, except Moraxella species. Taken together, these data indicate that some strains of Rothia produce factor(s) that specifically inhibit Moraxella at a distance.  Different species of Rothia are associated with inhibition and non-inhibition of M. catarrhalis. To confirm the identity of each Rothia isolate, we extracted and sequenced genomic DNA to generate draft genome sequences for each isolate used in the coculture assays. We then compared these genome sequences with 9 complete, high-quality Rothia genomes from GenBank, including R. aeria, R. dentocariosa, R. kristinae, R. mucilaginosa, R. nasimurium, and R. terrae. In addition, we constructed a core-genome phylogeny from 93 conserved single-copy genes using these same genome sequences. Based on gene content, average nucleotide identity (ANI), and phylogenetic placement (Fig. 3, Fig. S2), we identified the 14 Rothia isolates used in this study as R. aeria (n = 6, ANI = 97.7% to 99.5% relative to FIG 3 Pangenome of Rothia. This pangenome was generated from 9 reference genomes and 14 draft genome sequences of the strains used in this study using the anvi'o 7 pangenome pipeline. Each arc represents a single genome and is colored based on its species-level identification, as indicated by the key below. Genomes are ordered based on average nucleotide identity (ANI; top right). In total, this pangenome contains 48,332 genes within 8,099 gene clusters. The presence or absence of a gene cluster within a given genome is indicated by opaque and transparent colors, respectively. Black bar indicates the set of single-copy core gene (SCG) clusters.
Nasal Rothia Inhibit M. catarrhalis with SagA mBio reference strain FDAARGOS_1137), R. dentocariosa (n = 4, ANI = 96.5% to 96.8% relative to reference strain FDAARGOS_752), and a species closely related to R. mucilaginosa (n = 4, ANI: 91.5% to 91.6% relative to reference strain FDAARGOS_369). We note that though the isolates in the latter category were closely related to R. mucilaginosa, they did not reach the frequently used 95% ANI cutoff for species delineation (46). In addition, these four isolates did not reach this ANI threshold compared with any other characterized Rothia stain. Henceforth, for simplicity, we will refer to these isolates using the provisional species name "Rothia similmucilaginosa." A full biochemical characterization and formal classification of these R. similmucilaginosa isolates is forthcoming. We found that three R. dentocariosa isolates and all four R. similmucilaginosa isolates strongly inhibited M. catarrhalis. Two R. aeria isolates and one R. dentocariosa isolate weakly inhibited M. catarrhalis. The remaining four R. aeria isolates did not inhibit M. catarrhalis (Fig. 2B). We next sought to determine the identity of the anti-Moraxella inhibitory factor(s) produced by the inhibitory Rothia isolates.
Inhibition of M. catarrhalis by Rothia is not due to production of a siderophore. Rothia genomes contain biosynthetic gene clusters (BGCs) for the production of antibiotics and other secondary metabolites (47)(48)(49). We hypothesized that the Rothia strains with inhibitory activity against M. catarrhalis produce secondary metabolite(s) that the non-inhibitory strains lack. To test this hypothesis, we first identified BGCs contained within Rothia genomes using antiSMASH (50) (Fig. S3A). Consistent with a previous study that investigated the biosynthetic capacity of Rothia genomes (51), we identified a single BGC encoding the production of the non-ribosomal peptide siderophore enterobactin that was present in the genomes of all 14 isolates used in this study and 8 of the reference genomes (Fig. S2A). For all R. similmucilaginosa genomes, the enterobactin BGC was the only BGC we detected (Fig. S3A).
Siderophores are molecules produced by microbes to scavenge ferric iron and other minerals from the environment (52). In a previous study, we determined that R. aeria strains HSID18067 and HSID18069 did not produce siderophore activity under similar, but slightly different, assay conditions to those used in the present study (38). Here, these strains possessed no activity and partial activity against M. catarrhalis, respectively (Fig. 2B). Therefore, we wanted to determine if R. dentocariosa and R. similmucilaginosa strains produce siderophores as a potential mechanism to inhibit M. catarrhalis growth. Using chrome azurol S (CAS) assays, we detected no iron sequestration from R. aeria strain RSM41, R. dentocariosa strain RSM522, or R. similmucilaginosa strain RSM42 under the same conditions as used for the coculture inhibition assays (Fig. S3B), suggesting that these bacteria do not produce siderophores. In a parallel approach to test whether iron sequestration by Rothia is responsible for inhibition of M. catarrhalis, we performed inhibition assays using standard BHI agar plates 6200 mM ferric chloride (FeCl 3 ). We found that iron supplementation did not appreciably change the inhibition phenotype of M. catarrhalis strain O35E when cultured with R. similmucilaginosa strain RSM42 or R. dentocariosa strain RSM522 (Fig. S3C). Taken together, these data indicate that iron limitation via a siderophore or production of another metabolite by R. dentocariosa and R. similmucilaginosa are not responsible for the observed inhibition of M. catarrhalis.
Identification of secreted proteins associated with Rothia strains that inhibit M. catarrhalis. Given the previous finding that M. catarrhalis inhibition was likely not due to production of enterobactin or another secondary metabolite (Fig. S3), we hypothesized that Rothia may produce a protein that inhibits this pathobiont.
We compared the genomes of different Rothia to determine whether the inhibitory strains encode proteins that were absent in the genomes of the non-inhibitory strains. When we analyzed the pangenome of Rothia, we determined that there was a strong signal between overall gene content and Rothia species-level identity (Fig. 3). Because inhibition of M. catarrhalis by Rothia occurred at a distance, we reasoned that an inhibitory protein must be secreted. Therefore, we decided to focus our subsequent analysis on genes that encode proteins with identifiable secretion signals. Of the 4,086 total groups of homologous protein-encoding genes we identified in the Rothia pangenomes, 488 (12% total) encoded proteins with predicted secretion signals. We observed three clusters of Rothia based on the presence or absence of homologous groups of genes encoding secreted proteins. The first cluster consisted of R. aeria and R. dentocariosa, the second cluster consisted of R. kristinae, R. nasimurium, and R. terrae, and the final cluster consisted of R. mucilaginosa and R. similmucilaginosa (Fig. 4A). These clusters are broadly consistent with Rothia phylogeny (Fig. S2) but did not inform the identification of a protein secreted by both R. dentocariosa and R. similmucilaginosa that inhibits M. catarrhalis. In addition, we did not identify any homologous groups of genes encoding secreted proteins that were universally shared between R. dentocariosa and R. (simil)mucilaginosa but were absent in R. aeria. Further, given that R. aeria strains HSID18069 and RSM15 had weak activity against M. catarrhalis (Fig. 2B), it is possible that these strains also produce inhibitory protein(s), albeit at lower levels.
To narrow the list of candidates for the identity of M. catarrhalis inhibitory protein(s), we characterized the secreted proteomes of the Rothia isolates that we used for inhibition assays. We cultured R. aeria, R. dentocariosa, and R. similmucilaginosa as before, but excised the agar near the bacterial colonies where we would inoculate M. catarrhalis in a coculture inhibition assay and extracted proteins. We then used untargeted liquid chromatographytandem mass spectrometry (LC-MS/MS) to identify proteins secreted by each Rothia isolate. In total, we identified 181 homologous groups of secreted proteins from the three Rothia species. In contrast to the clustering observed when considering the content of genes encoding secreted proteins (Fig. 4A), we found that R. aeria, R. dentocariosa, and R. similmucilaginosa each formed their own species-based clusters (ANOSIM R = 1.00, P = 0.00001) (Fig. 4B). When we searched the secreted proteomes for proteins that were associated with both R. dentocariosa and R. similmucilaginosa, we identified six groups of homologous secreted proteins (Fig. 4B, Table 1). The proteins represented by homolog group_00865 emerged as candidates of interest for the anti-M. catarrhalis activity. These proteins were found in higher relative abundance in the secreted proteomes of R. dentocariosa (median: 2.4% of proteins with secretion signal) and R. similmucilaginosa (median: 3.5% of proteins  (Table 1), and their length is scaled according to the square root of their correlation coefficients such that longer vectors indicate stronger correlations. The vector for group_00865 (SagA) is labeled. The ANOSIM R and P values (100,000 permutations) for comparing species are shown in the top right of the plot. The stress value of the NMDS plot was 0.084. (C) The relative abundance of homolog group_00865, which corresponds to SagA (secreted antigen A), in the secreted proteomes of Rothia grouped by species. Horizontal black bars indicate the median relative abundance values. The points in all plots are colored based on the species-level identification and the shapes correspond to the M. catarrhalis inhibition phenotype, as indicated by the key below the plots.
Nasal Rothia Inhibit M. catarrhalis with SagA mBio with secretion signal) than in that of R. aeria (median: 0.75% of proteins with secretion signal) (Fig. 4C). In the R. mucilaginosa strain Num-Rm6536 reference genome, the gene encoding the group_00865 protein homolog is named secreted antigen A (sagA) (53). Therefore, we subsequently refer to this group of proteins as SagA proteins. SagA proteins from Rothia spp. are ;20 kDa (197 to 210 amino acids), containing a 30amino acid N-terminal signal peptide and a C-terminal endopeptidase, NlpC/P60 cysteine peptidase domain (InterPro accession no. IPR000064), and are homologous to members of the C40 peptidase family. Proteins with the NlpC/P60 domain have several known activities, including cleaving D-g -glutamyl-meso-diaminopimelate or N-acetylmuramate-L-alanine linkages in peptidoglycan cross-bridges (i.e., these proteins possess peptidoglycan endopeptidase activity) (54). Based on alignment of SagA from R. aeria strain RSM41, R. dentocariosa strain RSM522, and R. similmucilaginosa strain RSM42 to each other and to structurally characterized C40 family peptidases, including the family type protein (MEROPS accession no. MER0001322) (55), we identified the conserved cysteine and histidine residues involved in the cysteine peptidase reaction mechanism (Fig. S4A). However, SagA is classified as a nonpeptidase homolog in the MEROPS database because it has a serine (Ser-172), instead of a charged residue in the third position of the catalytic triad (Cys-His-[Asn/Gln/Glu/His]) (Fig. S4A). The charged residue does not directly participate in the reaction mechanism but is important for coordinating the basic histidine residue (corresponding to His-160 in SagA from R. similmucilaginosa strain RSM42). It is known that the eukaryotic endothelial protease vasohibin uses a Cys-His-Ser catalytic triad in which the serine residue orients the histidine base (56). We modeled 176 residues of SagA from R. similmucilaginosa strain RSM42 onto the C40 family peptidase RipA from Mycobacterium tuberculosis (Protein Data Bank [PDB] ID: 2XIV). Based on the homology model, we found that Ser-172 is positioned well to orient His-160 using a hydrogen bond (Fig. S4B). Therefore, we hypothesized that SagA is a functional peptidoglycan endopeptidase and may be responsible for inhibition of M. catarrhalis by Rothia.
SagA degrades peptidoglycan from M. catarrhalis and inhibits its growth. Currently, genetic manipulation of Rothia is challenging and the necessary conditions for transformation are not well-defined. Therefore, to further characterize SagA and determine its activity, we chose to use heterologous expression in E. coli.
We initially cloned a sequence encoding the C-terminal NlpC/P60 cysteine peptidase domain of SagA, without the N-terminal signal peptide, from R. similmucilaginosa strain RSM42 with an added Shine-Dalgarno sequence and start codon into an isopropyl b-D-1-thiogalactopyranoside (IPTG)-inducible expression vector. For simplicity, we subsequently refer to the truncated protein as SagA  . When we sequenced three different E. coli clones, we identified two different 32-bp deletions of the lac operator sequence. Given the propensity for IPTG-inducible systems to have leaky expression (57), we hypothesized that production of SagA  in E. coli was cytotoxic and the deletions in the lac operator sequence prevented expression. Therefore, we chose to use an L-arabinose-inducible expression system, as these systems are more tightly controlled than IPTG-inducible systems (57). We confirmed dosedependent production of the SagA  fusion protein from the L-arabinose-inducible expression system by E. coli as an ;17.2-kDa protein using sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) (Fig. S5A). During these experiments, we noticed that   tended to grow 2 to 3Â more slowly and formed smaller colonies when cultured on media containing L-arabinose, compared to uninduced controls (Fig. S5B). Taken together with the deletions observed in IPTG-inducible constructs, these data support the hypothesis that production of SagA  is cytotoxic to E. coli, consistent with the activity of a peptidoglycan endopeptidase.
For subsequent experiments, we chose to optimize the coding sequence for SagA for expression in E. coli and added a 6ÂHis tag for purification. We partially purified SagA from E. coli using nickel affinity chromatography and tested the protein for peptidoglycan degradation activity using zymography. We incorporated M. catarrhalis strain O35E peptidoglycan into a polyacrylamide gel during polymerization. Following electrophoresis, we renatured the proteins within the zymography gel, allowed them to hydrolyze M. catarrhalis peptidoglycan, and stained the remaining peptidoglycan using methylene blue. We confirmed that SagA degrades M. catarrhalis peptidoglycan, as indicated by a zone of clearing within the zymogram gel, corresponding to the same ;19.8-kDa band in the Coomassie-stained gel (Fig. 5A). Using LC-MS/MS, we confirmed that both of these bands corresponded to 6ÂHis-SagA  .
We then wanted to determine if SagA directly inhibits M. catarrhalis growth. We cultured M. catarrhalis strain O35E in the presence of 6ÂHis-SagA  and assessed its growth over time. We found that M. catarrhalis cultures exposed to SagA failed to grow, whereas cultures exposed to an equivalent purification from E. coli cells harboring an empty vector were not inhibited (Fig. 5B).
Rothia protect against M. catarrhalis colonization in cultures of respiratory epithelium. To determine if Rothia species were protective against M. catarrhalis in models that are more representative of their natural environments, we chose to perform coculture assays on air-liquid interface (ALI) cultures of respiratory epithelium. We differentiated residual tissue from donor lungs into epithelial respiratory cells and colonized these cells with R. aeria strain RSM41, R. dentocariosa strain RSM522, or R. similmucilaginosa strain RSM42. Subsequently, we infected these cultures with M. catarrhalis strain O35E. After 24 h of coculture, we washed the apical surface, collected the epithelium, Rothia, and M. catarrhalis cells together, lysed the cells, and assessed M. catarrhalis colonization levels using a quantitative PCR (qPCR) assay. We report our results in CFU equivalents (CFUe), based on calibration of the qPCR cycle threshold with known quantities of M. catarrhalis in pure culture. As a control, we included a treatment where the epithelial cells were not colonized with Rothia. However, post hoc comparisons did not initially reveal significant differences between any of the treatments. This non-significance was driven by a single outlying experiment in which M. catarrhalis strain O35E failed to colonize the cells (1.4 Â 10 4 CFUe detected [est. ,0.05% of the inoculum], compared to a median of 4.6 Â 10 5 CFUe [est. 1.5% of the inoculum]). Consequently, we repeated our analyses with this set of experiments excluded, including the matched Rothia-colonized ALI cultures. As expected, we found that there was a stronger effect of Rothia treatment on M. catarrhalis strain O35E levels (ANOVA F[3,32] = 4.50, P = 0.0096). Post hoc comparisons revealed that colonization of respiratory epithelial cells with either R. aeria strain RSM41 (Tukey's honestly significant difference [HSD] Q = 4.72, adjusted P = 0.011) or R. similmucilaginosa strain RSM42 (Tukey's HSD Q = 4.24, adjusted P = 0.026) decreased M. catarrhalis strain O35E colonization by ;0.8 log (Fig. 6). Unexpectedly, colonization of the epithelial cells with R. dentocariosa strain RSM522 did not result in a decrease in M. catarrhalis strain O35E levels (Fig. 6).

DISCUSSION
In this study, we identified specific nasal bacteria that inhibit the growth of the pathobiont M. catarrhalis. We initially compared the nasal cavity bacterial communities of children with and without Moraxella colonization, which led us to determine that Rothia was associated with healthy nasal microbiomes that were free of M. catarrhalis colonization (Fig. 1). We performed targeted isolation of Rothia from nasal lavage samples. We then determined that R. dentocariosa and R. similmucilaginosa strains inhibited the growth of M. catarrhalis in vitro, while R. aeria strains tended to be less inhibitory ( Fig. 2 and 3). We reasoned that the inhibitory factor produced by these Rothia spp. must be a secreted protein, so we characterized the secreted proteomes of these bacteria, which led us to identify a putative endopeptidase called SagA (Fig. 4). We heterologously produced SagA and confirmed that the enzyme can degrade peptidoglycan and inhibit M. catarrhalis growth in vitro (Fig. 5). Finally, we demonstrated that Rothia can reduce M. catarrhalis levels in an airway epithelial cell model (Fig. 6).
Rothia species, including two of the three highlighted in our study, are generally considered to be inhabitants of the oral cavity (58). In a recent survey, isolates of R. dentocariosa and R. mucilaginosa were commonly cultured from pharynx samples of healthy children. However, R. mucilaginosa was rarely cultured from nasopharynx samples and R. dentocariosa was not detected at this site (59). However, we were able to culture and identify Rothia isolates from approximately half of the nasal lavage specimens that we tested. In addition, Rothia are regularly identified by sequencing studies as lowabundance members of upper and lower respiratory tract microbiomes. The relative abundances we observed were similar to what has been reported for other 16S rRNA gene amplicon sequencing studies of the anterior nares, nasal cavity, and nasopharynx (60)(61)(62). Together, these findings support the conclusion that Rothia are indeed bona fide inhabitants of the human nasal microbiome.
Prior research on Rothia has largely focused on their potential as opportunistic pathogens, primarily in immunocompromised hosts (63). However, Rothia have more recently been viewed as commensal bacteria and perhaps even mutualists in the oral and respiratory tracts. For instance, colonization of either an alveolar epithelial cell model or mouse lung with R. mucilaginosa prevented a P. aeruginosa-mediated, lipopolysaccharide-induced inflammatory response, resulting in reduced tissue damage compared to controls (64). Within induced sputum from individuals with bronchiectasis, the absolute abundance of R. mucilaginosa was inversely correlated with the levels of the proinflammatory markers interleukin 1 beta and interleukin 8 (64). The enzymatic activity of SagA itself may also trigger beneficial immune signaling in hosts. In a subcutaneous melanoma model, mice whose gastrointestinal tracts had been colonized by SagA-producing enterococci had significantly reduced tumor volume when later administered the immune checkpoint inhibitor anti-PD-L1, compared to mice treated with anti-PD-L1 alone. This antitumor effect was shown to be dependent on activation of the innate immune sensor nucleotide-binding oligomerization domain-2 (NOD2). The NOD2 sensor is activated by N-acetylglucosamine muramyl dipeptides generated from peptidoglycan by the D,L-endopeptidase activity of enterococcal SagA, suggesting that the microbiota can modulate responses to immunotherapy (65). In addition, NOD2 plays a role in the immune response to otitis media (66), the pulmonary immune response (67), and pathogen clearance (68). These examples, together with our findings, suggest a new role for Rothia as mutualists that defend against the pathobiont M. catarrhalis and modulate the host immune response. Subsequent work will be required to validate the role Rothia spp. play in these processes in vivo.
We do not fully understand how SagA acts on M. catarrhalis. In general, the outer membrane of Gram-negative bacteria, including M. catarrhalis, excludes molecules as large as SagA (;17.2 kDa). However, human lysozyme C (;14.7 kDa) has been previously shown to be active against M. catarrhalis (69). Lysozyme targets the peptidoglycan, catalyzing hydrolysis of the 1,4-beta linkages between N-acetylmuramic acid and N-acetylglucosamine residues, whereas SagA presumably uses its endopeptidase activity to target the pentapeptide stem between layers of peptidoglycan. The precise mechanism used by lysozyme to overcome the outer membrane is not fully known, but potentially involves membrane disruption through its innate cationic property at the nasal mucosal pH of 6.5 (calculated charge [Z] = 17.8) (70). However, unlike lysozyme, the secreted fragment of SagA is anionic at pH 6.5 (calculated Z = -3.2). Perhaps Rothia produce their own cationic peptides that act to permeabilize the outer membrane to SagA, or take advantage of a host-produced protein in vivo. In the latter case, one such possibility is lactoferrin, which is known to permeabilize the outer membrane of Gram-negative bacteria (71) and is present at high concentrations within nasal secretions (72). Therefore, further work is required to determine how SagA gains access to the peptidoglycan of M. catarrhalis. Moreover, additional work is necessary to determine why Moraxella species were more susceptible to inhibition by Rothia than any of the other bacteria that we tested, including other closely related Gammaproteobacteria.
Here, we note an inconsistency between our in vitro and ALI cell culture results. We found that R. aeria tended to be non-inhibitory toward M. catarrhalis in vitro. However, R. aeria strain RSM41, which we chose as a representative of this species because it almost never inhibited M. catarrhalis in inhibition assays, was equally protective against M. catarrhalis as R. similmucilaginosa in ALI cell cultures. Although R. aeria tended to secrete less SagA than R. dentocariosa and R. mucilaginosa, it is possible that R. aeria produces sufficient quantities of the enzyme when cultured on epithelial cells, the anti-M. catarrhalis activity of SagA is potentiated by a host-produced factor (see above), or Rothia produces additional proteins that are responsible for M. catarrhalis inhibition on cells.
In conclusion, we report that Rothia species potentially inhibit colonization of the nasal cavity by M. catarrhalis using an interference competition mechanism by producing a secreted peptidoglycan endopeptidase that directly inhibits M. catarrhalis growth. Given the specificity of these Rothia against Moraxella spp., we suggest that these bacteria or SagA itself could be developed as therapeutics to selectively inhibit nasal colonization by M. catarrhalis. Furthermore, we suspect that additional efforts to screen the nasal microbiota for bacteria that produce antimicrobials will yield more examples of proteins or metabolites with anti-M. catarrhalis activity.

MATERIALS AND METHODS
Nasal specimen collection and cold severity scoring. The nasal lavage specimens used in this study were collected from children and banked as part of the RhinoGen study (25,43). Informed consent was obtained from guardians, and the Human Subjects Committee at the University of Wisconsin-Madison approved the study (institutional review board [IRB] approval no. H-2007-0136-CR008). The cold severity score was determined after sample collection. This score is the summation of daily symptom scores for the period of the illness. Specimens collected within 63 days of illness symptoms were considered to be associated with that illness. The daily scores ranged from 0 for absent to 3 for severe respiratory symptoms (25) (Table S1).
DNA extraction and 16S rRNA gene V4 region amplicon sequencing. We extracted total DNA from nasal lavage samples using the Bacteremia DNA isolation kit (MO BIO Laboratories Inc.). We included a negative extraction control of sterile water treated identically to the samples. All samples were quantified using a Qubit fluorometer with broad range standards (Invitrogen). Initially, we attempted to directly amplify the 16S rRNA gene V4 region from the extracted total DNA. However, we did not observe amplification after electrophoresis of PCR products on a 1% (wt/vol) agarose Tris-acetate-EDTA (TAE) gel. Therefore, we decided to use a nested PCR approach.
We first amplified the near full-length 16S rRNA gene in triplicate from the extracted samples with primers 27F (59-AGAGTTTGATCCTGGCTCAG-39) and 1492R (59-GGTTACCTTGTTACGACTT-39) using KAPA HiFi HotStart DNA polymerase (Roche). Each 25-mL reaction mixture contained 5 mL of extracted DNA (median: 34 ng, IQR: 3.8 to 110 ng) and 20 pmol of each primer. We used the following PCR cycling conditions: 95°C for 1 min; 15 cycles of 95°C for 30 s, 58°C for 30 s, and 72°C for 30 s; and a final extension at 72°C for 5 min. Subsequently, we amplified the 16S rRNA gene V4 region in triplicate using the PCR products as the template with primers 515F (59-[P5 Illumina Adapter]-[Index]-TATGGTAATTGTGTGCCAGCMGCCGCGG TAA-39) and 806R (59-[P7 Illumina Adapter]-[Index]-AGTCAGTCAGCCGGACTACHVGGGTWTCTAAT-39). Each 25-mL reaction mixture contained 2 mL of the unpurified first-round PCR product and 10 pmol of each primer. We used the same PCR cycling conditions as above. We electrophoresed 45 mL of pooled triplicated second-round PCR products on a 1.2% (wt/vol) agarose 0.5Â Tris-borate-EDTA (TBE) gel at 100 V for 1 h. Next, we purified the PCR product from the agarose gel using the Wizard SV Gel and PCR Clean-Up System (Promega). We quantified DNA as described above. In the negative extraction control, the DNA concentration was below the limit of detection. We pooled 2 ng of each sample together with 1 mL of the undiluted negative extraction control. The amplicon libraries were prepared and sequenced using the 2Â 300-bp pairedend Illumina MiSeq platform at the University of Wisconsin-Madison Biotechnology Center (Madison, WI).
Bacterial community analysis. We used fastp version 0.20.0 (73) for quality control and preprocessing of the raw amplicon sequencing reads. We used the mothur version 1.44.1 pipeline (74) for read processing, assembly, alignment, and classification. For sequence classification, we used the expanded human oral microbiome database 16S rRNA RefSeq version 15.1 (3). We defined OTUs using a 97% sequence similarity threshold (Table S2). There was .99% coverage of these samples, as assessed by Good's Coverage Metric (Table S1). For all subsequent analyses, we used R version 4.0.4 (75). We processed the OTU table generated by mothur and subtracted the reads from each OTU in each sample corresponding to the number of reads detected in the negative extraction control. If OTU counts became negative after subtraction, we replaced the negative value with 0. We removed OTUs represented by ,10 total reads from the data set. We then converted OTU counts to relative abundances and used the phyloseq package version 1.33.0 (76) and the vegan package version 2.5-Nasal Rothia Inhibit M. catarrhalis with SagA mBio 6 (77) in R to perform NMDS and ANOSIM based on the Bray-Curtis dissimilarity between nasal bacterial communities. We used the envfit function from vegan to fit OTU vectors onto the ordination plot. Bacterial isolation, culturing, and initial identification. To obtain isolates for culture-based work, we diluted 200 mL of thawed nasal lavage specimens with 800 mL of sterile phosphate-buffered saline (PBS) and then inoculated 100 mL onto 25-mL BHI (DOT Scientific) plates (diameter: 8.6 cm) solidified with 1.5% [wt/vol] agar (VWR). In addition to using standard BHI plates, we also included plates with 50 mg/mL lithium mupirocin (Sigma-Aldrich) to inhibit the growth of Staphylococcus spp. and allow for isolation of slower-growing organisms and 25 mg/mL vancomycin (Sigma-Aldrich) to inhibit the growth of Gram-positive bacteria and enrich for M. catarrhalis. Each sample was plated in triplicate onto both medium types. We incubated plates aerobically at 37°C for 1 week. We selected $2 colonies of each distinct morphotype per plate and passaged the isolates aerobically on BHI plates with no antibiotic supplementation at 37°C until we obtained pure cultures. All bacterial isolates were cryopreserved at 280°C in 25% (vol/vol) glycerol. Bacterial strains used in this study are listed in Table S3 in the supplemental material. For routine culturing and all following assays, we used BBL BHI (Becton, Dickinson and Co. [BD]) solidified with 1.5% (wt/vol) Bacto agar (BD) for plates because we observed poor or inconsistent growth of M. catarrhalis on BHI or agar produced by other manufacturers. For experiments testing the effect of iron on inhibition, we supplemented the BHI agar plates with 200 mM FeCl 3 (Fisher Scientific).
To identify isolates to the genus level, we performed colony PCR using the universal 27F (59-AGAGTTTGATCMTGGCTCAG-39) and 1,492R (59-CGGTTACCTTGTTACGACTT-39) primers (78). Briefly, we suspended a small portion of each bacterial colony or liquid culture in 20 mL of 0.02 N sodium hydroxide and incubated the samples at 95°C for 10 min. We then chilled the samples on ice and centrifuged them at 21,130 Â g for 5 min and used 1 mL of the supernatant as the template for PCR. We sequenced the PCR products using the Sanger method at the University of Wisconsin Biotechnology Center DNA Sequencing Facility (Madison, WI) and identified isolates to the genus level with the Ribosomal Database Project Classifier (79).
Inhibition assays. To determine whether Rothia isolates inhibited M. catarrhalis and other Gammaproteobacteria, we used coculture plate inhibition assays. We inoculated 3 mL BHI broth with single colonies of Rothia from 2-day-old plates and incubated the cultures at 37°C while shaking. After overnight growth, we diluted each culture to an optical density at 600 nm (OD 600 ) of 0.05 in 3 mL fresh BHI, incubated the cultures until they reached an OD 600 of 0.4 to 0.6, and spread 50 mL in a ;6 Â ;2 cm-wide line in the center of a 25-mL BHI agar plate (diameter: 8.6 cm). We incubated the plates aerobically for 3 days at 37°C. We prepared overnight cultures of M. catarrhalis and other Gammaproteobacteria by inoculating single colonies into 3 mL BHI broth and incubated the cultures at 37°C with shaking. After overnight growth, we diluted the cultures to an OD 600 of 0.5 in fresh BHI and spotted 5 mL approximately 0.5 cm away from the line of Rothia. When the spots dried, we returned the plates to 37°C, then scored them for inhibition after overnight incubation. All inhibition assays were experimentally replicated $3 times, with at least 2 technical replicates per experiment.
Genomic DNA extraction and sequencing. We used either the DNA Miniprep kit (ZymoBIOMICS) with 2Â 5-min bead beating steps or the MasterPure Yeast DNA purification kit (Lucigen) with the addition of 1 mL Ready-Lyse Lysozyme Solution (Lucigen) and 1 mL 5 mg/mL RNase A (Lucigen) to lyse Rothia isolates and purify genomic DNA. Libraries were prepared with a single library preparation method based on the Illumina Nextera kit and sequenced using the 2Â 150-bp paired-end Illumina NextSeq 2000 platform at SeqCenter (Pittsburgh, PA).
Genome assembly and annotation. We processed raw genomic reads using fastp 0.20.0 and assembled draft genome sequences using SPAdes 3.11.1 (80), using default parameters for both. To identify homologues present across Rothia genomes, we downloaded high-quality complete genome sequences of R. aeria, R. dentocariosa, R. kristinae, R. mucilaginosa, R. nasimurium, and R. terrae from GenBank. We identified protein-coding genes with Prokka (81) using default parameters and used BuildGroups.py from PyParanoid (82) to build groups of homologous gene families. To predict secreted proteins, we submitted the consensus sequences for each homologous family identified by PyParanoid to the SignalP-5.0 webserver (https:// services.healthtech.dtu.dk/service.php?SignalP-5.0) (83).
CAS assay. We used the overlay CAS assay to test for siderophore production as previously described (87). We cultured Rothia for 3 days at 37°C, as above. We then overlaid each plate with 5 mL of CAS reagent overlay (52.5 mM CAS, 50 mM FeCl 3 , 0.5 mM hexadecyltrimethylammonium bromide, 10 mM piperazine-N, N9-bis[2-ethanesulfonic acid], 1% [wt/vol] agarose [pH 6.8]) (88) and incubated the plates at ambient temperature in the dark for 4 h before scanning them for color change, which would indicate the production of a siderophore (89).
Extraction of secreted proteins. To identify proteins secreted by Rothia isolates into agar, we adapted a previously used protocol (90). We cultured Rothia onto six 25-mL BHI plates, as above. After incubation for 3 days at 37°C, we used clean razor blades to excise the agar from a distance of ;0.1 to ;1 cm away from both sides of the Rothia without disturbing the bacterial colony.
We centrifuged the agar slices at 4,000 Â g for 10 min at ambient temperature, then combined them with an equal volume of extraction buffer (50 mM Tris-HCl, 0.2% [wt/vol] N-lauroylsarcosine sodium salt [pH 7.5]) and incubated the mixture at ambient temperature on a shaking platform set at 250 rpm overnight. We then centrifuged the samples at 4,000 Â g for 20 min at ambient temperature and filtered the supernatant through cheesecloth to remove agar. We added trichloroacetic acid (TCA) to a final concentration of 7.5% (wt/vol) and precipitated protein overnight at 220°C. We pelleted the precipitated protein by centrifuging the samples at 4,000 Â g for 30 min at 4°C. We then washed the protein pellet three times with 1 mL acetone and centrifugation at 21,130 Â g for 5 min at 4°C. We evaporated residual acetone under a laminar flow hood. To remove residual agar from the samples, we first resuspended the protein pellets in 250 mL of 50 mM Tris-HCl (pH 7.5) and centrifuged the samples at 21,130 Â g for 30 min at 4°C. We then transferred the supernatant into a new tube, and centrifuged the samples overnight at 21,130 Â g at 4°C. Finally, the supernatant was transferred into a new tube and protein content was measured using a Coomassie protein assay (Thermo Fisher Scientific). We analyzed 3 mg of each protein sample by SDS-PAGE using a 12% Criterion TGX precast midi protein gel. We visualized proteins by staining with RAPIDstrain (G-Biosciences) for 1 h followed by 3Â 15-min destaining washes in distilled water.
Identification of secreted proteins. We submitted protein samples for digestion and LC-MS/MS analysis at the University of Wisconsin Biotechnology Center Mass Spectrometry Facility (Madison, WI). To precipitate proteins, 60 mL of each sample was treated with a mixture of 10% TCA/50% acetone/40% water (vol/vol/vol) for 30 min on ice. The samples were centrifuged at 16,000 Â g for 10 min and the pellet was washed with 800 mL acetone at 220°C. The samples were then centrifuged as described above and washed with 800 mL acetone at 220°C. Subsequently, the pellets were resolubilized and denatured in 50 mL of a solution containing 8 M urea, 50 mM ammonium bicarbonate (NH 4  The digested and acidified samples were cleaned using OMIX C18 SPE cartridges (Agilent) following the manufacturer's protocol, eluted with 20 mL of 50% (vol/vol) acetonitrile (ACN) with 0.1% (vol/vol) TFA, dried, and reconstituted in 40 mL 2% (vol/vol) ACN with 0.1% (vol/vol) formic acid. Peptides were analyzed by nanoLC-MS/MS using the Agilent 1100 nanoflow system (Agilent) connected to an LTQ-Orbitrap Elite (Thermo Fisher Scientific) equipped with an EASY-Spray electrospray source (held at constant 35°C). Chromatography of peptides prior to mass spectral analysis was accomplished using a capillary emitter column (PepMap C18, 3 mM, 100 Å, 150 Â 0.075 mm, Thermo Fisher Scientific) onto which 2 mL of the extracted peptides was automatically loaded. A NanoHPLC (high-performance liquid chromatography) system delivered 0.1% (vol/vol) formic acid (Solvent A), and 99.9% (vol/vol) ACN with 0.1% (vol/vol) formic acid (Solvent B) at a flow rate of 0.50 mL/min to load the peptides over a 30-min period. The flow rate was adjusted to 0.3 mL/min to elute peptides directly into the nano-electrospray with a gradual gradient of 0% to 30% B over 78 min and concluded with a 5-min fast gradient from 30% to 50% Bm at which time a 5-min flash-out from 50% to 95% B took place. As peptides eluted from the HPLC column/electrospray source survey, MS scans were acquired in the Orbitrap at a resolution of 120,000, followed by CID-type MS/MS fragmentation of the 30 most intense peptides detected in the MS scan from 350 to 1,500 m/z. Redundancy was limited by dynamic exclusion.
The raw MS/MS data files were converted to the mgf file format using MSConvert (ProteoWizard) and searched against user-defined databases for each Rothia strain along with a cRAP common lab contaminant database using in-house Mascot search engine version 2.7.0 (Matrix Science) with variable methionine oxidation plus asparagine or glutamine deamidation and fixed cysteine carbamidomethylation. The peptide mass tolerance was set at 10 ppm and fragment mass at 0.6 Da. Protein annotations, significance of identification, and spectral-based quantification was determined with Scaffold version 4.11.0 (Proteome Software Inc.). The protein identifications were accepted if they could achieve a false discovery rate of ,1.0% and contained at least 3 identified peptides. Protein probabilities were assigned by the Protein Prophet algorithm (91). Proteins that contained similar peptides and could not be differentiated based on MS/MS alone were grouped to satisfy the principles of parsimony. Proteins which shared significant peptide evidence were grouped into clusters.
Comparison of secreted proteomes. We mapped the spectral counts for each protein identified by LC-MS/MS onto the homolog groups generated using PyParanoid described above. In cases where we identified multiple proteins within the same homolog group from the same strain, we aggregated the spectral counts into a single count for that homolog group. We transformed spectral counts into a relative abundance by dividing each count by the total number of spectral counts identified in a sample. We then filtered out groups of potentially contaminating cytoplasmic proteins based on the presence of a signal peptide as determined by SignalP. We used the phyloseq and vegan packages in R to perform NMDS and ANOSIM and to fit homolog group vectors onto the ordination plot as above.
mL of a Rothia suspension for 40 h. We infected ALI cells with M. catarrhalis as previously described (98). Briefly, we infected the apical surface with 50 mL of the M. catarrhalis suspension and incubated the cells at 37°C for 24 h. At the end of the infection period, we gently washed the apical surface with 500 mL prewarmed PBS and collected the cells in 350 mL of RLT Plus Buffer (Qiagen) containing 0.5% Reagent DX (Qiagen).
Quantification of M. catarrhalis in airway epithelium cultures. We transferred the cells to PowerBead tubes (Qiagen) for bead beating (5 min, 50 oscillations/s, Qiagen TissueLyser LT). We then extracted RNA and DNA as separate fractions using the AllPrep DNA/RNA Minikit (Qiagen), following the manufacturer's instructions. We quantified M. catarrhalis copB DNA using real-time PCR on a 7500 Real Time PCR System (Applied Biosystems) using the specific primers copB-F (59-GTGAGTGCCGCTTTACAACC-39) and copB-R (59-TGTATCGCCT GCCAAGACAA-39), as previously described (99). We calibrated levels of copB to CFUe using a standard curve as previously described (98). Briefly, we cultured M. catarrhalis cultured overnight and quantified the CFU using standard methods. In parallel, we extracted DNA from an aliquot of the same culture using an AllPrep Power Viral kit (Qiagen) and prepared serial dilutions of the DNA corresponding to 10 2 to 10 6 CFU.
Statistical analysis and data visualization. All statistical analyses were performed in R with specific packages as described in the preceding sections. We generated graphics using ggplot2 (100) with some cleanup and assembly in InkScape.
Data availability. The raw amplicon sequences used for bacterial community analysis are deposited in the Short Read Archive under BioProject accession no. PRJNA866994. Genome sequences generated from this work are deposited under BioProject accession no. PRJNA867425. All other raw and derived data are available at https://doi.org/10.6084/m9.figshare.20444466. The scripts necessary to replicate this work are available at https://github.com/reedstubbendieck/rothia_moraxella.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.