ABSTRACT
Resources available in the human nasal cavity are limited. Therefore, to successfully colonize the nasal cavity, bacteria must compete for scarce nutrients. Competition may occur directly through interference (e.g., antibiotics) or indirectly by nutrient sequestration. To investigate the nature of nasal bacterial competition, we performed coculture inhibition assays between nasal Actinobacteria and Staphylococcus spp. We found that isolates of coagulase-negative staphylococci (CoNS) were sensitive to growth inhibition by Actinobacteria but that Staphylococcus aureus isolates were resistant to inhibition. Among Actinobacteria, we observed that Corynebacterium spp. were variable in their ability to inhibit CoNS. We sequenced the genomes of 10 Corynebacterium species isolates, including 3 Corynebacterium propinquum isolates that strongly inhibited CoNS and 7 other Corynebacterium species isolates that only weakly inhibited CoNS. Using a comparative genomics approach, we found that the C. propinquum genomes were enriched in genes for iron acquisition and harbored a biosynthetic gene cluster (BGC) for siderophore production, absent in the noninhibitory Corynebacterium species genomes. Using a chrome azurol S assay, we confirmed that C. propinquum produced siderophores. We demonstrated that iron supplementation rescued CoNS from inhibition by C. propinquum, suggesting that inhibition was due to iron restriction through siderophore production. Through comparative metabolomics and molecular networking, we identified the siderophore produced by C. propinquum as dehydroxynocardamine. Finally, we confirmed that the dehydroxynocardamine BGC is expressed in vivo by analyzing human nasal metatranscriptomes from the NIH Human Microbiome Project. Together, our results suggest that bacteria produce siderophores to compete for limited available iron in the nasal cavity and improve their fitness.
IMPORTANCE Within the nasal cavity, interference competition through antimicrobial production is prevalent. For instance, nasal Staphylococcus species strains can inhibit the growth of other bacteria through the production of nonribosomal peptides and ribosomally synthesized and posttranslationally modified peptides. In contrast, bacteria engaging in exploitation competition modify the external environment to prevent competitors from growing, usually by hindering access to or depleting essential nutrients. As the nasal cavity is a nutrient-limited environment, we hypothesized that exploitation competition occurs in this system. We determined that Corynebacterium propinquum produces an iron-chelating siderophore, and this iron-sequestering molecule correlates with the ability to inhibit the growth of coagulase-negative staphylococci. Furthermore, we found that the genes required for siderophore production are expressed in vivo. Thus, although siderophore production by bacteria is often considered a virulence trait, our work indicates that bacteria may produce siderophores to compete for limited iron in the human nasal cavity.
INTRODUCTION
Humans engage in symbioses with diverse sets of microbes at nearly every body site. Collectively, these microbes are referred to as the human microbiota. Members of the microbiota provide their human host with essential services. For example, within the gastrointestinal tract, mutualistic bacteria metabolize recalcitrant nutrients to release metabolites that are accessible to the human host (1, 2); synthesize many essential amino acids, cofactors, and vitamins (3, 4); and provide defense against pathogens (5, 6). However, despite providing critical services to their hosts, nutrient acquisition is often a principal challenge for bacteria colonizing humans and other animals. In part, this is due to host sequestration of resources as a means to control bacterial growth (7).
To form and maintain associations with humans and other eukaryotes, bacteria have evolved mechanisms to derive nutrition from their hosts (8, 9). Both commensal and pathogenic bacteria colonizing the gastrointestinal tract consume host-derived compounds (10). For instance, Bacteroides acidifaciens and Akkermansia muciniphila cells have been shown to incorporate carbon and nitrogen from host proteins (11), and the pathogen Salmonella enterica serotype Typhimurium uses tetrathionate derived from thiosulfate that is produced by the host inflammatory response during infection as a terminal electron acceptor (12). In addition to consuming host-derived products, bacteria colonizing the gastrointestinal tract can also acquire nutrients from meals that their host consumes (10). However, outside the gastrointestinal tract, bacteria have increasingly limited access to nutrients. Cutibacterium (Propionibacterium) acnes colonizing the sebaceous glands in human skin hydrolyzes sebum triglycerides to generate free fatty acids and glycerol (13, 14), which it may metabolize (15). Similarly, anaerobic oral bacteria that have spread to the lower respiratory tracts of cystic fibrosis patients may ferment host mucins to generate short-chain fatty acids and amino acids to feed other bacteria, including Pseudomonas aeruginosa (16).
To colonize the upper airway, bacteria must attach to the epithelial surface, accommodate or evade the host’s immune system, and cope with decreased access to freely available nutrients compared to bacteria colonizing the gastrointestinal tract (reviewed in reference 7). Within the human nasal cavity, there are minute concentrations of free amino acids, carbohydrates, organic acids, and minerals (17–19). For comparison, nasal secretions contain ∼65-fold-lower glucose concentrations than the lumen of the small intestine (19, 20). Furthermore, humans actively deplete the available glucose in the airway through polarized glucose importers in the membranes of airway epithelial cells (21). Similar to other environments such as the soil or the ocean, the bioavailability of iron in the nasopharynx is low (17, 18). As a means to acquire scarce iron, bacteria and other microbes release iron-chelating molecules called siderophores that can scavenge ferric iron and other minerals from the environment (22). However, human hosts have evolved countermeasures to circumvent bacterial siderophores. For instance, bacterial colonization of the nasopharynx triggers neutrophils to produce lipocalin-2, a protein that binds to enterobactin-type siderophores and prevents their uptake by bacteria (23, 24). Finally, in addition to experiencing nutrient and mineral limitation, bacteria that colonize the nasal cavity are exposed to oxygen stress, which may be either abiotic or produced by the action of host immune cells (25, 26). Therefore, the human nasal cavity is a low-resource and high-stress environment, requiring bacteria in this environment to engage in competitive interactions for survival (27, 28).
Competition is split into two modes, called interference and exploitation. Bacteria engaging in exploitation competition compete by preventing their competitors from accessing resources, by either rapidly consuming or sequestering these supplies. In contrast, bacteria using interference competition produce toxic effectors to directly inhibit their competitors (29). Bacteria isolated from the human nasal cavity are well known to use specialized (secondary) metabolites with antimicrobial properties to engage in interference competition. As examples, Staphylococcus lugdunensis produces a thiazolidine-containing cyclic peptide called lugdunin that inhibits the growth of Staphylococcus aureus in vitro. In a patient population, nasal colonization by S. lugdunensis was significantly associated with decreased S. aureus colonization, suggesting that lugdunin is produced in vivo (30). Similarly, Corynebacterium accolens secretes a lipase that cleaves human nasal triacylglycerols to produce antimicrobial free fatty acids that inhibit the growth of Streptococcus pneumoniae in vitro. In a 16S rRNA gene amplicon sequencing survey, the abundance of Corynebacterium species sequencing reads was increased in children negative for pneumococcal colonization, indicating that C. accolens may inhibit S. pneumoniae in vivo (31). Finally, under conditions of iron limitation and hydrogen peroxide-induced oxidative stress, which reflect conditions experienced by bacteria colonizing the nasal cavity, Staphylococcus epidermidis IVK45 increases the production of an antimicrobial peptide called nukacin IVK45 in vitro (32). Despite these examples of interference competition and the low-resource environment of the nasal cavity, mechanisms of exploitation competition are not well studied among members of the human nasal microbiota (33).
In this study, we investigated bacterial competition between Actinobacteria and Staphylococcus, which are among the most abundant members of the human nasal microbiota (33, 34). We found that Corynebacterium (phylum Actinobacteria) strains vary in their ability to inhibit the growth of coagulase-negative staphylococci (CoNS). Using a comparative genomics approach, we identified a gene cluster for siderophore biosynthesis that is present in the genomes of Corynebacterium propinquum strains that more strongly inhibit CoNS. We confirmed siderophore production and demonstrated that iron sequestration was the mechanism of CoNS inhibition. We identified the siderophore as dehydroxynocardamine. Finally, we detected the expression of the dehydroxynocardamine biosynthetic gene cluster (BGC) in metatranscriptomic reads from the human nasal cavity. Together, the data suggest that members of the human nasal microbiota engage in exploitation competition for limiting iron and may influence the composition of the human nasal microbiota in vivo.
RESULTS
Interaction assays reveal variation in staphylococcal inhibition by nasal Actinobacteria.The scarcity of nutrients and minerals in the human nasal cavity results in a stressful environment where the microbiota must compete to survive. To identify mechanisms of competition that occur among members of the nasal microbiota, we chose to use a culture-based approach and investigated interactions between Actinobacteria and Staphylococcus spp., which are major bacterial colonizers of the human nasal cavity. For this study, we isolated bacteria from frozen nasal lavage samples donated by healthy children as part of the Childhood Origins of Asthma (COAST) study (35, 36), which we identified by colony morphology, pigmentation, hemolysis, and 16S rRNA gene sequence (see Materials and Methods).
To identify differences in patterns among interactions between Actinobacteria and Staphylococcus spp., we assessed the ability of a subset of Actinobacteria isolates to inhibit the growth of a subset of Staphylococcus species isolates. In total, we tested 21 Actinobacteria isolates (10 Corynebacterium, 1 Curtobacterium, 1 Dermabacter, 3 Kocuria, 1 Microbacterium, 3 Micrococcus, and 2 Rothia isolates) against 39 Staphylococcus species isolates (15 S. aureus isolates, 5 Staphylococcus warneri/S. pasteuri isolates, and 19 other CoNS) (see Table S1 in the supplemental material), for a total of 812 pairwise combinations with ≥2 replicates. In the case of 7 combinations, the replicates were not in agreement, or the Actinobacteria isolate overgrew its well and prohibited inoculation of the Staphylococcus species isolate. We removed these 7 points from further analysis. We scored inhibition by visual inspection of the Staphylococcus species colonies after 1 week of coincubation. Strong inhibition corresponded to colonies with severe growth defects or total growth inhibition, whereas weak inhibition corresponded to colonies with moderate inhibition resulting in diminished growth. Staphylococcus species colonies that were indistinguishable from a monoculture control were considered uninhibited (Fig. 1A). To visualize patterns in the inhibition assays, we plotted the inhibition scores as a heat map that was clustered based on the phylogenetic relationships of the Actinobacteria and Staphylococcus species isolates on the horizontal and vertical dimensions, respectively. The Staphylococcus species phylogenetic tree was built from 16S rRNA gene sequence alignments, whereas the Actinobacteria phylogenetic species tree was built from alignments of 93 single-copy core bacterial genes (Fig. 1B).
Inhibition of nasal Staphylococcus spp. by nasal Actinobacteria. (A) Representative images of monocultures of Staphylococcus and cocultures of Actinobacteria (left) with Staphylococcus species isolates (right). The inhibition scoring system is depicted below the coculture images. (B) Twenty-one nasal Actinobacteria isolates (horizontal) were monocultured on BHI agar wells for 1 week before 39 Staphylococcus species isolates (vertical) were spotted adjacent to the Actinobacteria colony. The colonies were cultured together for 1 week before inhibition of the Staphylococcus species colony was scored. The heat map displays the inhibition scores of each Staphylococcus species isolate when paired with the corresponding Actinobacteria isolate. Each interaction was technically replicated at least twice. The gray cells indicate interactions where replicates were in disagreement or the Actinobacteria colony overgrew the well before Staphylococcus inoculation. (Left) Phylogenetic tree of Staphylococcus species isolates built from 685 bp of the 16S rRNA gene amplified with the universal primers 27F and 1492R. (Top) Core-genome phylogenetic tree of Actinobacteria built from 93 conserved, single-copy genes. Both phylogenies are rooted on B. subtilis 168, and nodes with ≥50% bootstrap support are indicated. The dashed yellow lines highlight strong inhibition of CoNS by one clade of Corynebacterium spp. Strains in boldface type are siderophore producers as measured by the CAS assay. Actinobacteria taxa marked with ● harbor a siderophore BGC within their genome.
By clustering the inhibition scores with respect to phylogeny, we immediately observed significant differences in the inhibition patterns between S. aureus and CoNS. Broadly, most CoNS isolates were inhibited by nasal Actinobacteria, but the S. aureus isolates were insensitive (χ2 = 371.73; df = 2; P < 2.2e−16) (Fig. 1B). However, the inhibition pattern was not solely dominated by the identity of the Staphylococcus species; we observed variation among Corynebacterium species isolates in inhibiting CoNS. In particular, we noticed three Corynebacterium species isolates forming a monophyletic clade (highlighted in orange in Fig. 1B) that more strongly inhibited the growth of CoNS than the other Corynebacterium species isolates (χ2 = 89.74; df = 2; P < 2.2e−16) (Fig. 1B). Given the limited variation in staphylococcal inhibition scores by other genera of Actinobacteria, we focused on Corynebacterium spp. to determine underlying factors responsible for the differences between weak and strong CoNS inhibitors.
Corynebacterium propinquum strongly inhibits CoNS.To identify the Corynebacterium strains, we constructed a core-genome phylogeny from the 10 Corynebacterium spp. that we isolated and 45 human-associated strains whose sequences were available from GenBank. All three of the strong inhibitors of CoNS were most closely related to Corynebacterium propinquum, while the weak inhibitors of CoNS were closely related to either Corynebacterium pseudodiphtheriticum (5 isolates) or “Corynebacterium genitalium” (1 isolate), which is currently not a recognized Corynebacterium species name (37) (Fig. S1). For clarity, we refer to the Corynebacterium strong inhibitors of CoNS as C. propinquum and the other isolates as C. genitalium or C. pseudodiphtheriticum. We note that all three C. propinquum isolates were obtained from the same individual and may represent replicate isolates instead of distinct strains (Table S1).
Iron acquisition genes are associated with increased CoNS inhibition.To identify differences among the Corynebacterium species strains that corresponded to differential CoNS inhibition, we took a comparative genomics approach. We generated draft genome sequences for each of the 10 Corynebacterium species isolates. Using the OMA algorithm, we predicted orthologs across the 10 genomes. In total, we predicted 2,702 ortholog groups, with 1,198 groups (44% total) universally shared across the 10 Corynebacterium species strains and 1,504 ortholog groups (56% total) that were absent in at least one strain. In particular, the gene content in C. genitalium HSID17239 was more divergent than those of the other isolates (Fig. 2A). Indeed, when C. genitalium HSID17239 was excluded from the analysis, 1,801 of the 2,702 (67% total) ortholog groups were shared universally across the remaining nine strains. Of the 2,702 ortholog groups, 1,106 were predicted to encode hypothetical or putative proteins, while the remaining 1,596 groups were annotated with functions (Table S2). There were 337 ortholog groups (12% of the total ortholog groups) that were uniquely shared among C. propinquum isolates (Fig. 2A). Although 223 of the orthologs were annotated as encoding hypothetical or putative proteins, a manual investigation of the 114 annotated orthologs revealed that 31 (27.2% of the subset) were annotated with functions related to iron acquisition and/or siderophore biosynthesis (Table S2). This result represents a 6.4-fold enrichment over the 4.3% abundance (68/1,596 orthologs) of the corresponding terms in all 10 Corynebacterium genomes (expected, 4.9%; P = 1.3e−19 [by a hypergeometric test]). As a second test, we identified Kyoto Encyclopedia of Genes and Genomes (KEGG) orthology (KO) terms that were enriched in C. propinquum genomes relative all 10 Corynebacterium species genomes. We annotated 169 of the 337 orthologs unique to C. propinquum (50% of the subset) with 117 KO terms. Compared to the total set of 357 KO terms, the 169 orthologs were enriched in 12 KO terms. In particular, we noted that among the KO terms, 2 KO terms were associated with iron acquisition and siderophore biosynthesis: (i) iron complex transport system permeases and (ii) bifunctional isochorismate lyase/aryl carrier proteins (Fig. 2B).
Comparative genomics of nasal Corynebacterium spp. (A, left) Core-genome phylogeny of nasal Corynebacterium from Fig. 1. The other Actinobacteria node is collapsed, as represented by a triangle. The C. propinquum clade is shaded in orange. (Right) Clustered presence-absence matrix of 1,504 orthologs predicted across the 10 Corynebacterium species genomes. The 1,198 orthologs that were conserved across all 10 genomes were excluded. Orthologs encoding functions pertaining to iron metabolism, iron transport, or siderophore biosynthesis are depicted in red. (B) KO term enrichment analysis of the orthologs uniquely shared among C. propinquum strains relative to all orthologs encoded within all 10 Corynebacterium species genomes. The odds ratios of all 12 significantly enriched (P < 0.05) KO terms are plotted. Note that odds ratios of infinity (Inf) correspond to KO terms where all orthologs that were annotated with the KO term were present in the enrichment set. KO terms in boldface type are related to iron transport or siderophore biosynthesis. The color of each point indicates the significance level, and the size of the point indicates how many of the orthologs were annotated with the KO term.
Corynebacterium propinquum produces a siderophore.The results of the gene and KO term enrichment analyses (Fig. 2) led us to hypothesize that the observed variation in inhibition of CoNS by Corynebacterium spp. (Fig. 1B) may be due to siderophore production. To directly detect siderophore production, we used the chrome azurol S (CAS) assay, which detects siderophore production as a color change from blue to yellow because siderophores are able to remove iron from the CAS dye complex (38). Consistent with our previous results, we confirmed siderophore production by all three C. propinquum strains, whereas none of the six C. pseudodiphtheriticum isolates or the single C. genitalium isolate produced detectable siderophore activity (Fig. 3A and Table S1).
Siderophore production by C. propinquum inhibits CoNS. (A) CAS assay results from C. propinquum HSID18034 (left) and C. genitalium HSID17239 (right), which strongly and weakly inhibit CoNS, respectively. A shift in the color of the CAS overlay from blue to yellow indicates siderophore production. (B) Coculture plate inhibition between the strains in panel A and CoNS on BHI medium (−Iron) and BHI medium supplemented with 200 μM FeCl3 (+Iron). (C) Coculture plate inhibition assays with C. propinquum (HSID18034 and HSID18036) and 5 Corynebacterium species non-siderophore producers (HSID17231, HSID17239, HSID17260, HSID17564, and HSID17575) against 22 CoNS strains on BHI medium (−Iron) and BHI medium supplemented with 200 μM FeCl3 (+Iron). Each pairing was duplicated with consistent results. The error bars represent the standard errors of the means. The photographs in panels A and B are representative of results for duplicate samples.
Iron supplementation rescues CoNS from inhibition by C. propinquum.The results from the CAS assay (Fig. 3A) indicated that siderophore production is correlated with strong inhibition of CoNS. To directly test if iron sequestration by C. propinquum was responsible for CoNS inhibition, we repeated the coculture plate inhibition assays on brain heart infusion (BHI) medium and on BHI medium supplemented with 200 μM FeCl3. We found that iron supplementation rescued CoNS from inhibition by C. propinquum, and non-siderophore producers were unable to inhibit CoNS under the conditions of this assay (Fig. 3B). We performed the assay with 2 C. propinquum siderophore producers and 5 non-siderophore producers (C. genitalium HSID17239 and 4 strains of C. pseudodiphtheriticum) against 22 strains of CoNS isolated from different nasal specimens. There were significant differences among the four experimental groups (χ2 = 291.16; df = 3; P < 2.2e−16 [by a Kruskal-Wallis test]), and post hoc analysis using Dunn’s pairwise comparisons indicated that the inhibition of CoNS by C. propinquum on normal BHI medium was significantly increased relative to that under all other conditions (P < 1.0e−40) (Fig. 3C). This result indicates that inhibition of CoNS in vitro by C. propinquum is likely due to iron sequestration. Alternatively, iron supplementation may alter the transcription profile and repress the production an alternative factor produced by C. propinquum that inhibits the growth of CoNS independent of siderophore production.
The C. propinquum-produced siderophore is dehydroxynocardamine.As a first step to identify the siderophore produced by C. propinquum, we used antiSMASH to identify BGCs for siderophore biosynthesis in the Corynebacterium species genomes that we sequenced for this study. We identified a single 12,553-bp siderophore BGC that was present within all three C. propinquum genomes but absent in C. genitalium HSID17239 and all six C. pseudodiphtheriticum genomes. The C. propinquum siderophore BGC is comprised of seven open reading frames (ORFs), which were annotated with functions consistent with iron acquisition and siderophore biosynthesis (39–41) (Fig. 4A). Specifically, in addition to transport-associated genes, this BGC encodes a pyridoxal phosphate (PLP)-dependent decarboxylase (dnoB), an l-lysine N6-monooxygenase (dnoC), and a siderophore synthetase (dnoD) that contains acyl-CoA N-acyltransferase (InterPro accession no. IPR016181 [http://www.ebi.ac.uk/interpro/entry/IPR016181]), aerobactin siderophore biosynthesis, IucA/IucC, N-terminal (InterPro accession no. IPR007310 [http://www.ebi.ac.uk/interpro/entry/IPR007310]), and ferric iron reductase FhuF (InterPro accession no. IPR022770 [http://www.ebi.ac.uk/interpro/entry/IPR022770]) domains. Although there were no experimentally characterized BGCs that were identical to the C. propinquum siderophore BGC, the order of the biosynthetic genes was identical to that of the desferrioxamine E dfo BGC from Pantoea agglomerans strain B025670 (MIBiG accession no. BGC0001572 [https://mibig.secondarymetabolites.org/repository/BGC0001572/index.html]). The biosynthesis enzymes encoded by both siderophore BGCs are highly similar (Table 1), and both synthetases belong to the type C′ nonribosomal peptide synthetase-independent family (40, 41), which indicated that the siderophore produced by C. propinquum is likely macrocyclic.
Corynebacterium propinquum produces the siderophore dehydroxynocardamine. (A) The dehydroxynocardamine BGC from C. propinquum. ORFs encoding biosynthetic and transport-related functions are filled with gray and white, respectively. See Table 1 for specific ORF annotation. The DtxR sequence motif is shown upstream of dnoB. (B) Subset of the molecular network of C. propinquum HSID18034 and C. genitalium HSID17239 agar core extracts. Gray nodes are metabolites shared by both strains, and black nodes are metabolites unique to C. propinquum. There were no detected metabolites that were unique to C. genitalium. The edges are weighted to the cosine score between the two features. A single cluster unique to C. propinquum is outlined in a dashed box. See Fig. S2 in the supplemental material for the full molecular network. (C) Zoomed-in view of the single cluster unique to C. propinquum with the nodes labeled by their corresponding m/z. This cluster contains single- and double-charged states of dehydroxynocardamine, dehydroxynocardamine B, and dehydroxynocardamine C. (D) Structures of dehydroxynocardamine, dehydroxynocardamine B, and dehydroxynocardamine C.
Corynebacterium propinquum siderophore BGC annotationsa
To identify the siderophore produced by C. propinquum, we used comparative mass spectrometry-based metabolomics. We cultured C. propinquum HSID18034 and C. genitalium HSID17239 on BHI agar plates. We sampled agar cores near the bacterial colonies, which we extracted using methanol and analyzed using data-dependent liquid chromatography-tandem mass spectrometry (LC-MS/MS). To visualize differences in the metabolomic profiles of the two strains, we used mass spectral molecular networking (42). Surprisingly, in a network containing 222 clusters and 1,531 singletons (Fig. 4B and Fig. S3), we observed only a single cluster that was unique to C. propinquum HSID18034 (Fig. 4C). The central node in this cluster corresponded to m/z 585.358 [M + H]+ (C27H48N6O8, Δppm = 3.44). The molecular weight and tandem mass spectrometry fragmentation pattern matched those of the siderophore dehydroxynocardamine (m/z 585.361 [M + H]+, C27H49N6O8), which is also called terragine E (Fig. 4D). Dehydroxynocardamine has been previously reported from marine sponge-associated Streptomyces spp. (43) and produced by the heterologous expression of soil DNA libraries in Streptomyces lividans (44) but has not been reported from Corynebacterium spp. or any other human-associated bacteria. The other nodes in the cluster correspond to two dehydroxynocardamine analogs with protonated ions of m/z 569.364 [M + H]+ and 553.368 [M + H]+ as well as the doubly charged states of each molecule (Fig. 4C). The high similarity in fragmentation patterns as well as the differences of m/z −16 and −32 suggest that these compounds contained one and two fewer N-hydroxy groups than dehydroxynocardamine, respectively (Fig. 4D and Fig. S3). Furthermore, through manual investigation of the mass spectra, we identified m/z 312.6592 [M + Fe + H]2+, corresponding to iron-bound dehydroxynocardamine B, indicating that dehydroxynocardamine functions as a siderophore. This parental mass was absent from the molecular network as its abundance was below the threshold for fragmentation. Based on the product, we chose to name the C. propinquum siderophore BGC dno, for dehydroxynocardamine.
The dno BGC is strongly associated with C. propinquum.To determine if other Corynebacterium spp. contain the dno BGC, we searched all available Corynebacterium species proteins in the NCBI nonredundant protein sequence (nr) database for the siderophore synthetase DnoD. We identified 17 hits in the nr database corresponding to 27 Corynebacterium species genome sequences. When we searched these genomes, we found that all seven ORFs of the dno BGC were contained by all four C. propinquum strains with deposited genomes, but the full dno BGC was not contained by any other Corynebacterium genome (Fig. S4A and B).
Next, we searched external naris metagenomes from 93 subjects in the Human Microbiome Project (HMP) WGS-PP1 (WGS production phase I) to determine if other nasal bacteria contain the dno BGC. We identified 258 contigs from a total of 718,888 contigs (0.04% total) that harbored at least one out of seven ORFs from the dno BGC. Of these 258 contigs, only three contigs from 2/93 total human subjects harbored all seven ORFs (Fig. S4C and D). These three contigs (lengths of 13,073, 97,360, and 178,113 bp) aligned to the C. propinquum HSID18034 genome with ≥98% sequence identity and ≥95% total sequence coverage, which suggests that these contigs were from C. propinquum. Together, these genomic and metagenomic searches indicate a strong phylogenetic signal for the association of C. propinquum with the dno BGC among Corynebacterium spp.
The dehydroxynocardamine BGC is expressed in vivo.We identified a 19-bp putative diphtheria toxin repressor (DtxR) iron box operator sequence (5′-TTATGCAAGGTTTTCCTAT-3′) (45) situated upstream of dnoB (Fig. 4A). When iron is abundant, DtxR forms a complex with ferrous iron and binds iron box operator sequences to repress the transcription of downstream genes, but this repression is relieved when iron is scarce (46). Transcriptional profiling of S. aureus from the nasal cavity indicated that bacteria colonizing the nose are iron starved in vivo (47). Therefore, we wanted to determine if C. propinquum strains express the dno BGC in vivo. As part of the HMP prediabetes study (48), nasal swabs from the anterior nares were collected from 16 adult individuals over multiple visits and processed for metatranscriptome sequencing. We analyzed these metatranscriptomes for the expression of the dno BGC and several housekeeping genes (rpoB, gyrB, sigA, and rpsL). We detected dno expression in 7/16 individuals from at least one visit (mean, 2.5 visits/subject; median, 1 visit/subject), with expression values ranging from 0.01 to 154.69 transcripts per kilobase per million (TPM) (mean, 26.94 TPM; median, 3.44 TPM). For comparison, we detected the expression of housekeeping genes at similar levels (Fig. 5), with the exception of rpsL, which is known to be more highly expressed than the other selected housekeeping genes in the model Gram-positive organism Bacillus subtilis (49). Together, these results indicate that the dehydroxynocardamine BGC is likely expressed in vivo.
The dehydroxynocardamine BGC is expressed in vivo. Ninety-five nasal metatranscriptomes from 16 subjects in the NIH HMP were pseudoaligned to the genome of C. propinquum HSID18034 to detect the expression of the dehydroxynocardamine BGC (dnoA-G) (white) and the housekeeping genes gyrB, rpoB, sigA, and rpsL (gray). The upper and lower bounds of the notched box plot indicate the 75th and 25th percentiles, respectively. The bars indicate the medians, and the notches represent the 95% confidence intervals of the medians. Each point indicates the expression of a single gene from a metatranscriptome in transcripts per kilobase per million (TPM).
DISCUSSION
In this study, we investigated inhibition of nasal Staphylococcus species isolates by strains of sympatric Actinobacteria. We uncovered two distinct patterns of interactions: (i) CoNS were significantly more sensitive than S. aureus to inhibition by nasal Actinobacteria, and (ii) there was variability among Corynebacterium spp. in their ability to inhibit the growth of CoNS (Fig. 1). Specifically, we noted that C. propinquum more strongly inhibited CoNS than did other Corynebacterium spp. Through a combination of comparative genomics (Fig. 2) and culture-based approaches (Fig. 3), we determined that the difference in inhibitory ability was due to the production of a siderophore, dehydroxynocardamine (Fig. 4). By analyzing metatranscriptome samples from the anterior nares of human subjects, we established that the dno BGC is expressed in vivo (Fig. 5), indicating that iron-mediated exploitation competition is potentially relevant to competition and organismal fitness within the human nasal cavity.
One broad explanation for the difference in inhibition patterns between S. aureus and CoNS isolates that we observed (Fig. 1B) is that CoNS lack the same mechanisms that S. aureus uses to obtain iron from the hosts and from the external environment. For instance, CoNS strains do not grow well on iron-limited media (19) and, unlike S. aureus, are unable to scavenge iron from human transferrin (50). Our finding that iron supplementation rescues CoNS from inhibition by C. propinquum (Fig. 3B and C) is consistent with previous results showing that CoNS are sensitive to inhibition by desferrioxamine siderophores and other iron-chelating agents in vitro (51, 52). Notably, strains of both S. aureus and CoNS are known to produce their own carboxylate siderophores, called staphyloferrins A (53) and B (54). Using the CAS assay, we determined that all 15 isolates of S. aureus and nearly every CoNS isolate (30/33 tested) used in this study produced siderophores when grown in monoculture (Fig. 1B; see also Table S1 in the supplemental material). However, it is known that CoNS have markedly lower levels of siderophore production than S. aureus (55). Furthermore, siderophore production by CoNS is under different regulatory control than in S. aureus (55), which may contribute to the increased susceptibility of CoNS to iron limitation imposed by C. propinquum. In addition, it was previously reported that Staphylococcus spp. are able to use catecholate and hydroxamate siderophores produced by Corynebacterium spp. and other organisms (56). Perhaps, in addition to higher levels of siderophore production, S. aureus is also better able to pirate dehydroxynocardamine produced by C. propinquum and circumvent inhibition. Further work is required to determine the exact mechanism that S. aureus uses to survive competition with C. propinquum compared to CoNS.
Here, we note that siderophore-mediated inhibition does not fully explain the patterns of inhibition of CoNS by nasal Actinobacteria. In addition to the 3 siderophore-producing C. propinquum strains, only 6 of the 11 other Actinobacteria contained siderophore BGCs in their genomes (Fig. 1B). Furthermore, while the genomes of Curtobacterium sp. strain HSID17257 and Kocuria sp. strain HSID16901 contain siderophore BGCs, these strains did not produce siderophores (Fig. 1B and Table S1). Regardless, the latter strain strongly inhibited CoNS under our assay conditions (Fig. 1B). Together, these data indicate that nasal Actinobacteria likely use multiple types of mechanisms to inhibit the growth of CoNS. Perhaps these Actinobacteria engage in interference competition mediated by antibiotic production, which is common among nasal bacteria (30–32), to inhibit the growth of CoNS. Subsequent work will be required to identify the molecules responsible for inhibition of CoNS that are produced by non-siderophore-producing nasal Actinobacteria.
Exploitation competition mediated by siderophore production is common among soil and marine bacteria (57–59), but within the context of host-associated systems, siderophore production is considered a pathogen virulence factor (reviewed in reference 60) because mutants that are unable to produce siderophores are often defective in colonizing hosts and causing disease (61, 62). The view of siderophores solely as pathogen-produced molecules is strengthened by lipocalin-2, a protein component of the innate immune response that binds to catechol siderophores to prevent their uptake by bacteria and whose production is induced by bacterial colonization of mucosal surfaces (23, 24). However, thus far, little work has considered siderophore-mediated competition among the members of the microbiota and between the microbiota and pathogenic bacteria (63). For instance, siderophore piracy is reported to occur between pathogens and beneficial or commensal bacteria in the human gastrointestinal tract (64, 65), but no such mechanisms within the human nasal cavity have hitherto been reported.
In conclusion, as (i) dehydroxynocardamine is not a known virulence factor, (ii) C. propinquum is considered a normal part of the nasal microbiota (33, 66, 67), and (iii) the expression of dnoA to dnoG (dnoA-G) was detected in vivo, we suggest that C. propinquum may produce dehydroxynocardamine as a means to mediate exploitation competition for iron with other bacteria within the human nasal cavity.
MATERIALS AND METHODS
Nasal lavage specimen collection.The nasal lavage samples that we used in this study were banked as part of the COAST study (35). Informed consent was obtained from parents, and the Human Subjects Committee at the University of Wisconsin—Madison approved the study (institutional review board [IRB] approval number H-2013-1044). Briefly, lavage samples were collected by spraying Deep Sea nasal spray (Major) into one of the participant’s nostrils. To collect the sample, the participant was then instructed to blow their nose into a plastic bag. Subsequently, the samples were stored at 4°C. Approximately 3 ml of either phosphate-buffered saline or Amies transport medium (Copan Diagnostics) was added to each sample before storage at −80°C.
Strain isolation and maintenance.For general strain propagation, we used brain heart infusion (BHI) medium (DOT Scientific). All plates were solidified using 1.5% agar (VWR). For iron supplementation experiments, we added 200 μM FeCl3 to the BHI agar after autoclaving and cooling the medium to 55°C before pouring the plates. Bacterial strains used in this study are listed in Table S1 in the supplemental material. To culture bacteria from the lavage samples, we spread 100 μl of each of the thawed samples onto BHI plates and incubated the 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 at 37°C until we obtained pure cultures. All bacterial isolates were cryopreserved at −80°C in 25% glycerol.
Bacterial isolate identification.To identify bacterial isolates, we sequenced the 16S rRNA gene. We used colony PCR with the universal 27F (5′-AGAGTTTGATCMTGGCTCAG-3′) and 1492R (5′-CGGTTACCTTGTTACGACTT-3′) primers to amplify the 16S rRNA gene (68). We sequenced PCR products using the Sanger method at the University of Wisconsin—Madison Biotechnology Center. We identified isolates to the genus level using the Ribosomal Database Project Classifier (69). To distinguish between S. aureus and CoNS (e.g., Staphylococcus capitis, Staphylococcus caprae, S. epidermidis, and Staphylococcus saccharolyticus) isolates, we used BLAST searches against the NCBI 16S rRNA sequence database with a threshold sequence identity of 99%. We corroborated the identity of Staphylococcus species isolates using colony pigmentation and hemolysis phenotypes (70–72).
16S rRNA gene phylogenetic analysis.We aligned the 16S rRNA gene sequences using SINA (v1.2.11) (73) through the Silva Web server (74) and discarded unaligned bases at the ends of the sequences. For phylogenetic analysis, we imported the alignments into MEGA7 (75). We inferred phylogenetic trees using the maximum likelihood method based on the general time-reversible (GTR) model with 100 bootstraps. All nucleotide positions containing gaps were removed in the final analysis.
Coculture plate inhibition assays.We used coculture plate inhibition assays to assess the activity of Actinobacteria isolates against Staphylococcus. We spread Actinobacteria from saturated overnight-grown cultures in BHI broth over one half of a well (diameter, 2.4 cm) containing 3 ml of BHI agar on a 12-well plate (Greiner Bio-One). We incubated the plates for 1 week at 37°C. Subsequently, we spotted ∼3 μl of a 10-fold-diluted saturated overnight-grown culture of each Staphylococcus species isolate in BHI medium adjacent to the Actinobacteria and returned the plates to the 37°C incubator. After 1 week of coculture, we removed the plates, scanned them, and scored the inhibition for each interaction pair. All interactions were tested with ≥2 replicates, with consistent results. An inhibition score of 0 indicated no inhibition, a score of 1 indicated weak inhibition, and a score of 2 indicated strong inhibition (Fig. 1A).
Whole-genome sequencing and assembly.We cultured bacterial strains for whole-genome sequencing in 3 ml of BHI broth supplemented with 0.5% glycine overnight at 37°C. We harvested the cells by centrifuging the cultures at 21,330 × g for 5 min. The cell pellets were washed with 10.3% sucrose and resuspended in 450 μl of a lysozyme solution (3 mg/ml lysozyme [Sigma], 0.3 M sucrose, 25 mM Tris-HCl [pH 8], 25 mM EDTA [pH 8]) for 30 min at 37°C. Subsequently, 13 μl of 20 mg/ml proteinase K (Thermo Fisher) was added to each sample, with additional incubation at 42°C for 15 min. Cells were lysed by adding 250 μl of 2% SDS and rocking the mixture for 15 min. DNA was purified using standard phenol-chloroform extraction and precipitated with 3 M sodium acetate and isopropanol. We visually assessed the quality of genomic DNA preparations by running samples on 0.5% Tris-borate-EDTA (TBE) gels. Genomic libraries for Illumina MiSeq 2- by 150-bp paired-end sequencing were prepared and sequenced by the University of Wisconsin—Madison Biotechnology Center. Raw reads were corrected with MUSKET v1.1 (76), and paired ends were merged with FLASH v1.2.7 (77). Reads were assembled into draft genome sequences with SPAdes v3.11.0 (78).
Core-genome phylogeny.We constructed a core-genome phylogenetic tree of nasal Actinobacteria as previously described (79). Briefly, we called genes in each genome using prodigal v2.6.0 (80) and used profile hidden Markov models in HMMER v3.1b2 (81) to search each genome for 93 full-length TIGRFAM amino acid sequences in the “core bacterial protein” set (GenProp0799). Each protein family was aligned using MAFFT v7.245 (82) and converted to codon alignments. We used RAxML v8.1.24 (83) to generate phylogenetic trees for each of the 93 codon alignments under the GTR gamma substitution model with 100 bootstraps. To generate the species phylogenetic tree, we used ASTRAL-II (84) from the individual phylogenetic trees with 100 bootstraps. We used FigTree v1.4.3 (http://tree.bio.ed.ac.uk/software/figtree/) to root the phylogeny on B. subtilis 168 and display the branch length based on proportional length to the root.
Comparative genomics.We annotated the assembled genomes using prokka v1.13 (85). For comparative genomics, we identified orthologs from the nucleotide sequences of each protein-coding open reading frame (ORF) with OMA v2.2.0 (86). To annotate orthologs with KEGG orthology (KO) terms, we used DIAMOND v0.9.21.122 (87) to query one sequence of each ortholog group against a custom database of KEGG-annotated protein sequences with an E value threshold cutoff of 1.0e−10. Each ORF was assigned the KO terms from the top BLAST hit. For KEGG KO term enrichment analysis, we used the GSEABase v1.42.0 package (88) and GOstats v1.7.4 (89) in R and performed a hypergeometric test with a P value cutoff of 0.05.
Identification of BGCs.To identify BGCs, we used antiSMASH 4.0.2 (90). Proteins from all identified BGCs were aligned via all-versus-all DIAMOND (query coverage, >75%; percent identity, >60%). BGCs that shared over 75% of proteins were called the same family. BGC families correlated with inhibition were identified through their presence in the strong-inhibition strains and absence in the weak-inhibition strains. Proteins from all identified BGCs were combined with all proteins from MIBiG v1.3 (91) and subsequently aligned with DIAMOND with the same parameters as the ones listed above to identify biosynthetic gene similarities to previously described BGCs.
Chrome azurol S assay.We used the overlay chrome azurol S (CAS) assay to test for siderophore production as previously described (92). We cultured Corynebacterium spp. overnight in 3 ml BHI medium at 37°C. We diluted the cultures grown overnight to an optical density at 600 nm (OD600) of 2 and spotted 5 μl onto 10-ml BHI agar plates (diameter, 5 cm). We incubated the plates at 37°C for 4 days to ensure robust Corynebacterium growth. Subsequently, we overlaid each plate with 6 ml of CAS reagent overlay [100 μM CAS, 200 μM hexadecyltrimethylammonium bromide, 10 μM FeCl3·6H2O, 10 mM piperazine-N,N′-bis(2-ethanesulfonic acid) (PIPES) (pH 6.8), 1% agarose] (38, 92) and incubated the plates at ambient temperature in the dark overnight before scanning plates for a color change from blue to yellow, indicating the production of a siderophore (38).
Siderophore identification and mass spectral molecular networking.To identify the siderophore, we used comparative mass spectrometry. We spotted 10 μl of either C. genitalium HSID17231 or C. propinquum HSID18034 cultures grown overnight onto the center of 25-ml BHI agar plates (diameter, 8.5 cm). After incubating the plates as described above, we removed agar cores (diameter, 0.6 cm) from an area near the Corynebacterium colony using the wide end of a P1000 pipette tip. We verified siderophore activity by directly placing an agar core onto CAS assay plates and observing the color change from blue to yellow. We washed a duplicate agar core in 2 ml of 50% methanol and 2 ml of 100% methanol, which we combined and dried with a gentle air stream under reduced pressure.
Data-dependent LC-MS/MS was performed on a Thermo Scientific Q Exactive Orbitrap instrument by the University of Wisconsin Analytical Instrumentation Center of the School of Pharmacy. The liquid chromatography method was performed on a Phenomenex XB C18, 2.1- by 100-mm, 2.6-μm-particle-size column with solvent A (water with 0.1% formic acid) and solvent B (acetonitrile with 0.1% formic acid). A gradient of 5% solvent B for 0.5 min to 30% solvent B over 16 min and then 97% solvent B for 2 min with a flow rate of 0.35 ml/min was used to separate the metabolites. We exported the files in mzXML format and uploaded them to Global Natural Products Social Molecular Networking (GNPS) (93). The data were filtered by removing all MS/MS peaks within ±17 Da of the precursor m/z. The MS/MS spectra were window filtered by choosing the top 6 peaks in the ±50-Da window throughout the spectrum. The data were clustered with MS-Cluster using a parent mass tolerance of 2.0 Da and an MS/MS fragment ion tolerance of 0.5 Da to create consensus spectra. Consensus spectra containing <2 spectra were discarded. A network was created where edges were filtered with a cosine score of >0.6 and more than 3 matched peaks. Edges between two nodes were kept in the network if each of the nodes appeared in the respective top 10 most similar nodes. The spectra in the network were then searched against GNPS spectral libraries. The library spectra were filtered in the same manner as the input data. All matches kept between network spectra and library spectra were required to have a score of >0.6 and at least 3 matched peaks. Analog search was enabled against the library, with a maximum mass shift of 100.0 Da.
Identification of the dehydroxynocardamine BGC in external naris metagenomes.We downloaded assembled metagenomes from the National Institutes of Health (NIH) Integrative Human Microbiome Project (iHMP) from the Human Microbiome Project Data Portal (https://portal.hmpdacc.org) using hmp_client. DNA was isolated from each of the 93 subjects between 1 and 3 times for a total of 227 external naris metagenomes (mean, 1.3/subject; median, 1/subject). We used the assembled metagenomes as a query against the amino acid sequences of each of the seven ORFs in the dehydroxynocardamine BGC using DIAMOND blastx (query coverage, ≥90%; percent identity, ≥40%).
Metatranscriptome analysis for dehydroxynocardamine BGC expression.As part of the iHMP, RNA from the anterior nares of 16 prediabetic subjects was isolated and depleted of 16S and 23S rRNAs. The RNA was converted to cDNA with random primers and paired-end sequenced using an Illumina platform (48). We downloaded the raw reads using hmp_client, as described above. We removed Illumina adapter sequences, low-quality reads, and unpaired reads using trimmomatic v0.36 (94) with the following parameter: SLIDINGWINDOW:5:20. RNA was isolated from each of the 16 subjects between 1 and 11 times for a total of 95 nasal metatranscriptomes (mean, 5.9/subject; median, 6.5/subject). For our analyses, we treated each individual metatranscriptome as a single sample. To determine the expression of the dno BGC, we used kallisto v0.44.0 (95) and pseudoaligned each of the metatranscriptomes onto the indexed ORFs of the C. propinquum HSID18034 genome. We chose to align the reads onto this genome because our genomic and metagenomic surveys (see Fig. S4 in the supplemental material) indicated that no other Corynebacterium species or nasal cavity-associated strain possessed the dno BGC. We extracted the expression data for the combined dehydroxynocardamine BGC (dnoA-G) and four housekeeping genes: rpoB, gyrB, sigA, and rpsL. We removed genes with <1 estimated count from the data set for a total of 148 expression counts (18 siderophore BGCs and 130 housekeeping genes).
Data availability.The genome sequences of Actinobacteria generated during this study have been deposited in the National Center for Biotechnology Information database under BioProject accession number PRJNA492917.
ACKNOWLEDGMENTS
We thank members of the Cameron R. Currie, James E. Gern, and Paul Straight (Texas A&M University) laboratories for feedback on the manuscript and project. We thank three anonymous reviewers for comments on the manuscript. We thank Aaron Stubbendieck for 3D printing stamps for rapid inoculation for coculture inhibition assays. We thank Samuel McInturf for thoughtful discussion and assistance in R scripting. We thank Robert Lemanske, Jr., and Daniel Jackson (University of Wisconsin—Madison), who are the principal investigators of the COAST birth cohort study, for providing nasal samples from which bacteria were isolated.
This work, including the efforts of Cameron R. Currie and James E. Gern, was funded by the NIH Centers for Excellence for Translational Research (U19-AI109673-01) and the NIH National Heart, Lung, and Blood Institute (P01 HL070831), respectively. Reed M. Stubbendieck was supported by a National Library of Medicine training grant to the Computation and Informatics in Biology and Medicine Training Program (NLM 5T15LM007359). The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
FOOTNOTES
- Received 1 October 2018.
- Accepted 14 December 2018.
- Accepted manuscript posted online 21 December 2018.
Supplemental material for this article may be found at https://doi.org/10.1128/AEM.02406-18.
- Copyright © 2019 American Society for Microbiology.
REFERENCES
- 1.↵
- 2.↵
- 3.↵
- 4.↵
- 5.↵
- 6.↵
- 7.↵
- 8.↵
- 9.↵
- 10.↵
- 11.↵
- 12.↵
- 13.↵
- 14.↵
- 15.↵
- 16.↵
- 17.↵
- 18.↵
- 19.↵
- 20.↵
- 21.↵
- 22.↵
- 23.↵
- 24.↵
- 25.↵
- 26.↵
- 27.↵
- 28.↵
- 29.↵
- 30.↵
- 31.↵
- 32.↵
- 33.↵
- 34.↵
- 35.↵
- 36.↵
- 37.↵
- 38.↵
- 39.↵
- 40.↵
- 41.↵
- 42.↵
- 43.↵
- 44.↵
- 45.↵
- 46.↵
- 47.↵
- 48.↵
- 49.↵
- 50.↵
- 51.↵
- 52.↵
- 53.↵
- 54.↵
- 55.↵
- 56.↵
- 57.↵
- 58.↵
- 59.↵
- 60.↵
- 61.↵
- 62.↵
- 63.↵
- 64.↵
- 65.↵
- 66.↵
- 67.↵
- 68.↵
- 69.↵
- 70.↵
- 71.↵
- 72.↵
- 73.↵
- 74.↵
- 75.↵
- 76.↵
- 77.↵
- 78.↵
- 79.↵
- 80.↵
- 81.↵
- 82.↵
- 83.↵
- 84.↵
- 85.↵
- 86.↵
- 87.↵
- 88.↵
- 89.↵
- 90.↵
- 91.↵
- 92.↵
- 93.↵
- 94.↵
- 95.↵
