Glycan-Dependent Corneocyte Adherence of Staphylococcus epidermidis Mediated by the Lectin Subdomain of Aap

ABSTRACT Staphylococcus epidermidis and other coagulase-negative staphylococci (CoNS) that colonize skin are known to promote skin immunity and inhibit colonization of pathogens that cause skin and soft tissue infections, including Staphylococcus aureus. However, S. epidermidis adherence to corneocytes, the cells that constitute the uppermost layer of the skin epidermis, remains poorly understood. Our study documents that S. epidermidis corneocyte adherence is dependent upon the accumulation-associated protein (Aap). Aap is composed of two distinct A and B domains. The A domain is comprised of a repeat region and a conserved L-type lectin domain, whereas the fibrillar B domain, which is comprised of G5 and E repeats, is linked to the cell wall in a sortase-dependent manner. Our studies revealed that adherence to corneocytes is dependent upon the lectin subdomain within the A domain. However, significant adherence was only observed when the lectin domain was expressed with both the A repeat and the B domain, suggesting further interactions between these three domains. Our data also suggest that the A repeat domain is important for stability or expression of Aap. Deglycosylation treatment suggested that glycans expressed in the host stratum corneum serve as potential binding partners for Aap-mediated corneocyte adherence. Last, bioinformatic analyses of the predominant commensal species of CoNS identified open reading frames (ORFs) homologous to aap, thus suggesting that Aap orthologues containing lectin-like domains may provide the basis for staphylococcal colonization of skin. Corroborating these observations, adherence to corneocytes in an S. aureus mgrA mutant was dependent upon SasG, the Aap orthologue in S. aureus.

was first identified by Schumacher-Perdreau et al. in 1994 (26) and was further studied by Hussain and colleagues in 1997 (27). Mitomycin mutagenesis led to the isolation of a biofilm-negative mutant, M7, which lacked a 140-kDa exoprotein (26). Hussain and colleagues found an association between this 140-kDa protein and enhanced biofilm production, and thus it was referred to as the accumulation-associated protein (Aap) (27).
Several studies have demonstrated that Aap is a rod-like fibril extending from the cell that can be observed by electron microscopy (28)(29)(30). Bioinformatic analyses suggest that Aap is a multidomain protein that consists of two distinct domains, A and B (31). A similar domain arrangement also exists in SasG, the Aap orthologue in S. aureus (22). Following extracellular export directed by the N-terminal signal sequence, the protein is anchored to the cell wall by sortase via the C-terminal LPDTG motif (32). Two independent studies found that the A domain, which is composed of a repeat region and an L-type lectin domain, functions to facilitate adherence to abiotic surfaces (32,33). Moreover, the N-terminal A domain is subjected to complete or partial proteolytic processing by the SepA metalloprotease (34,35). This enables dimerization of B repeats on adjacent bacteria, thus facilitating intercellular adhesion and bacterial accumulation (29,36,37). The C-terminal B domain is composed of B repeats, each of which contains a G5 module and an E spacer region. Structural analysis supports a "zinc zipper" model whereby interaction between G5 domains is zinc-dependent (29,30,37).
A study conducted by Macintosh et al. investigating S. epidermidis colonization of the skin identified Aap as one of the adhesins with a potential function in corneocyte attachment (38). The heterologous expression of Aap on the surface of Lactococcus lactis increased corneocyte adhesion 30-fold, suggesting that Aap facilitates adhesion to corneocytes independent of other adhesins. Additionally, the A domain of Aap was implicated in mediating this function, as the recombinant A domain of Aap partially blocked corneocyte binding by S. epidermidis NCTC 11047 and other aap-positive S. epidermidis clinical isolates. Moreover, another adhesin, serine-aspartate repeat-containing protein F (SdrF), has also been shown to bind recombinant proteins keratin 1 and 10, normal human epidermal keratinocytes (NHEKs), and human nasal epithelial cells (HNECs), suggesting a potential function in skin adherence (39).
In this study, we used isogenic mutants to document that S. epidermidis Aap mediates adherence to healthy skin. Furthermore, we constructed S. epidermidis strains expressing various Aap domains, and, in conjunction with blocking studies, documented that adherence to corneocytes is dependent upon the lectin subdomain within the A domain of Aap. Furthermore, we document that Aap-dependent adherence is significantly reduced following enzymatic cleavage of specific glycan structures present on the corneocytes of healthy skin. Last, we provide evidence of Aap orthologues in other staphylococcal species and document that S. aureus adheres to healthy skin in an SasG-dependent manner when sasG transcription is derepressed via an mgrA mutation.

RESULTS
Identification of the S. epidermidis factor responsible for adherence to corneocytes. To identify the factor(s) responsible for S. epidermidis adherence to skin, a corneocyte binding assay was developed and performed with an optical density (OD)-adjusted suspension of different S. epidermidis 1457 strains that constitutively express green fluorescent protein (GFP). Preliminary studies suggested that an OD at 600 nm (OD 600 ) of 0.15 (;10 7 CFU/ml) facilitated enumeration of adherent S. epidermidis cells (see Fig. S1 in the supplemental material). Since enhanced polysaccharide intercellular adhesin (PIA) production is a hallmark of strain 1457 (40), the function of PIA in skin adhesion was assessed using the 1457 and 1457 Dica strains. These experiments documented that binding of both 1457 and 1457 Dica strains to the corneocyte-adhered discs was similar, as assessed by confocal microscopy ( Fig. 1A and B). Thus, this finding suggests that S. epidermidis skin adhesion is independent of PIA and is consistent with the hypothesis that the primary bacterial attachment on the skin is a protein-dependent mechanism.
Sortase mediates the covalent binding of cell surface proteins containing an LPXTG motif to peptidoglycan (41). Indeed, inactivation of srtA reduces the localization of cell wall-anchored (CWA) proteins to the cell surface of S. epidermidis (32). To investigate the involvement of LPXTG-anchored surface protein adhesins in binding to the skin, a 1457 DsrtA mutant was constructed and investigated for its corneocyte binding efficiency. Both the 1457 DsrtA and 1457 Dica DsrtA mutants displayed negligible binding to corneocytes ( Fig. 1A and B), suggesting the involvement of a CWA protein in the skin adhesion phenotype. S. epidermidis expresses 11 proteins containing an LPXTG motif (31), of which Aap and SdrF have been previously shown to possess adhesive properties and have been implicated in facilitation of S. epidermidis colonization of skin (38,39). Similarly to our findings with the sortase mutant, corneocyte discs incubated with Aap-negative S. epidermidis 1457 Daap and 1457 Dica Daap mutants showed negligible adherence to corneocytes, in contrast to the Aap-positive 1457 and 1457 Dica mutant strains ( Fig. 1A and B), suggesting that the CWA protein Aap mediates bacterium-host interaction.
To test the function of SdrF in aiding Aap-dependent binding of 1457 to corneocytes, an sdrF allelic replacement mutant was constructed in the 1457 Dica mutant background. The corneocyte binding assay revealed that adhesion of the 1457 Dica DsdrF mutant was similar to that of the 1457 Dica mutant ( Fig. 1A and B), implying that in our assay, SdrF does not function to facilitate adherence to corneocytes. Furthermore, the finding that the 1457 Dica Daap DsdrF mutant phenocopied the 1457 Dica Daap mutant rather than the 1457 Dica DsdrF mutant ( Fig. 1A and B) strongly supports the importance of Aap in S. epidermidis skin colonization, irrespective of the presence of other CWA adhesins. A defective adherence phenotype of the aap mutation also diminishes the likelihood of two or more CWA proteins contributing collectively to the adhesive properties in this strain.
Identification of the Aap domain that mediates binding of S. epidermidis to corneocytes. Aap consists of structurally distinct domains (33). The N-terminal A domain of Aap, as well as its S. aureus orthologue SasG, have been previously implicated in promoting adhesion to mammalian cells and artificial surfaces (33,38,42). Therefore, we hypothesized that the Aap-mediated adherence of strain 1457 to corneocytes observed in Fig. 1 is facilitated by the A domain. To test this hypothesis, a bacterial blocking assay was performed using antiserum raised against the A or B domain of Aap (23). As shown in Fig. 2A, anti-AapA antiserum blocked adherence of 1457 to corneocytes, whereas anti-AapB antiserum did not, suggesting that Aap-mediated adherence of S. epidermidis 1457 to corneocytes from the skin primarily involves the A domain. A dose-dependent decrease of adherent bacteria was also observed with increasing concentration of anti-AapA antiserum (Fig. 2B), confirming the function of the Aap A domain in the facilitation of 1457 binding to corneocytes. In contrast, increasing the concentration of anti-AapB antiserum had no effect on adherence to corneocytes (Fig. 2C). Last, antibody raised against PIA (a noncontributing factor in bacterium-skin adherence) was used as a negative control and had no effect on the binding of bacteria ( Fig. 2A). Therefore, the 1457 Dica strain was subsequently used as the wild-type strain background for further studies.
The A domain is composed of a conserved 222-amino-acid globular subdomain that is predicted by bioinformatic analysis to be an L-type lectin-like domain. This lectin-like subdomain is positioned C-terminally to a strain-dependent, variable number of imperfect 16-amino-acid repeat regions (33). However, the particular subdomain that facilitates binding to corneocytes is yet to be identified. To further determine the contribution of the two subdomains, the ability of recombinantly expressed and purified rA repeat and rA lectin subdomain proteins to block adhesion was tested and compared to that of the complete rA rep1lec domain ( Fig. 3A to C). These data demonstrate that the rA lectin subdomain blocked the binding of 1457 to corneocytes in a concentration-dependent Adherence of S. epidermidis to Corneocytes ® manner, up to a maximum of 60% at 10 mM, similarly to the rA rep1lec domain (Fig. 3D). In contrast, the rA repeat subdomain was unable to block adherence. These results suggest that the L-type lectin-like subdomain is the key subdomain that contributes to the Aap A domain-mediated bacterium-host attachment.
Binding of 1457 Dica isogenic mutants expressing truncated variants of Aap to corneocytes. Isogenic mutants expressing truncated variants of Aap were generated to confirm the significance of the individual domains of Aap in adherence to corneocytes. Constructs expressing different combinations of the A and B domains were inserted via allelic replacement into the aap::tetM allele of the 1457 ica::dhfr aap::tetM strain, thus replacing aap::tetM with functional aap-expressing specific domains. All constructs contained the native aap promoter and included the N-terminal signal sequence for extracellular export of the translated protein. The proline/glycine-rich region (PGR) domain, which has been shown to form an extended stalk to resist compaction (43), was added, in addition to the C-terminal LPDTG sortase recognition motif for anchoring the exported protein to the cell wall. The generated knock-in mutants are shown in Fig. 4; pCM29 expressing GFP was transduced into each constructed mutant. To minimize proteolytic cleavage of Aap, the above mutants were grown in tryptic soy broth (TSB) containing the protease inhibitor a 2 -macroglobulin.
To confirm that the constructed knock-in mutants were able to express the appropriate Aap domains on the cell wall, Western blot analysis was performed using anti-AapA and anti-AapB antibodies. First, to determine if the polyclonal anti-AapA antibody could recognize the A repeat and A lectin subdomains individually, a Western blot analysis was performed using purified rA rep1lec domain and rA repeat and rA lectin subdomains. Although anti-AapA antibody could readily detect the rA rep1lec domain, the rA repeat subdomain was not detected, and the rA lectin subdomain was weakly detected ( Fig. 5A and B). Therefore, we were unable to confirm if the 1457 Dica aap::A-rep strain expressed the A repeat domain, due to the lack of antibody recognition when Western blot analysis was performed on its cell wall proteins using anti-AapA antibody (data not shown). Furthermore, we were unable to detect expression of the Aap A-lectin domain from the isolated cell wall proteins of the 1457 Dica aap::A-lec strain (data not shown). However, using anti-AapA antibody, we were able to detect the Aap A domain from the 1457 Dica aap::A-(rep1lec) strain (Fig. 5C). Last, cell wall proteins were isolated from the isogenic mutants expressing truncated Aap containing the B domain, and a Western blot analysis was performed using anti-AapB antibody (Fig. 5D). We found that anti-AapB antibody readily detected Aap from the 1457 Dica and 1457 Dica aap::A-rep1B strains. However, Aap was weakly detected in the 1457 Dica aap::B and 1457 Dica aap::A-lec1B strains. Interestingly, genetic insertion of the A repeat sequence of aap into the 1457 Dica aap::A-lec1B strain, generating the 1457 Dica aap:: A-(rep1lec)1B strain, restored expression of Aap to the levels observed in the 1457 Dica and 1457 Dica aap::A-rep1B strains. These data may suggest that the A domain, and the A repeat region in particular, may be required for stability of Aap. Importantly, it is well known that Aap does not migrate according to size in SDS-polyacrylamide gels, presumably due to the PGR repeats (43). Interestingly, the ;180-kDa band detected in the 1457 Dica aap::B strain was slightly larger in size than that detected in the 1457 Dica aap::A-lec1B strain, which was not predicted (Fig. 5D). Additionally, it is important to note that the presence of the Aap A repeat subdomain also resulted in aberrant migration of the rA rep1lec domain and rA repeat subdomain during SDS-PAGE, during which both migrated to a larger size than predicted (Fig. 5A). Accordingly, the major band detected near 300 kDa in the 1457 Dica and 1457 Dica aap::A-(rep1lec)1B strains was slightly larger in size than those in the 1457 Dica aap::A-rep1B strain ( Fig. 5D). However, the difference in size between the major bands of the 1457 Dica aap::A-rep1B and 1457 Dica aap::A-lec1B strains was much greater than predicted (Fig. 5D).
Our blocking data suggested that the A domain, and, in particular, the lectin subdomain, was required to mediate adherence to corneocytes. To further confirm the contribution of the A repeat region and the A lectin subdomain, the corneocyte binding assay was performed with the 1457 Dica strain expressing only certain domains of Aap (Fig. 6). We found that even though the expression of Aap in the 1457 Dica aap::A-rep1B strain was comparable to that in the 1457 Dica strain (Fig. 5D), no corneocyte adherence was detected in the 1457 Dica aap::A-rep1B strain (Fig. 6). In contrast, even though the expression of Aap in the 1457 Dica aap::A-lec1B strain was reduced in comparison to that in the 1457 Dica aap::A-rep1B strain (Fig. 5D), significant corneocyte adherence was observed for the 1457 Dica aap::A-lec1B strain (Fig. 6). These data confirm that the lectin subunit of the A domain, and not the A repeat region, is responsible for corneocyte adherence. The corneocyte binding ability of the 1457 Dica aap::A-(rep1lec)1B strain, expressing the B domain with the full-length A domain, was restored to the level of the 1457 and 1457 Dica strains (Fig. 6). These results suggest that the A-repeat domain may function to properly position the lectin subdomain for appropriate interaction with the corneocyte ligand in addition to the aforementioned possibility of conferring expression stability, as see in Fig. 5D. Furthermore, negligible corneocyte adherence was detected using the 1457 Dica aap::A-(rep1lec) strain, suggesting that the B domain may function to extend the A domain away from the cell wall to bind a ligand (Fig. 6). Additionally, no corneocyte adherence was observed for the 1457 Dica aap::A-lec,1457 Dica aap::A-rep, and 1457 Dica aap::B strains (Fig. 6). Glycan distribution on the surface of corneocytes. L-type lectin domains are protein domains with homology to the leguminous plant lectins which specifically bind carbohydrates. Therefore, we hypothesized that the predicted lectin domain of Aap has affinity for glycan moieties in the stratum corneum (SC) corneocytes that serve as binding partners. Eukaryotic cells have a rich and diverse surface glycan coat that constitutes the cell interface with the environment (44). However, the glycome (the sum of its diverse glycan structures) on every human tissue is distinct and is defined by expression levels of the enzymes responsible for glycan biosynthesis (45). The presence and distribution of various surface glycan moieties on corneocytes were thus detected utilizing a lectin-labeling technique. As noted in Fig. S2A and B in the supplemental material, commercially available rhodamine-labeled lectins that bind specific configurations of sugar molecules were able to detect different glycan moieties present on the surface of corneocytes from the upper layers of the stratum corneum in various distribution patterns. The results indicated that glycans persist on the surfaces of corneocytes taken from the forearm, and the labeling patterns suggested the presence of different sugar moieties in specific distribution patterns.
S. epidermidis binding of corneocytes after elimination of specific glycan structures by glycosidases. In human skin, N-linked glycosylation is the most common type of posttranslational modification on Asn-X-Ser/Thr motifs of epidermal proteins (46). A second major glycosylation found on the surface of keratinocytes is the covalent O-linkage to Ser/Thr residues of proteins (47). We reasoned that enzymatic treatment of corneocyte surfaces prior to addition of S. epidermidis with the endoglycosidases PNGase F (removes Nlinked oligosaccharides from glycoproteins, cleaving between the innermost GlcNAc and asparagine residue) and/or O-glycosidase (removes core 1 and core 3 O-linked disaccharides from glycoproteins) would result in reduced adherence of the 1457 Dica strain to corneocytes, Indeed, enzymatic treatment with both PNGase and O-Glycosidase reduced adherence of the 1457 Dica strain by approximately 50%, suggesting that both of these glycan linkages expressed in the host SC are important in S. epidermidis adherence ( Fig. 7A and B).
Based on these results, corneocytes were pretreated with specific exoglycosidases that catalyze hydrolysis of particular terminal sugar residues from glycoforms to assess the involvement of 5-N-acetylneuraminic acid (sialic acid), fucose, galactose, N-acetylglucosamine, Adherence of S. epidermidis to Corneocytes ® and mannose terminal sugars in binding S. epidermidis. Significant reduction in the 1457 Dica strain binding was observed in a2-3,6,8 neuraminidase-, a1-3,4 fucosidase-, and b1-4 galactosidase S-pretreated corneocytes ( Fig. 7C to E), confirming the importance of 5-N-acetylneuraminic acid, fucose, and galactose in contributing to ligand configuration, FIG 7 S. epidermidis corneocyte adherence following glycosidase treatment. Corneocytes were treated with specific glycosidases (A to G), and the number of adherent 1457 Dica/pCM29 bacteria after deglycosylation was normalized to the mean number of 1457 Dica/pCM29 bacteria adherent to corneocytes preincubated with buffer only (considered 100). Comparisons were performed by two-tailed unpaired Student's t test with 95% confidence intervals. ***, P , 0.0001.

Roy et al.
® either individually or together. In contrast, pretreatment of corneocytes with b-N-acetylglucosaminidase S and a1-2,3 mannosidase, to remove N-acetylglucosamine and mannose residues, respectively, did not result in any reduction in binding of the 1457 Dica strain to corneocytes ( Fig. 7F and G), suggesting that these residues either do not function in forming the ligand structure or their removal does not prevent the bacterial receptor from binding the ligand in the corneocytes. To confirm the activity of specific glycosidase treatment, corneocytes were stained with rhodamine-labeled lectins following enzymatic treatment and thus documenting reduced staining (see Fig. S3 in the supplemental material).
Genome analysis of predominant coagulase-negative staphylococci. The presence of protein sequences homologous to that of S. epidermidis 1457 Aap (GenBank accession no. ARG65518.1) in common staphylococcal skin commensals was investigated by bioinformatics analyses. The protein was considered present when a BLAST match with an alignment score of $200 was recorded, as shown in Table S3 in the supplemental material. Sequence similarity of the lectin domain of these homologous proteins was also compared to the 1457 Aap lectin domain to ensure elimination of Aap-like protein sequences having only G5-E domains and no lectin domain. As shown in Fig. 8A, S. hominis and S. haemolyticus strains have homologous Aap sequences with high percent identities to S. epidermidis 1457 Aap, as well as to the Aap lectin domain. Aap homologues, including the lectin domain, were also found in strains of Staphylococcus saprophyticus and S. capitis. On the other hand, Staphylococcus simulans has a protein in most strains annotated as "E domain-containing protein," with 27 to 47% homology to 1457 Aap. However, this predicted protein consists of only G5 domains. The lectin domain is not present in any of the Aap homologous protein sequences, nor as a part of any other proteins in S. simulans. Likewise, two strains of Staphylococcus warneri were also found to encode an Aap-like protein with only G5 domains and no lectin domain. Therefore, corneocyte binding assays were performed using crystal violet staining and light microscopy (Fig. 8B). These assays documented that S. hominis, S. haemolyticus, and S. saprophyticus bound to the corneocytes similarly to the 1457 Dica strain, whereas S. capitis, S. simulans, and S. warneri exhibited either decreased (S. capitis and S. simulans) or peripheral adherence (S. warneri) (Fig. 8B). Overall, these data suggest that the Aap orthologues may function to provide the basis for staphylococcal colonization of the skin, with the L-type lectin domain operating as receptors for specific ligands on the surface of corneocytes.
Corneocyte binding of S. aureus. Despite possessing the Aap orthologue SasG, S. aureus is not a normal component of the normal human skin microbiota. Indeed, many S. aureus strains, including JE2, have premature stop codons within sasG, which presumably result in a nonfunctional protein. Therefore, we chose to work with S. aureus MW2, which has been previously shown to express a functional SasG (48). As shown in Fig. 9, S. aureus MW2 and MW2 DsasG strains displayed negligible adherence to corneocytes. However, it has been previously demonstrated that the transcription of sasG, as well as that of seven other cell wall-associated proteins, is repressed by the global regulator MgrA, which in turn is regulated by the ArlRS two-component regulatory system (48). Thus, the contribution of MgrA was assessed in the corneocyte adherence assay. Indeed, the binding of MW2 increased significantly in the MW2 DmgrA mutant. This increase in adherence was confirmed to be SasG dependent, as the phenotype of the MW2 DmgrA DsasG mutant phenocopied that of MW2, suggesting that derepression of SasG in the absence of MgrA can significantly increase the binding of S. aureus to corneocytes. Thus, these data confirm that when SasG is expressed, it is able to mediate significant binding of S. aureus to corneocytes, similarly to S. epidermidis, reiterating the importance of Aap and its orthologues in the adherence of staphylococci to healthy skin.

DISCUSSION
There is emerging evidence that S. epidermidis and other commensal staphylococci ubiquitously present on cutaneous surfaces inhibit the colonization of harmful pathogenic microorganisms (9-14, 49). Although pathogen inhibition clearly involves production of inhibitory compounds by CoNS and other commensals, we surmise that adherence to corneocytes, and potentially intercellular aggregation, is a prerequisite for secretion of these compounds. Furthermore, S. epidermidis and other commensals may have evolved to adhere to specific ligands, thus successfully outcompeting S. aureus. Therefore, we propose that determining how S. epidermidis adheres to corneocytes is a missing component of understanding the pathogenesis of S. aureus skin and soft tissue infections. Indeed, understanding these interactions may reveal how S. aureus evolved to counteract inhibition by S. epidermidis and other CoNS.
This study demonstrated that Aap is a critical adhesin that mediates adherence of S. epidermidis 1457 to corneocytes. Our experiments confirmed that the Aap-mediated adherence of 1457 to corneocytes occurs primarily via interactions of the A domain with corneocytes. The conserved globular lectin-like subdomain was found to mediate adherence to surface ligands in the corneocytes and confer the A domain with optimum binding capacity. Inhibition of bacterial adherence using anti-Aap antiserum specific for either the A or B domains or the use of recombinant domain proteins provided results documenting the importance of the complete A domain and the lectin subdomain in the binding process. However, failure to completely prevent bacterial adherence even after blocking with a high concentration of the rA rep1lec domain (10 mM) (Fig. 3C), in contrast to the nearly complete blocking of adherence via anti-AapA domain antiserum ( Fig. 2A), suggested loss of structural integrity or reduced binding affinity in the recombinant soluble form of the A rep1lec domain. Interestingly, 50% blocking required 2.48 mM rA rep1lec domain compared to 7.54 mM rA lectin domain. Therefore, a lower concentration of the rA rep1lec domain than the rA lectin domain is required to bind half the available ligands present on corneocytes, indicating that the rA rep1lec domain has higher blocking efficiency than the rA lectin domain. Therefore, we speculate that the complete rA rep1lec domain might possess a higher affinity for the ligand on the corneocyte than rA lectin domain alone and hypothesize that the 16-amino-acid repeat region present N-terminally to the lectin domain provides conformational precision for the lectin domain to bind the ligand in the stratum corneum of the skin. Alternatively, since the Aap lectin domain is predicted to possess homodimer and homotetramer interaction sites for oligomeric interactions (predicted by the NCBI Conserved Domains Database), multiple copies of Aap lectin domains expressed together in S. epidermidis might show multivalent interactions with corneocyte ligands with higher binding avidity than that of the monovalent recombinant forms. Corneocyte binding assays performed using isogenic mutants expressing truncated variants of Aap confirmed that while the lectin subdomain binds to corneocytes, the N-terminal A-repeat region also functions to provide the lectin domain with the maximum binding ability, and the B domain functions, presumably, to extend the A domain, facilitating interaction with the corneocyte ligand.
Many studies have suggested that infection or colonization by bacteria is generally initiated by the specific recognition of host epithelial surfaces by adhesins and lectins. Glycans are present on all cell types and have specific biological functions in cell-cell recognition and cell-matrix interactions within the host (50). Pathogens have evolved to exploit these cell surface glycans by recognizing specific sequences of glycans as targets to bind host cells. Adhesins expressed on the surface of these microorganisms are often equipped with lectin domains that mediate binding to such glycans presented on proteins or lipids on the host surface (50). We hypothesize that the lectinlike domain of Aap binds glycans present on the surface of the corneocytes. The use of glycosidases suggested that the terminal glycans expressed in the host stratum Adherence of S. epidermidis to Corneocytes ® corneum serve as binding partners for the Aap-mediated colonization of S. epidermidis on the corneocyte surface. Indeed, antibodies against the highly expressed corneocyte proteins loricrin and cytokeratin-10 had no effect on S. epidermidis adherence (see Fig. S4 in the supplemental material). Given the terminal location of the negatively charged sialic acid in most human glycoforms and their widespread distribution, sialic acids are often the targets for binding by a large number of pathogenic organisms and their toxins (51). Hence, we hypothesize that Neu5Ac might present an important structure for S. epidermidis ligand in corneocytes. Galactose, GlcNAc, and fucose also form terminal "decorating" sugars, and thus we speculate the involvement of these sugars in S. epidermidis adherence. It should also be noted that the complete loss of S. epidermidis adherence to corneocytes was not observed following glycosidase treatment (Fig. 7). These results may suggest an incomplete removal of glycan moieties by the endoglycosidases, probably due to the difficulty in accessing the interior target in the presence of lipid-rich layers on the corneocyte sheets. Alternatively, it might suggest that glycolipids, in addition to glycoproteins, have a function in S. epidermidis adherence. Taken together, the results generated by this study demonstrate that the adherence of S. epidermidis to corneocytes is dependent on terminal sugar residues, with sialic acid, fucose, and galactose being the most promising binding ligands (Fig. 7).
Aap is a multifunctional protein that contributes to both primary attachment and subsequent biofilm accumulation (32,33,35). These well-defined functions are mediated by its structurally distinct A and B domains (29,36,37). Studies show that expression of fulllength Aap mediates attachment of S. epidermidis to uncoated abiotic surfaces, presumably through hydrophobic interactions via the A domain (32,33). Additionally, Schaeffer et al. found that S. epidermidis adherence to a rat jugular catheter was dependent upon Aap (33). It is unclear, however, if the same glycan expressed on corneocytes is also coated on catheters or whether Aap binds multiple glycan ligands. Additionally, Aap is subjected to proteolytic processing by the SepA metalloprotease (or by host-derived proteases), which results in distinct isoforms of the expressed protein (33)(34)(35). Based on Western blot analysis (34), we predict that about half of the cell wall-associated Aap molecules are processed by SepA. Thereafter, processed Aap can participate in Zn 21 -dependent intercellular adhesion via the B domains to facilitate cellular aggregation and/or biofilm accumulation in S. epidermidis. Moreover, SepA-mediated proteolysis of Aap within the A domain (at Leu 335) and at the junction between the A and B domains (at Leu 601) is under the regulation of the global transcriptional regulator, SarA, thus allowing S. epidermidis to fine-tune its biofilm-inducing properties in response to environmental cues (34). However, it is unclear if Aap processing via SepA has a specific function with regard to S. epidermidis adherence to corneocytes. We propose that, following A lectin-mediated adherence to the glycan, processed Aap facilitates additional cellular aggregation allowing for increased colonization of the skin surface (Fig. 10). Indeed, Gonzalez and colleagues recently found via scanning electron microscopy that bacteria colonizing skin were observed as aggregates of cells, suggesting that subsequent B domain-mediated aggregation may be important for skin surface colonization (52).
Commensal staphylococcal species possessing homologous Aap adhesins with a functional lectin-like domain may also function to bind corneocyte glycan ligands, similarly to S. epidermidis. We were able to confirm this in S. aureus MW2 strain expressing SasG, an orthologue of Aap. Though L-type lectins exhibit both sequence and structural similarity to one another, their carbohydrate binding specificities differ widely depending on the residues of the adjoining sequence of the carbohydrate binding site. Thus, it is possible that each CoNS species exhibits specificity for distinct sugar residues or moieties, which might explain the observed binding pattern differences between different strains. Alternatively, SasG and Aap (or other staphylococcal Aap orthologues) may bind similar glycan moieties, and competition for these binding sites may exist. This may suggest that S. aureus is selected to repress SasG via MgrA to avoid competition with S. epidermidis and/or other coagulase-negative staphylococcal species. Indeed, many strain backgrounds, including USA300, the most prominent S. aureus lineage causing infections in the United States, encode a truncated and presumably nonfunctional SasG protein, suggesting that certain populations of S. aureus may use alternative strategies to facilitate corneocyte adherence (48,53).
Conventionally, glycan recognition by proteins is of low affinity compared to proteinprotein interactions, which typically have high binding affinities. However, since adhesins and ligands often cluster in the plane of the membrane, the resulting combinatorial avidity can be high (50,54). It is also possible that S. epidermidis has adhesins that promote glycan-independent mechanisms of interaction with the host corneocyte proteins, in a similar manner as SdrF binding to cytokeratins (39). Nevertheless, glycan-dependent binding mechanisms often function to play a role in the first step of adhesion, prior to invasion during infections. It can also be advantageous in harsh environmental conditions where opportunities for contact and adherence may be transient and difficult (53,54).
In conclusion, S. aureus is the most significant cause worldwide of skin and skin structure infections. However, S. aureus is infrequently isolated from the skin from healthy subjects. This suggests that S. epidermidis and other bacteria colonizing the skin are highly effective at inhibiting colonization of this pathogen. We found that S. epidermidis binds to healthy corneocytes using Aap and that other coagulase negative staphylococci also contain an Aap orthologue, suggesting a potential common mechanism for corneocyte adherence. These data may suggest that S. epidermidis has the ability to outcompete S. aureus, potentially via adherence to carbohydrate molecules decorating corneocytes. Further studies will focus on competition between these two species, determine whether SasG and Aap bind similar ligands and, potentially, determine if S. aureus uses other adhesins to bind healthy or diseased skin.

MATERIALS AND METHODS
Bacterial strains and medium. Bacterial strains and plasmids used in these studies are described in Table S1 in the supplemental material. S. epidermidis strains were grown in tryptic soy broth (TSB) or agar (Difco). Escherichia coli DH5a, used for cloning purposes, and Escherichia coli BL21(DE3), used for Adherence of S. epidermidis to Corneocytes ® overexpressing recombinant proteins, was grown in lysogeny broth or agar (LB; Difco). Antibiotics and concentrations for cloning and selection purposes were as follows: ampicillin (50 mg/ml), kanamycin (100 mg/ml), tetracycline (5 mg/ml), erythromycin (10 mg/ml), chloramphenicol (10 mg/ml), and trimethoprim (10 mg/ml). Cultures were grown aerobically (1:10 medium-to-flask ratio, 250 rpm) at 37°C or at 30°C (for temperature-sensitive strains). Construction of recombinant His 6 -tagged fusion proteins. First, a 1,602-bp fragment containing the A domain sequence of aap from S. epidermidis 1457 (55) was amplified using primers r-aapAdom F and r-aapAdom R. Similarly, 915 bp of the A-repeat region of aap was amplified using the primers r-aapAdom F and r-aapArep R, and 681 bp of the L-type lectin domain of aap was amplified using the primers r-aapLec F and r-aapAdom R. PCR was carried out using chromosomal DNA from S. epidermidis strain 1457 as the template and Q5 high-fidelity DNA polymerase (New England Biolabs [NEB], Ipswich, MA) according to the manufacturer's instructions. The primers used are listed in Table S2 in the supplemental material. The resulting amplicons were cloned into pET28a using the NdeI and XhoI restriction sites, creating plasmids pNF319, pNF326, and pNF318 for N-terminal His 6 -tagged rA rep1lec domain, rA repeat subdomain, and rA lectin subdomain, respectively, in chemically competent E. coli DH5a cells. The primers T7Promoter F and T7Terminator R were used for confirming the inserts by PCR. Plasmid DNA was isolated from E. coli DH5a using the Wizard Plus SV Minipreps DNA purification system (Promega, Madison, WI) according to the manufacturer's recommendations. The plasmids were transformed into chemically competent E. coli BL21(DE3) cells for overexpression of the recombinant proteins.
Expression and purification of recombinant proteins. For recombinant expression of the A rep1lec domain and subdomains A repeat and A lectin , BL21(DE3)/pNF319, BL21(DE3)/pNF326, and BL21(DE3)/ pNF318 were induced using autoinduction medium (adapted from Studier [56]) for 18 h at 37°C under aerobic conditions. Induced cultures were pelleted and resuspended for lysis in 6 ml of lysis buffer containing BugBuster protein extraction reagent (EMD Millipore, Billerica, MA), EDTA-free cOmplete protease inhibitor cocktail (Roche, Mannheim, Germany), and benzonase nuclease (EMD Millipore), for 20 min at room temperature under gentle shaking conditions. Samples were then centrifuged at 4°C and supernatants incubated with 1-ml resins charged with divalent cobalt provided in HisPur cobalt spin columns (Pierce Biotechnology, Rockford, IL) overnight on an end-over-end mixer at 4°C for affinity purification of the recombinant proteins. Columns were then washed and eluted following the manufacturer's instructions. Affinity-purified proteins were concentrated at 20°C using a Vivaspin 10,000-molecular weight cutoff (MWCO) column (Sartorius AG; Göttingen, Germany) and stored in phosphate-buffered saline (PBS) with 30% glycerol at 220°C. SDS-PAGE and subsequent Coomassie staining tested the integrity of the respective protein preparations.
Construction of srtA and sdrF allelic replacement knockout mutants in strain 1457. To construct the sdrF allelic replacement plasmid, PCR was performed using chromosomal DNA from S. epidermidis 1457 as the template and Phusion high-fidelity DNA polymerase (NEB) according to the manufacturer's instructions. The primers used are listed in Table S2. A 1,001-bp fragment, containing the 39 region of sdrF and downstream sequence, was amplified using the primers sdrFDN F and sdrFDN R. The fragment was inserted into the XmaI and SalI sites of the shuttle vector pJB38 (57), creating plasmid pNF360. A 1,001-bp fragment, containing the 59 region of sdrF and upstream sequence, was amplified using the primers sdrFUP F and sdrFUP R. The fragment was inserted into pNF360 between the SacI and XmaI sites, creating pNF361. A 1,260-bp fragment containing the ermC cassette was amplified from pROJ6448 (58) with XmaI ends using the primers ermC-XmaI F and ermC-XmaI R. The fragment was inserted in the XmaI site of pNF361, creating the sdrF allelic replacement plasmid pNF362. The inserts were confirmed by restriction analysis and PCR with the primers pJB38seq F and pJB38seq R. Subsequently, pNF362 was electroporated into electrocompetent PS187 cells (59) (0.1-cm cuvette, 100 X, 25 mF, and 2.3 kV) (60), followed by transduction into the S. epidermidis 1457 icaADBC::dhfr aap::tetM strain (33) with bacteriophage U187 (59), using a previously described protocol (61). Allelic exchange was performed thereafter to generate single and double recombinants, as previously described (62). Single recombinants were selected for resistance to trimethoprim, tetracycline, erythromycin, and chloramphenicol, while the allelic replacement mutant generated from double recombination was selected based on loss of resistance to chloramphenicol alone (plasmid marker). Following allelic replacement, the sdrF mutant was backcrossed into the 1457 icaADBC::dhfr (63) and 1457 icaADBC::dhfr aap::tetM strains using the transducing phage U71 (64), to generate the 1457 icaADBC::dhfr sdrF::ermC and 1457 icaADBC::dhfr aap::tetM sdrF::ermC mutants, respectively. The allelic replacement mutants were confirmed by PCR with the primers sdrFseq F and sdrFseq R and pulsed-field gel electrophoresis (PFGE) as previously described (65).
To construct the srtA allelic replacement plasmid, PCR was performed using chromosomal DNA from S. epidermidis 1457 as the template and Taq DNA polymerase. The primers used are listed in Table S2. A 2,074-bp fragment containing the srtA gene and its upstream and downstream sequence was amplified with primers srtAUP F and srtADN R and cloned into pCR2.1 by TA cloning, creating plasmid pNF79. The insert flanked by EcoRI sites was excised from pNF79 and cloned into the EcoRI site of pUC19, creating plasmid pNF82. A 2,383-bp tetM cassette was amplified from pNF105 with ClaI ends using the primers tetM-ClaI F and tetM-ClaI R and inserted into the ClaI site in the middle of the srtA gene of pNF82, creating plasmid pNF107. Finally, pROJ6448, containing a temperature-sensitive origin of replication and conferring erythromycin resistance, was linearized with PstI and ligated into the PstI site of pNF107, generating the srtA allelic replacement plasmid pNF112. All of the above-described cloning were done in E. coli DH5a cells. Subsequently, pNF112 was electroporated into electrocompetent PS187 cells (0.1-cm cuvette, 100 X, 25 mF, and 2.3 kV) and then transduced into S. epidermidis 1457 using bacteriophage U187. Allelic exchange was performed thereafter to generate single and double recombinants. Single recombinants were selected for resistance to tetracycline and erythromycin. The allelic replacement mutant generated from double recombination was able to grow on tryptic soy agar with tetracycline (TSA1Tet) but not on TSA1Erm. Following allelic replacement, the srtA mutant was backcrossed into the 1457 and 1457 icaADBC::dhfr strains, using the transducing phage U71, to generate the 1457 srtA::tetM and 1457 icaADBC::dhfr srtA::tetM mutants, respectively. The allelic replacement mutants were confirmed by PCR and PFGE.
Construction of aap allelic replacement knock-in mutants expressing truncated variants of Aap in the 1457 Dica strain. Isogenic mutants expressing truncated variants of Aap from i to vi were generated by insertion of plasmid constructs containing specific combinations of aap domains (markerless) into the 1457 icaADBC::dhfr aap::tetM construct by allelic replacement of Daap::tetM. All of the allelic exchange constructs contained the upstream and downstream region of aap for upstream and downstream homologous recombination, respectively. Allelic exchange constructs for the following mutants were made as described below. PCR amplifications were carried out using chromosomal DNA from S. epidermidis 1457 as the template and one of the following polymerases: Q5 high-fidelity DNA polymerase (NEB), Pfurther long-range DNA polymerase (Monserate Biotechnology Group, San Diego, CA), or Phusion high-fidelity DNA polymerase (NEB). The primers used are listed in Table S2. (iv) The 1457 Dica aap::A-lec construct. The 1,369-bp insert [aap(PGR1LPDTG)1DN] amplified above with primers aapPGR F and aapDN R, was cloned into the XmaI and SalI site of the previously generated pNF334, creating pNF345. Next, 687 bp of the lectin domain sequence of aap was amplified with aapAlec F and aapAdom R and cloned into the XmaI site of pNF345. The directionality of the aapA-lectin insert was checked with aapAlecOut and aapUP F. The resulting generated construct was named pNF350 (aapA-lec allelic exchange vector). pNF350 was used to create the 1457 D ica aap::A-lec mutant for expressing a truncated variant of Aap with only the A-lectin subdomain attached to the PGR stalk and LPDTG Gram-positive anchor, devoid of the A-repeat region and the B domain.
(v) The 1457 Dica aap::A-rep1B construct. The 3,940-bp fragment containing [aap(B1PGR1 LPDTG)1DN], amplified previously with aapBdom F and aapDN R, was cloned into the XmaI and SalI site of the previously generated pNF346, creating pNF349 (aapA-rep1B allelic exchange vector). pNF349 was used to create an A-lectin subdomain-lacking 1457 Dica aap::A-rep1B mutant for expressing a truncated variant of Aap with the A-repeat region attached to the B domain, PGR stalk, and LPDTG Gram-positive anchor.
(vi) The 1457 Dica aap::A-lec1B construct. A 4,648-bp fragment comprising the A-lectin subdomain, B domain, PGR region, and LPDTG Gram-positive anchor of aap, along with aap DN, was amplified with primers aapAlec F and aapDN R. This fragment was cloned into the XmaI and SalI site of the previously made pNF334, generating pNF342 (aapA-lec1B allelic exchange vector). pNF342 was used to create an A-repeat region-lacking 1457 Dica aap::A-lec1B mutant for expressing a truncated variant of Aap with the A-lectin subdomain attached to the B domain, PGR stalk, and LPDTG Gram-positive anchor.
All of the above allelic exchange plasmids were constructed in E. coli DH5a cells. The inserts were confirmed by restriction analysis and PCR with the primers pJB38seq F and pJB38seq R. These plasmids were electroporated into electrocompetent PS187 cells (0.1-cm cuvette, 100 X, 25 mF, and 2.3 kV) and then transduced into the S. epidermidis 1457 icaADBC::dhfr aap::tetM mutant using bacteriophage U187. Allelic exchange was performed thereafter to generate single and double recombinants. Single recombinants were selected for resistance to trimethoprim and tetracycline and chloramphenicol, while the knock-in mutants generated from double recombination were selected based on loss of resistance to tetracycline and chloramphenicol. The mutants thus generated (i to vi) were confirmed by PCR with the primers aapUPseq F and aapDNseq R, PFGE, and immunoblotting with antibodies to Aap.
Adherence of S. epidermidis to Corneocytes ® (vii) The 1457 Dica aap::A-(rep1lec)1B construct. pNF344 containing the insert [UP1aap(SS1A-(rep1lec))] was used as the allelic exchange vector for introducing the A-repeat region back into the 1457 Dica aap::A-lec1B strain for regenerating the wild-type aap genetically. For this purpose, pNF344 was electroporated into electrocompetent PS187 and transduced into the 1457 Dica aap::A-lec1B mutant using bacteriophage U187. Allelic exchange was performed thereafter to generate single recombinants that were resistant to trimethoprim and chloramphenicol, while double recombinants were selected based on loss of chloramphenicol resistance on the plasmid. The knock-in mutants were selected from double recombinants based on PCR confirmation with the primers aapAlecOut and aapUP F. Mutant construction was also verified by PFGE and immunoblotting with Aap antibodies.
Preparation of cell wall proteins. Fresh TSB (12.5 ml in a 125-ml flask) was inoculated at an OD 600 of 0.02, and the cultures were grown to the exponential phase at 37°C with shaking at 250 rpm until an OD 600 of ;0.7 was reached. Cells were collected by centrifugation at 5,000 rpm for 10 min at 4°C, and the pellet was washed three times with 1 ml 1Â PBS (Dulbecco's PBS, no calcium, no magnesium). To isolate cell walls, cell pellets were resuspended in 200 ml lysis buffer (50 mM  , and lysozyme [100 mg/ml]) and incubated at 37°C with shaking for 30 min. Cell walls were separated from protoplasts by centrifugation at 6,000 Â g for 20 min. The supernatant (cell wall fraction) was removed, and samples were stored at 220°C.
Western blot analysis. The concentration of protein samples was determined using a NanoDrop 2000 spectrophotometer (Thermo Scientific, Waltham, MA) or with bicinchoninic acid (BCA) assay (Pierce Biotechnology). Samples were adjusted for equal protein amount (35 mg total protein for cell wall preparations; 50 ng for purified proteins), mixed with 4Â SDS sample loading buffer, denatured at 95°C for 10 min, and then loaded on a 7 to 10% SDS-PAGE gel. Current was applied at 60 V for 30 min and at 150 V until the dye front reached the bottom of the gel. Proteins were transferred onto an activated 0.45-mm polyvinylidene difluoride (PVDF) membrane in cold 1Â transfer buffer at 100 V for 2 h.
Collection of corneocytes. Sheets of corneocytes were collected by applying adhesive discs (D-Squame skin indicator D232; CuDerm Corp., Dallas, TX) with a 12.7-mm-diameter sensing area, on the inner forearm of healthy human volunteers with no history or physical signs of dermatological disease. Corneocytes were collected from the selected area after removal of the topmost epidermal layer using a regular adhesive tape and subsequent topical application of isopropyl alcohol with prep pads for sterilization. The outermost layers of the stratum corneum of the forearm were chosen for our experiments, as mature and rigid cornified envelopes were detected in the upper layers, in contrast to immature and fragile cornified envelopes from the deeper layers of stratum corneum (66).
For visualization of cytokeratin-10 and loricrin labeling on the corneocytes, corneocyte slides were incubated with 10 mg/ml rabbit polyclonal anti-cytokeratin-10 IgG (Abcam, Cambridge, MA) or rabbit polyclonal anti-loricrin IgG (Abcam) primary antibodies for 1 h. This was followed by incubation with Alexa Fluor 555-conjugated goat anti-rabbit IgG (Thermo Scientific) secondary antibody and subsequent confocal microscopy imaging.
Optimization of the corneocyte binding assay. The corneocyte binding assay was optimized based on previously published methods (67)(68)(69)(70), with modifications. For optimization, the effects of time and bacterial concentration were tested on corneocytes collected on 22-mm-diameter adhesive discs (D100 D-Squame standard sampling discs; CuDerm Corp.). Briefly, S. epidermidis 1457 cells grown to an exponential phase (OD 600 , ;0.7) were harvested, suspended in sterile PBS, and adjusted to an OD 600 of 0.15, 1.0, or 1.5. An OD 600 of 0.15 was found to correlate with ;10 7 CFU/ml, an OD 600 of 1.0 with ;10 8 CFU/ ml, and an OD 600 of 1.5 with ;10 9 CFU/ml. Each corneocyte sample disc was inoculated with one of three concentrations of bacterial suspension and incubated for one of three different time periods (45, 90, or 180 min). For incubation, the discs were placed with the corneocyte surface up in a small petri dish, and 500 ml of each bacterial suspension was placed over the corneocyte layer. Following incubation, the samples were rinsed with PBS 3 times by pipetting and stained with crystal violet for 10 s. Excess stain was removed by rinsing with running water for 10 s, and the discs were then attached to glass microscope slides using transparent glue. The slides were allowed to air dry prior to microscopic examination using light microscopy. An OD 600 of 0.15 was the chosen bacterial suspension concentration for incubation, due to visibly less clumping observed at this OD. Similarly, 45 min was chosen for the incubation, as 90-and 180-min incubation times resulted in more clumping.
Corneocyte binding assay. Overnight cultures of strains transduced with the green fluorescent protein (GFP)-expressing plasmid pCM29 were grown in 5 ml TSB containing 10 mg/ml chloramphenicol with shaking at 250 rpm at 37°C for 16 h. Fresh TSB (12.5 ml in a 125-ml flask) was inoculated at an OD 600 of 0.1, and the cultures were grown to the exponential phase at 37°C with shaking at 250 rpm until they reached an OD 600 of ;0.7. Cells were collected by centrifugation at 5,000 rpm for 5 min at 4°C, and the pellet was washed three times with 1 ml 1Â PBS (Dulbecco's PBS, no calcium, no magnesium). The final bacterial suspension was adjusted to an OD 600 of 0.15, corresponding to ;10 7 CFU/ml in PBS. A 300-ml aliquot of the OD-adjusted suspension was pipetted onto D232 D-Squame discs with adhered corneocytes to form a meniscus and incubated at 37°C for 45 min in a moist chamber. After incubation, the discs were washed with 1Â PBS three times by vigorous pipetting and air dried. A cover slip was placed for imaging using confocal laser scanning microscopy.
Subsequently, the corneocyte binding assay was performed with addition of bacteria to the suspension at a final OD 600 of 0.15.
Inhibition of bacterial binding using antibodies against cytokeratin-10 and loricrin was performed, in which corneocytes were preincubated separately with 10 mg/ml rabbit polyclonal anti-cytokeratin-10 IgG (Abcam) and rabbit polyclonal anti-loricrin IgG (Abcam) antibodies for 1 h at 37°C in blocking buffer (1% bovine serum albumin [BSA]) prior to incubation with bacteria.
Image acquisition and bacterial counting after corneocyte binding assay. Images were acquired by confocal laser scanning microscopy using an LSM 710 confocal microscope (Zeiss; Jena, Germany). GFP-expressing bacterial cells were observed with an oil immersion objective (Plan-Apochromat 63Â/ 1.40 oil differential interference contrast) using a maximum excitation wavelength (l ex ) of 488 nm and a maximum emission wavelength (l em ) of 542 nm. Additionally, the nonlabeled corneocyte sheets from the same view field were simultaneously imaged using phase contrast optics. For quantification of adherent bacteria on each disc, 10 images of 89.8 mm Â 89.8 mm microscopic fields containing unlabeled confluent corneocytes were captured at 63 Â 1.5 magnification, superimposing both the fluorescence and phase contrast image. These fields were randomly selected, and any field in which corneocytes were either not confluent or for which the entire area could not be brought into sharp focus was rejected. In the case of rhodamine-labeled corneocytes, an l ex of 553 nm and an l em 627 nm were used for visualization.
Quantitative analysis of the number of adherent bacteria was performed using ImageJ v. 10.0 (NIH, Bethesda, MD). The images captured by the fluorescence channel, which contained the GFP signals (GFexpressing bacteria), were converted from RGB to grayscale in ImageJ. For automated identification of bacteria in each image, thresholding of the image was accomplished by filtering within the hue-saturation-brightness (HSB) color space using the default mode. Subsequently, image noise was removed using "despeckle," brightness and contrast adjusted to make the bacterial cells more distinguishable from the background. The image was converted to a mask based on the set threshold to produce binary images with inverted lookup tables (LUTs), and "watershed" was used to create segmentation from one another in the case of clustered bacterial cells. The identified bacterial cells were finally counted using "analyze particles." The pixel counts for each circular area, as well as the average clump size within the area, were automatically entered into spreadsheet format by the software. For each strain tested, the number of adherent bacteria from all of the fields was plotted using GraphPad Prism 8. The mean number of adherent bacteria for each mutant was calculated along with standard error of the mean (SEM). Comparisons between individual mutants were performed by two-tailed unpaired Student's t test with 95% confidence intervals. A P value of ,0.05 was considered to be significant in each case.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.