Syndecan-1 Promotes Staphylococcus aureus Corneal Infection by Counteracting Neutrophil-mediated Host Defense*♦

Many microbial pathogens subvert cell surface heparan sulfate proteoglycans (HSPGs) to infect host cells in vitro. The significance of HSPG-pathogen interactions in vivo, however, remains to be determined. In this study, we examined the role of syndecan-1, a major cell surface HSPG of epithelial cells, in Staphylococcus aureus corneal infection. We found that syndecan-1 null (Sdc1−/−) mice significantly resist S. aureus corneal infection compared with wild type (WT) mice that express abundant syndecan-1 in their corneal epithelium. However, syndecan-1 did not bind to S. aureus, and syndecan-1 was not required for the colonization of cultured corneal epithelial cells by S. aureus, suggesting that syndecan-1 does not mediate S. aureus attachment to corneal tissues in vivo. Instead, S. aureus induced the shedding of syndecan-1 ectodomains from the surface of corneal epithelial cells. Topical administration of purified syndecan-1 ectodomains or heparan sulfate (HS) significantly increased, whereas inhibition of syndecan-1 shedding significantly decreased the bacterial burden in corneal tissues. Furthermore, depletion of neutrophils in the resistant Sdc1−/− mice increased the corneal bacterial burden to that of the susceptible WT mice, suggesting that syndecan-1 moderates neutrophils to promote infection. We found that syndecan-1 does not affect the infiltration of neutrophils into the infected cornea but that purified syndecan-1 ectodomain and HS significantly inhibit neutrophil-mediated killing of S. aureus. These data suggest a previously unknown bacterial subversion mechanism where S. aureus exploits the capacity of syndecan-1 ectodomains to inhibit neutrophil-mediated bacterial killing mechanisms in an HS-dependent manner to promote its pathogenesis in the cornea.

Microbial pathogens express a multitude of factors that interact with host components. Pathogens use these hostpathogen interactions to their advantage to survive in the host environment. Studies during the last several decades have proposed that many viral, bacterial, and parasitic pathogens bind to cell surface HSPGs 2 to facilitate their initial attachment and subsequent invasion of host cells (1)(2)(3). Evidence that the HSPG interaction is biologically important is provided by the finding that HS-binding pathogens show markedly attenuated attachment or invasion of host cells whose HS expression has been reduced by enzymatic treatment or mutagenesis. Furthermore, exogenous HS or heparin (pharmaceutical functional mimic of HS) inhibits pathogen attachment and entry. Furthermore, where examined, mutant strains lacking the HSPG adhesin are viable and show normal growth rates, suggesting that the ability to bind to HSPGs is strictly a virulence activity. However, the significance of HSPG-pathogen interactions in infectious diseases has yet to be clearly established in vivo.
Syndecans comprise a major family of cell surface HSPGs (1,4). Syndecans are type I transmembrane HSPGs composed of four members in mammals. At the cell surface, syndecans function primarily as a co-receptor for various HS-binding ligands and regulate cellular processes, such as adhesion, proliferation, migration, and differentiation. Although all syndecans harbor the ligand-binding HS chains in their extracellular domains, dramatic pathological phenotypes emerge when the single syndecan null mice are challenged with infectious or inflammatory stimuli (5)(6)(7)(8)(9)(10)(11)(12), indicating that certain postdevelopmental functions of each syndecan are specific and cannot be compensated by another syndecan or other HSPGs. How this is accomplished is incompletely understood, but syndecans likely perform specific functions in vivo because they are expressed on different cell types and locations at different levels and times (1,13). For example, in adult tissues, syndecan-1 is abundantly expressed by both simple and stratified epithelial cells and expressed to a lesser degree by other cell types (e.g. fibroblasts) (1,13,14).
The role of syndecans in microbial infections is currently an active area of research. For instance, Neisseria gonorrhoeae binds to the HS moiety of syndecan-1 and -4 through the Opa protein, and this interaction mediates both bacterial attachment and invasion (15). The intact syndecan cytoplasmic domain is essential in gonococcal invasion as N. gonorrhoeae attaches to but does not invade epithelial cells expressing syndecan mutant constructs lacking the cytoplasmic domain or those lacking specific signaling motifs in the cytoplasmic domain (15). These data suggest that binding of N. gonorrhoeae to syndecan-1 and -4 induces signaling through the syndecan cytoplasmic domain, leading to internalization of the bacteria. Alternatively, syndecan-2 and -3 expressed on the surface of dendritic cells have been shown to bind to HIV and facilitate viral transmission to CD4-positive T cells (16,17). Here, syndecans are thought to prolong the infectivity of HIV, increase infectivity of dendritic cells in cis, and promote transmission to T cells.
Syndecan-1 has also been proposed to modulate bacterial infections as a shed HSPG ectodomain. Syndecan-1 shedding is a highly regulated process that is stimulated in vitro by several inflammatory factors, and in vivo under certain pathological conditions (1,3,18,19). Bacterial pathogens, such as Staphylococcus aureus (20), Pseudomonas aeruginosa (21), Streptococcus pneumoniae (22), and Bacillus anthracis (23), secrete virulence factors that stimulate the host cell's metalloproteinase-mediated shedding mechanism at the cell surface. However, several strains of other Gram-positive and Gramnegative bacteria, including Streptococcus agalactiae (group B Streptococcus), Staphylococcus xylosus, Salmonella enteritidis, and Salmonella typhimurium, do not enhance shedding (21), suggesting that certain opportunistic bacterial pathogens selectively induce syndecan-1 shedding. The physiological function of syndecan-1 shedding in bacterial infections is not fully understood, but syndecan-1 shedding is induced by established virulence factors, suggesting that this is a pathogenic mechanism. Moreover, syndecan-1 shedding is induced in mouse models of P. aeruginosa lung (8) and burned skin infection (6), and administration of purified syndecan-1 ectodomains enhances P. aeruginosa virulence in the lung (8). Syndecan-1 ectodomains are thought to promote P. aeruginosa pathogenesis by interfering with innate host defense mechanisms, but the defense mechanisms inhibited by syndecan-1 ectodomains are not known. In addition, it is not known whether activation of syndecan-1 shedding is a general virulence mechanism used by various bacterial pathogens.
All these properties of syndecan-1 suggest that it may have a prominent role in infection, likely with a critical function as a cell surface attachment site or a shed anti-host defense factor. In either case, syndecan-1 could be a suitable drug target for treatment of infectious diseases. To address these issues, we examined the role of syndecan-1 in the pathogenesis of S. aureus corneal infection. Bacterial keratitis is a serious public health concern with significant ocular morbidity that can lead to reduced acuity and irreversible scarring (24 -26). It is one of the major causes of blindness worldwide. S. aureus is a leading cause of bacterial keratitis, accounting for 10 -25% of confirmed cases (27)(28)(29). S. aureus binds to HS (30), and syndecan-1 has been shown to enhance the attachment of S. aureus to several types of host cells (31). Furthermore, S. aureus induces syndecan-1 shedding in vitro through ␣and ␤-toxins (20), which are established exotoxin virulence factors in animal models of S. aureus keratitis (32,33). Our data surprisingly revealed that syndecan-1 does not bind to S. aureus and does not mediate S. aureus attachment to corneal epithelial cells. Instead, our results showed that S. aureus infection induces syndecan-1 shedding from the surface of corneal epithelial cells, and syndecan-1 ectodomains promote S. aureus corneal infection by interfering with the capacity of neutrophils to kill S. aureus.
Bacteria-S. aureus strains 8325-4 (20), 12598 (34), Woods (34), and P1 (35) were from our culture collection. S. aureus strains were grown to late log growth phase in tryptic soy broth (TSB), and the bacterial concentration was approximated by measuring absorbance at 600 nm. After washing, the concentration was adjusted to ϳ3-5 ϫ 10 8 cfu/5 l of TSB with or without the indicated test agents. The exact bacterial concentration in the inoculum was determined by immediately plating out dilutions of the initial inoculum onto TSB agar plates.
Mice-Unchallenged Sdc1 Ϫ/Ϫ mice on both the BALB/c and C57BL/6J backgrounds are healthy with normal growth, reproduction, tissue morphology, complete blood cell counts, and serum chemistry parameters (9 -11). Both female and male Sdc1 Ϫ/Ϫ mice and corresponding littermate WT mice on the BALB/c or C57BL/6J background (backcrossed 10 times) were used at an age of 5-8 weeks. Mice were maintained in microisolator cages under specific pathogen-free conditions in a 12-h light/dark cycle and fed a basal rodent chow ad libitum. All animal experiments were approved by the Institutional Animal Care and Use Committee of Children's Hospital, Boston, and complied with federal and Association for Research in Vision and Ophthalmology guidelines for research with experimental animals.
S. aureus Corneal Infection-A single vertical scratch was made with a 29-gauge needle in one of the corneas of each anesthetized mouse without penetrating beyond the superficial stroma. The other eye served as an uninjured control. A 5-l suspension of various doses of S. aureus with or without test agents was applied topically to injured or uninjured corneas. At various times post-infection, mice were euthanized, and the bacterial burden in whole enucleated eyes or isolated corneas was determined. Whole eyes or isolated corneas were homogenized in TSB containing 0.1% (v/v) Triton X-100, and serial dilutions of homogenates were plated onto TSB agar plates. The bacterial burden was consistently similar between isolated whole eyes and isolated corneas, indicating that S. aureus almost exclusively infects injured corneas when applied topically to the ocular surface.
S. aureus Colonization of Epithelial Cells-Mouse A6(1) corneal epithelial cells, derived from the corneal epithelium of 14-day-old Immorto-Mouse (Charles River Laboratories), were obtained from Dr. Joram Piatigorsky (NEI, National Institutes of Health) (37). Confluent A6(1) cells were grown for 1-2 days in 96-well plates at the nonpermissive temperature of 37°C in EpiLife culture medium supplemented with 20% FBS and human corneal growth supplement. Alternatively, confluent A6(1) cells were infected with adenovirus harboring shRNA knockdown constructs of U6 lamin or mouse syndecan-1 at a multiplicity of infection of 10 and incubated for 1 day with virus and for 2 more days after virus removal in Epi-Life culture medium with FBS and human corneal growth supplement. NMuMG epithelial cells were from our culture collection (21) and were used at confluency or infected with adenovirus shRNA constructs and used at confluency. Briefly, to knock down syndecan-1 expression, the top strand CAC CGG TCT ACT TTA GAC AAC TTC GAA AAG TTG TCT AAA GTA GAC C and bottom strand AAA AGG TCT ACT TTA GAC AAC TTT TCG AAG TTG TCT AAA GTA GAC C targeting mouse syndecan-1 were annealed and ligated into linearized pENTR/U6 by T4 ligase. The shRNA region was transferred into pAd/BlockIT/DEST by LR-clonase II, linearized by PacI digestion, and transfected into 293A virus packaging cells. Virus particles were collected, propagated in 293A cells, and stored at Ϫ80°C until use. A6(1) corneal epithelial cells or NMuMG cells transduced with or without adenovirus harboring shRNA constructs were incubated with various doses of S. aureus in 100 l of culture medium for 2 h at 37°C, washed, and lysed with 0.1% (v/v) Triton X-100 in TSB.
The bacterial colonization was quantified by plating serial dilutions of lysates onto TSB agar plates.
Syndecan-1 Shedding at the Ocular Surface-The corneas of anesthetized mice were scratched once with a 29-gauge needle, and 5 l of TSB or TSB with S. aureus (3 ϫ 10 8 cfu) was topically applied. At 3, 7, and 16 h post-infection, mice were euthanized, and their ocular surface was incubated with 5 l of 1% N-acetylcysteine in PBS for 5 min to break the mucous layer of tear film and facilitate recovery of ocular surface fluids (38). The ocular surface was then rinsed with 5 l of 1% N-acetylcysteine and 10 l of PBS, and the ocular surface fluids were collected from the inferior fornix. The recovered ocular surface fluids were combined and spun down, and the concentration of syndecan-1 and -4 ectodomains was determined by dot immunoblotting (20,21).
Neutrophil-mediated S. aureus Killing-Neutrophils were isolated from bone marrows of WT or Sdc1 Ϫ/Ϫ mice by Percoll density gradient centrifugation. Briefly, femurs and tibias were isolated from mice under anesthesia, cleaned, and flushed with Hanks' balanced salt solution (HBSS) containing 10 mM HEPES, pH 7.4 (HBSS/HEPES). Bone marrow cells were centrifuged at 300 ϫ g for 10 min, resuspended in 45% Percoll solution, layered on top of a 62 and 81% Percoll gradient solution, and centrifuged at 1500 ϫ g for 30 min. The neutrophil cell layer between 62 and 81% Percoll was collected, washed, resuspended in HBSS/HEPES, and counted. Neutrophils (10 6 cells) were incubated with S. aureus Woods strain (2-3 ϫ 10 3 cfu) pre-opsonized with S. aureus BioParticles opsonizing reagent in HBSS/HEPES containing 5% mouse serum for 2 h at 37°C. Bacterial killing was enumerated by incubating test samples with 0.1% Triton X-100 in HBSS/ HEPES for 30 min and plating serial dilutions onto TSB agar plates.
Histopathology-Eyes were enucleated at various times post-infection, fixed in 4% paraformaldehyde/PBS for 4 h at room temperature, embedded in paraffin, and sectioned horizontally. Eye sections (5 m) were stained with hematoxylin and eosin or immunostained with anti-mouse GR1, antimouse Mac3, 281-2 anti-mouse syndecan-1 or Ky8.2 antimouse syndecan-4 monoclonal antibodies, and Alexa 594 donkey anti-rat secondary antibodies. Stained tissue sections were visualized with the Zeiss Axiovert 40 CFL microscope, and pictures were taken with the AxioCam MRm high resolution camera. Adobe Photoshop CS4 was used to process the acquired images.
Statistical Analyses-All data are expressed as means Ϯ S.E. Differences between experimental and respective control groups were examined by Student's t test, and p values of less than 0.05 were considered statistically significant.

Sdc1 Ϫ/Ϫ Corneas Significantly Resist S. aureus Infection
Compared with WT Corneas-To determine whether syndecan-1 is an important host determinant for S. aureus corneal infection, we first compared the virulence of S. aureus in injured corneas of WT and Sdc1 Ϫ/Ϫ mice on the BALB/c background. The avascular cornea is remarkably resistant to bacterial infections, but when the corneal epithelial barrier is breached, it becomes susceptible to infection by several opportunistic bacterial pathogens. S. aureus (strain 8325-4) was inoculated topically onto WT and Sdc1 Ϫ/Ϫ corneas injured by a single scratch with a 29-gauge needle. The corneal bacterial burden was determined between 1 and 24 h post-infection by homogenizing the isolated corneas and plating out serial dilutions of the homogenates. The bacterial burden was similar between WT and Sdc1 Ϫ/Ϫ corneas at 1 h post-infection (Fig.  1A). However, the bacterial burden was significantly reduced by 5-fold in Sdc1 Ϫ/Ϫ corneas compared with WT corneas at 7 h (Fig. 1A) and reduced by 3-fold at 24 h after infection (data not shown), indicating that deletion of syndecan-1 is a gain of function mutation that enables mice to rapidly clear S. aureus in the cornea. Similar results were obtained with WT and Sdc1 Ϫ/Ϫ mice on the C57BL/6J background (Fig. 1B) and also by using a different strain of S. aureus (strain P1) with WT and Sdc1 Ϫ/Ϫ mice on the BALB/c background (Fig. 1C), indicating that the ability of Sdc1 Ϫ/Ϫ mice to resist S. aureus corneal infection is not specific to mouse background or bacterial strain. Furthermore, uninjured corneas of WT and Sdc1 Ϫ/Ϫ mice on both the BALB/c and C57BL/6J backgrounds rapidly cleared the S. aureus inoculum within 1 h post-infection (data not shown), consistent with the feature that the intact ocular surface is remarkably resistant to bacterial infections. In addition, Sdc1 Ϫ/Ϫ mice show delayed corneal wound repair (12), and impairment of corneal re-epithelialization is a prominent risk factor for bacterial keratitis, but Sdc1 Ϫ/Ϫ corneas nonetheless rapidly and significantly cleared S. aureus relative to WT corneas. These data suggest that potential epithelial cell defects are not contributing to the resistance seen in Sdc1 Ϫ/Ϫ mice. Together, these findings indicate that either bacterial or host defense mechanisms are significantly altered in Sdc1 Ϫ/Ϫ corneal tissues and suggest that syndecan-1 is an important host factor that promotes S. aureus corneal infection.

Syndecan-1 Is Not Essential in the Colonization of Corneal Epithelial Cells by S. aureus-To determine how Sdc1
Ϫ/Ϫ mice resist S. aureus corneal infection, we examined if S. aureus binds to syndecan-1 and if cell surface syndecan-1 facilitates the colonization of corneal epithelial cells by S. aureus. We unexpectedly found that all three strains of S. aureus examined do not bind avidly to radiolabeled purified syndecan-1 ( Fig. 2A). Compared with the binding of radiolabeled syndecan-1 to affinity beads conjugated with anti-syndecan-1 antibodies, binding to S. aureus strains was only slightly above background levels ( Fig. 2A), and the low level of binding observed was not inhibited by addition of excess unlabeled syndecan-1 ectodomain or HS (data not shown). Furthermore, shRNA-mediated knockdown of syndecan-1 had no effect on S. aureus colonization of cultured mouse corneal (A6(1)) or mammary gland (NMuMG) epithelial cells (Fig. 2B). The knockdown efficiency was ϳ40 and 50% in A6(1) and NMuMG cells, respectively, as determined by measurement of cell surface syndecan-1 by mild trypsin digestion and dot immunoblotting of trypsinates (21,39). Furthermore, consistent with the lack of direct interaction, purified ectodomains or HS did not affect the growth rate of S. aureus (data not shown). These data suggest that cell surface syndecan-1 does not mediate the initial attachment of S. aureus to injured corneas and are consistent with the similar bacterial burden in WT and Sdc1 Ϫ/Ϫ corneas at earlier times post-infection.

S. aureus Infection Induces Syndecan-1 Shedding in Corneal
Epithelial Cells-We next tested if S. aureus infection induces syndecan-1 shedding in the corneal epithelium. We reasoned that if syndecan-1 at the cell surface does not promote pathogenesis by binding to S. aureus and facilitating bacterial attachment, S. aureus may induce syndecan-1 shedding from the surface of corneal epithelial cells, and syndecan-1 ectodo-  FEBRUARY 4, 2011 • VOLUME 286 • NUMBER 5 mains may enhance S. aureus corneal infection. We first examined the expression of syndecan-1 in the cornea. Eye sections of WT mice before and 7 h after S. aureus infection were immunostained with the 281-2 anti-syndecan-1 ectodomain antibody. Uninfected corneas showed strong cell surface syndecan-1 expression in epithelial cells and very weak expression in stromal keratocytes and endothelial cells (Fig. 3A).

Syndecan-1 Promotes S. aureus Keratitis
Upon infection, the signal for cell surface syndecan-1 in corneal epithelial cells, especially in the apical compartment, was markedly diminished across the entire corneal epithelium (Fig. 3A). Steady-state syndecan-1 mRNA levels were not decreased in the cornea by S. aureus infection (data not shown), suggesting that syndecan-1 shedding caused the observed reduction in cell surface syndecan-1 levels. Indeed, levels of syndecan-1 ectodomains in ocular surface fluids increased rapidly and significantly upon S. aureus infection (Fig. 3B). Syndecan-1 ectodomains in ocular surface fluids were increased by over 10-fold at both 3 and 7 h post-treatment in groups whose corneas were injured and infected with S. aureus compared with those that were only injured (Fig. 3B), indicating that infection, and not injury, induces syndecan-1 shedding in corneal epithelial cells. Furthermore, the increase in syndecan-1 ectodomains was transient as it returned to near basal levels by 24 h post-infection (Fig. 3B). Notably, the kinetics of syndecan-1 shedding was associated closely with the clearance of S. aureus in mouse corneas. The bacterial burden was similar in WT and Sdc1 Ϫ/Ϫ corneas at 1 h postinfection where there was minimal syndecan-1 shedding, whereas it was significantly reduced in Sdc1 Ϫ/Ϫ corneas at 7 h post-infection where there was maximal shedding in WT corneas. Syndecan-4 ectodomain levels in ocular surface fluids also increased, but not as rapidly or significantly as syndecan-1 ectodomains (data not shown). Collectively, these observations indicate that S. aureus specifically induces syndecan-1 shedding in corneal epithelial cells and suggest that the lack of this mechanism enables Sdc1 Ϫ/Ϫ mice to efficiently clear S. aureus in their corneas.
Syndecan-1 Shedding Enhances S. aureus Virulence in Injured Corneas-To determine the physiological relevance of syndecan-1 shedding in S. aureus corneal infection, we assessed the effects of administering purified syndecan-1 ectodomains and inhibiting syndecan-1 shedding. If Sdc1 Ϫ/Ϫ mice indeed resist S. aureus corneal infection because they cannot shed their syndecan-1 ectodomains, then administering purified ectodomains should promote and inhibiting shedding should attenuate pathogenesis. To evaluate this hypothesis, injured Sdc1 Ϫ/Ϫ corneas were infected with S. aureus or infected with S. aureus and topically administered 170 ng of purified syndecan-1 ectodomains or HS at 3.5 h after infection, and the bacterial burden was measured at 10 h post-infection. Syndecan-1 ectodomains and HS were applied in a delayed manner to simulate the kinetics of syndecan-1 shedding (Fig. 3B). Administration of purified ectodomains or HS significantly increased the corneal bacterial burden by ϳ3-fold compared with those that were infected by S. aureus only and restored the wild type response to infection with S. aureus (Fig. 4A). Similar increases in the bacterial burden (3.5-fold) were seen when syndecan-1 ectodomains were given at an earlier time (1 h post-infection). Along with the observations that syndecan-1 does not bind to S. aureus and shRNA-mediated gene knockdown of syndecan-1 does not inhibit S. aureus colonization of cultured corneal epithelial cells, these data suggest that syndecan-1 ectodomains enhance S. aureus survival at the ocular surface by affecting mechanisms of the host and not those of the bacteria.  , n ϭ 4). B, A6(1) corneal epithelial cells or NMuMG epithelial cells transduced with adenovirus harboring a knockdown construct of U6 lamin or syndecan-1 were incubated with S. aureus 8325-4 for 2 h at 37°C, and bacterial colonization was assessed (mean Ϯ S.E., n ϭ 4 -6). A6(1) cells were incubated with 3.8 ϫ 10 6 cfu (low dose) or 1.4 ϫ 10 7 cfu (high dose), and NMuMG cells were incubated with 6 ϫ 10 5 cfu (low dose) or 6 ϫ 10 6 cfu (high dose) of S. aureus 8325-4.

Syndecan-1 Promotes S. aureus Keratitis
We next examined the effects of inhibiting syndecan-1 shedding with GM6001, a broadly acting metalloproteinase sheddase inhibitor, on S. aureus corneal infection. Injured WT corneas were infected with S. aureus or co-infected with S. aureus and GM6001. GM6001 significantly reduced levels of syndecan-1 ectodomains in ocular surface fluids of mice co-infected with S. aureus and GM6001 compared with those infected with S. aureus only (Fig. 4B), indicating that GM6001 inhibits corneal syndecan-1 shedding in vivo. Moreover, the bacterial burden was significantly decreased by ϳ2-fold in mice co-infected with S. aureus and GM6001 compared with those that were infected with S. aureus only (Fig. 4C). At the dose tested, GM6001 did not affect the growth rate of S. aureus in vitro, the overall corneal morphology, or mRNA levels of syndecan-1 in the cornea (data not shown). On the basis of these findings, we propose that S. aureus subverts syndecan-1 shedding to promote its pathogenesis in injured corneas.
Neutrophil Depletion Increases S. aureus Virulence in the Resistant Sdc1 Ϫ/Ϫ Corneas-To determine a potential cellular basis for the improved outcomes in Sdc1 Ϫ/Ϫ mice, we next examined the effects of immunodepleting neutrophils in these mice. Neutrophils are rapidly recruited to the site of infection in bacterial keratitis and they play a key role in the clearance of bacterial pathogens (40,41), but exaggerated or prolonged neutrophil responses can contribute to the pathogenesis of bacterial keratitis by causing excessive damage to corneal tissues. Sdc1 Ϫ/Ϫ mice were injected intraperitoneally with anti-GR1 antibodies to immunodeplete neutrophils (9), and injured corneas were infected with S. aureus at 24 h after anti-GR1 injection. Consistent with previous results, the majority of S. aureus was cleared by 7 h post-infection in the corneas of Sdc1 Ϫ/Ϫ mice that were pretreated with vehicle (Fig. 5A). However, the corneal bacterial burden was significantly increased in neutropenic Sdc1 Ϫ/Ϫ mice to the level of WT mice at 7 h post-infection, and the high bacterial burden was sustained at 14 h post-infection (Fig. 5A). These data indicate that neutrophils are essential for Sdc1 Ϫ/Ϫ mice to rapidly clear S. aureus in their corneas and to resist S. aureus corneal infection.
We next examined if neutrophil infiltration was altered in Sdc1 Ϫ/Ϫ corneas. CXC chemokines, such as KC (CXCL1), MIP-2 (CXCL2), lipopolysaccharide-induced CXC chemokine (CXCL5) (42), and complement component C5a (43), are the major chemotactic factors that induce neutrophil infiltration. Syndecan-1 shedding has been shown to moderate neutrophil recruitment either by generating a CXC chemokine gradient that orchestrates the transepithelial migration of neutrophils into lungs (7) or by facilitating the resolution of neutrophilic inflammation by removing CXC chemokines tethered to endothelial cells in various tissues (11). The transient increase in   FEBRUARY 4, 2011 • VOLUME 286 • NUMBER 5

JOURNAL OF BIOLOGICAL CHEMISTRY 3293
syndecan-1 shedding seen in infected corneas (Fig. 3B) was reminiscent of its effects on the generation of a KC gradient across the lung alveolar epithelium (7). Furthermore, several studies have shown that corneal epithelial cells, keratocytes, and resident macrophages can express CXC chemokines (44,45) and that they mediate the recruitment of neutrophils into corneal tissues in response to infectious and noninfectious tissue injury (46,47).
Based on these observations, we assessed whether chemokine expression and neutrophil infiltration were altered in infected Sdc1 Ϫ/Ϫ corneas. Scarified WT and Sdc1 Ϫ/Ϫ corneas were infected with S. aureus, and steady-state mRNA levels of chemokines in isolated corneas were measured. KC and MIP-2 mRNA levels were similarly induced at 7 h post-infection in both WT and Sdc1 Ϫ/Ϫ corneas (Fig. 5B), although KC mRNA levels were slightly decreased in infected Sdc1 Ϫ/Ϫ corneas relative to WT corneas. RANTES, a ligand of CCR5 and mediator of T cell recruitment, was slightly induced by infection but not as significantly as KC or MIP-2. These data indicate that the deletion of syndecan-1 does not increase the expression of CXC chemokines in response to S. aureus corneal infection.
Consistent with these findings, we found that a similar number of neutrophils are recruited to infected WT and Sdc1 Ϫ/Ϫ corneal tissues. Eye sections prepared from WT and Sdc1 Ϫ/Ϫ mice at 7 h after S. aureus infection were stained with hematoxylin and eosin or immunostained with anti-GR1 or -Mac3 antibodies. In both WT and Sdc1 Ϫ/Ϫ mice, abundant accumulation of neutrophils and minor edema were seen mostly in the upper half of the corneal stroma toward the basal epithelial aspect (Fig. 5C). Little to no accumulation of macrophages was observed (data not shown). These results indicate that syndecan-1 does not modulate the influx of neutrophils into S. aureus-infected corneas.
Syndecan-1 Ectodomain Inhibits Neutrophil-mediated Killing of S. aureus in an HS-dependent Manner-Increased susceptibility to S. aureus infections has been linked to a decrease in neutrophil infiltration or function. Because our data indicated that syndecan-1 does not affect neutrophil infiltration, we next examined if syndecan-1 regulates neutrophil function. Isolated WT and Sdc1 Ϫ/Ϫ neutrophils were incubated with pre-opsonized S. aureus, and bacterial killing was assessed by counting live S. aureus recovered from the neutrophil and S. aureus mixture at the end of the experiment. Sdc1 Ϫ/Ϫ neutrophils killed a slightly higher proportion of the bacterial inoculum compared with WT neutrophils (30 versus 25%), but the difference did not reach significance (Fig. 6A), indicating that Sdc1 Ϫ/Ϫ neutrophils do not have an inherent defect in their ability to kill S. aureus. These findings are consistent with previous observations that neutrophils do not express syndecan-1; WT and Sdc1 Ϫ/Ϫ neutrophils are similar in size, granularity, and pattern of GR1 staining; Sdc1 Ϫ/Ϫ neutrophils do not have an inherent defect in their ability to migrate; and Sdc1 Ϫ/Ϫ mice contain normal numbers of circulating neutrophils (7,11).
Based on these data, we next explored if syndecan-1 ectodomains enhance the survival of S. aureus by inhibiting the bacterial killing activity of neutrophils. Isolated neutrophils were incubated with pre-opsonized S. aureus in the absence or presence of purified syndecan-1 ectodomain, HS, or ectodomain devoid of HS, and bacterial killing was assessed. Addition of purified syndecan-1 ectodomain or HS significantly inhibited bacterial killing by 97 and 83%, respectively, whereas ectodomain devoid of HS chains had no inhibitory effect (Fig. 6B). At the doses tested, purified ectodomain and HS did not affect the viability of isolated neutrophils (data not shown), suggesting that the effects of syndecan-1 ectodomains are strictly on the bacterial killing mechanisms of neutrophils. Altogether, these data suggest a previously unknown pathogenic mechanism in S. aureus corneal infection where the bacterium induces syndecan-1 shedding to subvert the capacity of syndecan-1 ectodomains to inhibit neutrophilmediated bacterial killing mechanisms.

DISCUSSION
In this study, we addressed the physiological relevance of syndecan-1 in S. aureus keratitis using a mouse model of scarified corneal infection. We found that ablation of syndecan-1 has a pronounced positive effect on S. aureus clearance in the cornea. The bacterial burden was similar in WT and Sdc1 Ϫ/Ϫ corneas at earlier times post-infection, whereas it was significantly decreased in Sdc1 Ϫ/Ϫ corneas compared with WT corneas at later times post-infection. Furthermore, syndecan-1 did not bind to S. aureus, and gene knockdown of syndecan-1 did not decrease the colonization of cultured epithelial cells by S. aureus. Collectively, these data suggest that syndecan-1 affects mechanisms of the host and not of the bacteria to promote S. aureus pathogenesis in the cornea. Indeed, we found that depletion of neutrophils enhances S. aureus virulence in the resistant Sdc1 Ϫ/Ϫ corneas. However, the expression of KC and MIP-2 and extravasation of neutrophils into infected corneal tissues were similar between WT and Sdc1 Ϫ/Ϫ mice, indicating that a more potent neutrophil-mediated host defense occurred in Sdc1 Ϫ/Ϫ corneas despite a similar number of infiltrated neutrophils. Consistent with these data, we found that S. aureus infection induces syndecan-1 shedding from the surface of corneal epithelial cells and syndecan-1 ectodomains inhibit the capacity of neutrophils to kill S. aureus in an HS-dependent manner. These data reveal a new function of syndecan-1 in infectious diseases where it promotes pathogenesis by inhibiting the neutrophil arm of host defense and increasing bacterial survival in the host.
One of the surprising observations from our studies was that syndecan-1 is not essential for the initial attachment of S. aureus to the injured corneal epithelium, suggesting that other HSPGs or other matrix components mediate S. aureus attachment to corneal tissues. Indeed, the collagen-binding adhesin has been shown to be a virulence factor in a rabbit model of soft contact lens-associated S. aureus keratitis (48), and deletion of S. aureus fibronectin-binding proteins A and B has been shown to reduce S. aureus attachment and invasion of human corneal epithelial cells by 99% (49). These observations suggest that collagen and/or fibronectin, and not HSPGs, are the key host determinants that mediate the initial attachment of S. aureus to injured corneal tissues.
We also conclude that S. aureus takes advantage of the capacity of syndecan-1 ectodomains to inhibit neutrophil-mediated bacterial killing mechanisms by inducing syndecan-1 shedding from the surface of corneal epithelial cells. How S. aureus induces syndecan-1 shedding in corneal epithelial cells is not understood. However, we previously reported that S. aureus ␣and ␤-toxin induce syndecan-1 shedding in lung epithelial and mammary gland epithelial cells (20). Furthermore, studies of S. aureus keratitis in rabbit and mouse models showed that ␣-toxin mediates the majority of the virulence activities (33,50,51), whereas ␤-toxin is also important but contributes less to virulence in the cornea (33). Together, these observations suggest that S. aureus induces syndecan-1 shedding in the cornea through ␣and ␤-toxin, and the capacity to enhance syndecan-1 shedding and generate soluble ectodomains is an important virulence activity of these exotoxin virulence factors.
The pro-pathogenic effects of HS on S. aureus corneal infection suggest that other HSPGs may function similarly to syndecan-1. However, because other syndecans and HSPGs are intact in Sdc1 Ϫ/Ϫ mice, the significant difference in the capacity of WT and Sdc1 Ϫ/Ϫ corneas to clear S. aureus suggests that syndecan-1 ectodomains function specifically to increase bacterial survival. Precisely how syndecan-1 ectodomains act in this manner is not known, but previous studies have also shown that loss of syndecan-1 alone results in dramatic pathological phenotypes in animal models of various infectious and inflammatory diseases (6 -11), suggesting that other HSPGs cannot compensate for the loss of syndecan-1 in vivo. In the corneal epithelium, both syndecan-1 and -4 are expressed, but syndecan-4 is expressed at a lower level than syndecan-1. Furthermore, although syndecan-4 shedding was also induced during S. aureus corneal infection, syndecan-4 ectodomain levels did not increase as rapidly or significantly FIGURE 6. Syndecan-1 ectodomain inhibits the killing of S. aureus by neutrophils. A, WT and Sdc1 Ϫ/Ϫ neutrophils (10 6 ) were incubated with 3 ϫ 10 3 cfu of pre-opsonized S. aureus Woods strain for 2 h at 37°C. Percent bacterial killing was enumerated by plating serial dilutions of detergent lysates onto TSB agar plates (n ϭ 6). B, WT neutrophils (10 6 ) were incubated with 2 ϫ 10 3 cfu of pre-opsonized S. aureus Woods strain in the absence or presence of 1 g/ml purified syndecan-1 ectodomain, HS, or ectodomain devoid of HS for 2 h at 37°C, and bacterial killing was determined (n ϭ 5-6).
as those of syndecan-1. These observations suggest that the specificity of syndecan-1 functions in S. aureus corneal infection may be a reflection of its abundant expression and prominent shedding in corneal epithelial cells. Alternatively, because syndecan-1 acts in an HS-dependent manner, the potential unique structural features of syndecan-1 HS chains may mediate its specific functions in S. aureus corneal infection. However, the structure and function of syndecan-1 and -4 isolated from mammary gland epithelial cells are indistinguishable, and syndecan HS chains are considered to be cell type-specific and not core protein-specific (36), suggesting that the structure and function of syndecan-1 and -4 are also similar in the corneal epithelium. Nonetheless, these observations imply that despite similar cellular distribution and HS structures, corneal syndecan-1 and -4 have nonredundant roles in vivo. Future studies determining the response of Sdc4 Ϫ/Ϫ mice in S. aureus corneal infection and analyzing the fine structure of syndecan-1 and -4 HS chains should address these issues.
Both passive and active defense mechanisms effectively protect the cornea from infections. The mechanical action of the eyelid physically removes pathogen; the outermost lipid layer of the tear film serves as a physical barrier, and the washing effects of tears also remove pathogens (41). Furthermore, the aqueous tear fluid contains a wide array of antimicrobial factors (41). However, all of these defense mechanisms are likely not affected by syndecan-1 because S. aureus was cleared similarly in uninjured corneas of both WT and Sdc1 Ϫ/Ϫ mice. Once the passive host defense mechanisms have failed or the corneal epithelium is breached, neutrophils are actively recruited to the site of infection by various neutrophil-chemotactic factors, such as KC, MIP-2, and lipopolysaccharide-induced CXC chemokine (CXCL5). Syndecan-1 shedding has been shown to mediate the generation of a KC gradient across the lung alveolar epithelium (7) and to facilitate the removal of KC and MIP-2 tethered to endothelial cells in the lung, liver, and kidney (11). Hence, we were most surprised to find that syndecan-1 moderates neutrophil activity and not its infiltration into corneal tissues. Although increased mononuclear leukocyte adhesion onto retinal blood vessels was observed in TNF␣-stimulated Sdc1 Ϫ/Ϫ mice (52), we found no differences in the expression of KC and MIP-2 and infiltration of neutrophils into infected WT and Sdc1 Ϫ/Ϫ corneas, suggesting that mechanisms governing neutrophil influx in response to a single cytokine and an active bacterial infection may be different. Furthermore, because lumican and keratocan core proteins have been shown to bind to CXC chemokines and form a chemokine gradient that directs neutrophil migration during a corneal inflammatory response (44,53), mechanisms governing the generation of a CXC chemokine gradient in corneal tissues may be different from those of other tissues. In addition, because TLR2 and MyD88 are required for the induction of KC and MIP-2 in injured corneas and for the recruitment of neutrophils in response to S. aureus corneal infection (45), our results suggest that signaling events through TLR2 and MyD88 are also not affected by syndecan-1 in S. aureus corneal infection.
The underlying mechanisms of how syndecan-1 ectodomains inhibit the capacity of neutrophils to kill S. aureus in an HS-dependent manner remain unknown. Because syndecan-1 ectodomain does not bind to S. aureus and does not affect the viability of isolated neutrophils, ectodomains must be inhibiting key processes of neutrophils that kill bacteria. Furthermore, because our preliminary data suggest that sulfated domains in HS interfere with the killing of S. aureus by neutrophils, cationic defense factors and their mechanisms are presumably inhibited by syndecan-1 ectodomains. However, given that cell surface receptors for syndecan-1 or HS on neutrophils have not been reported and syndecan-1 ectodomains are not membrane-permeable, syndecan-1 ectodomains are likely acting on extracellular killing mechanisms of neutrophils. Activated neutrophils secrete a highly cationic heparin-binding protein (HBP, azurocidin/CAP37) that has antimicrobial activities against several bacterial pathogens, including S. aureus (54). HBP has also been shown to increase the phagocytosis of S. aureus by monocytes and macrophages (55). These observations suggest that syndecan-1 HS chains may bind to HBP and interfere with its bacterial killing and phagocytosis-potentiating activities, although it is not known if HBP affects bacterial phagocytosis by neutrophils.
Alternatively, syndecan-1 ectodomains may inhibit a recently described extracellular killing mechanism of neutrophils called neutrophil extracellular traps (NETs) (56). NETs are made by activated neutrophils, but not naive neutrophils, and consist of negatively charged, decondensed chromatin fibers with cationic antimicrobial factors, such as HBP, cathelicidins, defensins, and elastase, embedded in them. NETs trap pathogens and facilitate killing by bringing the pathogens and neutrophil-derived antimicrobial factors to proximity. Importantly, impaired NET formation in vivo predisposes the host to bacterial infections (57,58). Because functional NETs are formed by the ionic interactions between anionic chromatin fibers and cationic antimicrobial factors, the highly anionic HS chains of syndecan-1 ectodomains may bind to and displace the cationic antimicrobial factors from anionic NET fibers and inhibit NET-mediated killing of S. aureus. Altogether, these observations raise the possibility that syndecan-1 ectodomains may inhibit one of more bacterial killing mechanisms of neutrophils. Further studies will be required to elucidate the impact of syndecan-1 ectodomains on these host defense mechanisms of neutrophils in corneal tissues.
In summary, our findings extend the biological functions of syndecan-1 from that of a cell surface proteoglycan that regulates cell adhesion, proliferation, and migration to that of a soluble HSPG ectodomain capable of modulating neutrophilmediated host defense mechanisms crucial to the clearance of S. aureus in injured corneas. Interestingly, the majority of bacterial pathogens that can induce syndecan-1 shedding in vitro are major etiological agents of bacterial keratitis in vivo (e.g. S. aureus, P. aeruginosa, and S. pneumoniae), suggesting that subversion of the capacity of syndecan-1 ectodomains to counteract neutrophil-mediated bacterial killing is a general virulence mechanism of various ocular surface pathogens. Although the normal functions of syndecan-1 in the resting cornea remain to be defined, our findings suggest a possible Syndecan-1 Promotes S. aureus Keratitis beneficial role of inhibiting syndecan-1 shedding or neutralizing syndecan-1 ectodomains in treating bacterial keratitis.