Pseudomonas aeruginosa is an important opportunistic human pathogen that can cause infection at various body sites, including the cornea. In cases of bacterial keratitis that are contact lens associated, P. aeruginosa is the most common causative organism. ^1,2 P. aeruginosa keratitis is a rapidly progressive and destructive disease that can cause severe symptoms, such as ulcers, corneal perforation, and subsequent loss of vision. ³  

The pathogenesis of P. aeruginosa keratitis is complex and multifactorial, involving both bacterial factors and host components. As a result, eradication of the bacteria by antibiotics does not readily limit the pathological effects. Proteases produced by P. aeruginosa have long been studied as possible mediators of corneal damage and essential virulence factors during keratitis. Early studies with crude or purified Pseudomonas proteases demonstrated their toxicity to the rabbit cornea. ^4–6 Furthermore, a spectrum of proteins pertinent to host defense or tissue integrity was found to be susceptible to degradation by Pseudomonas proteases. ^7–9 The best evidence of a protease as an important virulence factor comes from studies of a protease-deficient mutant that, relative to the parent or rescue (complemented) strain, has reduced virulence but no decrease in its ability to grow in the rabbit cornea. 

Among the well-studied P. aeruginosa proteases are the three metalloproteases: alkaline protease (AP), elastase A (LasA), and elastase B (LasB). ^10–12 Infection of animal corneas with a mutant P. aeruginosa strain deficient in AP, LasA, or LasB has determined that none of these enzymes is essential for corneal virulence. ¹³ The secretion of active P. aeruginosa protease has been achieved by cloning into Pseudomonas putida , the specific P. aeruginosa protease gene and its auxiliary genes needed for protease maturation and secretion. Because P. putida lacks a secreted protease and grows in the rabbit cornea without causing pathological effects, infection with P. putida secreting a P. aeruginosa protease can reveal the specific contribution of the cloned protease to corneal virulence. P. putida secreting LasB caused significantly greater corneal pathology than its parent strain harboring the plasmid vector alone. ¹⁴ P. putida secreting AP was not significantly more virulent than its parent strain. ¹⁴ Furthermore, P. putida secreting AP was significantly less virulent than the P. putida strain that secreted LasB. ¹⁴ However, the importance of LasB to corneal infections is probably limited because multiple clinical strains of P. aeruginosa have been found to lack LasB expression. ^15,16 Also, a study by Kernacki et al. ¹⁷ demonstrated that LasB was not produced in vivo during corneal infection with P. aeruginosa . Taken together, the results in these experimental models indicate that none of the three metalloproteases of P. aeruginosa likely plays a crucial role in the pathogenesis of keratitis. 

P. aeruginosa also secretes a serine protease designated as protease IV (PIV). PIV has been demonstrated to be an important virulence factor in the rabbit cornea. ^18,19 PIV is found in all clinical isolates analyzed, a finding that differs from AP and LasB. ¹⁶ The loss of the piv gene has been shown to significantly reduce corneal virulence, and complementation (rescue) of this gene resulted in restored corneal virulence. ¹⁶ Also, the secretion of PIV by P. putida resulted in substantial corneal virulence, significantly more than the P. putida parent strain. ²⁰ PIV can degrade proteins of importance to host defense, including complement, fibrinogen, plasminogen, immunoglobulin, and surfactant proteins. ^9,21  

P. aeruginosa small protease (PASP) is a more recently identified protease. ²² The pasP gene, like the gene for PIV, has been found in all P. aeruginosa isolates analyzed. ²³ Injection of purified recombinant PASP into the rabbit cornea causes severe ocular pathology, including epithelial and stromal erosions, edema, and neutrophil infiltration into the corneal stroma. ²³ PASP has also been shown to degrade host proteins important to the corneal structure, such as collagens. ²³ Attempts to express PASP in P. putida did not achieve secretion of the active enzyme, indicating that yet to be identified auxiliary genes of P. aeruginosa are required for PASP secretion (Tang A, unpublished findings, 2008). 

The present study investigated the virulence contribution of PASP by analyzing the virulence of the PASP-deficient mutant, its parent strain, and a rescue strain in both rabbit and mouse keratitis models. Additionally, the ability of PASP to cleave several key host proteins is described herein. 

P. aeruginosa strain PA103-29 has been well characterized in the rabbit model of keratitis and as a source of PASP. ^16,23 PA103-29 is known to lack LasB and LasA production. ²⁴ Bacteria were grown in tryptic soy broth (TSB; BD Biosciences, Sparks, MD) at 37°C with shaking. The inoculum for corneal infections was grown overnight in TSB, subcultured in fresh TSB, and grown to log phase (0.2–0.3 OD at 650 nm) before being diluted in TSB. The number of bacteria in the inoculum was verified by culturing dilutions of the bacteria on tryptic soy agar (TSA) in triplicate. 

To construct the allelic exchange vector, as shown in Figure 1A, the flanking regions of the pasP gene and the Tet gene were PCR amplified with additions of restriction enzyme sites at both ends. These three PCR products were ligated and cloned into an intermediate plasmid, pGEM-T (Promega Corporation, Madison, WI). Then, the cassette containing the flanking regions and the Tet gene was excised and cloned into the suicidal vector pEX100T (kindly provided by Dr J. Hobden, Louisiana State University Health Sciences Center, New Orleans, LA), forming plasmid pEX100T-ΔPASP. The pEX100T vector expresses the sacB gene (levansucrase) that renders the host bacterium susceptible to sucrose. Introduction of pEX100T-ΔPASP into P. aeruginosa was accomplished by conjugal mating. ²⁵ Briefly, Escherichia coli carrying pEX100T-ΔPASP was mated with the E. coli strain harboring the conjugal plasmid (pRK2013) to generate a strain with both plasmids. Subsequently, this E. coli strain was mated with PA103-29 (recipient strain) and the transconjugants were selected on Pseudomonas isolation agar (BD Biosciences) supplemented with 10% sucrose and 150 μg/mL tetracycline. 

To construct the rescue strain, the pasP gene of PA 103-29 was cloned into the Pseudomonas/ E. coli shuttle vector pUCP20, conferring carbenicillin resistance (Fig. 1B). The resulting pUCP20-pasP plasmid was introduced into the PASP-deficient mutant by chemical transformation as previously described. ²⁰  

DNA from a P. aeruginosa colony was used as the template for PCR and the reactions for the pasP and Tet genes were conducted under routine PCR conditions (94°C for 1 minute; then 30 cycles of 94°C for 20 seconds, 55°C for 20 seconds, 68°C for 1 minute). To further confirm the correct replacement of the pasP gene of the PASP-deficient mutant, the affected chromosomal region was PCR-amplified and cloned into pGEM-T for sequencing. This chromosomal amplicon (∼4 kb) covering additional 150 bp outside the exchange cassette was sequenced in both directions (SeqWright, Inc., Houston, TX). 

Pseudomonas cultures were grown for 30 hours in M9 minimum medium as previously described. ⁹ For plasmid maintenance, the M9 medium was supplemented with 100 μg/mL carbenicillin. Bacteria were removed by centrifugation and the supernatant was filtered through a 0.22-μm filter, and concentrated 50-fold using an ultrafiltration cell (Amicon; Millipore, Billerica, MA) with a 10-kDa molecular weight cutoff filter. Western blot analysis on the concentrated culture supernatants was performed as previously described. ²³  

Specific pathogen-free New Zealand white rabbits were obtained from Charles River Laboratories International, Inc. (Wilmington, MA). Rabbits were maintained according to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. Rabbits were anesthetized by subcutaneous injection of a mixture of xylazine (100 mg/mL; Butler Co., Columbus, OH) and ketamine (100 mg/mL; Fort Dodge Animal Health, Fort Dodge, IA). Proparacaine (Bausch and Lomb, Tampa, FL) was topically applied to the eyes before injections. Pseudomonas cultures were grown to log phase and 1000 colony-forming units (CFU) of bacteria in 10 μL TSB were intrastromally injected into rabbit corneas. 

Male C57/B6 mice, 4 to 5 weeks old, were obtained from Harlan Laboratories, Inc. (Indianapolis, IN). To prepare the ketamine solution for mouse anesthesia, a 1.5-mL volume of xylazine was mixed with 10 mL ketamine and diluted 1:4 with saline. Mice were subcutaneously injected with this solution at a dose of 0.1 mL/20 g body weight. The mouse keratitis protocol was developed based on methods described by Kwon and Hazlett, ²⁶ Girgis et al., ²⁷ and McCormick et al. ²⁸ Pseudomonas cultures were grown to mid-log phase (OD₆₅₀ = 0.35, approximately 2 × 10⁸/mL). Mouse corneas were scarified by making four to six parallel abrasions in the corneal epithelium using a 30.5-gauge needle. An inoculum with 10⁶ CFU bacteria in 5 μL TSB was topically applied to each scarified eye. 

Rabbit eyes (n ≥ 8) underwent slit lamp examination (SLE) at 29 hours postinfection (PI) by two masked observers. Each of the seven parameters (injection, chemosis, iritis, hypopyon, corneal infiltrate, fibrin in the anterior chamber, corneal edema) was graded on a scale ranging from 0 (normal) to 4 (severe), as described by O'Callaghan et al. ²⁹ The sum of these grades for an eye, after averaging, determined the SLE score, which could range from 0 to a theoretical maximum of 28. 

Mouse eyes (n ≥ 8) were scored at 24 hours PI on the following scale: 0, clear and normal; 1, light corneal haze; 2, readily detectable corneal opacity; 3, infiltrate partially covering the pupil; 4, infiltrate covering the entire pupil and an ulcer fully formed; and 5, corneal perforation or panophthalmititis. ²⁸ Mean SLE scores (±SEM) were calculated. 

Rabbit corneas (n ≥ 4) were harvested at 29 hours PI and homogenized in 3.0 mL sterile PBS. A 0.1-mL aliquot of the homogenate was serially diluted (1:10) in PBS and plated on TSA in triplicate. 

Whole mouse eyes (n ≥ 4) were harvested at 24 hours PI and individually homogenized in 1.0 mL PBS. Mouse eye homogenates were diluted in PBS and cultured on Pseudomonas isolation agar. CFU was determined and expressed as base 10 logarithms. 

Rabbit corneas were harvested at 29 hours PI (n ≥ 3). After fixation, specimens were processed by Excalibur Pathology, Inc. (Oklahoma City, OK). The tissue sections were stained with hematoxylin and eosin. Each slide was examined microscopically and photographed under low magnification (×40). 

Recombinant PASP (rPASP) and native PIV were purified as previously described by Tang et al. ²³ and Engel et al., ⁹ respectively. Tears were collected using capillary tubes placed into the cul-de-sac of normal rabbit eyes. Human host proteins, including fibrinogen (Sigma-Aldrich, St. Louis, MO), complement C3 (Calbiochem-Novabiochem Corp., La Jolla, CA), and antimicrobial peptide LL-37 (Phoenix Pharmaceuticals, Inc., Burlingame, CA), as well as rabbit tears, were incubated with dH₂O, active rPASP, an equal volume of heat-inactivated rPASP, or PIV at 37°C for various periods of time. Aliquots of the reaction mixtures were electrophoresed on 12% SDS-acrylamide gels and stained with Coomassie blue. The degradation of proteins was indicated by changes in electrophoretic mobility of protein components. 

Mean and SEM were calculated using Microsoft Excel (Microsoft Corporation, Redmond, WA). For comparison of SLE scores, nonparametric one-way analysis of variance (Kruskal-Wallis test) was performed. ²⁹ Student's t-test was used for CFU determination data. P less than or equal to 0.05 was considered significant. 

The entire pasP gene of PA103-29 was replaced with the tetracycline resistance gene Tet via allelic exchange (Fig. 1A). This exchange event was confirmed by sequencing the affected chromosomal region from base 469,356 to base 472,785 (pasP: 470,662–471,237). There was no unintended change found in the flanking regions of the pasP gene. To construct the rescue strain, the pasP gene of PA103-29 was cloned into the expression vector pUCP20, forming a plasmid designated as pUCP20-pasP (Fig. 1B). The construct pUCP20-pasP was confirmed by DNA sequencing. Plasmid pUCP20-pasP was introduced by transformation into the mutant (PA103-29ΔPASP), forming a rescue strain. 

PCR analysis was performed on the parent, PASP-deficient, and rescue strains to verify the presence or absence of the pasP and Tet genes. As expected, the pasP gene was detected in the parent and rescue strains, whereas the Tet gene was detected in the PASP-deficient and rescue strains (Fig. 2A). In vitro, the PASP-deficient mutant grew as well as the parent or rescue strain; and the rescue strain, under carbenicillin selection (100 μg/mL), grew as well as the other two strains without antibiotic selection (data not shown). 

To analyze the production and secretion of the PASP protein, concentrated culture supernatants of the three strains were subjected to Western blot analysis (Fig. 2B). The PASP band of correct size (19 kDa) appeared only in the supernatants of the parent and rescue strains, but not in that of the PASP-deficient mutant. 

To examine the virulence contribution of PASP, experimental bacterial keratitis was initiated in both rabbits and mice by inoculating their corneas with the parent, PASP-deficient mutant, or rescue strain. SLE, bacteria enumeration, and/or histopathological analysis were performed in these animal studies. 

Intrastromal injection into the rabbit cornea of 1000 CFU of log-phase parent strain, PASP-deficient mutant, or rescue strain caused apparent pathology in all rabbit eyes at 29 hours PI. However, the disease in the eyes injected with the PASP-deficient mutant appeared milder than that of the eyes injected with the parent or rescue strain (Fig. 3A). The eyes injected with the parent or rescue strain, as compared with eyes injected with the PASP-deficient mutant, demonstrated more pronounced conjunctival swelling, corneal infiltration, and epithelial erosions. Also noted is that, at 16 hours PI, all of the infected rabbit eyes looked similar. 

At 29 hours PI, the eyes injected with the parent, PASP-deficient, or rescue strain produced a mean SLE score of 14.58 ± 1.09, 9.25 ± 1.49, or 17.75 ± 1.39, respectively (Fig. 3B). The SLE score of the eyes injected with the mutant was significantly lower as compared with the score of the eyes injected with the parent (P = 0.03) or the rescue strain (P = 0.011). The SLE score of the eyes injected with the rescue strain was not significantly higher than that of the eyes injected with the parent strain (P = 0.109). The parent, PASP-deficient, and rescue strains grew equally, reaching 7.23 ± 0.09, 7.11 ± 0.27, or 7.40 ± 0.08 log CFU at 29 hours PI, respectively (P = 0.971). 

The infected corneas underwent histopathological analysis at 29 hours PI (Fig. 4). In the corneas injected with the parent or rescue strain, the erosion of the corneal epithelium, the infiltration of neutrophils into the corneal stroma, and corneal edema were much more severe than that caused by the PASP-deficient mutant. 

The eyes of C57/B6 mice were scarified and topically inoculated with 1 × 10⁶ CFU of the strains. All three strains caused pathology in the mouse eyes by 24 hours after inoculation. However, the eyes inoculated with the PASP-deficient mutant demonstrated significantly less pathology than those eyes inoculated with the parent or rescue strain (Fig. 5A). The eyes inoculated with the parent or rescue strain had extensive corneal infiltration covering more than three-quarters of the cornea, whereas the infiltrate in the eyes inoculated with the PASP-deficient mutant was light, covering less than half of the cornea. The mouse eyes inoculated with the PASP-deficient mutant had a significantly lower SLE score than the eyes inoculated with the parent (P = 0.009) or rescue strain (P = 0.007) at 24 hours PI (Fig. 5B). The bacterial load for all three strains in the mouse eyes at 24 hours PI was equivalent, approximately 5.22 ± 0.08, 5.32 ± 0.14, or 5.54 ± 0.14 log CFU for the eyes inoculated with the parent, PASP-deficient, or rescue strain, respectively (P = 0.990). At 48 hours PI, mouse eyes of all three groups showed similar pathology and contained more than 6 logs CFU. By 72 hours PI, all eyes had more than 7 logs CFU and the pathology was severe in all three groups. 

Pseudomonas proteases are known to degrade some important components of the host defense system, such as complement and immunoglobulin. ^7,9 The degradation of several host defense molecules by PASP was determined by changes in electrophoretic mobility of protein components during SDS-PAGE. PASP was found to degrade fibrinogen α- and β-chains, but not γ-chain, after a 4-hour incubation (Fig. 6A). PASP degraded fibrinogen fully after a longer incubation (24 hours; data not shown). Human complement C3 α-chain was also susceptible to PASP degradation (Fig. 6B). Both PASP and PIV degraded the antimicrobial peptide LL-37 (Fig. 6C). When incubated with rabbit tears, PASP and PIV digested two ranges of tear proteins, one range above 80 kDa and the other between 25 and 45 kDa (Fig. 6D). Among the host proteins tested, none of the immunoglobulin types (IgA, IgG, and IgM) was susceptible to PASP degradation, and neither was fibrin, complement C1q, plasmin, or plasminogen (data not shown). 

The role of PASP in corneal virulence has been investigated in the present study by analyzing the virulence of the PASP-deficient mutant in two animal models of P. aeruginosa keratitis. The PASP-deficient mutant demonstrated significantly reduced virulence in both the rabbit intrastromal injection model and the mouse scratch model of keratitis. The PASP-deficient mutant in both models grew as well as the parent or rescue strain, but produced a significantly lower SLE score. Both histological analysis and gross examination of the eyes infected with the PASP-deficient mutant showed less inflammation and tissue damage than that of eyes infected with either the parent or rescue strain. PASP was found to degrade complement C3 and antimicrobial peptide LL-37, two members of the innate host defense system. Like PIV, PASP was also able to degrade multiple proteins of the rabbit tear film. 

The construction of the PASP-deficient mutant was achieved through allelic exchange with the entire open reading frame of the pasP gene being deleted and replaced with the Tet gene. Sequencing data confirmed the Tet replacement of the pasP gene occurred without any unintended DNA changes in the regions neighboring the pasP gene. Hence, the PASP-deficient mutant is a true isogenic mutant of PA103-29. As demonstrated by Western blot analysis, the PASP protein was absent in the culture supernatant of the mutant strain, but was present in culture supernatants of both the parent and rescue strains. Thus, the expected phenotypes of the strains were demonstrated. 

The present research has demonstrated that the deletion of the pasP gene did not alter the bacterial growth in vitro, or in vivo in the rabbit or mouse cornea. In the rabbit model, the rescue strain produced a higher SLE score than the parent strain, although it was not a statistically significant difference. There is a possibility that the increased production of PASP, which could be due to the multiple copies of the rescue plasmid in a cell, could have enhanced the virulence of the rescue strain. Previously, Caballero et al. ¹⁶ showed that a PIV-deficient mutation resulted in significantly reduced virulence in the rabbit model of keratitis, whereas a rescue strain had virulence equivalent to the parent strain PA103-29. When considered together, the findings of the Caballero et al. study ¹⁶ and the present study suggest that PIV and PASP are both important corneal virulence factors, but they cannot substitute for each other to achieve full corneal virulence when one protease is missing. Studies of PASP, coupled with those of PIV, strongly support the hypothesis that Pseudomonas proteases play a major role in keratitis. These two enzymes, unlike other Pseudomonas proteases, are the focus of this consideration because they are found in all strains tested. ^16,23,30  

There are now some ideas that could explain what the proteases accomplish during infection. PASP, unlike PIV, has a potent action on collagens, thus mediating the loss of epithelial cells and the formation of an ulcer extending down into the corneal stroma. ²³ The elimination of epithelial cells may remove a potent source of proinflammatory cytokines and other defensive molecules (e.g., LL-37). PIV is not efficient in cleaving collagens and does not cause erosions when expressed in nonpathogenic P. putida during experimental keratitis. ^20,23 However, PIV, like PASP, can cleave both C3 and LL-37, two very active members of the host defense system. PIV has also been demonstrated to digest surfactant proteins, and these molecules have been shown to protect the cornea from Pseudomonas infection. ^21,28,31 Complement mediates a protective inflammatory response that delivers both cellular and humoral components of the host defense system to the site of infection, so a reduction in the available pool of complement is considered a means to reduce the host defense against infection. Both C3 and LL-37 facilitate phagocytosis of bacterial pathogens, so the proteolytic destruction of these proteins could favor survival of the bacteria at the site of infection. ^32,33 LL-37 is also a proven nonspecific defensive molecule for the cornea that has bactericidal activity. LL-37, which corresponds to the C-terminal 37 amino acids (4.5 kDa) of the human antimicrobial protein cathelicidin (hCAP18), can form pores in bacterial cell membranes resulting in cell death. ³⁴ LL-37 is produced by neutrophils and macrophages, and, very importantly, by epithelial cells of the cornea. ³⁵ Huang et al. ³⁶ reported that mice deficient in cathelicidin exhibited increased susceptibility to P. aeruginosa keratitis, indicating the importance of LL-37 in defending the cornea from microbial infections. 

Degradation of fibrinogen, but not fibrin, by PASP may delay corneal wound healing by reducing the source of fibrin whose deposition in a wound is one of the initiating events in wound healing. ³⁷ P. aeruginosa primarily infects compromised tissue and corneal epithelial defects compromise the cornea to P. aeruginosa keratitis. By limiting corneal healing, a corneal defect could persist instead of being rapidly healed. Another aspect of wound healing that is inhibited by Pseudomonas proteases is the cleavage of plasminogen by PIV. ⁹ PIV, but not PASP, has been demonstrated to degrade plasminogen, thus removing the source of plasmin. Plasmin, a serine protease, is used to dissolve the meshwork of fibrin formed during wound healing. ³⁷ By digesting plasminogen, PIV possibly deprives the compromised tissue site of plasmin, thus delaying the process of normal wound healing. 

Although there are some similarities between the activities of PASP and PIV, there are some important differences that could explain why one enzyme cannot substitute for the other. Both can cleave C3 and LL-37 and both digested tear constituents in a similar manner. Among the differences between the two proteases is the ability of PIV, but not PASP, to cleave immunoglobulin and plasminogen, two molecules that contribute to the host defense system. However, PASP, unlike PIV, can cleave collagens favoring the loss of the epithelium and the formation of a corneal erosion. Thus, PIV and PASP share some activities, but there are important reactions contributing to virulence that can be performed by only one of these enzymes. 

Previous studies have shown that TLCK, a serine protease inhibitor, can inhibit both PIV and PASP activity. ^9,23 Unfortunately, TLCK is highly toxic to the eye and cannot be used as therapy in vivo. Because both proteases are important corneal virulence factors, there would be benefit in using an effective serine protease inhibitor as part of the therapy for Pseudomonas keratitis. In the past, researchers have tested a synthetic matrix metalloprotease inhibitor to treat P. aeruginosa keratitis. ³⁸ Studies also have shown that proteases of P. aeruginosa can activate host matrix metalloproteases and lead to host-derived damaging effects. ^39–41 As a result, this matrix metalloprotease inhibitor effectively inhibited the metalloproteases produced by P. aeruginosa , and topical treatment with the inhibitor prevented the corneal destruction caused by injection of sterile culture filtrate. However, when live bacteria were injected, the inhibitor could not prevent the corneal damage. ³⁸ These findings suggest that a protease inhibitor, or a cocktail of inhibitors, with a focus on serine protease inhibitors, could work better in preventing corneal pathology. 

The main finding of the present study is that the isogenic PASP mutant causes significantly reduced pathology in the rabbit or mouse cornea. PASP is also found to degrade proteins important to the corneal structure and in host defense. In conclusion, PASP is a commonly produced protease of P. aeruginosa that is essential for full corneal virulence. 

The authors thank Michael Bierdeman and Angela Arana for their kind assistance. 

Supported by funds from National Institutes of Health Grant EY12961. 

Disclosure: A. Tang, None; A.R. Caballero, None; M.E. Marquart, None; R.J. O'Callaghan, None