Pseudomonas aeruginosa PilY1 Binds Integrin in an RGD- and Calcium-Dependent Manner

PilY1 is a type IV pilus (tfp)-associated protein from the opportunistic pathogen Pseudomonas aeruginosa that shares functional similarity with related proteins in infectious Neisseria and Kingella species. Previous data have shown that PilY1 acts as a calcium-dependent pilus biogenesis factor necessary for twitching motility with a specific calcium binding site located at amino acids 850–859 in the 1,163 residue protein. In addition to motility, PilY1 is also thought to play an important role in the adhesion of P. aeruginosa tfp to host epithelial cells. Here, we show that PilY1 contains an integrin binding arginine-glycine-aspartic acid (RGD) motif located at residues 619–621 in the PilY1 from the PAK strain of P. aeruginosa; this motif is conserved in the PilY1s from the other P. aeruginosa strains of known sequence. We demonstrate that purified PilY1 binds integrin in vitro in an RGD-dependent manner. Furthermore, we identify a second calcium binding site (amino acids 600–608) located ten residues upstream of the RGD. Eliminating calcium binding from this site using a D608A mutation abolished integrin binding; in contrast, a calcium binding mimic (D608K) preserved integrin binding. Finally, we show that the previously established PilY1 calcium binding site at 851–859 also impacts the protein's association with integrin. Taken together, these data indicate that PilY1 binds to integrin in an RGD- and calcium-dependent manner in vitro. As such, P. aeruginosa may employ these interactions to mediate host epithelial cell binding in vivo.


Introduction
Pseudomonas aeruginosa is a Gram-negative, opportunistic pathogen prevalent in immunocompromised patients, burn victims, and people with cystic fibrosis. P. aeruginosa can infect any part of the body, but typically targets the respiratory tissues, skin abrasions, and the urinary tract [1]. This pathogen accounted for ,11% of the hospital-acquired infections reported in the U.S in 2002 [2,3], and has also been shown to infect other mammalian species, as well as insects and plants [4,5]. P. aeruginosa uses a range of methods to resist the effects of antibiotics, including efflux pumps, adaptive mutagenesis, and protective biofilms [6,7,8]. As such, Pseudomonas aeruginosa presents a significant challenge to human health.
P. aeruginosa employs type IV pili (tfp) for twitching motility and infection. The precise mechanism of host cell attachment has remained unclear, although evidence exists that loops of the major pilus structural subunit PilA exposed at the tip of the pilus fiber bind to gangliosides GM1 and GM2 [9]. In contrast, some data indicate that several P. aeruginosa clinical isolates, as well as laboratory strains, do not employ GM1 and GM2 during host cell attachment [10,11]. As such, it has been proposed that other factors on the Pseudomonas tfp are involved in binding to target cells [12,13].
Recent studies have suggested that host cell integrin proteins play a role in Pseudomonas infection. Anti-integrin antibodies were shown to reduce P. aeruginosa attachment to host cells [14]. Specifically, antibodies to the aVb5 integrin and aVb3 integrin were effective at disrupting P. aeruginosa binding to host cells, with antibodies to aVb5 integrin having the most pronounced effect [14]. Integrins are present on the epithelial cell surface of tissues infected by P. aeruginosa; indeed, aVb5 integrin is highly expressed in the lungs [15]. The presence of P. aeruginosa has also been shown to increase the expression of integrin subunits aV, a5, and b1 in epithelial cells [16,17]. Integrins and integrin-like proteins are found widely in animals, plants, and insects, all targets of P. aeruginosa infection [18]. Integrins are responsible for a range of cellular processes, including cell-cell attachment, cellular signaling, and angiogenesis [19]. The most common ligands for integrins are proteins that contain an arginine-glycine-aspartic acid (RGD) sequence, although other short peptides have also been found to mediate integrin-protein interactions (e.g., leucine-aspartic acidvaline or LDV).
The PilC class of proteins in Neisseria gonorrhoeae, Neisseria meningitidis, and Kingella kingae have been characterized as tfp biogenesis factors and as proteins involved in adhering to target tissues [20,21,22]. Generally, PilC N-terminal regions are associated with adhesion domains, while the C-terminal domains regulate tfp biogenesis. P. aeruginosa PilY1 shares sequence homology with the C-terminal regions of the PilCs proteins, and this domain in PilY1 has been shown to be a calcium-mediated tfp biogenesis factor (Table S1) [23]. Indeed, PilY1 is required for both twitching and swarming motility and adhesion to host cells [23,24,25]. The N-terminal domains of PilY1 and the PilCs share low sequence similarity, which might be expected if these domains are involved in target cell-specific activities. The target tissues of the Neisseria strains and Kingella are likely to be distinct from those targeted by P. aeruginosa. Here, we focus on putative regions of PilY1 involved in host cell attachment and we show that purified PilY1 binds to integrins in an RGD-and calcium-dependent manner in vitro.

PilY1 contains a conserved integrin binding motif
Given the importance of the tfp associated PilY1 protein in host tissue adherence and the ability of antibodies targeting the RGD binding domains of aVb5 and aVb3 integrins to reduce interactions between P. aeruginosa and host cells, we examined PilY1's sequence to locate possible integrin binding motifs (RGD or LDV) [14,25]. PilY1 was found to contain a conserved RGD motif ( Figure 1A, 1B). No other type IV pilus proteins or known surface proteins in P. aeruginosa contained an RGD or other putative integrin binding sequence [26,27]. One RGD motif was found in each of the PilY1s of P. aeruginosa strains of known sequence. For example, in the PilY1 from P. aeruginosa strain K (PAK), the RGD is located at residues 619-621; however, in the PAO1 P. aeruginosa strain PilY1, the RGD is at 657-659 ( Figure 1B). Based on these observations, we hypothesized that PilY1 binds to integrin in an RGD-dependent manner.

PilY1 binds integrin in an RGD-dependent manner
To test the hypothesis that PilY1's RGD motif mediates contacts with integrins, we created a purified form of PilY1 that starts at amino acid 532 and extends to C-terminus of the protein (residue 1163) and tested the ability of 532-1163 PilY1 to bind to aVb5 using an integrin binding assay. Titration of increasing concentrations of 532-1163 PilY1 showed that mid-nanomolar concentrations of this purified protein binds specifically to aVb5 ( Figure 2A). Next, we employed synthetic RGDS and GRADSP peptides to compete with PilY1 binding to aVb5 integrin. We found that that the RGDS peptide significantly reduced PilY1 binding to aVb5 integrin (p,0.10), while the GRADSP control peptide had no effect ( Figure 2B). Because previous reports also indicated that PAK binding to host cells can be reduced by using an anti-aVb3 integrin antibody, we also examined PilY1 binding to aVb3. We found that purified 532-1163 PilY1 binds to aVb3 integrin in an RGD-dependent manner, although with a lower apparent affinity than observed with aVb5 integrin ( Figure 2C) [14]. We therefore used aVb5 as the primary integrin of this study. Finally, we tested two RGD mutants of 532-1163 PilY1, D621A (RGA) and D619-621 (DRGD), for their ability to bind to aVb5 integrin in vitro. Deleting the RGD residues or mutating the aspartic acid residue at position 621 to alanine significantly aeruginosa PilY1 were aligned using the biology workbench server [26,27]. Blue residues and an ''*'' corresponds to identical residues throughout the 5 strains, green residues and a '':'' correspond to highly conserved residues, and navy residues and a ''.'' corresponds to mildly conserved residues (e.g. alanine and leucine). Reference numbers are for PAK_287 PilY1 (the strain used in this study). doi:10.1371/journal.pone.0029629.g001 reduced the association of PilY1 with aVb5 integrin ( Figure 2D). Importantly, these mutations did not alter the global secondary structure or the melting temperature of purified 532-1163 PilY1 ( Figure 2E, Table S2). Taken together, these data indicate that P. aeruginosa PilY1 binds to integrins in vitro in an RGD-dependent manner.

PilY1 has a second calcium binding site
A second potential calcium binding site was noted in the sequence of P. aeruginosa PilY1 between residues 600 and 608, which is in close proximity to the RGD motif (619-621) described above ( Figure 1A). Recall that a calcium binding site at residues 851-859 in PilY1 was shown in previous work to be critical to tfp biogenesis and twitching motility [23]. The sequences of the two calcium binding sites are closely related ( Figure 3A), and the putative 600-608 site is conserved in PilY1s of known sequence ( Figure 1B). Indeed, in the P. aeruginosa PilY1 sequences examined, this second potential calcium binding site was always found 8-10 residues upstream of the conserved RGD motif ( Figure 1B). To test the 600-608 site's ability to bind calcium in purified 532-1163 PilY1, we eliminated the previously published calcium binding site using either a D859A or D859K mutation [23], and measured calcium binding. The D859A and D859K forms of 532-1163 PilY1 exhibited K d 's for calcium binding of 412 and 266 nM, respectively ( Figure 3B). Thus, the 600-608 site appeared to bind calcium. We next created corresponding mutations (D608A, D608K) in the 532-1163 PilY1 construct, and found that these variants bound calcium with 2.3-2.4 mM affinity, similar to that reported previously for the 851-859 PilY1 site ( Figure 3C) [23]. Finally, we created a D608A/D859A double-mutant form of 532-1163 PilY1 and compared its calcium binding to wild-type 532-1163 PilY1. We found that the double-mutant exhibited only nonspecific calcium binding, while wild-type 532-1163 PilY1 associated with calcium with a 700 nM affinity ( Figure 3D). The variant forms of 532-1163 PilY1 examined in vitro in these studies maintained an overall wild-type fold and melting temperature, indicating that these changes did not alter the stability of the proteins (Figure S1A-E; Table S2). Additionally, mutations in the RGD sequence adjacent to the 600-608 site did not alter calcium binding by these purified forms of 532-1163 PilY1 ( Figure S2A). The 532-1163 PilY1 construct demonstrated 1.3 mM and 600 nM affinity binding to Zn and Mn, respectively, and non-specific binding to Mg ( Figure S3A-D). However, due to PilY1's structural similarity to established calcium binding sites, we consider calcium to be the primary metal bound to PilY1 [23,28,29,30]. Taken together, these observations support the conclusion that purified 532-1163 PilY1 contains two calcium binding sites, the previously characterized one at 851-859 and a newly identified one at 600-608 that is ten residues N-terminal to PilY1's RGD motif.

PilY1 binds integrin in a calcium dependent manner
Because the 600-608 calcium binding site in PilY1 is close in sequence to the RGD motif, we next sought to examine whether calcium binding impacted the association between 532-1163 PilY1 and integrin. We found that D859A, D859K, and D608K mutant forms of 532-1163 PilY1 exhibit wild-type or only slightly reduced binding to aVb5; in contrast, the D608A mutant of purified PilY1 was significantly inhibited in its binding to this integrin ( Figure 4A). Thus, eliminating the charged D608 side chain proximal to the RGD motif impacted PilY1's interaction with aVb5. We then examined double mutant forms of 532-1163 PilY1 and found that the two calcium binding sites interacted functionally with respect to integrin binding ( Figure 4B). For example, the D608A/D859A variant of 532-1163 PilY1 shows nearly wild-type levels of aVb5 binding ( Figure 4B); recall that the D608A form of PilY1 exhibited little integrin binding in vitro ( Figure 4A). Thus, D859A appears to act as a dominant positive over the D608A mutation. Similarly, the D608K/D859A PilY1 shows increased integrin binding relatively to wild-type ( Figure 4B). In contrast, adding the D859K mutation to either D608A or D608K in PilY1 had no impact on integrin binding relative to the 608 mutants alone ( Figure 4B). Taken together, these data suggest that both calcium binding sites in PilY1 impact integrin binding in vitro. Note that the single-and double-mutants considered here did not bind calcium ( Figure S2B), and exhibited wild-type CD spectra and melting temperatures ( Figure S1A-E, Table S2), indicating that the overall structure of these purified proteins was not altered.
The crystal structure of the C-terminal domain of PilY1 does not contain the 600-621 region; as such, it is not known whether the two calcium binding sites are in close structural proximity. However, it is possible that they interact physically, and that the functional impact of one upon the other is via the swapping of an aspartic acid from one site (e.g., D851 or D855) into the other site (600-608). To test this hypothesis, we created two triple mutant forms of 532-1163 PilY1, D608A/D851A/D859A and D608A/ D855A/D859A, and one quadruple mutant form, D608A/ D851A/D855A/D859A. Both triple mutants and the quadruple mutant show wild type levels of binding ( Figure S4). Thus, we conclude that an interaction between the two calcium binding sites in PilY1 is not primarily associated with aspartic acids in one site helping to coordinate calcium in the other site. However, taken together, these data indicated that a functional interaction between the two calcium binding sites in PilY1 impacts the ability of the RGD motif in this protein to bind to integrins in vitro.

Discussion
P. aeruginosa is an established and increasingly antibiotic resistant pathogen that predominantly infects patients with compromised defense mechanisms. Toward this end, we sought to understand how P. aeruginosa binds to host cells, a necessary first step in infection. We focused on the type IV pilus-associated protein PilY1, which is involved in bacterial motility and is related in sequence to factors in other bacteria required for host cell binding. We show that PilY1 contains an integrin binding RGD motif that mediates contact between purified PilY1 and integrin proteins in vitro. Additionally, we show that PilY1 contains two calcium binding sites and that each site impacts the protein's ability to associate with integrin ( Figure 4C). Both the calcium binding sites and the RGD motif are conserved in all PilY1 sequences from P. aeruginosa strains reported to date ( Figure 1B).
It was not anticipated that the PilY1 calcium binding sites, particularly the one distal (851-859) to the RGD sequence, would impact integrin binding. In fact, D859A can be seen as a dominant positive with respect to integrin binding ( Figure 4C). For example, while the D608A mutant form of PilY1 is significantly reduced in integrin binding, combining it with D859A to create a D608A/ D859A double-mutant produces a protein that is capable of binding to integrin ( Figure 4B). Further, a D608K/D859A doublemutant exhibits higher integrin binding than either single mutant or the wild-type protein ( Figure 4B). In spite of these in vitro data, however, the molecular contacts involved in these effects are not clear, in part because the region between residues 600 and 643 in PilY1 has not been elucidated structurally. It is tempting to speculate that a physical association exists between the two calcium binding sites that impacts RGD presentation for integrin binding. Such speculation is supported by the data outlined here, but remains speculation until future studies to examine the structure of this region of PilY1 are complete. Integrins and integrin-like proteins are present in plants, insects, and the animal kingdom. They are up-regulated during cellular stress and recovery [31,32,33]; indeed, the presence of P. aeruginosa itself up-regulates the expression levels of the aV integrin subunit [16]. In addition, integrins are established targets for bacterial pathogens. Bordetella pertussis protein pertactin, containing two RGD motifs, has been shown to adhere to Chinese hamster ovary cells in an RGD-dependent manner [34,35]. Pertactin is a member of the auto-transporter family, many of which contain RGD motifs mediating adherence to integrins [36]. However, there are conflicting data on whether auto-transporter RGDs are relevant in mouse models of infection [37]. Still, other in vitro and in vivo studies have shown that the following bacteria bind integrin in an RGD-dependent manner: Pyrenophora tritici-repentis (a wheat pathogen), Mycoplasma conjunctivae (sheep pathogen), and the mammalian pathogens Klebsiella pneumoniae and Helicobacter pylori [38,39,40]. Integrins are also known targets for viral pathogens. Proteins in hepatitis C virus, coxsackievirus A9, human herpes virus 8, Epstein-Barr virus, and adenovirus bind to host cells in an RGD dependent manner [41,42,43,44,45]. In summary, integrins have the potential to be important target proteins for PilY1mediated attachment of P. aeruginosa.
Interestingly, calcium has also been shown to play a vital role in host RGD substrate:integrin binding events ( Figure 5). Calcium binding helps in the folding of the blood clotting factor fibrinogen, and it protects against cleavage and degradation of the hemostasis protein von Wildenbrand factor and the elastin scaffold protein fibrillin ( Figure 5) [46,47,48,49]. Some integrin substrates, such as thrombospondin, use cooperative calcium binding at multiple sites to facilitate integrin binding, while calcium increases adhesion in tenascin, vitronectin, and osteopontin to target integrins ( Figure 5) [50,51,52,53,54]. Thus, there is precedent for calcium regulation of RGD-mediated contacts with integrins.
PilY1 is an established pilus biogenesis factor [23] and has been implicated in P. aeruginosa adhesion to host cells by functional studies [25] and due to its homology to the PilCs adhesin proteins of Neisseria [55]. Here, we provide evidence that PilY1 serves as an adhesin that binds to integrins in an RGD-dependent manner. Furthermore, we show that integrin binding is regulated by two distinct calcium binding sites in PilY1. These data provide an in vitro advance in our understanding of PilY1 structure and function. It will be of interest to determine in future studies whether PilY1 and integrins are involved in P. aeruginosa attachment to host cells in vivo.

PilY1 constructs and protein purification
Site-directed mutagenesis was performed to produce D608A, D608K, D621A, D619-621, D859A, D859K, D608A/D859A, D608A/D859K, D608K/D859A, D608K/D859K, D608A/ D851A/D859A, D608A/D855A/D859A, and D608A/D851A/ D855A/D859A mutants in a pDONR vector containing PilY1. Amino acids 532-1163 were cloned out of the pDONR vector for entry into pMCSG7 for protein expression. Vectors were transformed into BL21 Gold (Stratagene) on ampicillin plates overnight and a single colony was used to inoculate a 100 mL LB flask overnight containing 50 mM/mL ampicillin. Cell cultures were centrifuged at 30006g and the supernatant was discarded. The resultant pellet was used to inoculate a 1.5 L shaker flask of Terrific Broth with 50 mL of antifoam (Sigma-Aldrich) and 50 mM/mL ampicillin. Cells were grown at 37uC until OD 600 reached 0.6-0.8. Temperature was reduced to 18uC and protein expression was induced with 0.5 mM IPTG. Cells were grown overnight and harvested by centrifugation 60006g at for 15 minutes at 4uC, and pellets were stored at 280uC.
Cells pellets were thawed using buffer consisting of 25 mM HEPES pH 7.5, 150 mM NaCl, 10 mM imidazole, 5% glycerol, DNase and protease inhibitor tablets (Roche). Cells were sonicated and cell lysate was separated into soluble and insoluble fractions using high-speed centrifugation. The soluble fraction was filtered then nickel purified, buffer exchanged, and separated using an S200 gel filtration column on an Ä KTAxpress TM (GE Health-Care). If necessary, protein and storage buffers were chelated by Chelex-100 to remove bound calcium (Bio-Rad Laboratories). Purified proteins were concentrated to ,100 mM, frozen, and stored at 280uC.

Integrin binding assay
Costar EIA/RIA stripwell high binding plates were coated with either 2 mg of purified aVb5 integrin or 2 mg of purified aVb3 integrin (Millipore) for the experimental wells, or the molar equivalent of BSA for the control wells, at 100 mL in 100 mM sodium bicarbonate buffer pH 9.6 (Abcam) for 16 hours at 4uC. Plates were washed 16with 300 mL of PBS. Wells were then blocked with 1% BSA in PBS for one hour at room temperature. Plates were washed 16with PBS and 100 mg/mL of PilY1 (Wild Type,  D851A/D859A, D608A/D855A/D859A, or D608A/D851A/ D855A/D859A) in binding buffer (50 mM Tris pH 7.8, 100 mM NaCl 2 mM CaCl 2 , 1 mM MgCl 2 , 1 mM MnCl 2 , and 0.1% Triton) was added to each well including the BSA control wells for 6 hours at 16uC with or without RGDS peptide (inhibitor), GRADSP peptide (control), or no peptide. Metals are needed in the binding buffer for integrin structural integrity. Plates were then washed 36with PBST (0.1% Tween-20) and 36 with PBS before adding rabbit aPilY1 polyclonal antibody 1:500 per well in 1% BSA in PBS overnight for 16 hours at 4uC. Plates were then washed 3 times with PBST and 36 with PBS before adding the Alexa fluor 488 f(ab9) 2 fragment goat anti-rabbit (Invitrogen) 1:5000 in 1% BSA in PBS protected from light at room temperature for 1 hour. Plates were then washed 36 with PBST, 36 with PBS, then dried. Plates were read on a PHERAstar (BMG LabTech). The experimental and control well averages and standard error of the mean (SEM) were calculated, and then the control average was subtracted from the experimental values and the errors were compounded using the equation !((experimental SEM) 2 +(control SEM)) 2 .

Calcium binding assay
A binding curve for Oregon GreenH 488 BAPTA-5N, hexapotassium salt (Invitrogen) in 25 mM HEPES pH 7.5, 250 mM NaCl, 10 mM imidazole, 5% glycerol was measured on a PHERAstar (BMGLabtech) at 488 nm. 20 mM Oregon Green, either 2 mM CaCl 2 , 2 mM MnCl 2 , or 2 mM ZnCl 2 , and PilY1 constructs were serial diluted 1.5-3 fold from ,100 mM to ,1 nM to obtain an EC50 which was then used to calculate the respective PilY1 K d for calcium as described previously or PilY1 concentration was held constant and MgCl 2 concentration was titrated to determine if PilY1 had an affinity to magnesium [23].

Circular dichroism and thermal denaturation
Protein samples were buffer exchanged into chelated 10 mM K x H x PO 4 50 mM NaF pH 7.7 buffer and brought to 5 mM with or without the addition of CaCl 2 . A wavelength scan from 200-260 l was performed on a Circular Dichroism Spectrometer 62 DS (Aviv) at 16uC with a 10 second averaging time. Melting temperatures were measures at l 214 from 3uC to 95uC at one degree increments with a 10 second averaging time. Table S1 P. aeruginosa PilY1 is homologous to PilC family of bacterial adhesin and pilus biogenesis proteins. PilY1 was individually aligned to K. Kingae (Kk) PilC1 and PilC2, N. gonorrhea (Ng) PilC1 and PilC2, and N. meningitidis (Nm) PilC1 and PilC2 using http://blast.ncbi.nlm.nih.gov/. C-terminal homology indicates residues that are similar (e.g. leucine and isoleucine) while c-terminal identity corresponds to identical residues. Strains used for each protein are as follows: ACF19883 for Kingella kingae PilC1, ACF19886.1 for Kingella kingae PilC2, O05924_NEIME for Neisseria meningitidis PilC1 O05925_NEIME for Neisseria meningitidis PilC2, O05923_NEIGO for Neisseria gonorrhea PilC1, and Q51019_NEIGO for Neisseria gonorrhea PilC2. (TIFF)