Paxillin phosphorylation: bifurcation point downstream of integrin-linked kinase (ILK) in streptococcal invasion

Authors

  • Beinan Wang,

    Corresponding author
    1. Department of Microbiology, Medical School, University of Minnesota, 1460 Mayo Memorial Building, MMC 196, 420 Delaware Street SE. Minneapolis, MN 55455, USA.
      *E-mail beinan@mail.ahc.umn.edu; Tel. (+1) 612 625 0989; Fax (+1) 612 626 0623.
    Search for more papers by this author
  • Shaoying Li,

    1. Department of Microbiology, Medical School, University of Minnesota, 1460 Mayo Memorial Building, MMC 196, 420 Delaware Street SE. Minneapolis, MN 55455, USA.
    Search for more papers by this author
  • Shoukat Dedhar,

    1. BC Cancer Research Center and Department of Biochemistry and Molecular Biology, University of British Columbia, Vancouver, BC, V5Z1L3, Canada.
    Search for more papers by this author
  • P. Patrick Cleary

    1. Department of Microbiology, Medical School, University of Minnesota, 1460 Mayo Memorial Building, MMC 196, 420 Delaware Street SE. Minneapolis, MN 55455, USA.
    Search for more papers by this author

*E-mail beinan@mail.ahc.umn.edu; Tel. (+1) 612 625 0989; Fax (+1) 612 626 0623.

Summary

Efficient group A streptococcus (GAS) invasion of mammalian cells requires fibronectin (Fn) binding proteins, such as M1 and PrtF1/SfbI, that bridge bacteria to integrins and activate cellular signalling for ingestion. Previous studies of GAS invasion, mediated by both proteins, suggest a common signalling pathway. However, distinct cellular morphological changes at the port of bacterial entry suggest that different signals are also induced. Here we report that paxillin is phosphorylated in response to Fn-bound GAS that express either M1 or PrtF1/SfbI protein, but is not phosphorylated in response to a mutant deficient in both proteins. Inhibition of paxillin phosphorylation by a tyrosine kinase inhibitor, PP2, or by expression of a dominant negative form of paxillin significantly reduced invasion by M1+ but did not affect ingestion of PrtF1/SfbI+ strains. In contrast, another tyrosine inhibitor, genistein, did not significantly prevent paxillin phosphorylation and had no effect on ingestion of the M1+ strain, but reduced PrtF1/SfbI-mediated entry. This suggests that paxillin phosphorylation is induced by both proteins but only required for M1-mediated invasion. A bifurcation point, downstream of integrin-linked kinase (ILK) and phosphoinositide 3-kinase, likely accounts for the distinct morphological changes. Furthermore, ILK activity is indispensable for M1-induced paxillin recruitment and phosphorylation.

Introduction

Microbial pathogens are able to bind integrins of mammalian cells through a fibronectin (Fn) or other extracellular matrix (ECM) proteins and commandeer integrins as a receptor for their entry into cells (Joh et al., 1999). Many pathogenic bacteria produce Fn binding proteins (FnBPs) and some pathogens express multiple FnBPs, underscoring importance of FnBPs in establishment of bacterial infection. High-frequency internalization of group A streptococcus (GAS) strain 90-226 by epithelial cells is dependent upon the FnBP, M1 protein, Fn and α5β1 integrins (Cue et al., 1998; Dombek et al., 1999). Recently, we reported that components of the integrin signalling pathway, phosphoinositide 3-kinase (PI3K) (Purushothaman et al., 2003) and integrin-linked kinase (ILK) (Wang et al., 2006a), are required for invasion of mammalian cells by GAS strains that express different FnBPs, and also by other unrelated bacterial pathogens. Thus, integrin signalling is a common and critical cellular mechanism used by bacteria for their entry.

Integrin signalling is implicated in a variety of cellular functions, including cell adhesion, spreading, migration, cell survival and proliferation. It is initiated via engagement of integrins with ECM ligands such as Fn, laminin or collagen, and can lead to formation of focal adhesions, a structural and signal link between the ECM and actin cytoskeleton (Brakebusch and Fassler, 2003). Paxillin is an adaptor in the integrin pathway, providing multiple docking sites for signalling molecules and actin binding proteins (Turner, 2000a). The carboxyl terminus of ILK binds to cytoplasmic tails of β integrins and contains a paxillin binding subdomain that interacts directly with amino termini of paxillin (Nikolopoulos and Turner, 2001). This brings engaged integrins and paxillin into close proximity, allowing transfer of signals from integrins to paxillin. Phosphorylation of paxillin by tyrosine kinases such as Src and focal adhesion kinase (FAK) creates additional docking sites for cytoskeleton assembly and regulates the dynamics of focal adhesions, which are associated with cell morphological changes (Vindis et al., 2004), including zipper-like uptake of bacteria (Joh et al., 1999). Paxillin is phosphorylated in response to infection by a number of invasive pathogens, such as E. coli K1, Porphyromonas gingivalis, Campylobacter jejuni and Shigella flexneri (Watarai et al., 1996; Reddy et al., 2000; Yilmaz et al., 2002; Monteville et al., 2003). Ozeri et al. reported that GAS strain JRS4, which expresses PrtF1/SfbI, induces paxillin phosphorylation (Ozeri et al., 2001). However, whether paxillin phosphorylation is required for streptococcal invasion and how ILK is related to paxillin in bacteria-induced integrin signalling have not been characterized. Rohde et al. recognized that host cell caveolae are an entry-port for ingestion of PrtF1/SfbI+ strains (Rohde et al., 2003), a mechanism different from the zipper-like uptake mechanism employed by M1+ PrtF1/SfbI streptococcus (Dombek et al., 1999). These observations promoted us to postulate that strains that express different FnBPs at some point must induce different signals. Here, we report that Fn bound to M1 protein on the GAS surface induced paxillin phosphorylation. Inhibition of paxillin phosphorylation significantly reduced ingestion of M1+ streptococcus by HEp-2 cells, indicating that paxillin phosphorylation is required for efficient entry by this strain. In addition, the kinase activity of ILK is required for M1-mediated paxillin phosphorylation and recruitment to streptococcal focal adhesions, demonstrating a dual function of ILK in paxillin docking and signalling. In contrast, entry mediated by PrtF1/SfbI protein was not dependent on paxillin phosphorylation, although paxillin phosphorylation was induced by infection with PrtF1/SfbI+ streptococci. Therefore, a branch in this signalling pathway downstream of PI3K/ILK may account for the distinct cell membrane morphologies triggered by two GAS FnBPs.

Results

Chemical and genetic inhibition of paxillin phosphorylation reduced M1 protein-mediated GAS invasion

The role of paxillin in GAS entry was investigated to dissect the signalling events downstream of PI3K/ILK. Invasion assays were performed after epithelial cells were treated with a tyrosine kinase inhibitor, PP2, which is able to prevent phosphorylation of paxillin and FAK (Montiel et al., 2005) and also shown in Fig. 3A (line 4). Internalized M1+ 90-226 were significantly reduced in HEp-2 cells pretreated with PP2 (Fig. 1A). A recombinant Lactococcus lactis M1+ strain was used to determine whether only the streptococcal M1 protein is required for phospho-paxillin-dependent invasion. Expression of full-length M1 protein by L. lactis (pLM1) was previously shown to confer the ability to invade epithelial cells efficiently (Cue et al., 2001). Similar to the effect on M1+ 90-226 entry, PP2 decreased ingestion of M1+L. lactis, indicating that paxillin phosphorylation is implicated in M1-mediated signalling for invasion by these bacteria. As PP2 may also inhibit other tyrosine kinases, reduction of bacterial entry by impaired paxillin phosphorylation was verified by transient transfection of epithelial cells with a mutant gene of paxillin in which tyrosines at positions 31 and 118 are replaced with phenylalanine (Y31/118F). Previous studies showed that cells that express this mutant have retardant migration, a paxillin-dependent response (Petit et al., 2000). To reduce the background of paxillin phosphorylation caused by cell adherence to culture plates, invasion assays were performed on suspensions of transfected cells. As shown in Fig. 1B, significantly less efficient entry of M1+ 90-226 and M1+L. lactis was observed in cells transfected with the mutant gene than with a vector alone, confirming that phosphorylation of paxillin is required for M1-mediated entry.

Figure 3.

Paxillin phosphorylation and assembly induced by M1+ streptococci is prevented by inhibition of ILK kinase activity. HeLa cells were pretreated with inhibitors: KP-392 (KP) 50 μM; wortmannin (WM) 200 nM; PP2 10 μM; or DMSO overnight, plated on poly-l-lysine-coated plates for 2 h and infected with Fn-coated 90-226 M1+. Cell cytoskeleton was fractionated from cell lysate and examined by Western blots with (A) anti-phospho-paxillin and anti-actin; (B) anti-total paxillin and anti-actin (lanes 1–3); and anti-phospho-paxillin and anti-actin (lanes 4–6).

Figure 1.

Chemical and genetic inhibition of paxillin phosphorylation reduced M1 protein-mediated invasion.
A. HEp-2 cells were pretreated with PP2, or DMSO as a solvent control overnight. Invasion assays were performed with GAS strain 90-226 M+ or L. lactis M1+. Intracellular cfu from DMSO control cells was considered as 100%. In DMSO controls, 5.3 × 104 cfu of 90-226 M+ (5.3% of inoculum) and 8.8 × 104 cfu of L. lactis M1+ (8.1% of inoculum) were intracellular.
B. HeLa cells were transiently transfected with a dominant negative form of paxillin (Y31/118F) and the same amount of control vector DNA. Invasion assays were performed on suspended HeLa cells 48 h post transfection. Intracellular cfu from vector control cells was considered as 100%. In vector controls, 2.5 × 104 cfu of 90-226 M+ (11.4% of inoculum) and 4.3 × 104 cfu of L. lactis M1+ (5.1% of inoculum) were intracellular. Data are the means ± standard errors of three independent determinants.

Paxillin phosphorylation is induced by M1+ streptococcus and lactococcus

Paxillin phosphorylation is induced by a PrtF1/SfbI+ GAS strain (Ozeri et al., 2001). However, bacterial molecules responsible for paxillin phosphorylation were not defined. Therefore, effects of 90-226 M1+ on phosphorylation of paxillin were studied. Epithelial cells were grown on poly-l-lysine-coated plates to reduce background levels of focal adhesion components and then infected with Fn-coated streptococci. Infected cells were harvested at various times and cell lysates were analysed by Western Blot. While total amounts of paxillin protein remained unchanged during the infection time, phospho-paxillin was increased in cells infected with M1+ 90-226, which first became apparent at 30 min, peaked at 1–2 h and diminished thereafter by 4 h (Fig. 2A). In contrast, phospho-paxillin remained at basal levels throughout the time when cells were exposed to and associated with significant numbers of an isogenic M1 deletion mutant (ΔM1) (Cue et al., 1998). Similar to paxillin, phosphorylation of FAK was also induced by this strain, but with a peak at least 2 h later than phospho-paxillin. No concomitant changes of total FAK protein were observed. Strain 90-226 is poorly ingested by epithelial cells without bound Fn (Cue et al., 2001). To determine whether phosphory- lation of paxillin requires bound Fn, M1+ 90-226 were pre-incubated with or without Fn, and then incubated with epithelial cells for 2 h. As shown in Fig. 2B, without bound Fn, slightly increased levels of phospho-paxillin and phospho-FAK were observed in cells infected with M1+ 90-226 (lane 2) but not with uninfected cells (lane 1) or those infected with the M1 mutant (lane 3). Much higher levels of phospho-paxillin and phospho-FAK were detected when cells were infected with Fn-treated M1+ 90-226 (lane 4) than those infected with this strain without Fn treatment (lane 2). The Fn-treated M1 mutant (lane 5) also induced slightly higher amounts of the two phospho-proteins than without Fn (lane 3). These results suggest that M1 protein itself does not trigger enough paxillin and FAK signalling for efficient bacterial entry unless Fn is associated. Other GAS surface proteins with low affinity for Fn could be responsible for the slight increase of phospho-paxillin and FAK by Fn-treated M1 streptococci. The Fn requirement for optimal phosphorylation of paxillin and FAK indicates that ability of M1 protein binding of Fn is critical.

Figure 2.

Paxillin phosphorylation is induced by M1+ streptococcus and lactococcus. HeLa cells, grown on poly-l-lysine-coated plates in serum-free DMEM, were infected with Fn-treated bacteria. Cells were lysed after washing with PBS at indicated times, and total cell lysates were examined by Western blots with indicated antibodies.
A. Blots were stained with indicated antibodies. Equal amounts of loaded protein were ensured by staining anti-actin. Numbers on top of the panel represent intensity of the bands stained with anti-phospho-paxillin and measured by densitometry.
B. Cells were infected with serum-free DMEM (lane 1), bacteria treated with Fn (lanes 4 and 5) and without Fn (lanes 2 and 3).
C. Cells were infected with Fn-treated L. lactis M1+ or L. lactis M1.

The requirement for M1 protein in paxillin phosphorylation was confirmed by testing whether M1+ lactococcus would induce paxillin phosphorylation. Western blot showed more phospho-paxillin in cells infected with M1+L. lactis (Fig. 2C) than cells infected with parental strain M1L. lactis or cells treated with medium alone. This indicates that M1 protein interaction with Fn is sufficient to induce paxillin phosphorylation. It should be noted that less phospho-paxillin was associated with cells treated with the M1L. lactis, suggesting a possible negative effect of L. lactis on paxillin phosphorylation.

M1+ 90-226 streptococci-induced paxillin phosphorylation and recruitment into focal adhesions is prevented by inhibition of ILK kinase activity

Integrin-linked kinase binds to cytoplasmic tails of integrins and can also interact directly with paxillin (Nikolopoulos and Turner, 2001). This linkage permits transfer of signals from integrins to paxillin and downstream. Activation of ILK occurs in the presence of PI3K activity (Qian et al., 2005). We previously reported that inhibitors of ILK (KP-392) or PI3K (wortmannin) significantly prevent GAS invasion (Wang et al., 2006a). To understand the functional connection between ILK and paxillin in GAS invasion, the requirement for ILK kinase activity on paxillin phosphorylation was studied. We first assessed the effect of chemical inhibitors of ILK and PI3K on phosphorylation of paxillin in GAS-infected HEp-2 cells. As shown in Fig. 3A, KP-392 and wortmannin prevented phosphorylation of paxillin at Y118 induced by M1+ 90-226 streptococci, indicating that paxillin phosphorylation is ILK and PI3K dependent. As Y118 is known to be phosphorylated by tyrosine kinases such as Src and FAK, but not by ILK, a serine/threonine kinase, the impact of ILK on paxillin phosphorylation appears to be indirect. ILK kinase activity is known to be required for paxillin assembly into focal adhesion complexes (Attwell et al., 2003). Therefore we postulated that paxillin is recruited to streptococcal focal adhesions in a PI3K/ILK-dependent manner. To test this possibility, epithelial cells were pretreated with the ILK (KP) or PI3K inhibitor (WM), plated on poly-l-lysine-coated plates, and then infected with Fn-treated M1+ 90-226. Cytoskeleton fractions were prepared and paxillin was examined by Western blot. As shown in Fig. 3B, cells infected with M1+ 90-226 retained both more total (lane 2) and phospho-paxillin (lane 5) in the cytoskeleton fraction than non-infected cells (lane 1 and 4). The increased levels were diminished when cells were pretreated with KP-392 (lanes 3 and 6). Taken together, these experiments suggest that ILK kinase activity is required for paxillin recruitment to streptococcal focal adhesions and is also required for paxillin phosphorylation.

PrtF1/SfbI-mediated GAS entry is not sensitive to treatment with PP2 or a dominant negative form of paxillin

M1 and PrtF1/SfbI proteins are genetically unrelated, and their expression is controlled by different environmental signals. Although they both bind Fn the interaction induces different mechanisms of internalization. PrtF1/SfbI+ streptococci are ingested by caveolar endocytosis, characterized by formation of large invaginations in the cytoplasmic membrane (Rohde et al., 2003); whereas, strains that express only M protein induce elongated microvilli and microvilli-fused pseudopod-like structures that surround the bacteria, and are ingested by a zipper-like mechanism (Dombek et al., 1999). These differences in cellular morphology led us to postulate that intracellular signals downstream of PI3K/ILK may be divergent. Recombinant strain 90-226 M1, PrtF1/SfbI+ (Wang et al., 2006a) was used to analyse a possible divergent role of paxillin phosphorylation in PrtF1/SfbI-mediated entry. Invasion assays showed that after treatment of cells with PP2, invasion mediated by M1+ 90-226 was reduced by 80–90% in a dose-dependent manner; whereas, that mediated by PrtF1/SfbI was not significantly impaired (Fig. 4A). To confirm that ingestion of strain 90-226 M1, PrtF1/SfbI+ is not dependent on paxillin phosphorylation, HeLa cells were transfected with the dominant negative form of paxillin (Y31/118F) and infected with the recombinant 90-226 (M1, PrtF1/SfbI+) and a PrtF1/SfbI+ M6 type strain JRS4. In contrast to M1+ 90-226, entry of strain 90-226 (M1 PrtF1/SfbI+) or JRS4 was not prevented by expression of the mutant paxillin (Fig. 4B).

Figure 4.

PrtF1/SfbI-mediated GAS entry is not sensitive to treatment with PP2 or a dominant negative form of paxillin.
A. HEp-2 cells were treated as described in Fig. 1A. Invasion assays were performed with 90-226 M1+ PrtF1/SfbI or 90-226 M1 PrtF1/SfbI+. In DMSO controls, 9.1 × 104 cfu of 90-226 M1+ PrtF1/SfbI (9.1% of inoculum) and 9.7 × 105 cfu of 90-226 M1 PrtF1/SfbI+ (29.1% of inoculum) were intracellular.
B. HeLa cells were transiently transfected as described in Fig. 1B. Invasion assays were performed on suspended HeLa cells 48 h post transfection. In vector controls, 6.3 × 104 cfu of JRS4 (13.9% of inoculum) and 8.9 × 104 cfu of 90-226 M1 PrtF1/SfbI+ (25.9% of inoculum) were intracellular. Data are the means ± standard errors of three independent determinants.

Differential effects of tyrosine kinase inhibitors on paxillin phosphorylation induced by streptococci

Ozeri et al. reported that ingestion of PrtF1/SfbI+ strains is inhibited by genistein, another tyrosine kinase inhibitor (Ozeri et al., 2001); whereas, we reported that ingestion of M1+ 90-226 is not affected by this inhibitor at the same concentration (Purushothaman et al., 2003). To confirm that this difference is not caused by different backgrounds of the strains, invasion assays were performed with 90-226 strains that express either M1 or PrtF1/SfbI protein. Genistein prevented ingestion of M1 PrtF1/SfbI+ 90-226 in a dose-dependent manner but did not affect ingestion of the M1+ PrtF1/SfbI 90-226 strain (Fig. 5A). These results indicate that a genistein-sensitive tyrosine kinase is required for PrtF1/SfbI-mediated invasion, which is not required for M1+ 90-226 strain or can be bypassed. If PrtF1/SfbI-induced paxillin phosphorylation is required then genistein should block this reaction; therefore, effects of genistein on paxillin phosphorylation was assessed and compared with other chemical inhibitors. HeLa cells were pretreated with inhibitors and then infected with M1 PrtF1/SfbI+ 90-226. Cell lysates were fractionated and examined by Western blots. Consistent with the response to M1+ streptococci, phosphorylation of paxillin was induced by M1 PrtF1/SfbI+ 90-226 in both soluble (Fig. 5B, lane 3) and cytoskeleton fractions (Fig. 5C, lane 3). Unlike M1+ PrtF1/SfbI 90-226, significantly increased levels of phospho-paxillin were induced by this strain with or without Fn in the cytoplasmic fractions (Fig. 5B, lanes 2 and 3). Induction of paxillin phosphorylation was prevented when cells were pretreated with KP-392 and wortmannin (Fig. 5B, lanes 4 and 5), indicating again that ILK and PI3K are required for paxillin phosphorylation induced by Fn-bound PrtF1/SfbI protein. Though PP2 did not significantly alter invasion by this strain it did block paxillin phosphorylation. In contrast to those inhibitors, paxillin phosphorylation was not significantly inhibited by genistein, particularly in the cytoskeleton fraction (Fig. 5C, lane 7). Therefore, phosphorylation of paxillin is not required for efficient ingestion of PrtF1/SfbI+ streptococci. Instead an unknown genistein-sensitive target is required.

Figure 5.

Differential effects of tyrosine kinase inhibitors on paxillin phosphorylation induced by streptococci.
A. HEp-2 cells were treated as described in Fig. 1A. Invasion assays were performed with strain JRS4 M6+ PrtF1/SfbI+ or 90-226 M1 PrtF1/SfbI+. In DMSO controls, 1.8 × 104 cfu of 90-226 M1+ PrtF1/SfbI (3.6% of inoculum) and 1.4 × 106 cfu of 90-226 M1 PrtF1/SfbI+ (36.4% of inoculum) were intracellular. Data are the means ± standard errors of three independent determinants.
B and C. HeLa cells were pretreated with inhibitors: KP-392 (KP) 50 μM; wortmannin (WM) 200 nM; PP2 10 μM; genistein (Gen) 100 μM; or DMSO overnight, plated on poly-l-lysine-coated plates for 2 h and infected with Fn-coated 90-226 M1 PrtF1/SfbI+. (B) Soluble and (C) cytoskeleton fractions were prepared from cell lysates and examined by Western blots with anti-phospho-paxillin and anti-actin. Lanes 1–3: DMEM alone, 90-226 M1 PrtF1/SfbI+ alone, Fn-treated 90-226 M1 PrtF1/SfbI+.

Discussion

Efficient invasion of host cells by Fn-dependent mechanisms provides streptococci an intracellular niche, which may protect them from host defences and antibiotics, and allow dissemination and persistence in their host. A requirement for α5β1 integrin engagement with Fn bound to M1 protein for efficient GAS entry was recognized by Cue et al. (1998). Other studies of downstream events revealed that the intracellular molecules, PI3K and ILK, are required for this efficient mechanism (Purushothaman et al., 2003; Wang et al., 2006a). In this work we demonstrated that paxillin phosphorylation is induced by M1+ 90-226, but not by an isogenic M1 strain, when bacteria were pretreated with Fn. Paxillin phosphorylation was also induced when epithelial cells were infected with a L. lactis strain that expresses M1 protein, confirming that M1 is the key bacterial inducer of paxillin phosphorylation. Investigation on a functional connection between ILK and paxillin showed that disruption of ILK kinase activity impaired paxillin phosphorylation and recruitment to streptococcal focal adhesions. Inhibition of paxillin phosphorylation decreased invasion by the M1+ strain; whereas, it did not impair invasion by PrtF1/SfbI+ strains. Experiments also showed that tyrosine kinase inhibitors, PP2 and genistein, have different effects on paxillin phosphorylation and on streptococcal entry mediated by M1 and PrtF1/SfbI proteins. Results indicate that phosphorylation of paxillin is required for M1, but not for PrtF1/SfbI-mediated entry.

Direct interaction of ILK with both the integrin and paxillin permits extracellular signals to be transmitted from integrins to paxillin. We showed that ILK activity is required for paxillin phosphorylation at Y118, which is one of two key sites phosphorylated by tyrosine kinases Src and FAK. ILK is a serine/threonine kinase and may phosphorylate serine/threonine sites of paxillin, but was never shown to phosphorylate tyrosine (Turner, 2000b). Therefore it is reasonable to speculate that ILK has an indirect role in Y118 phosphorylation. However, ILK-dependent paxillin phosphorylation in both cytoskeleton and soluble fractions excludes the possibility that paxillin localization in focal adhesions is a prerequisite for phosphorylation. These experiments are the first to reveal requiring participation of ILK for paxillin phosphorylation and underlines dependence of paxillin on ILK in M1-mediated entry.

Paxillin is a common adaptor shared by different types of transmembrane receptors and has multidocking sites for many specific binding partners. This endows paxillin with the capacity to be involved in highly complex cellular reactions (Brown and Turner, 2004), and therefore, it is a frequent target of phosphorylation in response to diverse cellular stimuli. Despite the fact that paxillin is phosphorylated in response to a number of invasive bacterial pathogens, it has not been demonstrated that the phosphorylated form is required for bacterial uptake. Our experiments and others showed that both M1+ and PrtF1/SfbI+ strains are able to induce paxillin phosphorylation. However, impairing paxillin phosphorylation significantly prevented GAS entry mediated by M1, but not by that mediated by PrtF1/SfbI. Paxillin was found underneath attached PrtF1/SfbI+ strain (Ozeri et al., 2001). Thus, it is likely that paxillin scaffolding, but not phosphorylation is required. Paxillin phosphorylation at Y31 and Y118 is critical for recruitment of actin binding proteins into focal adhesions, leading to actin assembly (Brown and Turner, 2004). The different dependence on Y118 phosphorylation suggests that cytoskeleton assembly is critical in invasion mediated by M1 but is not important in that mediated by PrtF1/SfbI. These signalling differences are consistent with the distinct morphology at the entry port in cell membranes induced by these two FnBPs: PrtF1/SfbI+ strains enter host cells by inducing caveolae-mediated endocytosis without evidence of actin polymerization; whereas, those that express M1 protein are internalized by zipper-like mechanism with significant actin assembly (Dombek et al., 1999; Molinari et al., 2000; Rohde et al., 2003).

Genistein and PP2 are known inhibitors of a wide range of tyrosine kinases. The different sensitivities of paxillin to PP2 and genistein suggest that they have different preferential tyrosine kinase targets; thus the prevailing signal triggered by either of the FnBPs is impaired preferentially by PP2 or genistein in each circumstance. Genistein is known to block caveolae formation by inhibition of Src kinase, which phosphorylates cavolin, a structural and signal protein of caveolae and a prime requirement for caveolae formation (Li et al., 1996; Rohde et al., 2003). Therefore, it is likely that genistein targets Src more effectively and impairs caveolae formation. Although inhibition of Src by genistein could reduce paxillin phosphorylation indirectly, paxillin could be phosphorylated by other genistein-resistant tyrosine kinases; therefore, permitting an actin assembly based zipper-like mechanism of uptake. PP2 is known to inhibit Src activity; therefore, significant reduction of M1-mediated entry by PP2 would impair both paxillin and Src phosphorylation. Interestingly, PP2 also blocks caveolin-1 phosphorylation at the concentration we used (Labrecque et al., 2003). Insensitivity to PP2 of PrtF1/SfbI-mediated entry suggests that a PP2-insensitive tyrosine kinase is activated in response to PrtF1/SfbI+ GAS infection.

M1 and PrtF1/SfbI have different affinities for Fn and integrin receptors (Cue et al., 2001; Towers et al., 2003). These variations may permit them to bind Fn differently and trigger different or additional mechanisms for intracellular survival. Moreover, these FnBPs are also expressed under different conditions; PrtF1/SfbI expression is enhanced in an O2-rich environment, while M1 expression is greater at higher CO2 partial pressure (Caparon et al., 1992; Gibson et al., 1995). The evolved diversity of FnBPs in GAS may allow the bacterium to trigger alternative signalling for successful survival at various locations in a host and may also be linked to the diverse manifestation of streptococcal diseases. Ingestion of Listeria monocytogenes can also proceed by alternative mechanisms. Entry of this bacterium is mediated by internalin A-E-cadherin and InlB-HGF-R (hepatocyte growth factor receptor) interactions. Interestingly, mechanisms of caveolae formation are shared, but implicated in different steps of the two types of internalization (Seveau et al., 2004).

A requirement for PI3K and ILK in PrtF1/SfbI-mediated invasion suggests that they are also involved in caveolae endocytosis, a mechanism recognized for a growing number of microbial pathogens for intracellular survival. The observation of common and divergent signalling pathways triggered by two different FnBPs is shown in our model of GAS signalling (Fig. 6). In this model, M1 and PrtF1/SfbI associate with integrins through a Fn bridge, which activates a signalling pathways that includes PI3K, ILK, paxillin and FAK. In the case of M1, paxillin is phosphorylated and assembled into streptococcal focal adhesion causing recruitment of additional paxillin binding partners. This promotes actin polymerization-based zipper-like bacterial uptake. On the other hand, PrtF1/SfbI binds integrins differently and diverts the PI3K/ILK-dependent signalling to caveolin-1 phosphorylation by Src (Li et al., 1996), which leads to caveolae endocytosis, a flask-like membrane invagination highly dependent on large amounts of raft lipids (Pelkmans, 2005). Future experiments will attempt to identify the switch which directs the invasion signal to caveolae rather than to actin cytoskeleton rearrangement.

Figure 6.

Schematic diagram of postulated signalling pathways that are mediated by M1 and PrtF1/SfbI proteins leading to group A streptococcal entry. M1 inline image or PrtF1/SfbI inline image on the surface of GAS binds to soluble Fn molecules inline image, which in turn bind to α5β1 integrin. This causes PI3K-dependent ILK activation, which subsequently promotes paxillin phosphorylation and assembly, and leads to actin cytoskeleton rearrangement for uptake of GAS by a zipper mechanism in M1-mediated entry. Because PrtF1/SfbI protein binds to different domains of Fn with greater affinity than M1 protein. Moreover, the reported interaction of Fn-bound PrtF1/SfbI with αVβ3 integrins induces different signals. Combination of all these differences leads to predominant caveolae formation through Src-activated caveolin.

Experimental procedures

Reagents

PP2, wortmannin and genistein were obtained from Calbiochem (La Jolla, CA). KP-392 was obtained from Kintec Pharmaceuticals, Vancouver, BC, Canada (Wang et al., 2006b). Poly-l-lysine was from Sigma (St Louis, MO). The following antibodies were used: rabbit anti-paxillin (Santa Cruz Biotechnology, Santa Cruz, CA), anti-phospho-paxillin (Tyr-118) (Biosource, Camarillo, CA), anti-FAK, phspho-FAK (Tyr 397) (UpstateLake Placid, NY) and anti-actin (Sigma, St Louis, MO).

Bacterial strains and culture conditions

GAS strain 90-226 is a serotype of M1 strain (Cue et al., 1998). The M1 deletion variant of strain 90-226 (ΔM1) carries a complete non-polar deletion of emm1, is sensitive to phagocytosis and produces no M1 antigen or peptide fragment (Wang et al., 2006a). GAS strain JRS4 (emm6.1) (PrtF/SfbI+ and M6+) was a gifts from Dr M. Caparon (Hanski and Caparon, 1992; Jadoun et al., 1997). These strains were maintained on sheep blood agar and grown in THB-Neo peptone medium at 37°C for assays. The PrtF1/SfbI-expressing variant (90-226 M1 PrtF1/SfbI+) of strain 90-226 was constructed by introducing plasmid pPTF8, which contains prtF1/SfbI gene (Hanski and Caparon, 1992), into strain 90-226 ΔM1. It was cultured in THB-Neo medium containing kanamycin 500 μg ml−1. M1+L. lactis (pLM1) is a derivative of strain VELL122 containing plasmid pLM1 which encodes full-length M1 protein. L. lactis (pP59) is the same strain, but harbours the vector only and is used as an M1 control (Cue et al., 2001). L. lactis VELL122 was cultured in M17 medium containing 0.5% dextrose and erythromycin (5 μg ml−1) at 30°C.

Cell culture

The HeLa (human cervix epithelial) and HEp-2 (human larynx epithelial) cell lines were cultured in Dulbecco's modified Eagle's medium (DMEM) at a humidified atmosphere of 5% CO2 and 95% air. The media contained 10% fetal bovine serum (Invitrogen), penicillin (100 units ml−1) and streptomycin (0.1 mg ml−1).

Invasion assay

Invasion assays were performed as previously described (Dombek et al., 1999; Wang et al., 2006b). Bacterial invasion per well is colony-forming units (cfu) that survived antibiotic treatment and expressed as percentage of the inoculum cfu. For invasion assays with chemical inhibitors, cells were treated with inhibitors in serum/antibiotic-free medium overnight before infected with bacteria. Treated cells were normalized against controls which were considered 100% invasion. For invasion of suspended cells, HeLa cells were first transfected with paxillin mutant (Y31//118F) cDNA as described bellow. Cells were then detached with 2.5 mM EDTA-PBS, washed and suspended in DMEM. Fn-treated bacteria were added to cell suspensions and incubated for 2 h with occasional shaking. Unbound bacteria were removed by washing with DMEM at low speed centrifugation (500 g). Extracellular bacteria were killed by incubation with penicillin/gentamicin-DMEM for another 2 h. Cells were washed and lysed for measurement of intracellular cfu. Viability of host cells and bacteria were checked after treatment with chemical inhibitors by Trypan blue and cfu respectively. No significant difference was observed between treated and control samples.

Transient transfection of cells with dominant negative form of paxillin DNA

HeLa cells were transiently transfected with paxillin mutant (Y31//118F) cDNA encoded on plasmid pcDNA3, a kind gift from Dr C.E. Turne (SUNY Upstate Medical University, Syracuse, NY) (Troussard et al., 2003). The negative control was plasmid pcDNA3 (Invitrogen) without an insert. Transfection with plasmid cDNAs was promoted with LipofactamineTM 2000 reagent (Invitrogen) according to the manufacturer's guidelines. Cells were harvested 48 h post transfection. Invasion assays were performed in cell suspension and unbound and extracellular bacteria were treated with antibiotics and washed by low speed centrifugation.

Fn-treated bacteria and cytoskeleton fraction

Stationary phase culture of bacteria were washed with PBS, resuspended in DMEM containing human Fn 300 μg ml−1 and incubated at room temperature 30 min. After washing with PBS two times, Fn-treated bacteria were added to cells grown on poly-l-lysine-coated plates in serum-free DMEM. Cells were washed with PBS after 2 h incubation and treated with lysis buffer containing 1% Triton X-100, PMSF (1 mM), leupeptin (1 mM), aprotinin (1 mM), NaF (2 mM) and NaVO3 (1 mM). Cell lysate was centrifuged and supernatant was collected as Triton-soluble fraction. The remaining cytoskeleton-based Triton-insoluble fraction was suspended in extraction buffer containing 20 mM Tris-HCL, 80 mM KCL, 30 mM MgCl2, 1 mM EGTA, 0.25 M NaCl, 1 mM DTT, 1 mM leupeptin and 0.5 mM PMSF (Attwell et al., 2003). The cytoskeletal fraction was vortexed and then sonicated before protein quantification and Western blot analysis.

Acknowledgements

We gratefully acknowledge Dr C.E. Turner for dominant negative form of paxillin plasmids. This work was supported by National Institutes of Health Grant RO1-AI34503 (to P.P. Cleary).

Ancillary