A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling

The p21-activated kinase (PAK) family regulate a multitude of cellular processes, including actin cytoskeleton remodelling. Numerous bacterial pathogens usurp host signalling pathways that regulate actin reorganisation in order to promote Infection. Salmonella and pathogenic Escherichia coli drive actin-dependent forced uptake and intimate attachment respectively. We demonstrate that the pathogen-driven generation of both these distinct actin structures relies on the recruitment and activation of PAK. We show that the PAK kinase domain is dispensable for this actin remodelling, which instead requires the GTPase-binding CRIB and the central poly-proline rich region. PAK interacts with and inhibits the guanine nucleotide exchange factor β-PIX, preventing it from exerting a negative effect on cytoskeleton reorganisation. This kinase-independent function of PAK may be usurped by other pathogens that modify host cytoskeleton signalling and helps us better understand how PAK functions in normal and diseased eukaryotic cells.


Introduction
The p21 activated kinase (PAK) family of proteins are associated with numerous key cellular processes. Between them the Class I (PAK 1, 2 and 3) and Class II (PAK 4, 5 and 6) PAKs, which have both distinct and overlapping functions, regulate signalling pathways such as those involved in cell motility, cell cycle, cancer progression and gene expression [1,2]. Originally identified as serine/threonine kinases, all PAKs contain a C-terminal kinase domain, and a Cdc42/Rac interactive Binding (CRIB) domain [3,4]. Class I PAKs are autoinhibited as a result of an interaction between the kinase domain and the autoinhibitory domain (AID), though whether this interaction is inter-or intramolecular remains controversial [5,6]. Whichever model is correct, inhibition is relieved upon binding of Rac1 or Cdc42 to the CRIB (which partially overlaps with the AID), resulting in the kinase domain being exposed, stimulating its activity. Auto-phosphorylation of the protein then occurs (including within both the kinase domain itself and the CRIB/AID) stabilising PAK's open active state [7]. Class I PAKs can also be activated by binding to inositol phosphates and sphingolipids [8], and there is increasing evidence that dual binding to lipids and Rac1/Cdc42 is required for full activation [9]. Class II PAK regulation appears more complex, as they are constitutively phosphorylated whilst existing in a monomeric autoinhibited state. Numerous proteins including Cdc42 appear to be able to relieve the inhibition of the Class II PAK kinase activity upon binding [10]. Historically much PAK related research has concentrated on identifying substrates of the kinase domain, and a multitude of candidates have been discovered. These include both positive and negative regulators of actin turnover such as Myosin Light Chain (MLC) [11]), Filamin A [12], ArpC1b [13] and LIM Kinase [14]. PAKs also contain poly-proline rich regions (PPRs) which interact with SH3 domain-containing proteins such as the adaptor protein Nck [15] and the guanine nucleotide exchange factor (GEF) β-PIX [16]. PAK thus act as a scaffold for numerous signalling complexes, again leading to both stimulation and inhibition of actin rearrangement. This broad set of substrates and interactors highlight the complicated role PAK plays in actin cytoskeleton remodelling.
Many bacteria drive infection by manipulation of the host actin cytoskeleton. The Gramnegative facultative intracellular pathogen Salmonella enterica Serovar Typhimurium (hereafter "Salmonella") forces its own uptake into non-phagocytic host cells, a key step in infection ultimately leading to symptoms ranging from acute gastroenteritis to more serious systemic disease [17]. To gain entry into host cells Salmonella employs a cocktail of injected virulence factors that remodel the actin cytoskeleton to drive uptake through a process similar to macropinocytosis [18,19]. Previous research has identified a role for PAK in the cell's nuclear response to Salmonella infection [20], as well as in the generation of an acute inflammatory response [21]. However, no role for PAK in bacterial uptake has been demonstrated, with the overexpression of kinase dead PAK as well as PAK containing mutated proline residues being shown to have no negative effect on invasion [20]. PAK is known to play a key role in driving micropinocytosis in healthy cells [22], and has been reported to be subverted by a range of pathogens (both viral and bacterial) in order to rearrange the host cytoskeleton [23]. Indeed, we have recently identified a key role for Class I PAKs in the intimate attachment and subsequent actin-rich pedestal formation by the extracellular pathogens Enteropathogenic and Enterohaemorrhagic Escherichia coli (EPEC and EHEC) [24,25]. We therefore sought to examine the involvement of PAK in the initial remodelling of the actin cytoskeleton that drives Salmonella uptake.

Class I PAKs are crucial for Salmonella invasion
To determine whether PAK was required for the invasion of cells by Salmonella, Hap1 cells were treated with a range of commercially available PAK inhibitors, before being infected for 20 minutes. All of the Class I PAK inhibitors tested (AZ13705339, FRAX486, FRAX594, FRAX1036, G5555, IPA3 and NVS-PAK1) reduced Salmonella invasion by 75-85% compared to DMSO or untreated controls (Fig 1A). In contrast, neither of the Class II PAK inhibitors tested (PF3758309 or GNE 2861) had any significant effect on invasion. All of the inhibitors used interfere with ATP binding within the kinase domain except IPA-3 which binds to the PAK regulatory domain blocking Rac1/Cdc42 binding [26] (summarised in S1A Fig). Similar results were observed in both MEF (S1B Fig) and HeLa cells (S1C Fig), indicating that Class I, but not Class II PAKs are involved in Salmonella uptake into non-phagocytic cells.
To confirm the involvement of Class I PAKs, we used PAK1 (ΔPAK1) or PAK2 (ΔPAK2) knockout Hap1 cells. Although uptake was reduced in both cell lines (Fig 1B), ΔPAK1 cells saw the greater decrease (62%) compared to that in ΔPAK2 cells (37% decrease). Additional knockdown using siRNA (S1D Fig) of the PAK isoform not knocked out, (i.e. ΔPAK1 cells + PAK2 siRNA) resulted in invasion being impeded by approximately 75% (similar to the impairment seen in Class I PAK inhibitor treated cells). Therefore, Salmonella invasion involves both PAK1 and 2, with PAK1 seemingly making the greater contribution.
We have previously shown that EPEC also hijacks Class I PAKs, but unlike EPEC, which has the effector protein EspG [27], Salmonella has no effector known to directly interact with PAK [28]. We hypothesised that Salmonella could recruit PAK indirectly via injection of the effectors SopE1 and SopE2, which are able to activate the small GTPases Rac1 and Cdc42 [29], both potent activators of PAK. When ΔPAK1 Hap1 cells were transfected with Emerald-tagged PAK1 (Em-PAK1) prior to infection, Em-PAK1 was enriched at WT, but not ΔSopE1/E2, Salmonella entry foci (Fig 1C). Em-PAK1 was also recruited to sites of Salmonella entry in MEF and HeLa cells (S1E Fig). Em-PAK1 successfully restored invasion in ΔPAK1 cells infected with WT Salmonella (Fig 1D), confirming the functionality of Em-PAK1. As expected, ΔSopE1/E2 invasion was greatly reduced in all conditions, as SopE1/E2 are required to trigger the actin rearrangements that drive uptake.
To assess whether Salmonella activates PAK during infection, ΔPAK1 cells expressing Em-PAK1 were serum starved to reduce the starting level of phosphorylated (active) PAK, then either treated with control FBS, or infected for 20 minutes with WT or ΔSopE1/E2 Salmonella. Em-PAK1 was then extracted from cell lysates using a GFP Trap (Chromotek) and samples immunoblotted using a phospho-PAK1 Ser144 antibody. Serine 144 lies within the autoinhibitory domain of PAK, it is phosphorylated upon GTPase binding, and helps stabilise PAK in the open confirmation [7], thus is a good indicator of the levels of open active PAK in the cell. WT Salmonella caused a significant increase in the levels of phosphorylated PAK compared to the serum starved control, equivalent to that seen for FBS treatment, while the ΔSopE1/E2 strain did not ( Fig 1E). Collectively these results show that PAK is involved in Salmonella uptake and is recruited and activated in a SopE1/E2-dependent manner.

PAK Kinase domain is dispensable for Salmonella invasion
In an attempt to determine the role PAK plays during Salmonella invasion, various PAK1 constructs were transiently overexpressed in ΔPAK1 cells (S2A Fig), which were then infected with WT Salmonella (Fig 2A). The introduction of either WT or constitutively active PAK1 (PAK1 L107F , which is not autoinhibited and is instead constitutively "open") fully restored invasion, confirming that it was the loss of PAK1 causing the decrease in invasion. PAK1 H83/86L , which is unable to bind GTPases, was incapable of restoring invasion, however, remarkably, the expression of either a kinase-dead PAK1 construct (PAK1 K299R ), or a constitutively open, kinase-dead variant (PAK1 K299R, L107F ), completely restored Salmonella entry, suggesting that kinase activity is not required. This result was surprising as chemical inhibition of PAK kinase activity blocks Salmonella entry ( Fig 1A). As these inhibitors target all Class I PAKs, it may

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling therefore be necessary for there to be some PAK kinase activity in the cell, even if the PAK1 recruited by Salmonella does not have any kinase activity of its own. This activity is likely required for the phosphorylation of residues such as Serine 144 that maintains the open active Salmonella (20 mins) into WT and ΔPAK1 Hap1 cells, either mock-transfected (-) or transfected with Emerald-tagged PAK1 (+ Em-PAK1). Invasion values relative to WT Salmonella invasion into control WT Hap1 cells. (E) Immunoblot of total (Total PAK) and active (P-PAK1 Ser144 ) Emerald-PAK, isolated from transfected ΔPAK1 Hap1 cells using a GFP-Trap. Cells were serum starved overnight, then either treated with FBS, or infected with WT or ΔSopE1/E2 Salmonella (20 mins), as indicated. The level of active PAK relative to total PAK, calculated from band intensities (relative P-PAK1 Ser144 ), is indicated. All Error bars indicate SD. NS-no significant difference, �� -P <0.01, � P <0.05 (ANOVA followed by a post hoc Dunnett's comparison). https://doi.org/10.1371/journal.ppat.1009902.g001

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling conformation of PAK1. Consistent with this, both WT PAK1 and PAK1 K299R are phosphorylated in response to Salmonella invasion, but neither are in G5555-treated cells (S2C Fig). It is therefore possible that although kinase activity does not play a direct role in the signalling downstream of PAK, it does play a role in adopting an invasion-permissive conformation.
In order to explore this possibility further, and determine which domains of PAK are important, various PAK1 truncations were transiently overexpressed in ΔPAK1 cells (Figs 2B and S2B) and Salmonella invasion assays were carried out as before (Fig 2C). Consistent with the findings above, PAK1 lacking its kinase domain (PAK1 1-268 ) was able to fully restore Salmonella invasion to levels observed in WT cells, whereas the kinase domain alone (PAK1 269-545 ) could not. A construct encompassing the N-terminal polyproline region (NPP) and CRIB/AID (PAK1 1-149 ), could not restore invasion, however the CRIB/ AID and central polyproline region together (CPP; PAK1 71-268 ) was also able to fully restore invasion. Neither CRIB/AID nor CPP alone (PAK1 71-149 and PAK1 150-268 respectively) had any effect.
To examine the requirements for PAK recruitment to Salmonella entry foci, Emerald-tagged versions of the PAK1 truncations were expressed in ΔPAK1 Hap1 cells, and then infected with WT Salmonella for 10 minutes. PAK1  (lacking the kinase domain) was recruited to the sites of bacterial invasion equivalently to full-length PAK1 (Fig 2D), as were all forms of PAK1 containing the CRIB/AID (S2D Fig). Conversely, truncations of PAK1 without a CRIB/AID were not recruited, with the kinase domain and CPP alone appearing mostly cytosolic. Collectively these data show that the CRIB/AID is required to localise PAK1 to invasion foci, and that this domain along with the CPP is required to promote Salmonella entry, with kinase activity, and indeed the kinase domain, being dispensable.

PAK kinase and central polyproline domain have opposing effects on Salmonella invasion
Consistent with the kinase domain being dispensable, the kinase-dead mutant (PAK1 K299R ) not only restored invasion in ΔPAK1 cells but its over-expression also enhanced uptake into WT cells (Fig 3A). A kinase-active mutant (PAK1 T423E ) could not restore invasion in ΔPAK1 cells and surprisingly actually inhibited invasion when expressed in WT cells. This suggests that the kinase activity of PAK is not just dispensable, but that it is inhibitory for Salmonella uptake. The expression of the kinase domain alone (Fig 3B) failed to restore invasion in ΔPAK1 cells and despite kinase activity being inhibitory, its expression in WT cells had no significant effect on invasion ( Fig 3C). This may be because the kinase domain alone fails to be recruited by Salmonella (S2A Fig). To artificially localise it to the plasma membrane, the kinase domain was fused to either the CRIB/AID of PAK1, or as Salmonella is known to generate this lipid at sites of invasion, to the PI(3,4,5)P3-binding Pleckstrin homology (PH) domain of the Arf GEF ARNO (CRIB-kinase and PH-kinase respectively; Figs 3B and S3A). Both PH-kinase and CRIB-kinase failed to restore invasion in ΔPAK1 cells, and they also impeded invasion in WT cells by 55% and 35% respectively. The greater inhibitory activity of the CRIB-kinase construct could be due to it also blocking the binding of endogenous native PAK to Rac1 and Cdc42, i.e., acting like a dominant-negative. Indeed, expression of the CRIB alone (and not the PH alone) impeded invasion to a similar extent in WT cells. These data demonstrate that when localised to the bacterial entry site, the PAK1 kinase domain reduces Salmonella invasion significantly.
The CPP domain of PAK is needed for Salmonella entry (Fig 2C), however when expressed alone it failed to restore invasion in ΔPAK1 cells, and significantly inhibited invasion when expressed in WT cells (Fig 3C). This could suggest that when expressed in isolation it acts as a

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling cytoplasmic sink for the proteins normally recruited by endogenous PAK. Like the kinase domain, the CPP alone does not localise at the membrane. As shown above (Fig 2D), when fused to the CRIB domain (CRIB-CPP), this domain fully restored invasion in ΔPAK1 cells, and drove a small increase in invasion in WT cells. Interestingly, the expression of CPP fused to the ARNO PH domain (PH-CPP) not only significantly promoted uptake in WT cells, but restored invasion to 123% of control in ΔPAK1 cells. PH-CPP could also restore Salmonella entry in WT Hap1 cells treated with the PAK inhibitors IPA3 or G5555 (Fig 3D), as well as in MEF and HeLa cells treated with these inhibitors (S3B and S3C Fig). When expressed in Hap1 cells Emerald-tagged PH-kinase, PH-CPP as well as CRIB-kinase were all successfully recruited by Salmonella after a 10-minute infection, confirming that an exogenous PH domain (and also the CRIB/AID of PAK1 itself) could successfully localise proteins to bacterial entry foci (S3D Fig). It seems therefore that when appropriately localised to the membrane, the PAK1 CPP alone is capable of driving Salmonella invasion.

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling

The interaction between PAK and β-PIX is crucial during Salmonella invasion, yet β-PIX depletion restores invasion in the absence of PAK
Having identified the PAK CPP as crucial during Salmonella invasion we next sought to determine its function. The CPP contains a poly-proline motif (PAK1 183-205 - Fig 4A) which interacts with SH3 domains of other proteins. Expression of this motif fused to a PH domain for membrane localisation (PH-PAK1 183-205 ) fully restored Salmonella uptake in ΔPAK1 cells (as well as in WT cells treated with IPA3) to the same level as in WT control cells (Figs 4B and S4B). The expression of the same construct with a mutated PH domain (rendering it unable to bind membranes-PH R279C -PAK1 183-205 ) however, was unable to restore uptake. Interactions between this polyproline motif and SH3 domain-containing proteins can be disrupted with the mutation

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling P191G, R192A [16]. Expression of either PH-CPP P191G,R192A or CRIB-CPP P191G,R192A failed to restore uptake in both ΔPAK1 cells and IPA3-treated WT cells (S4A Fig). The expression of CRIB-CPP P191G,R192A also inhibited invasion in WT control cells, most likely due to it blocking Rac1 and Cdc42 from recruiting endogenous native PAK, as seen above for CRIB-containing constructs ( Fig 3C). These data suggest that interactions between this poly-proline motif and SH3 domain-containing proteins at the plasma membrane are essential for Salmonella invasion.
To determine which proteins were recruited by this polyproline motif, the CPP domain was fused to the C2 domain of lactadherin (a phosphatidylserine-binding domain, henceforth "PSB"), which would allow us to load the purified protein onto lipid bilayers. Phosphatidylcholine:phoshatidyserine (PC:PS) coated silica microspheres [30] were loaded with either control purified His-PSB alone or His-PSB-CPP. Loaded microspheres were incubated in porcine brain extract for 20 minutes, before being isolated, washed, and recruited proteins analysed by mass spectrometry (See S1 Data). The only proteins that showed significant specific binding to CPP were β-PIX, two isoforms of the β-PIX binding partner GIT (GIT1 and GIT2), and SCRIB, a scaffold protein also known to form a tight complex with β-PIX [31]. To confirm these interactions in Hap1 cells, we expressed Emerald-tagged PH-CPP and PH-CPP P191G,R912A , then isolated these proteins using a GFP trap. PH-CPP, but not the SH3-binding mutant PH-CPP P191G,R912A , was able to efficiently recruit β-PIX as well as the β-PIX binding partner GIT1. Recruitment of SCRIB however was not detected (Fig 4D).
Consistent with the importance of the PAK-β-PIX interaction, treatment of cells with the inhibitor PAK18 (a peptide corresponding to PAK residues 187-203, which disrupts binding between PAK and β-PIX, impeded Salmonella invasion by 55%, with a negative control peptide (PAK18 -ve ) having no effect (S4C Fig). Likewise, expression of the PAK1-binding β-PIX SH3 domain also impeded Salmonella invasion (by 75%), with a non-specific SH3 domain (Abi1-SH3) having no significant effect (S4D Fig) It is clear that PAK must interact with β-PIX in order to drive Salmonella uptake, yet despite this, to our surprise, depletion of β-PIX using siRNA not only failed to impede Salmonella invasion, but actually promoted it by 20% (Fig 4D). Even more intriguingly, when β-PIX was depleted in ΔPAK1 cells Salmonella invasion was restored to levels comparable to those in WT cells. Control and β-PIX depleted cells were also treated with the PAK inhibitors IPA3, G5555 or PAK18, and whilst all inhibitors impeded invasion by at least 70% in control cells, in β-PIX depleted cells each inhibitor failed to significantly reduce invasion. The depletion of β-PIX also successfully restored invasion in MEF and HeLa cells treated with G5555 or IPA3 (S4E and S4F Fig). Similarly, in a β-PIX knockout cell line (Δ-βPIX) Salmonella invasion was enhanced by 35% relative to the control, and the three PAK inhibitors had no significant negative effect on invasion (S4G Fig). This finding that β-PIX depletion restores invasion in the absence of Class I PAKs, suggests that the interaction between PAK and β-PIX may be necessary to impede some negative effect β-PIX has on invasion.

β-PIX GEF activity at focal adhesions negatively regulates Salmonella invasion
Consistent with the idea that β-PIX negatively regulates Salmonella uptake we found that its overexpression in WT Hap1 cells impeded invasion by over 50% (Fig 5B), and to a similar extent in both MEF and HeLa cells (S5A and S5B Fig). The isoform of β-PIX used here has five distinct domains, an N-terminal SH3 domain (which binds PAK's CPP) a Dbl-homolgy domain with Rac1 and Cdc42 GEF activity (DH), a lipid binding domain (PH), a GIT interaction domain (GID), and a C-terminal Region with a less well-defined function (CTR) [32]. To investigate which of these domains were responsible for inhibiting Salmonella invasion various

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling , full length β-PIX with a mutation in the DH domain that ablates GEF activity (β-PIX 1-625 L238R,L239S ), the SH3, DH and PH domains of β-PIX with the same mutation (β-PIX 1-400 L238R,L239S ) or β-PIX lacking its SH3 or DH domain (β-PIX 302-625 ). Invasion values are relative to control WT Hap1 cells. (D) Immunoblot of proteins that interact with PAK1. Emerald-tagged PAK1 was expressed in ΔPAK1 Hap1 cells 48 hrs prior to being isolated using a GFP-Trap. Cells were uninfected or infected for 20 mins with WT Salmonella. Total extracted protein was detected using a GFP antibody, and β-PIX, GIT1 and P-PAK ser144 using corresponding antibodies. (E) Fluorescence microscopy of WT and ΔPAK1 Hap1 cells expressing Emerald-tagged β-PIX (green). Cells were infected with WT Salmonella (10 mins) that had been pre-stained with Alexa fluor-350 (blue). Actin was stained with Texas-Red-phalloidin (red). Insets magnify the indicated area. Scale bar is 10 μm. (F) Fluorescence microscopy of cells expressing Emerald-vinculin (green) and Apple-actin

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling truncations were generated ( Fig 5A) and transiently expressed in both WT and Δβ-PIX cells (S5C Fig). The expression of constructs lacking either (or both) the SH3 and CTR domain still impeded uptake to a similar extent as full length β-PIX, in both WT ( Fig 6B) and Δβ-PIX cells (S5D Fig). Conversely the expression of constructs that lacked either the DH or GID domain failed to impede Salmonella uptake (summarised in Fig 5A). Therefore, in order to negatively regulate Salmonella uptake β-PIX requires both the DH and GID domain.

Fig 6. PAK also drives EPEC attachment in a CPP dependent and kinase independent manner. (A) Attachment of WT EPEC to ΔPAK1
Hap1 cells expressing the indicated PAK1 constructs, as described in (Fig 2A). Attachment values are relative to WT Hap1 cells. (B) Attachment of WT EPEC to ΔPAK1 Hap1 cells expressing the indicated PAK1 truncations, as described in (Fig 2C). Attachment values relative to WT Hap1 cells.

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling To determine whether the GEF activity of the DH domain was responsible for impeding Salmonella uptake we also constructed a GEF-dead mutant of β-PIX (β-PIX L238R,L239S ) [16]. The expression of β-PIX L238R,L239S not only failed to impede invasion in either WT or Δβ-PIX but actually promoted it by 22% in WT cells (S5E Fig), suggesting that the GEF-dead version can act as a dominant-negative for endogenous β-PIX. Consistent with this, expression of β-PIX L238R,L239S could fully restore Salmonella uptake to control levels in both ΔPAK1 and IPA3 treated cells (unlike WT full length β-PIX), as indeed could expression of β-PIX (302-625) , a construct lacking the GEF domain (Fig 5C). However, a GEF-dead construct lacking the GID (β-PIX 1-400(L238R,L239S) ) had no effect when expressed in these cells. Together this suggests that interaction with GIT is required for the β-PIX GEF activity to inhibit Salmonella entry, and for GEF-dead β-PIX derivatives to act as dominant-negative.
GIT, through an interaction with the protein Paxillin, is known to localise the β-PIX-GIT complex to Focal Adhesion (FA) complexes [33]. In agreement with this, Emerald-tagged full length β-PIX (Em-β-PIX) localised to adhesion complexes when expressed in Hap1 cells, as did all variants of β-PIX containing an intact GID (S5E Fig). In contrast, Em-β-PIX 1-400 (lacking a GID) appeared to be mostly cytosolic. We can therefore hypothesise that β-PIX exerts a negative effect on Salmonella invasion via its GEF activity when localised to FAs by GIT. To provide further evidence for this notion WT, ΔPAK1, and IPA3 treated WT Hap1 cells were pretreated with MOCK or GIT1 siRNA. Salmonella invasion was enhanced by 30% in GIT1depleted WT cells, and this depletion also restored invasion to control levels in ΔPAK1 and IPA3-treated cells (S5F Fig). GIT depletion restoring invasion in cells lacking class I PAK's strongly suggests that when β-PIX interacts with GIT1 (and therefore localises to the FA), it inhibits Salmonella invasion. This inhibition can be overcome when PAK interacts with β-PIX, possibly through the recruitment of the β-PIX-GIT complex to the bacterial entry foci, away from the FA. To confirm that full length PAK interacts with β-PIX, and to determine whether this interaction changes during Salmonella invasion, ΔPAK1 cells expressing Em-PAK1 were infected with WT Salmonella for 20 mins. Em-PAK1 was isolated using a GFP-Trap, and the recruitment of β-PIX and GIT1 were assessed by immunoblotting. Whilst the levels of active PAK (P-PAK ser144 ) increased relative to uninfected control cells, the levels of β-PIX and GIT bound to PAK did not noticeably change (Fig 5D). This would suggest that the mechanism that PAK employs to impede β-PIX may not rely on increased β-PIX-PAK binding, and instead could be due to a change in the localisation of the PAK-β-PIX complex.
β-PIX is known to modulate FA assembly [33,34], therefore its removal by PAK from the FA during Salmonella invasion may alter FA dynamics. Consistent with this, during Salmonella invasion Em-β-PIX is recruited to Salmonella-induced ruffles in WT cells, but not in ΔPAK1 cells (Fig 5E) and is recruited by Salmonella in MEF and HeLa cell (S5G Fig). Concomitant with this, a stark reorganisation of FAs occurs during the invasion process, with large Vinculin-and actin-rich FA complexes forming within the Salmonella induced ruffle (Fig 5F). To summarise, the data here demonstrates a novel pathway by which PAK, activated by SopE1/E2 during Salmonella invasion, recruits β-PIX to bacterial entry foci (away from FA's where its GEF activity is inhibitory) altering FA dynamics in order to drive the actin reorganisation required for efficient uptake of the bacteria.

PAK also drives EPEC attachment in a CPP-dependent, and kinaseindependent manner
As we have previously demonstrated the necessity of PAK in promoting the intimate attachment of EPEC to host cells [24], we sought to determine whether EPEC, like Salmonella, hijacks PAK in order to overcome the negative effect of β-PIX GEF activity at FAs. Almost

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling identically to Salmonella uptake, EPEC attachment is fully restored in ΔPAK1 cells when kinase-dead PAK (PAK1 K299R ) is expressed (as it is when constitutively-open PAK-PAK1 L107F , constitutively-open kinase-dead PAK-PAK1 K299R,L107F , but not PAK that cannot bind GTPases-PAK1 H83/86L is expressed- Fig 6A, representative image of attached EPEC- S6A Fig). Likewise, expression of either PAK1 lacking the kinase domain (PAK1 1-268 ), or a construct corresponding to the CRIB/AID and CPP domains (PAK1 71-268 ), also fully restored attachment in ΔPAK1 cells (Fig 6B). The expression of a kinase-active form of PAK (PAK1 T423E ) impeded attachment by 40% in WT cells and failed to restore it in ΔPAK1 cells (S6A Fig). Finally, the expression of PH-CPP fully restored EPEC attachment in ΔPAK1 cells, with PH-kinase or CPP alone exerting a negative effect on attachment in WT cells ( S6B Fig). Therefore, it appears that, as is the case for Salmonella invasion, PAK's role in EPEC attachment requires the CPP domain (correctly localised to the membrane), with the kinase domain not only being dispensable, but inhibitory. The ability to restore attachment by EPEC also translates into being able to form actin rich pedestals, with PAK1, PAK1 K299R and PAK1ΔKinase restoring pedestals formation in ΔPAK1 Hap1 cells, and PAK1 T423E failing to do so (S6C and S6D Fig).
The depletion of β-PIX using siRNA promoted the attachment of EPEC to WT cells by more than 40%, and the depletion of either β-PIX, or its binding partner GIT1 restored EPEC attachment in either ΔPAK1 cells or WT cells treated with the PAK inhibitor IPA3 to WT control levels ( Fig 6C). This result indicates that PAK's role during EPEC attachment is to impede the negative effect of the β-PIX-GIT complex, as it is during Salmonella invasion.
EPEC utilises the effector protein EspG in order recruit activated PAK and promote intimate attachment to host cells [24,27]. In the absence of EspG attachment is very poor, therefore we sought to confirm that this is due to the inability to impede or alter the localisation of β-PIX via PAK recruitment. Whilst not fully restoring attachment to the level of WT EPEC, the expression of PH-CPP greatly promoted the attachment of ΔEspG EPEC to WT Cells (from just 19% of control attachment to 68%- Fig 6D). Depletion of β-PIX fully restored the ability of ΔEspG EPEC to attach to WT cells, to a level similar to WT EPEC. Depletion of the β-PIX binding partner GIT1 also promoted attachment of ΔEspG EPEC, albeit to a lesser extent (Fig 6E). Depletion of either β-PIX or GIT1 also greatly promoted ΔEspG pedestal formation as expected (Fig 6F and 6G). This data strongly implies that in order to promote attachment to host cells, EPEC delivers the effector EspG, which recruits PAK, likely resulting in the recruitment of β-PIX-GIT complexes away from FAs, where they are normally inhibitory to the process.
In conclusion it seems that both Salmonella and EPEC utilise PAK in the same kinase-independent manner to drive the actin reorganisation required for efficient infection of host cells.

Discussion
Collectively the data presented here allows us to propose a model outlining how Salmonella hijacks PAK in order to promote uptake into host cells (Fig 7). Salmonella injects the effector proteins SopE1/E2 which activate the small GTPases Rac1 and Cdc42, resulting in the recruitment and activation of Class I PAKs at sites of bacterial entry. PAK could be recruited from the cytosol as well as directly out of nearby FAs (where it resides in resting cells). PAK, via its central polyproline region (CPP), is able to interact with the Rac1 GEF β-PIX, and either recruits the β-PIX-GIT complex out of the FA or modifies β-PIX directly within the FA. Either way, the interaction between PAK and β-PIX prevents β-PIX from impeding the uptake of Salmonella and creates a local cellular environment permissive to the actin reorganisation required. In the absence of PAK, β-PIX resident in FAs is able to exert a negative effect on

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling Salmonella uptake through its GEF activity. This is potentially caused by the mislocalisation of Rac1 away from invasion sites, as well as by the local manipulation of FA dynamics, both of which may impede the actin assembly necessary for Salmonella uptake.
Class I PAKs are serine/threonine kinases that are key regulators of the actin cytoskeleton, therefore it comes as no surprise that many pathogens subvert PAK to promote infection [23]. Its role in Salmonella uptake though does comes as a surprise since previous research, which outlined a key kinase-dependent role for PAK in the activation of c-Jun NH 2 -terminal kinase (JNK) during infection, found no obvious role for PAK in the invasion process [20]. These conclusions relied on the expression of dominant-negative PAK mutants which, as the authors themselves concede, are often not wholly effective. Indeed, one of the derivatives reported to have no effect on uptake was a kinase-dead version, a result we confirm (Fig 2A) and that is actually consistent with PAK's kinase activity being dispensable. Their research does however confirm that PAK is activated in a type three secretion system (T3SS) dependent manner during the first 30 minutes of infection, and that the expression of a variant of Cdc42 that cannot bind PAK does impede uptake, consistent with our findings here. More recent research has demonstrated that PAK1 activated during Salmonella infection plays a vital role in the stimulation of NF-κB, in order to establish pro-inflammatory signalling after cell invasion [21]. Throughout our study, we exploit a range of drugs, knockout cell lines and siRNA experiments

PLOS PATHOGENS
to clearly establish a role for PAK in the entry of Salmonella into non-phagocytic cells, we confirm many of our key results in HeLa and MEF cells also.
We demonstrate that the chemical inhibition of Class I but not Class II PAKs greatly impedes the invasion of Salmonella, something we previously found to be the case for EPEC and EHEC attachment [24]. We also outline a more prominent role for PAK1 over PAK2 through the use of knockout cells. Surprisingly, despite the kinase activity of PAK being previously shown to be an important regulator of the actin cytoskeleton [1]-as well as specifically of macropinocytosis [22], an event that shares many features with Salmonella invasion, our results clearly demonstrate that the role of PAK during Salmonella entry is independent of kinase activity, and instead requires only the CRIB and CPP domains. We show that chemical inhibitors of PAK that interfere with kinase activity actually impede Salmonella invasion, which appears at odds with the finding that the kinase domain of PAK is dispensable for uptake. These kinase inhibitors however target all Class I PAK kinase activity in the cell, and therefore prevent phosphorylation of residues such as Ser144 that leads to PAK becoming open and active. We would expect that these inhibitors block all PAK functions, and not just those associated with kinase activity. We therefore propose that whilst some PAK kinase activity is required for the initial opening and activation of PAK1, once recruited by Salmonella (or EPEC), the kinase domain plays no role in the downstream signalling required to drive uptake or attachment.
Although surprising, the kinase-independent role of PAK is not unprecedented. PAK has previously been proposed to function as a scaffold [16], mediating the recruitment of several effectors to sites of Rac1 activation. The expression of kinase-dead PAK can also induce lamellipodia formation [35][36][37], and in fact this mutant of PAK may be even more efficient at driving the generation of Rac1 dependent ruffles than wildtype [33]. Further to being dispensable for efficient uptake of Salmonella, we have shown that the kinase domain of PAK may actually inhibit invasion, partially opposing the kinase-independent pro-invasive function. Several PAK kinase substrates inhibit actin turnover such LIM-Kinase (LIMK), (which phosphorylates cofilin) [14], and MLC [38], which may explain why the kinase activity inhibits Salmonella invasion. Further work is required to define the mechanisms that control the balance between the potential stimulatory and inhibitory pathways that exist downstream of PAK.
We demonstrate that the CPP of PAK (when appropriately localised) is the sole requirement for PAK to promote efficient Salmonella uptake. It is well established that the CPP of PAK interacts with the Rac1/Cdc42 GEF β-PIX [16], and we too demonstrate this to be the case (Figs 4C and 5D). Initially we assumed that the GEF activity of β-PIX, when recruited to entry sites, could activate additional Rac/Cdc42, promoting invasion in a manner analogous to the role previously reported for the bipartite Rac1 GEF ELMO1/DOCK180 [39]. In fact, we demonstrate that β-PIX GEF activity impedes Salmonella uptake (Figs 5B, 5C, S5A, S5B and S5D). It could be that β-PIX activates Rac1 at the wrong location to promote uptake. Consistent with this, β-PIX GEF activity only inhibited uptake when β-PIX was able to interact with its well characterised binding partner, GIT1 [33]. As this interaction leads to localisation at FAs, it is therefore possible that β-PIX diverts large amounts of Rac1 away from the Salmonella entry foci. β-PIX has indeed been shown to localise Rac1 to FAs [40], and previous research has reported that in cells lacking Paxillin, a protein that links β-PIX-GIT to the FA [41], Salmonella uptake is enhanced [42], consistent with our findings that uptake is increased in β-PIX and GIT1 depleted cells. Salmonella-activated PAK, perhaps via competition for Rac1 binding by β-PIX [43], or by recruiting β-PIX (and indeed GIT1) out of the FA, may overcome this Rac1 mislocalisation. Whilst we demonstrate as expected that Rac1 is recruited to sites of Salmonella entry, determining whether PAK inhibition prevents this from occurring is difficult, since PAK inhibited cells in general do not form actin rich ruffles (S7A Fig). This aside, it is

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling still feasible that Rac1 mislocalisation caused by β-PIX is something that PAK is able to overcome during infection.
An alternative explanation is that by removing β-PIX from the FA, Salmonella may alter FA dynamics in a way conducive with the actin rearrangements required for uptake. Depletion of β-PIX [44,45] or overexpression of mutant variants lacking GEF activity [46,47] (both of which we have shown enhance Salmonella uptake) have been demonstrated to alter FA dynamics, and we show that FAs are remodelled during Salmonella invasion (Fig 5F), concomitant with β-PIX recruitment to entry foci (which we show occurs only in cells that express PAK1).
Consistent with the idea that FA dynamics are important, the FA components focal adhesion kinase (FAK), and p130Cas have previously been implicated in Salmonella uptake, with invasion being greatly impaired in FAK knockout cells [42]. Interestingly like PAK, the kinase activity of FAK was not required to restore Salmonella invasion, and we show that impeding FAK's kinase activity through the use of a chemical inhibitor [48]  Whether Salmonella directly alters adhesions in and around the invasion foci or indirectly via changes to the state of the actin cytoskeleton remains to be determined. It is of course possible that the effects seen on Salmonella entry are due to combination of FA alterations and potential changes in Rac1 availability outlined above.
Salmonella and EPEC both manipulate the actin cytoskeleton to drive very distinct outcomes, yet fascinatingly they both require kinase-independent functions of PAK to achieve this. This suggest that what PAK does may not directly affect the kind of actin structures generated. Instead, PAK seems to provide a suitable environment in which other effector proteins (such as SopB, SipA and SopE in Salmonella, and Tir in EPEC) can carry out actin assembly to achieve the desired result. We propose that the modification of FA dynamics, (and/or the change in Rac1 localisation) achieved through the recruitment of the β-PIX-GIT complex out of the FA by PAK's CPP, allows Salmonella and EPEC to reorganise the actin cytoskeleton to achieve uptake or intimate attachment respectively. This novel kinase-independent mechanism for PAK may not just be exclusive to the two pathogens we have investigated, with other bacteria and viruses potentially hijacking PAK to promote infection. It remains to be seen whether this pathway is also involved in normal cellular regulation of FA dynamics, or indeed more intriguingly in aberrant conditions, such as those in cancerous cells where PAK is often overexpressed or abnormally activated [50,51]. PAK's are important pharmacological targets, therefore our discovery of a novel kinase-independent pathway may be of considerable interest for researchers when designing new drugs.
It is certainly clear that a greater understanding of how PAK modifies both the microenvironment, and FA dynamics is required. What is also certain though, is that the relationship between PAK and β-PIX may be more complicated than previously believed, and at least in terms of pathogen manipulation of the actin cytoskeleton, be far more important than previously appreciated.

Bacterial strains
WT Salmonella (Salmonella enterica serovar Typhimurium, SL1344) was a kind gift from Jean Guard-Petter, the isogenic ΔSopE1ΔSopE2 was a kind gift from Wolf-Dietrich-Hardt. WT

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling EPEC (E2348/69) and the isogenic mutant ΔespG1/ΔespG2 (ΔEspG) were kind gifts from Feng Shao.

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling GAA), β-PIX (Mouse) (Mm_Arhgef7_1, CTCCTTGGTGACACTAAATAA; Mm_Arhgef7_2, CAGCAAACACTTCATATTTAA) and GIT1 (HS_GIT1_5, CAGCCTTGACTTATCCGAA TT; HS_GIT1_6, GCCGCTGAGGATGTCCCGAAA). siRNAs were transfected using Oligofectamine (Life-Technologies) following the manufacturer's instructions, a final concentration of 10 nM was used for each pair of siRNAs, and cells were treated 72 hrs prior to experiments being carried out. (Confirmation of knockouts and knockdown efficiencies can be found in S1D and S4H Figs)

Salmonella infection assays
Invasion was quantified using Salmonella carrying the pM975 plasmid (kind gift from Wolf-Dietrich-Hardt). Once internalised these bacteria express green fluorescent protein (GFP) via the SPI2 promoter, allowing for detection of internalised bacteria only. Unless stated otherwise, cells were infected for 20 minutes using Salmonella at a multiplicity of infection (MOI) of 100, washed twice in Phosphate Buffered Saline (PBS), then incubated in Gentamycin (100 μg/ ml) containing media for 90 minutes (to kill extracellular bacteria-a gentamycin protection assay). Cells were subsequently fixed, stained with Texas-Red Phalloidin (Life-Tech) and DAPI, then visualised using immunofluorescence (IF) microscopy. The number of GFP-positive (internalised) bacteria were counted per cell (minimum 500 cells per condition). All experiments were carried out at least three times, geometric means calculated, and significance determined, a P of <0.05 (as determined by ANOVA followed by a post hoc Dunnett's comparison) was deemed significant. When analysing recruitment of Emerald-tagged constructs, Salmonella were stained using Alexa-Fluor-350 carboxylic acid succinimidyl ester (Lifetech, 15 minutes, 25˚C), then washed in Tris (pH 7.4)-buffered saline before being used in a 10-minute infection (unless stated otherwise). Images were taken using wide-field IF, or total internal reflection (TIRF) microscopy as stated in the text.

Fluorescence microscopy
TIRF microscopy was performed on an Olympus IX83 inverted microscope equipped with a Uapon 100x/ 1.49NA objective and Olympus IX3-SSU automated xy stage and IX3 Z-Drift Compensator. Images were taken using CellSens software with a 100 nm penetration depth and recorded on a Hamamatsu imageEM x2 EM-CCD camera. Cells were imaged in Leibovitz L15 medium without phenol red (ThermoFisher) and maintained at 37˚C in a stage top incubator (Cell vivo, Pecon). Images for invasion, attachment and pedestal formations assays as well as Emerald tagged PAK and β-PIX recruitment by Salmonella were captured on an Olympus IX81 inverted microscope.

EPEC infection assays
For adhesion assays cells were infected with WT or ΔEspG EPEC for 90 minutes (MOI~50), cells were then washed twice in PBS, twice briefly in ice cold 200 mM glycine (pH 2), then again in PBS twice. Cells were fixed, stained with Texas-Red Phalloidin and DAPI, and adherent EPEC stained using an anti-intimin antibody. The number of adherent bacteria per cell were then counted using fluorescence microscopy (of at least 500 cells). To assess pedestal formation the same assay was carried out without the glycine wash step and the number of actin rich pedestals per cell were counted. All experiments were carried out at least three times, geometric means calculated, and significance determined by ANOVA followed by a post hoc Dunnett's comparison or post-hoc Tukey where appropriate, with a P of <0.05 considered significant.

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling

GFP trap
Constructs expressing Emerald-tagged proteins were transfected into Hap1 cells 48 hours prior to performing the GFP trap. The GFP trap was carried out using the manufacturer's instructions (Chromotek), resulting trapped proteins run on an SDS-PAGE gel before being transferred to PVDF membranes, and probed with appropriate antibodies. All immunoblots were visualised using a LI-COR Odyssey Fc imaging system, with band intensities determined using LI-COR Image studio software.

In-vitro pulldowns on lipid coated beads
His-tagged PSB and PSB-CPP were expressed in E. coli Rosetta (Novagen) at 18˚C before affinity purification [52]. Pulldowns using lipid coated silica micropsheres were subsequently carried out as described [30]. Briefly, silica microspheres (Bangs Laboratories) were coated in a lipid bilayer made up of phosphatidylcholine (PC) and phosphatidylserine (PS) at an 80:20 molar ratio (Avanti Polar Lipids). PSB or PS-CPP were anchored to these membranes and incubated in a cell-free porcine brain extract for 15 minutes, before being washed extensively.

PLOS PATHOGENS
A kinase-independent function of PAK is crucial for pathogen-mediated actin remodelling MEF cells or (C) WT HeLa cells, each either pretreated with DMSO (control), IPA3 or G5555, cells expressed either no construct (-) or PH-CPP. Invasion values are relative to those in control DMSO treated MEFs or HeLas respectively (D) Fluorescence microscopy of ΔPAK1 Hap1 cells expressing the Emerald-tagged PAK1 constructs indicated (green), infected (10 mins) with Alexa Fluor 350-stained WT Salmonella (blue). Actin stained with Texas-Red Phalloidin (red). Scale bar is 10 μm. All Error bars indicate SD. NS-no significant difference, �� -P <0.01, � P <0.05 (ANOVA followed by a post hoc Dunnett's comparison). (TIF) S4 Fig. (A) Invasion of WT Salmonella into WT, ΔPAK1 or WT Hap1 cells pretreated with IPA3. Cells expressed the indicated construct, as described in (Fig 4A)