Actin activates Pseudomonas aeruginosa ExoY nucleotidyl cyclase toxin and ExoY-like effector domains from MARTX toxins

Pseudomonas aeruginosa is a major cause of chronic infections in cystic fibrosis patients. The nucleotidyl cyclase toxin ExoY is a virulence factors injected by the pathogen and associated with severe damage to lung tissue. ExoY-like cyclases are also found in other Gram-negative pathogens and shown to contribute to virulence, although they remained poorly characterized. Here we demonstrate that filamentous actin (F-actin) is the hitherto unknown co-factor that activates P. aeruginosa ExoY within host target cells. Highly purified actin, when polymerized into filaments, potently stimulates (>10,000 fold) ExoY activity. ExoY co-localizes in vivo with actin filaments in transfected cells and, in vitro, it interferes with the regulation of actin assembly/disassembly-dynamics mediated by important F-actin-binding proteins. We further show that actin also activates an ExoY-like adenylate cyclase from a Vibrio species. Our results thus highlight a new sub-class within the class II adenylyl cyclase family, defined as actin-activated nucleotidyl cyclase (AA-NC) toxins.


INTRODUCTION 40
Pseudomonas aeruginosa is an opportunistic human pathogen that causes severe infections in 41 immune-compromised individuals and is a major cause of chronic infections in cystic fibrosis 42 patients. Equipped with a type III secretion system (T3SS), P. aeruginosa can inject effector 43 proteins directly into the host cells where they contribute to virulence of the pathogen [for 44 review see 1,2 ]. Four different T3SS delivered effectors have been characterized (exoenzyme 45 T, Y, U and S) , but new effectors were recently identified 3 . Exoenzyme Y (ExoY) is present 46 in 89% of clinical isolates 4 , and was originally identified as an adenylate cyclase in 1998 (ref. 47 5 ) due to amino acid sequence homology with two well characterized class II adenylate 48 cyclase toxins, CyaA from Bordetella pertussis and edema factor (EF) from Bacillus 49 anthracis. Recent results revealed that substrate specificity of these enzymes expressed in 50 cell cultures is not restricted to ATP: EF and CyaA were shown to use UTP and CTP as 51 substrate 6 while ExoY was shown to promote the intracellular accumulation of several cyclic 52 nucleotides 7, 8 with a preference for cGMP and cUMP over cAMP and cCMP formation 7 . 53 The physiological effects of ExoY resulting from accumulation of these cyclic nucleotides 54 include the hyperphosphorylation of tau and the disruption of microtubules causing the 55 formation of gaps between endothelial cells and increased permeability of the endothelial 56 barrier 8,9,10,11 . Most recent results showed that ExoY presence correlates with long term 57 effects on recovery after lung injury from pneumonia 12 . 58 Recent whole genome sequencing projects have identified ExoY nucleotidyl cyclase 59 modules among a variety of toxic Multifunctional Autoprocessing repeats-in-toxins 60 (MARTX) effector domains in multiple bacterial species of the Vibrio genus 13 that represent 61 emerging human or animal pathogens. These ExoY-like domains can be essential for 62 virulence 13 . Elucidating the cellular activities and specificities of ExoY and ExoY-like toxins 63

ExoY interacts with mammalian actin in vitro 113
To verify the interaction between ExoY and mammalian actin in vitro, we performed Ni-114 NTA agarose pulldowns. ExoY with a C-terminal Flag-His tag (ExoY-FH) and α-actin from 115 rabbit skeletal muscle (Cytoskeleton, Inc., designated here MA-99) were added at equimolar 116 concentrations to Ni-NTA agarose beads in batch. After 1 hr incubation the beads were 117 washed in Durapore columns and the retained proteins were eluted with imidazole. While 118 very little actin was bound unspecifically to the beads in the absence of ExoY, considerably 119 more actin was present in the eluate from the sample containing ExoY-FH, suggesting 120 specific binding between the two pure proteins (Fig. 2a). 121 Under our experimental conditions, which promote actin spontaneous polymerization 122 (300 mM NaCl, 2.5 μM ATP, 660 nM actin), actin should exist in solution both in its 123 monomeric (G-actin) and filamentous (F-actin) states. To investigate more specifically 124 whether ExoY could bind to F-actin, we performed high-speed co-sedimentation assays: we 125 polymerized G-actin-ATP and incubated F-actin at steady state subsequently with ExoY. 126 Samples were then centrifuged at high speed to separate F-actin present in the pellet from G-127 actin present in the supernatant. Fig. 2b shows that ExoY, which alone partitioned in the 128 supernatant fraction, was mostly found in the pellet fraction in the presence of F-actin, 129 indicating ExoY capacity to interact with F-actin. 130 131

Actin stimulates ExoY nucleotidyl cyclase activity in vitro 132
We next tested whether purified actin could activate ExoY in vitro. Highly pure non-muscle 133 (cytoplasmic) actin isolated from human platelets (Cytoskeleton, Inc., designated here A-99) 134 strongly stimulated the adenylate cyclase activity of ExoY (HF-ExoY, having 6xHis and Flag 135 tags at the N-terminus), with a maximal activity reaching 120 μmol of cAMP.min -1 .mg -1 . 136 Since in mammalian cells the transfection with a vector expressing ExoY led to accumulation 137 of cGMP to levels exceeding that of cAMP 7, 8 , we also tested GTP as substrate and found 138 that, in agreement with the preferential accumulation of cGMP over cAMP observed in vivo, 139 the guanylate cyclase activity of HF-ExoY was approximately 8 times higher than the 140 adenylate cyclase activity in the presence of actin in vitro (Fig. 3a). The background activity 141 without actin was estimated to be about 1 nmol and 10 nmol.min -1 .mg -1 for cGMP and 142 cAMP, respectively. Thus, the ExoY nucleotidyl cyclase activity was stimulated more than 143 10,000 fold by submicromolar concentrations of F-actin. Different mammalian actin isoforms 144 (A-99, a mixture of 85% β-and 15% γ-actin, or α-actin from rabbit skeletal muscle) led to a 145 similar activity for cGMP synthesis ( Supplementary Fig. 2). All together these data indicate 146 that actin is a specific activator of ExoY in eukaryotic cells. Subsequent experiments were 147 performed using highly pure skeletal muscle α-actin purified in one of our laboratory 148 (designated MA-L). 149 To examine a possible dependence of ExoY activation on the different states of actin 150 (ATP-versus ADP-bound, monomeric versus polymeric forms), we measured ExoY cGMP 151 synthesis activity at different actin concentrations below or above the critical polymerizing 152 concentrations in different actin nucleotide states. Measurements were performed at various 153 concentrations of actin that was initially loaded with either Mg-ATP or Mg-ADP. A similar 154 maximal activity of 1000 -1200 μmol of cGMP.min -1 .mg -1 was obtained with both ATP-155 bound and ADP-bound actin (Fig. 3b). In contrast, the actin concentrations required for half 156 maximal activation of ExoY (K 1/2 ) were dependent upon the bound nucleotides. The actin 157 concentration for half maximal ExoY activation with ATP-loaded actin was about 0.2 µM 158 ( Fig. 3b). This is just above the critical concentration of 0.1 μM above which Mg-ATP-actin 159 spontaneously polymerizes in solution with salt 19 . Conversely, the actin concentrations 160 required for half-maximal ExoY activation with ADP-loaded actin was about 2.4 µM (Fig.  161 3b), a shifted value that correlates with the alternative critical concentration of 1.7 μM 162 obtained with Mg-ADP-actin. Altogether, these results suggest that the maximal activation of 163 ExoY by actin was correlated with F-actin formation. 164 165

Activation of ExoY by actin is antagonized by latrunculin A or G-actin binding proteins 166
We then examined whether proteins or molecules that are known to bind to G-actin and to 167 inhibit its polymerization or spontaneous nucleation, could affect the activation of ExoY by 168 actin. We examined the activation of ExoY by G-actin in the presence of the drug latrunculin 169 A or G-actin binding proteins such as profilin or thymosin-β4 (Tβ4), which are among the 170 main monomeric actin-binding proteins in vertebrate cells 19 . These three molecules are 171 known to inhibit actin polymerization or spontaneous nucleation by binding to distinct G-172 actin interfaces. The small macrolide Latrunculin A from the Red Sea sponge Negombata 173 magnifica 20, 21 inhibits actin self-assembly by binding (K D ≈ 0.2 μM) to a cleft located on the 174 pointed face of G-actin. The protein profilin binds (K D ≈ 0.1 μM) in contrast to the opposite 175 face of monomers, called barbed face, and favors in vivo the unidirectional elongation of the 176 most-dynamic barbed ends of filaments. In vitro, G-actin:profilin complexes inhibit actin 177 spontaneous nucleation and thus polymerization in absence of actin nuclei or filament seeds. 178 Tβ4 is a small intrinsically disordered β-thymosin domain of 4 kDa that acts as a major G-179 actin-sequestering polypeptide in cells 22, 23 . Here, we used a chimeric β-thymosin domain 180 between Tβ4 and ciboulot from Drosophila called Chimera 2 (CH2), as it exhibits a higher 181 affinity for G-actin than Tβ4 (K D~0 .5 μM versus 2 μM) while retaining its sequestering 182 activity 22, 23 . Like Tβ4, CH2 displays an extended binding interface on actin monomers by 183 interacting with both their barbed and pointed faces 22, 23, 24 . 184 In ExoY activity measurements, actin monomers were saturated by the above 185 molecules to inhibit or significantly slow down the spontaneous nucleation or polymerization 186 of actin in solution. The inhibitory effect on actin assembly was verified in cosedimentation 187 assays ( Supplementary Fig. 3). In these conditions, all molecules tested inhibited ExoY 188 activation by micromolar actin concentrations (Fig. 4), which normally induce maximal 189 activation of the toxin. CH2 reduced at least 9 fold (up to 15) the ExoY activity measured in 190 the presence of 1-3 μM of actin (Fig. 4). Latrunculin A reduced at least 7 fold (up to 13) and 191 profilin decreased at least 4 fold (up to 7) the ExoY activity at similar ranges of actin 192 concentrations (Fig. 4). Latrunculin A, profilin, and CH2 did not affect the low background 193 activity of ExoY in the absence of actin. These data thus indicated that filamentous-actin is 194 the preferred activator of ExoY. 195 196 ExoY is an F-actin binding protein that can modify the intrinsic or regulated dynamics 197 of filaments by binding along filament sides 198 We next examined whether ExoY affects the intrinsic or regulated dynamics of actin self-199 assembly in vitro in assembly/disassembly assays with ExoY /ExoY K81M . The kinetics of 200 polymerization or depolymerization were monitored by following the increase or decrease, 201 respectively, in pyrene-actin fluorescence intensity (pyrenyl-labeled actin subunits exhibit 202 higher fluorescence when incorporated in filaments than free in solution). In polymerization 203 kinetics, ExoY slightly accelerated the rate of G-actin-ADP-Mg (Fig. 5a) and G-actin-ATP-204 Mg (Fig. 5b) self-assembly, confirming that ExoY can interact with actin without preventing 205 its self-assembly. Yet, this stimulation of G-actin-ATP/ADP polymerization by ExoY was 206 detectable only at high ExoY concentrations (in µM range). The dose-dependent acceleration 207 was independent of the ExoY adenylate cyclase activity since it was observed with the 208 inactive ExoY K81M variant and also with the wild-type ExoY when only ADP was present. 209 We further examined whether the ExoY-stimulation of actin polymerization was achieved by 210 increasing elongation rates on barbed-or pointed-end, or by severing filaments, but found no 211 effects of the toxin on these processes ( Supplementary Fig. 4). We were unable to isolate 212 stable G-actin/ExoY complexes in solution even at micromolar concentrations of both 213 proteins (and in presence of latrunculin A to prevent actin polymerization). Besides, the 214 ExoY-induced stimulation of actin polymerization was fully inhibited when actin was 215 saturated by profilin (Fig. 5b). These results confirm that ExoY is unlikely to stimulate actin 216 polymerization in host cells. Indeed, profilin-actin complexes form the major part of the 217 polymerization competent G-actin pool within eukaryotic cells 19, 24 . 218 To delineate the interaction of ExoY with F-actin, we performed dilution-induced 219 depolymerization assays monitoring filament disassembly from free barbed-and pointed 220 ends. As shown in Fig.5c, ExoY K81M inhibited the spontaneous disassembly of F-actin 221 induced by dilution. This indicates that ExoY directly binds to filaments and thus stabilizes 222 actin inter-subunit contacts. The inhibition of filament disassembly by ExoY K81M was also 223 observed when barbed ends were capped by gelsolin (Fig. 5c) were chosen based on sequence alignments of P. aeruginosa ExoY and ExoY-like 286 containing sequences from several Vibrio MARTX toxins ( Supplementary Fig. 6), the 287 signature of CPD cleavage sites present in some Vibrio ExoY-like modules and HCA 288 secondary prediction analysis. The MARTX-ExoY protein corresponding to residues 289 Y3412 to L3872 of Uniprot reference F0V1C5 was termed here VnExoY-L. The protein 290 carrying a C-terminal Flag-His tag (VnExoY-L-FH) was purified and tested for its 291 adenylate cyclase activity in the presence and absence of actin. Fig. 7b shows that 292 VnExoY-L displayed a potent adenylate cyclase activity in the presence of actin, which 293 stimulates the enzymatic activity more than 10,000 fold. In contrast to P. aeruginosa 294 ExoY, VnExoY-L did not exhibit any cGMP synthesizing activity. We conclude that actin 295 may be a common activator of the various ExoY-like cyclase modules, even though these 296 differ in their substrate selectivity. Here, we show that actin is a potent activator of a group of bacterial toxins that are 305 homologous to the P. aeruginosa ExoY effector and that display nucleotidyl cyclase 306 activities with different substrate selectivity. 307 We identified actin as a potential candidate for P. aeruginosa ExoY activation 308 through its enriched presence among the proteins that co-purified with TAP-tagged ExoY 309 expressed in S. cerevisiae. Actin is among the most abundant proteins in eukaryotic cells 310 with a large number of known interaction partners. It is also frequently retrieved un-311 specifically in "pull-down" experiments and it is ranked among the top contaminants in the 312 so-called "CRAPome", Contaminant Repository for Affinity Purification 35 . In our case, 313 however, we found that the interaction between ExoY and actin is highly specific and 314 conserved between yeast and different mammalian actin isoforms. We further showed that 315 ExoY is an F-actin binding protein (Fig. 5) and we demonstrated that F-actin is a potent 316 activator of ExoY, able to stimulate its adenylate and guanylate cyclase activity more than 317 10,000 fold. In accordance with this view, we found that ExoY activation by actin was 318 strongly antagonized by different G-actin binding proteins, such as profilin, or a Tβ4-319 derivative protein with similar activity as Tβ4 (CH2), or by latrunculin A that prevents actin 320 polymerization (Fig. 4). Profilin and latrunculin A interact with non-overlapping, opposite 321 binding sites on G-actin, which are mostly buried by actin:actin contacts in filaments. 322 Profilin may partially overlap and compete with ExoY binding sites as it inhibits as well 323 ExoY-mediated stimulation of actin self-assembly (Fig. 5b). In contrast, the antagonizing 324 effect of the small molecule latrunculin A is likely due to its inhibition of actin 325 polymerization rather than to a steric hindrance of the ExoY binding-sites. In the presence of 326 saturating concentrations (> 2-5 μM) of F-actin, the very low basal enzymatic activity of 327 ExoY was strongly stimulated to reach specific activities of about 120 μmol.min -1 .mg -1 and 328 900 μmol.min -1 .mg -1 for cAMP and cGMP synthesis, respectively. The higher guanylate 329 cyclase activity as compared to the adenylate cyclase one is in agreement with the 330 preferential accumulation of cGMP over cAMP observed in vivo 7, 8 . The corresponding kcat 331 for cGMP synthesis is approaching 1000 s -1 and therefore within the same order of 332 magnitude as the catalytic rates measured for cAMP synthesis for the related cyclase toxins 333 CyaA or EF, when activated by calmodulin, their common eukaryotic activator 36, 37 . 334 While ExoY represents to our knowledge the first example of a bacterial toxin that is 335 activated by F-actin, G-actin has been shown before to activate a bacterial toxin secreted by 336 the T3SS namely YopO/YpkA, a multidomain protein produced by pathogenic Yersinia ExoY-GFP with actin filaments in transfected NIH3T3 cells (Fig. 6). As many F-actin 357 binding proteins, ExoY can also weakly interact with actin monomers as indicated by the fact 358 that (i) ExoY is weakly activated (up to 5-10 % of maximal activity) in a dose dependent 359 manner by actin bound to the polymerization-inhibiting drug latrunculin A (Figs. 3b and 4) 360 and that (ii) ExoY weakly stimulated G-actin-ATP/-ADP polymerization in absence of 361 profilin (Fig. 5a,b). 362 While most of the ExoY related effects in infected cells likely depend on its catalytic 363 activity, the catalytically inactive ExoY K81M mutant has been observed to induce temporary 364 actin redistribution to the cell margins 10 as well as minimal intercellular gap formation 11 in 365 endothelial cells. These effects may be linked to a residual nucleotide cyclase activity of 366 ExoY K81M and/or to its direct binding to actin polymers in host cells. We showed that in vitro 367 Providencia, Burkholderia or Proteus (Fig. 7a). Here we showed that VnExoY-L, a rather 381 distantly related ExoY-like module from V. nigripulchritudo (38% sequence similarity with 382 P. aeruginosa ExoY and 28% with B. anthracis EF or B. pertussis CyaA, Fig. 7a and 383 Supplementary Table 3), was also strongly stimulated (over a 10,000 fold) by actin and 384 efficiently synthesized cAMP but not cGMP. The lack of guanylate cyclase activity is in 385 agreement with the results obtained with the V. vulnificus ExoY-like module 13 , a close 386 homolog of VnExoY-L (>75% sequence similarity, Fig. 7a and Supplementary Table 3), and 387 may thus reflect a more general difference regarding the nucleotide substrate specificities 388 between the P. aeruginosa ExoY and other ExoY-like proteins found in MARTX toxins like 389 those of the Vibrio genus ( Fig. 7a and Supplementary Fig. 6). 390 Actin may thus represent a common eukaryotic activator for a sub-group (Fig. 7a) of 391 the class II adenylyl cyclase toxin family (described in 16 ). This newly defined, actin-392 activated nucleotidyl cyclase (AA-NC) sub-family is also characterized by wider nucleotide 393 substrate specificity than the original class II members, the adenylate cyclase toxins CyaA 394 from B. pertussis and EF from B. anthracis. As calmodulin, the common cofactor of CyaA 395 and EF, actin is an abundant and highly conserved protein specific to eukaryotic cells. It 396 appears, therefore, to be a suitable molecular signal to indicate the arrival of the ExoY toxin 397 in the eukaryotic environment of the host target cells, where it should display its cyclic 398 nucleotide synthesizing activity. 399 Future studies should address the mechanism of activation of ExoY and ExoY-like 400 proteins by actin through structural analysis. This could eventually open several interesting 401 prospects in particular regarding the development of small molecules able to specifically 402 inhibit the activation of these toxins by actin, as a therapeutic approach against bacterial 403 infections, as well as the structural basis of the differential substrate selectivity of these AA-404

NCs. 405
Interestingly, Beckert et al. 7 have reported notable differences in accumulation of 406 various cNMPs in different cell lines exposed to P. aeruginosa ExoY, in particular, with 407 respect to the mode of delivery of the toxin (transfection versus infection). It will be 408 interesting to examine whether the relative efficacy in synthesizing different cNMPs depends 409 solely on availability of substrates or whether the actin dynamics and turnover in cells may 410 also play a role. 411 412 413

Strains, plasmids and growth conditions 415
Strains, plasmids and growth conditions are described in table S1. 416 417

Purification of ExoY, actin from rabbit skeletal muscle, and actin-binding proteins 418
ExoY-FH and VnExoY-L-FH were purified by nickel affinity chromatography under 419 denaturing conditions (in the presence of 8M urea) from the non-soluble protein fraction 420 obtained from 1 liter cultures of E. coli BLR (pUM460) or (pUM522), respectively. 421 Proteins were expressed from the λP L promoter controlled by the temperature sensitive cI 422 repressor (cI857), which was induced by shifting the temperature from 30ºC to 42ºC. were stopped by the addition of 450 μl stop solution (20 mM HEPES pH 7.5, 20 mM 531 EDTA, 0.5 mM cAMP) and the mixtures were filtered on Al 2 O 3 columns, which included 532 3 washes with 1 ml 20 mM HEPES pH 7.5 each to separate nucleotide substrates that were 533 retained in the columns from cyclic nucleotides present in the filtrates. Filtrates were 534 collected in 20 ml scintillation vials. 16 ml scintillation liquid (HiSafe3, Perkin Elmer) 535 were added before measuring 33 P in a TriCarb scintillation counter (Perkin Elmer). All 536 reactions were performed in duplicates. Differences between cpm values of most 537 duplicates were around or less than 10%. Standard deviations between duplicates are 538 indicated by error bars. 539 Muscle actin 99% pure (designated MA-99) from rabbit skeletal muscle 540 (Reference AKL99), or 99% pure non-muscle actin from human platelets (Reference 541 APHL99, designated A-99) was obtained from Cytoskeleton, Inc. Alternatively, we used 542 actin from rabbit skeletal muscle prepared in one of our laboratories (designated MA-L) 543 according to the procedure described above. For activity assays, all actin solutions were 544 diluted in G-buffer supplemented with BSA at 0.1 mg/ml. 545 Preliminary experiments to optimize reaction conditions showed that ExoY-FH 546 was most active at pH values between 8 and 9 and in Tris as compared to HEPES or Na-547 phosphate and shows a broad optimal NaCl concentration (between 100 and 300 mM 548 NaCl). 549 Extracts from HeLa cells for activation of ExoY were prepared as follows: Cells 550 grown in Dulbecco's modified Eagle's medium (DMEM) + 10% fetal bovine serum were 551 harvested after reaching 75% of confluence as follows: One wash with PBS was followed 552 by incubation in 10 ml PBS containing 0.01M EDTA for 5 min at 37 ºC before detaching 553 the cells by gentle tapping of the flasks. Cells were collected by centrifugation and washed 554 3 times in PBS. The cell pellet was resuspended in 2 ml of lysis buffer [50 ml Tris pH 7.5, 555 300 mM NaCl, 0.5 % NP50, complete EDTA-free protease inhibitor cocktail (Roche)] per 556 ml of cell pellet volume, after which the cells were snap frozen in liquid nitrogen and 557 stored at -80 ºC or processed immediately. Frozen cells were allowed to thaw on ice, 558 rotated at 4 ºC for 20 min and centrifuged for 1 h at 18,000 rpm in a SS34 rotor (Sorval). 559 The supernatant was centrifuged at 100,000 rpm at 4 ºC for 1 h in a TLA-110 rotor 560 against G-buffer to remove the KCl present in the storage buffer before adding 5 μl at 605 142.7 μM directly to undiluted actin (6 μl MA at 55.88 μM in G-buffer), incubated at 606 room temperature for 10 min and diluted by adding 14 μl G-buffer. Actin was not 607 converted into Mg-ATP-actin. 25 μl fresh F2 buffer was added and 11.2, 7.5, or 5.6 μl of 608 this mixture were combined with G-buffer to a total volume of 15 μl and used 609 immediately in activity assays. At final actin concentrations of 1.5, 1 and 0.75 μM, 610 profilin was present at 3, 2, and 1.5 μM, respectively. 611 Latrunculin A was purchased from tebu-bio (produced by Focus Biomolecules). 612 Studies with latrunculin were done similarly to those on profilin except that higher 613 concentrations of actin (between 5.25 and 1.0 μM final MA-L) were used, ME buffer (10 614 fold concentrated) was added to control reactions before polymerization but was added as 615 well to reactions containing latrunculin after combining latrunculin and actin and 10 min 616 incubation at room temperature. Samples containing latrunculin were incubated 2 hours at 617 room temperature as the corresponding controls. The mixture of lartunculin and actin was 618 diluted to different concentrations in G-buffer containing latrunculin to ensure a finale 619 concentration of 10 μM in all reactions. Control reactions lacking latrunculin contained 620 DMSO at concentrations equivalent to that introduced with latrunculin. 621 Sedimentation-assays to verify effectiveness of profilin or latrunculin in preventing 622 polymerization were performed at 2 or 1 μM actin (MA-L) in the exact same way as for 623 measuring activity except that GTP was not spiked with [α-33 P] GTP. 624 625