Molecular insights into Vibrio cholerae’s intra-amoebal host-pathogen interactions

Vibrio cholerae, which causes the diarrheal disease cholera, is a species of bacteria commonly found in aquatic habitats. Within such environments, the bacterium must defend itself against predatory protozoan grazers. Amoebae are prominent grazers, with Acanthamoeba castellanii being one of the best-studied aquatic amoebae. We previously showed that V. cholerae resists digestion by A. castellanii and establishes a replication niche within the host’s osmoregulatory organelle. In this study, we deciphered the molecular mechanisms involved in the maintenance of V. cholerae’s intra-amoebal replication niche and its ultimate escape from the succumbed host. We demonstrated that minor virulence features important for disease in mammals, such as extracellular enzymes and flagellum-based motility, play a key role role in the replication and transmission of V. cholerae in its aqueous environment. This work, therefore, describes new mechanisms that provide the pathogen with a fitness advantage in its primary habitat, which may have contributed to the emergence of these minor virulence factors in the species V. cholerae.

The severe diarrheal disease cholera is not extinct. Seven cholera pandemics have been 28 recorded in modern history, most notably in developing countries, and the latest is still 29 ongoing 1,2 . An estimated three million new cases of cholera occur every year leading to 30 almost 100,000 annual deaths 3,4 . Cholera is caused by ingestion of the bacterium Vibrio 31 cholerae. Toxigenic strains of this species are capable of damaging the host due to the 32 presence of so-called virulence factors, which refers "to the elements (i.e. gene products) that 33 enable a microorganism to colonize a host niche where the organism proliferates and causes 34 tissue damage or systemic inflammation." 5 Bacterial strains without these factors are usually 35 attenuated with respect to the infection process (in human or animal models). 36 For V. cholerae, the two major virulence factors, cholera toxin and the toxin-coregulated 37 pilus, play a pivotal role in the infection process, but additional minor virulence factors have 38 also been identified. These include factors such as outer membrane porins, a pore-forming 39 hemolysin, diverse proteases, N-Acetyl-glucosamine binding protein (GbpA), flagellum-40 based motility, to name a few 6-9 . The hemagglutinin/protease (HA/P or HapA), for example, 41 is a zinc-metalloprotease 10 , which was first identified due to its mucinase activity 11 . This 42 enzyme was later demonstrated to not only cause hemagglutination but to also hydrolyze 43 fibronectin, mucin, and lactoferrin, all of which were thought to contribute to the host defense 44 against V. cholerae 12 . HapA also causes cell rounding, loss of the barrier function of the 45 epithelial layer, and, ultimately, detachment of the cells under tissue culture conditions 13,14 . 46 Consistent with these in vitro activities, a 10-fold increase in the 50% lethal dose (LD 50 ) in 47 the absence compared to the presence of the HapA protease for V. cholerae strains that 48 otherwise lack cholera toxin was reported 15 . It was therefore suggested that HapA participates 49 in the attachment as well as in the detachment of V. cholerae from the gut epithelium 12,16 and 50 that tissue destruction through HapA-mediated cleavage of surface-exposed proteins may 51 contribute to fluid leakage and the stimulation of proinflammatory responses 7 . 52 To date, a limited number of studies have shown that minor virulence factors, such as 77 those mentioned above, are also advantageous in an environmental context 26,27 . Indeed, 78 despite V. cholerae being an aquatic bacterium that is well adapted for survival in this 79 environment 28 , the molecular details about V. cholerae's environmental lifestyle are still 80 lacking. To address this knowledge gap, we aimed to elucidate the molecular mechanisms 81 that V. cholerae would use as part of its environmental lifestyle, such as its interactions with 82 grazing amoebal predators. In this context, we recently demonstrated that V. cholerae 83 survives predation by Acanthamoeba castellanii, an aquatic species of amoebae, through the 84 evolution of two distinct phenotypes. Firstly, upon phagocytosis by the feeding amoebal 85 trophozoite, a fraction of the ingested V. cholerae population can resist amoebal digestion, 86 then exit the phagosomal pathway by exocytosis. Notably and in contrast to a previous 87 study 29 , we never observed free V. cholerae cells in the cytosol of such trophozoites 30 , 88 indicating that the bacteria are not released from food vacuoles intracellularly. Secondly, the 89 pathogen colonizes the amoeba's contractile vacuole, which is its osmoregulatory organelle 90 and, therefore, of prime importance for the survival of A. castellanii. We showed that the 91 bacterium proliferates within this replication niche, even upon amoebal encystement, before 92 it ultimately lyses the host 30 . Here, we address the underlying molecular mechanisms of this 93 host-pathogen interaction. We show that V. cholerae fine-tunes extracellular enzymes to 94 avoid premature intoxication of its host, allowing the bacterium to take full advantage of the 95 intra-amoebal replication niche for undisturbed and non-competitive proliferation before 96 finally lysing its amoebal host. We also demonstrate that flagellum-based motility is of prime 97 importance for V. cholerae to ultimately escape the ruptured host niche to return to the 98 aquatic environment. As the 3D reconstruction ( Fig. 1b and Movie S1) suggested that V. cholerae cells were 142 densely packed within the vacuole at this later point of the infection, we wondered how the 143 bacteria eventually escaped the vacuole after amoebal encystation and how the timing was 144 properly regulated. To answer these questions, we first took an educated guess approach and 145 tested a few diverse knockout mutants of V. cholerae for their intra-amoebal behavior. We 146 were especially interested in strains that lacked certain extracellular enzymes, as some of 147 those were known to be important in the pathogen's intra-human lifestyle and its transmission 148 to new hosts. 149 We therefore challenged A. castellanii with a HapA-deficient V. cholerae strain (Table S1) 150 and compared the amoebal response to the WT-challenged condition. By doing this, we 151 observed that a vast majority (79.8%) of infected amoebal cells showed an aberrant 152 morphology upon co-incubation with the hapA-minus strain for 20 hours, which occurred 153 significantly less often for WT-infected amoebae (11.6%, Fig. 2a and b). These 154 morphological abnormalities ranged from a shrinking or compacting phenotype towards 155 pseudopodia retraction and detachment, all of which ultimately abolished amoebal grazing 156 ( Fig. 2 and Fig. S1). Complementation assays, involving a genetically engineered derivative 157 of the hapA-minus strain with a new copy of hapA on the large chromosome (Fig. S2), were 158 performed using the amoebal infection protocol and fully restored the WT-infected amoebal 159 morphologies ( Fig. 2a and b). 160 Seeing these striking differences between WT-infected and hapA-minus-infected 161 amoebae, we wondered whether the latter might cause "amoebal constipation", meaning the 162 accumulation of undigested phagosomal V. cholerae without efficient exocytosis. Indeed, 163 based on the confocal microscopy images, we could localize the accumulated bacteria due to 164 their green fluorescence. However, recognition of the contractile vacuole in the transmitted 165 light channel was often difficult for hapA-minus mutant-infected amoebae due to their 166 compaction and malformation (Fig. 2a). Therefore, it seemed possible that the bacteria were 167 contained in digestive food vacuoles of the endosomal pathway, blocking amoebal digestion 168 as opposed to being localized to the contractile vacuole, as is the case for WT V. cholerae. To 169 distinguish between both scenarios (localization within a digestive vacuole or within the 170 contractile vacuole), we labeled the amoebal endosomal pathway with fluorescently labeled 171 dextran. As the dextran-and bacteria-derived fluorescent signals did not overlap, these 172 experiments confirmed the intra-contractile vacuolar localization of the hapA-minus strain 173 and excluded its massive accumulation within dextran-labeled digestive food vacuoles in 174 intoxicated amoebae (Fig. 2c).   Fig. 2e), the ability to form cellulose-positive cysts was severely impaired in 187 amoebae infected with the hapA-minus mutant strain. This impaired encystation contrasted 188 greatly to that of both the WT-infected A. castellanii and the complemented hapA-minus 189 strain ( Fig. 2d and e). We therefore concluded that the absence of the HapA protease leads to 190 premature amoebal intoxication and, consequently, a defect in the encystation process. 191 To better understand the aberrant morphologies and the defect in the encystation process 192 of the amoebae infected by the mutant strain, we again used CLEM to obtain high-resolution 193 images of colonized amoebae. Impaired membrane integrity around the contractile vacuole in 194 amoeba infected with hapA-minus V. cholerae was seen from the TEM images (Fig. 2f). The 195 WT-infected contractile vacuoles seemed to maintain their internal pressure, leading to a 196 clear separation between the content of the contractile vacuole and the amoebal cytosol ( Fig.  197 1a and 2f), though this evident distinction was lacking in host cells that were infected by the 198 mutant V. cholerae strain (Fig. 2f).

Uncontrolled hemolysis leads to premature amoebal intoxication 202
We speculated that the seemingly collapsed contractile vacuole membranes might have been 203 due to amoebal membrane rupture caused by the hapA-minus mutant strain. Given that 204 bacterial pathogens often cause membrane rupture through the secretion of pore-forming 205 toxins 34 , we wondered whether the secreted hemolysin of V. cholerae (HlyA) might be 206 involved in the observed amoebal intoxication. Consistent with this idea was a study by Tsou 207 and Zhu that showed that the HapA protease degrades HlyA in an in vitro assay 35 . This led us 208 to hypothesize that HapA protease-deficient strains of V. cholerae would display enhanced 209 hemolysis, which, ultimately, could cause the observed amoebal intoxication. To test this 210 hypothesis, we generated several V. cholerae strains that lacked hapA, hlyA, or both genes 211 simultaneously (Table S1) and tested them for hemolysis and proteolysis on blood and milk 212 agar plates, respectively. As shown in Fig. S2, we observed a significant correlation between 213 the absence of HapA-mediated protease activity and the presence of hemolysis. There was a 214 strong increase in hemolysis in strains lacking hapA, while the absence of hlyA fully 215 abolished this activity. Complementation assays, in which the mutant strain contained a new 216 copy of the missing gene elsewhere on the chromosome (Table S1), restored the system to 217 that of the WT ( Fig. S2a and b). Interestingly and consistent with the extracellular 218 localization of both enzymes, we showed that the secreted HapA protease from a WT strain 219 was also able to inactivate HlyA that was released by a co-cultured hapA-minus mutant strain 220 (Fig. S2). However, when we engineered an hlyA-overexpression strain (Table S1 and Fig.  221 S2), the protease activity exerted by the strain itself or from the co-cultured WT bacteria was 222 insufficient to abolish hemolysis (Fig. S2). 223 With these newly constructed strains in hand, we then tested their effect on A. castellanii. 224 Consistent with our hypothesis that a hapA mutant of V. cholerae would possess enhanced 225 hemolysin activity, which, ultimately, would result in amoebal intoxication, we found that the 226 absence of hlyA in the protease-minus background restored normal host morphology and 227 allowed cyst formation comparable to that observed for WT infection ( Fig. 3a and b and Fig.  228 S3a to d). Infection of A. castellanii by the V. cholerae double mutant (ΔhapAΔhlyA) also 229 resulted in an intact integrity of the contractile vacuole membrane ( Fig. 3c and Fig. S3e). 230 Overproduction of HlyA in a WT background strain, however, fully phenocopied or even 231 aggravated the protease-minus phenotypes ( Fig. 3a and b and Fig. S3a to d). These data, 232 therefore, support the notion that HlyA is causing amoebal intoxication and that the HapA  Table S2). This rescuing 243 phenotype was not observed when the second strain was the HlyA-overproducing strain ( To further support these findings and the interplay between the HapA protease and the 250 HlyA hemolysin, we generated a gene encoding a translational fusion between HlyA and 251 superfolder green fluorescent protein (sfGFP) 36 and used this allele to replace the endogenous 252 copy of hlyA. The resulting strain and a corresponding hapA-minus derivative (both also 253 containing a constitutively expressed dsRed gene on their chromosome; Table S1) were used 254 to confirm the functionality of the fusion construct, as judged by the hemolysis of blood cells 255 (Fig. S5a). Next, we infected amoebae with these strains. Due to the constitutively produced 256 red fluorescence we had a straightforwardly method for detecting colonized contractile 257 vacuoles. We further witnessed that green fluorescence from the HlyA-sfGFP fusion was 258 hardly detectable in the WT background, which contrasted with the protease-minus strain 259 points, it remained to be discovered how the pathogen eventually escaped from its amoebal 268 host to return to the environment. We reasoned that the pathogen would need to disrupt the 269 host's plasma membrane. The plasma membrane of A. castellanii contains lecithin, which is a 270 mixture of glycerophospholipids that includes a high percentage of phosphatidylethanolamine 271 and phosphatidylcholine 37 . We therefore speculated that V. cholerae might use its 272 lecithinase 25 to disrupt this membrane. We generated a lec-minus strain and confirmed its 273 impaired lecithinase activity on egg yolk plates (Fig. S6). We also complemented the lec-274 minus strain by placing a new copy of lec that was preceded by its native promoter onto the 275 chromosome (Δlec::lec). The same genetic constructs were also added into the WT 276 background to generate a lec merodiploid strain of V. cholerae (WT::lec). All of the strains 277 were tested for their lecithinase activity in vitro and behaved as expected (Fig. S6). 278 Next, we infected A. castellanii with this set of genetically engineered strains. We 279 observed that while the lec-deficient strains were able to colonize the contractile vacuole and 280 escape from it in the cyst stage, the bacteria were unable to ultimately lyse the amoebal host, 281 as visualized through the exclusion of extracellular dextran (Fig. 4). The complemented strain 282 had its cyst lysis capability restored, a phenotype that was enhanced for the lec merodiploid 283 bacteria (Fig. 4). We therefore concluded that the lecithinase enzyme indeed targets and 284 permeabilizes the plasma membrane of the host, thereby triggering the death of the cyst. 285

V. cholerae escapes the lysed amoebal compartments through flagellum-based motility 287
As we observed that V. cholerae is highly motile within the amoebal host (Movie S2), we 288 wondered whether this motility contributed to the pathogen's intra-amoebal lifestyle and its 289 escape from the succumbed host. We therefore generated non-motile mutants of V. cholerae 290 by deleting either the gene that encodes the major flagellin subunit FlaA or the flagellar 291 motor protein PomB (Table S1). While the first approach resulted in non-flagellated bacteria, 292 the latter approach led to rotation-deficient but fully flagellated bacteria (Fig. 5A). The 293 motility deficiency of the mutants was further confirmed in in vitro motility assays (Fig. S7). 294 Next, we infected A. castellanii with these mutant strains. While both mutant strains still 295 infected the amoebal contractile vacuole, the strains were more static within this niche (Fig.  296 5b and Movies S3 and S4). We therefore wondered whether motility would play a role in the 297 escape of V. cholerae from the lysed contractile vacuole and lysed cysts. To test this idea, we 298 infected amoebae with the WT and the flaA-minus strains, which we labeled with different 299 fluorescent proteins, namely dsRed and GFP, respectively. We then took time-lapse movies 300 over several hours to observe the colonizing bacteria of those amoebae that contained both 301 strains within the same contractile vacuole ( Fig. 5c and Movie S5). These experiments 302 showed that, compared to the WT strain, the non-motile mutant was retained in the lysed 303 contractile vacuole (Movie S6) and its escape back to the environment was severely impaired 304 after the final lysis of the cyst (Movie S7). We concluded, therefore, that motility plays a 305 Discussion 308 V. cholerae, the bacterial agent responsible for cholera, still poses a global threat to human 309 health. However, apart from its chitin-induced phenotypes, which include chitin catabolism, 310 inter-bacterial competition, and horizontal gene transfer 38 (reviewed by 39-42 ), we know very 311 little about its environmental lifestyle and its potential interactions with non-human hosts. 312 Such ancient host-pathogen interactions are, however, often considered as evolutionary 313 precursors to modern interactions that occur between bacteria and their human hosts 43 . 314 Moreover, aquatic predators are recognized for their contribution to pathogen emergence due 315 to the selection pressure they exert on their bacterial prey 44 . Here, we examined the molecular 316 mechanisms that V. cholerae uses to interact with the aquatic amoeba A. castellanii and to 317 maintain a favorable replication niche within the amoebal osmoregulatory organelle. Based 318 on the molecular checkpoints that were deciphered in the current study, we expanded our 319 model of the pathogen's intra-amoebal lifecycle (Fig. 6). Specifically, we showed the 320 importance of several extracellular enzymes in this host-pathogen interaction. The production 321 of the HapA protease, which cleaves the pore-forming hemolysin toxin, proved essential for 322 avoiding premature intoxication of the amoebal host (Fig. 6a). In contrast to intracellular 323 pathogens, such as Listeria monocytogenes or Shigella flexneri, that use pore-forming toxins 324 to escape from acidic vacuolar compartments to reach the host cell's cytosol, the situation 325 described herein is very different. In this case, the host-pathogen interaction relies on V. 326 cholerae residing in a non-digestive vacuole in which it can readily replicate 30 . This non-327 digestive contractile vacuole is an essential osmoregulatory organelle of the amoeba meaning 328 that HlyA-mediated rupture of the vacuolar membrane would, therefore, release V. cholerae 329 into the cytosol, though, at the expense of rapid host cell death. We speculated, therefore, that 330 this HapA-mediated disintegration of HlyA might have evolved to allow V. cholerae to 331 maximize its growth output within this intra-amoebal replication niche by avoiding the 332 premature death of its host. Notably, the HapA protease is considered a minor virulence 333 factor that contributes to disease outcome in animal models of cholera 15,45 , along with HlyA 334 itself, as described above. In this study, we did not observe any obvious phenotype for V. to ultimately kill the host and return the bacteria to the environment (Fig. 6e). This lecithinase 358 is encoded in close proximity to the hemolysin gene, a region that was previously speculated 359 to represent a pathogenicity island 23 . Unfortunately, apart from the absence of a phenotype 360 for the lec-minus mutant in rabbit-ligated ileal loops, a model of cholera that primarily aims 361 at judging cholera toxin-mediated fluid accumulation 25 , this enzyme had not been extensively 362 studied in vivo (for current 7 th pandemic O1 El Tor strains). However, a recent study based on 363 an activity-based protein profiling method showed that the lecithinase/phospholipase is active 364 in cecal fluids of infected infant rabbits 48 . Further studies are therefore required to 365 conclusively determine whether this enzyme should also be considered as a minor virulence 366 factor for animals, as we demonstrated here for amoebae. 367 Lastly, we showed that bacterial motility is required to efficiently escape from the lysed 368 contractile vacuole and the succumbed host (Fig. 6e), which, ultimately, allows the bacteria 369 to return to the aquatic environment. One could speculate that the lysed cyst might generate a 370 nutrient gradient that attracts novel predators, a disadvantageous scenario for any prey. 371 Indeed, we frequently observed that dead cysts were readily ingested by feeding trophozoites, 372 which could counter-select for non-motile mutants that cannot escape from the cellular 373 debris. Likewise, motility has also been demonstrated to be of importance for virulence in 374 animal models. Indeed, non-motile mutants exerted less fluid accumulation in rabbit ileal 375 loops 45 and were severely attenuated for the colonization of infant mouse small intestines 6 . 376 These correlations between the contribution to virulence in animal models and in the 377 environmental host-pathogen interaction described herein therefore support the "coincidental 378 evolution hypothesis". This hypothesis suggests that "virulence factors result from adaptation 379 to other ecological niches" and, in particular, from "selective pressure exerted by protozoan To generate a 3D model of the colonized contractile vacuole, serial TEM images were first 499 aligned manually using image software (Photoshop, Adobe). Each structure was then 500 segmented using TrakEM2 65 operating in the FIJI software 66 (www.fiji.sc). The 3D models 501 were then exported into the Blender software (www.blender.org) for smoothing 502 and rendering into a final image or movie. 503 504

Determination of enzymatic activities 505
Semi-quantitative assays in specific media were used to determine enzymatic activities. To 506 perform these assays, 5µl of the respective overnight bacterial cultures were spotted onto the 507 following three types of agar plates. (i) Milk agar was used to determine protease activity due 508 to the hydrolysis of casein. The modified recipe used in this study contained 0.5% tryptone, 509 0.25% yeast extract, 0.1% dextrose, 1% skim milk powder, and 1.25% agar. The bacteria 510 were incubated on these plates for 24 hours at 30ºC. (ii) Trypticase Soy Agar II with 5% 511 sheep blood (BD, Heidelberg, Germany) was used to test for hemolysis (24 hours at 30ºC). 512

Data availability 527
All data supporting the findings of this study are available from the corresponding author 528   Table S1. Bacterial strains and plasmids used in this study. 893 Table S2. Quantification of bacterial hemolysin-triggered aberrant amoebal morphologies.  Shown are a merged image of the transmitted light channel and the green channel (left) and the green channel image alone (second from left). After staining of the sample, the same amoeba was imaged at high resolution using transmission electron microscopy (TEM; right images). Scale bar in all images: 5 µm. (b) 3D reconstruction of the colonized contractile vacuole. The region containing the amoeba shown in panel (a) was serially thin sectioned (50 nm thickness) and serial images were taken with the TEM. These images were then aligned to generate a 3D model of the colonized amoeba. Shown are snapshots of the resulting 3D reconstruction movie (Movie S1).