Pseudomonas aeruginosa isolates co-incubated with Acanthamoeba castellanii exhibit phenotypes similar to chronic cystic fibrosis isolates

The opportunistic pathogen, Pseudomonas aeruginosa, is ubiquitous in the environment, and in humans is capable of causing acute and chronic infections. P. aeruginosa, when co-incubated with the bacterivorous amoeba, Acanthamoeba castellanii, for extended periods, produced genetic and phenotypic variants. Sequencing of late-stage amoeba-adapted P. aeruginosa isolates demonstrated single nucleotide polymorphisms within genes that encode known virulence factors, and this correlated with a reduction in expression of virulence traits. Virulence towards the nematode, Caenorhabditis elegans, was attenuated in late-stage amoeba-adapted P. aeruginosa compared to early stage amoeba-adapted and non-adapted counterparts. Late-stage amoeba-adapted P. aeruginosa lost competitive fitness compared to non-adapted counterparts when grown in nutrient rich media. However, non-adapted P. aeruginosa were rapidly cleared by amoeba predation, whereas late-stage amoeba-adapted isolates remained in higher numbers 24 h after ingestion by amoeba. In addition, there was reduced uptake by macrophage of amoeba-adapted isolates and reduced uptake by human neutrophils as well as increased survival in the presence of neutrophils. Our findings indicate that the selection imposed by amoeba on P. aeruginosa resulted in reduced virulence over time. Importantly, the genetic and phenotypic traits possessed by late-stage amoeba-adapted P. aeruginosa are similar to what is observed for isolates obtained from chronic cystic fibrosis infections. This notable overlap in adaptation to different host types suggests similar selection pressures among host cell types. Author Summary Pseudomonas aeruginosa is an opportunistic pathogen that causes both acute infections in plants and animals, including humans and also causes chronic infections in immune compromised and cystic fibrosis patients. This bacterium is commonly found in soils and water where bacteria are constantly under threat of being consumed by the bacterial predators, protozoa. To escape being killed, bacteria have evolved a suite of mechanisms that protect them from being consumed or digested. Here we examined the effect of long-term predation on the genotype and phenotypes expressed by P. aeruginosa. We show that long-term co-incubation with protozoa resulted in mutations in the bacteria that made them less pathogenic. This is particularly interesting as we see similar mutations arise in bacteria associated with chronic infections. Thus, predation by protozoa and long term colonization of the human host may represent similar environments that select for similar losses in gene functions.


Biofilm formation and planktonic growth of isolates from adapted populations 157
As flagella and pili also impact biofilm formation, biofilm biomass and planktonic growth rates 158 of adapted and non-adapted isolates were compared. Co-incubation of P. aeruginosa with A. 159 castellanii had a significant effect on P. aeruginosa surface colonization (Fig 3a; F2, 354 = 15.7, p 160 < 0.001). Post hoc analysis revealed no difference between treatments after 3 d (p = 0.998). 161 However, P. aeruginosa from amoeba-adapted day 24 populations formed 10-fold less biofilm 162 than isolates from non-adapted day 24 populations (p < 0.001). Although the average biomass 163 of biofilms formed by the amoeba-adapted isolates increased after 42 d of co-incubation with 164 amoeba, biofilm biomass remained 2-fold lower than that of the non-adapted population (p < 165 0.001). Additionally, the presence of amoeba exerted a strong negative effect on the planktonic 166 growth of P. aeruginosa isolates derived from the amoeba-adapted population (Fig 3b; F1, 354 = 167 29.6, p < 0.001). The planktonic growth rate of P. aeruginosa after 3 d was the same regardless 168 of the population (p = 0.56). However, after 24 and 42 d of amoebal-driven selection the 169 planktonic growth rate of amoeba-adapted derived isolates was significantly less than the non-170 adapted isolates (p < 0.05 and p < 0.001 respectively). 171 Quantification of pyoverdine and rhamnolipids 177 P. aeruginosa isolates obtained after co-incubation with A. castellanii produced reduced 178 quantities of pyoverdine compared to isolates from non-adapted populations (Fig 4a; F1, 174 = 179 45.74, p < 0.001). Although pyoverdine production was reduced in both amoeba-adapted and 180 non-adapted populations (F2, 174 = 12.08, p < 0.001), the concentration of pyoverdine in 181 supernatants from amoeba-adapted isolates from 3 d populations was reduced 2-fold compared to 182 non-adapted isolates (p < 0.001) and was further reduced after 24, and 42 days of selection (p < 183 0.001). 184 Rhamnolipid production varied within the 42 d amoeba-adapted and non-adapted isolates but 185 amoeba-adapted isolates produced less rhamnolipid overall when compared with the non-adapted 186 population (Fig 4b;  To investigate whether adaptation with amoeba confers a fitness advantage to P. aeruginosa 198 when grown with amoeba, we mixed fluorescent-tagged amoeba-adapted and non-adapted 199 isolates and grew them together with amoeba. After 48 h co-incubation with amoeba, the 200 proportion of amoeba-adapted cells is always higher when both amoeba-adapted::GFP (Fig 5a;  201 F1,4 = 95.27, p = 0.000617) and amoeba-adapted::mCherry (Fig 5b; F1,4 = 11.85, p = 0.0262) are 202 competed with the reciprocally tagged non-adapted strain, compared to no amoeba controls. The fluorescence ratios of Day 42 + A::GFP P. aeruginosa mixed with -A::mCherry (a,b,c) and 206 + A::mCherry with -A::GFP (d,e,f) after 48h of incubation with (black bars, b, e) and without 207 (grey bars, c, f) A. castellanii. Data are presented as Means ± SEM. * p < 0.05, *** p < 0.001. 208 Growth of amoeba-adapted and non-adapted P. aeruginosa in media or media 209 supplemented with amoeba supernatant 210 To investigate whether the amoeba-adapted strains were utilizing amoeba secretions to out-211 compete the non-adapted strains, 9 randomly selected amoeba-adapted and non-adapted isolates 212 were grown with and without the addition of amoeba supernatant to the growth media. The 213 addition of amoeba supernatant to the growth media resulted in specific growth rate increases of 214 -0.02 to 0.04 by 42 d amoeba-adapted and non-adapted P. aeruginosa isolates compared to the 215 minimal media supplemented with the same amount of glucose. There was no significant 216 difference in growth between amoeba-adapted and non-adapted isolates when grown in amoeba 217 supernatant (Fig 6;  Since the enhanced fitness of amoeba-adapted isolates in the presence of amoeba was not due to 227 increased growth rate, the intracellular survival of 42 d amoeba-adapted and non-adapted 228 populations were determined using a modified gentamicin protection assay. Intracellular CFUs 3 229 h after infection of non-adapted isolates within amoeba were higher than amoeba-adapted CFUs, 230 however, after 24 h the numbers of surviving intracellular non-adapted cells had decreased and 231 were comparable to the amoeba-adapted numbers (Fig 7a; Adaptation×Time F1,32 = 14, p < 232 0.001). The same trend was observed when the assay was conducted with raw 264.7 233 macrophages. There was a significant interaction of amoeba adaptation and incubation time ( changes and appeared similar to those infected with the wild type strain (Fig 7c). Propidium 242 iodide staining showed that many of these macrophages were dead. In contrast, macrophage 243 infected with amoeba-adapted P. aeruginosa exhibited a more normal morphology, with fewer 244 cells taking up the propidium iodide stain. non-adapted populations were determined using a modified gentamicin protection assay. 258 Intracellular CFUs 2 h after infection of non-adapted isolates within neutrophils were higher than 259 amoeba-adapted CFUs (Fig 8a). Amoeba-adapted strains also had increased survival against 260 human neutrophils when compared to non-adapted isolates (Fig 8b). Most opportunistic pathogens are not transmitted person to person but rather transit through the 288 environment between hosts and therefore, it is likely that the environment plays a significant role 289 in evolution of protective traits. Predation by protists is one of the major mortality factors for 290 bacteria in the environment and it is likely that traits that protect against predation may also 291 impact human hosts during infection (1-3). P. aeruginosa is responsible for a variety of 292 nosocomial acute as well as chronic infections (15) in particular chronic lung infections in CF 293 patients (16). Here we investigated the phenotypic and genotypic changes that occur during co-294 incubation of P. aeruginosa with A. castellanii for 42 days with a focus on virulence traits. 295 Populations were collected on days 3, 24 and 42 and individual isolates were randomly collected 296 at each time point. 297

Motility, biofilm formation and secretion of secondary metabolites 298
Sequencing of populations obtained from the day 42 co-incubation revealed 54 and 65 nsSNPs in 299 adapted and non-adapted populations, respectively. Gene functions of nsSNPs in coenzyme 300 metabolism, lipid metabolism, cell motility and secondary metabolite production occurred solely 301 within the amoeba-adapted population, while COGs representing inorganic ion transport and 302 energy production were over-represented in the non-adapted population (Fig 1). Phenotypic 303 assays of random isolates from the adapted and non-adapted populations confirmed loss of 304 function in many traits associated with these genes. For example, amoeba-adapted isolates 305 showed reductions in twitching and swimming motility from days 3 to 42 and a decrease in 306 swarming motility for day 3 and 24 isolates, but not for the day 42 isolates (Fig 2). In addition, 307 there was a decrease in planktonic growth rate and biofilm formation by amoeba-adapted isolates 308 20 compared to non-adapted isolates (Fig 3). It is possible that the loss of biofilm is a result of 309 selection against flagella or adhesive pili as these surface structures are immunogenic and act as 310 ligands for phagocytes, including amoeba (21, 22). In fact, a large number of P. aeruginosa 311 isolates from CF patients also lack flagella and pili and these mutants are resistant to 312 phagocytosis by macrophage (23). 313 There were also a number of nsSNPs in genes relating to secretion of secondary metabolites in 314 the adapted population. Phenotypic assays revealed that pyoverdine secretion was significantly 315 reduced in the adapted isolates as early as day 3 and remained low throughout the later time 316 points (Fig 4). Loss of pyoverdine in P. aeruginosa in chronic cystic fibrosis infections has also 317 been reported (24). In addition, we observed an overall decrease in production of rhamnolipids, 318 however, this was only a subset of the isolates tested. Unlike other traits that were largely or 319 completely lost, decreases in rhamnolipid secretion were variable, implying that there may be a 320 weaker selective pressure against this phenotype during co-adaptation. 321

Uptake by and survival in amoeba and immune phagocytes 322
Results presented here reveal that amoeba-adapted isolates are taken up less readily and have 323 increased intracellular survival in A. castellanii and RAW264.7 macrophage cells compared to 324 non-adapted isolates, although the exact mechanisms are not known (Fig 5 and 7). Similarly, we 325 saw reduced internalization of amoeba-adapted isolates by human neutrophils (Fig 8). The loss 326 of immunogenic flagella and pili may contribute to reduced uptake (25), as well as a reduction in 327 chemotaxis due to mutations in chemotaxis genes in all amoeba-adapted populations. 328 21 Protozoa have been reported to release dissolved free amino acids and other nutrients when 329 grazing on bacterial prey (26). A reduction in chemotaxis would decrease the encounter of 330 predator and prey since normally P. aeruginosa is attracted to amoeba where it attaches to the 331 surface of the amoeba. Likewise changes in LPS also contribute to host avoidance, enhancing 332 fitness of amoeba-adapted strains in the presence of amoeba or macrophage. Interestingly, 333 PA3349 encoding a protein necessary for flagella-mediated chemotaxis was mutated in all 334 amoeba-adapted lineages and is needed for acute but not for chronic P. aeruginosa infections 335 (27, 28). 336 The finding that fitness gains from adapting to amoeba could be similarly conferred to 337 interactions with macrophage cells, perhaps demonstrates the overlaps in traits used by P. 338 aeruginosa to interact with both host types. Much work has been done to show that specialist 339 intracellular pathogens such as Legionella pneumophila can evade the defenses of both amoeba 340 and macrophage cells during endocytosis in order to form an intracellular replication niche (29). 341 P. aeruginosa is more of an environmental generalist than an intracellular pathogen and has not 342 been observed to form intracellular replication vacuoles, even though it has been recovered from 343 within environmental amoeba (30). In addition, intracellular replication has been described in 344 non-phagocytic epithelial cells and is dependent on a functional type-3 secretion system (31). In 345 the amoeba co-adapted isolates, there was a general decrease rather than increase in intracellular 346 cell numbers, which indicates that it is unlikely that P. aeruginosa is replicating intracellularly 347 under the conditions used here.

P. aeruginosa populations co-incubated with A. castellanii show reduced virulence 357
In order to investigate how constant predation pressure affects P. aeruginosa phenotypic and 358 genotypic traits, we adapted P. aeruginosa to amoeba for 42 days. Interestingly, the amoeba-359 adapted isolates were less virulent in both fast and slow-kill C. elegans assays (Fig 9). Although 360 the initial 3-day virulence levels were comparable, co-adaptation with amoeba for 42 days 361 resulted in a loss of virulence against C. elegans by both fast kill and slow kill mechanisms. Fast 362 and slow killing involves different virulence factors. For example, fast killing is usually due to 363 the production and secretion of diffusible secondary metabolites, while slow killing occurs after 364 colonization and infection of the gut (33). The production of hydrogen cyanide is the main 365 virulence factor involved in the fast killing of C. elegans (34), while the regulated export of 366 proteins is needed for slow killing (33). 367 23 Consistent with the reduction in C. elegans killing, we observed reductions in key virulence 368 traits, such as motility, biofilm formation, and secondary metabolite production in days 24 and 369 42 amoeba-adapted isolates. The loss of motility and secondary metabolites probably 370 contributes directly to reduced C. elegans mortality. In P. aeruginosa, type-IV pili are involved 371 in adherence, swarming and twitching motility, and virulence. P. aeruginosa mutants deficient 372 in pilA and pilT demonstrate reduced pathogenesis in mice compared to the wild type (35). 373 Virulence towards nematodes has been partly attributed to pyoverdine. For example, when 374 Pseudomonas syringae interacted with C. elegans, the genes, pvdJ and pvdE that are involved in 375 the synthesis of pyoverdine, were significantly upregulated on fast kill agar (36). A similar 376 response was observed in P. aeruginosa, where 'red death' type killing of C. elegans was shown 377 to be partly due to the production of pyoverdine (37). Additionally, P. syringae ΔpvdJ and 378 ΔpvdL mutants were unable to produce pyoverdine and an unrelated toxin, tabtoxin, 379 demonstrated reduced AHL production and attenuated virulence against the tobacco plant host 380 In co-evolution with C. elegans, lasR and rhlR quorum sensing mutations occurred early in the 397 adaptation process. In addition, lasR mutations are prevalent across multiple CF isolates (43). 398 The regulatory genes lasR and rhlR control the expression of many virulence genes, and las and 399

P. aeruginosa and A. castellanii co-incubation 460
Overnight cultures of P. aeruginosa grown in was centrifuged at 4000 × g for 5 min and washed 461 twice with 1 × M9 salts solution (Sigma-Aldrich, USA; per litre, 6.78 g Na2HPO4, 3 g H2PO4, 1 g 462 NH4Cl, 0.5 g NaCl). A. castellanii, at a concentration of 1 × 10 3 cells mL -1 , was seeded onto the 463 surface of 25 cm 2 tissue culture flasks with 0.2 µm vented caps filled with 10 mL 1 × complete 464 M9 salts + 0.01% glucose. To maintain a strong selective pressure from amoeba, 100 µL of A. Reads were aligned to the P. aeruginosa PAO1 reference genome using CLC Genomics 495 Workbench 9 (CLC Bio, Aarhus, Denmark). SNPs and small insertions and deletions (indels) 496 were called using the probabilistic variant detection analysis, and mutations that were also in the 497 parental strain were filtered out. Genes that contained SNPs in > 25% of the total gene reads 498 were selected for functional analysis using the Database for Annotation, Visualization and 499 Integrated Discovery (DAVID) v6.8 functional tools (52, 53). Gene lists were uploaded onto the 500 DAVID website (https://david.ncifcrf.gov/) and annotations were limited to P. aeruginosa. A 501 functional annotation table was compiled for amoeba adapted populations obtained from day 42 502 of the experiment (Table 1). 503

Isolation of adapted strains 504
To facilitate further phenotypic analysis, P. aeruginosa cells obtained from amoeba-adapted and 505 non-adapted populations on days 3, 24 and 42 were plated onto LB10 agar and incubated 506 30 overnight at 37°C. Ten colonies were randomly selected from each population using a numbered 507 grid and a random number generator. These isolates were stored in 1 mL 70 % LB10 + 30 % 508 glycerol at -80 °C. 509

Assessment of surface colonization and planktonic growth 510
To determine if adaptation with A. castellanii altered surface colonization by P. aeruginosa, the 511 biomass of attached cells was quantified by crystal violet staining as previously described (54). The planktonic growth rate of each isolate was assessed by adding log phase bacterial cultures to 521 96 well plates containing 200 µL of 1 × M9 + 0.01 % glucose and incubating with agitation at 80 522 rpm for 18 h at 37 o C. The suspended biomass within each well was transferred to a fresh plate 523 and measured by spectrophotometry (OD 600 nm) (Infinite ® M200, Tecan, Switzerland). The 524 specific growth rate was determined by applying the formula µ = 2.303 ((log OD2) -(log 525 31 OD1)/(t2-t1)). Assessment of the planktonic growth rate was performed in triplicate for each 526 isolate, n = 30 for each time and treatment. 527

Motility assays 528
Twitching, swarming and swimming motility were assessed as previously described, using 529 motility agar (20 mM NH4Cl, 12 mM Na2HPO4, 22 mM KH2PO4, 8.6 mM NaCl,1 mM MgSO4, 530 100 µM CaCl2, 2 gL -1 Dextrose, 5 g L -1 casamino acids) containing 1, 0.5 or 0.3 % wt vol -1 531 agarose (Bacto™ , BD Biosciences, USA), respectively (55, 56). Five milliliters of motility agar 532 were added into the wells of 6 well plates and dried under laminar flow for 1 h. Isolates were 533 inoculated into the center of the well using 10 µL pipette tips, either to the base of the plate for 534 assessment of twitching motility or mid-agar for assessment of swimming and swarming. 535 Twitching and swarming plates were incubated at room temperature for 48 h and swimming 536 plates were incubated for 24 h prior to imaging with a digital camera (Canon EOS 600D digital 537 single-lens reflex (DSLR) mounted on a tripod, to allow for phenotypic characterization of the 538 resulting colonies and comparative endpoint twitch, swarm and swim distances. Determination 539 of the zone of motility was semi-quantitatively analyzed using ImageJ image analysis software. 540 Motility was assessed in triplicate (n = 3). 541

Quantification of pyoverdine 542
To determine if adaptation with amoeba (3, 24 and 42 d) affects the production of pyoverdine, 543 isolates were grown overnight in 1 mL LB10 media. Cells were removed by centrifugation at 544 32 5200 × g for 5 min and the absorbance of the supernatant was determined with a 545 spectrophotometer (Infinite ® M200, Tecan, Switzerland) at excitation 400 nm and emission 460 546 nm in triplicate. 547

Quantification of rhamnolipids 548
The orcinol method (57) was used to quantify the production of rhamnolipid biosurfactant of 549 nine randomly selected adapted and non-adapted isolates from the day 42 population. Briefly, 550 overnight P. aeruginosa LB cultures were diluted to OD600 0.01 in 25 mL of AB minimal media 551 (58) supplemented with 2 g glucose and 2 g casamino acids L -1 and grown overnight at 37 o C 552 with shaking at 200 rpm. The cell density was determined (OD600 nm) before filtration and 553 extraction of crude rhamnolipid from the supernatant two times using diethyl ether (7 mL). The 554 organic layer was collected, combined, and concentrated in a vacuum concentrator (SpeedVac, 555 Thermo Scientific) at 0 o C for 1 h followed by 2 h at 25 o C, until white solids formed. The solids 556 were resuspended in 500 µL of water and 50 µL of this solution was mixed with 450 µL of 557 freshly prepared orcinol (0.19 % in 53 % H2SO4). Samples were incubated at 80 o C for 30 min 558 and allowed to cool at room temperature for 15 min before quantification of absorbance (OD421). 559 The absorbance was normalized to cell concentration (OD600 nm) for each sample and a factor of 560 2.5 was applied to convert values from a rhamnose standard curve to rhamnolipid concentration 561 (59). 562

Nematode survival assay 563
To determine if P. aeruginosa adaptation to A. castellanii altered bacterial virulence, we tested 564 the survival of C. elegans sp. Bristol N2 after feeding on P. aeruginosa. Axenic C. elegans were 565 obtained via the egg-bleach synchronization method, plated onto NGM agar and fed with heat-566 killed Escherichia coli OP50. L4 stage worms were re-suspended in 1× M9 salts solution and 10 567 -30 worms were drop plated onto 35 mm dishes containing 2 mL fast or slow kill agar (33) 568 containing lawns of pre-established P. aeruginosa obtained from amoeba-adapted or non-569 adapted populations. Plates were incubated at 22 °C and worm numbers were scored by 570 microscopy at 0, 4, 8, 24 and 48 h for fast kill assays, and once per day for slow kill assays. 571 Nematode toxicity was tested using 9 randomly selected P. aeruginosa isolates from each 572 treatment and from times 3 and 42 d. Nematode survival assays were repeated twice 573 independently, and each experiment was performed in triplicate. 574

Competition assays 592
To determine if prior exposure of P. aeruginosa to amoeba increased competitiveness when 593 grown with amoeba, a competition assay was performed. One isolate from the 42 d amoeba 594