F-type Pyocins are Diverse Non-Contractile Phage Tail-Like Weapons for Killing Pseudomonas aeruginosa

Most Pseudomonas aeruginosa strains produce bacteriocins derived from contractile or non-contractile phage tails known as R-type and F-type pyocins, respectively. These bacteriocins possess strain-specific bactericidal activity against P. aeruginosa and likely increase evolutionary fitness through intraspecies competition. R-type pyocins have been studied extensively and show promise as alternatives to antibiotics. Although they have similar therapeutic potential, experimental studies on F-type pyocins are limited. Here, we provide a bioinformatic and experimental investigation of F-type pyocins. We introduce a systematic naming scheme for genes found in R- and F-type pyocin operons and identify 15 genes invariably found in strains producing F-type pyocins. Five proteins encoded at the 3’-end of the F-type pyocin cluster are divergent in sequence, and likely determine bactericidal specificity. We use sequence similarities among these proteins to define 11 distinct F-type pyocin groups, five of which had not been previously described. The five genes encoding the variable proteins associate in two modules that have clearly re-assorted independently during the evolution of these operons. These proteins are considerably more diverse than the specificity-determining tail fibers of R-type pyocins, suggesting that F-type pyocins emerged earlier or have been subject to distinct evolutionary pressures. Experimental studies on six F-type pyocin groups show that each displays a distinct spectrum of bactericidal activity. This activity is strongly influenced by the lipopolysaccharide O-antigen type, but other factors also play a role. F-type pyocins appear to kill as efficiently as R-type pyocins. These studies set the stage for the development of F-type pyocins as anti-bacterial therapeutics. IMPORTANCE Pseudomonas aeruginosa is an opportunistic pathogen that causes a broad spectrum of antibiotic resistant infections with high mortality rates, particularly in immunocompromised individuals and cystic fibrosis patients. Due to the increasing frequency of multidrug-resistant P. aeruginosa infections, there is great interest in the development of alternative therapeutics. One alternative is protein-based antimicrobials called bacteriocins, which are produced by one strain of bacteria to kill other strains. In this study, we investigate F-type pyocins, bacteriocins naturally produced by P. aeruginosa that resemble non-contractile phage tails. We show that they are potent killers of P. aeruginosa, and distinct pyocin groups display different killing specificities. We have identified the probable specificity determinants of F-type pyocins, which opens up the potential to engineer them to precisely target strains of pathogenic bacteria. The resemblance of F-type pyocins to well characterized phage tails will greatly facilitate their development into effective antibacterials.


35
With increasing antibiotic resistance, there is a strong incentive to identify alternative anti-36 bacterial therapeutics. To this end, interest in using phages or parts of phages to treat bacterial 37 infections has greatly increased in recent years (1), and phage treatments have proven effective 38 in clearing bacterial infections in humans (2)(3)(4). This success notwithstanding, there are potential 39 drawbacks to phage therapy, including the possibility that introduced phages may acquire and 40 transmit virulence or antibiotic resistance genes, and that negative outcomes may arise from 41 long-term phage reproduction within a patient. To circumvent these problems, the therapeutic 42 potential of phage tail-like bacteriocins, also referred to as tailocins, is also being explored. 43 Tailocin encoding operons, which are found in many diverse bacterial species, are likely derived 44 from prophages. The utility of tailocins as antibacterials has been amply demonstrated (5, 6). 45 Like phages, tailocins are highly specific for their target organism, but they possess additional 46 advantages. A single tailocin type can be engineered to kill a variety of bacterial species (7,8), 47 fiber. We compared the fiber and chaperone proteins of each of our sequenced clusters to those 118 of the characterized R-type pyocin types (15). Fiber sequences of the R2-, R3-and R4-types are 119 very similar to each other (> 98% identical). Hence, we considered groups R2, R3 and R4 as one 120 group and called it group R2, as was done in a previous study (26). For the R-type pyocin 121 clusters sequenced here, ten belonged to the R1 group, ten to the R2 group, and two to the R5 122 group. As R-type pyocins have been well characterized in previous studies (7,8,15), we focused 123 the present investigation on the F-type pyocins. 124 Conserved proteins encoded in the F-type pyocin cluster 125 The pyoF2 to pyoF10 genes encode confidently annotated functions required for formation of the 126 F-type pyocin tube and tip ( Table 1). The protein products of each of these genes are clearly 127 homologous to phage tail proteins (27), and these proteins are very similar (~ 95% pairwise 128 sequence identity) among the 21 F-type pyocin gene clusters that we have analyzed. Although 129 the F-type pyocin genes are arranged in an order that is syntenic with the genome of the well 130 characterized E. coli phage lambda (12), only the tail tip and central fiber proteins (PyoF6 to 131 PyoF10) of this phage share significant sequence identity with F-type pyocin proteins (31 to 38% 132 sequence identity). The phage tail region with the greatest similarity to the F-type pyocin cluster 133 across the tube and tail tip region is that from E. coli phage HK022. (Table 1). The HK022 134 proteins share 43% sequence identity, on average, to those of the F-type pyocin (Table 1). No 135 prophage tail region was more closely related to the F-type pyocin cluster than phage HK022 136 across the whole cluster, though some P. aeruginosa prophages were more closely related to the 137 3'-end of the cluster where genes encoding the tail tip proteins are located. 138 An unusual feature of F-type pyocin regions as compared to phage tails is the lack of any protein 139 with detectable similarity to a tail terminator. This protein is essential for phage tails because it is 140 required to join the tail to the head (28). The tail terminator also prevents uncontrolled 141 polymerization of the tails of some (28), but not all phages (29). Since F-type pyocins are not 142 joined to a head, the tail terminator appears to be dispensable. The pyoF1 gene lies in the 143 genomic position expected for a tail terminator gene. However, the 95 amino acid protein 144 encoded by this gene bears no detectable sequence similarity to tail terminators, has no homologs 145 outside of P. aeruginosa F-type pyocin clusters, and stop codons are observed in this ORF in several strains. Thus, we conclude that this is not a functioning protein as was also concluded in 147 a previous publication (12). 148 F-type pyocins can be grouped based on proteins encoded at the 3'-149 end of the cluster 150 The host range specificity of phages is determined by proteins located at the tail tip, which are 151 typically encoded by genes at the 3'-end of tail-encoding regions (27). The analogous proteins in 152 the F-type pyocin are encoded by genes pyoF10-pyoF15. Non-contractile tails resembling F-type 153 pyocins possess a long (> 700 residues) central fibre protein that projects directly below the tail 154 tip. In phage lambda, the region within the last 250 residues of the central fiber mediates host 155 cell specificity and surface binding (30, 31). The homologous protein in F-type pyocins is 156 encoded by pyoF10. We observed that the first 1160 residues of the PyoF10 proteins are highly 157 conserved among F-type pyocins (> 93% sequence identity), but the last 60 residues vary greatly, 158 with pairwise identities in this region often ranging below 35% (Fig. S1). This sequence 159 variability is consistent with a role for the C-terminus of PyoF10 in mediating host specificity. 160 In addition to the last 60 residues of PyoF10, the other five proteins encoded at the 3'-end of the 161 F-type pyocin cluster, PyoF11 to PyoF15, were found to vary considerably in sequence between 162 different F-type pyocin clusters. Based on pairwise comparison of homologous proteins encoded 163 in this region of the clusters (Fig. S1), the F-type pyocin regions found in different genomes were 164 divided into six groups, F1,F2,F4,F5,F6 and F7 (Fig. 2b while group F5 and F7 were previously described in P. aeruginosa strains PA14 and DK2, 171 respectively (16). The two most frequently occurring groups are F2 (11 members) and F7 (4 172 members). The F1, F5 and F6 groups were encoded only in pyocin clusters that also encoded R-173 type pyocins, while F4, F7, and F2 group clusters were found in the absence of R-type clusters 174 except in two instances (both F2 group). Further bioinformatic comparisons described below compare representative protein sequences from each of the six F-type pyocin groups that we 176 identified here. 177 PyoF11 and PyoF12 are newly recognized conserved proteins 178 PyoF11 and PyoF12 are proteins of unknown function that are encoded in every F-type pyocin 179 region. These are the most diverse proteins in the F-type pyocin clusters, often displaying 180 pairwise sequence identities of less than 25% (Fig. S1). Despite their diversity, the homologs of 181 these proteins from the six groups could be convincingly aligned (Fig. S2a,b). We used HMMer 182 (32) to create Hidden Markov Model (HMM) profiles from the PyoF11 and PyoF12 alignments. 183 Searching with these HMMs, we identified more than 50 occurrences each of pyoF11 and 184 pyoF12 gene homologs in diverse phages and prophages. These genes often occur together and 185 invariably lie immediately 3' to the central fiber gene (homolog of pyoF10). In some phage 186 genomes, the pyoF11 and pyoF12 genes are very likely the last genes in the tail operon as they 187 are followed immediately by lysis genes (e.g. Burkholderia phage Bcep176 and Xanthomonas 188 phage CP1). These observations suggest that PyoF11 and PyoF12 function in conjunction with 189 the central fiber protein, possibly binding to it or acting as chaperones to aid in folding of the 190 fiber. PyoF11 and PyoF12 had not been previously recognized as conserved proteins in the F-191 type pyocin cluster because these ORFs are not annotated as proteins in most P. aeruginosa 192 genomes. This is likely a result of the lack of annotation of these genes in the PA14 genome,193 which is commonly used as the reference genome for genome assembly and annotation. The 194 functions of PyoF11 and PyoF12 homologs in phages have never been investigated. 195 PyoF13, PyoF14, and PyoF15 are likely involved in host specificity 196 In addition to the central fiber, most non-contractile tailed phages possess genes downstream of 197 the central fiber gene that also encode cell surface receptor binding proteins These are known as 198 "side fibers" in E. coli phage lambda (33). The PyoF13 proteins, which share a common genomic 199 position with the lambda side fibers, are likely involved in determining host range specificity, 200 functioning as receptor binding proteins. A striking feature of the Pyo13 homologs is that their 201 N-termini (the first 140 residues) are very similar among the F-type pyocin groups with pairwise 202 sequence identities ranging from 55% to 90% while the pairwise identities in their C-terminal 203 regions generally range between 20% and 35% (Fig. S1, S3). We surmise that the more 204 conserved N-terminus of PyoF13 mediates binding of this putative receptor binding protein to 205 the F-pyocin tail tip, while the variable C-terminus mediates cell surface binding. The fibers 206 from different groups of R-type pyocins, which have been shown to determine bactericidal 207 specificity (15), display the same type of conservation pattern with N-terminal regions (first 450 208 residues) displaying greater than 95% pairwise sequence identity and C-terminal regions (last 209 250 residues) displaying pairwise sequence identities between 50 and 70% (Fig. S4). In contrast 210 to the F-type pyocins, there are only three distinct groups of R-type pyocins, as defined by fiber 211 sequences, and there is much less variability. 212 Homologs of PyoF13, which share sequence similarity with its N-terminal region, are found in 213 diverse phages and prophages and are located in similar genomic positions as pyoF13, 214 downstream from the central fiber gene. Genes encoding homologs of PyoF14 (~100 residues) 215 and PyoF15 (~75 residues) are also found in many phages and prophages, and they are invariably 216 located downstream of pyoF13 homologs or genes encoding other putative phage receptor-217 binding proteins. The sequences of PyoF14 and PyoF15 are variable, mirroring the sequence 218 variation seen in the C-terminal regions of PyoF13 (Fig. S1). We expect that PyoF14 and 219 PyoF15 are involved in host range specificity through interactions with PyoF13, or possibly by 220 acting as chaperones for the assembly of PyoF13 as is required for phage-encoded receptor 221 binding proteins (34). 222 Two F-type pyocin groups deviate from the others in the pyoF13 to pyoF15 region. The F2 223 group has a complete duplication of this region so that it possesses two copies of each gene. The 224 proteins encoded by the first copy, PyoF13 1 to PyoF15 1 , are distinct in sequence compared to the 225 homologs in other groups (Fig. S1, Fig S3). Conversely, PyoF13 2 to PyoF15 2 are very similar (> 226 90% identical) to homologs found in groups F4 and F5. In contrast to all other groups, group F7 227 lacks a pyoF14 gene. Consistent with this absence, its PyoF13 homolog displays a C-terminal 228 region that has no detectable sequence similarity to the other groups. 229

230
To determine if our bioinformatic groupings of the F-type pyocin clusters correlate with 231 bactericidal specificity, we examined the killing profiles of lysates produced from the 30 strains 232 following induction by mitomycin C. Serial dilutions of lysates of each of the 30 strains were 233 spotted onto lawns of the same 30 strains to produce an all-against-all matrix. Bactericidal 234 activities were detected as zones of clearing on the bacterial lawn. Analysis of these data was 235 complicated because P. aeruginosa produces other bactericidal entities in addition to R-type and 236 F-type pyocins, including S-type pyocins (11) and bacteriophages. Since the presence of any of 237 these can produce zones of clearing, further analyses were necessary to delineate the type of 238 activity present. Testing serial dilutions of lysates allowed us to distinguish clearings produced 239 by phages from those produced by pyocins (Fig. 3a). Due to their replicative nature, clearings 240 resulting from phage lysates resolved into individual plaques upon dilution, while the clearings 241 resulting from pyocins gradually disappeared without the appearance of individual plaques (Fig.  242 3a). Lysates were also spotted onto bacterial lawns containing proteinase K, which eliminated 243 clearings caused by protease sensitive S-type pyocins ( Fig. 3a) (11). By analyzing the activities 244 of the 30 lysates on 30 strains in this manner, we detected more than 450 bactericidal 245 combinations and found that greater than 90% were due to R-or F-type pyocins (Fig. S5). 246 All groups of R-type and F-type pyocins identified displayed bactericidal activities against 247 multiple strains. Notably, the killing spectra of lysates were invariably the same if they contained 248 pyocins of the same R-or F-type group (Fig. S5) . For example, lysates of four different strains 249 encoding F7 pyocins all displayed bactericidal activity against the same 11 strains (note that in a 250 single case the F-type pyocin activity was occluded by the presence of phage activity as denoted 251 by an orange color, Fig S5). These results demonstrate that our classification of pyocins based on 252 sequence analysis is predictive of biological activity. In Fig. 3b, a small subset of the bactericidal 253 data are shown to emphasize the differences in the killing spectra of the F-type pyocin groups. 254 No two groups kill exactly the same set of bacterial strains; however, considerable overlap exists 255 between some groups like F4 and F5. We also noted that no strain was susceptible to an R-or F-256 type pyocin that was encoded in its own genome, which is consistent with previous observations 257 that strains are resistant to their own pyocins (35). Since the F1 and F6 groups were encoded 258 only in strains that also encoded R1 pyocins, the killing caused only by the F-type pyocins could 259 not be discerned. However, comparison with results obtained using a strain encoding only an R1 260 pyocin revealed that strain S25 is susceptible to F1 pyocin as it was killed by a lysate containing 261 F1 and R1 pyocins, but not by a lysate containing only R1 pyocin (Fig. 3b). By the same logic, 262 strain S30 was found to be killed by F6 pyocin. The F5 group was found only in strains that also 263 encode an R-type pyocin. To assess the activity of this group we took advantage of a transposon insertion mutant of an essential R-type pyocin gene in PA14 (36) to detect the activity of the F5 265 pyocin alone (Fig. 3b). 266

F-type and R-type pyocins display similar levels of bactericidal
It was previously reported that one R-type pyocin particle is sufficient to kill a single cell, while 269 up to 280 F-type pyocin particles are required to kill the same cell (19). This implies that an F-270 type pyocin containing lysate would have considerably less killing activity than an R-pyocin 271 containing lysate. However, we observed many cases where lysates of F-type pyocins displayed 272 levels of killing activity as high R-type pyocin lysates. Although R-and F-type pyocin lysates 273 may contain different numbers of particles, we do not expect these numbers to deviate greatly as 274 all pyocin operons utilize the same transcriptional regulatory region. Most convincingly, we 275 tested the bactericidal activity of lysates of two PA14 mutants, one of which carried a transposon 276 insertion in an essential R-type pyocin gene (pyoR6) and one of which carried a similar insertion 277 in an essential F-type pyocin gene (pyoF10). It can be seen that the bactericidal activity of these 278 two lysates was the same, indicating that F-type pyocins and R-type pyocins are equally lethal to 279 a susceptible host (Fig. 3c). We observed that the same F-type pyocin lysate may display 280 different levels of activity on different strains. For example, lysates of F7 group pyocins 281 displayed greater than 10-fold greater bactericidal activity on strain PAO1 as on strain S14 (Fig.  282 3d). The previously observed low activity of F-type pyocins was likely caused by use of a non-283 optimal indicator strain. Overall, our data indicate that F-type pyocins have the potential to kill 284 bacterial cells as efficiently as R-type pyocins. 285 The genes downstream of pyoF10 are required for bactericidal activity 286 Although homologs of the proteins encoded at the 3'-end of the F-type pyocin cluster (PyoF11 to 287 PyoF15) are encoded in phages and prophages, the roles of these proteins have never been 288 investigated. To determine whether these proteins are essential for bactericidal activity, we tested 289 the activity of F-type pyocin mutants in strain PA14 (group F5). We tested transposon insertion 290 mutations in pyoF14, and pyoF15 from the PA14 non-redundant transposon mutant library (36). 291 We constructed in-frame deletion mutations in pyoF11 and pyoF12 and a nonsense mutation in 292 pyoF13 (Supplementary Materials and Methods). Mutations in each of these genes completely abrogated bactericidal activity, indicating that their protein products play essential roles in the 294 production of functional F-type pyocin particles (Table 2). To ensure that the loss of activity 295 resulting from these mutations was the result of abrogation of only the gene in which the 296 mutation was located, each gene was cloned into a plasmid expression vector (Supplementary 297 Materials and Methods) and we determined whether mutations could be complemented by the 298 plasmid expressed genes. The pyoF12 mutant could be complemented by a plasmid expressing 299 only pyoF12 (Table 2). However, complementation of the pyoF11 mutant required plasmid-300 based expression of both pyoF11 and pyoF12. A plasmid expressing only pyoF12 did not 301 complement the pyoF11 mutant. We conclude that both pyoF11 and pyoF12 are essential for 302 bactericidal activity, and that the pyoF11 in-frame deletion mutation also causes loss of pyoF12 303 activity, possibly through a polarity effect. Through a similar series of plasmid based 304 complementation experiments, we determined that pyoF13, pyoF14, and pyoF15 are also 305 essential for bactericidal activity, and that polarity effects are also manifested in this group of 306 genes ( Table 2) serotyping. Previous studies showed that the OSA acts a receptor for some R-type pyocins, while 316 it blocks killing by other R-type pyocins (26). To investigate the effect of the OSA on the 317 activity of F-type pyocins, we experimentally determined the serotypes of the 30 strains used in 318 this study by a slide agglutination assay. We observed a correlation between the serotype of a 319 strain and its F-type pyocin susceptibility profile (Fig. 4a). For example, all three strains of O2 320 serotype were resistant to all F-type pyocins, while the four O5 strains were killed only by F7 321 pyocins. Among the eight O6 serotype strains, the F2 pyocin killed all, but the F4, F5, and F7 322 pyocins were unable to kill some of these strains (Fig. 4a). The resistance of O6 strains S12 and 323 S27 to the activity of the F4 pyocin is expected as these strains encode an F4 pyocin. However, it 324 is not clear why the F7 pyocin fails to kill O6 strains S12, S27, and S5, or why strain S12, alone 325 among O6 strains, is resistant to F5 pyocin. Similarly, the F1 pyocin kills strain S25 but no other 326 O6 strains, and the F6 pyocin kills strain S6 but no other O13/O14 strains. These data show that 327 factors independent of OSA and pyocin type encoded within a strain contribute to F-type pyocin 328 susceptibility. 329 To directly assess the role of OSA in F-type pyocin activity, we tested ∆wbpM mutant strains, 330 which lack OSA in strains PAO1 and PA14 (Fig 4b). The F7-type pyocin is active against PAO1, 331 but was unable to kill the PAO1∆wbpM mutant, suggesting that this pyocin uses the OSA as a 332 receptor. By contrast, the ∆wbpM mutants of PAO1 and PA14 became susceptible to the F4 333 group, though the wild-type strains were not. In this case, the OSA appears to block the pyocin 334 from contacting its receptor. The F2 and F5 groups, which are unable to kill PAO1 or PA14, 335 were also not able to kill the mutants lacking OSA. Strains producing the F1 and F6 group 336 pyocins were unable to kill PAO1 with or without OSA, but PA14∆wbpM did become 337 susceptible to killing. However, this effect may have been due to the R1 pyocins produced by 338 these strains. From these experiments with strains lacking OSA, it is clear that the presence of 339 OSA affects the bactericidal activity of the F4 and F7 groups while the data are inconclusive for 340 the other groups. 341

342
To determine if this collection of F-type pyocin described above encompassed the full diversity 343 of F-type pyocins found across the P. aeruginosa species, we performed BLAST searches 344 against all P. aeruginosa genomes in the NCBI database using a PyoF13 sequence as the query 345 with the goal of identifying homologs with distinct sequences (i.e. share less than 90% sequence 346 identity with those in our established F-type pyocin groups). PyoF13 was chosen for these 347 searches because it is highly conserved among the F-type pyocin operons in its N-terminal 348 region, yet its C-terminal region varies depending on the pyocin group. We discovered three 349 PyoF13 homolog families encoded in F-type pyocin operons that shared less than 70% sequence 350 identity to any other PyoF13 group. F-type pyocins encoding these newly identified PyoF13 351 varieties were defined as groups F8 , F9, and F10. The F9 group is identical to a previously 352 identified group designated as the PA7 group (16). Another identified group, called F11, possessed PyoF13, PyoF14, and PyoF15 homologs that are greater than 95% identical to group 354 F2 2 , F4 and F5, but the PyoF10, PyoF11, and PyoF12 were unique. Finally, group F12 combined 355 PyoF10 to PyoF12 homologs that were 99% identical to group F11 with PyoF13 to PyoF15 356 homologs that were greater than 95% identical to group F10 (Fig. 5, S1). Group F12 was 357 previously identified in P. aeruginosa strain M18 (16). The sequences of proteins PyoF10 to 358 PyoF15 for all eleven F-type pyocin groups can be found in Appendix 1 (Supplementary 359 Material). 360 Pairwise sequence comparisons among all the F-type pyocin groups that we have identified 361 strongly supports the existence of two distinct specificity modules in F-type pyocins (Fig. S1). 362 Whenever the C-terminal regions of PyoF10 proteins in two groups are highly similar (> 90% 363 identity), then the PyoF11 and PyoF12 proteins are also highly similar. Similarly, when two 364 groups have PyoF13 proteins with highly similar C-terminal regions, then the PyoF14 and 365 PyoF15 proteins are also highly similar. Therefore, we have designated regions encoding the C-366 terminus of PyoF10, PyoF11, and PyoF12 as Specificity Module 1 and regions encoding 367 PyoF13, PyoF14, and PyoF15 as Specificity Module 2. Among our full set of pyocin groups, we 368 observed three instances where the same Specificity Module 1 region assorted with different 369 Specificity Module 2 regions (Fig. 5). We also observed three cases where identical Specificity 370 Module 2 regions assorted with different Specificity Module 1 regions. These data indicate that 371 recombination events have occurred between different F-type pyocin operons. 372

373
This study provides a comprehensive analysis of F-type pyocin operons present in P. aeruginosa 374 strains. We have defined the conserved genes in these operons and introduced a systematic 375 naming system for them. The initial 21 F-type clusters examined in strains from our collection 376 were categorized into six different groups, two of which were previously known (F1 and F2) and 377 four of which were named in this study (groups F4 to F7). The killing specificity of each of these 378 groups was shown to be distinct. An additional five F-type pyocin groups were discovered 379 bioinformatically, but the killing specificity of these groups remains to be tested. Importantly, we 380 identified two highly diverse F-type pyocin genes, pyoF11 and pyoF12, which are not annotated 381 as genes in many P. aeruginosa strains, yet are essential for bactericidal activity. The sequence 382 diversity in these two genes contributes to defining the F-type pyocin groups. An important 383 finding is that there are no genes in the F-type pyocin operons that are not also found in the 384 genomes of phages or prophages. Thus, the ability of these pyocins to efficiently kill bacteria 385 while isolated phage tails do not must be due to sequence modifications within their phage-386 derived proteins, not to the presence of unique toxin-encoding genes (unless such genes are 387 encoded elsewhere in the P. aeruginosa genome). Fully active R-type pyocins have been 388 produced heterologously in E. coli from a plasmid vector including only genes from the R-type 389 pyocin operon, indicating that their toxicity does not rely on genes outside of this region (9). 390 Analysis of the F-type pyocin operons clearly indicates the genes that are involved in killing 391 specificity. The proteins encoded by the pyoF2 to pyoF9 genes are highly similar in all the F-392 type pyocin groups. Divergence among the groups begins with the last 60 residues of PyoF10 393 and extends through PyoF15. We defined the F-type pyocin groups according to sequence 394 identity among these proteins (Fig. 5). Since F-type pyocins within the same group invariably 395 displayed the same killing spectra (Fig. S5), we conclude that some or all of the pyoF10 to 396 pyoF15 region determines killing specificity. By examining the patterns of recombination among 397 the specificity genes, we defined two specificity modules: Module 1 (pyoF10 to pyoF12) and 398 Module 2 (pyoF13 to pyoF15). The occurrence of highly similar Module 1 regions with distinct 399 Module 2 regions and vice versa in different F-type pyocin groups indicates that the two modules 400 act independently of one another (Fig. 5). This conclusion is supported by the appearance of 401 pyoF11 and pyoF12 homologs without adjacent pyoF13-pyoF15 homologs and vice versa in 402 phages and prophages. Since these families of proteins have not been characterized, defining 403 their roles in host specificity will be an important goal for further study. 404 The strong sequence similarity between most of the F-type pyocin genes implies that all the 405 groups are descended from a common ancestor that likely arose from a defective prophage. 406 However, it is also clear that some type of horizontal gene transfer mechanism has been 407 responsible for the evolution of the specificity regions, which are comprised of different 408 combinations of Specificity Modules 1 and 2. These reassortments could be caused by phages 409 carrying genes that are similar to these F-type pyocin genes occasionally recombining with the 410 homologous F-type pyocin genes. With respect to the evolution of the F-and R-type pyocins as a 411 whole, it is relevant that the F-type display considerably more divergence among their 412 specificity-determining genes as compared to the R-type (Fig. S1, S4). This suggests that the F-type pyocins may have arisen first and, thus, have had more time to diverge. Also supporting the 414 possibility that the F-type pyocin operon appeared first is that the R-type pyocin genes are 415 inserted in the middle of the lysis genes, which comprise an intact lysis cassette in strains 416 possessing only an F-type operon. 417 The bacterial cell surface receptors of F-type pyocins were previously unknown. Our 418 examination of the activity of the different F-type pyocin groups on 12 different serotypes of P. 419 aeruginosa revealed a clear correlation between bactericidal activity and O-antigen serotypes 420 (Fig. 4a). Further supporting a role for the OSA in the activity of at least some F-type pyocins is 421 the fact that activity of the F7 group required the presence of OSA, while group F4 activity was 422 blocked by OSA (Fig. 5b). While the OSA serotype clearly influences F-type pyocin host 423 recognition, this is not the only determining factor. For example, the F7 pyocin is able to kill 424 some but not all strains with the O6 serotype. In addition, F-type pyocins of a given type were 425 consistently unable to kill strains encoding the same type of F-type pyocin, regardless of 426 serotype. The mechanism of this self-immunity is not known. Many bacteriocins, such as S-type 427 pyocins and colicins, are encoded with specific immunity proteins (11). However, there is no 428 obvious immunity protein candidate encoded within F-type pyocin clusters as each gene is 429 homologous to phage tail proteins. It is possible that no specific immunity proteins exist for R-430 type or F-type pyocins. Rather, strains may have evolved to resist their resident R-and F-type 431 pyocins by altering their cell surface in subtle ways undetectable by the antibodies used in 432 serotyping. 433 Overall, our study shows that F-type pyocins are produced by a large number of P. aeruginosa 434 strains, they all possess antimicrobial properties, and they are promising candidates to study for 435 the development of new therapeutics. Our identification of the specificity determinants of F-type 436 pyocins points the way toward precisely engineering their killing as has been done with the 437 contractile R-type pyocins and non-contractile tailocins of Listeria (7, 8, 10). 438

440
Whole genome sequencing 441 Genomic DNA was isolated using a genomic DNA extraction kit (Bio Basic Inc). Next-442 generation whole genome sequencing was performed by the Donnelly Sequencing Center, 443 University of Toronto, using Illumina HiSeq2500. De novo assembly of reads into contigs was 444 performed using Velvet version 2.2.5 (38). Genes trpE and trpG were located and the region 445 between these genes was analyzed using Geneious (39). species, but was small enough to allow manual analysis of the protein sequences and the 452 genomic context of genes encoding proteins related to pyocin proteins. This work was aided by a 453 synteny viewing and phage gene annotation toolkit developed in our laboratory, which will be 454 described in detail elsewhere. Sequence alignment analysis was performed in Jalview (41). To 455 identify protein sequences similar to less frequently occurring proteins found in the pyocin 456 cluster (e.g. PyoR1, PyoF11, and PyoF12), alignments were constructed of the pyocin proteins. 457 HMMER3 (32) was then used to create Hidden Markov Models (HMMs). These HMMs were 458 used to detect proteins similar to a given pyocin protein. The genome context of genes encoding 459 these similar proteins within phage genomes was assessed to support a conclusion that the pyocin 460 protein possesses the same function as the phage protein. BLAST searches to identify new 461 groups of F-type pyocins were carried out against all P. aeruginosa genomes available at NCBI 462 in April, 2018. 463 HHpred searches were carried out using the online server 464 (https://toolkit.tuebingen.mpg.de/hhpred) (42). HMM-based searches were carried out using 465 HMMer (32) and analyzed by searching the Pfam (43) and TIGRfam (www.jcvi.org/cgi-466 bin/tigrfams/index.cgi) databases. 467

468
A continuous carbon film coated EM grid was made hydrophilic by glow discharge. 5 μl of 469 sample was applied to the surface of the grid and left for absorption for 2 minutes. Excess sample 470 was blotted away using the corner of a filter paper. The grid was washed three times with water 471 and stained with 2% (w/v) uranyl acetate. Grids were examined with a Hitachi H-7000 472 microscope. 473

474
To generate lysates containing pyocins and/or phages, 5 ml cultures started from overnights were 475 grown in LB at 30 °C until the cells reached an OD 600 of 0.4. Mitomycin C, was then added to a 476 final concentration of 2 μg/ml and shaking at 30 °C was resumed for 3 h or until cell lysis 477 occurred. Chloroform was added to all induced cultures (1-2 drops/ml) to ensure maximum 478 bacterial lysis. In experiments testing complementation from plasmids, 0.2% arabinose was 479 added to cells after 1 h of growth at 30 °C to induce the expression of proteins from the plasmid 480 prior to addition to mitomycin C to induce F-type pyocin induction from the genome. After lysis, 481 cultures were incubated at room temperature with DNase (10 μg/ml) for 30 min prior to 482 centrifugation at 10000 rpm for 10 min. For activity assays, 2 µl volumes of dilutions of these 483 lysates were spotted onto lawns of P. aeruginosa strains. Lawns of strains to be tested were made 484 by adding 150 μl of overnight culture to 3 ml of molten 0.7% top agar, which was immediately 485 poured onto an LB agar plate and allowed to harden. To distinguish S-type pyocin activity, 486 duplicate lawns were poured containing proteinase K (100 μg/ml). At least three biological 487 replicates were performed for each strain and lysate combination.   Shown are a lysate of strain S22 (left panel), which produces both R-and F-type pyocins; a lysate of strain S13 (middle panel), which produces just R-pyocins; and a lysate of strain S18 (right panel), which produces just F-pyocins. R-type pyocin particles are indicated by green arrows and F-type pyocin particles are indicated by red arrows. Grids were negatively stained with uranyl acetate. The scale bar shown applies to all three micrographs.

FIG 1
Transmission electron micrographs of lysates of cells producing R-and F-type pyocins.
Shown are a lysate of strain S22 (left panel), which produces both R-and F-type pyocins; a lysate of strain S13 (middle panel), which produces just R-pyocins; and a lysate of strain S18 (right panel), which produces just F-pyocins. R-type pyocin particles are indicated by green arrows and F-type pyocin particles are indicated by red arrows. Grids were negatively stained with uranyl acetate. The scale bar shown applies to all three micrographs. FIG 5 All F-type pyocin groups. A close up of genes encoded at the 3'-end of F-type pyocin clusters shows the 11 different groups identified in our sequenced strains and in the database. Groups of genes are colored the same if the proteins they encode display greater than 90% sequence identity. The extent of Specificity Modules 1 and 2 are shown at the top as are the names of the genes in these regions. P. aeruginosa strains where certain groups were previously identified are shown in parentheses.