Outer membrane changes enable evolutionary escape from bacterial predation

1 To combat antimicrobial resistant pathogens, natural predatory bacteria, like Bdellovibrio 2 bacteriovorus, represent potential alternatives. B. bacteriovorus could be particularly potent as it kills 3 a broad range of human bacterial pathogens, however, it remains unclear whether prey can evolve 4 genetically-determined resistance against predation. Here, we show that the model bacterium 5 Escherichia coli K-12 consistently evolves resistance against B. bacteriovorus during experimental 6 evolution. Selection for resistance scaled positively with predation pressure and was widespread after 7 two cycles of predator exposure. Like antibiotics, predation resistance was costly, manifesting in a 8 trade-off between predation resistance and fitness in the absence of predators. Genetic analysis 9 identified changes in outer membrane porin OmpF as common resistance mechanism, while a mutation 10 in cell envelope lipopolysaccharide-modifying enzyme WaaF was rarer but also conferred predation 11 resistance. Our study uncovers evolutionary and mechanistic aspects of prey escape from predation, 12 generating important knowledge on predator-prey interactions and to advance sustainable treatments. 13 14

predator-prey interaction, predatory bacteria, Escherichia coli, Bdellovibrio 15 bacteriovorus, predation resistance, evolution, prey escape 16 17 18 19 20 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 3

21 Bacterial resistance to conventional antibiotics is a major public health concern, that led to 1.27 million 22 deaths in 2019 1. This urgently calls for the discovery of novel antibiotics and alternative treatment 23 strategies2–4. One promising alternative involves the use of natural bacterial predators including 24 bacteriophages (viruses of bacteria)5 and predatory bacteria3,6, which are environmentally widespread7. 25 While phage therapy is currently extensively studied 2, much less is known about the potential of 26 predatory bacteria, like Bdellovibrio bacteriovorus, that kill and invade pathogenic bacteria and are 27 therefore considered ‘living antibiotic s’3,6. B. bacteriovorus is capable of kil ling growing and 28 stationary phase bacterial prey cells, and in addition irrespective of their antibiotic resistance8–10. 29 Similar to bacteriophages, B. bacteriovorus is neither toxic nor pathogenic to eukaryotic cells and is 30 therefore of great promise 3,6. However, key differences exist between the two predator types . For 31 example, bacteriophages are typically very specific to prey strains, whereas B. bacteriovorus is a 32 generalist with a broad prey range 11,12. This makes B. bacteriovorus suitable to treat polymicrobial 33 infections11,13. Moreover, there seem to be few defence and immunity mechanisms present among prey 34 species to defend themselves against B. bacteriovorus14,15. This contrasts with the o ver hundred 35 specific bacterial defence and immune systems reported against bacteriophages 16. Thus, 36 B. bacteriovorus might be a more effective treatment under certain circumstances, compared to 37 traditional antibiotics and bacteriophages. 38 Despite these promising aspects, little is known on the selection pressures that B. bacteriovorus may 39 impose on its prey and the evolutionary responses it may trigger17–19. Some evolutionary studies, 40 focussing on predator -prey interactions , reported evidence for resistance evolution 17–19 but did not 41 explore the underlying mechanisms on a phenotypic and genotypic level. While recognition of the prey 42 and prey invasion is likely multifactorial 20,21, potentially hindering fast resistance formation, prey 43 resistance has been described transient 3,22. One study described a form of phenotypic (plastic) 44 resistance, whereby several prey species developed a tra nsient yet reversible stage of predation 45 resilience22. Finally, several studies highlighted that secretion systems23 and secreted compounds such 46 as quorum-sensing signals24,25, indole26 and cyanide27 may offer some protection from B. bacterivorous 47 predation, while the presence of paracrystalline protein surface (S-) layers seems to prevent killing in 48 certain bacterial species14. 49 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 4 To obtain a systematic and in-depth understanding of the potential of prey to evolve resistance to 50 B. bacteriovorus, we performed an experimental prey evolution study with the model organism E. coli 51 K-12 as prey and B. bacteriovorus HD100 as predator. Upon repeated and intermittent exposure of 52 prey to different predator concentrations, we found strong evidence that E. coli K-12 predominantly 53 evolves genetically -based predation resistance under high predatory pressure , whereas increased 54 growth due to media adaptation and phenotypic effects played minor roles. Evolutionary escape from 55 predation evolved repeatedly across independently evolved prey lineages. Predation resistance is costly 56 and manifested in a negative association between the resistance level of a lineage and its growth in the 57 absence of predation. A combination of population sequencing with high genome coverage and single-58 gene deletion mutant experiments revealed that mutations in the outer membrane porin OmpF, and an 59 enzyme ( WaaF) involved in lipopolysaccharide (LPS) modification are involved with resistan t 60 phenotypes. These results suggest that the prey modifies different potential points of contact with the 61 predator15,20. Our study elucidates the ecological context, the evolutionary path , and the molecular 62 mechanisms of prey resistance evolution against B. bacteriovorus, generating key knowledge for the 63 development of bacterial predators as sustainable ‘living antibiotics’. 64 65

66 Experimental evolution leads to increased prey growth 67 We experimentally evolved E. coli K-12 prey under no, low and high predatory pressures of 68 B. bacteriovorus HD100 in eightfold replication . Experimental prey evolution was conducted for 15 69 alternate cycles of predation and prey recovery (Fig. 1a, Extended Data Fig. 1). After prey evolution, 70 we subjected all evolved prey lineages to the predation pressure they experienced during experimental 71 evolution. We found significant growth differences across the three predatory pressures (ANOVA: 72 F2,141 = 8780.2, p < 0.0001). All E. coli lineages evolved without predator grew significantly better 73 than the ancestor (red dotted line) , which indicates media adaptation (Fig. 1b+e). For prey lineages 74 evolved with predators, we observed that five low-predator evolved lineages (L1-L5, Fig. 1c+f) and 75 all high-predator evolved lineages (H1-H8, Fig. 1d+g) showed significant growth recovery relative to 76 the ancestor subjected to predation (black dotted line) . For most of these lineages (84.6%), growth 77 recovery was partial, while only two lineages (H4, H5, 15.4%) showed complete recovery, compared 78 to the unchallenged ancestor. Moreover, growth under predation was higher for prey lineages evolved 79 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 5 under high than under low predatory pressure. Altogether, t hese results suggest that prey evolved 80 partial resistance to predation and that higher predation pressure selected for higher levels of resistance. 81 82 Predation and not media adaptation drives resistance evolution 83 Media adaptation could be a simple explanation for the observed resistance phenotypes because faster 84 growth would allow the prey to outpace the predator’s killing speed . If this hypothesis holds true we 85 expect that no-predator evolved lineages, which showed highly improved growth, should be able to 86 resist predation. We found no support for this hypothesis as all no-predator evolved lineages were still 87 fully susceptible to predation (Fig. 2a, c). Instead, our results show that lineages L1-L5 (which evolved 88 partial resistance under low predation) were also able to grow under high predatory pressure (Fig. 2b, 89 d). We even observed slightly increased growth in a previously unnoticed lineage (L8). Together, these 90

indicate that adaptations are predation-specific and not driven by media adaptation. 91 92 Predation resistance arises early and remains persistent over time 93 Next, we examined the temporal pattern of predation resistance evolution. We noticed that prey optical 94 density (OD600nm) already increased at the second evolution cycle and fluctuated at a relatively high 95 level thereafter (Extended Data Fig. 1a). To verify that resistance arose early, we repeated the growth 96 experiments shown in Fig. 1b-g with all high-predator lineages (H1-H8) from the 2nd and 5th evolution 97 cycle ( Extended Data Fig. 1 and 2 ). These analyses confirmed that predation -resistant phenotypes 98 emerged early on during experimental evolution as all lineages could readily grow under high predation 99 pressure at evolution cycle 2, with barely any growth differences existing compared to later evolution 100 cycles 5 and 15 (Extended Data Fig. 1 and 2) . Hence, predation resistance evolved early and once 101 emerged within a prey lineage it remained persistent over generations. 102 103 Trade-off between predation resistance and prey growth in the absence of predation 104 Antibiotics resistance is often associated with fitness costs in the absence of antibiotics 28,29. Here, we 105 tested whether a similar trade-off occurs for E. coli in the context of predation resistance. To investigate 106 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 6 this, we grew all evolved lineages without predators and calculated their growth rate relative to the 107 ancestor. Relative growth rates differed significantly across predatory pressures (ANOVA: F2,141 = 108 460.42, p < 0.0001), and were highest in no-predator evolved E. coli lineages (Fig. 3a, d), followed by 109 low-predator evolved lineages with growth rates similar to the ancestor (Fig. 3b, e) and high-predator 110 evolved lineages with generally lower growth rate than the ancestor (Fig. 3c, f). When contrasting these 111 growth rates against relative prey resistance (quantified as relative growth integral under high predatory 112 pressure), we found a significant negative correlation between predation resistance and prey fitness, 113 indicating the predicted trade-off ( Fig. 3g). Thus, our findings corroborate that E. coli resistance 114 mechanisms are costly in the absence of B. bacteriovorus predation. 115 116 High consistency in resistance levels among single clones 117 Based on the detected trade-off between predation resistance and population prey growth, we wondered 118 whether there is heterogeneity in predation resistance among clones within prey population s. 119 Consequently, we analysed the predation resistance profile of six randomly isolated clones from each 120 of the eight high-predator lineages (15th cycle, 48 clones) and found that all clones showed consistently 121 high predation resistance . This shows that clones did not separate into susceptible and resistant 122 phenotypes (Extended Data Fig. 4 and 5 ) and thus did not segregate along a trade-off line within 123 populations. This result further indicates that the mechanisms conferring resistance to predation likely 124 operate at the individual and not at the group level of the prey . Such group-level mechanisms were 125 proposed to operate via the secretion of small soluble substances (quorum-sensing signals, indole, 126 cyanide)24–27 and are expected to protect not only the evolved producers but also susceptible wildtype 127 cells. 128 129 Mutation frequency in the prey genome peaks under high predatory pressure 130 To uncover the genetic basis of putative resistance mechanisms, w e sequenced all 24 evolved E. coli 131 lineages (N1-8: no predation, L1-8: low predation, H1-8: high predation) after the 15th evolution cycle 132 along with the original ancestor with high sequencing depth. The Illumina sequencing reads were 133 mapped to the reference genome of E. coli K-12 MG1655 and multi-sample variant calling with a 134 frequency cutoff of 5% was performed. We focused on E. coli genes that had mutations in evolved 135 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 7 lineages but not in the ancestor (Extended Data Fig. 6). We excluded mutated rRNA and transposase 136 genes because they did not correlate with any of the three evolution conditions (Extended Data Fig. 7). 137 Using these filters, we uncovered 43 mutated genes of which 17 occurred in N -lineages, 17 in L -138 lineages and 30 in H -lineages (Fig. 4a). A total of 26 genes mutated exclusively under predatory 139 pressure (Fig. 4a), and thus reflect the top candidates potentially explaining resistance phenotypes. H-140 lineages, had more exclusive gene mutations than L-lineages (18 vs. 5, Fig. 4b). Moreover, there were 141 fewer high-frequency mutations (>70%) among L-lineages than among H-lineages (Fig. 4b), which is 142 likely explained by the higher selection pressure. 143 144 Outer membrane porin OmpF mutations repeatedly arose and confer resistance to predation 145 Next, we conducted a cluster analysis (Fig. 4b) and analyzed the functions of the mutated gene products 146 including mutation types and (predicted) amino acid changes (Fig. 4c, Extended Data Table 1). These 147 analyses yielded ompF, encoding the outer membrane porin F, as the most frequently mutated gene. 148 Four H-lineages (H2, H4, H6 & H7) and two L-lineages (L1, L2) had mutations in this gene. Moreover, 149 lineage H3 had a point mutation in the upstream region of ompF (Extended Data Fig. 8 ) and lineage 150 H8 had an inframe insertion in envZ (Extended Data Table 1), which encodes a sensor histidine kinase 151 involved in OmpF regulation30. Mutation frequencies were high (73.6%-100%) in all lineages except 152 in L2. The sequencing of six individual clones per H-lineage showed that mutations in ompF and envZ 153 were present in all clones (Extended Data Fig. 9). Overall, there were no major differences in 154 mutational patterns between clonal and population level sequencing (Extended Data Fig. 9), reinforcing 155 the view of selective sweeps of a few beneficial mutations through populations. Mutation types in 156 ompF include missense variants (H2), frameshifts (H4, H6, H7) and direct stop gains (L1, L2) (Fig. 157 4b, c, Extended Data Fig. 8, Extended Data Table 1), indicating that most observed mutations probably 158 lead to OmpF function loss. 159 To directly test whether the loss of OmpF (a known attachment side for phages and known to be critical 160 for effective B. bacteriovorus predation15) confers resistance to predation, we generated a clean, single 161 deletion ompF mutant based on the E. coli KEIO library (details in methods, Supplementary Online 162

Table SOnline1). First, we confirmed that E. coli KEIO reference strain (BW25113) was fully 163 susceptible to B. bacteriovorus and grew well in the absence of predation (Fig. 5a,b ), thus showing 164 similar responses as the E. coli K-12 ancestor (Fig. 1b, d and 2a,b). We then subjected the ompF 165 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 8 mutant to no and high predation pressures and tracked its growth. We found this mutant to be 166 completely resistant to predation, showing full growth recovery under high predator exposure (Fig. 167 5a,b). We repeated the above experiments with a envZ clean deletion mutant and found that this 168 mutant was still susceptible to predation , in agreement with Mun et al. 15 (Fig. 5a, b) . This is 169 unsurprising because the experimentally evolved envZ mutant involved an inframe insertion [insertion 170 of Lys-Thr-Trp-Leu between Leu135 and Lys136] and not a deletion. We suspect that the insertional 171 mutation in our experiment might upregulate OmpR, resulting in lower levels of OmpF. We thus 172 conclude that functional changes in EnvZ, but not loss-of-function, could confer predation resistance. 173 In summary, our results indicate that predation resistance is primarily driven by OmpF loss-of-function 174 mutations. 175 176 Mutations in truA cause costly prey morphology change 177 In lineage H8, a high frequency mutation occurred in truA (100%, missense: Val200Ala) in addition 178 to the above -described envZ mutation (77.5%). The truA gene encodes the tRNA pseudouridine 179 synthase A important for protein synthesis 31. Phase-contrast microscopy revealed that cells from 180 lineage H8 no longer separated properly after cell division resulting in cells being thinner and longer 181 (Extended Data Fig. 3a-c). Experiments with a truA clean deletion mutant confirmed division 182 abnormalities (Extended Data Fig. 3), matching previously reported results on the effect of truA 183 disruption32. No such abnormalities in cell morphology occurred in the other H- and L -lineages 184 (Extended Data Fig. 3, Fig. SOnline1). However, deletion of truA did not result in predation resistance 185 but was rather associated with severe fitness costs in the absence of predation (Fig. 5). We hypothesise 186 that the resistance phenotype observed in the evolved truA mutant is masked in the clean deletion 187 mutant due the detrimental effects on cell division (Fig. 5) . Further, our mutation frequency data 188 suggest that the missense mutation in truA (100%) happened first and was associated with a drop in 189 final OD600nm after the recovery phase of early evolution cycles 1-5. Growth only recovered in later 190 cycles, likely due to the additional mutation in envZ (77.5%), probably partly compensating the fitness 191 drop (Extended Data Fig. 1b). Altogether, our results indicate that resistance to predation may occur 192 through cell morphology changes. However, such events might be rare due to the severe fitness costs 193 typically associated with changes in fundamental cell physiological processes. 194 195 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 9 Mutation in LPS synthesis gene waaF confers partial resistance to predation 196 In lineage H5, a high frequency mutation occurred in waaF (100%, missense: Ala255Glu), encoding 197 an heptosyltransferase involved in the biosynthesis of the lipopolysaccharide (LPS) inner core33. We 198 hypothesized that modification or abolishment of outer membrane structures could represent a 199 mechanism through which E coli can become resistant to predation. In support of our hypothesis, we 200 found that a waaF clean deletion mutant showed partial and significant resistance to predation (Fig. 201 5a,b). The same hypothesis could also hold true for the high frequency mutation observed in waaB 202 (lineage H4, 87.5%, stop codon gained) encoding a galactosyltransferase involved in LPS biosynthesis. 203 However, contrary to our expectation we did not find waaB (clean deletion mutation) to be resistant 204 to predation (Fig. 5a,b). Instead, it showed significant growth improvements in the absence of predation 205 (Fig. 5c,d). Because the mutation in waaB occurred together with an ompF mutation in lineage H4 we 206 conclude that WaaB loss-of-function does not confer predation resistance but is beneficial in an already 207 resistant lineage, showing the highest growth recovery under predation among all H-lineages (Fig. 1g). 208 209

210 Historical records on antibiotics tell us that bacterial pathogens have evolved resistance to every single 211 drug used in clinics. Nowadays, it is therefore imperative to understand putative mechanisms and paths 212 to resistance already during the development of new antibacterials34,35. In our study, we performed this 213 task for the predatory bacterium B. bacteriovorus, which is regarded as a potential ‘living antibiotic’ 214 preying on human pathogens. We found that the model prey E. coli K-12 MG1655 consistently evolved 215 resistance against predation through two main mechanisms: change in the outer membrane porin OmpF 216 (several independent events), and change in WaaF (one event), an enzyme involved in outer membrane 217 LPS biosynthesis. These mechanisms suggest that selection acts on cell -membrane associated traits 218 that form the first line of contact between prey and predators. The uncovered resistance mechanisms 219 are similar to the ones observed to evolve against bacteriophages36,37 (loss/alteration of host 220 receptors38,39) and against membrane-targeting polymyxin antibiotics (LPS modifications)40. 221 OmpF enables E. coli to import and export small molecules via passive diffusion. Osmolarity of the 222 environment likely effects prey susceptibility towards B. bacteriovous. It is known that E. coli 223 expresses high levels of OmpF under neutral pH, whereas high osmolarity leads to its 224 downregulation41. OmpF serves as an entry point for bacterial toxins (colicin) and antibiotics42–44 while 225 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 10 simultaneously serv ing as docking site for certain bacteriophages 38,39. However, its role in the 226 interaction with B. bacteriovorus is not yet resolved , apart from the fact that the deletion of OmpF 227 slows predation rate15 and that mutations in this gene are under selection in response to predation in 228 our study. We can thus only speculate about how loss or impairment of OmpF could foster predation 229 resistance, and we propose three scenarios. Firstly, change of OmpF could prevent predators to use the 230 porin as a docking site for attachment15, similar to phage invasion38,39. Secondly, OmpF might serve as 231 an ‘overpressure valve’ equilibrating the osmotic pressure building up when B. bacteriovorus enters 232 the prey periplasm. Loss of OmpF or a reduction in its small molecule transport rate might compromise 233 this function and presumably slow down predator entry. Thirdly, loss of OmpF could reduce the entry 234 of potential, hypothetical small molecules secreted by the predator. Such molecules could manipulate 235 the prey during attachment and recent evidence indicates that B. bacteriovorus can kill the prey even 236 without subsequent entry45. Altogether, our results combined with previous reports indicate that OmpF 237 is a mutational hotspot in the context of resistance evolution against antibiotics, bacteriophages, and 238 predatory B. bacteriovorus. 239 Important to note is that the relevance of OmpF for successful B. bacteriovorus predation seems to 240 depend on the prey species 15. For example, deletion of OmpF in Salmonella enterica did not reduce 241 prey killing15. Instead, it was reported that the LPS core has a major influence on attachment46,47. This 242 finding combined with our results (showing that mutations in LPS -modifying enzymes WaaF and 243 WaaB were under selection) strongly support the view that LPS structure takes on an important role in 244 the interactions with predators. Work with bacteriophages revealed two opposite roles for LPS 245 (particularly the long O -antigens): depending on their specific structure they may act as primary 246 receptors for attachment or block attachment to the outer membrane 38,39. Several lines of evidence 247 suggest that the dual role also applies to interactions with B. bacteriovorus20,46,48. First, a phage-tail 248 fiber protein homolog of B. bacteriovorus was indeed shown to bind to multiple bacterial surface 249 glycans/LPS in multiple prey species ( E. coli K-235, Serratia marcescens and Proteus mirabilis)20. 250 Second, attachment to prey is slowed down when predators are confronted with complex O -antigens 251 as opposed to shorter LPS only featuring the core oligosaccharides in several species ( S. enterica, 252 E. coli, Vibrio cholerae)46 48. In sum, our results indicate that evolutionary change in LPS structure can 253 modulate prey-predator interactions as observed for bacteriophage-bacteria interactions. 254 At first sight, our results on repeated resistance evolution seem to dampen the hope for predatory 255 bacteria to become sustainable antibacterial treatments . Nonetheless, our experiments also revealed 256 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 11 good news. Particularly, the strong fitness trade-off between resistance to predation and growth in the 257 absence of predation highlights that genetically manifested predation resistance is costly, 258 compromising cellular functionality. This phenomenon could limit the spread of resistant prey as they 259 are weak competitors and probably cleared away more efficiently by macrophages. There are concepts 260 to leverage such competitive constraints to increase treatment sustainability 28,49. A key condition for 261 leveraging on competitive constraints is that the pathogen is eliminated before compensatory mutations 262 arise. We observed two putative compensatory mutations in our experiments (mutation in waaB in the 263

of an ompF mutant; mutation in envZ in the background of a truA mutant). Clearly, more 264 research is required to understand the propensity of resistant prey mutants (with and without 265 compensatory mutations) to spread. Moreover, it will be important to test whether similar evolutionary 266 patterns (type of mutants, compensatory mutants and speed of resistance development) are detected 267 when using predatory B. bacteriovorus against pathogenic bacteria and clinical isolates as prey to 268 assess the sustainability of this ‘living antibiotic’. 269 To conclude, we like to emphasize two important points regarding prey resistance against bacterial 270 predation by B. bacteriovorus. First, resistance mechanisms do not need to offer complete protection 271 to escape predation. Instead, it suffices to alter the balance between prey growth and killing rate. In our 272 experiment, we detected a decrease in killing rate (by mutations in ompF and waaF), but no increase 273 in prey growth rate. Second, while we focused on predation resistance through mutation and selection 274 processes, previous work has reported ‘plastic’ resistance phenotypes, occurring after short timescale 275 exposures, offering transiently protection against B. bacteriovorus22. Such plastic effects also surface 276 in response to colistin exposure (i.e. change in membrane polarity) and might well have occurred in 277 our experiments (e.g., in resistant prey lineages exposed to low predatory pressure with few low 278 frequency genetic mutations). A key task for future studies is to integrate all these aspects and to 279 understand the dynamics of prey resistance and predation escape under situations that are closer to 280 clinical settings – for example using clinical pathogen isolates, assessing treatment efficacy in spatially 281 structured human cell cultures, and allowing predators to co -evolve with the pr ey. Only with such 282 integrative knowledge, we can ensure the long -term sustainability of B. bacteriovorus as a ‘living 283 antibiotic’, efficient in targeting multi-drug resistant pathogens, in clinically relevant contexts. 284 285 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 12

(Main text) 286 1. Murray, C. J., et al. Global burden of bacterial antimicrobial resistance in 2019: a systematic 287 analysis. The Lancet 399, 629–655 (2022). 288 2. Pirnay, J.-P., et al. Personalized bacteriophage therapy outcomes for 100 consecutive cases: a 289 multicentre, multinational, retrospective observational study. Nat Microbiol 9, 1434–1453 290 (2024). 291 3. Sockett, R. E. & Lambert, C. Bdellovibrio as therapeutic agents: a predatory renaissance? Nat 292 Rev Microbiol 2, 669–675 (2004). 293 4. Konwar, A. N., Hazarika, S. N., Bharadwaj, P. & Thakur, D. Emerging Non-Traditional 294 Approaches to Combat Antibiotic Resistance. Curr Microbiol 79, 330 (2022). 295 5. Koskella, B., Hernandez, C. A. & Wheatley, R. M. Understanding the Impacts of 296 Bacteriophage Viruses: From Laboratory Evolution to Natural Ecosystems. Annu. Rev. Virol. 9, 297 57–78 (2022). 298 6. Atterbury, R. J. & Tyson, J. Predatory bacteria as living antibiotics - where are we now? 299 Microbiology (Reading) 167, (2021). 300 7. Lai, T. F., Ford, R. M. & Huwiler, S. G. Advances in cellular and molecular predatory biology 301 of Bdellovibrio bacteriovorus six decades after discovery. Front. Microbiol. 14, 1168709 302 (2023). 303 8. Saralegui, C., Herencias, C., Halperin, A. V., De Dios-Caballero, J., Pérez-Viso, B., Salgado, 304 S., Lanza, V. F., Cantón, R., Baquero, F., Prieto, M. A. & Del Campo, R. Strain-specific 305 predation of Bdellovibrio bacteriovorus on Pseudomonas aeruginosa with a higher range for 306 cystic fibrosis than for bacteremia isolates. Sci Rep 12, 10523 (2022). 307 9. Upatissa, S., Mun, W. & Mitchell, R. J. Pairing Colicins B and E5 with Bdellovibrio 308 bacteriovorus To Eradicate Carbapenem- and Colistin-Resistant Strains of Escherichia coli. 309 Microbiol Spectr 11, e00173-23 (2023). 310 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 13 10. Sun, Y., Ye, J., Hou, Y., Chen, H., Cao, J. & Zhou, T. Predation Efficacy of Bdellovibrio 311 bacteriovorus on Multidrug-Resistant Clinical Pathogens and Their Corresponding Biofilms. 312 Jpn J Infect Dis 70, 485–489 (2017). 313 11. Dashiff, A., Junka, R. A., Libera, M. & Kadouri, D. E. Predation of human pathogens by the 314 predatory bacteria Micavibrio aeruginosavorus and Bdellovibrio bacteriovorus: Predation by 315 M. aeruginosavorus and B. bacteriovorus. Journal of Applied Microbiology 110, 431–444 316 (2011). 317 12. Jurkevitch, E., Minz, D., Ramati, B. & Barel, G. Prey Range Characterization, Ribotyping, and 318 Diversity of Soil and Rhizosphere Bdellovibrio spp. Isolated on Phytopathogenic Bacteria. Appl 319 Environ Microbiol 66, 2365–2371 (2000). 320 13. Im, H., Choi, S. Y., Son, S. & Mitchell, R. J. Combined Application of Bacterial Predation and 321 Violacein to Kill Polymicrobial Pathogenic Communities. Sci Rep 7, 14415 (2017). 322 14. Koval, S. F. & Hynes, S. H. Effect of paracrystalline protein surface layers on predation by 323 Bdellovibrio bacteriovorus. J Bacteriol 173, 2244–2249 (1991). 324 15. Mun, W., Upatissa, S., Lim, S., Dwidar, M. & Mitchell, R. J. Outer Membrane Porin F in E. 325 coli Is Critical for Effective Predation by Bdellovibrio. Microbiol Spectr 10, e03094-22 (2022). 326 16. Georjon, H. & Bernheim, A. The highly diverse antiphage defence systems of bacteria. Nat Rev 327 Microbiol 21, 686–700 (2023). 328 17. Varon, M. Selection of predation-resistant bacteria in continuous culture. Nature 277, 386–388 329 (1979). 330 18. Gallet, R., Alizon, S., Comte, P.-A., Gutierrez, A., Depaulis, F., van Baalen, M., Michel, E. & 331 Muller-Graf, C. D. M. Predation and Disturbance Interact to Shape Prey Species Diversity. 332 170, 143–154 (2007). 333 19. Gallet, R., Tully, T. & Evans, M. E. K. Ecological conditions affect evolutionary trajecory in a 334 predator-prey system. Evolution 63, 641–651 (2009). 335 20. Caulton, S. G., Lambert, C., Tyson, J., Radford, P., Al-Bayati, A., Greenwood, S., Banks, E. J., 336 Clark, C., Till, R., Pires, E., Sockett, R. E. & Lovering, A. L. Bdellovibrio bacteriovorus uses 337 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 14 chimeric fibre proteins to recognize and invade a broad range of bacterial hosts. Nat Microbiol 338 9, 214–227 (2024). 339 21. Chanyi, R. M., Ward, C., Pechey, A. & Koval, S. F. To invade or not to invade: two 340 approaches to a prokaryotic predatory life cycle. Can J Microbiol 59, 273–279 (2013). 341 22. Shemesh, Y. & Jurkevitch, E. Plastic phenotypic resistance to predation by Bdellovibrio and 342 like organisms in bacterial prey. Environmental Microbiology 6, 12–18 (2004). 343 23. Aharon, E., Mookherjee, A., Pérez-Montaño, F., Mateus Da Silva, G., Sathyamoorthy, R., 344 Burdman, S. & Jurkevitch, E. Secretion systems play a critical role in resistance to predation by 345 Bdellovibrio bacteriovorus. Research in Microbiology 172, 103878 (2021). 346 24. Hoshiko, Y., Nishiyama, Y., Moriya, T., Kadokami, K., López-Jácome, L. E., Hirano, R., 347 García-Contreras, R. & Maeda, T. Quinolone Signals Related to Pseudomonas Quinolone 348 Signal-Quorum Sensing Inhibits the Predatory Activity of Bdellovibrio bacteriovorus. Front. 349 Microbiol. 12, 722579 (2021). 350 25. Dwidar, M., Jang, H., Sangwan, N., Mun, W., Im, H., Yoon, S., Choi, S., Nam, D. & Mitchell, 351 R. J. Diffusible Signaling Factor, a Quorum-Sensing Molecule, Interferes with and Is Toxic 352 Towards Bdellovibrio bacteriovorus 109J. Microb Ecol 81, 347–356 (2021). 353 26. Dwidar, M., Nam, D. & Mitchell, R. J. Indole negatively impacts predation by B dellovibrio 354 bacteriovorus and its release from the bdelloplast. Environmental Microbiology 17, 1009–1022 355 (2015). 356 27. Mun, W., Kwon, H., Im, H., Choi, S. Y., Monnappa, A. K. & Mitchell, R. J. Cyanide 357 Production by Chromobacterium piscinae Shields It from Bdellovibrio bacteriovorus HD100 358 Predation. mBio 8, e01370-17 (2017). 359 28. Andersson, D. I. & Hughes, D. Antibiotic resistance and its cost: is it possible to reverse 360 resistance? Nat Rev Microbiol 8, 260–271 (2010). 361 29. Andersson, D. I. The biological cost of mutational antibiotic resistance: any practical 362 conclusions? Current Opinion in Microbiology 9, 461–465 (2006). 363 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 15 30. Cai, S. J. & Inouye, M. EnvZ-OmpR Interaction and Osmoregulation in Escherichia coli. 364 Journal of Biological Chemistry 277, 24155–24161 (2002). 365 31. Davis, D. R. & Poulter, C. D. Proton-nitrogen-15 NMR studies of Escherichia coli tRNAPhe 366 from HisT mutants: a structural role for pseudouridine. Biochemistry 30, 4223–4231 (1991). 367 32. Tsui, H. C., Arps, P. J., Connolly, D. M. & Winkler, M. E. Absence of hisT-mediated tRNA 368 pseudouridylation results in a uracil requirement that interferes with Escherichia coli K-12 cell 369 division. J Bacteriol 173, 7395–7400 (1991). 370 33. Gronow, S., Brabetz, W. & Brade, H. Comparative functional characterization in vitro of 371 heptosyltransferase I (WaaC) and II (WaaF) from Escherichia coli. European Journal of 372 Biochemistry 267, 6602–6611 (2000). 373 34. Aminov, R. I. A Brief History of the Antibiotic Era: Lessons Learned and Challenges for the 374 Future. Front. Microbio. 1, (2010). 375 35. Davies, J. & Davies, D. Origins and Evolution of Antibiotic Resistance. Microbiol Mol Biol 376 Rev 74, 417–433 (2010). 377 36. Fang, Q., Feng, Y., McNally, A. & Zong, Z. Characterization of phage resistance and phages 378 capable of intestinal decolonization of carbapenem-resistant Klebsiella pneumoniae in mice. 379 Commun Biol 5, 48 (2022). 380 37. Oechslin, F. Resistance Development to Bacteriophages Occurring during Bacteriophage 381 Therapy. Viruses 10, 351 (2018). 382 38. Maffei, E., Shaidullina, A., Burkolter, M., Heyer, Y., Estermann, F., Druelle, V., Sauer, P., 383 Willi, L., Michaelis, S., Hilbi, H., Thaler, D. S. & Harms, A. Systematic exploration of 384 Escherichia coli phage–host interactions with the BASEL phage collection. PLoS Biol 19, 385 e3001424 (2021). 386 39. Nobrega, F. L., Vlot, M., De Jonge, P. A., Dreesens, L. L., Beaumont, H. J. E., Lavigne, R., 387 Dutilh, B. E. & Brouns, S. J. J. Targeting mechanisms of tailed bacteriophages. Nat Rev 388 Microbiol 16, 760–773 (2018). 389 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 16 40. Jochumsen, N., Marvig, R. L., Damkiær, S., Jensen, R. L., Paulander, W., Molin, S., Jelsbak, L. 390 & Folkesson, A. The evolution of antimicrobial peptide resistance in Pseudomonas aeruginosa 391 is shaped by strong epistatic interactions. Nat Commun 7, 13002 (2016). 392 41. Kenney, L. J. & Anand, G. S. EnvZ/OmpR Two-Component Signaling: An Archetype System 393 That Can Function Noncanonically. EcoSal Plus 9, 10.1128/ecosalplus.ESP-0001–2019 (2020). 394 42. Thanassi, D. G., Suh, G. S. & Nikaido, H. Role of outer membrane barrier in efflux-mediated 395 tetracycline resistance of Escherichia coli. J Bacteriol 177, 998–1007 (1995). 396 43. Choi, U. & Lee, C.-R. Distinct Roles of Outer Membrane Porins in Antibiotic Resistance and 397 Membrane Integrity in Escherichia coli. Front. Microbiol. 10, 953 (2019). 398 44. Delcour, A. H. Outer membrane permeability and antibiotic resistance. Biochimica et 399 Biophysica Acta (BBA) - Proteins and Proteomics 1794, 808–816 (2009). 400 45. Tyson, J., Radford, P., Lambert, C., Till, R., Huwiler, S. G., Lovering, A. L. & Elizabeth 401 Sockett, R. Prey killing without invasion by Bdellovibrio bacteriovorus defective for a 402 MIDAS-family adhesin. Nat Commun 15, 3078 (2024). 403 46. Varon, M. & Shilo, M. Attachment of Bdellovibrio bacteriovorus to cell wall mutants of 404 Salmonella spp. and Escherichia coli. J Bacteriol 97, 977–979 (1969). 405 47. Schelling, M. & Conti, S. Host receptor sites involved in the attachment of Bdellovibrio 406 bacteriovorus and Bdellovibrio stolpii. FEMS Microbiology Letters 36, 319–323 (1986). 407 48. Duncan, M. C., Forbes, J. C., Nguyen, Y., Shull, L. M., Gillette, R. K., Lazinski, D. W., Ali, 408 A., Shanks, R. M. Q., Kadouri, D. E. & Camilli, A. Vibrio cholerae motility exerts drag force 409 to impede attack by the bacterial predator Bdellovibrio bacteriovorus. Nat Commun 9, 4757 410 (2018). 411 49. Wale, N., Sim, D. G., Jones, M. J., Salathe, R., Day, T. & Read, A. F. Resource limitation 412 prevents the emergence of drug resistance by intensifying within-host competition. Proc. Natl. 413 Acad. Sci. U.S.A. 114, 13774–13779 (2017). 414 415 416 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 17 Acknowledgments 417 The authors would like to thank Dr. Timothy Sykes and Dr. Weihong Qi from Functional Genomics 418 Center Zurich (FGCZ) for DNA library preparation, Illumina sequencing and initial data analysis by 419 multi-sample variant calling. Microscopy images were obtained with a Leica Thunder microscope 420 maintained by the Center for Microscopy and Image Analysis at University of Zurich. We would like 421 to thank Prof. R. E. Sockett (University of Nottingham, UK) for providing B. bacteriovorus HD100T. 422 Further, we would like to thank Prof. Alex Hall and Dr. Daniel Angst (both ETH Zürich, CH) for the 423 E. coli strains from the KEIO collection and pCP20. This study was enabled by FAN (Research Talent 424 Development Fund of the University of Zurich Alumni), the Ernst Göhner Stiftung, as well as the Kurt 425 und Senta Herrmann-Stiftung (all to S.M. and S.G.H.), all of which contributed towards research costs 426 and salary of S .M.. R.K. was funded by a project grant from the Swiss National Science Foundation 427 (no. 310030_212266), S .G.H. was funded by Swiss National Science Foundation Ambizione 428 Fellowship PZ00P3_193401. A CC-BY-NC-ND copyright license is applied to this manuscript. 429 430 Author contributions 431 S.M. and S.G.H. jointly conceived the project and obtained funding with input from R.K. . S.M. and 432 O.Ch.M. conducted the prey evolution experiment and isolated genomic DNA from prey populations 433 for sequencing of ancestor and evolved prey lineages. S.M. and O.Ch.M. isolated single clones of 434 evolved prey lineages and sent them for sequencing. S.M. and O.Ch.M. conducted predation assays 435 with prey of ancestor and evolved lineage populations at the end timepoint (15th evolution cycle). S.M. 436 made predation assays of evolved prey lineage populations at 2 nd and 5th evolution cycles as well as 437 evolved single clones (15th evolution cycle). S.L.L. performed predation assays of prey deletion strains 438 with help of S.I. . Data analysis and evaluation of predation assays was done by O.Ch.M., S.L.L., and 439 S.M. (lead). Genome sequence analysis was performed by O.Ch.M. (population), S.G.H. (population, 440 gene mutations), S.L.L. (single clones, OmpF population mutations) and S.M. (population, single 441 clones, all final heatmaps, lead). S.L.L. generated clean deletion strains ∆ompF, ∆waaF, ∆envZ, and 442 ∆truA. S.I. and S.G.H. helped S.L.L. to generate ∆waaB. S.L.L. took microscopy images of prey 443 ancestor and all evolved prey lineage populations (15 th evolution cycle), as well as E.coli BW25113, 444 ∆envZ, and ∆truA, with assistance of S.G.H. . S.L.L. performed image analysis with statistical input 445 by S.M. . S.G.H. advised on predatory bacterial culturing, lab experimentation, prey sequencing, and 446 molecular genetics aspects. R.K. advised on evolutionary and statistical aspects. S.M., S.G.H. and R.K. 447 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 18 wrote the paper with contributions from the other authors. All authors approved of the final version 448 before submission. 449 450 Competing interests 451 The authors declare no competing interests. 452 453 Additional information 454 Extended data is available for this paper at https://doi.org/XXX. 455 Supplementary information The online version contains supplementary material available at 456 https://doi.org/XXX. 457 Requests for materials should be addressed to Simona G. Huwiler. 458 459 460 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 19 Main figures 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 Fig. 1 │ Repeated exposure of predator B. bacteriovorus HD100 leads to growth of prey E. coli 482 K-12 under predation. a, Experimental evolution scheme: eight different prey lineages are exposed 483 repeatedly to different concentration of the predator: no (N1-N8, blue), low (L1-L8, yellow) and high 484 (H1-H8, red) predatory pressure . Each prey evolution cycle is composed of an alternating predation 485 and recovery phase, whereby surviving prey is separated, recovered and re-exposed to predation for 15 486 cycles. Growth dynamics of evolved prey from 15 th cycle: b-d, growth curves measured as OD600nm 487 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 20 over time of eight independently evolved prey lineages under no ( b), low (c) and high ( d) predatory 488 pressure when exposed to respective predation conditions where they evolved. They were compared with 489 ancestor under no (AN -no) or specific predatory pressures (AN -low & AN -high). Each point of the 490 growth curves represents average OD 600nm at that timepoint across replicates. e-g, Relative growth 491 integral of b-d, defined as growth integral (area under the curve) divided by the growth integral of 492 ancestor under no predatory pressure. Boxplots represent the median with 25th and 75th percentiles, and 493 whiskers show the 1.5 interquartile range. Red dotted line and shadow indicate the mean and standard 494 deviation range of AN-no. Black dotted lines and shaded areas indicate the mean and standard deviation 495 range of AN under respective predation conditions. The p -value significance levels (two -sided one 496 sample t-tests, p>0.05 = ns, p<0.05 = *, p<0.01 = **, p<0.001= ***) in comparison to AN-no is shown 497 by red asterisks and comparison to ancestor under respective predatory pressures are shown by black 498 asterisks. Experiments were repeated independently at least twice, N1-8 (n = 10 replicates), L1-8 (n = 499 10 replicates), H1-8 (n = 13 replicates), AN-no (n = 72 replicates), AN-low (n = 10 replicates) and AN-500 high (n = 62 replicates). 501 502 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 21 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 Fig. 2 │ Predation resistance evolves only under predation. Growth dynamics of evolved prey from 518 15th cycle: a & b, growth curves measured as OD600nm over time of eight independently evolved prey 519 lineages under no (a, N1-8, blue) and low (b, L1-8, yellow) predatory pressure when exposed to high 520 predatory pressure and compared with ancestor population under no ( AN-no, red line) and high (AN-521 high, black line) predatory pressure. Each point of the growth curves represents average OD600nm at that 522 timepoint across replicates. c & d, Relative growth integral of a & b, defined as growth integral (area 523 under the curve) divided by the growth integral of AN-no. Boxplots represent the median with 25th and 524 75th percentiles, and whiskers show the 1.5 interquartile range. Dotted lines and shaded areas indicate 525 the mean and standard deviation range of the ancestor under no (AN-no, red) and high (AN-high, black) 526 predatory pressure. The p-value significance levels (two-sided one sample t-tests, p>0.05 = ns, p<0.05 527 = *, p<0.01 = **, p<0.001= ***) in comparison to AN-no are shown by red asterisks, and to AN-high 528 are shown by black asterisks. Experiments were repeated independently at least twice, N1 -8 (n= 10 529 replicates), L1-8 (n= 10 replicates), AN-no (n= 72 replicates), and AN-high (n= 62 replicates). 530 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 22 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 Fig. 3 │ Prey resistance to bacterial predator is generally associated with negative fitness 547 consequences. Growth dynamics of evolved prey from 15 th cycle: a-c growth curves measured as 548 OD600nm over time, of prey lineages evolved under no (a, blue), low (b, yellow) and high (c, salmon) 549 predatory pressure, when exposed to no predatory pressure. Each point of the growth curve represents 550 the average OD600nm at that timepoint across replicates. Panel a is identical with Fig. 1a for a 551 complete overview. d-f, Relative growth rates of no ( d), low ( e), and high ( f) predatory pressure 552 evolved lines, when exposed to no predatory pressure. Relative growth rate is defined as maximum 553 growth rate divided by the maximum growth rate of the ancestor under no predatory pressure (AN-no). 554 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 23 Boxplots represent the median with 25th and 75th percentiles, and whiskers show the 1.5 interquartile 555 range. Red dotted lines and shaded areas indicate the mean and standard deviation range of ancestor 556 under no predatory pressure ( AN-no). The p-value significance levels (two -sided one sample t-tests, 557 p>0.05 = ns, p<0.05 = *, p<0.01 = **, p<0.001 = ***) in comparison to AN-no are shown by red 558 asterisks. Experiments were repeated independently at least twice, N1 -8 (n= 10 replicates), L1 -8 (n= 559 20 replicates), H1-8 (n= 13 replicates), AN-no (n= 72 replicates). Three outliers, one each from N6, 560 L4 and L7 were excluded from panels d & e for better visualisation but were included in the statistical 561 analysis. g, Trade-off between predation resistance and prey fitness. Fitness, quantified as mean 562 relative growth rate under no predatory pressure ( d-f) versus predator resistance, quantified as mean 563 relative growth integral under high predatory pressure (Fig. 1g & 2c-d). N1-8, L1-8, and H1-8 denotes 564 prey lineages that evolved under no (blue), low (yellow) and high (salmon) predatory pressure 565 respectively. A logistic regression function was fitted as a blue line to correlate the two variables 566 (resistance versus fitness, Wald z-statistics = -2.336, p- value = 0.0195). Black error bars represent the 567 standard error of mean for the respective variables. 568 569 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 24 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 Fig. 4 │Gene mutation al patterns in prey population s evolved under repeated predation 589 pressure. Genes mutated in evolved E. coli K-12 lineages after the 15th evolution cycle in comparison 590 to the ancestor. a, Venn diagram showing the number of mutated E. coli genes that reached frequencies 591 higher than 5% within a lineage , grouped according to no, low and high predatory pressure. Not 592 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 25 included in this Venn diagram are mutated rRNA and transposase genes, which are listed separately 593 (Extended Data Fig. 7). b, Hierarchically clustered heatmap of mutation frequency differences relative 594 to the ancestor of all the genes that exclusively mutated under high (H1-8) and/or low (L1-8) predatory 595 pressure. Heatmaps on top of the panel show fitness (blue) measured as average relative growth rate 596 of lineages under no predatory pressure, and resistance (red) measured as average relative growth 597 integral of lineages under high predatory pressure, according to the data in Fig. 3g . c, Overview of 598 mutated genes and mutation types together with a description of the encoded proteins, their functions 599 and cellular location (Loc) for all the genes that exclusively mutated under high and/or low predatory 600 pressure, according to b. 601 602 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 26 603 604 605 606 607 608 609 610 611 612 613 614 615 616 Fig. 5 │ OmpF deletion in E. coli prey results in high predation resistance. Different E. coli clean-617 deletion mutant strains ( ompF, waaF, waaB, envZ, truA) were generated based on the KEIO 618 library. E. coli BW25113 (BW) is the KEIO reference strain. a+c, Growth curve s measured as OD600nm 619 over time of the different E. coli deletion mutants exposed to either high predatory pressure (a) or no 620 predators (c). Each point of the growth curve represents the average OD 600nm per timepoint across 621 replicates. Experiments were independently repeated at least twice . Total replicate numbers for the 622 various strains: ompF (n= 14), waaF (n= 14), waaB (n= 14), envZ (n= 14), truA (n= 8), BW 623 without predatory pressure (BW-no) (n= 14), and BW with high predatory pressure (BW-high) (n= 624 15). b, Relative growth integrals from a. Relative growth integrals are calculated as growth integral 625 (area under the curve) of the clean -deletion mutant with predatory pressure divided by the growth 626 integral of BW-no. d, Relative growth rate s from c. Relative growth rates are calculated as the 627 maximum growth rate of the clean-deletion mutant divided by the maximum growth rate of BW -no . 628 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 27 Boxplots represent the median with 25 th and 75th percentiles, and whiskers show the 1.5 interquartile 629 range. Red and black dotted lines and shaded areas indicate the mean and standard deviation range of 630 BW-no and BW -high respectively. The p -value significance levels (two -sided one sample t-tests, 631 p>0.05 = ns, p<0.05 = *, p<0.01 = **, p<0.001= ***) in comparison to BW -no (red asterisks), and to 632 BW-high (black asterisks). 633 634 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 28

635 Growth of bacterial strains 636 In this study, Escherichia coli K-12 MG1655 (DSM 18039) as prey and Bdellovibrio bacteriovorus 637 HD100T as predator were used , unless indicated otherwise. B. bacteriovorus HD100T was obtained 638 from the laboratory of Prof. R. E. Sockett at University of Nottingham, UK. B. bacteriovorus HD100 639 was cultured and maintained on Ca/HEPES buffer (25 mM HEPES, 2 mM CaCl2, pH 7.6) with late -640 log E. coli K-12 MG1655 at 29°C for generally 24 h at 200 rpm shaking 1. General revival of 641 B. bacteriovorus and determination of plaque forming units (PFU) for enumeration was done according 642 to protocols by Lambert & Sockett1 and is described in more detail at the end of the Additional Online 643 materials. 644 All E. coli strains used are listed in Additional Online Materials Table SOnline1. To validate the 645 genetic effect on predation resistance, markerless single deletion strains envZ, ompF, truA, waaB, 646 and waaF were generated based on E.coli KEIO library strains2 containing an inserted kanamycin 647 resistance cassette (following section). The different E. coli strains were generally grown in YT both 648 (Yeast tryptone, 5 g l-1 NaCl, 5 g l-1 Bacto yeast extract, 8 g l-1 tryptone, pH 7.5) at 37°C with 200 rpm 649 shaking for generally 16 h overnight, and additional appropriate antibiotics as indicated. For ancestor 650 prey, E. coli K-12 MG1655 was retrieved from 20% glycerol stocks at -80°C and were streaked on a 651 YT agar plate (YT broth with 10 g l-1 agar) and incubated at 37 C for 24 h to obtain single colonies. 652 The overnight cultures from ancestor prey single colonies were used for 24-hr lysates (growth of 653 predator), revival and enumeration of the predator in our experiments. 654 655 Generation of markerless deletion strains of E. coli KEIO library 656 Markerless single deletion mutants in E. coli (envZ, ompF, truA, waaB, and waaF) were 657 constructed of the respective strain from the KEIO collection (based on E. coli BW25113)2. To achieve 658 this, the kanamycin resistance cassette (present in the KEIO collection to disrupt the individual genes) 659 was removed to enable predation assays with kanamycin-sensitive B. bacteriovorus HD100. The 660 kanamycin cassette was excised via pCP20 3, expressing an FLP -FRT recombinase. To generate 661 electrocompetent cells, and excise the kanamycin cassette, several protocols from the Barrick lab were 662 adapted4,5. To prepare electrocompetent cells the desired E. coli strain in YT broth with 30 μg/ml 663 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 29 kanamycin was incubated for 16 h at 37°C and 200 rpm. The stationary phase culture was used to 664 inoculate at least 10 ml of YT broth with 30 μg/ml kanamycin at optical density with absorbance at 665 600nm (OD600nm) of 0.05 and grown at 37°C and 200 rpm. At an OD 600nm of ~ 0.6 the culture was 666 centrifuged for 5 minutes at 6000 x g and 4°C, and the supernatant was discarded. The resulting pellet 667 was washed 4 times with 10% ice-cold glycerol of the same volume as the initial inoculation volume. 668 Finally, the pellet was resuspended in 10% glycerol at 100x concentration, and 50 μl aliquots stored at 669 -80°C. These electrocompetent E. coli cells were thawed on ice and maximally 100 ng of pCP20 added. 670 After gentle mixing, the sample was transferred into a cool ed 0.2-cm cuvette (Bio -Rad), and 671 electroporated at 2.5 kV with a MicroPulser electroporator (Bio -Rad). Then, the sample was 672 resuspended in 500 μl warm SOC medium and incubated for 30 min at 30°C and 120 rpm. The cells 673 were plated on YT agar plates with 100 μg/ml ampicillin and grown at 29°C for at least 16 h. To select 674 for loss of heat -sensitive plasmid pCP20 after recombination, a single colony from a YT agar plate 675 with 100 μg/ml ampicillin was picked to inoculate 5 ml of YT broth, and grown for 16 h at 43°C, 676 shaking at 200 rpm. 50 μl of a 10 -6 dilution of this stationary phase culture was plated on YT agar 677 without antibiotics for single clones. To screen for successful recombination and loss of plasmid, six 678 colonies were selected and streaked out on YT agar with 30 μg/ml kanamycin, YT agar with 100 μg/ml 679 ampicillin, and YT agar without antibiotics in this order. The YT plates without and with kanamycin 680 were incubated at 37°C, the YT plates with ampicillin at 29°C for at least 16 h. To screen for successful 681 excision of the kanamycin cassette, recombinants with sensitivity to both antibiotics were grown in 682 5 ml of YT broth, incubated for 16 h at 200 rpm and 37°C. Of these cultures, 20% glycerol stocks were 683 made and stored at -80°C. To control correct excision of the kanamycin cassette, PCR was used to 684 assess correct product size (Primers in Additional Online Materials Table SOnline 2). Additionally, 685 correct scar regions were confirmed by sanger sequencing (Microsynth AG, CH). 686 687 Experimental prey evolution 688 In the experimental prey evolution , E. coli K-12 MG1655 was used as ancestral prey strain and 689 subjected to three different predatory pressures by B. bacteriovorus HD100 involving 15 alternate 690 cycles of predation and recovery phase (Fig. 1a). The detailed protocol is described at the end of the 691 Supplementary Online Materials. The predation phase occurred in rich medium (containing ¾ YT, 1/4 692 25 mM HEPES and 2 mM CaCl 2) allowing growth of E. coli K-12 as well as predation of 693 B. bacteriovorus at 200 rpm and 29°C. E. coli K-12 prey were subjected to three different predatory 694 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 30 conditions: No predatory, low predatory pressure ( multiplicity of infection [MOI], ratio of predator 695 (PFU/ml) vs prey cells (CFU/ml), was ~ 2.5) and high predatory pressure (MOI ~ 10). For each 696 condition eight independent prey lineages were evolved from the E. coli K-12 ancestor. After the 24-697 hr predation phase surviving E. coli were separated from B. bacteriovorus by filtration to avoid co -698 evolution. In the subsequent 16-hr recovery phase retrieved E. coli from filtration were diluted ~1:100 699 and grown in YT at 200 rpm and 37°C. These overnight cultures were diluted back to an overall initial 700 OD600nm of 0.003 for the next predation phase and evolution cycle. 701 702 Isolation of prey single clones 703 Prior to the experiment we made overnight cultures of evolved lineages under high predatory pressure 704 (15th evolution cycle) stored at -80C, by extracting cell material (containing evolved population) from 705 cryo stocks and inoculating 50 ml YT medium in 250 ml Erlenmeyer flasks, incubated at 37 C, 706 200 rpm for 16 h. We harvested prey cells from 10ml of these overnight cultures by centrifugation at 707 5311 x g for 3 min and subsequently washed them in 10 ml of Ca -HEPES. Then prey cells were 708 resuspended in 6 ml of Ca-HEPES and adjusted to OD600nm = 3.0, then we diluted the cultures to 10 -6 709 and plated 100 l on YT agar plates and incubated them at 37 C for 24 h. We randomly selected 6 710 single colonies from each evolved lineage and overnight cultures were made with these single colonies 711 in 10 ml YT medium in 50 ml Falcon tubes incubated at 37C, 200 rpm for 16 h. From this culture we 712 prepared 20% glycerol stocks and stored them at -80C. Prior to experiments, we made overnight 713 cultures of the single colonies from cryo stocks, alongside their respective evolved lineage population 714 and ancestor in 50 ml YT medium in 250 ml Erlenmeyer flasks, incubated at 37 C, 200 rpm for 16 715 hours. These single clone cultures were used for predation assays and genome sequencing. 716 717 Predation assay 718 To evaluate predation resistance and growth dynamics of different prey, predation assays were 719 performed. Different E. coli prey strains were assessed: E. coli K-12 MG1655 population lineages 720 evolved under high (H1-8), low (L1-8) and no (N1-8) predatory pressure at the end 15th evolution cycle 721 (after recovery phase) were compared with ancestor E. coli K-12 MG1655. At first, we exposed prey 722 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 31 lineages that evolved under a particular predatory pressure to that pressure. We then focussed on 723 exposing evolved lines to the high predatory pressure condition. 724 To investigate the temporal aspect of evolution, we performed predation assays of prey lineages 725 evolved under high predatory pressure (H1 -H8) at the end of the 2 nd and 5 th evolution cycle and 726 compared them to the 15th cycle, when exposed to high predatory pressure. To test for cross resistance, 727 we assessed the performance of prey that evolved under no or low predator by exposing them to high 728 predatory pressure. The predation assays at the population level featured 5 replicates per prey-predator 729 combination and were repeated on two separate days. 730 To assess effects of specific prey genes on predation resistance, predation assay was conducted with 731 markerless single gene deletion mutants (envZ, ompF, truA, waaB, and waaF based on E. coli 732 BW25113 (reference strain from KEIO library2). These strains were exposed to high and no predatory 733 pressure conditions. 734 Prior to the predation assay, respective evolved prey lineages or prey strains (evolved lineages, 735 population- or single clone-based, single gene deletion, and KEIO reference strain) were grown from 736 cryo stocks by inoculation in 50 ml YT broth and incubation at 37 C and 200 rpm for 16 h. 737 Additionally, an overnight culture of ancestor prey was prepared from a single colony, at same growth 738 conditions. The eight independently evolved prey lineages under each predatory pressure were assessed 739 in parallel alongside ancestor prey. 10 -ml prey overnight cultures were centrifuged at 5311 x g for 3 740 min and subsequently washed with 10 ml of Ca-HEPES buffer. Then prey cells were resuspended in 741 6 ml of Ca-HEPES buffer and adjusted to OD600nm = 0.03. A 24-hr predator lysate was filtered through 742 a 0.8-m syringe filter (Whatman® Puradisc FP 30, Cellulose Acetate, sterile, WHA10462240). For 743 the predation assays, the same cultivation media was used as for the experimental evolution predation 744 phase, comprising 15 ml of YT broth, 5 ml of 25 mM HEPES and 2 mM CaCl2 overall. We set up the 745 predation assay in 96-well plates, in a total volume of 150 l. Initially 119 l of cultivation media was 746 preset, and different volumes of the filtered predator lysate were added to the assay: 0 l (no predator 747 exposure), 4 l (low predator exposure) or 16 l (high predator exposure), depending on the evolved 748 prey lineages and predatory pressure tested. Then we added 15 l of the respective prey solution 749 (OD600nm = 0.03 in Ca -HEPES). Then we adjusted the final volume in reaction wells to 150 l by 750 adding Ca -HEPES (as proxy for predator lysate and prey resuspension), resulting in a final 751 concentration of prey of OD 600nm = 0.003. We used the same ratio of prey and predator as in the 752 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 32 predation phase of the prey evolution experiment. The plate was incubated in a plate reader (Tecan 753 Infinite 200 Pro) at 29 C for 24 h. OD600nm of the reaction wells was measured with 10 flashes every 754 5 minutes after shaking at 6-mm amplitude for 120 secs. 755 We used OD600nm as proxy for the biomass of prey cells, as the predator cells only contribute to a very 756 minor fraction to OD600nm based on the small size. The kinetic measurement of OD600nm over 24 h, was 757 analysed using an interactive platform for analysis of microbial growth data 758 (https://statu.shinyapps.io/ExploreMicrobialGrowth/, version 2016.01.12) designed by Adin Ross -759 Gillespie at University of Zurich. We used Spline function to fit the kinetic growth data to extract the 760 integral (area under the curve) and maximal growth increase rate. For the predation assays in cultivation 761 media, we derived relative growth integral and growth rate of evolved prey lineages in comparison to 762 the ancestor prey. For this we divided the growth integral or growth rate of evolved lineages under all 763 predator conditions by that of ancestor when exposed to no predator, that is baseline. Values > 1 or < 764 1, indicate that evolved lineages perform better or worse than baseline. 765 766 Cell shape analysis by phase-contrast microscopy 767 E. coli K-12 lineages evolved under low, no, and high predatory pressure at the end of experimental 768 evolution (after the 15 th evolution cycle), E. coli K-12 MG1655 (ancestor), as well as E. coli K-12 769 BW25113, ∆envZ and ∆truA were regrown from cryo stocks in YT broth at 37°C and 200 rpm shaking 770 overnight. From each culture 10 µl was applied onto a 1% agarose in Ca/HEPES buffer pad on a clean 771 microscope slide for phase-contrast microscopy. Imaging was performed using a Leica DMi8 inverted 772 microscope equipped with a 100x objective (HC PL APO 1.40 oil PH3), a Leica monochrome 773 fluorescence DFC9000 gTC Camera System, and Leica Application Suite X Software Version 774 3.7.6.25997 (Leica microsystems). Exposure time was 100 ms with illumination intensity of 140. For 775 each evolved lineage and ancestor strain E. coli K-12 MG1655 a minimum of 15 images were 776 collected. For shape analysis of the evolved cells, software Fjij/ImageJ (Version 2.90 /1.53t, Java 777 1.8.0322, 64-bit)6,7 was used with the Plugin MicrobeJ (Version 5.13o (16) – beta)8. Area and length 778 restrictions were set to 0.5-3 p2, width to 0.5-2 p2. After initial shape detection, masks were manually 779 adjusted to observed cell size and a minimum of 300 cells per lineage/strain were analysed. 780 781 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 33 Statistical analyses 782 We used general linear models for statistical analysis in R 4.2.39,10. To assess whether relative growth 783 integral and relative growth rate vary in response to predatory pressure we used analysis of variance 784 (ANOVA) models (Fig. 1e-g, Fig. 2 c,d & Fig. 3d-f). We further used two-sided one-sample t-tests to 785 test for significant differences in relative growth integral or growth rate between evolved prey lineages 786 and that of ancestor when exposed to either no, low or high predatory pressure (i.e. being different 787 from mean relative growth integral or growth rate of ancestor, Fig. 1e-g & Fig. 2c-d & Fig. 3d-f & Fig. 788 5b,d & Extended Data Fig. 2d -f, j-l). Reported p-values for t-tests were adjusted for multiple testing 789 using the Holm method. We used a logistic regression model to test for a negative association (tradeoff) 790 between mean relative growth integral and mean relative growth rate (Fig. 3g). A detailed statistical 791 output of the data of this article is available in an Additional Online Material table. 792 793 Genome sequencing and analysis of evolved prey population lineages 794 We sequenced the genome of the 24 prey lineages (i.e. population) that evolved under no, low and high 795 predatory pressure condition (8 each) after the 15 th evolution cycle, at the population level and 796 compared it with the ancestor prey genome. Overnight cultures were grown in YT broth (200 rpm at 797 37°C) from cryo stocks of evolved prey lineages from the 15 th recovery phase at the end of 798 experimental evolution and separately from a single colony of the ancestor prey. From these cultures 799 we extracted genomic DNA using GenElute TM Bacterial Genomic DNA Kit (Sigma , NA2110). The 800 concentration of genomic DNA varied between 83.77 to 205. 407 ng/ l, which was sufficient for 801 standard library preparation. The samples were then transferred to the Functional Genomic Center 802 Zurich (FGCZ, Switzerland) for sequencing and base variant calling of the sequencing data of evolved 803 lineages against the ancestor. For library preparation the TruSeq DNA Nano Sample Prep Kit v2 804 (Illumina, Inc, California, USA) was used in the succeeding steps. DNA samples (100ng) were 805 sonicated with the Covaris using settings specific to the fragment size of 350 bp. The fragmented DNA 806 samples were size - selected using AMpure beads, end -repaired and adenylated. TruSeq adapters 807 containing Unique Dual Indices (UDI) for multiplexing were ligated to the size-selected DNA samples. 808 Fragments containing TruSeq adapters on both ends were selectively enriched by PCR. The quality 809 and quantity of the enriched libraries were validated using TapeStation (Agilent, Waldbronn, 810 Germany). An average fragment size of approximately 500 bp was achieved . The libraries were 811 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 34 normalized to 10 nM in Tris -Cl 10 mM, pH 8.5 with 0.1% Tween 20. For sequencing and cluster 812 generation the Novaseq 6000 (Illumina, Inc, California, USA) was used according to standard protocol. 813 Sequencing were paired end at 2 X150 bp. Technical quality of Illumina paired -end (PE) reads was 814 evaluated using FastQC (v0.11.9) and FastQScreen (v0.14.1). Raw reads were pre -processed using 815 fastp (v0.20.0) with the following parameter: hard trimming of 10 nucleotide (nt) and 1 nt at 5' and 3 816 ends, respectively; quality trimming of both ends using a sliding window of 4 nt and mean quality of 817 20; trimmed reads shorter than 18 nt, and trimmed reads with mean read quality lower than 20 were 818 filtered out. Pre-processed PE reads were aligned to the reference genome (E. coli K-12 MG1655 refseq 819 GCF_000005845.2) using bwa (v0.7.17) with the "mem" algorithm. PCR duplicates were marked 820 using Picard (v2.22.8, MarkDuplicates). Mapping results were quality controlled using SAMStat 821 (v1.5.1), Qualimap (v2.2.1), and Picard (v2.22.8, CollectGcBiasMetrics). Multi-sample variant calling 822 was performed using freebayes (V1.3.2) with frequency cutoffs of 5% and 1%, respectively. Only 823 unique reads in primary alignments with alignment quality 30 and above, and bases with quality 20 824 and above, were counted for variant analysis. Sites with multiple alternative alleles were split into 825 multiple rows using bcftools. Variant annotation was performed using snpEff (v4.3) against reference 826 gene models (NCBI, reference genome: ASM584v2, refseq: GCF_000005845.2). Called variants were 827 filtered against the ancestor strain using a customized Perl script. Markers were retained, where all 828 counted reads contain only the reference alleles in the ancestor strain. Per sample variant frequencies 829 were then calculated for these markers using the customized Perl script. The resulting excel file from 830 population variant calling was used for further analysis and representation by hierarchically clustered 831 heatmaps on absolute mutation frequencies and their differences compared to the ancestor strain. 832 Heatmaps were generated in R with Pheatmap11. We focused on gene mutations with a frequency more 833 than 5%. If a particular gene had multiple mutation effects, we selected the mutation with the highest 834 frequency. Mutation counts were added to a gene only if the mutation was located within the coding 835 region of a specific gene (direct effects). This excludes downstream effects on adjacent genes. 836 Mutations lying in intergenic regions were ignored due to their ambiguous impact on surrounding 837 genes. An exception was made for the mutation located in the intergenic region between asnS and 838 ompF. Since it is located 143 bp upstream the ompF transcription start site various ompF regulators 839 bind adjacent to this locus and is likely to influence ompF transcription. The mutations in ompF were 840 analysed in detail and the OmpF amino acid sequence with changes in the different prey lineages was 841 constructed in Excel (Microsoft Corporation, 2018) and compared with the ancestor strain. Uniquely 842 mutated genes (without transposases and rRNA genes) of all genomes of prey lineages at the respective 843 predatory pressure were plotted using R package VennDiagram12. 844 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 35 845 Genome sequencing and variant calling for prey single clones 846 48 cultures were prepared from six single prey clones each of all eight prey lineages evolved under 847 high predatory pressure (H1 -H8) at the end of the 15 th evolution cycle. Overnight cultures of these 848 single clones from lineages evolved under high -predatory pressure were inoculated from cryo stocks 849 into 10 ml YT medium in 50 ml Falcon tubes, and incubated at 37C, 200 rpm for 16 h. We harvested 850 prey cells equivalent of OD600nm = 10.0 from these cultures by centrifugation at 5000 g for 5 min, the 851 cells were then washed (5000 g for 5 minutes) in 1 ml Phosphate buffered saline (Gibco) and 852 resuspended into 0.5ml DNA/RNA shield buffer (Zymo Research) . This mixture was sent to 853 MicrobesNG (UK, https://microbesng.com, Birmingham, UK ), for DNA extraction, sequencing and 854 base variant calling against reference genome E. coli K-12 MG1655. 855 DNA was extracted by lysing 4 -5 µl cell suspension with 120 μL of TE buffer containing lysozyme 856 (MPBio, USA) , metapolyzyme (Sigma Aldrich, USA) and RNase A (ITW Reagents, Spain), and 857 incubated for 25 min at 37°C. Proteinase K (VWR Chemicals, Ohio, USA) (final concentration 858 0.1mg/mL) and SDS (Sigma -Aldrich, Missouri, USA) (final concentration 0.5% v/v) are added and 859 incubated for 5 min at 65°C. Genomic DNA was purified using an equal volume of SPRI beads and 860 resuspended in EB buffer (10mM Tris -HCl, pH 8.0). The DNA was quantified with the Quant -iT 861 dsDNA HS (ThermoFisher Scientific) assay in an Eppendorf AF2200 plate reader (Eppendorf UK Ltd, 862 United Kingdom) and diluted as appropriate. Genomic DNA libraries are prepared using the Nextera 863 XT Library Prep Kit (Illumina, San Diego, USA) following the manufacturer’s protocol with the 864 following modifications: input DNA is increased 2 -fold, and PCR elongation time is increased to 865 45 sec. DNA quantification and library preparation are carried out on a Hamilton Microlab STAR 866 automated liquid handling system (Hamilton Bonaduz AG, Switzerland). Libraries are sequenced on 867 an lllumina NovaSeq 6000 (Illumina, San Diego, USA) using a 250 bp paired end protocol. For projects 868 predating April, 2020, libraries were sequenced on an Illumina HiSeq 2500 using a 250 bp paired-end 869 protocol (Illumina, San Diego, USA). Reads are adapter trimmed using Trimmomatic version 0.30 13 870 with a sliding window quality cutoff of Q15. De novo assembly is performed on samples using SPAdes 871 version 3.7 14, and contigs are annotated using Prokka version 1.11 15. Reads were aligned to the 872

using BWA mem v0.7.17 16 and are then processed using SAMtools v1.9 17. Variants were 873 called using VarScan v2.4.0 18 with two thresholds, sensitive and specific, where the variant allele 874 frequency is greater than 10% respectively. The effects of the variants are predicted and annotated 875 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 36 using SnpEff v4.3 19. The base variant calling was performed using the criteria of a sequence coverage 876 of at least 3x and a frequency of at least 10% to generate a table reporting on the differences to the 877

genome ( NCBI, refseq: GCF_000005845.2) . The resulting Microsoft Excel table obtained 878 from Microbes NG was used to combine it with the population sequencing data based on the nucleotide 879 position. Gene annotations were obtained from both population and clone sequencing data, with 880 frequency more than 5%. 881 882 Data availability. The raw data (OD600nm measurements over time, microscopy images and raw data 883 on width and length, R code used for analysis) used to generate the findings of this project have been 884 deposited in the figshare repository (doi: XXX). The sequencing data for this study has been deposited 885 in the European Nucleotide Archive (ENA) at EMBL -EBI under accession number PRJEB78919 886 (https://www.ebi.ac.uk/ena/browser/view/PRJEB78919). 887 888 889

(Methods): 890 1. Lambert, C. & Sockett, R. E. Laboratory maintenance of Bdellovibrio. Curr Protoc Microbiol 891 Chapter 7, Unit 7B.2 (2008). 892 2. Baba, T., Ara, T., Hasegawa, M., Takai, Y., Okumura, Y., Baba, M., Datsenko, K. A., Tomita, 893 M., Wanner, B. L. & Mori, H. Construction of Escherichia coli K‐12 in‐frame, single‐gene 894 knockout mutants: the Keio collection. Molecular Systems Biology 2, 2006.0008 (2006). 895 3. Datsenko, K. A. & Wanner, B. L. One-step inactivation of chromosomal genes in Escherichia 896 coli K-12 using PCR products. Proc. Natl. Acad. Sci. U.S.A. 97, 6640–6645 (2000). 897 4. Barrick Lab. Barrick Lab :: ProtocolsElectrocompetentCells. at 898 899 5. Barrick Lab. Barrick Lab :: ProcedureFLPFRTRecombination. at 900 901 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 37 6. Schindelin, J., Arganda-Carreras, I., Frise, E., Kaynig, V., Longair, M., Pietzsch, T., Preibisch, 902 S., Rueden, C., Saalfeld, S., Schmid, B., Tinevez, J.-Y., White, D. J., Hartenstein, V., Eliceiri, 903 K., Tomancak, P. & Cardona, A. Fiji: an open-source platform for biological-image analysis. 904 Nat Methods 9, 676–682 (2012). 905 7. Rueden, C. T., Schindelin, J., Hiner, M. C., DeZonia, B. E., Walter, A. E., Arena, E. T. & 906 Eliceiri, K. W. ImageJ2: ImageJ for the next generation of scientific image data. BMC 907 Bioinformatics 18, 529 (2017). 908 8. Ducret, A., Quardokus, E. M. & Brun, Y. V. MicrobeJ, a tool for high throughput bacterial cell 909 detection and quantitative analysis. Nat Microbiol 1, 16077 (2016). 910 9. R Core Team. R: A language and environment for statistical computing. (Vienna, Austria, 911 2021). at 912 10. Wickham, H., Averick, M., Bryan, J., Chang, W., McGowan, L., François, R., Grolemund, G., 913 Hayes, A., Henry, L., Hester, J., Kuhn, M., Pedersen, T., Miller, E., Bache, S., Müller, K., 914 Ooms, J., Robinson, D., Seidel, D., Spinu, V., Takahashi, K., Vaughan, D., Wilke, C., Woo, K. 915 & Yutani, H. Welcome to the Tidyverse. JOSS 4, 1686 (2019). 916 11. Kolde, R. Pheatmap: pretty heatmaps. (R Package Version 1.0.10., 2019). at 918 12. Chen, H. VennDiagram: Generate High-Resolution Venn and Euler Plots. (R package version 919 1.7.3, 2022). at 920 13. Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina sequence 921 data. Bioinformatics 30, 2114–2120 (2014). 922 14. Bankevich, A., Nurk, S., Antipov, D., Gurevich, A. A., Dvorkin, M., Kulikov, A. S., Lesin, V. 923 M., Nikolenko, S. I., Pham, S., Prjibelski, A. D., Pyshkin, A. V., Sirotkin, A. V., Vyahhi, N., 924 Tesler, G., Alekseyev, M. A. & Pevzner, P. A. SPAdes: A New Genome Assembly Algorithm 925 and Its Applications to Single-Cell Sequencing. Journal of Computational Biology 19, 455–477 926 (2012). 927 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint E. coli escapes predation 38 15. Seemann, T. Prokka: rapid prokaryotic genome annotation. Bioinformatics 30, 2068–2069 928 (2014). 929 16. Li, H. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. 930 Preprint at http://arxiv.org/abs/1303.3997 (2013) 931 17. Danecek, P., Bonfield, J. K., Liddle, J., Marshall, J., Ohan, V., Pollard, M. O., Whitwham, A., 932 Keane, T., McCarthy, S. A., Davies, R. M. & Li, H. Twelve years of SAMtools and BCFtools. 933 GigaScience 10, giab008 (2021). 934 18. Koboldt, D. C., Chen, K., Wylie, T., Larson, D. E., McLellan, M. D., Mardis, E. R., Weinstock, 935 G. M., Wilson, R. K. & Ding, L. VarScan: variant detection in massively parallel sequencing of 936 individual and pooled samples. Bioinformatics 25, 2283–2285 (2009). 937 19. Cingolani, P., Platts, A., Wang, L. L., Coon, M., Nguyen, T., Wang, L., Land, S. J., Lu, X. & 938 Ruden, D. M. A program for annotating and predicting the effects of single nucleotide 939 polymorphisms, SnpEff: SNPs in the genome of Drosophila melanogaster strain w 1118 ; iso-2; 940 iso-3. Fly 6, 80–92 (2012). 941 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted October 17, 2024. ; https://doi.org/10.1101/2024.09.27.615459doi: bioRxiv preprint