Protective microbe enhances colonisation of a novel host species by modifying immune gene expression

Microbes that protect against infection inhabit hosts across the tree of life. It is unclear whether many protective microbes use or reduce the need for a host immune response, or how the immune system reacts when these microbes newly encounter a host species naturally and as part of a biocontrol strategy. We sequenced the transcriptome of a host (Caenorhabditis elegans) following its interaction with a non-native bacterium (Enterococcus faecalis) that has protective traits against the pathogen, Staphylococcus aureus. We show that microbe-mediated protection caused the differential expression of 1,557 genes, including the upregulation of many immune gene families conserved across the animal kingdom (e.g. lysozymes and c-type lectins). We found that this modulation of the host’s immune response was beneficial for both the protective microbe and the host. Given E. faecalis’ increased ability to resist lysozyme activity compared to S. aureus, our results indicate that the protective microbe could more easily invade and protect infected hosts by upregulating lysozyme genes. These results suggest that a protective microbe can exploit the host immune system even when introduced into a novel species. Microbes that protect via the host immune response in this way should favour continued investment into host immunity and avoid the evolution of host dependence. Author summary Organisms can be protected from infectious disease by the microbes they house. It is unclear, however, whether protective microbes affect the host immune response to infection, particularly in the early stages of symbiosis. In this study, we investigated the role of the host immune system in a novel protective interaction. We examined gene expression in a nematode after colonisation by a non-native microbe capable of suppressing the pathogen Staphylococcus aureus. The protective microbe altered the host immune response to infection in a way that it could exploit. By causing the host to increase the production of antimicrobials to which it itself is relatively resistant, the protective microbe was better able to colonise and defend infected hosts. These results indicate that protective microbes introduced into new host species can take advantage of the host immune system. Such a mechanism at the beginning of a protective symbiosis, formed either naturally or as part of a biocontrol strategy, could ensure continued investment in host-based defences over evolutionary time.

lectins). We found that this modulation of the host's immune response was beneficial for both 23 the protective microbe and the host. Given E. faecalis' increased ability to resist lysozyme 24 activity compared to S. aureus, our results indicate that the protective microbe could more 25 easily invade and protect infected hosts by upregulating lysozyme genes. These results suggest 26 that a protective microbe can exploit the host immune system even when introduced into a 27 novel species. Microbes that protect via the host immune response in this way should favour 28 continued investment into host immunity and avoid the evolution of host dependence. 29

Introduction 43
Microbes that defend against infection by pathogens (along with parasites and parasitoids) 44 inhabit a large diversity of plant and animal hosts(1, 2). Protection is most commonly found 45 when symbiotic microbes suppress pathogen growth and establishment by directly competing 46 for resources/space or producing antimicrobial compounds/toxins(2, 3). Microbes can also 47 control infection by modulating host immune systems(2, 3). Modification of the host's immune 48 response against infection is a less commonly documented protective mechanism(2), although 49 is unclear whether this pattern is real as immune-mediation may be more difficult to empirically 50 test in non-model systems. An example is seen in the tropical tree Theobroma cacao where the 51 foliar endophytic fungus Colletotrichum tropicale upregulates host immune and defence genes 52 increasing resistance to damage from pathogens(4). 53 54 Interest in the mechanisms underpinning microbe-mediated protection has been surging 55 because of its potential use in public health(5), species conservation(6) and agriculture (7). 56 Although new protective symbioses form naturally(8-11), their creation is being rapidly 57 pursued for the biocontrol of infectious disease, either by introducing existing symbionts into 58 new hosts(12), or by generating new symbionts, e.g. via paratransgenesis (13)(14)(15). For example, 59 Aedes and Anopheles mosquitoes have been artificially infected with strains of the inherited 60 symbiont Wolbachia that inhibit the replication of dengue and Zika viruses, as well as 61 malaria (5,(16)(17)(18). In addition to determining how a microbe might protect a novel host, the 62 specific mechanism of protection might predict the persistence of the microbial symbiont (2), 63 as well as the vulnerability of the host to infection should the mutualism break down (2,(19)(20)(21)(22). 64 If protective microbes directly suppress infection, there is less need for hosts to defend 65 themselves. This outcome could result in reduced investment into costly host-based immune 66 or damage response systems (1,2,23,24). Conversely, a mechanism of protection that involves 67 enhancing its protective ability. That this protective microbe was able to exploit the immune 93 system of a novel host species suggests the potential for the immune system to be involved in 94 the formation of protective symbioses more widely in nature or as part of a biocontrol strategy.

C. elegans' response to infection.
Populations of young adult worms were exposed to the pathogen S. aureus with or without E. faecalis-mediated protection. After 12h exposure, the RNA from approximately 1,000 worms per sample was sequenced. This was repeated five independent times for each treatment. We used RNA-sequencing to assess the transcriptional changes in C. elegans hosts during 108 infection by virulent S. aureus, with and without protection from E. faecalis (Fig.1). We found 109 that protection drove the significant differential expression of 1,557 genes (Supplementary File 110 1). Of these genes, 521 were downregulated in protected hosts, whilst 1,036 were upregulated. 111 We performed a GO-term enrichment analysis on this list of genes and identified significantly 112 enriched GO-terms in biological processes including "defence response to bacterium" 113 (GO:0042742, FDR-adjusted P= 6.12E-09, Supplementary File 2) and "innate immune

Novel protective microbe upregulates immune genes in infected hosts 144
We identified many lysozyme-encoding genes that were differentially expressed during 145 infection in E. faecalis-protected hosts (Table 1). Lysozymes are important in the host defence 146 response to bacteria as these enzymes break down bacterial cell walls(30). We found that 147 protection by E. faecalis caused the upregulation of the invertebrate lysozyme genes, ilys-2 and 148 ilys-5 compared to when worms were only infected by S. aureus (ilys-2 beta=1.171, P=6.28E-149 06; ilys-5 beta=0.85, P=0.00055). Previous studies found that RNAi knockdown of ilys-2 made 150 C. elegans hosts more susceptible to S. aureus infection(31). 151

152
We also found a significant upregulation in the lysozyme genes, lys-1 (beta=0.77, P=0.00028), 153 lys-2 (beta=0.78, P=3.01E-05), lys-7 (beta=2.12, P=1.15E-07) and lys-10 (beta=3.46, P=4.36E-154 05). Only lys-3 was significantly downregulated (beta=-1.2, P=0.00025). The most 155 significantly upregulated lysozyme gene was lys-7. Although both E. faecalis and S. aureus 156 have been reported to be resistant to lysozyme activity, E. faecalis is much more robust to 157 attack by these enzymes, with a minimum inhibitory concentration (MIC) of >62.5mg/ml 158 lysozyme(32). S. aureus has lower reported MICs of 15mg/ml(33). 159 160 Microbe-mediated protection also resulted in the differential regulation of 28 C-type lectin 161 (clec) genes which are carbohydrate-binding proteins that play a host defensive role against 162 gram-positive bacteria(34). We found that 20 clec genes were upregulated, whilst only seven 163 were downregulated (Table 1). Infection by S. aureus has been previously shown to upregulate 164 clec genes in C. elegans, e.g. clec-72 and clec-52 (35). In our study, clec-72 was upregulated 165 in protected hosts, whilst clec-52 was downregulated. 166 Table 1. Immune gene families differentially regulated by E. faecalis-mediated protection in C. elegans. Genes are listed in order of q-value (p-values corrected by Benjamini-Hochberg FDR for multiple testing). The beta value is the effect size of differential regulation. Blue indicates downregulation whilst red indicates upregulation. Genes in bold are represented twice as unique coding sequences.
Immune We additionally found numerous C. elegans cuticle genes that were differentially regulated by 167 protection. These genes were entirely upregulated and consisted of 44 col and 8 dpy genes 168 (Supplementary File 1). Both col and dpy genes are involved in the extracellular matrix and 169 the cuticle. It is unclear what role these genes play in response to microbe-mediated protection, 170 but genes encoding structural constituents of the cuticle are found to be significantly 171 downregulated during S. aureus infection alone (35). 172 173 Lysozyme expression determines protective microbe colonisation in novel infected host 174 Lysozymes are a well-known conserved antimicrobial defence(30). Given that E. faecalis is 175 reported to be more resistant to lysozyme activity than S. aureus(32, 33), we tested whether 176 the upregulation of lysozyme genes by E. faecalis played a role in protection against S. aureus. 177 We focused on the impact of the most significantly differentially expressed lysozyme gene, 178 lys-7, on bacterial colonisation and protection. 179

180
We obtained a lys-7 gene knock-out in an N2 background (strain CB6738, Caenorhabditis 181 Genetics Center, Minnesota) and tested whether lys-7 affected host mortality and bacterial 182 colonisation during microbe-mediated protection from infection. We found that host mortality 183 was higher in lys-7 knock-out hosts than the wild-type during co-colonisation with E. faecalis 184 and S. aureus, indicating that it plays a role in resistance amidst microbe-mediated protection 185 (Fig 3a. Quasibinomial GLM. host strain: F=14.7, df=1, P= 0.005). We found that E. faecalis 186 was significantly better at colonising wild-type hosts than the lys-7 knock-out mutant (Fig 3b.  187 t-test, t=-3.5, df=8, P= 0.0076). By contrast, S. aureus reached higher infection loads in the lys-188 7 knock-out mutant (Fig 3b. t-test, t=1.42, df=8, P=0.19). These data indicate that expression 189 of lys-7 yields a host-derived antimicrobial that negatively effects S. aureus more than E. 190 faecalis under co-colonisation. This differential effect is consistent with the finding that E. 191 faecalis has a much higher MIC for lysozyme than S. aureus (32,33), and this may have 192 enhanced E. faecalis' ability to colonise and compete within infected hosts.

A B
To test whether the lysozymes produced by lys-7 expression could affect host resistance to E. 207 faecalis and S. aureus, we measured the mortality of the wild-type N2 and lys-7 knockout 208 mutant when exposed to each bacterium. We found that lys-7 played a significant role in host 209 resistance to all bacterial exposure treatments (including to food), however we saw the largest 210 change in host mortality in response to S. aureus infection to almost 100% of the nematode 211 population after 24h ( Percentage mortality of the wild-type N2 and lys-7 knock-out mutant after exposure to pathogenic S. aureus, protective E. faecalis, and OP50 food. Each treatment was replicated five independent times. *P<0.05, **P<0.01, ***P<0.001.

Discussion 229
Microbes can protect hosts from pathogens via altering the host's immune response(4). They 230 can also suppress pathogens in more direct ways that diminish the need for host-based 231 immunity(1, 2). During the formation of a new symbiosis, these contrasting impacts of Although other mechanisms of protection may still play a role in our system at the origin of 242 the interaction, the transcriptomic evidence indicates that the immune system is a major factor. 243 Here, lys-7 was the most significantly upregulated lysozyme gene in protected hosts under 244 attack by pathogens. Lysozymes are well conserved across the animal kingdom(30) and aid in 245 resistance to S. aureus within C. elegans (31,39). It was previously demonstrated in C. elegans 246 that S. aureus drives the downregulation of lys-7 expression and when expression is restored 247 through chemical treatment, pathogen-induced host mortality significantly decreased along 248 with infection load(39). Given E. faecalis is relatively more resistant to lysozyme activity(32, 249 33), we hypothesized that lys-7 upregulation contributed to this bacterium's protective effect. 250 In support of this, we found that expression of this gene negatively affected S. aureus more 251 than E. faecalis, providing a competitive edge to the protective microbe. Increased E. faecalis 252 colonisation has previously been shown to relate to enhanced protection in our system (27,40). 253 These data are striking because the protective microbe was able to exploit host immunity in a 255 novel species. This result may have emerged because E. faecalis is widely distributed among 256 animal microbiotaas a protector (25), commensal and pathogen(41)and lysozymes are 257 conserved aspects of the animal immune response. Bacterial adaptations to these immune 258 components in one host species might therefore be relevant in another. Indeed, the transfer of 259 protective microbes to new hosts is more successful between host species that are more 260 phylogenetically similar, i.e. their immune system is likely to be more similar (42). 261

262
Our results suggest that immune-mediated protection can occur in novel interactions, and could 263 therefore be common at the beginning of a protective symbiosis. This finding has important 264 implications for predictions on the evolution of host dependence and can be used to inform 265 biocontrol design. Where protection involves the host immune system, it is predicted that hosts 266 will continue to invest in their immunity against pathogens over evolutionary time(23). For 267 biocontrol efforts, this is important to consider if the protective microbe should ever be lost 268 from the host population, such as via imperfect transmission(43), altered environmental 269 conditions (e.g. temperature change(43) or exposure to antibiotics (44)), or even the breakdown 270 of protection (e.g. if the symbiont evolves to become pathogenic(2, 22)). 271

Nematode host and bacteria: 274
Caenorhabditis elegans is a nematode that ingests microbes for nutrients (45-48) and is a well-275 established model for studying microbial colonisation and pathogenesis (37,(49)(50)(51)(52). We used 276 the simultaneous hermaphroditic N2 wild-type C. elegans strain from the Caenorhabditis 277 Genetics Centre (CGC, University of Minnesota) along with a knock-out mutant for the lys-7 278 gene that is outcrossed into the N2 genetic background (strain CB6738, CGC). We generated To examine host gene expression, we first exposed womrs to the protective microbe and 299 pathogen or the pathogen alone, followed by RNA extraction. Sterile and age-synchronised 300 eggs were collected using the bleach-sodium hydroxide solution. These eggs were kept in M9 301 buffer without food, shaking for ~8 hours at 88rpm and 20C to arrest development at L1. 302 Approximately 5,000 worms per replicate population were then transferred to 9cm NGM plates 303 seeded with E. coli OP50 and placed at 20C for two days. We chose this number to avoid 304 overcrowding and starvation. We then grew both bacteria species from frozen culture overnight 305 in 6ml THB in a shaking incubator at 30C at 200rpm and diluted the optical density of S. 306 aureus culture to be the same as the E. faecalis culture. We spread 120l culture per species 307 onto 9cm Tryptic Soy Broth (TSB) agar plates and incubated them overnight at 30C. Where 308 worms were to be exposed to both E. faecalis and S. aureus, we mixed 120l of each culture 309 together on the same TSB plate. We then removed worms from NGM plates, washed them in 310 50ml M9 buffer five times and placed approximately 2,000 young adults on the exposure plates 311 at 25C for 12 hours. 312

313
After 12 hours of exposure, we washed worms off each plate using M9 buffer within 10 minutes 314 in an order determined by a random number generator. We chose 12 hours of exposure to avoid 315 host mortality but provide sufficient time for C. elegans to respond to bacterial exposure and 316 infection. We washed the worms in 10ml M9 buffer five times and put approximately 1,000 317 worms in 50l into eppendorf tubes containing 1ml Trizol and vortexed for 20s. We then 318 freeze-thawed the samples of worms three times using dry ice and a heat block (40C) to break 319 the worm cuticle and stored at -80C. We extracted RNA using Zymo spin columns, following 320 the manufacturer's instructions with on-column DNA digestion using DNase I. 321

RNA sequencing 323
We quantified the resulting RNA using the Qubit® Fluorometer (Invitrogen) and all samples 324 were diluted to the same final concentration. The Oxford Genomics Centre then performed 325 library preparation and sequencing. The polyA signal was used to select the mRNA fraction 326 from the RNA and this was converted to cDNA. Second strand cDNA synthesis incorporated 327 dUTP and the cDNA was then end-repaired, A-tailed and adapter-ligated. Prior to 328 amplification, samples underwent uridine digestion. The prepared libraries were size selected, 329 multiplexed and checked for quality before paired-end sequencing using NovaSeq6000 with 330 150bp paired end reads. 331

332
We checked raw reads for quality using FastQC (0.11.5). Current (release 96) GTF and cDNA 333 FASTA files were downloaded from the ensemble database for C. elegans (WBcel235 version 334 of the C. elegans reference genome). We created a transcript index using Kallisto and the C. 335 elegans WBcel235 cDNA FASTA file. We then performed pseudoalignment using Kallisto 336 with 100 bootstraps and calculated transcript abundances. We then used the R package, Sleuth 337 to perform statistical analyses (see 'statistical analysis'). 338 339

Host mortality 340
We tested whether the lys-7 gene played a role in host mortality against S. aureus and E. 341 faecalis independently and during co-colonisation. We grew age-synchronised eggs to young 342 adult stage on 9cm NGM plates seeded with a lawn of E. coli OP50. Using the same protocols 343 as describe above, we grew S. aureus and E. faecalis in vitro overnight, standardised the 344 cultures to the same density, and made the exposure plates. We washed approximately 250 345 young adults of either CB6738 (lys-7 knock-out) or N2 wild-type in 50ml M9 five times before 346 exposing them at 25C for 24h. The proportion of dead worms were then counted as a measure 347 of host mortality. 348 349 Bacterial colonisation 350 We exposed CB6738 (lys-7 knock-out) and N2 wild-type worms to both S. aureus and E. 351 faecalis together following the same protocol as above (see 'Host mortality'). After 24 hours 352 of exposure to the bacteria, we collected 7-10 live worms per exposure plate and washed them 353 in 5ml of M9 five times under the microscope. To release the colonising bacteria, we placed 354 the worms in 2ml screwcap tubes with 50l M9 and 1.5mm Zirconium beads (Benchmark 355 Scientific) and broke the cuticle by shaking the tubes at 320rpm for 45s. We plated serial 356 dilutions onto Mannitol Salt Agar to isolate S.aureus and TSB with 100 μg/ml rifampicin 357 (Sigma-Aldrich) to isolate E. faecalis and incubated the plates at 30C overnight before 358 counting the resulting colony-forming units (CFUs) per host. 359 360 Statistical analysis: 361 We performed differential expression analysis on the transcript abundance outputs from 362 Kallisto using Sleuth in R v 3.2.0 (http://www.r-project.org/), using a likelihood ratio test of 363 fitted models. The significance of treatment was determined by a q-value of <0.05 (p-value 364 adjusted by means of the Benjamini-Hochberg false discovery date, FDR, correction for 365 multiple comparisons). We performed a GO-term enrichment analysis on the significant 366 differentially expressed genes using the g:Profiler online tool with the Benjamini-Hochberg 367 FDR correction for multiple comparisons(53). 368

369
We used parametric tests for all data which met the required assumptions. We used the Shapiro 370 test to detect whether data was normally distributed and F-tests to compare the variances of 371 two samples from normal populations. We compared bacterial CFUs per host using two-sample 372 t-tests. We used a binomial GLM to compare host mortality among host strains colonised by 373 E. faecalis and OP50. To account for overdispersion, we used quasibinomial GLMs to compare 374 host mortality among co-colonised host strains and host strains colonised by S. aureus. We