High-throughput phenotyping of infection by diverse microsporidia species reveals a wild C. elegans strain with opposing resistance and susceptibility traits

Animals are under constant selective pressure from a myriad of diverse pathogens. Microsporidia are ubiquitous animal parasites, but the influence they exert on shaping animal genomes is mostly unknown. Using multiplexed competition assays, we measured the impact of four different species of microsporidia on 22 wild isolates of Caenorhabditis elegans. This resulted in the identification and confirmation of 13 strains with significantly altered population fitness profiles under infection conditions. One of these identified strains, JU1400, is sensitive to an epidermal-infecting species by lacking tolerance to infection. JU1400 is also resistant to an intestinal-infecting species and can specifically recognize and destroy this pathogen. Genetic mapping of JU1400 demonstrates that these two opposing phenotypes are caused by separate loci. Transcriptional analysis reveals the JU1400 sensitivity to epidermal microsporidia infection results in a response pattern that shares similarity to toxin-induced responses. In contrast, we do not observe JU1400 intestinal resistance being regulated at the transcriptional level. The transcriptional response to these four microsporidia species is conserved, with C. elegans strain-specific differences in potential immune genes. Together, our results show that phenotypic differences to microsporidia infection amongst C. elegans are common and that animals can evolve species-specific genetic interactions.

that these two opposing phenotypes are caused by separate loci. Transcriptional analysis reveals 23 the JU1400 sensitivity to epidermal microsporidia infection results in a response pattern that 24 shares similarity to toxin-induced responses. In contrast, we do not observe JU1400 intestinal 25 resistance being regulated at the transcriptional level. The transcriptional response to these four 26 microsporidia species is conserved, with C. elegans strain-specific differences in potential 27 immune genes. Together, our results show that phenotypic differences to microsporidia infection 28 amongst C. elegans are common and that animals can evolve species-specific genetic 29 interactions. 30 31

Introduction 32
Obligate intracellular parasites only propagate inside of host cells, which places extreme pressure 33 on the host, resulting in genomic changes [1][2][3]. Animals have coevolved with pathogens and 34 genetic variation to pathogen susceptibility is common in animal populations [4,5]. Much of this 35 genetic and phenotypic diversity to infection is specific to individual species of pathogens or even 36 to different strains [6,7]. Hosts can evolve several ways to improve their fitness in the presence 37 of pathogens. They can become more resistant by preventing pathogen invasion or by limiting 38 pathogen growth [8]. Alternatively, hosts can evolve to better tolerate the infection, by limiting the 39 negative effects of infection without reducing the amount of pathogen [9,10]. Caenorhabditis elegans genotypes and resistant strains can outcompete sensitive ones in just a 46

A PhenoMIP screen of wild isolate strains reveals diverse responses to microsporidia 95 infection 96
To identify C. elegans strains that vary in their resistance or susceptibility to microsporidia 97 infection, we carried out a modified PhenoMIP screen. We chose 22 C. elegans wild isolate strains 98 that are genetically diverse and isolated from a range of geographic regions (Supplementary 99 Figure S1). We selected four pathogenic microsporidia species that differ in their impact on C. 100 elegans development and the tissues they infect (N. parisii, N. ausubeli, and N. ironsii infect the 101 intestine and N. ferruginous infects the muscle and epidermis) ( Figure 1A and Supplementary 102 Table S1) [15,16,19,43,51]. To synchronize infections, we treated adult populations with a sodium 103 hydroxide and sodium hypochlorite mixture (known as bleach synchronizing) and allowed the 104 recovered embryos to hatch in the absence of food. The process of bleach synchronizing animals 105 for microsporidia infection has been shown to have similar levels of infection between bleached 106 versus laid embryos, and short-term starvation is reported to have minimal impact on wild isolates 107 [19,20,40]. We pooled equal numbers of synchronized first larval stage (L1) wild isolates and a 108 laboratory reference-derived strain (VC20019) into a single population. VC20019, was chosen as 109 a reference because it has many genomic variants that can be used to distinguish it from N2 while 110 having developmental timing and fecundity similar to that of N2 [39,52]. We then plated these 111 absence of infection for 72 hours at 21°C to allow intergenerational immunity to reset [49]. Except 120 for the N. ausubeli medium-dose infection replicates (see Methods), worms in each condition 121 were infected for an additional three cycles for a total of eight generations. Adult animals were 122 collected every two generations at the end of the uninfected phase and genomic DNA was isolated 123 for further analysis (Figure 1B). 124 To determine how population composition changes due to exposure to microsporidia, we 125 calculated the relative abundance of each strain within each condition. We used MIPs to target 126 strain-specific variants for sequencing at high depth. In total we sequenced 127 samples, which 127 is the equivalent of 2162 infections and 736 mock/control infections. From each of the eight mock-128 infection controls and twenty-four infection replicates for every strain we calculated a mean fold-129 change rate (FCR) from across all timepoints. The distribution of FCRs for each strain was then 130 converted to a modified z-score distribution to identify candidate interactions (See Methods). 131 From this dataset, we identified 16 nematode/microsporidia pairings where the modified z-score 132 for both high-infection replicates was consistently outside the range of +/-1.5 (Figure 2, 133 Supplementary Figure S2 and Table 1). We also generated mean growth profiles across all 134 timepoints for each worm strain in environments that were either pathogen-free or infected by 135 each of the microsporidia species (Figure 3A and B). From our analyses we observed ten 136 interaction pairs with reduced population fitness (low-fitness strains), suggesting potential 137 susceptibility to infection. In addition, we observed six interactions with increased population 138 fitness (high-fitness strains), suggesting a potential resistance to infection conditions. Our high-139 fitness interaction set included the previously reported resistance interaction of CB4856 to N. 140 ironsii (Supplementary Figure S3) [19,51] which further suggested our scoring threshold was 141 appropriate for the detection of interactions in this assay format. Of the 15 novel candidate 142 interactions, we identified five interactions with N. ironsii (ERTm5), eight with N. ferruginous 143 (LUAm1), two with N. ausubeli (ERTm2), and no strong interactions with N. parisii (ERTm1). In 144 total, our screen yielded 13 C. elegans wild isolate strains with significant interactions, including 145 the strains JU360, JU1400, and MY2 that had interactions with more than one species of 146 microsporidia (Figure 2 and Table 1). 147

JU1400 fails to tolerate N. ferruginous infection 148
To validate a subset of our large-scale multiplex identification of microsporidia-strain phenotypes,  Figure S7). In total, of the 12 pairs that we observed 160 (Table 1), nine had altered relative fecundity profiles that could account for the fitness changes 161 observed from our PhenoMIP screen. Additionally, of the seven pairs we retested that had a z-162 score below -1.5, six had a significant reduction in embryos when infected, compared to N2. 163 To examine how a C. elegans strain responds to different microsporidia species, we focused on 164 JU1400, in which we observed reduced fitness to infection by both N. ferruginous and N. ironsii. 165 To determine why JU1400 appeared sensitive to N. ferruginous, we performed pulse-chase 166 experiments by infecting animals for 3 hours, washing away free spores, and replating animals 167 for an additional 69 hours [19]. We observed that the percentage of JU1400 and N2 animals 168 infected with N. ferruginous was similar ( Figure 3E and Supplementary Figure S8A and B). 169 Additionally, the amount of meronts formed during infection did not significantly differ between N2 170 and JU1400 (Figure 3F and Supplementary Figure S8C and 8D). Taken together, these results 171 demonstrate that although JU1400 is more sensitive to the effects of N. ferruginous, JU1400 is 172 not significantly more infected, suggesting a defect in tolerance. 173 To determine the specificity of JU1400 to N. ferruginous infection, we infected JU1400 with other 174 related microsporidia species and strains. We infected JU1400 with a second isolate of N. 175 ferruginous, LUAm3, as well as a separate muscle and epidermal-infecting microsporidian, N. 176 cider [44]. Infection by LUAm3 caused the same severe phenotype as LUAm1 (Supplementary 177 Figure S9A and B). JU1400 is infected by N. cider to a similar extent as N2, but infection by this 178 species resulted in a dose-dependent reduction in embryo numbers in JU1400 that was not 179 observed in N2 (Supplementary Figure S9C and D). 180

JU1400 specifically clears N. ironsii infection 181
In addition to the verified susceptibility to N. ferruginous, our PhenoMIP screen also identified a 182 more moderate fitness disadvantage for JU1400 in high-dose exposures to N. ironsii ( Figure 3A). 183 We infected N2 and JU1400 with N. ironsii for 72 hours and compared the relative change in 184 embryos for each strain versus mock-infected controls. We observed no significant difference in 185 this normalized number of embryos between the N2 and JU1400 strains ( Figure 4A). However, 186 JU1400 is significantly less infected, suggesting that JU1400 is resistant to N. ironsii infection 187 ( Figure 4C and Supplementary Figure S10A and C). To determine whether the observed 188 resistance conferred a fitness advantage at a later time point, we exposed JU1400 animals to N. 189 ironsii for 96 hours and observed significantly more embryos and significantly less infection suggest that although JU1400 appears to have a short-term disadvantage in our pooled 192 competition assays, this strain has a long-term advantage due to enhanced resistance towards 193 Several mechanisms of C. elegans resistance to microsporidia infection have been described, 195 including a block of invasion or clearance of infected parasite [19,20,49,51]. To determine how 196 JU1400 is resistant to N. ironsii infection, we performed a pulse-chase experiment by infecting 197 animals for 3 hours, washing away any external microsporidia spores, and examining animals at 198 3, 24, and 72 hours post infection (hpi). Both N2 and JU1400 had similar numbers of sporoplasms 199 at 3 hpi, but JU1400 had decreased levels of pathogen at both 24 and 72 hpi ( Figure 5A). This 200 experiment revealed that JU1400 does not block microsporidia invasion, but instead has the 201 capability to actively eliminate N. ironsii infection. To determine if the resistance in JU1400 was 202 dependent upon development stage, we performed pulse-chase infections at the L4 stage and 203 observed that JU1400 animals could no longer clear the infection (Figure 5B). 204 Resistance to N. ironsii has previously been observed in the C. elegans wild isolate, CB4856 [19]. 205 This resistance was shown to occur through a modest decrease in invasion coupled with the 206 ability to clear the infection at the L1 stage. To compare the properties of these two resistant 207 strains, we infected both strains with N. ironsii for 72 hours. This experiment shows that JU1400 208 is more resistant to infection ( Figure 5C). To determine the specificity of JU1400 resistance, we 209 tested infection with either N. parisii or N. ausubeli for 72 hours. CB4856 was slightly, but not 210 significantly less infected by N. parisii and N. ausubeli, whereas JU1400 was infected by these 211 species to the same extent as N2 ( Figure 5D). We also tested the ability of JU1400 and CB4856 212 to clear microsporidia infection, and both strains could only eliminate N. ironsii (Supplementary 213 Figure S10E and F). Together, these results demonstrate that JU1400 resistance is specific to 214 The immunity of JU1400 against N. ironsii could be a specific recognition of N. ironsii infection 216 that induces broad immunity towards microsporidia or the immunity itself could be specific for N. 217 ironsii. To distinguish between these possibilities, we set up coinfection experiments using 218 species-specific FISH probes and then coinfected JU1400 and N2 animals with N. ausubeli and 219 S13A) [53]. To determine if the same locus was responsible for N. ferruginous in MY1 and 245 JU1400, we tested their ability to complement. First, we determined that both MY1 and JU1400 246 display a recessive phenotype when crossed to our control strain. We then mated MY1 and 247 JU1400 and the resulting cross progeny were susceptible to N. ferruginous infection, indicating 248 that the two strains do not complement (Supplementary Figure S14A and B). We then infected 249 MY1 with N. ironsii. There were fewer MY1 animals containing spores compared to N2, and less 250 JU1400 animals had spores or meronts compared to MY1 animals (Supplementary Figure S14C  251 and D). Together this data suggests a region on the left side of chromosome I, is responsible for 252 sensitivity to N. ferruginous in multiple strains. 253 Our mapping of JU1400 with N. ironsii at 96 hours identified three regions that are linked with 254 resistance to infection, one on the left-hand side of chromosome II from 1,639,865-4,426,687, a 255 second in a broad area of chromosome V from 11,035,658-18,400,066, and a third on the right-256 hand side of chromosome X from 15,180,772-17,676,467 ( Figure 6B). A similar pattern was 257 observed when these mapping experiments were performed at 72 hours, with the exception that 258 the region on chromosome X was no longer observed (Supplementary Figure S13B). We also 259 observed a region on the left side of chromosome I where JU1400 variants underwent weak 260 negative selection upon over multiple N. ironsii infection rounds. Together, our data suggests that 261 JU1400 resistance to N. ironsii is multigenic and may be influenced by infection time. 262 To confirm that the region on chromosome I was responsible for sensitivity to N. ferruginous 263 infection, we generated Near Isogenic Lines (NILs). These NILs were constructed by crossing a 264 version of VC20019 that contained a CRISPR-integrated single-copy GFP on chromosome I 265 (AWR133) (Supplementary Figure S15A). We tested ten independent NILs that contained at 266 least some JU1400 sequence to the left of I:2,851,000. All ten of these NILs were sensitive to 267 infection by N. ferruginous (Supplementary Figure S15B). We used molecular markers to 268 genotype these NILs, mapping the susceptibility locus to I:1-1,051,810 (Supplementary Table  269 S4). We then outcrossed one of these NILs, AWR140 back to N2 six times. During this outcrossing 270 process, we recovered two lines, AWR144 and AWR145, that contained different responses to 271 N. ferruginous infection ( Figure 6D). We performed whole-genome sequencing of AWR138 272 (another independently isolated NIL), AWR140, AWR144, and AWR145 which narrowed the 273 critical region of association for our phenotype to I:1-701,078 ( Figure 6C). Within this region, we 274 identified 12 genes shared between JU1400 and MY1 (same complementation group) but absent 275 in ED3042 and JU360 which tested into separate complementation groups (Supplementary 276 Table S5). The NIL AWR145 is sensitive to N. ferruginous infection but is only partially resistant 277 to infection by N. ironsii (Figure 6D and E). Together, our data demonstrate that the N. 278 ferruginous sensitivity and N. ironsii resistance to JU1400 are determined by distinct genetic loci. 279

Transcriptional analysis of the JU1400 response to microsporidia infection 280
To determine how the JU1400 strain differs in its response to microsporidia infection, we 281 performed transcriptional analysis. We grew three biological replicates of N2 and JU1400 L1 282  We then sought to determine how the two C. elegans strains responded to the four different 291 microsporidia species (Supplementary Table S7). First, we used principal component analysis 292 (PCA), and projected samples back onto the two principal components which accounted for more 293 than 80% of the variance in our samples. PCA analysis revealed a clear clustering between the 294 two strains ( Figure 7A) and a secondary analysis by strain clustered the uninfected samples 295 against pathogen infected samples (Supplementary Figure S17). Second, we clustered the 296 differential expression profiles of the eight infected samples. This approach revealed several large 297 groups of genes that were differentially expressed either across most samples or within the same 298 strain ( Figure 7B and Supplementary Table S8). Although the largest difference in the samples 299 is between the two strains, we observe significant correlation between strains infected with each 300 microsporidia species (Supplementary Figure S18). 301 We then determined the properties of genes that are differentially regulated in each sample. The 302 number of differentially upregulated or downregulated genes ranged from 6-819, with the most 303 being in JU1400 infected with N. ferruginous ( Figure 7C). We then classified differentially 304 expressed genes into either conserved, microsporidia species (pathogen) specific, C. elegans 305 strain specific, or strain-pathogen specific categories ( Figure 7D). We identified 97 genes 306 upregulated in all samples, and no genes that were broadly downregulated. We identified ~30-50 307 differentially expressed genes specific to either N. ausubeli or N. ferruginous infection, one gene 308 specific to N. parisii infection, and no genes specific to N. ironsii infection. We also observed 45 309 or more differentially upregulated genes that were specific to either N2 or JU1400. To determine 310 the types of differentially expressed genes in these samples, we annotated them using a set of 311 domains previously observed to be regulated during microsporidia infection [47,48,54] 312 Table S9). About 35% of the conserved response were genes containing these 313 domains, with the most frequent being PALS, F-box, and MATH/BATH domains. About 25% of 314 the strain-specific genes also contained these domains, with each strain having at least one 315 PALS, F-box, and MATH/BATH, c-type lectin, and skr domain-containing gene expressed that is 316 specific to that strain ( Figure 7D and Supplementary Table S10). Skp-related proteins (skr) in 317 C. elegans were shown to bind to CUL-6 to ubiquitylate specific proteins [55]. skr-4 is upregulated 318 in both host strains infected by all microsporidia species. Among the upregulated genes, we also 319 observe strain-specific skr genes such as JU1400 specific skr-6, and N2 specific skr-5 320 Table S8). 321

(Supplementary
To understand how transcriptional responses correlated with resistant and susceptible 322 phenotypes, we analyzed genes that were specifically regulated in JU1400 when infected with 323 either N. ironsii or N. ferruginous ( Figure 7D). Surprisingly, we only observed three differentially 324 expressed genes specific to N. ironsii infection of JU1400. In contrast, we saw 181 upregulated 325 and 266 down-regulated genes specific to N. ferruginous infection of JU1400. Many of the 326 upregulated genes contained domains enriched in the IPR. The down-regulated genes contained 327 a few genes with IPR domains, as well as 23 genes with cuticle collagen domains. These proteins 328 with collagen domains were especially striking, as we never identified more than one of these 329 genes having altered expression in any of the other gene sets. 330 To determine if other stresses had similarities to the response of JU1400 to N. ferruginous, we 331 compared the set of genes specifically differentially expressed in this sample to previously 332 published C. elegans transcriptional datasets [56]. We identified the top ten chemical or microbial 333 treatment data sets that had the most significant overlap with JU1400 infected by N. ferruginous 334 Table S11). We then determined the overlap of these data sets to all our 335 samples ( Figure 7E-F and Supplementary Table S12). This revealed that genes differentially 336 downregulated in JU1400 in response to N. ferruginous infection overlapped significantly with 337 treatments of chemical poisons such as pesticides (fenamiphos and dichlorvos), organic 338 compounds (acrylamide and tert-butyl hydrogen peroxide), and heavy metals (mercury and 339 cadmium). All of these treatments were previously shown to inhibit development or the production 340 of progeny in C. elegans [57-60]. We also observed significant overlap of the response to these 341 chemicals with JU1400 infected with N. ironsii, and both strains infected with N. ausubeli. These 342 three infected samples along with JU1400 infected with N. ferruginous also contained the largest 343 number of significantly down-regulated genes ( Figure 7C and Supplementary Figure S19). 344

(Supplementary
Additionally, we observed significant overlap of these four microsporidia infected samples to 345 pathogenic bacteria or toxins including Staphylococcus aureus, Serratia marcescens, Bacillus 346 thuringiensis, Xenorhabdus nematophila, and Pseudomonas aeruginosa (Figure 7E-F). These 347 patterns of transcriptional overlap were the most striking with differentially downregulated genes, 348 although some overlap with upregulated genes was also observed (Supplementary Figure S20). 349 We also performed enrichment analysis of our transcriptional data sets, observing that the 350 upregulated genes are commonly enriched for metabolic processes and signalling pathways. In 351 contrast, down regulated genes are enriched for proteolysis, fatty-acid related metabolic 352 processes and drug metabolism (Supplementary Tables S13 and S14). We also analyzed the infection when compared to the population as a whole. We detected these interactions among 363 three of the four microsporidia species we measured, with N. ironsii and N. ferruginous accounting 364 for the bulk of the fitness interactions observed. We attributed these fitness changes as signatures 365 of susceptibility (fitness decreases) or resistance (fitness increases) with respect to microsporidia 366

infection. 367
So far, the causative gene for any host variants that influence microsporidia interactions has not 368 been identified. Previous genetic mapping in C. elegans, D. magna, and honey bees have 369 revealed 3-6 loci involved in resistance and sensitivity to microsporidia [19,23-26]. Additionally, 370 some loci in D. magna appear to be specific to different microsporidia species, which we show to 371 be the case for JU1400 [24]. Our genetic analysis of JU1400 identified that susceptibility to N. 372 ferruginous in both JU1400 and MY1 mapped to a single region on the left arm of chromosome I 373 while the resistance phenotype of JU1400 to N. ironsii appears to be caused by at least three 374 genomic regions. Taken together, our results add to the increasing evidence that suggests that 375 microsporidia have had a major impact on the evolution of animal genomes. 376 Microsporidia infect a wide diversity of animal phyla and infections are highly prevalent [12,13,61]. Pathogen tolerance is being increasingly appreciated as a valuable way for hosts to improve 387 fitness in the presence of pathogens, although these mechanisms are less understood [63]. 388 Tolerance is generally defined as the ability of a host to limit the consequences of a given level of 389 Here we show that JU1400 is very sensitive to infection by two different epidermal 390 infecting microsporidian species, but this strain does not display significantly higher levels of N. 391 ferruginous infection than non-sensitive control strains. RNA-seq analysis reveals that JU1400 392 has a unique response to microsporidia when infected with N. ferruginous. This response shares 393 similarities to several chemical toxins and pathogens that have been shown to damage C. 394 elegans. Although the mechanistic basis of this transcriptional similarity is likely different between 395 the various conditions, this result suggests that JU1400 infected with N. ferruginous has a 396 response similar to worms being exposed to toxic conditions. This signature is also detected to a 397 lesser extent in other strain-microsporidia pairings that cause developmental delay, including N. Our initial observation from our PhenoMIP experiment is that JU1400 has severely reduced 408 fitness when infected with N. ferruginous and a moderate reduction of fitness when infected with 409 N. ironsii ( Figure 3A). In our follow up experiments using embryo counts as a proxy for fitness, 410 we observed that at 72 hours there was no relative difference in reduction of embryos during 411 infection with N. ironsii ( Figure 4A) and significantly less reduction of embryos in JU1400 at 96 412 hours ( Figure 4B). However, we also observed a decrease in embryos at 72 hours when JU1400 413 was infected with N. ironsii compared to N2 or VC20019 (Supplementary Figure S4B). We also 414 performed MIP-MAP experiments using similar conditions to those in our initial PhenoMIP 415 experiment. These experiments showed that JU1400 regions were selected for at both 72 hours 416 Figure S13B) and at 96 hours ( Figure 6B). Regions on both chromosome II 417 and chromosome V were selected for at both times, but only the region on the X chromosome 418 was selected for at 96 hours. In addition to our experiments on host fitness, we also provide strong 419 evidence that there is resistance to N. ironsii in JU1400 due to its ability to clear the parasite. 420

(Supplementary
Thus, our evidence is supportive of a model where JU1400 is resistant to N. ironsii infection and 421 has increased fitness compared to control strains at a 96-hour time point. Our data is inconclusive 422 about the impact of N. ironsii on JU1400 at 72 hours as we have conflicting results depending on 423 the experiment. Our results suggest that a limitation of our PhenoMIP experiment is that 424 resistance to microsporidia infection doesn't always result in a fitness advantage under certain 425 conditions. Our data also suggests that future competitive selection experiments using 426 microsporidia could be improved using 96-hour selections to identify resistant strains. 427 JU1400 is resistant to N. ironsii infection by clearing the invaded parasite. This immune 428 mechanism is similar to previous results seen in N. ironsii infection of CB4856. However, the 429 immunity in JU1400 results in less infection than in CB4856. The JU1400 resistance mechanism 430 is specific, as the clearance we observe does not occur in even the closely related sister species 431 N. parisii, which shares 92% identical DNA [43]. A strain of mosquito that is resistant to infection 432 by a parasitic nematode was shown to rely on activation of an upregulated immune response [65]. 433 However, we see no evidence for an increased transcriptional response of JU1400 in response 434 to N. ironsii. A potential caveat to these experiments is that we infected JU1400 at the L1 stage, 435 but measured the transcriptional response at the L4 stage, a stage where if infections are initiated, 436 parasite clearance is not observed. We do, however, observe additional parasite clearance 437 between 24 and 72 hours (Figure 5A), which corresponds to between the L2/L3 stages and 438 adults, suggesting that if there was transcriptional signature of resistance, it would be observed 439 under our experimental RNA-seq conditions. Instead of a specific transcriptional response of 440 JU1400 to N. ironsii, we observe a C. elegans strain-specific response of immune genes including 441 upregulation of several CHIL genes, which have been observed to be upregulated in response to 442 epidermally infecting oomycetes, but have not previously been observed in response to 443 microsporidia infection [66,67]. This strain-specific-response of C. elegans has also been 444 observed with several bacterial pathogens, suggesting that there exists a diversity of innate 445 immune responses within the C. elegans population [68]. We also observe more strain-specific 446 regulated genes than pathogen-specific regulated genes, suggesting that C. elegans responds 447 similarly to infection by different species of Nematocida [69]. 448 Here we demonstrate that PhenoMIP can be used to identify variants of pathogen infection in a

Competing Interests 478
The authors declare that they have no competing interests.  Table S2). 495

PhenoMIP competitive assays 496
PhenoMIP competitive assays were based on a previously described protocol [39]. We modified 497 this method to use bleach-synchronization between generations to directly measure egg-498 production as an output of population fitness. At each generation, samples were bleach-499 synchronized as gravid adults ~72 hours after plating L1s and incubating at 21°C. Bleached 500 embryos were then allowed to hatch over 18-22 hours while incubating on a shaker at 21°C in 501 axenic M9 media. Each sample was gently pelleted and 3 -7 μl of each pellet was used to seed 502 the experimental plate for the next generation of animals. Overall, the competitive assay consisted To identify outliers and potential interaction candidates, we used a minimum z-score of ± 1.5 as 556 our threshold where both replicate samples were each required to meet the threshold. 557

Microsporidia infection assays 558
Continuous microsporidia infection assays were conducted using ~1,000 bleach-synchronized 559  EverBrite mounting medium with DAPI before mounting on slides for quantitation. 575

Embryo analysis and normalization 576
Unless otherwise noted, embryo analyses were completed using normalized values which were 577 calculated in the following manner. For each replicate experiment, the uninfected control for each 578 strain was used to establish a baseline mean embryo value ( � ). Embryo counts for individual 579 animals ( ) of the same strain within the same replicate experiment were normalized ( � ⁄ ) before 580 combining with the normalized values from matching replicate experiments. This normalization 581 step facilitated the analysis of relative changes to embryo production, rather than absolute 582 changes in embryo counts which could vary both between nematode strains and replicate 583 experiments. The combined normalized populations were compared using two-way ANOVA with 584 a Tukey HSD post hoc. 585

Microscopy and image quantification 586
All imaging was done using an Axio Imager.M2 (Zeiss) and captured via ZEN software under 587 identical exposure times per experiment. Animals containing clumps of spores visible in the body 588 after 72 hours by DY96 straining were considered as 'new spore formation'. Animals with one 589 sporoplasm or more in intestinal cells stained with FISH were considered infected. Gravidity of 590 worms was determined by counting the number of embryos per hermaphrodite animal. 591 Pathogen burden and body size was determined using ImageJ/FIJI software [76]. Individual 592 worms were outlined as "region of interest" with selection tools. Body size of worms were 593 determined using the "measure area" function. To quantify fluorescence (pathogen burden), 594 signal from FISH staining was subjected to the "threshold" function followed by "measure percent 595 area" tool. 596

Pulse-chase assays 597
Pulse-chase assays were performed on 6-cm NGM plates with roughly 6,000 bleach-598 synchronized L1s obtained from cultures rotating at 21°C for 18-22 hours. Worm populations were 599 first premixed in a 1.5 mL microcentrifuge tube with 5 µL of 10X OP50-1 and 1.25 million spores 600 per microsporidia strain. This dose was chosen to produce N2 controls with 85-90% infection 601 rates at 3 hpi. Samples were then spread onto individual 6-cm NGM plates and allowed to dry 602 before incubating at 21°C for 3 hours. To subsample the population at 3 hpi, plates were washed 603 at least two times with M9 in a 1.5 mL microcentrifuge tube to remove OP50-1 and residual 604 spores. Half the population was then fixed in acetone and stored at -20°C. The other half was 605 mixed with 50 µL of 10X OP50-1 and plated onto 6-cm NGM plates. Samples were then incubated 606 at 21°C for 21 hours (24 hpi), washed, subsampled, fixed, and replated (with 300 µL of 10X OP50-607 1) in the same manner as described above. The plates were then incubated at 21°C for 48hours 608 (72hpi) before washing and fixing in acetone. Pulse-chase co-infection assays were performed 609 as described above, except samples were infected with either one or two species of microsporidia 610 for 24 hours total. Pulse-chase assays initiated at the L4 stage were performed by using ~3000 611 bleach-synchronized L1s mixed with 200 µL of 10X OP50-1 and plated on unseeded 6-cm NGM 612 plates for 48 hours. Worm populations were then washed two times with M9 in 1.5 mL 613 microcentrifuge tubes. 50 µL of 10X OP50-1 and spores were added to each worm population 614 and then plated on unseeded 6-cm NGM plate for 3 hours at 21°C. Samples were then 615 subsampled, washed, and replated as described above. Chitin and FISH staining were performed 616 as described above for all pulse-infection experiments. 617

MIP-MAP and variant analysis 618
Wild isolates were mapped using a competitive fitness approach after crossing with males of the 619 mapping strain AWR133, a CRISPR-inserted Pmyo-2::GFP version of VC20019. Mapping was 620 completed using 2-4 competitive selection cycles on 10-cm NGM plates. Each cycle consisted of 621 2 generations of recombinant populations, each added to plates as bleach-synchronized L1 622 animals. In the first generation ~10,000 L1 cross-progeny animals were exposed to continuous 623 microsporidia infection (See Supplementary Table S3) for ~72 hours before bleach-624 synchronizing. Next, ~8,000 L1s were plated within 24 hours to "rest" on 10-cm OP50-seeded 625 NGM plates and grown for ~72 hours before being bleach-synchronized. L1 progeny of the "rest" 626 phase were then plated onto both an OP50-1-seeded plate with spores to begin the next cycle 627 and a secondary plate without spores to generate a sample to collect genomic DNA. Genomic (2,012,869). These markers were used for restriction fragment length polymorphism analysis with 636 the enzyme DraI. JU1400 hermaphrodites were crossed with AWR133 males. F1 GFP+ animals 637 at the L4 stage were chosen to produce F2 animals. 48 GFP+ animals were chosen and screened 638 at marker cewivar00066519 for JU1400 homozygosity and further screened at cewirvar00070489 639 for JU1400 homozygosity or heterozygosity. Seven F2 lines were chosen based on their 640 genotyping and four of these gave rise to five main progenitor NILs, each of which was used to 641 create a separate GFP+ and GFP-NIL strain. In total 10 NILs were generated and tested for 642 susceptibility against LUAm1. One line was chosen for outcrossing (AWR140) which gave rise to 643 the AWR144 and AWR145 NILs. 644

C. elegans Whole Genome Sequencing Analysis 645
Worm strains used for whole genome sequencing were grown on 10-cm plates with OP50-1 for to generate a transcriptome assembly for each sample, which was then combined with the 677 corresponding reference genome annotation using Cuffmerge of the Cufflinks 2.2.1 package to 678 produce a single annotation file. 679

C. elegans mRNA-seq analysis 680
The paired end reads generated for each strain were submitted as separate projects to Alaska

Strain-specific, pathogen-specific and strain-pathogen pair specific genes 703
Strain-specific genes were defined as genes that are significantly upregulated or downregulated 704 in at least three out of the four JU1400 or N2 samples. Pathogen-specific genes were defined as 705 those significantly upregulated or downregulated in both JU1400 and N2 upon infection by the 706 same microsporidia species, but not observed when infected by other microsporidia species. 707 Moreover, strain-pathogen pair specific genes were defined as genes that are significantly 708 upregulated or downregulated in a particular strain-pathogen sample, but not observed in the 709 other seven strain-pathogen samples. hmmsearch. Output genes with E-value <1e-5 were used as the gene list for the enrichment 723 analyses. In the output genes, MATH and BATH genes were combined into one list; fbxa, fbxb 724 and fbxc genes were merged into the gene list for F-box (Supplementary Table S10). Using the 725 R package gplots v3.1.1, the number of genes overlapping between the lists and samples were 726 computed, then plotted using R package ggplot2 v3.3.5. Genes that do not fall into any of these 727 gene classes or have any of these domains were categorised as "other". 728

Gene Set Enrichment Analysis 729
The WormExp [56] dataset, which contains published expression data for C. elegans, was 730 downloaded from the WormExp v1.0 website (https://wormexp.zoologie.uni-kiel.de/wormexp/). 731 The overlap between published datasets of C. elegans exposed to different chemicals or pathogens (Supplementary Table S12) and our samples, as well as the strain specific, 733 pathogen-specific, and strain-pathogen pair specific genes in our dataset were computed using 734 the R package gplots v3.1.1. For the statistical evaluation of the overlap, the p-values were 735 calculated using the adjusted Fisher exact test method from the program EASE, which removes 736 one gene in the set of intersected genes. Correction for multiple testing was implemented using 737 the Bonferroni method via the p.adjust() function in the R package stats v4.1.1. 738