Antagonistic paralogs control a switch between growth and pathogen resistance in C. elegans

Immune genes are under intense, pathogen-induced pressure, which causes these genes to diversify over evolutionary time and become species-specific. Through a forward genetic screen we recently described a C. elegans-specific gene called pals-22 to be a repressor of “Intracellular Pathogen Response” or IPR genes. Here we describe pals-25, which, like pals-22, is a species-specific gene of unknown biochemical function. We identified pals-25 in a screen for suppression of pals-22 mutant phenotypes and found that mutations in pals-25 suppress all known phenotypes caused by mutations in pals-22. These phenotypes include increased IPR gene expression, thermotolerance, and immunity against natural pathogens, including Nematocida parisii microsporidia and the Orsay virus. Mutations in pals-25 also reverse the reduced lifespan and slowed growth of pals-22 mutants. Transcriptome analysis indicates that pals-22 and pals-25 control expression of genes induced not only by natural pathogens of the intestine, but also by natural pathogens of the epidermis. Indeed, in an independent forward genetic screen we identified pals-22 as a repressor and pals-25 as an activator of epidermal defense gene expression. In summary, the species-specific pals-22 and pals-25 genes act as a switch to regulate a program of gene expression, growth, and defense against diverse natural pathogens in C. elegans.


Author summary
Infection by microbial pathogens imposes selective pressure on animal and plant hosts. For this reason, host immune genes tend to vary in DNA sequence over evolutionary time and become 'species-specific'. In this work we describe a pair of species-specific genes called pals-22/pals-25 that promote resistance against natural pathogens of the small roundworm Caenorhabditis elegans. This gene pair acts as an on/off switch for genes that are activated after infection and rewires the physiology of worms to either be in an immune state or a growth state. In particular, the pals- 22

Introduction
Evolutionarily ancient genes control core processes in diverse organisms. For example, the >500 million-year-old Hox gene cluster is required for establishing body plan polarity in animals as diverse as worms, flies and humans [1]. However, evolutionarily young genes can also play key roles in development. For example, the Drosophila Umbrea gene only evolved within the Drosophila lineage in the last 15 million years but is essential for chromosome segregation in Drosophila melanogaster [2]. In general, the functions of evolutionarily young genes are less well understood than the functions of evolutionarily ancient genes. New genes can arise through gene duplication and diversification [3]. Extensive gene duplication can lead to large, expanded gene families, which appear 'species-specific' if there is significant diversification away from the ancestral gene. The function of species-specific genes can provide insight into the pressures imposed upon organisms in their recent evolutionary past. Pathogen infection imposes some of the strongest selective pressure on organisms, and accordingly, many species-specific, expanded gene families are involved in immunity. One example is the family of mouse Naip genes, which encode sensor proteins in the inflammasome that detect bacteria to trigger cytokine release and cell death [4]. Another example is the plant R genes, which detect virulence factors from co-evolved pathogens to activate effectortriggered immunity [5]. Interestingly, a growing theme in plant R genes is that they can function as opposing gene pairs, with one R gene promoting host defense and the other R gene inhibiting host defense. Of note, both the Naip and R genes were identified through unbiased forward genetic screens for immune genes.
Recently, we described a forward genetic screen in C. elegans for genes that regulate the transcriptional response to natural intracellular pathogens [6]. From this screen we identified a C. elegans-specific gene called pals-22 that regulates expression of Intracellular Pathogen Response or IPR genes. Interestingly, we found that pals-22 also regulates proteostasis, potentially through ubiquitin ligase activity (see below). The 'pals' signature stands for protein containing ALS2CR12 signature, which is found in the single pals gene in humans called ALS2CR12. A genome-wide association study implicated ALS2CR12 in amyotrophic lateral sclerosis (ALS) [7], although this gene has no known role in ALS, and its biological function is unknown. The pals gene family has only a single member each in the mouse and human genomes but is substantially expanded in Caenorhabditis genomes: C. elegans has 39 pals genes; C. remanei has 18 pals genes; C. brenneri has 8 pals genes; and C. briggsae has 8 pals genes [8].
pals-22 mutants have several striking phenotypes in C. elegans. First, pals-22 mutants have upregulated expression of several IPR genes including the cullin gene cul-6, which is predicted to encode a component of a Cullin-Ring Ligase complex. Second, pals-22 mutants have increased tolerance of proteotoxic stressors, and this increased tolerance requires the wild-type function of cul-6. Third, pals-22 mutants have less robust health in the absence of stressors. In particular, they have slowed development and shorter lifespans compared to wild-type animals. Fourth, as shown by another group who identified pals-22 in an independent forward genetic screen, pals-22 mutants have increased transgene silencing and increased RNA interference (RNAi) against exogenous RNA [8]. Thus, loss-of-function mutations in pals- 22 appear to broadly reprogram the physiology of C. elegans.
Here we describe a forward genetic screen for suppressors of pals-22 and identify another pals gene called pals- 25. Interestingly, although it appears that pals-25 and pals-22 are in an operon together, these two genes function antagonistically and direct opposing phenotypes. We show that mutations in pals-25 strongly suppress all the physiological phenotypes seen in pals-22 mutants, including IPR gene expression, stress resistance, lifespan, development and transgene silencing. Furthermore, we find that pals-22 mutants have increased resistance against natural intracellular pathogens, like the microsporidian species Nematocida parisii and the Orsay virus. This increased resistance is suppressed by mutations in pals- 25. Also, we use RNA-seq analysis to show that the pals-22/pals-25 gene pair (hereafter referred to as pals-22/ 25) regulate expression of a majority of the genes induced by natural pathogens of the intestine and find that most of these genes are also induced by blockade of the proteasome. Interestingly, we observe that pals-22 and pals-25 also regulate expression of genes induced by natural eukaryotic pathogens infecting through the epidermis. Indeed, in an independent forward genetic screen to find regulators of epidermal defense gene expression we identified additional mutant alleles of pals-22 and pals-25. In summary, the species-specific pals-22/25 gene pair control an entire physiological program that balances growth with increased proteostasis capacity and resistance against diverse natural pathogens.

pals-25 is required for increased IPR gene expression in pals-22 mutants
Previously we found that wild-type pals-22 represses expression of IPR genes: pals-22 mutants have constitutive expression of several IPR genes including pals-5 [6]. A transcriptional reporter consisting of the 1273 bp upstream region of pals-5 fused to GFP, pals-5p::GFP, is a reliable marker of IPR gene expression [9] and exhibits strong GFP expression in a pals-22 mutant background [6] (Fig 1A-1C). To find positive regulators of IPR gene expression, we mutagenized pals-22; pals-5p::GFP strains and screened for loss of GFP expression in F2 animals. From one screen in the pals-22(jy1) mutant background and one screen in the pals-22 (jy3) mutant background, we screened a total of~23,000 haploid genomes and found eight independent mutant alleles that almost entirely reverse the increased pals-5p::GFP expression back to near wild-type levels in pals-22 mutants (Fig 1A-1F). All of these alleles are recessive, segregate in Mendelian ratios, and fail to complement each other, suggesting they are likely to represent loss-of-function mutations in the same gene.
We used whole-genome sequencing of two mutant strains (jy9 and jy100) to identify the causative alleles [10] and found predicted loss-of-function mutations in pals-25 in both strains. Further sequencing identified pals-25 mutations in the remaining six mutant strains (Fig 1G, S1 Table). pals-25 appears to be in an operon just downstream of pals-22 (CEOP3012, https:// www.wormbase.org), and while these two genes are paralogs, they share limited sequence similarity, with no significant similarity on the DNA level and only 19.4% identity on the amino acid level. Of note, neither pals-22 nor pals-25 have obvious orthologs in other Caenorhabditis species like Caenorhabditis briggsae and Caenorhabditis brenneri, and thus appear to be specific to C. elegans [8]. We used reciprocal BLAST analysis to search for orthologs in the genome of the newly described C. elegans sister species Caenorhabditis inopinata and were unsuccessful (see Materials and Methods). Interestingly, previous analysis of the pals gene family indicated that there was no conservation of the pals-22/25 region with 26 other nematode genomes, with perhaps very weak conservation to the right of pals-25 [8].  25(jy9) animals. Green is pals-5p::GFP, red is myo-2p::mCherry expression in the pharynx as a marker for presence of the transgene. Images are overlays of green, red and Nomarski channels and were taken with the same camera exposure for all. Scale bar, 100 μm. (F) pals-5p::GFP expression quantified in pals-22 suppressor mutants using a COPAS Biosort machine to measure the mean GFP signal normalized by length of individual animals, indicated by green dots. Mean signal of the population is indicated by black bars, with error bars as SD. Graph is To further confirm that pals-25 regulates pals-5p::GFP gene expression in pals-22 mutants, we performed RNAi against pals-25 in a pals-22; pals-5p::GFP strain. As expected, we found lowered expression of pals-5p::GFP (Fig 1H, S1A Fig), indicating that wild-type pals-25 is required for the increased expression of pals-5p::GFP seen in a pals-22 mutant background.
These observations suggest that pals-25 acts downstream of or in parallel to pals-22 to activate mRNA expression of IPR genes. We chose two pals-25 alleles to test this hypothesis in greater detail. Specifically, we used qRT-PCR to measure levels of endogenous mRNA in two different pals-22 pals-25 mutants compared to two different pals-22 mutants and wild-type animals (Fig 2A). We analyzed mRNA levels of pals-5, as well as seven other IPR genes including nematode-specific genes of unknown function (F26F2.1, F26F2.3, and F26F2.4) and predicted ubiquitin ligase components skr-3, skr-4, skr-5 and cul-6. Here we found that mutations in pals-25 reverse the elevated mRNA levels of all eight of these IPR genes in a pals-22 mutant background back to near wild-type levels. Importantly, a non-IPR gene, skr-1, is not affected by mutations in pals-22 or pals-25. These results indicate that in a pals-22 mutant background, wild-type pals-25 is required to activate IPR gene expression.
Previous analysis indicated that pals-22 is broadly expressed in several tissues in the animal, including the intestine and the epidermis [6,8]. Similarly, we found that pals-25 is broadly expressed. Using a transgenic strain with a fosmid containing pals-25 with endogenous cis regulatory control and tagged at the C terminus with GFP and 3xFLAG [11], we observed PALS-25::GFP expression throughout the animal, including expression in the neurons, epidermis, and intestine (S1B Fig). We did not see any change in PALS-25::GFP expression after pals-22 RNAi treatment (S1C Fig), suggesting that PALS-22 is unlikely to regulate mRNA or protein levels of PALS-25.

IPR genes are induced by infection and by proteasome blockade in pals-22 pals-25 mutants
As pals-25 is required to activate IPR gene expression in a pals-22 mutant background, we wondered whether pals-25 was also required for inducing IPR gene expression in response to external triggers. We originally identified IPR genes because of their induction by N. parisii infection [6,9], which is an intracellular pathogen in the Microsporidia phylum that invades and undergoes its entire replicative life cycle inside C. elegans intestinal cells [12]. We therefore infected animals with N. parisii and compared induction of IPR genes in pals-22 pals-25 mutants and wild-type animals at 4 hours ( Fig 2B). Here we found similar levels of IPR gene induction in pals-22 pals-25 and wild-type animals, suggesting that pals-22/25 regulate expression of IPR genes independently of infection. Next, we examined the role of pals-22/25 in the transcriptional response to proteasome blockade, which is another trigger of IPR gene expression [9] (Fig 2C). We used bortezomib, which is a small molecule inhibitor of the 26S proteasome. Here again, we found that bortezomib treatment induced IPR gene expression in pals-22 pals-25 mutants at levels similar to wild-type animals. Therefore, pals-22/25 appear to regulate IPR gene expression independently of infection and proteasomal stress.

pals-25 mutations reverse multiple physiological phenotypes caused by pals-22 mutations
pals-22 mutants have several striking physiological phenotypes, including slowed growth and shorter lifespans, as well as increased resistance to proteotoxic stress like heat shock [6]. a compilation of three independent replicates, with at least 100 animals analyzed in each replicate. ��� p < 0.001 with Student's t-test. (G) pals-22 and pals-25 gene coding structure (UTRs not shown), with blue exons for pals-22 and red exons for pals-25. See S1 Table "Mutations" sheet for residues altered in the mutated proteins. (H) pals-5p::GFP expression in animals treated with either L4440 RNAi control or pals-25 RNAi, quantified using a COPAS Biosort machine to measure the mean GFP signal normalized by length of animals. Parameters the same as in (F).
https://doi.org/10.1371/journal.ppat.1007528.g001 Therefore, we investigated whether mutations in pals-25 suppress these phenotypes of pals-22 mutants. First, we investigated developmental rate by measuring the fraction of animals that reach the fourth larval (L4) stage by 48 hours after embryogenesis. Nearly all wild-type animals are L4 at this timepoint, whereas less than 20% of pals-22 mutants are L4 (Fig 3A). We found that mutations in pals-25 completely reverse this delayed development of pals-22 mutants, with nearly all pals-22 pals-25 mutants reaching the L4 stage by 48 hours (Fig 3A). Next, we analyzed lifespan, as previous work showed that pals-22 mutants have a significantly shortened lifespan compared to wild-type animals [6,8]. Here again we found that pals-25 mutations reversed this effect, with pals-22 pals-25 mutants having lifespans comparable to wild-type animals ( Fig 3B, S2A and S2B Fig). Next, we investigated the effect of pals-25 mutations on the thermotolerance capacity of pals-22 mutants, which is greatly enhanced compared to wildtype animals. We found that pals-22 pals-25 double mutants have survival after heat shock at levels similar to wild-type animals (Fig 3C, S2C and S2D Fig), indicating that pals-25 is required for the enhanced thermotolerance of pals-22 mutants. Thus, these results show that in a pals-22 mutant background, wild-type pals-25 is required to delay development, shorten lifespan and enhance thermotolerance.
Previous work from the Hobert lab identified pals-22 in a screen for regulators of reporter gene expression in neurons [8]. They found that mutations in pals-22 led to decreased levels of GFP reporter expression in neurons and other tissues, and wild-type pals-22 thus acts as an 'anti-silencing' factor of multi-copy transgene expression. Therefore, we analyzed the effects of pals-25 mutations on transgene silencing in pals-22 mutants. Here we found that pals-25 mutations reverse the enhanced silencing of a neuronally expressed GFP transgene in pals-22 mutants (Fig 3D-3I), indicating that wild-type pals-25 activity is required to silence expression from multi-copy transgenes in a pals-22 mutant background. Of note, previous work found that a pals-25 mutation alone does not affect transgene silencing [8]. In summary, loss of function mutations in pals-25 appear to fully reverse all previously described phenotypes of pals-22 mutants.

pals-22 mutants have immunity against coevolved intestinal pathogens of the intestine, which is suppressed by pals-25 mutations
In addition to the previously described phenotypes of pals-22 mentioned above, we analyzed resistance of these mutants to intracellular infection. First, we analyzed the resistance of pals-22 mutants to N. parisii infection. We fed animals a defined dose of microsporidia spores and measured pathogen load inside intestinal cells. We analyzed pathogen load at 30 hours post infection (hpi), when N. parisii is growing intracellularly in the replicative meront stage, and found greatly lowered pathogen load in pals-22 mutants compared to wild-type animals ( Fig  4A-4F). We then tested pals-22 pals-25 double mutants and found these animals to have resistance comparable to wild type. One explanation for the altered levels of N. parisii observed in the intestines of pals-22 mutant animals is that these mutants have lowered feeding or accumulation of pathogen in the intestine, and thus simply have a lower exposure to N. parisii. To address this concern, we added fluorescent beads to our N. parisii infection assay and measured accumulation in the intestinal lumen. Here we found that pals-22 mutants and pals-22 pals-25 double mutants accumulated fluorescent beads at comparable levels to wild-type animals (S3A Fig), suggesting that their pathogen resistance to N. parisii is not simply due to lowered exposure to the pathogen in the intestinal lumen. As a positive control in this assay we tested eat-2(ad465) mutants and found that they had reduced fluorescent bead accumulation, consistent with their previously characterized feeding defect [16]. Altogether, these results indicate that pals-22 and pals-25 regulate resistance to infection by microsporidia.
We also investigated resistance of pals-22 mutants and pals-22 pals-25 double mutants to other pathogens. First, we measured resistance to infection by the Orsay virus. Like N. parisii, Orsay virus is a natural pathogen of C. elegans and replicates inside of C. elegans intestinal cells [17,18]. We used FISH staining of Orsay viral RNA to quantify the fraction of worms infected at 18 hpi. Here we found that pals-22 mutants had significantly decreased viral load when compared to wild-type animals ( Fig 4G-4L). This lowered viral infection in pals-22 mutants was fully reversed to wild-type levels in pals-22 pals-25 mutants. Importantly, we confirmed that pals-22 and pals-22 pals-25 mutants do not have altered fluorescent bead accumulation in the intestine compared to wild-type animals in the presence of virus (S3B Fig), indicating that their lowered viral load is not likely due to lowered exposure to the virus.
Interestingly, we found that pals-22 mutants did not have reduced pathogen load when infected with the Gram-negative bacterial pathogen Pseudomonas aeruginosa (clinical isolate PA14) ( Fig 4M). In fact, these mutants had increased pathogen load, which was suppressed by mutations in pals-25. To our knowledge, P. aeruginosa species are not common pathogens of nematodes in the wild, although under laboratory conditions, P. aeruginosa PA14 does accumulate in the C. elegans intestinal lumen and causes a lethal infection [19,20]. In summary pals-22 mutants have increased resistance to natural pathogens of the intestine but increased susceptibility to PA14, a 'non-natural' pathogen of the intestine.
The results above indicate that pals-25 acts downstream or in parallel to pals-22 to regulate IPR phenotypes (Figs 1-4). If pals-25 were acting downstream of pals-22, the prediction is that a pals-25 mutation in a wild-type background would have no effect. Because of the close genomic proximity of pals-22 and pals-25, it would be extremely difficult to recombine our existing pals-25 mutations away from pals-22 mutations into a wild-type background. Therefore, we used CRISPR/Cas9 editing to generate two independent pals-25 deletions in otherwise wildtype backgrounds (Fig 5A). In both cases we found that these pals-25 null mutants have wildtype levels of IPR gene expression, thermotolerance and N. parisii pathogen load (Fig 5B-5D). In addition, we used CRISPR/Cas9 editing to delete the entire pals-22 pals-25 locus and analyzed two independent alleles. In both cases we found that these pals-22/25 null mutants have wild-type levels of IPR gene expression, thermotolerance and N. parisii pathogen load (Fig 5B-5D), consistent with our previous analyses of pals-22 pals-25 mutants isolated from forward genetic screens. In addition, IPR gene expression was still induced by N. parisii or bortezomib treatment in these mutants (Fig 5E and 5F). These results are all consistent with pals-25 acting downstream of pals-22 and indicate that the entire pals-22/25 locus is dispensable for IPR induction, thermotolerance, and resistance to N. parisii.

RNA-seq analysis of pals-22/25-upregulated genes define the IPR
Previous work indicated that N. parisii and the Orsay virus induce a common set of genes, despite these being very different pathogens [9]. We called eight of these genes the IPR subset [6], and here we show they are regulated by pals-22/25 (Fig 2A). To identify additional genes regulated by pals-22/25, we performed RNA-seq analysis of pals-22 mutants, pals-22 pals-25 mutants, and wild-type animals. We performed differential gene expression analysis using a FDR<0.01 cutoff (see Materials and Methods for a complete description of criteria) and determined that 2,756 genes were upregulated in pals-22 mutants compared to wild-type animals ( Fig 6A, S4 Table). Next, we compared pals-22 mutants to pals-22 pals-25 double mutants and found that 744 genes were upregulated ( Fig  6A, S4 Table). Of these two comparisons, there are 702 genes in common that are upregulated both in pals-22 mutants compared to wild-type animals and in pals-22 mutants compared to pals-22 pals-25 double mutants ( Fig 6A). Therefore, these 702 genes are negatively regulated by wildtype pals-22 and require the activity of the wild-type pals-25 for their induction in the absence of wild-type pals-22. These 702 genes include pals-5 (Fig 1) and other IPR genes (Fig 2).
Because pals-22 is upstream of pals-25 in an operon, we investigated whether a pals-22 mutation might affect mRNA expression of pals-25. Consistent with the lack of PALS-25::GFP changes upon pals-22 RNAi treatment, we found no significant difference between pals-25 mRNA expression in pals-22(jy3) mutants compared to wild-type animals ( Fig 6B). To investigate whether a pals-25 mutation had an effect on pals-22 mRNA expression we performed qRT-PCR on a pals-25(jy81) deletion mutant and here we found there was no significant effect on pals-22 mRNA expression in these mutants compared to wild-type animals (S4 Fig).
We next compared the 702 pals-22/25-regulated genes to genes induced during N. parisii infection identified in a previous study [9] to expand our list of IPR genes. Out of 127 genes induced during N. parisii infection, we found that the pals-22/25 gene pair regulated mRNA expression of 80 of these genes ( Fig 6A). Specifically, of the 25 pals genes induced upon intracellular infection, all are induced in pals-22 mutants and reverted back to wild-type levels in pals-22 pals-25 double mutants (Fig 6B, S4 Table). Notably, all pals genes that are not regulated by pals-22/25 are also not induced by infection. Furthermore, the other nematode-specific genes F26F2.1, F26F2.3, and F26F2.4, which are induced by N. parisii and Orsay virus infection, were also found to be induced in pals-22 mutants and brought back to wild-type levels in pals-22 pals-25 double mutants (Fig 6C). In addition, we found that the ubiquitin ligase components are similarly regulated (Fig 6D). These studies thus define IPR genes as the 80 genes that are: 1) induced by N. parisii infection, 2) induced in a pals-22 mutant background, and 3) reversed back to wild-type levels in pals-22 pals-25 double mutants.

IPR genes are also regulated by proteasomal stress
Previous work indicated that blockade of the proteasome, either pharmacologically or genetically, will induce expression of a subset of IPR genes [9]. To determine the IPR genes that are a compilation of three independent replicates, with at least 100 animals analyzed in each replicate.  see S1 Table for more information. (B) qRT-PCR measurement of gene expression, shown as the fold change relative to wild-type control. (C) Survival of animals after 2 hour heat shock treatment at 37˚C followed by 24 hours at 20˚C. Strains were tested in triplicate, with at least 30 animals per plate. Assay was repeated three independent times with similar results, and data from a representative experiment are shown. Mean fraction alive indicates the average survival among the triplicates, errors bars are SD. �� p < 0.01, ns, not significant with Student's t-test as compared to wild-type control. (D) N. parisii FISH signal quantified using a COPAS Biosort machine to measure the mean red signal normalized by length of individual animals, indicated by red dots. Mean signal of the population is indicated by black bars, with error bars as SD. Graph is a compilation of three independent replicates, with at least 100 animals analyzed in each replicate. ��� p < 0.001, ns, not significant with induced by proteasome stress, we performed RNA-seq analysis to define the whole-genome response to this stress. Again, we conducted differential expression analysis and compared gene expression of animals after 4 hours of exposure to the proteasome inhibitor bortezomib compared to the DMSO vehicle control. From these experiments we determined that 988 genes are induced following bortezomib treatment, using the cut-off mentioned above and described in the Materials and Methods. Interestingly, nearly all of the IPR genes described above are also induced following bortezomib treatment (Fig 6E), an overlap that is highly significant (representation factor (rf) = 10.8; p<2.68E-74; rf is the ratio of actual overlap to expected overlap where rf>1 indicates overrepresentation and rf<1 indicates underrepresentation; see Materials and Methods). Previous work has shown that genes induced by N. parisii do not include the proteasome subunits induced by proteasome blockade as part of the bounceback response [9]. The bounceback response is induced via the transcription factor SKN-1. Consistent with these results, here we find that the IPR genes induced by bortezomib are distinct from those regulated by the transcription factor SKN-1, as defined by a previous study [21]. The overlap between SKN-1 regulated genes and IPR genes includes only one gene ( Fig  6E), a number expected by random chance (rf = 0.6; p<0.487).
As shown earlier, pals-22 mutants have increased resistance to heat shock, and previous work indicated that there is overlap between genes induced by chronic heat stress and genes induced by N. parisii and virus infection [9]. However, the genes in common are distinct from the canonical chaperones, or Heat Shock Proteins (HSPs), which are induced by the heat shock transcription factor HSF-1. To learn more about the connection between heat shock response, HSF-1, and the IPR, we compared the IPR genes with those induced by HSF-1 as defined in a previous study [22]. Here we found no genes in common between our set of 80 IPR genes and the set of 365 genes upregulated by HSF-1 (rf = 0; p<0.073) (S5 Table). We also compared the 365 genes upregulated by HSF-1 with the 702 genes that are regulated by pals-22/25 and found 16 genes in common (rf = 0.7; p<0.088) (S5 Table). These are approximately the numbers expected by random chance and indicate that pals-22/25 regulate a distinct set of genes compared to those regulated by HSF-1.

pals-22 and pals-25 regulate expression of genes induced by other natural pathogens
Next, we used Gene Set Enrichment Analysis (GSEA) to broadly compare pals-22/25-regulated genes to genes regulated by other pathogens, stressors, and stress-related pathways. Here we found that pals-22/25 does not significantly regulate expression of genes induced by the Gramnegative bacterial pathogen P. aeruginosa or the Gram-positive bacterial pathogens Staphylococcus aureus and Enterococcus faecalis as analyzed in previous studies (Fig 7). Notably, the strains used in these studies are clinical isolates. Furthermore, these pathogen species are not known to be natural pathogens of nematodes and are not found inside C. elegans intestinal cells before there is extensive tissue damage in the host [23]. (Refer to S7 Table for  Gene sets 1 and 2 were obtained from RNA-seq data outlined in this study, and Gene set 3 was obtained via RNA-seq in a previous study [9]. Gene sets 1 and 2 have a highly significant overlap (rf = 3.9; p<1.0E-317), gene sets 1 and 3 significantly overlap (rf = 2.8; p<1.41E-24), as do gene sets 2 and 3 (rf = 9.6; p<2.48E-63). We define the IPR genes as the 80 genes common across the three gene sets.  [21]. The overlap between IPR genes and genes induced following BTZ treatment is highly significant (rf = 10.8; p<2.68E-74), while the overlap between IPR genes and SKN-1 regulated genes is that expected by random chance (rf = 0.6; p<0.487). The overlap between BTZ-induced genes and SKN-1 regulated genes is considered significant (rf = 1.8; p<2.04E-4). See S2 and S3 Tables for detailed expression levels of genes discussed here. pathogen Drechmeria coniospora infects and penetrates the epidermis of nematodes, triggering a GPCR signaling pathway that upregulates expression of neuropeptide-like (nlp) genes including nlp-29 to promote defense [19]. Our transcriptome analysis shows that pals-22/25 regulates a number of genes in common with Drechmeria infection (rf = 0.8; p<0.009), though this number is approximately what is expected by chance (S5 Table). Notably, these genes do not include the well-characterized neuropeptide nlp defense genes, although they do include many of the pals genes. A more recently described natural pathogen of C. elegans is Myzocytiopsis humicola, which is an oomycete that also infects through the epidermis and causes a lethal infection [24]. Here, pals-22/25 regulate a significant number of genes in common (rf = 4.4; p<5.68E-11) with those induced by M. humicola 24h post infection, including the chitinase-like 'chil' genes that promote defense against this pathogen (S5 Table). Interestingly, these chil genes, like the pals genes, are species-specific [8,24].
We next used OrthoList 2 [25] to determine which genes identified from our RNA-seq analyses have predicted human orthologs. Of the 702 genes regulated by pals-22/25, 279 genes (39.7%) have predicted human orthologs (S8 Table), which is somewhat fewer genes than expected by chance (rf = 0.7; p<9.15e-20). In contrast, of the 368 genes induced in hsf-1 mutants, 190 (51.6%) have predicted human orthologs, which is approximately what is expected by chance (rf = 0.9; p<0.04). Therefore, more of the genes regulated by the conserved  Table. For details on the gene sets used see S6 Table. https://doi.org/10.1371/journal.ppat.1007528.g007 transcription factor hsf-1 have human orthologs compared to genes regulated by the C. elegans-specific pals-22/25 gene pair. Furthermore, when we restrict our analysis to just the 80 IPR genes, only 14 (17.5%) have predicted human orthologs, which is far fewer than expected by chance (rf = 0.3; p<9.82e-13), indicating that the transcriptional response to natural infection is significantly enriched for genes that are not well-conserved.

pals-22/25 control expression of epidermal defense genes induced by oomycetes
As described above, the RNA-seq analysis of genes regulated by pals-22/25 indicated that this gene pair controls expression of genes induced by diverse natural pathogens of C. elegans. Indeed, in a forward genetic screen for C. elegans genes that regulate expression of the M. humicola-induced chil-27p::GFP reporter, we isolated independent loss-of-function alleles of pals-22 (Fig 8A). These mutant alleles cause constitutive expression of chil-27p::GFP in the epidermis in the absence of infection (Fig 8B). RNAi against pals-22 also led to constitutive GFP expression (S5 Fig) in a manner that is indistinguishable from that observed upon infection with M. humicola. These results indicate that pals-22 acts as a negative regulator of chil-27 expression in the epidermis.
We then used a pals-22; chil-27p::GFP strain for a suppressor screen, analogous to the one described earlier for suppressors of GFP expression in pals-22; pals-5p::GFP (Fig 1). Interestingly, in this new screen we isolated two new alleles of pals-25 which fully suppress the constitutive expression of chil-27p::GFP seen in pals-22 mutants (Fig 8A and 8B), indicating that wild-type pals-25 acts as a positive regulator of chil-27 expression. These observations are consistent with our RNA-seq differential gene expression analysis, which determined that chil-27 is induced in a pals-22 mutant background and that pals-25 is required for this induction ( Fig  8C). Therefore, pals-22/25 act as a switch not only for genes induced in the intestine by natural intestinal pathogens, but also for genes induced in the epidermis by natural epidermal pathogens of C. elegans.

Discussion
In many organisms, there is a balance between growth and pathogen resistance. In particular, many studies in plants have indicated that genetic immunity to disease comes at a cost to the yield of crops [26]. Here we define a program in C. elegans that controls a balance between organismal growth with resistance to natural pathogens, which is regulated by the pals-22/25 species-specific gene pair. These genes act as a switch between a 'defense program' of enhanced resistance against diverse natural pathogens like microsporidia and virus, improved tolerance of proteotoxic stress and increased defense against exogenous RNA [8], and a 'growth program' of normal development and lifespan (Fig 9). We call this physiological defense program the "IPR" and it appears to be distinct from other canonical stress response pathways in C. elegans, including the p38 MAP kinase pathway, the insulin-signaling pathway, and the heat shock response, among others [6]. Our previous analyses indicated that ubiquitin ligases may play a role in executing the IPR program, as the cullin/CUL-6 ubiquitin ligase subunit is required for the enhanced proteostasis capacity of pals-22 mutants [6].
Our analysis indicates that pals-22 is genetically upstream of pals-25: pals-25 mutations have effects if pals-22 is mutated, but there is not an obvious effect of pals-25 mutations in a wild-type background. Interestingly, the pals-22 gene is also upstream of the pals-25 gene in an operon and the two genes are transcribed together. Whether the operon structure of pals-22/ 25 contributes to their genetic interactions is unknown, but neither appeared to have a substantial effect on mRNA expression of the other. Therefore, we favor a model whereby the https://doi.org/10.1371/journal.ppat.1007528.g008 epistatic interaction between these two genes is due to an interaction between their protein products, rather than because one gene affects the transcription of the other. Interestingly, pals-22/25 are part of a species-specific expansion of the pals gene family in C. elegans, and their genomic location is near F-box-containing genes that are among the most rapidly diversifying genes in the C. elegans genome. It has been proposed that these genes, as well as other species-specific gene family expansions, are involved in immune defense in C. elegans [27][28][29][30].
pals-22 mutants are highly resistant to the microsporidian pathogen N. parisii, which is the most common parasite found in wild-caught C. elegans [18,31]. Little is known about innate immune pathways that provide defense against N. parisii. Canonical immune pathways in C. elegans like the p38 MAP kinase pathway provide defense against most pathogens tested in C. elegans but do not provide defense against N. parisii [12]. The mechanism by which pals-22/25 regulate resistance to N. parisii is not clear. Our RNA-seq analysis demonstrates that pals-22/ 25 affect expression of hundreds of genes in the genome. In particular, most of the genes induced by the natural intracellular pathogens Orsay virus and N. parisii are controlled by pals-22 and pals-25, although the function of these IPR genes in defense is unknown. Interestingly, we found that pals-22/25 regulate expression of genes induced not only by natural intestinal pathogens but also of genes induced by natural epidermal pathogens, such as the oomycete species M. humicola. M. humicola induces expression of chil-gene family, and genetic analysis shows these genes promote defense against M. humicola [24]. Notably, we identified pals-22/25 in independent forward genetic screens for regulators of chil-27 and found that they regulate expression of this defense gene in the epidermis. Thus, pals-22/25 regulate expression of genes induced by diverse natural pathogens.
While pals-25 is required to activate IPR gene expression in a pals-22 mutant background, it is not required for activation of IPR gene expression in response to N. parisii infection or proteasomal stress. Therefore, pals-22/25 may not mediate detection of these triggers, although they might mediate detection and be redundant with other factors. Intriguingly, the pals-22/25 gene pair share evolutionary and phenotypic features with plant R gene pairs, which serve as sensors for virulence factors delivered into host cells by co-evolved plant pathogens. For example, the Arabidopsis thaliana gene pair RRS1 and RPS4 are species-specific, share the same promoter, and direct opposite outcomes, with RRS1 inhibiting and RPS4 promoting 'effectortriggered immunity' against natural pathogens [32,33]. Similarly, pals-22 and pals-25 are species-specific, appear to be in an operon together, and direct opposite physiological outcomes including defense against natural pathogens. RRS1 and RPS4 proteins directly bind to each other, and RRS1 normally inhibit RPS4 function until detection of bacterial virulence factors, at which point RRS1 inhibition is relieved and RPS4 is free to promote pathogen defense, although the steps downstream of RRS1/RPS4 are poorly understood. Although the pals genes do not share sequence similarity with the R genes, in this analogy PALS-22 would inhibit PALS-25 and serve as the 'tripwire' to detect virulence factors from natural pathogens and free PALS-25 to promote the IPR defense program. While this model is attractive, it is purely speculative as we currently have no direct evidence that PALS-22 detects virulence factors. Identification of such hypothetical virulence factors would be the focus of future studies.
The molecular events by which C. elegans detects infection are poorly understood, although nematodes do appear to use a form of effector-triggered immunity or 'surveillance immunity'. Studies with several distinct pathogens have indicated that C. elegans induces defense gene expression in response to perturbation of core processes like translation and the ubiquitin-proteasome system [34]. For example, studies with P. aeruginosa demonstrated that C. elegans detects the presence of the translation-blocking Exotoxin A through its effects on host translation, not through detection of the shape of the toxin [35,36]. In addition to this mode of detection, C. elegans may also detect specific molecular signatures like canonical Pathogen-Associated Molecular Patterns (PAMPs). In all likelihood, several types of pathogen detection are used by C. elegans. Surprisingly however, there have been no direct PAMP ligand/receptor interactions demonstrated for pattern recognition receptors (PRR) in the worm, although there has been a Damage-Associated Molecular Pattern (DAMP)/G-protein-coupled receptor interaction demonstrated to be critical for response to Drechmeria infection [37]. Indeed, C. elegans lacks many PRR signaling pathways that are well described in flies and mammals. For example, the C. elegans single Toll-like receptor tol-1 does not act canonically and worms appear to have lost its downstream transcription factor NFkB, which is critical for innate immunity in flies and mammals [38]. Perhaps conservation of immune genes is only reserved for defense against rare, 'non-natural' pathogens, because genes that are important for immunity are subject to attack and inhibition by microbes [39]. Thus, immune genes that provide defense against natural pathogens from the recent evolutionary past will not be broadly conserved but rather will be species-specific, like rapidly evolving R genes in plants. While R genes have been shown to encode proteins that detect virulence factors secreted into host cells by coevolved plant pathogens, the mechanism by which they activate downstream immune signaling is unclear. We propose that the IPR physiological program regulated by the pals- 22/25 antagonistic paralogs in C. elegans could be analogous to effector-triggered immunity regulated by opposing R gene pairs like RRS1/RPS4 used in plants for resistance against co-evolved pathogens.
Interestingly, an example of vertebrate-specific antagonistic paralogs has recently been described to play a role in regulating nonsense-mediated RNA decay (NMD) [40]. These studies provide a potential explanation to the long-standing question of how gene duplications are retained, when they are presumably redundant immediately following gene duplication. Specifically, this model predicts that gene duplication events can be rapidly retained if the proteins made from these genes are involved in protein-protein interactions. With just one non-synonymous nucleotide change that switches a wild-type copy to become dominant negative within a multimeric signaling complex, a gene duplication event can be selected for and retained in the heterozygote state-i.e. in one generation. Perhaps in this way, new genes can be born and survive, when gene pairs can evolve to direct opposing functions like the Upf3a/3b paralogs in NMD, and the RRS1/RPS4 and pals-22/25 paralogs in immunity/growth.

Strains
C. elegans were maintained at 20˚C on Nematode Growth Media (NGM) plates seeded with Streptomycin-resistant E. coli OP50-1 bacteria according to standard methods [41]. We used N2 wild-type animals. Mutant or transgenic strains were backcrossed at least three times. See S1 Table for a list of all strains used in this study.

EMS screens and cloning of alleles
pals-22 mutant worms (either the jy1 or jy3 allele) carrying the jyIs8[pals-5p::GFP, myo-2p:: mCherry] transgene were mutagenized with ethyl methane sulfonate (EMS) (Sigma) using standard procedures as described [42]. L4 stage P0 worms were incubated in 47 mM EMS for 4 hours at 20˚C. Worms were screened in the F2 generation for decreased expression of GFP using the COPAS Biosort machine (Union Biometrica). Complementation tests were carried out by generating worms heterozygous for two mutant alleles and scoring pals-5p::GFP fluorescence. For whole-genome sequencing analysis of mutants, genomic DNA was prepared using a Puregene Core kit (Qiagen) and 20X sequencing coverage was obtained. We identified only one gene (pals-25) on LGIII containing variants predicted to alter function in both mutants sequenced (jy9 and jy100). Additional pals-25 alleles were identified by Sanger sequencing. Screens carried out in the strains with the icbIs4[chil-27p::GFP, col-12p::mCherry] transgene were performed in a similar manner except that we used 24 mM EMS to recover the pals-22 alleles (icb88, icb90) and 17 mM EMS for the pals-22(icb90) suppressor screen and that in both cases we selected F2 animals manually using a Zeiss Axio ZoomV16 dissecting scope. The two pals-22 alleles (icb88, icb90) were identified by whole genome sequencing of GFP positive F2 recombinants after crossing to the polymorphic isolate CB4856 as previously described [43] whereas the two pals-25 alleles were found by direct sequencing of the mutant strains. The pals-22(icb89) allele was identified by Sanger sequencing. See S1 Table for a list of all mutations identified.

RNA interference
RNA interference was performed using the feeding method. Overnight cultures of RNAi clones in the HT115 bacterial strain were seeded onto NGM plates supplemented with 5mM IPTG and 1mM carbenicillin and incubated at 25˚C for 1 day. Eggs from bleached parents or synchronized L1 stage animals were fed RNAi until the L4 stage at 20˚C. For all RNAi experiments an unc-22 clone leading to twitching animals was used as a positive control to test the efficacy of the RNAi plates. The pals-22 RNAi clone (from the Ahringer RNAi library) was verified by sequencing and was effective in inducing pals-5p::GFP expression. The pals-25 RNAi clone was made with PCR and includes 1079 base pairs spanning the second, third, and fourth exons of pals-25. This sequence was amplified from N2 genomic DNA, cloned into the L4440 RNAi vector, and then transformed into HT115 bacteria for feeding RNAi experiments.

Quantitative RT-PCR
Endogenous mRNA expression changes were measured with qRT-PCR as previously described [6]. Synchronized L1 worms were grown on NGM plates at 20˚C to the L4 stage and then collected in TriReagent (Molecular Research Center, Inc.) for RNA extraction. For N. parisii infection, 7 x 10 6 spores were added to plates with L4 stage worms and then incubated at 25˚C for 4 hours before RNA isolation. Bortezomib (or an equivalent amount of DMSO) was added to L4 stage worms for a final concentration of 20 μM; plates were then incubated at 20˚C for 4 hours before RNA isolation. At least two independent biological replicates were measured for each condition, and each biological replicate was measured in duplicate and normalized to the snb-1 control gene, which did not change upon conditions tested. The Pffafl method was used for quantifying data [45].

Heat shock assay
Worms were grown on standard NGM plates until the L4 stage at 20˚C and then shifted to 37˚C for two hours. Following heat shock, plates were laid in a single layer on the bench top for 30 minutes to recover, and then moved to a 20˚C incubator overnight (dx.doi.org/ 10.17504/protocols.io.v6re9d6). Worms were scored in a blinded manner for survival 24 hours after heat shock; animals not pumping or responding to touch were scored as dead. Three plates were assayed for each strain in each replicate, with at least 30 worms per plate, and three independent assays were performed.

GFP fluorescence measurement
Synchronized L1 stage animals were grown at 20˚C to the L4 stage. The COPAS Biosort machine (Union Biometrica) was used to measure the time of flight (as a measure of length) and fluorescence of individual worms. At least 100 worms were measured for each strain, and all experiments were performed in biological triplicate.

Lifespan
L4 stage worms were transferred to 6 cm NGM plates seeded with OP50-1 bacteria and incubated at 25˚C. Worms were scored every day, and animals that did not respond to touch were scored as dead. Animals that died from internal hatching or crawled off the plate were censored. Worms were transferred to new plates every day throughout the reproductive period. Three plates were assayed for each strain in each replicate, with 40 worms per plate.

Microscopy
Worms were anesthetized with 10 μM levamisole in M9 buffer and mounted on 2% agarose pads for imaging. Images in S1B and S1C Fig were captured with a Zeiss LSM700 confocal microscope. All other C. elegans images were captured with a Zeiss AxioImager M1 or Axio Zoom V16.

P. aeruginosa pathogen load
Overnight cultures of a P. aeruginosa PA14-dsRed strain [47] were seeded onto SK plates with 50 μg/ml ampicillin, and then incubated at 37˚C for 24 hours followed by 25˚C for 24 hours. Worms at the L4 stage were washed onto the PA14-dsRed plates, incubated at 25˚C for 16 hours, and then assayed with a COPAS Biosort machine (Union Biometrica) for the amount of red fluorescence inside each animal.

RNA-seq sample preparation
Synchronized L1 stage worms were grown on 10 cm NGM plates seeded with OP50-1 E. coli at 20˚C until worms had reached the L4 stage. N2, pals-22(jy3), and pals-22(jy3) pals-25(jy9) strains were then shifted to 25˚C for 4 hours before harvesting for RNA extraction. Bortezomib (or an equivalent amount of DMSO) was added to plates with L4 stage N2 worms for a final concentation of 20 μM; plates were then incubated at 20˚C for 4 hours before RNA isolation. RNA was isolated with TriReagent purification, followed by RNeasy column cleanup (Qiagen), as described [48]. RNA quality was assessed by Tapestation analysis at the Institute for Genomic Medicine (IGM) at UC San Diego. Paired-end sequencing libraries were then constructed with the TruSeq Stranded mRNA method (Illumina), followed by sequencing on HiSeq4000 machine (Illumina). RNA-seq reads were uploaded to the NCBI GEO database with Accession number GSE118400.

RNA-seq analysis
Sequencing reads were aligned to WormBase release WS235 using Bowtie 2 [49], and transcript abundance was estimated using RSEM [50]. Differential expression analysis was performed in RStudio (v1.1.453) [51] using R (v3.50) [52] and Bioconductor (v3.7) [53] packages. As outlined in the RNAseq123 vignette [54], data was imported, filtered and normalized using edgeR [55], and linear modeling and differential expression analysis was performed using limma [56]. An FDR [57] cutoff of <0.01 was used to define differentially expressed genes; no fold-change criteria was used. Lists of upregulated genes used for comparisons were exported and further sanitized to remove dead genes and update WBGeneIDs to WormBase release WS263. Pseudogenes were annotated using a list obtained from WormBase release WS262.

Functional enrichments analyses
Functional analysis was performed using Gene Set Enrichment Analysis (GSEA) v3.0 software [58,59]. Normalized RNA-seq expression data were converted into a GSEA-compatible filetype and ranked using the signal-to-noise metric with 1,000 permutations. Gene sets from other studies were converted to WBGeneIDs according to WormBase release WS263. Independent analyses were performed for each of three comparisons: untreated pals-22(jy3) versus untreated N2 animals; untreated pals-22(jy3) versus untreated pals-22(jy3) pals-25(jy9) animals; bortezomib treated N2 versus DMSO vehicle control treated N2. Results were graphed based on their NES-value using GraphPad Prism 7 (GraphPad Software, La Jolla, CA).
Gene sets comparison lists were generated using WormMine and formatted in Excel. Representation factors and significance of overlaps were calculated using a hypergeometric test implemented at (nemates.org). For "total number of genes" input we use the most conservative number 11,387, which corresponds to the size of our RNA-seq dataset following the filtering of low-count and undetected genes.

Reciprocal BLAST search for pals-22/25 in C. inopinata
Amino acid sequences for PALS-22 and PALS-25 were searched against the C. inopinata (C. sp34 NK74SC) proteome using BLASTP with default settings (E-value<1.0E-10) (http://blast. caenorhabditis.org). PALS-22 returns no hit using default settings while PALS-25 returns a single hit, CSP34.Sp34_50171800.t1. When this is searched against C. elegans with BLASTP it returns PALS-39 as the top hit, while PALS-25 is the seventh hit. When C. elegans PALS-39 is searched against C. inopinata, CSP34.Sp34_50171800.t1 is returned as the top hit. Therefore, this protein appears to be a PALS-39 ortholog. Worms were fixed in paraformaldehyde after 30 minutes of feeding, and fluorescence of accumulated beads in each animal was measured using a COPAS Biosort machine to measure the mean red signal and length of individual animals, indicated by red dots. Mean signal of the population is indicated by black bars, with error bars as SD. Graph is a compilation of three independent replicates, with at least 100 animals analyzed in each replicate. Statistical analysis was performed using one-way ANOVA. ��� p < 0.001, ns, not significant.