The Pseudokinase NIPI-4 Is a Novel Regulator of Antimicrobial Peptide Gene Expression

Hosts have developed diverse mechanisms to counter the pathogens they face in their natural environment. Throughout the plant and animal kingdoms, the up-regulation of antimicrobial peptides is a common response to infection. In C. elegans, infection with the natural pathogen Drechmeria coniospora leads to rapid induction of antimicrobial peptide gene expression in the epidermis. Through a large genetic screen we have isolated many new mutants that are incapable of upregulating the antimicrobial peptide nlp-29 in response to infection (i.e. with a Nipi or ‘no induction of peptide after infection’ phenotype). More than half of the newly isolated Nipi mutants do not correspond to genes previously associated with the regulation of antimicrobial peptides. One of these, nipi-4, encodes a member of a nematode-specific kinase family. NIPI-4 is predicted to be catalytically inactive, thus to be a pseudokinase. It acts in the epidermis downstream of the PKC∂ TPA-1, as a positive regulator of nlp antimicrobial peptide gene expression after infection. It also controls the constitutive expression of antimicrobial peptide genes of the cnc family that are targets of TGFß regulation. Our results open the way for a more detailed understanding of how host defense pathways can be molded by environmental pathogens.


Introduction
Pathogenic microorganisms represent one of the most ubiquitous and powerful sources of selection for higher eukaryotes including humans [1]. Different pathogens have specific natural host tropisms, sometimes broad, as in the case of Pseudomonas aeruginosa [2,3], and in other cases, such as HIV, very narrow. Part of this tropism reflects the divergent mechanisms of host resistance, as exemplified by cultivar-specific resistance in plants [4]. The evolution of adaptive immunity is often cited as an extreme example of immune system evolution. But even among invertebrates that rely on their innate immune systems, there is evidence for considerable variation from the phylum to the species level. For example, in contrast to most other animal species, nematodes, including Caenorhabditis elegans, have lost NF-kB, a key transcription factor in immunity [5,6,7]. Studying the interaction of C. elegans with its natural pathogens therefore sheds light on NF-kBindependent defense pathways.
A number of natural pathogens of C. elegans have been identified, including viruses [8], microsporidia [9], and bacteria such as Microbacterium nematophilum [10] and Serratia marcescens [11,12]. Drechmeria coniospora is a nematophagous fungus that infects C. elegans and other species of nematodes [13]. When C. elegans is sampled from its natural environment, it is often found to be infected with D. coniospora (M-A. Felix, personal communication). D. coniospora produces adhesive conidia that attach to the worm's cuticle. These germinate to produce invasive hyphae that penetrate the cuticle and grow throughout the epidermis [14]. In C. elegans, infection with D. coniospora provokes an innate immune response in the epidermis involving the expression of a large number of genes including those encoding antimicrobial peptides (AMPs) of the NLP and CNC families [15,16,17].
Certain members of each family are found in 2 distinct genomic groups, comprising nlp-27, 28, 29, 30, 31 and 34, referred to as the nlp-29 cluster, and cnc-1, 2, 3, 4, 5 and 11, the cnc-2 cluster. The induction of expression of the genes of the nlp-29 cluster is strongly dependent on the p38 MAPK pmk-1, while that of the cnc-2 cluster genes requires the TGFß dbl-1. The expression of all the genes of the nlp-29 cluster, and some of those of the cnc-2 cluster is also strongly increased in the epidermis if worms are physically injured. In this case, the up-regulation of both the nlp genes and cnc-1, cnc-5 and cnc-11 (but not cnc-2 or cnc-4) is largely dependent upon pmk-1 [16,18,19].
We have shown that for nlp-29 cluster genes, following both infection and injury, inductive signaling passes via TPA-1, a protein kinase C delta (PKCh) that acts upstream of TIR-1, the nematode ortholog of SARM, and a MAPK cassette comprising the MAP3K NSY-1, the MAP2K SEK-1, and PMK-1 [20]. This cascade acts upstream of the STAT-like transcription factor STA-2 that physically interacts with the C-terminus of the SLC6 transporter SNF-12 [21]. SNF-12 is found in endosome-like vesicles in the epidermis, where it may act as a signaling platform during the innate immune response. The elements that contribute to signaling upstream of TPA-1/PKCh have only been partially characterized. Wounding and infection require G-protein signaling, involving the Ga protein GPA-12 and the Gß RACK-1, while infection specifically involves the Tribbles-like kinase NIPI-3 [18,20].
In addition to provoking the increased expression of AMPs, wounding also triggers a rise in intracellular Ca 2+ . This is controlled by an epidermal signal transduction pathway that includes the Ga(q) EGL-30. This pathway is required for actindependent wound closure, but not for injury-induced AMP expression [22]. On the other hand, the Death-associated protein kinase DAPK-1 negatively regulates wound repair and AMP gene expression [23]. Many, but not all, of the elements that act in the epidermis also mediate the innate defenses against intestinal pathogens and toxins [24,25,26,27,28,29,30,31,32]. Conversely, certain genes that participate in p38 MAPK signaling in the intestine, including dfk-2 [28] are not required for the induction of nlp-29 [20].
Our current understanding of both epidermal and intestinal innate immunity is far from complete. In the current study, we therefore undertook a large genetic screen for components of the signaling pathways that control AMP gene expression in the epidermis. We isolated and mapped 26 mutant alleles, uncovering 6 new genes required for AMP gene induction after D. coniospora infection. We cloned one of these genes, nipi-4 (nipi for ''no induction of peptide after Drechmeria infection''). We show here that nipi-4 encodes a nematode-specific protein with a kinase-like domain that is predicted to be a pseudokinase. It acts downstream of PKCh/TPA-1, which was previously shown to modulate the activity of a conserved p38 MAPK cassette [20]. This provides an illustration of an animal family-specific modulation of an innate immune signaling pathway.

A genetic screen for Nipi mutants
We undertook a large-scale genetic screen for mutants that prevented the normal induction of a Pnlp-29::GFP reporter transgene after infection with D. coniospora. From 130,000 mutagenized haploid genomes, we isolated 57 candidate mutant strains. These were then subjected to a confirmatory round of screening and outcrossing. We retained 44 mutant strains that had a sufficiently penetrant phenotype. All behaved as if they were carrying simple recessive alleles. To characterize these in further detail, we first quantified reporter gene expression in uninfected and infected worms. All mutants showed a reduction of Pnlp-29::GFP induction with the most penetrant alleles showing essentially a complete block of the reporter ( Figure 1A).
Three strains exhibited resistance to the phorbol ester PMA. As the only gene known to provoke PMA-resistance in C. elegans is tpa-1 [33,34], we sequenced this gene in one mutant and thus identified a G384E mutation. We presume that the other two mutants are also tpa-1 alleles, but did not characterize them as there are already more than 50 available tpa-1 alleles. For the other 41 strains, we performed classical SNP mapping to assign alleles to individual chromosomes, which was unambiguous for 26 of them. We then performed targeted complementation tests, between alleles, and with candidate genes on the appropriate chromosome. When a new allele failed to complement a candidate gene, the corresponding gene from the mutant was sequenced. This allowed the identification of new alleles for 4 previously characterized Nipi genes, 6 for snf-12, 3 for nsy-1, 2 for sek-1 and 1 for sta-2. These numbers give an indication of the degree of saturation of the screen. The remaining alleles appear to correspond to previously uncharacterized genes. They fall into 6 complementation groups, some represented by multiple alleles ( Figure 1B, Table 1).

Molecular identification of nipi-4
One complementation group was given the name nipi-4 and characterized in detail. Whole-genome resequencing of pooled recombinants [35] between nipi-4(fr106) and the polymorphic Hawaiian strain CB4856 clearly delineated a candidate region for the mutation on the center of chromosome V (Figure 2A). Within this region, only one nonsense mutation was found, in the gene F40A3.5, which is predicted to encode a 396 amino acid membrane-bound protein with a tyrosine kinase domain [36] ( Figure S1). Sequencing this gene from the other nipi-4 alleles revealed 3 independent mutations, a different nonsense mutation in fr71, an alteration of a splice acceptor site in fr99 that would be predicted to lead to a severely truncated protein, and a missense mutation in the kinase domain in fr68 ( Figure 2B, Table 1). This very strongly suggests that nipi-4 corresponds to F40A3.5. In contrast to fr68, which has a milder phenotype, the alleles fr71, fr99 and fr106 all provoke similarly penetrant phenotype and are predicted to correspond to null alleles ( Figure 2C). Transformation rescue confirmed the identity of nipi-4 as F40A3.5 ( Figure 2D).
Interestingly, we have only identified NIPI-4/F40A3.5 orthologs in nematodes. The Caenorhabditis proteins, from elegans, briggsae, brenneri, japonica and remanei species, are predicted to be kinasedead, since they lack the essential aspartic acid active site residue. In contrast, predicted NIPI-4 orthologs from non-Caenorhabditis species like Ascaris suum and Pristionchus pacificus are expected to be functional kinases. Conversely, only the Caenorhabditis proteins have a tyrosine in the predicted activation loop that could potentially be the target of phosphorylation ( Figure S1). We discuss the significance of these observations below.

nipi-4 acts cell autonomously in epidermal cells
To identify the cells in which nipi-4 is expressed, we generated transgenic animals carrying a GFP transcriptional reporter construct ( Figure 2B). We observed expression in the epidermis of C. elegans throughout development ( Figure 3A-E). This pattern overlaps with that of the previously characterized components of the PKCh/p38 MAPK pathway, including snf-12 and sta-2 [21] and suggests that nipi-4 may act in a cell-autonomous manner. To evaluate this directly, we generated transgenic animals in which the expression of nipi-4 was under the control of the col-19 promoter, which is expressed specifically in epidermal cells as animals enter adulthood [37]. In these worms we observed an essentially normal expression of Pnlp-29::GFP upon infection ( Figure 3F-G). On the other hand, expression of nipi-4 in the intestine under the control of the vha-6 promoter [38] did not give any rescue ( Figure S2). Together, this indicates that nipi-4 acts cellautonomously in the epidermis to regulate antimicrobial peptide gene expression.

nipi-4 regulates AMP gene expression after infection and wounding
To define further the function of nipi-4, we assayed the expression of the Pnlp-29::GFP reporter transgene in the nipi-4 mutant background under other conditions that normally lead to its expression, including injury, exposure to PMA and osmotic stress [16,18,20]. In a nipi-4(fr71) mutant, in addition to a nearcomplete block of Pnlp-29::GFP expression after infection, there was no induction of the reporter gene upon needle wounding or exposure to PMA. There was, however, a strong induction of Pnlp-29::GFP expression upon exposure to high salt, comparable to that seen in a sta-2 mutant. Similar results were obtained with nipi-4(fr99) and nipi-4(fr106) ( Figure 4A, and results not shown). This suggests that nipi-4 acts downstream of the PKCh TPA-1 to regulate nlp-29 expression specifically after wounding and infection.
We also analyzed the expression of other genes that have been shown to be induced upon D. coniospora infection [16]. We could confirm by qRT-PCR that in the nipi-4(fr106) mutant, just as in sta-2 or snf-12 mutants [21], the induction of nlp-29 after infection was essentially abrogated. Two other genes of the nlp-29 cluster, nlp-31 and nlp-34 were similarly affected ( Figure 4B). It is interesting to note that in the nipi-4 and sta-2 mutants the constitutive expression of nlp-34 was reduced by 10 fold whereas it was not greatly changed for nlp-29 and nlp-31. The genes of the cnc-2 cluster are regulated in a manner distinct from nlp-29 as their induction after D. coniospora infection is p38 MAPK independent. Rather their induction requires signaling via a non-canonical TGFb/DBL-1 pathway [19]. We found by qRT-PCR that loss of function of nipi-4 strongly affected the constitutive expression of cnc-1 and cnc-2 and to a lesser extent cnc-4. This parallels the phenotype due to loss of sta-2 function ( Figure 4B), as well as snf-12 [21]. Indeed, as discussed below, the constitutive expression of :GFP reporter (frIs7) following infection including sta-2(ok1860), nipi-3(fr4), tpa-1(k530) and 38 new alleles, 11 of which have been determined to define 6 new independent complementation groups. The average fold induction for each strain is represented after standardization across different independent experiments by normalizing to 10 the fold induction between the wild type strain infected versus non infected. (B) Genetic map of Nipi loci identified from screens or from candidate gene approaches. The map has been scaled to the genome sequence, as in [48]. The Nipi genes identified in the present mutagenesis and in previous studies [15,18,20,21]  these genes was reduced to such a degree that it is technically difficult to evaluate the extent of gene induction after infection. Thus, like snf-12 and sta-2, nipi-4 plays a role in innate immune signaling and influences targets of both the PKCh/p38 MAPK/ PMK-1 and TGFb/DBL-1 pathways.

Modulation of AMP gene expression by gpa-12 requires nipi-4
Our previous dissection of the innate immune signaling pathways that govern AMP expression in the epidermis relied on the use of PMA to activate TPA-1/PKCh and an active form of the Ga protein GPA-12 (GPA-12* [34]), produced under the control of a heat-shock promoter [20,21]. As both PMA and heatshock have pleiotropic effects on the physiology of C. elegans, we developed a more refined tool, with GPA-12* under the control of the col-19 promoter, driving its expression in the adult epidermis [37]. We injected this construct into worms carrying an integrated Pnlp-29::GFP reporter. In uninfected transgenic worms carrying the Pcol-19::GPA-12* construct, we observed a very marked increase in the expression of Pnlp-29::GFP in the epidermis from the late L4 stage onwards. The level of reporter gene expression was even further increased upon infection with D. coniospora ( Figure 5A & B). As expected, increased Pnlp-29::GFP expression was totally abrogated in a tpa-1 mutant (results not shown). We found by qRT-PCR that the transgenic strain exhibited an   Figure 5C). The results mirrored to a striking degree the pattern of gene expression changes induced by infection ( Figure 4B). When we crossed the Pcol-19::GPA-12* transgene into nipi-4(fr106) mutant, the elevated expression of the Pnlp-29::GFP reporter provoked by the active form of GPA-12 was abrogated, to a similar degree as in the sta-2 mutant background ( Figure 5A). The effect of GPA-12* on the expression of other cnc and nlp genes was also abolished in the nipi-4(fr106) mutant, as judged by qRT-PCR ( Figure 5C). Together, these results confirm the role of nipi-4 as a novel regulator of AMP gene expression during the infection of the worm, acting genetically downstream of PKCh TPA-1.

Discussion
To characterize the molecular pathways that underpin antifungal innate immunity in C. elegans, we previously undertook a small-scale genetic screen for genes required for the induction of an AMP reporter gene after infection. We isolated and characterized 5 alleles that fall into 4 complementation groups. This provided a framework to understand the regulation of antimicrobial peptide expression in the C. elegans epidermis [39]. In the current study, we chose to extend the approach, with the aim of performing a saturating screen. Counting just the alleles that were amenable to classical SNP mapping, we found a total of 11 complementation groups, of which 2 corresponded to genes that had been hit in the previous screen, and 3 to genes previously known to be involved in the regulation of nlp genes, leaving a total of 6 new complementation groups.
The current genetic screen has reinforced the importance of the p38 MAPK pathway as a central part of anti-fungal defenses in the epidermis as we recovered multiple alleles of the MAP3K nsy-1 and the MAP2K sek-1. A common p38 MAPK signaling cassette is also required for resistance to intestinal bacterial infection. In a screen for genes that both regulate the constitutive expression of the intestinal gene T24B8.5 and that are required for resistance to P. aeruginosa infection, 14 nsy-1, 8 sek-1, 7 pmk-1 and 3 tir-1 alleles were isolated, together with one allele of a new actor in the intestinal p38 pathway, the transcription factor atf-7 [29]. Our screens might have allowed the identification of many more distinct genes, presumably because they were less stringent. We may well find alleles of pmk-1 and tir-1 among the alleles that have yet to be mapped.
The genes of the cnc-2 and nlp-29 clusters share a common evolutionary origin [16]. It is reasonable to imagine that they were initially regulated by a common mechanism, but this is clearly no longer the case. Certain genes, however, participate in the regulation of both groups of genes. Indeed, the results we obtained using nipi-4 have reinforced observations that we previously made with snf-12 and sta-2 mutants [21]. These 3 genes are necessary for the induction of the genes of the nlp-29 cluster, but do not greatly affect their constitutive level of expression. On the other hand, loss of function of any one of the 3 essentially abolishes the constitutive expression of cnc-1 and cnc-2. The drop in the constitutive level of its expression is such that it is not technically possible for us to determine reliably by qRT-PCR whether D. coniospora infection still provokes an induction of these genes in the nipi-4, snf-12 or sta-2 mutants. In wild-type worms, the induction of the cnc genes is almost entirely dependent upon dbl-1. Changes in dbl-1 expression do not however affect their constitutive expression. Nor does dbl-1 have any effect on the induction of the genes of the nlp-29 cluster [19]. This suggests a model in which nipi-4, snf-12 and sta-2 are involved in two different processes, one that governs the constitutive expression of cnc-1 and cnc-2, and the other that controls the inducibility of genes of the nlp-29 cluster. The degree to which these two functions are interdependent remains to be established, as does the exact impact of nipi-4 on cnc-4, as it is less clear-cut.
In addition to these open questions, the structure of NIPI-4 itself raises a number of issues. As mentioned above, NIPI-4-like proteins are only found in a subset of nematodes, and are never duplicated. Certain defense mechanisms have been lost in parasitic nematodes such as Meloidogyne incognita [40], but a NIPI-4 homolog can be found in Meloidogyne species and in the animal parasite Ascaris suum, so this protein is not restricted to free-living nematodes. Interestingly, while orthologs in A. suum and Pristionchus pacificus possess the characteristic catalytic aspartate residue, this residue is absent in all Caenorhabditis species, so these proteins are predicted to be catalytically inactive pseudokinases. This suggests that NIPI-4 has evolved a kinase-independent function in Caenorhabditis species. NIPI-4 might compete with one or more active kinases for substrates or binding partners. In such a scenario, the putative kinase(s) would need to play a negative regulatory role. The loss of catalytic activity in NIPI-4 has, however, been mirrored by the acquisition of a potential activation loop phosphorylation site not seen in other species. It is therefore tempting to speculate that NIPI-4 in Caenorhabditis species is able to donate its activation loop to another kinase following heterodimerization, as is seen for example with STRAD and LKB1 [41]. NIPI-4 and all its identified orthologs possess a predicted transmembrane segment, N-terminal to the kinase domain that could allow its association with the membrane of one or more classes of intracellular vesicles. We previously showed that SNF-12 is found in endosome-like vesicles, and that endocytosis is indispensable for the transcriptional response to infection [21]. One could conjecture that NIPI-4, SNF-12 and STA-2 form a signaling complex on endosomes that is activated following physical association with the MAPK PMK-1. Understanding the function of NIPI-4 at the biochemical and cellular level, an objective of future studies, will give insights into how a speciesspecific host defense pathway can be molded by the natural pathogens found in a particular environmental niche.

Mutants Isolation
We mutagenized IG274 wild type worms carrying the frIs7 transgene with EMS using standard procedures [43]. 130,000 genomes were screened using the same criteria described in [18].
Briefly, synchronized F2 worms were infected at the L4 stage with D. coniospora. After 24 h at 25uC, we screened for worms that failed to show an elevated level of GFP expression after D. coniospora infection and transferred them onto nystatin containing NGM plates. Mutant alleles were mapped through standard genetic and bulk SNP mapping by analysis of 20 to 30 recombinants with the strain CB4856 [44]. Genetic complementation tests were done between mutants located on the same chromosome, defining 6 new independent complementation groups.
Whole Genome Sequencing nipi-4(fr106) mutation was further mapped and identified using a whole genome sequencing-SNP mapping protocol [35]. Briefly, nipi-4(fr106) was crossed with Hawaiian CB4856 males and 20 F2 mutant recombinant lines were isolated. The DNA of these pooled lines was prepared using a standard protocol with proteinase K lysis, RNAse A treatment and phenol/chloroform extraction. The pooled DNA was subjected to whole genome sequencing in multiplexed run with 4 samples in one sequencing lane of a v1.5 flowcell on HiSeq 2000 instrument, generating paired 100 nucleotide reads. The results were analyzed using Maqgene [45]. Infection, wounding, exposure to high salt and PMA Infections with D. coniospora and wounding were carried out at 25uC as described [18]. Briefly, animals were infected with D. coniospora at the L4 stage or exposed to high salt and incubated at 25uC. After 18 h, age-matched non-infected animals were used for wounding assays, exposure to PMA, or kept as control. Exposure of worms to high salt (350 mM NaCl) and PMA (1 mg/ml) were done on NGM plates as previously described [20].

Constructs and transgenic lines
Pnipi-4::GFP was obtained by Gateway cloning (Invitrogen TM ). A 2,521 bp fragment upstream of the nipi-4 start site was amplified (with primers JEP1974-JEP1975), cloned into the pDONRP4-P1R vector, then transferred into the destination vector pDEST-DD04-Neo a generous gift from D. Dupuy [46] so that it was cloned upstream of the GFP::unc-54_39UTR cassette. The Pnipi-4::GFP was injected at 20 ng/ml together with Pttx-3::DsRed2 at 70 ng/ml into N2 worms. Two independent lines were generated showing the same expression pattern IG1341 wt; frEx483 and IG1342 wt; frEx484.

Analysis with the COPAS Biosort
Analysis of Pnlp-29::GFP induction in the strain carrying the frIs7 integrated array for the different treatments were all done at the same time on worms 24 h after the L4 stage with the COPAS Biosort (Union Biometrica TM ) [18]. The frIs7 integrated array consists of two reporter transgenes, Pnlp-29::GFP and Pcol-12::DsRed2. As the latter exhibits a constitutive expression in the epidermis that is unaffected by infection or other tested conditions, the fluorescence ratio green/red represents the variation in Pnlp-29::GFP expression normalized for the size of individuals [18]. The mean values are shown normalized to the wild type control that is set to one. The number of animals used in each experiment is given below. As previously described [18], due to the nature of the distribution, standard deviations are not always an informative parameter when measuring fluorescent reporter gene expression using the Biosort. Data are, however, in all cases representative of at least 3 independent experiments. qRT-PCR L4 worms were infected for 6 h at 25uC with D. coniospora. 1 mg of total mRNA from infected and non-infected worms were used for reverse transcription (Applied Biosystems TM ). Quantitative real-time PCR were performed using 1 ml of cDNA in 10 ml of SYBERgreen Applied Biosystems TM and 0.1 mM of primers on a 7500 Fast Real-Time PCR System using act-1 (JEP538-JEP539) as a control, with nlp-29 (JEP848-JEP952), nlp-31 (JEP950-JEP953), nlp-34 (JEP969-JEP970), cnc-1 (JEP1087-JEP1088), cnc-2 (JEP944-JEP549) and cnc-4 (JEP1124-JEP1125), for primer sequences see [21]. Results were normalized to act-1, and then relative expression calculated using 2((A+10)2x), A being the normalized cycle number for nlp-34 in the non-infected sample and x the value of interest. Control and experimental conditions were tested in the same run. Means and SEMs were calculated from a minimum of 3 independent experiments.

Supporting Information
Figure S1 Alignment of the predicted NIPI-4 proteins. Accession numbers for the different proteins are the following: C. elegans NP_505028, C. remanei XP_003115465, C. brenneri EGT43601, C. briggsae CAP37545, C. japonica JA58647. The Ascaris and Pristonchius proteins present in Genbank (ADY47863, PP41334) appear to have been mis-predicted. The figure presents more plausible predictions based on manual editing, respecting splice consensus sequences, of the output from tblastn using the C. elegans NIPI-4 protein against the relevant genomic sequence. All included sequences were found as significant matches with a smallest sum probability of at least e-25. For Meloidogyne hapla, Oncocera volvulus, Strongyloides ratti only partial sequences are presented, no attempt to reconstruct complete sequences was made (*). Alignments were produced with Clustal W2 (http:// www.ebi.ac.uk/Tools/msa/clustalw2/) and Boxshade (http:// www.ch.embnet.org/software/BOX_form.html). Thanks to G.