Evolutionary plasticity in the innate immune function of Akirin

Eukaryotic gene expression requires the coordinated action of transcription factors, chromatin remodelling complexes and RNA polymerase. The conserved nuclear protein Akirin plays a central role in immune gene expression in insects and mammals, linking the SWI/SNF chromatin-remodelling complex with the transcription factor NFκB. Although nematodes lack NFκB, Akirin is also indispensable for the expression of defence genes in the epidermis of Caenorhabditis elegans following natural fungal infection. Through a combination of reverse genetics and biochemistry, we discovered that in C. elegans Akirin has conserved its role of bridging chromatin-remodellers and transcription factors, but that the identity of its functional partners is different since it forms a physical complex with NuRD proteins and the POU-class transcription factor CEH-18. In addition to providing a substantial step forward in our understanding of innate immune gene regulation in C. elegans, our results give insight into the molecular evolution of lineage-specific signalling pathways.


Author summary
When animals are infected, as part of their innate immune response, they switch on defence genes that encode proteins that help fight pathogens. We use the nematode Caenorhabditis elegans to understand the steps in this process. When infected, C. elegans can turn on clusters of antimicrobial peptide genes. We have discovered that the coordinated expression of these genes requires a particular chromatin remodelling complex (proteins that open up compact DNA), working in conjunction with a protein called Akirin. Akirin plays a central role in immune gene expression in insects and mammals, but we found that although it has conserved its role of bridging chromatin-remodellers and transcription factors needed for gene expression, the identity of its functional partners is different. Introduction A fundamental part of innate immune responses is the regulated expression of defence genes.
In both vertebrates and many invertebrates, including Drosophila, two of the key regulators controlling innate immunity are the Rel-homology domain (RHD) protein NF-κB and its protein partner IκB [1]. Across many species, NF-κB functions in concert with members of the conserved Akirin family (InterPro: IPR024132) to govern the expression of defence genes [2]. More specifically, in vertebrates, Akirin2 bridges NF-κB and the SWI/SNF chromatin-remodelling complex, by interacting with IκB-z and the BRG1-Associated Factor 60 (BAF60) proteins, downstream of Toll-like receptor (TLR) signalling [3,4]. In insects, an equivalent complex (including Relish and the Brahma-associated proteins BAP55 and BAP60 in Drosophila) governs antimicrobial peptide (AMP) gene expression upon infection by Gram-negative bacteria [4][5][6]. Infection of Caenorhabditis elegans by its natural pathogen Drechmeria coniospora [7] provokes an increase of AMP expression, but in the absence NF-κB and independently of the single TLR gene tol-1 [8,9]. It was therefore surprising that akir-1, the sole nematode Akirin orthologue was identified in a genome-wide RNAi screen for genes involved in the regulation of nlp-29 [10,11], an AMP gene that has been extensively used as a read-out of the epidermal innate immune response (e.g. [12][13][14][15][16]).
These previous studies have revealed surprising molecular innovation in the pathways that regulate AMP gene expression. To give one example, in other animal species, STAT-like transcription factors function in concert with Janus kinases (JAKs). But in C. elegans, although there are no JAKs [17], the 2 STAT-like proteins, STA-1 and STA-2, function in antiviral [18] and antifungal immunity [19], respectively. In the latter case, STA-2's function appears to be modulated by a nematode-specific member of the SLC6 family, SNF-12, acting downstream of the GPCR DCAR-1 and a p38 MAPK pathway to regulate nlp-29 expression [20]. Here, we undertook a focused study of akir-1, to understand how AMP gene expression is governed and also to gain insight into the evolution of lineage-specific signalling pathways. We have been able to identify Akirin's functional partners in C. elegans and thus reveal an unexpected molecular swap at the core of innate immune gene expression.

The C. elegans Akirin homolog is required for antifungal innate immunity
We previously conducted a semi-automated genome-wide RNAi screen [10] for genes that control the expression of the AMP reporter gene nlp-29p::gfp, following infection of C. elegans with D. coniospora [11]. In the screen, sta-1 was used as a negative control since its inactivation has no observable effect on nlp-29 reporter gene expression [19,20]. The candidates identified as positive regulators are collectively referred to as Nipi genes, for No Induction of Peptide expression after Infection. While akir-1(RNAi) caused a robust reduction in the induction of nlp-29p::gfp expression after infection (Fig 1A), it did not significantly affect the size of treated worms, nor the expression of the control col-12p::DsRed reporter transgene (S1A Fig), identifying it as Nipi gene and suggesting that it could have a specific function in innate immunity. When we used an available deletion allele, akir-1(gk528), which is predicted to be a molecular in IG274 worms carrying the integrated array frIs7 (containing nlp-29p::gfp and col-12p::DsRed) treated with RNAi against negative and positive controls (sta-1, dcar-1, respectively) or akir-1 and infected by D. coniospora (Infected), wounded (Wound), treated with a 5 mM dihydrocaffeic acid solution (DHCA) or exposed to 300mM NaCl (High salt). Here and in subsequent figures representing Biosort results, unless otherwise stated, graphs are representative of at least 3 independent experiments. The black bar represents the mean value for (from left to right), n = 189, 164, 179, 184, 183, 169; 23, 21, 30, 29, 36, 68; 97, 86, 114, 111, 104, 94; 96, 85, 100, 150, 129, 113; ÃÃÃÃ p<0.0001, ns p>0.05, Dunn's test. B. Fluorescent images of adult worms carrying frIs7, expressing a constitutively active Gα protein, GPA-12 Ã , in the epidermis and treated with RNAi against the indicated genes. Almost all of the residual GFP expression seen upon akir-1(RNAi), most prominent in the vulval muscle cells, comes from unc-53Bp::gfp used as a transgenesis marker. null, we recapitulated the effect on nlp-29p::gfp expression (S1B Fig). This analysis was, however, hampered by the mutants' pleiotropic phenotypes [21], including a developmental delay and very marked decrease in the expression of the control reporter transgene (S1C Fig). To avoid these confounding effects, and since RNAi of akir-1 gave robust and reproducible results, we used akir-1(RNAi) for our subsequent analyses.
The induction of nlp-29p::gfp expression upon D. coniospora infection is correlated to the infectious burden, which in turn reflects the propensity of spores to bind the worm cuticle [11,22]. There was no reduction in spore adhesion following akir-1(RNAi) (S1D Fig). Many genes required for the induction of nlp-29p::gfp expression after infection, including the GPCR gene dcar-1 [20] and the STAT transcription factor-like gene sta-2 [19], are also required for the transcriptional response of C. elegans to physical injury. We found that akir-1(RNAi) also abrogated reporter transgene expression upon wounding (Fig 1A). One trigger for the epidermal innate immune response is the increase in the tyrosine metabolite HPLA that accompanies infection with D. coniospora. HPLA acts via DCAR-1 to activate a p38 MAPK signalling cascade [20]. This GPCR can also be activated by the HPLA tautomer DHCA [23], a non-physiological ligand, which we use routinely as it is somewhat more potent and less toxic for worms than HPLA [20]. The induction of nlp-29p::gfp expression upon exposure to DHCA was greatly reduced upon akir-1(RNAi), to a degree that was comparable to dcar-1(RNAi) (Fig 1A). Together, these results suggest that akir-1 is required for the activation of the epidermal innate immune response, downstream of DCAR-1.

akir-1 likely acts in the epidermis to modulate AMP expression
To evaluate when and where akir-1 was expressed, we generated strains carrying a transcriptional reporter gene (akir-1p::gfp). Consistent with previous studies [26], expression of GFP was observed from the late embryo stage onwards, peaking at the late L4 stage. Expression was most evident in the lateral epithelial seam cells, the major epidermal syncytium, hyp7, as well as in multiple head and tail neurons (Fig 2A). The different components of the p38 MAPK pathway, including dcar-1, gpa-12 and sta-2, act in a cell autonomous fashion in the epidermis [13,19,20]. To determine whether this was also the case for akir-1, we knocked down its expression in the epidermis, using the previously characterized strain IG1502 [11,20]. This greatly decreased nlp-29p::gfp expression upon infection, and also, as judged by qRT-PCR substantially reduced the induction of all the genes of the nlp-29 locus, while not affecting their constitutive expression (Figs 2B, 2C & S2A and S2B). Although low levels of RNAi silencing in non-epidermal tissues have been reported for the strain JM43 [27] from which IG1502 was derived, overall our results suggest that akir-1 acts in a cell autonomous manner in the epidermis to modulate AMP gene expression upon infection.
To test the functional relevance of these observations, we assayed the effect of akir-1(RNAi) on the resistance of C. elegans to D. coniospora infection. Compared to the negative control, sta-1(RNAi), knocking down akir-1 principally in the epidermis (with strain IG1502) was associated with a significant reduction in survival (Fig 3). Interpretation of this result is complicated by the fact that the same RNAi treatment also caused a significant decrease in longevity on non-pathogenic E. coli (S2C Fig), so the reduced resistance to D. coniospora infection is not likely to result solely from the observed diminution in AMP gene expression.

akir-1 acts with NuRD and MEC chromatin remodelling complexes to modulate AMP expression
AKIR-1 is a member of the Akirin family. While invertebrates generally have just one Akirin protein, vertebrates can have up to 8 [28]. In mice and humans there are 2 paralogues [29]. AKIR-1 is much more similar to murine Akirin2 than Akirin1 (32.8% vs 10.9% overall sequence identity by BLASTP). While Akirin1 has been proposed to be involved in muscle regeneration and cell chemotaxis [30], as mentioned above, Akirin2 has a conserved function controlling innate immune gene expression through its interaction with BAF60/BAP60 and more generally the SWI/SNF chromatin-remodelling complex [2,3,5]. We therefore used RNAi to knock down the expression of components of the nematode SWI/SNF chromatinremodelling complexes, but also of the Nucleosome Remodelling and histone Deacetylase (NuRD) and MEC complexes, as well as related genes [31]. With the exception of swsn-1, which caused pleiotropic development defects and affected expression of the control col-12p:: DsRed reporter transgene, consistent with our previous results [11], none of the other SWI/ SNF genes appeared to be required for nlp-29p::gfp expression (S3A and S3B Fig). On the other hand, knocking down 6 genes dcp-66, hda-1, let-418, lin-40, lin-53, and mep-1, largely, and specifically, blocked the expression of nlp-29p::gfp upon D. coniospora infection (Figs 4A, S3C and S3D). Of note, the 3 RNAi clones that gave the most robust Nipi phenotype, those targeting hda-1/HDAC, lin-40/MTA and dcp-66/p66, had been identified in the previous genome-wide screen [11]. These 3 genes encode core subunits of the two canonical chromatin-remodelling (NuRD) complexes in C. elegans. The two complexes also share LIN-53/RbAp, but differ in their Mi-2 orthologs, having either LET-418 or CHD-3. LET-418 but not CHD-3, can interact with the Krüppel-like protein MEP-1 in a distinct complex, the MEC complex [31,32]. Our results suggest that both the LET-418-containing NuRD complex and the MEC complex are involved in defence gene expression. The 6 RNAi clones also strongly abrogated the elevated expression of nlp-29p::gfp normally seen in worms expressing GPA-12 Ã , in clear contrast to chd-3(RNAi) (Fig 4B). RNAi with the same 6 clones also blocked the induction of reporter gene expression in the strain IG1502 (S4A Fig). Under these conditions, (i.e. RNAi principally in the epidermis), the induction of expression of 5 endogenous nlp AMP genes normally provoked by D. coniospora infection was also severely compromised (S4B Fig). In contrast, there was no evidence for a role for HDA-2, RBA-1 or EGL-27 (Fig 4A), the respective homologues of the core subunits HDA-1, LIN-53 and LIN-40, that do not form part of either of the 2 biochemically characterized NuRD complexes in C. elegans [31]. Together these results suggest that both the LET-418 containing NuRD complex and the MEC complex act cell autonomously in the  epidermis, downstream of (or in parallel to) gpa-12, to control nlp AMP gene expression upon D. coniospora infection. Further, they suggest that in contrast to what has been described in flies and mammals, AMP gene expression is not dependent upon the SWI/SNF complex in C. elegans and raised the possibility that AKIR-1 might function together with the NuRD and MEC chromatin remodelling complexes.

AKIR-1 forms a complex with components of the NuRD and MEC chromatin remodelling complexes and CEH-18
To address this possibility, we took an unbiased biochemical approach to identify the in vivo protein partners of AKIR-1. From a mixed-stage population of worms carrying a functional akir-1p::AKIR-1::gfp construct (S5 Fig), we pulled down AKIR-1::GFP by immunoprecipitation from whole worm extracts and subjected the purified proteins to mass spectrometry analysis ( Fig 5A). Remarkably, all of the proteins that make up the NuRD and MEC complexes were found, i.e. LIN-40, LIN-53, LET-418, HDA-1, MEP-1 and DCP-66. The first 3, together with 6 other known or putative DNA-binding or transcription-related proteins [33], including CEH-18, were among the 53 high confidence protein partners (Fig 5B and 5C). Significantly, 9 of these 53 candidates (p = 2.7x10 -7 ), again including LIN-40 and CEH-18, had been identified in our previous RNAi screen for regulators of nlp-29p::gfp [11]. In the complete list of close to 1400 protein partners, there were a further 111 hits (S1 Table), so overall, fully 35% of the known candidate regulators of AMP gene expression (Nipi genes) were recovered through this independent biochemical approach when one includes the lower confidence candidates.
When we compared the list of 53 high confidence candidate AKIR-1 binding proteins with the 190 proteins identified as potential interactors of the nematode BAP60 homologue SWSN-2.2 [34], we found only 3 common proteins, none of which have been characterized as being specific regulators of nlp-29p::gfp expression (i.e. found as Nipi genes [11]; S2 Table). Using a less stringent list of 190 potential AKIR-1 binding proteins extended the overlap to 11 common partners, with just 2 corresponding to Nipi genes (arp-1 and dlst-1 that encode an actinrelated protein, and a predicted dihydrolipoyllysine succinyltransferase, respectively). The 11 common proteins did, however, also include SWSN-1 and SWSN-4 (S2 Table). This suggests that in some contexts, but not during its regulation of AMP gene expression, AKIR-1 might interact with the SWI/SNF complex. This functional dichotomy was further reinforced by examining the genes differentially regulated following knockdown of both swsn-2.2 and its paralogue ham-3 [34]. There were only a very small number (33/1521) of genes characterized as up-regulated by D. coniospora infection and among them, there were none encoding AMPs (S2 Table). Together these results support the idea that there is a specific AKIR-1-containing protein complex involving the NuRD and MEC chromatin remodellers, required for AMP gene regulation.
We therefore focused on the interaction between AKIR-1 and these chromatin-remodelling factors. We used available antibodies to validate the NuRD and MEC complex proteins LET-418 and HDA-1 as AKIR-1-interactors. Both could be detected together with AKIR-1::GFP, in samples from infected and control worms, derived from the strain used for mass spectrometric analysis, and importantly also from a strain of worms carrying a single copy akir-1::gfp insertion in the wild-type background ( Fig 6A). In the latter strain, AKIR-1::GFP exhibited an predominantly nuclear localization, including in the epidermis (S6 Fig). There was a clear reduction in the quantity of LET-418 that was pulled down with AKIR-1::GFP from the samples of infected worms compared to non-infected worms. The same tendency was observed for HDA-1. These results strongly support the existence of a physical complex between AKIR-1 and the NuRD and MEC complexes in uninfected worms that changes following infection.  Seeking to confirm the potential physical interaction between AKIR-1 and CEH-18, we made use of a strain expressing both AKIR-1::GFP and a doubly-tagged version of CEH-18 (CEH-18::GFP::3xFLAG; [35]). As the AKIR-1::GFP construct is a single-copy insert its expression is expected to be close to that of the endogenous protein; it was not detectable in the total protein extract. When we analysed the complex that was pulled-down together with CEH-18, however, we were readily able to detect AKIR-1::GFP (Fig 6B), lending further support to the proposed AKIR-1/NuRD/CEH-18 complex.

CEH-18 and AKIR-1 have overlapping non-redundant functions
As mentioned above, we previously reported a role for ceh-18 in the regulation of nlp-29p::gfp [11]; the results for 6 independent experiments assaying the effect of ceh-18(RNAi) on reporter gene expression following D. coniospora infection are available at http://bioinformatics.lif. univ-mrs.fr/RNAiScreen (clone sjj_ZC64.3). We were able to confirm this effect using a ceh-18 mutant strain (IG1714) carrying the frIs7 reporter gene: expression was abrogated upon infection compared to the wild-type (Fig 7A). We also demonstrated by qRT-PCR that ceh-18 was required for the increased expression of several nlp genes after infection (Fig 7B). In common with sta-2(RNAi), knocking down ceh-18 by RNAi did not reduce the induction of nlp-29p::gfp provoked by osmotic stress, but did strongly abrogate the elevated reporter gene expression normally seen in worms expressing GPA-12 Ã . RNAi against ceh-18 also significantly reduced the induction of reporter gene expression in the IG1502 strain ( Fig 7A). This is the same pattern of phenotypes as seen with akir-1(RNAi) (Figs 1A, 1B and 2B). The non-redundant function of CEH-18 resembles that of its binding partner, AKIR-1, supporting the hypothesis that they act together in a common complex.
When we assayed the effect of ceh-18(RNAi) on the resistance of C. elegans to D. coniospora infection, we observed a significant reduction in survival of IG1502 that was more pronounced than that seen upon akir-1(RNAi) or sta-2(RNAi) (Fig 7C). Unlike akir-1(RNAi) that reduces worm longevity (S2C Fig), ceh-18(RNAi) extends lifespan [36]. These results therefore support a specific role for ceh-18 in innate defence against D. coniospora infection, potentially via a regulation of immune gene expression.

AKIR-1 binds to AMP gene promoters
We then addressed the question of whether AKIR-1 (and by extension CEH-18) has the potential to interact with DNA, by chromatin immunoprecipitation (ChIP), using the strain of worms carrying a single copy akir-1::gfp insertion. We first tested the specificity of the ChIP by assaying the occupancy of AKIR-1::GFP on the promoter of act-1, an actin-encoding gene that is used as a control for qRT-PCR since its expression is unaffected by D. coniospora infection [37]. We detected a low and constant occupancy of the act-1 promoter using samples from uninfected or infected populations of worms ( Fig 8A). We take this to reflect non-specific binding. We then assayed the capacity of AKIR-1::GFP to associate with DNA fragments corresponding to the promoters of 3 AMP genes, or to their 3' UTRs. For all 3 AMP genes assayed, binding to the 3' UTRs appeared to be non-specific. On the other hand, we observed 1::GFP)] II) were probed with specific antibodies. The results for two independent pull-downs are shown. The presence of HDA-1 and LET-418 (NuRD complex components) could be confirmed. Anti-ACT-1 was used to control the total input for each immunoprecipitation. B. Complexes immunopurified using an anti-FLAG antibody, from a strain coexpressing AKIR-1::GFP and FLAG-tagged CEH-18 (wt; frSi12[pNP157(akir-1p::AKIR-1::GFP)] II; wgIs533[CEH-18:: TY1::GFP::3xFLAG + unc-119(+)]), were probed with anti-FLAG (top panel) and anti-GFP (bottom) antibodies. In addition to the immunopurified complex (IP), the extract before immunopurification (Input), the unbound fraction (flow-through: FT) and proteins immunopurified using an unrelated antibody (Mock) were also analysed.
https://doi.org/10.1371/journal.pgen.1007494.g006 markedly higher binding to the promoter regions relative to the 3' UTRs. The 3 genes, nlp-29, nlp-31 and nlp-34 are strongly induced by D. coniospora infection [24]. There was a >10-fold higher occupancy of AKIR-1::GFP on DNA in the samples from non-infected worms relative to the infected ones ( Fig 8A). Taken together, our results support a model, discussed further below, in which AKIR-1 plays 2 indissociable roles. First, in association with the NuRD and MEC complexes, it binds to the promoters of defence genes and potentially recruits transcription factors including CEH-18. Our results suggest that this does not influence the STA-2-independent basal expression of nlp genes. Secondly, AKIR-1 and its protein partners negatively regulate the STA-2-dependent transcription of defence genes, with this repression being relieved upon their removal from their binding sites following infection (Fig 8B). This could explain why loss of AKIR-1 (or CEH-18) function is associated with an incapacity to express AMP genes upon infection.

Discussion
We are interested in the mechanisms involved in the regulated expression of nlp-29, a representative of one class of AMP genes in C. elegans [24,38,39]. In common with many other AMP genes, the level of nlp-29 mRNA rapidly increases following either physical injury or infection with the nematophagous fungus D. coniospora. In both cases, the integrity of the cuticle and underlying epidermis is compromised. Although we have advanced in our understanding of how this triggers the innate immune response, and how the associated signal transduction pathway is organized, the details of the transcriptional regulation remain to be fully elucidated. We previously identified ELT-3, an epidermis specific GATA factor as being partially required, in a generic fashion, for nlp-29 expression [24]. The STAT-like transcription factor STA-2 plays a more specific role. It is largely dispensable for the constitutive expression of nlp-29, but is required for its induction upon wounding and infection [13,19]. In this work, we have made a considerable step forward by characterizing the key role of AKIR-1 and identifying its protein partners, including the NuRD and MEC complex chromatin remodelling proteins and the transcription factor CEH-18. All these factors are required for AMP gene expression after fungal infection of the nematode epidermis.
CEH-18 is a member of the POU subgroup of the Hox class of homeodomain transcription factors. These are regulators of cellular proliferation, differentiation and migration across species. In C. elegans, ceh-18 has primarily been characterized for its negative regulatory role in a somatic gonadal sheath cell-dependent pathway that governs oocyte meiotic arrest [40]. It has not been implicated in innate immunity previously. Among POU transcription factor genes in Drosophila, Dfr/Vvl, Pdm1/nub and Pdm2/miti were identified in a screen for transcriptional regulators that bind the NF-κB-family transcription factor Dif; they are important for the control of AMP gene expression [41][42][43]. The corresponding proteins were not, however, identified as physical interactors of Akirin in Drosophila [5]. Thus if POU transcription factors do Ratio of green fluorescence (GFP) to size (TOF) in wild-type (IG274) or ceh-18(mg57) mutant (IG1714) worms carrying frIs7, infected or not with D. coniospora for 16 h (yellow and blue, respectively; data for IG274 is as Fig 3B in [53]), and IG274 worms treated with RNAi against sta-1 (control) or ceh-18 and, from left to right, exposed to high salt (purple; cpsf-2(RNAi) is a positive control, sta-2(RNAi) a negative control [11]), in worms also expressing GPA-12 Ã in the epidermis, and in the rde-1(ne219);wrt-2p::RDE-1 background and infected by D. coniospora. For the latter 2 panels, sta-2 (RNAi) is a positive control. A minimum of 45 worms was used for each condition. The black bar represents the mean value; ÃÃÃ p<0.001, ÃÃÃÃ p<0.0001, relative to sta-1(RNAi), Dunn's test; for the other conditions there is not a significance decrease. The results of the 3 right panels are unpublished results from [11], representative of at least 4 independent experiments. B. Quantitative RT-PCR analysis of the expression of several genes in the nlp-29 cluster in wild-type worms, sta-2 and ceh-18 mutants infected by D. coniospora; results are presented relative to those of uninfected worms. Data (with average and SD) are from three independent experiments. ÃÃ , p < 0.001; Ã , p < 0.01; 1-tailed ratio paired t test. C. Results of 2 independent tests of survival of rde-1(ne219);wrt-2p::RDE-1 worms treated with RNAi against sta-1, sta-2, akir-1 or ceh-18, infected with D. coniospora and cultured at 25˚C (n>65 for all tests). The difference between the sta-1(RNAi) and ceh-18(RNAi) animals is highly significant in both trials (p<0.0001; one-sided log rank test).
Another transcription factor, LIN-40, a GATA protein, NuRD complex component and one of two C. elegans homologs of human metastasis-associated protein MTA1, was the top hit among AKIR-1's binding partners. Recent genome-wide ChIP-seq data from the "model organism encyclopedia of regulatory networks" project (via www.encodeproject.org), revealed the presence of LIN-40 at the nlp-29 promoter (binding site peak, V: 3984375) in DNA from uninfected young adult worms. This independent line of evidence supports the presence of a NuRD/AKIR-1 complex within this AMP gene cluster in the absence of infection. Consistent with our current understanding of its mechanism of action, we did not find STA-2 among the AKIR-1-interacting proteins. In the simplest model, a complex of AKIR-1, CEH-18 and the NuRD/MEC chromatin remodelling proteins is recruited to the nlp locus and opens it, but represses gene expression. Upon infection, the chromatin structure allows activated STA-2 access to the AMP gene promoters, and removal of the repressive NuRD/AKIR-1/CEH-18 complex permits gene expression. It is noteworthy that 3 of the 53 high-confidence AKIR-1 interactors are implicated in ubiquitin-mediated protein turn-over, and that in preliminary tests, in vitro ubiqutination activity could be detected within the purified AKIR-1 protein complex specifically after infection, not before.
Chromatin remodelling at the promoters of immune genes can prime them for enhanced activation [44]. Many AMP genes in C. elegans, as in other species, are arranged in genomic clusters [24]. AKIR-1-dependent modification of chromatin structure offers the possibility of coordinating a rapid increase in the expression of neighbouring AMP genes, potentially important when faced with a fast-growing pathogen like D. coniospora.
Akirin functions together with the SWI/SNF complex in other species. Although we excluded a role for the C. elegans SWI/SNF complex in nlp-29 expression, we did identify some SWI/SNF complex proteins, including SWSN-1, -3, -4 and -6, among the potential AKIR-1 binding partners. These were found through an unbiased whole-organism approach; it is likely that we sampled separate complexes from different tissues. Indeed AKIR-1 is known to be expressed widely and also to have essential functions in development; it is necessary for synaptonemal complex (SC) disassembly during meiosis [21]. These different candidates therefore merit investigation in the context of AKIR-1's other functions. It would also clearly be of interest to attempt to recover AKIR-1 interactors specifically from the epidermis, but this is a technical feat beyond the current study.
SC disassembly involves a conserved RAS/ERK (Extracellular signal-regulated kinase) MAPK cascade. Interestingly, the same pathway is required for the response of C. elegans to infection by the Gram-positive bacterium Microbacterium nematophilum [45]. Within the rectal epithelium, it cooperates with a Gαq signalling pathway to trigger changes in cell morphology. At the same time, in motor neurons, Gαq functions independently of RAS signalling to influence nematode behaviour in the presence of M. nematophilum [46]. Following infection, it also acts in the pharynx to regulate, non-cell autonomously, defence gene expression in the intestine [47]. These instances illustrate how the physiological response to infection is a mélange of interconnected signal transduction cascades. Further studies will be required to establish whether akir-1 is required for any or all of these processes.
Across species, MAPKs act as regulators of chromatin structure. In yeast, the p38-related MAPK Hog1 physically interacts with the RSC chromatin-remodelling complex. This association is increased upon osmotic stress and is thought to direct the complex to bind osmoresponsive genes, changing nucleosome structure, increasing RNA polymerase II binding and causing a burst of transcription [48]. In vertebrates, the SWI/SNF subunit BAF60 can be phosphorylated by p38 MAPK, also targeting it to specific loci [49]. It is not yet clear whether PMK-1 directly phosphorylates AKIR-1 or NuRD/MEC complex proteins; it was not found as a physical interactor of AKIR-1. The p38 MAPK PMK-3 on the other hand was. RNAi of pmk-3 does not inhibit nlp-29 expression [11]. Notably, pmk-3 does participate in adult axon regeneration, in a p38 pathway that while sharing some elements with the epidermal innate immune pathway [50] is clearly distinct. Our results therefore raise the possibility that AKIR-1 plays a role in axon regeneration, in association with PMK-3.
We established that the SWI/SNF complex does not play a major part in modulating AMP gene expression in the epidermis. Rather the NuRD and MEC complexes, in a physical complex with AKIR-1 and CEH-18 play an essential role. One possible cause of this evolutionary re-wiring of a regulatory circuit could be the loss of NF-κB from nematodes, which has also led to a restructuring of the TLR pathway [51]. The precise evolutionary trajectories that led to these changes can only be the subject of speculation, but these lineage-specific adaptations likely reflect the extreme selective pressure that is exerted by pathogens. This plasticity is even more remarkable when one considers the essential developmental processes that many of these factors are involved in, limiting the degree of change that can be tolerated. In conclusion, as well as substantially advancing our understanding of immune defences in C. elegans, our results illustrate how an organism can evolve novel molecular mechanisms to fight infection while conserving an overall regulatory logic.

Killing and longevity assays
For the experiments reported in Figs 3 and S2C, 50-70 worms at the L1 stage were cultured on the appropriate RNAi bacterial clone at 25˚C, and then (for Fig 3) infected at the young adult stage for 1h with D. coniospora and transferred to fresh RNAi plates and cultured at 15˚C (to accentuate differences in survival [61]), or transferred directly to fresh RNAi plates and cultured at 20˚C (for S2C Fig). In both cases, the surviving worms were counted every day as described elsewhere [62]. For the experiments reported in Fig 7, worms at the L1 stage were cultured on the appropriate RNAi bacterial clone at 25˚C for 32 hours and then infected with D. coniospora overnight. Groups of 20-30 worms were then transferred to wells in 12-well plates (3 wells per condition), and images of each well collected automatically at regular intervals (roughly every 20 minutes) using a custom system that will be described elsewhere. The images were then examined, and worms scored as dead when they no longer showed sign of any movement between images. Statistical analyses used one-sided log rank test within Prism (Graphpad software).

Analyses with the Biosort worm sorter
Expression of nlp-29p::gfp reporter was quantified with the COPAS Biosort (Union Biometrica). Generally, a minimum of 80 synchronized worms were analyzed for size (TOF), extinction (EXT), green (GFP) and red (dsRed) fluorescence. The ratio Green/TOF was then calculated to normalize the fluorescence. When only mean values for ratios are presented, the values for the different samples within a single experiment are normalized so that the control worms (WT) had a ratio of 1. As discussed more extensively elsewhere [38], standard deviations are not an appropriate parameter and are not shown on figures with the Biosort. The results shown are representative of at least 3 independent experiments.

RNA preparation and quantitative RT -PCR
RNA preparation and quantitative RT-PCR were done as described [24]. Results were normalized to those of act-1 and were analyzed by the cycling threshold method. Control and experimental conditions were tested in the same 'run'. Each sample was normalized to its own act-1 control to take into account age-specific changes in gene expression.

Liquid chromatography tandem mass spectrometry
Peptides were separated in an in-house packed analytical column (inner diameter: 75 μm; ReproSil-Pur C18-AQ 3-μm resin, Dr. Maisch GmbH) by online nanoflow reversed phase chromatography through an 8-50% gradient of acetonitrile with 0.1% formic acid (120 min). The eluted peptides were sprayed directly by electrospray ionization into a Q Exactive Plus Orbitrap mass spectrometer (Thermo Scientific). Mass spectrometry measurement was carried out in data-dependent acquisition mode using a top10 sensitive method with one full scan (resolution: 70,000, target value: 3 × 10 6 ) followed by 10 fragmentation scans via higher energy collision dissociation (HCD; resolution: 35,000, target value: 5 × 10 5 , maximum injection time: 120 ms, isolation window: 4.0 m/z). Precursor ions of unassigned or +1 charge state were rejected for fragmentation scans. Dynamic exclusion time was set to 30 s.

Mass spectrometry data analysis
Raw data files were processed by MaxQuant software package (version 1.5.5.0) [66] using Andromeda search engine [67]. Spectral data were searched against a target-decoy database consisting of the forward and reverse sequences of WormPep release WS254 (28,071 entries), UniProt E. coli K-12 proteome release 2016_02 (4,314 entries) and a list of 245 common contaminants. Trypsin/P specificity was selected. Carbamidomethylation of cysteine was chosen as fixed modification. Oxidation of methionine and acetylation of the protein N-terminus were set as variable modifications. A maximum of 2 missed cleavages were allowed. The minimum peptide length was set to be 7 amino acids. At least one unique peptide was required for each protein group. False discovery rate (FDR) was set to 1% for both peptide and protein identifications.
Protein quantification was performed using the LFQ label-free quantification algorithm [68]. Minimum LFQ ratio count was set to one. Both the unique and razor peptides were used for protein quantification. The "match between runs" option was used for transferring identifications between measurement runs allowing a maximal retention time window of 0.7 min. All raw mass spectrometry data have been deposited in the PRIDE repository with the dataset identifier PXD008074.
Statistical data analysis was performed using R statistical software. Only proteins quantified in at least two out of the three GFP pull-down replicates (or two out of two GFP pull-downs for the experiment using infected worms) were included in the analysis. LFQ intensities were log2-transformed. Imputation for missing values was performed for each pull-down replicate in Perseus [69] software (version 1.5.5.0) using a normal distribution to simulate low intensity values below the noise level (width = 0.3; shift = 1.8). The LFQ abundance ratio was then calculated for each protein between the GFP pull-downs and the controls. Significance of the enrichment was measured by an independent-sample Student's t test assuming equal variances. Specific interaction partners were then determined in a volcano plot where a combined threshold (hyperbolic curve) was set based on a modified t-statistic (t(SAM, significance analysis of microarrays); s 0 = 1.5, t 0 = 0.9 * 1.5) [70,71]. Proteins cross-reactive to the anti-GFP antibody were identified by a pull-down experiment using the non-transgenic rde-1 strain and were filtered out from the AKIR-1 protein interactor dataset.

Western blot analysis
Samples for western blot analysis were either prepared as per the co-precipitation protocol with the final elution performed in 50 μl 200 mM glycine pH 2.6 and immediately neutralisation by addition of 0.2 M Tris pH 10.4, or as per the immunoprecipitation protocol. Samples were then resolved on a 4-12% BisTris Gel (Invitrogen) and subjected to transfer to a membrane.

Chromatin immunoprecipitation
For extract preparations, N2 worms were grown on rich NGM seeded with HT115 bacteria, and young adult populations of worms were used to prepare about 3-4 gr of flash frozen worm popcorn. Worms were then fixed first with 1.5 mM EGS (ethylene glycol bis) for 20 min and then in 1.1% formaldehyde, with protease and phosphatase inhibitors, at room temperature with shaking, for 20 min. The fixing reaction was quenched by addition of glycine to a final concentration of 125 mM. Worms were then washed once with 10 ml FA buffer (50 mM HEPES/KOH (pH 7.5), 1 mM EDTA, 1% Triton X-100, 0.1% sodium deoxycholate, 150 mM NaCl) with protease inhibitors (Pierce), resuspended in FA buffer containing 0.1% sarkosyl and protease and phosphatase inhibitors, then dounce-homogenized on ice. Well-resuspended mixtures were then sonicated to shear chromatin (size rage 300-800 bp) using 12 cycles (30' on, 30' off) in a Bioruptor-Pico (Diagenode). Cellular debris was removed by centrifugation at 17,000 g for 15 min at 4˚C. Immunoprecipitation reactions contained approximately 3 mg of total protein, with 1% sarkosyl. Before addition of the antibody (NanoTrap-GFP, Chromotek), 5% of the material was taken as input. Immunocomplexes were collected and washed with 1 ml of the following buffers: FA buffer, two washes, 5 min each; FA buffer with 1 M NaCl, 5 min; FA with 500 mM NaCl, 10 min; TEL buffer (0.25 M LiCl, 1% NP-40, 1% sodium deoxycholate, 1 mM EDTA, 10 mM Tris-HCl, pH 8.0), 10 min, and TE (pH 8.0), two washes, 5 min each. Complexes were eluted in 1% SDS in TE with 250 mM NaCl at 65˚C for 30 min. Samples and inputs were treated with Proteinase K for 1 h, and cross-links were reversed at 65˚C overnight. DNA was purified with Qiagen PCR purification columns. Locus-specific ChIP qPCR reactions (SYBR Premix ExTaq II, TaKara) were done for each immunoprecipitation using specific elution (ChIP), negative control elution (nonspecific) and input samples, following a 50-fold dilution. Ct values were used to calculate the fold difference in DNA concentration between ChIP and nonspecific samples, normalized to the input.

Microscopy
Confocal images were captured using an inverted confocal spinning disk microscope (Yokogawa, Visitron Systems GmbH) associated with a 512 x 512 pixels EM-CCD camera (Hamamatsu). Worms were immobilized in 0.01% levamisole and visualized through a CFI Plan Fluor Nikon 40X oil, 1.3 NA objective and 1.5X lens, using a 488 nm laser. Z-stacks were acquired with a step size of 0.3 μm.
Supporting information S1 Fig. Akirin controls the expression of nlp-29, an AMP encoding gene. A. Quantification in arbitrary, but constant units of relative size (Time Of Flight; TOF; grey bars), optical density (Extinction; Ext; orange bars) and col-12p::dsRed expression (red bars) of wild type worms carrying the integrated array frIs7 (which contains the fluorescent reporter transgenes nlp-29p::gfp and col-12p::DsRed) treated with RNAi against sta-1, dcar-1 and akir-1. In all cases there are no significant differences between control and experimental values (paired two-sided student t test). B. Ratio of green fluorescence (GFP) to size (TOF) of wild type and akir-1(gk528) worms carrying frIs7 and assessed without further treatment (control) or after infection by D. coniospora. Data are representative of three independent experiments. C. Comparisons of growth (left panel), DsRed expression (middle panel) and optical density (right panel) between wild type and akir-1(gk528) worms carrying frIs7 on 3 successive days after hatching. Data are representative of three independent experiments. A minimum of 50 worms was analysed for each condition. D. Quantification of D. coniospora spore adhesion at the level of the nose and the vulva in wild type worms carrying the frIs7 array and treated with RNAi against sta-1, dcar-1 and akir-1. (PDF) S2 Fig. Akirin controls nlp-29 expression in the epidermis and longevity. A. Abundance of mRNA for genes in the nlp-29 cluster in rde-1(ne219); wrt-2p::RDE-1 worms treated with RNAi against sta-1, dcar-1 or akir-1, presented as the difference in cycling threshold (ΔCt) between each nlp gene and act-1. Data are from three independent experiments (average and SD). B. Fold induction of expression for nlp genes in each of 3 experiments in rde-1(ne219); wrt-2p::RDE-1 worms treated with RNAi against the indicated genes and infected by D. coniospora; results are presented relative to those of uninfected worms. The average values are represented in Fig 2C. C. Survival of rde-1(ne219);wrt-2p::RDE-1 worms treated with RNAi against sta-1 (n = 50) or akir-1 (n = 50). The difference between the sta-1(RNAi) and akir-1(RNAi) animals is highly significant (p<0.0001; one-sided log rank test). Data are representative of three independent experiments. (PDF) S3 Fig. nlp-29 expression is independent of the SWI/SNF nucleosome remodelling complex. Ratio of green fluorescence (GFP) to size (TOF) (A) and quantification in arbitrary, but constant units of relative size (TOF; grey bars), col-12p::dsRed expression (red bars) and optical density (Ext; orange bars) in wild-type worms carrying frIs7 and treated with RNAi against different genes (B, C). The genes corresponding to core (swsn-1, swsn-4, swsn-5), accessory (swsn-2.1, swsn-3, dpff-1), PBAF/PBAP (swsn-7 and pbrm-1) and BAF/PBAF (let-526 and swsn-9) elements of the SWI/SNF nucleosome remodelling complex are indicated. Populations of >100 worms were analysed for each condition. Data are representative of three independent experiments. D. Quantification of D. coniospora spore adhesion at the level of the nose and the vulva in wild type worms carrying frIs7 and treated with RNAi against the indicated genes.  Table. Comparison of protein-protein interactors for AKIR-1 and SWSN-2.2. Data was from [34] and S1 Table. The annotations come from Wormbase (WS257). Gene identifiers were made uniform using Wormbase Converter [37]. (XLSX) S3 Table. Raw data for figures. The numerical data underlying the different figures is presented in individual sheets, arranged by category (Biosort, survival and qRT-PCR). (XLSX)