Drosophila EYA Regulates the Immune Response against DNA through an Evolutionarily Conserved Threonine Phosphatase Motif

Innate immune responses against DNA are essential to counter both pathogen infections and tissue damages. Mammalian EYAs were recently shown to play a role in regulating the innate immune responses against DNA. Here, we demonstrate that the unique Drosophila eya gene is also involved in the response specific to DNA. Haploinsufficiency of eya in mutants deficient for lysosomal DNase activity (DNaseII) reduces antimicrobial peptide gene expression, a hallmark for immune responses in flies. Like the mammalian orthologues, Drosophila EYA features a N-terminal threonine and C-terminal tyrosine phosphatase domain. Through the generation of a series of mutant EYA fly strains, we show that the threonine phosphatase domain, but not the tyrosine phosphatase domain, is responsible for the innate immune response against DNA. A similar role for the threonine phosphatase domain in mammalian EYA4 had been surmised on the basis of in vitro studies. Furthermore EYA associates with IKKβ and full-length RELISH, and the induction of the IMD pathway-dependent antimicrobial peptide gene is independent of SO. Our data provide the first in vivo demonstration for the immune function of EYA and point to their conserved immune function in response to endogenous DNA, throughout evolution.


Introduction
In humans, innate inflammation is frequently linked to various types of diseases, namely to autoimmune diseases [1]. However, it is still unclear in many instances how inflammation is initiated and chronically maintained. Pathogen invasions and/or subsequent changes of cellular integrity can cause inflammation and DNA has been reported to be a strong immune stimulator under such conditions [2,3]. How cells recognize, and respond to, DNA are among the challenging questions in the field of inflammation, particularly in humans.
DNaseII is an evolutionarily conserved acid DNase localized in lysosomes [14]. Macrophages in the fetal liver and thymus of DNaseII 2/2 mice cannot digest DNA of engulfed dead cells, or DNA of nuclei expelled from erythroid precursors [15,16]. This results in their production of pro-inflammatory cytokines such as interferon-b (IFN-b) and tumor necrosis factor-a (TNF-a) [17,18]. As a result of excessive cytokine production, DNaseII 2/2 mice suffer from anemia during embryonic stages and later, as adults, from polyarthritis. Crosses of DNaseII 2/2 mice with mice deficient for various immune response-related signaling molecules conclusively showed that (1) the DNA-dependent IFN-b gene expression is TLR-independent and IRF-3/7-dependent and that (2) the DNA-dependent TNF-a gene expression is both independent of TLR-and IRF-3/7-signaling [19,20].
To elucidate the molecular mechanisms of DNA sensing in DNaseII 2/2 mice, Okabe et al. recently performed expression cloning and reported that Eyes absent 4 (EYA4) enhances the innate immune responses against DNA by activating IRF3 and NF-kB [21]. EYA was originally identified in Drosophila as a transcription factor that is essential for eye development [22]. Mammals have four EYA paralogs, EYA1-4. Mutations in human EYA1 cause branchio-oto-renal (BOR) syndrome, an autosomal dominant genetic disorder, affecting necks, ears, and kidneys [23,24,25]. EYA1 2/2 mice also show renal abnormalities and a conductive hearing loss similar to BOR syndrome [26]. EYA3 2/2 mice have a minor defect in locomotion with some deficits of respiratory, muscle and heart functions [27], and EYA4 2/2 mice show a defect in eustachian tube and middle ear [28]. EYA2 2/2 mice have not yet been generated. Several groups recently showed that EYA proteins carry C-terminally an evolutionarily conserved domain with tyrosine phosphatase activity [29,30]. The detailed biochemical analysis by Okabe et al. [21]. Further identified a Nterminal threonine phosphatase domain in mouse EYA4, and showed that this domain is responsible for the innate immune responses observed against DNA.
Drosophila has a single orthologue of mammalian DNaseII. Interestingly, Mukae et al. found that flies carrying a hypomorphic mutation in the DNaseII gene (DNaseII lo ) constitutively express the anti-bacterial peptides Attacin A and Diptericin, but not the antifungal peptide Drosomycin [31]. Transcription of antimicrobial peptides in Drosophila is essentially under the control of two regulatory pathways: the IMD pathway (largely similar to the TNF/TNF-R pathway in mammals) controls the transcriptions of Attacin A and Diptericin, whereas the TOLL pathway regulates the transcription of Drosomycin [32]. The observations by Mukae et al. therefore suggest that excess DNA in the DNaseII deficient flies could activate the IMD pathway, but not the TOLL pathway.
Induction of immune response to endogenous DNA had not been established before the study of Mukae et al. in 2002 [31] and few studies have been devoted to this aspect of the innate immune response since. We decided to utilize Drosophila DNaseII deficient flies as a model to study the immune response against DNA. By taking Attacin A induction as a read-out, we first have addressed here the function of the eya gene of Drosophila. The fact that Drosophila has a single eya gene, would allow for precise in vivo analysis of the function of this gene in the immune response. We now show that EYA indeed plays a significant role in the expression of the IMD pathway-dependent antimicrobial peptide Attacin A in DNaseII-deficient flies. Interestingly, EYA is not involved in this Attacin A induction by Gram-negative bacteria as most potent stimulus in this pathway activation. In addition to the tyrosine-phosphatase activity, Drosophila EYA also carries a threonine-phosphatase activity at its N-terminus, as its mammalian counterparts. We demonstrate that this threonine phosphatase domain of Drosophila EYA is responsible for the innate immune responses against DNA in the DNaseII deficient model. In contrast, our results show that the tyrosine phosphatase domain governs eye development. We further reveal that EYA is associated with IKKb and RELISH and AttacinA induction is independent of SO. Taken together with the Mukae et al and the Okabe et al studies in mice, our results point to a striking conservation of innate immune responses against endogenous DNA between Drosophila and mammals.

Results
DNaseII deficiency-dependent expression of Attacin A in eya mutant background We have first made use of three well-studied alleles of eya, which carry a 1.5 kb (eya 1 ) or a 322 bp (eya 2 ) deletion upstream of the transcription start site, and a nonsense loss-of-function mutation (eya cli-IID ) that causes truncation of the EYA protein at amino acid 335 [33]. Eya 1 or eya 2 flies develop to adults deprived of eyes. Eya cli-IID flies are lethal at the embryonic stage. The DNaseII lo allele has a missense G-to-A mutation at nucleotide 668, converting a Serine residue to an Aspargine residue at amino acid position 223, which results in a reduction of the acid DNase activity [33]. DNaseII mutants develop normally to the adult stage. Interestingly, these flies constitutively express the antimicrobial peptide Attacin A [31]. We generated eya 1 ;DNaseII lo , eya 2 ;DNaseII lo , and eya cli-IID ;DNaseII lo double mutant flies and monitored Attacin A expression. We noted that flies carrying homozygous eya and DNaseII lo double mutations died at the pupal stage and we therefore used adult eya/+;DNaseII lo flies to evaluate the effect of the eya gene in this context. As shown in Fig. 1A, DNaseII mutant flies constitutively expressed Attacin A, as previously reported [31]. In contrast, a heterozygous eya 1 mutation combined with DNaseII deficiency significantly reduced Attacin A mRNA levels. Similar results were obtained with the eya 2 and eya cli-IID alleles (Fig. 1A). To exclude possible effects of genetic background and/or aberrant development due to the gene deficiency/insufficiency, we generated an inducible eya/DNaseII knockdown system by RNA interference. In these experiments, we used established fly lines carrying dsRNAs for eya or/and DNaseII under a UAS promoter [34] and crossed them with a hsp-GAL4/tub-Gal80 ts fly line [35]. After heat shock treatment, we measured the Attacin A mRNA levels on day 8. As shown in Fig. 1B, heat shock itself did not induce the Attacin A gene in the wild-type and GAL4 driver lines. The knockdown of the DNaseII gene, but not that of the eya gene, strongly activated Attacin A: Interestingly the knockdown of eya resulted in a strong reduction of the DNaseII knockdown-induced Attacin A expression. We observed a very low level of expression of Drosomycin in the Oregon-R strain, and this expression remained unchanged in knockdowns of eya and/or of DNaseII (Fig. 1C). These results provide an in vivo demonstration that Drosophila eya is involved in the immune response to DNA. Attacin A can be induced by Gram-negative bacterial challenge through the Drosophila IMD pathway. We investigated the role of EYA in the immune response during bacterial infection. As Fig. 1D shows, the Attacin A mRNA level was increased by bacterial challenge. However, its level was not altered in eya 2 mutants. This result indicates that EYA plays a role in the immune response specifically against DNA. It is important to note that the genetic background affects the basal mRNA levels of Attacin A. However, Attacin A induction by DNaseII deficiency is consistent in both cases, mutant and RNAi transgenic flies. Two additional observations further support this conclusion: (1) cn bw-background DNaseII mutants established by backcrossing also show constitutive expression of Attacin A and (2) the rescue experiment showed that the Attacin A induction is dependent on DNaseII deficiency (X.L. and H.F., unpublished results).
In vitro functional analysis of the threonine phosphatase domain in recombinant Drosophila EYA mutant proteins As mentioned in the Introduction, the four mammalian EYAs carry a C-terminal tyrosine phosphatase domain, also called EYA domain [36], which is conserved in flies. In addition, the mammalian and fly EYA proteins have a well-conserved N terminal threonine phosphatase domain with six tyrosine residues [21].
To characterize the Drosophila EYA threonine phosphatase activity in the present context, we expressed in mammalian 293 T cells a series of recombinant Drosophila EYA proteins: (1) Q335*, a nonsense mutation that causes truncation of the EYA protein at amino acid 335 and corresponds to the eya cli-IID loss-of-function allele [37] in the literature; this protein carries neither the tyrosine nor the threonine phosphatase motifs; (2) T497M, a missense mutation of the EYA domain ( Fig. 2A), known as the eya E11 allele [37]; (3) D497N; this allele was experimentally generated for recombinant protein EYA4 by a D493N replacement, which abolishes the tyrosine phosphatase activity [21,29,30]; (4) Y4 is an experimentally introduced allele, in which four tyrosine residues have been replaced in the threonine phosphatase domain, resulting in the total loss of the threonine phosphatase activity [21]; (5) WT refers to the wild-type EYA. After expression in 293 T cells, the recombinant proteins were purified with anti-Flag mAb and served to determine in vitro threonine-phosphatase activities using a phospho-threonine peptide. The results were as follows (Fig. 2B): The threonine-phosphatase activities of D493N and T497M were similar to that of WT. In contrast the threoninephosphatase activities of the Y4 and that of Q335* were severely diminished. These results demonstrate that Drosophila EYA has a threonine phosphatase activity governed by the six-tyrosine-residues motif, and that this motif is not only evolutionarily but also functionally conserved from insects to mammals.

In vivo analysis of the role of the EYA threoninephosphatase domain in Attacin A expression
With this information at hand, we next examined which of the threonine versus tyrosine phosphatase domains of Drosophila EYA is responsible for the innate immune responses against DNA. We generated, through bacteriophage WC31 integrase-mediated transgenesis [38,39], fly lines expressing the various alleles of Drosophila EYA described above, i.e., WT, Q335*, T497M, D493N, and Y4. Expression was driven by a heat-shock promoter [22]. In brief, this transgenesis technique integrates a transgene (in our case, various forms of eya cDNAs) at the same specific site in the genome, allowing to make quantitative comparisons between different forms of EYA. We then crossed these eya transgenic flies to the homozygous eya 2 ;DNaseII lo strains. As mentioned above, Figure 1. EYA is involved in innate immune responses against DNA. mRNA levels were determined for antimicrobial peptide genes by quantitative RT-PCR and normalized to Rp49 expression(also known as RpL32). The relative values are indicated against Oregon-R as wild-type control. (A) Attacin A expression for DNaseII, eya 1 /+;DNaseII, eya 2 /+;DNaseII, and eya cli-IID /+;DNaseII mutant flies. One-way ANOVA was performed and followed by Dunnett's multiple comparison test. * indicates statistically significance (p,0.0001) by comparing to Oregon-R. (B) Attacin A expression for single eya, single DNaseII, and double DNaseII and eya knockdown flies at 8 days after heat-shock treatment. dsRNAs were expressed using the GAL4-UAS system. The GAL4 driver line alone was included. Two-tailed unpaired Student's t-test was performed for statistic analysis. (C) Drosomycin expression using the same sets of flies as for (B). (D) Attacin A expression by E.coli challenge for w 1118 as wild-type control and eya 2 mutant flies. The value represents the average and standard deviation of 3-6 independent experiments. A pool of 5-20 adult flies per genotype was collected for each experiment. doi:10.1371/journal.pone.0042725.g001 homozygous eya 2 ;DNaseII lo flies show nearly full lethality; in the mutant series we observed some degree of rescue of the lethality with the WT eya transgenes, but barely with the other transgenes. We therefore decided to continue our experiments in an eya 2 heterozygous and DNaseII lo homozygous background. Under these conditions most of the flies survived. In WT flies, we observed a high level of Attacin A induction. When we compared this level to the transgenic line carrying the dual phosphatase inactive form, Q335*, we noted a dramatically lowered level of expression. Flies carrying a transgene in which only the tyrosine phosphatase function was lost (D493N and T497M) showed no significant difference with those of WT flies. Finally and most interestingly in the present context, flies carrying the Y4 construct, i.e. with an inactive threonine phosphatase domain and a wild type tyrosine phosphatase domain, showed the same level of expression as the forms with dual inactive phosphatase domains. Since the different effects of eya transgenes on the Attacin A expression could not be explained by the expression levels of the transgenes (Fig. S1, the same set of RNA in Fig. 3B was used), our results clearly point to the threonine phosphatase domain as the domain up-regulating the level of expression of Attacin A induced in the DNaseII deficient background.
The threonine phosphatase domain is not required for eye development We next investigated the impact of the two different phosphatase activities of EYA on eye development. Although previous studies indicated that the tyrosine-phosphatase activity is required for eye development [29,40], the possibility that the threonine phosphatase activity also affects eye development could not be excluded. As shown in Fig. 4, the eyes did not develop in flies carrying the eya 2 mutation, as reported previously [22]. The inducible expression of the WT and Y4 forms of EYA at the stages from egg to pupae rescued the eyes absent phenotype. In contrast, the expression of the EYA forms of Q355*, D493N, and T497M could not rescue the phenotype. These results indicate that the tyrosine phosphatase activity of EYA is required for eye development, but that its threonine-phosphatase is dispensable.

EYA associates with IKKb and RELISH and AttacinA induction is independent of transcription factor SO
To obtain the mechanistic view of a link between EYA and the IMD pathway, we performed protein-protein association studies. Since the threonine phosphatase activity of EYA plays an important role for immune responses in DNaseII deficiency model, we reasoned to investigate two potential phospho-substrates of the IMD pathway: IKKb and RELISH [41,42]. When FLAG-MYCtagged EYA and hemagglutinin (HA)-tagged IKKb or RELISH were expressed in S2 cells, IKKb and RELISH associated with EYA (Fig. 5A).
On the other hand, EYA binds to the homeobox transcription factor Sine Oculis (SO) [43]. To determine whether the AttacinA can be a target of the So-Eya complex, we co-transfected eya and so in S2 cells with reporter plasmid ARE-luciferase [44]. As previously reported [44,45], co-transfection of eya and so activated ARE-luciferase 50-fold over the reporter alone (Fig. 5B). In this   condition, AttacinA induction was not observed while heat-killed E.coli challenge induced AttacinA by more than 200-fold (Fig. 5C). These results indicate that AttacinA is not a target of the complex of EYA and SO.

Discussion
We are interested in understanding how endogenous ligands can induce immune responses in Drosophila and whether the receptors and downstream signaling cascades are similar to those which are activated upon well-defined microbial stimuli (bacteria, fungi, virus) [46,47]. The discoveries that mice and flies deficient for lysosomal DNase activities mount an immune response, as evidenced by the constitutive expression of IFN-b or of the antibacterial peptide Attacin A respectively, was of great interest in this context. The recent report by Okabe et al. that this innate response to undigested DNA is regulated in mice by the eya gene(initially discovered in Drosophila eye development) stimulated our interest in the potential role of this gene in the immune response of Drosophila. We provide four essential findings: (1) the immune response induced by undigested DNA in DNaseII deficient flies requires the eya gene; (2) the N-terminal threonine phosphatase domain of the EYA protein is responsible for this function, whereas that of the C-terminal tyrosine-phosphatase domain is dispensable; (3) EYA associates with IKKb and RELISH and the So-Eya complex does not induce Attacin A; (4) the role of EYA proteins is conserved in this specific immune context between flies and mammals. We present the first in vivo demonstration for the role of the threonine-phosphatase domain of EYA proteins, which were so far surmised only on the basis of the in vitro studies in mice.
The two main questions, unanswered to date, pertain to (1) the identity of the DNA sensor (receptor) in flies; (2) the target molecule for the threonine-phosphatase activity of EYA.
As regards the first question, we have no firm data regarding DNA sensors in flies at present. In mammals, several molecules, play more or less well defined roles in DNA recognition, namely TLR9, AIM2, DAI, for which there are no homologues in flies. Identifying the DNA sensor in Drosophila is clearly a priority in the field. Of note, the sensor for DNA that accumulates in macrophages in DNaseII deficient mice has not yet been firmly identified. TLR9, which would appear as a good candidate, is not involved, as in DNaseII 2/2 TLR9 2/2 mice the innate immune response to accumulated DNA is unaffected.
How does EYA activate the IMD pathway to control expression of the Attacin A gene? Our transactivation and transcription assay indicates that AttacinA induction is not regulated by the So-Eya complex. Furthermore our protein-protein association studies suggest the link between EYA and the IMD pathway at the level of IKKb and RELISH. It is noteworthy that EYA can associate with full-length RELISH. In general, the full-length RELISH is located in the cytoplasm and activated via two events: (1) phosphorylation by IKKb and (2) cleavage by DREDD (similar to Caspase8/10), and eventually truncated N-terminal half of RELISH translocates to the nucleus to regulate transcriptions of target genes such as antimicrobial peptide genes [41,48]. We propose that EYA can make a complex with IKKb and RELISH in the cytoplasm, and activate the IMD pathway at this level. Interestingly, Eya 2 mutants can respond to Gram-negative bacteria. These observations clearly exclude possibilities that EYA directly involves the phosphorylation of Serine 528 and Serine 529 of RELISH, which are targets of IKKb and required for target gene expression. It is interesting to investigate the role of EYA on the phosphorylation status of Serines/Threonines in 107-aa C-terminal region in RELISH, which is also targets of IKKb and required for the interaction between RELISH and IKKb [41]. Of interest in the present context is the observation that both Drosophila and mammalian EYAs have two MAPK phosphorylation sites and that the Drosophila ERK and p38 MAPKs can phosphorylate Drosophila EYA in vitro [49]. Recently Morilo et al. demonstrated that NMO phosphorylates EYA and potentiates the transactivation function to enhance transcription of So-Eya target genes during eye specification and development [50]. Further, we now know that recombinant mouse EYA4 proteins produced in 293 T cells are phosphorylated (T.S., and S.N., unpublished results). The precise mechanism of dephosphorylation of target protein by EYA needs to be elucidated [51,52].
In the mammalian system, EYA4 has been reported to be recruited by the dsRNA homologue poly (I:C) to the IPS-1 complex to activate the IRF3 and NF-kB pathways. This complex consists of various regulating molecules, namely RIG-I, STING, and NLRX1 [21]. No clear-cut homologue of any of these molecules were found in Drosophila till now (although RIG-I and Drosophila Dicer-2 share a helicase domain with the significant amino acid sequence similarity [53]). The components of the complex responsible for DNA-sensing remain to be revealed. EYA is the first identified molecule that is found in both insects and mammals in DNA sensing cascade.
Lastly, a surprising result of this study was that while both DNaseIIand eya 1 -or eya 2 -deficient flies develop normally, combining these two deficiencies is lethal at the pupal stage. The elucidation of this developmental arrest in eya 1 ;DNaseII lo or eya 2 ;DNaseII lo flies will hopefully shed light on DNA sensing and signal transduction in flies.
We anticipate that comparative studies on DNA sensing using Drosophila and mouse DNaseII deficient animals will facilitate the understanding of the molecular mechanisms of DNA-triggered innate inflammation.

Fly strains and maintenance
All flies were maintained with standard corn meal and yeast extract medium at 25uC at a light cycle-and humidity controlledroom. Oregon-R and w 1118 were used as wild type controls. Flies used for AMP expression experiment were raised in antibiotics cocktail medium (100 mg/ml ampicillin, 50 mg/ml vancomycin, 100 mg/ml neomycin and 100 mg/ml metronidazole) [54] to reduce the risk of contaminating bacteria infection that could induce antimicrobial peptide gene expression [47] (Fig. S2). DNaseII[lo] (FBal0002709), In(2L)eya, eya [1] (FBst0003631) and eya [2](FBal0030759) mutant flies were obtained from the Bloomington stock center. Eya[cli-IID] (FBal0001705) mutant flies were gifted from Dr.Nancy N. BONINI. RNAi transgenic for DNaseII (NIG#7780R-3) and for eya (FBst0465312) were purchased from the Fly stocks of National Institute of Genetics, Japan and the Vienna Drosophila RNAi Center, Austria, respectively. Inducible ubiquitous driver, hsp-GAL4/tub-GAL80 ts recombined fly line, was established by Dr.Ferrandon and described previously [35]. Heat shock-mediated GAL4 induction was performed as follows: all crossings with the hsp-GAL4/tub-GAL80 ts driver were made at 18uC, three days old hatched flies were inoculated at 29uC for 2 days, at 37uC for 30 minutes, at 18uC for 30 minutes, at 37uC for 30 minutes, and then flies were kept at 29uC. Note that the incubation at 37uC was processed in the water bath. For WC31-mediated transgenesis previously described [38,39], we utilized X-linked attB landing fly line VK6-WC31 (y [1] (Fig. 3A).

Plasmids and Antibodies
The cDNA of eya was amplified from FBcl0108545 (Drosophila Genomics Resource Center) by PCR, cloned into pCRH8/GW/ TOPOH TA cloning vector (Invitogen). The cDNAs of Ikkb (also called as ird5) and Rel were amplified using a cDNA library from S2 cells by PCR and cloned into pDONR207 vector (Invitogen). These pENTRY clones were verified by DNA sequencing and used for further plasmid constructions. Series of mutations were introduced by QuikChangeH Lightning Site-Directed Mutagenesis Kit (Agilent Technologies) and the sequences were confirmed by DNA sequencing. These mutations includes: Y4 [21], D493N [21], T497M [37], and Q335* [37]. Primers used in this report are listed in Table S1. Based on pCaSpeR-hs (FBmc0000179) vector, the GatewayH-based destination vector pCaSpeR-attB-hsp-FW was generated for transgenesis. This vector containing the attB site for WC31-mediated transgenesis, GatewayH cassette fused with three repeated Flag sequence obtained from pTFW vector (the Drosophila Genomics Resource Center) for amino-terminal tagging. After LR reactions between pCaSpeR-attB-hsp-FW and pCRH8-N-dEYA (WT, Y4, D493N, T497M, and Q335*), series of pCaSpeR-attB-hsp-dEYA vectors were generated, followed by transgenesis using X-linked VK6-WC31. For the production of recombinant Drosophila EYAs, series of eya cDNAs in pCRH8/ GW/TOPOH were transferred to mammalian expression vector pEF-BOS [55]. For biochemical analysis, GatewayH 2 based destination vector pMT-HW was generated based on pHHW (the Drosophila Genomics Resource Center) and pMT/V5-His-A vector (Invitrogen). The vector contains Metallothionein-inducible promoter and hemagglutinin (HA) tag at amino-terminal. pAFMW destination vector contains Actin 5C promoter and dual tags of FLAG and MYC at amino-terminal. By LR reactions with these two destination vectors, we generated pMT-HA-IKKb, pMT-HA-REL, and pAFMW-dEYA. For transactivation and transcription assay, pARE-luciferase, pRmHa3-Flag-SO and pRmHa3-Flag-EYA are kindly provided by Dr. Ilaria REBAY [44]. pACH110 containing Actin 5C promoter and LacZ gene was used for normalization of transfection efficiency. Anti-FLAGH M2 antibody (SIGMA-ALDRICH) for immunoprecipitation, Rabbit polyclonal anti-HA antibody (SIGMA-ALDRICH), anti-MYC antibody (BOEHRINGER INGELHEIM), and Horseradish Peroxidase (HRP)-conjugated anti-Rabbit or anti-mouse IgG antibody (SIGMA-ALDRICH) as secondary antibodies for immunoblotting were used in this study.
Protein expression, purification, and thereonine phosphatase activity assay The Drosophila EYA proteins were produced in 293 T cells and purified using anti-Flag M2 affinity gel (SIGMA-ALDRICH) as described previously [21]. Briefly, series of pEF-BOS-dEYAs were transfected by Ca-Phosphate co-precipitation method, and cells were lysed and the supernatants after centrifugation was subjected to the purification. To quantify the phosphatase activity, the purified recombinant EYA was incubated with the phosphorylated synthetic peptide KR(pT)IRR at the concentration of 700 mM at 37uC for 60 min in 50 mM Tricine-KOH buffer (pH 8.0) containing 5 mM EDTA and 50 mM DTT using 0.5 pmol of a recombinant protein. The quantity of released phosphate was measured by colorimetric method using the malachite greenmolybdate. The malachite green-ammonium molybdate phosphate complex was detected at 620 nm using a Micro Plate Reader (BioLumin 960).

RNA analysis
Total RNA was extracted from adult flies using RNeasy Mini kit (Qiagen) or NucleoSpinH RNAII kit (Macherey-Nagel). TaqManH RNA-to-CT TM 1-Step Kit (Applied Biosystems) was used for quantitative RT-PCR with TaqManH Gene Expression Assays primers and probes (Applied Biosystems) using 7500 Fast Real-Time PCR System (Applied Biosystems). These assays include Attacin A (Dm02362218_s1), Drosomycin (Dm01822006_s1), and RpL32(also known as Rp49, Dm02151827_g1). A pool of 5-20 flies was collected for each experiment. The expressions of the antimicrobial peptide genes were normalized to the expression of the RpL32 gene for each sample. Each assay was performed in duplicated manner and the average of duplicates was used for a single experiment data. For gene expression of eya in flies and AttacinA in S2 cells, total RNA was extracted from adult flies or cells using NucleoSpinH RNAII kit (Macherey-Nagel) or TRI REAGENTH RT(Molecular Research Center). Reverse transcription was performed by RevertAid TM H Minus Reverse Transcriptase (Fermentas). Fast SYBRH Green Master Mix (Applied Biosystems) was used for Quantitative RT-PCR with 7500 Fast Real-Time PCR System (Applied Biosystems). eya or AttacinA mRNA levels were quantified and normalized to mRNA level of Rp49. Both primers (Q-eya or Q-AttA primers, and Q-Rp49 primers) were listed in Table S1.

Bacteria challenge
Escherichia coli (1106C) at 37uC shaker (INNOVA-44R) were inoculated for 7 to 8 hours until OD 600 reached 0.6-0.8. The pellet was collected and washed twice with PBS. Finally Escherichia coli was suspended in PBS and OD 600 was adjusted to 2.0 for injection. We injected 13.8 nl to each female fly by NANOJECT II (Drummond Scientific) with customized capillary needles (Drummond scientific) crafted by flaming/brown micropipette puller model P-97 (Sutter Instrument). Six hours after injection flies were collected for RNA extraction to determine AMP expression.

Eye phenotype rescue experiment
Eggs were laid in culture media for 24 hours, and then processed the 1-hour heat shock at 37uC for every 8 hours using heat-blocks in the chamber until the white pupae stage. Hatched adult flies were counted for appearance of eye phenotype and frozen at 280uC. Frozen and dried flies were subjected to scanning electron microscope analysis using HITACHI TM-1000 Tabletop SEM (HITACHI).

S2 cell transactivation and transcription assay
Drosophila S2 cells were transiently transfected by calcium phosphate method with 2.5 mg each of pRmHa3-Flag-SO and pRmHa3-Flag-EYA, 10 mg of the reporter plasmid ARE-luciferase, and 1 mg of pACH110 for normalizing transfection efficiency. After 16-hour incubation, cells were washed with PBS and CuSO 4 was added to the culture medium at the final concentration of 1 mM for induction. Cells were harvested 24 h later and lysed in 70 mL reporter lysis buffer (Promega). Ten micro liter or 2.5 mL of lysates were used for luciferase assay or b-galactosidase assay, respectively. Both assays were performed in duplicate manner using Mithras LB940 multimode microplate reader (Berthold technologies). Three or four independent transfections were made for each experimental set.

Immunoprecipitation and Immunoblotting
Six million of Drosophila S2 cells were transiently transfected with 5 mg each of pAFMW-dEYA and pMT-HA-IKKb or pMT-HA-REL by calcium phosphate method. After 16-hour incubation, cells were washed with PBS and CuSO 4 was added into the cell culture medium at the final concentration of 0.5 mM for induction. Cells were harvested 24 h later and then lysed with 200 mL of lysis buffer (50 mM Tris-HCl pH7.5, 10% Glycerol, 1% Triton X-100, 150 mM NaCl, 100 mM NaF, 5 mM ZnCl 2 , 1 mM Na 3 VO 4 , 10 mM EGTA pH8.0, 1 mM EDTA, protease inhibitor cocktails (ROCHE)) on ice for 30 min. After centrifugation, 10 mL of lysates were used for determining total protein expression level, and the rest was subjected to immunoprecipitation followed by immunoblotting. Anti-FLAG antibody was pre-incubated with 10 uL of DynabeadsH protein G (Invitrogen) in lysis buffer for 1 h at room tempreture followed by incubation with lysates for 2 h at 4uC. After washing 4 times by lysis buffer, the immunoprecipitates were boiled in 15 mL of 26 elution buffer (30% Glycerol, 0.15 M Tris-HCl PH6.8, 5% Sodium dodecyl sulfate, 0.02% Bromophenol blue, 0.72 M 2-Mercaptoethanol). For immunoblotting, proteins of immunoprecipitates and total lysates were resolved in NovexH 4-20% Tris-Glycine Gels (Invitrogen) followed by blotted to nylon membrane. Then the blot was incubated with anti-HA antibody(1:1,000) or anti-Myc antibody(1:1,000) followed by with Horseradish Peroxidase (HRP)-conjugated secondary antibodies(1:15,000). The image was acquired by Fusion FX7 system (Vilber Lourmat) after incubated with Super SignalH West Dura substrates (Pierce). Three independent experiments were performed.

Statistics
Two-tailed Student's t-test and one-way ANOVA were used for statistic analysis using Prism software.