A Short Receptor Downregulates JAK/STAT Signalling to Control the Drosophila Cellular Immune Response

Regulation of JAK/STAT signalling by a short, nonsignalling receptor in Drosophila modulates response to specific immune challenges such as parasitoid infestations.


Introduction
The innate immune response-the synthesis of antimicrobial peptides and mobilisation of dedicated immune cells-confers a broad protection against pathogens to all multicellular organisms. Drosophila has become a model for studying the role of hematopoietic (blood) cells and the evolution of cellular immunity (reviews by [1,2]). Similar to vertebrates, Drosophila hematopoiesis occurs in two waves during development [3]. A first population of hemocytes is specified in the embryo and gives rise to plasmatocytes involved in phagocytosis and crystal cells required for melanisation [4]. A second wave of plasmatocyte and crystal cell production occurs at the end of larval development. Larval hematopoiesis can also give rise to a third cell type, the lamellocytes, which are devoted to the encapsulation of foreign bodies too large to be phagocytosed. Lamellocytes only differentiate in response to specific immune challenges such as parasitisation by wasps, a common threat for higher order insects [1,2,5,6]. Larval hematopoiesis takes place in a specialised organ, the lymph gland (LG), which forms during embryogenesis and grows during larval development. In third instar larvae, the LG is composed of several lobes with the anterior-most lobes organised into a medullary zone (MZ) containing the hematopoietic progenitors, a cortical zone (CZ) containing differentiating hemocytes, and a so-called posterior signalling centre (PSC) ( Figure 1F) [7]. These three zones can be identified by the expression of different markers. Differentiating crystal cells and plasmatocytes in the CZ express prophenoloxidase (proPO) and the P1 Nimrod receptor [8], respectively. Hematopoietic progenitors can be distinguished by their expression of domeless (dome), which encodes the Drosophila receptor of the janus tyrosine kinase/signal transducers and activators of transcription (JAK/STAT) pathway and tep4, which encodes a thioester protein [7,9]. PSC cells express the transcription factors Antennapedia (Antp) and Collier/Knot (Col), the Drosophila Early B-cell Factor (EBF) ortholog [10], and the morphogen Hedgehog (Hh) [7,[11][12][13]. The PSC controls the balance between multipotent prohemocytes present in the MZ and differentiating hemocytes [9,12]. It acts in a non-cell-autonomous manner, perhaps via Hh signalling, to maintain JAK/STAT signalling in prohemocytes, thus preserving the multipotent character necessary for these cells to adopt a lamellocyte fate upon parasitisation. This key role of the PSC revealed that, in Drosophila, larval hemocyte homeostasis is dependent upon interactions between hematopoietic progenitors and their micro-environment, a role reminiscent of the vertebrate hematopoietic niche [9,12,14].
JAKs and STATs mediate intracellular signalling in response to secreted type I cytokines [15]. In mammals, large families of cytokines and single-pass transmembrane receptors, named type I cytokine receptors, which signal as either homodimers or heteromers, have been identified (for review [16][17][18][19]). JAK kinases are anchored to the intracellular part of signalling receptors. Binding of the cytokine induces conformational changes in the latter that bring two JAKs in close proximity. This allows JAK trans-phosphorylation and phosphorylation of the receptor, thereby creating a docking site for STAT transcription factors.

Author Summary
A specific microenvironment termed the ''niche'' supports long term maintenance of hematopoietic stem cells in vertebrates. A small group of specialised cells called the posterior signalling center (PSC) controls hemocyte (blood cell) homeostasis in the Drosophila larval hematopoietic tissue and thus fulfills a similar function to the vertebrate niche. The PSC acts at a distance to maintain JAK/STAT signalling in hematopoietic progenitors (prohemocytes), thereby ensuring their multipotent character. We report here that a short cytokine receptor encoded by CG14225/ latran is required to extinguish JAK/STAT signalling in prohemocytes and thereby ensures their mass differentiation into lamellocytes, an immune cell type required to fight specific threats such as wasp parasitism. Domeless, a related receptor in Drosophila, was previously the only known cytokine receptor that signals through the JAK/ STAT pathway. We show that Latran lacks the intracellular domains required for signal transduction and acts instead by antagonizing the function of Domeless in a dosedependent manner. The role of Drosophila Latran in the repression of JAK/STAT signalling under specific immune conditions raises the question of whether short, nonsignalling receptors that antagonize full-length receptors could also control specific aspects of vertebrate immunity. STATs become in turn phosphorylated, leading to their dimerisation and translocation into the nucleus where they function as transcriptional regulators. Recent finding in Drosophila also point to a noncanonical mode of JAK/STAT signalling, which could directly control heterochromatin stability (for review [20]). Altered JAK/STAT activity has been associated with several human diseases including leukaemia, myocardial hypertrophy, and asthma, while knock-out studies in mice point to a central role in hematopoiesis and regulation of immune functions [21,22]. In contrast to mammals, only one receptor, Domeless (Dome), one JAK (Hopscotch, Hop), one STAT (Stat92E or Marelle) and three cytokines, Unpaired (Upd), Upd2, and Upd3 have been functionally characterised in Drosophila (for review [23]). Sequencing of the D. melanogaster genome revealed, however, the existence of a dome-cognate gene (CG14225, renamed here latran [lat]) ( Figure 1) [24,25].
We report here that lat acts as a negative regulator of the JAK/ STAT pathway during larval hematopoiesis. lat is required for turning off JAK/STAT signalling in hematopoietic progenitors following wasp parasitisation, thereby allowing the massive differentiation of lamellocytes. In vivo and in vitro assays indicate that Latran (Lat) forms heteromers with Dome and antagonises Dome function in a dose-dependent manner. Our studies thus revealed a novel mode of regulation of JAK/STAT signalling, based on differential and tissue-specific expression of signalling and antagonist cognate receptors. The tight tissue-specific regulation of JAK/STAT signalling by latran is crucial for Drosophila to be able to mount a dedicated cellular immune defense. A negative regulation of JAK/STAT signalling by a nonsignalling receptor chain has, so far, only been reported in primary and cultured mammalian cells, for short versions of class I cytokine receptors [26,27]. However, the in vivo function of these short membrane receptors and how their expression is regulated and linked to tissue homeostasis remain to be established. The specific role of Drosophila lat in controlling a dedicated cellular immune response raises the possibility that nonsignalling receptors could control specific aspects of vertebrate immunity, prefiguring a new field of investigations on this pathway.

CG14225/latran Encodes a JAK/STAT Receptor-Like Protein
Vertebrate class I (one) cytokines bind to receptors composed of various single-pass transmembrane protein chains that form homo-and heteromeric complexes. Dome is the only class I cytokine receptor that has, so far, been characterised in Drosophila [28][29][30][31]. Existence of a D. melanogaster gene, CG14225/lat, coding for a protein structurally related to Dome was noticed several years ago [24]. dome and lat are adjacent to each other on the X chromosome and transcribed in the same orientation, suggesting that they originated from a gene duplication event ( Figure 1A).
The key role of JAK/STAT signalling in regulating larval hemocyte homeostasis [9,32] led us to ask whether lat was involved in this regulation. We first experimentally defined the 59 end of lat transcripts by RACE-PCR, using total RNA from LGs. We positioned the lat methionine initiation codon and established that 153 bp separate the 39 end of dome mRNAs from the lat transcription start site (Figures 1 and S1A). Dome and Lat show strong similarity in their extracellular domains, which include, from N-to C-terminal, a signal peptide, a cytokine binding motif (CBM) related to that of vertebrate receptors, and an approximately 200 amino acid region not found in vertebrate receptors ( Figures 1B and S2). We designate this region, whose function remains unknown, as LDHR for Lat-Dome Homology Region. Similar to the human class I cytokine receptor GP130, Dome contains three tandemly arranged fibronectin type III (Fn III) motifs. None Fn III motif are found in Lat. The intracellular region of Lat is shorter than that of Dome and shows no consensus STAT binding site (motif YXXQ, [25]) suggesting that lat encodes a nonsignalling form of cytokine receptor. A search for dome/latrelated genes in available genomes of other Drosophila species indicated that lat and dome are arranged in tandem, with an intergenic region varying from 150 bp (D. melanogaster) to 2.5 kb (D. virilis) ( Figure S3). A higher degree of sequence conservation between orthologous compared to paralogous genes was observed ( Figure S3), suggesting that lat sequences diverge at a much higher rate than those of dome.
lat Is Specifically Expressed in Hematopoietic Progenitors in the Drosophila LG In situ hybridisations show that unlike dome, lat is not expressed in embryos, a result confirmed by reverse transcriptase PCR (RT-PCR) (unpublished data). Thus, despite genomic proximity, the control of lat transcription is different from that of dome. In larvae, lat transcription was detected only in pericardial and LG cells. In situ hybridisation onto LGs expressing a membrane targeted green fluorescent protein (GFP) either in the MZ (dome-Gal4/UAS-mcd8 GFP (dome.GFP)) or the PSC (pcol85/UAS-mcd8GFP (pcol.GFP)) indicated that lat is only expressed in the MZ ( Figure 1C-1F). Coexpression of dome and lat in MZ cells, where JAK/STAT signalling is critically required to maintain a pool of hematopoietic progenitors [9], raised the question of the role of lat in controlling the activation of the JAK/STAT pathway in prohemocytes.
lat Mutants Are Fully Viable but Unable to Mount a Cellular Immune Response against Wasp Parasitisation To determine if lat plays a role in larval hematopoiesis, we generated a lat null allele by homologous recombination [33]. Several independent recombination events were obtained and homozygous mutant lines were established ( Figure S1). Homozygous and trans-heterozygous combinations of these lines produced fertile adults with no obvious morphological defects, indicating that lat is not essential for either viability or germ-line development. In particular, no phenotypic defect was observed in the eye, where the JAK/SAT pathway plays an important role in growth and patterning [34,35]. We then looked at the morphology of the LG in lat mutant larvae, using specific markers for the MZ (tep4), the PSC (col), or for differentiated hemocytes: crystal cells (doxA3) and plasmatocytes (P1). No obvious difference could be found between wild type (wt) and lat mutant larvae, suggesting that lat is neither required for the ontogeny of the LG, nor for the differentiation of plasmatocytes and crystal cells ( Figure  S4). The third type of Drosophila hemocytes, the lamellocytes (identified by the integrin a chain [a-PS4 marker]), massively differentiate at the expense of the pool of hematopoietic progenitors upon wasp parasitisation; they start to differentiate in the LG before being released into the hemolymph. In wt larvae, the number of circulating lamellocytes reaches its maximum 48 h after wasp egg-laying [5]. In sharp contrast to wt, virtually no circulating lamellocytes are found in the hemolymph of parasitised lat mutant larvae (Figure 2A, 2B), either 48 or 72 h after wasp egglaying. Several days later, adult wasps hatch from parasitised lat mutant pupae. srp-Gal4 driven lat expression in the LG (srp.lat) completely restored the ability of lat mutant larvae to produce lamellocytes following wasp parasitisation ( Figure 2C). We therefore conclude that lat is required for the massive differentiation of lamellocytes in response to wasp parasitisation.
Lamellocyte production upon parasitisation involves downregulation of JAK/STAT signalling in the MZ, thereby licensing hematopoietic progenitors to differentiate [9]. JAK/STAT activity in the LG can be monitored by the expression of a reporter transgene, dome-MESO-lacZ (dome-MESO), where LacZ is under the control of an intronic dome regulatory element [9,36,37]. Under normal conditions, LacZ expression is observed in the MZ of lat mutant as in wt larvae, indicating that the JAK/STAT pathway is active and that lat is not required for this activity (Figures 2 and S4). 30 h postinfestation a strong reduction of dome-MESO expression is observed in wt LGs ( Figure 2D, 2E) correlating with lamellocyte differentiation ( Figure 2E, insert) and premature LG dispersal [5]. In sharp contrast, dome-MESO remains expressed in lat mutant LGs and these, unlike wt LG, do not prematurely disperse ( Figure 2F, 2G), correlating with the absence of circulating lamellocytes in the hemolymph. This shows that lat is required for the downregulation of JAK/STAT activity in hematopoietic progenitors following parasitisation. The observation of few differentiated lamellocytes in lat mutant larvae ( Figure 2G and insert) indicates, however, that lat is not required for the lamellocyte differentiation program per se. In cells were the JAK/STAT pathway is activated, Stat has a predominantly nuclear localisation. In order to follow the activity of the pathway after parasitism, we analysed the subcellular localisation of a fluorescent Stat protein, Stat-GFP [38], expressed in the LG (srp-Gal4;STAT92E-GFP). In noninfectious conditions, Stat-GFP is mainly found in the nuclei, in either wt ( Figure 2H) or lat (unpublished data) mutant larvae, consistent with active signalling. 4-6 h after wasp egg-laying, Stat-GFP is found both in the cytoplasm and the nucleus in wt LG, whereas it remains predominantly localised in the nucleus in lat mutant LG ( Figure 2I, 2J). These data show a decreased activity of JAK/ STAT signalling in wt LG, already 4-6 h after wasp parasitisation, whereas no change can be detected in lat mutant LG.

Lat and PSC Activity in the LG: Robustness of Hemocyte Homeostasis
The PSC (niche) is critically required to maintain a balance between JAK/STAT-positive progenitors and JAK/STAT-negative differentiating hemocytes in third instar LG [9]. The function of lat in the MZ raised the question of the relative contribution of positive and negative regulation by the PSC and lat, respectively, in the maintenance of this balance. Therefore, we examined the proportion of prohemocytes (expressing dome-MESO) and differentiating hemocytes in LGs double mutant for lat and col ( Figure S5). Whereas in col mutant LGs, which lack a PSC, the MZ disappears and all prohemocytes differentiate [9], we observed a less severe phenotype in lat;col double mutants, namely the loss of an organised MZ with remaining prohemocytes intermingled with differentiated hemocytes ( Figure S5A). Intermingling of prohemocytes and differentiated hemocytes was also observed in lat;col double mutants following wasp parasitisation with, in this case, some lamellocytes among differentiated hemocytes ( Figure S5B). The persistence of prohemocytes in the lat;col double mutant LG underlines the crucial role of lat in the complete switch from progenitor to differentiated state that is observed either in col mutant larvae or following parasitisation.

Lat Is a Negative Regulator of JAK/STAT Signalling
The structural similarity between Lat and Dome together with lat function suggested that lat encodes a novel negative regulator of the JAK/STAT pathway. To test this hypothesis, we overex- pressed lat in the MZ and followed JAK/STAT activity using dome-MESO expression. lat overexpression led to a complete inhibition of JAK/STAT signalling in the MZ while both crystal cells and plasmatocytes were still able to differentiate ( Figure 3A-3D). To further investigate the possible mechanism behind this inhibition, we turned to reporter assay developed in cultured Drosophila Schneider (S2-NP) cells [29]. S2-NP cells display a basal level of endogenous JAK/STAT activity, as shown by transfection of a STAT reporter gene (10XStat92E-Luciferase reporter). A much stronger activity is observed upon coexpression of either of the cytokines Upd, Upd2 [29,30,36], or Upd3 ( Figure S6). To assess for lat function, we transfected S2-NP cells with 10XStat92E-luciferase, Actin promoter-driven Renilla luciferase, Upd expression vectors (Act-renilla and Act-upd), together with Actin promoter-driven Dome (Act-dome) and/or Lat (Act-lat) expression vectors at various relative concentrations. Since high level of forced Dome expression could act as a dominant-negative [39], we transfected low levels of Act-dome, which modestly increased JAK/STAT signalling ( Figure 3E). In contrast, transfection of similar amounts of Act-lat severely decreased signalling (.4-fold), confirming that lat acts as a negative regulator of the pathway ( Figure 3E) without affecting the level of Dome expression (unpublished data). Lat function is independent of the added cytokine ( Figure S6). Intermediate levels of JAK/STAT inhibition were observed for different relative amounts of Actdome and Act-lat indicating that the ratio between Lat and Dome is critical ( Figure 3E). These data both confirmed that lat is a negative regulator of JAK/STAT signalling and suggested that the ratio between Dome and Lat is a key factor in controlling JAK/ STAT activity.
Binding of Upd to Dome induces endocytosis of receptor-ligand complexes and their trafficking through the endosomal compartments, a trafficking required to activate JAK/STAT signalling [40]. We looked at the intracellular localisation of a tagged Dome (Dome-V5) in cells transfected with both Upd and Act-Dome-V5, where the pathway is active. Dome(-V5) was localised in cytoplasmic vesicles as previously described ( Figure 3F) [40]. This localisation was unchanged in cells cotransfected with a tagged Lat (Act-HA-Lat), which results in inactivation of the pathway ( Figure 3F, 3G). Both Dome-V5 and HA-Lat were localised in mostly overlapping intracytoplasmic vesicles, indicating that negative regulation of Dome activity by Lat is not linked to a defect in Dome internalisation.

Lat and Dome Can Form Heterodimers In Vivo
Long and short forms of vertebrate class 1 cytokine receptors can form heterodimers/multimers [17,41]. While Dome was previously shown to form homodimers [42], we tested the possibility that Lat could form heteromers with Dome, by using coimmunoprecipitation (co-IP) assays. S2-NP cells were transfected with equivalent amounts of plasmids encoding tagged Dome(-V5) and Lat(-HA). Cell lysates were prepared 72 h post-transfection and subjected to IP with either anti-V5 or anti-HA antibodies, followed by Western blot analysis with one or the other antibody. Lat and Dome co-IP in both directions indicated that they form heteromers in cell culture ( Figure 4A). We then tested whether Lat and Dome form heteromers in vivo, using the bblue-bblau b-galactosidase complementation technique developed to detect protein-protein interactions in vivo [43][44][45] and already applied to show that Dome forms homodimers [42]. Briefly, this technique uses two b-Gal mutant forms (Da and Dv) that are separately inactive but can complement each other if brought into proximity by fusing them to proteins that physically interact. Like for Dome [42], Da and Dv b-galactosidases were fused to the Lat C terminus. We used the da-Gal4 driver to coexpress different combinations of Lat and Dome fusion proteins in embryos ( Figure 4B-4E), because it leads to strong ubiquitous embryonic expression of the proteins ( Figure 4C and unpublished data) [42]. Contrasting with this ubiquitous expression, X-gal staining was only detected in the salivary glands and hindgut when Dome Da and Dome Dv were coexpressed as previously reported ( Figure 4B) [42]. A similar staining pattern was observed upon coexpression of LatDa/LatDv LatDa/DomeDv or DomeDa/ LatDv ( Figure 4D, 4E, and not shown), while no staining could be detected when only one fusion protein was expressed. These results indicate that Lat is able to form homodimers and heterodimers with Dome in vivo. We then tested b-gal complementation in the MZ using the dome-Gal4 driver because da-Gal4 is not expressed in the LG (unpublished data). Staining was observed upon coexpression of LatDa and DomeDv ( Figure 4F) or DomeDa and LatDv (unpublished data), while no staining could be observed upon expression of either fusion protein alone. These complemen- tation assays show that Dome and Lat form heterodimers in the LG. Unfortunately we could not determine whether Dome or Lat can also form homodimers in the LG since dome-Gal4 driven expression of two copies of either Dome or Lat leads to early larval lethality. We observed, however, that dome-Gal4-driven expression of one copy of either Lat, LatDa or LatDv resulted in an ''outstretched wing'' phenotype in adult flies, a phenotype previously described for upd mutant flies, hence the name outstretched (os/upd) given to these mutants [46,47]. This observation indicates that both the native and Lat fusion proteins are able to downregulate the JAK/STAT pathway. eyeless-Gal4 driven expression of lat in eye discs led to smaller eyes, another phenotype reminiscent of upd mutants ( Figure  S7) [39], confirming that lat is able to downregulate JAK/STAT signalling in tissues other than the LG when ectopically expressed. Taken together, cell culture and in vivo data indicate that Lat forms inactive heteromers with Dome and is a novel negative regulator of the JAK/STAT pathway.

Wasp Infestation Results in an Increased Lat/Dome Ratio
Both cell culture and in vivo data pointed to the Lat/Dome ratio as a key component in the regulation of JAK/STAT signalling. To determine whether this ratio is modified upon wasp parasitisation, we compared the levels of accumulation of lat and dome mRNAs in LGs relative to internal controls (rp49 and rpL17, ribosomal protein mRNAs). Quantitative RT-PCR (qRT-PCR) measurements were performed on total LG RNA from control larvae and larvae 4 h after wasp egg-laying in order to mainly detect primary changes that occur in response to parasitism [48]. We detected an about 2-fold increase in lat transcripts and 2-fold decrease in dome transcripts, which results in a significant drop in the dome/lat ratio ( Figure 5A). Since JAK/STAT signalling remains on after infestation in lat mutant larvae, we repeated the analysis on RNAs from lat mutant LGs. In this case, no decrease in the level of dome transcripts was observed upon wasp infestation, indicating that this decrease depends upon lat activity ( Figure 5B). In order to strengthen this conclusion, we tested whether decreasing JAK/STAT signalling in the LG by expressing dsRNA-hop (srp-Gal4 . dsRNA-hop) could rescue the lat mutant phenotype. We indeed observed that lowering the level of Hop activity restored the ability of lat mutant LG to massively produce lamellocytes following wasp parasitism ( Figure 5C), confirming that lat acts upstream of hop in the signalling cascade. Together, these data lead us to propose that the shift in the relative levels of dome and lat expression observed following wasp egg-laying operates as a switch leading to a complete extinction of JAK/ STAT signalling, thus allowing prohemocytes to differentiate.
upd3 Is Expressed and Required for JAK/STAT Signalling Activity in the MZ Three different ligands, Upd, Upd2, and Upd3, are known to activate JAK/STAT signalling in Drosophila [47,49]. In order to determine which of them are expressed in the LG, we first performed RT-PCR experiments, starting from LG mRNA. We detected upd3 and very low amounts of upd2 but no upd transcripts ( Figures 5, 6, and S8). Since upd2 mutants have no hematopoietic phenotype (unpublished data), upd2 was not further considered in this study. We then focused on upd3 expression and function. Since only genome annotation data were available, we determined the 59 end of upd3 transcript by RACE-PCR and repositioned the ATG initiation codon ( Figure S8A). In situ hybridisation of upd3 transcripts in Dome . GFP and pcol . GFP LGs indicated that upd3 is expressed in the MZ, the PSC, and in few scattered cells of the CZ (Figure 6A, 6B). While a upd3 loss of function mutant is not available, studies performed in vivo and in cell culture have established that upd3 dsRNA expression can efficiently suppress upd3 activity [50]. We looked at the consequence of upd3 dsRNA expression (UAS-upd3 dsRNA/dome-Gal4), which drastically reduces upd3 mRNA level in the LG (Figure 5D), on dome-MESO expression. No dome-MESO expression could be detected, showing that upd3 expression in the MZ is required to maintain JAK/STAT signalling active ( Figure 6C, 6D). When upd3 dsRNA expression was targeted to the PSC (pcol-Gal4), dome-MESO expression was unperturbed (unpublished data). We then determined whether upd3 levels are modified upon wasp parasitisation. The drastic decrease of upd3 transcripts observed 4 h after infestation ( Figure 5A) shows that upd3 downregulation is an immediate response to wasp parasitisation.
While JAK/STAT signalling is dependent upon the binding of Upd to Dome, dome is itself a target of the JAK/STAT pathway in the embryonic mesoderm [36], a regulatory loop reproduced by the dome-MESO enhancer in the LG [9,36]. To directly test whether the decreased amount of dome transcripts in the LG that follows wasp parasitisation could result from the drop of upd3 activity, we measured the relative amounts of dome and lat transcripts upon upd3 dsRNA expression in the MZ. Whereas the lat level was not affected, a 2-fold decrease was observed for dome transcripts ( Figure 5D). We conclude that the decrease in dome transcripts is a secondary response consecutive to decreased levels of upd3 mRNA. Unlike dome, however, upd3 downregulation is independent of lat function ( Figure 5B). Thus, we propose the following model: wasp parasitism results in a drastic decrease in upd3 levels, which in turn leads to a downregulation of JAK/  STAT signalling and a decrease of dome transcription. This, in turn, results in an increased lat/dome ratio, which subsequently leads to the complete shut down of the JAK/STAT pathway. The complete and efficient inhibition of JAK/STAT signalling in the LG thus requires lat function (Figure 7).

Discussion
The evolutionarily conserved JAK/STAT signalling pathway was discovered from studies on the role of interferon in the control of immune responses. Vertebrate genomes encode multiple forms of all major JAK/STAT pathway components, including multiple receptor subunits. As opposed to this, in Drosophila, only one functional receptor, Dome, had so far been characterised. However, sequence similarity between dome and the neighbouring gene CG14225/lat suggested a gene duplication event and raised the question of lat function.

Switching off JAK/STAT Signalling and Orienting Prohemocytes towards a Lamellocyte Fate: Two Facets of the Drosophila Immune Response to Wasp Parasitisation
In order to neutralise parasitoid wasp eggs, the Drosophila larval hematopoietic organ must rapidly release large amounts of lamellocytes in the hemolymph. In normal developmental conditions, the PSC maintains JAK/STAT signalling in the hematopoietic progenitor zone (MZ) thereby preserving their prohemocyte character. Upon wasp infestation, however, JAK/ STAT signalling is switched off, leading to a concerted differentiation of prohemocytes into lamellocytes [9]. In col mutant LGs, which are devoid of PSC cells, no lamellocytes differentiate after wasp infestation as a consequence of the premature loss of multipotent prohemocytes; conversely, in lat mutant larvae, prohemocytes are maintained and very few lamellocytes differentiate. Prohemocytes persist in lat;col double mutant LGs, suggesting that a basal level of JAK/STAT signalling subsists in this mutant context, leading to a stochastic rather than global differentiation of prohemocytes. The comparison of col, lat, and lat;col mutant phenotypes, therefore, allows to conclude that lat functions as a switch. In normal developmental conditions, PSC activity overrides lat function in the MZ. Upon wasp infestation, PSC activity is short-circuited and lat plays a decisive role in completely silencing the JAK/STAT pathway in all prohemocytes. Hemocyte homeostasis in the LG thus relies on both extrinsic signals from the niche and intrinsic JAK/STAT activity in progenitor cells (Figure 7). In lat mutants, some lamellocytes differentiate following wasp parasitisation indicating that lat is not strictly required for the lamellocyte differentiation programme per se. Thus, switching off JAK/STAT signalling and orienting prohemocytes towards a lamellocyte fate are two distinct responses to wasp parasitisation (Figure 7).

Differential Regulation of lat and dome Expression Warrants Inactivation of the JAK/STAT Pathway upon Wasp Infestation
In situ hybridisation and LG-targeted RNA interference experiments show that upd3 is expressed and required in the MZ to maintain JAK/STAT activity in prohemocytes, therefore acting in an autocrine and/or paracrine manner, as previously reported for Upd in embryos [47,51,52]. The drastic decrease of upd3 expression induced by wasp egg-laying is accompanied by a significant decrease in dome transcripts, showing that dome is both a component and a target of JAK/STAT signalling in the MZ ( Figure 5), as previously documented in the embryonic mesoderm [36]. lat and dome mRNA levels are not, however, coregulated in response to parasitisation even though the two genes lie very close to each other on the chromosome, a tandem organisation conserved in other Drosophila species. The uncoupling between dome and lat expression results in an increased lat/dome ratio following wasp infestation, which is determinant for the ability of Lat to antagonise Dome activity. Comparative analyses of RNAs from wt and lat mutant LGs show that the primary component of the JAK/STAT pathway that is affected by wasp infestation is the level of upd3 transcripts. Although we do not know yet how upd3 is downregulated, it is tempting to speculate that it could be at a post-transcriptional level, similar to the importance of posttranscriptional regulation for cytokine levels in vertebrates (for review [53]). In summary, our results show that a primary immune response to wasp egg-laying is a strong decrease in upd3 mRNA levels in the LG, which induces a downregulation of the JAK/ STAT pathway, followed by a decrease of dome and increase of lat levels. This results in an increased lat/dome ratio that further and completely turns off the JAK/STAT pathway. Since in the absence of lat the decrease in upd3 level does not completely switch off the JAK/STAT pathway. We conclude that Lat acts as a switch The PSC signal overrides lat function in the MZ (grey shades). In response to parasitisation (middle panel), there is a decrease of upd3 and dome and increase of lat transcripts, which ultimately lead to an increased lat/dome ratio. The PSC signal is short-circuited. As a result, JAK/STAT signalling is switched off, thus licensing prohemocytes to differentiate into lamellocytes. Lat activity is strictly required in the LG for this switch. In the absence of lat (right panel), residual upd3 levels maintain JAK/STAT activity, therefore preserving a pool of prohemocytes (grey shades). Upon wasp parasitisation some differentiating hemocytes become lamellocytes, however, indicating that lat is not required for this differentiation program per se. Arrows indicate activation, vertical bars repression. doi:10.1371/journal.pbio.1000441.g007 that is required for the total arrest of JAK/STAT signalling in hematopoietic progenitors in response to wasp parasitisation, a prerequisite to massive differentiation of lamellocytes and efficient immune response (Figure 7).

dome and lat, a Pair of Duplicated Genes with Antagonistic Functions
Dome is related to the human GP130 and cognate GP130-like (GPL) signalling receptors, which form heteromeric complexes with short, nonsignalling receptors such as IL-6R or Oncostatin M receptor (OSM-R) to mediate signalling ( Figure S9) [26,54,55]. lat encodes a short-type receptor that could either act as IL-6R and confer signalling specificity to Dome or as a dominant-negative receptor similar to what has been described ex vivo for short receptors such as GPL and IL13Ra2 [27]. Cell-culture and in vivo assays show that Lat antagonises Dome activity in a dosedependent manner and forms heteromers with Dome thereby acting as a dominant-negative receptor. Altogether, these data suggest that, following parasitism, which leads to decreased cytokine levels, Lat blocks Dome activity in the LG through the formation of inactive heteromers.
While our analysis indicates that lat is specifically required in the larval hematopoietic organ for massive lamellocyte production in response to an immune challenge, phenotypes induced by ectopic lat expression show that it can antagonise JAK/STAT activity in other tissues. Together, the phenotypic and protein interaction data suggest that LG specific lat expression has been selected during evolution to fulfil specific immune functions.
Cytokine signalling pathways are subject to extensive positive and negative feedback regulations, which are crucial to generate appropriate physiological responses [22]. Two genome-wide RNA interference (RNAi) screens for JAK/STAT signalling components were conducted in Drosophila cultured cells. While they identified large sets of putative positive and negative regulators, they failed to detect lat, either because its expression level in cell culture is too low to be functional or because the lat dsRNA constructs used in these screens were not efficient enough [29,30,56]. col/kn was identified in one of these RNAi screens, however, as a positive regulator acting downstream of Hop [30], suggesting another possible level of regulation of JAK/STAT signalling in the LG. Initial evidence for the involvement of JAK/STAT signalling in Drosophila cellular immunity came from the observation that a dominant gain-offunction mutation of the JAK kinase (hop Tum ) provokes the apparition of lamellocytes and melanotic masses in the absence of wasp infection. This finding led to the conclusion that upregulation of JAK/STAT signalling triggers lamellocyte differentiation, which is in apparent contradiction with our present data [32]. Whether constitutive JAK/STAT signalling in differentiating hemocytes could instruct them to become lamellocytes remains an interesting possibility. Of note, a STAT target, chimno, was recently shown to be expressed at higher levels in differentiating CZ cells as compared to undifferentiated MZ cells [57]. Recent studies further suggest a dual role of Wg signalling in the maintenance of prohemocytes and PSC cells [58]. A tight control of ROS levels in the MZ is also required to maintain a pool of prohemocytes [59]. How these different signalling pathways are interconnected with JAK/STAT signalling in order to maintain hemocyte homeostasis in the LG are important questions to be addressed in the future.

A Conserved Role for Dominant-Negative Short Receptors in the JAK/STAT Pathway?
The type I cytokine receptor family has considerably expanded in vertebrates [60]. This expansion results both from an increased number of receptor genes and from the generation of various protein isoforms that can act as either receptors or coreceptors [61]. Soluble versions of short receptors isolated from diverse body fluids have also been identified, which behave as antagonists by competing with membrane-associated receptors for ligand binding [62,63]. These soluble receptors are generated by either limited proteolysis or translation from differently spliced mRNAs. Finally, studies on IL13Ra2 [27] or GPL [64] suggested that short receptors anchored to the membrane could also behave as dominant negative receptors. However, the exact function of these receptors and how their expression is regulated and linked in vivo to tissue homeostasis, remains unknown. Our studies in Drosophila indicate that Lat acts as a dominant-negative receptor rather than a coreceptor, extending in vivo the few observations made in mammalian cell cultures [65,66]. Tissue-specific regulation of JAK/STAT signalling in response to environmental cues is crucial for the ability of Drosophila to mount a cellular immune defense. Our results bring to light a new mode of fine tuning of the JAK/STAT pathway, that is, differential expression of signalling and antagonist cognate receptors. Whether and when regulated expression of long and short receptor isoforms is employed in controlling specific aspects of immunity in vertebrates certainly deserves further investigation.

Generation of a lat Nul Mutant by Site-Directed Recombination
The procedure was adapted from the Ends out Knock Out method [33]. A lat KO ''donor'' transgene was constructed in pW25 [68] by inserting 4 kb of 59 and of 39 flanking sequences of the lat gene separated by the mini-white gene and used to transform white mutant flies ( Figure S1B). Two different inserts on the second chromosome were selected for the recombination-targeting protocol. Several independent lat KO lines were obtained and verified for deletion of lat and insertion of mini-white by PCR and Southern blot analyses ( Figure S1C). The lat 18A line was chosen for all the experiments.

Constructs
The mapping of lat and upd3 transcript 59 ends was performed by 59 RACE PCR (Marathon cDNA amplification kit, Clontech, and BD Smart RACE kit, BD Biosciences), using either polyA+ RNA from hop Tum-l larvae or total RNA from dissected w LGs. A full length lat cDNA was reconstructed and inserted in pUAS-T to generate UAS-Lat transgenic lines. UAS-LatDa and UAS-LatDv were constructed using the complete lat cDNA fused to the bgalactosidase Da and Dv fragments from pUAS-Dome-LacZDa and pUAS-Dome-LacZDv, respectively [42]. The fusion constructs were subcloned into pUAS-attB to generate transgenic flies using the ZH49B and ZH86F attP integration platforms [69]. Act-Lat, Act-HALat, and Act-DomeV5 plasmids were constructed and used for cell culture experiments.

RNA Probes
A 526-bp lat genomic fragment amplified using primers 6 and 8 ( Figure S1A) was cloned in the Invitrogen pCRBluntII-TOPO vector. Two different upd3 probes of 836 bp (primers 1 and 2) and 2,057 bp (primers 5 and 6) were designed for in situ hybridisation ( Figure S8).

X-Gal Staining
X-Gal staining was as described in [42].

Dissected
LGs were collected in trizol and total RNA was extracted using trizol reagent (Invitrogen) according to the manufacturer. Superscript Reverse Transcriptase II (Invitrogen) and oligo dT primers were used for reverse transcription. Realtime quantitative PCR was performed on a MyiQ single color real time PCR detection system (Biorad). CT values were collected and analysis was performed according to the 2 DDCT method [71] using rp49 and rpL17A to normalize estimates of relative expression. Primers used: 1 and 3 and 5 and 7 for dome and lat, respectively ( Figure S1), 3 and 4 for upd3 ( Figure S8). No significant differences were detected in the level of control RNAs in wt, lat and dome . upd3dsRNA experiments ( Figure 5). Primers sequences for rp49, rpL17A, upd, and upd2 are available on request. All qRT-PCR data are representative of three independent experiments and presented as means 6 standard deviation (SD). Statistical analyses were performed using Student's t test.

Cell Culture Experiments
Various amounts of Act-Lat, 0.2 ng of Act-Dome, and 1 ng of either Act-Upd, Upd2, or Upd3 were used to transfect S2-NP cells [29]. Luciferase assays were performed 4 d later, and the reporter activity was normalised as the ratio of firefly luciferase/Renilla. The results are from three independent experiments. For immunostaining, S2-NP cells were transfected with 1 ng of Actupd and 0.2 ng of Act-Dome-V5 with or without 1 ng of Act-HALat ( Figure 3E and 3F).

Immunoprecipitation of HALat/DomeV5 Complex
Drosophila S2-NP cells [29] were maintained in Schneider medium +10% FCS + penicillin-streptomycin (Sigma 1/100) at 25uC without supplemental CO2. Cells (3610 6 per well) were seeded and cultured in six-well plates. 24 h later, transfections using Effectene (Quiagen) were performed. Each well was transfected with 20 ng of plasmid encoding either HA-Lat, Dome-V5, or both, and completed with plasmid DNA encoding the empty vector (pHA vector) to a final amount of 400 ng of DNA. 48 h later, cells from each well were washed in PBS and lysed in 150 ml of ice-cold buffer containing 50 mmol Tris (pH 7.4), 150 mmol NaCl, 1 mmol EDTA, 1% NP40, and antiprotease cocktail (Roche) for 20 min. 140 ml of the crude lysate was used for IP. Protein G sepharose beads (Sigma) were first incubated with 1 mg of anti-HA or anti-V5 antibodies for 1 h at 4uC and then with the cleared supernatant for 2 h at 4uC. Beads were then boiled in denaturing sample buffer and the released proteins loaded on a gel with 3 ml of the crude lysate (1/50 of the total preparation) used as a control lane. The separated proteins were analysed by Western blotting with either mouse anti-V5 or mouse anti-HA antibodies.