dsRNAs were produced as described previously (31). The primers used for dsRNA synthesis are listed in Supplementary Table 1 .

The Drosophila S2 cells were cultured at 25°C in Schneider’s Insect medium (Sigma, Cat#S9895-1L, USA) containing 0.4 g NaHCO₃ (Sigma Cat#22350-6, USA) and 0.8 g CaCl₂ · 2 H₂O (Sigma S5791-500g, Germany), 10% heat-inactivated Fetal Bovine Serum, FBS (Sigma, Cat#F7524, France) and 50 U/ml Penicillin/Streptomycin (Biochrom AG, Cat#A2213, Berlin).

The S2 cells were transiently transfected using Fugene6 reagent (Roche, Basel, Switzerland) following the manufacturer’s instructions. The JAK/STAT pathway was activated by transfection with either the constitutively active pMT-hop^Tum-l or with the pMT-upd1-myc plasmid. To measure the JAK/STAT pathway activation, a Turandot M-luciferase (TotM-luc) reporter construct, a kind gift from Professor Jean-Luc Imler (University of Strasbourg, France), was used. For studying the effect of RNAi knockdown of selected genes on the expression of ET-V5, S2 cells were transfected with either the pMT-empty or pMT-Upd1 plasmids together with the pMT-ET-V5 plasmid and selected dsRNAs. For studying the effect of RNAi knockdown of selected genes on the expression of endogenous STAT92E, S2 cells were transfected with the pMT-Upd1 plasmid for induction of the JAK/STAT pathway. The pMT-empty plasmid was used as a control for uninduced conditions. The protein production from the pMT plasmids was induced 24-48h post transfection by adding CuSO₄ to each well to a final concentration of 250-500 µM. For Toll and Imd pathway reporter assays, a Spätzle or Imd overexpressing plasmid was used to induce the pathway and Drosomycin-luc or Attacin-luc was used as the reporter, respectively (33, 34). In all reporter assays, an Actin5C-β-galactosidase reporter plasmid was used to normalize the results for transfection efficiency and cell numbers as described previously (31). Luciferase and β-galactosidase reporter assays were carried out as previously described in (31).

The S2 cells were stably transfected with 9.5 µg of each plasmid construct (empty pMT-vector or pMT-ET-V5 ± pMT-upd1-myc) together with 0.5 µg of the coHygro plasmid to enable Hygromycin-B selection of transfected cells. Transfections were carried out on 6-well plates using the Fugene6 transfection reagent (Roche, Basel, Switzerland). Three days post-transfection the transfected cells were selected by adding medium containing 150-300 µg/ml of Hygromycin-B (HyClone Laboratories, Inc. USA). The Hygromycin-B concentration was increased in every passaging up to 300 µg/ml. Under Hygromycin-B selection, only cells carrying the coHygro plasmid survive. From here onwards, the cells were grown under Hygromycin selection.

24-48h after the protein production from the pMT-plasmids was induced with CuSO₄, the cells from three pooled wells, each containing 3 ml of S2 cell suspension, were collected per sample and centrifuged at 5000 x g for 3 min. The supernatants were discarded and 900 µl of lysis buffer (1 x PBS, 1% Igepal CA-630 [Sigma-Aldrich/Merck, Darmstadt, Germany], 0.5% Sodium deoxycholate, 0.1% SDS, 2 mM EDTA, Halt™ Protease and Phosphatase Inhibitor Cocktail, EDTA-free [Thermo Fisher Scientific Inc.]) was added to each sample. The samples were left to lyse on ice for 30 min. The lysates were centrifuged at 16 000 x g for 10 min and the cleared lysates were transferred into new tubes. The protein concentration of each lysate was measured with the Pierce™ BCA protein Assay Kit (Pierce^®, #23227, USA) according to the manufacturer’s instructions.

Putative ET-interaction partners were co-immunoprecipitated from S2 cell protein lysates using Anti-V5 Agarose Affinity Gel (Sigma, #A7345, Israel) according to the manufacturer’s instructions. The following lysates were used: S2 cells (neg. ctrl 1), empty pMT-vector (neg. ctrl 2), pMT-ET-V5 and pMT-ET-V5 + pMT-upd1-myc. 900 µl of each lysate, containing approximately 2.5 mg of protein, was incubated overnight with the Anti-V5 Agarose Affinity Gel, after which the complexes were washed 6 x 10 minutes with PBS containing protease and phosphatase inhibitor cocktail. All incubations and treatments were carried out at 4°C or on ice. Finally, loading buffer was added on the washed affinity gel containing protein complexes and the samples were boiled to release the proteins. Purified protein complexes were analyzed by SDS-PAGE and silver-staining.

For SDS-PAGE, Novex 10% NuPAGE Bis-Tris (Life Technologies, #NP0301BOX, USA) gels were used. Precision Plus Protein™ Dual Xtra Standard (Bio-Rad, #161-0377, USA) was used as a marker. The electrophoresis was carried out for 45 min using MOPS buffer and NuPAGE Gel program (200 V, 120 mA, 25 W). To stain the protein bands on the gel, the Pierce^® Silver Stain for Mass Spectrometry kit (Thermo Scientific, #24600, USA) was used according to the manufacturer’s instructions.

The anti-V5 antibody purified samples were analyzed by mass spectrometry with the help of Tuula Nyman at the University of Helsinki. To identify the proteins in complex with ET, the anti-V5 purified lysates were separated on SDS-PAGE gels and silver stained as described above. The whole sample lanes on SDS-PAGE gel were cut into equal sized pieces after which an in-gel trypsin digestion was performed. The digested peptides were analyzed as described in (35). Flybase Query and Blast tools were used to process the obtained data. The data from the LC-MS/MS was searched against the SwissProt database through ProteinPilot with MASCOT. The following MASCOT search parameters were used: Drosophila melanogaster as species, carbamidomethyl modification of cysteine as a fixed modification and oxidation of methionine as a variable modification, trypsin digestion allowing one missed cleavage, 50 ppm mass tolerance and 0.2 Da fragment tolerance for peptides. False discovery rates for the LC-MS/MS data were 1.5-4%.

For studying the effect of RNAi knockdown of selected genes on the expression of pMT-ET-V5, protein lysates were prepared as above and the production of the ET-V5 protein was analyzed by SDS-PAGE and Western blotting. The total protein content of the lysates was set to 50 µg/lane. For SDS-PAGE, Biorad TGX 4-20% mini gels (Bio-Rad, #456-1093, USA) were used. PageRuler™ Prestained protein ladder (Thermo Scientific #26616, Lithuania) was used as a marker. The electrophoretic separation of proteins was carried out for 20-25 min with 300 V using the Tris/Glycine/SDS buffer (Bio-Rad #1610732, USA).

For Western blotting, proteins were transferred from the SDS-PAGE gel either onto 0.45 µm nitrocellulose membrane (Hybond-C extra; Amersham Biosciences, RPN203E, United Kingdom) using the wet transfer technique with an in-house transfer buffer (25 mM Tris, 192 mM glycine, 8% methanol) and the NuPAGE Blot program (25 V, 160 mA, 1 h 15 min), or onto 0.2 µM nitrocellulose membrane using Trans-Blot turbo transfer packs (Bio-Rad, #1704158, USA) and equipment according to the manufacturer’s instructions for TGX mini-gels. The membranes containing the transferred proteins were first incubated with blocking buffer (5% skimmed milk powder in PBS + 0.05% Tween 20 [PBST]) for 2 h at RT or overnight at 4°C. After blocking, the membranes were incubated with primary antibodies in blocking buffer for either 1-2h at RT or overnight at 4°C. The following antibodies were used: anti-V5-HRP conjugate (1:3000, Invitrogen™, Life Technologies, P/N 46-0708), anti-V5 antibody (1:3000, Invitrogen™, Life Technologies P/N 46-0705) and α-tubulin antibody (1:1000, clone DM1A, Sigma-Aldrich). After incubation, the membranes were rinsed three times with PBST and further washed with PBST 4 x 5 min at RT. Goat anti-mouse IgG (H+L) HRP secondary antibody (1:5000, ThermoFisher #G-21040) was used for detecting the mouse monoclonal antibodies (V5 antibody, α-tubulin antibody). Secondary antibody was incubated for 1 h at RT, after which the membranes were washed as before. Immunostained proteins were visualized using an ECL Western Blotting Detection Reagent (either with #RPN2232 by GE Healthcare Amersham™, UK or Westernbright, Advansta, USA) and imaged with the BioRad ChemiDoc™ MP imager using the BioRad ImageLab software (Bio-Rad, California, USA) or by developing the signal on an X-ray film (Fuji Super RX-N film, Japan).

To estimate the amounts of the proteins of interest produced, band intensities on exposed films or imager images were analyzed with the FiJi-ImageJ-64-bit (ver1.51) software. Briefly, images were saved in grayscale in 32-bit mode, after which the bands to be analyzed were selected and their intensities plotted. The intensity of the ET-V5 protein band from the cell lysate was normalized to α-tubulin signal from that lysate. Plotted values were normalized to the control (GFP dsRNA-treated) protein band value. Per phenotype, 6-7 replicates were analyzed.

We utilized the GAL4/UAS system (36) to target the expression of transgenes into the fly blood cells, the hemocytes. To visualize larval hemocytes we used the eaterGFP (37) and msnCherry (38) fluorescent reporters for plasmatocytes and lamellocytes, respectively. These reporters were combined with two hemocyte drivers, P{Hml-GAL4.Δ}2 [Bloomington Drosophila Stock Center, BL #30139 (39)], and P{He-GAL4.Z}85 [BL #8699 (6)], in order to simultaneously visualize hemocytes and to express transgenic constructs in them, resulting in fly strain yw,msnF9mo-mCherry,eaterGFP;Hml^Δ-GAL4;He-GAL4, hereafter called mCherry,eaterGFP;Hml ^Δ-GAL4;He-GAL4. For suppression of gene expression by RNA interference (RNAi), we used the following transgenic lines obtained from the Vienna Drosophila Resource Center (VDRC) GD (P-element) and KK (phiC31) RNAi stocks: UAS-Prosα6^GD (VDRC ID #26653), UAS-Prosα6^KK (#100703), UAS-Stat92E^GD (#43867) and UAS-hep ^GD (#2968). The VDRC w¹¹¹⁸ strains #60100 (denoted as w^KK ) and #60000 (denoted as w^GD ) were used as genetic background controls for the KK and GD stocks, respectively. The following combination lines were generated: w; UAS-Prosα6^GD (#26653); UAS-Stat92E^GD (#43867) and UAS-hep^GD (#2968); UAS-Prosα6^GD (#26653); +. For testing the UAS-Prosα6-RNAi silencing efficiencies, we used a combination of the hemocyte drivers mentioned above, that also contain two inserts of UAS-eGFP (w; HmlΔ -GAL4, UAS-eGFP; He-GAL4, UAS-eGFP). Of note, to denote the presence of the GAL4/UAS in the figures, we use the symbol “>“.

For detecting the JAK/STAT activity in vivo, the 10xStat92E-GFP [BL #26197 (40)], reporter combined with He-GAL4 to get 10xSTAT92E-GFP; He-GAL4 was used (a gift from Prof. Dan Hultmark). To activate JAK/STAT signaling, the constitutively active form of Drosophila JAK, UAS-hop^Tum-l (41) was expressed in hemocytes. 15-20 females of a driver or genetic background control line were crossed with 10 males from transgenic construct lines and kept at 25°C and 12:12 light:dark cycle (LD12:12). Flies were placed daily in fresh vials, and vials with eggs were transferred to 29°C and LD12:12.

Late 3rd instar larvae were washed in water with a brush until clean, placed in a 20 µl drop of 8% bovine serum albumin (BSA) in 1 x phosphate buffered saline (PBS) and carefully ripped open using forceps. The hemolymph was allowed to bleed out and the carcass was removed. The hemolymph sample was pipetted into a 1.5 ml Eppendorf tube with 80 µl of 8% BSA in PBS. 30 µl of the sample was run with a BD Accuri C6 flow cytometer (Becton, Dickinson & Company). Each genotype was analyzed in triplicate (3 x 10 larvae). For the eaterGFP and msnCherry reporter analysis, a 488 nm 50 mW solid-state laser and 510 +/- 15 nm (FL1, GFP) and 610 +/- 20 nm (FL3, mCherry) optical filters were used to capture the fluorescence signal. GFP-only, mCherry-only and non-fluorescent hemocytes were used to check for the location of these hemocytes and to deduct fluorescence spill over into a wrong channel. By using these reporters, five hemocyte populations can be detected, consisting of single and double positive hemocytes. Following the naming strategy presented in (42), these hemocyte are: plasmatocytes (GFP^high), activated plasmatocytes (GFP^high,mCherry^low), lamelloblasts (GFP^low), prelamellocytes (GFP^low,mCherry^low) and mature lamellocytes (mCherry^high). The complete procedure of the hemocyte flow cytometry with gating strategies is described in (42).

For the 10xStat92E-GFP reporter fluorescence measurement, the hemocyte population was separated from the debris based on forward scatter (FSC-A) and side scatter (SSC-A). This population was gated and used in subsequent analyses. A 488 nm 50 mW solid-state laser was used to excite the green fluorescent protein (GFP) and the emission was detected using a 510 +/- 15 nm (FL1) optical filter. Non-fluorescent hemocytes (w^GD ) were used as a negative control and hemocytes with hop^Tum-l overexpression (10xStat92E-GFP;He-GAL4/UAS-hop^Tum-l ) as a positive control. Hemocytes with a fluorescence intensity above the autofluorescence of the hemocytes in the FL1 channel were considered GFP-positive. A forward scatter area vs. height plot (FSC-A vs. FSC-H) was used to define the population of round hemocytes, representing the majority of hemocytes in a healthy Drosophila larva, the plasmatocytes. These hemocytes were gated, and the GFP fluorescence intensity was measured for all hemocytes gated in the FSC-SSC plot and separately only for the hemocytes falling into the plasmatocyte gate.

3^rd instar larvae were bled in a drop of 1% BSA in PBS on a 12-well glass slide, pooling hemocytes from three larvae per well. Hemocytes were allowed to attach and spread on the slide for 50 min, after which they were fixed for ten minutes using 20 µl of 3.7% paraformaldehyde. After fixation, wells were washed three times using cold 1 x PBS and permeabilized for 5 min using 0.1% Triton-X100. After washing three times using cold 1 x PBS, cells were incubated for 30 min with 20 µl of Alexa Fluor 680 -conjugated Phalloidin (Invitrogen), in a 1:50 dilution in 1% BSA in PBS, to stain filamentous actin. The wells were washed, and the samples mounted using ProLong Gold antifade reagent with DAPI (ThermoFisher Scientific) and Zeiss coverslips (thickness no. 1 ½, 18 x 18 mm). Slides were left to cure overnight at room temperature protected from light and imaged with Zeiss LSM 780 laser scanning confocal microscope using a 40x oil immersion objective at several random locations on the next day. 405 nm, 488 nm and 628 nm lasers were used to excite DAPI, GFP and AlexaFluor 688 nm, respectively, and emission was collected at 410-488 nm, 490-544 nm and 661-759 nm. Multiple layers were imaged in Z-plane at 0.36 µm intervals. ImageJ version 2.1.0/1.53 c was utilized to create stacked images. ImageJ and Adobe Photoshop (release 22.5.1) were used to enhance the fluorescence signal for better visibility, keeping modifications constant across the images.

The effect of silencing Prosα6 in hemocytes on the number of successfully pupated and eclosed animals was assessed in msnCherry,eaterGFP;Hml^Δ-GAL4;He-GAL4 flies crossed with UAS-Prosα6^GD , UAS-Prosα6^KK and w^GD flies. Three to four replicate crosses of each genotype were made, and adults were transferred to fresh vials on three subsequent days. Eggs were counted from each vial and the vials were monitored daily for pupated/eclosed animals. At the same time, the number of pupae with melanotic nodules was recorded. Examples of pupae with melanotic nodules were imaged with a Nikon DS-Fi2 camera attached to a Nikon SMZ745T microscope and a Nikon DS-L3 camera control unit.

The prevalence of larval melanotic nodules was assessed for the same genotypes in three replicate crosses (100 animals each) by inspecting the larvae under a stereomicroscope. Example images of larvae (20x magnification) were taken with a Deltapix Invenio 10EIII camera (DeltaPix, Denmark) attached to a Nikon SMZ745T stereomicroscope using the DeltaPix InSight software.

The effect of silencing Prosα6 in hemocytes on the viability of the flies was assessed by monitoring the lifespan of the flies compared to controls. UAS-Prosα6^GD and UAS-Prosα6^KK RNAi lines were crossed with the msnCherry,eaterGFP;Hml^Δ -GAL4; He-GAL4 flies (described above) at 25°C. As a control, msnCherry,eaterGFP;Hml^Δ -GAL4; He-GAL4 flies were crossed with w^GD flies. The fly eggs were transferred to 29°C, and this temperature was used for the entire lifespan of the flies. Every 2-3 days the number of living flies was recorded, and the flies were transferred to fresh food.

Expression levels of selected genes in S2 cells and hemocytes were measured by quantitative reverse transcriptase PCR (qRT-PCR) from extracted RNA. For monitoring expression levels in S2 cells, the cells were grown on 24-well plates, treated with selected dsRNAs and transfected with selected plasmids as described above. For studying expression levels in hemocytes in vivo, hemocytes from 50 dissected larvae were collected for each replicate; three biological replicates from three independent crosses were used per genotype. The RNA extraction both from hemocytes and cultured cells was performed using the TRI reagent (MRC, Thermo Fisher Scientific). S2 cells were harvested from culture plates by pipetting and centrifugation (5000 x g, 3 min), after which cells were homogenized and lysed in TRI reagent by pipetting up and down at least 10 times. Extraction from hemocytes was started by adding TRI reagent onto the frozen hemocyte pellet, and cells were homogenized and lysed by pipetting up and down at least 10 times. Thereafter the extraction was performed according to manufacturer’s instructions. qRT-PCR was carried out from extracted RNAs (30-40 ng RNA/sample) using the iTaq Universal SYBR Green Onestep kit (Bio-Rad, Hercules, CA). The expression values obtained for selected genes were normalized to the expression of a gene encoding for ribosomal protein L32 (RpL32). The primers used are listed in Supplementary Table 1 .

Statistical analyses of luciferase assay measurements, qRT-PCR results and protein band intensity analyses were carried out using the two-tailed Student t-test for two arrays assuming equal variances. Statistical analyses of fly life span experiments were carried out with the log-rank (Mantel-Cox) test using Prism 6 (GraphPad) software. The difference was considered statistically significant if the p-value was < 0.05. Proportional data were analyzed using a Generalized linear model (GLM) with binomial distribution and combined with Tukey’s post-hoc test for pairwise comparisons. The data on hemocyte numbers were analyzed using a negative binomial GLM followed by pairwise comparisons of groups using estimated marginal means. A Kruskal-Wallis rank sum test combined with Dunn’s post hoc test was applied to GFP intensity comparisons in Figure 3D . R version 4.0.4 was used to perform GLM and Kruskal-Wallis rank sum tests.

In a previous study, we used a large-scale RNAi screen in Drosophila S2 cells to discover genes important for the Drosophila JAK/STAT pathway (31). In the screen we identified Eye transformer (ET) as a negative regulator of the JAK/STAT pathway (31). To screen for putative ET interactors, we created a Drosophila S2 cell-line stably overexpressing the ET protein with a V5 tag (ET-V5), and affinity-purified all the proteins in the ET complex with the V5 antibody. In Supplementary Figure 1A , it is shown that the ET-V5 signal can be detected with the V5-antibody (α-V5) in the lysates from ET-V5-expressing S2 cells. Since this control experiment showed that ET-V5 is successfully captured with this method, we next affinity purified proteins in complex with ET, separated them using SDS-PAGE and visualized the proteins by silver staining ( Supplementary Figure 1B ). Each line was cut, digested with trypsin and the protein composition was determined by mass spectrometry.

The potential ET interaction partners were studied in two situations, either with or without overexpression of the Dome receptor ligand upd1. When the ligand is present, the JAK/STAT pathway is activated, whereas without the ligand, the pathway stays inactive. In addition, untreated S2 cells and S2 cells transfected with an empty pMT vector were used as negative controls. After proteins that were also found in the negative controls were excluded, we identified in total 173 Drosophila melanogaster proteins in complex with ET under the conditions where the JAK/STAT pathway was inactive (ET-V5 alone) and 175 proteins when the ligand Upd1 was present (ET-V5 + Upd1-myc). Out of these, 136 proteins were found in both situations (ET with and without Upd1 induction). The mass spectrometry raw data, including putative candidates for JAK/STAT pathway regulation, is shown in Supplementary Table 2 .

The mass spectrometry screen identified in total 212 unique putative interaction partners of ET ( Supplementary Table 2 ). These genes/proteins were compared to the Kallio & coworker’s screen for JAK/STAT components (31) as well as other literature, to search for links to the JAK/STAT pathway, including immunity, hematopoiesis or stress response (25). In total, nine genes were selected for further study: four gene products where the interaction was detected with ET alone, three gene products where the interaction with ET was detected upon Upd1 induction and two gene products where the interaction was detected with ET both with and without Upd1 induction. The selected genes are listed in Table 1 , with a description of the gene function deduced from the www.flybase.org gene pages. To assess if the association of ET with these nine gene products signifies a regulatory effect on the JAK/STAT pathway, we carried out a JAK/STAT pathway reporter assay in S2 cells with dsRNAs targeting the genes ( Figure 1A ). As shown in Figure 1A , induction with overexpression of either upd1 (gray bars) or hop^Tum-l (black bars) causes an activation of the JAK/STAT pathway, as assessed by the production of the luciferase signal from the JAK/STAT target gene luciferase reporter construct Turandot M-luciferase (TotM-luc). GFP dsRNA was used as a negative control in both assays. Knockdown of ET was used as a positive control in measuring the Upd1-induced TotM-luc activity, but as ET is upstream of Hop, its knockdown does not have an increasing effect on Hop^Tum-l -induced TotM-luc activity, but instead, a decreasing effect, as discussed e.g., in (31). Of the nine RNAi treatments tested, only Proteasome α6 subunit (Prosα6) causes hyperactivation of the JAK/STAT pathway upon hop^Tum-l induction. Prosα6 is a component of the core particle of the proteasome complex [see below (43)]. The Upd1-induced TotM-luc signal is enhanced when G protein α o subunit (Gαo), Myosin light chain cytoplasmic (Mlc-c), Moesin (Moe), Prosα6 or Twins (tws) are knocked down by targeted RNAi.

We have previously carried out a genome-wide JAK/STAT pathway RNAi screen (31), and when analyzing unpublished putative negative regulators of the pathway found in the screen, we identified 16 genes whose RNAi caused more than a 10-fold hyperactivation of the JAK/STAT pathway ( Table 2 and Figure 1B ). Of these, 69% (11/16) had a described function in the proteasome pathway. The five non-proteasome-related putative negative regulator genes, are hephaestus (heph), karyopherin α3 (Kap-α3), Furin1 (Fur1), Integrator 2 (IntS2) and Integrator 6 (IntS6).

The proteasome is a large protein complex consisting of a 20S core particle capped by 19S regulatory particles (43). Among the seven alpha rings of the proteasome core particle, RNAi-mediated silencing of Prosα1, Prosα6 and Prosα7 caused hyperactivation of the JAK/STAT pathway as did silencing of four of the seven beta ring sub-particles and one related gene, namely Prosβ2, Prosβ3, Prosβ5, Prosβ7 and Prosβ2R1. Moreover, silencing of the Drosophila 19S regulatory particle genes Rpn1, Rpn2 and Rpn3, as well as proteasome-related chaperone protein Pomp, had similar effects ( Table 2 and Figure 1B ). In conclusion, RNAi-mediated silencing of the expression of proteasome genes causes hyperactivation of the JAK/STAT pathway in S2 cells [ Table 2 and Figure 1B (44)].

The total of twenty-five genes from the study of ET interaction partners ( Table 1 and Figure 1A ) and from the Kallio & coworkers’ RNAi screen for negative regulators of the JAK/STAT pathway ( Table 2 and Figure 1B (31), were selected for further study. To investigate if these genes act in the regulation of other immune-related pathways besides JAK/STAT, we tested if their silencing affects the activation of the Imd or Toll pathways using reporter assays. To investigate effects on the Imd pathway, the pathway was induced by transfecting S2 cells with the Imd plasmid and relevant dsRNAs. GFP was used as a negative control and Relish dsRNA as a positive control. As shown in Figure 1C , most of the genes studied do not regulate the Imd pathway; however, knockdown of effete (eff) and Prosβ2R1 leads to a slight decrease in pathway activation compared to GFP. Effete, the Drosophila homolog of the Ubc5 E2-ubiquitin conjugating enzyme, has previously been shown to be needed for Imd ubiquitination (45). Knockdown of Prosα6, Prosα7, heph and Fur1 leads to a slight elevation in the Imd pathway activation, whereas when Kap-α3, IntS2, and IntS6 are silenced, the pathway is markedly upregulated ( Figure 1C ).

To investigate the effects of silencing of the selected 25 genes on the Toll pathway, the pathway was induced by overexpression of the Spz^C106 plasmid. A MyD88 dsRNA treatment was used as a positive and GFP as a negative control of the pathway. As shown in Figure 1D , the silencing of several genes resulted in a reduction in Toll pathway activity, including mainly components of the proteasome: Rpn-genes, Pomp, Prosα- and Prosβ-genes (but not Prosβ2R1). A statistically significant reduction was seen also with Ced-6, Gαo, Twins (tws) and heph. Three genes that negatively regulate the Toll pathway were identified: IntS2, IntS6 and eff. The effect of IntS2 and IntS6 on the Toll pathway has been previously shown (46), but the effect of eff on the Toll pathway has not been previously studied and remains to be investigated.

As a summary, the integrator complex members (IntS2 and IntS6) appear to negatively regulate all pathways tested indicating that their effect is not specific to the JAK/STAT pathway. Although Effete co-localizes with ET, it does not seem to regulate the TotM-luc reporter gene activation mediated by the JAK/STAT pathway. Instead, Effete appears to negatively regulate the Toll pathway in S2 cells, which requires further exploration in the future. RNAi against the proteasome complex components does not have major effects on the Imd pathway, but silencing proteasome complex members decreases the activity of the Toll pathway indicating that the proteasome positively regulates the Toll pathway. Cactus, the Drosophila homologue of the mammalian Inhibitor of κB (IκB), is known to be degraded by the proteasome upon induction of the Toll pathway (47, 48), so it is plausible that silencing components of the proteasome complex leads to accumulation of the Cactus protein and reduction in the activity of the Toll pathway also in this context.

Because the proteasomal genes were identified in JAK/STAT pathway regulation in S2 cells ( Figures 1A, B ), we selected Prosα6 to study in more detail. To study the interplay between ET and Prosα6 in the JAK/STAT pathway, we carried out a double knockdown experiment with ET and Prosα6. As shown in Figure 2A , the TotM-luc activity was induced with the pMT-hop^Tum-l construct. As previously shown, ET knockdown reduces this hop^Tum-l -induced TotM-luc activity [(31) and Figure 1A ], while Prosα6 knockdown causes hyperactivation of this activity ( Figure 1A ). When both ET and Prosα6 are knocked down simultaneously (Prosα6+ET), the result is additive, indicating that the effect of ET and Prosα6 proteins on the JAK/STAT pathway activation is partially independent.

Next, we investigated what happens to the expression of ET upon Prosα6 knockdown. S2 cells were treated with Prosα6 dsRNA and control dsRNAs and either transfected with pMT-Upd1-myc, which induces the JAK/STAT pathway, or transfected with the pMT-empty plasmid for control. ET expression values from RNAs extracted from the cells were measured and normalized with RpL32 expression. As shown in Figure 2B , Prosα6 knockdown causes reduction in the transcription of ET both in conditions where the JAK/STAT pathway is inactive, and in those where it is activated with overexpression of Upd1.

Further, we investigated the effect of Prosα6 knockdown on the amount of ET protein, by treating S2 cells with Prosα6 dsRNA and control dsRNAs and transfecting them with pMT-ET-V5 together with pMT-Upd1-myc, which induces the JAK/STAT pathway. Treatment of cells with the pMT-empty vector and GFP dsRNA was used as a control. Cellular lysates were prepared and subjected to SDS-PAGE electrophoresis, Western blotting and antibody treatments and imaged. The ET-V5 protein band intensity was normalized to α-tubulin band intensity values in each sample. As shown in Figures 2C, D , the amount of ET-V5 protein is reduced in Prosα6 dsRNA-treated cell lysates compared to the negative control GFP dsRNA-treated lysates. Figure 2C shows an example of one experiment, and Figure 2D shows the quantification from three independent experiments including in total six or seven replicates per treatment. In ET dsRNA-treated cells, the ET-V5 band is not at all visible indicating that RNAi in S2 cells silences the protein production of the targeted gene very efficiently.

In conclusion, upon knockdown of Prosα6, ET transcription as well as the amount of the ET protein is decreased. As ET is the key negative regulator of the JAK/STAT pathway, these results in part explain the hyperactivation of the pathway by knockdown of Prosα6.

Since our previous results indicated involvement of the proteasome complex in regulating JAK/STAT signaling in S2 cells, we next tested if this was also the case in vivo in Drosophila larval hemocytes. We used a STAT reporter 10xSTAT92E-GFP combined with a hemocyte driver He-GAL4 as a read-out of the JAK/STAT activity in hemocytes using flow cytometry. As a positive control we overexpressed a constitutively active Drosophila JAK/STAT pathway component hopscotch (hop) in hemocytes (10xStat92E-GFP;He-GAL4/UAS-hop^Tum-l ), and as expected, detected an increase in the Stat92E-GFP signal compared to control hemocytes. We also detected a clear increase in the reporter expression when knocking down the Proteasome α6 subunit (10xStat92E-GFP;He-GAL4/UAS-Prosα6^GD ) compared to control hemocytes expressing only the reporter ( Figure 3A ).

The FSC-area (FSC-A) vs. FSC-height (FSC-H) plot is often used to exclude cell doublets from further flow cytometry analysis taking advantage of the fact that round particles will appear as a population at a 45° angle, while doublets have larger area than height. We observed that while the control animals had mainly round hemocytes (plasmatocytes) (49), appearing at a 45° angle in the FSC-A vs. FSC-H plot, Prosα6 silencing resulted in cells detected outside of this area, similar to when hop^Tum-l was overexpressed in hemocytes ( Figure 3B ). Overexpression of hop^Tum-l is known to cause hematopoietic neoplasia leading to the formation of melanotic nodules and a specific hemocyte type, the lamellocyte (6, 42). Furthermore, Avet-Rochex and coworkers (50) identified Prosα6 (then called Pros35) as a suppressor of melanotic nodules in Drosophila larval hemocytes. Therefore, we reasoned that at least a fraction of these hemocytes outside the 45° area might be lamellocytes, which, due to their flat and discoidal morphology (49), fall outside the area where the round plasmatocytes are detected in the control animals. Since lamellocytes can be easily separated from plasmatocytes by morphology, we stained hemocytes from 10xStat92E-GFP;He-GAL4/UAS-Prosα6^GD and 10xStat92E-GFP;He-GAL4/w^GD 3^rd instar larvae with the F-actin stain Phalloidin to visualize hemocyte morphology. We found that hemocytes from the Prosα6 knockdown larvae had more variation in their morphology when compared to the control hemocytes and observed the presence of hemocytes with a lamellocyte morphology ( Figure 3C ).

Next, we analyzed the GFP signal intensity of the Stat92E-GFP reporter and the proportion of fluorescent hemocytes separately in all hemocytes and in the plasmatocyte fraction ( Figure 3B ). In both cases, the GFP signal was significantly higher in the hemocytes where Prosα6 was silenced compared to controls ( Figure 3D , i). Also, the fraction of GFP-positive hemocytes increased significantly in both populations after Prosα6 knockdown ( Figure 3D , ii). These data indicate that silencing Prosα6 leads to activation of JAK/STAT signaling also in vivo.

Prompted by the observation of the appearance of lamellocytes and JAK/STAT reporter activation after Prosα6 silencing in hemocytes, we studied the hemocyte phenotype further. To investigate if silencing of Prosα6 in hemocytes causes hemocyte differentiation, we utilized hemocyte reporter constructs eaterGFP (expressed in plasmatocytes) and msnCherry (expressed in lamellocytes) combined with the Hml^Δ-GAL4 and He-GAL4 to silence Prosα6 in hemocytes and to quantify hemocyte types using a flow cytometer. Flow cytometry analysis of hemocytes showed that while in a wild-type larva, the majority of the hemocyte pool consists of eaterGFP-positive plasmatocytes, the silencing of Prosα6 caused aberrant lamellocyte differentiation and hemocyte activation ( Figure 4A , i-iii). The hemocyte response was stronger with the UAS-Prosα6^GD construct, where the numbers of all identified hemocyte types were increased, including lamelloblasts, the putative lamellocyte precursors, and prelamellocytes ( Figure 4B , i-vi). A similar, but milder, response was seen also with the UAS-Prosα6^KK construct. Furthermore, around 25% of Hml^Δ-GAL4;He-GAL4/UAS-Prosα6^GD larvae exhibited melanotic nodules, while in Hml^Δ -GAL4;He-GAL4/UAS-Prosα6^KK larvae this phenotype was again milder ( Figure 4C , i-ii).

Since silencing Prosα6 in hemocytes led to hemocyte activation and melanotic nodule formation at the larval stage, we further studied the effects of Prosα6 silencing on egg-to-adult development and the lifespan of the flies. We found that while egg-to-pupal development of the animals with Prosα6 silencing in hemocytes was comparable to the controls ( Supplementary Figure 2A ), the pupal eclosion rate was lower in animals expressing UAS-Prosα6^GD in hemocytes, whereas animals expressing UAS-Prosα6^KK eclosed at a normal rate ( Supplementary Figure 2A ). Inspection of the pupal cases revealed melanotic nodules in a fraction of animals with Prosα6 silencing (1% in msnCherry,eaterGFP;Hml^Δ-GAL4;He-GAL4//UAS-Prosα6^KK and 20% in msnCherry,eaterGFP;Hml^Δ-GAL4;He-GAL4//UAS-Prosα6^GD; Supplementary Figure 2B ), similar to the levels observed in larvae ( Figure 4C ). The lifespan of the msnCherry,eaterGFP;Hml^Δ-GAL4;He-GAL4/UAS-Prosα6^GD flies, both males and females, was shorter than that of the control flies, whereas the lifespan of msnCherry,eaterGFP;Hml^Δ-GAL4;He-GAL4/UAS-Prosα6^KK flies was not affected ( Supplementary Figures 2C, D ). When the expression of Prosα6 normalized to RpL32 was measured from hemocytes by qRT-PCR, it was shown that in the Hml^Δ-GAL4;He-GAL4/UAS-Prosα6^GD hemocytes, Prosα6 expression was 43 ± 14% from the respective control (w/Prosα6^GD ), whereas in the Hml^Δ-GAL4;He-GAL4/UAS-Prosα6^KK hemocytes, Prosα6 expression was 52 ± 17% from the control (w/UAS-Prosα6^KK ) hemocytes ( Supplementary Figure 2E ).

To conclude, knocking down Prosα6 in hemocytes causes immune activation characterized by the differentiation and activation of hemocytes and the formation of melanotic nodules, but also decreases the viability of the flies. It is possible that there is a threshold level of silencing needed for the full phenotypic effects, as with the KK line, only the mild activation of hemocytes, but not the other phenotypes, is seen.

To further verify that Prosα6 silencing induces hemocyte activation via JAK/STAT signaling, we opted for a genetic rescue experiment. We knocked down Prosα6 using the GD construct simultaneously with Stat92E and checked the effect on the hemocyte phenotype. Since also the JNK pathway has been shown to be important in lamellocyte formation (6), we also tested whether it has a role in hemocyte activation in the Prosα6 silencing background. To this end, we knocked down Prosα6 simultaneously with hemipterous (hep), a Drosophila JNK-pathway component and checked the hemocyte composition. Knocking down Stat92E alone in hemocytes resulted in a mild activating effect; a small number of lamelloblasts and lamellocytes were detected in an otherwise wild type background ( Figure 5A ). Knocking down hep led to a reduction in total hemocyte numbers compared to the control ( Figure 5B ). Neither Stat92E nor hep knockdown in the Prosα6-silenced background led to the rescue of total hemocytes, or of plasmatocytes, lamelloblasts or activated plasmatocytes, back to the levels detected in the control animals ( Figures 5A, B , i-iv). Both, however, reduced the formation of prelamellocytes and lamellocytes induced by Prosα6 silencing ( Figures 5A, B , v-vi). Even though knocking down neither Stat92E nor hep in the Prosα6 background rescued the hemocyte phenotype completely to the wild-type levels, the number of lamellocytes produced by the Prosα6 knockdown (600 on average) was reduced approximately seven-fold with Stat92E knock-down (51) and four-fold with hep knockdown (148). Taken together, these results suggest that Prosα6 silencing-induced hemocyte activation requires the function of the JAK/STAT pathway, and that the JNK pathway functions in parallel or downstream of JAK/STAT eliciting hemocyte activation.