Pelle Modulates dFoxO-Mediated Cell Death in Drosophila

Interleukin-1 receptor-associated kinases (IRAKs) are crucial mediators of the IL-1R/TLR signaling pathways that regulate the immune and inflammation response in mammals. Recent studies also suggest a critical role of IRAKs in tumor development, though the underlying mechanism remains elusive. Pelle is the sole Drosophila IRAK homolog implicated in the conserved Toll pathway that regulates Dorsal/Ventral patterning, innate immune response, muscle development and axon guidance. Here we report a novel function of pll in modulating apoptotic cell death, which is independent of the Toll pathway. We found that loss of pll results in reduced size in wing tissue, which is caused by a reduction in cell number but not cell size. Depletion of pll up-regulates the transcription of pro-apoptotic genes, and triggers caspase activation and cell death. The transcription factor dFoxO is required for loss-of-pll induced cell death. Furthermore, loss of pll activates dFoxO, promotes its translocation from cytoplasm to nucleus, and up-regulates the transcription of its target gene Thor/4E-BP. Finally, Pll physically interacts with dFoxO and phosphorylates dFoxO directly. This study not only identifies a previously unknown physiological function of pll in cell death, but also shed light on the mechanism of IRAKs in cell survival/death during tumorigenesis.


Introduction
Interleukin-1 receptor-associated kinases (IRAKs) are evolutionarily conserved Serine/Threonine (Ser/Thr) kinases that modulates Toll-like receptors (TLRs) and interleukin-1 receptors (IL-1Rs) signaling in the immune and inflammation response [1][2][3]. IRAKs transduce TLR/IL-1R signals to regulate subcellular distribution and activity of NF-κB transcription factor family via highly conserved downstream signaling molecules, including IκB kinase (IκK) complex and IκB [4,5]. While there are four IRAK family members (IRAK-1, IRAK-2, IRAK-4 and IRAK-M) in mammals, Pelle (Pll) is the only IRAK homolog in Drosophila [6]. Although Tube was initially characterized as a Drosophila IRAK homolog, it lacks the catalytic domain and thus functions as an adaptor protein in the Toll pathway [7]. As such, Drosophila melanogaster offers unique opportunities to investigate the physiological functions of IRAK with reduced genetic redundancy.
The Toll pathway in Drosophila has been implicated as a crucial regulator in embryonic Dorsal/Ventral (D/V) patterning [8,9], innate immune response against Gram-positive bacteria and fungi [10], proper muscle development and axon guidance [11,12]. This pathway is equivalent to the mammalian IL-1R/TLR pathway, with Toll/Cactus/Dorsal representing the respective homologs of IL-1R/IκB/NF-κB [13]. However, the Drosophila IκK is not involved in the Toll pathway [14,15], and Pll has been shown to be the only kinase implicated in Cactus phosphorylation [6,16]. Besides its role in the conserved pathway, it remains unknown whether Pll performs any Toll pathway independent functions in vivo.
Programmed cell death (PCD) or apoptosis is an important physiological regulator for tissue homeostasis and correct development. Cancer cells generally obtain the ability to evade cell death signaling pathways, which act as a protective tumor suppressive mechanism to remove damaged cells from the tissue. The central roles of caspases (cysteine aspartate-specific proteinases) have been widely studied in the context of apoptosis, and the core machinery of caspases signaling pathway, including the initiator caspases and the effector caspases, has been conserved throughout evolution [17,18]. In Drosophila, apoptosis is activated by the intrinsic and extrinsic cell death signals through transcriptional up-regulation of three pro-apoptotic genes reaper (rpr), head involution defective (hid) and grim [19]. Rpr, Hid and Grim (RHG) proteins subsequently bind to and antagonize the Drosophila inhibitor of apoptosis protein 1 (DIAP1), which in turn dampens caspases [20,21]. DIAP1 functions as an E3-ubiquitin ligase to block cell death by targeting the initiator caspase Dronc (Drosophila caspase-9) for degradation, and thus suppresses the activation of the downstream effector caspase Drice (Drosophila caspase-3) [22,23]. Hence, apoptosis is the consequence of a train of complex interactions between RHG proteins, DIAP1 and caspases in Drosophila.
Forkhead Box O (FoxO) transcription factors were originally identified in C. elegans as DAF-16 for its role in aging, and have been highly conserved in Drosophila (dFoxO) and mammals (FoxO1, FoxO3a, FoxO4 and FoxO6) [24][25][26][27][28]. FoxO proteins are known to modulate various physiological processes including cell proliferation and differentiation, DNA repair, apoptosis, metabolism, oxidative stress and lifespan [25,26,[28][29][30][31][32][33]. In general, the multiple functions of FoxO transcription factors are regulated by various post-translational modifications (PTMs), including methylation, acetylation, ubiquitination, glycosylation and phosphorylation [34]. For example, FoxOs are targeted for phosphorylation by a number of kinases in response to different cellular stresses, which in turn regulate the nuclear-cytoplasmic shuttling of FoxOs and their transcriptional activity [34]. In Drosophila, dFoxO was reported as a key modulator in the UV-induced caspase-mediated apoptotic response by directly regulating the expression of pro-apoptotic gene hid [30].
In this study, we examined the physiological function of pelle (pll) in Drosophila development, and found that pll negatively regulates FoxO-mediated apoptotic cell death in vivo. First, loss of pll in the developing wing generates Toll/NF-κB pathway independent phenotypes, which are caused by a reduction in cell number but not cell size. Secondly, depletion of pll triggers caspase-mediated cell death, whereas cell proliferation remains unaffected. Thirdly, pll modulates caspase activation and cell death through dFoxO. Fourthly, loss of pll results in activation of dFoxO, as manifested by its translocation from cytoplasm to nucleus and the expression of its target gene Thor/4E-BP. Finally, Pll physically interacts with and phosphorylates on dFoxO, which likely contributes to the cytoplasmic retention of dFoxO. Therefore, we have identified a previously uncharacterized physiological function of pll in cell death, which is independent of the canonical Toll/NF-κB pathway.

Results
Loss of pll produces Toll pathway-independent wing phenotype As a crucial component of the Toll pathway, Pll is known to regulate the dorsal-ventral polarity in early embryos [35] and the immune response in fat body [10]. Despite its ubiquitous lower level expression throughout development [6], the post-embryonic functions of Pll remain poorly understood. To explore the physiological function of pll in late development, we employed the UAS/Gal4 binary system to knockdown pll in a specific temporal and spatial manner. We found that expression of a pll RNA interference (RNAi) driven by ptc-Gal4 (ptc>pll-IR) along the anterior/posterior (A/P) compartment boundary resulted in a consistent loss of anterior cross-vein (ACV) phenotype (Fig 1C and 1M), compared with ptc-Gal4 control ( Fig 1A) or RFP expression (Fig 1B). To exclude the possibility that the phenotype is a result of RNAi's off-target effect, we examined two additional pll RNAi that target distinct regions of the pll transcript, and observed the same loss-of-ACV phenotype (S1A, S1F and S1H Fig). In addition, this phenotype could be rescued by expression of Pll (Fig 1D and 1M; S1G and S1H Fig), which by itself has no effect on ACV development (Fig 1E and 1M). The knock-down efficiencies of three pll RNAi lines were verified by their abilities to suppress the GMR>Pll rough eye phenotype (S2A-S2E Fig) and by real-time quantitative reverse transcription polymerase chain reaction (qRT-PCR) assay (S2L Fig). To further verify the physiological function of pll in ACV development, we examined two pll mutant alleles: pll 2 and pll 7 . While the heterozygous mutants show no obvious ACV defect, they could significantly increase the percentage of lossof-ACV phenotype in the ptc>pll-IR W background (S1A-S1E and S1H Fig). Together these data suggest that pll is physiologically required for ACV development in Drosophila wing.
Pll has been implicated in the Toll/NF-κB pathway that directly phosphorylates the Drosophila IκB factor Cactus [36,37]. To examine whether the Toll/NF-κB pathway is required for the development of ACV, we down-regulated the pathway by knocking-down the positive regulators such as Toll, tube, dorsal and Dif, or ectopically expressing the negative regulator Cactus. At least two independent RNAi lines were tested for each gene, including a previously reported RNAi for Toll [38] and tube [39]. The knock-down efficacies of the RNAi lines were verified by their ability to suppress the rough eye phenotype of GMR>Toll 10B (S2F- S2K Fig), which expresses the activated form of Toll in the developing eye. Intriguingly, down-regulation of the Toll/NF-κB pathway failed to recapitulate the loss-of-ACV phenotype produced by ptc>pll-IR (Fig 1F-1J and 1M), and this phenotype was not rescued by RNAi inactivation of cactus (Fig 1K-1M), the Drosophila I-κB homolog, implying that the physiological function of pll in wing development is independent of the canonical Toll/NF-κB pathway.
pelle regulates cell number, but not cell size in wing development To further characterize the physiological function of pll in wing development, we knocked down pll in distinct regions of the wing disc by using additional wing specific Gal4 drivers: Optomotor-blind (Omb)-Gal4, Scalloped (Sd)-Gal4 and engrailed (en)-Gal4. We noted that loss of pll in the distal part (Omb>pll-IR), wing pouch (Sd>pll-IR) or posterior compartment (en>pll-IR) of the wing disc caused severe reduction in corresponding areas of the adult wing, which were rescued by co-expression of Pll, but not that of GFP (Fig 2A-2N  To investigate the underlying mechanism by which loss of pll results in reduced wing tissue, we compared the cell number and cell size in the posterior (P) compartment with that in the anterior (A) compartment of the adult wing. We found that when pll was specifically knockdown in the P compartment by en-Gal4 (Fig 2L), the P/A ratio of cell size remained unaffected (Fig 2O), whereas the P/A ratio of cell number and total size decreased significantly ( Fig 2P  and 2Q), suggesting that pll regulates cell number, but not cell size, in wing development. To further verify this conclusion, we generated pll loss-of-function clones in larvae and found that the size of clones, labeled by GFP expression, was significantly smaller than that of wild-type (WT) controls in wing discs (Fig 3A-3C), confirming the role of pll in regulating tissue growth. However, the sizes of cells inside the clones (marked by GFP expression) were similar to that of wild-type controls in fat body (FB) and salivary gland (SG) (Fig 3G and 3H). Furthermore, wing margin bristles in pll knock-down clones, marked by loss of the yellow (y) gene, are not statistically different in size from their wild-type neighbors (Fig 3D-3F). Together, these data indicate that pll modulates tissue growth via regulating cell number but not cell size.

Loss of pll triggers caspase-mediated cell death
Loss of pll results in reduced cell number, which could be a consequence of decreased cell proliferation or increased cell death. To distinguish between the two possibilities, we first checked the cell proliferation rate by phospho-Histone H3 (pH3) staining in wing discs. We found that knock-down pll in the P compartment of wing discs (en>pll-IR) did not affect cell proliferation, as the P/A ratio of pH3 positive cells in such discs was not statistically different from that in control discs (Fig 3I-3K).
Apoptosis plays a crucial role in maintaining tissue homeostasis, making the decision between cell death and survival in response to various intracellular and extracellular stress [36,40]. To examine the role of pll in regulating cell death, we performed acridine orange (AO) staining and TdT-mediated dUTP nick end labelling (TUNEL) assay, both are commonly used to detect apoptosis [41,42]. We found that knock-down pll in the wing disc resulted in increased AO (   Apoptosis in Drosophila is triggered by transcriptional up-regulation of three pro-apoptotic genes rpr, hid and grim, and is mediated by the cleavage and activation of caspases [40]. Consistent with its role in cell death, loss of pll in the wing pouch results in up-regulated transcription of hid and rpr as revealed by X-gal staining of a hid-LacZ and an rpr-LacZ reporters ( . Furthermore, we found that a ubiquitous knock-down of pll (act>pll-IR) was able to activate the transcription of endogenous hid, rpr and grim, as compared to the act-Gal4 control, via executing a real-time qRT-PCR assay (S4M Fig). In addition, depletion of pll leads to enhanced antibody staining for the activated form of Caspase-3 (CC-3, Fig 5G and 5H), a read-out of the initiator caspase Dronc activity [43]. Moreover, we found that the loss-of-ACV phenotype induced by ptc>pll-IR was suppressed partially by the deficiency Df(3L)H99 that deletes rpr, hid and grim (S5A-S5C Fig), and significantly by expressing the inhibitor of apoptosis protein DIAP1 or a dominant-negative form of Dronc (Dronc DN ), but not LacZ (Fig 6). Accordingly, the phenotype of reduced distal-most area in Omb>pll-IR wing is considerably rescued by Df(3L)H99 (S5D- S5F Fig). Thus, we conclude from these data that depletion of pll is sufficient to activate caspase-mediated cell death.

dFoxO is required for loss-of-pll induced cell death
To understand the mechanism by which loss of pll induces caspase-dependent cell death, we considered the transcription factor dFoxO as a putative downstream factor negatively regulated by Pll. Firstly, dFoxO is known to be negatively regulated by other kinase, e.g. Akt [26,28]. Secondly, previous studies reported that dFoxO is required for the apoptotic response and regulates the expression of pro-apoptotic gene hid [30]. Thirdly, we have previously observed a similar loss-of-ACV phenotype resulted from dFoxO expression driven by ptc-Gal4 driver [33]. In support of the assumption, we found that loss-of-pll triggered cell death in the wing pouch, as detected by AO staining (Fig 5B and 5M) and cleaved caspase-3 (CC-3) antibody ( Fig 5H and 5N), was notably blocked by RNAi-mediated depletion of dFoxO, and in heterozygous or homozygous dFoxO Δ94 mutants (Fig 5D-5F and 5J-5N). Furthermore, depletion of pll induced various wing defects are suppressed by loss of dFoxO, either by mutation or expression of RNAi ( Quantifications of adult wing size/wild type (WT) ratio are shown for figures A-D and F-I respectively (n = 10). One-way ANOVA with Bonferroni multiple comparison test was used to compute P-values, significance is indicated with asterisks (*** P<0.001). (O and P) Quantifications of cell size (O) and cell number (P) in wings shown in K and L. The P/A ratio of cell size showed no difference while that of cell number decreased significantly when pll was knocked down in the P compartment by en-Gal4. Unpaired t test was used to calculate statistical significance, indicated with asterisks (*** P<0.001, n = 10 in each group). (Q) Statistic analysis of total size P/A ratio are shown for figures K-N. One-way ANOVA with Bonferroni multiple comparison test was used to compute P-values, significance is indicated with asterisks (*** P<0.001). ns stands for not significant.  Pll. Therefore, we conclude that loss of pll induces dFoxO-dependent cell death in wing development. pll is ubiquitously expressed throughout development [6]. To investigate whether Pll modulates dFoxO-dependent cell death in other tissues, we knocked down pll in the notum by the pnr-Gal4 driver, and observed a loss-of-bristle phenotype in the notum (Fig 8A, 8B and 8H). This phenotype, most likely caused by cell death as expression of the apoptotic gene reaper in the notum produces a similar bristle-ablation phenotype [46], is significantly suppressed by expressing a dFoxO RNAi, or deleting one or both copies of endogenous dFoxO, or expressing Pll, but not a GFP RNAi (Fig 8C-8H). Thus, Pll negatively regulates dFoxO-dependent cell death in a non-tissue-specific manner.

Pll regulates dFoxO subcellular localization and transcriptional activity
To address how Pll modulates dFoxO activity, we first checked whether Pll regulates dFoxO transcription by executing a real-time qRT-PCR assay. We found that a ubiquitous knockdown of pll (act>pll-IR) did not affect the mRNA level of dFoxO, as compared to the act-Gal4 control   (Fig 9A), suggesting that pll does not regulate dFoxO transcription. Previous studies indicate that the nuclear localization of dFoxO is regulated by a series of post-translational modifications, including phosphorylation by different kinases [34]. To examine whether Pll modulates the nuclear-cytoplasmic shuttling of dFoxO, we checked the sub-cellular localization of a dFoxO-GFP fusion protein in the fat body. We found that the nuclear localization of dFoxO-GFP was significantly increased when pll was knocked down in fat body cells by Cg-Gal4 ( Fig  9B-9D), which drives Gal4 expression specifically in fat body and hemocytes under the control of collagen (Cg25C) promoter [47]. To monitor the dFoxO activity directly, we detected the expression of its well-characterized target gene Thor/4E-BP by a Thor-LacZ reporter [26,28,48,49]. We found that Thor-LacZ expression was distinctly induced in the wing pouch by RNAi inactivation of pll under the control of Sd-Gal4 (Fig 9E and 9F). In addition, qRT-PCR assay confirmed that depletion of pll resulted in up-regulated Thor/4E-BP transcription ( Fig 9G). Thus, we conclude that Pll regulates dFoxO subcellular localization and transcriptional activity.

Pll physically interacts with and phosphorylates dFoxO
To investigate the underlying mechanism by which Pll regulates dFoxO activity, we expressed HA-tagged Pll (HA-Pll) and Flag-tagged dFoxO (Flag-dFoxO) in Drosophila S2R+ cells, and examined the phosphorylation of dFoxO through Calf Intestine Phosphatase (CIP) assay. We found that co-expression of Pll resulted in a mobility shift of dFoxO, which was abolished upon CIP treatment (Fig 10A), suggesting that dFoxO could be phosphorylated by Pll. Next, we performed the in vitro kinase assay to confirm the phosphorylation of dFoxO by Pll. Since Pll could be auto-phosphorylated (Fig 10B), to better distinguish the phosphorylated dFoxO from auto-phosphorylated Pll, we divided the full-length dFoxO (1-622aa) into two segments: the N-terminal (NT, 1-304aa) and the C-terminal (CT, 305-622aa). Indeed, bacterially purified GST-Pll was able to phosphorylate dFoxO in vitro, heavily on the NT and lightly on the CT (Fig 10B), suggesting Pll might phosphorylate dFoxO at multiple sites, most of which are located in the N-terminal half of dFoxO. Furthermore, Co-IP assay demonstrated a physical interaction between HA-Pll and Flag-dFoxO in S2R+ cells (Fig 10C). Finally, we performed the GST-pulldown assay and confirmed a direct binding between Pll and dFoxO ( Fig 10D). Together, these data suggest that Pll could phosphorylate dFoxO through direct interaction, thus negatively regulates dFoxO activity by preventing its nuclear translocation.

Discussion
Drosophila melanogaster has emerged as an excellent model organism to study apoptotic cell death and has made significant contribution to understand cell death regulation and its role in development. While mammalian IRAKs function as the mediator of IL-1Rs/TLRs signal transduction in the immune and inflammatory responses [1,2], Pll, the sole Drosophila orthologue of IRAKs, has been implicated as a central regulator of Toll pathway involved in embryonic suppressed by knocking-down dFoxO (D, J and P) or removing one or both copies of dFoxO (E, F, K, L, Q and R), but not by expressing a GFP-IR (C, I and O). In A-F, the lower panels are high magnification of the boxed areas in upper panels. (S-U) Statistical analysis of the ACV phenotype (S) and the adult wing size/wild type (WT) (T and U) as shown in figures A-F, G-L and M-R respectively. One-way ANOVA with Bonferroni multiple comparison test was used to compute P-values, significance is indicated with asterisks (*** P<0.001). ns stands for not significant. dorsal/ventral patterning, innate immune response, muscle development and axon guidance. In this work, we identified a Toll pathway independent function of Pll in modulating caspasemediated cell death in animal development.
Previous studies have suggested that Drosophila wing vein formation is a result of cell fate specification regulated by multiple signaling pathways including Notch, Hedgehog, EGF  (epidermal growth factor) and BMP (bone morphogenetic proteins) pathways [50,51]. In the present study, we found that knock-down pll along the A/P compartment boundary of the developing wing (ptc>pll-IR) resulted in extensive cell death in the wing disc ( Fig 4B) and a loss-of-ACV phenotype in the adult wing (Fig 1C; S1A and S1F Fig), implying a potential role of cell death in vein patterning. Consistent with this notion, the loss-of-ACV phenotype is rescued by blocking apoptotic cell death (Fig 6C-6E; S5B and S5C Fig), suggesting cell death is responsible for the loss of ACV. To investigate whether cell death is able to impede vein patterning, we initiated apoptosis by expressing the pro-apoptotic protein Grim under the control of ptc-Gal4. ptc>Grim caused extensive cell death in tissue ablation between L3 and L4 in the adult wing. In most cases, L3 and L4 were fused in the proximal area where ACV is located (S8A Fig). To adjust Grim expression and cell death, we added Tub-Gal80 ts that represses Gal4 activity in a temperature sensitive manner. At 25°C, Tub-Gal80 ts partially blocks ptc-Gal4 activity and allows limited Grim expression and therefore, cell death, between L3 and L4. Intriguingly, under this condition, we observed the loss-of-ACV phenotype accompanied by a slight reduction of area between L3 and L4 (S8B Fig), suggesting that both reduced area and loss-of-ACV phenotypes are consequences of cell death. From our experience, the loss-of-ACV phenotype is more sensitive to cell death, since weak cell death is sufficient to generate the phenotype, whereas stronger cell death is required to delete tissue between L3 and L4. Consistent with this notion, while ptc>pll-IR flies reared at 25°C only displayed the loss-of-ACV phenotype (Fig 1C), those raised at 29°C also showed reduced area between L3 and L4 (S8D and S8G  Fig), which was caused by a reduction in cell number, but not cell size (S8E and S8F Fig).
We show that Pll regulates caspase activation and cell death through dFoxO. Mechanistically, loss of pll promotes the nuclear translocation of dFoxO, which otherwise is retained in the cytoplasm by phosphorylation. A number of kinases, including AKT, IκK and JNK, have been reported to phosphorylate FoxO and regulate its nuclear-cytoplasmic trafficking [52][53][54].
Here we provide evidence that Pll is another dFoxO kinase that phosphorylates dFoxO and inhibits its nuclear localization. Thus, it would be very interesting to check whether a similar interaction is conserved between IRAKs and FoxOs in mammal.
The FoxO family proteins have been implicated in multiple important biological processes, including cell death and tumor suppression. It has been reported that conditional deletion of FoxO1, FoxO3 and FoxO4 simultaneously results in the development of hemangiomas and thymic lymphomas [55], and IκB kinase represses FoxO3a activity to promote human breast tumorigenesis and acute myeloid leukemia (AML) [52,53]. IRAKs also show altered expression level in tumors and surrounding stroma [43,56], and participate in tumor initiation and progression [57], yet the underlying mechanisms remain poorly understood. Thus, the inhibitory effect of Pll on dFoxO activity in Drosophila provides a beneficial framework for a better understanding of mammalian IRAKs' crucial roles in tumor development.
As the Toll/NF-κB pathway is not implicated in loss-of-pll triggered dFoxO-dependent cell death, we are curious about what are the pathways or factors act upstream of Pll to regulate its role in cell death? Since dFoxO has been reported as a downstream transcription factor in the JNK and Insulin pathways in Drosophila (S9P Fig; S10O Fig) [58][59][60][61], we wondered whether Pll is also involved in these pathways. Activation of JNK signaling between L3 and L4 by expressing Egr (Drosophila TNF) or Hep (Drosophila JNK Kinase), or depleting puc (encoding a JNK inhibitor), produced similar loss-of-ACV and reduced area phenotypes as that of pll depletion, yet the phenotypes were not affected by gain or loss of pll (S9 Fig). Inactivation of the Insulin pathway by expressing a dominant negative form of PI3K, or knocking-down PI3K or Akt, resulted in diminished area between L3 and L4, which remained unaffected by gain or loss of pll (S10 Fig). We also examined the Hippo pathway known to play a crucial role in regulating cell death and organ size [62]. Up-regulation of Hippo pathway in distinct wing areas by different Gal4 drivers led to various small wing or wing tissue ablation phenotypes, which were not altered by changing Pll level (S11 Fig). Finally, we checked dMyc, the fly homolog of c-Myc that regulates cell growth and cell death in Drosophila [63][64][65], and the cell polarity gene scribble (scrib), whose depletion promotes cell death [66][67][68][69]. We found that depletion of dMyc triggered wing phenotype and loss of scrib induced cell death are both independent of Pll (S12 Fig). Thus, while Pll directly regulates dFoxO-mediated caspase-dependent cell death in development, the upstream factors modulating Pll activity remain unknown, which deserve further investigation.

AO staining
Wing discs were dissected from third instar larvae in PBST and incubated in 1×10 −5 M AO for 5 min at room temperature prior to imaging as described [79].

TUNEL staining
The wing discs of wandering third-instar were dissected out in PBS. Wing discs were fixed in 4% Paraformaldehyde for 30 min at room temperature (RT) and washed with PBS-Tx (0.3% Triton100) three times for 30 min. TUNEL staining was performed using the Fluorescein Cell Death Kit produced by Boster Company. Imaging of prepared sample was conducted by a Leica confocal microscope (Leica SP5).

qRT-PCR
Five third instar larvae of indicated genotypes were collected and total RNA was isolated using TRIzol (Invitrogen). qRT-PCR was performed as previously described [81] using following primers:

GST-pulldown assays
The GST-pulldown assays were performed as previous described [83]. Recombinant GST-Pelle-λPPase and 6χHis-dFoxo proteins were produced in BL21 cells and purified with glutathione-Sepharose 4B (GE Healthcare) or Ni-NTA resin (GE Healthcare) respectively according to standard protocols. 2 μg GST or 10 μg GST-fusion proteins was incubated at 4°C for 5 hours with 10 μg purified 6χHis-dFoxo and 50 μl of glutathione-Sepharose beads. Supernatants were collected as input and the Sepharose beads were then extensively washed 5 times with lysis buffer and resuspended in SDS loading buffer and boiled. A quarter of the sample buffer was loaded in 12% SDS-PAGE for detection with anti-His antibody (Sungene Biotech). Unpaired t test was used to calculate statistical significance. ns stands for not significant.

Supporting Information
in reduced wing tissue in the corresponding areas, which were not altered by changing Pll level or expressing LacZ (C-E, I-K, O-Q and U-W). (F, L, R and X) Quantifications of total wing size/wild type (WT) ratio or wing size L3-L4/total ratio are shown for figures A-E, G-K, M-Q and S-W respectively (n = 10). One-way ANOVA with Bonferroni multiple comparison test was used to compute P-values, significance is indicated with asterisks ( ÃÃÃ P<0.001). ns stands for not significant.