We reasoned that the overexpression of a Dilp-binding protein that impinges on the ligand-receptor interaction should counteract the effects of receptor overexpression. dInR overexpression during eye development (by means of a GMR-Gal4 strain, in which the Gal4 protein is overexpressed in photoreceptor neurons, and a UAS-dInR, which expresses dInR when activated by Gal4) results in hyperplasia of the eyes, a phenotype that is sensitive to the levels of the Dilps 17. A collection of enhancer-promoter (EP) elements, which allow the overexpression of nearby genes (F.W., W.B., H.S., D. Nellen, K. Basler and E.H., unpublished work), was screened for suppressors of the dInR-induced hyperplasia (Figure 1a). A strong suppressor (EP5.66, Figure 1b) carried an EP element 8.5 kb upstream of the Imp-L2 coding sequence (Figure 1f). Two different UAS transgenes, both containing the Imp-L2 coding sequence but varying in strength, confirmed that the suppression was caused by Imp-L2. Whereas the weaker UAS-Imp-L2 (containing 5' sequences with three upstream open reading frames) only partially suppressed the dInR-induced overgrowth (Figure 1c), UAS-strong.Imp-L2 (UAS-s.Imp-L2, lacking the 5' sequences) completely reversed the phenotype (Figure 1d). In addition, a point mutation in the Imp-L2 coding sequence (see below) abolished the suppressive effect of EP5.66 (Figure 1e). Imp-L2 is therefore a potent antagonist of dInR-induced growth.

Imp-L2 has previously been shown to be upregulated 8–10 hours after ecdysone treatment 2021. It encodes a secreted member of the immunoglobulin (Ig) superfamily containing two Ig C2-like domains. Whereas several orthologs of Imp-L2 are present in invertebrates such as arthropods and nematodes, the homology in vertebrates is confined to the second Ig C2-like domain, which is homologous to the carboxyl terminus of human IGFBP-7 (Figure 1g). The carboxy-terminal part of IGFBP-7 differs considerably from the other IGFBPs, possibly accounting for the affinity of IGFBP-7 for insulin 6. Interestingly, Imp-L2 has been shown to bind human insulin, IGF-I, IGF-II and proinsulin, and its homolog in the moth Spodoptera frugiperda, Sf-IBP, can inhibit insulin signaling through the insulin receptor 22.

To further assess the function of Imp-L2 as a secreted inhibitor of insulin signaling, we ectopically expressed Imp-L2 using various Gal4 drivers. Strong ubiquitous over-expression of Imp-L2 by Act-Gal4 led to lethality with both UAS transgenes. Whereas driving UAS-s.Imp-L2 by the weaker ubiquitous arm-Gal4 driver also resulted in lethality, driving UAS-Imp-L2 generated flies that were decreased in size and weight (-15% in males and -29% in females, data not shown) but eclosed at the expected ratio and had wild-type appearance. By generating clones of cells that over-express Imp-L2, we confirmed that cell specification and patterning were normal in Imp-L2-overexpressing ommatidia (Figure 2a). However, a reduction of cell size was observed in the clones. This reduction seemed to be non-autonomous because wild-type ommatidia close to the clone were also reduced in size. Given the convex nature of the eye we were unable to quantify the effects of Imp-L2 overexpression on more distantly located ommatidia. Eye-specific overexpression of both UAS-Imp-L2 and UAS-s.Imp-L2 by GMR-Gal4 led to a strong reduction in eye size (data not shown). Whereas the GMR-Gal4, UAS-Imp-L2 flies were of normal size, body weight was reduced by 38.3% and development was delayed by one day in GMR-Gal4, UAS-s.Imp-L2 male flies (Figure 2b). Next, we used the ppl-Gal4 driver to over-express Imp-L2 in the fat body, a tissue that can be expected to produce and secrete Imp-L2 more efficiently than the eye. Driving UAS-s.Imp-L2 by ppl-Gal4 was lethal, whereas ppl-Gal4, UAS-Imp-L2 flies showed a pronounced reduction in body size (Figure 2c) and were delayed by 2 days. Both the size decrease and the developmental delay are characteristic phenotypes of reduced IIS such as in chico mutants 23, supporting the hypothesis that Imp-L2 acts as a secreted negative regulator of this pathway.

Next, we assessed the effect of Imp-L2 overexpression on phosphatidylinositol(3,4,5)trisphosphate (PIP₃) levels using a green fluorescent protein-pleckstrin homology domain fusion protein (tGPH) that specifically binds PIP₃ and serves as a reporter for PIP₃ levels in vivo 24. The amount of membrane-bound tGPH reflects signaling activity in the phosphoinositide 3-kinase/protein kinase B (PI 3-kinase/PKB) pathway. Overexpression of dInR resulted in a severe increase of membrane PIP₃ levels (Additional data file 1, Figure S1A,B). Co-overexpression of Imp-L2 together with dInR reduced the PIP₃ levels (Additional data file 1, Figure S1D), similar to the effect caused by PTEN (Additional data file 1, Figure S1C), a negative regulator of IIS. Therefore, Imp-L2 inhibits PI 3-kinase/PKB signaling upstream of PIP₃, without affecting dInR levels (Additional data file 1, Figure S1B',D').

We used two strategies to generate loss-of-function mutations in Imp-L2. First, we performed an ethylmethane-sulfonate (EMS) reversion screen in which we selected mutated chromosomes carrying EP5.66 that no longer suppressed the dInR overexpression phenotype (Figure 1a). One allele (Imp-L2^MG2) containing a point mutation resulting in a premature stop at amino acid 232 was identified in this way (Figure 1e,f). This truncation destroys the conserved cysteine bridge of the second Ig domain (Figure 1g). Overexpression of the truncated Imp-L2 version had no inhibitory effect on size (Figure 1e), suggesting that Imp-L2^MG2 is a functional null allele.

Second, we generated additional Imp-L2 alleles by imprecise excision of GE24013 (GenExel), a P-element located 349 bp upstream of the ATG start codon of the Imp-L2-RB transcript (Figure 1f). We obtained Imp-L2 deletions (Def20, Def42) lacking the entire coding sequence. Heteroallelic combinations of the mutant alleles increased body size: whereas mutant males showed a 27% increase in body weight, mutant females were 64% heavier (Figure 2d,e). Introducing one copy of a genomic rescue construct (Figure 1f) 25 into homozygous mutant flies reverted the weight to the level of Imp-L2^+/- flies, which were already heavier (+14% in males, +44% in females, Figure 2e) than the controls. By measuring the cell density in the wing, the size increase could be attributed primarily to an increase in the number of cells, because cell size was only slightly affected (Figure 2e). Apart from the size increase, the flies lacking Imp-L2 appeared completely normal, eclosed with the expected frequency and were not delayed. Thus, under standard conditions, Imp-L2 loss-of-function dominantly increases growth by augmenting cell number without perturbing patterning, developmental timing or viability.

The weight difference was more pronounced in mutant females than in males, although the increases in wing area and cell number were similar (Figure 2e and data not shown). This differential effect was caused by enlarged ovaries in Imp-L2 mutant females (data not shown).

The facts that Imp-L2 is a secreted protein and that removal of Imp-L2 function did not rescue either chico or PI3K mutant phenotypes (data not shown) are consistent with the hypothesis that Imp-L2 acts upstream of the intra-cellular IIS cascade at the level of the ligands. Immunohistochemistry in larval tissues revealed that, besides strong expression in corpora cardiaca (CC) cells (Figure 3a and Additional data file 1, Figure S2D), Imp-L2 protein was also weakly expressed in the seven m-NSCs that produce Dilp1, Dilp2, Dilp3 and Dilp5 (Figure 3b) and project their axons directly to the subesophageal ganglion, the CC, the aorta and the heart 1926. Thus, Imp-L2 potentially interacts with some of the Dilps directly at their source. We therefore tested for genetic interactions of Imp-L2 with the dilp genes. A deficiency (Df(3L)AC1) uncovering dilp1-5 not only dominantly suppressed the dInR-mediated big eye phenotype 17, but also dominantly enhanced the small eye phenotype caused by eye-specific overexpression of Imp-L2 (Additional data file 1, Figure S3). dilp2 is the most potent growth regulator of all dilp genes 18. Weak ubiquitous overexpression of dilp2 by arm-Gal4 caused an increase in body and organ size 18, and this phenotype was dominantly enhanced by heterozygosity for Imp-L2 (Figure 3c). In homozygous Imp-L2 mutants, expression of dilp2 under the control of arm-Gal4 caused lethality, reminiscent of strong dilp2 expression 18. Expressing Imp-L2 and dilp2 individually at high levels in the fat body also caused lethality, but coexpression resulted in viable flies of wild-type size (Figure 3d). Thus, Imp-L2 decreases the sensitivity to high insulin levels and is sufficient to rescue the lethality resulting from dilp2-induced hyperinsulinemia.

It has previously been shown that Imp-L2 can bind human insulin and insulin-related peptides 22. To address whether Imp-L2 binds Dilp2, we constructed a Flag-tagged version of Dilp2, which is functional (data not shown). Using in vitro translated, ³⁵S-labeled Imp-L2 together with Flag-Dilp2 extracted from stably transfected S2 cells, we could show that Imp-L2 binds Dilp2 in vitro (Figure 3e). A truncated form of Imp-L2 lacking a functional second Ig domain (like that produced by the MG2 allele) failed to bind Dilp2 (Figure 3e).

Despite being a potent inhibitor of Dilp2 action, Imp-L2 is not essential under standard conditions. Hyperactivation of the dInR pathway leads to increased accumulation of nutrients in adipose tissues, precluding them from circulating and thus resulting in starvation sensitivity at the organismal level 24. We therefore tested whether Imp-L2 functions as an inhibitor of IIS under stress conditions. We exposed wild-type and Imp-L2 mutant early third instar larvae to various starvation conditions and scored for survival. Larvae lacking Imp-L2 showed a massive increase in mortality rate when exposed to 1% glucose or PBS for 24 hours (Figure 4c). To test whether the inability of the mutant larvae to cope with starvation was due to a failure in adjusting IIS, we monitored PIP₃ levels under these conditions. Whereas control flies showed a decrease of PIP₃ levels when exposed to complete starvation for 4 hours (Figure 4a), Imp-L2 mutant larvae still contained PIP₃ levels that were comparable to those of control larvae reared on normal food (Figure 4b), suggesting that Imp-L2 is necessary to adjust IIS under starvation conditions. The fact that PIP₃ levels were also slightly reduced in Imp-L2 mutants upon starvation could be attributed to the downregulation of dilp3 and dilp5 at the transcriptional level 18.

The dampening of IIS upon starvation could be achieved either by enhanced secretion of stored Imp-L2 or by an upregulation of Imp-L2 production. Indeed, expression profiling revealed a slight upregulation of Imp-L2 after 12 hours complete starvation 27. We could not detect a change in Imp-L2 protein expression in the brain, the ring gland or the gut after complete starvation for 24 hours (data not shown). However, Imp-L2 was induced in fat body cells, where it appeared in vesicle-like structures (Figure 4d). Thus, under adverse nutritional conditions, Drosophila larvae weaken IIS by upregulating Imp-L2 expression in the fat body.

The following fly stocks and transgenes have been used: y w; w¹¹¹⁸; arm-Gal4; Act5C-Gal4; UAS-GFP; UAS-lacZ (all from the Bloomington Drosophila stock center); GMR-Gal4 (a gift of M. Freeman); ppl-Gal4 (a gift of M. Pankratz); UAS-dInR 17; Df(3L)AC1 17; tGPH 24; GMR>w⁺>Gal4 17; UAS-dPTEN 31; UAS-dilp2 17; GE24013 (GenExel). All crosses were performed at 25°C unless stated otherwise.

The EP screen that led to the identification of Imp-L2 will be described elsewhere (F.W., W.B., H.S., D. Nellen, K. Basler and E.H., unpublished work). A double-headed EP element (containing ten Gal4-binding sites at each end) suppressing the GMR-Gal4, UAS-InR big eye phenotype was identified in the Imp-L2 locus. Plasmid rescue of EP5.66 revealed that it was inserted 6,969 bp upstream of the first exon of the Imp-L2-RB (CG15009-RB) transcript.

To obtain loss-of-function alleles of Imp-L2, we performed an EMS mutagenesis screen in which we selected mutated chromosomes carrying EP5.66 that could no longer suppress the dInR overexpression phenotype in the eye. EP5.66 males were fed with 25 mM EMS and subsequently crossed to GMR-Gal4, UAS-dInR virgins. 39,000 F1 flies were screened for a reversion of the suppressive effect of EP5.66 on the growth phenotype caused by GMR-Gal4, UAS-dInR. Only one of the identified reversion lines, Imp-L2^MG2, could be confirmed. Sequencing the genomic DNA of Imp-L2^MG2 revealed a point mutation that resulted in a truncation (Trp232Stop).

In order to generate additional Imp-L2 mutants, the P-element GE24013 (marked with white⁺) inserted 102 bp upstream of the first exon of the Imp-L2-RC transcript was mobilized by supplying Δ2–3 transposase. Jump starter males were mated with balancer females, and single F1 w^- males were recrossed to balancer virgins. Stocks (350) were established and molecularly tested for deletions by single-fly PCR using several primer pairs, leading to the identification of the alleles Imp-L2^Def42, Imp-L2^Def20, Imp-L2^Def35, Imp-L2^Def223 and Imp-L2^Def29.

In order to generate the UAS-Imp-L2 construct, a BglII/XhoI fragment of Imp-L2 was excised from the Imp-L2-RB containing cDNA clone LP06542 and inserted into pUAST 32. To obtain UAS-s.Imp-L2, the second and third exons of Imp-L2 were amplified by PCR from genomic DNA. The fragment was subcloned into pCRII-Topo (Invitrogen). The insert was then excised with EcoRI and cloned into pUAST 32. Because of the lack of the first exon of the Imp-L2-RB transcript (containing three upstream open reading frames), UAS-s.Imp-L2 has a stronger phenotype than UAS-Imp-L2. The EP element contains ten UAS sites, whereas the UAS transgenes contain only five.

For the generation of the genomic rescue construct, the genomic fragment L2G314 (kindly provided by J. Natzle) was used. The fragment (5 kb of genomic sequence upstream of the first exon of the Imp-L2-RB transcript and 1 kb downstream of the third exon) was excised with BamHI and Asp718 and inserted into the pCaSpeR-4 transformation vector 33.

The Flag-dilp2 construct was created by PCR amplification of the dilp2 coding sequence without the signal peptide sequence from the full-length cDNA clone, EST GH11579 (obtained from Research Genetics). The resulting PCR product was then equipped with the hemagglutinin signal peptide sequence and a Flag tag and inserted into pUAST 32.

Drosophila embryonic S2 cells were grown at 25°C in Schneider's Drosophila medium (Gibco/Invitrogen) supplemented with 10% heat-inactivated fetal-calf serum (FCS), penicillin and streptomycin.

For the construction of the stably expressing Flag-dilp2 cell line, S2 cells were co-transfected with UAS-Flag-dilp2, Act-Gal4 and a third vector containing a blasticidin-resistance gene, using effectene transfection reagent (Qiagen). Two days after the transfection, the selection medium (Schneider's containing 10% FCS and 25 μg/ml blasticidin) was added to the cells. After 10 days the selection medium was replaced by Schneider's containing 10% FCS and 10 μg/ml blasticidin.

S2 cells expressing Flag-dilp2 were grown to confluence in 175 cm² culture flasks, washed with ice-cold PBS and extracted in immunopreciptiation (IP) buffer (120 mM NaCl, 50 mM Tris pH 7.5, 20 mM NaF, 1 mM benzamidine, 1 mM EDTA, 6 mM EGTA, 15 mM Na₄P₂O₇, 0.5% Nonidet P-40, 30 mM β-glycerolphosphate, 1× Complete Mini protease inhibitor (Roche)). After incubation for 15 min on an orbital shaker at 4°C, solubilized material was recovered by centrifugation at 13,000 rpm for 15 min and supernatants were collected. Anti-Flag antibody (5 μg, Sigma M2, F3165) was added and incubated over night at 4°C while rotating. Protein G sepharose beads (Amersham Biosciences) were added for 2 h and the beads were washed four times with IP buffer. Cell lysate from native S2 cells was subjected to the same procedure and the resulting beads were used as control. To verify the immunoprecipitation, a fraction of the beads was incubated with SDS loading buffer (62.5 mM Tris-HCl pH 6.8, 20 mM DTT, 2% SDS, 25% glycerol, 0.02% bromophenol blue) for 5 min at 90°C and the proteins were separated by SDS-PAGE. The presence of Flag-Dilp2 was confirmed by immunoblotting.

For the in vitro translation the Imp-L2-RC cDNA (SD23735) was cloned into pCRII.1 (Invitrogen) downstream of the SP6 polymerase promoter. As a control, the point mutation encoding a non-functional, truncated version of Imp-L2 (identified in the EMS reversion mutagenesis) was inserted into Imp-L2-RC (in pCRII.1 see above) using the Quick-Change site-directed mutagenesis protocol (Stratagene). Both the Imp-L2 and the Imp-L2^MG2 constructs were translated in vitro using the TNT Quick coupled transcription/translation system (Promega) according to the manufacturer's protocol. Briefly, 2 μg of DNA was incubated with 20 μCi [³⁵S]methionine and 20 μl TNT Quick Master Mix in a total volume of 25 μl for 90 min at 30°C. The product (2.5 μl) was used in the in vitro pulldown assay together with Flag-Dilp2 bound to beads or with control beads in IP buffer containing 0.05% NP-40. The reaction was rotated overnight at 4°C, the beads were washed six times with IP buffer (0.05% NP-40) and incubated with SDS loading buffer containing 100 mM DTT for 10 min at 80°C. The dissociated proteins were separated using SDS-PAGE and detected by autoradiography.

Freshly eclosed flies were collected, separated according to sex, placed on normal fly food for 3 days and anesthetized for 1 min with ether before weighing. Weight was determined using a Mettler Toledo MX5 microbalance. Wing size was analyzed as described 34. ImageJ 1.32j software was used to determine the pixels of the wing area. Scanning electron microscope pictures were taken from adult flies that were critical-point dried and coated with gold.

Heat-shock induced overexpression clones (y, w, hs-Flp; GMR>w⁺>Gal4) were induced 24–48 h after egg-laying by a 1 h heat shock at 37°C. Tangential sections of adult eyes were generated as described 35.

For all starvation experiments, eggs were collected for 2 h on apple agar plates supplemented with yeast. After 72 h, larvae were quickly washed in PBS and transferred either to a new apple agar plate with yeast (normal food, called 'yeast' henceforth), a solution containing 20% glucose in PBS, or a filter paper soaked with 1% glucose in PBS or PBS only. After 24 h, dead larvae were counted.

For the tGPH reporter analysis under starvation, the 'PBS' or 'yeast' conditions were used (see above). After 4 h starvation, larvae were dissected in PBS, fixed and stained with Hoechst. Pictures were taken using a Leica SP2 confocal laser scanning microscope.

The antibody against Imp-L2 was described earlier 25 and kindly provided by J. Natzle (Department of Molecular and Cellular Biology, University of California, Davis, USA). Antibody staining against Imp-L2 was performed using the following dilutions: rat anti-Imp-L2 (1:500), donkey anti-rat-FITC (1:200, Jackson). Other antibodies used were: anti-β-galactosidase (1:2,000, polyclonal, rabbit), an antibody against the carboxyl terminus of dInR (INRcT, 1:10,000) 36. Nuclei were either stained with 4',6-diamidino-2-phenylindole (DAPI) or Hoechst. Pictures were taken using a Leica SP2 confocal laser scanning microscope.

RNA in situ hybridization using digoxigenin-labeled probes was performed as described 17. The probes against Imp-L2 were derived from s.Imp-L2 in a pBluescript SK⁺ vector.