A novel function of Drosophila eIF4A as a negative regulator of Dpp/BMP signalling that mediates SMAD degradation

Abstract

Signalling by the TGF-β superfamily member and BMP orthologue Decapentaplegic (Dpp) is crucial for multiple developmental programmes and has to be tightly regulated. Here, we demonstrate that the Drosophila Dpp pathway is negatively regulated by eukaryotic translation initiation factor 4A (eIF4A), which mediates activation-dependent degradation of the Dpp signalling components Mad and Medea. eIF4A mutants exhibit increased Dpp signalling and accumulation of Mad and phospho-Mad. Overexpression of eIF4A decreases Dpp signalling and causes loss of Mad and phospho-Mad. Furthermore, eIF4A physically associates with Mad and Medea, and promotes their degradation following activation of Dpp signalling in a translation-independent manner. Finally, we show that eIF4A acts synergistically with, but independently of, the ubiquitin ligase DSmurf, indicating that a dual system controls SMAD degradation. Thus, in addition to being an obligatory component of the cap-dependent translation initiation complex, eIF4A has a novel function as a specific inhibitor of Dpp signalling that mediates the degradation of SMAD homologues.

To understand the regulation of Dpp signalling, we have previously identified a dominant-negative mutation in eukaryotic translation initiation factor 4A (eIF4A), which acts as a suppressor of dpp haploinsufficiency¹. This allele, eIF4A^R321H (also known as eIF4A^YE9), is associated with an increased number of amnioserosa cells¹, the fate of which is determined by Dpp in the early Drosophila embryo². To investigate whether eIF4A mutations cause increased Dpp signalling in general, we examined the effects of eIF4A mutations on Dpp signalling in other developmental or genetic contexts. First, we found that eIF4A^R321H and eIF4A¹⁰⁰⁶ (a null allele^1,3) dominantly suppressed the sterility of females that were doubly heterozygous for null mutations of Mad and saxophone (sax) (see Methods), which encode homologues of mammalian Smad1/5/8 and type I activin receptor, respectively^4–8. Embryos that are produced by Mad sax¹/+ female flies die with a partially ventralized phenotype (data not shown; also see refs 5, 6). In contrast, Mad sax¹/eIF4A^R321H and Mad sax¹/eIF4A¹⁰⁰⁶ females laid morphologically normal eggs that hatched to viable adult progeny (Table 1), indicating that reducing the amount of eIF4A can compensate for a reduced maternal Mad and sax dosage for embryonic viability. Second, eIF4A^R321H and eIF4A¹⁰⁰⁶ dominantly suppressed the partial lethality and ‘thick vein’ phenotype that was associated with reduced levels of the type I receptor Thickveins^6,9 (Fig. 1a; Table 1), indicating that the effect of eIF4A mutations on Dpp signalling is not limited to embryogenesis. Third, transient ectopic expression of dpp induced by mild heat shock was tolerated in wild-type flies, but resulted in lethality in a eIF4A^R321H/+ background, indicating that eIF4A^R321H dominantly enhances the effects of dpp ectopic expression (Table 1). eIF4A^R321H heterozygosity similarly enhanced the effects of overepxressing dpp by the eye-specific GMR–Gal4 (see below and Fig. 1h). Therefore, eIF4A mutations seem to augment the Dpp signalling strength.

Table 1.

Mutations in eIF4A augment Dpp signalling

Genotype	Viability (n)	Egg hatching rate (n)	Phenotype
Mad sax1/+	100%	0% (209)	Eggs weakly ventralized
Mad sax1/eIF4A1006	100%	24% (195)	Eggs normal
Mad sax1/eIF4AR321H	100%	95% (187)	Eggs normal
tkv1/tkv4	23% (230)	ND	Wing veins thickened
tkv1/tkv4 eIF4A1006	94% (304)	ND	Minor wing-vein defects
tkv1/tkv4 eIF4AR321H	101% (226)	ND	Wing veins normal
hs–Gal4/+;UAS–dpp/+ (plus heat-shock)	98% (203)	ND	Ectopic wing veins
hs–Gal4/+; eIF4AR321H/+; UAS–dpp/+ (plus heat-shock)	0% (349)	ND	Lethality
hs–Gal4/+; UAS–eIF4A/+; UAS–dpp/+ (plus heat-shock)	100% (128)	ND	Wing veins normal

The viability of flies of a particular genotype was determined by comparing with siblings from the same cross from which they were derived. The egg hatching rate was calculated for eggs produced by viable females of the indicated genotypes that were mated with wild-type males. Percentage viability was calculated by comparing with siblings without inherited eIF4A mutations or transgenes from the same cross. n indicates the number of embryos or adult flies examined. ND, not determined. Heat-shock was administered by incubating pupae at 37 °C for 30 min at 18 h after puparium formation.

Figure 1.

Levels of eIF4A affect Dpp signalling and Mad protein levels. (a) Adult wings of different genotypes are shown anterior side up. Note the ‘thick veins’ (arrows) in the tkv¹/tkv⁴ wing and the suppression by eIF4A¹⁰⁰⁶ or eIF4A^R321H. (b–e) Whole-mount embryos stained with anti-pMad (brown in b, green in c), dpp antisense RNA (blue in d) and anti-Mad (brown in e) are shown anterior to the left and dorsal side up. (b) Higher levels of pMad signals in the dorsal epidermal cells (arrowhead) and posterior midgut (arrow) of stage 10 eIF4A^R321H embryos (right) than in wild-type (WT) embryos (left). (c) At stage 13, pMad signals were detected in parasegment 7 (ps7), gastric caeca (gc) and the epidermis (epd) wild-type embryos (left). In eIF4A^R321H embryos, pMad signals were detected at higher levels and expanded domains in all these tissues, such that pMad staining was seen in significantly larger areas of in gc and ps7 and broader spatial extent in the epidermis (right). (d) Increased dpp transcription is indicated by wider domains of dpp mRNA signals in gc and ps7 in eIF4A^R321H embryos (right). (e) Little or no Mad protein is detected in ps7 of wild-type embryos (left), but is detected in discrete cells of ps7 of eIF4A^R321H embryos (right). (f–g) Adult presumptive gut tissues of the 3^rd instar larvae were doubly stained for GFP (green; marks nuclei of eIF4A⁺ cells), and pMad (magenta in f) or Mad (red in g) are shown partially. Note that increased levels of pMad and Mad were detected within eIF4A¹⁰⁶⁹ mutant cells (lack of green nuclei). Arrows point to nuclei lacking GFP. (h) Expressing dpp by GMR–Gal4 resulted in rough and bulging eyes (left), which was completely suppressed by coexpressing eIF4A⁺ (centre). Expressing dpp by GMR–Gal4 in a eIF4A^R321H/+ background resulted in lethality (4% viability; n = 188). A small number of survivors exhibited more severe roughness and outgrowth of eye tissues (right). (i) Expressing dominant-negative Put or eIF4A by the wing-margin-specific C96–Gal4. (j) Loss of pMad in wing-margin primordial cells (arrow) of the 3^rd instar imaginal discs stained with anti-pMad (green). All scale bars equal 100 μm.

To investigate at which step loss of eIF4A increases Dpp signalling, we examined the levels of active or phosphorylated Mad (pMad). In early wild-type embryos, pMad signals are detected in spatial and temporal patterns that are correlated with dpp expression¹⁰. At stage 10, pMad signals had dissipated from the procephalon and posterior midgut of wild-type embryos (Fig. 1b; also see refs 10, 11), but were still detected in eIF4A^R321H embryos (Fig. 1b) — a phenotype that is very similar to that caused by loss of DSmurf¹¹ — indicating a prolonged duration of Dpp signalling. Pronounced elevation and expansion of pMad signals were also detected in stage 11 (see Supplementary Information, Fig. S1) and stage 13 eIF4A^R321H embryos (Fig. 1c). Consistent with the higher levels and expanded domains of pMad signals, we found that dpp expression in gc and ps7, which is subject to positive autoregulation, was also expanded in eIF4A^R321H embryos (Fig. 1d). As we detected expanded domains but little or no ectopic pMad, and the initial pattern of dpp or pMad at stages 5–6 was not obviously changed in eIF4A^R321H embryos (data not shown; see Supplementary Information, Fig. S1), the presence of higher levels and expanded domains of pMad signals in later embryonic stages indicate a prolonged Mad activation in eIF4A^R321H embryos.

To understand the cause of the prolonged Mad activation, we examined Mad protein levels in eIF4A^R321H embryos. Mad protein levels are generally low in all tissues, including ps7, in wild-type embryos (Fig. 1e; left). However, in eIF4A^R321H embryos, elevated levels of Mad protein were found in ps7 (Fig. 1e; right). The observed accumulation of Mad protein is consistent with the idea that Mad protein levels are normally negatively controlled by protein degradation, and eIF4A mutations disrupt this process. Interestingly, the higher levels of Mad protein that were detected in eIF4A^R321H embryos seemed to be nuclear localized, indicating that eIF4A^R321H may interfere with the degradation of activated Mad.

To investigate whether eIF4A mutant cells autonomously increase Mad phosphorylation and protein levels, we generated cell clones that were mutant for a weak allele of eIF4A, eIF4A¹⁰⁶⁹, as cells that are mutant for strong eIF4A alleles are not viable³. We examined multiple larval tissues and found increased pMad and Mad levels associated with eIF4A¹⁰⁶⁹ mutant clones, most prominently in the presumptive adult gut (Fig. 1f, g). These results indicate that eIF4A mutations cause prolonged Mad phosphorylation, which might be due to reduced Mad degradation.

To test whether wild-type eIF4A antagonizes Dpp signalling, we expressed eIF4A using the Gal4/UAS system¹². First, expressing one copy of UAS–eIF4A by the eye-specific GMR–Gal4, which by itself had little effects, completely neutralized the effects of ectopic dpp expression on eye development (Fig. 1h), indicating that overexpression of eIF4A antagonizes Dpp signalling. Conversely, expressing dpp by GMR–Gal4 in eIF4A^R321H heterozygotes caused lethality, and the escapers had enlarged and deformed eyes (Fig. 1h), which is consistent with the idea that endogenous eIF4A antagonizes Dpp signalling. Second, expressing eIF4A by the wing-margin-specific C96–Gal4 resulted in ‘notches’ in the wing margin (Fig. 1i; right) — a phenotype that is similar to those produced by expressing dominant-negative molecules of the Dpp pathway, such as Put and Sax (Fig. 1i; middle). We examined pMad levels in 3^rd instar imaginal discs and found a prominent loss of pMad signals in wing-margin primordial cells in which C96–Gal4 is expressed (Fig. 1j). These results are consistent with the notion that wild-type eIF4A negatively modulates Dpp signalling at the level of Mad protein regulation.

To investigate the molecular mechanism by which eIF4A negatively modulates Dpp signalling, we co-expressed Mad, Medea and eIF4A^WT (or eIF4A^R321H) in 293T cells with or without the activated form of Dpp receptor Tkv^QD, which is able to phosphorylate and activate Mad in a ligand-independent manner¹³. Medea is a co-Smad that is homologous to Smad4 (refs 14, 15). It has been reported that transfected Mad and Medea are localized in the cytoplasm, and cotransfection with Tkv^QD results in their nuclear translocation¹⁵. We found that, in the absence of Tkv^QD, eIF4A^WT (or eIF4A^R321H) and Mad were localized in the cytoplasm in the same nonuniform patterns (Fig. 2a, rows 1, 3). However, following co-transfection with Tkv^QD, in cells transfected with eIF4A^WT, the levels of Mad were reduced and its nuclear translocation was not detectable (Fig. 2a, row 2). This is in contrast to cells without transfected eIF4A (see Supplementary Information, Fig. S2; also see ref. 15) or to those transfected with eIF4A^R321H (Fig. 2a, row 4), which exhibit prominent Mad nuclear translocation. These results are consistent with the notion that eIF4A^WT antagonizes and eIF4A^R321H enhances Dpp signalling.

Figure 2.

Physical association of eIF4A with Mad and Medea. (a) Flag–Mad, Myc–Medea, and eIF4A^WT or eIF4A^R321H were co-transfected into 293T cells with or without Tkv^QD. Cells were stained with anti-Flag (green), anti-eIF4A (magenta) and propidium iodide (PI; red) to reveal the subcellular localization of transfected Mad and eIF4A. (b) 293T cells were co-transfected with Flag–Mad, Myc–Medea, eIF4A^WT and eIF4A^R321H in different combinations, as indicated. Cell lysates were immunoprecipitated (IP) with anti-His (for eIF4A), and the immunoprecipitates were subjected to SDS–PAGE and blotted with antibodies, as indicated. Note that co-immunoprecipitation of Mad and Medea with eIF4A occurred. (c) Drosophila embryo extracts from wild-type or eIF4A^R321H heterozygotes were immunoprecipitated with rabbit anti-eIF4A or goat anti-Mad, or with rabbit anti-βgal or goat anti-HA antibodies (negative controls, respectively), and the immunoprecipitates were subject to SDS–PAGE and blotted with indicated antibodies. (d) As the R321H mutation is located in the carboxy-terminal half of the eIF4A protein¹, which consists of substrate (RNA)-binding domains, the protein was divided into amino- and C-terminal halves, such that eIF4A^WT and eIF4A^R321H shared the same N-terminal fragment but different C-terminal fragments, with a different amino acid at position 321. His-tagged N-terminal, C-terminal, or full-length eIF4A^WT or eIF4A^R321H were expressed from Escherichia coli. The proteins were incubated with Drosophila embryo extracts and were then pulled down by Ni-NTA agarose beads, subjected to SDS–PAGE, and blotted with anti-Mad and anti-histidine. Note that Mad is pulled down by all species of eIF4A proteins, but in higher amounts by eIF4A^WT full-length (lane 6) and C terminus (lane 4). The anti-histidine antibody also recognizes a non-specific band that co-migrated with full-length eIF4A. For each binding assay above, the membrane was blotted with an antibody against the target protein of interest, and was then stripped of bound antibodies and re-blotted with an antibody against the bait protein. See Supplementary Information Fig. S5 for full-length gels.

As transfected eIF4A and Mad both reside in the cytoplasm in similar patterns prior to Mad activation, we tested whether they physically interact with the following assays. First, in co-transfected 293T cells, we found that eIF4A was able to co-immunoprecipitate with either Mad or Medea (Fig. 2b), indicating that eIF4A physically interacts with both Mad and Medea in transfected cells. Second, we performed immunoprecipitation studies using embryo extracts, and found that eIF4A^WT and, to a lesser extent, eIF4A^R321H were co-immunoprecipitated with Mad from embryo extracts (Fig. 2c). Finally, we confirmed the physical interaction between Mad and eIF4A using bacterially expressed eIF4A protein fragments and embryo lysate. The results indicated that Mad binds to the full-length eIF4A (Fig. 2d, lane 6) mainly through the carboxyl terminus (Fig. 2d, lane 4), and has a lower affinity for the amino terminus or eIF4A^R321H (Fig. 2d, lanes 5, 7). These results indicate that eIF4A^WT is able to physically associate with Mad and that eIF4A^R321H is less able to do so.

We next examined the protein levels of Mad and Medea in the presence or absence of eIF4A and/or Dpp signalling in transfected cells. We found that low levels of Dpp signalling (low concentration of Tkv^QD) induced a dramatic degradation of Mad and Medea, as well as eIF4A itself, in the presence of eIF4A^WT (Fig. 3a, lane 6) but not eIF4A^R321H (Fig. 3a, lane 7), or in the absence of Tkv^QD (Fig. 3a, lane 3). We further confirmed that eIF4A-induced Mad and Medea degradation depends on the dosage of Tkv^QD and the proteosome system. Degradation of Mad and Medea was detected in the absence of exogenous eIF4A at high doses but not low doses of Tkv^QD (Fig. 3b, lanes 2, 3). However, the presence of exogenous eIF4A^WT or, to a lesser extent, eIF4A^R321H resulted in the disappearance of Mad and Medea even at low doses of Tkv^QD (Fig. 3b, lanes 4, 6), and this effect was inhibited by the proteosome inhibitor MG-132 (Fig. 3b, lane 8), indicating that eIF4A promotes proteosome-dependent degradation of both Mad and Medea.

Figure 3.

eIF4A promotes Dpp signalling-dependent Mad and Medea degradation. (a–b) eIF4A induces Tkv^QD-dependent degradation of Mad and Medea. (a) 293T cells were transfected with Flag–Mad, Myc–Medea and eIF4A^WT (or eIF4A^R321H) with or without low-level Tkv^QD (0.5 μg) in combinations, as indicated. Cell lysates were subject to SDS–PAGE and were blotted with anti-Flag, anti-Myc and anti-eIF4A consecutively. Note the apparent degradation of Medea, Mad and eIF4A^WT in the presence of Tkv^QD (lane 6), whereas no degradation was detected in the absence of Tkv^QD (lanes 2–4) or eIF4A^WT (lanes 5 and 7). (b) eIF4A-stimulated degradation depends on the proteosome. 293T cells were co-transfected with Flag–Mad, Myc–Medea, eIF4A^WT (or eIF4A^R321H), and different amounts of Tkv^QD, and treated with the proteosome inhibitor MG-132, as indicated. Note the degradation of Medea and Mad in the presence of eIF4A^WT and, to a lesser extent, eIF4A^R321H at indicated Tkv^QD concentrations, which was inhibited by MG-132. (c) eIF4A promotes Tkv^QD-dependent ubiquitination of Mad. 293T cells were transfected with His–eIF4A^WT, Flag–Mad, Myc–Medea and HA–ubiquitin with or without low levels of Tkv^QD in the presence of MG-132. Mad was immunoprecipitated with anti-Flag and its ubiquitination was detected by anti-HA. Note the increased Mad ubiquitination in the presence of eIF4A (lane 4, compare with lane 2). (d) Ubiquitination of eIF4A depends on the presence of Mad, Medea and Tkv^QD. 293T cells were transfected, as indicated, without added MG-132. eIF4A was immunoprecipitated by anti-histidine and its ubiquitination was detected by anti-HA. Note the dramatic increase of ubiquitinated eIF4A^WT in the presence of Mad, Medea and Tkv^QD (lane 4), compared with either lack of Mad and Medea (lane 2) or Tkv^QD (lane 3). See Supplementary Information Fig. S5 for full-length gels.

It has been shown that Mad is ubiquitinated prior to its degradation¹⁶. We found that eIF4A increased Dpp signal-dependent (+Tkv^QD) ubiquitination of Mad (Fig. 3c). Moreover, we found that eIF4A itself also appeared to be ubiquitinated and degraded following activation of Dpp signalling, as Tkv^QD caused a dramatic increase in eIF4A ubiquitination and degradation only in the presence of Mad and Medea, but not in their absence (Fig. 3d; compare lanes 2 and 4).

As the known function of eIF4A has been in translation initiation, it is formally possible that eIF4A might preferentially promote the translation of an inhibitor of the Dpp pathway. However, this is unlikely to be the case based on the following observations. First, when overexpressed in cultured cells, eIF4A^R321H also promoted Mad and Medea degradation especially at high doses of Dpp signalling, albeit less efficiently than eIF4A^WT (see Fig. 3b). This is presumably due to its ability to bind to Mad and Medea (see Fig. 2b–d) and to be ubiquitinated (see Fig. 3d). However, eIF4A^R321H behaves genetically as a dominant-negative molecule in causing lethality and growth delay¹, possibly by inhibiting general translation. Indeed, purified eIF4A^R321H protein potently inhibited protein translation in an in vitro system in a dose-dependent manner (see Supplementary Information, Fig. S3a). These results indicate that the ability of eIF4A to inhibit Dpp signalling is separable from its function as a translation initiation factor. Second, inhibiting the eIF4F complex does not lead to increased Dpp signalling. Mutations in eIF4E cause lethality and larval growth delay¹⁷, phenotypes that are essentially identical to those of eIF4A. eIF4E is an essential regulatory subunit of the eIF4F complex¹⁷. However, none of the eIF4E mutants we tested caused increased Dpp signalling in a sensitized assay (see below and Fig. 4a; data not shown). We further examined the number of amnioserosa cells, which reflects the levels of Dpp signalling in the early embryo^1,2. Unlike eIF4A or other Su(dpp) mutations¹, eIF4E mutant embryos had a normal number of amnioserosa cells (see Supplementary Information, Fig. S3c). These results indicate that the negative modulation of Dpp signalling by eIF4A is independent of its function in protein translation.

Figure 4.

eIF4A acts synergistally with and independently of DSmurf. (a) eIF4A mutations, but not those of eIF4E or activated TKV, interact synergistically with DSmurf mutations. The effects of maternal mutations on the viability of 1xdpp⁺ versus 3xdpp⁺ progeny are plotted as the ratios of these two categories. Females of the indicated genotypes were mated to 0xdpp⁺/2xdpp⁺ males. The F1 progeny would inherit either 1xdpp⁺ or 3xdpp⁺. Note the ratio is infinity (∞) for progeny of eIF4A¹⁰⁰⁶/+;lack^KG07014/Df mothers, and no progeny were produced by eIF4A^R321H/+;lack^KG07014/Df mothers. Df = Df(2R)Exel7149, which removes DSmurf. (b). eIF4A mutations further enhance the phenotypes of flies lacking maternal and zygotic DSmurf. The cuticles of embryos derived from females of the indicated genotype mated to lack^KG07014/Df males. Dorsalization was quantified by the number of denticles in the first row of the 6^th ventral denticle belt (arrow). Note that embryos heterozygous for eIF4A¹⁰⁰⁶ or eIF4A^R321H and maternally and zygotically null for DSmurf exhibited narrower or no ventral denticle belts, respectively. Inset, high magnification of boxed area showing dorsal hairs. eIF4A^R321H/+;lack/Df females are sterile. They lay very few eggs, most of which seem to be unfertilized (white eggs with no cuticles), regardless to which males they were mated. However, a small percentage (<5%) of the eggs appeared to be fertilized and secreted cuticles. These eggs are completely dorsalized (bottom panel). (c). Mutations in eIF4A further increase Mad protein accumulation in lack/Df embryos. Stage 13 embryos of the indicated genotypes were stained by anti-Mad (dark stains). Note that lack/Df embryos exhibited elevated levels of Mad protein compared with the wild type, similar to eIF4A^R321H/+ embryos (compare with Fig. 1e), and eIF4A¹⁰⁰⁶/+;lack/Df embryos accumulate Mad protein in even more cells, such as in the germ bands and guts (arrows). Right panels are higher magnifications of the boxed area (ps7). Scale bar, 100 μm.

The ubiquitin E3 ligase DSmurf (encoded by lack) has been shown to negatively regulate Dpp signalling¹¹ by specifically binding to and promoting degradation of phospho-Mad¹⁶. To investigate whether the function of eIF4A in mediating Mad degradation depends on DSmurf, we tested genetic interactions and the epistatic relationship between eIF4A and DSmurf mutations in a sensitized genetic assay. Flies with one copy and three copies of dpp⁺ (1xdpp⁺ and 3xdpp⁺, respectively) are inversely sensitive to maternal-effect mutations that increase Dpp signalling strength^1,5,6. Increasing Dpp signalling promotes the viability of 1xdpp⁺ animals, which are normally not viable in an otherwise wild-type genetic background^1,5,6, and decreases that of 3xdpp⁺ animals. Thus, when fathered by 0xdpp⁺/2xdpp⁺ males (see Methods), the ratio of F1 siblings inheriting 1xdpp⁺ versus 3xdpp⁺ is a sensitive measurement of the effects of genetic mutations carried by the mother on Dpp signalling.

This sensitized assay allowed us to make the following two findings with regards to eIF4A. First, we found that eIF4A and DSmurf mutations interact synergistically in trans-heterozygotes. Heterozygous mutations in eIF4A and DSmurf individually either did not rescue or moderately rescued the lethality of 1xdpp⁺ animals, such that when females that were heterozygous for eIF4A¹⁰⁰⁶, eIF4A^R321H or lack^KG07014 (a strong allele) were mated to 0xdpp⁺/2xdpp⁺ males, the ratios of 1xdpp⁺ versus 3xdpp⁺ progeny in the F1 were 0 (no rescue), 0.15 and 0.29, respectively (Fig. 4a). However, transheterozygosity of eIF4A and DSmurf mutations produced synergistic effects on the viability of F1 progeny, dramatically increasing the number of 1xdpp⁺ offspring at the expense of their 3xdpp⁺ siblings, resulting in ratios of 1.2 and 1.4, respectively (Fig. 4a). Generally increasing Dpp signalling did not result in synergistic interaction with DSmurf. Specifically, low-level expression of the activative Tkv^QD molecule (using hs–Gal4/UAS–Tkv^QD females under mild heat-shock conditions) can suppress dpp haplo-lethality, resulting in a 1xdpp⁺ versus 3xdpp⁺ progeny ratio 0.01 (8/896). This is significant because 1xdpp⁺ survivors were never found in the progeny of wild-type females (n > 5000) or eIF4A¹⁰⁰⁶/+ females (n > 2000). However, in the progeny of lack^KG07014/+;hs–Gal4/UAS–Tkv^QD females, the ratio was not significantly different from that of lack^KG07014/+ females (Fig. 4a). Taken together with previous results^16,11, the synergy between eIF4A and DSmurf strongly indicates that they function in the same biological process — that is, promoting Mad protein degradation.

Second, we tested whether the effects of eIF4A mutations on Dpp signalling depend on DSmurf. Flies without DSmurf (lack^KG07014/Df) were viable and fertile, and the females produced progeny in which the ratio of 1xdpp⁺ versus 3xdpp⁺ animals was >1.5 (Fig. 4a). However, animals with a half dose of eIF4A⁺ in this genetic background (eIF4A¹⁰⁰⁶/+;lack^KG07014/Df), although perfectly viable, resulted in complete lethality of 3xdpp⁺ offspring (ratio = 8; Fig. 4a). Flies heterozygous for eIF4A^R321H in the absence of DSmurf (eIF4A^R321H/+;lack^KG07014/Df) were barely viable (10.5% viable; n = 228) and completely sterile, whereas either eIF4A^R321H/+ or lack^KG07014/Df flies separately were perfectly viable and fertile (data not shown). Thus, eIF4A mutations further enhance the mutant phenotypes of flies without DSmurf.

To further substantiate this finding, we examined eIF4A embryos lacking both maternal and zygotic DSmurf. Increased Dpp signalling can lead to expansion of dorsal/lateral cell fates at the expense of ventral cell fates. We found that although embryos produced by either eIF4A¹⁰⁰⁶/+ or lack^KG07014/Df females (crossed to lack^KG07014/Df males) exhibited no significant dorsalization, those from eIF4A¹⁰⁰⁶/+;lack^KG07014/Df females were moderately (yet significantly) dorsalized, and embryos from eIF4A^R321H/+;lack^KG07014/Df females were completely dorsalized (Fig. 4b). Consistent with these results, embryos from eIF4A¹⁰⁰⁶/+;lack^KG07014/Df females exhibit more pronounced Mad accumulation (Fig. 4c). Thus, reducing eIF4A can further increase Dpp signalling and Mad accumulation in the absence of DSmurf. These results indicate that eIF4A and DSmurf can function independently in controlling Mad degradation.

In conclusion, our results indicate that eIF4A plays an essential role in limiting the duration and spatial distribution of Dpp signalling by promoting Mad/Medea degradation. We propose that in the absence of Dpp signals, eIF4A is able to physically associate with both Mad and Medea. Following the activation of the Dpp pathway, eIF4A promotes the ubiquitination and degradation of Mad and Medea, leading to attenuation and termination of Dpp signalling. eIF4A may function as an adaptor that links SMAD proteins to a degradation system that can be independent of the ubiquitin ligase DSmurf. We found no evidence that eIF4A mutations affect the Hedgehog and Wingless signalling pathways by genetic interaction analyses (see Supplementary Information, Fig. S4), thus eIF4A seems to be a specific regulator of Dpp signalling.

METHODS

Fly stocks and genetics

All crosses were carried out at 25 °C on standard cornmeal/agar medium. eIF4A^R321H and transgenic flies carrying UAS–eIF4A and UAS–eIF4A^R321H were as described previously¹. UAS–Tkv^QD, UAS–Put^DN and UAS–Sax^DN were as described previously^13,18. The P-element-associated eIF4A alleles (eIF4A¹⁰⁶⁹ and eIF4A¹⁰⁰⁶) were as described previously³. Sp dppH ⁴⁶/CyO-23 flies have the 0xdpp⁺/2xdpp⁺ genotype, as they carry a dpp null mutation (0xdpp⁺) and two doses of dpp⁺ on the CyO-23 balancer chromosome: the endogenous dpp⁺ and a p[dpp⁺] transgene, as previously described^1,5. Crossing females (dpp⁺/dpp⁺) to Sp dppH ⁴⁶/CyO-23 males produces 1xdpp⁺ and 3xdpp⁺ animals in the F1 progeny, allowing determination of their ratios. The Mad sax double-mutant stock Df(2L)JS17 sax¹/CyO-23 is as described previously⁵. Df(2L)JS17 is a deficiency that removes Mad, thus was used as a null allele of Mad. Somatic mutant clones for tkv⁴ and eIF4A¹⁰⁶⁹ were generated using the FLP–FRT method by crossing FRT^40A tkv⁴/CyO or FRT^40A eIF4A¹⁰⁶⁹/CyO flies to hs–flp;FRT^40A or hs–Myc [Au: OK?], respectively. The resulting larvae were heat-shocked for 2 h at 37 °C during the 2^nd and early 3^rd instar larval stages and dissected at the late 3^rd instar stage. A brief heat-shock was administered 15 min before dissection to increase hs–Myc expression for marking the clones. All other stocks, including DSmurf mutants (lack^KG07014 and Df(2R)Exel7149), eIF4A–lacZ and dpp–lacZ enhancer traps (eIF4A^k01501, eIF4A⁰²⁴³⁹ and dpp^P10638), and various tissue-specific Gal4 lines, including C96–Gal4 (ref. 19), hs–Gal4, GMR–Gal4 and UAS–dpp were from public Drosophila stock centres or as described previously¹. lack^KG07014 seemed to be a stronger allele than the reported null allele lack^15C (ref. 11), based on the observation that, when mated to dppH ⁴⁶/CyO-23 males, lack^KG07014/+ mothers produced 25% viable 1xdpp⁺ progeny, whereas <3% 1xdpp⁺ animals survived from lack^15C/+ mothers (also see ref. 11).

Plasmids and antibodies

His-tagged full-length and N- and C-terminal eIF4A constructs were made by ligating PCR-amplified eIF4A or eIF4A^R321H cDNA fragments¹ to pQE vectors (Stratagene, La Jolla, CA) and subsequently subcloned into pcDNA3 (Invitrogen, Carlsbsd, CA). PCR primers for eIF4A-N and eIF4A-C were cgatGAGCTCggatgaccgaaatgagata (forward) and cgcaTCTAGAtaa tgggggcagcatcttgaa (reverse), and ctatGAGCTCtttcaagatgctgccccca (forward) and cgcaTCTAGAgcggctcagcagaaaaaa (reverse), respectively. The R321H substitution was located in the C terminus. Flag–Mad, Myc–Medea and TKV^QD constructs (generous gifts from M. Kawabata) were as described previously²⁰.

The primary antibodies (dilutions) used for whole-mount immunostaining and western blots were: rabbit anti-eIF4A (a gift from P. Lasko, McGill University, Canada; 1:300), sheep anti-eIF4A-c (a gift from C. Proud, University of British Columbia, Canada; 1:300), rabbit anti-pMad (PS1; a gift from P. ten Dijke, The Netherlands Cancer Institute, The Netherlands; 1:100), mouse anti-Myc (9E10; Developmental Hybridoma Bank, Iowa City, IA; 1:50), anti-Flag (M2; Sigma, St Louis, MO; 1:1000), mouse anti-β-gal (Promega, Madison, WI; 1:1000), goat anti-Mad (Santa Cruz, Santa Cruz, CA; 1:250) and anti-polyhistidine (Sigma; 1:2000). Different fluorophore-conjugated secondary antibodies were obtained from Molecular Probes (Carlsbad, CA; 1:250). HRP-conjugated secondary antibodies (Promega; 1:1000) were used for western blots.

Cell culture and immunoprecipitation

293T cells were cultured in DMEM (Invitrogen) supplemented with 10% fetal bovine serum in six-well plates. Cells were transfected with appropriate expression plasmids using FuGene 6 (Roche, Basel, Switzerland) according to the manufacturer’s instructions. Cells were harvested 48 h after transfection and were stained with the indicated antibodies or lysed in cell lysis buffer (20 mM Tris-HCl at pH 7.5, 150 mM NaCl, 1 mM Na₂EDTA, 1 mM EGTA, 1% Triton-X-100, 2.5 mM sodium pyrophosphate, 1 mM β-glycerophosphate, 1 mM Na₃VO₄, 1 μg ml⁻¹ leupeptin) (Cell Signaling Technology, Danvers, MA) for immunoprecipitation or SDS–PAGE. For proteosome inhibition, MG-132 (Sigma) was added at 50 μM 4 h prior to lysis. To immunoprecipitate endogenous Mad or eIF4A, extracts from 200 Drosophila embryos (0–24 h old) of appropriate genotype were used.

Pull-down and in vitro translation assays

1 μg of bacterially expressed His-tagged eIF4A protein was incubated with 200 μl of embryo extracts (in cell lysis buffer; see above) and 20 μl of Ni-NTA agarose beads for 2 h at room temperature. The Ni-NTA agarose beads were spun down and washed in the same buffer and subjected to SDS–PAGE to analyze bound proteins by western blotting with anti-Mad and anti-eIF4A.

In vitro translation was performed using the TNT^® T7 Coupled Reticulocyte Lysate System in conjunction with the Transcend™ Colorimetric Non-Radioactive Translation Detection System (Promega) as per the manufacturer’s instructions. His-tagged eIF4A proteins expressed in Escherichia coli and quantified by SDS–PAGE, followed by Coomassie Blue staining, were included in the assays. The translation rate was measured by the amount of firefly luciferase (internal positive control provided by the Kit) produced, which had incorporated biotinylated lysine and was detected by the Colorimetric (BCIP/NBT) reagents following SDS–PAGE and transfer per the manufacturer’s instructions.