Drosophila Ik2, a member of the IκB kinase family, is required for mRNA localization during oogenesis

Protein kinases of the IκB kinase (IKK) family are known for their roles in innate immune response signaling pathways in both mammals and Drosophila (Ghosh and Karin,2002; Peters and Maniatis,2001; Silverman and Maniatis,2001). Mammalian IKKs all have roles in immune responses, but have a variety of targets. IKKα and IKKβ were identified in a protein complex that phosphorylates IκB and targets it for degradation, thereby allowing the nuclear localization and activation of NF-κB transcription factors (DiDonato et al.,1997; Mercurio et al.,1997; Zandi et al.,1997). Gene targeting experiments in the mouse demonstrated that IKKβ, but not IKKα, is required for NF-κB activation by pro-inflammatory stimuli through receptors such as TLR4(Li et al., 1999a; Li et al., 1999b; Tanaka et al., 1999). IKKα activates the Rel/p52 transcription factor, because it activates proteolytic processing of the p100 precursor of p52 in an IκB-independent process (Senftleben et al., 2001). IKKϵ and TANK binding kinase 1 (TBK1) are required to phosphorylate and activate the transcription factor Interferon regulatory factor 3 (IRF3) in response to viral infection(Fitzgerald et al., 2003; McWhirter et al., 2004; Sharma et al., 2003). In addition to these immune response functions, IKKα has an NF-κB-independent role in epidermal differentiation and limb development(Hu et al., 1999; Hu et al., 2001; Sil et al., 2004; Takeda et al., 1999).

Dorsoventral patterning of the Drosophila embryo relies on the activation of Dorsal, a Rel-family transcription factor, by a signaling pathway that is homologous to mammalian TLR pathways(Anderson, 2000). In response to activation of the receptor Toll, Cactus (the Drosophila IκB)is degraded, which allows Dorsal to move to embryonic nuclei and activate genes, such as twist, that are required for specification of ventral cell types. Phosphorylation of Cactus is required for its degradation(Fernandez et al., 2001), but the responsible kinase has not been identified. The Drosophila genome encodes two members of the IKK family. DmIkkβ (ird5 -FlyBase) is essential for the response to bacterial infection(Lu et al., 2001). DmIkkβ is required for proteolytic processing and activation of Relish, a p100-like Rel/ankyrin-repeat protein, like the role of mammalian IKKα in the activation of p100. The function of the second Drosophila protein kinase of the IKK family, ik2(IκB kinase-like 2), has not been characterized, but it was a good candidate to control the phosphorylation and degradation of Cactus.

To test whether ik2 encodes a Cactus kinase, we characterized the phenotypes caused by loss of ik2 function. Here, we present data showing that Ik2 is essential for dorsoventral and anteroposterior embryonic patterning of the Drosophila embryo. However, Ik2 does not act as a Cactus kinase, but exerts its effects on embryonic patterning through the localization of specific mRNAs during oogenesis. The data indicate that Drosophila Ik2 regulates RNA localization through regulation of the cytoskeleton and define a novel function for this protein family.

The l(2)38Ea complementation group(Kozlova et al., 1998)consists of five alleles and were renamed accordingly: Ea³⁶ (ik2¹), Ea⁴¹(ik2²), Ea⁴⁶(ik2³), Ea⁴⁷(ik2⁴), Ea⁶⁶(ik2⁵). The following mutations in the kinase domain were identified from sequenced ik2 genomic DNA: ik2¹(G250D), ik2² (N69I), ik2³ (G19D), ik2⁴ (G109A), ik2⁵ (D160N). ik2^alice was recovered in a maternal effect genetic screen(Luschnig et al., 2004), and also had a point mutation in the kinase domain (F297I).

The X chromosome FLP recombinase stock P[ry⁺,hsFLP], the second chromosome FRT stock P[w⁺ FRT 40A] (Chou and Perrimon,1996), Df(2L)Ketel^RX32(Erdelyi et al., 1997), BicD¹ and BicD²(Mohler and Wieschaus, 1986),and the tubulin-Gal4 and daughterless-Gal4 drivers were obtained from the Bloomington Stock Center. The recombinant chromosomes ik2¹P[w⁺ FRT 40A] and ik2^aliceP[w⁺ FRT 40A] were provided by F. Schnorrer and C. Nüsslein-Volhard (Tübingen). The following stocks were also used:Tau-GFP line 2.1 (Micklem et al.,1997); Kinesin-β-galactosidase insertion line KZ503(Clark et al., 1994); and Nod-β-galactosidase insertion line NZ143.2(Clark et al., 1997). To induce expression of FLP recombinase, flies were mated for 24 hours, and second instar larvae were heat shocked in a 37°C water bath for two hours on two consecutive days.

To visualize the chorion under the microscope, eggs were washed with 0.7%NaCl and 0.1% Triton X-100, and mounted in Hoyer's medium(Van der Meer, 1977). For cuticle preparations, embryos were collected on apple juice agar plates,washed with 0.7% NaCl and 0.1% Triton X-100, and bleached to remove the chorion. Embryos were fixed for 1 hour at 65°C in 1:4 glycerol:acetic acid and mounted in Hoyer's medium.

Whole-mount ovaries were hybridized with digoxigenin-labeled grk,bcd and osk RNA probes, all described previously(Berleth et al., 1988; Ephrussi et al., 1991; Neuman-Silberberg and Schüpbach,1993). Ovaries were fixed and stained according to Suter and Steward (Suter and Steward,1991). Embryonic fixation and hybridization were performed as described previously (Tautz and Pfeifle,1989). Fluorescent ovarian in situ hybridization to detect osk mRNA was performed as described previously(Cha et al., 2002).

For Twist staining of embryos, 0- to 2-hour embryos were collected, aged 2 hours at 25°C, and fixed with 4% paraformaldehyde. Embryos were rehydrated with 1×PBS and incubated in 0.3% BSA in PBST (1×PBS with 0.3%Triton X-100) for 30 minutes. After overnight incubation at 4°C with primary antibody (in 0.3% BSA in PBST) and washing with PBST, the samples were incubated for 1 hour with biotin goat anti-rabbit secondary antibody (Jackson ImmunoResearch) and signal was amplified using the Vector Elite ABC kit(Vector Laboratories). Rabbit anti-Twist (1:5000) was kindly provided by Siegfried Roth (Cologne).

Ovaries from 24- to 48-hour-old females were dissected and fixed as previously described (Verheyen and Cooley,1994a). Antibodies were used at the following concentrations:mouse monoclonal P1H4 anti-dynein heavy chain, 1:500(McGrail and Hays, 1997);anti-β-galactosidase monoclonal antibody, 1:2000 (Promega). For Grk antisera (1:10), ovaries were fixed and stained as described previously(Queenan et al., 1999). For visualization of actin, ovaries were incubated with either FITC-phalloidin or rhodamine-phalloidin (Molecular Probes) for 2 hours. Images were captured using a Leica TCS SP2 confocal microscope system and Leica Confocal Services software (version 2.61).

Fly heads were fixed with 4% formaldehyde overnight, then dehydrated with a graded ethanol series. Samples were critical-point dried in CO₂,then sputter coated with 30 nm of gold palladium and examined with a scanning electron microscope.

An ik2 genomic DNA fragment from 95 bp before the first codon to 350 bp following the stop codon was cloned from Oregon R genomic DNA using the primers 5′-GCTCTAGAGTCACAATCGAGAAGGCGCTT-3′ and 5′-GCTCTAGAGCTCAATGGCGTCGAG-3′, which incorporated an XbaI site on both the 5′ and 3′ ends. The resulting 3.1 kb fragment was inserted into the XbaI site of the UASp vector to drive maternal expression (Rørth,1998). The rescue plasmid was injected into yw embryos and transgenic lines were selected for expression of the whitegene.

We identified a gene on the Drosophila second chromosome, ik2, as a member of the IKK family. The closest mammalian homologs of Drosophila Ik2 are IKKϵ and TANK binding kinase 1 (TBK1), which are 60-61% identical to Ik2 in the kinase domain and 51% identical across the entire protein; by contrast, Ik2 is only 28% identical to the other Drosophila IKK, DmIkkβ. A saturation mutagenesis experiment had identified lethal complementation groups in polytene chromosome region 38E(Kozlova et al., 1998), the region that includes the ik2 gene. We identified missense mutations in the ik2 kinase domain in all five alleles of the l(2)38Eacomplementation group. A sixth allele, ik2^alice,identified in a genetic mosaic screen for maternal effect mutations(Luschnig et al., 2004), had a missense mutation in the C-terminal end of the kinase domain. All six ik2 alleles caused recessive lethality, and the majority of the mutants died as first instar larvae. At low temperature and in uncrowded culture conditions, rare escaper adults (<1%) were observed, but they died shortly after eclosion.

If ik2 encoded the Cactus kinase, embryos produced by females that lack ik2 would not be able to degrade Cactus, so Dorsal would not enter embryonic nuclei to activate genes required for ventral cell fate specification and the embryos would be dorsalized. Because ik2mutations were lethal, we used FRT/FLP recombination combined with the ovo^D dominant female-sterile mutation to generate mutant clones in the female germline (Chou and Perrimon, 1992). More than 95% of the embryos laid by ik2^alice and ik2¹ mutant females did not hatch; however, larval cuticle preparations showed that none of the embryos were dorsalized. Instead, the majority of embryos produced by ik2^alice and ik2¹ mutants had a bicaudal phenotype, ranging from headless embryos(Fig. 1B) to embryos with a duplicated abdomen in place of the head and thorax(Fig. 1C). In addition to this anteroposterior patterning defect, a large number of embryos from both ik2^alice (Fig. 2B) and ik2¹(Fig. 2C) germline clones had expanded ventral cuticular structures, the opposite of the expected phenotype. Some embryos were both ventralized and bicaudal, with expanded ventral denticle bands and filzkörper (a posterior structure) in both the tail and the anterior of the embryo. Both ik2 alleles produced bicaudal and ventralized embryos, but 89% (n=190) of the embryos produced by ik2^alice mutant females were bicaudal with no apparent dorsoventral abnormalities, whereas only 47% (n=110) of embryos produced by ik2¹ mutant females were bicaudal, and the remainder of the embryos appeared to be too ventralized to score for ectopic posterior cuticular structures. We observed a similar range of phenotypes in embryos produced by ik2², ik2³ and ik2⁵ females with mutant germline clones.

The anteroposterior pattern of the embryo depends on the localization of maternal mRNAs. The bicoid (bcd) mRNA is localized to the anterior end of the egg, and specifies anterior cell fates including the head and thorax (Berleth et al.,1988). nanos mRNA is localized at the posterior pole of the egg, a process that depends on prior posterior localization of oskar (osk) mRNA, and specifies the pattern of the abdomen(Nüsslein-Volhard et al.,1987; Ephrussi et al.,1991).

The best-characterized mutations that lead to a high frequency of bicaudal embryos are gain-of-function alleles of the Bicaudal C(BicC) and Bicaudal D (BicD) genes(Mohler and Wieschaus, 1986; Suter and Steward, 1991; Wharton and Struhl, 1989; Schüpbach and Wieschaus,1991; Mahone et al.,1995; Castagnetti and Ephrussi,2003). Bicaudal embryos produced by BicC and BicD females have ectopic anterior nanos and oskarmRNA, leading to duplicated posterior structures(Mahone et al., 1995; Ephrussi et al., 1991). We examined the localization of these key patterning mRNAs in embryos produced by ik2 germline clones to determine whether the cause of the ik2 bicaudal phenotype was similar to that of the BicC and BicD mutants.

In wild-type embryos prior to gastrulation, bcd mRNA is localized to the anterior pole (Fig. 1D)(Berleth et al., 1988) and osk mRNA is found exclusively at the posterior pole(Fig. 1H)(Ephrussi et al., 1991). bcd mRNA was localized normally in all ik2^alice(Fig. 1E) and ik2¹ (Fig. 1F) mutant embryos examined, similar to BicD¹/BicD² mutants(Fig. 1G). By contrast, oskar mRNA was present at both the anterior and posterior poles in 100% of the embryos produced by ik2^alice(Fig. 1I) and ik2¹ (Fig. 1J) mutant germline clone females, and 100% of the embryos produced by BicD¹/BicD² mutant females(Fig. 1K). Thus even though a bicaudal phenotype could not be detected in all cuticle preparations of the ventralized class of embryos, all embryos produced by ik2^alice and ik2¹ mutant females had anteriorly localized oskar mRNA. While gain-of-function BicCand BicD alleles produce bicaudal embryos at high frequency, ik2 is the first locus to be described where loss of function causes a completely penetrant bicaudal phenotype.

In embryos produced by BicD mutant females, ectopic anterior oskar is sufficient to recruit the posterior determinant nanos to the anterior pole to specify abdominal development(Ephrussi et al., 1991). We found ectopic nanos mRNA at the anterior pole of all ik2^alice (Fig. 1M) and ik2¹(Fig. 1N) mutants, which indicated that oskar localized at the anterior of ik2embryos was sufficient to mislocalize nanos to the anterior pole.

The bcd and osk mRNAs are transcribed in the nurse cells,transported into the oocyte and targeted to the anterior and posterior ends of the oocyte during stages 9-10 of oogenesis(Schnorrer et al., 2000; St Johnston et al., 1989; Ephrussi et al., 1991). In BicD mutant females, osk mRNA is transported efficiently from the nurse cells into the oocyte, but accumulates at both the anterior and posterior poles of the oocyte (Ephrussi et al., 1991). We examined osk mRNA in both BicD¹/BicD² (data not shown) and ik2mutant oocytes using a probe conjugated directly to fluorescein. At a stage when osk mRNA was localized tightly to the posterior pole of wild-type oocytes (Fig. 3A), osk mRNA was present in all regions of the ik2 mutant oocyte cortex (Fig. 3B,C) and was enriched at both the anterior and posterior poles.

While osk mRNA is being transported to the posterior pole of wild-type oocytes, the message is translationally repressed, ensuring that the protein is only active once it reaches its correct subcellular location(Kim-Ha et al., 1995; Markussen et al., 1995; Rongo et al., 1995). Females that carry mutations in BicC produce bicaudal embryos as a result of ectopic anterior osk mRNA and premature translation of Osk at ectopic sites in the oocyte (Mahone et al.,1995; Saffman et al.,1998). In ik2 oocytes, Osk protein was localized exclusively at the posterior of the oocyte at stage 9 and was not detectable at earlier stages (data not shown). Thus the ectopic localization of osk mRNA in ik2 mutant oocytes, and not premature translation, is the cause of the ik2 bicaudal phenotype.

The bcd mRNA is transported from the nurse cells to the oocyte and then restricted to the anterior of the developing oocyte by stages 8-9 of oogenesis (Berleth et al.,1988; St Johnston et al.,1989). The bcd transcript was tightly localized to the anterior cortex of the oocyte in most ik2 egg chambers. However, in approximately 20% of ik2¹ mutant egg chambers the expression domain of bcd mRNA was expanded toward the posterior of the oocyte (data not shown), a pattern that has been seen when dynein-mediated RNA transport is disrupted (Duncan and Warrior, 2002; Januschke et al., 2002; Schnorrer et al.,2000).

More than half of the embryos produced by ik2¹ germline clone females had a ventralized cuticle pattern, with a range of phenotypes that included expanded and disorganized ventral denticle belts, ventral holes and a reduced dorsal cuticle. To determine whether the cuticle defects reflected an early disruption of dorsoventral patterning, we examined the expression of Twist, a direct target of the Rel protein Dorsal. At the cellular blastoderm stage, Twist is expressed in cells in the ventral 25% of the embryo (Fig. 2G). Consistent with the ventralized cuticle pattern, the Twist expression domain expanded to more dorsal positions in the majority of ik2 mutant embryos (Fig. 2H,I). The Twist domain was expanded in 61% of ik2^alice germline clone embryos (n=62). Ninety-four percent of ik2¹mutant embryos (n=106) had an expanded Twist domain, and Twist was often expressed in cells around the entire dorsoventral circumference of the embryo, similar to the strong ventralized phenotype caused by the strongest constitutively active Toll allele(Schneider et al., 1991).

Establishment of the dorsoventral pattern of the embryo is dependent on two sequential signaling pathways, the Gurken/Egf receptor pathway, which acts during oogenesis, and the embryonic Toll-Dorsal signaling pathway(Morisato and Anderson, 1995). Loss of the Tgfα ligand Gurken (Grk) or its receptor, the Drosophila Egf receptor (Egfr) causes ventralization of both the embryo and the surrounding eggshell(Schüpbach, 1987). The eggshell, which is made by the somatic follicle cells that surround the developing oocyte, has a clear dorsoventral polarity, marked by a pair of dorsal appendages at a dorsoanterior position on the eggshell(Fig. 2D). The majority of eggs produced by ik2¹ germline clones had a single, fused dorsal appendage (Fig. 2E), and some lacked dorsal appendages altogether(Fig. 2F), such as grkor Egfr mutants. Because both the embryos and the eggshells produced by ik2 mutants were ventralized, like grk or Egfrmutants, it seemed likely that ik2 affected the Grk/Egfr pathway.

By stage 9 of oogenesis, grk mRNA is localized as a cap around the oocyte nucleus, in the dorsoanterior corner of the oocyte(Fig. 3D)(Neuman-Silberberg and Schüpbach,1993). Grk protein is translated adjacent to the oocyte and signals to the Egfr on nearby follicle cells to instruct those cells to adopt a dorsal cell fate, which sets up the dorsoventral axis of the eggshell and embryo. In the majority of the ik2^alice oocytes(Fig. 3E) and in all ik2¹ oocytes (Fig. 3F), grk mRNA was localized to the anterior margin of the oocyte but was never concentrated dorsally. The pipe gene, which is required to activate the Toll ligand, is expressed only in ventral follicle cells, as the result of repression by Egfr signaling in dorsal follicle cells(Sen et al., 1998). During stages 9-10 of oogenesis, pipe is expressed in ventral follicle cells, corresponding to the future ventral side of the embryo where the Toll-Dorsal signaling pathway will be activated(Fig. 3G). By contrast, pipe mRNA was expressed in both ventral and dorsal follicle cells in the majority of the ik2¹ egg chambers analyzed(Fig. 3H,I). Thus, the absence of a source of grk mRNA at the dorsoanterior corner of ik2mutant oocytes prevents the specification of dorsal follicle cells and causes ventralization of the eggshell and embryo.

To confirm that the anteroposterior and dorsoventral defects observed in ik2 mutants are the result of a genetic requirement for ik2in oogenesis only, we also performed an epistasis analysis of pipe. Females that are homozygous for recessive mutations in pipe produce dorsalized embryos (Fig. 4C),as the Toll ligand is never activated and ventral cell fates are never specified (Sen et al., 1998). We examined the embryos produced by mothers that carry germline clones of ik2¹ in a pipe¹/pipe¹ and a pipe¹/pipe² mutant background. A hundred percent of the embryos examined were dorsalized in a manner similar to pipe mutants (Fig. 4D), suggesting that ik2 affects dorsoventral patterning at a step upstream of the Toll-Dorsal pathway.

Localization of mRNAs to the correct position within the oocyte depends on microtubules and microtubule motors(Brendza et al., 2000; MacDougall et al., 2003; Pokrywka and Stephenson, 1991; Pokrywka and Stephenson,1995). Although recent re-examinations of the oocyte microtubule cytoskeleton have demonstrated that the bulk oocyte microtubules are non-polar(Cha et al., 2001; Cha et al., 2002), it has been proposed that subsets of microtubules are used by microtubule motors to localize mRNAs to specific subcellular regions(MacDougall et al., 2003; Januschke et al., 2006). Localization of osk mRNA to the posterior pole depends on kinesin-based transport, whereas grk mRNA localization to the anterior dorsal corner depends on dynein-based transport. We therefore evaluated the organization of the microtubule cytoskeleton and microtubule motors in ik2 mutants.

Examination of a Drosophila Tau-GFP reporter(Micklem et al., 1997), which decorates all microtubules, revealed that the microtubule cytoskeleton is abnormal in ik2 mutants. In wild-type stage 9 oocytes(Fig. 5A) fixed to preserve the microtubule cytoskeleton, Tau-GFP highlighted the non-polar microtubules in the cytoplasm, but was excluded from the oocyte nucleus. In stage 9 ik2 mutant egg chambers (Fig. 5B), we observed staining throughout the ooplasm, but there were abnormal aggregates of Tau-GFP at the circumference of the nucleus, which suggested that a subpopulation of microtubules was disrupted in the mutant.

After stage 7 of oogenesis, the plus-ends of microtubules are concentrated at the posterior end of the oocyte, and kinesin, the plus-end directed motor,is enriched at the posterior. In wild-type stage 9 egg chambers, the Kinesin Heavy Chain-β-galactosidase fusion(KHC-β-lacZ) transgene was localized to the posterior pole of the oocyte (Fig. 5C)(Clark et al., 1994). At stage 8-9, ik2^alice (Fig. 5D) and ik2¹ (data not shown) mutant egg chambers were indistinguishable from wild type. This indicates that microtubule plus-ends are correctly concentrated at the posterior of the oocyte in ik2 mutants at the time when the kinesin motor is actively transporting osk mRNA to the posterior pole.

During mid-oogenesis, Dynein heavy chain (Dhc) accumulates at the posterior of the oocyte (Fig. 5E)(McGrail and Hays, 1997),where it presumably is stored to allow repeated rounds of minus end-directed transport. In ik2^alice, ik2¹ and BicD¹/BicD² mutant egg chambers, Dhc was distributed along the lateral cortex of the oocyte(Fig. 5F and data not shown)and was sometimes enriched in the dorsal and ventral corners of the anterior margin.

Although the plus-ends of the microtubules appeared normal in ik2mutants, the position of the minus-ends was not. In most cells, microtubule minus-ends are anchored in a microtubule organizing centre (MTOC) at the centrosome, whereas the minus-ends in the Drosophila oocyte are distributed along the anterior and lateral cortex, and are enriched in the area around the oocyte nucleus (Theurkauf et al., 1992; Cha et al.,2002). A transgene that encodes a fusion protein of the motor-like domain of Nod, a kinesin-related protein, with the coiled-coil domain of KHC and β-galactosidase has been used in previous studies as a marker of microtubule minus-ends (Clark et al.,1997). Although Nod preferentially binds microtubule plus-ends in vivo (Matthies et al., 2001; Cui et al., 2005), the Nod-β-gal transgene reliably localizes to the oocyte microtubule minus-ends, probably as a result of sequences outside of the Nod motor-like domain. At stage 8-10 in wild type, Nod-β-gal is localized to the anterior margin of the oocyte and is enriched in the dorsoanterior corner adjacent to the oocyte nucleus (Fig. 5G). Nod-β-gal was not detected at the anterior of stage 8-9 ik2¹ oocytes (Fig. 5H), and instead was present at low levels at the posterior and lateral cortex; at later stages, Nod-β-gal could not be detected in ik2¹ oocytes (Fig. 5J). Thus, microtubule minus-ends are not properly distributed at the anterior of ik2 mutant oocytes during mid-oogenesis.

In addition to the microtubule abnormalities in ik2 mutant oocytes, all of the rare ik2 escaper adults had abnormal bristles with a morphology that suggested defects in the actin cytoskeleton(Fig. 6). Bristles are constructed with rings of membrane-attached, cross-linked actin bundles(Tilney et al., 1995; Tilney et al., 1996). At high magnifications, the actin footprints along the length of the bristle shaft in ik2 escaper adult eyes (Fig. 6F) appeared less organized than those in wild-type bristles(Fig. 6E). Loss of activity of specific actin-associated proteins, including Profilin (chickadee)(Verheyen and Cooley, 1994b)and the β-subunit of capping protein(Hopmann et al., 1996), causes abnormal bristles similar to those seen in ik2 mutants. All of the ik2 allelic combinations analyzed, as well as the alleles in trans to the deficiency Df(2L)Ketel^RX32, which removes the ik2 genomic region, displayed similar bristle phenotypes. Expression of a UASp-ik2 transgene under the direction of either the ubiquitous tubulin-Gal4 or daughterless-Gal4 driver rescued both the viability and bristle abnormalities of ik2 allelic combinations completely (data not shown), which confirmed that these phenotypes are caused by loss of ik2.

We also examined the actin cytoskeleton in oocytes derived from females containing ik2 germline clones. Beginning at stage 7, we identified ectopic sites of actin polymerization in the ooplasm in ik2¹ (Fig. 7G) and ik2^alice (data not shown) mutant egg chambers. Grk protein colocalizes with the oocyte cortical F-actin at stages 7-8 (Neuman-Silberberg and Schüpbach,1996), and we observed that Grk protein colocalized with the ectopic actin in affected ik2 egg chambers(Fig. 7H). Thus, both the actin cytoskeleton and the anchoring of microtubule minus-ends are disrupted in ik2 mutants.

Contrary to the prediction based on its sequence, ik2 does not act as a Cactus/IκB kinase in the Drosophila embryo. The embryonic ventralization caused by loss of ik2 is the opposite of the phenotype predicted for a Cactus kinase, and all effects of ik2 on the dorsoventral pattern of the embryo can be explained by a loss of activity of the Grk/Egfr pathway during oogenesis. Embryos that lack maternal activity of both Drosophila IKKs, Ik2 and DmIkkβ, are ventralized and are indistinguishable from ik2 single mutants (data not shown), which rules out the possibility that the Drosophila IKKs act in both the Grk/Egfr and Toll pathways, and indicates that an unidentified kinase of another family is required to target Cactus for degradation. Additional experiments will be required to test whether ik2 plays other roles in the immune response.

We found that instead of playing a role in Cactus degradation, Drosophila ik2 is required for the localization of specific mRNAs during oogenesis. Both the actin and microtubule cytoskeletons are disrupted in ik2 mutants, and defects in microtubule-based transport are sufficient to account for the defects in mRNA localization seen in ik2 mutants. Because Drosophila ik2 is specifically required for organization of the oocyte cytoskeleton, our results raise the possibility that some of the NF-κB-independent roles of the mammalian IKKs may act through the cytoskeleton.

The embryonic patterning defects caused by the loss of ik2function are due to the failure to transport all osk mRNA to the posterior pole of ik2 mutant oocytes, which leads to bicaudal embryos, and failure to localize grk mRNA to the dorsal anterior of the oocyte, which leads to ventralized embryos. Loss of ik2 has a milder effect on bcd mRNA localization; bcd is correctly localized in most oocytes, but is not tightly restricted to the anterior pole in a minority of cases.

Many lines of evidence indicate that osk localization to the posterior pole depends on kinesin and gurken localization to the dorsoanterior corner depends on dynein(Brendza et al., 2000; Cha et al., 2002; MacDougall et al., 2003). However, the kinesin and dynein motors in the oocyte are interdependent. For example, posterior localization of dynein(Fig. 5) and the anterodorsal localization of gurken are both disrupted in Khc mutants(Brendza et al., 2002), and hypomorphic Dhc mutants have a reduced amount of Khc-β-gal at the posterior pole(McGrail and Hays, 1997). Both motor systems are at least partially functional in ik2 mutants: most oskar is localized to the posterior pole of the oocyte (a kinesin-dependent process) and grk mRNA is localized anteriorly (a dynein-dependent process).

Several lines of evidence suggest that the RNA localization defects seen in ik2 oocytes are associated with defects in a subset of dynein-mediated, minus-end-directed transport processes. The movement of grk mRNA to the dorsoanterior corner of the oocyte depends on two sequential dynein-based movements: grk mRNA moves first to the anterior of the oocyte along microtubules with plus-ends at the posterior pole and minus-ends at the anterior, and then moves dorsally on microtubules with minus-ends that form a cage around the oocyte nucleus(MacDougall et al., 2003). The dorsal movement of grk mRNA is specifically blocked in ik2mutants, which would be consistent with a failure in this dynein-based movement. Restriction of bcd mRNA to the anterior margin of the oocyte, which is disrupted in some ik2 oocytes, depends on the swallow gene product, which binds dynein light chain(Schnorrer et al., 2000). Overexpression of dynamitin disrupts dynein function and causes changes to the localization of grk and bcd mRNA that are similar to the phenotype of ik2 oocytes (Duncan and Warrior, 2002; Januschke et al., 2002). In addition, BicD mutations produce a maternal effect phenotype similar to that of ik2. BicD is part of a protein complex with dynein light chain in early oocytes, neuroblasts and the early embryo (Bullock and Ish-Horowicz,2001; Hughes et al.,2004; Navarro et al.,2004), and has been proposed to link cargo to microtubules in both Drosophila and mammalian cells(Hoogenraad et al., 2003; Matanis et al., 2002).

Although this evidence links ik2 to a dynein transport system, the most penetrant phenotype in ik2 mutants is oskmislocalization and subsequent production of bicaudal embryos, a kinesin-dependent process. However, loss of ik2 function, like the BicD gain-of-function mutations, does not eliminate kinesin function,because the majority of osk mRNA accumulates at the posterior pole. Because the kinesin and dynein motors in the oocyte are interdependent, osk mRNA mislocalization could be caused by a decreased kinesin activity that is secondary to dynein disruption.

In addition to defects in minus-end-directed transport, the organization of the microtubules is also perturbed in ik2 oocytes(Fig. 5). The plus-ends of microtubules are localized correctly to the posterior pole of the oocyte. However, there are abnormal aggregates of microtubules around the oocyte nucleus, where a population of microtubule minus-ends is normally anchored,and the microtubule minus-end marker, Nod-β-gal, is not localized at the anterior of the oocyte. These defects suggest that abnormal organization of microtubule minus-ends during mid-oogenesis could be the basis of the defect in minus-end-directed transport.

The adult bristles and ovaries of ik2 mutants also displayed abnormalities in the actin cytoskeleton. The bristle defects are nearly identical to those caused by mutations in actin-associated proteins(Hopmann et al., 1996; Verheyen and Cooley, 1994b),or to bristles that were treated with F-actin-inhibitors(Tilney et al., 2000). Bristles contain a central core of microtubules, but mutations in the dynein heavy chain gene Dhc64C or treatment with drugs that disrupt microtubule dynamics do not cause bristle phenotypes like the thick, branched bristles seen in ik2 mutants(Gepner et al., 1996; Tilney et al., 2000) (data not shown). The actin cytoskeleton of the oocyte is also disrupted in ik2mutants, with ectopic sites of actin polymerization in the ooplasm(Fig. 7). These actin defects are distinct from those caused by mutations that affect nurse cell ring canal actin (Hudson and Cooley,2002), which suggests that actin organization is not globally disrupted in ik2 mutants and that the actin defects are restricted to the oocyte cortex.

Recent data have defined two sets of microtubules in the oocyte that are both nucleated from minus-ends at the centrosome associated with the oocyte nucleus; one set remains associated with the oocyte nucleus, whereas the remaining microtubules shift their minus-ends from the oocyte to the cortex(Januschke et al., 2006). It was suggested that translocation of the minus-ends of the latter set of microtubules to the cortex could depend on actin and motor proteins(Januschke et al., 2006). Our data suggest that anchoring of microtubule minus-ends to the oocyte cortex depends upon an Ik2-dependent interaction of microtubule minus-ends with the F-actin network, analogous to the interaction of microtubule plus-ends with the actin cytoskeleton through microtubule tip proteins(Gundersen et al., 2004).

The phenotypes of ik2 in the ovary and adult bristles are very similar to those caused by mutations in spn-F(Abdu et al., 2006). Like ik2 mutations, null mutations in spn-F affect the localization of osk and grk mRNAs during oogenesis, and cause bicaudal and ventralized embryos. spn-F mutant oocytes have ectopic sites of F-actin polymerization, and spn-F bristles are similar to ik2 mutant bristles. Ik2 and Spn-F have been shown to interact in a yeast two-hybrid screen(Giot et al., 2003), which suggests that these proteins can form a complex. Spn-F associates specifically with microtubule minus-ends (Abdu et al.,2006). We therefore propose that Ik2 and Spn-F act together to regulate interactions between the minus-ends of microtubules and the actin-rich cortex.

We thank Nina Lampen for help with SEM, the MSKCC Molecular Cytology Core Facility for help with confocal microscopy, Byeong Cha for sharing protocols,Tatiana Kozlova, Frank Schnorrer, Christiane Nüsslein-Volhard, Trudi Schüpbach, Caryn Navarro, Ira Clark, Daniel St Johnston, and Tom Hays for Drosophila stocks, and Siegfried Roth, Tom Hays, Caryn Navarro, Ruth Lehmann, Trudi Schüpbach and Anne Ephrussi for antibodies and probes. We thank Ed Espinoza for assistance with fly work, Nina Matova for helpful comments on the manuscript, and Uri Abdu and Trudi Schüpbach for sharing unpublished results. This work was supported by NIH grant AI45149 to K.V.A. and the Lita Annenberg Hazen Foundation.