Annexin B9 binds to β_H-spectrin and is required for multivesicular body function in Drosophila

The spectrin-based membrane skeleton (SBMS) is a ubiquitous membrane-associated cytoskeletal network (Bennett and Baines, 2001). Roles for the SBMS have been demonstrated in multiple organisms and tissues including: in neuronal structure, function and membrane organization (Hammarlund et al., 2007; Hulsmeier et al., 2007; Ikeda et al., 2006; Lacas-Gervais et al., 2004; Pielage et al., 2006); in epithelial structure and stability (Lee et al., 2010; Lee et al., 1997; Thomas et al., 1998); and in muscle function (Bennett and Healy, 2008; Mohler et al., 2005). Widely regarded as a static structural element, emerging results have indicated that these proteins actually have dynamic roles in transport processes including: trans-Golgi network to plasma membrane (Kizhatil et al., 2007a); ER to Golgi (Stabach et al., 2008; Stankewich et al., 2010); at the early endosome (Phillips and Thomas, 2006); and endosome to lysosome transport (Johansson et al., 2007). Some of these roles arise through its interaction with the dynactin complex (Johansson et al., 2007; Lorenzo et al., 2010; Muresan et al., 2001), and together these data suggest that regulation of trafficking is another core function of the spectrins. The SBMS is involved in both apical and basolateral membrane organization, and growth in response to polarity cues (Johnson et al., 2002; Kizhatil et al., 2007b; Pellikka et al., 2002). Precisely how spectrin makes these contributions remains unknown.

β_Heavy-spectrin (β_H) is apically restricted in most tissues (Thomas and Kiehart, 1994) and is required for epithelial morphogenesis (Lee et al., 2010; Thomas et al., 1998; Zarnescu and Thomas, 1999). In primary epithelia, β_H is recruited by the apical polarity determinant Crumbs (Medina et al., 2002; Pellikka et al., 2002) through the FERM-binding motif in the Crumbs cytoplasmic domain (Medina et al., 2002). This motif is required to stabilize the Cadherin-based zonula adherens (ZA) (Klebes and Knust, 2000), and loss-of-function mutations in karst (which encodes β_H) result in a mild disruption of the ZA (Zarnescu and Thomas, 1999) and of the Ig-CAM, Roughest (Lee et al., 2010). This disruption probably results in the morphogenetic defects seen in karst mutants; however, the specific role played by β_H in stabilizing these junctions remains unknown. A primary role for β_H in protein and membrane trafficking is indicated by the observations that karst mutant cells exhibit endosomal defects (Phillips and Thomas, 2006), that overexpression of the C-terminus of β_H causes membrane expansion (Williams et al., 2004) and that β_H collaborates with Crumbs to regulate apical membrane size (Johnson et al., 2002; Pellikka et al., 2002). These data lead to the hypothesis that the karst phenotype arises from defects in the trafficking of multiple cargoes in the endomembrane system, including the adhesion molecules DE-Cadherin (encoded by shotgun) and Roughest.

In a genome-wide yeast two-hybrid screen, Giot and colleagues previously reported an interaction between β_H and Annexin B9 (AnxB9, also known as AnnIX) (Giot et al., 2003). We have pursued this interaction and demonstrate that AnxB9 binds to specific β_H isoforms and is responsible for intermembrane adhesion generated by expression of the β_H C-terminus, and that reduction in the level of AnxB9 causes an elevation of basolateral β_H and the appearance of elevated levels of multivesicular body (MVB) markers that overlap with β_H on cytoplasmic vesicles. β_H can also be driven into these structures either by upregulation, as found in pak mutants, or through the use of a dominant-negative β_H protein. Reduction in AnxB9 or β_H results in the accumulation of cargoes in cytoplasmic vesicles and elevated EGFR signaling, consistent with a defect in cargo progression through the MVB. We also show that reduction in AnxB9 degrades the apical–lateral domain boundary. Taken together, our data support a model in which AnxB9 is required for efficient cargo progression through the MVB and high-fidelity segregation of the apical and lateral domain. By contrast, β_H has an earlier role in cargo trafficking that is disrupted by the absence of AnxB9.

In a genome-wide yeast two-hybrid (Y2H) screen, Giot and colleagues previously identified a protein interaction between β_H and AnxB9 (AnxB9 bait #CT17989) (Giot et al., 2003). We pursued this interaction because overexpression of the 33rd segment of β_H (βH33) results in the production of intermembrane ‘junctions’ that are similar to those produced by vertebrate annexins [compare Williams et al. (Williams et al., 2004) with Lambert et al. (Lambert et al., 1997)]. Although Curagen was no longer able to supply the β_H clone originally used in the study by Giot et al., they were able to supply a short sequence indicating that it included the C-terminal portion of the protein. We therefore began by extending the previous Y2H result. FlyBase predicts four splice variants in βH33 (Fig. 1A; http://www.flybase.org), confirmed by RT-PCR and sequencing Fig. 1B). Y2H mapping revealed that AnxB9 binds to the β_H-C and β_H-D spliceoforms (Fig. 1C). Mapping experiments with AnxB10 and AnxB11 indicated that this binding is specific for AnxB9 (supplementary material Fig. S1A,B).

Antibody #182Y, raised against AnxB9, recognized both AnxB9 and AnxB10 (Fig. 1D); however, AnxB9-specific reactivity could be achieved by subtraction of cross-reacting antibodies using an AnxB10 fusion protein (Fig. 1D, antibody ‘B9’). AnxB9 showed a punctate cytoplasmic distribution during embryogenesis, with occasional cortical concentrations (Fig. 1E,F), and does not generally colocalize with β_H (Fig. 1E–F″). This suggests that the wild-type in vivo interaction between AnxB9 and β_H is transient, and it might be associated with an internal membrane compartment.

We have previously shown that expression of βH33 causes membrane expansion and intermembrane adhesions (‘bimembranes’) (Williams et al., 2004) that bear a striking resemblance to the intermembrane junctions induced by some vertebrate annexins (Lambert et al., 1997). Co-staining for βH33 and AnxB9 in the salivary gland of the developing embryo showed that the two proteins colocalize in bimembranes (Fig. 1H–H″) indicating a close relationship between the two proteins independently from the Y2H assay. Interestingly, the minimal region that will recruit AnxB9 into bimembranes is amino acids 3560–3920 [the construct βHPH+5-3 in (Williams et al., 2004), extending to the arrowhead in Fig. 1A, ‘splicing 33’], which does not contain the AnxB9-binding site, so there must be both a direct and an indirect mechanism for AnxB9 to bind to βH33.

To test the role of AnxB9 in β_H function and bimembrane formation we used an inducible RNAi line (UAS-AnxB9^RNAi, hereafter AnxB9^RNAi) to knockdown expression of this protein. We chose to examine the role of AnxB9 in the salivary gland because this tissue does not express AnxB10 (see below), eliminating any ambiguity due to antibody or RNAi cross-reaction, and because the large size of these cells permits the visualization of cytoplasmic compartments. The AB1-GAL4 (hereafter AB1) and 185Y-GAL4 drivers used initiate expression shortly after gland invagination and persist throughout larval life. Immunoblot analysis of third-instar glands from AB1>AnxB9^RNAi individuals indicated that we could achieve substantial knockdown of AnxB9 using this construct (Fig. 2A). RT-PCR analysis on all fly Annexins, indicated that AnxB10 was not expressed in the third-instar gland and that no knockdown of AnxB11 was observed (Fig. 2B).

We next tested to see whether AnxB9^RNAi expression would modify βH33-induced bimembranes. In 185Y-GAL4> AnxB9^RNAi + βH33 embryonic salivary glands, bimembranes appeared rapidly but subsequently faded away with time, presumably owing to the eventual knockdown and turnover of AnxB9 (Fig. 2C–C″). This suggests a role for AnxB9 in the βH33–membrane interaction. Examination of late-stage 185Y-GAL4> AnxB9^RNAi + βH33 glands by transmission electron microscopy (TEM), showed that reducing AnxB9 resulted in the loss of all bimembranes (Fig. 2D,E). This suggests that AnxB9 is responsible for the intramembrane adhesion induced by βH33 expression and that bimembranes are an exaggerated manifestation of a normal functional partnership.

Because no mutations are available in AnxB9, we stained for β_H in glands where AnxB9 was knocked down, to test the role of AnxB9 in β_H localization. Whereas β_H is exclusively apical in most epithelia (Thomas and Kiehart, 1994), it also exhibited weak lateral and basal staining in wild-type glands, where it was confined to below the septate junctions (SJ; Fig. 2F). β_H was also seen on inward membrane folds on the basal surface (Fig. 2F,G). In AB1>AnxB9^RNAi glands β_H was still seen at the apical membrane but was increased on the basolateral surfaces (Fig. 2H). In addition, β_H was present on a number of internal vesicular structures in AB1>AnxB9^RNAi glands (arrows in Fig. 2H). These varied in morphology from small puncta or vesicles to larger more complex structures (broken line in Fig. 2I) that often appeared to have connections to the lateral or basal membranes (arrows in Fig. 2I). The overall levels of β_H in the gland were not detectably perturbed by the reduction in AnxB9 (Fig. 2J), and these effects were not seen when we knocked down AxnB11 using any of three different lines available from the Vienna stock center (data not shown), indicating that these effects are specific to AnxB9.

α-Spectrin and basolateral β-spectrin were also present on the internalized structures (Fig. 2K–M″). Antibody incompatability prevents co-staining for β_H and β-spectrin; however, co-staining for each β-chain with α-spectrin showed that α-spectrin and β_H are predominantly in separate domains (Fig. 2K–K″; supplementary material Fig. S2A,B,D), whereas α-spectrin and β-spectrin colocalization is fairly precise (Fig. 2M–M″; supplementary material Fig. S2C,E). Thus, β_H might not be associated with α-spectrin on these structures, and segregation of the SBMS is occurring into α+β and β_H-only domains. These results indicate that AnxB9 has a role in the apical restriction of β_H and that a subpopulation of β_H accumulates on internal vesicles and tubules in the absence of AnxB9.

Previous data has suggested a role for β_H in the endosome pathway (Phillips and Thomas, 2006), and vertebrate annexins are required for endosome function and organization (Futter and White, 2007). We therefore stained the AB1>AnxB9^RNAi glands for various endosomal compartments. Four markers for the multivesicular body (MVB)-late endosome pathway exhibited altered distributions: Hrs, which functions during MVB formation as part of the endosomal sorting complex required for transport-0 (ESCRT-0) (Lloyd et al., 2002); EPS15, a molecular partner of Hrs with functions both at MVBs and during endocytosis (Roxrud et al., 2008); and Vps16, as well as Rab7, which controls endosome to lysosome trafficking (Pulipparacharuvil et al., 2005; Nickerson et al., 2009). Other compartment markers, such as Rab4 and Rab5 (early endosome), and Rab11 (recycling endosome), were not visibly perturbed (data not shown). Although previous data linked β_H to the Golgi-resident protein Lava lamp (Sisson et al., 2000), the distribution of this protein was also unaffected by AnxB9^RNAi (supplementary material Fig. S3A–A′).

Hrs was normally present in peripheral puncta (Fig. 3A–A‴) where it colocalized with Vps16 (supplementary material Fig. S3B–B″). In AB1>AnxB9^RNAi glands, larger Hrs vesicles were seen throughout the cytoplasm, with smaller puncta seen in discontinuous circular patterns (Fig. 3B–B″), suggesting that Hrs is present on subdomains of larger structures. Hrs was also present on structures that resemble β_H internalizations (see below). Vps16 exhibited a similar perturbation upon AnxB9 depletion (compare Fig. 3C with 3D) and continued to be associated with Hrs (supplementary material Fig. S3C–C″). Vps16 was also present in the nuclear region in a single elongated structure that was often branched and appeared to protrude into the nucleoplasm (Fig. 3C′). This was largely unperturbed in AB1>AnxB9^RNAi cells (Fig. 3D′). In wild-type cells, Rab7 was present in puncta that were concentrated at the cell periphery (Fig. 3E). In AB1>AnxB9^RNAi glands, puncta were more commonly seen deep within the cytoplasm and the signal was also seen on larger circular structures, although not with the frequency of Hrs or Vps16 (Fig. 3E′). EPS15 exhibited a slightly different response to AnxB9 knockdown. In wild-type cells, EPS15 was present in peripheral puncta and faintly at the apical surface (Fig. 3F,G). In AB1>AnxB9^RNAi cells these puncta were found throughout the cytoplasm but not in the distinctive circular groupings seen with Hrs and Vps16 (Fig. 3H). However, EPS15 was also present on basal and lateral structures like β_H, Hrs and Vps16 (arrowheads in Fig. 3H).

Taken together, these data indicate that reducing the levels of AnxB9 causes the abnormal accumulation of vesicular structures labeled with MVB markers, suggesting a role for AnxB9 in the creation and progression of this compartment.

The karst endosome phenotype (Phillips and Thomas, 2006) and the appearance of cytoplasmic β_H in cells expressing AnxB9^RNAi (Fig. 2), suggests that AnxB9 interacts with β_H during protein recycling and/or degradation. Co-staining for β_H and Hrs or EPS15 in AB1>AnxB9^RNAi glands reveals that both Hrs- and EPS15-positive structures overlap with cytoplasmic β_H (Fig. 4B–C″). The lack of detectable cytoplasmic β_H in wild-type cells (Fig. 4A) suggests that β_H is recruited de novo to these vesicles upon AnxB9 knockdown or that the equilibrium level of β_H on such structures in wild-type cells is low or hard to fix. We therefore investigated other methods to probe the role of β_H in these processes.

The Rac effector Pak is a negative regulator of β_H and the viable genotype pak⁶/pak¹¹ causes an increase in apical β_H levels (Conder et al., 2007). Upregulation of β_H in this genotype revealed β_H to be present on conspicuous vesicular structures, where it colocalized with Hrs (Fig. 4D–D″). Further evidence for an association of β_H with endosomal structures comes from the analysis of the dominant-negative β_H construct Minikarst, a β_H derivative that lacks only segments 14–28 and is tagged with mCherry (Fig. 4E). The Minikarst transgene encodes all four C-terminal spliceoforms, but cannot rescue the karst mutation and results in lethality or a variety of visible phenotypes when expressed in a wild-type background (data not shown). In the salivary gland (AB1>Minikarst), Minikarst protein accumulated in distinct patches at the membrane (arrows in Fig. 4F; a facing view of a lateral membrane, created by projection, is shown in Fig. 4G) and also on vesicular structures in the cytoplasm (arrowhead in Fig. 4F). Minikarst vesicles also labeled with Hrs (arrowhead in Fig. 4F–F″) and with Vps16 (Fig. 4H–H″), again suggesting that they are endosomal in origin and that this protein causes a block in the endosomal pathway at a similar step to AnxB9 knockdown.

These data represent three diverse and independent treatments, all of which result in β_H accumulation on cytoplasmic structures positive for MVB markers. Taken together, this evidence suggests that β_H has a normal role in the endosomal system but is present transiently or at levels that are too low to detect with our current reagents. The normal role for AnxB9 might be to release β_H from such compartments.

The accumulation of vesicles labeled with MVB markers suggests that AnxB9 is involved in MVB function. Perturbation of MVB formation by mutations in the ESCRT 0–III complexes causes large accumulations of specific cargoes and overproliferation of imaginal tissue (Vaccari et al., 2009). We did not detect overproliferation in any tissue examined upon AnxB9 knockdown (data not shown), and we did not see large accumulations of specific cargoes (see below). This suggests that the roles for β_H and AnxB9 at the MVB are modulatory rather than as a key part of the core machinery, leading to milder defects. Nonetheless, if AnxB9 does modulate traffic through the MVB, expression of AnxB9^RNAi should have a detectable affect on many cargoes. We therefore stained for Ubiquitin as a proxy for proteins entering endosomes; because Ubiquitin is removed during intralumenal vesicle formation at the MVB, its accumulation is a sensitive indicator of MVB defects. In wild-type glands Ubiquitin was present in a few dispersed puncta and at the apical surface in both the anterior and posterior part of the gland (Fig. 5A–A‴). Ubiquitin was also concentrated in puncta in the perinuclear region. In AB1>AnxB9^RNAi glands, Ubiquitin puncta were elevated in number and size in both the anterior and posterior gland (Fig. 5B,C). In the posterior gland, Ubiquitin puncta were concentrated in the apical cytoplasm (Fig. 5B), whereas in anterior regions the puncta were concentrated in the central cytoplasm (Fig. 5C). These results are consistent with the hypothesis that AnxB9 is required for cargoes to progress to the de-ubiquitylation step at the MVB.

If AnxB9 partners with β_H in the endosome system, a reduction in the levels of β_H should also cause accumulation of Ubiquitin. In glands expressing β_H^RNAi, we indeed observed accumulation of Ubiquitin, consistent with a role for β_H in cargo progression to and through the MVB (Fig. 5D). However, the pattern was strikingly different from that in cells expressing AnxB9^RNAi, as Ubiquitin was concentrated in large prominences at or near the apical membrane. The distinct nature of these results permitted a simple epitstasis test; in AB1>AnxB9^RNAi + β_H^RNA glands Ubiquitin accumulated at the apical membrane (Fig. 5E), formally putting β_H upstream of AnxB9.

These observations are consistent with the hypothesis that MVB function is decreased when AnxB9 is reduced, causing the accumulation of ubiquitylated cargoes. Our observations also demonstrate that β_H probably has an earlier role in cargo movement than AnxB9.

An increase in receptor signaling is a hallmark of mutations in endosome and MVB functions because receptors signaling from endosomes do not progress promptly to the MVB for final inactivation (Vaccari and Bilder, 2009). The EGF receptor (EGFR) is one such receptor (Vaccari et al., 2009); hence, we examined changes in EGFR activity when β_H or AnxB9 were reduced. Because we believe that β_H and AnxB9 do not provide core MVB functions, and they lead to mild MVB phenotypes, we looked for increased EGFR activity in the sensitized background provided by the rhomboid^ve (rho^ve) allele. rho^ve is a regulatory mutation that reduces the production of the primary EGFR ligand, Spitz (Sturtevant et al., 1993), and homozygous rho^ve flies exhibit truncated wing veins (Fig. 6B) because EGFR plays an essential role in wing vein formation (Shilo, 2005). Introduction of one copy of the kst¹ allele into this genetic background results in complete suppression of the vein defect for L2–L4 (Fig. 6C′,D′) and partial suppression for L5 (Fig. 6F–F″). The kst² allele exhibited a similar, but milder suppression (Fig. 6D″). This suggests that reduction in the amount of β_H results in an increase in EGFR activity. Similarly, if we reduce the levels of AnxB9 in the wing blade (MS1096-Gal4>AnxB9^RNAi) in a rho^ve background, vein formation is also restored (Fig. 6E′).

To specifically link wing vein restoration to MVB formation, we took advantage of the fact that kst¹ does not fully suppress the formation of vein L5, and introduced representative alleles in genes encoding endosomal and MVB functions to the rho^ve kst¹/rho^ve+ flies to see whether further suppression would be achieved. Whereas <20% of L5 veins were suppressed by kst¹ alone, alleles at several loci resulted in complete restoration of L5 in a large majority of flies. None of these alleles caused any visible suppression of rho^ve when present in the absence of a karst allele (data not shown). This synergy strongly suggests that the suppression of the rho^ve phenotype by karst alleles is as a result of a role for β_H in endosome progression through the MVB.

β_H generally colocalizes with the ZA and can also be present on the apical surface (Thomas and Kiehart, 1994; Thomas et al., 1998), and karst (β_H) mutations cause a mild variable disruption of the ZA (Zarnescu and Thomas, 1999; unpublished results). Given the emerging role for β_H in protein recycling, this phenotype might arise from inappropriate trafficking of DE-Cadherin in the absence of β_H. We therefore examined the effects of AnxB9^RNAi and β_H mutations on the distribution of DE-Cadherin. Because this aspect of the karst phenotype is relatively weak and variable, we sensitized the system to β_H and AnxB9 defects by overexpressing DE-Cadherin.

Imaging of DE-Cadherin–GFP in AB1>DE-Cadherin–GFP glands revealed a lateral and basolateral accumulation of this protein and caused a distinct bulging of the basal surface (Fig. 7A–A″). In such glands, β_H exhibited extensive colocalization with DE-Cadherin–GFP in the basolateral domain, and is more highly concentrated where there is more DE-Cadherin–GFP (Fig. 7A–B″). However, β_H on the apical surface retained its independence.

In AB1>AnxB9^RNAi + DE-Cadherin–GFP glands there was a striking accumulation of DE-Cadherin–GFP on internal vesicles (Fig. 7C–D″). Some of these structures were very large, Hrs-negative vesicles (e.g. arrowhead in Fig. 7D′), which we have not yet been able to identify. A second population is much smaller and is Hrs-positive, identifying them as endosomal (insets Fig. 7D–D″). We interpret this result to indicate that DE-Cadherin–GFP recycling and/or degradation is being slowed owing to the loss of AnxB9, leading to its accumulation to high levels in intermediate compartments.

We next reduced the levels of β_H by introducing one copy of various karst alleles or β_H^RNAi into an AB1>DE-Cadherin–GFP background. AB1>DE-Cadherin–GFP; kst¹/+ glands exhibited a striking accumulation of small punctate DE-Cadherin–GFP signal in the subapical cytoplasm (Fig. 7E), and AB1>DE-Cadherin–GFP + β_H^RNAi glands, where β_H is substantially eliminated (data not shown), accumulated larger DE-Cadherin–GFP vesicles (Fig. 7F). The effect of reducing β_H is thus very similar to loss of AnxB9 and is consistent with the notion that these two proteins collaborate in DE-Cadherin trafficking.

Finally, we tested the effects of Minikarst on DE-Cadherin–GFP distribution. In AB1-Gal4>DE-Cadherin–GFP + Minikarst glands, there was a striking suppression of the basal bulging phenotype and a distended lumen was often observed (Fig. 7G). Again DE-Cadherin–GFP is present in two classes of vesicular structures: large DE-Cadherin–GFP vesicles were Minikarst negative (Fig. 7H′), whereas the smaller ones (often associated with basal invaginations) were Minikarst positive (inset in Fig. 7H–H″) and endosomal (see Fig. 4 for Minikarst and Hrs colocalization). These data are also consistent with the hypothesis that β_H and AnxB9 both act to modulate DE-Cadherin trafficking. Interestingly, the Minikarst interaction was somewhat distinct from AnxB9 and karst or kst^RNAi, in that the basal domain had approximately the same level of DE-Cadherin–GFP as the lateral membranes and the bulging phenotype was suppressed. This suggests that β_H has more than one role in conjunction with DE-Cadherin and that Minikarst disrupts more than one of these roles.

Vertebrate AnxA2 has a major role in apical domain development in MDCK cells (Martin-Belmonte et al., 2007). Cells expressing AnxB9^RNAi remain polarized and retain a lumen despite lowering AnxB9 to undetectable levels. However, the elevation of β_H in the basolateral domains along with the appearance of β_H-positive endosomal structures associated with those domains, and the effects on basolateral DE-Cadherin–GFP trafficking, all point to a role for AnxB9 in basolateral trafficking despite being identified as a partner of an overtly apical protein (i.e. β_H). We therefore wondered whether AnxB9 might be important for maintaining the basolateral restriction of polarity markers.

Co-staining for β_H and the basolateral group protein Coracle, showed that Coracle was concentrated at the SJ, which lacks β_H (Fig. 8A–A″). On lateral membranes below the SJ, both β_H and Coracle were present (Fig. 8B–B″). Coracle was also present on small cytoplasmic puncta (Fig. 8B″), and β_H was concentrated at the basal edge of the SJ (arrows in Fig. 8A–A″). The two proteins exhibited a precise segregation at the apical–lateral margin (Fig. 8C–C″). In AB1>AnxB9^RNAi cells, β_H and Coracle still resided in separate domains, but staining at the apical–lateral boundary showed some intermingling (Fig. 8D–D″). In addition, Coracle now showed overlapping staining with internal β_H structures (Fig. 8E–E″). A similar blurring of the apical–lateral boundary was seen with Discs large protein (Dlg; Fig. 8F,G), and Dlg was also seen on basolateral internalizations (Fig. 8I).

Taken together, these results suggest that AnxB9 has a role in maintaining not only the apical bias of the β_H network, but in the segregation of the apical and lateral domains. Furthermore, the appearance of both β_H and basolateral markers in a close proximity on internal compartments, suggests that a role of AnxB9 is to facilitate their segregation at or in conjunction with a compartment where apical and basal proteins are normally sorted from one another.

Here, we describe a physical and genetic relationship between AnxB9, β_H and markers of MVBs. β_H is primarily apical, and loss of AnxB9 results in an increase in basolateral β_H and its appearance on cytoplasmic structures that overlap with the MVB markers. Similar colocalizations are seen when β_H internalization is generated either in pak mutants or through the expression of a dominant-negative version of β_H. We also show that there is an accumulation of ubiquitylated proteins in the absence of AnxB9, and that loss of AnxB9 or β_H causes elevated EGFR signaling and the redistribution of DE-Cadherin to endosomal vesicles. We also demonstrate that reduction of AnxB9 results in degradation of the apical–lateral boundary and the appearance of the basolateral proteins Coracle and Dlg on vesicles adjacent to β_H-spectrin.

Annexins have been widely associated with the endomembrane system (Futter and White, 2007; Gerke et al., 2005; Grewal and Enrich, 2009), where AnxA1 is required for inwards vesiculation of intralumenal vesicles at the MVB (White et al., 2006). AnxA6 participates in late endosome to lysosome transport (Grewal et al., 2000; Grewal et al., 2010; Pons et al., 2001), binds to spectrin and associates with this protein on endosomes (Grewal et al., 2000; Kamal et al., 1998; Watanabe et al., 1994). In the absence of AnxB9 in Drosophila, we see accumulation of MVB protein markers and a failure to de-ubiquitylate protein cargoes, showing that these are conserved functions. However, it is not possible to say which vertebrate annexin(s) are true orthologs of AnxB9 becuase the annexins underwent independent expansions in different phyla (Fernandez and Morgan, 2003; Moss and Morgan, 2004). In addition, the presence of 12 annexins in vertebrates, but only three in the fly, suggests that each fly protein might have functions that overlap with multiple vertebrate isoforms. The observation that reducing the levels of AnxB9 does not lead to overproliferation or to large elevations in the level of apical proteins, such as Crumbs and EGFR, as seen with mutations in core endosome and MVB functions (Chanut-Delalande et al., 2010; Lu and Bilder, 2005; Vaccari et al., 2009), suggests a modulatory role for this protein in MVB biology. The loss of annexins from the Saccharomyces cerivisiae genome (Fernandez and Morgan, 2003) and the viability of AnxA6-knockout mice (Hawkins et al., 1999) also suggest that these proteins are not part of the core MVB machinery.

The precise role for AnxB9 in MVB formation and function remains to be elucidated. The trapping of elevated levels of α-, β- and β_H-spectrins on MVB-related structures with conspicuous connections to the plasma membrane in the absence of AnxB9, suggests that spectrin has traveled in from the plasma membrane during internalization and that AnxB9 is required to release it from endosomal structures. In support of this hypothesis, quick-freeze deep-etch images have shown that spectrin remains associated with freshly internalized vesicles for some distance below the plasma membrane (Hirokawa et al., 1983), and AnxA6 has been suggested to induce spectrin proteolysis to facilitate clathrin-coated pit release (Kamal et al., 1998). We speculate that AnxB9 similarly triggers spectrin proteolysis to release it from endosomes during MVB formation (Fig. 9). In addition, our observation that AnxB9 is responsible for intermembrane adhesion on the cytoplasmic leaflet (see Fig. 2) (Williams et al., 2004) suggests that it could also participate in inter-endosome adhesion and fusion, or directly in intralumenal vesicle formation, as suggested for other annexins (Fig. 9) (Futter and White, 2007). The epistasis test between AnxB9 and β_H knockdown suggests that β_H acts upstream of AnxB9 in the endosome pathway and is fully consistent with the model proposed in Fig. 9. Future work with appropriate transport assays will permit direct mechanistic testing of this model.

Historically, the SBMS has been seen to modulate protein endocytosis as a physical barrier to coat protein assembly (e.g. Marshall et al., 1984) or as an anchor to increase protein half-life at the membrane (e.g. Hammerton et al., 1991). Although these roles undoubtedly exist (>50 proteins bind to the SBMS) (De Matteis and Morrow, 2000), these mechanisms are largely passive and inspired by the omnipresent erythrocyte model. Hints of a more dynamic life were obtained through visualization of spectrin bound to endocytic vesicles in the terminal web (Hirokawa et al., 1983) and the identification of βIII spectrin as the anchor for the dynactin complex (Holleran et al., 2001; Muresan et al., 2001). However, spectrin is now emerging as a significant modulator of trafficking processes. Spectrin binding is an essential step for Rab7-stimulated dynein activation during transport to lysosomes (Johansson et al., 2007) and for the localization of dynactin to costameres in muscle (Ayalon et al., 2011), suggesting that spectrin is an integral part of the dynein-dynactin system. Furthermore, expression of mutant β-spectrin isoforms in the fly leads to dynein-dynactin-based axonal transport defects (Lorenzo et al., 2010). Finally, β₂ spectrin is required for trans-Golgi network to lateral membrane transport in HBE cells (Kizhatil et al., 2007a), a pathway that is sufficiently vigorous that its disruption leads to a significant shortening of the lateral domain (Kizhatil et al., 2007b). Our data significantly adds to the view that spectrin represents an important interface between the actin cytoskeleton and endomembrane transport processes.

β_H is conspicuously associated with the ZA (Lee et al., 2010; Thomas and Kiehart, 1994; Thomas et al., 1998; Zarnescu and Thomas, 1999) and is required for its integrity (Zarnescu and Thomas, 1999). Given the growing association of β_H with protein recycling (Phillips and Thomas, 2006; Williams et al., 2004) (this paper), we hypothesize that this phenotype arises from defective or inappropriate trafficking of DE-Cadherin. The observation that DE-Cadherin relocates to vesicular structures when levels of β_H are reduced supports this idea. In addition, the karst ZA phenotype is quite variable, and the requirement for AnxB9 for stress resistance suggests that this variability arises because DE-Cadherin recycling is less robust in the absence of these proteins. The degradation of the apical–lateral boundary in AnxB9^RNAi cells might also reflect a problem with this lateral diffusion barrier.

The observations that the apical–lateral margin is degraded when levels of AnxB9 are reduced, and that cytoplasmic spectrin in AnxB9^RNAi cells segregates into apical (β_H) and basolateral (α-spectrin, β-spectrin Coracle, Dlg) domains, suggests that the annexin-mediated process we have uncovered has some relationship to apical–lateral sorting. We note that reduction of PATJ, another apical Crumbs partner, has also been shown to cause mislocalization of lateral TJ proteins when it is knocked down in CACO2 cells (Michel et al., 2005) and that there is an increasing realization of the significance of protein endocytosis and recycling in apicobasal polarity maintenance (Harris and Tepass, 2010). It will be interesting to clarify this relationship in future work.

In conclusion, we have shown that AnxB9 is a molecular partner of β_H-spectrin that is required for efficient protein trafficking through the MVB and for robust segregation of apical and lateral proteins in Drosophila. Our results also provide a strong rationale for the stabilization of the ZA through modulation of protein trafficking by β_H and strengthen an accelerating body of evidence that spectrin is intimately involved in protein trafficking events.

To produce an antibody to AnxB9, exon 2 was amplified from genomic DNA using the primers 5′-cggaattctctagagcCGTCGAGGATGCGGCTATTCTGC-3′ and 5′-ccatcctcgaggctctagacGCCGCTAAACTCCCGCTTGATGG-3′. This fragment was expressed as a GST fusion protein in pGex-4T1, purified by standard methods and used to immunize a leghorn chicken. IgY antibodies (#182Y) were purified from egg yolks and recognize AnxB9 and AnxB10.

To specifically detect AnxB9 we subtracted antibodies that cross-reacted with AnxB10. To generate the AnxB10 fusion protein, exons 2–4 were amplified by RT-PCR using the primers 5′-cgtcccccgggGCCCACGGTTAAGGACGCAG-3′ and 5′-gcccgctcgagCAGGGCCCGCTTGTAGTCA-3′. This fragment was expressed as a GST fusion protein in pGEX-4T1, purified, immobilized onto nitrocellulose and used to remove all detectable cross-reaction with AnxB10.

Other antibodies used were: rabbit anti-β_H antibody, which was affinity purified by standard methods using the immunogen for serum #243 (Thomas and Kiehart, 1994) (1:100); mouse anti-Actin antibody (#C4; 1:25,000), which was obtained commercially; mouse anti-Crumbs antibody (#Cq4; 1:25), which was obtained from Elisabeth Knust (Max Planck Institute of Molecular Cell Biology and Genetics, Dresden, Germany) (Tepass et al., 1990); mouse monoclonal anti-Myc antibody (1:100; Oncogene Research Products, San Diego, CA); mouse anti-Ubiquitin antibody (1:1000) was obtained from Enzo Life Sciences (Plymouth Meeting, PA); Guinea pig anti-Hrs antibody (1:800) and anti-EPS15 antibody were obtained from Hugo Bellen (Baylor College of Medicine, Houston, TX); Chicken anti-Avalanche antibody (1:500) was obtained from David Bilder (University of California Berkeley, Berkelely, CA); mouse anti-Coracle antibody (1:50) was obtained from Richard Fehon (University of Chicago, Chicago, IL); pabbit anti-Lava-lamp antibody (1:5000) was obtained from John Sisson (University of Texas, Austin, TX); rabbit anti-dVps16A antibody (1:1000) was obtained from Helmut Kramer (UT Southwestern Medical Center at Dallas, Dallas, TX); rabbit anti-Rab7 antibody (1:100) was obtained from Patrick Dolph (Dartmouth College, Hanover, NY); Alexa-Fluor-labeled secondary antibodies were obtained from Invitrogen and were used at 1:250 following pre-adsorbtion against fixed wild-type embryos.

One- and two-dimensional PAGE followed standard protocols. All blotting was onto nitrocellulose filters which, were blocked and probed in Tris-buffered saline containing 5% dried milk powder and 0.1% Tween 20. Immunoblot detection utilized horseradish-peroxidase-conjugated secondary antibodies from Jackson Immunoresearch and chemiluminescent substrates from Pierce Biotechnology.

For embryo immunostaining, appropriate collections of embryos were fixed using 4% paraformaldehyde (PFA) for 20 minutes with shaking, as previously described (Thomas and Kiehart, 1994).

Third-instar salivary glands were dissected and chilled as rapidly as possible without separating glands from other organs until after fixation. Thus, dissection in PBS and transfer to ice-cold PBS was performed in 2–5 seconds per larvae. Fixation was performed on ice in 4% (w/v) PFA in PEM buffer on ice for 30 or 60 minutes with gentle agitation, followed by five PBS rinses, and blocking, extraction, staining and washing in incubation solution (10% normal goat serum, 0.2% Saponin, 0.3% deoxycholate and 0.3% Triton X-100 in PBS). For Crumbs staining, samples were post-fixed in 100% methanol for 10 minutes and rehydrated through a methanol-PBT (PBS with 0.1% Tween 20) series before staining as above.

Embryos were imaged using a Zeiss LSM 510 META confocal. Salivary glands were imaged on a CARV II spinning disc confocal (BD Biosystems). Embryos were prepared for electron micrcoscopy according to Tepass and Hartenstein (Tepass and Hartenstein, 1994) and imaged on a JEOL JEM 1200 EXII transmission electron microscope.

Oregon-R or the transformation host yellow white were used as control lines. karst stocks were as described previously (Thomas et al., 1998; Zarnescu and Thomas, 1999). Stocks carrying the mutant alleles rab5², rabenosyn^40-3, vps28^B9, vps25^A3, vps20^I3, and vps32^G5 were obtained from David Bilder (University of California, Berkeley, CA). hrs^y28 was obtained from Hugo Bellen (Baylor College of Medicine, Houston, TX). eps15^E75 (#24900), Df(3L)ru-22 (#4214; uncovering rhomboid) and Df(3L)Ar14-8 (#439; uncovering rhomboid), as well as the driver lines AB1-Gal4 (#1824), 185Y-Gal4 (#3731) and MS1096-Gal4 (#8860) were obtained from the Bloomington Stock Center (Bloomington, IN). RNAi knockdown lines specific for karst (#37074 and #37075), AnxB9 (#27493 and #106867) and AnxB11 (#29693, #36186 and #101313) were obtained from the Vienna Drosophila RNAi Center (Dietzl et al., 2007). In addition, we also made our own AnxB9 RNAi line by cloning two copies of the above fragment, in opposition, into a modified pUAST vector (Brand and Perrimon, 1993), containing an intron from OAMB (a gift from Kyung-An Han, University of Texas at El Paso, TX). The intron separates the two inserts and promotes stability of the clone. Transformed lines (UAS-AnxB9^RNAi) were produced by standard methods (Rubin and Spradling, 1982). RNAi specificity was confirmed by semi-quantitative RT-PCR for AnxB9, AnxB10 and AnxB11 using the primer pairs: 5′-CAAAATGAGTTCCGCTGAGT-3′ and 5′-AATGGTCTTGATGCCGTAGTT-3′; 5′-CGGCACCGACGAGCAGGAAATC-3′ and 5′-GGGCCCGCTTGTAGTCACCAGAGG-3′; and 5′-CCAACGAGCAGCGCCAGGAGAT-3′ and 5′-CGCAGTTCGCCGGCTTTCAGTAG-3′, respectively (all pairs span introns to detect DNA contamination). Primers 5′-TACAGGCCCAAGATGGTGAA-3′ and 5′-ACGTTGTGCACCAGGAACTT-3′ were used to detect ribosomal protein Rps49, as a control.

Minikarst, a dominant-negative β_H construct, was made by deleting segments 14–28 (see Thomas et al., 1997). Minikarst was built from a cDNA segment encoding amino acids 1–1605 and a genomic clone starting at amino acid 3200 through to the normal stop codon. To facilitate cloning, two amino acids (Ser-Arg) were inserted at the joint between these two segments. The Minikarst transgene therefore lacks repeats 14–28 but retains the actin-binding domain, the SH3 domain, the dimer nucleation site and the tetramerization site and encodes all spiceoforms in segment 33. The stop codon was deleted and the protein fused to mCherry (Shu et al., 2006). This construct was cloned into pUASTattB and transformed lines (UAS-Minikarst) were produced by integration to the chromosome 3R attP landing site at 99F8 (stock #BL24867) by Rainbow Transgenic Flies (Newbury Park, CA).

We thank many investigators for supplying antibodies, as well as the Bloomington and Vienna Stock Centers. We also thank Missy Hazen and Ruth Haldeman for electron microscope training and assistance, Richard Cyr and Simon Gilroy for use of their confocal early in this study, David Gilmour for the rp49 primers, Curagen for supplying their AnxB9 clone and Scott Selleck for the use of Imaris. This work was funded by NSF grant #0644691 to C.M.T.