Protection of Armadillo/β-Catenin by Armless, a Novel Positive Regulator of Wingless Signaling

This study uses an RNAi screen in Drosophila to identify a UBX protein, Armless, as a novel positive regulator of the important Wingless/Wnt signaling pathway, acting to stabilize Armadillo/?-Catenin by antagonizing its turnover.


Introduction
The wingless (wg) gene was found nearly forty years ago with the characterization of a Drosophila mutant without wings [1]. The gene encodes a secreted glycoprotein, the founding member of the Wnt family of signaling proteins [2]. In the decades following its discovery, Wg/Wnt signaling has been shown to be essential during embryogenesis. Indeed, it is important throughout an organism's life, controlling also the homeostasis of different organs, for example, regeneration of epithelial cells in the intestine-the aberrant behavior of these cells in cancer is caused by constitutive Wg/Wnt signaling, which is consequently a key focus of medical and translational research [3,4].
The relay of the Wg signal is controlled at different levels. However, the pivotal step is the regulation of the levels of Armadillo (Arm)/b-Catenin, the key transducer of the Wg/Wnt pathway. A multiprotein complex consisting of the scaffold proteins Axin and APC and the kinases Shaggy/GSK3b and Casein kinase I (CKI) recruits and phosphorylates Arm/b-Catenin. This marks Arm/b-Catenin for ubiquitination by the SCF/Slimb/bTRCP E3 ubiquitin ligase and subsequent degradation by the ubiquitin-proteasome system (UPS). When Wg/Wnt binds its receptors at the cell membrane, degradation of Arm/b-Catenin is prevented, presumably by protein interactions that lead to the dissociation of the E3 ubiquitin ligase from the Arm/b-Catenin destruction complex [5]. As a consequence, Arm/b-Catenin translocates into the nucleus, where it adopts its role as a transcriptional effector of Wg/Wnt signaling. Although this step is crucial, and is a potential point of regulation, little is known about the players involved in the processing of Arm/b-Catenin and its ultimate degradation.
In a genome-wide RNA interference (RNAi) screen we isolated Armless (Als) as a regulator of proximodistal growth of Drosophila limbs, and show in subsequent analyses that it exerts its function in the Wg pathway. Detailed genetic studies demonstrate that Als acts downstream of the destruction complex, at the level of the SCF/Slimb/bTRCP E3 Ub ligase. Cells depleted for Als exhibit strongly reduced Arm protein levels. Importantly, the activity of a constitutively active form of Arm, Arm S10 , which cannot be phosphorylated and hence escapes ubiquitination and proteasomal degradation, is insensitive to depletion of Als. Using immunopurification and mass spectrometry analysis we found that Ter94 interacts with Als. Ter94 is an AAA ATPase associated with protein turnover and proteasomal degradation [6]. In sum, our data suggest that Als acts downstream of the Arm/b-Catenin destruction complex to positively regulate Arm protein levels, possibly by rescuing Arm from ubiquitination via Slimb. The human ortholog of Als, UBXN6, can substitute for Als in Drosophila, and Wnt target gene expression was impaired upon knock-down of UBXN6 in HEK-293 cells. We thus infer that Als and UBXN6 represent regulators of a conserved mechanism that ensures appropriate levels of Armadillo/b-Catenin by antagonizing its entry into the UPS.

CG5469/armless Is Required for Wing Growth
Among its numerous developmental roles, Wg controls growth and patterning of the Drosophila wing. A hallmark phenotype of reduced wg function is nicked wings with reduced size. Positively acting components of the Wg pathway mostly display similar phenotypes when depleted via RNAi during larval stages. We established and used a tester system that allowed screening for genes involved in wing growth and margin formation. Our screen covered approximately 83% of all the predicted 14,306 genes of Drosophila. The details of the screen will be published elsewhere. CG5469 behaved as expected of a putative Wg signaling component: RNAi targeting its transcripts affected wing size in a non-allometric manner and caused a notched margin and loss of sensory bristles. This phenotype could be reproduced by the expression of nine independent UAS-CG5469 RNAi lines with different Gal4 drivers ( Figures 1B-1E and S1A-S1J). Expression of CG5469 RNAi effectively reduced mRNA levels of CG5469 ( Figure S1L, cf. Figure 1B-1E), and haplo-deficiency exacerbated the wing phenotypes ( Figure S1M and S1N); these phenotypes were rescued upon overexpression of CG5469, as detailed in Text S1 and shown in Figure S2.
Wing notches could be caused by increased apoptosis or reduced cell proliferation. To discriminate between these possibilities, we experimentally accelerated cell divisions by string/cdc25 overexpression [7], and inhibited apoptosis by Diap1 overexpression. Only string/cdc25 overexpression restored marginal wing tissue loss caused by depletion of CG5469 ( Figure  S3F). This finding is consistent with a role of CG5469 in Wg signaling, as impaired Wg pathway activity affects primarily proliferation (Figure S3A-S3C; [8,9]). Because of its mutant phenotypes, which are caused by severely reduced levels of Arm (see below), we refer to CG5469 as armless (als; for the nomenclature see also Text S1).

Author Summary
The Wg/Wnt signaling pathway, found in most animals, is essential for regulating tissue growth and the formation of different cell types during development. Defects in the Wg/Wnt signaling relay can have serious consequences, ranging from aberrant organ patterning to malignant tumor formation. A pivotal step in the transmission of the Wg/Wnt signal is the stabilization of the protein Armadillo/ b-Catenin, a key component of the pathway. However, the means by which the levels of this protein are regulated remain unclear. Here, we describe a novel control point of Armadillo/b-Catenin levels. Using RNA interference, we performed a screen in the fruit fly Drosophila melanogaster and identified Armless, a protein whose biological function was previously unknown, as a novel regulator of Wg/Wnt signaling, essential for the Wg/Wnt-dependent expression of downstream target genes. Our experiments suggest that Armless interferes with the tagging of Armadillo/b-Catenin with ubiquitin, thereby sparing it from proteasomal degradation. We also show that Armless directly interacts in vivo with Ter94, a ubiquitous ATPase involved in protein turnover. Our results suggest that Armless antagonizes Ter94's function in protein turnover, thereby acting as a positive regulator of Wg/Wnt signaling by promoting the stabilization of Armadillo/b-Catenin.

als Interacts with the Wg Pathway and Is Required for Wg Target Gene Expression
To further test whether als interacts with Wg signaling, we analyzed the effects of als RNAi in different, sensitized backgrounds. Interfering with the Wg pathway at various levels leads to typical wing margin notches and loss of distal wing tissue ( Figure S3G-S3J). These phenotypes were enhanced upon als depletion ( Figure  S3G9-S3J9). To test whether als is also involved in the Wg pathway in other developmental contexts, we expressed als RNAi in the primordia of the dorsal thorax and the legs. Again, consistent with a role of als in the Wg pathway ( Figure S4D and S4H; [10,11]), thorax and scutellum sizes were affected ( Figure S4B, S4C, S4F, and S4G, cf. Figure S4A), and showed reduced or missing microchaete, a slight misorientation of remnant microand macrochaete, and mild thoracic clefts ( Figure S4B and S4C). Depletion of als in the leg discs caused a reduction of tarsal segments along the proximodistal axis and a dorsolateral shift of the sex combs-phenotypes characteristic of reduced Wg signaling (Figure S4J-S4M; [12,13]). Stronger als RNAi expression caused a dorsalization of the leg as judged from the mirror-image patterning of leg trichomes ( Figure S4K and S4M). To analyze the phenotype at earlier stages of development, we expressed als RNAi or Als HA (a dominant-negative-like version; see Text S1) in the embryo with da-Gal4. We found that cuticles were shorter than in control embryos, with ventral denticle belts partially fused, which is indicative of a segment polarity defect ( Figure S4S-S4V, S4X, and S4Y). These phenotypes are reminiscent of milder Wg loss-of-function phenotypes caused by arm RNAi or UAS-Lgs 17E expression ( Figure S4P-S4R and S4W) by da-Gal4.
To further test whether als is required for Wg signaling output, we monitored the expression of well-known Wg target genes. Wg is secreted from a stripe of cells at the dorsoventral (D/V) border and triggers a transcriptional response in cells along the D/V axis of imaginal wing discs (the future proximodistal axis of the adult wing). Wg activates the proneural gene senseless (sens) in sensory organ precursors, which develop into the bristles of the adult wing margin [8,14]. als RNAi expression in the P-compartment resulted in a loss of Sens therein (Figure 2A and 2A9). Expression of Distalless (Dll), frizzled3 (fz3), and wingful (wf), which are activated by Wg in a broader region centered on the D/V border [15,16], was strongly reduced or lost in response to reduction of als function ( Figure 2B-2E). Conversely, arrow (arr), which is normally repressed by Wg signaling in central pouch regions ( Figure 2F; [17]), became ectopically activated towards the central wing pouch ( Figure 2G). In sum, these observations strongly suggest that als encodes a component with a positive role in the Wg pathway.
To try to gain insight into the pathways als might regulate, we further looked at als expression. In situ hybridization and lacZ reporter transgenes indicated that als is expressed at early third instar wing imaginal discs at the center of the pouch adjacent to the D/V border and, at faint levels, throughout the wing primordium ( Figure S5A-S5C). Furthermore, the expression of UAS-als RNAi in the P-compartment of wing imaginal discs caused a strong decrease in als expression ( Figure S5B, inset). Interestingly, although normal when Wg signaling is inhibited, als expression could be upregulated when Wg signaling was activated by the overexpression of either Wg or Arm S10 , and stimulation of Wg signaling by the inhibition of GSK3b caused Figure 2. als is required for Wg-dependent target expression. L3 wing imaginal discs; anterior is to the left, posterior is to the right. als RNAi expression in the P-compartment of wing imaginal discs (visualized by the co-expression of UAS-GFP, green fluorescence) (A-C9) or in the entire wing pouch (D-G) causes loss of expression of Wgsignaling-dependent target genes-Senseless (A and A9) (line 39000), Distalless (B and B9) (line oligo3), frizzled3 (C and C9) (line oligo3), or wingful (D and E; white arrows indicate the Wg-dependent expression domain of wf at the D/V border which is lost in E.) (line 39000)-and ectopic expression of arrow, which is a negative target (G c.f. F; white arrows indicate the extent of the domain that is free from arrow-lacZ expression) (line oligo3). doi:10.1371/journal.pbio.1001988.g002 elevated Als and als levels ( Figures S5D and 8A). This is consistent with als functioning in the Wg pathway and having its activity fine-tuned by feedback mechanisms that ensure maximum pathway activity.

als Is Not Required for Notch or Hedgehog Signaling
Wing notching is a phenotype characteristic not only of impaired Wg signaling, but also of reduced Notch signaling. The activities of both pathways are required to establish the D/V border and subsequent proximodistal wing growth. To determine whether als also plays a role in the Notch pathway, we analyzed the expression of the following Notch targets: (1) the proneural gene e(spl)m8 ( Figure S6A; [18]), (2) a Notch responsive element (NRE) reporter ( Figure S6A inset; [19]), and (3) the vestigial (vg) ''boundary enhancer'' ( Figure S6B; [20]). In contrast to what we observed for the Wg targets above, the expression patterns of these Notch reporters were unaltered upon als depletion ( Figure S6A9-S6B0). Moreover, neither did reduction of als function suppress the phenotypes caused by elevated Notch signaling ( Figure S6D and S6F, cf. Figure S6C and S6E), nor did it ever mimic a neurogenic thorax phenotype or fused tarsal segments ( Figure S4), which are both typical for reduced Notch signaling. Finally, we monitored the expression of the wg gene itself, which is under control of Notch activity [21][22][23]. Neither wg transcription nor Wg protein expression were affected at the D/V boundary upon als RNAi expression ( Figure S7A-S7G).
We also tested other pathways involved in wing growth and patterning, such as Hedgehog (Hh) signaling. Increased Hh signaling causes wing overgrowth predominantly along the anteroposterior (A/P) axis, along with ectopic veins in the Acompartment ( Figure S6G). als depletion impaired wing growth along the proximodistal axis, which is under control of Wg signaling, but not along the A/P axis, and it did not alter the vein patterning defects caused by increased Hh signaling ( Figure S6H). Furthermore, expression of the Hh target gene patched (ptc) was not affected by als depletion (Figures S6I, S6J, and 8E). In addition, we tested Upd/Jak/Stat signaling and EGFR/Ras signaling and found that none of these pathways was affected by als depletion (Figure S6K-S6N). Our findings therefore suggest a specific role for als in Wg signaling in the tissues that we have tested.

Mapping Als Action by Cell Autonomy and Epistasis Analyses
To investigate whether Als functions in Wg producing or receiving cells, UAS-als RNAi was expressed in clones of wing imaginal cells marked by GFP co-expression. Expression of the Wg target Sens was lost in early induced clones ( Figure 3A-3A0). This effect is cell-autonomous, i.e., it also occurs when mutant cells are close to wild-type cells that produce Wg. Adult wing clones, marked by UAS-forked RNAi co-expression, had wing notches and thickened veins ( Figure 3B-3E), both of which were tightly associated with mutant cells. We observed similar phenotypes when Wg signaling was cell-autonomously impaired ( Figure 3F-3I).
To map where Als acts in the Wg pathway, we used phenotypic assays based on the activation or inactivation of Wg signaling in the eye or the wing. Expression of Wg under control of the sevenless (sev) enhancer causes small, narrow eyes that lack interommatidial bristles ( Figure 4B; [24,25]). This phenotype is partially suppressed when Wg signal transduction is hindered, for example, by the expression of a dominant-negative form of Lgs (Lgs 17E [26]; Figure 4G). Impaired als function also causes a suppression of the sev-wg phenotype ( Figure 4C-4F). In contrast, depletion of als did not alter the small eye phenotype caused by the overexpression of Eiger ( Figure 4H-4J), a trigger of the apoptotic pathway [27].
We manipulated the Wg signaling cascade in the wing and the eye at successively more downstream positions and assessed the requirement of Als for the manifestation of the respective Wg gainof-signaling phenotypes (see legend of Figure 5 for a detailed description of phenotypes). First, we found that Als depletion suppressed Wg pathway activation caused by RNAi of notum/ wingful (encoding an extracellular inhibitor of Wg [28,29]) or by overexpression of Arrow (co-receptor of Wg, [17]) or Dishevelled ( Figure 5A-5B9, 5L, and 5L9). Next, we tested knock-downs of components of the destruction complex, both in the wing and the eye, and found that aberrant Wg output was suppressed by als RNAi for Axin, APC, and Shaggy/GSK3b ( Figure 5C, 5C9, 5D, 5D9, 5H-5K9, 5M, and 5M9). Only the phenotypes of RNAi against the SCF/Slimb/bTRCP E3 ubiquitin ligase [30] and expression of Arm S10 , a constitutively active form of Arm/b-Catenin [31], could not be ameliorated by als depletion ( Figure 5E-5F9 and 5O-5Q9). Interestingly, wild-type Arm could not overcome the als RNAi phenotypes in the wing and the eye, and did not lead to the emergence of ectopic sensory bristles ( Figure 5G, 5G9, 5R, and 5R9). We interpret these genetic findings to indicate that Als functions upstream of Arm, but downstream of the Arm/b-Catenin destruction complex, possibly at the level of the SCF/ Slimb/bTRCP E3 Ub ligase.

als Encodes a UBX Domain Protein That Localizes in the Cell Cortex and Depends on a Functional PUG-Domain
Three different protein domains can be distinguished within Als: (1) an N-terminal coiled-coil domain, (2) a PUG domain, and (3) a UBX domain ( Figure S9). Coiled-coil domains function in protein oligomerization, whereas the PUG und UBX domains are found in ubiquitin regulatory proteins [32]. To find out more about its molecular role, we tested where Als localizes within cells and which of its protein domains are functionally required. Nterminally tagged HA Als localized close to the cell membrane in a punctate ring pattern ( Figure S8A-S8B9), with strongest expression in the apical region of the cell ( Figure S8C, S8D, S8E). This localization is reminiscent of that of E-Cadherin, Arm, and the Wg receptor Frizzled2 (Fz2, Figure S8C-S8E0). We then expressed Als protein variants that lacked particular domains and assessed whether these could rescue the als RNAi phenotypes. For this we created tagged and untagged constructs that were either sensitive or insensitive to the RNAi used. Expression of Als DN and Als DPUG did not alter the als RNAi phenotype ( Figure S9A-S9I), irrespective of their tag or targetability. However, expression of Als DUBX as well as full-length Als protein substantially rescued the wing phenotypes caused by als RNAi (Figure S9J-S9L). This suggests that the UBX domain is dispensable for Als function, whereas the N-terminal and the PUG domains are both important. The requirement of the PUG domain implicates Als in playing a role in a ubiquitin-mediated protein degradation process.

Als Interacts with Ter94, a Conserved ATPase Involved in Protein Turnover
To further our understanding of Als function, we expressed HAtagged variants of Als in Drosophila Kc-167 cells, and after affinity purification we subjected the samples to shotgun liquid chromatography-tandem mass spectrometry (LC-MS/MS). In three independent experiments using Als HA , HA Als, and HA Als DUBX , but not in the negative control sample, we identified Ter94 as  the highest ranking protein interactor of Als (see Text S1), irrespective of the stimulation of the Wg pathway by Wg. Ter94 is the Drosophila ortholog of p97, an AAA ATPase found to be associated with ubiquitin-dependent protein degradation processes, and that can interact with different UBX domain proteins [33,34]. To further validate the interaction between Als and Ter94 we co-expressed Ter94 HA together with FLAG Als, FLAG Als DN , and FLAG Gal4, which served as a negative control, in Kc-167 cells. Western blot analyses following reciprocal immunoprecipitations showed that (1) Ter94 HA and endogenous Ter94 co-immunoprecipi-tated with FLAG Als and FLAG Als DN , and (2) FLAG Als, FLAG Als DN , and endogenous Als co-immunoprecipitated with Ter94 HA ( Figure 6A). When we expressed Ter94 HA in wing imaginal discs, we found a strikingly similar sub-cellular distribution and co-localization with FLAG Als ( Figure 6B-6B0). Importantly, we demonstrated a physical interaction between Als and Ter94 in vivo with bimolecular fluorescence complementation (BiFC) analysis ( Figure 6C and 6C9; [35]; see also Materials and Methods), where the co-overexpression of Als-VC and Ter94-VN resulted in a Venus-YFP fluorescent signal. In sum, our biochemical and genetic results indicate that Als molecularly interacts with Ter94. The UBX domain is dispensable for Als to bind to Ter94, which is in agreement with our finding that Als lacking the UBX domain could still rescue the als RNAi phenotype.

Armadillo Levels Depend on als Activity
Our biochemical and genetic studies suggest that Als is functionally linked to the proteasomal degradation pathway. Since Als is positively required for Wg signaling, it could exert its function in the pathway either by suppressing the degradation of a positive component, or by enhancing the degradation of a negative component. We first analyzed the protein levels of key negative pathway components in wing imaginal discs. We expressed different als RNAi lines in the wing primordium with nub-Gal4; however, neither Axin nor Shaggy nor Apc2 levels were affected upon als depletion (Figure S10A-S10L); [36]). In contrast, we found that the protein levels of Arm, the key positive signaling component, were dependent on als: when als RNAi was expressed in the P-compartment under the control of hh-Gal4, Arm levels were strongly reduced in P but not A cells ( Figure 7A-7A0). Even when we overexpressed Arm under the control of nub-Gal4, coexpression of als RNAi led to strongly reduced Arm levels ( Figure 7C and 7C9, cf. Figure 7B and 7B9). DAPI staining shows that all cells have normally shaped nuclei and are thus not apoptotic ( Figure 7A9, 7A0, 7C, and 7C9). Constitutively active, non-degradable Arm S10 localizes like Arm to the apical membrane upon overexpression ( Figure 7D and 7D9), and was also observed at higher levels in the cytoplasm ( Figure 7F and 7F9). Importantly, we found that Arm S10 levels were not altered in als-depleted wing imaginal disc cells ( Figure 7E, 7E9, 7G, and 7G9). This is consistent with our finding that Arm S10 overexpression could completely overcome the phenotypes caused by als depletion ( Figure 5F, 5F9, 5Q, and 5Q9). Other positively acting components of the Wg pathway, Arrow and Fz2, did not exhibit alterations in protein levels upon als RNAi expression ( Figure  S10M-S10P). The observation that Arm is subject to degradation in the absence of als function is in agreement with our observation that overexpressed Arm-in contrast to Arm S10 -is not able to rescue als RNAi phenotypes in wing and eye primordia ( Figure 5G9 and 5R9).
The rescue experiments with Arm and Arm S10 suggested that Als acts upstream of Arm's proteasomal degradation. To further corroborate this idea, we reduced the levels of ubiquitin in addition to als depletion. This led to a suppression of the als RNAi phenotypes in the wing and the eye ( Figure S11I, S11L, and S11O, cf. Figure  S11G, S11J, and S11M). Reduction of ubiquitin also suppressed Wg loss-of-function phenotypes that were based on elevated activity of the destruction complex by Axin overexpression (Figure S11A and S11C). In contrast, depletion of ubiquitin did not suppress the Lgs 17E overexpression phenotype, which intersects the Wg pathway downstream of the proteasome ( Figure S11D and S11F). These findings support the idea that Als acts upstream of Arm's proteasomal degradation.
To check whether Als directly interacts with Arm, we carried out selected reaction monitoring, a highly sensitive mass spectrometry approach that allows the detection of low amounts of peptides in a complex sample [37]. Als and Ter94 were used as bait. In none of the studies was an interaction between Arm and Als detected. Additionally, in shotgun affinity purification mass spectrometry using Arm or Arm S10 as a bait we did not find Als, whereas, as expected, components of the degradation complex such as Apc2, Axin, Sgg, and CKI were detected in the cytoplasm (Text S1).
We further analyzed the consequences of als depletion for Wg signaling in Kc-167 cells (Figure 8). Stimulation of the Wg pathway by the inhibition of GSK3b led to strong transcriptional upregulation of the Wg targets naked (nkd) and wingful (wf) ( Figure 8B and 8C, bar 2), and an increase in Arm levels ( Figure 8D, lanes 3 and 11). Consistent with the in vivo data, double-stranded RNA (dsRNA)-mediated knockdown of als ( Figure 8A) caused a reduction in nkd and wf expression in stimulated cells ( Figure 8B and 8C, bar 6, cf. bar 2). Importantly, the increase in Arm levels based on GSK3b inhibition was significantly reduced when als was knocked down ( Figure 8D, lane 4, cf. lane 3, and lane 13 and 14, cf. lane 11). When the pathway was stimulated by inhibiting E1 Ub ligase activity, which caused a milder upregulation of nkd and wf than the GSK3b inhibition ( Figure 8B and 8C, bar 3; see also Materials and Methods), transcriptional upregulation was still decreased upon als depletion ( Figure 8B and 8C, bar 7), but to a far lesser extent; at the same time, Arm was reduced upon als depletion ( Figure 8D, lane 6, cf. lane 5). Inhibition of the 26S proteasome increased Arm levels in both control and als-dsRNA conditions ( Figure 8D, lanes 9 and 10).
When we depleted ter94 in wing imaginal discs and in Kc-167 cells, we found an upregulation of the positive Wg target genes nkd and wf, and a downregulation of the negative Wg target arr ( Figure 8E). Moreover, in als-depleted cells, ter94 reduction led to an amelioration of the changes in Wg target expression ( Figure 8E), and to a partial restoration of Arm levels ( Figure 8D, lane 12). In contrast, the Hh target gene ptc was not altered ( Figure 8E). Overexpression of ter94 RNAi in wing imaginal discs, however, caused pupal lethality. Because of this, it was difficult to analyze the adult wing phenotype caused by the combined depletion of ter94 and als; because of ter94's pleiotropic role, the analysis of the consequences of ter94 depletion is challenging [38].
In addition, we found that overexpression of Ter94 HA suppressed the strong dominant-negative-like effect of Als HA overexpression in the wing and the eye primordium ( Figure 8F-8I), while overexpression of Ter94 HA could not suppress phenotypes caused by als RNAi (Figure S2M). This is consistent with the finding that overexpression of Ter94 HA alone did not cause any phenotype ( Figure S2N); however, it suggests that an excess of Ter94 HA can neutralize the excess of the dominant-negative Als HA .
Together, these results suggest that Als functions downstream of GSK3b and downstream or at the level of E1 Ub ligase and that Als is required to antagonize Ter94 to allow normal Wg signaling. This is in agreement with our genetic epistasis experiments ( Figure 5) and is corroborated by the further finding that overexpression of CSN6, an essential de-ubiquitinating enzyme and negative regulator of SCF/Slimb/bTRCP E3 Ub ligase [39], reverted the als RNAi phenotypes in the wing and the eye ( Figure  S11H-S11N). Of note, CSN6 overexpression suppressed the Wg signaling defects caused by Axin overexpression (Figure S11A and S11B), but did not suppress Wg signaling defects caused by Lgs 17E overexpression ( Figure S11D and S11E).

Human UBXN6 Is the Functional Ortholog of Als
To address the evolutionary conservation of Als's function, we tested whether the human ortholog of Als, UBXN6, can substitute for Als in wing and eye development. Overexpression of HA UBXN6 in imaginal discs could largely rescue the wing phenotypes and fully rescue the eye phenotypes caused by als RNAi (Figure 9A-9D). Further, we found that HA UBXN6 localizes in cells of imaginal discs in a manner similar to HA Als ( Figure 9E and 9E9, cf. Figure S8A-S8E). The UBXN6 and Als proteins share 33% identity and belong to the same subfamily of UBX domain proteins, which is specified by the presence of an UBX and a PUG domain ( Figure 9F). We next asked whether UBXN6 is required for Wnt signaling in human cells. We stimulated the Wnt pathway in HEK-293 cells with mouse Wnt3a (mWnt3a) while targeting UBXN6 by small interfering RNA (siRNA). siRNA against CTNNB1/b-Catenin served as a control. To monitor the output of the Wnt pathway, we analyzed three established Wntresponsive genes, SP5, AXIN2, and FZD1, by real-time PCR. All three targets were upregulated upon pathway stimulation ( Figure 9G-9G0, green bars), and this response was abolished upon siRNA treatment of UBXN6 ( Figure 9G-9G0, red bars). The depletion of b-Catenin caused a similar downregulation of Wnt target gene expression. Further, we found b-Catenin levels reduced upon UBXN6 depletion in HEK-293 cells ( Figure 9H, lane 6, cf. lane 1). Similar to mWnt3a, the stimulation of the pathway by GSK3b inhibition caused an increase of Wnt target genes. Depletion of UBXN6 reduced this effect ( Figure 9I). These findings suggest that UBXN6 is a functional ortholog of Als and that UBXN6 might play a similar role for Wnt signaling in human cells.

Discussion
A prevalent mechanism for controlling information flow in signaling pathways is the alteration of the protein levels of key components. In the Wg/Wnt pathway, the Arm/b-Catenin destruction complex targets Arm/b-Catenin for ubiquitination by the SCF/Slimb/bTRCP E3 Ub ligase, resulting in proteasomal degradation and low cytoplasmic levels of Arm/b-Catenin in the Wnt pathway off state. If the pathway is turned on, Slimbmediated ubiquitination is prevented, thus rescuing Arm from its proteasomal fate and causing a concomitant increase in Arm protein levels [5]. Here we describe Als as a new component of this control system; we found that Als is required to prevent the degradation of Arm/b-Catenin.

Als/UBXN6 Is a Novel Regulator of the Wnt/Wg Pathway
We identified als in a genome-wide in vivo RNAi screen in Drosophila. Because we did not succeed in isolating an EMS-or Pelement-induced null allele and because another gene overlaps with als, we demanded, and obtained, particularly thorough evidence validating als gene function. (1) The als RNAi phenotypes could be reproduced by nine different UAS-RNAi transgenes encoding independent RNA target sites. Together with an extended off target analysis, we could rule out unintentional RNAi as a cause for the als phenotypes. (2) RNAi-mediated inhibition of als expression was ascertained by monitoring als mRNA expression via real-time PCR and antisense mRNA in situ hybridization. (3) Expression of HA Als with different RNAiinsensitive rescue transgenes, as well as with its human ortholog UBXN6, rescued als RNAi phenotypes.
Our analyses show that als encodes an essential positive Wg signaling component. This conclusion is based on the following evidence. als depletion caused wings with notched wing margins and loss of sensory bristles, which is characteristic of impaired Wg signaling. The distal wing region is most sensitive to als levels, as is the case for other positive components of Wg signaling. In agreement with this, we found increased als expression in the central wing pouch, at least in earlier L3 larval stages. Stimulation of the Wg pathway in wing imaginal discs or Kc-167 cells caused higher als expression, suggesting that als can be positively controlled by Wg signaling. However, Als levels must be precisely controlled since already mild overexpression of UAS-als elicits a dominant-negative effect on Wg signaling. The function of als for Wg signaling is not restricted to the wing: also in other tissues, such as the thorax, eyes, legs, and the embryo, als RNAi phenotypes are identical to those seen when Wg signaling is disturbed. Also in human HEK-293 cells we found UBXN6/UBXD1, the ortholog of Als, to be required for Wnt signaling, and human HA UBXN6 largely rescued the als RNAi phenotypes in Drosophila, which suggests their functional conservation. Depletion of als also enhanced Wg-sensitized phenotypes, further supporting the notion that its product is a Wg pathway component. Moreover, the expression of positively regulated Wg target genes is reduced or abolished upon loss of als function, while Wg-repressed target gene expression is ectopically activated. Importantly, while interfering with als function suppressed Wg signaling, it did not affect other pathways, such as Notch and Hh, Jak/Stat, or EGFR signaling. However, we cannot rule out that als is not required in another pathway in a different biological context. In humans, UBXN6 is reported to play a role in diverse scenarios: for example, it was shown to play a role in Caveolin turnover in human osteosarcoma U2OS cells [40]. This might indicate a broader role of UBXN6 in mammalians.

Als: A Protector of Armadillo
Our data show that Als regulates Armadillo protein levels. Based on our epistasis experiments, Als acts downstream of Shaggy/GSK3b and upstream of the SCF/Slimb/bTRCP E3 Ub ligase, which is known to ubiquitinate Arm, a prerequisite for proteasomal degradation. Consistent with this, the degradationresistant form of Arm, Arm S10 , could completely bypass the requirement for als, in contrast to the wild-type form of Arm. This suggests that proteasomal degradation acts downstream of als; however, we would like to point out that this cannot be taken as an unambiguous proof. Importantly, depletion of ubiquitin and overexpression of CSN6, a negative regulator of SCF/Slimb/ bTRCP E3 Ub ligase, could ameliorate the als phenotype (as well as phenotypes based on the overexpression of Axin or Shaggy, which overactivate the destruction complex, thus resulting in enhanced Arm degradation). In contrast, altering these factors did not ameliorate the Lgs 17E phenotype, which is caused by interfering more downstream in the Wg pathway. These findings suggest that als works upstream of proteasomal degradation. A further informative experiment was monitoring Wg pathway components with respect to protein levels: Arr, Fz, Axin, APC, Sgg, and Arm. The only change in the absence of Als function was Arm: its levels were strongly reduced upon als depletion. The effects on Arm levels could be due either to a direct effect on Arm or to an indirect effect on a negative component. Importantly, the rate-limiting factor Axin as well as other key negative components of the Arm/b-Catenin destruction complex were unaltered upon als depletion.

Als Interacts with Ter94 ATPase
Some further mechanistic insight was obtained with our finding that Ter94 interacts in vitro and in vivo with Als. Interestingly, we found that Als-Ter94 localizes at the cell cortex, as was similarly observed for the Arm/b-Catenin destruction complex [41]. Our studies are consistent with earlier work that showed that the human ortholog of Ter94, p97, interacts with UBXN6 [42]. Ter94/p97/Cdc48 is a conserved and highly abundant AAA ATPase that was found to associate with SCF/Slimb/bTRCP E3 Ub ligases or proteasomal shuttle factors to mediate UPSmediated protein degradation ( [33,[42][43][44][45]; reviewed in [6,34]). Specifications of the diverse activities of Ter94/p97 and the fate of its substrates are mainly exerted by UBX domain protein cofactors, which eventually either promote or hinder p97's function in protein turnover; an example of the latter involves the dissociation of the SCF/Slimb/bTRCP E3 Ub ligase complex, eventually leading to its inactivation [45]. Interestingly, it was recently reported that inactivation of the E3 ligase complex upon Wnt signaling is achieved by its dissociation from the destruction complex [5]. Based on our experiments and what is known about Ter94/p97, we suggest as a possible mechanism that Als antagonizes Ter94's positive effect on E3 ligase function, thereby rescuing Arm levels. We, however, did not observe increased protein levels of Slimb, Axin, Shaggy, or APC in our analyses; thus, our results favor a model in which Als antagonizes Ter94 to hinder the transfer of Arm to the proteasome by interfering with the SCF/Slimb/bTRCP E3 Ub ligase function or its assembly. Importantly, we did not find an interaction between Arm and Als. This is consistent with the finding that UBX domain family members lacking an UBA domain, such as UBXN6/Als, do not directly interact with substrate proteins [44], but are necessary for the activity or fate of the Ter94/p97. Interestingly, Zhang et al. [38] found that ter94 depletion affected the partial proteolysis of Ci. However, they observed neither any typical consequence of disturbed Hh signaling per se (i.e., no alteration of Hh target gene expression in genes such as ptc) nor any phenotypical consequence upon overexpression of a dominant negative form of Ter94 (i.e., aberrant wing patterning and growth typical for Hh signaling). This is consistent with our data that neither Ci target expression nor Hh signaling was affected upon als or ter94 depletion ( Figures S6G-S6J and 8E).
p97/Ter94 is known as a highly pleiotropic AAA ATPase associated with many cellular functions. Further, p97/Ter94 acts in multifaceted and large protein-protein complexes, and it is its regulatory co-factors, including UBX domain proteins, that render  (lanes 7-9). In Kc-167 cells (middle bar diagram), a co-depletion of als and ter94 caused an upregulation of the positive Wg targets nkd and wf (lanes 3 and 6), which were reduced upon als depletion (lanes 1 and 4). No alteration was observed for the Hh target gene ptc (lanes 7-9). Reduced nkd expression of als-depleted, GSK3b-stimulated Kc-167 cells (right bar diagram) could be ameliorated upon co-depletion of ter94. (F-I) The wing and eye phenotypes caused by the expression of the dominant-negative-like Als HA were suppressed by the overexpression of Ter94 HA . co-i, control siRNA. doi:10.1371/journal.pbio.1001988.g008 p97/Ter94 specific for a certain task in a particular cellular context. For example, p47/Shp1 is a co-factor of p97/Ter94 that blocks other co-factors from Ter94 binding [46,47]. Interestingly, in our Kc-167 cell mass spectroscopy experiments, we found p47 in Ter94/Als protein complexes, but only in the absence of Wg stimulation (Text S1). On the other hand, we found als transcript and Als protein levels elevated upon Wg signaling (Figures 8A and  S5D). These findings suggest a dynamic regulation of the Ter94 complex upon signaling inputs. The identification and functional analysis of all key components of the Als-Ter94 complex will be needed to obtain a refined insight into Als-Ter94's molecular mechanism.

Concluding Remarks
Critically, our work spotlights an underappreciated facet in the control of the output of the entire canonical Wg/Wnt pathwayhow Arm/b-Catenin is handed over to the proteasome-and the potential for regulating this step; our works also indicates that this step, in contrast to the conventional wisdom, is tunable. Our identification and characterization of the UBX protein Als as a positive regulator of Wg/Wnt signaling contributes to this layer of pathway control.

Drosophila Strains
UAS-als RNAi lines. als was identified in a genome-wide RNAi screen based on the Vienna Drosophila Resource Center (VDRC) RNAi library [48] with the VDRC line 28998 (II/CyO).
pNE3-based RNAi lines contain five copies of UAS; pWAL20based RNAi lines contain ten copies of UAS. The pNE3 vector was obtained from B. Haley and is a derivative of the pUAST-attB vector [49,50]; the pWAL20 (pWALIUM20) vector was obtained from J. Ni (Transgenic RNAi Resource Project [TRiP], Drosophila RNAi Screening Center).
RNAi off target analysis. We analyzed the mRNA levels of computer-predicted, putative off targets of the VDRC lines upon expression of UAS-VDRC-RNAi transgenes in wing imaginal discs by real-time PCR. We did not find specific alterations of the putative off targets. For the generation of our UAS-oligo-RNAi transgenes, we employed the DRSC off-target-finder algorithm (16-mer allowed) to exclude off-target-prone RNAi target sites.
Lines for overexpression: UAS-als lines. Transgenic lines are based on the WC31 transgenesis system, using the line ZH-86Fb attB as the landing site on Chromosome 3.
cDNA for the 1.327-kb long coding region of als was obtained by PCR on cDNA from L3 wing imaginal disc after total RNA extraction. A Drosophila Kozak consensus region was attached in front of the ATG. N-and C-terminally epitope-tagged versions are as follows: UAS-als HA (C-terminal triple HA-tag), UAS-HA als (Nterminal triple HA-tag), UAS-FLAG als (N-terminal double FLAGtag), UAS-HA als oligo3-mut (N-terminal triple HA-tag); within the oligo3 target site we introduced seven conservative basepair mutations by alternative codon usage (59-CTGATGAGAGTC-TAGACTTCA-39).
Als protein domain deletion variants. Deletions of the PUG, UBX, and N-terminal domains were generated by overlap extension or conventional PCR.
UAS-HA UBXN6 line. The coding region of human UBXN6 was isolated by PCR on cDNA derived from HEK-293 cells after total RNA extraction.
Cell clone analysis. For cell clone analysis in wing imaginal discs ( Figure 4) and adult wings, somatic clones were generated in animals of the following genotype: yw, hsflp/+; +/+; actin5c. CD2.Gal4, UAS-GFP, UAS-forked RNAi /UAS-als RNAi (VDRC line 39000 or line oligo3 10UAS ) or UAS-als HA . Heat shock flippase (hsflp) was induced 2 and 3.5 d after egg deposition (30 min, 37uC). Third instar larvae were fixed and stained for Wg target Figure 9. Human UBXN6 is the functional ortholog of Als. (A-D) Human UBXN6 expression could largely or completely rescue als RNAi wing and eye phenotypes. (E and E9) When expressed in wing imaginal discs HA UBXN6 localization resembles that of HA Als (cf. Figure S8A-S8E). (F) UBXN6 and Armless belong to the same protein subfamily of UBX domain proteins (blue). (G-G0) Expression of the WNT target genes SP5, AXIN2, and FZD1 was upregulated upon stimulation of the pathway with mWnt3a (green bar), and was reduced upon siRNA against human UBXN6 in HEK-293 cells. Depletion of b-Catenin served as a control. (H) b-Catenin levels were reduced upon siRNA against human UBXN6 in HEK-293 cells (lane 6, cf. lane 1; and lane 3, cf. 2), whereas depletion of UBXN6 had no effect on b-Catenin levels upon inhibition of the proteasome (MG132; lane 5, cf. lane 4). (I) WNT target genes were upregulated upon GSK3b inhibition (CHIR; green bars), and were reduced upon depletion of UBXN6 (blue bars). doi:10.1371/journal.pbio.1001988.g009 expression by standard immunohistochemical techniques. UASforked RNAi was used to mark adult cell clones (trichomes of the wing epithelium are shorter and adopt a forked-like shape).
als-lacZ transgenic lines. 4,978 kbp (HindIII-XbaI) and 2,927 kbp (StuI-XbaI) of als's upstream genomic DNA (down to the start codon) were cloned into a pattB-lacZ vector. Wing imaginal disc expression of the lines was analyzed by X-Gal staining (brightfield microscopy) and by anti-bGal antibody staining, visualized by Alexa Fluor 594 secondary antibody (UV confocal microscopy).

In Vivo Interaction of Als with Ter94 (BiFC Analysis)
Ter94-VNm9 (Ter94 fused to the N-terminal part of Venus-YFP) and Als-VC155 (Als fused to the C-terminal part of Venus-YFP) were co-expressed and analyzed for interaction by BiFC [34]. Direct interaction of two proteins causes a reconstitution of the two non-fluorescent YFP subfragments into a functional YFP, resulting in a yellow fluorescent signal.

Imaging
Images of immunostainings were generated with a Zeiss Lsm710 confocal microscope using 406 and 606 oil objectives. Images were analyzed with ImageJ software. Adult fly wings were mounted in Euparal, and images were generated with a Zeiss Axioplan microscope.
In Situ mRNA Hybridization RNA probes were generated by run-off transcription of antisense mRNA with T7 RNA polymerase (Promega) from linearized plasmid templates with the PCR-amplified als coding region (full-length template as well as an EcoRV-truncated template, generating the first half of the coding sequence). Antisense mRNA was generated with a 39 primer containing the T7 RNA polymerase binding sequence (59-GAATTTAATAC-GACTCACTATAGG-39).

Real-time PCR
UAS-als RNAi was expressed under the control of c765-Gal4 ( Figure S1L) in wing imaginal discs, which were collected at the late third instar larval stage for total RNA isolation (RNeasy Kit, Qiagen, or TRI-Reagent, Sigma-Aldrich). 1 mg of total RNA was used as a template for reverse cDNA transcription (Superscript-RT-II, Roche). Quantitative PCR reactions were carried out with J of the real-time PCR reaction and als-specific primers, 25 cycles, in the presence of SYBR Green (Roche), which enabled quantitative detection of the amplicons. Primers were designed by Roche Universal ProbeLibrary and are intron-spanning (als primer pair a: forward: ggaaaagacactatatgactgcaaact; reverse: aaggagcgggtgtatcattg; als primer pair b: forward: gaaaccatgtccaagattaagaagt; reverse: atgccgctgccagttaaa). Real-time PCR was carried out on an Eppendorf Realplex Mastercycler. mRNA levels were calculated with the comparative CT (threshold concentration) method and were normalized for three external standards (actin5C, TBP, and gpdh). For each RNAi condition, data were obtained from three biological replicates (i.e., three independent but identical Gal46UAS-RNAi fly crosses), and real-time PCR was carried out in technical triplicates per experiment.
Drosophila Cell Culture and Immunoprecipitation of Epitope-Tagged Proteins Immunoprecipitation. Cells were harvested 48 h posttransfection. Per each experimental condition, cells of two T75 flasks were pooled and lysed in lysis buffer (86 of pellet volume) containing 20 mM Na 2 H 2 PO 4 (pH 7.4), 200 mM NaCl, 0.5% Nonidet P-40, 2 mM EDTA, 1 mM DTT, 1.5 mM Na 3 VO 4 , and protease inhibitor (Complete Mini; Roche). Cell suspension was dounced 206, incubated 10 min at room temperature, and centrifuged with 13,000 rpm at 4uC. The supernatant (total cell lysates) was incubated either with rabbit anti-HA antibody (Abcam) or mouse anti-FLAG antibody (Sigma), conjugated to Protein-A Sepharose beads (Sigma), and allowed to rotate 3 h at 4uC. The beads were then collected (3,600 rpm, 10 min at 4uC), washed 36 with the lysis buffer, and transferred to a column to allow gravity flow. Protein complexes were eluted with 3650 ml of glycine (0.2 M, pH 2.5) and neutralized with 50 ml of 8 M urea (containing 100 mM NH 4 CO 3 ). The pH of the eluted protein complex was adjusted with NH 4 HCO 3 to a final pH of 8.8 and stored at 220uC prior to sample preparation for mass spectroscopy (see below).

Mass Spectrometry Analysis of Epitope-Tagged Als and Ter94
Sample preparation for mass spectrometry. Proteins isolated via immunoprecipitation against the hemagglutinin-tag were reduced with 5 mM TCEP and treated with 10 mM iodoacetamide to modify cysteine residues. Tryptic digestion was performed overnight using 2-5 mg of trypsin per sample. Digested peptides were purified by reverse phase C-18 chromatography (Sep-PaK, Waters). For mass spectrometry analysis, samples were resuspended in buffer containing 5% acetonitrile and 0.2% formic acid.
Mass spectrometry parameters and data analysis. For each sample, 4 ml was loaded on a LTQ-Orbitrap XL ETD mass spectrometer (Thermo Fisher Scientific). The instrument was coupled to an Eksigent NanoLC system. Samples were automatically injected into a 10-ml sample loop and loaded onto an analytical column that was packed in-house with Magic C18 AQ beads (3 mm, 200 Å , Bischoff Chromatography) and was 8 cm in length, with an internal diameter of 75 mm. Peptide mixtures were delivered to the analytical column at a flow rate of 500 nl/min (3% acetonitrile, 0.2% formic acid) for 16 min and then eluted using a gradient of acetonitrile (3%-35%; 0.53%/min) with 0.2% formic acid at a flow rate of 250 nl/min. The samples were measured in a survey scan from 300 to 2,000 amu, followed by six datadependent LC-MS/MS scans with dynamic exclusion (isolation width 2 amu, three repeat counts, exclusion list size 500, dynamic exclusion duration 60 s). LC-MS/MS spectra were assigned to peptide sequences using Mascot software, version 2.4 (Matrix Science). All peptide assignments where the Mascot ion score was greater than the homology score and where the Mascot expect score was smaller than 0.05 were accepted.

Human Cell Culture
HEK-293 cells were seeded into a 96-well plate (40,000 cells/well). 1 d after seeding cells were transfected (Lipofectamine 2000, Invitrogen) with siRNAs targeting human b-Catenin, human UBXN6, or a negative control siRNA (Qiagen FlexiTube siRNA). The siRNA concentration per well was 0.05 mM. Co-transfected histone2B-RFP served as the control of the transfection efficiency. The Wnt pathway was stimulated by mWnt3a from co-transfected pcDNA3::mWnt3a. Each experimental condition was made in triplicate. 48 h after transfection, cells were harvested, and an aliquot was used for real-time PCR (1) to analyze the siRNA-mediated knockdown of b-Catenin or UBXN6 and (2) to analyze the transcriptional level of the Wnt targets SP5, AXIN2, and FRIZZLED-1. To analyze b-Catenin levels, siRNA (Qiagen FlexiTube siRNA, Thermo Scientific SMART pool) was applied two times (the second siRNA application was done 1.5 d after the first siRNA application), and cells were harvested 86 h after the first transfection. Simultaneously depleting als (G0-J0) leads to an enhancement of the notched wing margin phenotype observed with depleting als alone (G9-J9). (TIF) Figure S4 als depletion in the thorax and legs. (A-H) als RNAi expression in the notum, the primordium of the thorax, by pannier-Gal4, which is active in the mesothorax (E). als depletion causes mild thoracic clefts (B and C), a slight polarity defect of thoracic bristles, reduced or lost dorsocentral microchaete, and an overall shortening of the thorax and scutellum (B, C, F, and G). Text S1 Supporting information.