Zasp52, a Core Z-disc Protein in Drosophila Indirect Flight Muscles, Interacts with α-Actinin via an Extended PDZ Domain

Z-discs are organizing centers that establish and maintain myofibril structure and function. Important Z-disc proteins are α-actinin, which cross-links actin thin filaments at the Z-disc and Zasp PDZ domain proteins, which directly interact with α-actinin. Here we investigate the biochemical and genetic nature of this interaction in more detail. Zasp52 is the major Drosophila Zasp PDZ domain protein, and is required for myofibril assembly and maintenance. We show by in vitro biochemistry that the PDZ domain plus a C-terminal extension is the only area of Zasp52 involved in the interaction with α-actinin. In addition, site-directed mutagenesis of 5 amino acid residues in the N-terminal part of the PDZ domain, within the PWGFRL motif, abolish binding to α-actinin, demonstrating the importance of this motif for α-actinin binding. Rescue assays of a novel Zasp52 allele demonstrate the crucial importance of the PDZ domain for Zasp52 function. Flight assays also show that a Zasp52 mutant suppresses the α-actinin mutant phenotype, indicating that both proteins are core structural Z-disc proteins required for optimal Z-disc function.


Introduction
Like most animals, invertebrates have three main types of muscles, body wall, heart, and visceral muscle, but in contrast to vertebrates, all of them are striated. A particularly highly organized muscle is the indirect flight muscle (IFM) in insects, a stretch-activated fibrillar muscle [1]. Sliding filaments mediate muscle contraction in the sarcomere, the smallest functional contractile unit of muscle. Many proteins contribute to the elastic and contractile properties of muscles, most notably myosin thick filaments, which are anchored at the M-line, and actin thin filaments, which are anchored at the Z-discs. Z-discs border the sarcomere and are multiprotein complexes that transmit tension during contraction, and maintain structure and function of the myofibril, in part by serving as signaling centers [2]. In addition, Z-discs and their precursors, Z-bodies, play a crucial role in myofibril assembly. A major component of Z-discs is α-actinin, which cross-links slightly overlapping barbed ends of actin filaments at the Z-disc. In addition, proteins of the Zasp PDZ domain family function in maintenance of Z-discs and have also been proposed to play an important role in myofibril assembly. They have a unique N-terminal PDZ domain in common containing a conserved PWGFRL motif proposed to be required for α-actinin binding [3]. In vertebrates, the Zasp PDZ domain family comprises the Alp/Enigma family members ZASP/Cypher/Oracle/LDB3/PDLIM6, ENH/PDLIM5, PDLIM7/ ENIGMA/LMP-1, CLP36/PDLIM1/Elfin/hCLIM1, PDLIM2/Mystique/SLIM, ALP/PDLIM3, and RIL/PDLIM4, and in addition Myopodin/SYNP2, and CHAP/SYP2L. The first three members (ZASP, ENH, and PDLIM7) are called Enigma family proteins and have one N-terminal PDZ domain and three C-terminal LIM domains [4][5][6][7][8]. The next four members (CLP36, PDLIM2, ALP and RIL) are called Alp family proteins and comprise one N-terminal PDZ domain with only one C-terminal LIM domain [6,[9][10][11][12][13]. Myopodin and CHAP have only the N-terminal PDZ domain in some isoforms [14][15][16]. In Drosophila, Zasp52 has a PDZ, ZM (Zasp-like motif) and four LIM domains; while Zasp66 and Zasp67 only feature the N-terminal PDZ domain and a weakly conserved ZM domain. Zasp52 colocalizes with α-actinin at Z-discs and plays a role in myofibril assembly and maintenance [3,17,18]. Many different Zasp52 splice isoforms have been identified resulting in up to 61 different proteins, some of them restricted to specific muscle types [19,20]. Furthermore, Zasp52, Zasp66, and Zasp67 cooperate in myofibril assembly and play partially redundant roles at the Z-disc [3]. Mutations of Zasp52 orthologs in vertebrates cause similar defects, ranging from improper formation of somites and heart in zebrafish to fragmented Z-discs in skeletal and cardiac muscles in mice [4,7,21]. Similar to Drosophila, a ZASP/Cypher and ENH double knock-out in mice demonstrates partial redundancy in myofibril assembly [22]. The single C. elegans ortholog ALP-1 displays defects in actin filament organization, but motility defects are much milder than in vertebrates or Drosophila [23][24][25]. Mutations in the human ortholog ZASP result in phenotypes of variable severity from congenital myopathy with fetal lethality to late-onset cardiomyopathy [26,27].
In this study, we explore the relationship of Zasp52 and α-actinin. We show that even though different Zasp52 deletion transgenes co-immunoprecipitate α-actinin and localize to Zdiscs, only an extended PDZ domain mediates direct interaction of Zasp52 with α-actinin. Through site-directed mutagenesis we also demonstrate the importance of the PWGFRL motif in α-actinin binding. A rescue assay confirms the importance of the PDZ domain of Zasp52 for myofibril assembly. Finally, we show genetically that the Zasp52 α-actinin interaction is required for IFM function, because the α-actinin heterozygous flight defect is suppressed by removal of one copy of Zasp52. Our data indicate that Zasp PDZ domain family proteins are core scaffold proteins of muscles.

Different Zasp52 transgenes all localize to the Z-disc
We have previously shown that Zasp52 binds directly to α-actinin via an N-terminal construct containing the PDZ, ZM, and LIM1 domain [3]. Both PDZ and ZM domain bind α-actinin in human ALP, whereas in ZASP only the PDZ domain binds α-actinin [28][29][30]. Furthermore, even though the ZASP ZM domain is not required for direct binding to α-actinin, it is required for localization to Z-discs [29]. To clarify precisely which domain of Zasp52 is required for localization to Z-discs and which for direct interaction with α-actinin, we performed a range of in vivo and in vitro experiments. We started with generating a full-length Zasp52 transgene corresponding to the most common isoform (Zasp52-PR containing PDZ, ZM, LIM1 and LIM234 domains), as well as transgenes deleting the PDZ and ZM domain, respectively, to test if these domains are necessary for Z-disc localization (Fig 1A). We expressed the transgenes with Mef2-Gal4 in muscles. As expected, Zasp52-PR localized to Z-discs in IFM (Fig 1B and  1C). Both Zasp52-PRΔPDZ and Zasp52-PRΔZM also localize to Z-discs in IFM (Fig 1D and  1E), indicating that these transgenes can still be recruited by Z-disc proteins. To test α-actinin binding, we pulled down all three proteins from thorax extracts using Flag beads and tested for the presence of α-actinin. All three proteins co-immunoprecipate α-actinin, whereas Mef2-Gal4 control extracts do not (Fig 2A). Next we generated two additional transgenes encompassing the N-terminus and the C-terminus of Zasp52 ( Fig 1A). As expected Zasp52-PK localizes to Z-discs ( Fig 1F), but surprisingly, Zasp52-LIM234 localizes to Z-discs, as well ( Fig 1G). Zasp52-LIM234 also localizes to Z-discs when expressed with the weak driver UH3-Gal4, ruling out an overexpression artefact (S1 Fig). We further confirmed the α-actinin interaction by co-immunoprecipitation. Both Zasp52-PK and Zasp52-LIM234 co-immunoprecipitate α-actinin, whereas Mef2-Gal4 control extracts do not (Fig 2B).
We noticed that Mef2-Gal4-mediated expression of transgenes results in various muscle defects such as wavy myofibrils (Fig 1C-1G). To assess if myofibril defects are due to overexpression, we expressed Zasp52-PRΔPDZ with the driver UH3-Gal4, which strongly reduces transgene expression compared to Mef2-Gal4 ( Fig 3A). Zasp52-PRΔPDZ still localizes to Zdiscs, but dominant phenotypes are absent, indicating that defects caused by Mef2-Gal4-mediated expression are owing to strong overexpression ( Fig 3B). The same effect is seen with Zasp52-LIM234 (S1 Fig).
Our data show that multiple domains of Zasp52 are able to mediate Z-disc localization irrespective of expression level, and therefore raise the question if different domains of Zasp52 can bind directly to α-actinin.

An extended Zasp52 PDZ domain is required for α-actinin interaction
We have previously shown a direct interaction of Zasp52-PK containing a PDZ, ZM, and LIM1 domain with α-actinin [3]. To determine if Zasp52 directly interacts with α-actinin via different protein domains, we generated, overexpressed and purified various GST-Zasp52 constructs and tested their interaction with α-actinin ( Fig 4A and S2 Fig). We first tested the Nand C-terminal halves of Zasp52. Zasp52-LIM234 or GST control extract cannot interact directly with α-actinin, whereas Zasp52-PK, as shown previously, robustly interacts with αactinin ( Fig 4B). Next we analysed Zasp52-PP consisting only of PDZ and ZM domain, and Zasp52-LIM1. Only Zasp52-PP can interact with α-actinin ( Fig 4B). Finally, we tested Zasp52-PDZ and Zasp52-ZM individually. Only Zasp52-PDZ can directly interact with α-actinin, albeit very weakly (Fig 4C). This suggests that additional non-conserved amino acids are required for optimal interaction of Zasp52 with α-actinin. We therefore generated a series of truncated proteins and assessed their interaction with α-actinin ( Fig 4A). A 142 and a 154 amino acid-long protein interacted as well with α-actinin as the 233 amino acid-long Zasp52-PP ( Fig 4D). In contrast, a 111 amino acid-long protein interacted as weakly with αactinin as the 92 amino acid PDZ-only protein (Fig 4D). This shows that in addition to the PDZ domain 20-50 amino acids C-terminal to the PDZ domain, but not the ZM domain, are  Zasp52 PDZ Domain Is Essential for Myofibril Assembly required for optimal interaction with α-actinin. Our data indicate that the PDZ domain plus a C-terminal extension is the only Zasp52 domain in direct contact with α-actinin.

Zasp52 PWGFRL motif essential for α-actinin binding
We previously identified 16 highly conserved amino acids in the N-terminal half of the Zasp52 PDZ domain as the defining feature of Zasp PDZ domain proteins (Fig 5A), and proposed its requirement for α-actinin binding [3]. We now call these 16 amino acids the PWGFRL motif. By site-directed mutagenesis, we first changed P18W19 to DF in Zasp52-PK to mimic the   Zasp52 PDZ Domain Is Essential for Myofibril Assembly LMO7 PDZ domain, which does not bind α-actinin through its PDZ domain [31]. α-Actinin binding is completely abolished ( Fig 5B). Next, we generated three single amino acid mutations in the shorter Zasp52-PP construct containing only the PDZ and ZM domain (R22T, G25D, G26W), also modelled on the amino acids found in LMO7. While Zasp52-PP binds α-actinin, all three mutants abolish α-actinin binding ( Fig 5C). These data establish the PWGFRL motif as required for α-actinin binding and also show that the C-terminal extension on its own cannot bind α-actinin.

Novel Zasp52 alleles show variable IFM defects
In order to better analyse IFM defects of Zasp52, we were looking for viable Zasp52 alleles. Recently, a large number of MiMIC insertions based on the Minos transposon were created in Drosophila inserting splice acceptor sites followed by stop codons at various positions in the genome [32]. Three of them are inserted in the Zasp52 locus: MI02988 after exon 2, MI07547 after exon 8, and MI00979 after exon 15 (Fig 6A, exons numbered according to [20]). Zasp52 MI00979 truncates the last three LIM domains similar to the RNAi line iZasp52ex20 ( Fig  6B) [3]. Zasp52 MI07547 does not affect the shortest isoform, Zasp52-PP, and truncates the other isoforms just before the first LIM domain resulting in proteins containing a PDZ and ZM domain ( Fig 6B). Lastly, Zasp52 MI02988 truncates Zasp52 within the PDZ domain evenly disrupting most splice isoforms (Fig 6B). The splice trap is not fully efficient, because we can detect some residual protein at higher loading concentrations (shown for Zasp52 MI02988 in Fig  6C). In addition, both Zasp52 MI07547 and Zasp52 MI02988 should not disrupt LIM-only isoforms like Zasp52-PQ, which we cannot detect with our N-terminal Zasp52 antibody (Fig 6A and  6B). Consistent with the presence of residual protein and unaffected isoforms, IFM defects are stronger in Zasp52 MI02988 /Df(2R)BSC427 than homozygously (S3A Fig). We therefore regard Zasp52 MI02988 as a hypomorph. Zasp52 MI02988 /Zasp52 MI02988 is semiviable with lethality at all developmental stages and only 20% adult escapers (Fig 6D). In Zasp52 MI02988 /Df(2R)BSC427 adults myofibril assembly in IFM is severely disrupted. In many myofibrils, Z-discs and Mlines are no longer distinguishable (Fig 7A). Zasp52 MI00979 /Df(2R)BSC427 looks similar to the previously described RNAi lines iZasp52ex20 and iZasp52ex16 (Fig 7C) [3]. Surprisingly, Zasp52 MI07547 /Df(2R)BSC427 shows no obvious defects (Fig 7B), which is perhaps related to the presence of Zasp52-PP and Zasp52-PQ. In summary, Zasp52 MI02988 is most suitable for further analysis, because most Zasp52 isoforms are strongly reduced and it shows the most pronounced IFM defects. Having now all the tools in hand, we attempted to rescue the Zasp52 MI02988 /Df(2R)BSC427 IFM defects by expressing our transgenes with UH3-Gal4. We started with the most common isoform, Zasp52-PR, which can indeed fully rescue all IFM defects and localizes normally to Zdiscs (Fig 8A and 8B and S3B Fig). In contrast, Zasp52-PRΔPDZ cannot rescue any aspect of myofibril assembly (Fig 8C). A deletion of the 26 amino acid ZM domain can still fully rescue (Fig 8D), and surprisingly, Zasp52-PK consisting of PDZ, ZM and LIM1 domain, can also fully rescue ( Fig 8E). As expected, Zasp52-LIM234 cannot rescue, even though it localizes properly to Z-discs (Fig 8F). We also expressed Zasp52-PR in Zasp52 MI00979 /Df(2R)BSC427 mutants, which resulted in no rescue (S3C Fig). This suggests that LIM-only isoforms like Zasp52-PQ, which are disrupted in MI00979 together with all full-length isoforms confer unique functions that cannot be rescued by Zasp52-PR. Overall our data indicate that with regard to myofibril assembly, the PDZ domain fulfils crucial functions.
Removal of one copy of Zasp52 suppresses Actn 8 /+ defects The α-actinin null mutant Actn 8 has previously been shown to show a mild IFM defect heterozygously [52], and is therefore a good candidate to test for a genetic interaction with  Zasp52 PDZ Domain Is Essential for Myofibril Assembly Zasp52 MI02988 . We first analysed IFM of heterozygous Zasp52 MI02988 /+, Actn 8 /+ and the transheterozygous Actn 8 /+; Zasp52 MI02988 /+ flies by antibody stainings, but could detect no obvious differences compared to wild type (S4 Fig). We therefore employed an infrared laser tachometer to measure wing beat frequency in wild type and heterozygous flies. In y w flies, wing beat frequency is around 200 Hz with only minimal deviations (Fig 9A and 9E). Actn 8 /+ and Zasp52 MI02988 /+ flies similarly reach 200 Hz, but wing beat frequency often dips to lower frequencies (Fig 9B and 9C), although only Actn 8 /+ is significantly different from wild type ( Fig  9E). Intriguingly, the Actn 8 /+ defects are suppressed to wild type levels in transheterozygous Actn 8 /+; Zasp52 MI02988 /+ flies (Fig 9D and 9E).
This genetic interaction of Zasp52 and α-actinin is consistent with the biochemical interaction and confirms the importance of both proteins in IFM.

Discussion
Here we have analysed the domains of Zasp52 necessary for myofibril assembly and for direct interaction with α-actinin. Furthermore, the genetic interaction indicates Zasp52 is a core structural muscle protein owing to its suppression of α-actinin mutant phenotypes.

Zasp PDZ domain versus Alp/Enigma family
The nomenclature of proteins similar to Zasp52 is complex and consequently rather confusing. Originally, proteins similar to Zasp52 were classified according to their LIM domains and called Alp or Enigma family proteins depending on the number of C-terminal LIM domains [33]. As they were closely related, they were united in the Alp/Enigma family of proteins [34]. The Alp/Enigma family is also called PDZ-LIM family [35], but this is potentially confusing, because there are additional proteins (LMO7, LIMK1, LIMK2) with PDZ and LIM domains that feature a different domain order and a very divergent PDZ domain [3,34]. Additionally, in Drosophila, two genes, Zasp66 and Zasp67, encode no LIM domains, and in vertebrates several Alp/Enigma family members have functional splice isoforms without LIM domains [36][37][38]. Furthermore, by focusing on the PDZ domain, we have uncovered two additional proteins, myopodin and CHAP, featuring a very similar PDZ domain with an otherwise unrelated C-terminus. We therefore coined the term Zasp PDZ domain proteins for all proteins containing the conserved PWGFRL motif in the N-terminal half of the PDZ domain encompassing the Alp/Enigma family as well as myopodin and CHAP [3].

α-Actinin binding of Zasp PDZ domains
Most Zasp PDZ proteins except PDLIM7 and Zasp67 have been shown to bind directly or indirectly to α-actinin. However, only for ALP, RIL, CLP36, and ZASP/Cypher, the PDZ domain was confirmed to interact directly with α-actinin [7,28,[39][40][41]. To further complicate the issue, the ZM region of ALP has been shown to bind directly to α-actinin [28,42] and the ZM region, as well as the LIM domains of ZASP/Cypher colocalize with α-actinin [7,29]. Moreover, LIM domains of CRP proteins can bind α-actinin [43,44]. We therefore set out to clarify the precise biochemical and genetic interaction of Zasp52 and α-actinin. We first showed that all domains of Zasp52 can localize to Z-discs and can co-immunoprecipitate α-actinin (Figs 1  and 2). This is consistent with results obtained for ZASP [7,29], and suggests that multiple domains of Zasp52 bind to Z-disc proteins forming part of a larger complex involving α-actinin. However, only the PDZ domain can bind directly to α-actinin (Fig 4C), again consistent with yeast two-hybrid assays and in vitro biochemistry from ZASP/Cypher [7,30,41]. Still, the PDZ domain on its own interacted very weakly with α-actinin (Fig 4C), and only inclusion of a C-terminal extension provides optimal binding to α-actinin (Fig 4D). Similar results have been obtained for CLP36 and ALP, where only a 119 amino acid extended PDZ domain could bind full-length α-actinin, whereas the 85 amino acid PDZ domain proper could not [28]. PDZ domain extensions are known to play important roles in determining binding specificities, e.g. by stabilizing PDZ domain folding [45], and we here demonstrate their importance for the  Zasp52 PDZ Domain Is Essential for Myofibril Assembly non-canonical PDZ α-actinin interaction. Next, we used site-directed mutagenesis to determine the importance of the PWGFRL motif for α-actinin binding. This motif was previously implicated by mutating the central GF to AA, which disrupts ZASP binding to α-actinin [7]. The amino acids GF are, however, highly conserved across PDZ domains, for example in LMO7 (see Fig 5A), raising the possibility that mutating them disrupts PDZ domain function or folding in general. Furthermore, modelling using the NMR structure of the ZASP PDZ domain and α-actinin EF-Hand 34 suggested binding to the PWGFRL motif, but again, this complex could not be detected experimentally [46], presumably due to the lack of the C-terminal extension. We therefore introduced 5 mutations in the PWGFRL motif. The P18DW19F mutation in Zasp52-PK, and the R22T, G25D and G26W mutation in Zasp52-PP are modelled on LMO7, a PDZ-LIM domain protein unable to bind α-actinin via its PDZ domain [31]. These changes should therefore not disrupt PDZ domain folding or general function. All of these amino acids completely abolish α-actinin binding without affecting protein stability ( Fig  5 and S2 Fig), thus establishing the importance of the PWGFRL motif for α-actinin binding. Moreover, the complete absence of α-actinin binding in these point mutations is evidence that the C-terminal PDZ domain extension only enhances α-actinin binding, but does not provide any α-actinin binding capabilities on its own.

The in vivo importance of the Zasp PDZ domain
In order to perform rescue assays of myofibril assembly in IFM, we characterized three novel Zasp52 alleles. Zasp52 MI02988 evenly reduces most Zasp52 isoforms and is semiviable, potentially because of residual expression of disrupted isoforms or because of the presence of LIMonly isoforms. It has strong myofibril defects over a deficiency deleting Zasp52. Transgenes containing a PDZ domain fully rescued myofibril assembly, transgenes lacking the PDZ domain failed to rescue (Fig 8). This establishes the Zasp PDZ domain as a critical contributor to myofibril structure. Deletion of the three C-terminal LIM domains did not affect myofibril rescue, which could indicate that they carry only signalling roles, which is well established for vertebrate LIM domains [6,35,41]. Alternatively, LIM-only isoforms, which are not affected in Zasp52 MI02988 , can independently provide LIM234 functions. This is not unlikely, because several LIM-only proteins, e.g. FHL2, CRP3/MLP, and PINCH play important roles in muscle development [47][48][49]. A structural function of C-terminal LIM domains is further supported by the stronger phenotype of Zasp52 MI00979 compared to Zasp52 MI07547 . Zasp52 MI00979 disrupts LIM-only proteins and all full-length isoforms; Zasp52 MI07547 does not disrupt Zasp52-PP and Zasp52-PQ, which together reconstitute a full-length Zasp52 protein lacking only the LIM1 domain, albeit in two parts (see Fig 6). This suggests first that a PDZ ZM domain protein like Zasp52-PP can carry out important structural functions; second, that a LIM-only Zasp52 protein carries some unique function. Additional evidence for the latter is that we could not rescue Zasp52 MI00979 with the full-length Zasp52-PR transgene (S3C Fig). Surprisingly, deletion of the ZM domain had no effect (Fig 8D). However, it is likely that the ZM domain carries signalling or subtle mechanical functions that we have not detected with our assay.

Suppression of Actn 8 /+ defects by Zasp52
Important structural muscle proteins like actin, myosin heavy chain, α-actinin, and troponin, are very sensitive to dosage, showing a phenotype with only one wild type copy of the gene [50][51][52]. Two lines of evidence suggest an important structural role for Zasp52 as well. First, increasing Zasp52 dosage with a strong Gal4 driver line, but not a weak Gal4 driver line, causes IFM defects. Second, when testing for heterozygous defects of Zasp52 MI02988 with a wing beat assay, we could detect a trend towards stronger deviations in wing beat frequency, although no statistical significance (Fig 9C and 9E). In contrast, Actn 8 /+ shows clear deviations in wing beat frequency (Fig 9B and 9E), consistent with defects observed previously by electron microscopy of IFM [52]. We propose that a Zasp52 null mutant would show heterozygous defects similar to Actn 8 /+. Importantly, suppression of the Actn 8 /+ phenotype by removing one copy of Zasp52 confirms the importance of the biochemical interaction of α-actinin and Zasp52 for IFM function. It also indicates that an imbalance of these two proteins causes a stronger phenotype than low levels of both proteins. This suggests they cooperate closely in Z-disc assembly and are required at fixed levels. This genetic interaction is reminiscent of the actin gene Act88F and myosin heavy chain gene Mhc interaction in IFM. There, single heterozygotes show IFM defects not seen in transheterozygotes [50]. Zasp PDZ domain proteins are therefore versatile multifunctional proteins with many structural roles residing in and close to the PDZ domain. Some of these structural roles are mediated by direct binding to α-actinin, but given the strong phenotype of Zasp52 alleles, it is highly likely that additional binding partners are required for Zasp52 to fulfil its function in myofibril assembly.
To test for genetic interaction between Zasp52 and α-actinin, Actn 8 /FM7 was crossed to y w, and Zasp52 MI02988 /CTG and Zasp52 MI00979 /CTG were crossed to Actn 8 /FM7 or y w and incubated at 20°C. Transgenic flies were generated by injecting sequence-verified plasmids into y w (for RK and LIM234) and M{3xP3-RFP.attP}ZH-86Fb (for RR, RR ΔPDZ , and RR ΔZM ) embryos.
For GST pulldown assays, E. coli strain BL-21 bacteria expressing GST-tagged recombinant proteins were lysed by sonication in 20 mM Tris-HCl pH 8, 200 mM NaCl, 1 mM MgCl 2 , 1 mM DTT, 5% glycerol, 0.2% Triton X-100, 1 mg/ml lysozyme and complete EDTA-free protease inhibitor. The clarified cell extract after centrifugation was filtered with a 0.45 μm filter and coupled to prewashed glutathione-agarose beads (Santa Cruz Biotechnology) for 3 hours at 4°C. The beads retaining the GST-tagged proteins were washed three times with above buffer with 250 mM NaCl and 0.5% Triton X-100. Subsequently, rabbit skeletal muscle α-actinin (Cytoskeleton) was added and incubated for another 3 hours at 4°C. Final washes were in above buffers with 150 mM NaCl and 0.2% Triton X-100. Beads were resuspended in SDS sample buffer and analyzed by SDS-PAGE and immunoblotting.
Images were acquired on a LSM 510 Meta laser scanning confocal microscope using a 63x 1.4 NA Plan Apo oil immersion objective (Carl Zeiss).

Wing beat Frequency Assay
The wing beat frequency of a fly was determined using an optical tachometer as previously described [55]. In this study, 32 7-day old female flies were first glued on pipette tips, followed by measurement of wing beat frequency for 5 minutes using a tachometer (Model UT372, Uni-Trend Technology). A one-minute continuous flight window was selected from the 5-minute flight record, and the variation of that was determined by calculating the standard deviation. One-way ANOVA followed by Tukey's multiple mean difference post hoc tests were performed to determine statistically significant differences between genotypes using Prism 7 software (GraphPad).