Novel perspectives of target-binding by the evolutionarily conserved PP4 phosphatase

Protein phosphatase 4 (PP4) is an evolutionarily conserved and essential Ser/Thr phosphatase that regulates cell division, development and DNA repair in eukaryotes. The major form of PP4, present from yeast to human, is the PP4c-R2-R3 heterotrimeric complex. The R3 subunit is responsible for substrate-recognition via its EVH1 domain. In typical EVH1 domains, conserved phenylalanine, tyrosine and tryptophan residues form the specific recognition site for their target's proline-rich sequences. Here, we identify novel binding partners of the EVH1 domain of the Drosophila R3 subunit, Falafel, and demonstrate that instead of binding to proline-rich sequences this EVH1 variant specifically recognizes atypical ligands, namely the FxxP and MxPP short linear consensus motifs. This interaction is dependent on an exclusively conserved leucine that replaces the phenylalanine invariant of all canonical EVH1 domains. We propose that the EVH1 domain of PP4 represents a new class of the EVH1 family that can accommodate low proline content sequences, such as the FxxP motif. Finally, our data implicate the conserved Smk-1 domain of Falafel in target-binding. These findings greatly enhance our understanding of the substrate-recognition mechanisms and function of PP4.

We and others have shown that the EVH1 domain of the PP4R3 orthologues confers substrate specificity by directly interacting with the target proteins. For example, the EVH1 domain of the yeast PP4R3 Psy2 interacts with the MPPP short linear motif (SLiM) of the glucose signal transducer protein Mth1 [30], suggesting that PP4R3 Psy2 binds to proline-rich sequences (PRS), similar to canonical EVH1 proteins [33,34]. However, we reported previously that in Drosophila the EVH1 domain of PP4R3 Flfl specifically binds to the low-proline 19-mer Falafel-Interacting Motif (FIM) of the key centromeric protein, CENP-C [8]. This work revealed for the first time that the EVH1 domain of Falafel can also recognize ligands with low proline content, such as the FKKP (Phe-Lys-Lys-Pro) sequence within the FIM [8]. Recently, it has been reported that the EVH1 domain of the human PP4R3α SMEK1 recognizes both poly-proline-like (MxPP) and low-proline (FxxP) consensus binding motifs [32]. Here, we report on the identification of novel interacting partners of the Drosophila PP4R3 Flfl and demonstrate that Falafel EVH1 directly binds to the targets' FxxP and MxPP SLiMs, similar to its human counterpart. This further supports the idea that PP4 uses the same SLiMs for substrate recognition across species. This interaction is dependent on the Leu70 (Leu69 in humans) residue of the EVH1 domain, which is exclusively conserved in all PP4R3 orthologues. Finally, we resolve the previously unknown function of the Smk-1 domain of PP4R3 and demonstrate that it is physically involved in target recognition in flies.

Identification of novel targets of PP4R3 Flfl
PP4 is involved in the regulation of cell division, which includes centrosome function, microtubule organization, spindle assembly checkpoint activity and kinetochore integrity [5]. However, the molecular mechanisms involved in its regulatory activities remain poorly defined. Therefore, we aimed to use affinity purification coupled to mass spectrometry (AP-MS) to identify novel targets of the Drosophila PP4R3 Flfl (hereafter Falafel or Flfl) involved in cell cycle control. Our repeated attempts to identify new binding partners of transgenic Falafel purified from D.Mel-2 cultured cells (representing interphase cells) or Drosophila syncytial embryos, which characteristically undergo rapid and synchronous nuclear divisions (representing mitotically active tissue), were not successful, with only few previously identified interactors detected. As transgenic Falafel forms a functional complex with R2 and PP4c [8], we have hypothesized that the interaction between the target protein and PP4c-R2-Falafel could be transient, with the substrate being released upon dephosphorylation.
To overcome this possible limitation, we engineered truncated forms of Falafel: Flfl 1-168aa (containing the EVH1 domain, hereafter EVH1) and Flfl 169-361aa (containing the Smk-1 domain, hereafter Smk-1) each fused to glutathione S-transferase (GST). Then we performed GST pull-down from wild-type syncytial embryos followed by mass spectrometric (MS) analysis of the samples. This strategy resulted in the identification of several putative interacting partners of Falafel (electronic supplementary material, table S1). These included expected known PP4 substrates such as barrier-to-autointegration factor (BAF) [10,12], CENP-C [8], Miranda [22] and gamma-tubulin [35] as well as previously unrecognized proteins, many of which are predicted to be phosphoregulated and involved in cell division/cell cycle control, development or DNA repair. This included all four members of the chromosomal passenger complex (CPC), a key regulator of mitosis (reviewed in [36,37]). Consistent with previous work most interactors were bound to the EVH1 domain.
Surprisingly, the Smk-1 domain also pulled-down several putative partners of Falafel (electronic supplementary material, table S1), including the gamma-tubulin ring complex proteins known to be essential for spindle microtubule nucleation at the centrosomes (reviewed in [38]). In addition, we identified all three members of the conserved Rod-Zwilch-Zw10 (RZZ) complex. The RZZ is regulated by phosphorylation, and its activity stabilizes the kinetochore association of Mad1 and Mad2 spindle assembly checkpoint factors, thus ensuring accurate chromosome segregation. Smk-1 also co-purified with the heavy chain of the motor protein dynein, a previously identified RZZ interactor and a factor involved in the transport of Mad2 from kinetochores [39,40]. These findings suggest that PP4 directly regulates the spindle assembly checkpoint activity for faithful chromosome segregation and that the conserved Smk-1 domain may also be responsible for target recognition.

The EVH1 domain preferentially binds to the FxxP and MxPP consensus motifs
Canonical EVH1 proteins establish interactions with their ligands by recognizing polyproline motifs [34]. Accordingly, the EVH1 domain of the yeast PP4R3 Psy2 binds to the proline-rich MPPP SLiM of the Mth1 protein for glucose signalling regulation [30]. By contrast, we previously demonstrated that the EVH1 domain of PP4R3 Flfl could bind to the low proline content FKKP motif localized in the FIM region of CENP-C and that mutation of either the F (Phe) or P (Pro) residues abolished the interaction [8]. We therefore searched for motifs similar to FKKP and MPPP in the primary sequence of the newly identified direct interacting partners of EVH1. We found various numbers (from 1 to 7) of sequences similar to FKKP or MPPP (electronic supplementary material, figure S3A), suggesting that the consensus recognition sequences might be FxxP and MxPP. Most EVH1-targets contain several putative SLiMs, such as Prp16, which has 6 FxxP and 1 MxPP motif. However, CG8478 has a single MAPP (aa 504-507) motif. Our analysis of the adjacent sequences of several FxxP and MxPP motifs failed to establish any pattern that predicts which motif is recognized by the EVH1 domain. We therefore investigated the putative binding regions of 7 different EVH1-target proteins (Prp16, Psc, Incenp, Stwl, Sowah, CG8478 and Miranda) that interact  royalsocietypublishing.org/journal/rsob Open Biol. 10: 200343 with EVH1, in vitro. We generated a variable number of 35 S-methionine-labelled partially overlapping sequences of the target proteins by IVTT: 6 for Prp16, 7 for Psc, 6 for Incenp, 5 for Stwl, 5 for Sowah and 5 for CG8478. We made a single recombinant protein for Miranda spanning aa 1-280, a region previously reported to bind Falafel [22] and that contains 2 FxxP motifs. We were unsuccessful at expressing Centrobin pieces by IVTT. Next, we performed in vitro PPI experiments using the different target fragments as prey and GST or GST-EVH1 as bait. The binding results were analysed by SDS-PAGE followed by autoradiography (electronic supplementary material, figure S4), allowing us to identify the minimally required EVH1-interacting fragments in most cases.
To further determine which motifs are responsible for EVH1-binding we mutated the FxxP or MxPP motifs to AxxA (A, alanine) (for FxxP) and AxPA (for MxPP) in the full-length proteins and assayed them by PPI. We found that CG8478's MAPP (aa 504-507), Miranda's FRTP (aa 92-95), Prp16's FKKP (aa 98-101), Sowah's MPPP (aa 538-541) and Stwl's MVPP (aa 882-885) sequences are essential for EVH1 binding, whereas mutations of other FxxP/MxPP sequences did not alter the interaction (figure 3b-f). Interestingly, each of these targeting sequences lay in structurally disordered regions (electronic supplementary material, figure S3A). We further found that out of 2 putative SLiMs, Centrobin uses the MPPP (aa 36-39) for EVH1 binding (figure 3a). Examination of Incenp revealed that although it has 3 putative SLiMs, only mutation of the MPPP (aa 120-123) to AxPA weakened the interaction with EVH1 (figure 3g). We also managed to map the 2 interacting fragments of Psc that bind to EVH1 (aa 721-810 and aa 786-876; electronic supplementary material, figure S4F). However, mutagenesis of the 3 putative SLiMs within, and 1 outside of these regions failed to alter their interaction with EVH1 (figure 3h). This suggests that Psc binds to EVH1 in an FxxP/MxPP SLiM-independent manner. We were unable to challenge the function of the fifth motif (1201-MTPP-1204) as the protein did not express in IVTT reaction. These experiments unequivocally reveal that most of the EVH1-interacting partners we have identified bind to EVH1 via their FxxP or MxPP short linear motifs.
Intriguingly, Prp16 and Centrobin are found to be targets of the EVH1 domain of Drosophila Falafel (figure 2a) and human SMEK1, as well [32]. Although Centrobin is functionally conserved in both species, there is no significant primary sequence homology between the two orthologues (electronic supplementary material, figure S5A). Similarly, besides the conserved RNA helicase domain occupying the C-terminal region of Prp16, the N-terminal unstructured halves of the fruit fly and human orthologues do not show a high level of primary sequence conservation (electronic supplementary material, figure S5B). Prp16 is recognized by the PP4R3-EVH1 domain via its N-terminally located FKKP (aa 98-101) SLiM in Drosophila and via its FKAP (aa 39-42) motif in humans, respectively (figure 3b,i; electronic supplementary material, figure S5B). Neither SLiM is conserved. While Centrobin is bound by the EVH1 domain via its N-terminally located MPPP (aa 36-39) motif in Drosophila, it is recognized via the C-terminally positioned FRVP (aa 741-774) SLiM sequence in humans (figure 3a,i; electronic supplementary material, figure S5A). However, the human EVH1 of SMEK1 is able to bind to Drosophila Centrobin and Prp16 proteins (figure 4b).
These observations confirm that PP4 has conserved targets and it recognizes FxxP or MxPP SLiMs via the EVH1 domain of its R3 subunit. However, the position and composition of the SLiMs may differ even in the same orthologues.
Taken together with previous [8] findings, we propose that the EVH1 domain of the Drosophila PP4R3 Flfl subunit preferentially binds to the FRTP and FKKP as well as the MPPP, MAPP or MVPP short linear motifs. In humans, the EVH1 domain of the SMEK1 also preferentially binds FxxP and MxPP short motifs [32], suggesting that the substratebinding mode of the PP4R3 orthologues' EVH1 domain is conserved throughout species.

A conserved leucine in EVH1 of PP4R3
orthologues is critical for SLiM-binding in fruit flies and humans Crystallographic studies have revealed that the low proline content CENP-C FIM (1048-PDESSADVVFKKPLAPAPR-1068) is recognized by the EVH1 domain of Falafel in a polyproline II (PPII [33]) conformation and that the essential Phe1057 (F1057) occupies a hydrophobic pocket in EVH1, whereas Pro1060 (P1060) is sandwiched between the Leu70 and the conserved Trp20 residues of Falafel [8]. Through sequence alignments of EVH1 domains we demonstrated that Phe77 (1EVH numbering of murine mena [41]), which is highly conserved and invariant in all known canonical EVH1 domains and essential for ligand-binding [33], is replaced by leucine (Leu70) in Falafel EVH1 (figure 4a). More interestingly, the Leu70 (Falafel numbering) is exclusively conserved in PP4R3 orthologues, from yeast (Leu91 in Psy2) to human (Leu69 in SMEK1 and SMEK2) (figure 1b; electronic supplementary material, figure S5C). This suggests that the EVH1 domain of PP4R3 orthologues deviates from the canonical EVH1 domains and uses an invariant leucine residue in binding to atypical (low proline content) ligands, such as the FxxP (or MxPP) SLiM.
To test this assumption, we substituted the Leu70 in Falafel's EVH1 to alanine (hereafter GST-Dm-EVH1 L70A ) and Leu69 in human SMEK1's EVH1 to alanine (hereafter GST-Hs-EVH1 L69A ) and tested their interactions with 9 different Falafel EVH1 -interacting Drosophila proteins, in vitro. We found that not only was Drosophila EVH1 able to bind to these targets, but the human EVH1 also interacted with each of the Drosophila proteins (figure 4b). Thus, target recognition is conserved across species. In both cases the binding was dependent upon the conserved Leu70/69 residues with the alanine substitutions completely abolishing interaction with the target FxxP or MxPP SLiM-containing proteins (figure 4b). Strikingly, Drosophila Prp16 and Centrobin were bound to the human SMEK1's EVH1 despite their human counterparts having different SLiM recognition sequences (figure 3i; electronic supplementary material, figure S5A-B). Psc was also able to bind to Dm-EVH1 and Hs-EVH1, which was dependent on Leu70/ Leu69 residues, respectively (figure 4b).

The Smk-1 domain of Falafel is involved in target-binding
The Smk-1/DUF625 domain was first identified in the SMK-1 protein, a component of the IIs longevity pathway in C. elegans [31]. SMK-1 is the R3 subunit of PP4, which together with Daf-16 contributes to innate immunity in adult roundworms [42]. The SMK-1 domain is present in PP4R3 royalsocietypublishing.org/journal/rsob Open Biol.  orthologues from yeast to humans and its primary sequence is highly conserved in Metazoa (figure 1c). However, its function and structure are currently unknown. Our AP-MS experiment identified several putative targets of Falafel's Smk-1 domain including components of the RZZ complex, additional spindle assembly checkpoint (SAC) proteins and DNA repair determinants. We selected several candidates and tested whether they directly interact with the Smk-1 domain of Falafel using IVTT and GST-Smk-1 immobilized onto beads. This confirmed that 8 of the proteins physically interact with Smk-1 ( figure 5a,b). Of the remaining candidates, one was not expressed in IVTT (electronic supplementary material, table S1) and a few failed to bind in vitro (electronic supplementary material, figure  S2B). Although among the RZZ components both Zw10 and Zwilch showed weak but specific interactions with GST-Smk-1 in vitro (figure 5a), in D.Mel-2 cultured cells Flag-Zw10 showed stronger interaction with GFP- Smk-1 (figure 5c,d),

G S T -E V H 1 G S T I V T T in p u t G S T -E V H 1 G S T I V T T in p u t G S T -E V H 1 G S T I V T T in p u t G S T -E V H 1 G S T I V T T in p u t G S T -E V H 1 G S T I V T T in p u t G S T -E V H 1 G S T I V T T in p u t G S T -E V H 1 G S T I V T T in p u t G S T -E V H 1 G S T I V T T in
suggesting that Zw10 may link Falafel/PP4 to the RZZ complex. Even though we used equal amounts of GST-EVH1 or Smk-1 in the in vitro PPI experiments (electronic supplementary material, figure S1A-C), we noticed that Smk-1 shows a weaker binding to its targets (figure 5a) compared to the binding of EVH1 to its own targets ( figure 2a). However, we consider most of the Smk-1 interactions specific, because neither GST nor GST-EVH1 could interact with the binding partners of Smk-1 (electronic supplementary material, figure S2A).
Although neither the 3D structure nor mode of ligandbinding of the Smk-1 domain has been revealed so far, it is not reasonable to suppose that it recognizes the same SLiMs as the EVH1 domain does. Indeed, among the 8 identified direct interactors of Smk-1, 7 proteins do not contain FxxP or MxPP motifs at all (electronic supplementary material, figure  S3B). Zw10 is the only one to contain FxxP sequences (aa 238-241 and 438-441), whose mutation to AxxA did not 16 23   5d). This suggests that Smk-1 uses a completely different, but still unknown mechanism for target recognition. Our attempts to identify consensus binding-motifs within the full-length proteins or narrow-down Smk-1-interacting regions of the target proteins (electronic supplementary material, figure S6) have failed so far.

Discussion
According to Paul Ehrlich's key-lock principle, agents only work when they are bound (Corpora non agunt nisi fixata) [43]. This rule also applies to phosphatases, which must recognize and physically bind to their targets to catalyse their dephosphorylation. It is well established that PP1, G S T -S m k -1 G S T i n p u t i n p u t C o I P C o I P I V T T i n p u t  royalsocietypublishing.org/journal/rsob Open Biol. 10: 200343 PP2A and PP2B/Calcineurin Ser/Thr PPPs use their catalytic or regulatory domains to recognize short linear motifs (SLiMs) in their target substrate proteins. The catalytic subunit of PP1 directly interacts with the RVxF-type docking motif [44,45], which might be supported by the N-terminally localised SILK or MyPhoNe sequences [46]. The catalytic subunit of PP2B/Calcineurin directly binds to its SLiMs (LxVP and PxIxIT), too, independently of the regulatory domain [47,48]. However, the SLiM preference of PP2A is determined by its various regulatory subunits: B56 recognizes the LxxIxEx/LSPIxE motif [49,50], while B55 binds its substrates differently [51]. By contrast, comparatively little is reported about the mechanisms employed by other PPP family members (PP4-7) to target their substrates. In higher eukaryotes, the PP4c catalytic subunit of PP4 can form multiple, mutually exclusive complexes with differing regulatory subunits, including the heterodimeric PP4c-R1, PP4c-R4 as well as the heterotrimeric PP4c-R2-R3, which is the major form and is present from yeast to humans (reviewed in [5]). It is believed that the regulatory R3 subunit directly interacts with PP4 substrates through its N-terminal EVH1 domain, but the target SLiM has not been identified. Since canonical EVH1 domains bind to proline-rich sequences, it was reasonable to suppose that the R3 orthologues bind to high proline content SLiMs, too. The discovery that yeast R3 Psy2 interacts with the MPPP sequence of the glucose signal transducer protein Mth1 [30] had partially supported this idea. However the MPPP SLiM was slightly different from typical EVH1 ligands [33,34]. We previously reported that the EVH1 domain of the R3 subunit (Falafel) of the Drosophila PP4 can specifically bind to low-proline sequences, as well. For the first time, we demonstrated that the EVH1 preferentially bound to the FKKP-containing FIM motif of CENP-C, in which the F and P were essential for the interaction [8]. Here, we describe the discovery of novel physical interactors of the EVH1 domain of Falafel. We have identified several putative SLiMs similar to the FKKP and MPPP sequences and demonstrate that the consensus binding motifs recognized by PP4's EVH1 domain are FxxP and MxPP. A recent finding was also made using the human PP4R3 SMEK1 [32], revealing that the binding mechanism of PP4 has been highly conserved throughout evolution. This is further supported by our observation that the human SMEK1-EVH1 is able to bind to all Drosophila proteins that are direct interactors of Falafel-EVH1, including Prp16 and Centrobin. Remarkably, Drosophila and human Centrobin are only functional homologues, and the PP4-bound FRVP SLiM present in the human protein [32] is different from the MPPP sequence we identified in its fruit fly counterpart. We have identified a similar interchangeability for Prp16. From these and other complementary data, we propose that not only the binding mechanism of PP4 but also its substrate set and accompanying regulatory roles are conserved and that this is irrespective of sequence similarity or target protein SLiMs. Indeed, FxxP and MxPP sequences are quite frequent even in the simple proteomes, raising the likelihood of PP4-mediated regulation. Given the short sequence length it is unclear how FxxP or MxPP motifs are selected by PP4. It is unlikely to require unique topological cues as these sequences commonly reside in disordered regions of the protein. Future work will be required to define this mechanism.
In the three already identified classes of canonical EVH1 domains belonging to the Ena/Vasp/Homer/Spred families [33,34] a conserved phenylalanine and tryptophan are essential for the formation of the recognition surface that accommodates the proline-rich ligand. We discovered that in Falafel the conserved phenylalanine is replaced by leucine and experimentally proved that this residue is essential for binding to FxxP and MxPP motif-containing target proteins. Moreover, we demonstrated that this leucine is invariant in all EVH1 domains of the PP4R3 orthologues, from yeast to humans. Accordingly, mutation of Leu70 (in fruit fly Falafel) or Leu69 (in human SMEK1) to alanine altered the interaction with the target proteins. These findings reveal a novel mode of PP4 binding, in which a conserved Leu is essential. They further suggest that the EVH1 domain of PP4 may represent a new type (or class) within the EVH1 family that is able to accommodate low proline content sequences, such as the FxxP motif. Given the increased flexibility offered by PP4R3 orthologues that use leucine instead of phenylalanine it is unclear why the former is not more prevalent. It will be a future challenge to unveil the background for this evolutional selection.
Finally, our work sheds light on the function of the Smk-1 domain of Falafel. Surprisingly, although Smk-1 is present in human SMEK1 and SMEK2, too, its function was not described in the recent work by Ueki and colleagues [32]. Indeed, it is not clear whether Smk-1 represents an independent domain or is just part of the ARM/Heat repeats occupying the middle regions of R3 orthologues. Nevertheless, it has been shown that the Smk-1 domain of mammalian SMEK1 physically interacts with Par3 in neural progenitor cells [29], suggesting that PP4R3 uses at least two different substrate-binding mechanisms through its conserved domains, EVH1 and Smk-1. Our findings unequivocally support this model. We identified several Smk-1 interacting proteins in Drosophila, including components of the spindle assembly checkpoint, a process that is believed to be regulated by PP4. We showed that Smk-1 binds to its targets in an FxxP-and MxPP-independent manner, however, we were not able to identify the regions specific to Smk-1. It would be important to identify more binding partners of Smk-1 in flies or in mammalian models by AP-MS or other techniques, and clarify the crystal structure and mode of binding of this domain. This would contribute to a better understanding of how the Smk-1-mediated targetbinding of PP4 works and why PP4R3 orthologues have different binding mechanism.
One may assume that the two target-binding domains with different structure in a single regulatory subunit allow the PP4c-R2-R3 holoenzyme to recognize and thus regulate different types of molecules in different biological processes and in various tissues. Having two independent domains in a single protein can give maximum flexibility but also maximum stringency to the enzyme. Future studies are required to investigate this hypothesis in detail. 4 1-168)) or Drosophila Zw10 and Zwilch were respectively cloned into the pDONR221 vector using the Gateway Cloning System (Thermo Fisher Scientific). Entry clones of Falafel domains (Flfl 1-168aa (EVH1) and Flfl 169-361aa (Smk-1)) were previously described [8]. Entry clones containing Dm-EVH1 L70A , Hs-EVH1 L69A , Zw10-238 AxxA 241 or Zw10-438 AxxA 441 were created according to standard procedures using the QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies). Expression constructs were made by LR reaction using the following destination vectors: pUGW (N-terminal eGFP fusion under the regulation of the constitutively active polyubiquitin promoter in D.Mel-2 cultured cells (Drosophila Gateway Vector Collection, DGVC)), pAFW (N-terminal 3xFLAG fusion under the regulation of the constitutively active actin5c promoter in D.Mel-2 cells, DGVC), pF3A_3xFlag-GW (N-terminal 3xFLAG fusion under the regulation of the SP6 promoter for wheat germ-based in vitro expression system, in house modified pF3A-WG (Promega)) and pDEST15 (N-terminal GST fusion in Escherichia coli, Thermo Fisher Scientific). Drosophila cDNAs in pOT2 vector or cDNAs sub-cloned into the pHY22 plasmid were used directly in the T7 promoter-based IVTT expression system to generate 35

Recombinant protein expression and purification
GST, GST-EVH1, GST-Dm-EVH1 L70A , GST-Smk-1, GST-Hs-EVH1 and GST-Hs-EVH1 L69A were expressed in E. coli SixPack strain [52] and purified to homogeneity as follows. Cells were grown in 50 ml LB medium to A 600 approximately 0.6 and expression was induced with 0.5 mM isopropyl 1-thio-β-D-galactopyranoside for 5 h at 25°C. Cells were lysed by ultrasound disruption in 30 ml phosphate buffered saline (PBS) supplemented with 1 mM PMSF and 0.2 mg ml −1 lysozyme, followed by centrifugation at 16 000×g at 4°C for 15 min. The cleared supernatant was loaded onto pre-equilibrated glutathione sepharose 4B resin (GE Healthcare) and incubated at 4°C for 1.5 h. Then, beads were washed five times with PBS supplemented with 0.05% Triton X-100, immobilized proteins were kept on beads and stored at −20°C in PBS supplemented with 50% glycerol.

IVTT and in vitro binding assays
For in vitro PPI experiments 35 S-methionine-labelled prey proteins (full-length wild-type or mutant and truncated forms) were produced in vitro using TnT T7 Quick IVTT (Promega, L1170). Fifty nanograms of purified PCR products or recombinant plasmids were added to a 15 µl reaction, containing TnT Quick Master Mix, RNasin Plus RNase Inhibitor, T7 TnT PCR Enhancer, EDTA-free protease inhibitor cocktail (Merck) and 0.3 MBq Methionine-L [ 35 S] (Perkin Elmer). After incubating the mixture at 30°C for 1 h, samples were centrifuged at 12 000×g at 25°C for 5 min. The supernatant (IVTT input) was divided into equal parts and used for PPI experiments in which GST served as a negative control and GST-fused proteins were used as bait. GST or GST-fused bait proteins were immobilized on glutathione sepharose 4B resin (GE Healthcare), equilibrated with 1 ml WB1 (50 mM HEPES pH 7.5, 150 mM NaCl, 2 mM MgCl 2 , 1 mM EGTA, 1 mM DTT, 0.1% Triton X-100), beads were settled by centrifugation (600×g, 4°C for 3 min), re-suspended in 800 µl of binding buffer (WB1 supplemented with EDTA-free protease inhibitor cocktail (Merck) and 0.5 mg ml −1 BSA), mixed with an equal amount of 35 S-methionine-labelled prey proteins and incubated for 90 min at 4°C. Beads were washed three times with WB1 and three times with WB2 (WB1 supplemented with 50 mM NaCl and 0.1% Triton X-100), transferred to new tubes and boiled in 10 µl Laemmli sample buffer. Proteins (5% IVTT input and 100% bound) were separated by SDS-PAGE. Resultant gels were stained with Coomassie Brilliant Blue, scanned, dried and directly used for autoradiography. Exposure to autoradiography film (Kodak) was carried out at −80°C.
Flag-ROD (from pF3A_3xFLAG-ROD plasmid) was expressed in TnT SP6 High-Yield Wheat Germ Protein Expression System according to the manufacturer (Promega), separated by SDS-PAGE, blotted to nitrocellulose (Amersham) and visualized by immunoblotting using an anti-FlagM 2 antibody.

Co-Immunoprecipitation from D.Mel-2 cells
D.Mel-2 cells (Life Technologies) were grown in Insectagro DS2 serum-free medium (Corning) supplemented with 2 mM stable-glutamine (Biosera) and PenStrep (Gibco). Cells were transiently co-transfected using Cellfectin II reagent (Thermo Fisher Scientific) with GFP or GFP-Smk-1 (in pUGW plasmid) and Flag-Zw10 or Flag-Zwilch (in pAFW plasmid, wild-type or mutant), respectively and collected 48 h post-transfection. Cells were lysed in 500 µl EB buffer (supplemented with 25 µM MG132 and 0.1 µl ml −1 Benzonase Nuclease) by passing the cell suspension through a G25 needle (ten times) followed by centrifugation at 16 000 × g at 4°C for 10 min. The supernatants were incubated with GFP-Trap magnetic agarose beads (Chromotek) at 4°C for 90 min. The beads were washed five times with EB buffer and boiled in Laemmli sample buffer. Inputs and bound proteins were separated by SDS-PAGE, transferred to polyvinylidene fluoride (PVDF) membrane and analysed by immunoblotting using anti-GFP and anti-FlagM 2 antibodies.