A new perspective on the evolution of the interaction between the Vg/VGLL1-3 proteins and the TEAD transcription factors

The most downstream elements of the Hippo pathway, the TEAD transcription factors, are regulated by several cofactors, such as Vg/VGLL1-3. Earlier findings on human VGLL1 and here on human VGLL3 show that these proteins interact with TEAD via a conserved amino acid motif called the TONDU domain. Surprisingly, our studies reveal that the TEAD-binding domain of Drosophila Vg and of human VGLL2 is more complex and contains an additional structural element, an Ω-loop, that contributes to TEAD binding and in vivo function. To explain this unexpected structural difference between proteins from the same family, we propose that, after the genome-wide duplications at the origin of vertebrates, the Ω-loop present in an ancestral VGLL gene has been lost in some VGLL variants. These findings illustrate how structural and functional constraints can guide the evolution of transcriptional cofactors to preserve their ability to compete with other cofactors for binding to transcription factors.


Introduction
The Hippo pathway, which controls organ size and tissue regeneration via the regulation of cell division and apoptosis (Ma, Meng et al., 2019, Wang, Yu et al., 2017, is an intense field of research in regenerative medicine (Fu, Plouffe et al., 2017, Hong, Meng et al., 2016, Moya & Halder, 2018, Wang, Liu et al., 2018. Furthermore, deregulation of Hippo signalling in several cancers suggests that molecules targeting this pathway may have antineoplastic properties (Janse van Rensburg & Yang, 2016, Maugeri-Sacca & De Maria, 2018, Park, Shin et al., 2018, Santucci, Vignudelli et al., 2015, Zanconato, Cordenonsi et al., 2016 (but see (Kakiuchi-Kiyota, Schutten et al., 2019, Moya, Castaldo et al., 2019). The Hippo pathway regulates, via a signalling cascade involving the MST1/2 and LATS1/2 protein kinases, the TEAD (TEA/ATTS domain) transcription factors that, upon binding to YAP (Yes associated protein) or VGLL1-3 (Vestigial-like), become transcriptionally active (Landin-Malt, Benhaddou et al., 2016, Lin, Park et al., 2017, Liu, Wong et al., 2012, Ma et al., 2019. The accurately determine its affinity using a standard SPR procedure. Therefore, we used the kinetic titration method, which allows Kd determination for high affinity ligands (Karlsson, Katsamba et al., 2006). To validate this methodology for our proteins, we purified Yki 30-146 , which according to results obtained with the equivalent YAP 50-171 fragment should have a high affinity for wt Sd (Hau et al., 2013). Since the Kd values of Yki 30-146 measured by both SPR methods are similar (Table 1), kinetic titration was applied to Vg 288-337 and a Kd KT = 1.6 nM was determined (Table 1). Vg 288-337 binds 190 times more tightly to wt Sd than the TONDU domain alone  ), showing that the region Vg 312-337 (including the putative Ω-loop) dramatically enhances the affinity for wt Sd . To determine whether additional regions outside Vg 288-337 contribute to the binding to wt Sd , the longer protein fragment Vg 263-382 was purified. The affinity of Vg 263-382 -Kd KT = 0.38 nM (Table 1) -is similar (fourfold difference) to that of Vg 288-337 , suggesting that no other regions from Vg 263-382 contribute significantly to the interaction with wt Sd . The residues in the β-strand of the Sd/TEAD-binding domain of YAP/Yki are variable across species and their contribution to the interaction with Sd/TEAD is limited (Hau et al., 2013, Li et al., 2010. In contrast, the amino acids from the β-strand in the Sd/TEAD-binding domain of Vg/VGLL1-3 are well conserved (Simon et al., 2016) suggesting that they might contribute more to the interaction with these transcription factors. The Kd values for wt Sd of Yki 41-79 and Vg 298-337 , which both lack the putative β-strand region (Fig 1A), are 69 nM and 61 nM, respectively ( Table   1). As observed for YAP, the deletion of the putative β-strand of the Sd-binding domain of Yki has little effect on binding to wt Sd (only a threefold change in Kd). In contrast, the deletion of the putative β-strand from the Sd-binding domain of Vg leads to a 40-fold reduction in binding, indicating that this secondary structure element plays an important part in the interaction with wt Sd .

Structural characterization of the Vg:Sd interaction. To study the interaction between Vg and
Sd at the atomic level, we set up co-crystallization experiments between wt Sd and Vg 288-337 as well as Vg 298-337 and we obtained crystals of the Vg 298-337 :wt Sd complex diffracting at 1.85 Å (PDB 6Y20; Table S1). In agreement with their high amino acid sequence homology and in line with our CD data (Fig S4A), Sd and TEAD have a very similar three-dimensional structure ( Fig S6). The superimposition of the structure of the Vg:Sd complex on that of the YAP:TEAD and VGLL1:TEAD complexes show that the α-helix of the Sd/TEAD-binding site of these three co-factors bind to a similar location at the surface of Sd/TEAD (Fig 2A). The axes of the αhelix of Vg and YAP are not aligned, as also previously noticed between VGLL1 and YAP (Mesrouze, Erdmann et al., 2016), but the α-helices of Vg and VGLL1 bind in a similar fashion ( Fig 2B). The three conserved residues in the VxxHF motif from the TONDU domain of VGLL1 and Vg occupy the same position and His305 Vg forms the same two hydrogen bonds with Sd (with Ser342 Sd and Val395 Sd ) as His44 VGLL1 does with TEAD in the VGLL1:TEAD complex (Pobbati et al., 2012) (Fig 2C). The structure of the Vg:Sd complex shows that the Sdbinding domain of Vg indeed contains an Ω-loop, which binds to Sd in a similar fashion than the Ω-loop of YAP binding to TEAD (Fig 2A). The linker region connecting the α-helix and the Ω-loop of Vg is not completely resolved in our structure probably because of its flexibility (Fig 2A). The three hydrophobic residues Met324 Vg :Phe329 Vg :Phe333 Vg present in the Ω-loop of Vg occupy the same position at the binding interface as Met86 YAP :Leu91 YAP :Phe95 YAP do ( Fig 2D). They also create a hydrophobic core within the bound Ω-loop, which is stabilized in the YAP:TEAD complex by Phe96 YAP that makes a π-cation interaction with Arg87 YAP ( Fig   2E). In the Vg:Sd complex, Trp334 Vg is at the position of Phe96 YAP but it does not engage in a π-cation interaction because Ser325 Vg replaces Arg87 YAP . This suggests that the presence of a larger aromatic residue at this position might be sufficient to stabilize and shield the hydrophobic core of the bound Ω-loop from solvent. The key salt bridge between Arg89 YAP and Asp272 TEAD4 found in the YAP:TEAD complex  is also observed between Arg327 Vg and Asp276 Sd in the Vg:Sd complex ( Fig 2F). Finally, Ser332 Vg is within hydrogen bond distance of Glu267 Sd :Tyr435 Sd , as is the case for Ser94 YAP and Glu263 TEAD4 :Tyr409 TEAD4 in the YAP:TEAD complex ( Fig 2F). Overall, the structure of the Vg 298-337 :Sd 223-440 complex clearly demonstrates the presence of an Ω-loop in the Sd-binding domain of Vg and that it interacts with Sd in a manner similar to that of the Ω-loop of YAP with TEAD.

Effect of mutations in the Ω-loop binding pocket of Sd.
The β-strand:α-helix region of YAP (YAP 44-74 ) has a low affinity for TEAD (Mesrouze et al., 2014) while its Ω-loop (YAP 85-99 ), binds more tightly to it (Hau et al., 2013, Zhang, Lin et al., 2014. Consequently, mutations in the Ω-loop binding pocket of TEAD are more destabilizing than mutations in the β-strand:αhelix binding pocket (Li et al., 2010. Since the β-strand:α-helix region of Vg (Vg 288-311 ) has a high affinity for wt Sd (Table1), we hypothesized that the Vg:Sd complex might be more permissive to mutations in the Ω-loop binding pocket of Sd than the Yki:Sd complex, which should behave like the YAP:TEAD complex. To test this hypothesis, we engineered mutations disrupting key interactions at the Ω-loop binding interface. The Asp276Ala Sd mutation prevents the formation of a salt bridge with Arg327 Vg in the Vg:Sd complex ( Fig 2F) and probably with Arg69 Yki in the Yki:Sd complex. The same mutation in TEAD4, Asp272Ala TEAD4 , has a strong destabilizing effect on the YAP:TEAD complex (Mesrouze, Bokhovchuk et al., 2018). The Tyr435His Sd deletes the hydrogen bond with Ser332 Vg in the Vg:Sd complex ( Fig 2F) and probably with Ser74 Yki in the Yki:Sd complex.
This mutation in TEAD1, Tyr421His TEAD1 , which is at the root of Sveinsson's chorioretinal atrophy (Fossdal, Jonasson et al., 2004), has a major negative impact on the YAP:TEAD complex (Bokhovchuk, Mesrouze et al., 2019). Asp246Ala Sd and Tyr435His Sd were purified, and their CD spectrum shows that the mutations do not affect the structure of Sd ( Fig S4B) as previously observed with the corresponding TEAD4 mutations (Bokhovchuk et al., 2019. The affinity of Vg 263-382 and of Yki 30-146 for Asp276Ala Sd is reduced by 1.58 and 3.6 kcal/mol, respectively (Table 2) and the Tyr435His Sd mutation lowers the binding of Vg 263-382 and of Yki 30-146 by 2.44 and 4.11 kcal/mol, respectively (Table 2). Therefore, the Vg:Sd complex is less destabilized by these two mutations than the Yki:Sd complex, suggesting that the higher affinity of the β-strand:α-helix region of Vg for wt Sd mitigates the impact of disruptive mutations at the Ω-loop binding pocket.

Presence of an Ω-loop in other proteins of the Vg/VGLL1-3 subfamily. The presence of an
Ω-loop in the Sd-binding domain of Vg from D. melanogaster could be specific for this species or genus. In agreement with the initial observation made by Ohde et al. (Ohde et al., 2009), our analysis shows that the Vg proteins of other species belonging to the same or different insect orders contain a sequence that has strong homology with the residues located in the Ω-loop of Vg from D. melanogaster ( Fig S7). Therefore, Vg may not be the only member of the Vg/VGLL1-3 family to possess an Ω-loop. This prompted us to turn our attention to the VGLL proteins of higher animal species and more particularly to human VGLL1-3. As previously reported (Simon et al., 2016), the β-strand:α-helix region is very well conserved between Vg and VGLL1-3 ( Fig 1B). VGLL1 has no homology with Vg 323-334 ( Fig 1B) indicating that it does not possess an Ω-loop in agreement with the published structural data (Pobbati et al., 2012).
VGLL3 140-151 shows some level of homology with Vg 323-334 , but several residues (e.g. Met324 Vg or Arg327 Vg ) required for the interaction between Vg and Sd are missing in VGLL3 ( Fig 1B) suggesting that VGLL3 140-151 should only bind very weakly (if at all) to TEAD. In contrast, VGLL2 135-146 shows significant sequence homology with Vg 323-334 ( Fig 1B) indicating that this region of VGLL2 may form an Ω-loop upon binding to TEAD. To measure the affinity for TEAD4 of VGLL2 [135][136][137][138][139][140][141][142][143][144][145][146][147][148][149] (Ω-loop) and Vg 323-337 (Fig 1B), we used a TR-FRET assay (Hau et al., 2013, Mesrouze et al., 2014 because VGLL2 135-149 and VGLL3 140-154 at high concentration bind in a non-specific manner to SPR sensor chips. In the TR-FRET assay, we found that VGLL3 140-154 has a very weak potency (IC50 > 500 μM;  Table 3) and Vg 323-337 (157 μM; Table 3). This finding suggests that VGLL2, in contrast to VGLL1 and VGLL3, possesses an Ω-loop. Since VGLL2-derived peptides are rather unstable in solution (presence of three cysteines in their sequence), making them difficult to handle in crystallization experiments, we built up a molecular model of the VGLL2:TEAD complex. This model suggests VGLL2 to form an Ω-loop upon binding to TEAD, and the superimposition on the Vg:Sd structure shows that it could bind to TEAD4 in a manner similar to the way Vg binds to Sd (Fig 3). The main structural difference between bound Vg and VGLL2 resides in the linker region connecting the β-strand:α-helix and the Ω-loop. As the linker is much longer in VGLL2 (27 residues instead of 12), it adopts a more extended conformation in the VGLL2:TEAD4 complex. On the basis of these observations, we next measured the affinity of VGLL2 85-108 (βstrand:α-helix) and VGLL2 85-149 (β-strand:α-helix:Ω-loop) ( Fig 1B) to TEAD4. VGLL2  has an affinity (590 nM, Table 3) similar to that of its Vg counterpart, Vg 288-311 (291 nM, Table   1). The Kd of VGLL2 85-108 is in the low nanomolar range (34 nM, Table 3), showing that the presence of the Ω-loop substantially increases complex formation between VGLL2 and TEAD4. However, VGLL2 85-149 binds less tightly to TEAD4 (Table 3) than Vg 288-327 to Sd or TEAD4 (Tables 1 and 3). Since the isolated β-strand:α-helix and Ω-loop regions of VGLL2 and Vg have a similar affinity for TEAD or Sd, we hypothesized that the longer linker region connecting these elements in VGLL2 (Fig 1B and 3) has a negative impact on the interaction with TEAD4. To test this hypothesis, we designed a hybrid peptide (hereafter referred to as Flyman) formed of the β-strand:α-helix and Ω-loop of VGLL2 connected by the linker from Vg ( Fig 1B). Flyman is more potent than VGLL2 85-149 and its potency is similar to that of Vg 288-337 (Table 3) suggesting that the linker found in VGLL2 has a negative effect on the binding of VGLL2 85-149 to TEAD4.
To determine whether the Ω-loop region of VGLL2, VGLL2 135-146 also contributes to the interaction with TEAD in cells in the context of the full-length proteins, N-terminally V5tagged wt TEAD4 and (FLAG)3-wt VGLL2 were co-transfected into HEK293FT cells harboring a genomic deletion of YAP and TAZ (transcriptional co-activator with PDZ-binding motif, a paralog of YAP) (see below). Following V5-mediated immunoprecipitation, wt TEAD4 and bound wt VGLL2 were monitored using Western blot with V5 and FLAG-antibodies, respectively. wt VGLL2 was efficiently co-immunoprecipitated with wt TEAD4 , while it was not detected in similar experiments carried out in the absence of V5-tagged TEAD4 (empty vector control) ( Fig 4A). To test whether VGLL2 interacts with TEAD4 via its Ω-loop region, we used the delΩ VGLL2 mutant protein, in which the residues 135-146 have been deleted. The deletion significantly decreases the amount of VGLL2 protein co-immunoprecipitated with wt TEAD4 indicating that the region 135-146 is required for efficient binding to TEAD4 (Fig 4A). To determine whether VGLL2 interacts with TEAD4 via its Ω-loop binding pocket, we used the Asp272Ala TEAD4 mutant. According to our molecular model, this mutation should prevent the formation of a salt bridge with Arg139 VGLL2 and it should have the same disruptive effect on the VGLL2:TEAD4 complex as the Asp246Ala Sd mutation on the Vg:Sd complex ( Table 2).
The amount of wt VGLL2 co-immunoprecipitated with Asp272Ala TEAD4 is lower than that detected in the presence of wt TEAD4 ( Fig 4A). Altogether, these findings show that the Ω-loop region, VGLL2 135-146 , is required for an efficient interaction between full-length VGLL2 and TEAD4 in cells. Upon quantification of the immunoprecipitated fraction (IP) over input ratio across three separate experiments which exhibited a similar pattern, the Asp272Ala TEAD4 mutation or the Ω-loop deletion of VGLL2 led to a 2 to 2.5-fold reduction in the VGLL2:TEAD4 interaction ( Fig 4B). The combination of these two alterations ( Fig 4A) did not further decrease the extent of co-immunoprecipitation (~ 2-fold reduction, Fig 4B). This supports the notion that the two mutations affect the same binding interface; if they disrupted interactions at distinct interfaces, then an additive or a synergistic effect would be expected.
Similar results with regard to VGLL2:TEAD4 complex formation and the impact of the Asp272Ala TEAD4 and delΩ VGLL2 mutations were observed in wild-type HEK293FT (i.e. with intact YAP and TAZ). However, the inter-experimental reproducibility across the four probed co-IP conditions was less robust. We hypothesized that this was a consequence of the competition between YAP, TAZ and VGLL2, which bind with different affinities to the Ω-loop binding pocket of TEAD, leading to an increased susceptibility to multi-parametric experimental variations in complex stoichiometry. To facilitate the mechanistic dissection and increase experimental robustness, we therefore opted for the use of double YAP/TAZ knockout cells in our studies. We also repeatedly observed that the expression levels of FLAG-tagged VGLL2 were significantly higher in co-transfection experiments performed with TEAD4 proteins (Fig 4). In contrast to our previously reported observations for FAM181A , this effect happened with wild-type and mutant forms of TEAD4 and VGLL2, suggesting a the formation of a complex formation between these two proteins. This is presumably mediated through the interactions taking place at the β-strand:αhelix binding site, which are unaffected by the Asp272Ala TEAD4 or the deletion of the Ω-loop and may enhance the stability of co-overexpressed VGLL2 in this experimental setting.

Functional role of the Ω-loop from Vg in Drosophila.
To investigate the functional contribution of the Ω-loop in an in vivo context, we generated several transgenic drosophila fly lines that conditionally express different HA-tagged forms of Vg. We obtained UAS lines containing cDNAs encoding either full-length Vg, or mutant versions bearing a deletion of the Ω-loop (delΩ), a deletion of the β-strand:α-helix (delβ:α) or the deletion of the whole Sdbinding domain of Vg (delβ:α:Ω). In the larval wing primordium, the wing imaginal disc, Vg-Sd interaction plays a major role in the specification of the wing cell fate; the deletion of Vg leads to a complete loss of the wing pouch and subsequently the adult wing (Williams, Bell et al., 1991). Strikingly, ectopic expression of Vg outside its wild-type expression pattern leads to ectopic wing outgrowths in the wing imaginal disc, whereas ectopic expression in other imaginal discs leads to their transformation into wing fate (Baena-Lopez & Garcia-Bellido, 2003, Simmonds et al., 1998. We used such in vivo ectopic expression assays to probe the function of the different forms of Vg mentioned above. Upon 24 h ectopic expression using an en-gal4 driver line expressed in the posterior compartment of the wing imaginal disc, we detected comparable level of expression of the four HA::Vg proteins ( Fig 5A). Expression of HA::Vg or HA::Vg delΩ resulted in tissue outgrowth and repression of the hinge specific protein Homothorax (Hth) in the posterior compartment ( Fig 5B), suggesting hinge to wing transformation via these two functional Vg proteins.
Expression of HA::Vg delα:β and HA:Vg delα:β:Ω did not lead to morphological defects and no change in the distribution of Hth were observed. Consistently, the presumptive wing region, delimited by a Wingless (Wg) ring pattern, was expanded upon expression of HA::Vg or HA::Vg delΩ , but not when HA::Vg delα:β and HA:Vg delα:β:Ω where expressed ( Fig 5C). These results show a strong dependency on the β-strand and α-helix for full Vg function, but fail to reveal differences between wild-type Vg and Vg delΩ . Therefore, we turned our attention to a different ectopic expression assay and expressed the four Vg constructs under the eye-specific driver line ey-Gal4. As expected based on previous studies (Takanaka & Courey, 2005), expression of HA::Vg resulted in different eye phenotypes ( Fig 5B), including wing-like tissue outgrowths in the eye, with a high phenotypic penetrance ( Fig 5C). Expression of HA::Vg delΩ was also able to generate such outgrowths, but the size and penetrance of these was lower than upon HA::Vg expression. Expression of HA::Vg delα:β and HA::Vg delα:β:ω failed to change the fate of eye cells. Together, these results show again a strong dependence on the TONDU domain (β-strand:α-helix region), and also reveal a moderate but significant role of the Ω-loop in the function of Vg.

Discussion
VGLL1 and YAP bind via a similar β-strand:α-helix motif to an overlapping region at the surface of the TEAD transcription factors , Li et al., 2010, Pobbati et al., 2012. This motif, which is highly conserved amongst the Vg/VGLL1-3 proteins (TONDU domain) (Simon et al., 2016), is the only structural element known to be involved in the interaction between these co-factors and the TEAD transcription factors. However, YAP (and its paralog TAZ (Kaan, Sim et al., 2017)) requires in addition to the β-strand:α-helix motif an Ω-loop for efficient binding to TEAD (Hau et al., 2013, Li et al., 2010. Therefore, the main difference between the TEAD-binding domain of the Vg/VGLL1-3 and YAP-like proteins is the presence of an Ω-loop in the latter (Gibault, Coevoet et al., 2018, Pobbati et al., 2012, Santucci et al., 2015. The data presented in this article challenge this view, revealing that the Sd/TEAD-binding domain of some Vg/VGLL1-3 proteins also contains an Ω-loop. We demonstrate that Vg from D. melanogaster has an Ω-loop that is required for tight binding to Sd and our in vivo assays show that in some ectopic assays (such as the transformation of the eye to wing-like structures), full function of the Vg protein requires the Ω-loop. Furthermore, we show that human VGLL2, in contrast to human VGLL1 and VGLL3, also possesses an Ωloop that contributes to its interaction with TEAD. Altogether, these findings indicate that the Sd/TEAD-binding domain of the Vg/VGLL1-3 proteins exists in two different forms: a βstrand:α-helix motif or a β-strand:α-helix:Ω-loop motif. This suggests that to be fully functional, all the Vg/VGLL1-3 proteins must possess a TONDU domain but that the presence of an Ω-loop is not required for some of them (VGLL1, 3) while for others (Vg, VGLL2) it might be needed to reach complete activity. In the following, we shall discuss a hypothesis to explain the evolution of this protein family.
The whole genome duplications that occurred at the origin of the vertebrate lineage (Dehal & Boore, 2005, Panopoulou, Hennig et al., 2003 probably led to the presence of duplicates of an ancestor Vg gene in the genome of early vertebrates. As VGLL1-3 from vertebrates derive from Drosophila Vg (Faucheux, Naye et al., 2010), it is likely that the duplicated VGLL proteins present in early vertebrates contained an Ω-loop. The presence of multiple forms of the same gene in genomes enables the accumulation of mutations that can lead to the modification of the ancestral gene (Huminiecki & Wolfe, 2004, Lynch & Katju, 2004, Seoighe, Johnston et al., 2003. The nanomolar affinity of the isolated TONDU domain of Vg/VGLL2 and the triple-digit micromolar affinity of their isolated Ω-loop indicate that the TONDU domain is the "hot spot" for the interaction of these proteins with Sd/TEAD. Since isolated TONDU domains and the TEAD-binding domain of Yki/YAP (or TAZ) have a similar affinity for TEAD (here and (Bokhovchuk et al., 2019)), the VGLL proteins need to contain only a TONDU domain to bind to TEAD and to compete with YAP (or TAZ) for binding to TEAD, as is observed in cells with VGLL1 and VGLL3 (Figeac, Mohamed et al., 2019, Pobbati et al., 2012. Therefore, mutations in the Ω-loop region of the VGLL proteins from early vertebrates could take place and be tolerated during evolution because the mutated proteins would still be able to bind with sufficient affinity to TEAD via their TONDU domain. We propose that the accumulation of mutations during this evolutionary process led to the loss of the Ω-loop in VGLL1 and VGLL3. VGLL2 followed a different trajectory since it still contains a functional Ω-loop. Nonetheless, the linker connecting this Ω-loop to the TONDU domain has a negative contribution to the interaction with TEAD. It is difficult to determine whether the ancestral VGLL variant leading to VGLL2 had a similar linker or whether mutations have progressively transformed it from a shorter Vg-like linker to a longer one. In the second case, the transformation of the linker could be the first step in leading to the loss of the Ω-loop and VGLL2 could be an evolutionary intermediate between an ancestral VGLL protein and VGLL1/VGLL3.

The duplication of genes and their subsequent transformation by mutations is a well-
accepted concept in the literature, and the Vg/VGLL1-3 and Yki/YAP families provide a beautiful example of how structural and functional requirements have contributed to the evolution of two co-factor families that regulate the same transcription factors through interaction with an overlapping binding site. The loss of the β-strand:α-helix motif dramatically decreases the affinity of Vg/VGLL1-3 ("hot spot" loss) while in Yki/YAP it prevents the ability to compete with Vg/VGLL1-3 and reduces affinity. Therefore, this motif must be preserved in both families. The loss of the Ω-loop generates Vg/VGLL1-3 variants that are still able to bind to Sd/TEAD and to compete with Yki/YAP, while in the latter it abolishes binding ("hot spot" loss). Therefore, the Vg/VGLL1-3 proteins but not the Yki/YAP proteins can lose their Ω-loop.
These structural and functional constraints may explain why a β-strand:α-helix or a β-strand:αhelix:Ω-loop motif is present in the Sd/TEAD-binding domain of the Vg/VGLL1-3 proteins while, to our knowledge, only a β-strand:α-helix:Ω-loop motif is found in the Yki/YAP proteins.

Material and methods
Peptides. The synthetic peptides (both N-acetylated and C-amidated) were purchased from Biosynthan (Germany). The purity (>90%) and the chemical integrity of the peptides were determined by LC-MS from 10 mM stock solutions in 90:10 (v/v) DMSO:water. VGLL2 85-149 , which contains three cysteines (Fig 1B), was dissolved in 50:50 (v/v) acetontrile:(water+1mM TCEP) to avoid oxidation mechanisms induced by DMSO. VGLL2 85-108 was also prepared in the same solvent. The experiments conducted with VGLL2 85-149 were done with freshly made solutions (less than 72 h old). The peptide solutions were stored at -20°C and for each experiment the solutions were centrifuged to remove potential aggregates; the supernatant was dosed by HPLC and the integrity of the peptide was evaluated by LC-MS. To determine whether acetonitrile affects the output of the biochemical assays, Vg 288-337 was also dissolved in 50:50 (v/v) acetonitrile:(water+1mM TCEP). The Kd value of this peptide preparation -4.1 ± 0.1 nM (SPR) -is similar to that obtained with a DMSO solution of peptide (3.1 nM; Table 3) showing that the acetonitrile (2% final concentration) does not interfere with our assays.

Cloning, expression and purification of the proteins for the biochemical and biophysical assays.
The amino acid sequences of Scalloped (Sd, UniProt P30052, amino acids 223-440), Yorkie Structural biology. The DNA fragment encoding for wt Sd , obtained as described above, was PCR amplified and inserted into a pET28-derived vector providing an N-terminal His-HRV3Ctag by HiFi DNA Assembly Cloning (New England Biolabs, Ipswich, MA) according to the instructions of the manufacturer. The construct encoding for wt Sd was expressed in E. coli BL21(DE3) cells (Novagen, Madison, WI) without addition of biotin to the medium and purified according the protocol identical to the one described above for the biotinylated Sd proteins. Crystals of the complex between untagged wt Sd and Vg 298-337 were grown at 293 o K using the sitting drop vapour diffusion method. wt Sd (8.3 mg/ml) in 50 mM Tris pH 8.0, 250 mM NaCl, 2 mM MgCl2, 1 mM TCEP and 5% glycerol was pre-incubated with 0.5 mM Vg 298-337 (molar ratio Vg 298-337 / wt Sd ~1.5). For crystallization, the peptide-protein complex was mixed with an equal volume of the reservoir solution (0.3 µL + 0.3 µL). Initial crystals were obtained using 0.1 M Bis-Tris pH 6.5 and 25% PEG3350 as reservoir solution. Micro seeds were prepared from these conditions and used for further crystallization trials. Diffraction quality crystals were obtained using 0.2 M MgCl2,6H2O, 0.1 M Bis-Tris pH 6.5 and 25% PEG3350 as reservoir solution. Prior to shock cooling in liquid nitrogen, the crystals were soaked for a few seconds in reservoir solution containing 30% glycerol. X-ray diffraction data were collected at the Swiss Light Source (SLS, beamline X10SA) using an Eiger pixel detector.
Raw diffraction data from two crystals originating from the same crystallization drop were analysed and processed using the autoPROC (Vonrhein, Flensburg et al., 2011)  superimposed on this partial model. TEAD4 from this structure was deleted and the β-strand:αhelix region of VGLL1, VGLL1 27-51 , was mutated to mimic VGLL2 85-108 . Finally, an arbitrary conformation allowing the α-helix and the Ω-loop sequences to be connected in 3D was given to an amino acid stretch corresponding to the linker region VGLL2 109-136 . Molecular dynamics simulation of 10 ns with an explicit water solvation model was run using the Desmond module (default parameters) in the molecular modelling package Maestro (Schrödinger Inc. Cambridge, MA). The β-strand:α-helix and Ω-loop regions of VGLL2 were quite stable during the simulation, maintaining their secondary structure. At the end of the simulation, the linker region converged towards a loop conformation with no regular secondary structure that appeared stabilized by an aromatic stacking interaction between residues Tyr115 VGLL2 and Phe196 VGLL2 .

Data availability
The refined coordinates of the Vg:Sd complex structure have been deposited in the Protein Data Bank (www.wwpdb.org) under accession code 6Y20.