Structural mechanism for versatile cargo recognition by the yeast class V myosin Myo2

Class V myosins are actin-dependent motors, which recognize numerous cellular cargos mainly via the C-terminal globular tail domain (GTD). Myo2, a yeast class V myosin, can transport a broad range of organelles. However, little is known about the capacity of Myo2-GTD to recognize such a diverse array of cargos specifically at the molecular level. Here, we solved crystal structures of Myo2-GTD (at 1.9–3.1 Å resolutions) in complex with three cargo adaptor proteins: Smy1 (for polarization of secretory vesicles), Inp2 (for peroxisome transport), and Mmr1 (for mitochondria transport). The structures of Smy1- and Inp2-bound Myo2-GTD, along with site-directed mutagenesis experiments, revealed a binding site in subdomain-I having a hydrophobic groove with high flexibility enabling Myo2-GTD to accommodate different protein sequences. The Myo2-GTD-Mmr1 complex structure confirmed and complemented a previously identified mitochondrion/vacuole-specific binding region. Moreover, differences between the conformations and locations of cargo-binding sites identified here for Myo2 and those reported for mammalian MyoVA (MyoVA) suggest that class V myosins potentially have co-evolved with their specific cargos. Our structural and biochemical analysis not only uncovers a molecular mechanism that explains the diverse cargo recognition by Myo2-GTD, but also provides structural information useful for future functional studies of class V myosins in cargo transport.

Myo4. All of these are composed of an N-terminal motor domain followed by six IQ motifs, a central coiled-coil (CC) 3 region, and a C-terminal globular tail domain (GTD, also called cargo-binding domain or CBD) (1-3) (Fig. 1A). As unconventional myosins, class V myosins possess an amazing ability to recognize numerous cargos, such as various organelles, vesicles, protein complexes, and even mRNA (4 -6). The selection, recognition, and binding of those cargoes mainly rely on the GTDs of class V myosins (5,(7)(8)(9)(10).
Myo2 is the major player in the transport of organelles in budding yeasts (11). During the budding process, Myo2 directionally delivers organelles from the mother cell to the bud tip in an efficient way. The GTD of Myo2 (Myo2-GTD) was found to be responsible for recognizing a broad array of cargo adaptor proteins in S. cerevisiae (1). So far, a dozen cargo adaptors have been well characterized to directly interact with Myo2-GTD in cargo loading and transport, including mitochondria receptor Myo2 receptor-related protein 1 (Mmr1) (12,13), vacuole-related protein Vac17 (14,15), inheritance of peroxisomes gene 2 (Inp2) (16 -18), and also as Rab GTPase family proteins, Sec4, Ypt32, and Ypt11 (19 -21). Additionally, a kinesin-like protein Smy1 (suppressor of myosin) was shown to bind with Myo2-GTD and cooperate with Sec4 in the secretory vesicle transport (22)(23)(24)(25).
Previous mutagenesis studies suggested some key amino acids as potential cargo-binding sites for transporting mitochondria, vacuoles, and secretory vesicles (26) (Fig. 1B). However, for a long time only the apo-structure of Myo2-GTD was solved (10). Therefore the molecular mechanisms underlying the various cargo recognitions by Myo2-GTD remain elusive. Recently, several structures of MyoVa-GTD in complex with cargos were solved. Considering the significant differences in the GTD structures between MyoVa and Myo2 (Fig. 1B), and the fact that most cargo adaptors of Myo2 have no homologues in mammals, transferring the knowledge gained from the cargo-bound structures of mammalian MyoVa to yeast Myo2 would be difficult.
In this study, we determined crystal structures of Myo2-GTD in complex with three cargo adaptor proteins, Mmr1, Smy1, and Inp2. These complex structures reveal two independent binding sites in Myo2-GTD. One of the sites for interaction with Mmr1 was consistent with the previously reported mitochondrion/vacuole-specific binding region. Importantly, the other binding site identified for both Smy1 and Inp2 is a novel cargo-binding site in Myo2-GTD. This new site exhibits major structural differences between vertebrate MyoVa and Myo2, suggesting it may be specific to yeast myosin V. Moreover, the conformational flexibility of the Smy1/Inp2-binding site allows different cargo recognition in an overlapping site, providing a mechanistic explanation of the diversified cargo recognition by Myo2.

Modification of Myo2-GTD for crystallization
To explore Myo2-GTD-mediated cargo recognition, we systematically studied the interaction between GTD and reported binding partners. First, we successfully expressed and purified the C-terminal GTD of Myo2, comprising residues 1087-1574, in a bacterial expression system. To crystallize the GTD/cargo complexes, limited trypsin digestion was applied to treat the GTD sample as previously reported (10,27). However, the trace amount of residual protease was observed to digest the highly flexible binding targets (Fig. S1A), and thereby to disrupt the complex formation. To solve this problem, we designed a new GTD construct (GTD⌬L) by removing a flexible N-terminal loop (residues 1087-1151) and a central loop (residues 1342-1347), which contains a potential trypsin-cleavage site (27) (Fig.  1A). Such a modification did not impair the binding of Myo2 to its binding partner, Mmr1 (Fig. S1B), but promoted crystallization in many screen conditions without the trypsin treatment. We determined the Myo2-GTD⌬L structure at 1.9 Å resolution (Table 1). GTD⌬L adopts the same conformation as the previously reported apo-structure (10) with an overall r.m.s. deviation of 0.9 Å. Therefore, we used GTD⌬L for the following structural and biochemical characterizations.
The GTD⌬L structure contains two ␣-helical subdomains (Fig. 1B). In subdomain-I, the N terminus and C terminus are packed together by the interaction between ␣1 and ␣16. Such a conformation is largely different from the corresponding regions of MyoVa (Fig. 1B). Another significant structural difference has been found at the ␣6/␣7 region. Compared with MyoVa-GTD, the ␣6 and ␣7 helices in Myo2-GTD are much longer. These different structural elements in subdomain-I together with ␣4 and a long loop connecting subdomains-I and -II (loop-I/II) create a groove unique to Myo2-GTD ( Fig. 1B and  Fig. S2).

Smy1 occupies a hydrophobic groove in Myo2-GTD
As the binding site for Smy1 on Myo2-GTD is still unknown, we chose Smy1 as our first target for uncovering the cargo recognition mechanisms of GTD. Smy1 was first reported to be a kinesin-like protein, which suppresses a temperature-sensitive phenotype of cells carrying a motility-deficient mutation in Myo2 (22,28). By interacting with both actin and Myo2, Smy1 enhances the motility of Myo2 (29) and cooperates with Sec4 in transporting secretory vesicles (24). The Myo2-interacting site (MIS) in Smy1 was mapped onto its C-terminal region (25) (Fig.  2A).
Consistent with the previous result, the quantitative analysis using isothermal titration calorimetry (ITC) indicated that Smy1-MIS (residues 564 -656) interacts with Myo2-GTD⌬L with a binding affinity of ϳ3 M (Fig. S3). To understand how Myo2 can specifically recognize Smy1, we aimed to solve the Myo2-GTD⌬L/Smy1-MIS complex by protein crystallography. However, extensive crystallization trials failed to yield any complex crystals presumably due to the highly flexible nature of MIS. To overcome this problem, we split the MIS sequence into two halves based on the sequence analysis: a highly conserved C-terminal half (Smy1-MIS C , residues 615-650) and a less conserved N-terminal half (Smy1-MIS N , 564 -614). Because Smy1-MIS C showed a moderate binding affinity of ϳ20 M to GTD⌬L, whereas Smy1-MIS N did not bind with GTD⌬L ( Fig. 2 and Fig. S3), we used the Smy1-MIS C protein for the complex preparation with GTD⌬L. To avoid the dissociation between Smy1-MIS C and GTD⌬L during purification, we covalently linked the two fragments together as a chimera (Smy1-MIS C ::GTD⌬L) by fusing the Smy1-MIS C sequence to the N Crystal structures of the Myo2/cargo complex terminus of GTD⌬L connected by a tobacco etch virus (TEV) cutting site. Afterward, TEV protease was added to the purified sample during crystallization. By using this strategy, we successfully obtained complex crystals and solved the crystal structure of the Myo2-GTD⌬L/Smy1-MIS C complex at a resolution of 3.06 Å using the molecular replacement method (Table 1).
In the complex structure, the peptide (residues 633-647) containing 15 residues of Smy1-MIS C can be clearly assigned (Fig. S4A). The Smy1-MIS C peptide occupies the Myo2-unique groove in subdomain-I, burying ϳ800 Å 2 of solvent-accessible surface area (Fig. 2C), which was not previously recognized as a cargo-binding site. To tightly hold Smy1-MIS C , the groove adopts a hand-like conformation. In the complex structure, ␣6, ␣7, and a helix (␣A) between them together form a "thumb," ␣1 and ␣16 are "fingers," whereas ␣4 and the C-terminal part of the loop-I/II act as a "palm." The groove is largely hydrophobic with hydrophobic residues in the groove clustered into two patches. The first hydrophobic patch, mainly formed by aromatic residues from the thumb and palm (e.g. W1213 Myo2 , F1261 Myo2 , F1264 Myo2 , F1275 Myo2 , Y1287 Myo2 , F1542 Myo2 ), recognizes L637 Smy1 and L639 Smy1 in the N-terminal part of Smy1-MIS C (Fig. 2D). The second hydrophobic patch, formed by smaller residues from the fingers and the palm (e.g. I1207 Myo2 , V1545 Myo2 , V1565 Myo2 , A1566 Myo2 , V1569 Myo2 , V1570 Myo2 ), interacts with I642 Smy1 and V644 Smy1 in the C-terminal part of Smy1-MIS C (Fig. 2D).
Consistently, disruption of the hydrophobic interactions by mutating L637 Smy1 , L639 Smy1 , or I642 Smy1 to a charged residue (glutamic acid) abolishes the Myo2-GTD⌬L/Smy1-MIS C interaction (Fig. 2B). Likewise, the F1264E or F1275E mutation in the hydrophobic groove of Myo2 dramatically weakens the binding of Myo2-GTD⌬L to Smy1-MIS C ( Table 2). In contrast, mutations in the mitochondrion/vacuole-specific binding site, away from the Smy1-binding groove, show little effects on the binding (Table 2). Indeed, previous yeast two-hybrid assay of truncation mutants of Myo2-GTD indicated that deletion of the C-terminal ␣16 finger is sufficient to eliminate its binding to Smy1 (10). In addition to the hydrophobic interactions, the binding of Smy1-MIS C to GTD⌬L is strengthened by hydrogen bonds, which help fixing the backbone conformation of Smy1-MIS C and positioning the residues, such as L637 Smy1 , L639 Smy1 , and I642 Smy1 , with their side chains facing to the hydrophobic groove of Myo2-GTD⌬L (Fig. 2D). The interacting residues in either Smy1 or Myo2 are highly conserved across various budding species ( Fig. 2A and Fig. S2), suggesting that the Smy1mediated vesicle transport of Myo2 is very likely a common feature of budding yeasts.

Inp2 possesses a similar GTD binding mode to Smy1
Given the large number of known cargos for Myo2, it was important to probe whether the Smy1-binding groove in GTD could accommodate cargo proteins other than Smy1. Because the Myo2/Smy1 interaction requires the three highly conserved residues (Leu-637, Leu-639, and Ile-642) in MIS C (Fig. 2, A and B), a (L/F)XLXX(I/V/L) sequence pattern found in other cargo proteins, where X denotes any of the 20 common amino acids, may use the similar GTD-binding mode. To test our hypothesis, we first searched the sequence pattern from previously identified cargo proteins by using the SCANSITE server. Interestingly, Inp2, which is a peroxisome-specific receptor for Myo2 (17), was found to have the above sequence pattern in the middle of the peptide sequence (residues 531-543) (Fig. 3A). In line with our findings, this highly conserved region (Inp2-MIS) where I i is the intensity of the measured reflection and I m is the mean intensity of all symmetry related reflections. b R cryst ϭ ⌺ʈF obs ͉ Ϫ ͉F calc ʈ/⌺͉F obs ͉, where F obs and F calc are observed and calculated structure factors. R free ϭ ⌺ T ʈF obs ͉ Ϫ ͉F calc ʈ/⌺ T ͉F obs ͉, where T is a test data set of about 5% of the total reflections randomly chosen and set aside prior to refinement.  Next, ITC was employed to test the potential interaction between Inp2-MIS and Myo2-GTD⌬L. The results indicate that Inp2-MIS directly interacts with GTD⌬L with a relatively low binding affinity of ϳ30 M (Fig. 3A), and neither an extension at the N-terminal or C-terminal end of the Inp2 would significantly increase the binding affinity ( Fig. 3A and Fig. S6). By applying the similar strategy that we used for crystallization of the Myo2-GTD⌬L/Smy1-MIS C complex, we succeeded in solving the Myo2-GTD⌬L/Inp2-MIS complex structure at 2.85 Å resolution (Table 1).

Crystal structures of the Myo2/cargo complex
Fully consistent with our prediction, the Inp2-MIS peptide also binds to the hydrophobic groove in subdomain-I of Myo2-GTD (Fig. 3B). Although only nine residues (residues 532-540, Fig. S4B) were clearly assigned in the structure, the short Inp2-MIS peptide adopts a similar conformation to Smy1-MIS C (Fig.  3B). In detail, three highly conserved hydrophobic residues (Phe-534, Leu-536, and Leu-539) in Inp2-MIS reside in the same positions as the corresponding residues in Smy1-MIS C and form hydrophobic interactions with Myo2-GTD⌬L (Fig. 3, A and C). Besides, Ile-538 in Inp2-MIS is also involved in the hydrophobic interactions. Similarly to the Myo2/Smy1 interaction, several hydrogen bonds that are mainly formed between the side chains of Myo2-GTD⌬L and the main chains of Inp2-MIS further enhance the Myo2/Inp2 interaction. To validate our structure, because the Inp2-binding site was reported to locate at subdomain-II of Myo2-GTD (18), we measured the binding of the cargo proteins to the Smy1-MIS C ::GTD⌬L fusion protein, in which Smy1-MIS C is covalently linked to the N terminus of GTD⌬L and presumably covers the Smy1/Inp2binding groove. Our results indicate that neither Inp2-MIS nor Smy1-MIS C show a detectable binding to the fusion protein (Table 2), thus confirming our structural findings.

Mmr1 interacts with the mitochondrion/vacuole-specific binding site of GTD via both hydrophobic and charge-charge interactions
Genetic studies indicate that Myo2 is required for Mmr1mediated mitochondrial distribution (12,13,26,30,31). Additionally, using yeast two-hybrid and co-immunoprecipitation assays, Eves et al. (26) revealed that a region containing residues 398 -430 in Mmr1 is necessary and sufficient for the binding of Mmr1 to Myo2-GTD (Fig. 4A). By performing analytical gel filtration and ITC-based binding assays, we confirmed that this region of Mmr1 and Myo2-GTD forms a 1:1 complex with binding affinity of ϳ7 M (Figs. S1B and S7). Moreover, we found that including an additional ϳ40 -50 residues in the N terminus of the reported Myo2-interacting site increases the binding affinity ϳ3-fold (Fig. 4, A and B). Thus, we named these N-terminal extensions and the original boundaries as Mmr1-MIS (residues 349 -430), Mmr1-MISЈ (residues 360 -430), and Mmr1-MIS C (residues 398 -430), respectively. The three Mmr1 boundaries were screened for crystallization with Myo2-GTD⌬L and only the Mmr1-MIS C and GTD⌬L mixture could be crystallized. The structure of the Myo2-GTD⌬L/Mmr1-MIS C complex was determined at 2.73 Å resolution (Table 1). Each asymmetric unit contains two essentially identical complexes. To our surprise, in one complex, two Mmr1-MIS C molecules were found to interact with GTD⌬L at two different sites ( Fig. 4C and Fig. S4, C and D).
As shown in Fig. 4C, one Mmr1-MIS C molecule (yellow) adopts a conformation with an N-terminal loop (N-loop) and a C-terminal ␣-helix (␣ Mmr1 ). The binding interface of this

Crystal structures of the Myo2/cargo complex
Mmr1-MIS C molecule on Myo-GTD agrees very well with the mitochondrion/vacuole-specific binding site, as it was mapped in previous mutagenesis studies (7, 10, 14, 15, 26). More specifically, the interface includes a narrow groove between the two parallel helices, ␣5 and ␣7, for the binding of the N-loop, and a pocket formed by ␣5, ␣7, and the loop-I/II for the binding of ␣ Mmr1 (Fig. 4D and Fig. S2). Compared with the Smy1/Inp2binding site that is predominantly hydrophobic, this interface

Crystal structures of the Myo2/cargo complex
on GTD⌬L is much more hydrophilic. Thus, besides hydrophobic interactions, the binding of the first Mmr1-MIS C molecule to GTD⌬L largely depends on polar interactions. For example, N1304 Myo2 forms two strong hydrogen bonds with main chain atoms in the N-loop of Mmr1-MIS C , thereby stabilizing the conformation of R412 Mmr1 for the cation-interaction with Y1303 Myo2 (Fig. 4D). Interestingly, the ␣5/␣7 groove and the ␣ Mmr1 -binding pocket show reverse charge potentials, negative for the former and positive for the latter (Fig. 4E). Such a charge distribution on Myo2-GTD allows the specific recognition of corresponding charged residues in Mmr1-MIS C by forming two charge pairs, R409 Mmr1 -E1293 Myo2 and E422 Mmr1 -K1311 Myo2 (Fig. 4, D and E). Considering the ␣ Mmr1 -helix as a dipole, the placement of the negatively charged C-terminal end of ␣ Mmr1 near the positively charged pocket of Myo2-GTD further enhances the charge-charge interaction. Together, this charge-mediated guidance is likely to play a role in orientating the Mmr1-MIS C peptide to the proper position for the GTD binding. Consistent with our structural finding that both hydrophobic and charge-charge interactions are required for the Myo2/Mmr1 interaction, replacing the positively charged residue R409 Mmr1 or the hydrophobic residue L410 Mmr1 with a glutamic acid abolishes the binding of Mmr1-MIS to Myo2-GTD⌬L (Fig. 4B).
Notably, despite the fact that E1293 Myo2 and D1297 Myo2 form salt bridges and hydrogen bonds with R409 Mmr1 and T408 Mmr1 (Fig. 4D), neither a E1293K nor D1297N mutant of Myo2 lost the ability to bind with Mmr1 (26). Because E1293 Myo2 and R409 Mmr1 are highly solvent exposed, the salt bridges formed between them are unlikely very stable, as indicated by the weak density of R409 Mmr1 (Fig. S4C). Furthermore, the charge potential change caused by the E1293K or D1297N mutation may be partially compensated by nearby negatively charged residues, like E1222 Myo2 and D1297 Myo2 . Therefore, these two mutations may impair but do not eliminate the binding of Myo2 to Mmr1. Indeed, the D1297N mutant interacts with Mmr1-MIS C albeit 3-fold decreased binding affinity (Table 2).
Unexpectedly, the second Mmr1-MIS C molecule (Fig. 4C, light pink) binds with the Smy1/Inp2-binding groove by forming a long loop (Fig. 4, A and C, and Fig. S8A). Although this Mmr1-MIS C molecule adopts a different conformation from Smy1-MIS C and Inp2-MIS (Fig. S8B), two hydrophobic residues, L410 Mmr1 and V413 Mmr1 , occupy a similar position to L639 Smy1 and I642 Smy1 allowing an interaction with the two hydrophobic patches, respectively (Fig. S8A). Furthermore, such a conformation places R409 Mmr1 in a position to form salt bridges with E1211 Myo2 (Fig. S8A).
The structural finding of two Mmr1-MIS C molecules interacting with one GTD⌬L molecule contradicts the 1:1 binding stoichiometry indicated by the biochemical data (Figs. S1B and S7). To test whether the Smy1/Inp2-binding groove in Myo2-GTD participates in Mmr1 binding, we measured the binding of the Smy1-MIS C ::GTD⌬L fusion protein, and the Myo2-GTD⌬L E1211A and F1275E mutants to Mmr1-MIS C . In contrast to the above three designed disruptions in the Smy1/Inp2binding groove, another four mutants on the mitochondrion/ vacuole-specific binding site (D1297N, Y1303A, K1311E, and K1312A) show negligible effects on the Myo2-GTD/Mmr1-MIS C interaction ( Table 2), indicating that the observation of the second Mmr1-MIS C molecule in the complex is likely a crystallization artifact. Nevertheless, this artifact implies that the Smy1/Inp2-binding groove has the potential to recognize cargo with different binding modes. As Mmr1 was found to compete with the vacuole-specific receptor Vac17 for binding to Myo2 (26), we tested the binding of Vac17-MIS to Myo2-GTD⌬L and its mutants (Fig. S9). The Myo2 mutations on the mitochondrion/vacuole-specific binding site disrupting the Mmr1 binding were also found to interfere with the Vac17 binding. On the other hand the mutations in the Smy1/ Inp2-binding site show little impact on the binding affinities ( Table 2).
The considerably higher binding affinities of Mmr1-MIS or Mmr1-MISЈ and Myo2-GTD⌬L indicate that the N-terminal sequence (residues 360 -397, MIS N ) facilitates the binding of Mmr1 to Myo2-GTD⌬L. Because the N terminus of the first Mmr1-MIS C molecule in the complex extends toward the Smy1/Inp2-binding groove (Fig. 4C), we speculated that Mmr1-MIS N interacts with the groove. Consistently, both the Smy1-MIS C ::GTD⌬L fusion and the E1211A mutation weaken the GTD binding to Mmr1-MISЈ by 2-3-fold (Fig. S7). These results suggest that Mmr1-MIS may employ the two sites to boost its binding to Myo2-GTD.

Conformational changes of the Smy1/Inp2-binding groove in Myo2-GTD for cargo binding
The overall structures of cargo-bound forms of Myo2-GTD are highly similar with that of the apo-form, except for the Smy1/Inp2-binding groove. Binding-induced conformational changes are observed in ␣1/␣16 (the fingers), the loop-I/II, and the ␣6/␣7 region (the thumb) (Fig. 5A). Specifically, the unstructured region that connects ␣6 and ␣7 forms a short ␣A-helix, due at least in part to the involvement of Phe-1275 in the formation of the hydrophobic patch upon cargo binding (Figs. 2D and 3C, and Fig. S8B). To understand the thermal motion of the Smy1/Inp2-binding groove, the distributions of B-factors (atomic displacement parameters) in the Myo2-GTD⌬L structures of the cargo-free and cargo-bound forms Crystal structures of the Myo2/cargo complex were investigated. In general, residues in peripheral loop regions usually have higher B-factors or higher degrees of motion. However, compared with the majority of the helices, the whole Smy1/Inp2-binding groove in the apo-structure shows much stronger motion (Fig. 5B, upper panel), suggesting that the groove is highly flexible without bound cargos. Interestingly, the motion of the groove is remarkably decreased in the Smy1-bound structure (Fig. 5B, lower panel). Taken together, we propose that high flexibility is a genuine property of the Smy1/Inp2-binding groove, which can accommodate various flexible sequences in the cargo proteins to achieve the versatile binding.
An overlap of binding sites between different cargos was also found in mammalian MyoVa-GTD (9,32). For example, the cargo adaptors Melanophilin and Spir interact with MyoVa-GTD on an identical surface (Fig. 6A). Compared with the Smy1/Inp2-binding site in Myo2-GTD, the Melanophilin/Spirbinding site in MyoVa-GTD shows little conformational flexibility (32). Although two different cargo-binding sites have also been identified in subdomain-I of MyoVa-GTD, they overlap only to a small extent with the cargo-binding sites of Myo2-GTD (Fig. 6, A and B). For example, the Smy1/Inp2-binding groove does not exist in MyoVa, whereas the small helix in MyoVa-GTD that is a RILPL2-binding site is replaced by a flexible loop in Myo2-GTD (Fig. 1B). These differences between MyoVa and Myo2 suggest that class V myosins in mammals and yeasts have evolved in very different ways to recognize their own cargos. Because the cargo adaptors of MyoVa and Myo2 share little similarity in amino acid sequences, the different cargo recognition mechanisms adopted by MyoVa and Myo2 are likely to be selected by their cargos. Nevertheless, the identified cargo-binding sites in MyoVa and Myo2 are highly conserved within vertebrates (9) and budding yeasts (Fig. 6C), respectively.

Discussion
Class V myosins recognize numerous cargos mainly via direct bindings of motor's GTD to cargos or cargo-specific adaptors with diverse amino acid sequences and broad physiological functions. Despite many years of extensive studies, more efforts are still required to decode the diversified yet specific recognitions by GTD. Recent progress was made mostly on mammalian class V myosins (9,32,33) revealing several cargobinding modes by GTD. However, because the low sequence and structural similarity between GTDs of mammalian MyoVs and yeast Myo2 (9), those discoveries provide very limited knowledge for yeast studies. The series of structures of Myo2-GTD in complex with different cargo proteins described here advance our understanding of Myo2 function. The cargo-binding sites on Myo2-GTD described here do not only confirm previous biochemical and functional studies, but also provide key information for future in vivo experiments dissecting the Myo2/cargo interactions.
The structures of Myo2 in complex with Smy1 and Inp2 uncover a novel cargo-binding groove in GTD. Our structural and biochemical analysis indicates that the Smy1/Inp2-binding groove recognizes the sequence motif through conserved hydrophobic residues arranged in a specific pattern (Fig. 3A). The structural comparison of cargo-bound Myo2-GTD with the apo-form reveals that the Smy1/Inp2-binding groove is relatively flexible. Compared with a rigid binding surface, such a highly flexible conformation exhibiting makes the groove a versatile binding site for the MIS sequences in cargo proteins, like Smy1 and Inp2, which are largely flexible as well. Like a human hand, the flexible thumb (␣6, ␣A, and ␣7), palm (␣4 and the loop-I/II), and fingers (␣1 and ␣16) work together to grasp various binding targets. Interestingly, the vesicle-binding site in subdomain-II of Myo2-GTD, which is also a versatile binding site for Ypt31/32, Sec4, and Kar9 (26), is surrounded by flexible helices and loops (Figs. 1B and 5B). It may thus be speculated that the high flexibility of the vesicle-binding site contributes to the different cargo recognitions by Myo2-GTD.
To maintain the high efficiency of Myo2-mediated cargo transport, it is very important to keep cargos associated with Crystal structures of the Myo2/cargo complex the motor stably. This raises a question about the relatively low affinities measured in this work between Myo2-GTD and the MIS sequences of Smy1, Inp2, and Mmr1 (Table 2). A plausible explanation might be that these cargo adaptors contain multiple Myo2-binding sites. Consistently with this explanation, the central coiled-coil region of Smy1 was also reported to bind with Myo2 (24). Although the Inp2-binding site was mapped onto subdomain-I of Myo2-GTD in our work (Fig. 3B), surface residues in subdomain-II were also found to be involved in the binding of Myo2 to Inp2 (18). As the boundary used in a previous study covers a much larger region than Inp2-MIS (18), it is very likely that Inp2 interacts with Myo2-GTD by using at least two binding sites in subdomains-I and -II, respectively. Likewise, Mmr1 may also contain two Myo2-binding segments, Mmr1-MIS N and -MIS C , to interact, respectively, with the Smy1/Inp2-binding site and the mitochondrion/vacuole-binding site in Myo2-GTD as suggested by our findings (Fig. 4 and Fig. S7). In addition to strengthening the motor/cargo interaction, another advantage for employing a multi-binding site is to provide better regulations, beneficial for delicate control during cargo loading and unloading. For example, the binding of GTD to its cargos helps to release the head-to-tail autoinhibition (37). Because subdomain-I has been reported to participate in the head-to-tail interaction (32,33), the cargo-binding modes identified in this study have the potential for the activation of Myo2.
We noted with interest that despite barely showing sequence similarity, the cargo adaptors including Smy1, Inp2, Mmr1, and Vac17 all contain at least one CC domain (Figs. 2A, 3A, and 4A, and Fig. S9). Similarly, several cargo adaptors for MyoVa, such as Melanophilin and RILPL2, also contain CCs in their sequences (9,34). Class V myosins are known to form a dimer via their long coiled-coil domain located after the IQ repeat motifs (35,36). Moreover, the CC domain of Melanophilins has been demonstrated to be essential for MyoVa-mediated melanosome transport in melanocytes (34). Given the possibility of CC-mediated self-association, the potential oligomerization of the cargo adaptors is likely to enhance their binding to the dimerized Myo2, and thus to increase motor's transport activity. Consistently, the dimerization of She3, an adaptor for Myo4, was found to synergistically boost the binding of She3 to Myo4 (38). A recent study demonstrated that the CC domain of Mmr1 was sufficient for self-interaction and contributed to the function in mitochondrial inheritance (31).

DNA constructs and site-directed mutagenesis
DNA encoding the GTD domain of Myo2 and the fragments of cargo adaptors were PCR-amplified from the yeast cDNA library. The coding sequences were cloned into a home-modified vector pET32a with a N-terminal thioredoxin-His 6 tag for protein expression and purification. All point mutations were introduced using a site-directed mutagenesis kit and confirmed by DNA sequencing.

Protein sample preparation
For the expression of the protein samples, transformed Escherichia coli C Plus (Novagen) were grown in LB medium at 37°C to an optical density with A 600 nm of ϳ0.8 -1.0, then induced by the addition of isopropyl 1-thio-␤-D-galactopyranoside with a final concentration of 0.2 mM, and were further grown for an additional 14 h at 16°C. The proteins were purified using Ni-affinity chromatography followed by size-exclusion chromatography with a HiLoad Superdex 200-pg 16/600 column (GE Healthcare).

ITC assay
ITC measurements were carried out on a VP-ITC Microcal calorimeter (Malvern) at 25°C. All protein samples were dissolved in the general buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA, and 2 mM DTT). Titrations were performed by injecting a protein solution containing a fragment from a cargo adaptor or its mutants at a concentration of 400 M into a protein solution containing Myo2-GTD⌬L or its mutants at a concentration of 40 M. A time interval of 2 min between injections was used to ensure that the titration peak returned to the baseline. The titration data were analyzed using the program Origin7.0 and fitted by a one-site binding model.

Analytical gel filtration chromatography
Analytical gel filtration chromatography was carried out on an ÄKTA pure system (GE Healthcare). Protein samples at a concentration of 40 M were loaded onto a Superdex 200 Increase 10/300 GL column (GE Healthcare), equilibrated with the general buffer.

Crystallization
To prepare the Myo2-GTD⌬L/Smy1-MIS C or Myo2-GTD⌬L/Inp2-MIS complexes, Smy1-MIS C (residues 615-650) and Inp2-MIS (531-543) were fused to the N terminus and C terminus of Myo2-GTD⌬L, respectively, between a (GS) 5 linker and TEV cleaved site (ENLYFQS). The thioredoxin-His 6 tags were cleaved by HRV 3C protease at 16°C overnight and removed by size-exclusion chromatography. The fusion proteins were concentrated to ϳ30 mg/ml. Prior to crystallization, TEV protease was added into the protein solutions at 4°C overnight to cleave the fusion linkers between the cargo peptides and Myo2-GTD. To prepare the Myo2-GTD⌬L/Mmr1-MIS C complex, Myo2-GTD⌬L and Mmr1-MIS C were purified separately and then mixed in a 1:3 molar ratio to a final concentration of ϳ20 mg/ml. Crystallization trials were set up using the sitting drop vapor diffusion method at 16°C. Crystals were obtained in the conditions with 1 M potassium chloride, 1 M ammonium sulfate, and 0.1 M HEPES, pH 7.0, for the Myo2-GTD⌬L apo-form; 8% (w/v) PEG8000 and 0.1 M sodium dihydrogen phosphate, pH 6.5, for the Smy1-MIS C bound form; 8% (w/v) PEG4000 and 0.2-0.3 M lithium/ammonium sulfate, pH 7.5, for Inp2-MIS bound form; and 20% (w/v) PEG3350 and 0.2 M sodium acetate for the Mmr1-MIS C bound form. Before diffraction, the crystals were soaked in crystallization solution containing an additional 30% (v/v) glycerol for cryoprotection.

Structure determination and analysis
The X-ray diffraction datasets were collected at Shanghai Synchrotron Radiation Facility beamlines BL17U, BL18U, and BL19U1. The diffraction data were processed and scaled using Crystal structures of the Myo2/cargo complex HKL2000 (39). The initial phase of the complex structure was determined by molecular replacement using the Myo2-GTD apo-structure (PDB code 2F6H) as the search model. The Mmr1 peptide was further built into the model and refined in PHENIX (40). COOT was used for model rebuilding and adjustments (41). In the final stage, an additional TLS refinement was performed in PHENIX. The model quality was checked using MolProbity (42). The model of Myo2-GTD⌬L/ Smy1 and Myo2-GTD⌬L/Inp2 were built and refined using the same strategy as that used for the Myo2-GTD⌬L/Mmr1 complex structure. Data collection and refinement statistics can be found in Table 1. The structures were deposited in the PDB with accession codes 6IXO (apo-form), 6IXQ (Smy1-MIS C bound), 6IXR (Inp2-MIS bound), and 6IXP (Mmr1-MIS C bound). The binding interface in the complex structures was analyzed using PISA (43). The conservation surface was generated using the Consurf server (http://consurf.tau.ac.il/2016/). 4 All structural figures were created using the PyMOL Molecular Graphics System.