Structure and comparison of the motor domain of centromere-associated protein E

Crystallization and structure determination of the motor domain of centromere-associated protein E in complex with its inhibitor was performed. In the determined structure, endogenous ADP was observed in the nucleotide-binding site instead of the inhibitor.


Introduction
Antimitotic anticancer drugs, such as taxanes and vinca alkaloids, have been widely used in the clinical therapy of human malignancies (Wood et al., 2001;Jordan & Wilson, 2004). They cause serious side effects such as toxicity in nondividing cells such as peripheral neurons. On the other hand, antimitotic agents that target mitotic kinesins are expected to be more likely to act on dividing cells but not on nondividing cells, and thus antimitotic agents that inhibit the functions of kinesin motor domains minimize the toxicity to nondividing cells, causing decreased side effects. Therefore, such inhibitors will be promising candidates for the development of cancer drugs (Sakowicz et al., 2004). To date, a large number of Eg5 inhibitors, such as STLC and PVZB1194, have been reported (Ogo et al., 2007;Matsuno et al., 2009). A large amount of structural information on Eg5 in complex with its inhibitors has also been reported (Yokoyama et al., 2015(Yokoyama et al., , 2018Myers & Collins, 2016). However, Eg5 inhibitors have not found clinical use because they target not only cancer cells but also other actively proliferating cells.
A recent new target for mitotic inhibition is centromereassociated protein E (CENP-E), which is a member of the kinesin-7 subfamily. CENP-E plays important roles in proper chromosome segregation during mitosis. The role of CENP-E in chromosome congression has been extensively described in various models (McEwen et al., 2001;Putkey et al., 2002;Schaar et al., 1997;Wood et al., 1997;Yao et al., 2000). After ISSN 2059-7983 entering mitosis, CENP-E locates on the kinetochores during spindle formation, and is required for the proper alignment of mitotic chromosomes at the spindle midzone (Wood et al., 1997;Schaar et al., 1997;Kim et al., 2010). The motor domain of CENP-E plays an important role in transporting peripheral polar chromosomes towards the spindle midzone (Barisic et al., 2014;Zhang et al., 2017). CENP-E is composed of three domains: tail, stalk and motor. The motor domain, located at the N-terminus, consists of 339 residues and is approximately 40 kDa in size; it is the active site for ATPase activity. CENP-E uses the N-terminal motor domain to gain driving energy by hydrolyzing adenosine triphosphate (ATP) to adenosine diphosphate (ADP), and moves along microtubules. The release of ADP from CENP-E is significantly slower than from other kinesins and is the rate-limiting step in ATP turnover (Sardar & Gilbert, 2012).
To date, only the crystal structure of the CENP-E motor domain in complex with MgADP (CENP-E-MgADP 1t5c) has been reported (Garcia-Saez et al., 2004;PDB entry 1t5c). It is difficult to perform rational drug design by fragmentbased drug discovery (FBDD) or structure-based drug design (SBDD) owing to a lack of structural information on CENP-E. Therefore, it is necessary to determine crystal structures of the CENP-E motor domain in complex with its inhibitors.
Here, in order to elucidate the mechanism by which the CENP-E motor domain binds to its inhibitors, we tried to cocrystallize the CENP-E motor domain in complex with the ligand 3-chloro-4-isopropoxyl benzoic acid (CIBA; Qian et al., 2010), an ATP-competitive inhibitor, and determined the structure at 1.9 Å resolution. Endogenous ADP instead of CIBA was observed in the nucleotide-binding site, even though ATP or ADP had not been added. The determined structure of the CENP-E motor domain was compared with those of other kinesin motors. Based on the characteristic structure of CENP-E, the mechanism by which ADP is retained in CENP-E is discussed.

Preparation of inhibitor and construction of plasmids
CIBA was synthesized as described by Qian et al. (2010). The cDNA of CENP-E 1-339 (residues 1-339 of CENP-E) was cloned into pCold III bacterial expression vector to construct pCENP-E 1-339 , similarly to as described by Yamane et al. (2019). The recombinant protein consisted of the CENP-E motor domain (Met1-Ser339) extended with MNHKVH at the N-terminus and GSHHHHHH at the C-terminus.
The recombinant protein was purified in three steps involving nickel-affinity, cation-exchange and gel-filtration chromatography. The harvested cells were resuspended in buffer consisting of 50 mM Tris-HCl pH 7.5, 0.5 M NaCl, 2 mM MgCl 2 , 0.2 mM EGTA, 5 mM -mercaptoethanol, 25 mM imidazole, 10%(w/v) sucrose and protease-inhibitor cocktail (Roche) and adjusted to pH 7.4, and were disrupted by sonication. The cell lysate was centrifuged and the supernatant was loaded onto 1 ml Ni-NTA agarose resin (Qiagen) equilibrated with buffer consisting of 20 mM Tris-HCl, 0.3 M NaCl, 2 mM MgCl 2 , 5 mM -mercaptoethanol, 20 mM imidazole, 10%(w/v) sucrose and adjusted to pH 7.4. After washing with buffer containing 25 mM imidazole, the proteins were eluted with buffer consisting of 500 mM imidazole, 50 mM piperazine-1,4bis(2-ethanesulfonic acid) (PIPES)-NaOH, 0.1 M NaCl, 2 mM MgCl 2 , 5 mM -mercaptoethanol, 10%(w/v) sucrose and adjusted to pH 6.8. The eluted protein was loaded onto a 1 ml HiTrap SP HP cation-exchange column (GE Healthcare) equilibrated with buffer consisting of 50 mM PIPES-NaOH pH 6.8, 2 mM MgCl 2 , 1 mM EGTA, 1 mM tris(2-carboxy-  i jI i ðhklÞ À hIðhklÞij= P hkl P i I i ðhklÞ, where I i (hkl) is the intensity of an individual reflection and hI(hkl)i is the mean intensity of that reflection. ‡ R = P hkl jF obs j À jF calc j = P hkl jF obs j, where |F obs | and |F calc | are the observed and calculated structure-factor amplitudes, respectively. R free is calculated for 10% of the reflections that were randomly excluded from refinement. ethyl)phosphine (TCEP), 5%(w/v) sucrose and adjusted to pH 6.8, and was eluted with a linear gradient of 0-0.4 M NaCl. The eluted fractions were further purified by gel-filtration chromatography using a HiLoad 16/600 Superdex 200 prep-grade column equilibrated with buffer consisting of 50 mM PIPES-NaOH pH 6.8, 2 mM MgCl 2 , 1 mM EGTA, 1 mM TCEP, 5%(w/v) sucrose, 0.3 M NaCl and adjusted to pH 6.8. The eluted proteins were concentrated with a Vivaspin 20 centrifugal concentrator (Sartorius) with a 10 kDa molecular-mass cutoff. The concentration of CENP-E was determined with a NanoDrop One (Thermo Scientific) using an extinction coefficient of 3.186 Â 10 4 M À1 cm À1 . The purity of the CENP-E protein during the purification procedure was confirmed by SDS-PAGE analysis.

Crystallization
The purified protein was at 11 mg ml À1 in 50 mM PIPES-NaOH pH 6.8, 300 mM NaCl, 2 mM MgCl 2 , 1 mM EGTA, 1 mM TCEP, 5%(w/v) sucrose. Crystallization was performed using the hanging-drop vapor-diffusion method at 4 C. After approximately ten days, imperfect crystals appeared. It has been reported that good crystals can be obtained using the microseed matrix seeding method (D'Arcy et al., 2014). The 11 mg ml À1 protein solution was mixed with CIBA in a molar ratio of 1:10 (at least 277 mM CENP-E 1-339 and 2.77 mM CIBA). Crystallization was performed using the sitting-drop vapor-diffusion method at 4 C. Crystallization drops were prepared by mixing 0.9 ml of the CENP-E 1-339 -CIBA solution described above, 0.8 ml reservoir solution and 0.3 ml seed solution. The seed solution was prepared using the reservoir solution consisting of 90 mM Tris-HCl pH 7.5, 18%(w/v) PEG 3350. Hexahedron-shaped crystals appeared with approximate dimensions of 0.1 Â 0.2 Â 0.05 mm.

Structure determination
We tried to determine the structure of the CENP-E motor domain in complex with its inhibitor CIBA. The structure was determined at 1.9 Å resolution, which was higher than that of the previously reported structure CENP-E-MgADP 1t5c (Garcia-Saez et    al., 2004; PDB entry 1t5c; 2.5 Å resolution; Fig. 1a). Unfortunately, electron density for ADP instead of CIBA was observed in the nucleotide-binding site, although ADP had not been added during protein preparation (Fig. 1b). The determined structure was of the CENP-E motor domain in complex with MgADP. There are two molecules (chains A and B) in the asymmetric unit. The structure from this study is almost identical to the previously reported structure CENP-E-MgADP 1t5c (Garcia-Saez et al., 2004; PDB entry 1t5c), with a root-mean-square deviation (r.m.s.d.) for corresponding C atoms of 0.71 Å . Gel-filtration analysis has suggested that the CENP-E motor domain is monomeric (Garcia-Saez et al., 2004). Hereafter, molecule A is used to discuss the structure of the CENP-E motor domain.
Molecule A includes residues Glu4-Asn17, Ala27-Asn159, Asn161-Tyr191, Asn197-Lys216, Gly224-Ala243 and Leu252-Ser339 and MgADP. Molecule B comprises residues Glu4-Ser18, Ala27-Tyr191, Gln198-Lys216, Ser225-Ala243, Leu252-Gln276 and Phe280-Ser339 and MgADP. The C atoms of 301 residues in the two monomers were superposed by a least-squares fit using PDBeFold and their final r.m.s.d. was 0.28 Å . The average B factor of the protein was relatively high compared with the Wilson B factor (Table 1). This may be because the structure contains a large number of disordered and missing residues. Fig. 1(a) shows a front view of the CENP-E-MgADP structure reported in this study. It has a mixed eight-stranded -sheet core with flanking solvent-exposed -helices and a small three-stranded antiparallel -sheet in the N-terminal region (1a, 1b and 1c). The long linker region of 9 and 10 ( Fig. 1a) has the same docked conformation as described in the previously reported structure of the CENP-E motor domain (Garcia-Saez et al., 2004).

Structural comparison with known structures
The structure of chain A of CENP-E-MgADP reported in this study was compared with the previously determined structures of CENP-E-MgADP (Garcia-Saez et al., 2004;PDB entry 1t5c), the motor domain of Eg5 in complex with MgADP (Eg5-MgADP; Turner et al., 2001; PDB entry 1ii6) and Eg5-AMPPNP (Parke et al., 2010; PDB entry 3hqd) (Fig. 2). The CENP-E-MgADP structure determined at 1.9 Å resolution in this study will provide significantly more structural information than the previously determined CENP-E-MgADP 1t5c structure at 2.5 Å resolution. Several similar and differing features of the two structures are described. The structure in this study differed slightly from that of CENP-E-MgADP 1t5c. Loop L1 has been claimed to be unique to CENP-E (Garcia-Saez et al., 2004). The 0 helix is a conserved structure in kinesins such as Eg5. The structure in this study is nearly identical to CENP-E-MgADP 1t5c but these residues are disordered (Figs. 2 and 4a). The region containing 0 and L1 of CENP-E seems to be flexible, which indicates that this structure is unique to the CENP-E motor domain.
The structure at the beginning of L2 (residues 41-45 of chain A) differs from those in Eg5-MgADP and Eg5-MgAMPPNP but is nearly identical to that in CENP-E-MgADP 1t5c (Figs. 2 and 4a). L2 in CENP-E-MgADP is smaller than those in Eg5-MgADP and Eg5-AMPPNP. The loop itself has a double conformation. Thus, each of the residues Asp34-Asn36 of L2 in CENP-E-MgADP from this study has a double conformation. His54 of L3 in CENP-MgADP from this study has a double conformation.
The orientation of the side chain of His111 in the middle of 2 is the same as that in CENP-E-MgADP 1t5c but differs from those in Eg5-MgADP and Eg5-AMPPNP. Therefore, it is unique to CENP-E. The B factors of the main chain and side chain of His111 are below 40 Å 2 (Fig. 3). Plots of average B factors of each residue of chain A of CENP-E-MgADP from this study. Residue numbers are given in parentheses.
helix is similar to that in other kinesins such as Eg5 (Figs. 2b, 2c and 4b). The end of 2 of CENP-E-MgADP from this study is away from the eight-stranded -sheet core. Although the number of residues in 2 is the same, the large r.m.s.d. value indicates that the -helix is slightly shrunk compared with those of Eg5-MgADP and Eg5-MgAMPPNP. The r.m.s.d. for CENP-E-MgADP 1t5c is up to 0.6 Å , indicating that the structure of 2 is unique to CENP-E (Fig. 2a). The structure of L5, which is located in the middle of 2 in kinesin motor domains and is involved in the binding of inhibitor by Eg5, is nearly identical to that of CENP-E-MgADP 1t5c and is smaller than those of Eg5-MgADP and Eg5-AMPPNP, as described in the previous report (Garcia-Saez et al., 2004;Fig. 4b). L5 of Eg5 regulates both nucleotide and microtubule binding through a set of reversible interactions with 3 ( Muretta et al., 2013). His102 of CENP-E-MgADP from this study also has a double conformation.
Helix 3 is relatively similar in the structures of CENP-E-MgADP 1t5c and . The orientation of 3 is similar to that in CENP-E-MgADP 1t5c, but the positions of both ends of 3 are closer to the -sheet core and are intermediate between the previously reported structures of Eg5-MgADP and Eg5-MgAMPPNP. Helix 3 of CENP-E-MgADP from this study has a larger number of residues than that of Eg5-MgAMPPNP.
L9 and L11, which correspond to switch I and II, respectively, of CENP-E-MgADP from this study, show high B factors (Fig. 3) and were not modeled. Owing to high flexibility, the structure differs from those of CENP-E-MgADP 1t5c, Eg5-MgADP and Eg5-MgAMPPNP. The r.m.s.d. values for the region 199-203 between the structure of CENP-E-MgADP from this study and Eg5-MgAMPPNP are high (Fig. 2c) Structural comparison with other structures. The structures of CENP-E-MgADP from this study (green), CENP-E-MgADP 1t5c (cyan), Eg5-MgADP (PDB entry 1ii6; magenta) and Eg5-MgAMPPNP (PDB entry 3hqd; yellow) and Mg 2+ (white) are shown. (a) Ribbon representations of the structure of the CENP-E-MgADP from this study superposed with the previously reported structures of CENP-E-MgADP 1t5c, Eg5-MgADP (PDB entry 1ii6) and Eg5-MgAMPPNP (PDB entry 3hqd). The view is the same as in Fig. 1(a). (b) CENP-E has a unique structural orientation of L5 and helix 3 compared with ADP-bound and ATP-bound forms of Eg5 (stereoview). (c) Stick representations of superposed residues 299-301 of CENP-E-MgADP from this study and CENP-E-MgADP 1t5c. that this region is involved in the nucleotide-binding site. The structure of the end of L11 was also unique to other kinesin structures. The average B factors of residues 252-258 (L11) were high, so these regions were flexible (Fig. 3). However, the structure of residues 252-254, which interacted with the symmetry-related molecule, was located in a slightly different position compared with CENP-E-MgADP 1t5c. This is a possible reason why the B factor of residue 253 is relatively low (Fig. 3).
The structure of L13 between 5 and 8 in this study is nearly identical to that in CENP-E-MgADP 1t5c, but not to that in Eg5-MgADP (Fig. 2), indicating that the main chain of residues 287-301 in this long region was unique to the CENP-E motor domain.
CENP-E-MgADP 1t5c contained a P300A mutation, whereas the CENP-E motor domain in this study was expressed as the wild type. Pro300 is unique to CENP-E in kinesins. The main chain of Pro300 and the peptide bond between residues 299 and 300 of CENP-E-MgADP from this study is located in almost the same position as that in CENP-E-MgADP 1t5c (Fig. 4c).

Nucleotide-binding site
MgADP and four water molecules are located in the nucleotide-binding pocket (Figs. 1b, 4b and 5). The Mg 2+ ion interacts with a -phosphate oxygen moiety, four water molecules and Thr93 O 1 at the end of the P-loop (Fig. 5). The adenosine ring moiety of ADP makes van der Waals interactions with the side chain of Tyr94 at the beginning of 2 of CENP-E (Fig. 5) or Phe in Eg5. At the nucleotide-binding site of CENP-E, hydrophobic interactions such asstacking between ADP and Tyr94 are stronger than those in other kinesins. This is expected to be one potential reason why CENP-E tends to retain ADP for a long time. 2 and L5 located near ADP are shorter than in other kinesins, which is unique to CENP-E (Figs. 4a and 4b).
Owing to the larger number of residues in loop L1, which includes residues 18-26, compared with other kinesins, we think that the conserved residues 16-27 are disordered and are then located where they are able to interact with ADP. Further studies will be needed to support the possibility.
The CENP-E motor domain used in this study exhibited ATPase activity in the presence and absence of microtubules in an ATPase assay performed at 25 C. For crystal structure analysis, the CENP-E motor domain was prepared at 4 C without adding ATP/ADP; CIBA was then added after several days of purification. However, the structure of CENP-E-MgADP from this study contained ADP but not CIBA (Figs. 1,  4b and 5a). The ADP was derived from the bacteria used to express the CENP-E motor domain, and CENP-E is thought to have retained ADP throughout the purification procedure. This suggests that ADP release may be a control point in the role of CENP-E in chromosome congression during mitosis.

Conclusion
The structure was determined by crystallizing the CENP-E motor domain with its inhibitor CIBA. However, electron density for ADP instead of CIBA was observed in the nucleotide-binding site of CENP-E, even though ADP was not added during protein preparation. The determined structure at 1.9 Å resolution shows the CENP-E motor domain in complex with MgADP, and also gives much more structural information than the previously reported structure at 2.5 Å resolution. The results of this study support the biochemical view that the release of ADP from CENP-E is a rate-limiting step in the ATPase cycle (Sardar & Gilbert, 2012). In order to elucidate the structure of the CENP-E motor domain bound to its ATP-competitive inhibitor, it will be necessary to perform an additional experimental procedure to dissociate ADP from CENP-E before forming a CENP-E-inhibitor complex. Further studies will be needed to clarify the mechanism by which ADP dissociates from CENP-E. This structure will contribute to understanding and clarifying the function of the kinesin CENP-E. Studies of CENP-E will also lead to the development of anticancer drugs and will be of considerable interest for future antimitotic therapies.