|Home | About | Journals | Submit | Contact Us | Français|
The spindle checkpoint maintains genomic stability and prevents aneuploidy. Unattached kinetochores convert the latent open conformer of the checkpoint protein Mad2 (O-Mad2) to the active closed conformer (C-Mad2), bound to Cdc20. C-Mad2–Cdc20 is incorporated into the mitotic checkpoint complex (MCC), which inhibits the anaphase-promoting complex/cyclosome (APC/C). The C-Mad2-binding protein p31comet and the ATPase TRIP13 promote MCC disassembly and checkpoint silencing. Here, using nuclear magnetic resonance (NMR) spectroscopy, we show that TRIP13 and p31comet catalyze the conversion of C-Mad2 to O-Mad2, without disrupting its stably folded core. We determine the crystal structure of human TRIP13, and identify functional TRIP13 residues that mediate p31comet–Mad2 binding and couple ATP hydrolysis to local unfolding of Mad2. TRIP13 and p31comet prevent APC/C inhibition by MCC components, but cannot reactivate APC/C already bound to MCC. Therefore, TRIP13–p31comet intercepts and disassembles free MCC not bound to APC/C through mediating the local unfolding of the Mad2 C-terminal region.
The spindle checkpoint is a cell-cycle surveillance system that ensures proper chromosome segregation in mitosis1–3. Activated by unattached kinetochores, checkpoint proteins collaborate to inhibit the anaphase-promoting complex or cyclosome bound to its mitotic activator Cdc20 (APC/CCdc20)4–7 until all chromosomes achieve bipolar attachment to spindle microtubules. A major effector of the spindle checkpoint is the mitotic checkpoint complex (MCC), which consists of BubR1–Bub3, Cdc20, and Mad2. MCC binds to the substrate-binding site of APC/CCdc20, preventing it from recognizing and ubiquitinating securin and cyclin B1. Stabilization of securin and cyclin B1 delays sister-chromatid separation and mitotic exit.
A rate-limiting step in MCC formation is the conformational activation of the checkpoint protein Mad28, 9. Mad2 is an unusual protein with multiple folded conformers, including the latent open conformer (O-Mad2) and the active closed conformer (C-Mad2)10–14 (Fig. 1a). C-Mad2 can bind to Mad1, Cdc20, and other proteins through the Mad2-interaction motif (MIM)4, 11, 12, 15. O-Mad2 and C-Mad2 share the same stably folded core. They differ in their N- and C-terminal regions. When bound to a ligand, the C-terminal region of C-Mad2 wraps around the MIM and stretches across the core like a “safety belt” (Fig. 1a, right panel). Mad2 forms a constitutive complex with Mad1, in which Mad2 adopts the C-Mad2 conformation12. The C-terminal region of O-Mad2 does not form the safety belt and blocks the ligand-binding site (Fig. 1a, left panel). Because C-Mad2 is thermodynamically more stable than O-Mad2, O-Mad2 alone, in the absence of a ligand, can spontaneously convert to the unliganded C-Mad2 conformation, in which the safety belt is formed but does not trap a ligand13, 16 (Fig. 1a, middle panel).
Unattached kinetochores promote Mad2 activation and MCC formation through a series of signaling and recruitment events1–3. The KMN network of proteins (consisting of Knl1–Zwint, the Ndc80 complex, and the Mis12 complex) serves both as a major microtubule receptor and as a docking platform for checkpoint proteins at outer kinetochores, and is required for spindle checkpoint signaling17. When kinetochores are not attached by microtubules, the Ndc80 complex (Ndc80C) is phosphorylated by Aurora B and recruits the checkpoint kinase Mps118, 19. Mps1 phosphorylates Knl1, providing docking sites for the Bub1–Bub3 complex20–27. Bub1–Bub3 phosphorylated by Mps1 and Cdk1 further recruits the Mad1–C-Mad2 complex, and other MCC components BubR1–Bub3 and Cdc2028–36. The Mad1–C-Mad2 complex at kinetochores recruits additional O-Mad2 from the cytosol and converts it to an intermediate form (I-Mad2)14, 37, 38. Through a high-energy transition state in which its C-terminal region is locally unfolded, I-Mad2 binds to Cdc20 to form the Cdc20–C-Mad2 complex, which then associates with BubR1–Bub3 to produce MCC38–40. Once formed, the MCC diffuses away from kinetochores and inhibits APC/CCdc20 (refs. 29, 41).
Kinetochore attachment by microtubules inhibits binding of Mps1 to Ndc80C18, 19. This and other mechanisms reduce the levels and activities of Mps1 and downstream checkpoint components at the kinetochore. The C-Mad2-binding protein p31comet binds to the Mad1–C-Mad2 core and limits Mad2 conformational activation, at kinetochores when the checkpoint is on and in the cytosol after Mad1–C-Mad2 is removed from kinetochores during checkpoint inactivation42–45. These mechanisms collectively inhibit MCC formation. The existing MCC is then disassembled through APC/C-dependent ubiquitination of Cdc20 and through p31comet-mediated MCC disassembly, leading to APC/CCdc20 activation and anaphase onset46–52. Recently, the AAA+ ATPase TRIP13 has been shown to collaborate with p31comet to promote MCC disassembly and checkpoint inactivation53–57. p31comet acts as an adaptor to recruit C-Mad2 to TRIP13. TRIP13 then induces the conformational change of C-Mad2 to O-Mad2 in a process that requires ATP hydrolysis55. The structures of both human TRIP13 and its Caenorhabditis elegans homolog PCH-2 have been determined55, 58. Chemical crosslinking and functional studies further suggest that TRIP13 might promote the structural transition of Mad2 through engaging and unfolding the N-terminal region of Mad258.
In this study, we confirm that, in the presence of p31comet, TRIP13 catalyzes the conversion of C-Mad2 to O-Mad2 using nuclear magnetic resonance (NMR) spectroscopy. Proton/deuterium (H/D) exchange experiments further show that TRIP13-catalyzed structural transition of Mad2 does not involve the global unfolding of Mad2, but entails local unfolding of the C-terminal region. We have determined the crystal structure of human TRIP13 in its ADP-bound state, and identified TRIP13 residues critical for p31comet–Mad2 binding and for coupling ATP hydrolysis to Mad2 unfolding. Finally, we show that TRIP13–p31comet prevents inhibition of APC/C by MCC components in vitro, but does not restore APC/C activity after APC/C has already been inhibited by MCC. Our studies provide insight into the mechanism of TRIP13-catalyzed conformational change of Mad2 and the molecular processes leading to spindle checkpoint silencing.
To confirm that TRIP13 catalyzed the structural transition of Mad2, we performed NMR analysis of 15N-labeled Mad2 in the presence or absence of TRIP13 and p31comet. For this purpose, we used the monomeric Mad2R133A mutant that retained strong binding to p31comet (K d 240nM) and the ability to undergo the O–C structural transition13, 43. The 1H–15N heteronuclear single quantum coherence (HSQC) spectrum of Mad2R133A showed that it had adopted the C-Mad2 conformation (Fig. 1b, Supplementary Fig. 1a). We then added ATP and sub-stoichiometric amounts of TRIP13 (1:325 molar ratio of TRIP13 hexamers to Mad2 monomers) and ΔN35-p31comet (1:33 molar ratio of p31comet to Mad2) to the C-Mad2R133A sample, and acquired an HSQC spectrum after 5h of incubation. The HSQC spectrum of the TRIP13-treated sample was different from that of the starting C-Mad2R133A sample, and matched that of O-Mad2R133A (Fig. 2c, Supplementary Fig. 1b). Therefore, TRIP13–p31comet catalyzes the conversion of unliganded C-Mad2 to O-Mad2 in the presence of ATP.
Because C-Mad2 and O-Mad2 share a common folded core (Fig. 1a), TRIP13–p31comet catalyzed C-Mad2–O-Mad2 conversion does not necessarily require the global unfolding of Mad2. Local unfolding of the C-terminal region of C-Mad2 can complete this conversion and, in the case of liganded C-Mad2, release the bound ligand. To test whether TRIP13-catalyzed structural transition of Mad2 involved global or local unfolding of Mad2, we performed H/D exchange NMR experiments. Amide protons that form hydrogen bonds in folded structural elements exchange slowly with solvent, and their NMR signals remain visible for hours after the protein is lyophilized and dissolved into D2O. In contrast, amide protons in loop regions or transiently folded regions that do not form stable hydrogen bonds exchange rapidly with solvent, and their NMR signals disappear in D2O.
The HSQC spectrum of C-Mad2R133A in a D2O solution for 2.5h showed a set of slow-exchanging peaks that mostly belonged to residues in the stably folded core and strand β9 (Fig. 2a, b). A set of slow-exchanging peaks remained at 6h after the addition of TRIP13, p31comet, and ATP (Fig. 2c). The positions of these peaks matched those of peaks in O-Mad2R133A, indicating that C-Mad2 had been converted to O-Mad2 during the course of experiment. We mapped these slowing-exchanging amides onto the structures of both C-Mad2 and O-Mad2. The slow-exchanging amides in the presence of TRIP13 and p31comet belonged exclusively to residues in the folded core of Mad2 (Fig. 2d, e). The slow-exchanging amides on β9 and alternating amides on β5 that participated in hydrogen bonding with β9 were no longer protected, indicating that the C-terminal region of C-Mad2 (β8 and β9) were expectedly unfolded during the structural transition. On the other hand, many amide protons in the Mad2 core (β4-6 and αA-C) were protected from exchange with solvent during the TRIP13-catalyzed structural transition from C-Mad2 to O-Mad2. Our results therefore demonstrate that Mad2 does not undergo global unfolding during TRIP13-catalyzed structural conversion.
Because the C-terminal region has to unfold and refold to allow C-Mad2 to convert to O-Mad2, the C-Mad2–O-Mad2 transition catalyzed by TRIP13–p31comet minimally involves the local unfolding of the C-terminal safety belt of C-Mad2. Because not all structural elements of C-Mad2 contain slow-exchanging residues (Fig. 2b), we cannot exclude the possibility of local unfolding of other structural elements, aside from the C-terminal region.
The structure of PCH-2, the C. elegans homolog of TRIP13, has previously been determined, revealing an asymmetric hexameric architecture55 (Fig. 3a). When viewed from the side, the PCH-2 hexamer forms a slightly curved structure with a convex face and a concave face (Fig. 3a), similar to ClpX59. The convex face in ClpX has been shown to bind to substrates, and has generally been referred to as the top face59. We will adopt this convention and refer to the convex surface of PCH-2/TRIP13 as the top face.
To understand how TRIP13 recognized p31comet and Mad2, we attempted to determine the structure of the human TRIP13–p31comet–Mad2 complex. We subjected human TRIP13 E253A mutant (which was deficient in ATP hydrolysis) bound to ATP and the p31comet–Mad2 complex to crystallization, and obtained diffracting crystals. Structure determination, however, revealed that the crystals contained TRIP13 alone in the ADP-bound state (Fig. 3b, Supplementary Fig. 2a). Even though the ATPase-deficient mutant of TRIP13 was used in our studies, the residual ATPase activity of TRIP13 E253A presumably converted ATP to ADP, and contributed to the dissociation of the complex.
Structure of human TRIP13 bound to ADP is highly similar to that of the C. elegans PCH-2 promoter bound to ADP, with an RMSD of 1.57Å for all atoms (Fig. 3b, c). Both contain an N-terminal domain (NTD), a large AAA+ subdomain, and a small AAA+ subdomain. ADP is bound at the interface between the large and small AAA+ subdomains. The structures of the large and small AAA+ subdomains and their relative orientation are virtually identical between human TRIP13 and the ADP-bound protomer of C. elegans PCH-2. The only discernable difference is located at the so-called pore loop. This region forms a helix in PCH-2, but does not have visible electron density and is likely flexible in human TRIP13 (Fig. 3c). The pore loop has been implicated in substrate binding and unfolding in other AAA+ ATPases59, 60. The structural differences between human and C. elegans TRIP13 are consistent with the intrinsically dynamic nature of this loop.
Our structure of human TRIP13 bound to ADP is virtually identical to those of human apo-TRIP13 or TRIP13 bound to ATP58 (Supplementary Fig. 2b). Because the human TRIP13 structure does not form a closed hexamer (Supplementary Fig. 2c) and has a disordered pore loop, we used the structure of the C. elegans PCH-2 hexamer to guide our subsequent mutagenesis and biochemical studies.
To gain insight into the mechanism of TRIP13, we systematically mutated conserved residues in TRIP13 that were surface exposed or partially exposed, and studied the effects of these mutations on the ATPase activity, Mad2 unfolding, and binding to p31comet–Mad2. We describe our findings on the following mutants: W221A, F222A, K227A, K231E, N278A, and G320A. We developed a high-throughput TRIP13 ATPase assay. ATP hydrolysis by TRIP13 wild type (WT) was dose-dependent, and the TRIP13 E253A mutant was deficient in ATP hydrolysis (Supplementary Fig. 3a, b). Consistent with a previous report55, the ATPase activity of TRIP13 was stimulated by the p31comet–C-Mad2 complex, but not by either protein alone (Supplementary Fig. 3c). Similar results were obtained when Mad2 was locked into the C-Mad2 conformation with the L13A mutation, with binding of the artificial peptide ligand MBP1, or with binding of the Cdc20 N-terminal 170-residue fragment (Cdc20N) (Supplementary Fig. 3d). Mad2L13A stimulated TRIP13 more effectively, presumably because it could not stably adopt the O-Mad2 conformer and was expected to go through futile binding/unfolding cycles catalyzed by TRIP13. The ATPase activity of TRIP13 was greatly stimulated at 1:1 molar ratio of TRIP13 hexamer to p31comet–Mad2L13A (Supplementary Fig. 3e), suggesting that one TRIP13 hexamer likely engages one p31comet–Mad2 heterodimer at a given time.
As reported previously for mouse TRIP1355, human TRIP13 W221A had much higher basal ATPase activity, nearly equal to the wild-type TRIP13 activity stimulated by p31comet–Mad2 (Fig. 4a). The ATPase activity of TRIP13 W221A was not further stimulated by p31comet–Mad2. TRIP13 W221 is located in the pore loop (Fig. 4b), a structural element of AAA+ ATPases thought to be responsible for substrate translocation into the pore of the hexamer59. In contrast to W221A, mutation of F222, a neighboring residue in the pore loop, did not reduce the basal ATPase activity of TRIP13, but abolished the stimulation by p31comet–Mad2 (Fig. 4a). Thus, the pore loop of TRIP13 serves two function: (1) restraining the basal ATPase activity of TRIP13 and (2) coupling ATP hydrolysis to p31comet–Mad2 binding or the conformational change of Mad2. F222 is specifically involved in the second function.
The α2 helix in the large AAA+ subdomain is located on the top face of TRIP13 and is adjacent to the pore loop55, 61 (Fig. 4b). Mutations of K227 and K231, two residues in the α2 helix, both reduced the stimulated activity of TRIP13 by p31comet–Mad2, despite having opposite effects on the basal ATPase activity of TRIP13 (Fig. 4a). Intriguingly, mutation of N278, a residue in the α3 helix in the large AAA+ subdomain61 (Fig. 4b), also abolished the stimulation of TRIP13 by p31comet–Mad2 (Fig. 4b). Finally, mutation of G320 in the hinge between the large and small AAA+ subdomains greatly reduced the ATPase activity of TRIP13 and its stimulation by p31comet–Mad2 (Fig. 4a, b). Thus, we have identified a set of residues in diverse structural elements of TRIP13 that is critical for coupling ATP hydrolysis to substrate binding or unfolding.
We developed a beads-based assay to monitor the disassembly of C-Mad2–ligand complexes by TRIP13 (Fig. 5a). TRIP13 WT was capable of efficiently dissociating Mad2 from MBP1 in the presence of ATP and p31comet (Fig. 5b) and in a dose-dependent manner (Supplementary Fig. 4a). The ATPase-deficient mutant TRIP13 E253A was unable to dissociate Mad2, but bound strongly to p31comet–Mad2–MBP1 in the presence of ATP. Interestingly, TRIP13 at 25nM exhibited considerable ATPase activity (Fig. 4a), but no dissociation of Mad2 was observed at this concentration (Supplementary Fig. 4a). For unknown reasons, only at concentrations above 500nM, TRIP13 was able to efficiently dissociate Mad2. We next tested whether the TRIP13 mutations affected the ability of TRIP13 to release C-Mad2 from bound ligands. TRIP13 W221A, F222A, K227A, K231E, and N278A, which were deficient in p31comet–Mad2 stimulation, all failed to dissociate Mad2 from MBP1 (Fig. 5c) or Cdc20N (Supplementary Fig. 4b). Thus, the coupling between ATP hydrolysis and p31comet–Mad2 binding or Mad2 unfolding is critical for releasing Mad2 from ligands.
Surprisingly, TRIP13 G320A, which at 25nM showed little ATPase activity (Fig. 4a), was able to dissociate Mad2 from MBP1 or Cdc20 (Fig. 5c, Supplementary Fig. 4b). This apparent discrepancy prompted us to perform further experiments on TRIP13 G320A. TRIP13 G320 is located at the hinge between its large and small AAA+ subdomains (Fig. 5d). Because the Mad2 dissociation assay was performed with 500nM TRIP13, we tested whether TRIP13 G320A had ATPase activity at higher concentrations. Indeed, even though TRIP13 G320A at 25nM had undetectable ATPase activity in the presence or absence of p31comet–Mad2, this mutant had substantial basal activities at 100nM and 200nM concentrations (Fig. 5e), which were further stimulated by p31comet–Mad2. This likely explains the ability of G320A at high concentrations to dissociate Mad2 from ligands.
One possibility for the lack of ATPase activity of G320A at low concentrations is that the mutation weakens the oligomerization of TRIP13 and prevents it from reaching/maintaining the functional hexameric state. To test this possibility, we compared the elution profiles of TRIP13 WT and G320A in the presence of ATP on a gel filtration column. Both proteins had very similar fractionation profiles, and predominantly formed oligomers (Supplementary Fig. 5). The fractionation profiles of TRIP13 WT in the presence of ATP were very similar to those reported previously58. Thus, in the presence of ATP, the G320A mutation does not appear to disrupt the oligomerization of TRIP13.
We next subjected TRIP13 WT to analytical ultracentrifugation sedimentation velocity studies in the presence of 100μM ATP. While interpretable concentration profiles could be collected from such samples, the resulting c(s) distributions demonstrated a large degree of polydispersity in s 20,w values greater than 5s. Thus, consistent with the broad fractionation profiles on gel filtration, ATP binding does not induce the formation of discrete TRIP13 oligomers in vitro, possibly due to the nucleotide-driven conformational dynamics of TRIP13. Interestingly, in the absence of ATP, TRIP13 WT at high concentrations displayed sedimentation behavior consistent with an equilibrium between discrete species (Fig. 5f). By contrast, TRIP13 G320A was monomeric under all concentrations tested. Thus, TRIP13 G320A is deficient for oligomerization in the nucleotide-free state, even though the hinge is not located in the hexamer interface.
Previous studies have shown that not all six protomers of AAA+ ATPases bind to and hydrolyze nucleotides at the same time59. The relative positions of the small and large AAA+ subdomains in a given protomer are different between its nucleotide-bound and nucleotide-free states55. Thus, the flexibility of the hinge might enable relative movements between the two subdomains during the ATPase cycle. These movements not only drive substrate remodeling, but also maintain the functional hexamer state of the enzyme. We propose that the G320A mutation interferes with conformational dynamics of TRIP13, weakens TRIP13 oligomerization, and reduces the ATPase activity of TRIP13. Sequence alignment of the hinge regions of TRIP13 and two other eukaryotic AAA+ ATPases, p97/VCP and NSF, revealed a conserved motif with the consensus of ϕGXP (ϕ, a hydrophobic residue; X, any residue) (Fig. 5g). Mutations affecting the length of the hinge in ClpX uncouple the ATPase activity from substrate unfolding62. We speculate that hinge flexibility in AAA+ ATPases may be generally important for the conformational dynamics between the large and small AAA+ subdomains during the ATPase cycle.
The TRIP13 mutants deficient for stimulation by p31comet–Mad2 and for Mad2 release from ligands might have lost their ability to bind p31comet–Mad2 or be defective in coupling ATP hydrolysis to Mad2 unfolding. To distinguish these possibilities, we wished to directly assay the binding of p31comet–Mad2 to TRIP13. We observed that TRIP13 E253A, which was deficient in ATP hydrolysis and Mad2 dissociation, was able to bind to p31comet–Mad2–MBP1 beads in the Mad2 dissociation assay (Fig. 5b, c). We hypothesized that the ATP-bound state of TRIP13 might bind more strongly to p31comet–Mad2. Indeed, addition of ATPγS, a slowly hydrolyzable ATP analog, did not support Mad2 dissociation, but greatly enhanced the binding of TRIP13 WT to beads bound with p31comet–Mad2–MBP1 (Supplementary Fig. 6a). Addition of ADP did not enhance TRIP13 WT binding to p31comet–Mad2–MBP1, and reduced the binding of TRIP13 E253A, presumably because ADP displaced some ATP bound to this mutant. These results show that p31comet–Mad2 prefers to bind to the ATP-bound state of TRIP13. The use of ATPγS in place of ATP converts the Mad2 dissociation assay into an assay that can monitor the binding between TRIP13 and p31comet–Mad2.
We next tested the binding of the TRIP13 mutants to p31comet–Mad2 with this assay. TRIP13 W221A, F222A, K227A, and K231E were deficient in binding to p31comet–Mad2–MBP1 (Fig. 6a) or p31comet–Mad2–Cdc20N (Supplementary Fig. 6b), suggesting that these residues at the pore loop and the α2 helix mediate binding of p31comet–Mad2 to TRIP13 (Fig. 6b). Therefore, p31comet–Mad2 likely binds to the top face of TRIP13 through charged interactions with the α2 helix and through interactions with the pore loop.
Recently, Corbett and workers suggested that TRIP13 engages the N-terminal region of Mad2 through the pore loop and unfolds Mad2 from the N-terminus58. The most crucial evidence for Mad2 N-terminal unfolding by TRIP13 is the crosslinking data performed with the TRIP13 W221C mutant. This crosslinking appeared to show that the N-terminal region of Mad2 makes direct contact with the TRIP13 pore loop. On the other hand, we have shown that the W221A mutation abrogates binding of TRIP13 to p31comet–Mad2 in the presence of ATPγS (Fig. 6a, Supplementary Fig. 6b). We wondered whether the W221C mutant used in the Corbett study was similarly defective in binding to p31comet–Mad2. We thus made the TRIP13 W221C mutant (Supplementary Fig. 6c). Expectedly, the W221C mutant behaved very similarly to the W221A mutant. It had higher basal ATPase activity, which was not further stimulated by p31comet–Mad2 (Supplementary Fig. 6d). W221C also could not dissociate Mad2 from its ligand MBP1 (Supplementary Fig. 6e). More importantly, W221C was defective in binding to p31comet–Mad2 (Supplementary Fig. 6f). Therefore, the installation of a crosslinkable residue in the pore loop destroys the function of this loop. Although the crosslinking experiments were performed with sub-stoichiometric amount of W221C mixed with TRIP13 WT, it remains possible that the mutated pore loop does not form functional contact with Mad2 even when it is present as a single copy in the hexamer.
TRIP13 G320A expectedly retained binding to p31comet–Mad2 (Fig. 6a, Supplementary Fig. 6b), as it was capable of dissociating Mad2 from ligands in the presence of ATP. The weakened binding of TRIP13 G320A to p31comet–Mad2 is likely an indirect consequence of altered conformational dynamics of the mutant.
Surprisingly, the TRIP13 N278A mutant, which was deficient in promoting Mad2 dissociation from ligands and whose ATPase activity was not stimulated by p31comet–Mad2, retained substantial binding to p31comet–Mad2. Thus, the N278 mutation uncouples p31comet–Mad2 binding from ATP hydrolysis and subsequent conformational change of Mad2. N278 is therefore critical for harnessing the energy of ATP hydrolysis to switch the conformation of Mad2. TRIP13 N278 is equivalent to PCH-2 N278, which is located at the interface between adjacent PCH-2 protomers. It is in close proximity to the pore loop at four such interfaces between one nucleotide-bound protomer and one nucleotide-free protomer (I and II), but is far away from the pore loop at the other two interfaces between two nucleotide-free protomers (III) (Fig. 6c). We speculate that N278 engages with the pore loop in specific nucleotide states of PCH-2/TRIP13 when it is bound to p31comet–Mad2, and facilitates the dynamic movements of the pore loop during ATP hydrolysis.
MCC disassembly has been proposed to be the key function of TRIP13 in checkpoint silencing53, 54, 56. We tested whether TRIP13, with the help of p31comet, could inactivate MCC in vitro. As shown previously28, 29, 63, incubation of recombinant BubR1N (residues 1–370), dimeric Mad2 WT, and full-length Cdc20 produced a minimal, functional MCC, which inhibited the ubiquitination of securin by APC/CCdc20 (Fig. 7a, b). Addition of TRIP13, p31comet, and ATP to the mixture of MCC components (prior to the addition of APC/CCdc20) completely restored the ubiquitination activity of APC/CCdc20 (Fig. 7b), indicating that TRIP13 was able to inactivate MCC. As expected, TRIP13-dependent MCC inactivation required p31comet and ATP. The ATPase-deficient TRIP13 E253A was unable to restore APC/CCdc20 ubiquitination activity. The TRIP13 mutants defective for Mad2 dissociation from peptide ligands were also unable to inactivate MCC. TRIP13 G320A, which was capable of releasing Mad2 from ligands, was able to inactivate MCC to allow APC/CCdc20-mediated ubiquitination of securin. The perfect correlation between the abilities of TRIP13 mutants to release Mad2 from ligands and to inactivate MCC strongly suggest that TRIP13 inactivates MCC through promoting the C-Mad2 to O-Mad2 conversion and disrupting the C-Mad2–Cdc20 interaction.
We next tested whether the N-terminal region of Mad2 was required for TRIP13-dependent inactivation of MCC. Mad2 ΔN10 could be incorporated into a functional MCC that inhibited APC/CCdc20 (Fig. 7c). In contrast to MCC containing Mad2 WT, MCC containing Mad2 ΔN10 was not efficiently inactivated by TRIP13, and APC/C activity was not fully restored. Furthermore, Mad2 ΔN10 retained its binding to p31comet, but the Mad2 ΔN10–p31comet complex was deficient in TRIP13 binding (Supplementary Fig. 6g). Therefore, consistent with the results of Corbett and coworkers58, the N-terminal region of Mad2 is required for TRIP13 binding and proper MCC disassembly. This result does not, however, indicate that the N-terminus of Mad2 directly engages the pore loop of TRIP13 or that the unfolding of Mad2 by TRIP13 starts at the N-terminus. We note that our Mad2 ΔN10 protein contained a 12-residue tag in place of the native N-terminal 10 residues of Mad2. It is formally possible that the extra two residues in length in the Mad2 ΔN10 protein, not the change in amino acid sequence, accounts for the weakened binding to TRIP13.
Finally, we tested whether TRIP13–p31comet could inactivate MCC that was already bound to APC/CCdc20. We pre-incubated the MCC components with APC/CCdc20 and then added TRIP13, p31comet, and ATP, before performing the ubiquitination assay. TRIP13–p31comet was unable to relieve the inhibition of APC/CCdc20 by the pre-bound MCC (Fig. 7c). Therefore, our results indicate that TRIP13 can catalyze the disassembly of free MCC, but cannot effectively inactivate MCC already bound to APC/CCdc20.
The asymmetric hexameric ring structure of TRIP13 is reminiscent of that of the bacterial unfoldase ClpX55, 59, 64. ClpX binds to substrates with its convex, top surface, and is believed to completely unfold substrates and thread them through an inner pore that is lined by the pore loop conserved in AAA+ ATPases59. The unfolded substrates that emerge from the pore are then degraded by the ClpP protease, which is bound to the concave, bottom surface of ClpX59. The substrates of ClpX are bacterial proteins, which are not structurally related to Mad2. We have found that, similar to ClpX, TRIP13 binds to p31comet–Mad2 on its convex, top surface through charged interactions with the α2 helix in the large AAA+ subdomain and through engaging the pore loop. Because blocking ATP hydrolysis by mutations or ATPγS enhances p31comet–Mad2 binding to TRIP13, p31comet–Mad2 likely binds to the ATP-bound promoters of TRIP13.
Unlike ClpX, however, TRIP13-catalyzed conformational change does not involve the complete, global unfolding of Mad2, as slowly exchanging amide protons are protected from D2O during the entire process. We propose that the p31comet–Mad2–MIM complex docks on TRIP13 protomers in the ATP-bound state and at a site involving the α2 helix (Fig. 8a). The C-terminus of Mad2 then engages the pore loop of TRIP13 and relieves the autoinhibition of the ATPase activity. ATP hydrolysis drives the movement of the pore loop, which tugs at the Mad2 C-terminus. The translocation of the Mad2 C-terminus is further supported by other structural elements, such as N278 in α3. The local unfolding of the C-terminal safety belt of Mad2 releases the bound MIM and disrupts the p31comet–Mad2 interaction. As TRIP13 does not bind strongly to either Mad2 or p31comet, the breakup of the p31comet–Mad2 complex causes both proteins to be released from TRIP13. TRIP13 can then bind another p31comet–Mad2–MIM complex to repeat the process.
At present, we do not know whether the local unfolding of Mad2 C-terminal region involves the threading of this region through the central pore of the TRIP13 hexamer. If it does, mechanisms must exist to release the partially unfolded Mad2 from TRIP13 so that the stable core of Mad2 is not threaded through the pore and unfolded. Alternatively, ATP-driven translocation events of the Mad2 C-terminus may suffice to unravel the C-terminal safety belt of Mad2 and occur solely on the top face of TRIP13.
A recent study suggests that the N-terminal region of Mad2 directly engages the pore loop of TRIP13 and that Mad2 unfolding by TRIP13 starts at the N-terminus58. The key evidence supporting these claims involves crosslinking experiments with the TRIP13 W221C mutant as a sub-stoichiometric mixture with WT. We have now shown that this pore loop mutant as a hexamer is defective in p31comet–Mad2 binding and in Mad2 unfolding. It is possible that the mutated pore loop, even as a single copy in a hexamer, does not make functional contact with Mad2. On the other hand, consistent with the previously reported cellular data58, the Mad2 N-terminus is required for TRIP13 binding and for TRIP13-catalyzed disassembly of MCC. Although it makes more intuitive sense that TRIP13 initiates Mad2 unfolding locally from the C-terminus, our results do not disprove that TRIP13 unfolds Mad2 locally from the N-terminus.
Our reconstitution of TRIP13–p31comet-dependent MCC inactivation in vitro has provided insight into the mechanism by which TRIP13 promotes spindle checkpoint silencing. Perhaps our most important finding is that TRIP13–p31comet only inactivates free MCC prior to its binding to APC/CCdc20, but not MCC already bound to APC/CCdc20. We propose that TRIP13–p31comet disassembles free MCC in a step-wise fashion (Fig. 8b). Because p31comet and BubR1 bind to overlapping interfaces on C-Mad2 and compete for C-Mad2 binding45, 65, 66, p31comet binding to C-Mad2 disrupts the integrity of MCC and releases BubR1. The p31comet–C-Mad2–Cdc20 complex is then recognized by TRIP13. TRIP13-catalyzed conversion of C-Mad2 to O-Mad2 dissociates Mad2 from Cdc20, causing the complete disassembly of MCC. In this way, TRIP13–p31comet reduces the cellular concentration of free MCC and helps to turn off the spindle checkpoint.
Our results suggest that additional mechanisms are required to inactivate MCC bound to APC/C and promote checkpoint silencing. One such mechanism is APC/CCdc20-dependent ubiquitination of the Cdc20 molecule that is part of the MCC47, 48. Cdc20 ubiquitination is likely sufficient to release MCC from APC/C, which can then be acted on by TRIP13–p31comet (Fig. 8b). Alternatively, degradation of ubiquitinated Cdc20 by the proteasome can cause the complete disassembly of MCC.
In addition to promoting MCC disassembly, TRIP13–p31comet is needed to revert the spontaneously converted C-Mad2 back to O-Mad2. Mad2 released by Cdc20 degradation that is still in the C-Mad2 conformer may also be converted to O-Mad2 by TRIP13–p31comet (Fig. 8b). TRIP13–p31comet may therefore be required to maintain the pool of O-Mad2 in the cell by converting free, unliganded C-Mad2 back to the O-Mad2. Failure to do so causes premature accumulation of C-Mad2 and spindle checkpoint defects67, as O-Mad2 is the Mad2 conformer that is activated by Mad1–Mad2 at kinetochores to become I-Mad2, which is subsequently incorporated into MCC.
In conclusion, using a combination of biophysical and biochemical approaches, we have demonstrated that TRIP13–p31comet catalyzes the conversion of C-Mad2 to O-Mad2 without the global unfolding of Mad2. We have defined the functional roles of multiple structural elements of TRIP13 in coupling ATP hydrolysis to Mad2 unfolding. Finally, we have reconstituted TRIP13–p31comet-dependent MCC inactivation in vitro, leading to new testable models of how TRIP13 and p31comet promote spindle checkpoint silencing and the metaphase–anaphase transition.
BL21 (DE3) Escherichia coli cells were transformed with plasmids encoding His6-tagged human TRIP13 WT or mutants, cultured in Terrific Broth until the OD600 reached 1.0, and induced overnight at 18°C with 0.2mM isopropyl β-D-1-thiogalactopyranoside (IPTG). Collected cells were resuspended in the TRIP13 Purification Buffer (50mM NaH2PO4·2H2O pH 8.0, 300mM NaCl, 5mM MgCl2, 10% glycerol, 5mM β-mercaptoethanol (BME), 200μM ATP) supplemented with 20mM imidazole and 1mM phenylmethylsulfonyl fluoride (PMSF) and lysed. The cleared lysate was incubated with Ni2+-NTA Agarose beads (Qiagen). Beads were washed with the TRIP13 Purification Buffer supplemented with 50mM imidazole, and eluted with the TRIP13 Purification Buffer supplemented with 250mM imidazole. The proteins were incubated with TEV for overnight to remove the His6 tag, and fractionated on a Resource Q column (GE) with the TRIP13 Purification Buffer containing 50mM NaCl as the starting buffer, and the TRIP13 Purification Buffer containing 1M NaCl as the elution buffer. TRIP13 proteins were further purified on the HiLoad 16/600 Superdex 200 gel filtration column (GE) in the buffer containing 20mM HEPES, pH 7.7, 100mM KCl, 5mM MgCl2, 5% glycerol, 1mM DTT. The proteins were concentrated with Centriprep Concentrators (Millipore) and stored at −80°C.
His6-ΔN35-p31comet–Mad2L13A complex and His6-Mad2 (containing the C79S and C106S mutations to avoid disulfide formation in vitro) were expressed in BL21 (DE3) cells. Collected cells were lysed in the Mad2 Purification Buffer (50mM Tris, pH 8.0, 300mM NaCl, 5mM BME) supplemented with 1mM PMSF. The cleared lysates were incubated with Ni2+-NTA Agarose beads (Qiagen). Beads were washed with the Mad2 Purification Buffer supplemented with 10mM imidazole, and with the MOPS Buffer (50mM MOPS pH 7.2, 300mM NaCl, 5mM BME) supplemented with 25mM imidazole. Proteins were eluted with the MOPS Buffer supplemented with 200mM imidazole, and exchanged into the MOPS Buffer with no imidazole with PD-10 desalting columns (GE). The His6-tag of the ΔN35-p31comet–Mad2L13A complex was cleaved by TEV, whereas His6-Mad2 was not cleaved. The proteins were purified with the Resource Q column with the MOPS Buffer containing 75mM NaCl as the binding buffer, and the MOPS Buffer containing 1M NaCl as the elution buffer. The proteins were concentrated and further purified with the Superdex 200 Increase 10/300 GL column (GE) in a buffer containing 25mM Tris, pH 7.5, 200mM NaCl, 10% glycerol, and 1mM Tris(2-carboxyethyl)phosphine (TCEP).
His6-ΔN35-p31comet was expressed in M15[pRep4] E. coli cells. Cells were lysed in the buffer containing 50mM MOPS, pH 7.2, 150mM NaCl, 5mM BME, 10mM imidazole. After purification with Ni2+-NTA Agarose beads and TEV cleavage, the protein was further purified with the Resource Q column with the binding buffer of 20mM HEPES, pH 7.5, 50mM NaCl, 10% glycerol, and 1mM DTT and the elution buffer of 20mM HEPES, pH 7.5, 1M NaCl, 10% glycerol, and 1mM DTT. The protein was further purified by gel filtration in the buffer containing 20mM HEPES, pH 7.5, 200mM NaCl, 10% glycerol, and 1mM DTT.
His6-Mad2R133A was expressed in M15[pRep4] cells in M9 minimal media containing 15N-NH4Cl as the sole nitrogen source. The cells were collected in the R133A Buffer (50mM sodium phosphate, pH 8.0, 300mM NaCl, 5mM BME) supplemented with 10mM imidazole, lysed by sonication, and cleared by centrifugation. The lysate was incubated with Ni2+-NTA Agarose beads (Qiagen). Beads were washed with the R133A Buffer supplemented with 10mM imidazole, followed by the R133A Buffer supplemented with 20mM imidazole. The protein was eluted with the R133A Buffer supplemented with 250mM imidazole. The protein was dialyzed into the buffer containing 50mM Tris, pH 8.0, 100mM NaCl, 1mM DTT, and cleaved overnight with TEV to remove the His6-tag. The cleaved protein was purified with the Resource Q column (GE) with the binding buffer of 20mM Tris, pH 8.0, 50mM NaCl, and 1mM DTT, and the elution buffer of 20mM Tris, pH 8.0, 1M NaCl, and 1mM DTT. The protein was further purified with the Superdex 75 10/300 GL column (GE) in a buffer containing 50mM sodium phosphate, pH 6.8, 300mM KCl, and 1mM DTT.
Expression of Cdc20N-His6 (residues 1–170 of human Cdc20) in BL21 (DE3) cells were induced with 1mM IPTG for 6h at 37°C. The collected cells were lysed in 8M urea, 100mM NaH2PO4, and 10mM Tris at pH 8.0 for overnight at room temperature. The lysate was sonicated, cleared by centrifugation, and incubated with Ni2+-NTA Agarose beads (Qiagen). The beads were washed with 8M urea, 100mM NaH2PO4, and 10mM Tris at pH 6.3, and eluted with 8M urea, 100mM NaH2PO4, and 10mM Tris at pH 4.5. The protein was dialyzed into the buffer containing 25mM Tris, pH 8.0, 300mM NaCl, 5% glycerol, and 1mM BME.
All NMR experiments were recorded on Agilent DD2 600MHz spectrometer at 30°C. To obtain uniformly 15N-labeled Mad2R133A, we used M9 minimal media with 15NH4Cl as the sole nitrogen source (1gl−1). Purified 15N-Mad2 R133A was incubated overnight at room temperature before the NMR experiments to ensure that the two forms of Mad2 have reached equilibrium13. The final NMR samples contained 140µM 15N-Mad2R133A (after overnight pre-incubation) with or without 4.3µM ΔN35-p31comet and 0.43µM TRIP13 dissolved in a buffer containing 25mM Tris, pH 7.5, 300mM KCl, 1mM MgCl2, 1mM DTT and 5% D2O. ATP was added to the 15N-Mad2R133A–ΔN35-p31comet–TRIP13 sample at a final concentration of 1mM. For monitoring the conformational change of Mad2, a series of 2D 1H−15N HSQC spectra each lasting 2h 34min were acquired for a total acquisition time of 40h.
For 1H/2H (H/D) exchange experiments, the final NMR samples were lyophilized and dissolved in D2O, and contained 150µM 15N-Mad2R133A in a buffer containing 25mM Tris, pH 7.5, 200mM NaCl, 5mM MgCl2, and 1mM DTT. After the acquisition of a 2D 1H−15N HSQC spectrum lasting 2h 34min, ATP (0.1M in D2O) and the TRIP13–N35-p31comet complex were added into the Mad2R133A D2O sample to final concentrations of 5mM and 1.5µM, respectively. A series of 2D 1H-15N HSQC spectra each lasting 2h 34min were acquired for a total acquisition time of 24h. All data were processed with NMRpipe68 and analyzed with NMRView69.
His6-TRIP13E253A ΔC2 (with the C-terminal two residues deleted) was expressed in BL21(DE3) cells, and purified using Ni2+-NTA Agarose resin pre-equilibrated with the lysis buffer (50mM phosphate, pH 7.5, 300mM NaCl, 40mM imidazole, and 5mM BME). After cleavage by TEV to remove the His6-tag, the TRIP13E253A protein was further purified with a resource Q column followed by a Superdex 200 gel filtration column equilibrated with a buffer containing 20mM phosphate, pH 8.0, 300mM NaCl, and 0.5mM TCEP. The purified protein was concentrated to 2mgml–1. The ΔN35-p31comet–Mad2L13A complex was expressed and purified as previously described45. TRIP13E253A and ΔN35-p31comet–Mad2L13A were mixed at 1:2 molar ratio in the presence of 200μM ATP. The protein mixture was fractionated on a Superdex 200 column equilibrated with a buffer containing 20mM Tris, pH 8.0, 100mM NaCl, and 0.5mM TCEP. The TRIP13E253A–ΔN35-p31comet–Mad2L13A complex was then concentrated to 5.5mgml–1, and subjected to crystallization by the sitting-drop vapor diffusion method at 20°C. The best crystals were obtained with a reservoir solution containing 0.1M HEPES, pH 7.5, 0.2M NaCl, and 4% (v/v) isopropanol. Crystals were cryo-protected by increasing the final concentration of glycerol to 30% (v/v) in a step-wise fashion and flash-cooled in liquid nitrogen.
The X-ray diffraction data were collected at beamline 19-ID (SBC-CAT) at the Advanced Photon Source (Argonne National Laboratory, Argonne, Illinois, USA). The data were processed in the program HKL300070. Initial phases were obtained by molecular replacement in the program Phaser71, using the structure of C. elegans PCH-2 as a search model (PDB ID: 4XGU). Iterative model building and refinement were performed in the programs Coot72 and PHENIX73, respectively. Although the TRIP13E253A–ΔN35-p31comet–Mad2L13A complex was used for crystallization in the presence of ATP, the final crystals only contained TRIP13E253A bound to ADP. The data collection and refinement statistics are summarized in Supplementary Table 1.
All reactions were performed in white OptiPlate-384 plates (Perkin Elmer) at room temperature. TRIP13 and ΔN35-p31comet–Mad2L13A were diluted in the Reaction Buffer (25mM Tris, pH 7.5, 200mM NaCl, 10mM MgCl2, 1mM DTT, 5% glycerol, and 0.05% Tween) to a final volume of 5µl, and incubated for 15min. ATP was added to the reaction to a final volume of 10µl and a final concentration of 0.1mM, and incubated for 1h. Volume of 10µl Glo Reagent from the ADP Glo Kinase Assay kit (Promega) was added to the reaction, and incubated for 40min. Volume of 20µl Detect reagent from the kit was added to the reaction, and incubated for 30min. Luminescence was detected with the VICTOR 3V Multilabel Plate Reader (Perkin Elmer). A standard curve was prepared according to the ADP Glo kit instructions to calculate the concentration of ADP based on luminescence. All data were normalized to the reading of the sample containing the Reaction Buffer alone with no proteins. The best results were obtained when the ADP Glo kit reagents were diluted twofold with the Reaction Buffer. Different standard curves were used for undiluted vs. diluted reagents.
The MBP1 peptide (with the sequence of SWYSYPPPQRAVC) was synthesized at the Protein Chemistry Technology Core at University of Texas Southwestern Medical Center (UTSW). MBP1 or Cdc20N was coupled to SulfoLink Coupling Resin (Thermo Fisher) according to the manufacturer’s instructions. Beads were washed in the ATPase Reaction Buffer. His6-Mad2 (0.5mgml−1) was incubated with the beads for 30–60min. After washes with the ATPase Reaction Buffer, TRIP13 and ΔN35-p31comet at indicated concentrations and 1mM ATP were added to the beads and incubated for another 30–60min. Proteins bound to beads were dissolved in SDS sample buffer, analyzed by SDS-PAGE, and stained with Coomassie blue. The Mad2-binding assay was performed in a similar manner except that ATPγS was used in place of ATP. All reactions were performed in the Reaction Buffer at room temperature.
All analytical ultracentrifugation (AUC) experiments were performed in the sedimentation velocity (SV) mode at 4°C in an Optima XL-I ultracentrifuge (Beckman-Coulter, Indianapolis, IN). Samples (400μl) were introduced into the sample sectors of chilled, charcoal-filled dual-sector Epon centerpieces that had been sandwiched between sapphire windows in standard AUC cell housings. An equal volume of AUC buffer (10mM HEPES, pH 7.4, 300mM NaCl, 5mM MgCl2, 1.357M glycerol, and 1mM TCEP) was placed in the reference sectors. One sample of monomeric bovine serum albumen (BSA, Sigma-Aldrich, St. Louis, MO) was also centrifuged to serve as a molar-mass standard, and to determine whether the calculated (see below) solution parameters were realistic. BSA had been purified over a Superdex 200 16/60 size-exclusion. When present, ATP or ATPγS was included at a concentration of 100μM in both the sample and reference buffers. The assembled cells were positioned in a chilled An50-Ti rotor, which was then put into the centrifuge and incubated at the experimental temperature under vacuum overnight. Next, the rotor was accelerated to 201,600g and concentration profiles were collected using the onboard absorbance optics tuned to 280nm. The data were collected until there was no evidence of sedimentation. Solution parameters (density, viscosity) were calculated in SEDNTERP74. Using SEDFIT75 and the c(s) methodology, we calculated the molar mass of BSA, which was within 10% of the known value (66,430Da). Thus, the calculated solution parameters were deemed acceptable, and SEDFIT was used to analyze the TRIP13 samples. Maximum-entropy regularization with a confidence level of 0.683 was used, and time-invariant noise elements in the data were subtracted76. An s-resolution of 100 was used. All sedimentation coefficients were converted to standard conditions (water, 20°C) using GUSSI77.
BubR1N (residues 1–370), His6-Cdc20 full-length, and dimeric Mad2 used in the APC/C assays were purified as described29. Uba1, UbcH10, and securin-Myc were purified as described78. Ube2S was a gift from Dr. Michael Rape (University of California at Berkeley). APC/C was isolated from Xenopus egg extracts with anti-APC3 antibody-coupled Protein A beads (Bio-Rad). Each reaction contained 5µl of beads that were incubated with 50µl of extract. The beads were supplemented with 3.8μM His6-Cdc20 for 1h and washed. The minimal functional MCC was formed by an incubation of BubR1N, Cdc20, and Mad2 for 20min. APC/CCdc20 beads were incubated with the pre-formed MCC and washed. The ubiquitination reaction was performed in the presence of securin-Myc for 30min, and quenched with the SDS loading buffer. The reaction mixtures were analyzed by SDS-PAGE followed by immunoblotting with the anti-Myc antibody (monoclonal; clone 9E10).
For TRIP13 inactivation of MCC, TRIP13 was diluted and mixed with p31comet and added to the MCC mixture in the presence of 5mM MgCl2 and 1mM ATP. The final reaction volume was 8µl. This reaction was incubated for 1h, and diluted to a final volume of 25µl by the XB buffer (10mM HEPES, pH 7.7, 100mM KCl, 0.1mM CaCl2, 1mM MgCl2, and 50mM sucrose) supplemented with 2mgml−1 BSA. The APC/CCdc20 beads (5µl) were then added, and the APC/C ubiquitination assay was performed. In the case of MCC bound to APC/C, the MCC was first incubated with the APC/CCdc20 beads, and then the TRIP13 reaction was performed.
The coordinates of the human TRIP13–ADP structure have been deposited to the Protein Data Bank with the PDB ID: 5WC2. All other relevant data are available upon request from the authors.
We thank Haydn Ball at the UTSW Protein Chemistry Technology Core for peptide synthesis, Shih-Chia Tso in the UTSW Macromolecular Biophysics Resource for assistance with analytical ultracentrifugation, and Michael Rape at the University of California at Berkeley for providing recombinant Ube2S. Use of Argonne National Laboratory Structural Biology Center beamlines at the Advanced Photon Source was supported by the US Department of Energy under contract DE-AC02-06CH11357. This study is supported by grants from the National Institutes of Health (GM107415 to X.L.), Cancer Prevention and Research Institute of Texas (RP160255 to X.L.) and the Welch Foundation (I-1932 to X.L. and I-1441 to H.Y.). H.Y. is an Investigator with the Howard Hughes Medical Institute.
M.L.B. conceived the project, expressed and purified TRIP13 proteins, performed all ATPase, Mad2 dissociation, and protein-binding assays, and wrote the initial draft of the paper. B.-C.J. performed crystallization and structure determination of human TRIP13. F.L. performed the NMR and H/D exchange experiments. B.L. performed the APC/C ubiquitination assays. E.B.Y. purified 15N-labeled Mad2 and performed the initial NMR experiment. Q.W. assisted in the acquisition and analysis of NMR data. C.A.B. performed the analytical ultracentrifugation experiments. H.Y. and X.L. supervised the project, analyzed the results, and wrote the paper with input from all authors.
The authors declare no competing financial interests.
Electronic supplementary material
Supplementary Information accompanies this paper at 10.1038/s41467-017-02012-2.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Hongtao Yu, Email: ude.nretsewhtuostu@uY.oatgnoH.
Xuelian Luo, Email: ude.nretsewhtuostu@ouL.naileuX.