Cep57 is a Mis12-interacting kinetochore protein involved in kinetochore targeting of Mad1–Mad2

The spindle assembly checkpoint (SAC) arrests cells in mitosis by sensing unattached kinetochores, until all chromosomes are bi-oriented by spindle microtubules. Kinetochore accumulation of the SAC component Mad1–Mad2 is crucial for SAC activation. However, the mechanism by which Mad1–Mad2 accumulation at kinetochores is regulated is not clear. Here we find that Cep57 is localized to kinetochores in human cells, and binds to Mis12, a KMN (KNL1/Mis12 complex/Ndc80 complex) network component. Cep57 also interacts with Mad1, and depletion of Cep57 results in decreased kinetochore localization of Mad1–Mad2, reduced SAC signalling and increased chromosome segregation errors. We also show that the microtubule-binding activity of Cep57 is involved in the timely removal of Mad1 from kinetochores. Thus, these findings reveal that the KMN network-binding protein Cep57 is a mitotic kinetochore component, and demonstrate the functional connection between the KMN network and the SAC.

Kinetochore targeting of Cep57 via Mis12. To further investigate whether the interaction between Mis12 and Cep57 is required for the kinetochore localization of the latter, we sought to disrupt the interaction with mutations in Mis12. In vitro pull-down assays using Mis12 mutants and Cep57 (1-242 amino acids; all expressed in E. coli and purified) showed that deletion of the amino acids 111-140 region of Mis12 abolished its interaction with Cep57 ( Supplementary Fig. 2a,b). In yeast two-hybrid assays, we further narrowed down the region and found that deletion of amino acids 131-140 was sufficient to disrupt the interaction ( Supplementary Fig. 2c,d). Then, we set out to determine the critical residues within the region or nearby by mutating some conserved and characteristic amino acids to glycine ( Supplementary Fig. 2e). We found that single amino-acid substitution (L132G) in Mis12 was enough to abolish its interaction with Cep57 in yeast two-hybrid assays ( Supplementary Fig. 2f); this was confirmed by in vitro pull-down assays with recombinant Mis12 point mutant (L132G) and Cep57 (1-242 amino acids) expressed in E. coli and purified (Fig. 2a). The L132G mutation of Mis12 did not markedly affect its kinetochore localization and that of some other KMN network components (DSN1, KNL1 and Hec1; Fig. 2b-i), which had been considered to require Mis12 for their kinetochore targeting 40 . However, the L132G mutant specifically led to decrease in the localization of Cep57 at kinetochores (by B79%), but not at spindle poles ( Fig. 2j-l), suggesting that interaction with Mis12 is required for Cep57 to efficiently anchor kinetochores.
Cep57 is involved in activation of the mitotic checkpoint. The KMN network is an important scaffold for the kinetochore accumulation of SAC components 1,26,27 . To determine whether Cep57, a KMN-associated protein, functions in the SAC, we monitored the mitotic progression of cells transfected with Cep57-siRNA followed by nocodazole treatment, which induced long-term activation of the SAC by unattached kinetochores in mitotic cells. Control HeLa cells were arrested in mitosis for a median time of 1,524 min (Fig. 3a,b; Supplementary Movie 1), while in Cep57-depleted cells, the time was reduced to 1,016 min ( Fig. 3a,b; Supplementary Movie 2); a reduction also occurred in  Mad1-or Mad2-depleted cells ( Fig. 3b; Supplementary Fig. 3a,b). Similar results were obtained when cells were transfected with Cep57-siRNA and treated with taxol, a drug that inhibited spindle dynamics and activated the SAC (Fig. 3c). Furthermore, the mitotic index was decreased in Cep57-depleted cells after treatment with nocodazole (10-100 nM; Supplementary Fig. 3c). Under treatment with 100 nM nocodazole, the mitotic index of Cep57-depleted cells was reduced to B49% from B77% in control cells, and siRNA-resistant Cep57 rescued this index to B65% ( Supplementary Fig. 3d,e). Consistently, we found an increased percentage of cells with multiple small nuclei after Cep57 depletion, and this was also rescued by siRNA-resistant Cep57 ( Supplementary Fig. 3f,g). Collectively, these data suggest that depletion of Cep57 attenuates the SAC activation induced by drugs that affect microtubules. Unattached kinetochores trigger a conformational change in cytoplasmic Mad2, and induce its binding to Cdc20 (ref. 8). This binding prevents the activation of APC/C and the degradation of its substrates 10,12 . Therefore, we investigated whether the weakened mitotic arrest in Cep57-depleted cells is mediated by the inhibition of Mad2-dependent Cdc20-APC/C. HeLa cells were synchronized by double-thymidine block and released into nocodazole after siRNA transfection (Fig. 3d). After released into nocodazole for 12 h, the mitotic cells were collected for immunoprecipitation with anti-Cdc20 antibody. The results showed that Cep57 depletion decreased the binding of Cdc20 to Mad2 (Fig. 3e). Then, after double-thymidine block, we collected cells after different durations of treatment with nocodazole to probe the APC/C substrates securin and cyclin B1. Compared with control cells, Cep57-depleted cells showed a more rapid decrease in the protein levels of both securin and cyclin B1 (Fig. 3f,g), indicating that Cep57 depletion relieves the nocodazole-induced and Mad2-dependent inhibition of APC/C.
Role of Cep57 in mitotic progression/chromosome segregation. Activation of the SAC delays anaphase initiation and ensures the equal distribution of chromosomes into two daughter cells 1,2 . To further determine whether Cep57, functioning during nocodazole-induced SAC activation, contributes to mitotic progression and chromosome segregation, we monitored (j) Immunostaining of Cep57 (red) and g-tubulin (purple) in metaphase HeLa cells expressing RNAi-resistant WT or L132G GFP-Mis12 after treatment with Mis12 siRNA. DNA was stained with DAPI (blue). Scale bars, 5 mm. (k,l) Quantification of the kinetochore (k) and spindle pole (l) signals of Cep57 from (j). The signal of cells expressing wild-type GFP-Mis12 was normalized to 1.0. More than 100 kinetochores from 10 cells were measured. The experiment was repeated three times. Data are mean±s.e.m. **Po0.01; NS, not significant (unpaired two-tailed Student's t-test).
Cep57-depleted HeLa and RPE1 cells by live-cell imaging microscopy, and found that with Cep57 depletion the average time from nuclear envelope breakdown to anaphase onset was shortened by 7 min in HeLa cells, and by 4 min in RPE1 cells relative to controls ( Considering that Cep57 is also localized to spindle poles, and its depletion results in an increased percentage of cells with multipolar spindles 34 , which may also induce chromosome lagging. To define whether kinetochore-localized Cep57 contributes to avoiding chromosome segregation errors, we first labelled g-tubulin in HeLa cells transfected with Cep57-siRNA and calculated the percentage of bipolar segregated cells with chromosome lagging, and found that it was significantly raised by B24%, and the siRNA-resistant Cep57 restored it by B15% (Fig. 4e,f). Similar results were obtained in RPE1 cells ( Fig. 4g; Supplementary Fig. 4b). Furthermore, the point mutant of Mis12 (L132G) that specifically reduced the kinetochore-localized Cep57 Taken together, our results suggest that Cep57 is required for mitotic timing control and correct chromosome segregation. Cep57 contributes to the recruitment of Mad1 to kinetochores. SAC signalling is considered to be initiated by the accumulation of the Mad1-Mad2 complex at kinetochores 4,8 , so we tested whether Cep57 is necessary for the kinetochore recruitment of this complex. Cep57 depletion not only reduced its own kinetochore signal by B90%, but also the signal of Mad1 (by B53%) and Mad2 (by B51%) in HeLa cells with nocodazole treatment (Fig. 5a-d), though the total protein levels of Mad1 and Mad2 did not change ( Supplementary Fig. 5a). SiRNA-resistant Cep57 restored the kinetochore signal of Mad1 to B88% (Fig. 5e,f). However, neither Mad1 nor Mad2 depletion affected the either protein level or kinetochore localization of Cep57 (Fig. 5a,b,g,h; Supplementary Fig. 5b,c). In taxol-treated cells, the kinetochore-targeting efficiency of Mad1 was also reduced by Cep57 depletion (by B51%; Fig. 5i,j). Thus, Cep57 is involved in the kinetochore recruitment of Mad1-Mad2.
Cep57 interacts with Mad1. To investigate the mechanism by which Cep57 is responsible for the kinetochore accumulation of Mad1-Mad2, we determined whether Cep57 was associated with Mad1-Mad2 using yeast two-hybrid assays, and the results showed that Mad1 bound to Cep57 ( Supplementary Fig. 6a,b). Immunoprecipitation assays with both endogenous and ARTICLE exogenous proteins also showed the interaction between Cep57 and Mad1 ( Fig. 6a; Supplementary Fig. 6c). The interaction of Cep57 with Mad2 was barely detectable unless Mad1 was present (Fig. 6b), which suggested that Cep57 is associated with the Mad1-Mad2 complex via Mad1. We further used purified recombinant Cep57 and Mad1 from HEK293T cells (Fig. 6c) and E. coli (Fig. 6d) to perform binding assays and the results showed that Cep57 directly bound to Mad1 in vitro. To determine which regions of Cep57 are responsible for the binding, we performed pull-down assays using truncated mutants of Cep57, and found that its C terminus (195-500 amino acids) interacted with Mad1 (Fig. 6e). Purified GST-Cep57 (195-500 amino acids) and Flag-Mad1 co-immunoprecipitated, further confirming the interaction of the C-terminal region of Cep57 with Mad1 (Fig. 6f). We also constructed truncated mutants of Mad1 to map the regions responsible for the binding to Cep57 (Fig. 6g). Immunoprecipitation assays showed that the N terminus of Mad1 (1-530 amino acids) was crucial for this interaction (Fig. 6h). Furthermore, short truncated mutants of the N terminus of Mad1 (1-175 and 351-530 amino acids) barely precipitated with Cep57, and the internal region mutant (176-350 amino acids) also showed a very weak binding affinity (Fig. 6i). In addition, among these mutants of Mad1, only the N-terminal mutant (1-530 amino acids) showed weak kinetochore localization, but its three short truncated mutants (1-175, 176-350 and 351-530 amino acids) or the C terminus (531-718 amino acids) did not show the localization (Fig. 6j; Supplementary Fig. 6d), suggesting that the structural integrity of Mad1 is important for its kinetochore targeting.

Role of Cep57 microtubule-binding activity in SAC silencing.
Given that Cep57 is a microtubule-binding protein, the binding is mediated by its C terminus 33 , and the same region binds to Mad1 (Fig. 6e,f), we sought to investigate whether the interaction of Cep57 with Mad1 is affected by the binding of Cep57 to microtubules. First, we performed microtubule co-sedimentation assays with purified Cep57, Mad1 and tubulin in vitro. The C terminus of Cep57 showed distinct co-sedimentation with microtubules ( Supplementary Fig. 7a), whereas Mad1 was not detected in the co-sediment (Fig. 7a). Then, we coupled the C-terminal Cep57 (expressed in bacteria and purified) to beads and performed pull-down assays in vitro. The interaction of Cep57 with Mad1 was reduced with increased binding of Cep57 to microtubules (Fig. 7b), indicating that the C-terminal Cep57 strongly binds to microtubules, and this binding inhibits the interaction of Cep57 with Mad1. These results suggest that microtubules competitively replace Mad1 binding to Cep57. In addition, the interaction of Cep57 with Mis12 did not alter the binding activity between Mad1 and Cep57 ( Supplementary  Fig. 7b), and the binding of Cep57 to microtubules did not affect the interaction of Cep57 with Mis12 ( Supplementary  Fig. 7c).
To further determine whether the microtubule-binding activity of Cep57 is required for checkpoint silencing in human cells, we generated a mutated Cep57, in which 12 positively charged residues in the C terminus were replaced with alanine (named Cep57-12A) to abolish its microtubule-binding activity and retain its interaction with Mad1 (Fig. 7c, Supplementary Fig. 7d). This mutant was not co-localized with microtubules and diffused throughout the cytoplasm of interphase HeLa cells after overexpression (Fig. 7d), while its kinetochore localization was not significantly reduced compared with that of the wild type in metaphase cells ( Supplementary Fig. 7e,f). In Cep57-12Aexpressing cells after endogenous Cep57 depletion, we still observed the Mad1-positive immunostaining signal at the metaphase kinetochores, but the signal almost disappeared at this stage in the wild-type Cep57-expressing cells (Fig. 7e-g). Furthermore, with Cep57-12A expression, the percentage of metaphase cells in prometaphase and metaphase cells was increased (by B11% relative to wild-type Cep57 expression; Fig. 7h). Together, these data suggest that the loss of microtubulebinding activity of Cep57 delays the removal of Mad1 from kinetochores and results in extended metaphase arrest.

Discussion
In this paper, we show that Cep57 is localized to kinetochores, and its N terminus binds to Mis12. Cep57 also interacts with Mad1 via its C terminus and participates in the accumulation of Mad1-Mad2 at kinetochores, while the microtubule-binding activity of Cep57 may contribute to the timely removal of Mad1 from kinetochores (Fig. 8).
We previously reported that Cep57 is a component of the spindle pole and midzone, and functions in spindle pole architecture and central spindle microtubule organization 34,35 . Previous findings have revealed that xCep57 is a kinetochore component 31 . Mass spectrometry has shown that chicken Cep57 occurs in isolated mitotic chromosomes 44 . We show here that human Cep57 is localized at kinetochores, suggesting the conservation of kinetochore localization in the Cep57 family.
At kinetochores, the KMN network is considered to contain core microtubule-binding sites 1 . Here we provide evidence that Cep57, as a KMN network-binding protein, is involved in the kinetochore recruitment of Mad1-Mad2 in human cells. Our results showed that significant depletion of Cep57 (B90%) still left Mad1-Mad2 signals (B50%) at kinetochores (Fig. 5a-h), indicating that Cep57 is not the unique kinetochore recruiter of Mad1-Mad2. In fact, multiple proteins are involved in the kinetochore accumulation of Mad1-Mad2, and the underlying molecular mechanisms are thought to be complicated and to vary in different species 14,[23][24][25]45 . In Caenorhabditis elegans and yeast, Mad1 has been shown to interact with Bub1, which contributes to the kinetochore accumulation of Mad1 in Caenorhabditis elegans cells 24,25 . In human cells, Hec1 is required for kinetochore localization of Mad1-Mad2 (refs 23,46). It has been identified as a Mad1-interacting candidate by a yeast two-hybrid screen, but recombinant Hec1 and Mad1 proteins barely bind to each other in vitro 23 . Nuf2, a binding partner of Hec1 and a subunit of the Ndc80/Hec1 complex, has also been reported to function in the kinetochore accumulation of Mad1-Mad2 (refs 14,46). Though there is no evidence of an interaction between Nuf2 and Mad1-Mad2, depletion of Nuf2 reduces Mad1-Mad2 at kinetochores 14 . Moreover, depletion of the RZZ complex protein ZW10 has also been found to weaken Mad1-Mad2 signals at kinetochores 19 . Similar to Nuf2, ZW10 is not a direct binding partner of Mad1-Mad2, but it may affect the kinetochore localization of Mad1, as Mad1-binding site(s) created by related kinetochore component(s) 19 . Cep57 is a Mis12-binding protein that is closely localized to RZZ and Ndc80/Hec1 complexes. xCep57 is found to show interaction with Ndc80/Hec1 and Zwint-1 (a binding factor and a kinetochore recruiter of ZW10) in Xenopus 43,47 . So it is possible that Hec1, Nuf2, Bub1, ZW10, Cep57 and other related proteins cooperate to recruit Mad1-Mad2 at the outer kinetochores. This speculation is also supported by previous 48 and our results that the structural integrity of Mad1 is essential for its fully efficient kinetochore targeting. The N terminus of Mad1 (1-530 amino acids) is essential for kinetochore localization, but the localization is weak, and its shorter truncated mutants are barely detectable at kinetochores. The C terminus of Mad1 (531-718 amino acids) is not localized to kinetochores, but is critical for the efficient accumulation of Mad1 at kinetochore, in line with the report that deletion of the C-terminal domain diminishes Mad1 kinetochore targeting 48 . Therefore, it is likely that multiple regions of Mad1 are required for its kinetochore accumulation via multiple recruiters. Whether Hec1, Nuf2, ZW10, Bub1, Cep57 and even other proteins at the outer kinetochores cooperate to participate in the kinetochore targeting of Mad1-Mad2 in human cells is an interesting issue.
Our results show that Cep57 is associated with Mad2 via Mad1. While in yeast two-hybrid assays, the yeast expressing both bait-Cep57 and prey-Mad2 proliferated slowly in selective medium ( Supplementary Fig. 6a), suggesting a possible weak direct interaction between Cep57 and Mad2, though we could not exclude the possibility that yeast Mad1 mediated the interaction. The current model of the SAC activation shows that closed Mad2 is recruited to unattached kinetochores by Mad1, and acts as a catalytic template to trigger the transition from cytoplasmic open Mad2 to closed Mad2 (refs 1,2,6,8,10). However, it is still unclear how kinetochore-localized Mad2 induces the transformation of cytoplasmic Mad2, and whether other factors are required for Mad2 to function as a template 49,50 . The possible connection between Cep57 and Mad2 needs further investigation.
Checkpoint silencing is mainly induced by the removal of essential checkpoint components (such as Mad1, Mad2 and BubR1) from kinetochores, and these components may be driven forward to the spindle pole along kinetochore-connected microtubules by the dynein-dynactin complex and other related proteins [51][52][53][54][55][56] . When the spindle microtubules are searching for kinetochores and the kinetochore-microtubule attachment has The RZZ and Ndc80/Hec1 complexes are responsible for the accumulation of Mad1-Mad2 to kinetochores by direct or indirect means in human cells 14,15,19,23 . Cep57 also acts as one of the anchoring factors of the Mad1-Mad2 complex by the interaction with Mad1 at the kinetochores (top). When the kinetochores are captured by microtubules, Cep57 binds to microtubules along with other outer kinetochore proteins including Hec1, Nuf2 and KNL1 (refs 1,30). Thus, Mad1-Mad2 is disassociated from the kinetochores, which results in SAC silencing (bottom). NT, N terminus. CT, C terminus.
Considering our results that Cep57 shows stronger binding activity to microtubules than to Mad1, it is possible that with the establishment of kinetochore-microtubule attachment, the kinetochore-localized Cep57 binds to microtubules, which facilitates the dissociation of Mad1-Mad2 from kinetochores and contributes to the SAC silencing.
Many kinetochore proteins (such as Hec1, Nuf2 and KNL1) show microtubule-binding activity, and they may have diverse and redundant mechanisms to ensure proper microtubulekinetochore attachment. Hec1 and Nuf2 are thought to be the most important for load-bearing attachments and microtubule plus-end dynamics, since their N termini have a calponinhomology microtubule-binding domain that is similar to the plus-end tracking protein EB1 (end-binding 1) 1,30,61 . However, KNL1 microtubule-binding activity, achieved by a short motif enriched with several positively charged residues, is dispensable for the formation of microtubule-kinetochore attachment and instead functions in silencing the spindle checkpoint 28 . The results present here indicate that Cep57 shows a microtubulebinding mode similar to KNL1, and Cep57 depletion results in no significant defects in microtubule-kinetochore attachment in human cells, which suggests that Cep57 does not play the central role in microtubule-kinetochore attachment, but more likely functions as one of the mediators of checkpoint silencing.
The SAC is associated with mosaic variegated aneuploidy syndrome (MVA), a disease characterized by growth retardation, childhood tumorigenesis, microcephaly and constitutional mosaicism induced by chromosomal gains and losses 62,63 . A previous report showed that gene mutations of the spindle checkpoint component BubR1 causes MVA due to defective spindle checkpoint activation 62 . Biallelic loss-of-function CEP57 mutations also cause MVA, but the molecular mechanism by which such mutations induce aneuploidy syndrome remains obscure 64 . Our findings on the recruitment of Mad1-Mad2 by Cep57 reveal that it functions in the SAC activation and ensures correct chromosome segregation. This may provide the molecular mechanism by which the CEP57 mutation, similar to BubR1, causes MVA. Antibodies. His-tagged Cep57 protein was injected into mice and rabbits to generate polyclonal antibodies. The anti-Cep57 mouse antibody was used for western blotting (WB, 1:500) and immunofluorescence (IF, 1:50). The anti-Cep57 rabbit antibody was raised and purified using the GST-Cep57 (332-500 amino acids) recombinant protein 34  Cell cultures and treatments. HeLa, RPE1 and HEK293T cells were cultured in 10% fetal bovine serum (Gibco) containing DMEM (Gibco) at 37°C under 5% CO 2 . JetPEI (Polyplus transfection) or Lipofectamine 2000 (Invitrogen) was used to transfect cells according to the manufacturer's instructions.

Methods
For the double-thymidine and nocodazole block, HeLa cells were treated with thymidine for 16 h, released for 12 h and were blocked again for 16 h, and then released into nocodazole.
GST pull-down, immunoprecipitation and western blotting. Cell lysates used for GST pull-down and immunoprecipitation assays were obtained as follows: cells were washed with PBS and then lysed in lysis buffer (150 mM NaCl, 1 mM MgCl 2 , 50 mM Hepes, pH 7.4, 1 mM EGTA and 0.5% Triton X-100) containing protease inhibitors. The lysates were obtained by collecting the supernatant after centrifugation for 10 min at 20,000g at 4°C.
For immunoprecipitation, the cell lysates were incubated with antibodies at 4°C for 4 h, and were mixed with protein A-Sepharose beads (Amersham Biosciences) for 2 h. The beads were further washed with lysis buffer, and were boiled for 5 min in protein-loading buffer (with SDS).
For GST pull-down assays, the GST and GST-tagged proteins were incubated with Glutathione Sepharose 4B beads (Amersham Bioscience) for purification. Then, the beads were washed five times with lysis buffer, and incubated with the lysates for 4 h at 4°C. The beads were pelleted and washed five times, and then the sample was boiled with protein-loading buffer (with SDS) for 5 min.
For western blotting, SDS-polyacrylamide gel electrophoresis was used to separate the proteins, and then they were transferred to a polyvinylidene difluoride membrane (Millipore). The membrane was sequentially incubated with primary antibodies and horseradish peroxidase-conjugated protein A (Jackson ImunoResearch) or secondary antibodies. Uncropped scans of typical blots are presented in Supplementary Fig. 8.
For in vitro binding assays, B0.2 mg of MBP, MBP-Cep57 (1-242 amino acids) and MBP-Cep57 (151-500 amino acids) were coupled with beads and were incubated with B0.2 mg of purified and eluted GST-Mis12 or GST-Mad1 (176-718 amino acids), and then the beads were washed five times with lysis buffer. Then, the samples were boiled for 5 min and analysed with western blotting.
Microtubule polymerization and co-sedimentation assays. Microtubules were assembled at 30°C for 30 min in BRB80 buffer (100 mM PIPES, 1 mM MgSO 4 , 2 mM EGTA, 1 mM GTP, pH 6.8) stabilized with taxol (20 mM). For microtubule pull-down assays, the taxol-stabilized microtubules were incubated with beadcoupled Cep57 proteins (expressed in bacteria and purified) in BRB80 buffer at room temperature for 1 h, and then the beads were washed three times with BRB80 buffer before boiling. For microtubule co-sedimentation assays, Cep57 expressed in bacteria and purified, and was incubated with assembled microtubules, and the samples were centrifuged at 100,000g for 10 min at 25°C. The supernatant and pellet were boiled and separated by SDS-polyacrylamide gel electrophoresis.
Immunofluorescence and time-lapse microscopy. Cells were fixed and permeabilized in methanol for 5-10 min at À 20°C, and incubated overnight at 4°C with primary antibodies in PBS containing 4% bovine serum, followed by staining with secondary antibodies and 1 mg ml À 1 4,6-diamidino-2-phenylindole. A confocal microscope (LSM-710 NLO, Zeiss) equipped with a Â 100/1.40 numerical aperture (NA) objective lens and a super-resolution confocal microscope (Leica TCS SP8 STED Â 3) equipped with a Â 100/1.4 NA objective lens were used to observe fixed cells. Three-dimensional super-resolution images were captured using a three-dimensional structured illumination microscope with the N-SIM System (3D-SIM, Nikon). Two time-lapse microscopes, a spinning-disk PerkinElmer UltraView VoX (Nikon) equipped with a Â 40/0.9 NA objective lens and an API Delta Vision Elite (Applied Precision) equipped with a Â 60/1.40 NA objective lens, were used to visualize the live cells. For time tracking, cells were observed in a chamber at 37°C under 5% CO 2 and images were post-processed with Velocity (Nikon) and Delta Vision Softworx (Applied Precision) software. Yeast two-hybrid assays. The yeast two-hybrid assays were performed according to the Matchmaker Two-Hybrid System Handbook. Mad1, Mad2, Mis12, Cep57 and Cep57R cDNA were amplified from HeLa cell cDNA library and inserted into bait vector pGBKT7 (Clontech Laboratories) and prey vector pGADT7 (Clontech Laboratories). The bait and prey vectors were co-transformed into the AH109 yeast strain, and sequentially plated onto double-(Trp À and Leu À ) and quadruple-(Trp À , Leu À , His À and Ade À ) selective media for 2-5 days at 30°C.
Statistical analysis. The fluorescence intensity was measured using Scion Image and Image J software (National Institutes of Health). Statistical analyses were performed using SPSS and GraphPad Prism 5 software. The unpaired two-tailed Student's t-test was used to calculate statistical significance.