Two-step colocalization of MORC3 with PML nuclear bodies

Eukaryotic cells have several organelles that are segregated by a lipid bilayer, including the endoplasmic reticulum, Golgi complex and lysosomes, amongst others. The nucleus is the largest of these organelles and contains many functional compartments unprotected by a lipid bilayer, such as chromosome territories, nucleoli, speckles, Cajal bodies and promyelocytic leukemia nuclear bodies (PML NBs) (Handwerger and Gall, 2006; Lamond and Earnshaw, 1998). These nuclear subdomains are systematically arranged to react effectively during cellular responses to external stimuli, transcription, cellular stress and DNA repair. However, little is known about the molecular mechanisms involved in the formation of the nuclear domains (NDs) and the recruitment of proteins into the NDs. Only nucleostemin is known to shuttle between the nucleolus and nucleoplasm in a GTPase-dependent manner to regulate cell-cycle progression (Tsai and McKay, 2005).

The PML NB is a nuclear domain mainly formed by the PML protein, originally identified in acute promyelocytic leukemia (APL), in which it is fused to the retinoic acid receptor-α (RARA) (de The et al., 1991; Kakizuka et al., 1991; Pandolfi et al., 1991). Many proteins are constitutively, but more often transiently, localized in PML NBs, where they regulate a variety of cellular processes such as DNA-damage response, transcription, apoptosis and cellular senescence (Bernardi and Pandolfi, 2007). PML is a member of a tripartite motif (TRIM) family containing an RBCC motif (a RING domain, two B-boxes and a coiled-coil domain). PML is also covalently modified at three lysine residues by three members of the small ubiquitin-like modifier (SUMO) family (SUMO1, SUMO2 and SUMO3), and it has a short SUMO-interacting motif (SIM), which contains a hydrophobic core [(V/I/L)-x-(V/I/L)-(V/I/L)] flanked at the N- or C-terminal by acidic and/or serine residues (Geiss-Friedlander and Melchior, 2007; Minty et al., 2000; Shen et al., 2006). SUMO is attached to most substrates at the ε-amino group of a lysine residue in the ψKxE/D sequence (ψ, hydrophobic amino acid; K, lysine; x, any amino acid; E, glutamic acid; D, aspartic acid). The SUMOylation of PML is necessary for the formation of mature PML NBs. SUMOylated PML interacts with PML itself and with other proteins such as Daxx that are recruited to PML NBs through its SIM domain (Bernardi and Pandolfi, 2007; Lin et al., 2006; Zhong et al., 2000).

MORC3 (microrchidia family CW-type zinc-finger 3), also known as NXP-2, is one of the MORC-family proteins, which are characterized by three conserved domains consisting of a GHL (Gyrase B, Hsp90 and MutL)-ATPase domain at the N-terminus (Dutta and Inouye, 2000), a CW-type zinc-finger domain containing four conserved cysteines and two tryptophans in the middle portion (Perry and Zhao, 2003) and a coiled-coil dimerization domain at the C-terminus (Inoue et al., 1999; Kimura et al., 2002). We have reported previously that the MORC3 protein is localized to PML NBs in an ATPase-dependent manner (Takahashi et al., 2007). MORC3 recruits p53 into PML NBs and activates p53 to induce cellular senescence in normal human and mouse fibroblasts (Takahashi et al., 2007). The GHL-ATPase component has diverse functions, but the basic architecture and ATP-dependent conformational changes in the proteins of the GHL-ATPase family are well conserved (Dutta and Inouye, 2000; Sacho et al., 2008; Shiau et al., 2006). The ATPase cycle drives protein conformational changes from an ATP-free ‘open’ state to a ‘closed’ state where the ATPase domain interacts intermolecularly by ATP-binding and returns the protein to the ‘open’ state by ATP-hydrolysis and release of the bound adenine nucleotide (Richter and Buchner, 2006; Sacho et al., 2008). The GHL-ATPase has a variable middle domain, which in Hsp90 binds to client proteins (Richter and Buchner, 2006), and contains the active site tyrosine responsible for DNA cleavage in type II DNA topoisomerase (Baird et al., 1999). MORC3 also has a CW-type zinc-finger domain between the N- and C-terminal regions. The CW domain is structurally similar to the plant homeo domain (PHD) finger, which has been identified as a module for binding methylated histone H3 (Iyer et al., 2008). Therefore, the CW domain is predicted to have a role in DNA binding (Perry and Zhao, 2003), in reading the histone code or in recruiting other proteins to the nucleosome (Iyer et al., 2008).

In this study, we found that MORC3 can form PML-independent nuclear domains, and we examined the molecular mechanisms involved in the PML-independent formation of MORC3 NDs and the association of MORC3 with PML.

We have previously shown that the MORC3 nuclear domain (ND) formation is suppressed in the APL cell line, in which PML NBs are broken down by the expression of PML-RARA, and also in PML-knockdown Saos-2 cells (Takahashi et al., 2007). However, we detected the formation of MORC3 nuclear domains (NDs) in Pml^−/− as well as Pml^+/+ mouse embryonic fibroblasts (MEFs) (Fig. 1A). Overexpressed MORC3 also formed NDs in Pml^+/+ and Pml^−/− MEFs (Fig. 1B). In MEFs, only 5% of MORC3 NDs formed independently of PML and 8% of cells had PML-independent MORC3 NDs (Fig. 1A,D). Interestingly, in mouse bone marrow cells, 69% of MORC3 NDs formed independently of PML and 78% of cells had PML-independent MORC3 NDs (Fig. 1C,D). In bone marrow cells containing PML-independent NDs, 35% of cells also showed MORC3 NDs formation in PML NBs. A similar percentage of total splenocytes and CD11b⁺ cells also had PML-independent and PML-dependent MORC3 NDs that formed in a cell-type-independent manner (supplementary material Fig. S1). These results suggest that MORC3 can form PML-independent NDs and that it colocalizes with PML via a two-step molecular mechanism: step one involving the formation of MORC3 NDs and step two involving the association of MORC3 with PML NBs.

To analyze the molecular mechanisms involved in MORC3 ND formation, as well as a previously constructed ATPase domain mutant (E35A) (Takahashi et al., 2007), we first prepared expression vectors carrying a ΔCC mutant deficient in the C-terminal coiled-coil domain, and an EΔCC mutant that has both the E35A and the coiled-coil domain deficiency (Fig. 2A), and examined the localization of the MORC3 mutants in HeLa cells (Fig. 2B). Cells carrying the wild-type (WT) construct formed NDs, but all of the mutants showed diffuse localization over the entire nucleus, regardless of the formation of intact PML NBs.

To characterize the diffuse localization of the MORC3 mutants further, we treated the mutant-expressing cells step-by-step with the buffers indicated to extract the nuclear fractions and detected the results by immunostaining (Fig. 2B) and western blotting (Fig. 2C). Although the ND structure of MORC3 WT was maintained upon treatment with CSK buffer containing 0.5% Triton X-100 and high-salt buffer containing 250 mM ammonium sulfate, these domains were disrupted by DNaseI treatment, which resulted in the diffusion of MORC3 throughout the nucleus; however, PML was still able to form NBs (Fig. 2B). MORC3 WT was washed out of the nucleus by further treatment with RNaseA (data not shown). These results suggest that MORC3 WT associates with DNA or nucleosomes and that when the domains are disrupted by DNaseI treatment, MORC3 binds to the nuclear matrix. Most of the ΔCC mutant MORC3 was washed out of the nucleus with CSK buffer (Fig. 2B) and was extracted in the CSK buffer fraction. By contrast, most of the E35A mutant MORC3 could not be extracted with buffer containing DNaseI and was contained in the pellet (the nuclear matrix) fraction (Fig. 2C). The EΔCC mutant showed the same localization pattern as E35A and was just detectable in the nuclear matrix fraction. Therefore, the ATPase domain containing the E35A mutation binds to the nuclear matrix.

Because MORC3 NDs were disrupted by DNaseI treatment, we predicted that MORC3 must have a DNA-binding or nucleosome-binding domain. Other GHL-ATPase proteins have a functional middle domain such as the client binding domain in Hsp90 (Richter and Buchner, 2006) and a DNA cleavage domain in Type II DNA topoisomerase (Baird et al., 1999). We predicted that the CW domain should be a domain implicated in DNA or nucleosome binding, and so we constructed a MORC3 expression vector carrying a CW-domain mutation in which alanine 419 was substituted for tryptophan (W419A) (Fig. 2A). Despite containing the intact ATPase and coiled-coil domains, the CW-domain mutant W419A showed diffuse localization over the whole nucleus and was almost entirely washed out of the nucleus by CSK buffer, but a small part of the W419A mutant still colocalized with PML NBs (Fig. 2B). Most of the W419A protein was extracted in the CSK buffer fraction similarly to the ΔCC mutant (Fig. 2C). Another CW-domain-conserved cysteine mutant (C416A) also exhibited the same localization as W419A (data not shown).

We have indicated previously that MORC3 WT shows diffuse localization upon depletion of ATP (Takahashi et al., 2007). To confirm whether diffuse MORC3 WT binds to the nuclear matrix in the same way as MORC3-E35A, we incubated FLAG-MORC3-expressing HeLa cells in ATP-depletion medium and treated them with CSK buffer (Fig. 2D). Similarly to the E35A mutant, the diffuse MORC3 WT was not washed out of the nucleus by CSK buffer. Therefore, ATP-free MORC3 is localized diffusely throughout the nucleus and binds to the nuclear matrix. A mutation of the conserved glutamic acid in motif I in other GHL-ATPases has been shown to eliminate ATP hydrolysis, with little or no effect on ATP binding (Ban et al., 1999; Ban and Yang, 1998; Tran and Liskay, 2000). However, MORC3-E35A was still localized diffusely to the nuclear matrix by ATP depletion (data not shown). Furthermore, we analyzed the localization of D67N and G101A, mutations of the conserved residues in motifs II and III, respectively, which are known to affect ATP binding in other GHL-ATPases (supplementary material Fig. S2) (Panaretou et al., 1998; Tran and Liskay, 2000). D67N and G101A were localized to the NDs and the entire nucleus, respectively. ATP depletion caused D67N to diffuse more quickly than the WT, whereas the diffuse localization of G101A did not change. Thus, we predicted that ATP binding is essential for the formation of NDs. We also confirmed that W419A binds to the nuclear matrix upon ATP depletion because of its intact ATPase domain (data not shown).

To confirm whether ATP binding was required to form MORC3 NDs, we performed in situ MORC3 ND diffusion analysis using the non-hydrolysable ATP analogue AMP-PNP. When FLAG-MORC3 WT-expressing HeLa cells were incubated in buffer containing AMP-PNP, we found that MORC3 NDs were maintained (Fig. 3A). By contrast, ND-localized MORC3 was eliminated from nuclei by treatment both with, and without, ATP. Interestingly, AMP-PNP also blocked the diffusion of MORC3 during incubation in DNaseI-containing buffer (Fig. 3B). These results suggest that ATP binding is essential for the formation of NDs and that MORC3 detaches from NDs upon subsequent ATP hydrolysis.

In most GHL-ATPase family members, the C-terminal coiled-coil domain constitutively homodimerizes, and the ATPase domains interact with each other in an ATP-binding-dependent manner (Dutta and Inouye, 2000; Richter and Buchner, 2006). To test whether MORC3 forms a homodimer, we coexpressed FLAG-MORC3 WT or its mutants (Fig. 4A, green) with HA-MORC3 WT (Fig. 4A, red) in HeLa cells and examined the interaction by immunostaining (Fig. 4A) and immunoprecipitation analyses (Fig. 4B). HA-MORC3 formed NDs in FLAG-MORC3 WT-expressing cells (Fig. 4A, top row of images), but showed diffuse localization in E35A-mutant-expressing HeLa cells (Fig. 4A, second row). HA-MORC3 was effectively coimmunoprecipitated with FLAG-MORC3 (Fig. 4B), as previously reported (Takahashi et al., 2007). Although the ΔCC mutant was still colocalized with HA-MORC3 WT in the NDs (Fig. 4A, third row), the deficiency in the coiled-coil domain clearly diminished their mutual interaction (Fig. 4B). Furthermore, the EΔCC mutant revealed a different localization pattern to HA-MORC3 WT (Fig. 4A, forth row), and HA-MORC3 WT was not absolutely coimmunoprecipitated with the FLAG-EΔCC mutant (Fig. 4B). These results suggest that the coiled-coil domain is required for homodimerization and that the ATPase domains interact with each other in an ATP-binding-dependent manner, as in other GHL-ATPases.

The CW-domain mutant W419A interacted with MORC3 WT in NDs in a manner that was similar to the ΔCC mutant (Fig. 4A, fifth row of images). Therefore, the results in Figs 2 and 4 suggest that W419A exists in a soluble nucleoplasmic space because of its inability to bind DNA or nucleosomes.

Many proteins localized in PML NBs are SUMOylated (Bernardi and Pandolfi, 2007), and proteomic analysis has been used to show that MORC3 is also SUMOylated (Zhao et al., 2004). We have also cloned SUMO1 and a SUMO E2-conjugating enzyme, UBC9/UBE2I, using MORC3 as bait in a yeast two-hybrid screen (data not shown). To examine whether MORC3 was indeed SUMOylated, FLAG-MORC3 was coexpressed with HA-SUMO1 in HeLa cells. We detected three major SUMOylated bands (Fig. 5B), which were also detected with an anti-HA antibody in samples immunoprecipitated with anti-FLAG-antibody (supplementary material Fig. S3). To identify the SUMOylation sites of MORC3, we looked for a typical SUMOylation motif within the MORC3 peptide sequence and replaced each lysine of five candidate SUMOylation sites with arginine (K139R, K650R, K740R, K794R and K855R). We detected a reduction in SUMOylation signals in only the K650R and K740R mutants (Fig. 5B); however, SUMOylation was not completely abolished in the 2KR mutant (double mutant of K650R and K740R) (Fig. 5C). Next, we introduced new mutations at candidate sites and at other predicted sites in the 2KR double mutant to generate 3KR (K650R, K740R and K794R) and 4KR mutants (K650R, K651R, K740R and K794R), finally resulting in the 5KR mutant (K597R, K650R, K651R, K740R and K794), which was mutated at five SUMOylation sites. The 5KR mutant was not completely modified by SUMO1 (Fig. 5C). All sites mutated in the 5KR mutant were essential for SUMOylation. In mammalian cells, some proteins are also modified with SUMO2 or SUMO3 as well as SUMO1 (Geiss-Friedlander and Melchior, 2007). We found that MORC3 WT was modified with SUMO3 in HeLa cells coexpressing FLAG-MORC3 WT and HA-SUMO3 (supplementary material Fig. S3). To determine whether MORC3 was modified with SUMO1, we generated an anti-SUMO1 polyclonal antibody that specifically recognizes SUMO1 but not SUMO3 (supplementary material Fig. S3), which confirmed that FLAG-MORC3 was modified with endogenous SUMO1 in samples immunoprecipitated with the anti-FLAG antibody (Fig. 5D). Interestingly, the FLAG-MORC3 5KR mutant formed a larger number of smaller NDs than the WT (Fig. 5E).

SUMOylated proteins often interact with other proteins through their SUMO-interacting motif (SIM) (Minty et al., 2000). PML has a SIM site (amino acids 556-559) and three SUMOylation sites (K65, K160 and K490) (Kamitani et al., 1998) (supplementary material Fig. S4A) and it interacts with other SUMOylated proteins such as Sp100, and SIM-containing proteins such as Daxx, through the SUMO-SIM interaction (Bernardi and Pandolfi, 2007; Shen et al., 2006). To clarify whether the SUMOylation of MORC3 is involved in the colocalization of MORC3 with PML, we constructed a PMLI-AS mutant with a mutation at the SIM site (supplementary material Fig. S4A,B) and examined the interactions between FLAG-MORC3 WT and HA-PMLI WT, the FLAG-MORC3-5KR mutant and HA-PMLI WT, FLAG-MORC3 WT and the HA-PMLI-AS mutant, or FLAG-MORC3-5KR and HA-PMLI-AS using immunofluorescence (Fig. 6A). Whereas MORC3 WT completely colocalized with PMLI WT (Fig. 6A, top row), MORC3-5KR (Fig. 6A, second row) and PMLI-AS (Fig. 6A, third row) formed NDs independently of PMLI and MORC3, respectively. Next, we confirmed the MORC3-PMLI interaction using immunoprecipitation analysis (Fig. 6B). The association of MORC3-5KR with PMLI WT, and of MORC3 WT with PMLI-AS, was reduced compared with that between MORC3 WT and PMLI WT. Although MORC3-5KR colocalized with PMLI-AS for unknown reasons (Fig. 6A, forth row), the association of MORC3-5KR with PMLI-AS could not be detected by immunoprecipitation analysis (Fig. 6B). Unlike PMLI, PML splicing variants III and IV, which retain the SIM site, partially colocalized with MORC3 WT, and PMLIII- and PMLIV-independent MORC3 NDs and MORC3-independent PMLIII and PMLIV NBs were observed (supplementary material Fig. S5A).

In addition, we investigated the influence of diffusely localized MORC3-E35A on PMLI localization. Endogenous PML and other overexpressed PML variants III and IV formed NBs in cells expressing the MORC3-E35A mutant (Fig. 2B and supplementary material Fig. S5B) (Takahashi et al., 2007). However, PMLI WT colocalized diffusely with the E35A mutant throughout the nucleus (Fig. 6C; top row of images). This colocalization was almost completely disrupted in E-5KR and PMLI WT-expressing cells (Fig. 6C; second row of images), in E35A and PMLI-AS-expressing cells (Fig. 6C; third row of images), and in E-5KR and PMLI-AS-expressing cells (Fig. 6C; forth row of images). To confirm whether MORC3 associated with PMLI via a SUMO-SIM interaction, we disrupted the SUMO-SIM interaction by immunoprecipitation without the isopeptidase inhibitor, N-ethylmaleimide (NEM) (Fig. 6D). HA-PMLI was not coimmunoprecipitated with FLAG-MORC3 in the absence of NEM. Furthermore, to examine whether SUMOylation was essential for the association of MORC3 with PMLI, SUMOylation was suppressed by UBE2I siRNA. This resulted in the efficient suppression of UBE21 protein expression (supplementary material Fig. S6A) and a marked reduction in the level of SUMOylated MORC3 (supplementary material Fig. S6B). In the siRNA-transfected cells, the diffuse localization of MORC3 increased and MORC3 formed NDs independently of PMLI (Fig. 6E). We also observed cells in which MORC3 was dispersed throughout the nucleus (data not shown). These results suggest that the SUMOylation of MORC3 is important for the association between PMLI and MORC3. However, the PMLI SUMOylation mutant colocalized with MORC3 WT and coimmunoprecipitated with FLAG-MORC3 (supplementary material Fig. S7). Therefore, the SUMOylation of PMLI was not strongly involved in the PMLI-MORC3 interaction. To examine whether MORC3, unassociated with PML NBs, was SUMOylated in mouse bone marrow cells, we coimmunostained cells with anti-MORC3, anti-PML and anti-SUMO1 antibodies (Fig. 6F). Most of the SUMO1 colocalized with PML NBs, but did not accumulate in MORC3 NDs that did not colocalize with PML NBs.

Finally, to analyze the dynamics of MORC3 NDs in live cells, we followed the re-formation of MORC3 NDs by allowing ATP recovery in cells that showed diffuse nuclear localization of MORC3 after ATP depletion. As previously described, EGFP-MORC3 forms NDs more rigidly than FLAG-MORC3 and is not dispersed from NDs upon ATP depletion (Takahashi et al., 2007). We generated a vector expressing MORC3 tagged with monomeric EGFP. Although mEGFP-MORC3 was not dispersed in HeLa cells by sodium azide alone in glucose-free medium, we observed diffuse localization of mEGFP-MORC3 in some of the mCherry-PMLI-expressing CHO-K1 cells after a 3 hour incubation with 10 mM sodium azide. The cells exhibiting diffuse localization of mEGFP-MORC3 in ATP-depletion medium were then transferred into normal medium to allow ATP recovery. In many of the cells that exhibited diffuse localization of MORC3, most MORC3 ND gradually reformed in the PML NBs that had accumulated MORC3 during ATP depletion (Fig. 7). After ATP recovery, MORC3 ND appeared in the PML-NB-independent region (arrow in Fig. 7) and the ND signal gradually increased in brightness. Forty minutes later, MORC3 ND fused to PML NB in which MORC3 had already localized.

In this study, because we found that MORC3 can form NDs in a PML-independent manner in Pml^−/− fibroblasts and mouse hematopoietic cells, we analyzed the molecular mechanisms controlling the PML-independent formation of MORC3 NDs. We reported previously that MORC3-ATPase is involved in the localization of MORC3 to PML NBs (Takahashi et al., 2007). In this study, we determined that the colocalization of MORC3 with the PML NBs occurs in two distinct steps: the formation of MORC3 NDs, and the association of MORC3 NDs with PML NBs (Fig. 8). The binding of ATP to the ATPase domain stimulates the formation of NDs, whereas the subsequent ATP hydrolysis, and release of the adenine nucleotide, induces the diffusion of MORC3. The association of MORC3 NDs with PML NBs is regulated via the interaction between SUMO1 covalently linked to MORC3 and the SIM in PMLI. In fact, we found that in mouse bone marrow cells, SUMO1 did not accumulate in MORC3 NDs unless they colocalized with PML NBs. This result suggests that the SUMOylation of MORC3 is required for its colocalization with PML. Furthermore, in cells expressing mEGFP-MORC3 and mCherry-PMLI, we also found that MORC3 ND colocalized with PML NBs in two steps (Fig. 7). In the first step, the ND appeared in the PML-independent nuclear region before gradually increasing in brightness and size. In the second step, the ND fused to the PML NB to which MORC3 had already localized. This result strongly supports a new model for MORC3 localization, which we have termed the ‘two-step colocalization mechanism of MORC3 with PML NBs’. However, many MORC3 proteins accumulated in PML NBs without the formation of PML-independent MORC3 NDs. Therefore, MORC3 proteins might first be SUMOylated before accumulating in NDs that colocalize with PML NBs upon ATP binding (Fig. 8, alternative step). Although most MORC3 proteins were localized to PML NBs, only a few MORC3 proteins were SUMOylated (Fig. 5D). We predict that SUMOylation of only a small number of MORC3 proteins are involved in the association of MORC3 NDs with PML NBs.

We have shown that MORC3 interacts with itself constitutively via the C-terminal coiled-coil domain and via the N-terminal ATPase domain in an ATP-binding-dependent manner (Fig. 8). ATP-free MORC3 homodimerizes through the coiled-coil domain to form its ‘open-state’ structure, which is diffused throughout the nuclear matrix. ATP binding induces homodimerization of the N-terminal ATPase domain to form a ‘closed-state’ structure, which forms the NDs. The subsequent hydrolysis of ATP by the MORC3 ATPase domain releases the bound adenine nucleotide to form the open structure again and MORC3 moves from the NDs to the nucleoplasmic nuclear matrix. During this transfer, our results suggest that MORC3 passes transiently through one more step. Our in situ ND-diffusion analysis showed that MORC3 was washed out of the nucleus by treatment with CSK buffer only, or buffer containing ATP, and did not bind to the nuclear matrix (Fig. 3A). The CW and ΔCC mutants were also washed out of the nucleus with CSK buffer (Fig. 2B). We predict that MORC3 first localizes to a soluble nucleoplasmic space after its detachment from the NDs, and then attaches to the nuclear matrix (Fig. 8).

Although we have not yet measured MORC3 ATPase and ATP-binding activities directly, we found that the non-hydrolysable ATP analogue AMP-PNP blocked the diffusion of MORC3, but ATP did not (Fig. 3A). This result indicates that MORC3 has both ATPase and ATP-binding activities.

Amino acid residues essential for the ATP-hydrolysis and ATP-binding activities have been identified based on the crystal structures of many bacterial and yeast GHL ATPase proteins (Dutta and Inouye, 2000; Sacho et al., 2008; Shiau et al., 2006). The conserved glutamic acid residue in GHL-ATPase motif I has been identified as the catalytic base involved in the activation of a water molecule for nucleophilic attack on the γ-phosphate of ATP. Mutation of this residue to alanine abolished ATP hydrolysis, but not binding, in all other members of the GHL-ATPase protein family (Panaretou et al., 1998; Tran and Liskay, 2000). The conserved asparagine in motif II (Panaretou et al., 1998) and glycines in motif III (Tran and Liskay, 2000) interact directly with the N6-amine of the adenine moiety and with the α- and γ-phosphates of ATP, respectively. Mutation of these residues (D to N and G to A) abolished the ATP-binding activity (Panaretou et al., 1998; Tran and Liskay, 2000). In this study, we generated three MORC3 mutants with mutations in the equivalent residues (E35A, D79N and G101A). However, our data suggest that the MORC3 E35A mutant does not have ATP-binding activity. First, the diffuse localization of E35A is similar to another ATP-binding mutant, G101A, and to WT in ATP-depleted cells. Second, the ATP analogue, AMP-PNP, which is not hydrolyzed by GHL-ATPase, blocked the diffusion of MORC3 WT during treatment with Run-on buffer or with DNaseI. Furthermore, although the D67N mutant should have no ATP-binding activity, as predicted by similar mutations in other GHL-ATPase proteins (Panaretou et al., 1998), the MORC3 D67N mutant formed nuclear domains and was dispersed upon ATP depletion more quickly than the WT (supplementary material Fig. S2). The double mutant, D67N plus E35A, showed diffuse localization throughout the nuclei (data not shown). These results indicate that D67N has weak ATP-binding activity and that E35A and G101A have little or no ATP-binding activity. As we have not measured ATP-hydrolysis and ATP-binding activities of MORC3 directly in this study, the essential residues for these processes in MORC3 might be slightly different to other GHL-ATPase proteins. Crystal structure analysis of MORC3 will be required to identify the essential residues for MORC3 ATPase activity and ATP binding in future experiments.

We found that MORC3 is modified with SUMO1 at five sites in the coiled-coil domain (Fig. 5B,C), and that the SUMOylation mutant, 5KR, strongly decreased the interaction with PMLI (Fig. 6A,C). Furthermore, the interaction between MORC3 and PMLI was completely abolished in the absence of the isopeptidase inhibitor NEM (Fig. 6D) and UBE2i siRNA induced the formation of separate NDs of MORC3 and PMLI (Fig. 6E). Therefore, the SUMOylation of MORC3 is essential for the interaction between MORC3 and PMLI. However, this interaction was mildly, but reproducibly, reduced by mutation of the SIM of PMLI (Fig. 6B). In Fig. 6A, MORC3 and PMLI-AS formed NDs independently but were still in close contact with each other. In addition to the SIM site of PMLI, PMLI might have another domain that interacts with other SIM-containing proteins that are involved in the interaction with SUMOylated MORC3. We previously reported that endogenous PML NBs are formed in E35A-expressing cells (Takahashi et al., 2007), whereas in this study, PMLI was diffusely localized throughout the nucleus, with E35A (Fig. 6C). Since other alternative splicing variants, PMLIII and PMLIV, can form NBs in MORC3-E35A-expressing cells (supplementary material Fig. S5), they might be detected in endogenous PML NBs of MORC3 E35A-expressing cells. PMLIII and PMLIV contain the SIM but not the C-terminal region encoded by exon 9 in PMLI. Therefore, in addition to the SUMO-SIM interaction, this region might be involved in mediating a stronger interaction of MORC3 with PMLI than with PMLIII or PMLIV, and might interact with SIM-containing proteins that bind to SUMOylated MORC3, as mentioned above.

The CW domain is a mononuclear zinc finger composed of four cysteine and two tryptophan residues that are well conserved in a limited range of species including vertebrates, vertebrate-infecting parasites, and higher plants (Perry and Zhao, 2003). The CW domain is structurally similar to the PHD domain, which is known to bind to methylated histone H3. In addition, several CW-domain-containing families contain other domains that interact with methylated DNA or chromatin, such as SNF2 (sucrose non-fermenting 2) helicases, SET [Su(var)3-9, Enhancer-of-zeste, Trithorax] domains, PWWP (Pro-Trp-Trp-Pro) domains, and MBDs (methyl-binding domains). Thus, the CW domain was predicted to have a role in DNA binding, reading the histone code or recruiting other proteins to the nucleosome (Iyer et al., 2008; Perry and Zhao, 2003). However, the function of the CW domain has not yet been elucidated. In this study, we found that MORC3 forms NDs in a DNA-dependent manner because DNaseI treatment disrupted MORC3 NDs but not PML NBs (Fig. 2B). This result suggests that binding of MORC3 to DNA or DNA-associated proteins is required for the localization of MORC3 in NDs. Second, although the CW-domain mutant W419A has intact ATPase and coiled-coil domains and is predicted to form the ‘closed-state’ structure, the W419A mutant was diffused throughout the nucleus and was mostly washed out of the nucleus by treatment with CSK buffer (Fig. 2B). It is possible that a small amount of the W419A mutant that interacts with endogenous MORC3 was localized to PML NBs. Most of the W419A protein was extracted in the CSK buffer fraction (Fig. 2C), and thus might be present in the soluble nucleoplasmic space. Furthermore, the W419A mutant bound to the nuclear matrix during incubation in the ATP-depletion medium because W419A contains an intact ATPase domain (data not shown). Therefore, we demonstrated that MORC3 might associate with DNA or nucleosomes through its CW domain. Interestingly, AMP-PNP also blocked the diffusion of MORC3 NDs during treatment with DNaseI (Fig. 3B). This suggests that ATP-binding MORC3 might interact with proteins in the nuclear matrix, as well as with DNA or nucleosomes.

The size and number of MORC3 NDs vary among different cell types and species. MEFs form a small number of tiny MORC3 NDs, whereas human and mouse hematopoietic cells form a small number of larger MORC3 NDs. We found that MORC3 NDs are formed in different places to PML NBs in a large percentage of mouse bone marrow cells. MORC3 5KR formed a large number of small NDs (Fig. 6) and MORC3 WT completely colocalized with PMLI, but often formed PMLIII- or PMLIV-independent NDs (supplementary material Fig. S5A). Therefore, the SUMOylation of MORC3 and the expression profile of PML splicing products appear to regulate the number, size and PML-independent localization of MORC3 NDs to modulate cellular processes such as the cell cycle, stress and differentiation. We have already created a Morc3-knockout mouse, which dies at birth or within 1 day, for unknown reasons (Takahashi et al., 2007). This will help us to determine the molecular functions of MORC3 NDs in the future.

The expression vectors for FLAG-tagged MORC3 WT and the E35A mutant have been described previously (Takahashi et al., 2007). The constructs pLNCX HA-PMLI and HA-PMLI 3KR were kindly provided by Issay Kitabayashi (National Cancer Center, Tokyo, Japan) (Nguyen et al., 2005). We cloned PMLI and PMLI 3KR fragments into the pME-HA-I-P vector and pME-HA vector (Takahashi et al., 2007). The cDNA fragment of the PMLI AS mutant, in which the SUMO-interacting motif VVVI (amino acids 556-559) was substituted by AAAS, as described previously (Shen et al., 2006), was amplified by PCR with the primers containing the mutations and cloned into the pME-HA-I-P vector. SUMO1 (NM_152699) and SUMO3 (NM_006936) cDNA fragments were amplified by PCR and subcloned into the pME-HA vector. The coiled-coil deficient mutant of MORC3 (ΔCC, C-terminal 350 amino acids deleted) and the SUMO1 ΔGG mutant (C-terminal six amino acids deleted) were constructed by cloning the PCR-amplified fragments into the pME-FLAG-I-P vector and the pME-HA vector, respectively. The MORC3 SUMOylation mutants, whose lysines in the SUMOylation sites were substituted by arginine, and the MORC3 CW mutant, W419A, in which a conserved tryptophan (W419) in the CW domain was substituted by alanine, were generated using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA). The expression vector of mEGFP-MORC3 containing monomeric EGFP was generated from pME-EGFP-MORC3-I-P (Takahashi et al., 2007) by introducing the mutation A206K into EGFP (Zacharias et al., 2002). The pME-mCherry-PMLI vector was generated by replacing HA-tag of pME-HA-PMLI with mCherry cDNA (a kind gift from Roger Tsien, University of California at San Diego, San Diego, CA) carrying a SalI site instead of the stop codon.

HeLa cells and mouse embryonic fibroblasts (MEFs) were cultured in DMEM (Sigma) and CHO-K1 cells were cultured in Ham's F12 (Sigma) at 37°C in an atmosphere of 5% CO₂. All media were supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin and 100 μg/ml streptomycin. HeLa cells were transfected with the pME-I-P vector using FuGENE 6 (Roche) according to the manufacturer's instructions, selected with 1 μg/ml puromycin for 2 days, and used in the following experiments. In siRNA experiments, HeLa cells were transfected with the UBE2i siRNA (siGENOME SMARTpool M-004910-00-0005, Thermo Fisher Scientific, Lafayette, CO) and control siNRA (siCONTROL Non-Targeting siRNA#3, Thermo Fisher Scientific) using Lipofectamine2000 (Invitrogen), and the following day, the cells were further cotransfected with the FLAG-MORC3 and HA-PMLI expression vectors. Two days later, the cells were stained with anti-FLAG and anti-HA antibodies or sampled for immunoblotting. The retroviral vectors, pLNCX2-FLAG-MORC3-I-P (Takahashi et al., 2007) and pLP/VSVG (Invitrogen), were cotransfected into Plat-GP packaging cells (kindly provided by Toshio Kitamura, The University of Tokyo, Tokyo, Japan) (Morita et al., 2000) with FuGENE 6. The retrovirus was collected from the culture supernatant and used to infect the cells.

The National Cancer Institute (Mouse Models of Human Cancer Consortium) Mouse Repository provided the Pml-knockout mice, 129/Sv-Pml^tm1Ppp (Wang et al., 1998). Animal experiments were conducted in accordance with our institutional guidelines. Pml^+/− and Pml^−/− MEFs were generated from 14.5 day post coitum littermate embryos derived from an intercross between Pml^+/− mice.

The following antibodies and antiserum were used in this study: mouse monoclonal anti-FLAG M2 (Sigma), rabbit polyclonal anti-HA (MBL, Nagoya, Japan), mouse monoclonal anti-PML (PG-M3; Santa Cruz Biotechnology, Santa Cruz, CA), rabbit polyclonal anti-PML (H-238; Santa Cruz), goat polyclonal anti-mouse Pml (E-15; Santa Cruz), rabbit polyclonal anti-human UBE2I/UBC9 (ab33044, Abcam, Cambridge, MA) and mouse monoclonal anti-γ-tubulin (GTU-88, Sigma). The anti-MORC3 mouse monoclonal antibody (17A9; MBL) was generated as previously described (Takahashi et al., 2007). Histidine-tagged SUMO1-dc4 (His-SUMO1-dc4), deficient in the four C-terminal amino acids, was purified from Rosetta DE3 (Novagen, Madison, WI) carrying pET16-His-SUMO1-dc4. Anti-SUMO1 rabbit polyclonal antiserum was raised against the recombinant His-SUMO1-dc4.

Proteins were transferred from SDS-PAGE gels to Hybond-P (GE Healthcare) or Immobilon-P (Millipore, Bedford, MA) membranes, incubated with various primary antibodies and peroxidase-conjugated secondary anti-mouse IgG or anti-rabbit IgG antibodies (MBL) and visualized using ECL or ECL-Plus detection reagents (GE Healthcare).

Immunofluorescence studies were performed using a standard procedure, as previously described (Takahashi et al., 2007; Watanabe et al., 1996). Cells cultured on Lab-Tek II Chamber Slides (Nalge Nunc) or collected on glass slides by using a Cytospin4 centrifuge (Thermo Scientific, Waltham, MA), were fixed, permeabilized, and incubated with the primary antibodies indicated and the appropriate Alexa Fluor 488-, Alexa Fluor-594- or Alexa Fluor-647-conjugated, cross-adsorbed secondary antibodies (Invitrogen, Carlsbad, CA). Confocal images were collected on a Radiance 2000 laser-scanning system (Bio-Rad) connected to a BX51 microscope (Olympus) using the 60× oil-immersion objective for double staining and a confocal laser-scanning biological microscope FV10i (Olympus) for triple staining.

Cells were lysed in high-salt extraction buffer [50 mM Tris-HCl, pH 7.5, 5 mM EDTA, 0.5% Nonidet P-40, 300 mM NaCl, 150 mM KCl, 1 mM dithiothreitol, with Complete Protease Inhibitor Cocktail Tablets (Complete Mini, EDTA free, Roche) and 20 mM N-ethylmaleimide]. After removal of cell debris by centrifugation, lysates were then diluted three times with no-salt buffer (50 mM Tris-HCl, pH 7.5, 5 mM EDTA, 0.5% Nonidet P-40, 1 mM dithiothreitol, with complete protease inhibitor and 20 mM N-ethylmaleimide) and incubated with anti-FLAG-M2 affinity agarose gel (Sigma). The immunoprecipitated proteins were washed with 1× buffer (50 mM Tris-HCl, pH 7.5, 5 mM EDTA, 0.5% Nonidet P-40, 100 mM NaCl, 100 mM KCl, 1 mM dithiothreitol and 20 mM N-ethylmaleimide) and detected by western blotting. To detect SUMOylated MORC3, cells were solubilized with RIPA buffer (10 mM Tris-HCl, pH 7.4, 1 mM EDTA, 1% Nonidet P-40, 0.1% sodium deoxycholate, 0.1% sodium dodecylsulfate, 150 mM NaCl, with Complete Protease Inhibitor and 20 mM N-ethylmaleimide) and incubated with anti-FLAG-M2 affinity agarose gel. The immunoprecipitated proteins were immunoblotted with anti-SUMO1 antiserum and anti-FLAG antibody.

Nuclear-matrix binding analysis was performed as described elsewhere (Spector, 1998; Kimura et al., 2002). HeLa cells were transfected with FLAG-MORC3 WT or mutants, and treated step by step with the buffers indicated. First, the cells were treated with CSK buffer [10 mM PIPES, pH 6.8, 300 mM sucrose, 100 mM NaCl, 3 mM MgCl₂, 1 mM EGTA, 0.5% Triton X-100, with Complete Mini Tablets and 2 mM vanadyl riboside complex (VRC)] for 3 minutes at 4°C. Next, the cells were treated with high-salt buffer (CSK buffer containing 250 mM ammonium sulfate) for 4 minutes at 4°C. Finally, the cells were treated with CSK buffer, DNaseI buffer [CSK buffer containing 250 U/ml RNase-free DNaseI (Takara Bio, Otsu, Japan)] or RNase buffer [CSK buffer without VRC containing 250 U/ml RNase-free DNaseI and 2 μg/ml RNaseA (Wako Chemicals, Richmond, VA)] for 40 minutes at 37°C. The treated cells were fixed with 4% paraformaldehyde for more than 6 hours and analyzed by immunostaining with an anti-FLAG antibody. Proteins, extracted with each buffer, were analyzed by immunoblotting with an anti-FLAG antibody.

ATP-depletion analysis was performed as previously described (Platani et al., 2002). HeLa cells expressing FLAG-MORC3 WT or its mutants were incubated in glucose-free DMEM medium containing 10 mM sodium azide (Wako Chemicals) and 50 mM 2-deoxyglucose (Sigma) for 30 minutes, and then fixed and immunostained with an anti-FLAG antibody, as described above.

MORC3-ND-diffusion analysis was performed by modifying the in situ analysis previously described (Kimura et al., 2006; Saitoh et al., 2006). HeLa cells were transfected with FLAG-MORC3 WT and semipermeabilized with run-on buffer (10 mM HEPES-KOH, pH 7.3, 100 mM potassium acetate, 1 mM MgCl₂, 1 mM dithiothreitol, with Complete mini tablets) containing 50 μg/ml digitonin for 5 minutes on ice. The permeabilized cells were rinsed with run-on buffer and incubated in run-on buffer in the presence, or absence, of 10 mM ATP (Sigma) or 10 mM AMP-PNP (Adenylyl-imidodiphosphate, Roche) for 30 minutes at 37°C. The incubated cells were fixed and analyzed by immunostaining with an anti-FLAG antibody.

mEGFP-MORC3 and mCherry-PMLI expression vectors at a 9:1 ratio were cotransfected into CHO-K1 cells in a 35-mm-diameter glass-base dish (Asahi Glass, Tokyo, Japan) with FuGENE 6. One day later, to deplete ATP from the cells, the transfected cells were incubated in glucose-free medium containing 10 mM sodium azide for 3 hours at 37°C. ATP recovery was performed by replacing the ATP-depletion medium with normal medium. Live-cell images were collected every 10 minutes on a fluorescent inverted microscope, Eclipse TE300 (Nikon) equipped with a high-quality cooled CCD camera system, VB-7000 (Keyence, Osaka, Japan) and a 0.75 NA 40× Plan Fluor objective (Nikon).

We thank Issay Kitabayashi (National Cancer Center) for HA-PMLI expression vectors, Toshio Kitamura (The University of Tokyo) for Plat-GP, Roger Tsien (University of California at San Diego) for the mCherry plasmid, Toshiko Yasuda and Fomiko Mori for technical assistance, Junji Takeda (Osaka University), Kikuya Kato (Osaka Medical Center) and Tatsuo Takeya (Nara Institute of Science and Technology) for critical suggestions, and Kimura (Osaka University) for technical suggestions. This work was supported by a Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science, and Technology of Japan (N.I.).