Histone Deacetylase 7 Promotes PML Sumoylation and Is Essential for PML Nuclear Body Formation

Abstract

Promyelocytic leukemia protein (PML) sumoylation has been proposed to control the formation of PML nuclear bodies (NBs) and is crucial for PML-dependent cellular processes, including apoptosis and transcriptional regulation. However, the regulatory mechanisms of PML sumoylation and its specific roles in the formation of PML NBs remain largely unknown. Here, we show that histone deacetylase 7 (HDAC7) knockdown reduces the size and the number of the PML NBs in human umbilical vein endothelial cells (HUVECs). HDAC7 coexpression stimulates PML sumoylation independent of its HDAC activity. Furthermore, HDAC7 associates with the E2 SUMO ligase, Ubc9, and stimulates PML sumoylation in vitro, suggesting that it possesses a SUMO E3 ligase-like activity to promote PML sumoylation. Importantly, HDAC7 knockdown inhibits tumor necrosis factor alpha-induced PML sumoylation and the formation of PML NBs in HUVECs. These results demonstrate a novel function of HDAC7 and provide a regulatory mechanism of PML sumoylation.

The tumor suppressor promyelocytic leukemia protein (PML) is extensively localized in discrete nuclear punctate structures called PML nuclear bodies (NBs; also known as Kremer bodies, nuclear domain 10 [ND10], and PML oncogenic domains). Many proteins are recruited to, released from, or posttranslationally modified in these dynamic structures (2). The recruitment of certain proteins to PML NBs in response to various extracellular stimuli is important to mediate particular cellular responses (4, 8). Therefore, PML NBs serve as organizing centers within the cell that are involved in multiple cellular processes, including immune response, cell proliferation, and apoptosis (reviewed in reference 2).

PML NBs vary in number and size depending on cell type, cell cycle phase, and the differentiation status of the cell (7). While other proteins may shuttle in and out of NBs, PML remains a constitutive component of these structures. In fact, PML NBs are absent in PML^−/− mouse embryonic fibroblasts but can be restored by the expression of PML (29, 49). Due to the requirement of PML for the formation and maintenance of NBs, it is important to understand how PML is regulated and, in turn, how its regulation affects the structure and function of PML NBs.

PML is subject to multiple posttranslational modifications, including sumoylation, which is intimately involved in PML NB regulation. PML is sumoylated on three lysine residues, and a PML mutant protein that cannot be sumoylated is unable to form proper PML NBs (20, 49). It is of interest that several PML NB proteins are also sumoylated, and it has been proposed previously that the PML NB represents a major active site for protein sumoylation (37). In addition to the three sumoylation sites, PML contains a SUMO-binding domain in its C terminus which was reported previously to be important for interacting with nearby sumoylated PML, thereby initiating the nucleation of PML NBs and promoting the recruitment of other PML NB components, like Daxx, p53, and Sp100, etc. (reviewed in reference 2). Thus, the identification of the proteins responsible for regulating PML sumoylation is critical to understanding how extracellular stimuli regulate PML NB formation and protein composition and how PML NBs signal in numerous cellular responses.

SUMO conjugation can occur in the absence of an E3 ligase in vitro and can be stimulated by a SUMO E3 ligase (22). There are three distinct types of SUMO E3 ligases, including the protein inhibitor of activated STAT (PIAS) family, the Ran-binding protein 2 (RanBP2), and the polycomb protein Pc2 (21, 23, 35). While they do not share common structural features, they all have distinct motifs that bind the E2 enzyme, Ubc9, and their substrate. Several extracellular stimuli have been shown to control PML sumoylation (29, 33), but the cellular factors that modulate PML sumoylation remain largely unknown.

Mammalian class IIa histone deacetylases (HDACs) are structurally distinct from class I HDACs and include HDAC4, HDAC5, HDAC7, and HDAC9 (10, 11, 17, 18, 25, 26, 44, 46, 50). In addition to its deacetylase activity, HDAC4 has been shown previously to promote the sumoylation of MEF2 (16, 47), liver X receptor (15), and HIC1 (41), raising the possibility that other HDACs, such as HDAC7, may also stimulate protein sumoylation. We have recently demonstrated that HDAC7 partially localizes to PML NBs and that endogenous HDAC7 and PML interact in mammalian cells (13). We hypothesize that HDAC7 may stimulate PML sumoylation and play a role in regulating its function.

In this report, we show that HDAC7 knockdown decreases PML sumoylation and prevents PML NB formation in human umbilical vein endothelial cells (HUVECs). Conversely, the overexpression of HDAC7 potently stimulates PML sumoylation in a deacetylase-independent manner. Furthermore, HDAC7 associates with Ubc9 and recombinant HDAC7 stimulates PML sumoylation in vitro, indicating that HDAC7 may act as an E3 SUMO ligase for PML. Importantly, HDAC7 knockdown in HUVECs significantly decreases tumor necrosis factor alpha (TNF-α)-induced PML sumoylation, suggesting that HDAC7-mediated PML sumoylation participates in the cellular response to extracellular stimuli. Taking these results together, we demonstrate that HDAC7 acts as a SUMO E3 ligase to promote PML sumoylation and plays an important role in regulating the formation of PML NBs.

MATERIALS AND METHODS

Plasmids and DNA constructs.

PML and HDAC7 expression plasmids were described previously (14, 45). A cytomegalovirus-based promoter (CMX)-hemagglutinin (HA)-PML expression plasmid was generated by PCR and subcloned into the CMX-HA (1H) vector (27). PML and HDAC7 cDNAs with truncations and deletions were PCR amplified using full-length PML4 and HDAC7 genes as templates. HA-PML (3KR) was generated by site-directed mutagenesis (Strategene). Expression plasmids for yellow fluorescent protein (YFP)-HDAC7, HA-HDAC4, HA-HDAC5, FLAG-SUMO1, HA-Ubc9, and glutathione S-transferase (GST)-HDAC7 fusions were generated by PCR. All plasmids were checked by sequencing.

Reagents and antibodies.

TNF-α was purchased from Promega. N-Ethylmaleimide was purchased from Sigma-Aldrich. HDAC7 antibody has been described previously (14) and does not cross-react with HDAC4 or HDAC5. The following antibodies were purchased from Santa Cruz: anti-PML (sc-966 and sc-5621) and anti-Daxx (sc-7152). Anti-HA (E6779) and anti-FLAG (F3165) antibody-conjugated beads and anti-FLAG antibodies were purchased from Sigma. Anti-HA antibodies were purchased from Roche (2013819). Anti-SUMO2/SUMO3 (401900) antibodies were purchased from Invitrogen. Anti-SUMO1 antibodies were generated using His₆-SUMO1 as an antigen (Chemicon) and purified using GST-SUMO1 and GST sequential affinity columns.

Cell culture, transfection with siRNA, and immunoblotting.

HeLa cells were grown in Dulbecco's modified Eagle medium supplemented with 10% fetal bovine serum, 50 U of penicillin G/ml, and 50 μg of streptomycin sulfate at 37°C in 5% CO₂. HUVECs were isolated from freshly obtained umbilical cord samples by collagenase digestion of the umbilical vein. HUVECs were a gift from S. Matsuyama (Case Western Reserve University) and were maintained in endothelial cell basal medium (EGM-2; Cambrex) containing EGM-2 SingleQuot growth supplements (Cambrex). HUVEC passages 2 to 5 were used in this study. TNF-α (20 ng/ml) was added for 20 h. Control, PML, and HDAC7 SMART pool small interfering RNAs (siRNAs) were purchased from Dharmacon, and HUVECs were transfected according to the manufacturer's protocol. For the detection of endogenous PML sumoylation, whole-cell extracts were prepared 72 h posttransfection and subjected to immunoprecipitation with anti-PML antibodies, followed by Western blotting with anti-PML or anti-SUMO1 antibodies. For the determination of mRNA expression levels, total RNAs were isolated 48 h posttransfection and analyzed by reverse transcription-PCR with a kit from Invitrogen. For immunostaining and confocal microscopy, 48 h after transfection with siRNA, HUVECs were treated with or without TNF-α for 20 h prior to immunostaining and confocal microscopy.

Protein-protein interaction assays.

GST and GST-HDAC7 fusion proteins were expressed in the Escherichia coli DH5α strain, affinity purified, and immobilized on glutathione-Sepharose 4B beads. In vitro pulldown assays in which immobilized GST-HDAC7 proteins were incubated with whole-cell extracts expressing Ubc9 or PML for 1 h at 4°C were carried out according to our previously published protocol (25). After extensive washes, sodium dodecyl sulfate-polyacrylamide gel electrophoresis sample buffer was added to the beads and the mixture was boiled and separated by sodium dodecyl sulfate-7.5% polyacrylamide gel electrophoresis. Western blots were probed with antibodies against HA or FLAG.

For immunoprecipitations, cells were grown on 10-cm plates and transfected with the appropriate plasmids (10 μg of total DNA) by using Lipofectamine 2000 (Invitrogen). After 48 h, the cells were washed in 1× phosphate-buffered saline (PBS) and resuspended in radioimmunoprecipitation assay buffer with protease inhibitors. After incubation on ice for 1 h, the lysed cells were centrifuged at 4°C and 8,100 × g for 10 min; supernatant was collected and kept at −80°C. The whole-cell extracts were incubated with appropriate antibodies for 4 h at 4°C and then with protein A/G beads for 2 h at 4°C. The immunopellets were washed three to four times and analyzed in Western blots probed with the antibodies indicated below. For endogenous immunoprecipitation, HUVECs were treated with TNF-α as described above. Whole-cell lysates were prepared and immunoprecipitated with anti-PML antibodies, and the immunoprecipitates were analyzed by immunoblotting with anti-PML or anti-SUMO1 antibodies.

Detection of PML sumoylation in mammalian cells.

HeLa cells were transfected with HA-PML and FLAG-SUMO1 expression constructs in the absence or presence of class II HDACs. To detect PML sumoylation in mammalian cells, whole-cell extracts in the presence of N-ethylmaleimide were prepared. Whole-cell lysates were subjected to immunoprecipitation with anti-HA or anti-FLAG antibodies, followed by immunoblotting with anti-HA or anti-FLAG antibodies as indicated below. For the sumoylation of endogenous PML, HUVECs were transfected with HDAC7 siRNA and treated with or without TNF-α and cell lysates were prepared. Immunoprecipitations were performed with anti-PML antibodies, and the immunopellets were immunoblotted with anti-PML or anti-SUMO1 antibodies.

In vitro sumoylation assays.

In vitro sumoylation assays were performed as described previously (5, 31) with some modifications. Briefly, 4 μg of His-tagged PML recombinant protein was incubated with 50 or 100 ng of GST or GST-HDAC7 proteins for 30 min at room temperature, after which a sumoylation reaction mixture (0.1 μg of SUMO E1, 0.1 μg of Ubc9, and 4 μg of SUMO1) in reaction buffer (50 mM NaCl, 20 mM HEPES, 0.1% Tween 20, 5 mM MgCl₂, pH 8.0) was added and the preparation was maintained at 37°C for 1 h. Reaction mixtures were immunoblotted with anti-PML or anti-SUMO1 antibodies.

Confocal microscopy.

Transfected cells were fixed in 3.7% paraformaldehyde in PBS for 30 min at room temperature and permeabilized in PBS with the addition of 0.1% Triton X-100 and 10% goat serum for 10 min. The cells were washed three times with PBS and incubated in a mixture of PBS with goat serum (10%) and 0.1% Tween 20 solution (ABB) for 60 min. Incubation with primary antibodies in ABB was carried out for 120 min. The cells were washed three times in PBS, secondary antibodies were added to the ABB mixture, and the mixture was kept in the dark at room temperature for 30 to 60 min. Coverslips were mounted onto slides using Vectashield mounting medium with DAPI (4′,6-diamidino-2-phenylindole; Vector Laboratories, Inc.). All confocal images were acquired using a Zeiss LSM 510 inverted laser scanning confocal microscope. A 63× oil immersion planapochromat objective with a numerical aperture of 1.4 was employed for all experiments. For endogenous PML and PML introduced by transient transfection, images of Alexa Fluor 594 were collected using a 633-nm-wavelength excitation light from an He/Ne2 laser, a 633-nm dichroic mirror, and a 650-nm long-pass filter. For endogenous HDAC7, images of Alexa Fluor 488 were collected using a 488-nm-wavelength excitation light from an argon laser, a 488-nm dichroic mirror, and a 500- to 550-nm band-pass barrier filter. All DAPI-stained nuclear images were collected using a coherent Mira-F-V5-XW-220 (Verdi 5W) Ti:sapphire laser tuned at 750 nm, a 700-nm dichroic mirror, and a 390- to 465-nm band-pass barrier filter. The primary antibodies are described above. The secondary antibodies (anti-mouse and anti-rabbit Alexa Fluor 488 and anti-rabbit and anti-mouse Alexa Fluor 594) were from Molecular Probes. The applied laser power (measured in percentages) of the Zeiss LSM 510 system has a linear range. For the quantification of the intensity, if two samples were acquired with similar parameters, e.g., detector gain and scan speed, etc., varying the percentage of laser power can be used to semiquantify the two images in terms of their average pixel intensity. See Fig. 6A for an example in which the laser power used to acquire the image in panel b was seven times higher than that used to acquire the image in panel g or q. In other words, the average pixel intensity or the level of proteins seen in panel g is seven times greater than that seen in panel b in this case.

FIG. 6.

TNF-α promotes PML sumoylation and PML NB formation in HUVECs in an HDAC7-dependent manner. (A) HUVECs were treated with or without TNF-α. Twenty hours posttreatment, immunostaining with anti-PML (α-PML), SUMO1, or Daxx was carried out, followed by confocal microscopy. The laser used for panel b was seven times (7×) more intense than that used for panel g. Similarly, the laser used for panel l was three times more intense than that used for panel q. (B) HUVECs were treated with (+) or without (−) TNF-α. Twenty hours posttreatment, immunoprecipitation (IP) with anti-PML antibodies was carried out, followed by immunoblotting (IB) with anti-SUMO1. WCE, whole-cell extracts; Mr(K), molecular size in kilodaltons. (C) HUVECs were transfected with control siRNA (siControl) or the HDAC7 siRNA SMART pool (siHDAC7). Forty-eight hours posttransfection, cells were treated with TNF-α as described in Materials and Methods and whole-cell extracts were prepared and subjected to immunoprecipitation with PML antibodies and Western blotting with anti-SUMO1 antibodies. (D) HUVECs were treated with TNF-α and analyzed by immunostaining and confocal microscopy as described in the legend to panel A. HDAC7 staining was significantly decreased in four cells marked with asterisks or the number 2. Cells numbered 1 are those in which HDAC7 was moderately knocked down or in which HDAC7 was not knocked down.

RESULTS

HDAC7 knockdown reduces PML sumoylation and PML NB formation in HUVECs.

To test whether endogenous HDAC7 regulates PML sumoylation and PML NB formation in vivo, we knocked down HDAC7 in HUVECs by using siRNA (Fig. 1A; also see Fig. S1 in the supplemental material). We found that the size and number of PML NBs were significantly reduced in cells transfected with HDAC7 siRNA (Fig. 1A, panels h to k versus panels d to g). In addition, the elimination of PML NBs by PML siRNA treatment was accompanied by the redistribution of SUMO1 (Fig. 1B). This result is most easily observed by comparing panels e, f, and g in Fig. 1B. Panel f shows the PML distribution in both transfected and nontransfected cells. Coimmunostaining with anti-PML and anti-SUMO2/SUMO3 antibodies revealed that PML and SUMO2 and SUMO3 do not colocalize (data not shown). These results are consistent with previous reports that PML sumoylation plays a critical role in PML NB formation and suggested that HDAC7 may regulate PML sumoylation. To directly examine the effect of HDAC7 knockdown on PML sumoylation, PML was immunoprecipitated from cell lysates and analyzed for sumoylation (Fig. 1C). Our results revealed that HDAC7 knockdown resulted in decreased accumulation of the major PML-containing band (Fig. 1C, lanes 1 and 2). This PML species was sumoylated by SUMO1, as shown in the immunoblot (Fig. 1C, lanes 3 and 4), and the level of its sumoylation was significantly reduced by HDAC7 knockdown. Indeed, the decreased PML protein levels correlated with a loss of PML sumoylation. These results demonstrated that the loss of HDAC7 led to reductions in the sumoylation and steady-state accumulation of PML. Consistent with the immunostaining results, anti-SUMO2/SUMO3 antibodies did not yield any detectable signal (data not shown). These data strongly suggest that HDAC7 plays an important role in PML sumoylation and PML NB formation.

FIG. 1.

HDAC7 knockdown abolishes PML sumoylation and PML NB formation. (A) HUVECs were treated with HDAC7 siRNA. Forty-eight hours posttransfection, cells were subjected to immunostaining using anti-PML (α-PML) and anti-HDAC7 antibodies, followed by confocal microscopy. Note that cells 1 and 2 express different levels of HDAC7 due to different efficiencies of knockdown by HDAC7 siRNA. They also display different intensities of PML or PML NBs. (B) HUVECs were transfected with PML siRNA (siPML). Forty-eight hours posttransfection, cells were subjected to immunostaining with anti-PML and anti-SUMO1 antibodies, followed by confocal microscopy. Cells transfected with PML siRNA are marked with arrows. siControl, control siRNA. (C) HUVECs were treated with siRNA against HDAC7. Whole-cell lysates were immunoprecipitated with anti-PML antibodies. Immunopellets and whole-cell lysates were probed with anti-PML, antiactin, anti-HDAC7, or anti-SUMO1 as indicated. A quantitative analysis revealed that HDAC7 knockdown led to a 60% loss of PML sumoylation and a loss of 58% of total PML protein levels. IB, immunoblotting; IP, immunoprecipitation; +, present; −, absent.

HDAC7 stimulates PML sumoylation in mammalian cells.

To further examine whether HDAC7 promotes PML sumoylation, we overexpressed HDAC7 and monitored the amount of steady-state PML sumoylation. We cotransfected HeLa cells with plasmids expressing FLAG-SUMO1 and HA-PML in the presence or absence of HDAC7 expression plasmids and analyzed PML sumoylation by immunoblotting (Fig. 2A). We found that there was an increase in the amount of slower-migrating PML species in the presence of exogenous HDAC7 (Fig. 2A, lanes 2 and 3), suggesting that HDAC7 promotes the sumoylation of PML. To confirm that the slower-migrating species observed in lane 3 were actually sumoylated PML species, we carried out reciprocal immunoprecipitations with anti-HA (PML) or anti-FLAG (SUMO1), followed by immunoblotting to monitor the level of sumoylated PML (Fig. 2A, lanes 4 to 9). We found that the fraction of coprecipitated PML contained more sumoylated forms when HDAC7 was coexpressed (Fig. 2A, lanes 6 and 9). We also analyzed the effects of HDAC7 on PML sumoylation in the absence of exogenous SUMO1. Our results confirmed that HDAC7 promotes the sumoylation of PML (Fig. 2B). Furthermore, increasing the amount of HDAC7 increased the intensities of the slower-migrating PML species in a dose-dependent manner (Fig. 2C, lanes 1 to 5). Immunoprecipitation and Western blotting analysis further confirmed that these slower-migrating PML species were indeed sumoylated PML (Fig. 2C, lanes 6 to 10). In summary, these results demonstrated that HDAC7 promotes PML sumoylation and accumulation in mammalian cells.

FIG. 2.

HDAC7 promotes PML sumoylation in mammalian cells. (A) HeLa cells were transfected with HA-PML, FLAG-SUMO1, and YFP-HDAC7 expression plasmids. Immunopellets precipitated with anti-FLAG (α-FLAG) and anti-HA antibodies were immunoblotted with anti-HA and anti-FLAG antibodies as indicated. The band migrating at 80 kDa is likely to be nonsumoylated HA-PML that is coprecipitated with sumoylated PML (FLAG-SUMO1-tagged HA-PML) due to the ability of PML to form dimers. It is marked with an asterisk. IB, immunoblotting; IP, immunoprecipitation; +, present; −, absent. (B) HeLa cells were transfected with HA-PML and YFP-HDAC7 expression constructs. Whole-cell lysates or immunopellets precipitated with anti-HA antibodies were immunoblotted with anti-HA and anti-SUMO1 antibodies as indicated. Note that not only was the intensity of sumoylated PML increased in cells cotransfected with an HDAC7 construct, but also slower-migrating sumoylated PML species were detected. (C) Cells were cotransfected with plasmids expressing increasing concentrations of YFP-HDAC7 (lanes 3 to 5), FLAG-SUMO1, and HA-PML. Whole-cell extracts (WCE) were immunoprecipitated with anti-FLAG antibodies (lanes 6 to 10), and the immunoprecipitates were analyzed by immunoblotting with anti-HA antibody. Nonsumoylated PML is marked with an asterisk. The expression levels of endogenous YFP-HDAC7 (lower band in lower panel, marked with an arrowhead) and YFP-HDAC7 introduced by transfection (upper band in lower panel) are shown.

Mapping the HDAC7 domain sufficient to promote PML sumoylation.

Similar to other class IIa HDACs, HDAC7 contains multiple protein-protein interaction domains that include a MEF2/ACTN4-binding motif, three 14-3-3 binding sites, and the C-terminal zinc-binding motif unique to class IIa HDAC (Fig. 3A). It has been shown previously that amino acids 67 to 257 of HDAC4, corresponding to amino acids 2 to 254 of HDAC7, are capable of stimulating MEF2 sumoylation (47). To define the regions of HDAC7 required for the stimulation of PML sumoylation, we investigated whether the N-terminal or C-terminal domain of HDAC7 was sufficient to stimulate PML sumoylation. As shown in Fig. 3B, the coexpression of the C-terminal fragment of HDAC7 [amino acids 500 to 938; referred to hereinafter as HDAC7 (500-938)] with HA-PML and FLAG-SUMO1 dramatically increased PML sumoylation. However, the expression of the N-terminal fragment of HDAC7 [amino acids 2 to 254; referred to hereinafter as HDAC7 (2-254)] decreased PML sumoylation. In order to test further whether the catalytic activity of HDAC7 is essential for the stimulation of PML sumoylation, a catalytically dead HDAC7 (D692A/D694A) (25) was coexpressed with HA-PML and FLAG-SUMO1. The results show that the HDAC7 mutant form still potently stimulated HA-PML sumoylation by FLAG-SUMO1, indicating that the HDAC catalytic activity is dispensable for the potentiation of PML sumoylation (Fig. 3C). Together, these data demonstrate that HDAC7 stimulates PML sumoylation independent of its deacetylase activity.

FIG. 3.

The C terminus of HDAC7 stimulates PML sumoylation in mammalian cells and in vitro. (A) Schematic representation of HDAC7 functional domains. The structure of the HDAC7 C terminus has recently been resolved and shows that a CCHC zinc-binding motif is present at the N terminus of the HDAC enzymatic domain. WT, wild type. (B) HeLa cells were cotransfected with HA-PML and FLAG-SUMO1 expression plasmids and a plasmid expressing full-length HA-HDAC7, HDAC7 (2-254), or HDAC7 (500-938). Whole-cell lysates and immunoprecipitates were immunoblotted with anti-HA (α-HA) or anti-FLAG antibodies. IB, immunoblotting; IP, immunoprecipitation; +, present; −, absent. (C) HeLa cells were cotransfected with HA-PML and FLAG-SUMO1 expression plasmids and a wild-type HA-HDAC7 or mutant (Mut) HA-HDAC7 expression plasmid. Immunoprecipitations were carried out with anti-FLAG or anti-HA antibodies, and immunoprecipitates were immunoblotted with anti-HA or anti-FLAG antibodies as indicated. (D) HeLa cells were cotransfected with expression plasmids for FLAG-PML and HA-HDAC7 truncations. Whole-cell lysates were immunoprecipitated with anti-FLAG antibodies. The resulting immunopellets, along with the whole-cell lysates, were analyzed by immunoblotting with anti-HA and anti-FLAG antibodies as indicated. WCE, whole-cell extracts. (E) HeLa whole-cell lysates expressing HA-PML were used in GST pulldowns with either GST alone or GST fused to the indicated truncations of HDAC7. The precipitated fractions were analyzed by immunoblotting with anti-HA antibodies. 500-ter, HDAC7 (500-938). (F) Recombinant His-tagged PML proteins were subjected to in vitro sumoylation assays with GST or GST-HDAC7 (500-938) fusion protein as described in Materials and Methods. The arrows and solid circles mark SUMO1-modified PML (sumo), and the asterisk marks nonsumoylated PML (non-sumo). Note that the amount of sumoylated PML compared to non sumoylated PML, as well as the relative sumoylation level, was increased when GST-HDAC7 (500-938) was added. It is estimated that five- to sixfold stimulation of PML sumoylation was observed when 100 ng of GST-HDAC7 (500-938) was used. For lanes 7 to 12, the intensity of sumoylated PML in lane 8 was used for normalization and was set at 1. The ratio of the intensity compared to that in lane 8 is shown and represents the increase (n-fold) in sumoylation stimulation (lanes 11 and 12).

These results suggested that HDAC7 may interact with PML via the C-terminal region. To test this hypothesis, we mapped the HDAC7 minimal interaction domains by coimmunoprecipitation experiments. Cells were cotransfected with HA-HDAC7 truncation and deletion expression plasmids and a FLAG-PML expression plasmid and analyzed by immunoprecipitation with anti-FLAG antibodies and Western blotting with anti-FLAG and anti-HA antibodies. We found that, indeed, the C-terminal catalytic domain (amino acids 500 to 938) of HDAC7 was sufficient for association with PML in mammalian cells (Fig. 3D). To confirm these observations, GST pulldown assays were performed by incubating whole-cell lysates expressing HA-PML with bacterially expressed and purified GST-HDAC7 fusion proteins (Fig. 3E). Consistent with the coimmunoprecipitation results, we found that HDAC7 (500-938) interacted with PML (Fig. 3E, lane 5) but that the other fragments were unable to bind PML (Fig. 3E, lanes 3 and 4).

To investigate whether the HDAC7 C terminus is capable of directly stimulating PML sumoylation in a reconstituted system, we performed in vitro sumoylation assays using purified recombinant SUMO E1 and E2, His-tagged PML (His₆-PML), SUMO1, GST, and GST-HDAC7 fusion proteins (see Fig. S2 in the supplemental material). In the absence of SUMO E1 and E2, purified bacterially expressed His₆-PML was detected (Fig. 3F, lane 1). The incubation of His₆-PML with SUMO E1 and E2 resulted in the appearance of a major slower-migrating PML band accompanied by an increase in the sumoylated/nonsumoylated His₆-PML ratio (Fig. 3F, lanes 2 versus 1 and 8 versus 7), indicating that the SUMO E1 and E2 were sufficient for PML sumoylation in vitro (Fig. 3F, lanes 2, 4, 7, and 9). Consistent with our data from HeLa cells, the addition of GST-HDAC7 (500-938), but not GST alone, led to five- to sixfold increases in PML sumoylation (Fig. 3F, lanes 6 versus 4 and 12 versus 10). These data further support our hypothesis that HDAC7 enhances PML sumoylation and suggest that HDAC7 acts as a SUMO E3 ligase for PML.

Class IIa HDACs promote PML sumoylation.

The class IIa HDAC family members show significant homology in the C terminus. HDAC4 has been found to be capable of stimulating the sumoylation of MEF2 (47, 48). We further tested whether other class IIa HDACs were also capable of potentiating PML sumoylation. Cells were cotransfected with HA-PML and FLAG-SUMO1 expression plasmids and with HA vector or a HA-HDAC4, HA-HDAC5, or HDAC7 expression plasmid (Fig. 4, lanes 1 to 4). Western blotting following immunoprecipitation verified that all three HDACs increased the amount of sumoylated PML when overexpressed in HeLa cells (Fig. 4, lanes 5 to 12). These data suggest that class IIa HDACs can promote PML sumoylation in HeLa cells.

FIG. 4.

Other class IIa HDACs also potentiate PML sumoylation. HA-PML, FLAG-SUMO1, and HA-HDAC4, HA-HDAC5, or HA-HDAC7 expression plasmids were singly expressed or coexpressed in HeLa cells. Whole-cell extracts were prepared, and immunoprecipitations were performed as described in the legend to Fig. 2A. Asterisks indicate the expression of HA-HDAC4, HA-HDAC5, and HA-HDAC7 in lanes 2, 3, and 4, respectively. The arrow marks HA-PML. We did not observe the sumoylation of HDAC4, HDAC5, or HDAC7 under our assay conditions (data not shown). IB, immunoblotting; IP, immunoprecipitation; +, present; −, absent; α-HA, anti-HA antibody.

HDAC7 associates with Ubc9.

The above data imply that HDAC7 may function as a SUMO E3 ligase for PML sumoylation. Since all known SUMO E3 ligases interact with the SUMO E2 enzyme Ubc9, we hypothesized that HDAC7 might also interact with Ubc9. To investigate this possibility, we examined whether HDAC7 can interact with Ubc9 by coimmunoprecipitation. Whole-cell lysates from cells expressing FLAG-HDAC7 and HA-Ubc9 were subjected to immunoprecipitation with anti-HA antibodies, followed by immunoblotting with anti-FLAG antibodies. Indeed, Ubc9 and HDAC7 were coimmunoprecipitated when the proteins were coexpressed (Fig. 5A, lane 6). To determine the region of HDAC7 that interacts with Ubc9, we performed similar coimmunoprecipitation experiments utilizing different HDAC7 fragments. The results demonstrated that both HDAC7 (2-254) and HDAC7 (500-938) were able to interact with Ubc9 (Fig. 5B, lanes 9 and 12) and that fragments encompassing amino acids 72 to 172 [HDAC7 (72-172)] and 241 to 533 [HDAC7 (241-533)] were unable to interact with Ubc9 (Fig. 5B, lanes 10 and 11). In accordance with these data, when these HDAC7 fragments were expressed as GST fusion proteins and used in GST pulldown experiments with whole-cell lysates expressing HA-Ubc9, both GST-HDAC7 (2-254) and GST-HDAC7 (500-938) interacted with Ubc9 but no interaction with GST-HDAC7 (241-533) was detected (Fig. 5C). Altogether, these data indicate that the interaction between HDAC7 and Ubc9 requires residues in both the N- and C-terminal regions of HDAC7, supporting the hypothesis that HDAC7 may promote PML sumoylation as a SUMO E3 ligase by scaffolding with both the substrate and the SUMO E2 enzyme Ubc9.

FIG. 5.

HDAC7 associates with Ubc9. (A) HeLa cells were cotransfected with expression plasmids for HA-HDAC7 and FLAG-Ubc9, whole-cell extracts (WCE) were prepared and immunoprecipitated with anti-HA antibodies (α-HA), and the immunoprecipitates were analyzed by immunoblotting with anti-HA and anti-FLAG antibodies as indicated. IP, immunoprecipitation; +, present; −, absent. (B) HeLa cells were cotransfected with FLAG-Ubc9 and HA-HDAC7 full-length or truncation expression plasmids. Whole-cell lysates and immunopellets from immunoprecipitations with anti-FLAG antibodies were analyzed by immunoblotting with anti-FLAG and anti-HA antibodies as indicated. WT, wild type; 500-ter, HDAC7 (500-938). (C) Immobilized GST-HDAC7 fusion proteins were incubated with extracts from cells expressing HA-Ubc9 and analyzed by immunoblotting with anti-HA antibodies.

HDAC7 is essential for TNF-α-induced PML sumoylation in HUVECs.

Our lab has recently shown that TNF-α increases PML protein accumulation in HUVECs (13). The physiological significance of PML sumoylation in PML NB formation led us to investigate whether TNF-α enhances PML sumoylation and the formation of PML NBs in HUVECs and whether this effect occurs in an HDAC7-dependent manner. We first examined PML NBs with or without TNF-α treatment by immunostaining and confocal microscopy. Figure 6A shows that TNF-α treatment significantly induced PML NB formation, as seven (panel b) and three (panel l) times the laser intensity used in vehicle treatment was required to detect equal amounts of signal used in TNF-α-treated cells (panels g and q). SUMO1 and Daxx have been shown to be constitutive components of PML NBs. In accordance with an increase in the formation of PML NBs, the amounts of SUMO1 and Daxx in PML NBs were also increased in the presence of TNF-α (Fig. 6A, panels c versus h and m versus r).

In addition, we have also found that HDAC7 is recruited to PML NBs (13) in response to TNF-α treatment. This recruitment was accompanied by increased PML sumoylation (Fig. 6B). To investigate whether HDAC7 plays a role in TNF-α-induced PML sumoylation, we knocked down HDAC7 by using siRNA and assessed PML sumoylation in response to TNF-α. We found that the knockdown of HDAC7 decreased the amount of the major sumoylated PML species (Fig. 6C, lane 3 versus lane 4). Together, these data suggest that the knockdown of HDAC7 inhibits TNF-α-stimulated PML sumoylation and subsequent accumulation. Furthermore, a correlation between a reduction in the expression of HDAC7 and a decrease in the number and intensity of PML NBs was observed (Fig. 6D), implicating a role of HDAC7 in the formation of PML NBs in response to TNF-α. These data demonstrate that HDAC7 is critical for TNF-α-mediated PML sumoylation in HUVECs.

DISCUSSION

Our major finding is that HDAC7 is capable of stimulating PML sumoylation both in mammalian cells and in a reconstituted purified system, supporting the notion that HDAC7 is a direct physiological modulator of PML sumoylation. Furthermore, we have shown that in HUVECs, TNF-α treatment induced PML sumoylation and PML NB formation in an HDAC7-dependent manner.

We employed three independent assays to demonstrate that HDAC7 controls PML sumoylation. First, we found that the knockdown of endogenous HDAC7 by siRNA decreased PML accumulation and sumoylation and PML NB formation. Furthermore, the amount of PML lost correlated with a decrease in the level of sumoylation. Second, the coexpression of HDAC7 with PML, in the absence or presence of exogenous SUMO1, was observed to increase PML sumoylation. Third, in vitro sumoylation assays showed that purified HDAC7 reconstituted this sumoylation-stimulating activity. To our knowledge, HDAC7 is the first member of the class IIa HDACs that fulfills these criteria and the first protein shown to stimulate PML sumoylation in a purified system.

Although PML has been proposed previously to be sumoylated by SUMO1, SUMO2, and SUMO3 (1, 12, 24), Western blotting analyses revealed that PML was sumoylated only by SUMO1 in HUVECs, even after TNF-α treatment (Fig. 1C and 6B and data not shown). This observation is also supported by the facts that SUMO1, but not SUMO2 and SUMO3, colocalized with PML NBs (data not shown) and that the knockdown of PML redistributed SUMO1 (Fig. 1B) but not SUMO2 and SUMO3 (data not shown). Nonetheless, we cannot exclude the possibility that PML can be sumoylated by different SUMO proteins, depending on the cell type and distinct extracellular stimuli.

Many proteins that interact with PML have been identified, including those that localize to PML NBs as either transient or constitutive residents (2). It is reasonable to speculate that some of these proteins may modulate PML sumoylation and/or the formation of PML NBs. One such candidate is Daxx, which binds to sumoylated PML and constitutively localizes to PML NBs (30). However, the knockdown of Daxx did not affect PML NB formation (6). We have recently reported that the stable knockdown of the peptidyl-prolyl isomerase Pin1 increases the size and number of PML NBs in breast cancer cell lines (36). To date, HDAC7 is the only known PML-interacting partner that, when knocked down, inhibits both PML sumoylation and the formation of PML NBs.

What is the mechanism by which HDAC7 stimulates PML sumoylation? As HDAC7 is a transcriptional corepressor, the knockdown of HDAC7 may potentially affect the expression of the genes encoding the sumoylation machinery components, such as SUMO1. However, we detected no differences in SUMO1 mRNA accumulation with and without HDAC7 knockdown (see Fig. S3A in the supplemental material). In addition, no decrease in SUMO1 conjugation was observed when HDAC7 was knocked down (see Fig. S3B in the supplemental material). These observations indicate that knocking down HDAC7 did not affect SUMO1 conjugation machinery. Protein sumoylation can occur in the absence of an E3 SUMO ligase in vitro. Although it has been speculated that PML may function as a SUMO E3 ligase, no evidence indicates that PML is capable of stimulating the sumoylation of other substrates in vitro or in mammalian cells. Structural and functional studies of known SUMO E3 ligases PIAS, RanBP2, and Pc2 indicate that these E3 SUMO ligases do not share conserved functional domains and may stimulate protein sumoylation in distinct manners (23, 35, 38). A recent structural study indicated that amino acids 518 to 605 of HDAC7 harbor a class IIa HDAC-specific zinc-binding motif (a CCHC motif) which may provide a docking site for protein-protein interactions and multiprotein complex formation so as to bring Ubc9 and PML into close proximity (Fig. 3) (39). By doing so, HDAC7 may act as a scaffold for the stimulation of PML sumoylation. Indeed, HDAC7 (500-938) binds both PML and Ubc9 and is sufficient to stimulate PML sumoylation in vivo and in vitro. It will be interesting to determine whether the zinc-binding CCHC motif is critical for HDAC7-induced PML sumoylation.

Although sumoylation has been recognized to be essential for PML NB formation, the mechanism underlying this observation is not completely understood. We and others have recently demonstrated that the SUMO-binding domain present in PML and Daxx is required for PML NB formation (32, 40). Our data that the knockdown of HDAC7 resulted in a decrease in PML NBs and that HDAC7 stimulated the SUMO1 conjugation of PML in vitro further support the notion that the sumoylation of PML promotes PML NB formation. Interestingly, HDAC7 knockdown also led to a concomitant decrease in PML protein accumulation (Fig. 1C). Together, our data suggest that sumoylation promotes not only PML NB formation but also PML protein accumulation. This conclusion is consistent with our recent finding that SUMO1 modification blocks PML from interacting with Pin1 and likely prevents Pin1-mediated phosphorylation-dependent PML degradation (36).

It is interesting that the polysumoylation of PML by SUMO2 and SUMO3 leads to PML degradation in response to arsenic through the recruitment of RNF4, a RING finger ubiquitin E3 ligase. Ubiquitin and proteasomes are also recruited to PML NBs, leading to subsequent PML degradation (28, 42). These findings, together with ours, provide a possible explanation for the differential consequences of stabilization and PML monosumoylation and of degradation and PML polysumoylation. They suggest that PML sumoylation, ubiquitination, and degradation are tightly linked.

When PML and SUMO1 were coexpressed in HeLa cells, multiple species of sumoylated PML were detected. However, only singly SUMO1-conjugated PML was detected in HUVECs. As described above, SUMO2 and SUMO3 conjugations generate polysumoylation, whereas SUMO1 conjugation generates monosumoylation. Because PML can be conjugated with SUMO1, SUMO2, or SUMO3, we suspect that different cell types may generate distinct sumoylation ladders. It is possible that different cell types may express distinct members of the sumoylation-desumoylation machinery which controls the conjugation of PML with SUMO1, SUMO2, or SUMO3.

Cellular stresses including UV irradiation, viral infection, and reactive oxygen species are known to regulate the size and number of PML NBs (3, 9, 34, 43). Nonetheless, their relationships to PML sumoylation and PML protein levels are not clearly understood. Our data show that TNF-α treatment increases the size and number of PML NBs (Fig. 6A), as well as PML sumoylation and protein levels (Fig. 6B). The knockdown of HDAC7 not only decreased PML sumoylation, but also inhibited PML NB formation. Together, our data establish a positive correlation between PML sumoylation and the size and number of PML NBs in response to extracellular stimuli.

Taken together, our work highlights a novel mechanism by which the formation of PML NBs is controlled by HDAC7 through the regulation of PML sumoylation. Furthermore, TNF-α treatment promotes PML sumoylation through HDAC7. It is possible that HDAC7 may also regulate PML sumoylation in response to other extracellular stimuli. For example, TNF-α, along with other apoptotic inducers, may signal synergistically to PML through HDAC7. This pathway could then be exploited to promote PML NB formation in cases in which PML expression is low, as has been shown previously for many cancers (19). In summary, our work sheds new light on the regulation of PML NBs. The importance of PML NBs is underscored by the identification of the many proteins recruited to these structures, increasing their capacity to regulate a variety of cellular processes. Further investigation into the components and dynamics of PML NBs is integral to a full understanding of the cellular responses to extracellular stimuli.