The ND10 Component Promyelocytic Leukemia Protein Acts as an E3 Ligase for SUMOylation of the Major Immediate Early Protein IE1 of Human Cytomegalovirus

PML positively affects the SUMOylation of IE1, but not of IE2.

ND10 bodies are hypothesized to function as SUMOylation hot spots within the nucleus, with PML itself representing one of the various SUMO E3 ligases that can be found in these subnuclear structures (8, 17). Given the fact that the two important viral regulatory proteins IE1p72 (IE1) and IE2p86 (IE2), which are both known to be posttranslationally modified by SUMO, transiently localize at ND10 during the initial stage of infection, we questioned whether their SUMO modification might be affected by PML. To address this question, we infected PML knockdown (kd) human foreskin fibroblasts (HFFs) with a small interfering RNA-mediated knockdown of PML (siPML2 cells) in which PML expression was highly reduced, along with control cells (vector) expressing endogenous levels of PML, as well as HFFs overexpressing PML isoform I (HFF/PMLI), with HCMV at a multiplicity of infection (MOI) of 0.1 (Fig. 1). In accordance with previous data, we found that the extent of IE1 SUMOylation is directly correlated with the abundance of PML (Fig. 1, second gel from top) (34). The IE2 SUMOylation status, on the other hand, remained unaffected irrespective of whether PML was present or absent (Fig. 1, third gel from top). Thus, PML specifically enhances SUMOylation of IE1. This correlates with the observation of a direct protein interaction between PML and IE1 (23).

Similar results were obtained in transient-transfection experiments with HEK293T cells, where IE1 was expressed alone or in combination with PML isoform VI (Fig. 2). As shown in Fig. 2A, coexpression of PML significantly augmented the steady-state level of SUMOylated IE1. In addition, we observed that the enhanced conjugation of SUMO to IE1 exhibited PML dose dependency; however, at high concentrations of PML, SUMOylation of IE1 could not be further increased (Fig. 2B, lanes 4 to 7). Since E3 ligases are thought to serve as adaptor proteins, limiting amounts of additional components of the SUMO conjugation machinery may explain this effect. In contrast, quantification revealed that SUMO conjugation of IE2 did not increase but was slightly diminished after coexpression of large amounts of PML (Fig. 2C). Hence, PML expression, intriguingly, has contrary effects on the SUMOylation status of IE1 and IE2.

In this context, it is notable that we regard HEK293T cells as an advantageous cell line to study the effects of exogenously introduced PML, since in these cells very little endogenous PML is present compared to primary HFFs (Fig. 3A). In line with this finding, knockdown of PML in HEK293T cells did not alter the outcome of any of our transfection experiments (see Fig. 9C). In addition, HEK293T cells are nearly devoid of the major ND10 factor Sp100 (Fig. 3B), while hDaxx protein levels, on the other hand, are clearly elevated in comparison to HFFs (Fig. 3C). Thus, the transformed phenotype of HEK293T cells goes along with dramatic changes in ND10 protein abundance and composition, and as a consequence, no genuine ND10 structures are detectable by immunofluorescence staining in these cells (data not shown).

PML is capable of stimulating IE1 SUMO modification in vitro.

To determine whether PML has a direct effect on the SUMO modification of IE1, an in vitro SUMOylation assay was performed (Fig. 4). To this end, glutathione S-transferase (GST)-tagged variants of wild-type (wt) PML or a SUMOylation-negative PML mutant (both PML isoform VI), as well as of IE1, were expressed in Escherichia coli, which lacks a SUMOylation system. Thereafter, prokaryotically expressed PML and IE1 were affinity purified, and in the case of IE1, the GST tag was removed by enzymatic cleavage followed by chromatographic purification. When incubated in a SUMOylation reaction mixture, the addition of either wt PML (PML) or the SUMO-deficient PML mutant (SUMOmut) upregulated IE1 modification by both SUMO-1 and SUMO-3 (Fig. 4A and B). In contrast, after incubation of IE1 alone without any components of the SUMO conjugation pathway, no higher-molecular-mass band of IE1 was detectable (Fig. 4C, lane 1). The effect of PML on IE1 SUMOylation in these in vitro reactions was comparable to the previously reported E3 ligase function of PML on p53 (8). In a reaction with increasing amounts of PML, the enhanced SUMOylation of IE1 was confirmed; however, no linear correlation of the intensity of SUMOylation signals with the amount of PML was detected (Fig. 4D, lanes 3 to 6, and E). This is consistent with our observations from cotransfection analyses (Fig. 2B). Taken together, these data indicate that PML has the ability to stimulate the SUMOylation of IE1, not only in vivo, but also in vitro. This further strengthens the assumption that IE1 is a direct target protein for the SUMO E3 ligase activity of PML.

PML isoforms differ in their abilities to SUMOylate IE1.

The PML gene consists of nine exons, and by alternative splicing, six nuclear PML isoforms (PML I to VI) are expressed (35). All these PML variants contain a common N terminus harboring the TRIM motif, but they vary in their C termini due to individual C-terminal extensions (Fig. 5A). Consequently, the question arose as to whether the observed stimulation of IE1 SUMOylation by PML isoform VI is a shared feature of all nuclear PML species. To assess the roles of the individual PML isoforms for IE1 SUMOylation, we expressed IE1 alone or in combination with PML I to VI in vivo (Fig. 5B). As shown in Fig. 5B, irrespective of which PML isoform was expressed (Fig. 5B, top), in each case an IE1-SUMO band was detectable (Fig. 5B, middle, lanes 3 to 8), which was absent in the PML-negative sample (Fig. 5B, middle, lane 2). Interestingly, however, although all PML variants showed rather comparable expression levels (Fig. 5B, top), the degree of IE1 SUMOylation varied significantly depending on the PML isoform expressed (Fig. 5B, middle; quantified in panel C). While PML III and VI induced large amounts of SUMO-modified IE1 (Fig. 5B, lanes 5 and 8), the presence of PML IV resulted in only very weak IE1-SUMO signals (Fig. 5B, lane 6), whereas expression of PML I, II, or V gave rise to an intermediate IE1 SUMOylation phenotype (Fig. 5B, lanes 3, 4, and 7).

Direct protein interaction is required for PML-mediated SUMOylation of IE1.

To investigate whether the physical interaction between PML and IE1 is a prerequisite for PML-induced posttranslational SUMO modification of IE1, we took advantage of an IE1 point mutant (L174P) that has been reported to fail to associate with PML (36, 37). By performing coimmunoprecipitation experiments, we could confirm a loss of interaction between PML and the IE1 mutant L174P, since, in contrast to wt IE1, the mutant was no longer able to coprecipitate PML (isoform VI) (Fig. 6A, bottom, compare lanes 2 and 3). Furthermore, the mutant was deficient in the deSUMOylation of PML (Fig. 6B, top, compare lanes 2, 5, and 6). As can be seen in Fig. 6B, the IE1 mutant L174P could also no longer be SUMOylated by PML compared to wt IE1 (Fig. 6B, third panel from top, compare lanes 5 and 6). This indicates that direct interaction is essential for PML-induced SUMOylation of IE1. In addition, by mutating the lysine residue at position 450 of IE1 to arginine (K450R), which has been mapped as the SUMO acceptor site of IE1, we could likewise abolish the appearance of the high-molecular-weight form of IE1 upon coexpression of PML (Fig. 6D, middle, lane 6), although this K450R mutant interacted as efficiently with PML as wt IE1 (Fig. 6C, bottom, compare lanes 2 and 4). Thus, this unambiguously identifies the slower-migrating IE1 band that runs at around 100 kDa as the SUMO-modified version of IE1.

PML SUMOylates IE1 in a RING finger-dependent manner.

Next, we set out to elucidate the structural determinant for PML-mediated SUMOylation of IE1 (Fig. 7). Since the RING finger domain of PML is proposed to confer the E3 ligase activity, we disrupted the RING structure by mutating two key zinc-chelating cysteine residues at positions 57 and 60 to serines (Fig. 7A). As shown by Shen and coworkers, such a PML mutant is still able to interact with wt PML (27). This led the authors to conclude that introduction of the respective mutations does not disturb the overall structure of the protein. In addition, we could demonstrate that wt IE1, but not the L174P mutant, was still capable of interacting with the PML RING finger mutant (PML RINGmut) (Fig. 7B). This further strengthens the assumption of correct folding of the PML RING finger mutant. Inactivation of the SUMO E3 ligase activity of PML abrogated PML-induced SUMOylation of IE1 in vivo (Fig. 7C): comparably to the situation where IE1 was expressed alone (Fig. 7C, lane 4), no IE1-SUMO variant could be detected upon coexpression of PML RINGmut (Fig. 7C, lane 6). Furthermore, the experiment revealed that SUMOylation of the RING mutant PML protein was drastically reduced compared to wt PML (Fig. 7C, top, compare lanes 2 and 3 or 5 and 6). This is in agreement with previous findings by Shen and coworkers, which fostered the idea that PML acts as a SUMO E3 ligase that is also responsible for catalyzing its auto-SUMOylation (27). By performing an in vitro SUMOylation assay, we additionally confirmed the importance of PML's RING domain for IE1 SUMOylation (Fig. 7D). In contrast to the positive-control PML SUMOmut (Fig. 7D, lane 3), a GST-tagged PML mutant (isoform VI) with a deletion of the RING domain failed to stimulate IE1 SUMO conjugation above levels seen without PML or in the presence of GST alone (Fig. 7D). Notably, deletion of the PML RING domain does not affect the interaction between PML and IE1, as previously demonstrated by coimmunoprecipitation analyses (38). Taken together, these data demonstrate that an intact RING finger domain is not only necessary for PML self-SUMOylation, but also absolutely essential for PML-based SUMO conjugation of IE1.

A PML RING finger mutant fails to orchestrate ND10 formation.

Since SUMOylation of PML is a prerequisite for the nucleation of ND10 structures, we next tested the capacity of the SUMOylation-defective RING domain mutant (PML RINGmut) to assemble ND10 bodies. To this end, the subnuclear localization pattern of PML RINGmut was analyzed in normal HFFs, as well as PML-kd cells (siPML2), by confocal immunofluorescence microscopy (Fig. 8). Consistent with the results of Western blot experiments demonstrating a lack of PML auto-SUMOylation, introduction of PML RINGmut into siPML2 cells lacking endogenous PML did not result in a re-formation of ND10 structures (Fig. 8A, e to m, and B). Instead, PML RINGmut and other ND10-related factors, like Sp100 or hDaxx, displayed a microdispersed distribution throughout the nucleus (Fig. 8A, f, g, k, and l). In contrast, expression of PML RINGmut in normal HFFs resulted in a partial recruitment of the mutant into the preexisting ND10 bodies, presumably through interaction with endogenous PML via the coiled-coil domain (Fig. 8A, a to d, and B). Collectively, our data demonstrate that the PML RING domain is critical for PML auto-SUMOylation and ND10 formation.

Intact ND10 structures are not necessary for PML-based SUMOylation of IE1.

In light of the fact that the PML RING finger mutant lacks the ability to assemble ND10 bodies, the question arose as to whether intact ND10 structures are required for PML-mediated SUMOylation of IE1. This is important, since other components of the SUMO conjugation machinery that are involved in the modification of IE1, like the E1-activating enzyme, the E2-conjugating enzyme, or SUMO itself, accumulate at ND10. To investigate this, we took advantage of a SUMOylation-negative mutant of PML (SUMOmut) with lysine-to-arginine substitutions at residues 65, 160, and 490 (Fig. 9A) that has been reported to be defective in ND10 formation (28). In the next step, we coexpressed IE1, together with either wt PML, PML RINGmut, or PML SUMOmut in HEK293T cells (Fig. 9B). Interestingly, although the auto-SUMOylation of both PML mutants was highly compromised (Fig. 9B, top), only the RING finger mutant failed to stimulate IE1 SUMO conjugation (Fig. 9B, middle, lane 7). In contrast, IE1 was as efficiently SUMOylated by SUMOmut as by wt PML (Fig. 9B, middle, compare lanes 6 and 8), which is in accordance with the findings obtained by in vitro SUMOylation reactions (Fig. 4). In this context, it is noteworthy, that we ruled out the possibility that the small amount of endogenous PML present in HEK293T cells (Fig. 3) has any impact on the outcome of this experiment. Identical experiments were performed in both PML-kd HEK293T (PML-kd/293T) and genuine HEK293T cells, yielding identical results (compare Fig. 9B and C). Therefore, we conclude that SUMOylation of PML and the accompanying capacity to form genuine ND10 structures are not required for SUMOylation of IE1, while an intact RING finger domain of PML is indispensable for the activity. This assumption is consistent with the observation that IE1 SUMOylation is detectable throughout the HCMV replication cycle, even after ND10 bodies are disrupted (Fig. 1, second panel from top).

Finally, we investigated the subnuclear localization pattern of PML SUMOmut in normal HFFs and PML-kd cells (siPML2) (Fig. 9D). In cells expressing endogenous PML (HFFs), the SUMO-deficient mutant (SUMOmut) localized exclusively to ND10 (Fig. 9D, top row). Hence, the targeting of PML SUMOmut to ND10 bodies occurs much more efficiently than that of PML RINGmut (Fig. 8). Intriguingly, even in a PML knockdown background (siPML2), PML SUMOmut was capable of aggregating in distinct foci (Fig. 9D, middle and bottom rows), which is in clear contrast to the microdispersed distribution pattern of PML RINGmut in siPML2 cells (Fig. 8A, e to m). However, the number of PML SUMOmut foci formed in PML-depleted cells was clearly reduced compared to normal HFFs (Fig. 9E), indicating that these SUMOmut aggregations do not represent genuine ND10. The latter notion is further supported by our observation that these nuclear foci are completely devoid of hDaxx while Sp100 still efficiently localized to these SUMOmut-formed structures (Fig. 9D, e to m). Consequently, PML RINGmut and SUMOmut seem to differ in their capabilities to nucleate punctate PML-based complexes and to associate with other ND10-related factors.

Oligonucleotides and plasmid constructs.

The oligonucleotide primers used for this study were purchased from Biomers GmbH (Ulm, Germany) and are listed in Table 1. The PML isoform VI plasmid was a generous gift from Gerd Maul (Philadelphia, PA, USA). Lentivirus vectors expressing FLAG-tagged versions of PML isoforms I to VI were kindly provided by Roger Everett (Glasgow, United Kingdom) (47). All cDNAs encoding the PML isoforms were made resistant to the anti-PML shRNA siPML2 (48) by introduction of five silent point mutations within the siPML2 target sequence (47). The altered sequence is 5′-AGATGCTGCAGTTAGCAAG-3′. Derivatives of Myc-PML isoform VI with point mutations in the RING domain (RINGmut; C57S and C60S) or of the SUMO modification sites (SUMOmut; K65R, K160R, and K490R) and IE1 mutants L174P (a PML interaction-negative mutant) and K450R (a SUMOylation-deficient mutant) were constructed by site-directed mutagenesis using a QuikChange site-directed mutagenesis kit, as instructed by the manufacturer (Stratagene, La Jolla, CA). For bacterial expression of GST-tagged variants of wt PML or PML SUMOmut, the respective coding sequences were cloned into the prokaryotic expression vector pGEX-6P-1 (GE Healthcare, Munich, Germany) via BamHI/EcoRI using primers 5′PML-1597 and 3′PML-1597/2691. GST-IE1 was obtained by cloning of a codon-optimized version of the IE1 cDNA sequence into pGEX-6P-1 (GE Healthcare, Munich, Germany) via BamHI/XhoI using the primer pair 5′IE1codonopt_BamH1 and 3′IE1codonopt_Xho1. For generation of GST-RINGmut containing a deletion of the RING domain (amino acids [aa] 1 to 102), a codon-optimized version of PML isoform VI was used for PCR amplification of aa 103 to 560 using primers 5′Bam_PMLco_aa103 and 3′Sal_PMLVIco, followed by cloning of the PCR product into pGEX-6P-1 via BamHI and SalI. The integrity of all the newly generated plasmids was confirmed by automated DNA sequence analysis.

Cells and viruses.

HEK293T cells were cultivated in Dulbecco's minimal essential medium (DMEM) containing 10% fetal calf serum. HEK293T cells with a small interfering RNA-mediated knockdown of PML (PML-kd/293T) were cultured in DMEM supplemented with 10% fetal calf serum and 5 μg/ml puromycin (22). Primary HFFs were prepared from human foreskin tissue as described previously and were maintained in Eagle's minimal essential medium (GIBCO/BRL, Eggenstein, Germany) supplemented with 5% fetal calf serum (32). siPML2 cells and the respective control cells (designated vector) were cultured in Dulbecco's minimal essential medium (GIBCO/BRL, Eggenstein, Germany) supplemented with 10% fetal calf serum and 5 μg/ml puromycin (48). Infection experiments were performed with the HCMV laboratory strain AD169. Viral stocks were titrated via IE1p72 fluorescence (49). For this purpose, HFFs were infected with various dilutions of virus stocks. After 24 h of incubation, the cells were fixed and stained with monoclonal antibody (MAb) p63-27, directed against IE1p72 (50). Subsequently, the number of IE1-positive cells was determined and used to calculate viral titers, expressed as IE protein-forming units (IEU).

Retrovirus transduction and selection of stably transduced cells.

HFFs stably expressing FLAG-PML isoform I were generated via retrovirus transduction using the ViraPower lentiviral system (Invitrogen, Karlsruhe, Germany) according to the manufacturer's protocol and maintained in Dulbecco's minimal essential medium supplemented with 10% fetal calf serum plus 50 μg/ml Geneticin.

Antibodies.

MAb p63-27, which recognizes IE1, has been described elsewhere (50). The polyclonal antiserum raised against exon 5 of IE2 (referred to as anti-pHM178) was generated by immunizing rabbits with the prokaryotically expressed protein. PML was detected by using either the rabbit polyclonal antibody H-238 (Santa Cruz Biotechnology, Santa Cruz, CA) or the mouse monoclonal antibody 5E10 (kindly provided by Roel van Driel, University of Amsterdam, Amsterdam, The Netherlands). For the detection of Myc- or FLAG-tagged versions of PML or SUMO-1, the monoclonal antibodies MAb-Myc (1-9E10.2; ATCC) and MAb-FLAG M2 (Sigma-Aldrich, Deisenhofen, Germany) were utilized. Sp100 was detected by applying either the rabbit polyclonal antiserum GH3 (a kind gift from Hans Will, Heinrich Pette Institute for Experimental Virology and Immunology, University of Hamburg, Hamburg, Germany) or a rabbit polyclonal antibody raised against aa 209 to 227 of Sp100 (kindly provided by Peter Hemmerich, Leibniz Institute for Age Research, Fritz Lipmann Institute, Jena, Germany). A rabbit monoclonal antibody from Epitomics (Burlingame, CA) or the mouse monoclonal antibody MCA2143 (Serotec, Duesseldorf, Germany) was used for the detection of hDaxx. Monoclonal antibody AC-15, which recognizes beta-actin, was purchased from Sigma-Aldrich (Deisenhofen, Germany). Horseradish peroxidase-conjugated anti-mouse and anti-rabbit secondary antibodies for Western blot analysis were obtained from Dianova (Hamburg, Germany), while Alexa Fluor 488- or Alexa Fluor 555-conjugated secondary antibodies for indirect immunofluorescence experiments were purchased from Molecular Probes (Karlsruhe, Germany).

Indirect immunofluorescence analysis.

For indirect immunofluorescence analysis, 3 × 10⁵ HFFs were grown on coverslips for transfection using the transfection reagent FuGene HD (Promega, Mannheim, Germany) according to the manufacturer's instructions. At 48 h posttransfection, the cells were washed three times with phosphate-buffered saline (PBS), followed by fixation with 4% paraformaldehyde for 10 min at room temperature. Then, the cells were permeabilized with PBS–0.2% Triton X-100 on ice for 20 min, followed by incubation with the respective primary or secondary antibodies for 30 min at 37°C. Finally, the cells were mounted by using Vectashield mounting medium plus 4′,6-diamidino-2-phenylindole (DAPI) (Vector Laboratories, Burlingame, CA). The samples were examined by using a Leica TCS SP5 confocal microscope with 488-nm and 543-nm laser lines, scanning each channel separately under image capture conditions that eliminated channel overlap. The images were exported as tagged-image file format (TIFF) files and were then processed using Photoshop.

Western blotting and co-IP assay.

For Western blot analysis, extracts from infected HFFs or transfected HEK293T cells were lysed, diluted in sodium dodecyl sulfate (SDS)-Laemmli buffer, and boiled at 95°C for 10 min (32). Proteins were separated by SDS-polyacrylamide gel electrophoresis (PAGE) on 8% to 12.5% polyacrylamide gels and transferred onto nitrocellulose membranes (GE Healthcare, Munich, Germany), followed by chemiluminescence detection according to the manufacturer's protocol (ECL Western blotting detection kit; Amersham Pharmacia Europe, Freiburg, Germany). The expression of SUMOylated and non-SUMOylated proteins was quantified using Image J and AIDA software, followed by the calculation of the ratio of protein SUMOylation relative to the overall protein expression. Coimmunoprecipitation (co-IP) analysis was performed as described previously (51). Briefly, HEK293T cells were transfected in 6-well plates (5.0 × 10⁵ cells/well) via calcium phosphate coprecipitation. Two days posttransfection, the cells were lysed for 20 min at 4°C in 800 μl of co-IP buffer (50 mM Tris-HCl [pH 8.0], 150 to 300 mM NaCl, 5 mM EDTA, 0.5% NP-40, 1 mM phenylmethylsulfonyl fluoride [PMSF], 2 μg/ml of aprotinin, 2 μg/ml of leupeptin, and 2 μg/ml of pepstatin). After centrifugation, the supernatant was incubated with the appropriate antibody (MAb p63-27) coupled to protein A Sepharose beads for 1.5 h at 4°C. The Sepharose beads were collected and washed six times in co-IP lysis buffer. Antigen-antibody complexes were recovered by boiling in SDS sample buffer and analyzed via Western blotting.

Protein generation and purification.

PML, SUMOmut, RINGmut, and IE1 were expressed in fusion with GST using E. coli strain BL21(DE3). The fusion proteins were purified from crude lysates by affinity chromatography using glutathione–Sepharose columns according to the specifications of the manufacturer (GE Healthcare, Bio-Sciences AB, Uppsala, Sweden). In the case of IE1, the GST tag was cleaved by addition of PreScission protease (GE Healthcare) at a ratio of 1:100 (wt/wt). Free GST and GST-tagged PreScission protease was removed by glutathione Sepharose affinity chromatography. Afterward, IE1 was purified by gel filtration using a HiLoad 16/60 Superdex 200 Prep Grade column (GE Healthcare, Bio-Sciences AB, Uppsala, Sweden) according to the manufacturer's protocol. Eluted proteins were analyzed by SDS-PAGE and Coomassie blue staining.

In vitro SUMOylation assays.

In vitro SUMOylation reactions were performed at 37°C for 1 h in 20-μl volumes containing prokaryotically expressed IE1 or GST-PML/GST-SUMOmut/GST-RINGmut (based on PML isoform VI), along with 50 nM the heterodimeric human E1-activating enzyme (SAE1/SAE2), 62.5 μM the E2-conjugating enzyme UbcH9, 1 mM Mg-ATP, 25 μM either SUMO-1 or SUMO-3, and reaction buffer (500 mM HEPES, pH 8, 1,000 mM NaCl, 10 mM dithiothreitol [DTT]; BostonBiochem, Cambridge, MA) according to the manufacturer's protocol. After termination of the reaction using SDS sample buffer containing beta-mercaptoethanol, the reaction products were fractionated by SDS-8% PAGE for immunoblot detection of IE1, PML, and GST.