Role of SUMO in RNF4-mediated Promyelocytic Leukemia Protein (PML) Degradation

Promyelocytic leukemia protein (PML) is a tumor suppressor acting as the organizer of subnuclear structures called PML nuclear bodies (NBs). Both covalent modification of PML by the small ubiquitin-like modifier (SUMO) and non-covalent binding of SUMO to the PML SUMO binding domain (SBD) are necessary for PML NB formation and maturation. PML sumoylation and proteasome-dependent degradation induced by the E3 ubiquitin ligase, RNF4, are enhanced by the acute promyelocytic leukemia therapeutic agent, arsenic trioxide (As2O3). Here, we established a novel bioluminescence resonance energy transfer (BRET) assay to dissect and monitor PML/SUMO interactions dynamically in living cells upon addition of therapeutic agents. Using this sensitive and quantitative SUMO BRET assay that distinguishes PML sumoylation from SBD-mediated PML/SUMO non-covalent interactions, we probed the respective roles of covalent and non-covalent PML/SUMO interactions in PML degradation and interaction with RNF4. We found that, although dispensable for As2O3-enhanced PML sumoylation and RNF4 interaction, PML SBD core sequence was required for As2O3- and RNF4-induced PML degradation. As confirmed with a phosphomimetic mutant, phosphorylation of a stretch of serine residues, contained within PML SBD was needed for PML interaction with SUMO-modified protein partners and thus for NB maturation. However, mutation of these serine residues did not impair As2O3- and RNF4-induced PML degradation, contrasting with the known role of these phosphoserine residues for casein kinase 2-promoted PML degradation. Altogether, these data suggest a model whereby sumoylation- and SBD-dependent PML oligomerization within NBs is sufficient for RNF4-mediated PML degradation and does not require the phosphorylation-dependent association of PML with other sumoylated partners.

Promyelocytic leukemia protein (PML) 5 is a tumor suppressor (1) whose gene is translocated in cases of acute promyelocytic leukemia (2). PML functions as the organizer of PML NBs, which are dynamic structures harboring numerous transiently and permanently localized proteins (3). The importance of PML NB structural integrity was first revealed in acute promyelocytic leukemia because, in this malignancy, the abnormal fusion protein PML/RAR␣ leads to NB disruption. Patient treatment with As 2 O 3 induces the reversion of the acute promyelocytic leukemia phenotype as well as PML/RAR␣ degradation and PML NB reformation (4).
PML is a target for the post-translational modification by SUMO, an ubiquitin-like protein that is covalently coupled to PML lysine residues 65, 160, and 490 via a process called sumoylation (5,6). Among the four human SUMO paralogs identified, SUMO1, -2, and -3 were found to be conjugated to target proteins. It involves an enzymatic cascade for the transfer of the mature SUMO and the formation of an isopeptide bond between the COOH-terminal glycine of SUMO and a lysine from the target protein. Sumoylation is a reversible process due to the existence of several deconjugating enzymes.
PML NB formation requires both the covalent linkage (sumoylation) (reviewed in Ref. 7) and the non-covalent interactions of SUMO with PML through a SUMO binding domain (SBD also named SIM for SUMO interacting motif) (8). Interestingly, PML SBD contains specific serines, acting as substrates for the caseine kinase-2 (CK2), which are implicated in PML ubiquitination and degradation (9) and which phosphorylation status could regulate the function of the SBD.
Because sumoylation of proteins is dynamic and reversible, this post-translational modification is difficult to follow in vivo and its detection mainly relies on the identification of sumoylated protein species by Western blot following cell lysis. Recently, we used bioluminescence resonance energy transfer (BRET) to detect covalent linkage of ubiquitin (ubiquitination) in living mammalian cells and in real time (10). In brief, BRET monitors the interaction between a protein fused to a luciferase and a protein fused to yellow or green fluorescent protein (YFP or GFP), upon addition of a luciferase substrate; it is a proximity-based assay that requires that the donor of energy (luciferase fusion) and the acceptor (YFP or GFP fusions) are within 50 to 100 Å for an efficient energy transfer (11)(12)(13). However, a demonstration that BRET may provide a method of choice to follow the dynamics of protein sumoylation in living cells is lacking. Here, we developed a sensitive and quantitative SUMO BRET assay for the detection of PML interactions with SUMO in living cells. We proved that BRET can be used to detect both SUMO covalent and non-covalent interactions with PML (model, Fig. 1H). For this purpose, we used the PMLIII isoform in which sumoylation is induced by As 2 O 3 and triggers a proteasomedependent PML degradation (14); the degradation process involves the ubiquitination of poly-SUMO covalently coupled to PML by the poly-SUMO-specific E3 ubiquitin ligase RNF4 (15)(16)(17). Altogether, our BRET results indicate that, As 2 O 3 and/or RNF4-induced PML degradation are dependent on the integrity of both PML sumoylation target sites and the PML SBD core sequence but not on the CK2 serine phosphorylation sites within the SBD. However, phosphorylation of these serines is required for most PML SBD-dependent non-covalent interactions. This phospho-regulation of PML SBD ("SBD phospho-switch") establishes another link between the phosphorylation and SUMO, different from the phospho-sumoyl switch (18).

Constructs and Expression Vectors
The sequence of all cDNA constructs was confirmed by DNA sequencing.

BRET Constructs
YFP-SUMO1, YFP-SUMO1G, YFP-SUMO2, and YFP-SUMO3 (all cloned in pEYFP (C1), BD Biosciences Clontech) were a kind gift of Dr. Dasso M. (National Institutes of Health, Bethesda, MD) (19). GFP-SUMO1 and GFP-UBI were previously described, respectively, in Refs. 20 and 10. The myctagged SUMO1 construct used was subcloned as an XhoI fragment in the XhoI site of a myc-tagged version of pcDNA3.1 vector. UBC9 (accession number U45328) was cloned in the SacI/KpnI site of pGFP10 C3 (Perkin Elmer Life Sciences). The NLS used for the Luc-NLS construct was already described (13) and Luc-NLS was generated by inserting the NLS sequence within the ApaI/BamHI site of pHRLuc-C3 (Perkin Elmer Life Sciences).
To generate the other BRET constructs, Renilla Luciferase (Luc) or YFP were cloned together with the cDNA of interest by a three-piece ligation in the BamHI/XhoI or BamHI/XbaI site of pcDNA3.1(ϩ) (Invitrogen). Luc was amplified by PCR from the pcDNA3-CXCR4-Luc vector (21) and cloned either at the COOH-terminal of the cDNA of interest as an EcoRI-XhoI fragment (primers used: "Luc_EcoRI_ATG_C-term sense" and "Luc_Stop_XhoI_C-term antisense" primers) or at the N-terminal as a BamHI-EcoRI fragment (primers used: "Luc_Bam-HI_ATG_N-term sense" and "Luc_no-Stop_EcoRI_N-term antisense"). YFP was amplified by PCR from pcDNA3-CXCR4-YFP (21), digested with EcoRI and XhoI, and cloned at the COOH-terminal of the cDNA of interest (primers used: "YFP_EcoRI_ATG_C-term sense" and "YFP_Stop_XhoI_Cterm antisense"). The cDNA of interest was amplified by PCR for cloning in phase either at the N-terminal of the fusion clone as a BamHI-EcoRI fragment or at the COOH-terminal as an EcoRI-XhoI or EcoRI-XbaI fragment. In each case, the sequence joining the cDNA of interest and either Luc or YFP sequence encodes VPVNSGGGGS as a linker.
PML ⌬CC was constructed by replacing a 411-bp fragment coding for the PMLIII coiled-coil domain (nucleotides 669 to 1069 of PMLIII coding sequence) by a short unrelated 18-bp sequence, which includes a NotI site. The PML ⌬CC sequence was generated by PCR amplification of 5Ј-and 3Ј-overlapping fragments, which both include the NotI site from the 18-bp sequence; the 5Ј-PCR fragment includes the initiator ATG of PMLIII and an upstream BamHI cloning site and the 3Ј-PCR fragment includes the PMLIII Stop codon and a downstream XhoI cloning site. The 5Ј and 3Ј PCR fragments (BamHI-NotI and NotI-XhoI fragments, respectively) were ligated together in a three-piece ligation reaction in the BamHI/XhoI site of pcDNA3.1(ϩ) (primers used: "PML_BamHI_ATG sense" and "PML_⌬CC1_NotI antisense" for the 5Ј-fragment and "PML_⌬CC2_NotI sense" and "PML_ Stop_XhoI antisense" for the 3Ј-fragment).

PML Expression Constructs
PML WT and mutants were amplified by PCR from pBluescript SKϩ for subcloning in appropriate vectors. To generate untagged expression constructs, PML PCR fragments digested with BamHI and XhoI were cloned into the BamHI/XhoI site of either pcDNA3.1(ϩ) for transient expression or HIV-1-based lentiviral pTRIP vector for stable expression (a gift from Dr. P. Charneau, Groupe de Virologie moléculaire et de Vectorologie, Institut Pasteur, Paris) (primers used: "PML_BamHI_ATG sense" and "PML_ Stop_XhoI antisense"). To generate Luc and YFP BRET constructs, PML PCR fragments digested with BamHI and EcoRI were cloned NH 2 -terminal to Luc or YFP in the BamHI/XhoI site of pcDNA3.1(ϩ) as part of a three-piece ligation (primers used: "PML_BamHI_ATG sense" and "PML_No Stop EcoRI_Nterm" antisense).

Antibodies
The following antibodies were used: mouse monoclonal anti-PML (clone PGM3) (used for immunofluorescence microscopy), rabbit polyclonal anti-PML (clone H-238) (used for the Western), rabbit polyclonal anti-actin (clone C-11) were from Santa Cruz Biotechnology and mouse monoclonal anti-FLAG (clone M2) from Sigma.

Arsenic and Epoxomycin Treatments
As 2 O 3 (Sigma number A1010, 330 mM solution stock resuspended in 1 N NaOH) and Epoxomycin (Sigma number E3652, 100 M solution stock resuspended in dimethyl sulfoxide) were used at the concentrations and for the times indicated in the text.

Stable Transduction, Transient Transfection, and Cell Culture
Stable Transduction-A stable expression of PML constructs was obtained by a lentiviral-based strategy as described (23). The efficiency of the HIV-1-based lentiviral pTRIP vector, in which PML constructs are subcloned, relies on the presence of a triple-stranded DNA structure that acts as a cis-determinant of HIV-1 DNA import. The stable integration of the transgenes into the host DNA allows efficient and long term transgene expression without clone selection. Virus stock production and infection were as described in Ref. 23 to transduce and express PML wild type and mutants in the HEK293T cell line. Briefly, virus particles were produced after transient co-transfection of HEK293T cells, with the p8.91 encapsidation plasmid, the pHCMV-G vector encoding the vesicular stomatitis virus envelope and the pTRIP vector encoding PML WT and its mutants, using a standard calcium phosphate method.
BRET 1 Transient Transfections-HEK293T cells were maintained in Dulbecco's modified Eagle's medium supplemented with GLUTAmax (Invitrogen), 10% fetal bovine serum, 100 units/ml penicillin and streptomycin, 1 mM sodium pyruvate (all from Invitrogen). Cells were seeded at a density of 5 ϫ 10 5 cells per well in 6-well dishes, 24 h before transfection. Transient transfections were performed using polyethylenimine (PEI, linear, M r 25,000; catalog number 23966 Polysciences, Inc., Warrington, PA) in Opti-MEM medium (Invitrogen). PEI powder was dissolved in water to a concentration of 1 mg/ml in water that was heated to 80°C. Usually, 0.1 g of the PML-Luc construct was transfected alone or with increasing quantities of YFP-tagged SUMO1. The amount of transfected DNA was completed to a total of 2 g with pcDNA3.1(ϩ) empty vector. PEI (10 g in 100 l of Opti-MEM medium) was added on the DNA and the samples were incubated for 20 min at room temperature. The PEI-DNA suspension was then added to the attached cells in 2 ml of fresh culture media. Following an overnight incubation, transfection medium was replaced with fresh Dulbecco's modified Eagle's medium for 3 h to allow cell recovery. Transfected cells (8 ϫ 10 4 ) were then detached and replated in 96-well white plates with clear bottoms (Costar) pre-treated with D-polylysine (Sigma) and left in culture for 24 h before being processed for BRET 1 assay. When required, the plated cells were treated with As 2 O 3 for the appropriate time within this 24 h.
BRET 2 Transient Transfections-HEK293T cells were maintained in Dulbecco's modified Eagle's medium supplemented with 1 mM L-glutamine, 10% fetal bovine serum (Wisent), and 100 g/ml penicillin and streptomycin. Cells were seeded at a density of 1 ϫ 10 6 cells/100-mm dish, 24 h before transfection. Transient transfections of plasmids were performed using the calcium phosphate precipitation method. Usually, 1 g of the PML-Luc construct was transfected alone or with increasing quantities of GFP-tagged SUMO1. The amount of transfected DNA was completed to a total of 10 g with pGEM empty vector as described in Ref. 13. The transfection medium was replaced with the supplemented Dulbecco's modified Eagle's medium after 26 h and cells were left in culture for an additional 22 h before being processed for BRET 2 assay. When required, As 2 O 3 treatment was done on the cells in suspension (see BRET 2 detection assay).

BRET Detection Assays
BRET 1 Assay-BRET 1 measurement was done on attached cells as previously described (21). Before measurement, culture media was replaced by PBS. Coelenterazine H (Nanolight Technology) was then added to a final concentration of 5 M in PBS. Readings were then collected using a multidetector plate reader MITHRAS LB940 (Berthold) allowing the sequential integration of the signals detected in the 480 Ϯ 20 and 530 Ϯ 20 nm windows, for luciferase and YFP light emissions, respectively. The BRET 1 signal was determined by calculating the ratio of the light intensity emitted by the YFP fusion over the light intensity emitted by the Luc fusion. The values were corrected by subtracting the background BRET 1 signal detected when the Luc fusion construct was expressed alone.
For BRET 1 titration experiments, BRET 1 ratios were expressed as a function of the [acceptor]/[donor] expression ratio (YFP/Luc). Total fluorescence and luminescence were used as a relative measure of total expression of the acceptor and donor proteins, respectively. Total fluorescence was determined with MITHRAS using an excitation filter at 485 nM and an emission filter at 535 nM. Total luminescence was measured in the MITHRAS, 10 min after the addition of coelenterazine H and the reading was taken in the absence of emission filter.
BRET 2 Assay-The BRET 2 assays were conducted on cells in suspension as previously described in Ref. 13. In brief, transiently transfected cells were resuspended in PBS. Cells were then distributed in 96-well microplates (white Optiplate from Packard) (20 g corresponding to ϳ1-2 ϫ 10 5 cells). When required, cells were treated with 10 M As 2 O 3 or vehicle for the appropriate time at 37°C. For the BRET 2 assay, upon addition of the cell permeant luciferase substrate (coelenterazine Deep Blue, PerkinElmer Life Sciences), the bioluminescence signal resulting from the degradation of the substrate was detected using a 370 -450-nm band pass filter (donor emission peak 400 nm). The transferred energy resulted in a fluorescence signal emitted by the GFP acceptor (excitation peak 400 nm and emission peak 510 nm) that was detected using a 500 -530-nm band pass filter. The BRET 2 signal was quantified by calculating the acceptor fluorescence/donor bioluminescence ratio as previously reported. Expression level of each construct was determined by direct measurements of total fluorescence and luminescence on aliquots of transfected cell samples. The GFP total fluorescence was measured using a FluoroCount (PerkinElmer Life Sciences) with an excitation filter at 400 nm and an emission filter at 510 nm. The total luminescence was measured using the same cells incubated with coelenterazine H for 10 min (Molecular Probes) (emission peak 485 nm). The BRET 2 was plotted as a function of the GFP/Luc fusion protein expression ratio, both fusion protein expressions being assessed with the same cells as described above, to take into account the potential variations in the expression of individual constructs from transfection to transfection. The values were corrected by subtracting the background BRET 2 signal detected when the Luc fusion construct was expressed alone. All BRET 1 and BRET 2 results presented are derived from two to 10 independent experiments done in duplicate.
Confocal Microscopy-For confocal microscopy, 3 ϫ 10 5 HEK293T cells were seeded 24 h after their transfection onto glass coverslips pre-treated with poly-D-lysine (Sigma) placed within 6-well plates. Cells were then washed three times in PBS 24 h later and fixed in 4% paraformaldehyde for 10 min at room temperature. Following several washes, cells were permeabilized with blocking buffer (PBS containing 0.3% Triton X-100 and 2% bovine serum albumin (w/v)). Monoclonal anti-PML (1/250) and anti-RNF4 (1/100) antibodies were then added for 2 h in blocking buffer. Cells were washed twice and incubated for 1 h with the appropriate Alexa Fluor 488-or 594-conjugated secondary antibody (1/500 dilution, Molecular Probes, Inc.). Coverslips were mounted using Vectashield (Vector Laboratory). Confocal images were obtained on a Leica TCS-NT/SP inverted confocal laser-scanning microscope using an Apochromat ϫ63/1.32 oil-immersion objective. Co-localization was performed by overlay of the images using Leica Confocal Software LCS (Heidelberg, Germany). Excitation and emission filters for the different labeled dyes were as follows: YFP (green): ex ϭ 488 nm, em ϭ 540/25 nm; Texas Red (red): ex ϭ 568 nm, em ϭ 610/30 nm. Western Blot Analysis-For total cell extracts, cells were washed in cold PBS, lysed in hot Laemmli buffer, and boiled for 10 min. Proteins were separated by 10% SDS-PAGE and transferred to nitrocellulose membranes (Optitran BA-S 83 NC, Schleicher and Schuell). The membranes were blocked in 5% gelatin from cold water fish (Sigma) for 2 h and then incubated overnight at 4°C with the first antibody (H-238 anti-PML, anti-FLAG, or anti-actin). Following washes, the membranes were incubated for 1 h with the appropriate Alexa Fluor 680-coupled secondary antibody (Molecular Probes) that was then revealed by an Odyssey Infrared Imaging System (LI-COR Biosciences).

PML/SUMO1 Interactions Are Detected by BRET-To better
understand the role of SUMO in PML function and its involvement in the dynamic of PML NB formation, we developed a BRET assay for the detection of PML interactions with SUMO in living cells. For this purpose, PMLIII was tagged at its carboxyl terminus with a luciferase (PML WT -Luc), whereas SUMO1 was tagged at its amino terminus with YFP or its GFP green variant. We performed a classic BRET titration assay (13,21) by co-transfecting HEK293T cells with a constant amount of PML WT -Luc and increasing amounts of either YFP-SUMO1 (BRET 1 , Fig. 1A) or GFP-SUMO1 (BRET 2 , Fig. 1B) (differences in the BRET assay sensitivity, depending on the fluorescent protein acceptor used, were previously described (24)). In each case, the occurrence of BRET was quantified by measuring the ratio of light emitted by the acceptor (YFP or GFP) and the luciferase donor upon addition of the membrane permeable luciferase substrate, coelenterazine. The BRET ratio was plotted as a function of the YFP/Luc or GFP/Luc fusion protein expressions to take into account the potential variations in the expression of individual constructs.
As shown in Fig. 1, A or B, a strong BRET signal occurred between PML WT -Luc/YFP-SUMO1 or PML WT -Luc/GFP-SUMO1 BRET pairs and increased as a hyperbolic function that is indicative of a specific interaction (10). Similar results were obtained using GFP-SUMO2 as an acceptor protein (data not shown). Negative controls, which exhibited a linear nonspecific BRET signal resulting from random collision ("bystander" BRET (10)), ascertained this specificity: (i) a donor targeting Luc to the nucleus by fusion to a nuclear localization signal (Luc-NLS, Fig. 1B) and (ii) a RING finger PML mutant (PML C57A,C60A -Luc, Fig. 1A) with altered SUMO binding properties (Fig. 3C) (8) that is unable to interact with the only known SUMO E2 ligase UBC9 (25). Altogether, these results validate the use of BRET to detect PML/SUMO1 interactions.
Both Covalent and Non-covalent PML/SUMO1 Interactions Are Detected by BRET-Because BRET is a proximity-based assay, the BRET signal observed with the PML WT /SUMO1 pair may result not only from the covalent (sumoylation) but also from the non-covalent interaction between PML and SUMO (model, Fig. 1H). To evaluate the contribution of the covalent linkage of SUMO to PML WT -Luc relative to its SUMO noncovalent interactions, we compared the BRET titration curve obtained with PML WT -Luc with that of PML 3KR -Luc, its sumoylation-less mutant (with lysines 65, 160, and 490 mutated to arginine), in pairs with YFP-or GFP-SUMO1 (Fig. 1, C and  D). Because the covalent linkage of SUMO is abrogated within the PML 3KR mutant, the difference in the BRET signal between the PML WT and PML 3KR saturation curves depends on PML sumoylation (Fig. 1C); furthermore, the PML 3KR saturation curve mainly revealed the contribution of non-covalent interactions to the BRET signal (Figs. 1C and 7B, model). Thus, it appears that non-covalent interactions represent a major part of the BRET signal in BRET 1 (Fig. 1C) and BRET 2 (Fig. 1D) assays. Strikingly, a linear and low nonspecific BRET signal was obtained with PML WT -Luc and YFP-SUMO1G, a SUMO mutant with a single glycine residue at its COOH terminus, which is unable to be covalently conjugated to proteins (19) (Fig. 1, A and C). Similar results were obtained with YFP-SUMO2G (data not shown). Interestingly, this shows that SUMO1 and -2 paralogs do not interact in their free form in vivo with PML, in contrast with data obtained in vitro (8). Thus, the non-covalent PML/SUMO1 interactions detected by BRET involve SUMO1-modified proteins recruited by PML-Luc as illustrated in Fig. 7B, model.
To further establish the capacity of the BRET assay to detect variations in the sumoylation status of PML in living cells, we tested PML WT -Luc or PML 3KR -Luc/YFP (or GFP)-SUMO1 pairs in the presence of As 2 O 3 . Clearly, a significant increase in the BRET signal was obtained with PML WT upon As 2 O 3 treatment as seen in titration (Fig. 1, C and D), time course (Fig. 1E), and dose-response (Fig. 1F) experiments. In contrast, no or very limited change in the BRET signal was observed with PML 3KR as revealed by the saturation curves (Fig. 1, C and D) and confirmed by the dose-response curve (Fig. 1F). Consequently, the As 2 O 3 -enhanced BRET signal was concluded to be dependent on PML sumoylation. To evaluate the relative contribution of each of the three PML sumoylation target lysines, we compared the As 2 O 3 -induced sumoylation of single PML Lys 3 Arg mutants. Although K490R and K65R single mutants exhibit a partially reduced As 2 O 3 -induced sumoylation compared with PML WT , mutation of Lys 160 alone was sufficient to prevent the As 2 O 3 -induced sumoylation of PML to the same extend as the triple 3KR mutant (Fig. 1G). Thus, Lys 160 is a prerequisite for As 2 O 3 -induced PML modification by SUMO1. These results are in agreement with previous Western blot data (14). Additional experiments revealed that As 2 O 3 exhibited a similar potency to induce PML sumoylation with SUMO1 (log EC 50 ϭ Ϫ6,430 Ϯ 0.181 M), SUMO2 (log EC 50 ϭ Ϫ6,618 Ϯ 0.203 M), or SUMO3 (log EC 50 ϭ Ϫ6,626 Ϯ 0.272 M) as shown by comparing the EC 50 in dose-response curves (Fig. 4A). Thus, we proved here that the sensitivity of both BRET 1 and BRET 2 , which assay conditions differ as described under "Experimental Procedures," was sufficient to monitor dynamic changes in PML sumoylation in living cells. Fig. 7A, is involved in non-covalent recruitment of sumoylated proteins, among them PML itself, as demonstrated by co-immunoprecipitation studies (8). To evaluate the contribution of PML SBD in non-covalent PML/ SUMO1 interactions detected by BRET, we compared the BRET 1 signal of PML WT with that of PML SBD that is mutated in its SBD core sequence (VVVI hydrophobic amino acids; see Fig.  7A) (8). As seen in Fig. 2A, a significant decrease in the BRET signal was obtained with PML SBD relative to PML WT indicating that non-covalent PML/SUMO1 interactions are largely dependent on the PML SBD core sequence. Upon As 2 O 3 treatment, the increase in the BRET profile was quite similar with both PML WT and PML SBD as shown by saturation and doseresponse curves (Fig. 2, A and B). This demonstrates that SBDdependent PML/SUMO non-covalent interactions are not required for As 2 O 3 -induced PML sumoylation.

PML SUMO Binding Domain Is Not Required for As 2 O 3 -induced PML Sumoylation-The identified PML SBD, with the sequence represented in
The limited contribution of sumoylated endogenous proteins to the non-covalent BRET signal is estimated using PML 3KR -Luc fusion (Fig. 1C), which can interact with endogenous sumoylated proteins via its intact SBD. Interestingly, As 2 O 3 induced a slight increase in PML 3KR -Luc BRET signal (Fig. 1C, BRET 1 assay) likely resulting from the enhanced sumoylation of endogenous PML by YFP-SUMO1. In contrast, the BRET signal observed with PML 3KR -Luc was not increased by As 2 O 3 in the BRET 2 assay (Fig. 1D), consistent with the fact that sensitivity of BRET 2 required a much higher level of expression of the donor luciferase fusion than the BRET 1 assay (24) rendering the contribution of the YFP-SUMO1-modified endogenous PML negligible.
Altogether, our results indicate that in the absence of a sumoylation inducer such as As 2 O 3 (basal conditions), most of the PML-Luc/YFP-SUMO1 BRET signal is accounted for by the SBD-dependent non-covalent interaction of endogenous sumoylated proteins with the PML-Luc donor, whereas in the presence of the inducer, the large increase in the BRET signal only reflects a change in PML sumoylation. Our data also indi-cate that the As 2 O 3 -induced PML sumoylation does not require the non-covalent interaction of SUMO1-modified proteins with PML SBD.
Serine Phosphorylation Sites within the SBD Are Required for PML SBD Interactions with SUMO-modified Proteins-Recently, serines acting as targets for CK2 kinase phosphorylation were identified on PMLIV (9). Interestingly, these serines (Ser 560 , Ser 561 , Ser 562 , and Ser 565 in PMLIII are indicated in green in Fig. 7A) are adjacent to the SBD hydrophobic core sequence within PML SBD (8). We hypothesized that these serines, as phosphorylation targets, could functionally regulate the interaction properties of PML SBD. To determine whether these serines may affect the binding of SUMO1-modified proteins to PML SBD, a mutant of these serines (PML S560A,S561A,S562A,S565A -Luc, herein called PML S560 -565A -Luc) was tested by BRET in pair with YFP-SUMO1 (Fig. 3A). Based on the titration curves (Fig. 3A, left panel), the BRET signals calculated for PML WT , PML S560 -565A , and PML SBD at an identical YFP/Luc ratio are presented as a bar graph (Fig. 3A,  right panel). Interestingly, a similar reduction in the BRET signal was observed with PML SBD and PML S560 -565A when compared with PML WT . This indicates that the serines within the SBD are important determinants for the SBD-dependent noncovalent interactions of PML with most of its SUMO-modified protein partners (if not all). Consistent with the hypothesized involvement of the phosphorylation of these serines in these interactions, a BRET signal comparable with that of PML WT was restored when PML S560D,S561D,562D,S565D , a serine phosphomimetic mutant (herein called PML S560 -565D ), was used (Fig. 3B). Altogether, these results indicate that serine phosphorylation sites within the SBD regulate PML SBD-dependent interactions with SUMO1-modified proteins.
It is noteworthy that, as for PML SBD , sumoylation of the PML S560 -565A mutant remains As 2 O 3 -inducible (Fig. 3A). The amplitude of the As 2 O 3 -induced signals was similar for PML WT and PML S560 -565A indicating that the integrity of the polyserine motif is dispensable for As 2 O 3 -induced PML sumoylation. This is consistent with the sumoylation profile obtained by Western blot using extracts from HEK293T cells stably expressing PML WT or its mutants (Fig. 3C). An increase in sumoylation of PML WT , PML SBD , and PML S560 -565A was detected upon addition of As 2 O 3 by a change in the sumoylation pattern toward higher molecular weight SUMO-modified species (as expected, no SUMO1-modified forms could be detected with PML 3KR , PML C57/60 , and the sumoylation-deficient double mutants, PML 3KR/SBD and PML 3KR/S560 -565A ).
Interestingly, the serine phosphorylation-deficient mutant, PML S560 -565A , is localized like PML WT in multiple and dense nuclear dots characteristic of PML NBs as shown by confocal microscopy (Fig. 3D). In contrast, PML SBD is distributed in large and diffuse aggregates with disruption of PML NBs as previously described (8). This result most likely suggests that the polyserine motif (from Ser 560 to Ser 565 ), in contrast to the VVVI core sequence of the SBD, is not required for the forma-  tion of PML NBs. It also suggests that some non-covalent interaction(s) required for the integrity of PML NBs is lost in PML SBD but maintained in PML S560 -565A .
As suggested (26), negative charges imposed either by a stretch of neighboring acidic amino acids and/or phosphorylated serine residues within the SBD of some proteins favors the recruitment of SUMO1 but not SUMO2. Thus, we used the SUMO BRET assay to determine whether differential SBD-dependent non-covalent recruitment of sumoylated proteins by PML could be evidenced with the three SUMO paralogs. Hyperbolic curves of a similar range of amplitudes were obtained with PML WT and the three SUMO paralogs as evidenced by BRET titration experiments (Fig. 4B). Furthermore, the BRET signal was drastically and similarly decreased with all SUMO paralogs for the PML S560 -565A mutant to a level close to those obtained with the PML SBD mutant, as evidenced by BRET titration curves or corresponding bar graphs (Fig. 4, B and C). This indicates that comparable binding of the three SUMO paralogs occurred with PML SBD and that this binding depends on PML from Ser 560 to Ser 565 residues and their phosphorylation.
PML SBD and PML Sumoylation Target Sites Are Both Critical for the As 2 O 3 -induced PML Degradation-It is now known that PML is polyubiquitinated and degraded in a proteasomedependent manner, a process that requires the Lys 160 sumoylation site (14). Consequently, we studied the possible link between PML sumoylation, ubiquitination, and the As 2 O 3 -enhanced PML degradation.
In an attempt to follow the ubiquitination of PML and its mutants in basal conditions and upon exposure to As 2 O 3 , we conducted a ubiquitin BRET 2 assay (10) using PML-Luc fusions in pairs with GFP-UBI. As seen in Fig. 5C, the BRET titration curves obtained with PML SBD , PML 3KR , and PML 3KR-SBD were similar in the presence or absence of As 2 O 3 and show that the BRET signal increased as a function of the amount of GFP-UBI transfected. In contrast, with PML WT and PML S560 -565A , a clear shift of the BRET curves to the right was seen upon As 2 O 3 treatment (Fig. 5A). This is indicative of a change in the expres-  ). B, comparison of the BRET 1 signal between PML WT -Luc and PML S560 -565D -Luc phosphomimetic mutant in pair with YFP-SUMO1 at a fixed YFP/Luc expression ratio. C, Western blot of extracts from HEK293T cells, stably expressing PML WT or its various mutants, submitted or not to As 2 O 3 treatment (5 g, 4 h). PML and its SUMO-modified species are revealed by anti-PML antibody. D, localization by confocal microscopy of PML WT -Luc or its various mutants in transiently transfected HEK293T cells.
sion of the fusions in the presence of As 2 O 3 , because BRET ratios are plotted as a function of the GFP/Luc fusion protein expression. To directly assess how the BRET ratio varies in relation to the expression of PML WT -Luc and GFP-UBI fusions, the BRET ratio, Luc expression, and GFP expression are presented as separate bar graphs for each sample (11 samples) analyzed with the ubiquitin BRET assay in the presence or absence of As 2 O 3 (Fig. 5B). Clearly, neither the BRET signal nor the GFP-UBI expression are significantly changed upon As 2 O 3 addition, whereas a 3-7-fold decrease in the PML WT -Luc expression was observed. Although similar results were obtained with the PML S560 -565A mutant (data not shown), no significant difference in PML SBD -Luc, PML 3KR-Luc, and PML 3KR-SBD -Luc expression was detected upon As 2 O 3 treatment (Fig. 5D and data not shown). The significant decrease in PML WT -Luc and PML S560 -565A -Luc expression, seen as early as 30 min after the addition of As 2 O 3 , is not normally seen in cells where GFP-UBI is not overexpressed. This indicates that overexpression of ubiquitin accelerates the kinetics of As 2 O 3 -induced PML degradation.
Consistent with the results obtained with the ubiquitin BRET assay in transiently transfected cells, PML 3KR and PML SBD stably expressed in HEK293T cells were shown to be resistant, whereas PML WT and PML S560 -565A were sensitive to degradation by a long incubation (16 h) with As 2 O 3 , as revealed by Western blot (Fig. 5E). Altogether these results indicate that As 2 O 3 -induced PML degradation is not dependent on serine phosphorylation sites within the SBD but requires the integrity of both PML sumoylation target sites and the PML SBD core sequence, as also needed for PML NB formation (8).
Both PML Interaction with RNF4 and Its RNF4-induced Degradation Are Reinforced by As 2 O 3 -RNF4 is a poly-SUMO-specific E3 ubiquitin ligase involved in the control of PML stability and implicated in As 2 O 3 -induced PML degradation (15,16). As 2 O 3 , which enhances PML sumoylation, was also suggested to favor PML SUMO-dependent polyubiquitination and proteasome-mediated degradation (15,16).
To further study the mechanisms involved in PML degradation, we assessed the effect of RNF4 on the expression of PML WT and its mutants exhibiting a differential sensitivity to the As 2 O 3 -induced degradation as described above (Fig. 5). First, we overexpressed RNF4 in HEK293T transiently transfected with PML WT . Clearly, RNF4 alone induced a decrease in PML expression (Fig. 6A, lanes 1 and 4), a process accelerated by As 2 O 3 and observed as early as 1 h after addition of this agent (Fig. 6A, lanes 4-6). In contrast, no change in PML level was observed in the absence of overexpressed RNF4 upon a similar As 2 O 3 treatment (Fig. 6A, lanes 1-3). Then, we compared the fate of PML WT , PML S560 -565A , PML 3KR , or PML SBD expressed in HEK293T cells in the presence or absence of overexpressed  PML-Luc fusions are divided into two groups depending on their response to As 2 O 3 treatment: degradation (A and B) and no degradation (C and D). BRET 2 titration curves are in A and C. As confirmed by the bar graphs in B, a shift of the BRET 2 curves to the right seen in the presence of As 2 O 3 in A is indicative of a change in the expression of the Luc fusions (change in the GFP/Luc expression ratio upon As 2 O 3 treatment). Bar graphs for PML WT -Luc (B) and PML SBD -luc (D) in pairs with GFP-UBI as derived from the data in A and C, respectively. For each sample treated or not with As 2 O 3 , the BRET 2 and the expression levels of the PML-Luc fusion are shown (B and D); the expression level of GFP-UBI that essentially remains unchanged upon addition As 2 O 3 for all pairs tested is also shown in B for PML WT -Luc. E, Western blot on extracts from HEK293T cells stably expressing PML WT , PML 3KR , PML SBD , or PML S560 -565A and submitted or not to a 16-h treatment with 1 M As 2 O 3 .
RNF4 and the proteasome inhibitor epoxomycin (Fig. 6C). The RNF4-induced decrease in the PML WT level was inhibited by epoxomycin (Fig. 6C, lanes 1, 2, and 4), indicating that RNF4 alone leads to PML degradation via the proteasome pathway. Similar experiments indicated that RNF4 also induced PML S560 -565A degradation but did not alter the level of PML 3KR and PML SBD (Fig. 6C, lanes 1, 2, and 4). However, in overexpression experiments where significantly higher RNF4/PML ratios were used, a forced degradation of the much more resistant PML SBD and to some extent PML 3KR was also observed (data not shown). Taken together, our results show that in addition to PML sumoylation target sites, efficient proteasome-dependent RNF4-mediated degradation depends on the integrity of the PML SBD core sequence but not of the adjacent phosphorylable serines.
To determine whether the ability of PML mutants to interact with RNF4 can explain their differential sensitivity to RNF4induced degradation, we tested their interaction by a BRET 1 assay using a limiting concentration of a Luc version from PML WT or its mutants in pairs with increasing and saturating concentrations of RNF4-YFP. Clearly, BRET titration curves indicated that PML WT , PML S560 -565A , and unexpectedly PML SBD , interacted with RNF4 and that their interaction was strongly increased by As 2 O 3 , indicating that this drug enhances the affinity of PML and its mutants for RNF4 (Fig. 6B). In contrast, lower BRET signals indicative of weak or no interaction were obtained with PML 3KR and these signals were only slightly increased by As 2 O 3 (Fig. 6B), possibly due to the interaction of PML 3KR with endogenous PML as previously noted (Fig. 1C). Interestingly, the differential sensitivity of PML and its mutants to RNF4-mediated degradation was also evidenced by the clear shift of PML, PML S560 -565A , and PML SBD BRET curves to the right upon As 2 O 3 treatment. Analysis of each individual sample treated or not with As 2 O 3 confirmed that these curve shifts were due in each case to a decrease in the level of expression of the Luc fusions upon As 2 O 3 exposure; in the conditions used, no change in the expression of PML 3KR was seen after As 2 O 3 treatment at any of the RNF4/PML 3KR ratios used. Considering that PML SBD was resistant to degradation upon As 2 O 3 exposure (Figs. 5 and 6C), it was unexpected to find that the interaction of PML WT and PML SBD with RNF4 were similarly increased by As 2 O 3 . Thus, it appears that the capacity to interact with RNF4 is not the only determinant in the differential sensitivityofPMLanditsmutantstoRNF4-inducedproteasomedependent degradation.

Covalent and Non-covalent PML/SUMO Interactions Are Required for As 2 O 3 -and RNF4-induced PML Degradation-
Our results indicate that stimulation of the sumoylation or ubiquitination machinery through overexpression of SUMO1 (not shown), ubiquitin, or RNF4 ligase as well as through As 2 O 3 treatment enhances PML degradation; regulation of the PML level is thus dependent on the relative abundance of the various components of the ubiquitin machinery but also, as less usual, of the SUMO machinery. We also demonstrate here that, in addition to be necessary for NB formation (8), both the covalent linkage of SUMO to PML and the integrity of PML SBD, which mediates non-covalent interactions of PML with SUMO-modified proteins, are required for As 2 O 3 -and RNF4-induced PML degradation (see model in Fig. 7C). The requirement of the integrity of PML SBD for PML degradation provides an explanation for our previous data showing that a COOH terminus mutant, lacking amino acids 555 to 641 of PML and thus PML SBD, which starts at residue 556, completely impairs the As 2 O 3 -induced PML degradation (4).
Interestingly, a functional PML SBD is, however, not required for the As 2 O 3 -enhanced PML sumoylation and for the basal or As 2 O 3 -enhanced PML/RNF4 interaction. We suggest that the increased PML interaction with RNF4 in the presence of As 2 O 3 is most likely due to the As 2 O 3 -enhanced formation of higher molecular weight poly-SUMO chains onto PML (slower migrating species most often observed on blots upon addition of As 2 O 3 , data not shown) and to higher affinity of RNF4 for poly-SUMO chains over monomeric SUMO, as recently reported based on in vitro studies (16). Because inactivation studies revealed that both SUMO2/3 and SUMO1 are required for As 2 O 3 -induced PML degradation (15), we suggest that SUMO2/3 inactivation, by preventing the formation of the polymeric sumoylation chain onto PML, most likely impairs the PML/RNF4 interaction and thus the As 2 O 3 -induced PML degradation. Although As 2 O 3 is suggested to increase the ratio of SUMO1 over SUMO2/3 incorporated in the sumoylated PML (15), it is not yet clear, however, how SUMO1 participates in the  . Models for PML covalent and non-covalent interactions with SUMO dissected by BRET and their implication in As 2 O 3 and/or RNF4-induced PML degradation. A, schematic representation of PMLIII (used in this study) and PMLIV isoforms. The three sumoylation target lysines (K), the conserved RBCC/TRIM motif (RING finger, B box 1, B box 2, coiled-coil region), the NLS, and the SBD (8) are illustrated. The whole sequence defining the SBD with its core hydrophobic amino acids (VVVI, in red) (8) and a region including adjacent phosphorylated serines (green) (9) are represented. B, covalent (sumoylation) and SBD-dependent non-covalent interactions are illustrated as two BRET components discriminated using the PML WT -Luc donor and its PML 3KR and PML SBD mutants in pairs with the YFP-SUMO acceptor. The black arrow indicates the BRET occurring between the Luc donor and YFP acceptor. The loss of transfer of energy resulting from mutation of the sumoylation target lysine or the SBD is highlighted by red dashed arrows. C, model for the involvement of PML and its sumoylation, SBD, and phospho-SBD mutants in NB formation and As 2 O 3 -and/or RNF4-induced PML degradation. The molecular events leading to PML NB maturation are illustrated in a sequential order as an oversimplification to point to the steps prevented by the mutants used in this study and their resistance or sensitivity to As 2 O 3 -and/or RNF4-induced PML degradation.