Promyelocytic Leukemia Protein (PML) Requirement for Interferon-induced Global Cellular SUMOylation*

We report that interferon (IFN) α treatment at short and long periods increases the global cellular SUMOylation and requires the presence of the SUMO E3 ligase promyelocytic leukemia protein (PML), the organizer of PML nuclear bodies (NBs). Several PML isoforms (PMLI-PMLVII) derived from a single PML gene by alternative splicing, share the same N-terminal region but differ in their C-terminal sequences. Introducing each of the human PML isoform in PML-negative cells revealed that enhanced SUMOylation in response to IFN is orchestrated by PMLIII and PMLIV. Large-scale proteomics experiments enabled the identification of 558 SUMO sites on 389 proteins, of which 172 sites showed differential regulation upon IFNα stimulation, including K49 from UBC9, the sole SUMO E2 protein. Furthermore, IFNα induces PML-dependent UBC9 transfer to the nuclear matrix where it colocalizes with PML within the NBs and enhances cellular SUMOylation levels. Our results demonstrate that SUMOylated UBC9 and PML are key players for IFN-increased cellular SUMOylation.

Interferons (IFNs) 1 have been first recognized for their antiviral properties, but also have important immunomodulatory, anti-proliferative and apoptotic activities (1)(2)(3). They act on cells by interacting to their respective receptors, activating the JAK/STAT pathways and inducing the transcription of IFNstimulated genes (ISGs) (3,4). IFNs are classified in three types (1)(2)(3). In humans, type I IFN includes IFN␣, IFN␤, IFN, IFN, and IFN, type II IFN comprises only IFN␥ whereas type III IFN contains IFN1 to IFN4. Although type I and type III IFNs bind to different cell surface receptors, they initiate the same signal transduction pathway leading to the activation of the JAK tyrosine kinases, Tyk2 and JAK1, that in turn phosphorylate STAT1 and STAT2, which heterodimerize and form, with the DNA binding protein IFN regulatory factor 9 (IRF9), a complex called IFN-stimulated growth factor 3 (ISGF3) (3).
ISGF3 then translocates to the nucleus and binds to ISG promoters containing an IFN-stimulated response element (ISRE). There are also growing evidences that type I IFN activates a STAT2/IRF9 complex that forms an ISGF3-like complex in the absence of STAT1, and is sufficient to activate ISRE-driven transcription (5,6). The interaction of type II IFN to its specific receptor induces the phosphorylation of STAT1 by JAK1 and JAK2, the homodimerization of activated STAT1, their migration to the nucleus and binding in the promoter regions of ISGs to an element termed gamma-activated sequence (GAS).
It has been established that many of the ISGs or key regulators of IFN signaling are modified by ubiquitin or ubiquitinlike modifiers such as Small Ubiquitin-like Modifier (SUMO) or ISG15 (7,8). Protein modification by SUMO termed SUMOylation involves the covalent binding of SUMO to a target protein K residue that is regulated by a unique SUMO E2-conjugating enzyme named UBC9. There are five SUMO family members; the best studied being SUMO1, which shows ϳ50% sequence homology with SUMO2 and SUMO3 that are themselves 97% identical to each other. Although SUMO1, 2, and 3 have a broad tissue distribution, the expression of SUMO4 is limited to certain organ types (9) and contains a proline residue close to the diglycine motif, which prevents its maturation and conjugation (10). SUMO5, a novel tissue-specific member of the SUMO family has been shown to facilitate the formation of promyelocytic leukemia (PML) nuclear bodies (NBs) in human cells but does not appear to be expressed in mice (11).
Although SUMO1 and SUMO2/3 are conjugated via the same set of enzymes, they can preferentially target different substrates and demonstrate distinct dynamics and distribution within the cell (12). SUMO2/3 is more abundant in a large free pool compared with SUMO1 and is also more dynamically conjugated and deconjugated from substrate proteins (13). Also, the covalent modification of proteins with SUMO2/3 may have varying consequences compared with SUMO1 modification (7,14). In addition to being covalently attached to substrates, SUMO can interact in a noncovalent manner via the SUMO interaction motifs (SIMs) (15).
It has been reported that STAT1 is conjugated to SUMO1 at K703 and that mutations in the SUMO attachment site in STAT1 result in increased transcriptional activity in response to IFN␥ (16). More recently, we reported that the stable expression of the different SUMO paralogs leads to STAT1 SUMOylation and to a decrease in IFN␣and IFN␥-induced STAT1 phosphorylation that result in an inhibition of IFN␥ transcriptional response without affecting the IFN␣ pathway (7). Also, protein SUMOylation contributes to the antiviral effects of IFN␣ against HSV1 or HIV, though only a limited number of SUMO targets are presently known (17). So far, some ISG products (STAT1, IRF-1, IRF-2, IRF-3, IRF-7, PML, Sp100, p53, TRIM5␣) have been shown to be SUMOylated in transfected cells (18), and cell treatment with IFN for 24h was previously reported to increase SUMOylation (17). However, it is unknown whether IFN␣ treatment for short or long periods could alter the SUMOylation of proteins that regulate IFN pathway. Therefore, we sought to identify proteins conjugated to SUMO3 in response to IFN␣ using a novel proteomics approach based on SUMO remnant immunoaffinity purification (19). Remarkably, Kinetics studies revealed that cellular SUMOylation was enhanced at short (0.75 h) and long (16 h) periods following IFN␣ treatment, and necessitated the presence of PML protein, the key organizer of NBs (20). Also, we found that several SUMO sites were regulated in response to IFN␣ including UBC9 K49. These observations support the notion that the recruitment of UBC9 to PML NBs upon IFN treatment cooperated with PML to enhance global cellular SUMOylation.
Experimental Design and Statistical Rationale-The quantitative analysis of changes in the SUMO proteome on IFN␣ stimulation was conducted in technical triplicates. The light channel was used as an internal nonstimulated control whereas the medium and heavy channels were treated for 0.75 h and 16 h with IFN␣, respectively. The combined SILAC samples were split into three technical replicates after the cell fractionation step. The Ni-NTA purification, tryptic digestion, peptide desalting and SUMO peptide IP experiments were performed in a parallel fashion. Each technical replicate was injected once on the LC-MS system. Technical triplicates were conducted to obtain statistically robust data that could be assessed using onesample Student T-tests because of the normal Gaussian distribution of the log 2 of the SILAC fold-change values. Because of the amount of starting material needed per biological replicate (500 million cells per condition per replicate) and the cost of the SILAC media and anti-NQTGG antibody, multiple biological replicates was not possible.
For each of the technical triplicates, 24 mg of cytoplasmic or nuclear protein were incubated with 960 l of Ni-NTA beads for 16 h at 4°C. Ni-NTA beads were washed once with 10 ml of Ni-NTA denaturing incubation buffer, 4 times with 10 ml of Ni-NTA denaturing washing buffer (8 M urea, 100 mM NaH 2 PO 4 , 10 mM Tris-HCl, 20 mM imidazole, 5 mM 2-Mercaptoethanol, 20 mM 2-Chloroacetamide) and finally twice with 5 ml of 100 mM ammonium bicarbonate. The protein content was determined by micro Bradford assay. On beads protein digestion was performed by adding trypsin to a ratio 1:50 (w:w) Trypsin/Protein extract for 4 h at 37°C. Resulting Ni-NTA enriched digests were acidified by adding TFA to a final concentration of 1%, desalted on 1cc HLB cartridges as per the manufacturer's instructions and eluted into microfuge tubes prior to their lyophilization by SpeedVac.
For each sample, 163 l of PureProteome protein A/G magnetic beads (1 l of slurry per 4 g of Ni-NTA purified protein) were equilibrated with 326 g of anti-K(NQTGG) antibody (2 g of antibody per 1 l of protein A/G slurry) for 1 h at 4°C in PBS. Saturated beads were washed 3 times with 200 mM triethanolamine, pH 8.3. 1.63 ml of 5 mM DMP in 200 mM triethanolamine, pH 8.3 was added per sample and incubated 1 h at room temperature to crosslink the antibody to the beads. The reaction was quenched for 30 min by adding 81 l of 1 M Tris-HCl, pH 8. The beads were washed 3 times with ice cold PBS and once with PBS containing 50% glycerol and stored at Ϫ20°C until future use. The lyophilized peptides were reconstituted in 500 l of PBS containing 50% glycerol by vortexing for 10 min at the highest speed. The anti-K(NQTGG) bound beads were added to the peptide mixture and incubated 1 h at 4°C. The beads were washed once with 1 ml PBS containing 50% glycerol, 6 times with 1 ml of ice cold 1X PBS, once with 1 ml of 0.1ϫ PBS and once with water. Peptides were eluted from the beads with 3 successive elutions using 200 l of 0.2% formic acid in water and filtered through a 0.45 m spin tube. Eluted peptides were dried down by speed vac and stored at Ϫ80°C for MS analysis.
Mass Spectrometry Analysis-Peptides were reconstituted in 25 l of water containing 0.2% formic acid and 10 l of this mixture was injected by nanoflow-LC-MS/MS using an Orbitrap Fusion Mass spectrometer (Thermo Scientific) coupled to a Proxeon Easy-nLC 1000. Samples were separated directly on a 150 m ϫ 20 cm nano LC column (Jupiter C18, 3 m, 300 A, Phenomenex) without a trapping column. The separation was performed on a linear gradient from 7% to 30% acetonitrile, 0.2% formic acid over 105 min at 600 nl/min. Full MS scans were performed on ion from m/z 350 to m/z 1500 at resolution 120,000 at m/z 200, with a target AGC of 5E5 and a maximum injection time of 200 ms. MS/MS scans were acquired in HCD mode with a normalized collision energy of 25 and resolution 30,000 using a Top 3 s method, with a target AGC of 5E3 and a maximum injection time of 3000 ms. The MS/MS triggering threshold was set at 1E5 and the dynamic exclusion of previously acquired precursor was enabled for 20 s within a mass range of Ϯ 0.8 Da. Ions with charges Ͼ10 and Ͻ2 were excluded from triggering MS2 events.
Data Processing-MS/MS spectra were searched against Uniprot/ SwissProt database including Isoforms (released on March 10, 2015, 42 084 entries) using MaxQuant (version 1.5.1.2) (22). The precursor ion tolerance was set to 20 and 4.5 ppm for the first and main searches, respectively. The MS/MS spectra search were set to a mass tolerance of 20 ppm. The maximum missed cleavage sites were set to 2 using trypsin/P as protease. Carbamydomethylation (C) was set as fixed modification and acetylation (Protein N term and K), phosphorylation (S), oxidation (M), deamination (NQ), and NQTGG (K) were set as variable modifications. Searches were conducted with the match between runs function enabled with a 20 min alignment window and 0.7 min match time window. Identified proteins and SUMO sites were filtered with a 1% FDR using the reverse database as the decoy. MS/MS spectra for modified peptides with an Andromeda score below 40 (default values) were discarded from further analysis.
The MaxQuant output files were processed using the R-software. Possible improper SUMO site identification was filtered out by removing all "potential contaminants," "reverse" sites from the list. Further, the SUMO site list was filtered with a SUMO site probability score of 0.75 or greater, which is routinely used for PTM site identification methodologies. The reported "normalized" SILAC ratios generated by the MaxQuant software were Log 2 transformed prior to their statistical analysis. Imputations were employed for SUMO sites that were quantified in 2 of 3 triplicates using normally distributed values with a randomized 0.3 width (log 2 ) and a 1.8 down shift (log 2 ). Sites were deemed statistically regulated by IFN␣ if their p values/fold-change combination met the permutation-based FDR of 5%, which was corrected using the significance analysis of microarrays method (S 0 correction factor) to consider the standard deviation of the data set (23). The volcano plots shown in Fig. 2B-2E, were created by using the one-sample t test results on the y axis (-log 10 (p value)) and the Log 2 FC between the IFN␣ treated samples compared the control samples. The corresponding dashed lines on the volcano plots depict the boundaries of statistically regulated SUMO sites based on an FDR of 5% that is adjusted using the S 0 correction factor.
Bioinformatic Analysis-Protein networks were created using STRING database with experimentally mapped interaction with a medium confidence of 0.4 (24). Networks were generated using all identified SUMOylated proteins in this study. Cytoscape 3.2.0 was used to visualize the network (25). Gene ontology (GO) term enrichments were performed in cytoscape with Bingo 3.0.3 using a Benjamini-Hochberg corrected p value below 0.01 as the cutoff for statistically significantly regulated terms (26). The following terms were analyzed versus the human proteome: Biological Processes (GOBP), Molecular Functions (GOMF), and Cellular Compartments (GOCC).
Immunofluorescence Analyses-Cells were grown on 12 mm slides and fixed with 4% paraformaldehyde for 15 min, rinsed in PBS, incubated in NH 4 Cl 50 mM for 10 min and permeabilized with 0.5% BSA/0.3% Triton X-100/2% Normal Goat Serum for 30 min. Cells were then incubated with primary antibodies at room temperature for 30 min. Slides were rinsed in PBS 0.5% BSA and incubated at RT in the dark for 30 min with secondary antibodies and rinsed in PBS 0.5% BSA. Finally, slides were washed in PBS, counterstained with Hoechst 33342 and mounted in Fluoromount-G medium. Images were digitally acquired with a Zeiss LSM 710 confocal Microscope.
Western Blot Analysis of Total Cell Extracts and Cytoplasmic, RIPA Soluble and Insoluble Fractions-For total cell extracts, cells were washed in PBS, lysed in hot Laemmli sample buffer, and boiled for 10 min. The cytoplasmic fraction was obtained by lysing the cells in Tris-HCl 10m M pH 7.6, MgCl 2 1.5 mM 0.5%, NEM 20 mM, DTT 1 mM 20 min at 4°C. Cells were then centrifuged at 500 ϫ g for 15 min to separate nuclei (pellet) from the cytosol (C). The RIPA soluble fraction was extracted by incubating the pellet for 20 min on ice in RIPA buffer (50 mM Tris, pH 7.5, 200 mM NaCl,1% Triton X-100/1%, deoxycholate 0.5%, SDS 0.1%, 1 mM EDTA), followed by centrifugation at 15000 ϫ g for 15 min to separate the RIPA soluble fraction (R) from the pellet (P). The RIPA insoluble fraction (P) was washed two times in RIPA buffer, suspended in PBS and boiled in Laemmli buffer. Protein extracts were separated by SDS-PAGE followed by electroblotting onto nitrocellulose membrane. After blocking of nonspecific binding sites with 5% nonfat milk, the membranes were incubated with primary antibody (SUMO1, SUMO2/3, PML, PKR, pSTAT1, pSTAT2, STAT1, STAT2, p53, UBC9, p21 or Actin) followed by horseradish peroxidase-conjugated secondary antibody. The secondary antibody was detected with the ECL chemiluminescence detection system.
Real-time PCR-Total RNAs were extracted using RNeasy Mini Kit (Qiagen) following manufacturer's instructions. RNA samples were converted to cDNA using RevertAid H Minus First Strand cDNA Synthesis Kit (Thermo Scientific). Real-time PCR reactions were performed in duplicates using 5 l of cDNA diluted 10 times in water using Takyon ROX SYBR MasterMix blue dTTP (Eurogentec). The following program was used on a 7900HT Fast Real-Time PCR System (Applied Biosystems): 3 min at 95°C followed by 35 cycles of 15 s at 95°C, 25 s at 60°C and 25 s at 72°C. Values for each transcript were normalized to expression levels of RPL13A (60S ribosomal protein L13a) using the 2-⌬⌬Ct method. Primers used for quantification of transcripts by real time quantitative PCR are as followed: 5Ј-AGGGATGGGGTGGATGAGG-3Ј and 5Ј-GGGGTATAT-GATGGGGGAGTAG-3Ј for p21.
Apoptosis-Cell apoptosis was assessed using Annexin V-FITC/PI Kit (BD Biosciences). Briefly, HEK293-wt and HEK293-SUMO3m were untreated or treated with 1000 U/ml of IFN␣ for 72 h. Cells were collected, resuspended in 100 l of PBS and stained with Annexin V for 15 min at 4°C, followed by PI staining. Fifty thousand cells were analyzed by flow cytometry on a FACSCalibur (BD Biosciences).

Enhancement of Global Cellular SUMOylation in Response
to IFN␣-Initially, we treated HEK293-SUMO3m cells with IFN␣ for short (0.75 h) or long periods (16 and 24 h) to determine the impact of this cytokine on global cellular SUMOylation. Ni-NTA purification of total extracts from HEK293-6xHis-SUMO3m cells revealed that SUMO3-modified proteins increased at 0.75 h and were more abundant at 16 h following IFN␣ treatment (Fig. 1A). Western blot analysis of cell extracts indicated that SUMO3 expression decreased IFN␣-induced STAT1 phosphorylation without affecting the abundance of phosphorylated STAT2 (Fig. 1B), consistent with a previous report (7). Indeed, phosphorylation of STAT2 in response to IFN␣ was similar in both in HEK293-wt and HEK293-SUMO3m cells, whereas phosphorylation of STAT1 was lower at 0.75 h in HEK293-SUMO3m cells compared with HEK293-wt cells. Also, ISG products such as STAT1 or PKR were enhanced in both cell lines after 16 h of IFN␣ treatment showing that IFN␣ response was not inhibited in SUMO3mexpressing cells.
Quantitative proteomics analyses were performed using stable isotope labeling with amino acids in cell culture (SILAC). Cells were incubated with PBS (Control cells), or IFN␣ for short (0.75 h) or long (16 h) time periods using cultures grown in either light, medium or heavy SILAC media, respectively ( Fig. 2A). After IFN␣ treatment the cell pellets were pooled during the collection process prior to fractionation into cytoplasmic and nuclear fractions under hypotonic conditions. Cytoplasmic and nuclear protein extracts were divided into 3 aliquots to produce technical triplicates. Each nuclear and cytoplasmic triplicate was subjected to Ni-NTA purification to enrich SUMOylated proteins. The Ni-NTA bound material was digested with trypsin to release the NQTGG epitope produced by the SUMO3m construct. SUMOylated peptides were further purified using the anti-K(NQTGG) antibody prior to LC-MS/MS analyses.
A summary of the identified and quantified SUMO sites is reported in supplemental Table S1. In total, we identified 558 SUMO sites with a FDR Յ 1% and a SUMO site localization confidence Ն 0.75 across all time points and fractions. Quantitative proteomics analyses enabled the profiling of 264 and 346 sites in cytoplasm and nuclear fractions for the two time points examined. Several SUMO sites were regulated in response to IFN␣ treatment (Figs. 2B-2E). In contrast to the Western blot analysis that depicts a substantial increase in global SUMOylation upon IFN␣ treatment (Fig. 1A), the proteomics analysis revealed that a similar number of SUMO substrates were up-regulated and down regulated because of the treatment. These seemingly conflicting results can be consolidated when looking at the intensity or abundance of the SUMOylation events that are up-regulated in response to IFN␣ (supplemental Fig. S1). The RankN plots that rank the peptide in ascending order of intensity highlight that SUMOylation sites that are up-regulated during the IFN␣ treatments   (Figs. 2D, 2E) and in the cytoplasm at 16 h (Fig. 2C). We also observed the decrease in SUMOylation of STAT1 at K 703 in the cytoplasm at 0.75 h and its increase at 16 h following IFN␣ stimulation (Figs. 2B, 2C). To distinguish changes in protein SUMOylation (Fig. 3A) from those associated with protein abundance, we analyzed the cytoplasmic and nuclear extracts by Western blotting (Fig. 3B). PML and STAT1 are ISG products and their protein expression increased following extended IFN␣ treatment in HEK293-wt and HEK293-SUMO3m cells (Fig. 3B). Therefore, the enhancement of PML and STAT1 SUMOylation at 16 h (Fig. 3A) could be due in part to the increase of their protein levels. In contrast, IFN␣ treatment significantly increased UBC9 SUMOylation without altering its total protein level (Fig. 3B).
Because of the variation in the mixing of the three SILAC channels during the sample preparation stage the changes in SUMOylation site abundance in response to IFN␣ were obtained from the MaxQuant normalized SILAC ratios, which may have led to the loss of some regulated SUMO sites. Indeed, the normalization of the SILAC ratios may have forced SUMO sites that were weakly regulated in response to the treatment to have a greater p value. However, we found that the error in SILAC mixing was consistent in all three technical replicates, which suggests that the workflow adopted here was reproducible (supplemental Figs. S2-S4).
SUMOylation of Endogenous PML, STAT1, and pSTAT1-Although we did not detect an increase in PML SUMOylation at short IFN treatment by quantitative proteomics analyses (Fig. 2), we evaluated changes in endogenous protein SUMOylation in Ni-NTA-purified extracts from wt and SUMO3m cells treated with IFN␣ or IFN␥ for 0.75 h. PML immunoblots of Ni-NTA affinity-purified extracts from HEK293-SUMO3m cells showed that PML was SUMOylated and that IFN␣ or IFN␥ enhanced PML SUMOylation (Fig. 3C), with a further increase in PML modification when IFN␣ or IFN␥ was combined with the proteasome inhibitor MG132.
Recently, we reported that SUMO3 expression in HeLa cells showed a decrease in STAT1 phosphorylation without affecting the transcriptional response when cells were stimulated with IFN␣ (7). Here, we observed a similar decrease in IFN␣-induced STAT1 phosphorylation in HEK293-SUMO3m compared with HEK293-wt cells (Fig. 3D). STAT1 was found conjugated to SUMO3 in HEK293-SUMO3m cells, consistent with our proteomic analysis (Fig. 2B), and a short IFN␣ treatment (0. 75 h) reduced slightly the level of STAT1 SUMOylation (Fig.  3D). As expected, phosphorylation of STAT1 was observed only in response to IFN␣, and immunoblots of Ni-NTA-purified extracts from HEK293-SUMO3m using anti-pSTAT1 antibodies revealed that pSTAT1 was conjugated to SUMO3 (Fig. 3D).
To determine the stability of the corresponding proteins, we performed Ni-NTA-purification of cytoplasmic and nuclear extracts from HEK293-SUMO3m cells with and without proteasome inhibition for 4 h before adding IFN␣ for 0.75 h (Fig. 3E). Analysis of the inputs revealed that pSTAT1 was detected in the cytoplasm and the nucleus (Fig. 3E), though pSTAT1 migrated at a lower molecular weight in the latter fraction. This shift was associated with pSTAT1 SUMOylation because the corresponding band was only observed in the cytoplasmic fraction of the Ni-NTA purified extract of IFN␣-treated cells (Fig. 3E). Furthermore, SUMOylated pSTAT1 was not stabilized with MG132, suggesting that SUMOylation did not lead to the degradation of pSTAT1. Immunoblots analyses of the inputs and Ni-NTA fractions also revealed that most SUMOylated substrates in IFN␣-treated SUMO3 cells were nuclear proteins (Fig. 3E).
Previous reports indicated that IFN␣ induced p53-dependent apoptosis (27,28) and that modification of p53 at K386 by SUMO1 induces p53-dependent transcription (29). Immunoblot of p53 following Ni-NTA purification of control and IFN␣treated HEK293-SUMO3m proteins revealed that p53 was SUMOylated in HEK293-SUMO3m cells, and that this modification was enhanced after 0.75 h of IFN␣ treatment (supplemental Fig. S5A). The increased p53 SUMOylation in SUMO3m-expressing cells was also accompanied by a positive regulation of IFN␣-induced activation of p53 target gene p21 (supplemental Figs S5B, S5C) and a higher IFN␣-induced apoptosis (supplemental Fig. S5D), suggesting that enhanced p53 SUMOylation in response to IFN␣ plays a role in this process.
PML is Required for IFN-induced Global SUMOylation and UBC9 Transfer to the Nuclear Matrix-PML has been reported to exert a SUMO E3 ligase activity and may mediate the SUMOylation of many PML NB interacting proteins (30). As PML plays a key role in the IFN response (31, 32), we next examined whether the increase of protein SUMOylation in response to IFN requires PML. Accordingly, we analyzed the  Table S1. changes in SUMOylation following IFN treatment in wt mouse embryonic fibroblasts (MEFs) and PMLϪ/Ϫ MEFs untreated or treated with IFN␣ for a short period (0.75 or 1 h) (Fig. 4A) or with IFN␣ or IFN␥ for a longer period (16 h) (Fig. 4B). In all treatments, IFNs enhanced SUMO2/3 conjugates in wt MEFs but not in PMLϪ/Ϫ MEFs. Similarly, IFN␣ and IFN␥ promoted the increase of cellular SUMO1 modification in the presence of PML, and extended IFN treatment increased the levels of PML in wt MEFs (Fig. 4B).
These results suggest that endogenous PML is required to enhance cellular SUMOylation in response to type I and II IFNs. Next, we investigated whether a specific PML isoform is implicated in the increase of SUMOylation in response to IFN. Several PML isoforms generated by alternative splicing from a single gene are designated PMLI to PMLVII (33,34). They share the same N-terminal region, which encodes the RBCC/ TRIM (RING finger, B-box, and Coiled-Coil) motif, but differ in their C-terminal region because of the alternative splicing. PMLϪ/Ϫ MEFs transduced with retroviral vectors expressing each one of the seven human PML isoforms (PMLI to PMLVII) were then treated with IFN␣ to determine the role of the various C-terminal regions of PML. The extent of protein SUMOylation was found to vary significantly for the different PML isoforms examined (Figs. 4C, 4D). IFN␣ was unable to stimulate global cellular SUMOylation in PMLϪ/Ϫ cells expressing PMLI, PMLII, PMLV, PMVI or PMLVII (Fig. 4C). Also, none of these PML isoforms increased SUMO2/3 conjugates in untreated cells. In contrast, cellular SUMOylation was stimulated in untreated PMLϪ/Ϫ cells expressing PMLIII or PMLIV with a further increase upon IFN treatment (Fig. 4D). Collectively, these results show that the enhancement of cellular SUMOylation by IFN required endogenous PML and that this process is orchestrated by PMLIII and PMLIV.
UBC9, the unique E2 SUMO conjugating enzyme, is known to interact with the RING finger domain of PML (35), a protein acting as a SUMO E3 ligase (30). The observation that UBC9 is differentially SUMOylated upon IFN␣ stimulation and that this treatment enhanced protein SUMOylation via PML, prompted us to examine the interrelationship between PML and UBC9. Accordingly, we analyzed by immunofluorescence (Fig. 5A) and immunoblotting (Fig. 5B) the localization and expression of UBC9 in wt MEFs and PMLϪ/Ϫ MEFs untreated or treated with IFN␣. Analysis by confocal microscopy shows that UBC9 was found in the nucleus and the cytoplasm of untreated cells (Fig. 5A). In untreated wt MEFs expressing endogenous PML, UBC9 formed nuclear dots distinct from those of PML NBs. IFN␣ treatment, which enhanced PML SUMOylation at 0.75 h (Fig. 3C) and 16 h (Fig. 3A) as well as PML expression at 16 h (Figs. 3C, 4B), resulted in the increase of PML NB size, where PML and UBC9 partly colocalized (Fig.  5A). In contrast, UBC9 was localized in the nucleoplasm and the cytoplasm of control and IFN␣-treated PMLϪ/Ϫ MEFs (Fig. 5A).
To determine whether IFN␣ treatment induced the transfer of UBC9 to the nuclear matrix, and whether PML is implicated in this process, we examined the subcellular localization of UBC9 and PML in untreated and IFN␣-treated wt MEFs and PMLϪ/Ϫ MEFs. Cells were fractionated into cytoplasmic, RIPA soluble (nucleoplasm) and RIPA-insoluble fractions (nuclear matrix and some chromatin components). In PMLϪ/Ϫ MEFs, UBC9 was found in the cytoplasmic and nucleoplasmic fractions of control or IFN␣-treated cells (Fig. 5B). Similar localization of UBC9 was observed in untreated wt MEFs, where PML was mainly localized in the nucleoplasm fraction with a small fraction in the nuclear matrix (Fig. 5B). Remarkably, IFN␣ treatment, which enhanced PML expression and its conjugation to SUMO, shifts UBC9 toward the nuclear matrix (Fig. 5B). To determine whether the expression of PMLIII was able to recruit UBC9, PMLϪ/Ϫ MEFs expressing PMLIII were treated with murine IFN␣. As observed in Fig. 5C, PMLIII was able to recruit endogenous UBC9 within PML NBs, where both proteins colocalized in untreated and IFN␣-treated MEFs. PML NBs became larger upon IFN␣ treatment because of PML SUMOylation and the recruitment of PMLIII partners. Taken together, our results show that IFN␣ increased PML expression, its conjugation to SUMO3, and the transfer of UBC9 to the nuclear matrix where UBC9 and PML colocalized within PML NBs. DISCUSSION IFNs play essential roles in modulating immune response against host infections via the induction of ISGs through the JAK/STAT pathway. This is achieved in part through protein modifications of key regulator of the IFN signaling transduction machinery. SUMOylation is an important modification implicated in intrinsic and innate immunity regulating IFN production, IFN signaling, the localization and the activity of many ISG products (18). Although SUMO decreases STAT1 phosphorylation in response to IFN␣, it does not alter STAT2 phosphorylation, or the formation of an ISGF3-like complex responsible for binding to ISRE and inducing transcriptional and biological responses (7). Accordingly, accumulating evidences support the existence of alternative STAT2 signaling pathways that are independent of STAT1 (36,37). We show here first that in HEK293-SUMO3m cells, STAT1 but not STAT2 activation is lower in response to IFN␣ compared with wt cells, and second that p53 SUMOylation increases upon IFN␣ treatment, correlates with higher levels of p21 protein expression, enhanced apoptosis and as previously shown induction of cellular senescence (38). Taken together these experiments confirmed the functionality of the His 6 -SUMO3 mutant in response to IFN␣. Importantly, we report here that in response to IFN␣, cellular SUMOylation was enhanced at a short period (0. 75 h) and reached a maximum 16 h post-treatment. The global increase in protein SUMOylation requires PML. Previous reports indicated that SUMO3 expression alone did not alter PML protein levels, though its proteasome-dependent degradation was noted for IFN␣ treatment beyond 18 h (7). This observation is consistent with the decreased protein SUMOylation observed here when cells were exposed to IFN␣ for 24 h (Fig. 1). The identification of SUMOylated proteins in the cytoplasm and the nucleus at short and long IFN␣ treatment revealed that a large proportion of the SUMO proteome was regulated by IFN␣ (172 out of 558 SUMO sites). Among the identified proteins, some are ISG products (e.g. PML, STAT1, ADAR1, Vimentin) and others are regulators of IFN signaling or IFN production (e.g. PML, STAT1, Tif1␣/TRIM24, TRIM28, TRIM33). In addition, we found that most of the proteins from the SUMOylation machinery (e.g. UBC9, SUMO1, UBC9 is SUMOylated at K14, K18, K49, K65 and K154 in vitro and its SUMOylation at K14 displayed enhanced binding to SIM-containing proteins (39,40). However, from this list only SUMOylation at K49 was identified in vivo. We identified SUMOylation at K48 and K49 of UBC9 and have shown that IFN␣ enhanced UBC9 SUMOylation at K49 at an early time point in the nucleus and increased SUMOylation at K49 in both the cytoplasm and nucleus during prolonged IFN␣ treatment without altering its protein level. In untreated cells, UBC9 is localized in the cytoplasm and the nucleoplasm, but migrates to the nuclear matrix in response to IFN␣ in a PMLdependent manner. PML and PML NBs play a key role in the IFN response (25,(41)(42)(43). SUMOylation of PML is critical for the formation and function of PML NBs and it affects PML localization, stability and ability to interact with other partners. In addition, PML has SUMO E3 ligase activity, which may mediate the SUMOylation of many PML NB-associated proteins (30). It is noteworthy that PML negative cells have a defect in IFN-induced biological response (21) as well as in IFN-induced global cellular SUMOylation. Many restriction factors and key regulators of IFN pathway are SUMOylated and required this modification for their functions (18). Remarkably, introducing PMLIII or PMLIV in PML negative cells restores IFN-enhanced global cellular SUMOylation and therefore IFN functions.
The present study indicates that endogenous PML is required for increased SUMOylation in response to IFN␣ or IFN␥, and that this process is mediated by PMLIII and PMLIV. The mechanism by which the isoform-specific protein sequences enhance cellular SUMOylation is unclear. It is likely that they bind factors that enhance or interfere with the IFN pathway. Nuclear PML is the organizer protein of the PML NBs. All six human nuclear PML isoforms (PMLI-PMLVI) are able to form NBs when expressed in PML-negative cells (44). Although PML isoforms may have related functions because of their common functional RBCC/TRIM domain, increasing evidences suggest that the variability in the C-terminal region confers specific functions to each PML isoform (34). It has been reported that PMLIV acts as a SUMO E3 ligase enhancing the SUMOylation of various proteins (30). We show here that PMLIII or PMLIV expression stimulated overall SUMOylation in untreated PMLϪ/Ϫ cells and are critical for IFN increased global cellular SUMOylation. The RING domain and the B-boxes, shared by all PML isoforms, are likely required for the E3 activity of PML (30). A SUMO E3 ligase is expected to bind both UBC9 and its substrates. PML interacts with UBC9 via its RING motif, the C-terminal region specific to PMLIII and PMLIV could be implicated in the interaction with the various substrates and/or to a higher recruitment of UBC9 to PML NBs. We report here that expression of PMLIII in PML negative cells was able to recruit UBC9 to PML NBs where both proteins colocalized. Further investigations are needed to determine whether the UBC9 recruitment to PML NB is specific to PMLIII and PMLIV.
IFNs directly induce the PML gene resulting in the increase of different PML isoforms (45,46). This effect could be partly responsible for the enhanced cellular SUMOylation observed during longer periods of IFN treatment. In addition, we show that IFN␣ induced the SUMOylation of PML and UBC9 and favored the translocation of UBC9 to the nuclear matrix to promote their co-localization within PML NBs. These findings suggest that PML and UBC9 act in a cooperative manner to enhance cellular SUMOylation upon IFN␣ stimulation, further demonstrating that PML NBs are a hub for protein SUMOylation.
Collectively, our findings lead to the following conclusions which are illustrated in Fig. 6: (i) IFN enhances cellular SUMOylation in a PML-dependent manner as early as 0.75 h post-treatment, with an increase of PML and UBC9 SUMOylation; (ii) IFN induces PML gene transcription resulting in an increase of PML isoforms 16 h post-treatment; (iii) IFN induces PML-dependent UBC9 translocation to the nuclear matrix, leading to its recruitment to PML NBs; The increase of PML expression and the recruitment of UBC9 within PML NBs promotes the enhancement of SUMOylation in response to IFN; (iv) PMLIII and PMLIV are key players of the IFN-induced increase of cellular SUMOylation. These novel findings provide further biological insights into the SUMO pathway, the contribution of PML and IFN response.