Promyelocytic Leukemia Protein Isoform II Promotes Transcription Factor Recruitment To Activate Interferon Beta and Interferon-Responsive Gene Expression

To trigger type I interferon (IFN) responses, pattern recognition receptors activate signaling cascades that lead to transcription of IFN and IFN-stimulated genes (ISGs). The promyelocytic leukemia (PML) protein has been implicated in these responses, although its role has not been defined. Here, we show that PML isoform II (PML-II) is specifically required for efficient induction of IFN-β transcription and of numerous ISGs, acting at the point of transcriptional complex assembly on target gene promoters. PML-II associated with specific transcription factors NF-κB and STAT1, as well as the coactivator CREB-binding protein (CBP), to facilitate transcriptional complex formation. The absence of PML-II substantially reduced binding of these factors and IFN regulatory factor 3 (IRF3) to IFN-β or ISGs promoters and sharply reduced gene activation. The unique C-terminal domain of PML-II was essential for its activity, while the N-terminal RBCC motif common to all PML isoforms was dispensable. We propose a model in which PML-II contributes to the transcription of multiple genes via the association of its C-terminal domain with relevant transcription complexes, which promotes the stable assembly of these complexes at promoters/enhancers of target genes, and that in this way PML-II plays a significant role in the development of type I IFN responses.

T ype I interferons (IFN) IFN-␣/␤ are a large group of cytokines that play a major role in innate antiviral responses. These responses provide a crucial initial defense against invading viruses and also aid in commissioning an effective adaptive response; when a host's ability to mount an IFN response is impaired or when a virus has acquired particularly effective countermeasures against that response, severe pathogenesis typically ensues (1).
IFN-␣/␤ expression is induced in response to various pathogen-associated molecular patterns (PAMP) (1, 2) including double-stranded RNA (dsRNA), which is produced by many viruses during their replication (3). dsRNA is recognized by pattern recognition receptors (PRRs) including RIG-I and Mda5, triggering a signaling cascade that leads ultimately to the activation of transcription factors (TFs) such as IFN regulatory factor 3 (IRF3) and NF-B (4,5). IRF3 is phosphorylated and moves into the nucleus (6), while NF-B is released from its inhibitor IB, allowing its accumulation in the nucleus (7). These TFs, together with c-jun/ ATF-2, interact with the IFN-␤ promoter to form an enhanceosome (8). The assembled TFs recruit coactivator CREB-binding protein (CBP) or its homologue p300 (9)(10)(11), which are histone acetyltransferases (HATs) whose action promotes the assembly of the basal transcriptional machinery at the promoter.
Secreted IFN-␤ stimulates both the producer cells and other cells to produce IFN-␣, which acts like IFN-␤ and so amplifies the response, as well as a large number of IFN-responsive gene (ISG) products (12). Both IFN-␣ and IFN-␤ are recognized by receptors IFN〈R1 and IFN〈R2 and activate the JAK-STAT signaling pathway. The phosphorylated signal transducer and activator of transcription 1 (STAT1)-STAT2 complex then associates with IRF9 to form the ISGF3 heterotrimer, which binds to the IFN-stimulated response element (ISRE) located within the promoters of most ISGs (13,14). The ISG products establish an antiviral state in the cell (12).
The promyelocytic leukemia (PML) gene was originally identified through its involvement in a chromosomal translocation associated with acute PML (15,16). PML protein currently has seven principal isoforms, designated PML I to VII. These isoforms contain an identical N-terminal region that includes a RING finger domain, two B-boxes, and a coiled-coil domain (RBCC, also known as the tripartite motif, TRIM) and divergent C termini as a result of differential RNA splicing (17)(18)(19).
At least some PML isoforms are essential for maintaining the integrity of nuclear substructures termed PML nuclear bodies (PML-NB), but they also play an important role in the recruitment and localization of other proteins to PML-NB. To date, more than 100 such proteins have been identified either transiently or constitutively associated with PML-NB, including CBP, p53, Sp100, Daxx, and the small ubiquitin-like modifier (SUMO) (20); strikingly, nearly one-half of these proteins are involved in transcriptional regulation.
PML proteins and PML NBs are strongly implicated in a wide variety of cell activities (21), including DNA damage and repair (22), apoptosis (23), senescence (24), and antiviral responses, including the interferon response in particular (25)(26)(27). The rela-tionship between PML and IFN is supported first by the evidence that the PML gene itself is an ISG with ISRE and IFN-␥-activated sequence (GAS) elements in its promoter that mediate induction by type I and II IFN (28,29). Second, ectopic expression of some isoforms of PML protein can inhibit the growth of IFN-sensitive viruses (27,30). Third, various viruses encode proteins that disrupt PML and/or PML bodies, including the E4 Orf3 protein of human adenovirus type 5 (HAdVC-5 or Ad5) and ICP0 of herpes simplex virus 1 (HSV-1), and viruses lacking these functions are unable to overcome IFN responses (31)(32)(33)(34)(35). Ad5 E4 Orf3 targets PML-II specifically (36), suggesting that this isoform in particular might have a role in the IFN response, and PML-II is also one of two isoforms shown to be inhibitory to HSV-1 infection (37). Most recently, PML-IV was shown to enhance IFN-␤ synthesis during virus infection (25).
PML has been linked with transcription regulation in several contexts (38). A growing number of studies have demonstrated that PML participates in the regulation of cytokine signaling (39). PML protein as a whole has been implicated in type II IFN signaling, since it affected STAT-1 DNA binding (40,41); various PML isoforms also affected IFN-␥-induced gene expression in the major histocompatibility complex class I (MHC-I) and MHC-II loci (42,43), with PML-II binding the transcription factor CIITA (43).
We therefore sought to examine the role of PML protein in type I IFN signaling. PML-II specifically was found to play a key role in the induction of IFN-␤ and ISG expression. PML-II positively regulated IRF3, NF-B, and STAT1 activities; PML-II was also found to be associated with transcriptional complexes involving these factors, and the absence of PML-II impaired the stability and DNA binding of these complexes at promoters. Finally, specific sequences in the isoform-specific C terminus of PML-II were necessary both for interaction with TFs and for ISG expression. Thus, we propose that PML-II has a role in the induction of genes of the innate immune response by mediating transcriptional complex assembly at their promoters.
Luciferase reporter assays. Cultures were transfected with 125 pmol/ml siRNA targeting PML-II or PML-V or control siRNA for 24 h and then cotransfected with 225 ng of luciferase reporter plasmid and 25 ng pcDNA3.1-HisB::lacZ (CMV-␤gal). After 24 h, cells were transfected or mock transfected with 1 g/ml poly(I·C) for stimulation of IFN responses and harvested and lysed with passive lysis buffer (Promega) 16 h later. Lysates were assayed for luciferase and ␤-galactosidase activities as previously described (52). Raw data were converted to relative luciferase activity (RLA) by normalizing to the corresponding ␤-galactosidase activity, to correct for variation in transfection efficiency.
Confocal immunofluorescence. Cells grown on coverslips in 12-well culture plates were treated with siRNA and poly(I·C) as above and then processed for immunofluorescence as described previously (53) and viewed by confocal microscopy using a 63ϫ, 1.4 numerical-aperture (NA) objective and a Leica SP2 system. All images shown are representative single images from a z stack, taken through the thickest part of the cell. Each fluorescence channel was imaged separately, and images were merged subsequently using Leica software. In some experiments, images collected in the red channel were false-colored green to improve contrast.
ChIP assays. HEK293 cells were grown to 30 to 40% confluence and then treated with siRNA and poly(I·C) as described above. Chromatin immunoprecipitation (ChIP) was carried out broadly as described previously (65). Briefly, cells were cross-linked with 1% formaldehyde for 10 min at room temperature and then harvested and resuspended in cell lysis buffer (20 mM HEPES [pH 7.9], 25% glycerol, 420 mM NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA) on ice for 20 min. Nuclei were pelleted by centrifugation in a microcentrifuge at 13,000 rpm for 10 min and lysed with nuclear lysis buffer (50 mM Tris-HCl [pH 8.0], 1 mM EDTA, 150 mM NaCl, 1% SDS, 2% Triton X-100) on ice for 10 min. DNA was sheared by sonication to a fragment size of 200 to 1,000 bp, and then lysates were diluted with IP dilution buffer (50 mM Tris-HCl [pH 8.0], 1 mM EDTA, 150 mM NaCl, 0.1% Triton X-100). After incubation with 2 g specific antibody or control IgG at 4°C for 6 h or overnight, protein-chromatin complexes were collected on protein A-Sepharose beads, cross-linking reversed, and DNA eluted with 62.5 mM Tris-HCl [pH 6.8], 200 mM NaCl, 2% SDS, and 10 mM dithiothreitol (DTT) at 65°C for 5 h or overnight. DNA was then extracted by phenol-chloroform and analyzed by SYBR green qPCR using promoter-specific primers for IFN-␤ (TGCTCT GGCACAACAGGTAG and CAGGAGAGCAATTTGGAGGA; amplicon, 82 bp), ISG15 (CGCCACTTTTGCTTTTCCCT and ATAAGCCTGAGG CACACACG; 158 bp), ISG56 (TTGGGTTTCTGCAGCACTAGA and AC CTAGGGAAACCGAAAGGG; 150 bp), protein kinase R (PKR) (TACCC CAATCCCGTAGCAGA and CGTTTTCCCCTTGGACTCCG; 82 bp), and p21 (ATCCCTATGCTGCCTGCTTC and TCTCCTACCATCCCCT TCCT; 184 bp). An amplicon located 10.3 kbp from the ISG56 promoter ISRE was also tested (CTCTGCCTATCGCCTGGATG and CCTGCCTT AGGGGAAGCAAA; 77 bp). All primers were designed with the NCBI primer designing tool and were validated for amplification efficiency and for specificity using standard curve and melt curve analyses, respectively.

Depletion of PML-II reduces IFN-␤ expression.
In order to investigate the function of PML-II in interferon expression, we first established conditions for transient knockdown of PML-II using previously described small interfering RNA (42). Endogenous PML-II mRNA was significant reduced, to about 20 to 30% of control levels, by siRNA treatment (Fig. 1A). At the protein level, CMV-␤gal plasmids for a further 24 h, and then transfected with 1 g/ml poly(I·C) for 16 h before being harvested for reporter assays. Error bars indicate the standard deviations (SD) from the means for at least three biological replicates within an experiment. (E) HEK293 cells were treated with the siRNAs indicated for 48 h and then stimulated with poly(I·C) as described for panel A. IFN-␤ mRNA levels in total RNA were determined by RT-qPCR and normalized to ␤-actin mRNA levels and are shown here relative to values obtained from control siRNA-treated, poly(I·C)-induced cells. In this experiment, poly(I·C) treatment achieved an 8,000-fold stimulation of IFN-␤ mRNA levels. Data are the means Ϯ SD of three replicate values from one representative of three experiments. (F) HeLa cells were treated and assayed for IFN-␤ mRNA levels, and data were analyzed and presented as described for panel E. Poly(I·C) treatment achieved a 1,000-fold stimulation of IFN-␤ mRNA levels. (G) MRC5 cells were treated with 50 pmol/ml siRNA for 72 h prior to transfection with 1 g/ml poly(I·C). Twenty-four hours later, total RNA was analyzed for IFN-␤ expression by RT-qPCR and data were analyzed as described for panel E. Data are the means Ϯ SD from five replicates in two independent experiments. A fold stimulation value for the effect of poly(I·C) could not be obtained, as basal IFN-␤ expression was undetectable. this knockdown protocol reduced the prominent foci of PML-II nuclear fluorescence to undetectable levels in Ͼ90% of cells (Fig.  1B). To confirm the isoform specificity of the knockdown, we overexpressed Flag-tagged PML-II or PML-V and observed the reduction in each protein following specific siRNA treatment. Substantial reductions were seen in amounts of the major 106-kDa exogenous PML-II protein form in siPML-II-transfected HEK293 cells compared with either control siRNA or siPML-V treatment (Fig. 1C, top); conversely, PML-V was efficiently removed by siPML-V while being relatively more resistant to siPML-II (Fig. 1C, bottom). The bands in the empty vector (EV) lane are background detected by the anti-FLAG antibody.
The effect of reducing PML-II levels on IFN-␤ expression was first tested using a luciferase reporter assay. As expected, upon stimulation with poly(I·C), the activity of the IFN-␤ promoter was significantly increased, but prior depletion of PML-II resulted in a significant decrease in this level of induced promoter activity, to levels approximately 25% that of the control (Fig. 1D). In contrast to PML-II, the selective removal of PML-V by siRNA ( Fig. 1C) had no effect on IFN-␤ promoter activity (Fig. 1D). To confirm that the effects of PML-II siRNA in reporter assays reflected the behavior of the endogenous IFN-␤ promoter, endogenous IFN-␤ mRNA levels were tested by reverse transcriptase (RT) qPCR. Amounts of mRNA induced by poly(I·C) in HEK293 cells ( Fig.  1E) and in HeLa cells (Fig. 1F) were significantly reduced by depleting PML-II. Involvement of off-target effects in this reduction was excluded by demonstrating that a second siRNA targeting an unrelated sequence in the PML-II mRNA [siPML-II (B)] had the same effect on IFN-␤ mRNA levels as the initially tested siPML-II RNA (A) (Fig. 1E). To extend these findings to normal cells, a strain of human lung fibroblast cells (MRC5 cells) was also tested. As in immortalized cells, depleting PML-II significantly reduced the induced IFN-␤ mRNA level (Fig. 1G). This effect was specific to removal of PML-II, as depletion of PML-V, which is expressed from a distinct but closely related mRNA, had no effect on IFN-␤ expression in either HeLa or MRC-5 cells (Fig. 1F and G). Taken together, these data indicate that expression of the IFN-␤ gene upon poly(I·C) stimulation is significantly dependent on the presence of PML-II.
Depletion of PML-II impairs IRF3 and NF-B activities. The stimulation of IFN-␤ expression requires the coordinated activation of multiple transcription factors. Among these, IRF3 and NF-B are critically important, acting through promoter subelements PRDIII/I and PRDII, respectively (5,8,66,67). Given the effect of PML-II depletion on IFN-␤ expression, we tested the activities of IRF3 and NF-B individually under PML depletion conditions using PRD-III/I-and PRDII-driven luciferase reporters. As expected, the activity of the IRF3-responsive reporter, PRDIII/I, was greatly increased upon stimulation with poly(I·C); remarkably, prior depletion of PML-II almost completely abolished this increase ( Fig. 2A). To confirm the effect of PML-II depletion on IRF3 activity, we also measured the induction of mRNA from a series of endogenous IRF3-responsive genes, including ISG15, ISG54, and ISG56 (11,68), with or without PML-II depletion ( Fig. 2B to D); mRNA expression from these genes was significantly dependent on PML-II.
Unlike the IRF3-responsive construct, the NF-B-responsive reporter had significant basal activity in HEK293 cells, and this activity was only marginally increased by poly(I·C) stimulation. However, removal of PML-II still reduced its activity substantially, with or without stimulation (Fig. 2E). Similarly, the depletion of PML-II reduced poly(I·C)-induced IL-6, TNF-␣, and IL-8 mRNA levels (Fig. 2F to H), which are established NF-B-dependent genes (60,69). Collectively, these results indicate that depletion of PML-II impairs promoter activation by both IRF3 and NF-B. To exclude the possibility that PML-II was generally required for efficient gene expression, we also tested the expression of a number of control genes. Under conditions of PML-II depletion and poly(I·C) stimulation, expression of a representative ISG was greatly reduced as expected, while expression of three housekeeping genes was unaffected (Fig. 2I).
PML-II plays a role in the JAK/STAT signaling pathway. Impaired IFN-␤ expression due to knockdown of PML-II should affect downstream signaling and thus the activation of IFN-stimulated genes. Therefore, it was expected that activation of an ISRE-dependent promoter by poly(I·C) should be indirectly inhibited by PML-II depletion, as was observed (Fig. 3A). More surprisingly, when IFN-␣, which directly activates the JAK-STAT signaling pathway, was used as the inducer, activation of the ISRE was again strongly inhibited by PML-II depletion (Fig. 3A). This effect was specific to depletion of PML-II since depleting PML-V had no effect on ISRE activity when stimulated by either inducer (Fig.  3B). Thus, a significant part of the effect of PML-II depletion during poly(I·C) induction is attributable to a direct effect on the JAK-STAT pathway.
To further investigate the importance of PML-II to IFN-␤ downstream signaling, induced mRNA levels from a series of ISGs including ISG20, RANTES (CCL5), IP-10 (CXCL10), and IRF7 were measured. Expression induced by either poly(I·C) (Fig. 3C to F) or IFN-␣ (Fig. 3G to M) was in all cases significantly reduced by depleting PML-II. Thus, PML-II depletion reduces the expression of numerous ISGs, both via its effect on IFN-␤ expression and due to a direct impact on signaling via the JAK/STAT pathway.
PML-II has little effect on cytoplasmic signaling events leading to IFN-␤ induction. In principle, the role of PML-II in the induction of IFN-␤ expression could be at any point from PAMP sensing through the resulting signaling cascade to the assembly of the enhanceosome at the IFN-␤ promoter. To investigate where PML-II acted in this pathway, we first examined IRF3 phosphorylation and IRF3 and NF-B nuclear translocation, which have been described as key steps in signal transduction following PAMP recognition (5,6,70). Neither poly(I·C) stimulation nor siRNA transfection affected the expression of total IRF3 (Fig. 4A). The amount of phosphorylated (activated) IRF3 increased gradually with the duration of poly(I·C) stimulation; depletion of PML-II did reduce the accumulation of phospho-IRF3 somewhat (Fig.  4A), although this effect was modest and was apparent only with longer periods of stimulation, suggesting that it was a secondary consequence of the changes in ISG expression observed. Some phosphorylated IRF3 was also detected in unstimulated cells. This may reflect a requirement to maintain a basal level of ISG expression in order to give a rapid antiviral response in the initial stages of infection, though NF-B has been shown to be more important than IRF3 for this (71,72).
Stimulation with poly(I·C) also caused the expected movement of IRF3 into the nucleus; the number of IRF3-positive nuclei increased from 0% to between 30% and 40% by 16 h; consistent with the limited effect on pIRF3 level, depletion of PML-II had no significant effect on the fraction of cells showing IRF3 nuclear localization following poly(I·C) treatment ( Fig. 4B and D). For NF-B, the effect of poly(I·C) stimulation on localization was less striking. Only 10% of nuclei stained positive for NF-B (p65) after 16 h, and in these cells, there was still predominantly cytoplasmic fluorescence ( Fig. 4C and D). As with IRF3, there was no significant difference in the extent of NF-B nuclear translocation in the presence or absence of PML-II ( Fig. 4C and D). This is consistent with the lower activity of NF-B relative to IRF3 in the IFN-␤ subdomain reporter assays in this cell type ( Fig. 2A and E). Collectively, these results show that PML-II has only limited effects on the cytoplasmic signaling events leading to IFN-␤ induction; the more profound effects of PML-II depletion on gene expression must therefore be accounted for by events in the nucleus.
PML-II interacts with transcriptional complexes. Interaction with different partner proteins such as specific transcription factors is important for PML involvement in various cellular activities (38,(73)(74)(75)(76). The assembly of transcription factors (TFs), including IRF3 and NF-B, and coactivators CBP/p300 into a transcription complex at the IFN-␤ promoter is a key step for IFN-␤ induction (5,6,10,11). It is also known that CBP can be recruited to PML-NB and that it interacts with PML (73,77), suggesting that PML might also associate with the transcriptional complex involved in IFN-␤ activation. We therefore examined the association of PML-II with IRF3, NF-B, and CBP. As expected from earlier reports, poly(I·C) induced the association of IRF3 and NF-B with CBP ( Fig. 5A and B) (11,78). Overexpressed Flag-tagged PML-II also associated with CBP in both stimulated and unstimulated cells, consistent with previous observations (Fig. 5C) (73,77). Importantly, in these Flag-PML-overexpressing cells, poly(I·C) stimulation induced a greater association of NF-B with complexes precipitated by Flag beads than what was seen with no stimulation or in empty vector-transfected cells, although some nonspecific background precipitation was observed (Fig. 5C). Similarly, poly(I·C) stimulation induced a strong interaction of Flag-PML-II with complexes that contained STAT1, a component of the transcription factor ISGF3, whereas in unstimulated cells there was no detectable interaction (Fig. 5C). In light of these results, we next tested whether depletion of PML-II affected the stability of the induced association between these transcriptional complex components. PML-II depletion substantially reduced the association between CBP and IRF3 (Fig. 5D, compare lanes labeled IRF3) or STAT1 (Fig. 5E, compare lanes labeled CBP); quantitation of these effects showed that the CBP-IRF3 interaction was reduced to 66% and the CBP-STAT1 interaction to 62% of their respective controls. Taken together, these data indicate the formation of transcriptional complexes during poly(I·C) stimulation that involve PML-II, CBP, and three DNAbinding TFs, IRF3, NF-B, and STAT1. They further show that PML-II contributes to the stable association of TFs with CBP.
PML-II alters transcription factor binding at promoters. TF binding at promoters or enhancers is essential for gene transcription. We next examined the effect of PML-II depletion on the binding of TFs at the IFN-␤ promoter by chromatin immunoprecipitation (ChIP). In line with the observed effects on promoter activity, depletion of PML-II led to significant reductions in IRF3 and NF-B binding at the IFN-␤ promoter ( Fig. 6A and B). Previous studies have demonstrated that the TFs assembled at the IFN-␤ promoter can recruit the coactivators CBP/p300 and that this is important for activation (5,6,10,11). Therefore, the absence of this PML isoform might also affect CBP binding to the IFN-␤ promoter. Indeed, there was a significant loss of CBP binding at the promoter in the absence of PML-II (Fig. 6C). Thus, the knockdown of PML-II affects TF and CBP recruitment to the IFN-␤ promoter, which accounts for the severely reduced expression of the gene when induced under conditions of PML-II depletion.
The expression of ISGs requires ISGF3 assembly and binding to ISRE elements. We therefore considered that PML-II might also affect the binding of ISGF3 to the ISRE, which was tested by ChIP analysis for its STAT1 component on three representative ISG promoters. Consistent with the effect of PML-II depletion on ISG expression, STAT1 binding to these promoters was decreased in the absence of PML-II (Fig. 7A). Finally, to test whether PML-II more generally affected TF binding to chromatin, we examined the binding of p53 to a well-characterized target, the p21 promoter; p53 is known to be bound to this promoter even in the uninduced state (79). PML-II knockdown had no effect on this interaction (Fig. 7B). Thus, taken together, our results indicate that PML-II specifically and positively regulates TF binding at ISRE elements and at the IFN-␤ promoter, correlating with its effects on expression of these genes.
The PML-II unique C-terminal domain is required for TF binding and enhanced gene expression. PML isoforms have dis- were mock stimulated or transfected with 1 g/ml poly(I·C) for 16 h and then lysed with 0.5% NP-40 lysis buffer. A portion of each cell lysate was retained for direct analysis by Western blotting, while equal parts of the remainder were incubated with IRF3, NF-B (p65), or CBP antibodies or with control IgG and precipitated with protein A-Sepharose beads; proteins in total or immunoprecipitated samples were separated by SDS-PAGE, and IRF3, NF-B, or CBP was detected by Western blotting. (C) HEK293 cells were transfected with 250 ng/ml Flag-PML-II plasmid or pCI-neo empty vector for 24 h and then stimulated with poly(I·C); lysates were prepared and immunoprecipitated with anti-Flag beads, and precipitates and total lysates were analyzed for Flag-PML, CBP and NF-B, and STAT1 as described for panel A. (D, E) HEK293 cells were treated with siRNA and then stimulated with poly(I·C) as described for Fig. 1A. Lysates were prepared and immunoprecipitated with IRF3 or CBP antibodies, and precipitates and total lysates were analyzed for IRF3, STAT1, and CBP as described for panel A. tinct functions due to their different C-terminal domains, which are important for their interaction with partner proteins (80). For example, previous work in our laboratory showed that a 40-amino-acid residue segment of the PML-II C-terminal domain conferred an interaction between this protein and Ad5 E4 Orf3 (47). The role of PML-II unique C-terminal sequences in binding cellular TFs was tested by coimmunoprecipitation using a set of Flagtagged PML-II deletion mutants, PML-II-⌬1, PML-II-⌬2, and PML-II-⌬3 (47), recloned into a full-length PML-II background (Fig. 8A). Full-length PML-II associated with NF-B and STAT1 (Fig. 8B), as expected from the results shown in Fig. 5C. Deletion ⌬3 was at least as effective as the full-length protein in these interactions; however, both ⌬1 and ⌬2 mutants essentially lost the association with NF-B, while ⌬1 had somewhat reduced STAT1 binding (Fig. 8B).
The N-terminal RBCC domain of PML protein mediates multimerization with other PML isoforms (reviewed in reference 19).  Cells were fixed and subjected to ChIP with STAT1 antibodies in parallel with a normal IgG control precipitation on an equal volume of extract. Precipitated DNA was assayed for ISG15, ISG56, and PKR promoter sequences by SYBR green qPCR, and data were processed as for Fig. 6. (B) HEK293 cells treated with siRNA as described for panel A were fixed and subjected to ChIP analysis using p53 antibodies and qPCR primers specific for the p21 promoter.  (47). (B) HEK293 cells were transfected with 250 ng/ml Flag-PML-II wild-type or deletion mutant plasmid or pCI-neo empty vector as indicated for 48 h and then stimulated with poly(I·C), and lysates were prepared and immunoprecipitated with anti-Flag beads. Precipitates and total lysates were analyzed for Flag-PML, NF-B (p65), and STAT1 by Western blotting. (C) As described for panel B but using ⌬RBCC variants of each PML-II plasmid. (D, E) HEK293 cells were transfected with plasmids as described for panels B and C. Following 1 g/ml poly(I·C) (B) or 1,000 units/ml IFN-␣ stimulation (C) for 16 h, total RNA was analyzed for specific mRNA levels by RT-qPCR. mRNA levels are displayed relative to those observed in control siRNA-treated cells without stimulation. Quantification was performed as described for Fig. 1. Data shown are the means Ϯ SD for at least three replicates within an experiment.
To exclude the possibility that PML-II binding with other isoforms was contributing to the observed interactions with NF-B and STAT1, TF binding to ⌬1 to ⌬3 mutants was reanalyzed using constructs with the PML RBCC domain deleted. As expected for a function unique to PML-II, deletion of the N-terminal RBCC domain did not affect binding to NF-B or STAT1 (Fig. 8C). Deletion ⌬1 abolished and ⌬2 significantly diminished specific binding to NF-B, while the deletion ⌬3 mutant again retained full interaction capability (Fig. 8C) and the ⌬1 mutant also showed a significantly reduced association with STAT1. These results match closely those seen for full-length PML-II.
PML-II overexpression only modestly potentiates activation of endogenous IFN-␤ expression by poly(I·C), but this effect is increased when the RBCC domain is deleted (Fig. 8D). Consistent with the TF binding data (Fig. 8B and C), PML-II ⌬1 in a ⌬RBCC background had impaired ability to stimulate (or to hyperstimulate in the presence of IFN-␣) the expression of a representative ISG, ISG56 (Fig. 8E). Collectively, these results indicate that the PML-II unique C-terminal domain is essential for its binding with TFs and for activation of gene expression in the IFN response.
PML-II associates with target promoters. PML-II interacts with specific TFs and CBP, and this complex is induced by poly(I·C) stimulation (Fig. 6). Having shown that the ⌬RBCC form of PML-II could both bind specific TFs and increase the activity of target promoters (Fig. 8), we finally sought to test whether PML-II ⌬RBCC was recruited to ISG promoters by ChIP analysis. It was necessary to use exogenously expressed PML for this experiment to provide a Flag tag for immunoprecipitation, and we elected to use the ⌬RBCC form to avoid effects of potential interactions between exogenous and endogenous PML via their RBCC domains. These experiments thus involve significant overexpression of the FLAG-tagged protein relative to endogenous levels of PML-II, though the ratio in individual cells will vary (45). As predicted from our protein coprecipitation studies, results showed that in poly(I·C)-stimulated cells there was significant association of Flag-PML-II ⌬RBCC with the ISG15 and ISG56 promoters ( Fig. 9A and B). This interaction was specific to PML-II since an equivalent Flag-tagged PML-V protein, the knockdown of which had no effect on IFN-␤ or ISG expression ( Fig. 1D and  3B), showed no association with the ISG56 promoter (Fig. 9B). The interaction with PML-II was focused on the promoter region, since amplification of a target taken from elsewhere in the ISG56 gene gave a significantly lower signal (Fig. 9C). Thus, the unique C-terminal region of PML-II associates specifically with target promoters.

DISCUSSION
PML protein has been demonstrated to be involved in a large number of cellular processes, including antiviral defense and transcriptional regulation (21,26,81). However, although a role for PML in the type I IFN response has been implied by many previous studies, a mechanism has remained elusive. In this study, we investigated the function of PML in controlling the type I interferon response. Our data show that one isoform, PML-II, acts as a positive regulator of both IFN-␤ transcription and transcription of downstream response genes that are activated by type I IFN. This positive effect is achieved via PML-II interacting with multiple transcriptional complexes and increasing their binding at target promoters.
Our study used siRNAs to specifically target particular PML isoforms. PML-II is expressed from an mRNA that has been spliced to remove an intron that, when retained as part of the mRNA, encodes the unique C terminus of PML-V. Thus, in a PML-V mRNA the sequence encoding the C terminus of PML-II remains unused in the 3= untranslated region. Therefore, any siRNA targeted toward the PML-II unique 3= exon necessarily also targets PML-V mRNA. However, our results show that while depletion of PML-II significantly reduced expression of IFN-␤ and ISGs, depletion specifically of PML-V had no effect (Fig. 1D, F, and G and 3B). Thus, the biological effect of PML-II siRNA is not due to effects on PML-V expression. The PML gene also encodes PML-III from an mRNA that is very similar to PML-II mRNA, differing from it only by the retention of an additional 40 nucleotides (nt) of sequence at the 5= end of its unique 3= exon. We have not specifically depleted PML-III alone and therefore cannot formally exclude that it plays a role in the expression of IFN-␤ and ISGs. However, the PML-III unique C terminus is short and poorly conserved in comparison with PML-II (71 versus 259 residues) (47), it is expressed at very low levels, particularly in normal cells (82), in which we still observed a strong negative effect of PML-II siRNA treatment on gene expression, and its overexpression has been reported to have no effect on IFN-␣-activated gene expression (41), so we consider it unlikely that PML-III is significant in this context. We found that PML-II was necessary for efficient IFN-␤ transcription in response to stimulus with poly(I·C), a form of dsRNA that is recognized by cells as a PAMP, an effect mediated through increased TF binding at the IFN-␤ promoter. Recently, exogenous PML-IV (but not PML-II) was reported to potentiate IFN-␤ expression in response to infection with vesicular stomatitis virus by sequestering at PML-NBs the peptidyl-prolyl cis/trans-isomerase (Pin1) that is needed for the degradation of the key activating TF, phosphoIRF3 (25). This study found, as we do, that exogenous PML-II overexpression has little positive effect on the production of IFN-␤ mRNA in response to stimulus; however, the study did not test whether selective removal of PML-II would abrogate the potentiating effect of PML-IV overexpression, which our study would predict. Thus, the two studies reveal different PML functions in the IFN response that are associated with distinct PML isoforms.
Our data show that PML-II promotes the binding of STAT1 to ISG promoters, which occurs as part of the ISGF3 complex with STAT2 and IRF9. A role for PML protein in the regulation of transactivation by STAT1 homodimer in response to type II IFN-␥-mediated signaling was reported previously; however, contradictory results as to an activating or inhibitory role for PML were obtained by three different groups (40,41,43). Choi et al. found that activity of a GAS reporter plasmid induced by IFN-␥ was increased in PML-null mouse embryonic fibroblasts (MEFs), as was the amount of phospho-STAT1 and GAS-binding activity of STAT1 homodimer from nuclear extracts in vitro, while, using the same cell system as well as siRNA knockdown in human cells, El Bougrini et al. found the exact opposite, and they further found that overexpression of any nuclear PML isoform potentiated the response to IFN-␥. In agreement with the latter work, Ulbricht et al. found that PML depletion reduced the transcriptional upregulation of MHC-II expression in response to IFN-␥ in primary human fibroblasts or Hep2 cells. Our present study does not address or resolve this conflict; however, we did find that PML-II positively regulated the JAK/STAT signaling pathway in response to both poly(I·C) and IFN-␣ stimulation. El Bougrini et al. also tested the response of an ISRE reporter to IFN-␣ in 293 cells following knockdown of PML; in contrast to our results, they found only a modest reduction in response. It is likely that this difference relates to their use of a total PML knockdown strategy, whereas we targeted PML-II specifically; the absence of multiple PML isoforms is likely to affect the cells in a complex way, with the net result being the sum of multiple effects.
We also found that the activity of NF-B was impaired by depleting PML-II in both stimulated and unstimulated cells. The latter result supports and extends a recent microarray analysis that showed that the knockdown of all PML isoforms suppressed the expression of a group of NF-B-dependent genes such as IL-6 and IFN-␥-inducible protein 10 (IP-10) genes (83). The weak activation of NF-B compared with that of IRF3 in response to poly(I·C) stimulation in HEK293 cells may reflect the inhibitory effect on its activation of the Ad5 E1B 19K protein present endogenously in these cells (84). The difference may also reflect the observation that IRFs are principally involved in IFN-␤ production while NF-B p65 is more important for the induction of proinflammatory genes upon virus infection (71). In the present study, the expression of a number of NF-B-dependent genes and ISGs, including IRF7, was found to be depressed by PML-II depletion. IRF7 is essential in the positive-feedback loop of IFN-␤ expression (85)(86)(87). Impaired IRF7 expression would be expected to reinforce the effect of PML-II depletion on IFN-␣/␤ expression and thus amplify the negative effect of this depletion on ISG expression.
Several previous studies have shown that PML proteins generally, or PML-II in particular, play a role in antiviral responses and are targets for viral proteins that are involved in combatting innate immune responses (81). Ad5 E4 Orf3 rearranges PML NB through an interaction with PML-II (47) and is also necessary for the virus to replicate in cells with an established IFN response (34,35). HSV-1 causes global degradation of PML protein, including PML-II, which is one of two isoforms that are inhibitory to HSV-1 infection (37), via its ICP0 protein. The growth defect of ICP0 mutants in vivo is greatly reduced in mice lacking a functional IFN response (32), and the inhibitory effect of IFNs on the replication of ICP0 mutants is largely abolished when they are grown in PMLnull cells, in contrast to what is seen in normal cells (88). Furthermore, HSV-2 alters PML RNA splicing to favor PML-V expression over that of PML-II, suggesting a particular significance of PML-II in antiviral responses (89). PML was also shown to limit the replication and speed/extent of pathogenesis of lymphocytic choriomeningitis virus (LCMV) and vesicular stomatitis virus (VSV) in mice, effects suggested to be via enhanced innate immune responses (90). Our findings here are consistent with these previous reports that PML is important for aspects of the antiviral response and provide a mechanistic explanation for this effect.
Numerous proteins that can be physically and/or functionally linked to PML protein have been found (20,21); in the present study, PML-II was found to associate with transcription factors IRF3, NF-B, and STAT1 and coactivator CBP. CBP plays an important role in the transcription of a large number of genes, including IFN-␤ and ISGs (78,(91)(92)(93). It provides histone acetyltransferase (HAT) activity, which modifies chromatin and supports recruitment of the general transcriptional machinery, including RNA pol II (94)(95)(96). Our finding that the depletion of PML-II causes a significant effect on CBP recruitment to the IFN-␤ promoter under conditions where its activity should be stimulated therefore contributes to an explanation of its reduced expression. Modulation of CBP function by PML has been suggested previously from studies of hormone signaling (73). We further showed that the ability of PML-II mutants to associate with STAT1 and NF-B correlated with their ability to potentiate gene expression activated by these factors.
It has been reported that the Ad5 E1A 13S transcriptional activator protein interacts specifically with PML-II to potentiate E1A transcriptional activation (97). In that study, while total PML depletion modestly increased virus yield, restoration of PML-II in a PML-null background caused a substantial further increase. Interpreting these results in light of our findings, we suggest that Ad5 has evolved to make positive use of a PML-II function that intrinsically has antiviral effects when expressed in the context of a full set of PML isoforms. E1A binds CBP and its homologue p300 via its CR1 region and binds various DNA-binding transcription factors via its CR3 region (98,99). The interaction of E1A with PML-II also required CR3 sequences (97). As these findings are analogous to the effects of PML-II on TF/CBP association with IFN-␤/ISG promoters that we observed, we therefore suggest that PML-II also acts in E1A activation by bridging these factors to support their binding to DNA.
The interaction of CBP with PML has been reported to involve sequences from the N-terminal coiled-coil domain that is present in all the principal nuclear PML isoforms (73), indicating that CBP binding should not be restricted to PML-II. However, while we have not tested a full range of isoforms, our data show that PML-II has a functional role in CBP activity that at least one other isoform, PML-V, does not possess. Possibly, the differing PML isoform C-terminal domains affect the conformation or availability of the CBP binding site so that its activity is manifest only in certain isoforms. However, we favor an interpretation in which CBP binding by PML is not sufficient for it to exert a positive effect on transcriptional activity, other sequences that are unique to PML-II also being required. As discussed below, these include sequences that we have shown are necessary for interaction with specific DNA-binding transcription factors.
Different PML isoforms can have distinct functions mediated by their unique C-terminal domains (80). Bioinformatic analysis showed that this part of PML-II was likely to be unstructured but with the propensity to become ordered upon interaction with partner proteins; mutational analysis revealed that interaction with Ad5 E4 Orf3 required one particular molecular recognition element in this region (47). Based on its chemistry and its localization properties, the PML-II unique C-terminal domain was more recently speculated to interact with transcription factors (100). Here we have shown that specific sequences in the unique PML-II C-terminal domain are essential for its interaction with two transcription factors, NF-B and STAT1, while the N-terminal RBCC domain is dispensable. It was recently reported that the unique C-terminal domain of PML-II can bind to PML-NBs independent of the shared N-terminal region (100), suggesting that C-terminal PML-II might replace full-length PML-II for some functions. In this context, we found that removal of the RBCC domain gave a protein with enhanced activity in the expression of IFN-␤ and ISGs; the biological significance and mechanism of this effect remains to be determined, but it could reflect a greater functional availability of the PML-II C-terminal domain when less tightly tethered to PML-NB.
In summary, our results show that PML-II positively regulates the expression of genes involved in the IFN response, reflecting a positive effect on the formation of relevant transcription factor complexes and their association with the promoters of these genes. Our data support a model in which activators of the innate immune response cause PML-II to associate with multiple transcriptional complexes and that this interaction facilitates their loading onto, or stable association with, target promoters. The observed role of PML-II in regulating expression of a large number of cytokines and chemokines strongly supports a broad function for PML-II in the innate immune response.

ACKNOWLEDGMENTS
We are grateful to S. Goodbourn, Li Yong, and J. Mankouri for their gifts of reporter plasmids used in this work and to Kun-Sang Chang for generously providing anti-PML antibodies.
Y. Chen was funded in part by a Warwick Postgraduate Research Fellowship and by scholarships from the Sir Richard Stapley Education Trust and the Great Britain-China Educational Trust; J. Wright was supported by The Wellcome Trust, grant WT094713MA; X. Meng was funded by a Warwick-China Council Scholarship and by the Sir Richard Stapley Education Trust.