Promyelocytic leukemia (PML) nuclear bodies selectively associate with transcriptionally active genomic regions, including the gene-rich major histocompatibility (MHC) locus. In this paper, we have explored potential links between PML and interferon (IFN)-γ–induced MHC class II expression. IFN-γ induced a substantial increase in the spatial proximity between PML bodies and the MHC class II gene cluster in different human cell types. Knockdown experiments show that PML is required for efficient IFN-γ–induced MHC II gene transcription through regulation of the class II transactivator (CIITA). PML mediates this function through protection of CIITA from proteasomal degradation. We also show that PML isoform II specifically forms a stable complex with CIITA at PML bodies. These observations establish PML as a coregulator of IFN-γ–induced MHC class II expression.
Promyelocytic leukemia (PML) nuclear bodies, also known as the PML oncogenic domain or ND 10 (nuclear domain 10), are multiprotein subnuclear structures that were first observed 50 yr ago by electron microscopy as nuclear-dense granular bodies (de Thé et al., 1960). PML bodies are present in cell nuclei of almost all human tissues (Gambacorta et al., 1996; Cho et al., 1998). Usually, nuclei of cultured cells contain ∼5–30 PML bodies, ranging in size from ∼0.2 to 1 µm (Ascoli and Maul, 1991). It has been suggested that PML bodies act as dynamic nuclear organizing centers for the coordinated regulation of different cellular functions, including DNA damage response, apoptosis, senescence, and antiviral responses (Sternsdorf et al., 1997; Negorev and Maul, 2001; Bernardi and Pandolfi, 2007). This variety of functions may be attributable to regulated traffic at PML bodies of a selected number of nuclear proteins, which play key roles in these processes (Boisvert et al., 2001; Wiesmeijer et al., 2002; Weidtkamp-Peters et al., 2008; Brand et al., 2010). Six nuclear PML isoforms sharing the same N terminus but containing variable C termini generated by alternative splicing have been identified (Fagioli et al., 1992; Jensen et al., 2001) and are present at all PML bodies (Condemine et al., 2006).
Because components of the ubiquitin–proteasome and SUMO pathways accumulate in or adjacent to a subset of PML bodies, they have also been proposed as sites of specific protein turnover within the nucleus (Fabunmi et al., 2001; Lallemand-Breitenbach et al., 2001; Rockel et al., 2005; Saitoh et al., 2006; Scharf et al., 2007; Chen et al., 2008; Sharma et al., 2010). The expression of the PML gene and other genes encoding PML body components is strongly increased by interferons (IFNs) in vitro and in vivo. Therefore, PML isoforms and PML bodies have also been implicated in IFN response and immune surveillance pathways, including antiviral responses and inflammatory processes (Terris et al., 1995; Tavalai and Stamminger, 2008).
According to a recently curated PML body interactome, 166 different proteins can be physically and/or functionally linked to PML. Strikingly, one half of these factors are involved in transcriptional regulation (Van Damme et al., 2010). Many transcription factors transiently associate with PML bodies, in which their activity is modulated by posttranslational modifications (PTMs). This may result in either activation or repression of specific genes, depending on the context (Zhong et al., 2000). In addition, sites of nascent mRNA transcription and transcriptionally active specific genomic loci were found to be associated with the periphery of PML bodies (Grande et al., 1996; Boisvert et al., 2000; von Mikecz et al., 2000; Fuchsová et al., 2002; Kießlich et al., 2002, Wang et al., 2004; Xie and Pombo, 2006).
Most cell types do not express major histocompatibility (MHC) II genes constitutively, but their expression can be stimulated by IFN-γ (Collins et al., 1984). The MHC II expression pattern is controlled primarily by transcriptional regulation at the conserved MHC II gene promoters. The MHC II enhanceosome contains the constitutively expressed transcription factors regulatory factor X (RFX), nuclear factor-Y, and cAMP response element binding protein (Choi et al., 2011). MHC II gene expression is completely dependent on the expression of the MHC class II transactivator (CIITA), which is recruited in a coactivator-like fashion to the MHC II promoters through protein–protein interactions with RFX, cAMP response element binding protein, and nuclear factor-Y (Steimle et al., 1993; Reith et al., 2005). CIITA expression is induced by IFN-γ through the Janus kinase (JAK)–signal transducer and activator of transcription (STAT) pathway, and expression of exogenous CIITA is sufficient to induce MHC class II expression in many different cell types (Steimle et al., 1994). Shiels et al. (2001) and Wang et al. (2004) have previously demonstrated a highly nonrandom spatial association of the MHC locus with PML bodies, raising the hypothesis that PML (bodies) may be directly involved in transcriptional regulation of selectively associated genes. Subsequently, Gialitakis et al. (2010) have found that the MHC II gene DRA relocates to PML bodies upon induction with IFN-γ and that PML protein isoform IV may function to maintain a sustained transcription-permissive epigenetic state of the DRA gene. Here, we analyze potential functional relationships between PML (nuclear bodies) and CIITA-dependent MHC class II gene expression.
IFN-γ increases the spatial proximity between PML bodies and the MHC II gene cluster
IFN-γ induces a stable relocation of the human leukocyte antigen (HLA) class II DRA gene to PML bodies in HeLa cells (Gialitakis et al., 2010). Here, we confirm and extend these observations for HEp-2 cells, the primary human lung fibroblast line MRC-5, and primary human foreskin fibroblasts (HFFs) using a MHC II–specific FISH probe (Fig. 1 A). To quantify the spatial proximity between PML bodies and the MHC II gene cluster, we assessed the frequency of lacking association (no contact), association (contact), and colocalization (considerable overlap) between the two signals in confocal images (Fig. 1, A and B). The definition colocalization included both considerable overlap and complete colocalization. This approach revealed an increase of association and colocalization events between PML bodies and the MHC II probe in HEp-2, MRC-5, and HFF after IFN-γ stimulation (Fig. 1 B). Quantification was based on the MHC II signal closest to the next PML body. However, the second MHC class II signal was also associated or colocalized with the PML body signal in >80% of MRC-5 cells or HFF in which the first signal was already associated or colocalized (Fig. 1 A, white circles; and not depicted). We also calculated the theoretical probability for random associations between PML bodies and the MHC class II signals by applying a self-developed algorithm that considers nuclear volume and number and size of objects in the nucleus. The probabilities for random signal overlap varied between 0.30 and 0.59% in MRC-5 and 4.48 and 8.77% in HEp-2 cells (Table S1) and were thus substantially lower than the measured values. Centromeric chromatin of chromosome 6 was not in spatial proximity to PML bodies, although IFN-γ induced a low degree of nonrandom signal overlap in HFFs (Fig. 1 B). Currently, we have no explanation for this phenomenon. It might be related to a dynamic, cell cycle–regulated connection between centromeres and PML bodies as described previously (Everett et al., 1999). To exclude that elevated numbers of PML bodies after IFN-γ treatment caused random proximity between PML bodies and genomic regions, we determined the association frequency between PML bodies and centromeres in MRC-5 fibroblasts (Fig. 1 C). Although the number of PML bodies increased three- to fourfold during IFN-γ treatment (Fig. 1 D), there was no increase of their signal overlap rate with the centromeres (Fig. 1 E). Signal overlap events between PML bodies and centromeres in IFN-γ–treated HEp-2 cells, or HFFs never exceeded 1 or 2%, respectively (unpublished data). The latter result appears to be contradictory to the data for HFF in Fig. 1 B, in which PML bodies were not found to be in close proximity to the centromere 6 region. Although we have no ready explanation for this phenomenon, the discrepancy might be explained by using HFFs from different population doublings. Despite this inconsistency, it is safe to conclude that IFN-γ specifically induces an increase of the spatial proximity between the MHC II gene cluster and PML bodies in HEp-2 cells as well as in primary human lung and skin fibroblasts.
PML is required for efficient IFN-γ–induced MHC II expression
To investigate whether PML plays a role in MHC II expression, HFF stably expressing control or PML small hairpin RNA (shRNA) were treated with IFN-γ, and MHC II expression was monitored at different time points by indirect immunofluorescence microscopy using monoclonal antibodies against various class II proteins. In the presence of PML, IFN-γ–induced expression of HLA-DR/DP/DQ became visible after 24 h, and after 48 h, most cells were positive (Fig. 2 A). Coimmunostaining with anti-PML antibodies confirmed an IFN-γ–induced increase in the number of nuclear bodies. In contrast, cells expressing the PML-specific shRNA had no PML bodies and showed a strongly reduced induction of class II molecule expression (Fig. 2 A). A similar result was obtained in HEp-2 cells (Fig. S1). Quantification revealed that the number of MHC II–positive cells was diminished by a factor of ∼4 in HFF in the absence of PML (Fig. 2, B and C). To determine whether PML depletion interferes with MHC II expression at the transcript level, RT-PCR was performed. As expected, CIITA and MHC II mRNAs were undetectable in unstimulated cells. Transcription of IFN-γ–stimulated genes within the MHC II locus, including TAP1, LMP7, and MHC II antigens DRA and DMB, was clearly detected after 24 h of IFN-γ treatment (Fig. 2 D). TAP1 is activated by IFN-γ but not via CIITA (Wright, et al., 1998). LMP7 expression appears to be regulated by CIITA-dependent and -independent pathways (Londhe et al., 2012). Depletion of CIITA strongly diminished IFN-γ–induced transcription of the MHC II genes but not of TAP1 or LMP7. Importantly, IFN-γ–induced MHC II mRNA expression was also significantly reduced in PML-depleted HFFs, whereas transcription of TAP1 and the MHC class I gene HLA-F was not affected (Fig. 2 E). Interestingly, expression of LMP7 mRNA was up-regulated twofold in PML-depleted cells, which may be explained by PML-mediated selective gene regulation within the MHC class I cluster (Kumar et al., 2007). Collectively, these results demonstrate that down-regulation of PML leads to a decrease in IFN-γ–induced MHC II molecule expression at the transcriptional level.
PML regulates CIITA mainly at the posttranslational level
MHC class II mRNA expression is quantitatively dependent on the amount of CIITA (Otten et al., 1998). CIITA was up-regulated by IFN-γ at concentrations as low as 10 U/ml (Fig. 3 A), and induction peaked between 6 and 24 h (Figs. 3 B and 4 B). By Western blotting, various PML isoforms were detected (Fig. 3 C, lane 1). IFN-γ induced up-regulation of all PML-specific bands with similar efficiency, indicating no increase of particular PML protein products through IFN-γ (Fig. 3 C, lanes 2, 4, and 5). The IFN-γ–induced CIITA protein level was reduced to 59 ± 11% in HEp-2 cells, in which PML was down-regulated by siRNA to 13 ± 6% of control levels (Fig. 3 D). Down-regulation of CIITA had no effect on PML protein levels (Fig. 3, C and D). SP100, another constitutive IFN-γ–inducible PML body component, was also efficiently depleted in HEp-2 cells using specific siRNA, but this had no effect on the CIITA protein level (Fig. 3 E). In HFFs stably expressing shRNAs against PML, the amount of PML protein is reduced to a very low level (3.4 ± 0.9%; Fig. 3 F). Under these conditions, the CIITA protein level decreased to 41 ± 15% compared with control shRNA-treated fibroblasts (Fig. 3 G). Consistent with published data, we also found a reduction of the Sp100 level in PML knockdown fibroblasts (Fig. 3 G; Everett et al., 2006). We also analyzed the protein level of RFX5 and NFY-C, two of the constitutive components of the MHC II enhanceosome. This showed that PML depletion has no effect on the protein amount of these factors (Fig. S2), suggesting that PML knockdown does not result in reduced protein levels of all components of the MHC II enhanceosome.
Next, we asked at which level PML might regulate the CIITA protein amount. IFN-γ induces transcription of the CIITA gene through the JAK–Stat pathway (Dong et al., 1999). Because it was demonstrated recently that PML depletion leads to reduced STAT1 phosphorylation (El Bougrini et al., 2011), we tested whether CIITA mRNA expression is affected through the JAK–STAT signaling pathway. Although overall STAT1 levels remained unchanged during the first 2 h of IFN-γ activation, phosphorylation of STAT1 at Serine 727 and Tyrosine 701 was reduced in PML-depleted cells (Fig. 4 A). These results confirm and extend previous observations (El Bougrini et al., 2011). At later time points (6–48 h), IFN-γ induced an up-regulation of STAT1, which was similar in shControl and shPML cells (Fig. 4 B). This observation suggests that the IFN-γ activation site/IFN-stimulated response element–dependent activation of the STAT1 gene is independent of PML. However, the CIITA protein amount was higher between 6 and 36 h of IFN-γ stimulation in cells proficient for PML (Fig. 4 B). The increase in CIITA level was accompanied by a subsequent induction of HLA-DR protein expression, which was clearly reduced in shPML cells compared with shControl cells (Fig. 4 B). Measuring the CIITA mRNA level showed that the PML status of cells is directly linked to the induction of CIITA mRNA in the early response to IFN-γ stimulation for ≤3 h (Fig. 4 C). This observation shows that PML regulates transcription of the CIITA gene by modulating the early STAT1 phosphorylation level. However, between 6 and 24 h after IFN-γ stimulation, the CIITA mRNA level was not significantly different in PML-proficient and -deficient cells (Fig. 4 C). It is during this period of time that the lack of PML leads to decreased protein level of CIITA (Fig. 4 B, blot CIITA, compare lanes 2–4 with lanes 8–10). This strongly suggested that PML stabilizes CIITA at the posttranscriptional level, in addition to its early regulatory effect on CIITA mRNA transcription via the JAK–STAT pathway.
To test whether PML can regulate the stability of CIITA, we performed cotransfection experiments. When constant amounts of the GFP-CIITA plasmid were cotransfected with increasing amounts of RFP-tagged PML isoforms, expression of GFP-CIITA was increased in a PML-dependent manner (Fig. 4 D). PML isoforms I to IV induced increased expression of GFP-CIITA, with PML isoform II yielding the strongest effect (Fig. 4 D). This result was obtained in three independent assays (unpublished data). Increased expression of RFP alone resulted in a decrease of the amount of GFP-CIITA indicating that PML and not RFP or increased plasmid concentration is responsible for the observed effects (Fig. 4 E). CIITA has a short half-life caused by rapid proteasomal degradation (Schnappauf et al., 2003). This allowed us to test whether PML also stabilizes CIITA at the endogenous level. The protein amount of CIITA was analyzed at various time points in cycloheximide-treated cells. This approach revealed rapid disappearance of the CIITA protein band within 2 h in control cells (Fig. 4 F). The decrease of the CIITA protein level after cycloheximide was stronger in PML-depleted fibroblasts (Fig. 4 F). Quantitation of signals confirmed that degradation of CIITA was significantly stronger in cells lacking PML (Fig. 4 G). These data demonstrate that also at the endogenous level, PML can regulate the amount of CIITA through protection from degradation.
CIITA forms a complex with PML isoform II at nuclear bodies
We then analyzed CIITA distribution in the nucleus in more detail (Fig. 5). All anti-CIITA antibodies tested in this study resulted in a cellular background immunofluorescence staining pattern even without IFN-γ treatment and IFN-γ induced only a slight increase of the anti-CIITA immunofluorescence signals (unpublished data). These observations indicated that the epitopes recognized by these antibodies are only partially accessible. We therefore tested several different established immunofluorescence protocols to find conditions, which result in more accessible epitopes (Guillot et al., 2004). This approach showed that only those protocols, which included a prepermeabilization step produced specific anti-CIITA immunofluorescence staining in IFN-γ–treated cells (Fig. 5 A and not depicted). When cells were permeabilized with Triton X-100 before fixation, endogenous CIITA was found in a dotlike pattern in the nucleus in addition to a microspeckled diffuse distribution (Fig. 5 A). The dotlike pattern of CIITA completely colocalized with PML bodies, showing that a subpopulation of endogenous CIITA resides at PML bodies (Fig. S3). We therefore conclude that epitopes recognized by the anti-CIITA antibody used here are inaccessible in normally fixed PML bodies but become accessible by detergent treatment before fixation. The same method did not lead to accumulation at PML nuclear bodies of the unrelated protein Bcl-x (Fig. 5 B).
To circumvent the limitations of anti-CIITA antibody staining, we used GFP-tagged CIITA. Expression of GFP-CIITA induced a dose-dependent expression of MHC class II genes (Fig. S4 A). Thus, the GFP-CIITA fusion protein is functional, confirming our published data (Hake et al., 2000; Camacho-Carvajal et al., 2004). Within the nucleus, GFP-CIITA was distributed homogenously (Fig. 5 C, left). Upon IFN-γ treatment, GFP-CIITA also accumulated in a dotlike pattern in the nucleus of some but not all cells (Fig. 5 C, right; and not depicted). The dots represented PML bodies as revealed by anti-PML coimmunostaining (Fig. 5 D). Coexpression experiments revealed that all PML isoforms colocalized with GFP-CIITA in the nuclear bodies (Fig. 5 E). The accumulation of GFP-CIITA in PML bodies varied with overexpression of different PML isoforms but appeared most efficient in cells coexpressing RFP–PML II (Fig. 5 E). We obtained similar results in PML knockout mouse fibroblasts transfected with the different PML isoforms (Fig. 5 F). Quantification demonstrated that RFP–PML II most efficiently attracted GFP-CIITA into nuclear bodies of HEp-2 cells. Because the same result was obtained in PML knockout cells (Fig. 5 G), we conclude that the observed interaction is specific for PML isoform II. Coimmunoprecipitation experiments were performed to confirm complex formation. IFN-γ–treated HEp-2 cells were cotransfected with GFP-tagged CIITA and RFP-tagged PML isoform II. RFP–PML II was coimmunoprecipitated by GFP-CIITA (Fig. 6 A, blot 2, lane 4), and reciprocally, GFP-CIITA was coimmunoprecipitated by RFP–PML II (Fig. 6 A, blot 1, lane 7). As controls, immunoprecipitates of untagged GFP did not contain RFP–PML II (Fig. 6 A, blot 2, lane 5), and immunoprecipitates of untagged RFP did not contain GFP-CIITA (Fig. 6 A, blot 1, lane 8). Another set of coimmunoprecipitation experiments revealed that GFP-CIITA most efficiently formed a complex with GFP-tagged PML II (Fig. 6 B, blots 2 and 3, lane 8). Nevertheless, small amounts of the other RFP-tagged PML isoforms were also detected in GFP-CIITA immunoprecipitates (Fig. 6 B, blot 3, lanes 7 and 9–12).
FRAP was used to determine protein complex formation at PML nuclear bodies in living cells. GFP-CIITA fluorescence was bleached at PML nuclear bodies or within the nucleoplasm (Fig. 7 A). Fluorescence recovery of GFP-CIITA after the bleach pulse was rapid within the nucleoplasm as well as at PML bodies (Fig. 7 A). Quantification of the FRAP data demonstrated that GFP-CIITA fluorescence recovered to ∼90% already after 5 s, and prebleach values were reached after 20 s in cells cotransfected with RFP-tagged PML isoform I (Fig. 7 B). This indicates that the complete pool of GFP-CIITA molecules exchanges rapidly at nuclear bodies and at chromatin. The latter is characteristic of typical transcription factor exchange at chromatin (Hager et al., 2009; Hemmerich et al., 2011). GFP-CIITA dynamics at PML bodies containing overexpressed RFP–PML I were very similar to those measured in the nucleoplasm, but the FRAP curve at PML bodies indicated a measurable decrease in GFP-CIITA mobility (Fig. 7 B). IFN-γ did not significantly change the mobility of GFP-CIITA in RFP–PML I-cotransfected cells (Fig. 7 B). Similar FRAP curves were observed in cells cotransfected with RFP-tagged PML isoforms III to VI (unpublished data).
In contrast, exchange of GFP-CIITA at PML bodies was substantially reduced in cells coexpressing RFP–PML II (Fig. 7 C). To quantify this difference, FRAP curves were fitted with exponential functions assuming two differently mobile GFP-CIITA populations as described previously (Hemmerich et al., 2008). The relative fractions and the recovery halftimes of GFP-CIITA in RFP-PML–coexpressing cells were determined from the two-component exponential fit functions (Table S2). The existence of at least two differently mobile fractions is consistent with previous analyses of transcription factor mobility in living cells. The initial fast recovery in FRAP experiments represents freely diffusing molecules, whereas the late phase of the FRAP curve is dominated by association/dissociation kinetics at chromatin. We determined the residence time of the slow exchanging population of GFP-CIITA from its exponential term of the fit function as previously described (Sprague et al., 2004). This approach revealed residence times of a few seconds for GFP-CIITA in chromatin as well as at nuclear bodies after cotransfection with RFP-tagged PML I and III to VI (Fig. 7 D, red bars). Similar results were obtained in IFN-γ–treated cells (Fig. 7 D, green bars). In contrast, the residence time of GFP-CIITA at nuclear bodies containing overexpressed RFP–PML II was 38 (±7) s, and this value increased to 226 (±22) s after IFN-γ treatment (Fig. 7 D). These data confirm that CIITA can directly or indirectly interact with PML isoform II at nuclear bodies and that this interaction is strongly enhanced by IFN-γ. All PML isoforms exchange at nuclear bodies with mean residence times of several minutes in living cells (Weidtkamp-Peters et al., 2008, Brand et al., 2010). We tested whether IFN-γ also had an effect on the exchange rate of PML II at nuclear bodies. Evaluation of the FRAP results indeed revealed that the exchange rate of GFP–PML II was decreased upon IFN-γ treatment compared with untreated cells (Fig. 7 E). The residence time of the slow exchanging fraction of GFP–PML II was decreased almost eightfold in the presence of IFN-γ (Fig. 7 E). Thus, the interaction between GFP-CIITA and RFP–PML II at nuclear bodies occurs on a hyperstable RFP–PML II scaffold to which GFP-CIITA can bind with residence times of several minutes.
The MHC II gene cluster is subject to rapid and massive conformational and epigenetic changes upon IFN-γ treatment in cells without constitutive MHC class II molecule expression. IFN-γ treatment induces a rapid looping of the MHC gene cluster away from CT-6 followed by complete decondensation of the locus (Volpi et al., 2000; Müller et al., 2004; Branco and Pombo, 2006; Christova et al., 2007). Evidence for a direct role of PML bodies in transcriptional regulation of MHC II genes was recently established by the observation that the HLA class II DRA gene is specifically relocated to PML bodies in HeLa cells upon IFN-γ induction (Gialitakis et al., 2010). PML bodies were proposed in this study as coactivators of DRA transcription by maintaining H3K4me2 levels at the DRA gene promoter through complex formation between the K4 methyltransferase mixed lineage leukemia (MLL) protein and PML isoform IV (Gialitakis et al., 2010). Here, we confirm that IFN-γ specifically evokes an increase of the spatial proximity between PML bodies and the MHC II gene cluster in human cells lines and extend these observations to primary human fibroblasts. We show for the first time that PML regulates IFN-γ–induced MHC II mRNA expression mainly through stabilization of the CIITA at the protein level. We also demonstrate that PML II, but no other PML isoform, forms a complex with CIITA at PML bodies. Our quantitative FRAP data indicate that the (overexpressed) PML II–CIITA complex can be stabilized at PML bodies for several minutes by IFN-γ. Compared with other isoforms, PML II also exhibited the strongest effect with respect to CIITA stabilization. These observations suggest that PML bodies may act as coregulators of IFN-γ–induced MHC II expression through a functional interaction between PML II and CIITA.
There is accumulating evidence for transcriptional control of nuclear body–associated genes through PML: During programmed in vitro differentiation of mouse embryonic stem cells via neural precursor cells into postmitotic neurons, PML bodies are tightly associated with the Oct3/4 locus when this gene is active in embryonic stem cells. Shut down of Oct3/4 mRNA transcription in neural precursor cells and postmitotic neurons is directly correlated with the loss of contact between PML bodies and the OCT3/4 locus (Aoto et al., 2006). A high association rate (50%) was also observed between PML bodies and the p53 gene in interphase Jurkat cells, but the functional relevance of this interaction was not investigated (Sun et al., 2003). Specific associations of genomic regions with PML bodies has led to the idea that they act as anchoring points for specific chromatin regions, where they have a function analogous to that postulated for matrix attachment regions (Wang et al., 2004; Ching et al., 2005). This idea has been substantiated recently by the demonstration of the organization of the MHC class I locus into distinct higher-order chromatin loop structures based on the physical and functional interaction between PML and the matrix attachment region–binding protein SATB1. These studies indicated that a SATB1–PML body network intersects at the MHC I locus to coordinate expression of MHC I genes (Kumar et al., 2007). Recently, it was shown that a CTCF–CIITA–RFX network complex mediates an IFN-γ–inducible long-range interaction between HLA-DRB1, HLA-DQA1, and the XL9 enhancer positioned between these two MHC II genes. These interactions are believed to form 25- and 15-kb loop-type structures connecting the promoter regions of HLA-DRB1 and HLA-DQA1 (Majumder et al., 2008). Subsequently, the same group showed that CTCF is required for a structural reorganization of the complete MHC II locus into distinct loop structures in a CIITA-dependent manner (Majumder and Boss, 2010). However, endogenous CTCF (MacPherson et al., 2009) and GFP-tagged CTCF (our unpublished data) do not accumulate at PML bodies. The structural and functional relationship between CTCF-mediated loop formation and CIITA–PML-mediated transcriptional control of MHC II mRNA expression remains to be determined.
The nonrandom close proximity between PML bodies and the MHC II cluster is not restricted to genomic subregions of this locus because FISH probes from different regions of the MHC II cluster are in similarly close proximity to PML bodies (Shiels et al., 2001). We also observed that in >80% of IFN-γ–treated human fibroblasts showing a close proximity between a PML body with one MHC II allele, the second allele was also in close proximity to another PML body. MHC II genes are not subject to allelic exclusion (Majumder and Boss, 2010). Because of the stochastic nature of transcription initiation (Suter et al., 2011) and dynamic conformation changes of the activated MHC II genomic region (Müller et al., 2004; Christova et al., 2007), its close proximity to PML bodies may be detected in only a fraction of cells at a given time. In addition, PML body–associated genes within the MHC II locus are transcriptionally active as revealed by RNA FISH (Wang et al., 2004) and quantitative PCR (qPCR; Kumar et al., 2007), indicating that PML bodies have no default transcription-repressive effect on genes in close proximity. Collectively, it is safe to say that PML bodies are in very close proximity to transcriptionally active MHC II genes in a considerable fraction of IFN-γ–treated cells. However, it should also be pointed out that a physical association between PML bodies and the MHC II locus is not an absolute requirement for MHC II gene transcription because cells without PML nuclear bodies still express detectable amounts of MHC II mRNAs (Fig. 2 E) from which MHC II molecules are being translated (Fig. 4 B).
In contrast to our study, Wang et al. (2004) did not detect differences in the expression level of genes in the proximity of PML bodies after PML knockdown. In the current study, we have specifically analyzed the expression of CIITA-dependent genes, which have not been investigated in the Wang et al. (2004) study. Therefore, the results obtained by Wang et al. (2004) cannot be directly compared with our findings. Furthermore, the PML knockdown in the Wang study yielded ∼50% of the cells without PML bodies, and a considerable amount of PML protein was still detected by Western blotting after siRNA (Wang et al., 2004). It cannot be fully excluded that the PML knockdown was not efficient enough to reveal a PML-mediated influence on the expression of the genes analyzed in this study.
The transcription of MHC II genes is quantitatively controlled by the level of CIITA (Fig. S4 A; Otten et al., 1998). We uncovered here that PML is able to modulate CIITA expression at the transcriptional level through the JAK–STAT signaling pathway at an early time point (3 h) after IFN-γ induction (Fig. 4 A). Subsequently, there was no significant difference of the CIITA mRNA level between control and PML knockdown cells. Because the protein half-life of CIITA is only ∼30 min as a result of rapid proteasomal degradation (Schnappauf et al., 2003), one can assume that the reduced CIITA mRNA level after 3 h of IFN-γ induction has only a little effect on the CIITA protein amount produced from mRNA during the 6–24-h interval. Consistent with the idea of a PML-mediated stabilization of CIITA at the protein level, we observed substantially reduced CIITA protein amounts in PML-depleted cells between 6 and 24 h after IFN-γ addition and increased degradation of endogenous CIITA in the absence of PML. These observations indicate that PML functions to protect CIITA from proteasomal degradation. In a similar fashion, PML also protects other nuclear factors, such as the transcription factors p53 and p73, and the DNA damage response protein TopBP1 from proteasomal degradation by sequestration into PML nuclear bodies (Louria-Hayon et al., 2003; Xu et al., 2003; Bernardi et al., 2004; Bernassola et al., 2004). This is solid evidence to suggest that one major function of PML bodies is stabilization of specific nuclear factors at the protein level.
We have shown previously that GFP–PML II is stably bound to PML bodies with a mean residence time of 6.2 min (Weidtkamp-Peters et al., 2008; Brand et al., 2010). Because PML II is able to physically tether CIITA to PML bodies for several minutes (Fig. 5), they might serve as a structural hub or surface at which CIITA is stabilized through protection from degradation. Currently, we do not know whether the IFN-γ–induced relocation of the MHC II gene cluster to PML bodies is functionally connected to the stabilization of CIITA. PML bodies may facilitate efficient assembly of CIITA into MHC II enhanceosome complexes, and the transcriptional activity of these complexes may be increased by more spatial proximity between MHC II genes and the nuclear bodies. This model is consistent with our observation that GFP-tagged RFX-AP, RFX-B, and NFYa can also bind to PML bodies and that a small but significant amount of MHC II promoter chromatin coimmunoprecipitates with endogenous PML (our unpublished data). Interestingly, IFN-γ also induced tighter binding of GFP–PML II to PML bodies (Fig. 7 E), whereas exchange of GFP–PML V and GFP–PML VI was unaffected (not depicted). The IFN-γ–induced increased residence time of PML II at nuclear bodies may be one mechanism for prolonged retention and stabilization of CIITA at PML bodies. The activity of CIITA is regulated by PTMs, including phosphorylation, acetylation, and ubiquitination (Wu et al., 2009). PML bodies are potential hot spots for such PTMs (Bernardi and Pandolfi, 2007). Acetylation of CIITA by CBP at K141 and K144, which up-regulates CIITA activity at MHC II promoters (Spilianakis et al., 2000), could be mediated by the acetyltransferase CBP, which is also a PML body component (Boisvert et al., 2001). It will be interesting to reveal which PML body–localized PTMs contribute to CIITA stability.
The SATB1–PML, the MLL–PML IV, and the CIITA–PML II interaction networks (Kumar et al., 2007; Gialitakis et al., 2010; and this study, respectively) provide compelling evidence for a direct role of PML bodies in the transcriptional regulation of the human MHC locus (Shiels et al., 2001). In all of these studies, specific PML isoforms were shown to contribute to transcriptional regulation of MHC genes. The underlying mechanisms share common features such as close proximity between the PML body and MHC genes, but they are clearly different in nature, including regulated chromatin loop formation (SATB1–PML), epigenetic marking (MLL–PML IV), and regulation of CIITA protein levels (CIITA–PML II). Future studies will further unravel the molecular details of how PML or the PML nuclear bodies participate in these multilevel mechanisms underlying the control of MHC molecule expression.
Materials and methods
Cell culture and transfection
The following cell lines and primary cells were used: human HEp-2 cervix carcinoma epithelial cells (CCL23; American Type Culture Collection), human MRC-5 lung fibroblasts (CCL171; American Type Culture Collection), primary human dermal fibroblasts from foreskin or biopsies (provided by R.W. Kinne, Experimental Rheumatology Unit, University Hospital Jena, Jena, Germany), primary human dermal fibroblast lines stably transduced with control or PML shRNA vectors were established as described previously (Tavalai et al., 2006), and mouse 3T3-PML+/+ and 3T3-PML−/− cells (provided by T. Hofmann, University of Heidelberg, Heidelberg, Germany; Brand et al., 2010). Cells were cultured in DME supplemented with 10% fetal calf serum in a 9.5% CO2 atmosphere at 37°C. IFN-γ (I-3265; Sigma-Aldrich) was added to the cells as indicated. For live-cell imaging experiments, cells were seeded on 42-mm glass dishes (H. Saur Laborbedarf) and transfected with plasmid DNA 1–2 d before observation using transfection reagent (FuGENE HD; Roche) according to the manufacturer’s protocol.
The following primary antibodies were used: anti-PML rabbit antibody (ABD-030; Jena Biosciences), anti-PML mAb (PG-M3; Santa Cruz Biotechnology, Inc.), antitubulin mAb (T-9026; Sigma-Aldrich), anti-CIITA rabbit antibody (serum 21; Bontron et al., 1997), anti-Sp100 rabbit antibody (ABD-031; Jena Biosciences), anti-Sp100 rabbit antibody (GH3; provided by H. Will, Heinrich Pette Institute, Hamburg, Germany; Milovic-Holm et al., 2007), anti–HLA-DR mAb (G46-6; BD), anti–HLA-DR/DP/DQ mAb (TÜ-39; BD), anti–Bcl-x rabbit mAb (#1018-1; Epitomics; Enzo Life Sciences), anti–RFX-5 rabbit antibody (NB100-195; Novus Biologicals), and anti–NFY-C rabbit antibody (NBP10-19147; Novus Biologicals). The following secondary antibodies were used: Alexa Fluor 488-anti–mouse IgG (Invitrogen), Alexa Fluor 546-anti–sheep (Invitrogen), Cy2-, Cy3-, or Cy5-anti–rabbit IgG (highly cross-adsorbed; Dianova), and horseradish peroxidase–coupled anti–rabbit IgG and anti–mouse IgG (highly cross-adsorbed; Dianova).
The following plasmids were used for transient transfection assays: pEGFP–PML I to VI and pmRFP–PML I to VI were described in detail previously (Weidtkamp-Peters et al., 2008; Brand et al., 2010); pmRFP-C1/2/3 vectors were cloned by replacing the GFP cDNA in pEGFP-C1/2/3 with a cDNA encoding monomeric RFP (provided by R. Tsien, Stanford University, Stanford, CA) as a PCR fragment via the AgeI and BsrGI sites; pEGFP-CIITA is described in Schnappauf et al. (2003); plasmids pEGFP–RFX-5, pEGFP–RFX-AP, and pEGFP–RFX-B were provided by J. Boss (Emory University School of Medicine, Atlanta, GA). All pEGFP and pmRFP vectors contain full-length cDNAs of these proteins fused to the 3′ end of GFP or RFP, respectively, giving rise to N-terminally fluorescently labeled fusion proteins. NFY-expressing GFP plasmids were provided by D. Doenecke (University of Göttingen, Göttingen, Germany).
Whole-cell extracts were electrophoresed on SDS-PAGE and transferred to nitrocellulose membrane (Protran; Schleicher & Schuell). The membrane was blocked in 5% skimmed milk/PBS-T (PBS + 0.1% [vol/vol] Tween 20 [Carl Roth]) and incubated with protein-specific primary and subsequently horseradish peroxidase–conjugated species-specific secondary antibodies (Jackson ImmunoResearch Laboratories, Inc.). Signal was detected using the ECL reagent (GE Healthcare) on imaging film (BioMax; Kodak).
Immunocytochemistry and microscopy
Cells grown on 15-mm-diam coverslips were fixed with 4% formaldehyde for 10 min and permeabilized with 0.25% Triton X-100 for 3 min. Primary antibodies were incubated on cells for 45 min. After washing steps with PBS, secondary antibodies coupled to appropriate fluorophores were incubated on cells for 45 min followed by a DNA staining step using TO-PRO-3 (Invitrogen) for 10 min and mounting with antifade mounting medium (ProLong gold; Invitrogen). Other cell fixation methods included different protocols described in detail previously by Guillot et al. (2004). In brief, protocols were as follows: for protocol 1, cells on coverslips rinsed in PBS (10 mM Na2HPO4/NaH2PO4, 137 mM NaCl, and 2.7 mM KCl, pH 7.4), fixed/permeabilized in ethanol (20 min at −20°C); for protocol 2, cells on coverslips were rinsed in PBS-Mg, fixed in 1.75% formaldehyde in PBS-Mg (10 mM Na2HPO4/NaH2PO4, pH 7.4, 150 mM NaCl, and 2 mM MgCl2; 20 min at 4°C), rinsed in PBS-Mg (15 min), and permeabilized with 0.5% Triton X-100 in PBS-Mg (15 min); for protocol 3, cells on coverslips were rinsed in PBS-Mg, permeabilized in 1.5% Triton X-100 in PBS-Mg (10 min at 4°C), rinsed, and fixed in 1.75% formaldehyde in PBS-Mg (20 min at 4°C); for protocol 4, cells on coverslips rinsed in PBS, fixed in 4% formaldehyde and 250 mM Hepes, pH 7.6 (10 min and 4°C), and refixed in 8% formaldehyde and 250 mM Hepes, pH 7.6 (50 min at 4°C) followed by permeabilization in 0.5% Triton X-100 in PBS (30 min with gentle rocking); for protocol 5, cells on coverslips were rinsed in PBS, fixed/permeabilized in 0.1% Triton X-100 in 4% formaldehyde and 250 mM Hepes, pH 7.6 (10 min at 4°C), refixed in 8% formaldehyde and 250 mM Hepes, pH 7.6 (50 min at 4°C), and permeabilized as in protocol 4; for protocol 6, cells on coverslips were rinsed in PBS, fixed in 4% formaldehyde and 125 mM Hepes, pH 7.6 (10 min at 4°C), and refixed in 8% formaldehyde and 125 mM Hepes, pH 7.6 (50 min at 4°C) followed by permeabilization in 0.5% Triton X-100 in PBS (30 min with gentle rocking). For microscopy of fixed cells, a laser-scanning confocal microscope (LSM 510 Meta or LSM 710/ConfoCor 3; Carl Zeiss) was used using a 63×-fold oil objective (numerical aperture of 1.3) and ZEN software (Carl Zeiss).
For immuno-FISH analyses, cells were fixed in 4% and then in 8% formaldehyde in 250 mM Hepes, pH 7.6 (10 min and 2 h, respectively), according to previously established protocols (Guillot et al., 2004). Permeabilization was performed in 0.5% saponin/0.5% Triton X-100/20 mM glycin/PBS for 1 h. After extensive PBS washes, cells were stored in 20% glycerol/PBS at −80°C. Indirect immunofluorescence was performed as described in the previous section, and cells were fixed again in 4% formaldehyde/0.2% Triton X-100/PBS followed by 20 nm glycin/PBS incubation for 30 min, and another 30-min incubation in 0.1 M HCl. Cellular RNA was digested in a solution containing 500 µg/ml RNase A (Roche) at 37°C for 2 h. DNA was denatured in 70% deionized formamide in 2× SSC for 8 min at 75°C. Hybridization using a MHC II–specific probe was performed at 37°C for 48 h in 50% deionized formamide/2× SSC/10% dextrane sulfate. The probe to detect the MHC II locus (bacterial artificial chromosome clone RP11-10A19; BACPAC Resources, Children’s Hospital Oakland Research Institute) was characterized previously (Müller et al., 2004). The probe was labeled with digoxigenin-11–deoxy-UTP using a Nick Translation kit according to the manufacturer’s protocol (Roche), denatured for 10 min at 70°C, and coprecipitated with human Cot1 and herring sperm DNA. Posthybridization washes were as follows: 50% formamide in SSC (1 h at 37°C), 0.1× SSC (30 min at 60°C), and 4× SSC/0.1% Tween 20 (20 min at 42°C). The probe was detected using a sheep antidigoxigenin antibody (Roche). Blocking and washing steps were performed in 0.5% blocking solution (Tris-HCl/150 mM NaCl at 37°C; Roche). Nuclei were counterstained with 2 mM TOTO-3 (Invitrogen). Coverslips were mounted and polymerized in MobiGLOW (MoBiTec).
HEp-2 cells were transfected and treated as indicated. After 2 d, ∼107 cells were harvested, shock frozen in liquid nitrogen, suspended in 200 µl lysis buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 5 mM EDTA, 0.5% NP-40, freshly added 40 mM N-ethylmaleimide, 0.1% benzonase nuclease, 2 mM Pefabloc, 2 µg/ml aprotinin, 2 µg/ml leupeptin, and 2 µg/ml pepstatin) for ∼15 min on ice, and subjected to sonication (10 cycles of 15 s on/off at high intensity and 4°C; Bioruptor Plus; Diagenode). After centrifugation for 15 min at 16,100 g and 4°C, the supernatant was diluted with 800 µl of ice-cold washing buffer (10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.5 mM EDTA, freshly added 2 mM Pefabloc, 20 mM N-ethylmaleimide, and protease inhibitor cocktail [Roche]), keeping 25 µl of the supernatant as input for Western blotting. For each precipitation, ∼25 µl GFP- or RFP-Trap bead slurry (ChromoTek) was equilibrated in washing buffer, added to the dilution, and incubated for ∼3 h with gentle end-over-end mixing at 4°C. After centrifugation (2 min at 2,700 g and 4°C) and discarding the supernatant, the beads were washed twice, resuspended in 50 µl SDS sample buffer, and boiled for 10 min at 95°C.
FRAP experiments were performed at 37°C in Hepes-buffered medium on a confocal microscope (LSM 710/ConfoCor3) essentially as described before (Weidtkamp-Peters et al., 2008; Brand et al., 2010), using a 63× oil objective (numerical aperture of 1.3) and ZEN software. 5–10 images were taken before the bleach pulse, and 50–200 images were taken after bleaching of regions of interest containing one nuclear body each at 0.05% laser transmission to minimize scan bleaching. Image acquisition frequency was adapted to the recovery rate of the respective GFP fusion protein. The pinhole was adjusted to 1 airy unit. Quantitation of relative fluorescence intensities was performed according to Weidtkamp-Peters et al. (2008) using Excel (Microsoft) and Origin software (OriginLab). Recovery halftimes and relative amounts of the two differently mobile populations of molecules exchanging at PML bodies were determined from the fitted two-component exponential functions (Sprague et al., 2004). Assuming that FRAP of the slower exchanging component is unaffected by free diffusion and therefore reaction dominant, the dissociation rate (koff) of the slower component is identical to the decay constant (kslow) of its exponential term (Sprague et al., 2004). Therefore, the residence time was calculated as (kslow)−1.
HEp-2 cells at low density in 6-well plates were transfected with SMARTpool siRNAs against PML, Sp100, CIITA, or a control siRNA (Thermo Fisher Scientific). Transfection was performed using HiPerFect (QIAGEN) at 50-nM RNAi concentration. 3 d after transfection, siRNA-treated cells were harvested. Each experiment was performed at least three times.
RNA extraction and cDNA synthesis
Total RNA was extracted from whole cells using an isolation system (peqGOLD RNAPure; Peqlab). RNA samples were resuspended in sterile diethylpyrocarbonate pyrogen-treated, RNase/DNase-free water (Invitrogen). RNA concentration was measured using the spectrophotometer (NanoDrop ND-1000; Thermo Fisher Scientific). Reverse transcription synthesis of cDNA was performed on 1 µg of total RNA using the SuperScript III First-Strand Synthesis SuperMix (Invitrogen) with random hexamer primers following the manufacturer’s instructions.
Quantitative real-time PCR
5 ng cDNA was used for quantitative real-time PCR of a 25-µl total volume. Each independent PCR reaction contains 12.5 µl Platinum SYBR Green qPCR SuperMix uracil DNA glycosylase (Invitrogen) and 0.2 µM of each primer. The run and analysis were performed using a real-time PCR instrument and software (iQ5; Bio-Rad Laboratories). Primer sequences are as follows: LMP7 forward, 5′-CAGCTATTCTGGAGGCGTTG-3′, and reverse, 5′-GCTGAGCCCGTACTCTCTCTT-3′; TAP1 forward, 5′-CAGGAGACGGAGTTTTTCCA-3′, and reverse, 5′-CAGAGCATGATCCCCAAGAG-3′; HLA-F forward, 5′-AGGAACAGACCCAGGACACA-3′, and reverse, 5′-ACCACAGCTCCAAGGACAAC-3′; HLA-DRA-A forward, 5′-TAAGGCACATGGAGGTGATG-3′, and reverse, 5′-GTACGGAGCAATCGAAGAGG-3′; HLA-DMB forward, 5′-CTCTCACAGCACCTCAACCA-3′, and reverse, 5′-TAGAAGCCCCACACATAGCA-3′; and CIITA forward, 5′-AGCCTTTCAAAGCCAAGTCC-3′, and reverse, 5′-TTGTTCTCACTCAGCGCATC-3′. For the normalization of the qPCR values, glyceraldehyde 3-phosphate dehydrogenase was used as a housekeeping gene (forward, 5′-GTCAGTGGTGGACCTGACCT-3′, and reverse, 5′-AAAGGTGGAGGAGTGGGTGT-3′). All primers were validated, and serial dilutions of cDNA were run to generate a standard curve to analyze the efficiency of the primers.
Primer for semiquantitative RT-PCR (Fig. 2 D)
Primers were as follows: CIITA forward, 5′-AACATCACTGACCTGGGTGCCTAC-3′, and reverse, 5′-CCCACGTCGCAGATGCAGTTATTG-3′; TAP1 forward, 5′-AGGGCTATGACACAGAGGTAGAC-3′, and reverse, 5′-CTGGTGGCATCATCCAGGATAAG-3′; LMP7 forward, 5′-GGTCCTACATTAGTGCCTTACG-3′, and reverse, 5′-AGTACAGCCTGCATTCCTTG-3′; HLA-DRA forward, 5′-TCCGCAAGTTCCACTATCTC-3′, and reverse, 5′-ACACCACGTTCTCTGTAGTC-3′; HLA-DMB forward, 5′-TGTCCTCTCACAGCACCTCAAC-3′, and reverse, 5′-ACATAGCAGGCCAGCATCACAG-3′; and glyceraldehyde 3-phosphate dehydrogenase forward, 5′-TGCACCACCAACTGCTTAG-3′, and reverse, 5′-GATCGCAGGGATGATGTTC-3′.
Theoretical association probability
To calculate the theoretical probability for a colocalization event between sphere-shaped objects (PML bodies and gene loci) in the nucleus, the volume of the nucleus and the number and volume of the objects were taken into account. These data were determined from the 3D image stacks of FISH experiments. In case of a colocalization (or association) event of a genomic locus and a PML body, the distance between the centers of the two colocalizing distinct spheres within the nucleus is equal or less than the sum of the sphere’s radii: d < rg + rp, in which rg is the radius of the genomic locus, and rp is the radius of a PML body. Considering one single nucleus of radius R with m PML bodies of radius rp each, and n genomic loci with radius rg each, we assume that the positions of all genomic loci are equally distributed within the nucleus and that no two genomic loci are closer than 2 rg + 2 rp, which guarantees that a specific PML body does not colocalize with more than one genomic locus at a time. Then, considering the volume of the nucleus and the total volume of the spheres around all genomic loci, which define a colocalization event, the colocalization probability p for one PML body is given by p = n (rg + rp)3/R3. As for m PML bodies, the probability pm that no PML body colocalizes with any genomic locus is pm = (1 − p)m. We get the final result for the probability pA to find at least one colocalization between a PML body and a genomic locus within the whole nucleus with m PML bodies and n genomic loci: pA = 1 − (1 − n (rg + rp)3/R3)m. This model contains the explicit assumption that two genomic loci have a minimum distance of 2(rp + rg), which is a plausible assumption regarding the size of genomic elements and of PML bodies. On the contrary, PML bodies are allowed to intersect in this probability model, which does not coincide with our experimental findings. Nevertheless, for the numbers of PML bodies in our experiments, the effect of this approximation on the true probability is very small.
From the colocalizing probability pA within one nucleus, one can compute the probability pk(j) for the event to find among k investigated cells j cells with at least one colocalization caused by chance: pk(j) = (k/choose j) pAj (1 − pAj)(k − j). If in an experiment with k cells there are q cells with at least one colocalization, the p-value for this event is the sum of the probabilities pk(j) to find q cells or more with at least one colocalization: p-value = SUM(q ≤ I ≤ k) pk(i).
Online supplemental material
Fig. S1 shows by immunostaining that a transient siRNA-mediated knockdown of PML in HEp-2 cells leads to reduced cellular IFN-γ–induced MHC class II molecule expression. Fig. S2 shows by Western blotting that lack of PML protein does not affect the protein level of the MHC class II enhanceosome components RFX5 or NFY-C. Fig. S3 shows by immunostaining that a subfraction of CIITA accumulates at all PML bodies at the endogenous level. Fig. S4 shows by qPCR and immunofluorescence that the fusion protein GFP-CIITA is functional with respect to induction of MHC II mRNA and protein expression, respectively. Table S1 shows the theoretical probability of association between PML bodies and genomic loci based on our self-developed software model. Table S2 provides a complete list of all dynamic parameters obtained from FRAP analyses of GFP-CIITA at PML bodies in nuclei of living cells dependent on coexpression of individual PML isoforms.
We are grateful to J. Boss, D. Doenecke, and H. Will for providing expression plasmids and/or antibodies. P. Hemmerich and A. Horch wish to thank Ana Pombo for extensive help with FISH experiments.
This work was supported by grants HE 2484/3-1 and SFB796 from the Deutsche Forschungsgemeinschaft.
class II transactivator
human foreskin fibroblast
human leukocyte antigen
mixed lineage leukemia
regulatory factor X
small hairpin RNA
signal transducer and activator of transcription