P-STAT1 mediates higher-order chromatin remodelling of the human MHC in response to IFNγ

The human major histocompatibility complex (MHC) located on chromosome band 6p21 is the most important genomic region with respect to immunity to infectious agents, autoimmunity and transplantation (Horton et al., 2004). The products of the classical MHC class I (HLA-A, HLA-B, HLA-C) and class II (HLA-DR, HLA-DP, HLA-DQ) genes present processed antigens to cytotoxic and helper T-cells, respectively. The MHC class III region encodes complement proteins and inflammatory cytokines. Classical MHC class II genes are expressed constitutively only in antigen-presenting cells such as B-lymphocytes, macrophages and dendritic cells. Treatment of other cell types with interferon-γ (IFNγ), however, induces expression of the classical class II and several other genes in the MHC via the JAK/STAT signalling pathway (Boehm et al., 1997; Stark et al., 1998).

IFNγ binding to its receptor at the cell membrane leads to phosphorylation of JAK1 and JAK2. These proteins then phosphorylate the transcription factor STAT1 at Tyr701. Phosphorylated STAT1 (P-STAT1) dimerises and moves rapidly into the nucleus, where it binds to the GAS (gamma activating sequence) element of the promoters to initiate transcription of IFNγ primary response genes. In the MHC, these include TAP1 (Min et al., 1996), Hsp70/90 (HSPA1) (Stephanou et al., 1999) and tapasin (TAPBP) (Herberg et al., 1998). The primary response genes on other chromosomes include IRF1 (Hobart et al., 1997) and the class II transactivator gene (CIITA) (van den Elsen et al., 2004), which are required for subsequent activation of the HLA genes (Reith et al., 2005). HLA class I and II genes share well-conserved promoter modules to which the constitutively expressed transcription factors, RFX, X2BP and NF-Y, bind cooperatively to form the enhanceosome. In response to IFNγ, CIITA is synthesised, and then binds and stabilises the enhanceosome (Gobin et al., 1999; Krawczyk et al., 2004). Histone hyperacetylation of HLA class II genes, including HLA-DRA, is detected 4 hours after the start of IFNγ treatment, and is followed by transcription 2 hours later (Spilianakis et al., 2003).

The cell-type-specific and inducible expression of the MHC makes it a powerful model system for analysing the relationship between transcription and chromatin architecture. We previously found that transcriptional activation by IFNγ of the human HLA class II genes in fibroblasts is preceded by massive remodelling of the chromatin fibre, which manifests as a rapid looping-out from the chromosome 6 territory (CT6) (Volpi et al., 2000). Similar giant chromatin loops have been observed for the human epidermal differentiation complex (Williams et al., 2002), the human and mouse β-globin loci (Ragoczy et al., 2003) and the mouse Hox gene cluster (Chambeyron and Bickmore, 2004). The significance of this chromatin remodelling over large genomic regions is not known. However, important insights into the relationship between chromatin movement and gene expression have been obtained from studies on artificial transgene arrays where targeting of strong transcriptional activators initiates recruitment of histone acetyl transferases (HATs) and other chromatin modifiers, such as BRG1, accompanied by large-scale chromatin decondensation (Carpenter et al., 2005; Muller et al., 2007).

To understand the mechanism and significance of higher-order chromatin remodelling of the MHC, we examined the architecture of the locus together with DNA-protein interactions and histone acetylation at the early stages of IFNγ induction. We provide evidence for a direct link between the JAK/STAT signalling pathway and higher-order chromatin remodelling across the MHC. We further show that the conformational changes reflect a statistically significant level of decondensation in the MHC but not in the surrounding genomic regions. Taken together, our findings suggest that the induction of an `open' or transcriptionally competent chromatin conformation is a crucial early step in cytokine-mediated activation of the MHC.

We previously found a clear difference in the conformation of the chromatin fibre carrying the MHC in cells with different profiles of HLA class II expression (Volpi et al., 2000). In B-lymphoblastoid cells, in which the HLA class II genes are constitutively expressed, the MHC is present on a giant external chromatin loop in ∼35% of CT6s. By contrast, in fibroblasts that do not express HLA class II genes, the frequency is ∼10%, increasing after IFNγ treatment to ∼35%. It is of particular interest that this conformational change is seen in fibroblasts after only 10 minutes exposure to IFNγ, several hours before the HLA class II gene cluster is transcribed.

To understand the significance of these observations, we addressed the following questions. Is the massive chromatin conformational change induced through the JAK/STAT pathway, which is required for subsequent HLA class II transcription? Which DNA-protein interactions are significant in this process? Does histone acetylation play a role? Finally, does the visible conformational change reflect an altered level of chromatin condensation?

To determine whether higher-order chromatin remodelling occurs through the JAK/STAT signal transduction pathway, we analysed a well-characterised set of STAT1-mutant cell lines derived from the fibrosarcoma cell line HT1080 (Chatterjee-Kishore et al., 2000; Muller et al., 1993). We first verified that IFNγ was able to induce looping-out of the MHC in HT1080 cells by analysing the location of MHC-specific probes relative to the chromosome-6-territory paint by FISH. As in our previous work, `external chromatin loops' were defined as configurations where the MHC probe signal was outside the painted chromosome 6 domain, without touching the border of the domain. This included observations in which a faint stalk-like projection was seen extending from the chromosome domain towards the MHC signal. Each chromosome homologue was counted separately and a score derived for the percentage of loci that were located on an external chromatin loop. These criteria were applied to all cell types studied and all FISH experiments were repeated at least twice, and analysed by two independent researchers to ensure objective, reproducible results. The frequency with which the MHC locus was present on an external chromatin loop in untreated HT1080 cells, 13% of CT6s (n=306), almost doubled within the first 10 minutes of IFNγ treatment, reaching a peak of 34% at 24 hours (n=219, P<0.05) (Fig. 1A, supplementary material Fig. S1). This time-course is identical to our previous findings on MRC5 fibroblasts treated with IFNγ (Volpi et al., 2000), indicating that HT1080 cells provide a suitable model system for studying higher-order chromatin architecture in the MHC.

The role of STAT1 in higher-order chromatin remodelling of the MHC was then examined in HT1080-derived U3A cells, which are STAT1-negative and not responsive to IFNγ (Chatterjee-Kishore et al., 2000; Muller et al., 1993). No higher-order chromatin remodelling of the MHC was detected in U3A cells in response to IFNγ (MHC locus on an external chromatin loop before treatment in 16% of CT6s, n=493; and after treatment in 15% of CT6s, n=481; P=0.59; Fig. 1B). A complemented U3A cell line (U3A/STAT1) that stably expresses STAT1 (Kumar et al., 1997; Muller et al., 1993) was then examined. STAT1 is phosphorylated in these cells to normal levels and HLA class II genes are expressed in response to IFNγ (Improta et al., 1994). IFNγ treatment was able to induce MHC looping-out from the CT6 in U3A/STAT1 cells (MHC locus on an external chromatin loop before treatment in 18% of CT6s, n=782, and after treatment in 26% of CT6s, n=449, P<0.05; Fig. 1C). However, complementation of U3A cells with a Tyr to Phe point mutant of STAT1 (STAT1^Y701F) that cannot be phosphorylated by JAKs, did not restore MHC looping-out in response to IFNγ (MHC locus on an external chromatin loop before treatment in 17% of CT6s, n=393, and after treatment in 15% of CT6s, n=448; P=0.45; Fig. 1D). These findings implicate P-STAT1 in both transcriptional upregulation of HLA class II genes and higher-order chromatin remodelling of the MHC in response to IFNγ.

The rapid induction of higher-order chromatin remodelling in the MHC by IFNγ suggested that CIITA is not involved. To test this and to distinguish a possible direct role for STAT1 in chromatin remodelling from its role in activating CIITA transcription, U3A cells were transfected with a constitutively expressed CIITA vector (pEBS-PL-CIITA form III). Although CIITA was detected at both mRNA and protein levels (see supplementary material Fig. S2A), neither HLA-DRA expression (see supplementary material Fig. S2B) nor looping-out of the MHC were observed (MHC locus on an external chromatin loop in 17% of CT6s, n=314, P=0.80). These findings indicate that expression of CIITA alone is not sufficient for IFNγ induced changes in chromatin structure across the MHC locus.

To determine the order of transcription factor interactions at individual genes, chromatin immunoprecipitation (ChIP) experiments were performed on HT1080 and STAT1-null U3A cells at different times after the start of IFNγ treatment. In vivo recruitment of STAT1, RNAP II, TFIIB, BRG1, and histone acetylation changes to promoters of the TAP1, HSPA1 and HLA-DRA genes located in the MHC (see supplementary material Fig. S3), and the IRF1 gene located on chromosome 5, were then determined by quantitative real-time (RT)-PCR. Since the classical MHC class II genes are coordinately regulated, HLA-DRA was used, as elsewhere, as a model for changes that occur at all of the class II genes. GAPDH promoter sequences were amplified from the same immunoprecipitated material as a control for the efficiency of immunoprecipitation. IFNγ-induced changes were calculated relative to non-induced levels for each time point.

Phosphorylation of STAT1 and its translocation to the nucleus, where it binds to the target sequences with very high specificity, occurs within 5 minutes of IFNγ treatment (Haspel et al., 1996). We found that, during IFNγ treatment of HT1080 cells, P-STAT1 becomes associated with the promoters of the primary response genes TAP1, HSPA1 and IRF1, reaching a maximum at 30 minutes and decreasing after 1 hour. P-STAT1 did not associate with the GAPDH promoter, which has no STAT1-response element (Fig. 2). As expected, the same experiments performed on U3A cells showed no P-STAT1 binding. The slight delay in the detection of P-STAT1 in the ChIP assay, as compared with FISH observations, is probably owing to the characteristics of the ChIP method, which detects the average promoter occupancy in a population of cells at any given time.

These findings are consistent with the activation-deactivation cycle of STAT1 revealed by EMSA (Haspel and Darnell, 1999). We used for our study a highly specific antibody that recognises only the IFNγ-activated form of STAT1, phosphorylated on the Tyr701. Therefore, we cannot exclude the possibility that STAT1 is dephosphorylated on the target sequence and then continues to play a role in the activation of other genes, such as LMP2 (PSMB9) (Chatterjee-Kishore et al., 2000). When combined with the observation that higher-order chromatin architecture of the MHC is unaffected by IFNγ in the U3A cells complemented with the phosphorylation-defective STAT1^Y701F protein, these data strongly suggest that P-STAT1 is the factor that transmits the signal for chromatin modification to specific sites.

A significant increase in RNAP II recruitment was found at the TAP1, HSPA1 and IRF1 promoters in HT1080 cells at 1 hour of IFNγ treatment, and the level escalated thereafter (Fig. 2). By contrast, there was no enrichment of RNAP II at the HLA-DRA promoter during the first 6 hours of IFNγ induction, as expected from the lack of transcriptional activity of CIITA-dependent MHC genes during this time period (Spilianakis et al., 2003). RNAP II was not found at the promoters tested at any time in STAT1-null U3A cells, consistent with the absence of TAP1, IRF1 and HLA-DRA expression (Chatterjee-Kishore et al., 1998; Muller et al., 1993).

TFIIB was found to be associated with the promoter regions of the primary response genes TAP1, HSPA1 and IRF1, and the CIITA-dependent HLA-DRA gene at the same levels before and after IFNγ treatment (data not shown). The enrichment at these promoters was well above (20-30 times) that for the β-globin gene, which is not expressed in fibroblasts (see supplementary material Table S1), and did not differ between HT1080 and U3A cells, indicating that the binding of TFIIB to MHC promoters is not dependent on STAT1. The presence of TFIIB and other general transcription factors (Spilianakis et al., 2003) at the promoters of these genes before they are expressed might reflect the presence of partially assembled pre-initiation complexes to help maintain a transcriptionally poised chromatin conformation.

ATP-dependent chromatin remodelling complexes play a significant role in altering chromatin structure during mammalian differentiation, cell cycle and recombination (reviewed in de la Serna et al., 2006). They act by disrupting histone-DNA interactions, leading to the exposure of DNA sequences to regulatory proteins. BRG1, a catalytic subunit of the chromatin-remodelling SWI/SNF complex, has been implicated in transcriptional activation of the heat shock-induced mouse gene hsp70 (homologous to human HSPA1) (Corey et al., 2003) and certain human IFNα-induced genes (Huang et al., 2002). Analysis of BRG1 recruitment during IFNγ treatment in HT1080 cells revealed enrichment relative to non-induced levels at the TAP1 and IRF1 promoters reaching a maximum at 30 minutes after the start of IFNγ treatment (Fig. 2). The levels then decreased at 1 hour, similar to the kinetics of P-STAT1 recruitment. BRG1 was not associated with the HSPA1 or HLA-DRA promoters in the first 6 hours of IFNγ treatment, indicating that its recruitment to HSPA1 might be activator- or species-dependent. BRG1 was not enriched at any of the promoters examined in STAT1-negative U3A cells. These findings suggest that BRG1 is involved in STAT1-dependent remodelling of the TAP1 and IRF1 promoters in response to IFNγ.

To determine whether BRG1 participates in IFNγ-induced higher-order chromatin remodelling of the MHC, we analysed the cell line SW13, which lacks BRG1 and BRM, another catalytic subunit of the SWI/SNF complex (Pattenden et al., 2002). SW13 cells do not express HLA class II genes but have an intact JAK/STAT pathway. The induction of at least one IFNγ response gene, CIITA, can be restored after introduction of exogenous BRG1 (Pattenden et al., 2002). Expression of TAP1, HSPA1, IRF1, HLA-DRA and CIITA was found by RT-PCR to be unaffected in SW13 by treatment with IFNγ (see supplementary material Fig. S4A). FISH analysis showed that IFNγ treatment also had no significant effect on higher-order chromatin conformation of the MHC (MHC locus on an external chromatin loop before treatment in 19% of CT6s, n=762, and after treatment in 16% of CT6s, n=743, P=0.10; Fig. 3). However, complementation of SW13 cells with BRG1 (retroviral expression vector pBabe-IRESpuroBRG1) restored induction of TAP1, IRF1, HLA-DRA and CIITA in response to IFNγ (see supplementary material Fig. S4A). In these cells, IFNγ induced looping-out of the MHC (MHC locus on an external chromatin loop before treatment in 15% of CT6s (n=294), and after treatment in 29% CT6s; n=294, P<0.05; Fig. 3), as in HT1080 cells. SW13 cells stably infected with BRM or an empty vector showed no increase in the number of external chromatin loops. These findings suggest that BRG1 is involved in higher-order chromatin remodelling of the MHC upon IFNγ induction.

Histone acetylation is recognised as a hallmark of transcriptionally competent chromatin. At 4-8 hours after the start of IFNγ treatment, histone acetylation is reported to increase at the promoter of the HLA-DRA gene in a CIITA-dependent manner (Beresford and Boss, 2001; Spilianakis et al., 2003). Since STAT1 interacts with CBP/p300, which has histone acetylase activity (Zhang et al., 1996), we set out to determine whether higher-order chromatin remodelling arose from earlier histone hyperacetylation within the MHC. ChIP experiments on HT1080 cells and MRC5 fibroblasts following IFNγ treatment revealed a progressive increase in histone H3 acetylation at the promoters of the primary response genes TAP1 and HSPA1, reaching a maximum at 2-4 hours of IFNγ treatment (Fig. 2). At the HLA-DRA promoter, an increase in histone H3 acetylation was observed after 2 hours of IFNγ treatment, increasing further within 24 hours. The same experiments performed on STAT1-deficient U3A cells showed no IFNγ induced acetylation at any of the promoters tested. Histone H3 acetylation occurring at both primary and CIITA-dependent promoters within the first 4 hours of IFNγ treatment is therefore STAT1-dependent. As no significant histone hyperacetylation was observed at the genes examined immediately after the start of IFNγ treatment, these experiments indicate that histone hyperacetylation is unlikely to play a role in the onset of higher-order chromatin remodelling in the MHC.

Next, we investigated whether unspecific hyperacetylation by histone deacetylase (HDAC) inhibitors could induce higher-order chromatin remodelling in the MHC. MRC5 fibroblasts were treated with 10 mM sodium butyrate (SB) for 1 hour in order to avoid induction of apoptosis and cell cycle arrest associated with prolonged exposure to HDAC inhibitors. ChIP analysis of HLA-DRA indicated that treatment with SB for 1 hour leads to the maximum level of acetylation in H4 histones and two-thirds of the maximum acetylation in H3 histones (data not shown). No transcriptional activation of HLA genes was observed.

SB was found to induce looping-out of the MHC. Treatment with SB resulted in an increase in the number of external loops across the entire MHC locus. External chromatin loops increased from 11% (n=163) to 22% (n=127) P=0.02, for the class I region; from 11% (n=135) to 32% (n=228) P<0.05, for the class II region; and from 12% (n=125) to 24% (n=109) P=0.02, for the class III region (Fig. 4A,B). Only a small increase of loop induction from 5% (n=108) to 8% (n=109) P=0.44, of CT6s was found in the gene-poor 6p24 region. Treatment of HT1080 with SB showed identical results. A time-course analysis of SB treatment of MRC5 cells showed that external chromatin loops were formed within 10 minutes of the start of treatment (Fig. 4C, n=232, P=0.01). Furthermore, treatment with IFNγ and SB together did not enhance the number of external chromatin loops (Fig. 4D; P>0.07), suggesting that once external loops have formed SB cannot induce additional higher-order chromatin modifications.

Chromatin decondensation has been associated with induction of transcription in both transgene arrays and gene clusters (Carpenter et al., 2005; Sproul et al., 2005). To examine whether IFNγ induces alterations in chromatin decondensation in MRC5 cells, interphase distances were measured from a series of six probe pairs in the MHC separated by genomic distances ranging from 0.63-3.4 Mb (see supplementary material Fig. S3). The mean values of the interphase distances were in the range of 0.9-1.94 μm in untreated cells and 1.21-2.52 μm in IFNγ-treated cells (see supplementary material Table S2). IFNγ treatment led to an increase in interphase distances in the MHC of up to 25%, indicating that the chromatin becomes decondensed, as observed previously for part of the MHC class II region (Müller et al., 2004). No statistically significant increase was observed for probe pairs in the control gene-poor 6p24 region, values were 0.52 μm and 0.54 μm (P>0.05) for untreated and IFNγ treated cells, respectively. We then measured the interphase distances in the genomic regions flanking the MHC, using probe pairs with genomic separations up to ∼3Mb. No significant change was observed with IFNγ treatment (supplementary material Table S2), indicating that decondensation induced by IFNγ treatment is limited to the MHC.

A linear relationship was found in the MHC between the mean square interphase distance and genomic distance between probes tested for both IFNγ-treated and -untreated cells (Fig. 5A). The slope of the regression line for IFNγ-treated cells (2.26 μm²/Mb) was found to be ∼1.5 times greater than for the untreated cells (1.50 μm²/Mb), demonstrating that the chromatin in the MHC becomes decondensed after IFNγ treatment. Evaluating the shape of the statistical distribution measured for each probe pair gave us ratios for the standard deviation to its mean (s.d.: mean) and for the median to its mean (median: mean), close to the ideal values of a Raleigh distribution before and after IFNγ treatment (data not shown). These findings demonstrate that chromatin behaves as a random polymer (van den Engh et al., 1992).

The distance measurements described above were taken irrespective of their position relative to the chromosome territory, i.e. we did not discriminate whether a probe signal was located on an external loop. Therefore, we compared chromatin within the painted chromosome domain and within loops. In untreated cells, comparison of the slope of the regression line for probe pairs within a painted chromosome domain (1.23 μm²/Mb), with that for probe pairs on an external chromatin loop (3.49 μm²/Mb) reveals a 2.7 times increase in decondensation. Comparison of the slope of the regression line for probe pairs within a painted chromosome domain before (1.23 μm²/Mb) and after IFNγ treatment (1.88 μm²/Mb) revealed a moderate 1.5 times increase in decondensation (Fig. 5B). No significant difference was found in the slope of the regression line for probe pairs on an external chromatin loop before (3.26 μm²/Mb) and after IFNγ treatment (3.49 μm²/Mb) (Fig. 5C). Since the level of chromatin condensation in external loops was relatively unchanged after IFNγ treatment, which suggests that once an external loop has formed no further decondensation can occur.

The MHC is a large gene cluster whose physiological activation by IFNγ is a crucial step in the cell-mediated immune response. The classical HLA class I and class II genes are highly inducible by IFNγ in almost all cell types, yet the mechanism by which this occurs is different for these two sets of genes. Nevertheless, within a few minutes of IFNγ treatment, chromatin carrying the entire MHC undergoes massive higher-order chromatin remodelling (Volpi et al., 2000). Here, we assess the role of transcriptional activators and chromatin remodelling factors in chromatin organization across the MHC, taking advantage of well-characterised STAT1-deficient cell lines. We integrate FISH, to visualise chromosome architecture, with ChIP, to examine the ordered assembly of the transcription machinery and the acetylation changes at the early time points of transcriptional activation by IFNγ. Our findings demonstrate that the higher-order chromatin remodelling observed over the entire MHC in response to IFNγ is mediated by STAT1, and that it reflects decondensation of the chromatin.

IFNγ treatment induced visible chromatin changes of the MHC in HT1080 fibrosarcoma cells with the same dynamics as found previously in fibroblasts (Volpi et al., 2000). The basal level (∼10%) of external loops carrying the MHC in cells that do not express the HLA class II genes rises to 35% after IFNγ induction. Even within a population of cells that is expressing the locus at a very high level we still detect a maximum of 35%. Hence, external loops might not be an absolute requirement for transcription per se, but may enhance transcriptional competence of the locus. They might also reflect the intermittent transcription of a locus at any given moment in time. The significance of these observations is likely to be understood only when techniques are developed to visualise higher-order chromatin alterations in live cells. Remarkably, higher-order chromatin remodelling was not detected in U3A, the STAT1-null derivative of HT1080. Complementation of U3A cells with wild-type STAT1 restores chromatin remodelling, strongly supporting a role of STAT1 in chromatin decondensation of the locus. Furthermore, a single point mutation in STAT1 that prevents phosphorylation also abolishes higher-order chromatin remodelling, providing a clear indication that the cytokine signal is indeed transmitted through the JAK/STAT signalling cascade. In HT1080 cells treated with IFNγ, P-STAT1 binds to the promoters of the primary response genes TAP1 and HSPA1 in the MHC, and IRF1 on chromosome 5. Recent genome-wide profiling of STAT1 recruitment in HeLa cells reveals high-affinity binding to predicted target sites located predominantly at the promoter and enhancer regions (Heintzman et al., 2007; Robertson et al., 2007). Our data are the first to demonstrate a direct role of the JAK/STAT signalling pathway in altering higher-order chromatin architecture in mammalian cells. A role for JAK/STAT signalling in chromatin architecture has recently been reported in Drosophila, where overactivation of the pathway disrupts the stability of heterochromatin leading to transcription (Shi et al., 2006).

The STAT family of transcriptional activators is implicated in the regulation of a variety of cellular processes far beyond the IFNγ response (Levy and Darnell, 2002). Cell-type-specific responses are achieved in cooperation with a variety of transcription factors, co-activator proteins and chromatin-remodelling complexes (reviewed in Platanias, 2005). In response to IFNγ, the SWI/SNF chromatin remodelling component BRG1 is recruited to TAP1 and IRF1 around the same time as P-STAT1. IFNγ-dependent BRG1 recruitment does not occur in STAT1-null cells, suggesting that the sequence-specific binding of STAT1 to the GAS elements brings BRG1 to the TAP1 and IRF1 promoters in response to IFNγ. BRG1 and STAT1 are reported to cooperate during induction of IFNγ-responsive genes (Ni et al., 2005). We suggest that STAT1 and BRG1 interact with each other, as shown for STAT2 and BRG1 (Huang et al., 2002). A possible role for BRG1 in higher-order chromatin unfolding of the MHC is suggested by our finding that complementation of the BRG1-BRM-deficient cell line SW13 with BRG1 restores higher-order chromatin remodelling across the MHC locus in response to IFNγ. The activator-dependent binding of BRG1 does not exclude further recruitment of SWI/SNF complexes by acetylated histones or HATs at later stages of MHC induction. Although the role of ATP-dependent chromatin-remodelling complexes in modulating higher-order chromatin structure is still not defined, components of these complexes have been shown to be involved in regulating chromatin structure and gene expression over large distances in T-cell differentiation (Yasui et al., 2002).

Comparison of the kinetics of P-STAT1, BRG1 and RNAP II recruitment reveals a slight delay of RNAP II binding to the target promoters. Since there is no evidence for a direct interaction between STAT1 and RNAP II subunits, this delay might be explained by an intermediate interaction of P-STAT1 with the TRAP-mediator complex (Zakharova et al., 2003) and/or CBP, which tethers RNAP II to the promoters of STAT1 activated genes. The amount of RNAP II increases further with time after induction despite the decrease in P-STAT1. This finding indicates that although the initial recruitment of RNAP II is STAT1-dependent, it is probably maintained at the promoters by other factors. The presence of TFIIB at the promoters of the classical HLA genes before they are expressed might reflect the presence of a preinitiation complex that helps retain an open chromatin conformation even without transcription.

The higher-order chromatin changes in response to IFNγ are followed by an increase of histone H3 acetylation at the promoters of primary response genes as well as HLA-DRA, which is only expressed several hours later. Interestingly, we find that inhibition of deacetylation by SB in the absence of IFNγ induces external chromatin loops carrying the MHC. The inducible genes within the MHC have high basal levels of histone acetylation and, hence, inhibition of histone deacetylase inhibitors results in hyperacetylation, which might maintain an open chromatin structure. Similar chromatin conformational changes and an increase in acetylated histone H3 and dimethylated H3 have been found in other cell types upon treatment with HDAC inhibitors (Bartova et al., 2005). As the developmentally regulated Hox genes do not show chromatin decondensation in response to the deacetylase inhibitor TSA (Chambeyron and Bickmore, 2004), inducible and developmentally regulated genes may have different levels of basal histone acetylation.

The timing of recruitment of transcription factors and chromatin modifying activities to the MHC, as observed here by ChIP, agrees well with immuno-fluorescence studies of VP16-induced transcriptional activation and decondensation of artificial lac-operator arrays (Carpenter et al., 2005). Local chromatin changes induced by IFNγ at specific positions within the MHC, a natural array of coordinately regulated genes, appear to be multiplied rapidly throughout the locus leading to an altered chromatin configuration. Using distance measurements between probe pairs in the MHC, we show here a linear correlation between interphase distance and genomic separation over a range of 3.4 Mb. Our measurements provide quantitative evidence that IFNγ leads to decondensation of the MHC region both within the chromosome territory and, even more so, when the locus is present on an external chromatin loop where maximum decondensation is achieved. We have demonstrated previously that the MHC class II region and other gene-rich, transcriptionally active regions are closely associated with PML nuclear bodies (Wang et al., 2004). It remains to be determined whether decondensed chromatin carrying the MHC is directed towards transcription factories (Cook, 2002; Osborne et al., 2004) or towards other co-regulated genes, as shown for genes involved in T-cell differentiation (Spilianakis and Flavell, 2004).

In conclusion, our findings suggest that large-scale chromatin remodelling represents an important early step in transcriptional activation of the MHC (Fig. 6). IFNγ activates the JAK-STAT cascade, which signals to the chromatin and RNAP II machinery. Transcriptional upregulation of primary response genes in the MHC coincides with remodelling of the entire locus within minutes of IFNγ treatment. It is striking that the entire MHC locus becomes decondensed, including the classical HLA genes, which are not direct targets for STAT1. Although it is unclear how the higher-order chromatin modification spreads so rapidly across the MHC, it seems possible that the acquisition of a decondensed chromatin state generates a transcriptionally permissive environment for the subsequent HLA class II response. The human transcriptome map reveals a clustering of highly expressed genes in specific chromosomal regions, suggesting that this arrangement is important for the regulation of certain genes (Caron et al., 2001). The clustering of MHC genes has been suggested to provide a selective advantage in a number of respects, including the co-inheritance of advantageous haplotypes (Trowsdale, 2002). We propose that, because the entire locus is subject to a rapid higher-order chromatin modification in response to STAT1, clustering is also beneficial in facilitating an efficient immune response to infection.

The human fibrosarcoma cell line HT1080, its STAT1-null derivative U3A, and complemented U3A cells, U3A/STAT1 and U3A/STAT1^(Y701F), were grown as described previously (Muller et al., 1993). Normal embryonic fibroblast MRC5 cells (CCL-171) and a BRG1-BRM-deficient cell line SW13 (CCL-105) derived from the human small-cell carcinoma of the adrenal cortex were obtained from the ATCC and cultured in Dulbecco's modified Eagle's medium supplemented with 10% foetal calf serum, 2 mM L-glutamine, 50 U/ml penicillin and 50 mg/ml streptomycin, at 37°C in a 10% CO₂ atmosphere.

Cells were grown to 70% confluence, and 200 IU/ml of IFNγ (recombinant human IFNγ, R&D Systems) added to the culture medium for times between 10 minutes and 24 hours. Untreated cells from the same culture were used as a control. Cells were exposed to 10 mM SB (Sigma-Aldrich) for up to 1 hour prior to harvest. For combined treatments, MRC5 cells were treated with SB for 1 hour and IFNγ was added for the final 10 minutes for the first time point. For the final time point, cells were treated with IFNγ for 24 hours and SB added for the last hour. Cell lines were harvested by standard techniques using methanol-acetic acid fixation to produce nuclear preparations for fluorescence in situ hybridisation (FISH).

Retroviral vectors pBabe-IRESpuro, pBabe-hBRG1-IRESpuro and pBabe-hBRM-IRESpuro were obtained from Hideo Iba (Mizutani et al., 2002). Stable populations of transfected human SW13 cells were generated by retroviral infection as described previously (Costa-Pereira et al., 2005). U3A cells were transfected with 20 μg CIITA (pEBS-CIITA form III, obtained from Walter Reith, Université de Genève, Geneva, Switzerland) using GeneJuise transfection protocol (Novagen). Stable transformants were selected with 200 mg/ml hygromycin B. Western blot analysis was performed with 10 μg nuclear extract using a standard protocol with anti-BRG1 (Upstate), anti-β-Actin (Sigma) and anti-CIITA (Abcam) antibodies.

FISH was performed using standard methods (Volpi et al., 2000). Preparations were examined with a Zeiss Axiophot microscope equipped for epifluorescence using a Zeiss plan-neofluar 100× objective and an optivar set at 1.25× (for chromatin conformation analysis) or 2× (for chromatin condensation analysis). Separate grey-scale images were recorded with a cooled CCD-camera (Photometrics). They were then pseudocoloured and merged. SmartCapture 2.1.1 software (Digital Scientific, Cambridge, UK) was used for image analysis and processing. A binomial test was performed for statistical analysis using R software. P<0.05 was considered to be significant.

For analysing the conformation of the chromatin fibre carrying the MHC, the PAC probe RP1-172K2 for the MHC class II gene HLA-DRA was co-hybridised with FITC-labelled chromosome 6 paint (Cambio) to fixed nuclear preparations and detected with rhodamine-conjugated anti-digoxigenin (Vector). For analysing the response of the chromatin fibre carrying the MHC to sodium butyrate (SB), the class I cosmids P1454 and C0426 (Goldsworthy et al., 1996) and the class III cosmid K101 (Kendall et al., 1990) were also used. Nuclei were counterstained with DAPI (200 ng/ml) and mounted in Cityfluor antifade solution.

For analysing chromatin condensation, the distances between pairs of probes within and outside the MHC were measured in randomly selected interphase nuclei using the programme ImageJ (http://rsb.nih.gov) (Yokota et al., 1995). One-hundred measurements were taken for each pair tested. Correlation analysis, linear regression analysis and other statistical evaluations were performed with Microsoft Excel.

ChIP experiments were performed as described (Christova and Oelgeschlager, 2002) with the following antibodies against P-STAT1 (Tyr701) from Cell Signalling, RNA polymerase II (clone 8WG16) from Covance, TFIIB (C-18) from Santa Cruz, Brg-1 (H-88) from Santa Cruz and acetyl-histone H3 and Histone H4 from Upstate. Real-time (RT)-PCR was performed with SybrGreen master mix from Sigma on an MJ Chromo 4 Robocycler (Bio-Rad) with immunoprecipitated samples and corresponding input genomic DNA. The amounts of immunoprecipitated material were normalised to the relevant genomic DNA to allow direct comparison between different antibodies. The fold enrichment was calculated relative to the non-induced levels and more than twofold enrichment counted as significant. GAPDH promoter sequences were amplified from the same material as a control for the IP in each sample and results were corrected accordingly. Primer sequences are shown in supplementary material Table S1.

We thank Walter Reith for the CIITA expression vector, Hideo Iba and C. Muchardt for the BRG1 and BRM expression vectors, Gavin Kelly for statistical evaluation of results, and Facundo Batista, Julie Cooper, Stephan Beck, Alistair Newall and Petra Gross for critical discussions. We thank the reviewers for their insightful and helpful comments. A.B. was supported in part by a Marie Curie Research Fellowship (QLG1-CT-2002-51704). This work was supported by Cancer Research UK.