Interleukin‐33 (IL‐33) is an IL‐1‐like ligand for the ST2 receptor that stimulates the production of Th2‐associated cytokines. Recently, we showed that IL‐33 is a chromatin‐associated factor in the nucleus of endothelial cells in vivo. Here, we report the identification of a short IL‐33 chromatin‐binding peptide that shares striking similarities with a motif found in Kaposi sarcoma herpesvirus LANA (latency‐associated nuclear antigen), which is responsible for the attachment of viral genomes to mitotic chromosomes. Similar to LANA, the IL‐33 peptide docks into the acidic pocket formed by the H2A–H2B dimer at the nucleosomal surface and regulates chromatin compaction by promoting nucleosome–nucleosome interactions. Taken together, our data provide important new insights into the nuclear roles of IL‐33, and show a unique example of molecular mimicry of a chromatin‐associated cytokine by a DNA tumour virus. In addition, the data provide, to the best of our knowledge, the first demonstration of the existence of non‐histone cellular factors that bind to the acidic pocket of the nucleosome.
Interleukin‐33 (IL‐33; previously known as NF‐HEV) is the most recent addition to the IL‐1 family (Baekkevold et al, 2003; Schmitz et al, 2005; Carriere et al, 2007). It has been shown to function as a ligand for the ST2 receptor that stimulates the production of Th2‐associated cytokines in mast cells and Th2 lymphocytes (Schmitz et al, 2005; Ali et al, 2007; Allakhverdi et al, 2007), and protects the heart in response to cardiac stress and atherosclerosis (Sanada et al, 2007; Miller et al, 2008). Recently, we discovered that IL‐33 is an abundant chromatin‐associated factor in the nucleus of endothelial cells in vivo (Carriere et al, 2007). In addition, we showed that IL‐33 has transcriptional regulatory properties (Carriere et al, 2007). This suggested that IL‐33 is a dual‐function protein that might act both as a cytokine and as an intracellular nuclear factor. As a chromatin‐associated cytokine, IL‐33 is similar to HMGB1 (high‐mobility group box1), an architectural chromatin‐binding nuclear factor that functions extracellularly as a cytokine when released by necrotic cells or when secreted by activated macrophages during inflammation (Wang et al, 1999; Scaffidi et al, 2002).
Although the association of IL‐33 with chromatin has been shown previously, its nuclear partners and mechanisms of targeting have not yet been described. Here, we show that IL‐33 tethers to chromatin by docking into the acidic pocket formed by the histone H2A–H2B dimer at the surface of the nucleosome. A similar mechanism has previously been shown to be used by Kaposi sarcoma herpesvirus (KSHV) for attachment of viral genomes to mitotic chromosomes (Barbera et al, 2006). Our results indicate that the virus pirated the chromatin‐binding motif (CBM) of IL‐33 for the establishment of latent infection in human cells.
Identification of a short IL‐33 CBM
Previously, we identified a chromatin and mitotic chromosome association domain at the amino terminus of IL‐33 (aa 1–65) that is not conserved in other IL‐1 family members (Carriere et al, 2007). Further deletion mutagenesis showed that IL‐33 residues 40–58 are sufficient for tethering N‐ or carboxy‐terminal green fluorescent protein (GFP) to mitotic (Fig 1A) or interphase (Fig 1B) chromatin in living human embryonic kidney (HEK) 293T cells. The individual residues essential for chromosome association, within the IL‐33 CBM (aa 40–58), were then identified by alanine scanning mutagenesis (Fig 1C). Six residues were required for binding to mitotic chromatin: human IL‐33 residues M45, L47, R48, S49, G50 and I53. A triple alanine mutation within the IL‐33 CBM, IL‐3340–58(47AAA49), also abrogated chromosome association (data not shown). The results were confirmed in the context of full‐length IL‐33 (IL‐33FL). GFP‐IL‐33FL and GFP‐IL‐33FL(F44A) were found to associate with mitotic chromatin but not the GFP‐IL‐33FL(47AAA49) and GFP‐IL‐33FL(R48A) mutants (Fig 1D). We conclude that the panel of mutations introduced within the IL‐33 CBM has the same effects in the context of full‐length IL‐33 and IL‐3340−58 peptide.
Viral mimicry of the IL‐33 CBM by KSHV
The IL‐33 CBM was found to be evolutionarily conserved in murine and canine IL‐33 orthologues (Fig 2A,B). Sequence alignment of the IL‐33 CBM peptides (Fig 2C) showed striking homologies to the N‐terminal CBM (aa 5–14) of KSHV LANA (latency‐associated nuclear antigen; Piolot et al, 2001; Wong et al, 2004; Barbera et al, 2006), which also binds to mitotic chromosomes in human and mouse cells (Fig 2A,B). On the basis of the sequence alignment with LANA CBM, the minimal IL‐33 CBM peptide was reduced further to amino acids 44–53 (Fig 2D). Interestingly, the conserved residues of the MXLRSG motif (Fig 2C) have been found to be individually required for mitotic chromosome association of both IL‐33 (Fig 1C) and LANA (Wong et al, 2004; Barbera et al, 2006). This suggested that, similar to LANA, IL‐33 might tether to chromatin through the histone H2A–H2B dimer. Thus we examined the ability of IL‐33 CBM and single‐point mutants to associate with H2A and H2B in vivo, using LANA CBM as a control. GFP‐IL‐3340–58, GFP‐LANA5–22 and GFP‐IL‐3340–58(F44A), which associate with chromosomes, precipitated H2A and H2B, whereas GFP and GFP‐IL‐3340–58(R48A), which do not associate with chromosomes, did not (Fig 2E). Next, we analysed the interaction of IL‐33 CBM with H2A and H2B in vitro by using acid‐extracted histones from HeLa cells. Glutathione S‐transferase (GST)‐IL‐3340–58 precipitated H2A and H2B, similar to GST‐LANA5–22, whereas GST did not (Fig 2F). By contrast, H2A and H2B were not precipitated by GST‐IL‐3340–58(47AAA49) or GST‐IL‐3340–58(R48A) fusion protein (Fig 2G). Essentially, identical results were obtained with purified or recombinant H2A–H2B dimer in vitro (supplementary Fig S1 online) and with full‐length IL‐33 in vivo (supplementary Fig S2 online). Taken together, these findings show a tight link between chromatin association of IL‐33 CBM and its binding to H2A–H2B in vitro and in vivo, providing strong evidence that the H2A–H2B dimer mediates attachment of IL‐33 to chromatin.
The IL‐33 CBM recognizes the H2A–H2B acidic pocket
Next, we performed extensive molecular modelling studies of the IL‐33–nucleosome complex. Modelling of IL‐33 CBM on the basis of the X‐ray structure of the LANA CBM complexed with the nucleosome core particle (Barbera et al, 2006) suggested that the IL‐33 CBM can adopt a tight hairpin structure, stabilized by an array of intramolecular hydrogen bonds similar to that of the LANA CBM (Fig 3A,B). Remarkably, L51 and I53 in IL‐33 replace LANA R12 and T14, respectively, which gives rise to a much denser network of intramolecular hydrophobic interactions (supplementary Fig S3 online). The strict conservation of the MXLRSG motif in the IL‐33 and LANA CBM hairpins suggested that the IL‐33 CBM might dock into the same cavity as the LANA CBM on the nucleosomal surface (Barbera et al, 2006). Molecular modelling showed excellent shape and charge complementarity between the IL‐33 CBM and the nucleosomal surface (Fig 3C,D), and the presence of bulkier residues at the N and C termini of IL‐33 CBM could be accommodated without steric clash. Similar to the LANA CBM, the IL‐33 CBM is predicted to dock into the negatively charged acidic pocket formed by the H2A–H2B dimer on the nucleosomal surface (Fig 3E,F). The model suggests that hairpin binding is mediated by hydrogen bonds between IL‐33 residues R48–S49 and negatively charged residues of the H2A–H2B acidic pocket (H2A residues E61, E64, D90; Fig 3G,H), and by stacking interactions involving the MXL submotif and H2A Y57 residue (supplementary Fig S3 online). To validate the molecular modelling results experimentally, we used histones carrying mutations in the putative interaction domain (Fig 4A). We found that IL‐33 does not bind to H2A carrying mutations in the acidic pocket and binds only very poorly to histone variant H2A.Bbd (Burr body deficient), which has a divergent pocket sequence (Fig 4B). By contrast, IL‐33 associated with histone variant H2A.Z, which has an extended acidic patch. Taken together, these data indicate that, similar to the LANA CBM, IL‐33 CBM recognizes the acidic pocket formed by the H2A–H2B dimer at the surface of the nucleosome.
The IL‐33 CBM regulates chromatin compaction
The recent demonstration of the crucial role of the H2A–H2B acidic pocket in chromatin compaction (Chodaparambil et al, 2007; Zhou et al, 2007) suggested that the IL‐33 CBM might influence the higher order structure of chromatin. To test this hypothesis, we used a previously established in vitro chromatin model system based on the 208‐12 DNA template (Simpson et al, 1985; Schwarz et al, 1996; Chodaparambil et al, 2007; Zhou et al, 2007), which consists of 12 repeats of a 208‐bp nucleosome positioning sequence (Fig 5A). Nucleosomal arrays were assembled and chromatin reconstitution was verified by digestion with micrococcal nuclease (Fig 5B). Chromatin compaction was then assayed by measuring the abundance of monomeric arrays with increasing divalent cation (MgCl2) concentrations known to promote chromatin condensation (Schwarz et al, 1996; Chodaparambil et al, 2007). The Mg50 value has previously been defined as the concentration of MgCl2 at which 50% of nucleosomal arrays oligomerize into large, condensed higher order chromatin structures that are easily pelleted by microcentrifugation (Chodaparambil et al, 2007). We found that addition of the IL‐33 CBM peptide lowered the Mg50 value needed for chromatin condensation (Mg50<1 mM), whereas a control peptide (IL‐33 cont) had no effect compared with control arrays (Mg50 ∼3 mM; Fig 5C; supplementary Fig S4 online). Interestingly, the LANA CBM had the same effect in this assay as that of the IL‐33 CBM, as recently reported (Chodaparambil et al, 2007). On the basis of these findings, we conclude that the IL‐33 CBM regulates chromatin compaction by promoting oligomerization of nucleosomal arrays into the higher order structures of chromatin.
The LANA CBM has been shown to alter nuclear architecture and to influence chromatin condensation in vivo (Chodaparambil et al, 2007). Interestingly, when tested in the same cellular assay in U2OS cells, the IL‐33 CBM provoked similar changes in chromatin organization in the nucleus (Fig 5D,E). More than 85% of the cells expressing high levels of GFP‐IL‐3340–58 or GFP‐LANA5–22 showed nuclei with large regions of Hoechst exclusion. These cellular effects were not observed with the IL‐33 CBM mutant GFP‐IL‐3340–58(47AAA49) or with GFP alone. Therefore, similar to the LANA CBM, the IL‐33 CBM alters nuclear architecture in vivo.
Previously, we have shown that IL‐33 has transcriptional repressor activity when tethered to a promoter by a heterologous Gal4‐DNA‐binding domain (Carriere et al, 2007). Interestingly, we found that a single‐point or triple mutation within the IL‐33 CBM greatly reduced the transcriptional repressor activity of IL‐33 (Fig 5F,G). These results indicate that binding to the acidic pocket of H2A–H2B is important for the transcriptional regulatory function of IL‐33.
Finally, we examined the possibility that IL‐33 might share transcriptional targets with LANA. However, we found no change in the expression of principal LANA target genes after either IL‐33 knockdown or overexpression (supplementary Fig S5 online), suggesting that IL‐33 mimics certain functions of LANA (targeting to chromatin through H2A–H2B acidic patch and modulation of chromatin structure) but not all.
Here, we have identified the mechanism responsible for IL‐33 attachment to chromatin. We showed that IL‐33, similar to KSHV LANA, uses a short motif containing the crucial hexapeptide MXLRSG for docking into the acidic pocket formed by the H2A–H2B dimer at the surface of the nucleosome. Although many factors have been shown to recognize the flexible histone tails, IL‐33 is the first cellular factor shown to bind to the acidic pocket of nucleosome, a region that is important for chromatin compaction (Chodaparambil et al, 2007; Zhou et al, 2007). This suggested that docking of IL‐33 into the acidic pocket of H2A–H2B might modulate chromatin structure. Accordingly, we found that the IL‐33 CBM alters nuclear architecture in vivo and regulates chromatin compaction in vitro by enhancing the self‐association and oligomerization of the nucleosome into higher order chromatin structures. Similar effects on chromatin structure were recently reported for the LANA peptide (Chodaparambil et al, 2007). Interestingly, the acidic pocket of the nucleosome has been shown to couple chromatin compaction with transcriptional repression (Zhou et al, 2007), and this might explain our observation that a single‐point or triple mutation within the IL‐33 CBM that abrogates interaction with the H2A–H2B acidic patch greatly reduced the transcriptional repressor properties of IL‐33.
Our data show that the IL‐33 CBM is conserved in murine and canine IL‐33 orthologues. By contrast, the motif is found in KSHV and RFHV (retroperitoneal fibromatosis herpesvirus) LANA but not in LANA equivalents from other primate and murine rhadinoviruses such as herpesvirus saimiri and MHV68 (murine herpesvirus 68). This suggests that the KSHV LANA CBM has been pirated from cellular IL‐33. To the best of our knowledge, this is the first time a human DNA tumour virus has been shown to mimic a cellular factor for attachment to mitotic chromatin, an essential process for the maintenance of viral genomes in latently infected tumour cells. Although there are previous examples of viral mimicry of cellular cytokines (including mimicry of IL‐6 and CC‐chemokines by KSHV; Moore et al, 1996; Boshoff et al, 1997), the piracy by KSHV of the IL‐33 CBM, rather than the IL‐1‐like domain, is unique. This surprising example of viral mimicry is likely to have been one of the crucial events during KSHV evolution for the establishment of latent viral infections in human cells. Further characterization of the IL‐33/LANA CBM might lead to the identification of new therapeutic agents for the treatment of KSHV‐associated diseases.
Plasmid constructions. IL‐33 deletion mutants were amplified by PCR using the human IL‐33/NF‐HEV cDNA (NM_033439) as a template. IL‐33‐CBM and LANA‐CBM constructs, and IL‐33‐CBM alanine scanning point mutants were generated by cloning linker oligonucleotides into the EcoRI and BamHI sites of plasmid pEGFP.C2 (Clontech, Mountain View, CA, USA).
Fluorescence microscopy. For live cell imaging, HEK293T cells were transfected with GFP fusion protein expression vectors using the calcium phosphate procedure. At 2 days after transfection, DNA was counterstained with Hoechst 33342, and living cells were observed by fluorescence microscopy on an inverted fluorescence microscope equipped with a digital camera (Eclipse TE300; Nikon, Tokyo, Japan).
GST pull‐down and immunoprecipitation assays. Details on the in vitro (GST pull‐down) and in vivo (co‐immunoprecipitation) binding assays used to analyse association of the IL‐33 CBM with H2A–H2B are described in the supplementary information online.
Nucleosomal array reconstitution, analysis of oligomerization and transcriptional reporter assay. Nucleosomal arrays were prepared using the 208‐12 DNA template, and their ability to oligomerize at various MgCl2 concentrations was determined using a differential centrifugation assay. The effects of the IL‐33 CBM mutations on the transcriptional repressor activity of Gal‐4‐IL‐33 were determined using a Gal4‐luciferase reporter. The details of the protocols are available in the supplementary information online.
Molecular modelling. Modelling was performed using the Accelrys modules InsightII, Homology, Discover, Docking and Delphi, run on a Silicon Graphics Fuel workstation. The structure of the IL‐33 CBM and IL‐33 CBM–nucleosome complex was modelled on the basis of the structure of LANA CBM in complex with the nucleosome (Protein Data Bank entry code 1ZLA).
Supplementary information is available at EMBO reports online (http://www.emboreports.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
We are grateful to K. Bystricky for providing the 208‐12 DNA template and G. Bouche for help with the in vitro chromatin assays. We thank F. Amalric for critical reading of the manuscript. This study was supported by grants from Ligue Nationale Contre le Cancer (Equipe Labellisée), ANR‐Programme Blanc ‘Cuboïdale’ and MAIN European Network of Excellence (FP6‐502935).
- Copyright © 2008 European Molecular Biology Organization