|Home | About | Journals | Submit | Contact Us | Français|
Cellular senescence is an irreversible proliferation arrest, tumor suppression process and likely contributor to tissue aging. Senescence is often characterized by domains of facultative heterochromatin, called senescence-associated heterochromatin foci (SAHF), which repress expression of proliferation-promoting genes. Given its likely contribution to tumor suppression and tissue aging, it is essential to identify all components of the SAHF assembly pathway. Formation of SAHF in human cells is driven by a complex of histone chaperones, namely, HIRA and ASF1a. In yeast, the complex orthologous to HIRA/ASF1a contains two additional proteins, Hpc2p and Hir3p. Using a sophisticated approach to search for remote orthologs conserved in multiple species through evolution, we identified the HIRA-associated proteins, UBN1 and UBN2, as candidate human orthologs of Hpc2p. We show that the Hpc2-related domain of UBN1, UBN2, and Hpc2p is an evolutionarily conserved HIRA/Hir-binding domain, which directly interacts with the N-terminal WD repeats of HIRA/Hir. UBN1 binds to proliferation-promoting genes that are repressed by SAHF and associates with histone methyltransferase activity that methylates lysine 9 of histone H3, a site that is methylated in SAHF. UBN1 is indispensable for formation of SAHF. We conclude that UBN1 is an ortholog of yeast Hpc2p and a novel regulator of senescence.
Cell senescence is an irreversible proliferation arrest that is triggered by activated oncogenes and, consequently, is an important tumor suppression process in vivo (7, 8, 10, 11, 13, 17, 33, 49, 52, 56, 62, 65). Senescence is also caused by telomere attrition that results from repeated rounds of cell division (8, 45). Consequently, senescence in vivo is thought to contribute to tissue aging, through exhaustion of renewable tissue stem cell populations (8, 28, 30, 34). Other triggers of senescence include cellular stresses, such as oxidative stress, and inadequate growth conditions (8, 45).
One cellular characteristic of senescence in many human cell types is formation of specialized domains of facultative heterochromatin called senescence-associated heterochromatin foci (SAHF) (15, 38, 71). SAHF result from condensation of individual chromosomes into isolated heterochromatic domains (19, 69). SAHF are thought to result in repression of proliferation-promoting genes, thereby contributing to senescence-associated cell cycle arrest. SAHF contain several molecular indicators of transcriptionally silent heterochromatin, including heterochromatin proteins 1 (HP1α, -β, and -γ) and histone variant macroH2A. In addition, SAHF contain increased amounts of HMGA proteins (19, 37).
Two chromatin regulators, HIRA and ASF1a, drive formation of SAHF in human cells (71). HIRA and ASF1a are the human orthologs of proteins that create transcriptionally silent heterochromatin in yeasts, flies, plants, and mammals (20, 21, 35, 41, 53-55, 61). HIRA is a histone chaperone that preferentially deposits the histone replacement variant H3.3 in nucleosomes (32, 36, 58, 61). Likewise, yeast Asf1p has histone deposition activity (60). Consistent with their overlapping properties, yeast Asf1p and Hir proteins physically interact, and this interaction is necessary for telomeric silencing (14). Likewise, formation of SAHF in human cells depends upon a trimeric complex of HIRA, ASF1a, and histone H3 (58, 59, 69, 71), most likely in part due to the ability of this complex to facilitate nucleosome assembly and increased nucleosome density. In addition, the HIRA and ASF1a proteins and orthologs in other species have other cellular and molecular functions. For example, the HIRA/ASF1a complex is thought to mediate transcription-coupled deposition of histone H3.3 (26, 39, 43, 46, 58), and ASF1 proteins play a variety of roles in DNA replication, transcription, and DNA repair-coupled nucleosome disassembly and assembly and histone modification (47). To generalize, ASF1 proteins appear to be histone-binding proteins with diverse functions in histone and chromatin metabolism, and their functional specification is, in part, achieved through their interaction with more role-specific partners, such as HIRA or the DNA replication histone chaperone complex, CAF-1 (58, 59).
Formation of SAHF by HIRA and ASF1a depends upon prior localization of HIRA to acute promyelocytic leukemia (PML) nuclear bodies (67, 71), subnuclear organelles enriched in PML and many other proteins (48). PML bodies have been previously shown to play a role in onset of cell senescence (16, 18, 40). At a molecular level, PML bodies are thought to serve as sites of assembly of macromolecular regulatory complexes and/or protein modification (48). Thus, it seems likely that PML bodies are a molecular “staging ground” where HIRA-containing complexes are assembled or modified prior to their translocation to chromatin and formation of SAHF. Recently, we showed that recruitment of HIRA to PML bodies and formation of SAHF are initiated by repression of Wnt signaling in presenescent cells (66), thereby establishing a direct link between senescence and Wnt signaling, two important determinants of tissue homeostasis and tumor progression (9, 31). Another observation further underscores the impact of this pathway on tumor progression. Specifically, inactivation of HMGA proteins, components of mature SAHF, abrogates SAHF formation and facilitates cell transformation and tumor formation (37). Formation of SAHF might also contribute to tissue aging. Fibroblasts from skin of aging baboons exhibit several molecular markers of cellular senescence (27), and expression of HIRA in these cells shows a striking increase with age (29). These links to cancer and aging make it critical to define all the components of this senescence-associated chromatin-remodeling pathway.
In yeast, the human orthologs of HIRA and ASF1a are found in a larger complex, comprised of Hir1p, Hir2p, Asf1p, Hir3p, and Hpc2p. These five proteins copurify as a complex, and yeast cells lacking individual members of the complex have similar phenotypes (22, 44, 54, 57, 64). Hir1p and Hir2p are both orthologs of HIRA, with Hir1p being homologous to the N terminus and Hir2p being homologous to the C terminus of HIRA. Asf1p is the yeast ortholog of human ASF1a. However, the human orthologs of Hir3p and Hpc2p have gone unrecognized to date. To further our understanding of the SAHF assembly process, we set out to identify a human ortholog of yeast Hpc2p. By searching sequence databases for remote orthologs of Hpc2p in multiple species simultaneously, we identified a HIRA-binding protein, UBN1, as a likely human ortholog of Hpc2p and demonstrated an essential role for this protein in formation of SAHF.
For the identification of mammalian counterparts of Hpc2, we used a script for remote orthology detection (Hugo Ceulemans, unpublished). This script first identifies orthologs as sets of three or more proteins from different species that mutually recognize each other as the best hit in interspecies Blast searches, after masking for low-complexity regions. Starting with the query and following a predefined phylogenetic order, the script then aligns each identified ortholog to the query and its previously aligned orthologs. At each alignment step, only the set of nonoverlapping maximally scoring local (fragment) alignments is kept for the ensuing steps. For each of the resulting multiple sequence alignments of fragments, the script builds a calibrated hidden Markov model with the HMMer package and uses this model to scan for sufficiently significant hits in proteomes of interest. In the case of Hpc2p, one of the constructed hidden Markov models detected one (or in two cases two) significant hits (E < 0.001) in each of 18 studied fungal, animal, plant or protozoan proteomes. The putative human Hpc2 orthologs are UBN1 and a previously uncharacterized related protein (FLJ25778), for which we propose the name UBN2. In Fig. Fig.1A,1A, the groups of similar amino acids (separated by commas) to be used for the similarity calculation are as follows: GAVLI, FYW, CM, ST, KRH, DENQ, P.
WI38 and IMR90 fibroblasts and U2OS cells were grown as described by ATCC. Human mesenchymal progenitor cells (hMPCs) were grown as described previously and according to the supplier's instructions (Lonza) (42).
pBABE-RasV12 was a gift of William Hahn (Dana-Farber Cancer Institute). pQCXIP-HA-UBN1 was made by standard molecular biology procedures; details are available on request. pLKO1 and pLKO1-shUBN1 were obtained from OpenBiosystems (shUBN1 hairpin sequence, CCGGGCCAGCTCAATCTCCAAACATCTCGAGATGTTTGGAGATTGAGCTGGCTTTTT). Small interfering RNA (siRNA) to UBN1 was purchased from Dharmacon (catalog no. L-014195). The following reagents have been described previously: anti-HIRA (24), anti-ASF1a (pAb88 and pAb87), anti-ASF1b (pAb87) (71), and mouse monoclonal anti-UBN1 (3). The mouse monoclonal antibody to UBN1 was a gift of Sirpa Aho and Jouni Uitto (both at Thomas Jefferson University). A rabbit polyclonal antibody to UBN1 (pAbUBN1C) was raised to glutathione S-transferase (GST)-UBN1(692-1134) as described previously (25). This was used after affinity purification on a GST-UBN1(692-1134) column (rabbit anti-mouse immunoglobulin G [Sigma M7023] as a negative control) or after partial purification over melon gel (Pierce) (partially purified antibody from rabbit preimmune serum as a negative control). The anti-H3Ac antibody was a rabbit polyclonal supplied by Upstate (06-599).
In vitro protein-binding assays were performed as described previously (2), using 35S-labeled proteins prepared in vitro from plasmid DNA using the Promega TNT kit and purified recombinant GST fusion proteins.
Extracts from Sf9 cells individually infected or coinfected with baculoviruses to express recombinant His-HIRA(1-405) and His-UBN1(1-175) were prepared by Dounce homogenization in 1× phosphate-buffered saline. His-tagged proteins were recovered from clarified extracts by immobilized metal affinity chromatography (IMAC). Peak fractions of recombinant protein, eluted from the IMAC column with 250 mM imidazole, were pooled and resolved on a Superdex 200 10/30 size exclusion column equilibrated in 1× phosphate-buffered saline.
Retrovirus-mediated gene transfer was performed as described previously (71), using Phoenix cells to make the infectious viruses (Gary Nolan, Stanford University). Cells infected with viruses encoding resistance to puromycin or neomycin were selected in 2 μg/ml and 500 μg/ml, respectively, of the appropriate selection agent. Lentivirus-mediated gene transfer was performed as described in the Virapower kit (Invitrogen), using 293FT cells to make the infectious viruses.
Total RNA was prepared using Trizol (Invitrogen), according to the manufacturer's instructions. Reverse transcription-PCR (RT-PCR) was performed using the Qiagen one-step RT-PCR kit, according to the manufacturer's instructions. Primer sequences are available on request.
UBN1 chromatin immunoprecipitation (ChIP) assays were performed using a standard protocol (70). ASF1a ChIP assays were performed similarly, but using an additional ethylene glycol disuccinate bis(sulfo-N-succinimidyl)ester cross-linking step (68).
For ChIP assays, either affinity purified UBN1 polyclonal antibody at a concentration of 5 μg/immunoprecipitation reaction mixture or melon gel (Pierce) purified antibody at a concentration of 10 μg/immunoprecipitation reaction mixture was used. The PCR primers amplified a 343-bp region of the cyclin A2 gene that includes sequences from exon 1 and intron 1 of the gene (between CGACCGGCGGCTACG and AAAGGGATGCGGGAT).
Histone methyltransferase (HMTase) assays were performed as described previously (50).
To identify candidate human orthologs of yeast Hpc2, we performed a bioinformatics search for Hpc2p-related protein sequences that are conserved through evolution, from yeast to human (see Materials and Methods for details of the method). This search uncovered a short sequence motif of about 50 amino acids that is conserved in at least 18 species from yeast to human. The human genome codes for two proteins with this Hpc2p-related domain (HRD), ubinuclein 1 (UBN1) and the hypothetical protein FLJ25778 (hereafter referred to as UBN2) (Fig. (Fig.1A).1A). UBN1 was previously identified as a protein that binds to the Epstein-Barr virus EB1 protein, a member of the basic-leucine zipper family of transcription factors (3). Significantly, in HeLa cells ectopically expressing epitope-tagged histone variant H3.3, Tagami et al. showed that UBN1 copurifies with a histone H3.3 chaperone complex containing human HIRA and ASF1a, but they did not comment on its potential similarity to Hpc2p (58). This underscores the power of our search approach, which simultaneously scans for low-level but significant conservation in multiple species. Outside of the HRD, UBN1 and UBN2 are not closely related to Hpc2p (Fig. (Fig.1B),1B), but UBN1 and UBN2 do share other regions of similarity to each other. Both UBN1 and UBN2 were expressed at the mRNA level in a range of primary and transformed human cell types (Fig. (Fig.1C).1C). In immunoprecipitation-Western blot analysis, a polyclonal antibody and a mouse monoclonal antibody both recognized a 120-kDa polypeptide that was abolished by siRNA to UBN1 (Fig. (Fig.1D),1D), confirming the specificity of the two antibodies. We used these antibodies to show, for the first time, that UBN1 coimmunoprecipitates with endogenous human HIRA and ASF1a in a panel of primary and transformed human cell types, including primary WI38 and IMR90 fibroblasts, primary hMPCs, and transformed human osteosarcoma cells (Fig. 1E and F). Conversely, HIRA and ASF1a both coimmunoprecipitated with UBN1 (Fig. (Fig.1F).1F). Together with the earlier glycerol gradient sedimentation data on ectopically expressed histone H3.3-containing protein complexes (58), these data show that HIRA, ASF1a, and UBN1 are members of the same protein complex, like yeast Hir1p, Hir2p, Asf1p, and Hpc2p (22, 44).
Consistent with this, we observed that siRNA-mediated knockdown of HIRA decreased the steady-state abundance of UBN1 and vice versa. However, knockdown of either protein did not affect mRNA abundance of the other and did not affect abundance of ASF1a (Fig. (Fig.2A).2A). This result is consistent with a close physical interaction between HIRA and UBN1, where removal of either protein destabilizes the other. Indeed, under conditions where we could detect a stable in vitro interaction between HIRA and UBN1, we were unable to detect an interaction between ASF1a and UBN1 (Fig. (Fig.2B).2B). Previous studies showed that ASF1a interacts directly with the evolutionarily conserved B domain of HIRA (59, 71). Taken together, these results suggest that ASF1a and UBN1 each interact independently with HIRA, consistent with the idea that HIRA forms a scaffold for the ASF1a/HIRA/UBN1 complex.
Since UBN1 and Hpc2p share the ability to bind to HIRA/Hir-containing complexes and UBN1 interacts through HIRA but not ASF1a, we hypothesized that the HRD of UBN1 and Hpc2p is an evolutionarily conserved HIRA/Hir-binding domain. To test this, we proceeded to define the interaction domains between HIRA and UBN1. First, deletion analysis of HIRA showed that in vitro binding of HIRA to UBN1 does not require the ASF1a-binding B domain but does require the N-terminal WD repeats (Fig. (Fig.3A,3A, lanes 3 and 4). A protein lacking the WD repeats still bound to ASF1a (Fig. (Fig.3A,3A, lanes 4 and 9), showing that the failure of this fragment to bind to UBN1 was not due to gross misfolding of the protein. The WD repeats alone were sufficient for binding to UBN1 but not ASF1a (Fig. (Fig.3A,3A, lanes 5 and 10). Next, we defined the region of UBN1 and UBN2 that mediates binding to HIRA. Deletion of the N-terminal region of UBN1 containing the HRD abolished binding to HIRA (Fig. (Fig.3B,3B, lane 4), and this same region of both UBN1 and UBN2 (residues 1 to 166) was sufficient for binding to HIRA (Fig. (Fig.3B,3B, lanes 2 and 3). In sum, these results show that the HRD of UBN1 and UBN2 mediates an interaction with the WD repeats of human HIRA.
Because all the binding studies described above were done using unpurified proteins in cell extracts or rabbit reticulocyte lysates, these results do not indicate whether HIRA and UBN1 directly interact with each other. To test this, a His-tagged version of the WD repeat domain of HIRA [His-HIRA(1-405)] and a His-tagged version of the HRD-containing domain of UBN1 [His-UBN1(1-175)] were coexpressed in insect Sf9 cells using baculovirus. Expressed proteins were recovered by nickel affinity chromatography and then fractionated by gel filtration chromatography. When coexpressed in Sf9 cells, both His-HIRA(1-405) and His-UBN1(1-175) coeluted over the chromatography column. The Coomassie blue-stained gel revealed no other copurifying proteins of comparable abundance (Fig. (Fig.4A4A and data not shown). In contrast, when either protein was expressed alone in Sf9 cells, the proteins were less abundant and eluted in the void volume of the column, consistent with the “free” proteins being unfolded and aggregated (Fig. (Fig.4A).4A). Regardless of the exact state of the singly expressed proteins, the facts that the coexpressed proteins coelute at a different apparent molecular weight than the singly expressed proteins and that no other copurifying equally abundant proteins are detectable very strongly suggest that HIRA and UBN1 directly interact with each other to form a homogeneous complex.
Finally, we asked whether the HRD of yeast Hpc2p mediates the interaction between yeast Hpc2p and yeast Hir1p. Indeed, we found that the HRD of yeast Hpc2p was necessary for binding to the WD repeats of yeast Hir1p (Fig. (Fig.4B).4B). In sum, these results show that the HRD of UBN1, UBN2, and Hpc2p is an evolutionarily conserved domain that interacts with the WD repeats of human HIRA and yeast Hir proteins.
Next, we asked whether the interaction between the UBN1 HRD and the HIRA WD repeats depends on the most conserved residues of the HRD. A series of HRD substitution mutants with mutations in the minimum HIRA-binding domain (residues 1 to 166) of UBN1 was made. These mutants included double- and triple-residue substitution mutants and more extensive substitution mutants in which the sequence NAAIRS was substituted for a longer contiguous tract of amino acids (Fig. (Fig.5A).5A). NAAIRS is thought to be relatively nondisruptive and compatible with alternate secondary structures based on its appearance in both α-helical and β-sheet structures, and it has been successfully used for scanning mutagenesis (51). A NAAIRS substitution mutation (M1) that removed the most N-terminal conserved residues of the UBN1 HRD (Y132 and D133) did not affect binding to HIRA (Fig. (Fig.5B).5B). Likewise, another double substitution (M2) of two other nonconserved residues in this region (L125 and I126) did not affect binding to HIRA (Fig. (Fig.5B).5B). Substitution of the conserved FID sequence (residues 138 to 140) by either EEL (M3) or NAAIRS (M4) gave different results. The former substitution did largely abolish binding to HIRA, whereas the latter did not (Fig. (Fig.5B).5B). This difference may be due to the relatively conservative substitution of A for I139 in the M4 NAAIRS mutant. Regardless, any of several substitutions at the most C-terminal block of conserved residues abolished HIRA binding (Fig. 5B and C). These included a NAAIRS substitution mutation (M5), FYI to EYE (M6), F160E (M7), and I162E (M8). We conclude that this C-terminal block of evolutionarily conserved amino acids is particularly important for interaction with HIRA, consistent with the notion that these residues contribute to a conserved, and therefore physiologically significant, HIRA/UBN1 interface.
To determine whether UBN1 binds to a functionally important part of the HIRA protein, we turned to the Drosophila mutation sesame (ssm). This mutation results in female sterility, due to a defect in replacement of sperm chromosomal protamines by maternally provided histones after fertilization of the egg. Recent reports have demonstrated that this sterility is due to a point mutation, R225K, of the fly HIRA protein between the fourth and fifth repeats of the WD domain (6, 32). Significantly, the fly ssm protein is still stably expressed (32), and the corresponding mutation [HIRA(R227K)] of the human protein does not destabilize the protein and does not affect binding of HIRA to ASF1a (Fig. (Fig.5D5D and data not shown). Thus, the ssm mutation likely defines a functionally important protein-protein interaction surface on the HIRA protein. Since UBN1 binds to the HIRA WD repeats, we asked whether the ssm R227K mutation disrupts the interaction of human HIRA with UBN1. Indeed, in the context of full-length human HIRA, the R227K mutation specifically disrupted the interaction of HIRA with UBN1 (Fig. (Fig.5D).5D). Moreover, both UBN1 and UBN2 failed to bind to a variant of the human HIRA WD fragment that harbors the R227K substitution (Fig. (Fig.5E).5E). Since the fly ortholog of UBN1 (yemanuclein) is specifically expressed in fly oocytes (4), it is tempting to speculate that the ssm female sterility phenotype results from disruption of the dHIRA/yemanuclein interaction, although we obviously cannot exclude the possibility that the ssm mutation also disrupts interactions of HIRA with one or more other proteins. Regardless, UBN1 and UBN2 clearly bind to a functionally significant domain of the HIRA protein.
Having shown that the evolutionarily conserved HRD of UBN1 binds to the functionally important WD domain of HIRA, we next asked whether UBN1 is involved in formation of SAHF in senescent human cells. Since HIRA's localization to PML bodies is a prerequisite for formation of SAHF (67), we asked whether UBN1 is also recruited to SAHF in presenescent cells. We found that, like HIRA, UBN1 is localized throughout the nucleoplasm in proliferating primary human fibroblasts in a fine speckled pattern (Fig. (Fig.6A).6A). However, UBN1 colocalized with both HIRA and PML in PML bodies in senescent cells (Fig. 6A and B). Confirming UBN1's recruitment to PML bodies, in senescent cells UBN1 also colocalized with SP100 (Fig. (Fig.6C),6C), another protein known to be contained in PML bodies (48). Also, ectopically expressed hemagglutinin (HA)-tagged UBN1 was recruited to PML nuclear bodies (Fig. (Fig.6D).6D). We confirmed that the foci observed in these assays are formed by UBN1, by showing that siRNA-mediated knockdown of UBN1 abolished the foci (Fig. (Fig.6E).6E). In sum, these results show that, like its binding partner HIRA, UBN1 is localized to PML bodies only in senescent cells, thus implicating this protein in SAHF assembly together with HIRA.
The HIRA/ASF1a complex is thought to contribute to repression of expression of proliferation-promoting genes, such as the cyclin A gene, by formation of repressive heterochromatin at such genes (1, 38). If UBN1 is involved in this process, then it might physically interact with such genes. We tested this hypothesis by ChIP. An antibody to UBN1 immunoprecipitated the cyclin A2 gene, but a control antibody did not (Fig. (Fig.7A).7A). Reinforcing the specificity of this interaction, the cyclin A2 gene was also immunoprecipitated by an anti-HA antibody, but only in cells ectopically expressing HA-UBN1 (Fig. (Fig.7B).7B). We also confirmed the specificity of the interaction by showing that siRNA- or shRNA-mediated knockdown of UBN1 abolished the ChIP signal (Fig. 7C and D). Using a modified ChIP protocol with a longer cross-linker, ethylene glycol disuccinate bis(sulfo-N-succinimidyl) ester, we also detected ASF1a bound to the cyclin A2 gene for the first time (Fig. (Fig.7E).7E). Thus, both UBN1 and ASF1a can bind to the cyclin A2 gene. Formation of SAHF at the cyclin A2 gene is accompanied by methylation of histone H3 lysine 9 (H3K9Me) (38). Thus, if UBN1 is involved in heterochromatinization of this gene and formation of SAHF, it might be associated with histone H3K9 HMTase activity. To test this hypothesis, cell extracts from primary human fibroblasts were immunoprecipitated with an antibody to UBN1 or a control antibody and assayed for HMTase activity toward purified recombinant wild-type histone H3 and toward substitution mutants in which in vivo methylated lysine residues are converted to arginine. HMTase activity was coprecipitated with UBN1 antibody but not with the control antibody (Fig. (Fig.7F).7F). This activity methylated wild-type histone H3 and the K4R and K27R mutants but failed to methylate the histone H3K9R substrate, showing that UBN1 is specifically associated with an HMTase activity targeted toward H3K9 (Fig. (Fig.7F).7F). Together, these data are consistent with a role for UBN1 in formation of transcriptionally silent chromatin at the cyclin A gene, a feature of SAHF.
To directly test a role for UBN1 in formation of SAHF, we asked whether ectopic expression of UBN1 is able to accelerate cell senescence and induce SAHF formation. Primary human fibroblasts were infected with a retrovirus encoding wild-type UBN1 or a control virus and assayed for cellular and molecular hallmarks of senescence, including formation of SAHF. Ectopic expression of UBN1 markedly reduced proliferation of primary human fibroblasts over a 10-day period (Fig. (Fig.8A).8A). This was accompanied by a dramatic increase in the number of cells displaying senescence-associated β-galactosidase activity (Fig. 8B and C), an empirical marker of senescence (17). In addition, cells ectopically expressing UBN1 showed enhanced localization of HIRA to PML bodies and formation of SAHF (Fig. 8D to G). We conclude that ectopic expression of UBN1 activates the HIRA-driven SAHF assembly pathway and accelerates cell senescence.
Next, we tested the consequence of UBN1 knockdown for the SAHF assembly pathway. Specifically, we asked whether knockdown of UBN1 affected formation of SAHF. Primary IMR90 fibroblasts were infected with a retrovirus encoding oncogenic Ras together with a control virus or a lentivirus encoding an shRNA to UBN1. The shRNA to UBN1 knocked down expression of UBN1 but did not affect expression of activated Ras (Fig. (Fig.9A).9A). Knockdown of UBN1 also strikingly impaired formation of SAHF, as judged by DAPI (4′,6′-diamidino-2-phenylindole) staining (Fig. 9B and C). We conclude that UBN1 is required for activation of the SAHF assembly pathway and formation of SAHF that is triggered by an activated oncogene.
We have shown that yeast Hpc2p and human UBN1 and UBN2 are members of a protein family that contain a conserved HRD domain that binds to the WD repeats of HIRA/Hir proteins. In addition, we have demonstrated a role for human UBN1 in a chromatin-remodeling pathway that constitutes an interface between cell senescence and Wnt signaling, two important physiological determinants of cell proliferation, stem cell renewal, and tumor progression (9, 31). Other lines of evidence also implicate this chromatin-remodeling pathway in tumor suppression and tissue aging (27, 29, 37). In sum, UBN1 is an ortholog of yeast Hpc2p and a newly found component of a chromatin-remodeling pathway and is likely important for tumor suppression and tissue aging.
The sophisticated simultaneous search for Hpc2p orthologs in the genomes of multiple species (see Materials and Methods) was crucial to identify UBN1 and UBN2 as Hpc2p orthologs and the HRD as a conserved HIRA-binding domain. The evolutionarily conserved HRD is comprised of two core blocks of conservation, each of 5 to 10 amino acids and separated by 15 to 20 amino acids. Mutational analysis highlighted the role of the C-terminal part of this HRD conserved region in the interaction with the HIRA WD domain. Significantly, this stretch of amino acids has similarity to two other previously defined WD domain interaction motifs. The sequence FYINSGT from UBN1 is similar to the FXIXXIL sequence of the evolutionarily conserved WD-binding EH1 motif (identical or similar residues are in bold) (12). Significantly, in UBN1 we showed that the conserved F and I residues are required for binding to HIRA. The UBN1 sequence FYINSGT is also similar to the WD-binding sequence SYLDSGI from β-catenin (63). Whether the UBN1/HIRA interaction is structurally analogous to either of these interactions remains to be determined by structural studies. Outside of the HRD, Hpc2p is not closely related to UBN1 and UBN2. This suggests that there has been significant divergence of function of the proteins aside from HIRA/Hir binding, although the yeast and mammalian complexes are all involved in aspects of chromatin metabolism.
Different histone chaperone complexes mediate nucleosome assembly in conjunction with specific physiological processes. For example, a histone chaperone that contains the heterotrimeric (p150, p60, and p48) CAF-1 complex and ASF1a (or a related protein, ASF1b) mediates DNA replication-coupled nucleosome assembly (23). In contrast, the HIRA-containing histone chaperone complex studied here mediates DNA replication-independent chromatin assembly (58). Previous structural, functional, and mutagenesis studies have deduced a model of the HIRA/ASF1a/histone H3/H4 complex in which the B domain of HIRA and histone H3 (bound to H4) proteins bind to opposite faces of ASF1a to form a quaternary complex (59). The current work demonstrates the existence of a UBN1/HIRA/ASF1a complex in which UBN1 binds to the WD repeats of HIRA. In this model, UBN1 and ASF1a bind independently to distinct domains of HIRA, suggesting that HIRA is a scaffold for the UBN1 and ASF1a proteins, perhaps in association with the histone H3/H4 complex. Interestingly, previous studies showed that p60CAF-1 binds to ASF1a through a B domain-like motif (59), making the p60CAF-1/ASF1a interaction of the CAF-1/ASF1a/histone H3/H4 complex analogous to the HIRA/ASF1a interaction. Also like HIRA, p60CAF-1 contains WD repeats, and the WD repeats of HIRA and p60CAF-1 are mutually more closely related to each other than any other human protein, reinforcing the parallel between HIRA and p60CAF-1. However, we have been unable to detect an interaction between p60CAF-1 and UBN1, suggesting that another, unknown protein binds to the WD repeats of p60CAF-1 in the p60CAF-1/ASF1a/histone H3/H4 complex. In this light, the presence of the UBN2 gene, and its expression at least at the RNA level, is perhaps significant. Obviously, the function of this additional Hpc2p ortholog remains to be investigated. We have also identified candidate Hpc2p orthologs in other model organisms that have previously been used to define the role of the Hir/Asf1 protein complex, such as Drosophila melanogaster, Caenorhabditis elegans, Arabidopsis thaliana, and Schizosaccharomyces pombe (Fig. (Fig.1A).1A). This identification will allow each of these species to be better exploited to properly analyze the function of this important chromatin-remodeling complex.
Several lines of evidence show that UBN1 contributes to SAHF formation in senescent cells. First, UBN1 associates with HIRA in vitro and in vivo and with the HIRA/ASF1a complex in vivo. Second, UBN1 colocalizes with HIRA in PML bodies in senescent cells and cells entering senescence. Third, UBN1 can be detected bound to the proliferation-promoting cyclin A2 gene, and it is also associated with HMTase activity toward lysine 9 of histone H3. Since H3K9Me is predominantly associated with transcriptional repression (5), these observations are consistent with a role in transcription repression of proliferation-promoting genes. Fourth, ectopic expression of UBN1 accelerates formation of SAHF. Fifth, RNA interference-mediated knockdown of UBN1 blocks formation of SAHF. However, we previously showed that in ectopic expression assays, residues 421 to 729 of HIRA are sufficient to drive formation of SAHF (71). This fragment of HIRA does not contain the UBN1-binding domain defined here, arguing against an essential role for UBN1 in HIRA-mediated formation of SAHF. Currently, we hypothesize that UBN1 potentiates formation of SAHF at endogenous levels of protein expression but is not required for formation of SAHF when HIRA is ectopically overexpressed. More sophisticated experiments are required to test this hypothesis. Regardless, the results reported in this work indicate that in the absence of ectopic expression, UBN1 is an indispensable component of the HIRA/ASF1a SAHF assembly pathway.
Cellular senescence is a known tumor suppressor mechanism (7, 8, 10, 11, 13, 17, 33, 52, 56, 62, 65) and is implicated in tissue aging (8, 28, 30, 34). Therefore, as a driver of senescence, the UBN1/HIRA/ASF1a chromatin-remodeling pathway might contribute to tumor suppression and the normal aging process. Indeed, recent studies are consistent with roles in both processes (27, 37). Future studies should address the role of normal and aberrant regulation of all components of the pathway, now including UBN1, in both aging and cancer.
We thank Brad Johnson and Shelley Berger for critical comments during the course of this work.
The lab of P.D.A. is funded by grants R01 GM062281, P01 AG031862, and R01 CA129334-01 and the Mary Kay Ash Foundation. P.D.A. is a Leukemia and Lymphoma Society Scholar. The contribution from the lab of R.M. was funded by grant P01 AG031862.
Published ahead of print on 24 November 2008.