|Home | About | Journals | Submit | Contact Us | Français|
Hdac3 is essential for efficient DNA replication and DNA damage control. Deletion of Hdac3 impaired DNA repair and greatly reduced chromatin compaction and heterochromatin content. These defects corresponded to increases in histone H3K9,K14ac, and H4K5ac and H4K12ac in late S phase of the cell cycle, and histone deposition marks were retained in quiescent Hdac3-null cells. Liver-specific deletion of Hdac3 culminated in hepatocellular carcinoma. While HDAC3 expression was down regulated in only a small number of human liver cancers, the mRNA levels of the HDAC3 cofactor NCOR1 were reduced in 1/3 of these cases. siRNA targeting of NCOR1 and SMRT (NCOR2) increased H4K5ac and caused DNA damage, indicating that the HDAC3/NCOR/SMRT axis is critical for maintaining chromatin structure and genomic stability.
Histone deacetylases (HDACs) play major roles in modulating chromatin accessibility during transcription, replication, recombination and repair (Gallinari et al., 2007; Goodarzi et al., 2009), yet the role of individual HDACs in these processes is still unclear. Deacetylation of histones is required for re-establishing chromatin structure on a local basis after transcription of a gene or after the repair of a DNA double strand break (Tsukamoto et al., 1997). On a global scale, HDACs act during DNA replication when the cellular histone content is doubled, as these newly synthesized histones are acetylated prior to their deposition onto nascent DNA. The residues most commonly associated with this process are H4K5ac and H4K12ac (Sobel et al., 1995; Taddei et al., 1999). These modifications presumably allow histone chaperones to configure the nucleosome correctly before deacetylation stabilizes the nucleosome and/or allows higher order compaction of the chromatin and the formation of heterochromatin (Luger et al., 1997; Luger and Richmond, 1998; Neumann et al., 2009; Verreault et al., 1996).
This process of histone acetylation/deacetylation is required for genomic stability and cell viability, as perturbations in the acetyl transferase or components of this pathway cause genomic instability and result in a failure to recover from genotoxic stress (Clarke et al., 1999; Han et al., 2007; Smith et al., 1998; Yuan et al., 2009). This is a dynamic process that occurs across the entire genome and the role of HDACs in the re-establishment of chromatin structure after replication is one of the least explored areas of their action. As such, genetic methods have been the most informative approaches to understand the physiological role of these critical regulatory enzymes.
Targeting enzymes that control chromatin structure and topography has been an extremely valuable tool in cancer therapy. A wide variety of general and specific small molecule inhibitors targeted towards HDACs are currently in clinical trials and are used as therapies for both solid and hematological tumors (Bolden et al., 2006). At therapeutic doses, histone deacetylase inhibitors (HDIs) not only cause cell cycle-dependent DNA damage, but also affect DNA repair, which sensitizes cells to ionizing radiation (IR), topoisomerase inhibitors and cisplatin (Baschnagel et al., 2009; Marchion et al., 2004; Suzuki et al., 2009). However, the molecular mechanism for inefficient DNA repair following HDI treatment is still not clear. Given the high levels of histone acetylation that accumulate in the context of these inhibitors, it is reasonable to assume that disruption of chromatin structure may contribute to cell death. As more selective HDAC inhibitors are moving into clinical trials, it is important to elucidate the function of individual HDACs to design better and more specific drugs for cancer therapy and to understand the mechanism(s) of action or side effects.
Hdac3, a class I HDAC, associates with the nuclear hormone co-repressors (NCoR and SMRT) (Codina et al., 2005) and is generally thought of as a locus-specific co-repressor that is recruited to promoters to repress genes regulated by nuclear hormone receptors and other transcription factors (Jones and Shi, 2003). In yeast, Snt1 and Hos2 have features of NCoR/SMRT and Hdac3, respectively (Pijnappel et al., 2001). This suggests a more ancestral and fundamental role of these proteins perhaps in the cell cycle, and that this machinery is also used for gene-specific transcriptional regulation. In agreement with this hypothesis, conditional deletion of Hdac3 in mouse models demonstrated that murine embryonic fibroblasts (MEFs) required Hdac3 for cell viability (Bhaskara et al., 2008). The observed apoptosis was associated with an impaired S-phase progression and DNA double strand breaks, rather than altered transcriptional programs (Bhaskara et al., 2008). The DNA damage was blocked when cells were taken out of the cell cycle by serum starvation, which suggested that Hdac3 acts during the S phase (Bhaskara et al., 2008). We propose that the cell cycle functions of HDAC3 and its regulatory factors NCoR and SMRT may be the ancestral role for and that disruption of these cell cycle functions may have dramatic consequences for the regulation of chromatin structure and genomic stability. These roles may impact the usefulness of HDAC3 as a therapeutic target in cancer and other diseases.
The inactivation of Hdac3 resulted in increased sensitivity of quiescent MEFs to ionizing radiation, suggesting a defect predominantly in non-homologous end joining (NHEJ)-mediated repair in these cells (Bhaskara et al., 2008). To test whether these defects were due to altered DNA repair functions or due to altered histone modifications and the ensuing changes in chromatin structure, we examined whether the inactivation of Hdac3 increases the sensitivity of Hdac3-null cells to other DNA damaging agents. For this purpose, we treated Hdac3FL/+ and Hdac3FL/− MEFs carrying a tamoxifen-inducible ER-Cre allele with either increasing concentrations of doxorubicin or cisplatin 48 hr after the addition of 4-hydroxy tamoxifen to inactivate Hdac3. Doxorubicin inhibits topoisomerase II (Swift et al., 2006) and triggers S-phase associated DNA double strand breaks (DSBs) that are repaired by the homologous recombination (HR) pathway, whereas cisplatin crosslinks DNA to form intra-strand adducts (Siddik, 2003). Inactivation of Hdac3 increased the sensitivity of MEFs to doxorubicin (Fig. 1A) and cisplatin (Fig. 1B), suggesting that in the absence of Hdac3, these DNA repair pathways are inefficient.
Given that Hdac3 deletion appeared to affect two independent types of DNA repair, we examined whether Hdac3 plays a role in the two major NHEJ and HR DSB repair pathways. We employed a chromosomally integrated reporter allele system established in HEK 293 cells (Fig. 1C and Fig. 1D) and used siRNAs to deplete the endogenous levels of HDAC3 (Fig. 1E) prior to cutting the reporter site with the I-Sce1 homing endonuclease. The efficiency of rejoining I-Sce1-cleaved sites was measured using quantitative PCR for NHEJ and by flow cytometry for reconstituted GFP expression for HR. In both cases, the reduction in Hdac3 levels caused a 50–60% decrease in DNA repair (Fig. 1F and Fig. 1G), indicating that Hdac3 is essential for efficient NHEJ- and HR-mediated repair.
Although Hdac3 has been linked to NHEJ due to its association with the SMRT/Ku70 complex (Yu et al., 2006), our DNA damage sensitivity and repair data suggested that Hdac3 loss affects DNA repair by targeting an element that is common to multiple types of DNA repair. Moreover, Hdac3 is not recruited to the sites of double strand breaks following IR treatment (Fig. S1), nor did its loss affect the localization of other members of the DNA damage response (Rad50, Brca1, Mdc1 and Mre11; data not shown). One of the key histone modifications that contributes to the DNA damage response, the first step in double strand break repair, is H3K9 trimethylation (H3K9me3), which recruits the histone acetyltransferase Tip60 (Sun et al., 2009) and other factors involved in the damage response (e.g., HP1β; (Ayoub et al., 2008)). Therefore, we tested whether siRNA targeting of HDAC3 would alter histone acetylation at the site of an integrated reporter. Chromatin immunoprecipitation employing anti-H3K9,K14ac showed that histone acetylation was increased at the substrate locus at a level consistent with global changes in H3K9,K14ac (see Fig. 3 below), along with a concomitant decrease in H3K9me3 (Fig. 1H), suggesting that global changes in histone modifications could contribute to the defects in DNA repair.
H3K9me3 is one of the best marks of heterochromatin. The decrease in H3K9me3 upon siRNA-mediated suppression of HDAC3 or Hdac3 deletion in the liver (Fig. 1H, and Fig. 3 below) prompted us to use the Albumin-Cre transgene to delete Hdac3 in vivo to examine chromatin structure. Initially, we used transmission electron microscopy to examine the nuclei from Alb-Cre:Hdac3+/− and Alb-Cre:Hdac3−/− p17 livers (Fig. 2A). As expected, in control nuclei the electron-dense heterochromatin was found at the nuclear periphery (Fig. 2A). In contrast, Alb-Cre:Hdac−/− p17 liver nuclei showed a significant decrease in the amount of heterochromatin, especially at the periphery (Fig. 2A; note that the residual electron dense material remaining is consistent with the presence of nucleoli). A similar result was obtained by enumerating the Hoechst staining of heterochromatic foci that are evident in fluorescence microscopy in mouse cells. Alb-Cre:Hdac−/− liver nuclei showed significantly fewer foci (Fig. 2B). Moreover, when cell fractionation was used to isolate chromatin, immunoblot analysis demonstrated that Alb-Cre:Hdac3−/− hepatocytes contained roughly 2-fold less HP1β on chromatin (Fig. 2C). Finally, at the gene-specific level we used ChIP to examine H3K9,K14ac at the p53 locus (Su et al., 2009). Inactivation of Hdac3 caused the accumulation of acetylated H3K9,K14 at the promoter, upstream of the promoter, within intron 1, and within the body of the gene (Fig. S1), which is consistent with global changes in chromatin structure and histone modifications (see Fig. 3 below). Thus, Hdac3 is required for maintaining chromatin structure in vivo.
Given the reduction in heterochromatin, we used micrococcal nuclease (MNase) digestion to examine nucleosomal compaction. Digestion of isolated nuclei with increasing concentrations of MNase demonstrated that the bulk chromatin from Hdac3−/− hepatocytes was more sensitive to MNase digestion when compared to the chromatin from control hepatocytes (Fig. 2D), indicating that global chromatin structure was altered and is more “open” in the absence of Hdac3. In addition, we noted that the amount of DNA associated with mono-nucleosomes did not increase with increasing concentrations of MNase, suggesting that the nuclosomal DNA was more accessible in the absence of Hdac3. Southern blot analysis of these same samples using major and minor satellite probes or quantitative PCR for these regions showed modest sensitivity without a change in nucleosomal spacing (Fig. S2) (Gilbert and Allan, 2001; Sugimura et al., 2010). Given the apparent sensitivity of mono-nucleosomal DNA to MNase (Fig. 2D), we tested the sensitivity of nucleosomes to ionic conditions by extracting histones with differing concentrations of salt. Consistent with prior results (Li et al., 1993), Histone H3 was resistant to NaCl concentrations up to 1.2M in control hepatocytes, but in the absence of Hdac3, Histone H3 was nearly completely soluble in 900 mM NaCl, suggesting that nucleosome integrity was altered in Hdac3−/− hepatocytes.
Structural determinations of the nucleosome show that the basic lysine residues in histone tails can potentially associate with DNA, but more likely mediate contacts with acidic surfaces on adjacent nucleosomes to allow nucleosome compaction and fiber formation (Luger et al., 1997; Luger and Richmond, 1998; Schalch et al., 2005). Therefore, we performed western blot analysis of nuclear extracts prepared from Alb-Cre:Hdac3+/− or Alb-Cre:Hdac3−/− p17 livers to examine global histone acetylation and methylation marks that neutralize the lysine charges and that regulate chromatin structure. Hdac3 deacetylates H4K5ac and H4K12ac in vitro (Johnson et al., 2002) and loss of Hdac3 resulted in an accumulation of H4K5ac, H4K12ac, and H4K16ac, as well as H3K9,K14ac with little affect on the acetylation of other residues (Fig. 3) (Knutson et al., 2008). Examination of histone methylation demonstrated that H3K9me3 (Lachner et al., 2001) and H3K79me2 were reduced (Fig. 3), while H3K4me3, H3K27me3 and H4K20me2 were unaffected (Fig. 3).
H4K5ac and H4K12ac are commonly associated with histone deposition onto newly synthesized DNA. Given that hepatocytes are generally quiescent, our liver-specific deletion of Hdac3 implies that Hdac3 is required for removing these marks during or after DNA synthesis. Therefore, we used Hdac3FL/− NIH 3T3 cells infected with Adenovirus expressing Cre to examine these marks during the cell cycle. Cells were infected with Adeno-Cre and synchronized in G0/G1 by serum starvation and then released into the cell cycle by the addition of serum to the culture medium. The level of H4K5ac and H4K12ac was then examined using western blot analysis of extracts prepared from cells at various time points after serum addition. After 48 hr of culture in the absence of serum, the percentage of cells in G0/G1 approached 90% (Fig. S3) and H4K5ac and H4K12ac were reduced to low levels in control cells, but remained high in Hdac3-depleted cells (Fig. 4A). After serum addition, both cultures re-entered the cell cycle with cells beginning to enter S phase at 12 hr and over 50% of the cells progressing through S phase by 18 hr as measured by BrdU incorporation (Fig. S3). At 18 and 24 hr after serum addition, H4K5ac and H4K12ac increased in the control cells consistent with the acetylation of these residues on newly synthesized histones that are deposited on new DNA (Fig. 4A). In cells lacking Hdac3, the levels of acetylation of these residues were already high and increased only modestly during S phase (Fig. 4A). While H3K9,K14 acetylation has not been linked to histone deposition in mammalian cells in the manner that H4K5ac and H4K12ac have, its levels were low in G0/G1 phase control cells and increased during S phase. In the absence of Hdac3, H3K9,K14ac was not reduced upon serum starvation, suggesting that acetylation of these residues was not removed after S phase, which could account for the loss of H3K9me3 and reduced heterochromatin in Hdac3-null livers (Fig. 2 and and33).
To better define the requirements for Hdac3 in the removal of cell cycle-associated marks, we directly examined H4K5ac in S phase cells, as this is a classical deposition mark. The punctate immunofluorescence pattern of PCNA 18 hr after release from serum starvation was used to identify cells in late S phase, which are characterized by foci of PCNA at the nuclear periphery (Celis and Celis, 1985; Madsen and Celis, 1985; Taddei et al., 1999). Although, a general increase in H4K5ac was observed in the absence of Hdac3 (data not shown), a pronounced increase in H4K5ac was found in late S phase cells (11% of cells in controls vs. 57% of Hdac3-null cells, Fig. 4B), especially at the nuclear periphery.
Even relatively modest over expression or reduced expression of histones is sufficient to affect genomic stability by causing DNA double strand breaks (Gunjan and Verreault, 2003; Olive and Banath, 1995). Given the apparent high level of H4K5ac in the nucleus of Hdac3-null cells after S phase (Fig. 4A and 4B), we asked what proportion of the total histone H4 is acetylated in the absence of Hdac3. Two sequential rounds of immunoprecipitation of lysates prepared from quiescent Hdac3-null NIH 3T3 cells were performed using anti-H4K5ac to identify the acetylated histone. Western blot analysis using anti-H4K5ac confirmed that most of the acetylated histone was immuno-purified in the first immunoprecipitation (Fig. 4C, upper panel). The total histone H4 that was acetylated at K5 was then determined by comparing the amount of total histone H4 in the immunoprecipitation versus H4 that remained in the supernatant (Fig. 4C, lower panel). For the western blot analysis, 1/5th the amount of the supernatant was loaded as compared to the precipitated H4K5ac. Thus, roughly 15–20% of the total histone H4 was acetylated at K5 in control cells and 30–40% was acetylated in the absence of Hdac3 (Fig. 4C).
Previously, we noted S phase-associated DNA double strand breaks in Hdac3-null cells (Bhaskara et al., 2008). Given the requirement for Hdac3 in removing histone acetylation marks that are added during the S phase of the cell cycle and the links between loss of H3K9me3 and genomic instability, we examined the consequences of inactivation of Hdac3 to chromosomes as they progress through mitosis. Metaphase spreads were prepared from Hdac3FL/+ and Hdac3FL/− MEFs following either Ad-Cre infection or following tamoxifen treatment of MEFs carrying the ER-Cre transgene. Both chromosome breaks and gaps were quantified using cytogenetic analysis. Using either Ad-Cre (Fig. S4) or ER-Cre to delete Hdac3 (Fig. 5A) led to a 5–8 fold increase in the average number of breaks and gaps in metaphase chromosomes when compared to control cells (Fig. 5B), indicating a crucial role for Hdac3 in the maintenance of genome stability.
To test whether the DNA damage and genomic instability phenotypes found in MEFs lacking Hdac3 was recapitulated in vivo, we examined Alb-Cre:Hdac3+/− and Alb-Cre:Hdac3−/− livers for DNA double strand breaks using immunofluorescence to detect γH2AX and 53BP1, which localize to sites of DNA double strand breaks (Iwabuchi et al., 2003). While there was little or no endogenous DNA damage in Alb-Cre:Hdac3−/− livers at post-natal day 17 (data not shown), by p28 the Hdac3-null livers displayed an increased number of cells with γH2AX and 53BP1 foci when compared to the control hepatocytes (0 Gy panels, Fig. 5C; see Fig. S4 for quantification). Subsequently, we examined DNA repair in Hdac3-null hepatocytes at p28 following a non-lethal dose of IR (3Gy). An increased percentage of cells with a substantial amount of DNA damage was detected both 1hr and 6hr after IR in Alb-Cre:Hdac3−/− when compared to the control Alb-Cre:Hdac3+/− hepatocytes (Fig. 5C). Quantification of 53BP1 foci in ~100 cells from two independent experiments following a 6 hr recovery period revealed that Hdac3-null cells have a greater percentage of cells with 5 to 10 foci when compared to the control cells (Fig. S4). We also observed DNA damage even after a 24 hr recovery period in some Alb-Cre:Hdac−/− hepatocytes, whereas control hepatocytes had repaired the damage caused by IR treatment (data not shown).
Analysis of gene expression data obtained previously from Alb-Cre:Hdac3−/− livers at p28 identified an up-regulation of genes that belong to the p53 network (Fig. S4), suggesting the presence of DNA damage. Likewise, quantitative RT-PCR analysis revealed an up-regulation of miRNAs regulated by p53 in Hdac3-null livers (Fig. S4), which is also consistent with the activation of a DNA damage response (Rokhlin et al., 2008). In addition, hepatocellular carcinoma (HCC) progression markers, such as gamma-glutamyl transpeptidase 1 (Pavesi et al., 1989) and insulin-like growth factor II (Qiu et al., 2008) were up-regulated in the microarray analysis of Alb-Cre:Hdac3−/− livers by 2.2- and 2.7-fold respectively (Knutson et al., 2008). These data, coupled with the observed genomic instability, prompted us to age cohorts of 20 control and 20 Alb-Cre:Hdac3−/− mice. By 15–16 weeks of age, the livers were very pale due to the dramatic accumulation of neutral lipids and fat caused by inactivation of Hdac3 (Knutson et al., 2008) and contained “white nodules” when examined by gross morphology (Fig. 6A). These nodules were encapsulated with a fibrous lining that positively stained with Massion’s trichrome (data not shown) and appeared to be benign “adenoma-like” structures with the cytoplasm of the cells filled with microvesicular fluid and an abundance of mitochondria (Fig. 6A, and data not shown). By 8–10 months of age, most of the mice began to show signs of distress and necropsy identified the presence of tumors in the liver. The experiment was humanly terminated for all mice by 14 months of age (Fig. 6B). Pathological analysis indicated that 20 of 20 mice succumbed to low-grade hepatocellular carcinoma (HCC) at a mean age of 10.2 months (Fig. 6B). Immunohistochemical staining for Hdac3 confirmed that the tumors lacked expression of Hdac3 (Fig. 6C) and Ki-67 staining confirmed a high proliferative index in the tumors (Fig. 6D). The HCC displayed a loss of normal architecture, a trabecular patterning of cells, a lack of ductal morphology, and very disorganized features (Fig. 6D).
The loss of genomic stability and the impaired response to DNA damage suggested that a high mutation rate stimulated the development of HCC (Fig. 1, Fig. 5, Fig. 6). To begin to assess what pathways were involved in the formation of HCC, we performed gene expression analysis using cDNA microarrys (Fig. 7A). In the array data, we noted the enhanced expression of c-Myc, a commonly over expressed oncogene, which was confirmed using quantitative RT-PCR (Fig. 7B). Signatures consistent with activation of the Ras pathway and impairment of the p53 pathway were also identified in this analysis (Fig. 7A). The Wnt pathway has been identified as a key regulatory node in human HCC and we also noted that this pathway was affected in the Hdac3-null tumors. Therefore, we examined β-catenin localization using both cell fractionation and immunohistochemical staining. As early as p28, we found increased amounts of β-catenin localized to the nucleus and there was prominent nuclear localization of β-catenin in the Hdac3-null HCCs (Fig. 7B, C), confirming that this oncogenic pathway was up regulated in this mouse model of HCC.
Analysis of HDAC3 mRNA levels in 4 independent human HCC datasets from the GEO database indicated that HDAC3 was reduced in some cases (Fig. 8 and Fig. S5). HDAC3 is highly expressed in cycling cells (e.g., in the crypt regions of the colonic epithelium) (Spurling et al., 2008; Wilson et al., 2006), so comparing quiescent normal tissue to cycling cancer cells may under estimate the level of its loss of expression. Nevertheless, in the largest dataset, HDAC3 levels were reduced by 1.5-fold in 13% of the cases and over 2-fold in 3.3% of the cases as compared to normal controls (data not shown and Fig. 8A). There was no association with hepatitis C or hepatitis B viral infection or survival (data not shown). Given that HDAC3 is dormant until activated by association with NCoR or SMRT (Guenther et al., 2001), and that it is recruited to chromatin through association with other factors, we probed the GEO HCC datasets for changes in expression of the 57 direct interaction partners of HDAC3 identified in the Human Protein Reference Database (Table S1). Four of these genes (STAT3, GTF2I, GCM1, and NCOR1) showed reduced expression in HCC. NCOR1 is not only down regulated, but also NCOR1 is located on a region of chromosome 17p that is deleted in human HCC (Xu et al., 2001). The levels of NCOR1 were reduced by 2-fold or greater in nearly 1/3 of HCC samples in the largest dataset (Fig. 8A) and was similarly down-regulated in the majority of the samples in smaller HCC gene expression datasets (Fig. S6). Using immunohistochemistry to detect nuclear NCOR1, we found that 2/5 human HCCs tested had reduced levels of NCOR1 (Fig. 8B), which is consistent with the mRNA expression results.
Given that NCoR/SMRT directly control HDAC3 functions (Guenther et al., 2001), we used siRNAs to probe the requirements for these HDAC3 cofactors in the regulation of histone acetylation. In HeLa cells that express both family members, depletion of either NCoR or SMRT alone had only modest effects on global histone acetylation (data not shown). Targeting both family members together caused a small increase in H4K5ac (Fig. 8C). However, even though NCoR/SMRT levels were only reduced by about 50%, there was a significant accumulation in the levels of global H4K5ac in both HeLa cells (Fig. 8C) and NIH 3T3 cells (Fig. S6B). These increases in histone acetylation were associated with a decrease in the levels of HDAC3 detected in both HeLa and NIH3T3 cells (Fig. 8C and Fig. S6C). Immunofluorescence using anti-PCNA to identify S phase cells, demonstrated a 5-fold increase in the number of cells with high levels of H4K5ac at the nuclear periphery in late S-phase cells targeted with siRNAs to both co-factors (Fig. 8D and Fig. S6C). Given this alteration in histone marks, we examined these cells for DNA double strand breaks using anti-53BP1. Cells depleted of NCOR1 and SMRT showed a dramatic increase in the number of cells with greater than 10 foci (Fig. 8E and Fig. S6D). Collectively, these results show that the NCoR/SMRT/HDAC3 axis is required for removing histone marks globally and maintaining genomic stability.
The increase in acetylation of H4K5, H4K12, H4K16ac, and H3K9,K14 that was observed upon inactivation of Hdac3, along with the concomitant loss of H3K9me3, provides a likely mechanism for the failure to maintain chromatin structure in Hdac3-null mice (Fig. 2). H4K5ac and H4K12ac are associated with histone deposition (Sobel et al., 1995), especially in heterochromatic regions that are replicated late in S phase (Taddei et al., 1999) and this pattern was accentuated in the absence of Hdac3 or when NCoR/SMRT were targeted using siRNAs (Fig. 4 and and8).8). The removal of these marks is required for propagation of heterochromatin in yeast (Zhou et al., 2009), which suggests that the accumulation of these marks (and the reduced levels of H3K9me3) may be the underlying cause for the reduction in heterochromatin in the Hdac3-null livers. Indeed, loss of H3K9me3 impaired the DNA damage response, and the inactivation of the murine H3K9 methyltransferases or mutation of HP1β caused genomic instability, defects in DSB repair, and an increased tumor risk (Aucott et al., 2008; Kondo et al., 2008; Luijsterburg et al., 2009; Peters et al., 2001; Sun et al., 2009). This suggests that the failure to maintain a normal chromatin structure underlies the Hdac3−/−-associated defects in two distinct types of DNA repair (NHEJ and HR), and in genomic stability (Fig. 1, Fig. 5), which ultimately led to tumor development (Fig. 6).
The siRNA-mediated knockdown of NCoR/SMRT, like deletion of Hdac3, caused the accumulation of histone deposition marks, suggesting that these Hdac3 activating factors also play an intrinsic role during the S phase. N-CoR and SMRT were initially identified as transcriptional co-repressors associated with nuclear hormone receptors (Chen and Evans, 1995; Horlein et al., 1995; Karagianni and Wong, 2007), as well as with a variety of DNA binding factors (Perissi et al., 2004). However, the function of these co-repressors during the cell cycle may represent the ancestral activity of these complexes. We speculate that during evolution this cell cycle machinery was recruited in higher organisms to regulate gene expression patterns in a cell type-specific manner (e.g., in response to nuclear hormones) or to form heterochromatin to more permanently silence gene expression.
While HDAC3 has been suggested to be over expressed in colorectal carcinoma (CRC) (Spurling et al., 2008; Wilson et al., 2006), it is not amplified at the DNA level. In addition, HDAC3 is expressed at higher levels in the proliferating cells of the colonic crypts, which might suggest that its levels are higher in CRC because the cells are cycling. Conversely, it is notable that HDAC3 lies within a region of chromosome 5q31.3 that is frequently deleted in breast cancer (Johannsdottir et al., 2006) and myelodysplastic syndromes (Ebert, 2009). Intriguingly, NCOR1 lies in a region of chromosome 17 that is frequently deleted in HCC (Mahlknecht et al., 1999), and an analysis of expression profiles indicated that down regulation of NCOR1 expression is common in a subset of human HCC (Fig. 8A and Fig. S6). Our data are consistent with the inactivation of the HDAC3/NCOR/SMRT axis possibly contributing to a subset of human cancer by allowing the increase of histone acetylation during the S phase leading to DNA damage and further accumulation of mutations.
Nearly all non-targeted cancer therapeutics (i.e., those that do not target a mutant protein that initiates a cancer) develop a therapeutic window by acting on cycling cells to cause DNA damage (Ashwell and Zabludoff, 2008; Lieberman, 2008). However, one side effect is that these agents, when given at too high a dose or for too long, also cause genomic instability in normal cells leading to therapy-associated secondary cancers. Our results raise this possibility for HDIs, all of which currently target HDAC3. However, compounds such as SAHA appear to be well tolerated, possibly owing to their short half-life in vivo (Butler et al., 2000). That is, SAHA may cause S phase-associated DNA damage for those cancer cells in S phase during the 4–6 hr window in which the daily dose of SAHA is active, but only cause mild problems for the majority of normal cells that are not cycling. In addition, normal cells that are proliferating such as in the gastrointestinal tract and in the bone marrow, can either repair the DNA damage or their chromatin is “reset” after the SAHA is metabolized. Thus, we predict that while continuous inhibition of Hdac3 is detrimental (e.g., Fig. 6), transient inhibition, even when frequently repeated, may be safe.
Please see the supplemental experimental procedures for additional methods used.
Mice harboring either a conditional floxed (fl) allele or a null (−) allele were created as described previously (Knutson et al., 2008). To create liver-specific Hdac3 knockout mice, mice with a floxed or a null allele were crossed to transgenic mice expressing Alb-Cre (Knutson et al., 2008) to obtain Alb:Cre:Hdac3+/− and Alb:Cre:Hdac3−/− offspring mice. All experiments using mice were approved by the Vanderbilt University Institutional Animal Care and Use Committee.
DNA repair assays were performed using HEK293 cells engineered with integrated reporters for both types of double strand break repair (Zhuang et al., 2009). Briefly, cells (1 × 105) were transfected twice with HDAC3 siRNA using Oligofectamine (Invitrogen, CA). pCMV-I-SceI or the control vector were transfected into the cells using FuGene 6 (Roche). The cells were either analyzed by two-color FACS analysis to examine homologous recombination or PCR was used to determine the rate of NHEJ-mediated repair (see supplemental experimental procedures for primer sequences).
Liver tissue was minced into fine pieces and fixed in 2.5% glutaraldehyde in 0.1M cacodylate buffer at room temperature for 1–2 hr. Samples were then washed in 0.1M cacodylate buffer and treated with 1% aqueous osmium tetraoxide. Tissues were then washed, dehydrated and embedded in Spurr resin. Thin sections (100 nm) were viewed with an FEI CM-12 transmission electron microscope operated at 80KeV.
Four HCC and normal control experiments (GSE14323, GSE6764, GSE6222, and GSE5975) were selected from the Gene Expression Omnibus database (GEO, http://www.ncbi.nlm.nih.gov/geo/)(Barrett et al., 2009). The description of this analysis is elaborated in the supplement. For GSE6764 (Wurmbach et al., 2007), we compared the normal liver samples with HCC samples of different stages. For GSE6222 (Liao et al., 2008), we excluded the HuH7 cell line data. GSE1898 was excluded from this analysis due to variations in the probe sets. Expression data generated from murine tumors lacking Hdac3 is found in GSE22457.
For preparation of whole cell protein extracts, the cell pellet was washed with phosphate buffered saline (PBS) and sonicated in RIPA buffer (0.5% Triton-X-100, 0.5% deoxycholic acid, and 0.5% SDS in PBS) with protease inhibitors (0.5mM PMSF, 2µg/µl leupeptin and 15µg/µl aprotinin) prior to western analyses. Antibodies used and the preparation of nuclear extracts are described in the Supplementary Materials.
Broad-spectrum histone deacetylase inhibitors (HDI) are being used to treat a variety of cancers, and more selective inhibitors are being developed for therapeutic uses. Genetic analysis of individual histone deacetylases (HDACs) is essential to understand the action of these inhibitors and their potential side effects. We show that inactivation of Hdac3, a central target of all currently used HDIs, causes genomic instability, and deletion of Hdac3 in the liver leads to hepatocellular carcinoma. These phenotypes correlate with global increases in the acetylation of specific histone residues, disruptions in chromatin structure, and a loss of heterochromatin. Our results genetically link the HDAC3/NCOR/SMRT axis to the maintenance of critical cell cycle functions and genomic stability.
We thank all the members of Hiebert lab for helpful discussions, reagents and advice. We thank the Vanderbilt Imaging, Human Tissue Acquisition and Pathology, and Functional Genomics Shared Resources for services and support. We thank Drs. Nick Gilbert and James Allan for providing plasmids containing the major and minor satellite probes. This work was supported by the T. J. Martell Foundation, the Robert J. Kleberg, Jr. and Helen C. Kleberg Foundation, National Institutes of Health grants (R01-CA64140, RO1-CA77274, RO1-CA109355) and core services performed through Vanderbilt Digestive Disease Research grant NIDDK P30DK58404 and the Vanderbilt-Ingram Cancer Center support grant NCI P30CA68485. SB was supported by a fellowship (1F32CA138091-01) from the NCI.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Expression data generated from murine tumors lacking Hdac3 is found in the Gene Expression Omnibus accession number GSE22457.