The histone deacetylase 9 gene encodes multiple protein isoforms.

Histone deacetylases (HDACs) perform an important function in transcriptional regulation by modifying the core histones of the nucleosome. We have now fully characterized a new member of the Class II HDAC family, HDAC9. The enzyme contains a conserved deacetylase domain, represses reporter activity when recruited to a promoter, and utilizes histones H3 and H4 as substrates in vitro and in vivo. HDAC9 is expressed in a tissue-specific pattern that partially overlaps that of HDAC4. Within the human hematopoietic system, expression of HDAC9 is biased toward cells of monocytic and lymphoid lineages. The HDAC9 gene encodes multiple protein isoforms, some of which display distinct cellular localization patterns. For example, full-length HDAC9 is localized in the nucleus, but the isoform lacking the region encoded by exon 7 is in the cytoplasm. HDAC9 interacts and co-localizes in vivo with a number of transcriptional repressors and co-repressors, including TEL and N-CoR, whose functions have been implicated in the pathogenesis of hematological malignancies. These results suggest that HDAC9 plays a role in hematopoiesis; its deregulated expression may be associated with some human cancers.

In eukaryotes, the ability to dynamically form and maintain distinct functional domains of chromatin is fundamental to nuclear processes, including regulation of gene transcription (1). The basic repeating unit of chromatin is the nucleosome, which is a complex consisting of 1.75 superhelical turns of DNA wrapped around a core histone octamer comprising two subunits each of H2A, H2B, H3, and H4 (2). The N-terminal tails of histones H3 and H4 protrude from the nucleosome and interact with the negatively charged DNA phosphate backbone when in their highly basic, unmodified state (3). These tails contain specific amino acids that are targets for a variety of enzymes, producing diverse modifications including acetylation, methylation, and phosphorylation. Acetylation is thus far the most widely studied and involves substitution of the ⑀amino group of specific lysines in a process catalyzed by histone acetyltransferases. This leads to a more acidic residue and an overall decreased affinity for DNA by the histone octamer. For the transcriptional machinery, the packaging of DNA into nucleosomal arrays presents a major physical obstacle in gaining access to the DNA template, and there has long been evidence that unwinding of nucleosomes because of the acetylation of histone tails plays an important role in the activation of transcription (4). As expected, enzymes that remove these modifications, histone deacetylases (HDACs), 1 are important in gene silencing, and recent studies have implicated abnormal HDAC function in a number of human cancers (5).
Here, we report the cloning and characterization of HDAC9, a member of the Class II histone deacetylase family. HDAC9 contains 1069 amino acids and functions as deacetylase both in vitro and in vivo, and this activity is essential for its associated repression of gene expression. Furthermore, HDAC9 is alternatively spliced to generate multiple protein isoforms that may harbor distinct biological activities and may be associated with human cancer. One of these isoforms consists of the noncatalytic N-terminal region of HDAC9 (conserved in HDACs 4, 5, and 7) and has been previously identified in Xenopus (MEF2interacting transcriptional repressor (MITR)) (35) and human (histone deacetylase-related protein (HDRP)) (36,37).
Double-stranded oligonucleotide primers were cloned into both ends of the HDAC9 cDNA according to Ausubel et al. (38), utilizing BclI (5Ј end) and BbsI (3Ј end) sites to create a N-terminal FLAG-tagged cDNA containing 5Ј-BamHI-XhoI-3Ј ends. The cDNA was then subcloned into the pSG5 vector (Stratagene) containing a modified polylinker to create F-HDAC9. FLAG-tagged HDAC9⌬CD (F-HDAC9⌬CD) was created by subcloning a 5Ј-PstI-XhoI-3Ј 3Ј fragment from HDAC9⌬CD into F-HDAC9. Plasmids containing the various FLAG-tagged alternatively spliced variants of HDAC9 and HDAC9⌬CD were constructed by subcloning appropriate fragments from the partial cDNAs described above into F-HDAC9 and F-HDAC9⌬CD.
GST expression vectors were derived from pGEX-5X1 (Amersham Biosciences) by subcloning indicated cDNAs in-frame with the coding region for glutathione S-transferase. Mammalian two-hybrid expression vectors were derived from pM (Clontech) by subcloning the indicated cDNAs in-frame with the coding region for the GAL4 DNAbinding domain. The GAL4 uas x5-Tk-Luc reporter was derived from the pT109luc (39) plasmid by inserting five copies of the GAL4 DNAbinding site upstream of the minimal HSV-Tk promoter and luciferase (Luc) gene. The MEF2 RE x3-Tk-Luc reporter was also derived from the pT109luc plasmid by inserting three copies of the consensus MEF2binding site. The TEL RE x3-Tk-Luc reporter, as well as mammalian and in vitro expression vectors for AML1, TEL, and TEL-AML1, have been previously described (40). Mammalian and in vitro expression vectors for BCL-6 (41), MEF2D (42), PLZF (43), full-length and partial N-CoR (27,44), HDAC1 (45), HDAC3 (10), HDAC4 (13), as well as Sin3A and B (30), and SUMO-1 and -2 (46) proteins have been described by others.
Cell Culture-293T, COS-7 and tsCOS cells (47) were maintained in Dulbecco's modified Eagle's medium with 10% fetal calf serum (Sigma). Hematopoietic cell lines and the colon cancer cell line SW-620 were maintained in RPMI 1640 medium with 10% fetal calf serum (Sigma).
Isolation of Cell Populations-Adult peripheral blood was taken with informed consent, and the low density cells (Ͻ1.077 g/ml) were separated using Lymphoprep (Nycomed Pharma AS) and resuspended in phosphate-buffered saline (PBS) containing 1% fetal calf serum (Sigma). For sorting, the cells were stained separately with either mouse monoclonal anti-CD19 (Dako) or fluorescein isothiocyanate-conjugated anti-CD14 (Caltag-Medsystems) followed by enrichment of positive cells using the standard MACS system protocol (Miltenyi Biotec). Rat anti-mouse IgG 1 beads were used to isolate CD19 ϩve cells and antifluorescein isothiocyanate beads for CD14 ϩve cells. Nonviable cells were excluded by the addition of the dye To-pro-3 iodide (Molecular Probes) to presorted cells. Enriched cells were analyzed and sorted to Ͼ99% purity using a fluorescence-activated cell sorter (FACS Vantage SE; B.D. Biosciences) at two wavelengths (using 530/30-and 660/20-nm band passes).
Histone Deacetylase Assays-Immunoprecipitations for histone deacetylase assays were essentially performed as described (16). 293T cells (5 ϫ 10 7 cells) were transfected with PolyFect (Qiagen). One 10-cm dish was used per four HDAC assays. After 24 h, the cells were harvested and lysed in low stringency lysis buffer (50 mM Tris-HCl, pH 7.5, 120 mM NaCl, 0.5 mM EDTA, 0.5% Nonidet P-40) in the presence of a protease inhibitor mixture (Roche Molecular Biochemicals). To control for expression of different constructs, concentrations of protein extracts were normalized with a modified Lowry assay (Bio-Rad), and Western blot analysis was performed with the ECL procedure (Amersham Biosciences) as described (38). Extracts were precleared by incubation with protein 10% v/v protein A/G-Sepharose (Sigma) overnight at 4°C. Precleared lysates were immunoprecipitated by incubation with 10% v/v M2 anti-FLAG-agarose (Sigma) overnight at 4°C. Immune complexes were recovered by washing three times with low stringency lysis buffer, twice with lysis buffer containing 0.5 M NaCl (high stringency), and twice with HDAC buffer (10 mM Tris-HCl, pH 8.0, 10 mM NaCl, 10% glycerol). For inhibition studies, the immunoprecipitated complexes were preincubated with trichostatin A (400 nM) in HDAC buffer for 30 min at 4°C. Peptides corresponding to the N-terminal sequences of histone H3 (ARTKQTARKSTGGKAPRKQLC) and H4 (SGRGKGGKGLGKG-GAKRHRC) were [ 3 H]acetylated according to the manufacturer's instructions (Upstate Biotechnology, Inc.), with 20,000 cpm used as substrate per reaction. The beads were resuspended in 200 l of HDAC buffer containing acetylated peptide and 1 mM phenylmethylsulfonyl fluoride. Histone deacetylase activity was determined after incubation for 4 h at 37°C, according to the peptide manufacturer's instructions.
Chromatin Immunoprecipitation Assay-293T cells (5 ϫ 10 7 cells) were transfected by CaPO 4 (Profection, Promega) with 5 g of GAL4 uas x5-Tk-Luc reporter plasmid plus 10 g of expression vector containing heterologous fusions of the GAL4 DNA-binding domain with the indicated cDNAs. Immunoprecipitation of plasmid DNA plus associated histones was carried out ϳ40 h after transfection according to a previously published protocol (48), with the following modifications. Histone-DNA complexes were cross-linked by the addition of 1% formaldehyde to the medium and incubation at 37°C for 10 min. After lysis, the chromatin was sonicated to 0.2-1.0 kb and diluted 10-fold in IP buffer (0.01% SDS, 1% Triton X-100, 2 mM EDTA, 20 mM Tris-HCl, pH 8.0, 150 mM NaCl, plus protease inhibitors). Protein samples for Western blotting were taken prior to dilution; control samples for assaying input DNA were taken after dilution and de-cross-linked. Anti-acetylated histone H4 polyclonal antibody (Upstate Biotechnology, Inc.) was used for the immunoprecipitation, and the DNA-histone complexes were collected overnight with protein A/G-Sepharose beads (Santa Cruz). Sequences spanning the GAL4-binding site in the reporter were detected by semi-quantitative PCR using the forward primer 5Ј-ATTG-CAGCTTATAATGGTTA-3Ј and the reverse primer 5Ј-TTCGAAT-TCGCCAATGACAA-3Ј. The number of cycles was determined empirically to give results that fall within the linear range of this particular PCR assay, and the reactions were visualized by agarose gel electrophoresis followed by ethidium bromide staining.
RNA Isolation and Reverse Transcriptase-PCR Analysis-Total RNA was extracted from purified cells and cell lines using an RNeasy kit (Qiagen). cDNA was made from total RNA by reverse transcription using Moloney murine leukemia virus (Invitrogen) or Omniscript (Qiagen) reverse transcriptases in accordance with the manufacturers' instructions. HDAC9 cDNA was primed with 5Ј-TAAAACAGAGGCAG-CAGGGGAAGAGTGAGT-3Ј and HDAC9⌬CD/MITR cDNA was primed with 5Ј-TCAGATAATGACTTTAATTACAAATCCTGG-3Ј. Glyceraldehyde-3-phosphate dehydrogenase cDNA was primed with random hexamers (Roche Diagnostics). 2 g of total RNA was used per 20-l reaction, and 2 l of cDNA was used per subsequent PCR. The primer pairs were used for semi-quantitative PCR are shown in Fig. 5A. PCRs were visualized by agarose gel electrophoresis and ethidium bromide staining.
Preparation of an HDAC9-specific Antiserum-A rabbit antiserum specific to the C terminus of human HDAC9 was raised using a synthetic peptide corresponding to amino acids 1046 -1060 of the protein (DVEQPFAQEDSRTAG), which was conjugated to diptheria toxoid (Mimotopes). This C-terminal region is not conserved in any other HDACs closely related to the HDAC9 protein.
GST Pull-down Assays-All of the GST fusion proteins were prepared using standard procedures (43). [ 35 S]Methionine-labeled proteins were synthesized in vitro using a rabbit reticulocyte-coupled transcription-translation system (Promega), following the supplier's directions. 35 S-Labeled proteins were incubated with 1 g of GST or a given fusion protein. The assays were performed in NETN buffer (20 mmol/liter Tris, pH 8.0, 100 mmol/liter NaCl, 1 mmol/liter EDTA, 0.5% Nonidet P-40) at 4°C for 1 h with gentle rocking. Glutathione-Sepharose beads were washed five times with H buffer (20 mmol/liter HEPES, pH 7.7, 50 mmol/liter KCl, 20% glycerol, 0.1% Nonidet P-40). The bound proteins were eluted in Laemmli loading buffer and separated by SDS-PAGE. The gels were fixed in 25% isopropanol and 10% acetic acid, dried, and exposed to Biomax film (Kodak).
Co-immunoprecipitation Assays-10-cm dishes containing tsCOS cells were transfected with 2 g each of expression constructs encoding N-CoR, TEL, and FLAG-tagged HDAC9 isoforms using PolyFect (Qiagen). After 36 h the cells were washed in PBS and harvested, and the immunocomplexes were isolated using the Catch and Release immunoprecipitation system (Upstate Biotechnology, Inc.) according to the manufacturer's instructions. Co-immunoprecipitated HDAC9 isoforms were resolved by SDS-PAGE, blotted, and detected with anti-FLAG M2 antibody. Endogenously expressed proteins were co-immunoprecipitated as above and detected with affinity-purified anti-HDAC9 rabbit polyclonal antibody.
Double Immunofluorescence Microscopy-COS-7 cells growing on coverslips were transfected using PolyFect (Qiagen). After 36 h, the cells were fixed with 100% methanol at Ϫ20°C for 10 min and washed three times for 5 min in PBS. The cells were permeabilized with 0.5% Triton X-100 in PBS and washed three times for 5 min in PBS. The cells were blocked for 30 min at room temperature in 2% bovine serum albumin (Vector Laboratories) and 5% relevant serum (Jackson Immunoresearch Laboratories, Ltd.) for the secondary antibody to be used. The primary antibodies were diluted to the appropriate concentrations in blocking buffer, centrifuged at 4°C for 20 min, and then incubated at 4°C overnight. The cells were washed three times for 15 min in PBS and incubated with the appropriate fluorescein isothiocyanate-or TRITC-conjugated secondary antibody (Jackson Immunoresearch Laboratories, Ltd.) for 2 h at room temperature. The cells were washed three times for 15 min in PBS and mounted on slides using Vectashield with DAPI (Vector Laboratories). Imaging was performed on a Zeiss Axioplan2 microscope with a cooled CCD camera (Photometrics Quantix) using Smartcapture 2 software (Digital Scientific).
Confocal Immunofluorescence Microscopy-10 4 REH cells in log phase growth were cytospun and fixed in 4% paraformaldehyde for 15 min at room temperature. The samples were subsequently processed as described above. The cell nuclei were stained with To-pro-3 iodide (Molecular Probes), and slides were mounted with Vectashield (Vector Laboratories). Affinity-purified anti-HDAC9 rabbit polyclonal antibody was used for detection, and imaging was performed on a Leica TCS-SP2 system.
In Vivo Sumoylation Assays-In vivo sumoylation assays were carried out using the nickel affinity pull-down technique (49). Briefly, tsCOS cells in 10-cm dishes were transfected with 2 g each of expression constructs encoding F-HDAC9, F-HDAC9⌬exon12, and polyhistidine-tagged SUMO-1 or -2 using PolyFect (Qiagen). After 36 h the cells were washed in PBS and harvested directly in 1 ml of guanidine lysis buffer (6 M guanidine HCl, 100 mM NaCl, 10 mM Tris, 50 mM NaH 2 PO 4 , pH 8.0), sonicated, and centrifuged. The cleared samples were incubated for 2 h with 20 l (packed volume) of Talon nickel affinity beads (Clontech). The bound proteins were washed twice in lysis buffer, three times in urea buffer (8 M urea, 100 mM NaCl, 50 mM NaH 2 PO 4 , pH 6.5), and once in cold PBS before being eluted by boiling in Laemmli loading buffer. 20% of each guanidine lysis sample was removed and precipitated for 15 min on ice with 5% trichloroacetic acid. Samples were centrifuged, washed in 100% ethanol, and resuspended in Laemmli loading buffer. The proteins were separated by SDS-PAGE and detected with affinity-purified anti-HDAC9 rabbit polyclonal antibody.

RESULTS
Cloning of Full-length HDAC9 cDNA-The GenBank TM (50) high throughput genomic sequence data base (htgs) was searched with the amino acid sequences corresponding to the deacetylase domains of various Class I and II HDACs. Several DNA sequences encoding peptides with significant homology to HDAC5 were found on a human BAC clone RP11-8I15 (accession number AC016186), which contains 70 unordered contigs and maps to chromosome 18. When the GenBank TM nucleotide data base was searched with a composite of the novel sequences showing homology to HDAC5, it aligned exactly with BAC clone CTB-13P7 (accession number AC002088), which mapped to chromosome 7p21.1, indicating that clone RP11-8I15 had been submitted incorrectly to the htgs data base. A search of the expressed sequence tag data base (dbEST) revealed that cDNA corresponding to the putative histone deacetylase had been identified in germinal center B cells (accession number AA287983); therefore several hematopoietic cell lines were analyzed by RT-PCR with oligonucleotide primers specific for the putative HDAC sequence and also HDRP/MITR, which had been mapped to 7p15-p21 (35). In addition to the amplifying the putative HDAC sequence itself; it was found that this primer set also amplified cDNA from HDRP/MITR through to the putative HDAC domain. The Ensembl Genome Server (51) indicated an open reading frame (ORF) containing 22 exons (ϳ2700 bp of sequence), and RT-PCR was performed using oligonucleotide primers from within the known sequence of the putative HDAC9 cDNA established by us and the expected stop codon of the ORF, but no products were observed. These data, together with the observed conservation of C-terminal amino acid residues among other Class II HDACs indicated the absence of the entire ORF. 3Ј-Rapid amplification of cDNA ends was performed but failed to yield the remaining sequence, so GenBank TM was searched for overlapping expressed sequence tags to "walk" along the cDNA (Fig. 1B). This analysis revealed that the sequence corresponding to the final five exons of the putative HDAC gene contained in BAC clones RP5-1194E15 and GS1-465N13 (accession numbers AC004994 and AC004744, respectively) had been submitted in the antisense direction in relation to the rest of the gene (Fig. 1C). These clones accounted for almost 160 kb of DNA, and subsequent analysis with them in the correct orientation provided a genomic sequence that matched the ORF that we had generated from overlapping expressed sequence tags and, when translated, showed significant homology with the amino acid sequences for HDACs 4, 5 and 7 (see Fig. 3A). We designated this novel gene HDAC9. An ORF mapping to the HDAC9 locus has been independently cloned by another group (7); however, it does not encode the full-length protein (see further below).
The HDAC9 Gene Is Differentially Spliced to Encode Multiple Isoforms-The full-length product of the HDAC9 gene comprises 1069 amino acids as shown in Fig. 2 and is encoded by exons 2-26 (exon1 is untranslated). The 26 exons that form the HDAC9 cDNA span ϳ500 kb of genomic sequence on 7p21.1. The isoform that lacks the catalytic domain, HDAC9⌬CD (HDRP/MITR), is 593 amino acids in length and contains 16 residues of unique sequence encoded by a region of exon 12, which is 3Ј to the splice donor site used to generate the HDAC9 ORF (Fig. 2). There are several exon deletions that may occur that naturally preserve the open reading frame of HDAC9 (Fig.  1C), and two have been identified and cloned. It has been previously suggested that HDAC9 is potentially alternatively spliced at exon7 (7), and this is indeed the case. HDAC9⌬exon7 is 1025 amino acids long and contains an Ala 3 Glu substitution at position 222 as the result of the deletion of exon 7. This isoform lacks two serines (Ser 223 and Ser 253 ), which when phosphorylated have been implicated in 14-3-3 protein-dependent shuttling of HDAC4 and 5 from the nucleus to the cytoplasm, and also a tripartite nuclear localization signal (21,23,52). Exon 12 may also be deleted in-frame to generate an alternate protein isoform that is 981 amino acids long. The region encoded by exon 12 contains a conserved sumoylation site identified in HDAC4 (53) and a potential leucine zipper motif that may mediate interactions between HDAC9 and other proteins. cDNA that encodes an isoform possessing neither exon 7 nor exon 12 has also been cloned. HDAC9⌬exon15 is 1027 amino acids long and lacks a region within the catalytic domain adjacent to the active site. This isoform does not naturally conserve the ORF of HDAC9 and may have undergone RNA editing.
Phylogenetic analysis of HDAC9 shows that it is a member of the Class II histone deacetylases and is most closely related to HDAC5. The recent discovery of HDACs 10 and 11 indicate that, based on analysis of HDAC catalytic domain (Fig. 3B) and whole protein (Fig. 3C), there is a subdivision of the Class II histone deacetylase group consisting of HDACs 6, 10, and 11.
HDAC9 Possesses Deacetylase Activity-To determine whether HDAC9 possesses histone deacetylase activity, an in vitro assay was performed using anti-FLAG immunoprecipi-tated HDAC9. As shown in Fig. 4A, HDAC9 catalyzes the deacetylation of peptides corresponding to the N-terminal tails of both histone H3 and H4 with overall activity comparable with that of HDAC4. Additionally, the Class II HDACs 4 and 9 appear to deacetylate the histone H4 peptide substrate less effectively when compared with HDAC1. The value for deacety- FIG. 1. Isolation of HDAC9 cDNA and analysis of the HDAC9 gene. A, schematic representation of the two major HDAC9 isoforms together with the oligonucleotide primers (indicated as arrows) used in cloning of their respective cDNAs. B, identification of the 3Ј sequence of the HDAC9 open reading frame. The shaded letters represent the open reading frame derived by using overlapping expressed sequence tags (indicated by GenBank TM accession number) to walk along chromosome 7p21.1. C, genomic organization of HDAC9. The accession numbers of the clones that comprise the HDAC9 gene are shown together with the exon positions on chromosome 7, which were established using the BLAT alignment tool (84). HDAC9 exon/intron splice junctions are also detailed with consensus splice donor and acceptor sequences between exons (uppercase letters) and introns (lowercase letters) underlined. Exons highlighted in gray have the potential to be spliced out in-frame. ⌬exon7 and ⌬exon12 cDNAs have been detected by RT-PCR. lation of histone H4 peptide is 44% for HDAC4 and 36% for HDAC9, respectively, of the value observed for histone H3 peptide. This contrasts with HDAC1, which deacetylated histone H4 peptide with 70% of the activity seen for histone H3 peptide. These data are specific for a given HDAC, because a high stringency wash removes endogenous Class I and II HDACs present in the lysate that are not directly bound by anti-FLAG M2 agarose beads (Fig. 4B). Moreover, isoforms of HDAC9 that either lack (HDAC9⌬CD) or possess a deletion (HDAC9⌬exon15) in the catalytic domain do not deacetylate histone H3 peptide effectively (Fig. 4A). This indicates that co-immunoprecipitation of uncharacterized proteins possessing histone deacetylase activities or known HDACs at levels not detected by our analysis is not likely to account for the observed in vitro activity of HDAC9. The in vitro assay results are corroborated in vivo by chromatin immunoprecipitation analysis (Fig. 4C). There is a difference in the amount of target DNA sequence that may be co-immunoprecipitated with acetylated histone H3, compared with acetylated histone H4, when proteins containing a heterologous fusion of GAL4 DNA-binding domain to the catalytic domain of HDAC9 or to a lesser extent the whole HDAC9 protein are tested. This result is not observed for HDAC9⌬CD and is most likely a reflection of the ability of this N-terminal isoform of HDAC9 lacking the catalytic domain to bind Class I and II HDACs in vivo (54).
Distribution of HDAC9 mRNA Is Tissue-specific and May Be Deregulated in Human Cancers-Analysis of HDAC4, HDAC9, and HDAC9⌬CD expression in various human tissues and cell lines reveals that not only are there differences in the pattern of expression between HDAC4 and HDAC9 but also between the transcripts encoding the two major isoforms of HDAC9 (Fig. 5B). For example, all three transcripts were expressed in skeletal muscle and the adult and fetal brain (Fig. 5B, lanes 3  and 12), although HDAC9 and HDAC9⌬CD were considerably more abundant in fetal tissue. HDAC4 alone is expressed in kidney, liver, and the myeloid leukemic cell line HL60 (Fig. 5B,  lanes 7, 8, and 16). Although expression of HDAC9 was very low or absent in heart, T cell leukemia cell line MOLT-3, and early myeloid leukemia cell line KG-1, both HDAC4 and HDAC9⌬CD were co-expressed in these tissues (Fig. 5B, lanes  2, 14, and 17). To examine this further, we have carried out RT-PCR analysis for multiple HDAC9 isoforms in a large number of hematopoietic cell lines.
When HDAC9 and HDAC9⌬CD transcripts are analyzed for expression of exon 7 and 12 deletion splice variants in hematopoietic tissues and cell lines (Fig. 5C), mRNA encoding the FIG. 3. Phylogenetic analysis of HDAC9. A, amino acid sequence alignment of HDAC9, HDAC4, HDAC5, HDAC7, and a bacterial deacetylase, HDLP. The indicated sequences were aligned using Clustal W. Identical residues are boxed and highlighted in dark gray; similar residues are shaded in light gray. B, evaluation of amino acid identities and similarities of HDAC9 deacetylase domain compared with those of HDAC1 and other Class II histone deacetylases. The histone deacetylase domains are shown in light gray. The percentage values were obtained by comparing the deacetylase domain of HDAC9 with the indicated protein sequences on the BioEdit Sequence Alignment Editor using the Blosum62 matrix. C, phylogenetic tree of HDAC1 through to HDAC11. The sequences were aligned using the Clustal W server at the Center for Molecular and Biomolecular Informatics (University of Nijmegen, Nijmegen, Holland). The PHYLIP notation output was used to construct an unrooted tree with Unrooted (Manolo Gouy, University Claude Bernard, Lyon, France).
isoform lacking the catalytic domain appears to be expressed in many cell types. Although where co-expressed with full-length HDAC9, HDAC9⌬CD is considerably more abundant, we cannot exclude the possibility that this may be due to the efficiency of amplification. In normal tissues, HDAC9 transcripts encoding the catalytic domain were found at low levels in the bone marrow, spleen, and thymus (Fig. 5C, lanes 2-4). The highest levels, however, were observed in cells expressing CD14 ϩve (monocyte/macrophage) and, to a lesser extent, CD19 ϩve (B cell) surface markers (Fig. 5C, lanes 5 and 6). Further, inspection of the RT-PCR data reveal that HDAC9 is generally expressed in pre-B cell acute lymphoblastic leukemia cell lines FIG. 4. HDAC9 possesses histone deacetylase activity. A, HDAC9 deacetylates histone H3 and H4 peptide in vitro. 293T cells were transfected with FLAG-tagged HDACs as indicated. Whole cell lysates were produced, and the HDACs were precipitated with anti-FLAG agarose. The precipitates were thoroughly washed and assayed for their ability to deacetylate [ 3 H]acetyl-labeled peptides corresponding to the N terminus of histones H3 or H4. Free acetate was extracted, and the number of cpm were measured by liquid scintillation. The HDAC inhibitor trichostatin A was added to control reactions as indicated. The cells were transfected with pSG5 empty vector (Stratagene) as a negative control. This is taken to be the background level of deacetylase activity measured (dark gray). B, under high stringency conditions, HDAC9 does not co-immunoprecipitate with other previously characterized members of the histone deacetylase family. 293T cells were transfected with FLAG-tagged HDAC9 and immunoprecipitated as described above. The precipitates were then immunoblotted with antibodies raised against the indicated HDACs. Aside from HDAC9, input values for the indicated HDACs show levels of the endogenous enzyme. C, HDAC9 causes hypoacetylation of target genes in vivo. 293T cells were transfected with GAL4 uas x5-Tk-Luc reporter together with GAL4 DNA-binding domain fusions as indicated. GAL4DBD-HDAC9CD contains only the catalytic domain of HDAC9. The GAL4 reporter is derived from the pT109luc plasmid (39). This plasmid contains a SV40 origin of replication, which has been shown to induce chromatinization of plasmid DNA in cell lines expressing large T antigen (85). Soluble chromatin preparations from the transfections were immunoprecipitated with antibodies against acetylated histone H3 (␣AcHistone H3) or H4 (␣AcHistone H4) (Upstate Biotechnology, Inc.) and analyzed by semi-quantitative PCR. As negative and positive controls, soluble chromatin preparations were immunoprecipitated with IgG or anti-GAL4 (DBD) (RK5C1) antibody, respectively. Aliquots of the chromatin were also analyzed before immunoprecipitation (DNA input). The numerical values, which reflect relative abundance of acetylated chromatin, were obtained by densitometric analysis using Labworks analysis software (Ultra-violet Products). (Fig. 5C, lanes 7-9), B cell lymphoma cell lines (Fig. 5C, lanes  10 -13), and also the plasma cell line U-266 (Fig. 5C, lane 14). HDAC9 is also expressed in some T cell lines (Fig. 5C, lanes  15-18). HDAC9 is not expressed in various acute myeloid leukemia cell lines (Fig. 5C, lanes 19 -22), with the exception of KG1 (a multilineage lymphomyelocytic cell line) (55), which expresses HDAC9⌬CD. Two of four acute monocytic leukemia cell lines analyzed expressed isoforms of HDAC9, and low levels are also found in the erythroleukemia cell line, HEL (Fig.  5C, lane 27). Overall, the expression data suggest a selective although not absolute bias of HDAC9 expression toward lymphoid and monocytic cells within the hematopoietic system.
Samples from colon cancer cell lines showed increased expression of HDAC9 relative to normal colon tissue HDAC9 mRNA (Fig. 5C, lanes 31-34), especially variants lacking exons 7 and 12 in SW-620. In general, there is greater variety of expression of relative amounts of differentially spliced HDAC9 transcripts in samples containing the catalytic domain of HDAC9 than those lacking it. One exception is bone marrow (Fig. 5C, lane 2), where there is very little expression of HDAC9⌬CD transcripts containing exon 7.
HDAC9 Isoforms Interact Differentially with Co-repressors and Proteins Implicated in Hematalogical Malignancies-Previously, we and others have demonstrated involvement of histone deacetylases in the pathogenesis of both lymphoid and myeloid neoplasms (40,56,57). Given the expression of HDAC9 in cell lines and samples derived from B cell tumors, we examined whether it could interact with any oncogene products known to recruit HDAC-containing complexes and to be involved in B cell malignances. As anticipated, HDAC9 was found to associate in vitro with the TEL protein, which is fused to AML1 in pre-B cell childhood acute lymphoblastic leukemia (58,59), and also BCL-6, which is frequently associated with B cell neoplasias (60) (Fig. 6A, lanes 3 and 4). HDAC9 also interacted with HDACs 3 and 4 (Fig. 6A, lanes 1 and 2) and HDAC1 and PLZF (data not shown). In addition, HDAC9 was found to interact with the co-repressors mSin3A, mSin3B, and N-CoR, whose activities have been implicated in the mechanism of action of several human cancers (61) (Fig. 6A, lanes 5-8). These data indicate that N-CoR contacts HDAC9 at multiple points, FIG. 5. Tissue distribution of HDAC9 transcripts. A, the oligonucleotide primer sequences used for RT-PCR analysis are detailed together with a schematic representation of the primer positions in HDAC9 and HDAC9⌬CD (arrows). B, differential expression of Class II HDACs in normal tissues and leukemic cell lines. RT-PCR was performed on cDNA derived from normal human tissue (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13) and cell line (14 -19) total RNA as indicated. PCR products were resolved by agarose gel electrophoresis and visualized by ethidium bromide under UV light. C, HDAC9 isoforms are differentially expressed in the B cell lineage and cell lines derived from B cell malignancies. RT-PCR was performed on RNAs derived from normal human tissues (1)(2)(3)(4)(5)(6)(7)(8), primary CLL samples (9 -12), primary childhood acute lymphoblastic leukemia samples (13)(14)(15)(16), or hematopoietic cell lines  total RNA, as indicated. NLB (7)(8) refers to normal B cell. Where indicated (SAC/IL2), the cells were stimulated with Staphylococcus aureus Cowan I strain (1/5000) and interleukin 2 (50 units/ml). The identity of different isoforms (indicated on the left) was confirmed by isoform-specific and general oligonucleotide probes (data not shown).
with the C-terminal region of N-CoR also interacting with the catalytic domain of HDAC9. Additionally, the catalytic domain of HDAC9 was found to associate with mSin3A but not mSin3B.
The ability of HDAC9 expressed in vivo to interact with N-CoR or TEL was also examined. Consistent with the in vitro data, full-length HDAC9 isoforms were found to interact with N-CoR and TEL (Fig. 6B, lanes 12 and 17). However, TEL co-precipitated poorly with HDAC9⌬exon7 compared with N-CoR (Fig. 6B, lanes 13 and 18), indicating that the major site of interaction between TEL and HDAC9 lies within exon 7. Cotransfection of N-CoR or TEL with isoforms of HDAC9 lacking the catalytic domain produced a markedly weaker interaction relative to that observed for full-length HDAC9 when compared with the in vitro data (Fig. 6B, lanes 14, 15, 19, and 20). This suggests that the catalytic domain may be interacting with other proteins in vivo that act to stabilize the protein complex. Finally, the interactions between endogenously expressed HDAC9 and BCL-6, N-CoR, and TEL were examined in cell lines known to express these proteins (40, 62, 63) (Fig. 6C).
The results demonstrate that HDAC9 and the above factors can associate under physiological levels.
The in vitro and in vivo interaction data between HDAC9 and N-CoR and TEL are corroborated by specific patterns of cellular localization between various isoforms of HDAC9 and its interacting partners. Fig. 7A shows the cellular localization of various isoforms of FLAG-tagged HDAC9 when visualized either with anti-FLAG antibody or an antibody to an N-terminal epitope of HDAC9. Note that HDAC9⌬exon7 is completely excluded from the nucleus. As previously reported for the mouse homologue, HDAC9⌬CD punctate nuclear localization (64). The cellular distribution of HDAC9 isoforms containing the catalytic domain was also investigated (Fig. 7B) using an antibody raised against a peptide corresponding to a C-terminal region of HDAC9 (see "Experimental Procedures"). HDAC9 displayed a diffuse pattern of distribution within the nuclei of the REH cells and was also found in the cytoplasm, which corroborates RT-PCR data for this cell line showing expression of HDAC9⌬exon7 mRNA (Fig. 5C, lane 8).
To further analyze the interactions of HDAC9 and its iso-  35 S-radiolabeled proteins as indicated, precipitated with glutathione-Sepharose, and visualized by autoradiography. GST-HDAC9, GST-HDAC9CD, and GST-HDAC9⌬CD were produced from Escherichia coli DH5␣. [ 35 S]Methionine-labeled proteins were synthesized in vitro using a rabbit reticulocyte lysate-coupled transcription-translation system. B, the indicated FLAG-tagged isoforms of HDAC9 were co-transfected with N-CoR and TEL expression vectors. Total protein levels were normalized, and protein complexes were immunoprecipitated with anti-N-CoR (C-20) and anti-TEL (C-20) antibodies as indicated. Co-immunoprecipitated HDAC9 isoforms were resolved on a 5% SDS-polyacrylamide gel and detected by immunoblotting with anti-FLAG M2 antibody. The arrowhead indicates suspected degradation or incompletely translated products that appear with the FLAG-tagged HDAC9 expression vectors containing the catalytic domain. The cells transfected with pSG5 empty vector (Stratagene) were included as a negative control (lanes 1 and 10). C, endogenous HDAC9 was immunoprecipitated from NALM-6, RAJI, and REH cell lines with the indicated antibodies. A specific blocking peptide for anti-N-CoR (C-20) and an irrelevant antibody, anti-GAL4 (DBD) (RK5C1) were included as negative controls. Samples containing co-immunoprecipitated HDAC9 were resolved on a 7.5% SDS-polyacrylamide gel and detected by immunoblotting with affinity-purified rabbit serum containing antibodies raised against the C-terminal region of HDAC9.
forms in the context of the cell, we performed double immunofluorescence assays. When HDAC9 was co-expressed with N-CoR (Fig. 7C), they co-localize within the nucleus, and N-CoR assumes a more diffuse nuclear distribution that is identical to that seen with HDAC9 alone. However, when N-CoR was coexpressed with HDAC9⌬exon7, there was a dramatic change in FIG. 7. Immunofluorescence analysis of HDAC9. A, HDAC9 is alternatively spliced to generate multiple isoforms. COS-7 cells were transiently transfected with F-HDAC9, F-HDAC9⌬exon7, F-HDAC9⌬exon12, F-HDAC9⌬exon15, F-HDAC9⌬CD, or F-HDAC9⌬CD⌬exon7 as indicated. After methanol fixation, the cells were stained with DAPI (blue) and FLAG antibody (green). B, HDAC9 isoforms containing the catalytic domain are expressed endogenously in the childhood acute lymphoblastic leukemia cell line, REH. After fixation with 4% paraformaldehyde, the cells were stained with To-pro-3 iodide (blue) and affinity-purified anti-HDAC9 antibody. C, HDAC9 interacts with the co-repressor N-CoR. COS-7 cells were transiently transfected with N-CoR and F-HDAC9, F-HDAC9⌬exon7, F-HDAC9⌬CD, or F-HDAC9⌬CD⌬exon7 as indicated. After methanol fixation, the cells were stained with DAPI (blue) and anti-FLAG M2 (green) or N-CoR (C-20) (red) antibodies. D, HDAC9 interacts with TEL. The COS-7 cells were transiently transfected TEL and F-HDAC9, F-HDAC9⌬exon7, F-HDAC9⌬CD, or F-HDAC9⌬CD⌬exon7 as indicated. After methanol fixation, the cells were stained with DAPI (blue) and anti-FLAG M2 (green) or TEL (N-19) (red) antibodies. the distribution of the co-repressor to the cytoplasmic location of the HDAC9⌬exon7 isoform. Recruitment or retention of N-CoR to the cytoplasm was not observed upon co-expression with HDAC9⌬CD⌬exon7, indicating that the domains of interaction between HDAC9 and N-CoR lay within both exon 7 and the catalytic domain. This is supported by the in vitro data showing that the catalytic domain of HDAC9 interacts with the Cterminal region of N-CoR (Fig. 6A, lane 8). Similar results were also observed for BCL-6 (data not shown). In contrast to N-CoR, TEL was not excluded from the nucleus when co-expressed with either HDAC9⌬CD or the full-length HDAC9 lacking exon 7 (Fig. 7D), further indicating that the main domain of interaction between HDAC9 and TEL is mediated through exon 7. It appears from the above results that different HDAC9 isoforms display individual interaction profiles with various proteins, and this is reinforced by the observation that AML1 co-localizes exclusively with HDAC9⌬CD (data not shown), suggesting that the domain of interaction lies within the C-terminal sequence unique to this isoform.
HDAC9 Represses Transcriptional Activity in Vivo-It has been previously established that HDACs repress transcription when tethered to DNA as Gal4 fusion proteins (11). As expected from earlier experiments (Fig. 4, A and C), this effect is also observed with HDAC9, HDAC9⌬CD, and HDAC9CD alone (Fig. 8A). A GAL4 uas x5-TK-Luc reporter gene was transiently transfected into 293T cells together with the expression vectors for the indicated GAL4 fusion proteins. Although HDAC9⌬CD lacks an HDAC domain, it associates with other HDACs and co-repressors (Figs. 6 and 7), and this may be reflected in its ability to repress reporter gene expression.
The ability of HDAC9 to associate with transcriptional repressors, such as TEL for example, suggested that HDAC9 could play a role in repression of promoter activities by these proteins. To evaluate this for the TEL protein, HDAC9 and just the HDAC9 catalytic domain were co-transfected with TEL expression vector and a luciferase reporter gene containing three copies of the TEL consensus binding site (5Ј-TAAACAG-GAAGT-3Ј). As expected, the addition of HDAC9 potentiated repression by TEL, and the degree of repression observed was greater for TEL plus full-length HDAC9 than for TEL plus HDAC9⌬CD (Fig. 8B).
When Sparrow et al. (35) first identified the Xenopus homologue of HDAC9⌬CD, they reported that it associated with myocyte enhancer factor 2D (xMEF2D) and repressed xMEF2D-mediated transcription. In the human hematopoietic system, MEF2D is found in both B and T cells (65), and HDAC9 was tested for its ability to affect MEF2D-mediated transcription. As expected, co-transfection of HDAC9 with MEF2D abolished MEF2D-mediated transcriptional activation and repressed a MEF2 RE x3-Tk-Luc reporter. Co-transfection of HDAC9⌬CD with MEF2D produces a similar result, but the level of MEF2D-mediated activation was only returned to the basal level (Fig. 8C). This reflects the results observed with TEL promoter activity and indicates that full-length HDAC9 forms a more powerful repressor complex in this context.
Deletion of HDAC9 Exon 12 Results in the Loss of a Site Modified by Sumoylation-To test whether the different HDAC9 isoforms containing the catalytic domain are endogenously expressed protein, Western blot analysis was performed on whole cell lysates from tsCOS cells transfected with different isoforms of HDAC9 (Fig. 9, lanes 1-3) and selected hematopoietic and colon cancer cell lines that were used in RT-PCR analysis, plus the T-cell leukemia cell line JM-1 (Fig.  9, lanes 4 -9). The anti-HDAC9 rabbit polyclonal antibody used specifically recognizes endogenous proteins expressed in cells positive for HDAC9 mRNA isoforms (Fig. 5C). Although overall protein and mRNA levels are comparable between the different samples examined, the full-length protein isoform is the most abundant species. This could be due to decreased efficiency of translation and/or protein isoform stability. The Western analysis shows that HDAC9 has an apparent molecular mass of around 160 kDa, which is greater than the predicted molecular mass of 117.5 kDa and indicates that HDAC9 has undergone post-translational modifications. Interestingly, the HDAC9⌬- exon12 isoform, which only encodes 80 residues, shows an unexpectedly large decrease in apparent size (relative to ⌬7 and ⌬15 deletions), suggesting the presence of a site of posttranslational modification. This was thought most likely to be sumoylation, which has been detected on the corresponding region of HDAC4 (46,53). To examine this possibility, tsCOS cells were transfected with HDAC9 and HDAC9⌬exon12 together with polyhistidine-tagged SUMO-1 and -2. Upon nickel affinity precipitation of His-tagged protein complexes, only the full-length HDAC9 isoform containing exon 12 was visible as a SUMO-1 or -2 conjugate (Fig. 10, lanes 8 and 9). SUMO-2 may itself be sumoylated to form polymeric chains (53), and the multiple bands observed in lane 8 are due to endogenous SUMO-2 present in the cells. This is illustrated by the fact that the polymeric chains containing His-tagged SUMO-2 in lane 9 migrate slightly more slowly than those observed in lane 8, where only the terminal SUMO-1 may be His-tagged. Note that mono-sumoylated HDAC9 precipitates with markedly less efficiency compared with di-and tri-sumoylated HDAC9. This is most likely due to a combination the presence of multiple His tags and/or improved access to the tag. DISCUSSION In this manuscript, we describe the full and complete cloning and characterization of the ninth member of the histone deacetylase family, HDAC9. The HDAC9 gene is located at 7p21.1, a region implicated in neurological disorders (66 -69) and a variety of cancers including colorectal cancer (70), fibrosarcoma (71), childhood acute lymphoblastic leukemia (72), Wilms' tumor (65), and peripheral nerve sheath tumors (73). The 3Ј end of the HDAC9 ORF is located only approximately 150 kb from the TWIST gene, which is implicated craniosynostosis-associated Saethre-Chotzen syndrome (74 -76). Investigation of the genomic neighborhood of the TWIST gene showed that deletions of DNA that encompassed the HDAC9 locus led to a more severe phenotype with significant learning difficulties (68). The potential involvement of HDAC9 in central nervous system development and the pathogenesis of Saethre-Chotzen syndrome is corroborated by its expression in the developing brain.
The possibility that gene dosage effects involving HDAC9 are important is also underlined by the correlation of increased HDAC9 expression in colon cancer cell lines with the finding that around 37% of colon cancers possessed gains of DNA sequence corresponding to 7p21 (70). The HDAC9 gene comprises 26 exons and spans ϳ500 kb (more than 12 times the size of HDAC5 (77) and almost 40 times that of HDAC3 (78)), and this may render it more susceptible to the effects of genomic instability. Comparison of the mouse open reading frame with human HDAC9 shows a high degree of conservation (data not shown), including the potential for the same in-frame alternative splicing. This point is reinforced when a comparison is made of the noncoding genomic DNA contained within the mouse and human HDAC9 genes. There are zones of high homology present, especially in the regions adjacent to the regulated exons 7 and 12.
A sequence representing an incomplete open reading frame of full-length HDAC9 has been recently reported by Zhou et al. (7). It is possible that this sequence is a truncated isoform of HDAC9 lacking exons 25 and 26. Because sequences encoded by exons 25 and 26 are largely conserved among Class II HDACs and comprise the very C terminus of the catalytic domain, it is not surprising that the truncated HDAC9 product reported by Zhou et al. (7) possessed very little deacetylase activity (only 10% of the activity of HDAC4). This clearly contrasts with the activity level of the full-length HDAC9 protein reported here, which is close to that of HDAC4 (Fig. 4A).
Until recently, expression analyses indicated that both Class I and II HDACs existed as single isoforms (11)(12)(13)16). This is no longer the case with respect to Class II HDACs because major variants have been detected for HDAC7 and 10 (8, 15, 18 -20). It should be noted that Zhou et al. (7) also reported the presence of an alternate isoform of HDAC9 that consisted of a transcript lacking exons 21-26 and encoding 19 residues of unique sequence, in much the same way as HDAC9⌬CD. Recent studies on HDAC4 have shown that the region encoded by exon 7 contains residues that form a powerful nuclear localization signal and two serines, which when phosphorylated facilitate the binding of 14-3-3 signaling proteins (52). These residues are conserved in HDAC9, and proteins lacking the exon7 region may have a dramatic effect on the cellular localization of interacting factors such as N-CoR (Fig. 7C). The interactions between HDAC9 isoforms and interacting proteins such as FIG. 9. Expression of endogenous HDAC9 proteins. Whole cell lysates were prepared from tsCOS cells transfected with F-HDAC9, F-HDAC9⌬exon7, F-HDAC9⌬exon12, and selected cell lines as indicated. Total protein levels were normalized, and the samples were resolved on a 5% SDS-polyacrylamide gel. Western blot analysis was performed with affinity-purified rabbit serum containing antibodies raised against the C-terminal region of HDAC9. Enhanced chemiluminescence was used for detection.
FIG. 10. SUMO-1 and -2 are conjugated to HDAC9 but not HDAC9⌬exon12 in vivo. Whole cell lysates were prepared from ts-COS cells co-transfected with F-HDAC9, F-HDAC9⌬exon12, and His 6tagged SUMO-1 or -2 as indicated. pSG5 empty vector (Stratagene) was used as a control plasmid. Samples precipitated with trichloroacetic acid to show input protein levels (lanes 1-6) and nickel pull-downs (lanes 7-12) were analyzed by Western blotting using anti-HDAC9 rabbit polyclonal antibody. The open arrowheads indicate mono-, di-, and tri-sumoylated HDAC9. The solid arrowhead refers to a nonspecific band observed with the trichloroacetic acid precipitates.
BCl-6, TEL, or N-CoR merit more detailed investigation, but it seems likely that many of the interactions are mediated through multiple regions in HDAC9 that are specific for a given partner protein. This introduces a level of complexity with regard to HDAC9 association with partner proteins not thus far reported for other family members. It has been suggested that HDACs may act as recruiting centers for specific protein complexes as well as being functional enzymes (79), and these data would tend to be in line with this notion.
The other differentially spliced isoform investigated was HDAC9⌬exon12, which also contains potential functional domains such as a leucine zipper motif and a sumoylation site. The HDAC9⌬exon12 isoform cannot be sumoylated and may prove to exert an important and separate function because SUMO is believed to alter the interaction properties of its targets, often affecting their localization within the cell (80). Moreover, recent studies have shown that sumoylation can affect HDAC1 and HDAC4 catalytic activity (46,81).
In summary, the HDAC9 gene encodes multiple, functionally distinct, and differentially expressed protein isoforms. Gene regulation arising from alternative splicing is recognized as an increasingly important factor in the creation of proteome diversity (82,83). The genomic size and degree to which HDAC9 is regulated is unprecedented among other HDACs, and it perhaps reflects the complexity of a system controlling its activities and may indicate a wider role for its function than just histone and/or protein modification.