Role of hPHF1 in H3K27 Methylation and Hox Gene Silencing

HeLa nuclear proteins were separated into nuclear extract and nuclear pellet fractions as previously described (28). After solubilization, proteins derived from nuclear pellets were resuspended in buffer D (50 mM Tris-HCl [pH 7.9], 0.1 mM EDTA, 2 mM dithiothreitol [DTT], 0.2 mM phenylmethylsulfonyl fluoride [PMSF], and 25% glycerol) containing 20 mM ammonium sulfate (BD20) and loaded onto a DEAE52 column equilibrated with BD20. Proteins bound to DEAE52 column were step eluted with BD350 and BD500. The BD350 fraction was dialyzed against buffer C (50 mM Tris-HCl [pH 7.9], 0.2 mM EDTA, 1 mM DTT, 0.2 mM PMSF, and 20% glycerol) containing 100 mM potassium chloride (BC100) and loaded onto a phosphocellulose P11 column. The column was step eluted with BC300, BC500, and BC1000. The BC500 fraction was then dialyzed against BD20 and subjected to a DEAE-5PW column (TosoHaas). The bound proteins were eluted with a 10-column-volumn (cv) linear gradient from BD20 to BD500. HMTase activities were separated into two peaks on this column. The first peak fractions were pooled and adjusted to BD500 using saturated ammonium sulfate and were then loaded onto a phenyl-Sepharose column (Pharmacia). The bound proteins were eluted with a 15-cv linear gradient from BD500 to BD0. HMTase activities were again split into two peaks. The fractions from both peaks were pooled separately and further fractionated following the same scheme described below. The pooled fractions were dialyzed into BC50 and then subjected to a Sephacryl S300 gel filtration column (Pharmacia). The HMTase activities eluted with a relative molecular mass of between 670 and 443 kDa from both pools. The fractions containing the HMTase activities were pooled, dialyzed against buffer P (5 mM HEPES-KOH [pH 7.5], 0.2 mM EDTA, 40 mM KCl, 0.01% Triton X-100, 0.01 mM CaCl₂, 1 mM DTT, 0.05 mM PMSF, and 10% glycerol) containing 10 mM potassium phosphate (BP10), and loaded onto a hydroxyapatite column. The bound proteins were eluted with a 12-cv linear gradient from BP10 to BP600. Fractions containing HMTase activity were combined and dialyzed into buffer BC150. The bound proteins were then eluted with a 10-cv linear gradient from BC150 to BC500. Finally, the active fractions were subjected to an affinity column containing protein A beads coated with SUZ12 antibodies to purify the EED-EZH2 complex. For protein identification, the candidate polypeptides were digested with trypsin and identified as previously described (28).

hPHF1 cDNA was PCR amplified from an I.M.A.G.E cDNA clone, and the sequence was verified by DNA sequencing. The baculovirus construct was generated by insertion of the open reading frame of hPHF1 into the pFASTBAC vector (GIBCO) between the EcoRI and XhoI sites. The virus was generated and amplified according to the manufacturer's protocol. The baculoviruses for other components of the EED-EZH2 complex were previously described (4). The procedure for purification of the EED-EZH2 complex with or without hPHF1 was also described previously (4). The eluted complexes from the Flag affinity column were further purified over a gel-filtration Superose 6 column (Pharmacia).

Oligonucleosome, mononucleosome, and core histone substrates used for HMTase assays were purified from HeLa cells as described previously (4). Wild-type and mutant recombinant histone H3 were generated and purified as described previously (4). HMTase assays were performed essentially as previously described (28).

The histone methylation assay was performed with a series of reactions containing increasing concentrations of ³H-labeled S-adenosylmethionine (³H-SAM). The reactions were allowed to proceed for 15 min and then stopped by addition of sodium dodecyl sulfate (SDS) loading buffer. Histones were separated by SDS-polyacrylamide gel electrophoresis (PAGE). Gels were exposed to film, and histone methylation was quantified by scintillation counting of bands containing histone 3 (H3) excised from gels. A histone sample with a known amount of radioactivity (cpm) was run at the same time to calibrate the incorporated methyl groups.

Full-length hPHF1 cDNA was cloned into EcoRI and XhoI sites of pGEX-KG vector for the production of glutathione S-transferase (GST)-fusion protein. Antibodies against hPHF1 were generated in rabbit using hPHF1(230-567) as an antigen. Antibodies against SUZ12, H3K27me1, H3K27me2, H3K27me3, H3K4me2, and H3K9me3 have been previously described (4).

3T3 and GC1spg cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum at 37°C. The lentiviral vector pTY was requested from NIH AIDS Research & Reference Reagent Program and was modified by replacing LacZ with an internal ribosome entry site followed by either puromycin or enhanced green fluorescent protein (eGFP) for selection. The pTY-U6-mPcl1 small interfering RNA (siRNA) vector was cotransfected with pHP, pVSVG, and pCEP4tat into 293T cells with Superfect (Qiagen). The supernatant containing lentivirus was collected 36 h posttransfection and transduced into NIH 3T3 or GC1spg cells with 8 μg/ml Polybrene (Sigma). Stably transfected cells were selected in the presence of 2 μg/ml puromycin. To generate rescued cell lines, the stable KD cells were further transduced with lentiviruses expressing eGFP and Flag-mPcl1, which contains mutations on the siRNA targeting sequence. eGFP-positive cells were sorted by fluorescence-activated cell sorting and expanded for further analysis. RNAs were extracted from the above cell lines and analyzed by reverse transcription-PCR (RT-PCR) and real-time qPCR. The primer sequences for the Hox gene analysis were described previously (2). The oligonucleotides used to target mPcl1 mRNAs are RNA interference 1 (RNAi-1) sense primer GATGTGCTGGCCAGATGGA and antisense primer TCCATCTGGCGAGCACATC and RNAi-2 sense primer GGTCACCTCTGGGACTTCA and antisense primer TGAATGCCCAGAGGTGACC. The oligonucleotides used to target mEzh2 mRNAs are RNAi-1 sense primer GTATGTGGGCATCGAACGA and antisense primer TCGTTCGATGCCCACATAC. The primers used in the quantitative PCR (qPCR) were mPcl1 cDNA primers TGTTGTGTGTGTCGCTCTGA and AAATGTCCAGCATCCCAGTC and mEzh2 cDNA primers AACCCTGTGACCATCCACGGC and ATCAGACGGTGCCAGCAGTAAG. ChIP assays were performed with indicated antibodies as previously described (4). The primer pairs across the mouse Hox A10 genomic locus were designed by Array designer (Premierbiosoft) and listed as follows: 1, AAGTGTGTGAGCGAAAATTGTG and TCCAGCATTAACACAGTTTCAG; 2, TCTCCCAGGGATGGTGAATCTC and ACTTGCTACCAGCCTCACAGAC; 3, AGTAGAGGCAGCCGTTGTAGTG and TCCTGAGCCGTCCCTGTCTG; 4, GCATAGCCTCCTGGGTGTGG and AGGCTGAGCTGGGTTTGGG; 5, AAATGGCTGGGAAAAGGACTGC and GCCGATGATCAATGCCTGGATC; 6, TGGCCTCGACTTAACCTTCC and AACAAACACCAAGCAAACAGAC; 7, TCAGTGTCAAGTCCTGAATGGG and GAAGGATTTTAGCCAGGCAAGC; and 8, TCCATTTTATCCTGTCCACCAC and GTGGCCTAGCGGAGGACC. Platinum Taq polymerase (Invitrogen) and Sybr green master mix (Applied Bioscience) were used for PCR and real-time qPCR, respectively.

By monitoring HMTase activity, we previously purified and characterized an H3K27-specific methyltransferase EED-EZH2 complex (3). In the same purification, we noticed that the nucleosomal histone H3-specific methyltransferase activity split into two peaks on the DEAE-5PW column (Fig. 1B). From peak 2, we purified and characterized the EED-EZH2 complex, which contains EZH2, EED, SUZ12, RbAp48, and AEBP2 (3). To identify and characterize the enzymatic activity in peak 1, the proteins were fractionated on a phenyl Sepharose column, which further split the enzymatic activity into two peaks (Fig. 1C, top panel). Western blot analysis of the column fractions indicated that EZH2, SUZ12, RbAp48, and RbAp46 all cofractionate with the two enzymatic activities (Fig. 1C, bottom three panels). To define the composition of the two enzymatic activities, the samples were further fractionated sequentially on S300, hydroxyapatite, and MonoQ columns before being affinity purified using an anti-SUZ12 antibody column (Fig. 1A). Silver staining followed by mass spectrometry analysis of the affinity purified samples indicated that both enzymatic protein complexes contain EZH2, SUZ12, EED, RbAp48, and RbAp46. However, hPHF1 was only present in complex 1 (Fig. 1D).

Sequence analysis indicated that hPHF1 is highly related to the Drosophila PcG protein PCL. The fly Pcl gene was identified in a genetic screen as an enhancer of Polycomb (Pc), and both homozygous and heterozygous mutants exhibit lethal and homeotic phenotypes (7). Previous studies have demonstrated that the Drosophila PCL protein associates with the ESC-E(Z) complex during early embryogenesis and can directly interact with E(Z) through its PHD fingers (21, 26). In contrast to Drosophila, where only a single Pcl gene exists, three genes that code for protein homologs to fly Pcl have been identified in both human and mouse genomes (Fig. 1E). hPHF1 was mapped to chromosome 6p21.3 by fluorescence in situ hybridization and shows 42% similarity to Drosophila Pcl (5). Its closest mouse homolog, mPcl1, originally named Tctex3 (T-complex testis expressed 3), is expressed specifically in the testis (12), but its function remains unknown. mPcl2 was initially cloned as a metal response transcription factor, Mtf2/M96 (11). Loss of functional mutation in mice causes posterior transformation of axial skeletons with low penetrance that is less severe than the defects exhibited with Pcl mutation in Drosophila, indicating that there might be functional redundancy among the Pcl homologs in mice (30). hPCL3 was first identified in humans and has been shown to be upregulated in multiple cancers (31). The function and expression pattern of mPcl3 remain to be characterized.

To assess the potential role of hPHF1 in modulating the enzymatic activity of EED-EZH2 complex, we reconstituted the EED-EZH2 complex in the presence or absence of hPHF1 following the scheme outlined in Fig. 2A. A homogenous EED-EZH2 complex with hPHF1 was purified by affinity chromatography followed by gel filtration to remove unincorporated free Flag-EED and Flag-EED-containing subcomplexes (Fig. 2B). Silver staining and Western blot analysis of the column fractions confirmed that hPHF1 copurifies with other components in a protein complex between 440 and 670 kDa, which behaves similarly to the native complex (Fig. 1A). Using the same strategy, we also reconstituted and purified a complex without hPHF1 (Fig. 2C). We note that hPHF1 is not stained well by silver (Fig. 2B) but is readily detected by Coomassie staining (Fig. 2C).

Previous studies indicated that the EED-EZH2 complex prefers oligonucleosome substrates (4). To test whether incorporation of hPHF1 in the EED-EZH2 complex affects its substrate preference, equal amounts of histone H3 in various forms were subjected to methylation by equal amounts of EED-EZH2 complex in the presence or absence of hPHF1. The results shown in Fig. 2D indicate that incorporation of hPHF1 might increase its relative activity for oligonucleosome substrates. In addition, a time course experiment demonstrated that incorporation of hPHF1 into the complex improved the kinetics of the methylation reaction (Fig. 2E). However, incorporation of hPHF1 does not change the site specificity (H3K27) of the EED-EZH2 complex (data not shown).

To understand the exact effect of hPHF1 on the reaction kinetics, we performed methyltransferase reactions using a wide range of SAM concentrations and allowed the reaction to proceed for only 15 min to keep it in the linear range (Fig. 3). The maximal velocity (V_max) and the Michaelis-Menten constant (K_m) were then derived from the Lineweaver-Burk plot (or double-reciprocal plot) (Fig. 3, middle panels). Based on the calculated V_max and K_m, accurate Michaelis-Menten plots were generated. This analysis revealed that incorporation of hPHF1 increased the reaction V_max from 70 to 189 methyl histones/min/enzyme, while it decreased the K_m from 0.88 to 0.60 μM. Together, these data allow us to conclude that hPHF1 stimulates the activity of EED-EZH2 complex by increasing the V_max 2.7-fold and decreasing the K_m 1.5-fold.

To assess the function of hPHF1 in H3K27 methylation in vivo, we used a lentivirus-derived siRNA approach in NIH 3T3 cells to generate two stable knockdown cell lines, KD1 and KD2, which target the N-terminal or C-terminal region of mPcl1, respectively. Semiquantitative and real-time RT-PCR analysis demonstrated over 60 to 70% reduction in the mRNA levels of the two knockdown cell lines (Fig. 4A). Because mPcl1 only presents in a subpopulation of the Eed-Ezh2 complexes, we also generated a stable Ezh2 knockdown cell line using a similar strategy for purposes of comparison. Characterization of the Ezh2 knockdown cell line by semiquantitative and real-time RT-PCR indicates that about 70% knockdown at the mRNA level was achieved (Fig. 4B). To evaluate the respective effects of mPcl1 and Ezh2 knockdown on Hox gene expression, we analyzed the expression levels of all 39 mouse Hox genes in the knockdown cells and compared them to that in the control cells by RT-PCR. Due to the low expression levels of some Hox genes in NIH 3T3 cells, we were able to detect expression of only 18 Hox genes. Results shown in Fig. 4C indicate that Ezh2 knockdown resulted in a broad and significant change in Hox gene expression. In contrast, knockdown of mPcl1 only affected expression of a limited number of genes. Interestingly, most genes that are affected by mPcl1 knockdown are down-regulated, but HoxA10 is up-regulated in response to Ezh2 and mPcl1 knockdown. Given that PcG proteins are known to be involved in transcriptional silencing, genes that are down-regulated by Ezh2 or mPcl1 knockdown might be a result of indirect effects.

Because very few Hox genes analyzed are affected by mPcl1 knockdown in NIH 3T3 cells, it is possible that NIH 3T3 is not the appropriate cell type for analysis of mPcl1 function. Since mPcl1 (also named Tctex3) was first identified due to its unique and restricted expression pattern in male germ cells (12), we speculated that knockdown of mPcl1 in male germ cells might cause a broader effect on Hox gene expression. Therefore, we generated knockdown of mPcl1 in GC1Spg cells, an immortalized mouse male germ cell line that constitutively expresses the simian virus 40 large T antigen (15), using the KD2 lentivirus-delivered siRNA. Expectedly, more efficient knockdown was achieved in this cell line, with a more than 80% reduction in the mRNA level as determined by real-time PCR (Fig. 5A). Interestingly, the expression levels of multiple Hox genes were altered and most of them, including HoxA10, are upregulated by knockdown of mPcl1 (Fig. 5B and C). Because a higher knockdown efficiency for mPcl1 and a broader effect on Hox gene expression are observed in the GC1spg cell line, we performed further analysis of mPcl1 function in this cell line. Because mPcl1 knockdown resulted in significant up-regulation of the HoxA10 gene in two different cell lines, subsequent analysis was focused on this gene.

To exclude possible off-target effects of mPcl1 knockdown and to overcome the problem that no Pcl1 antibody capable of working for ChIP is available, we generated a cell line that stably expresses Flag-mPcl1 using the GC1spg-derived Pcl1 knockdown cell line (Fig. 6A). RT-PCR analysis indicated that the expression of Flag-mPcl1 restored the mPcl1 level to that of the control (Fig. 6B, middle panel). Importantly, rescue of mPcl1 knockdown resulted in recovery of HoxA10 silencing (Fig. 6B, top panel), demonstrating that mPcl1 plays an important role in HoxA10 silencing. To address whether mPcl1 knockdown affects the global H3K27 methylation level, we compared the H3K27 methylation levels of histones isolated from control, knockdown, and rescued cell lines. Results shown in Fig. 6C indicate that mPcl1 knockdown results in a global decrease in H3K27me2 and H3K27me3 levels but has no obvious effects on the levels of other modifications, such as H3K27me1, H3K4me2, or H3K9me3 (Fig. 6C, lanes 1 and 2). Quantification of the methylation levels indicates that knockdown of mPcl1 resulted in decreases of genome-wide H3K27 di- and trimethylation levels to 30% and 70% of the control levels, respectively (Fig. 6D). Apparently this effect is directly related to mPcl1 knockdown, as rescue of mPcl1 expression by Flag-mPcl1 restored the H3K27 methylation levels (Fig. 6C and D).

To further characterize the relationship between mPcl1 binding, H3K27 methylation and HoxA10 silencing, we performed ChIP analysis across the entire HoxA10 gene using control, mPcl1 knockdown, and Flag-mPcl1-rescued cells. Antibodies used for ChIP assays include anti-Suz12, anti-Flag, and all three forms of anti-H3K27me. An equal amount of IgG was used as a control for antibody specificity. Results shown in Fig. 7B indicate that Flag-mPcl1 is localized to the promoter and upstream flanking region (amplicons 3 to 5; bottom panel). H3K27me2 and H3K27me3 levels are also enriched in the same region in the Flag-mPcl-rescued cells compared to that in the knockdown cells. However, H3K27me1 signal was not detected in all regions examined. Interestingly, knockdown of mPcl1 resulted in a significant decrease in H3K27me2 levels, while only a mild decrease in H3K27me3 levels is observed (Fig. 7B, compare left and middle panels). Notably, in region 4 and the surrounding region, H3K27 dimethylation and trimethylation were decreased to less than 40% and 70%, respectively (Fig. 7C). Importantly, the decreased H3K27 methylation caused by mPcl1 knockdown was restored to the control level in the rescued cells (Fig. 7B and C). Moreover, knockdown of mPcl1 appears to cause a subtle, but consistent, decrease in binding of Suz12 to the promoter region (Fig. 7B, amplicons 4 and 5, and C), which is also restored in the rescue cells (Fig. 7B, amplicons 4 and 5, and C). mPcl1 binding appears to correlate with Suz12 recruitment and H3K27 methylation. This raises the possibility that mPcl1 may contribute to Eed-Ezh2 recruitment to a subset of Hox genes. This possibility is consistent with a previous study demonstrating a direct interaction between the PHD domains of Drosophila PCL and E(Z) (21).