A Novel Plant Homeodomain Finger 10–Mediated Antiapoptotic Mechanism Involving Repression of Caspase-3 in Gastric Cancer Cells

Although the incidence of gastric cancer has declined over the past 2 decades, the disease has a high death rate (700,000 per year), making it the second most common cause of cancer mortality (1). For advanced carcinomas, therapeutic options are limited to radiation therapy and chemotherapy, and these modalities do not lead to a cure. The identification of biomarkers that distinguish cancer cells from normal cells is important for the identification of therapeutic targets in human malignancies. Several strategies have been developed to identify relevant tumor antigens that can be used for active immunotherapy strategies in tumor patients. Recently, a novel method termed serologic analysis of antigens by recombinant expression cloning (SEREX; ref. 2) has permitted direct molecular determination of new tumor antigens that elicit an IgG antibody response in tumor patients. This method has enabled the discovery of several novel genes with tumor specificity, such as MLAA-2 (3), CAGE (4), and T21 (5). Using sera from cancer patients, we have previously reported the identification of immunologically recognized proteins, such as plant homeodomain (PHD) finger 10 (PHF10; refs. 6, 7)and FRZB (8–10), which belong to the zinc finger protein family and are likely candidates for cancer-specific immunotherapy.

The PHD is also called the leukemia-associated protein domain, which was first identified as HAT3.1, a protein involved in plant root development (11). Similar PHD domains have been found in >400 eukaryotic proteins, including transcription factors and other proteins implicated in chromatin-mediated transcriptional regulation (12, 13). Proteins from the PHD zinc finger superfamily are well documented to be capable of translocating to the nucleus and regulating transcription. In the PHF family, the PHD class of transcription factors (e.g., RING, PHF1, PHF2, PHF11, and HBXAP) has been shown to be particularly interesting in cancer and autoimmune diseases for their transcriptional function on targeted proteins. These factors are expressed in the nucleus and act in cell proliferation and differentiation in a variety of diseases (14, 15). Potential clues about the relationship between aberrant expression of PHF10 and cancer were observed in a series of studies that found prevalent overexpression of PHF10 in many types of cancers, including colon cancer (16). However, the molecular event in the progression of gastric cancer that involves PHF10 remains unclear.

In this study, we explored the role of PHF10 in the malignant phenotype of gastric cancer and mapped the relationship between PHF10 and its target molecules, which may offer new therapeutic avenues for the treatment of this severe disease.

Two gastric cancer cell lines, MKN45 and MKN28, were obtained from the Japanese Cancer Research Resources Bank, and the other gastric cancer cell lines were obtained from the American Type Culture Collection. GES-1, an immortalized gastric epithelial cell line, was a gift from Prof. Feng Bi (Institute of Digestive Disease, Xi'jing Hospital). Mouse monoclonal antibodies against cytochrome c; β-actin; mouse polyclonal antibody against PHF10; rabbit monoclonal antibodies against procaspase-3, Bid, caspase-8, poly(ADP-ribose) polymerase (PARP), NF-κB, survivin, FasL, Bcl-2, and Bax; and rabbit polyclonal antibody to active+procaspase-3 were purchased from Abcam. The rabbit polyclonal antibody again caspase-9 was from BD Biosciences. Monoclonal mouse anti-actin and anti–glyceraldehyde-3-phosphate dehydrogenase (GAPDH) antibodies were purchased from Sigma.

RNA isolation was done using the Trizol reagent according to the manufacturer's instructions. Total RNA was then subjected to quantitative reverse transcription-PCR (q-RT-PCR) analysis. q-RT-PCR was done using an iCycler IQ Real-Time PCR Detection System (Bio-Rad). Reactions contained 5 pmol forward and reverse primers, 1× iQ SYBR Green Supermix (Bio-Rad), and 2 μL template cDNA from investigated cell lines. The raw quantifications were normalized to the GAPDH values for each sample. The primer sequences used in the q-RT-PCR analyses are available in Supplementary File 1.

The M-PER reagents and Halt Protease Inhibitor Cocktail kits (Pierce Biotechnology) were used for cell extract preparation according to the manufacturer's instructions. The protein concentration of the cell extracts was quantified using the bicinchoninic acid protein assay (Pierce Biotechnology). Western blot analysis was carried out as previously described (17, 18). Labeled bands were detected by the ECL chemiluminescent kit (Pierce Biotechnology). Images were captured and the intensity of the bands was quantitated with the VersaDoc image system (Bio-Rad).

SGC7901 cells were washed twice with PBS, fixed, and permeabilized with 4% formaldehyde and 0.5% Triton X-100 in PBS for 10 minutes. After washing with PBS, cells were blocked with 1% bovine serum albumin and incubated with mouse polyclonal anti-PHF10 (1:50). After an incubation period of 1 hour, cells were washed thrice with PBS and incubated for 1 hour with secondary antibody (FITC-conjugated rabbit anti-mouse; 1:500; Sigma). Following three washes with PBS, cells were incubated with an actin-specific marker, phalloidin (Sigma). Coverslips were washed thrice, mounted using mounting medium with 4′,6-diamidino-2-phenylindole (DAPI; Molecular Probes), and viewed using a Leica TCS-SP2 confocal system.

By amplifying cDNA of SGC7901 using the hPHF10-F1 (5′-TACAAGCTTCTTCAAGAACAAGTCAGTGAA-3′) and hPHF10-R1 (5′-TACGAATTCTTATCCCTCTTTGCTGT-3′) primers, full-length PHF10 was obtained and inserted into plasmids, pFLAG-CMV4 and pCMV-BD, to generate fusion proteins. Alignment done in the European Molecular Biology Laboratory protein database showed that the MYST-RELATED PROTEINS subdomain (or PTHR10615:SF8), which ranges from +44 to +391, is evolutionarily conserved among various genes for 2PHD-containing proteins and across different species. Consequently, we hypothesized that this region may be required for PHF10 to exert transcriptional regulation. The remainder of the sequence in the MYST-RELATED PROTEINS subdomain upstream of 2PHD domains was designated as the LEAD domain (Fig. 3B). A series of FLAG-tagged motif fusion proteins were generated using pFLAG-PHF10 expression plasmids as the template. All resulting constructs were designated as indicated in Fig. 4B. The GAL4/LexA promoter activity assay was done using a pL8G5-luc plasmid with or without cotransfection of a pLexA-VP16 construct (17, 18).

The putative 700-bp caspase-3 promoter sequence was amplified by PCR using human genomic DNA as the template as described in previous studies (19). The promoter sequence was cloned into pGL4-Basic with Xhol and HindIII sites added to each end (Promega). In addition, differently sized fragments of the caspase-3 promoter, which were generated by progressively deleting from the 5′ end, were cloned into the promoterless pGL4-Basic luciferase reporter vector. The sequences of all promoter fragments derived by PCR were confirmed by sequencing. The luciferase plasmids were transformed into JM109-competent cells (Promega) using the standard protocol for mass production. Short hairpin RNA (shRNA) specific to PHF10 mRNA was designed and prepared as described previously (7) using Psilencer 2.0 vector (Psi-V). The shRNA used in this study is relatively the most effective one after examined via q-RT-PCR. All constructs were verified by sequencing.

Cells were transfected at 60% to 70% confluence with pFLAG-CMV4 empty vector (pFLAG-V), pFLAG-CMV4-PHF10 (pFLAG-PHF10), Psilencer2.0 vectors carrying shRNA for PHF10, and the scrambled sequence using Lipo2000 (Roche) according to the manufacturer's protocol. Cells were cultured in medium supplemented with 1 mg/mL G418 (Promega) for 4 weeks and then maintained in medium containing 350 μg/mL G418. Stably transfected clones were then picked. Psilencer2.0-1333 (Psi-1333) was selected as the most effective vector used for PHF10 knockdown. Clones that had been stably transfected with pFLAG-V and Psilence2.0-scrambled were used as controls.

Animal studies were conducted with the approval of the Committee on the Ethics of Animal Experiments at our institution. To examine the growth rate of gastric cancer cells expressing upregulated and downregulated PHF10 in animals, 4 × 10⁶ stably transfected cells in 0.2 mL PBS were injected s.c. into the dorsal flank of 5-week-old male severe combined immunodeficient mice. Mice were monitored weekly, and the tumor volume was measured with a linear caliper. Tumor volume was calculated using the following equation: volume = (width + length)/2 × width × length × 0.5236. Mice were sacrificed at 5 weeks after inoculation of the cells. All tumor grafts were dissected, weighed, harvested, fixed, and embedded. Histologic procedures and terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) assays were done as described previously (7). All experiments were repeated twice.

To detect DNA fragmentation, we first did DNA fragmentation assays. Briefly, cells were collected at indicated times, suspended in lysis buffer [10 mmol/L Tris-HCl (pH 8.0), 10 mmol/L NaCl, 10 mmol/L EDTA, 100 mg/mL proteinase K, 1% SDS], and incubated overnight at 55°C. DNA was extracted with phenol/chloroform (1:1) followed by precipitation with 0.3 mol/L sodium acetate in ethanol. DNA was then washed with 70% aqueous ethanol. The DNA pellet was resuspended in TE buffer [10 mmol/L Tris-HCl (pH 8.0), 1 mmol/L EDTA] and incubated for 2 hours at 37°C with 0.5 mg/mL DNase-free RNase (Roche). After phenol/chloroform reextraction by the same method, DNA was analyzed on a 1% (w/v) agarose gel supplemented with ethidium bromide. Fluorescence-activated cell sorting (FACS) analysis for each transfectant was done as described previously (20). Additionally, to detect the morphologic changes in apoptosis, we conducted TUNEL assays on cells seeded on coverslips using the Apo-BrdU In Situ DNA Fragmentation Assay kit according to the manufacturer's instructions. Similar to confocal analysis, phalloidin and DAPI are used to stain actin and DNA, respectively, and anti-bromodeoxyuridine-FITC was used to detect Br-dUTP ligated to the cleaved DNA.

Cells were cotransfected with L8G5 luciferase, pLexA-VP16, and pCMV-BD-PHF10 or other PHF10 truncated bodies according to the protocol provided for the Dual Luciferase kit (Promega) to investigate the effect of PHF10 on transcriptional activity. Determination of luciferase activity was done with a luminometer (Turner Designs TD-20/20). To precisely determine the region involved in the interaction between the caspase-3 promoter region and PHF10, cells were also cotransfected with fragments of pGL4 caspase-3 and other truncated pFLAG-PHF10 fusion constructs along with pRL-TK to investigate the effects of individual domains of PHF10 on the transcriptional activity of caspase-3.

For these experiments, SGC7901 and MKN28 cells were subjected to chromatin immunoprecipitation (ChIP) with the ChIP Assay kit (Upstate Cell Signaling Solutions). Briefly, cross-linking of proteins with DNA was done with 4% formaldehyde at 37°C for 15 minutes and quenched with glycine. Cell lysates were sonicated (Branson Sonifier) to shear the DNA to 400- to 1,000-bp length fragments. Chromatin samples were then precleared with a salmon sperm DNA/protein A agarose 50% slurry for 30 minutes at 4°C and immunoprecipitated overnight in the absence of antibody or with an anti-PHF10 antibody or an anti-FLAG antibody (M5; Sigma). The 3′-untranslated region (3′-UTR) region of the caspase-3 gene was amplified as a control for the ChIP assay using the following primer set (designated as P1): 5′-TCCCAAGTCCTCGCAGAA-3′ (sense) and 5′-TGGCCTTAAAGAACTCATT-3′ (antisense). The region between −270 and −170 nucleotides of the caspase-3 promoter was amplified with two pairs of primers (designated as P2 and P3): P2, 5′-TTTCCAAGTCTCCCTCAATT-3′ (sense) and 5′-GGATTTGAAATCTAGG-3′ (antisense); P3, 5′-GGAAGACCTAGATTTC-3′ (sense) and 5′-CGTCTGCACTGCTTCCG-3′ (antisense). The PCR products were separated on a 2% agarose gel, stained with ethidium bromide, and visualized under UV light.

Each experiment was done independently at least twice with similar results. Findings from one representative experiment are presented. Results are expressed as mean ± SD. Significant differences from in vitro and in vivo experiments were assessed with the Student's t test (two-tailed). Analysis was done using the Statistical Package for the Social Sciences statistical software (version 11.05; SPSS, Inc.). A P value of <0.05 was deemed significant.

The expression levels of PHF10 in SGC7901 and MKN28 cells were drastically different (data not shown). SGC7901 cells, which were derived from a poorly differentiated gastric cancer, had a relatively higher abundance of PHF10 protein than MKN28 cells, which were derived from a lesion with moderate malignancy. To explore the role of PHF10 on cell viability, vectors were used to reduce or overexpress PHF10 expression in SGC7901 and MKN28 cells (SGC790^siPHF10 and MKN28^PHF10), respectively. As shown in Fig. 1A, knockdown of PHF10 expression in SGC7901 cells for 10 or 12 hours dramatically increased cell apoptosis as determined by internucleosomal DNA cleavage. Moreover, DNA cleavage was further promoted by H₂O₂ treatment in SGC7901^siPHF10 cells (Fig. 1A). Similar results were also observed using a FACS assay of apoptosis by propidium iodide and Annexin V staining (Fig. 1B), consistent with TUNEL assays (Supplementary Fig. S1). These results suggest that the reduced expression of PHF10 induced apoptosis in SGC7901 cells. Next, to determine whether expression of PHF10 plays a role in gastric cancer progression in vivo, SGC790^siPHF10, MKN28^PHF10, and respective controls were s.c. implanted into the dorsal flanks of nonobese diabetic/severe combined immunodeficient mice. Tumors generated from SGC7901^siPHF10 were significantly smaller than tumors derived from the control cells (490 ± 150 mg versus 1,480 ± 260 mg; Fig. 1C). In contrast, tumor growth was dramatically increased in tumors established from MKN28^PHF10 cells that had overexpressed PHF10 compared with MKN28^control (850 ± 370 mg versus 1,750 ± 150 mg; Fig. 1C). To determine whether tumor growth inhibition in SGC7901^siPHF10 is due to the induction of apoptosis in vivo, TUNEL assays of tumor tissues were done. The results revealed that knockdown of PHF10 expression induced apoptosis in SGC7901 cells (6 ± 1 versus 23 ± 2 cells per field). In parallel, overexpression of PHF10 decreased apoptosis of MKN28 cells compared with the control cells (7 ± 2 versus 1 ± 2 cells per field; Fig. 1D).

Based on these in vitro and in vivo studies, we further identified potential downstream targets of PHF10. Analysis of protein profiles by Western blotting assays of SGC7901^siPHF10 cells showed that, in line with previous findings, procaspase-3 expression was decreased. This decrease in expression was followed by generation of a 17-kDa active form, caspase-3, and cleavage of its substrates, PARP, which appeared with significantly reduced intensity or derivation of additional bands, as expected (Fig. 2A, right and left). However, procaspase-9 displayed less obvious but detectable degradation. As shown in Fig. 2A, there was a slight but detectable accumulation of procaspase-8, which activates Bid and facilitates cytochrome c release; however, caspase-8 substrate, Bid, displayed no significant change. In addition, the expression levels of Fas-L, Bcl-2, cytochrome c, NF-κB, and survivin were not significantly altered.

Because siPHF10 could induce apoptosis in SGC7901, we sought to determine whether caspase inhibitors antagonized the effect of siPHF10. For this experiment, cells in the exponential growth phase were transiently transfected with shRNA for PHF10 or the control vector and sequentially treated with caspase inhibitors. Treatment of cells with a caspase-3 inhibitor, z-DQMD-fmk (Sigma), or a combination of z-DQMD-fmk and a caspase-8 inhibitor, Z-IETD-fmk (Calbiochem), at indicated doses significantly reduced the percentage of apoptotic SGC7901^siPHF10 cells as determined by FACS analysis (Fig. 2B). However, Z-IETD-fmk alone did not completely inhibit apoptosis, which suggests that caspase-8 might also contribute to apoptosis in SGC7901^siPHF10 cells. In addition, inhibitors for caspase-1, caspase-2, caspase-5, caspase-6, and the proteasome were ineffective at rescuing the time-dependent loss of viability induced by shRNA (data not shown).

Time-dependent procaspase-3 accumulation and the resultant generation of caspase-3 and cleavage of caspase-3 substrates were evident in SGC7901 ^siPHF10 cells without pretreatment with z-DQMD-fmk, whereas the opposite results were observed in MKN28^PHF10 cells. In the context of z-DQMD-fmk, procaspase-3 was significantly accumulated but with little alteration in its substrates in SGC7901 ^siPHF10 cells (Fig. 2C; Supplementary Fig. S2A), except for caspase-9. We found that caspase-9 varied time dependently in nonpretreated cells, whereas this trend disappeared when SGC7901 ^siPHF10 cells are pretreated with z-DQMD-fmk. The variation trend might come from either transcriptional repression or induced cleavage by overexpressed caspase-3, consistent with the fact that caspase-9 acts as one substrate of caspase-3. However, we also noticed that enforced expression of PHF10 in MKN28 cells (Fig. 2C) led to the downregulation of procaspase-9 in a time-dependent manner. Other than transcriptional regulation, one explanation may be that PHD finger proteins could function as an E3 ligase in the ubiquitination complex although this depends on further conclusions (21).

To investigate the effect of PHF10 overexpression on stimuli-induced apoptosis, MKN28 cells were transfected with pFLAG-PHF10 for 12 hours, washed thrice, and then exposed to Fas, H₂O₂, or etoposide. As shown in Supplementary Fig. S2B, pFLAG-PHF10 transfection significantly decreased the ratio of the apoptotic MKN28^PHF10 cells induced by Fas, H₂O₂, or etoposide, whereas no effect on cellular apoptosis was detected in pretreated MKN28^control cells, which further confirms that PHF10 might antagonize the common apoptosis-related molecule caspase-3.

As shown in Fig. 3A (top), PHF10 protein labeled with anti-mouse FITC was heterogeneously distributed in the nucleus. To examine the potential function of PHF10 in transcriptional regulation, we cotransfected 293T cells with pCMV-BD-PHF10, pL8G5-luc, and pLexA-VP16. As shown in Fig. 3A (middle), coexpression of GAL4-PHF10 with pLexA-VP16 inhibited VP16-activated L8G5 luciferase activity by 75%, suggesting that PHF10 may function as a transcriptional repressor. Together, these observations provide evidence to indicate that PHF10 may function as a transcriptional repressor of caspase-3.

As shown in Fig. 3B, expression of GAL4-BD-P2 (without tail) inhibited L8G5 luciferase activity by ∼89%, similar to the full-length protein GAL4-BD-P1, suggesting that tail sequence functioned little in transcriptional activity. Repression of GAL-BD-P3 (without the second PHD domain) on pL8G5 was slightly but more significantly presented than both GAL-BD-P4 (without both PHD domains) and GAL-BD-P5 (head without domains). These observations confirmed that both terminal sequences beside the tandem repeat three domains exerted almost no function in transcriptional repression. Similar results were obtained from experiments that used GAL-BD-P3, GAL-BD-P6, GAL-BD-P7, GAL-BD-P8, GAL-BD-P15, GAL-BD-P16, GAL-BD-P17, and GAL-BD-P18, suggesting that the two PHD motifs and the leading chain (LEAD domain) together represent the basal repression domain of PHF10. In addition, the observation that combination of any one single PHD domain and the LEAD domain did not contribute as much as the tandem repeat containing three domains in transcriptional suppression suggests that the second PHD domain exerts a greater effect on luciferase activity repression than the first PHD domain. That no obvious repressive activity was observed from GAL-BD-P9 to GAL-BD-P14 further suggested that deletion of any one PHD domain and/or the LEAD domain would completely abolish the transcriptional suppressive activity of PHF10.

Procaspase-3 accumulated in SGC7901^siPHF10 pretreated with z-DQMD-fmk without evident variation of substrates, which is consistent with the substrates being downstream mediators of procaspase-3. Because PHF10 functions in transcriptional regulation, this finding indicates that procaspase-3 may be transcriptionally modulated by PHF10. To confirm this and distinguish the binding motif for PHF10 on caspase-3, we did systematic luciferase assays in 293T cells. PHF10 repression of caspase-3 promoter activity was confirmed by cotransfecting 293T cells with caspase-3 luciferase and wild-type PHF10 or the shRNA for PHF10 (Fig. 4A). The promoter activity of caspase-3 was decreased by 3.5-fold in wild-type PHF10-transfected 293T cells compared with cells transfected with the control vector (pFLAG-V), which is consistent with the presence of a negative regulatory element in the region of caspase-3. Opposite results were observed following combined transfection of shRNA and deletions of caspase-3. As shown in Fig. 4B, loss of the regions spanning ¡©588 to −270 bp (D1–D3) in the caspase-3 construct did not affect the suppression of caspase-3 promoter activity by PHF10; however, further deletion beyond −170 completely abolished the repression of caspase-3 luciferase activity, suggesting that the D9 region (−270 to −170) contains negative regulatory elements that are required for PHF10 to repress caspase-3 promoter activity. Many transcription factors have more than one binding site in the promoter region. Therefore, deletions in D7 to D8 (¡©588 to −270) and D4 to D6 (−170 to +127) were also included in our assays. Based on the results of these assays, we conclude that the regions between −270 and −170 contained the repressor site(s).

To rule out a minimal effect due to use of an individual cell line, we did additional luciferase assays by cotransfecting CHO, AGS, SGC-7901, CRL5974, GES-1, and MKN28 cells (Supplementary Fig. S3A). Results from all these cells were similar to those obtained in 293T cells. However, in AGS and SGC-7901 cells, the luciferase activities were significantly different from all other cell lines investigated. High expression of PHF10 in both cell lines might serve as one possible explanation. Furthermore, results in Supplementary Fig. S3B, which were consistent with the L8G5 luciferase assay system results, confirmed that the fragment of the PHF10 protein encompassing three domains in tandem was essential for PHF10 to interact with the target sites. Caspase-3 mRNA levels (tested via q-RT-PCR) were increased after knocking down normal expression of PHF10 by shRNA (Fig. 4C) but were decreased following pFLAG-PHF10 transfection. Furthermore, q-RT-PCR experiments showed alteration of procaspase-3 expression over time (data not shown). The PHF10 protein was expressed at different levels by different cell lines. More specifically, PHF10 was expressed at higher levels in AGS, SGC7901, N87, and KATOIII cell lines than in SNU-1, SNU-16, MKN28, MKN45, and GES-1 cell lines (Fig. 4D). Most gastric cell lines showed an inverse correlation between PHF10 and caspase-3 expression, which was consistent with the speculation that PHF10 transcriptionally represses caspase-3 expression. However, a highly similar level of caspase-3 and PHF10 expression was observed in AGS and MKN28 cell lines, which suggests the presence of additional mechanisms contributing to tumorigenesis in these cell lines.

We sought to determine whether PHF10 binds to the promoter region of the caspase-3 gene in vivo using ChIP assays. The PCR forward and reverse primers flanking the essential regulatory region are shown in Fig. 5A. Using these primers for the region from −270 to −170, the two fragments derived by PCR overlapped with each other so as not to neglect any possible binding sites. First, to determine whether endogenous PHF10 binds to the caspase-3 promoter, we conducted ChIP assays on SGC7901 cells, which overexpressed PHF10 (Fig. 5B). Compared with control IgG, we found that PHF10 was clearly bound to caspase-3 promoter sequence using anti-PHF10 antibody (Fig. 5B, lanes 9 and 10). Next, to determine whether exogenous PHF10 (i.e., that produced via gene transfection) binds to the caspase-3 promoter, we immunoprecipitated chromatin fragments from MKN28^PHF10 with control IgG and a specific anti-PHF10 or anti-FLAG antibody and found that the DNA fragment of the caspase-3 promoter was immunoprecipitated by the specific anti-PHF10 or anti-FLAG antibody (Fig. 5C, lanes 9, 10, 13, and 14), whereas the DNA fragment of the caspase-3 3′-UTR was not immunoprecipitated by the specific anti-PHF10 or anti-FLAG antibody. In Fig. 5D, exogenous PHF10 clearly bound to the caspase-3 promoter (lanes 16 and 19) but not to 3′-UTR of caspase-3 (lane 13) compared with MKN28^control and MKN28^PBS cells. Thus, both exogenous and endogenous PHF10 bound to the caspase-3 promoter in gastric cancer cells in vivo.

Cancers arise from the rare simultaneous acquisition of the two cooperating conditions that permit deregulated cell proliferation and suppressed apoptosis (22, 23). This event, like cancer itself, happens rarely because the two processes are obligatorily interdependent. Deregulated proliferation on its own triggers expeditious cell death, whereas suppression of apoptosis confers no selective advantage in the absence of cell proliferation. Therefore, these two conditions must arise together in the same cell at the same time. This cooperative hypothesis for the emergence of cancer not only is in keeping with the well-characterized synergy exhibited by oncogenes but also has the great benefit of solving a great conundrum of vertebrate biology (22). To date, the extrinsic and intrinsic apoptotic pathways that ultimately lead to activation of effectors (caspase-3, caspase-2, and caspase-7) have been well characterized. The extrinsic pathway is initiated by ligation of death receptors (CD95/Fas, tumor necrosis factor receptor, and tumor necrosis factor–related apoptosis-inducing ligand receptor) to stimulate activator caspases (caspase-8 and caspase-10), which in turn cleave and activate effectors. The intrinsic pathway requires disruption of the mitochondrial membrane and release of proteins. Disruption of this pathway is extremely common in cancer cells (24). Among the apoptosis participants, the caspase-3 gene, a key factor in the apoptosis cascade, is widely expressed in tissues and cell lines and has been shown to have negative relationship with cancer malignant behaviors. Its expression is altered during the induction of apoptosis by drugs and during differentiation and development (25). Despite these studies, the mechanism by which the caspase-3 gene is transcriptionally modulated in cancers remains unknown.

To date, a wide variety of transcription factors have been shown to be involved in the modulation of caspase-3 expression. Candidate transcription factors involved in caspase-3–independent apoptotic pathways have surfaced from recent studies. Previous studies have indicated that an Ets element for Sp1-binding functions in the regulation of caspase-3 promoter activity (26). In addition, several other putative binding sites for the Sp1 family identified in the caspase-3 promoter have been shown to cooperate with p73 to activate the caspase-3 promoter (19, 26). In rats, a hypoxia-inducible factor-1–binding site has been identified in the caspase-3 promoter (27). Recently, further evidence has shown that the family of PHF proteins is overexpressed in many types of malignant tumors and might be involved in the progression of cells toward malignancy (28–30). PHD zinc finger proteins probably constitute the largest individual family of such nucleic acid–binding proteins; however, few target genes have been identified for PHD zinc finger proteins. In this study, we isolated and characterized a novel human PHD-containing zinc finger gene, PHF10. shRNA-mediated knockdown of PHF10 expression inhibited tumor growth and induced apoptosis both in vivo and in vitro, and overexpression of PHF10 did the opposite. These observations are supported by our findings that downregulation of PHF10 induced caspase-3 accumulation and activation of its downstream substrates, whereas PHF10 overexpression decreased the abundance of caspase-3. By assessing the protein profile of the apoptosis pathway, we identified a negative correlation between PHF10 and caspase-3 at the protein level. As we know, function of caspase-3 on its substrates relied on both the form activation and adequate amount. However, previous studies have given evidence that upregulation of the procaspase-3 could lead to the induction of apoptosis factor activation and thus trigger apoptosis, consistent with our results (19, 27, 31, 32).

Based on the results stated above, the next question to be addressed is how PHF10 modulates caspase-3. Indeed, the homeodomain (PHD) has been previously shown to interact with specific DNA sequences as a DNA-binding unit to function in maintaining chromatin stability (29, 30). In this investigation, using systematic transcriptional analysis, we showed that PHF10 might function as a transcriptional repressor, in which the LEAD domain and the PHD domains act in tandem. Knockdown of PHF10 induced apoptosis and gave rise to the accumulation of caspase-3 and activation of its substrates, which raised the possibility that PHF10 might participate in gastric cancer tumorigenesis by its function as a transcriptional repressor of caspase-3.

To extend these observations, we used a combination of transcriptional assays and showed that PHF10 seemed to be a transcriptional repressor, which led to the downregulation of caspase-3 protein levels. PHF10 bound to the promoter region of caspase-3; the intact PHF10-binding site in the caspase-3 promoter region is required for the full induction of caspase-3 promoter activity by PHF10. Furthermore, considering the dynamic nature of the interaction between PHF10 and the caspase-3 promoter, we conducted assays to determine the transcriptional activities of truncated forms of PHF10. We found that the integrity of the three domains in tandem was essential for PHF10 to achieve its role in transcriptional regulation, which is consistent with observations in the L8G5-luc system (33). The present mechanistic investigations found that downregulation of PHF10 in gastric cancer cells led to significant induction of caspase-3 mRNA and protein and that the opposite trend was achieved in gastric cancer cells with PHF10 overexpression. Evidences are beginning to emerge that some PHD fingers can bind to nucleosomes and chromatin binding is a more widespread property of PHD fingers (34). The role of the PHD fingers seems to be to consolidate, or to strengthen, a separate chromatin-binding activity of either the same or an associated protein, although the interactions between PHD fingers and chromatin need further biochemical characterization (29, 35). PHD fingers also bind to proteins other than histones and rely on the structural integrity of the domain, as originally proposed. Findings that PHD domain participated in transcriptional modulation of caspase-3 in our study suggested that this activity could be a property of this slightly unusual PHD module. This might reveal whether PHD fingers can interact simultaneously with nucleosomes, specific protein ligands, and DNA sequence, thereby tethering these ligands to interaction complex. Anyway, the actual relationship between them needs further studies. Because the regulation of gene expression is complex, involving transcriptional regulators, as well as DNA methylation, histone acetylation, methylation, sumoylation, and other modifications, these results may provide only a limited understanding of PHF10 as a transcriptional regulator of apoptosis in gastric cancer. Moreover, further analysis of the role of PHF10 in human cancers is necessary to assess the involvement of PHF10 in tumorigenesis. In summary, our findings support the notion that PHF10 acts as an oncogene, and the present report provides the basis to develop PHF10 as a potential therapeutic target for the treatment of gastric cancer.

No potential conflicts of interest were disclosed.

We thank Prof. Jianhua Wang for the polishing of this manuscript and Dr. Wu Zhang for the FACS analysis in Shanghai Institute of Hematology under the direction of Prof. Jiang Zhu.

Grant Support: National High Technology Research and Development Program of China (863 Program No. 2006AA02A301 and No. 2007AA02Z179); Science and Technology Commission of Shanghai Municipality (No. 07jc14041, 09DZ1950100); National Natural Science Foundation (No. 30471961, 30772107, 30670939, and 30872476); Key Projects in the National Science & Technology Pillar Program (No. 2008BA152B03); Shanghai Leading Discipline-Surgery (S30204); Shanghai Key Laboratory of Gastric Neoplasms (09DZ2260200); and Doctoral Fund of School of Medicine, Shanghai Jiao Tong University (BXJ0812).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.