Homeodomain-Interacting Protein Kinase-2 Restrains Cytosolic Phospholipase A₂–Dependent Prostaglandin E₂ Generation in Human Colorectal Cancer Cells Free

Homeodomain-interacting protein kinase-2 (HIPK2) belongs to a family of corepressors for homeodomain transcription factors but, unlike other transcriptional corepressors, it has protein kinase activity (1). Its role as transcriptional regulator, often in complex with histone deacetylases (HDAC), has been shown for several different transcription factors including Smad1/4 and Brn3a (2–4). Moreover, HIPK2 plays a role in regulating p53 oncosuppressor functions through different posttranslational modifications. Thus, HIPK2 phosphorylates p53 at Ser⁴⁶ in response to severe DNA damage, inducing apoptosis (5, 6). On the other hand, HIPK2 mediates p53 acetylation in response to moderate genotoxic damage, selectively inducing growth arrest (7). The role played by HIPK2 during apoptosis is also becoming evident in a p53-independent way. Thus, HIPK2 phosphorylates the antiapoptotic transcriptional corepressor, CtBP, targeting it for proteasomal degradation (8, 9), activates the c-Jun NH₂-terminal kinase signaling pathway participating in transforming growth factor-β-induced c-Jun NH₂-terminal kinase activation and apoptosis (10), and induces caspase-dependent apoptosis in sympathetic neurons (11). Finally, silencing of endogenous HIPK2 by RNA-interference has been shown to increase cell survival after genotoxic damage (12), suggesting an involvement in restraining tumor progression.

Several lines of evidence record that prostaglandin E₂ (PGE₂) plays a leading role in carcinogenesis (13, 14). This is coherent with the biological traits of PGE₂, i.e., trigger of proliferation, angiogenesis, and invasiveness, and inhibitor of apoptosis (15–21). The generation of the prostanoids is carried out through three consecutive enzymatic steps: (a) the release of arachidonic acid from membrane phospholipids by phospholipases, mainly cytosolic phospholipase A₂α (cPLA₂); (b) the transformation of arachidonic acid to the unstable endoperoxide PGH₂ by prostaglandin H synthase, popularly known as cyclooxygenase-1 and cyclooxygenase-2 (COX-1 and COX-2); (c) its metabolization to PGE₂ by several PGE synthases (PGES; ref. 22). The rate-limiting step in PGE₂ generation is the availability of arachidonic acid for COX-isozymes, which is dictated by cPLA₂ expression and activity. cPLA₂ is part of a complex gene family which consists of eight secretory and three cytoplasmic phospholipases. cPLA₂ is the Ca²⁺-sensitive form which is ubiquitously expressed in most tissues and preferentially hydrolyzes phospholipids in the sn-2 position where arachidonic acid is esterified. Cellular cPLA₂ activities are tightly regulated by different factors, including Ca²⁺ and phosphorylation (23, 24). Calcium ions drive the membrane translocation of cPLA₂, whereas its phosphorylation of Ser⁵⁰⁵ by mitogen-activated protein kinases enhances the cellular activities of cPLA₂ under physiologic conditions by elongating its membrane residence and thereby improving its overall affinity for the perinuclear membranes (25). Increasing evidence shows that cPLA₂ expression could also be regulated at a transcriptional level (26, 27). Differently from the downstream enzymes, COX-2 and the microsomal PGES-1 (mPGES-1), which are widely expressed in precancerous and cancerous lesions of the intestine (28–31), cPLA₂ overexpression is less frequently detectable in human colorectal cancer and often does not correlate with COX-2 expression (32, 33). This suggests that cPLA₂ expression is tightly regulated during tumorigenesis; however, the mechanism involved is not yet known.

In the present study, we addressed (a) whether HIPK2 modulates PGE₂ biosynthesis by constraining cPLA₂ expression in human colon cancer cell line RKO using RNA-interference for HIPK2 function and (b) the consequence of HIPK2 depletion on tumor cell growth in vivo. Finally, we correlated the expression of HIPK2 and cPLA₂ in human colorectal cancers. Our findings reveal the novel mechanism of HIPK2 to restrain progression of human colon tumorigenesis, at least in part, by turning off cPLA₂-dependent PGE₂ generation.

Cells and reagents. Human colon carcinoma cell line RKO-pSuper and HIPK2-interfered (HIPK2i) cell lines (12) and human lung cancer cell line H1299 were cultured in RPMI 1640 (Life Technologies, Grand Island, NY) supplemented with 10% heat-inactivated fetal bovine serum (FBS, Life Technologies) plus glutamine and antibiotics in a humidified atmosphere with 5% CO₂ at 37°C.

cPLA₂ selective inhibitor arachidonyl trifluoromethyl ketone (Cayman Chemical, Ann Arbor, MI; ref. 34) was dissolved in DMSO and stored at −20°C. [³H]-PGE₂, [³H]-PGF_2α, [³H]-6-keto-PGF_1α, and [³H]-TXB₂ were purchased from Perkin-Elmer Life Science Products (Brussels, Belgium).

Prostanoid analyses. RKO-pSuper and HIPK2i cells, grown to subconfluence in 60 mm Petri dishes, were cultured for 24 hours with 0.5% FBS before adding fresh medium with 10% FBS for the indicated time with or without the abovementioned inhibitor. Cell culture media were used to assess PGE₂, PGF_2α 6-keto-PGF_1α, and TXB₂ biosynthesis by specific RIA techniques, as previously described (35–37). PGD₂ biosynthesis was measured in cell culture medium by using the PGD₂-methoxylamina enzyme immunoassay kit (Cayman Chemical) following the manufacturer's instructions. Rabbit polyclonal anti-PGE₂, anti-PGF_2α, anti-6-keto-PGF_1α, and anti-TXB₂ antibodies were described elsewhere (36, 37).

Western blot analyses. Total cell extracts were prepared by incubating at 4°C for 30 minutes in lysis buffer [50 mmol/L Tris-HCl (pH 7.5), 50 mmol/L NaCl, 5 mmol/L EDTA, 150 mmol/L KCl, 1 mmol/L DTT, 1% Nonidet P-40] plus a mix of protease inhibitors (Sigma Chemical Company, St. Louis, MO). Proteins were then separated by SDS-PAGE, blotted onto nitrocellulose (Bio-Rad, Richmond, CA) and immunoreacted with mouse monoclonal anti-cPLA₂, anti-β-actin (both from Santa Cruz Biotechnology, Santa Cruz, CA), rabbit polyclonal anti-phospho-cPLA₂ (Ser⁵⁰⁵; Assay Designs Technology, Ann Arbor, MI; refs. 38, 39), antitubulin mouse monoclonal (Sigma BioSciences), and anti-HIPK2 rabbit antiserum (kindly provided by M.L. Schmitz, University of Bern, Switzerland). Immunoreactivity was detected with the enhanced chemiluminescence reaction kit (Amersham Corp., Arlington Heights, IL).

Cell transfection and luciferase assays. H1299 cells were transiently transfected by using the modified calcium phosphate precipitation method as described earlier (5). The amount of plasmid DNA was equalized in each sample by supplementing with empty vector. The expression vectors used in this study were: cPLA₂-luc (kindly provided by R.A. Nemenoff, Department of Medicine, University of Colorado Health Sciences Center, Denver, CO), Myc-HDAC1 (kindly provided by C.Y. Choi, Department of Biological Science, Sungkyunkwan University, Suwon, Republic of Korea), and HIPK2-Flag (5). Transfection efficiencies were normalized with the use of a cotransfected β-galactosidase construct. Luciferase activity was assayed as previously described (5).

RNA extraction and reverse transcriptase-PCR analysis. RKO-pSuper and HIPK2i cells were grown to subconfluence in 100 mm dishes and collected. Total messenger RNA was extracted using the RNeasy mini kit (Qiagen S.P.A., Milan, Italy). Reverse-transcription reactions and PCR assays were done using the MuLV reverse transcriptase and the AmpliTaq DNA polymerase (Gene Amp RNA PCR kit, Perkin-Elmer, Roche Molecular System, Branchburg, NJ). The sequence of the primers used were as follows: human cPLA₂ upstream, 5′-CTC TTG AAG TTT GCT CAT GCC CAG AC-3′; and downstream, 5′-GCA AAC ATC AGC TCT GAA ACG TCA GG-3′. Primers for human glyceraldehyde-3-phosphate dehydrogenase and HIPK2 have been described elsewhere (12). DNA products were run on 2% agarose gel and visualized by ethidium bromide using UV light.

Chromatin immunoprecipitation assay. Chromatin immunoprecipitation was done essentially as described (7). Protein complexes were cross-linked to DNA in living nuclei by adding formaldehyde (Carlo Erba, Milan, Italy) directly to the cell culture medium at 1% final concentration. Antibodies used were rabbit polyclonal anti-acetylated histone H4 (Cell Signaling) and no specific immunoglobulins (Santa Cruz) were as negative controls. In each experiment, the linearity of the signal was insured by amplification of increasing amounts of template DNA. Immunoprecipitation with no specific immunoglobulins (Santa Cruz) was done as a negative control.

In vivo tumorigenicity assay. Six-week-old female CD-1 nude (nu/nu) mice (Charles River Laboratories, Calco, Italy) were used for in vivo studies. They were housed in specific pathogen-free conditions and fed standard cow pellets and water ad libitum. Experiments were done in accordance with institutional standard guidelines for animal experiments. Each experimental group included eight animals. Solid tumors were obtained by injecting (i.m.) 3 × 10⁶ viable RKO-pSuper and HIPK2i cells (three different interfered cell populations) suspended in 0.1 mL PBS into the right leg muscles. The mice were examined every day after injection. The tumors were detectable 10 days after the injection. Their dimensions were measured every other day and their volumes were calculated from caliper measurements of two orthogonal diameters (x and y, larger and smaller diameters, respectively) by using the formula: volume = xy² / 2. Tumor growth was monitored for 1 month and the animals were sacrificed in accordance with institutional standard guidelines.

Analysis of HIPK2 and cPLA₂ expression in human colorectal cancers. Tissue specimens from nine colorectal cancers were obtained with informed consent from previously untreated patients who underwent surgical resection at the National Institute for Cure and Study of Tumours (Milan, Italy). Immediately after tumor resection, an experienced pathologist selected and sampled the tumor specimen and its surrounding normal mucosa, which were subsequently stored at −80°C. The cohort of patients encompasses five consecutive sporadic colorectal cancers (mean age, 65 years; two with Duke's B, two with Duke's C, one with Duke's D) and four consecutive colorectal cancers (mean age, 47 years; two with Duke's A, two with Duke's C, and one with Duke's D) from patients with familial adenomatous polyposis (FAP; Apc germ line detected mutations) for which frozen material was available. Total RNA (1 μg), extracted from snap-frozen tumor tissue samples, stored at −80°C, was reverse-transcribed into cDNA using oligo (dT) primers and reverse-transcriptase (Superscript II, Life Technologies) according to the manufacturer's recommendations. The integrity of cDNA was detected by the amplification of the housekeeping β-actin gene. cDNA (1 μL) was used as template for each reverse transcriptase-PCR (RT-PCR) reaction.

Statistical analysis. The data are expressed as mean ± SE. Statistical comparisons were made by ANOVA or t test followed by Student-Newman-Keuls test for in vitro and in vivo studies in experimental animals. In contrast, the Mann-Whitney U test was used for comparisons in human cancers.

HIPK2 silencing induces prostanoid biosynthesis. We initially analyzed arachidonic acid metabolites released from RKO-pSuper and HIPK2i cells. We found that RKO-pSuper cells cultured in the presence of 10% FBS up to 24 hours did not release detectable levels of thromboxane (TXB₂; a nonenzymatic metabolite of TXA₂), PGF_2α, PGD₂, 6-keto-PGF_1α (a nonenzymatic metabolite of prostacyclin), whereas only scanty levels of PGE₂ and PGF_2α (from <10 to 15 and 21 pg/10⁶ cells, respectively) were detected. Conversely, depletion of HIPK2, confirmed by Western blot analysis (Fig. 1A, inset), was associated with the production of high levels of PGE₂ and to a smaller extent of PGF_2α (at 24 hours, PGE₂ and PGF_2α averaged 1,800 and 379 pg/10⁶ cells, respectively), whereas the other prostanoids were undetectable (data not shown). These results suggest that HIPK2 silencing was associated with rousing prostanoid biosynthesis and that PGE₂ was the major product of arachidonic acid metabolism. In Fig. 1A, the time-dependent production of PGE₂ in RKO-pSuper cells and HIPK2i cells is shown. We next assessed the role of cPLA₂ in mediating PGE₂ biosynthesis. As shown in Fig. 1B, arachidonyl trifluoromethyl ketone, a potent and selective slow binding inhibitor of the cPLA₂ (34), profoundly suppressed PGE₂ biosynthesis demonstrating the dominant role of cPLA₂ in prostanoid production in HIPK2-depleted cells and suggesting that HIPK2 might function as a repressor of cPLA₂ in tumors.

Immunoblot analysis of total cell extracts (Fig. 1C and D) showed that the scanty levels of prostanoids produced by RKO-pSuper cells were associated with low levels of cPLA₂ protein, which was scarcely phosphorylated at Ser⁵⁰⁵. Conversely, enhanced PGE₂ generation in HIPK2i cells was associated with a strong induction of cPLA₂ protein, which is highly phosphorylated especially at 3 and 24 hours (Fig. 1C and D).

HIPK2 is involved in cPLA₂ gene regulation. Previous reports have established a role for HIPK2 in transcriptional regulation acting as transcriptional corepressor (1–4). Here, we aimed to evaluate whether HIPK2 plays a role in the cPLA₂ gene regulation. To this end, we first investigated the effect of HIPK2 overexpression on the transcriptional activity of the cPLA₂-luc promoter. As HIPK2 functions as a transcriptional corepressor interacting, among others, with HDAC1 corepressor (2), we transfected H1299 cells with HDAC1 and HIPK2 expression vectors, along with the cPLA₂-luciferase reporter vector. As shown in Fig. 2A, the efficient transcriptional activity of the cPLA₂-luc promoter was reduced by HIPK2 overexpression and further inhibited by coexpressing HDAC1. Interestingly, HIPK2 overexpression did not affect transcriptional activity of COX-2 and mPGES1 promoters (data not shown). In agreement with the luciferase data, interference of the endogenous HIPK2 resulted in a strong increase of the acetylated histone H4 levels on cPLA₂ promoter, as shown by chromatin immunoprecipitation analysis, compared with the pSuper control cells (Fig. 2B). These findings suggest a recruitment of the HDAC complex to the cPLA₂ target promoter only when HIPK2 is expressed, allowing deacetylation of histone H4 that represses chromatin in vivo. To verify whether the results of transcription and chromatin immunoprecipitation analyses corresponded to an in vivo regulation of the cPLA₂ gene, RT-PCR analyses were carried out using cDNAs derived from RKO-control and HIPK2i cells. As shown in Fig. 2C, cPLA₂ gene was easily detected only in HIPK2i cells, suggesting that HIPK2 is required for cPLA₂ transcriptional regulation. Altogether, these data support the notion that HIPK2 is a transcriptional corepressor that can suppress the cPLA₂ gene in colon cancer cells likely through HDAC1 interaction.

HIPK2 depletion increases in vivo tumor growth. To ascertain the involvement of HIPK2 in cell growth, proliferation and viability assays were first done. RKO-pSuper and HIPK2i cells were plated in the presence of 10% serum and cell number was determined at daily intervals. We found that both cell types grew in a similar extent, suggesting that HIPK2-depletion does not modulate the capability of the cells to grow in vitro (data not shown). Interestingly, we found that HIPK2i cells showed a longer capability to survive in in vitro growth culture conditions, compared with the pSuper control cells (data not shown). Morphologic observations of the cells show that HIPK2-depleted cells grow in clusters and easily detach from the plate compared with the pSuper control cells, suggesting the acquirement of anchorage-independent growth as a further index of transformation (Fig. 3A).

We next studied, for the first time, the involvement of HIPK2 in colon tumorigenicity in vivo. RKO-pSuper cells and -HIPK2i cells derived from three different interfered populations were injected i.m. in nude mice. The tumors were detectable 10 days after injection and the tumor sizes were determined every other day for ∼20 days in addition. As shown in Fig. 3B, the tumors derived from HIPK2i cells showed noticeably increased growth compared with the pSuper control cells, with a median tumor mass 2.5 times bigger than in control cells. These data are in agreement with the potentiality of HIPK2i cells to survive in vitro and reinforces our hypothesis that loss of HIPK2 confers in RKO cells the capability to activate a survival pathway that is inhibited by HIPK2.

HIPK2 and cPLA2 gene expression in human colorectal cancers. To better address the physiologic relevance of the role played by HIPK2 in restraining tumor progression, we compared, for the first time, the gene expression of HIPK2 and cPLA₂ in human colorectal cancer specimens collected from patients with FAP and sporadic colorectal cancer, using RT-PCR analysis. As shown in Fig. 4A and B, HIPK2 mRNA levels were different, in a statistically significant fashion, between the two groups of tumoral specimens, i.e., higher in colorectal cancers of patients with FAP than in those of sporadic colorectal cancer. Interestingly, we found that high expression of HIPK2 was associated with undetectable cPLA₂ mRNA levels in the colorectal cancers of patients with FAP (Fig. 4A, left), whereas reduced expression of HIPK2 was associated with detectable levels of cPLA₂ mRNA in sporadic colorectal cancers (Fig. 4A, right). These results might suggest that HIPK2 also restrains cPLA₂ expression in vivo in humans. Interestingly, we found that HIPK2 expression in human colorectal cancers tended to correlate inversely with the staging of the tumors (Fig. 4C), suggesting that HIPK2 expression takes part in the control of tumor progression.

In the present study, we showed evidence of a novel pathway by which HIPK2 protein kinase may curb tumor growth. HIPK2 silencing de-repressed cPLA₂ gene transcription, thus unleashing the biosynthesis of PGE₂. Observational and randomized controlled studies in many different population cohorts and settings have shown the protective effects of nonsteroidal anti-inflammatory drugs for colorectal cancers (40). Inhibition of PGE₂ by nonsteroidal anti-inflammatory drugs has been proposed to be involved in these effects (41–45). In fact, PGE₂ is the predominant prostanoid found in most colorectal cancers and is known to promote colon carcinoma growth by stimulating proliferation, angiogenesis, invasiveness, and by inhibiting apoptosis (3–9). A direct evidence of its involvement in colorectal tumorigenesis has been provided by studies of disruption of PGE₂ receptor genes (46, 47). The rate-limiting step in PGE₂ generation is the availability of arachidonic acid for COX-isozymes, which is dictated by cPLA₂ expression and activity. cPLA₂ can be induced by several cytokines and oncogenic Ras through transcriptional mechanisms (26). In fact, the cPLA₂ promoter contains a number of putative regulatory elements including activator protein sites, nuclear factor κB sites, and glucocorticoid regulatory elements (48) that are critical for induction by oncogenic Ras (27). However, until now, no information is available on the possible regulation of cPLA₂ expression by chromatin remodeling. Here, we show that HIPK2 could act as a corepressor for the cPLA₂ promoter in agreement with the function of HIPK2 as transcriptional corepressor along with Groucho and HDAC (2). The role of HIPK2 as transcriptional regulator has been shown for several different transcription factors including homeodomain transcription factors, Smad1/4, and Brn3a (2–4).

The mechanism implicated in cPLA₂-dependent intestinal tumorigenesis through the supply of arachidonic acid to COX-2—thus rousing PGE₂ biosynthesis—is distinctly shown by the finding that deletion of cPLA₂ suppresses APC^Min-induced polyp number similarly to COX-2 deletion (49, 50). This finding shades the proposal that the level of free arachidonic acid and not PGE₂ is the critical factor in tumorigenesis. Moreover, it suggests that cPLA₂ is the predominant source of arachidonic acid for COX in the intestine. These results confirm our findings on a restrainable effect of HIPK2 on cPLA₂ expression and provide incentives to perform further studies aimed to ascertain the molecular mechanisms involved in the regulation of PGE₂ biosynthetic machinery by HIPK2 in more detail.

RKO-HIPK2i cells—releasing PGE₂—were characterized by enhanced tumorigenicity in vivo versus parental cells when injected into nude mice. Our results obtained in experiments done in vitro and for the first time in vivo on tumor growth, suggest that HIPK2 depletion in tumor cells might likely activate a survival pathway that, in the presence of HIPK2, is negatively regulated. This is coherent with our results that HIPK2 gene expression tended to correlate inversely with the staging of colorectal cancers and with the expression of cPLA₂ (Fig. 4B and C). Interestingly, high expression of HIPK2 was associated with undetectable cPLA₂ mRNA levels in colorectal cancers of patients with FAP, whereas reduced expression of HIPK2 was associated with detectable levels of cPLA₂ mRNA in sporadic colorectal cancers. Although a reduced expression of HIPK2 genes has been found in thyroid and breast cancers (51), further studies are required to clarify whether HIPK2 expression is a determinant of the type and/or the staging of neoplasia. Moreover, our results pave the way for investigating whether HIPK2 restrains tumorigenesis by turning off cPLA2-dependent PGE₂ generation in different types of human neoplasias.

Our finding that cPLA₂ is hardly detectable in patients with FAP might suggest that PGE₂ generation has only a scanty contribution in tumorigenesis in this setting despite a wide expression of the downstream enzymes. This is nicely confirmed by the results of randomized clinical trials with selective COX-2 inhibitors in patients with FAP showing only a modest clinical efficacy (52, 53). Our observation is based on small numbers but should spur the performance of larger studies comparing clinical outcomes with molecular biomarkers (such as the concurrent expression of cPLA₂, COX-2, and mPGES-1). The development of biomarkers to predict the efficacy of coxibs is this setting will enable us to implement a personalized therapy to restrict the drugs to patients likely to benefit from them, and to avoid the useless exposure of patients to the risk of cardiovascular hazards attached to coxibs (54).

In conclusion, our findings reveal the possible involvement of HIPK2 in restraining the progression of human colon tumorigenesis, at least in part, by turning off cPLA₂-dependent PGE₂ generation. This new property of HIPK2 is notable because cPLA₂ expression dictates prostanoid biosynthesis despite the expression of the downstream enzymes (i.e., COX-2 and mPGES-1). Our observation of the tight control of cPLA₂ expression in FAP may explain the modest efficacy of COX-2 inhibitors in polyp regression, shown in randomized clinical trials in this setting, despite the wide expression of COX-2 detected in the lesions (52, 53).

We thank all the people cited in the text for their generous gifts. We also thank M.P. Visconti for technical support.