Ketogenic HMGCS2 Is a c-Myc Target Gene Expressed in Differentiated Cells of Human Colonic Epithelium and Down-Regulated in Colon Cancer

Mitochondrial 3-hydroxy-3-methylglutaryl-CoA synthase (HMGCS2) is a potential regulatory point in the pathway that converts acetyl-CoA to ketone bodies (1-3). In rats, hepatic ketogenic capacity increases rapidly during postnatal development (4) or fasting (5) when the mRNA, protein, and activity of HMGCS2 increase in liver (6-8). The factors responsible for the postnatal and starvation-associated induction of HMGCS2 gene expression have been identified and shown to involve changes in circulating hormone and fatty acid levels. Administration of dibutyryl-cyclic AMP to adult rats increases mRNA, protein levels, and enzymatic activity of HMGCS2 in the liver (6, 9). Similar effects are produced by a high-fat diet (10), whereas insulin causes the opposite effect (6).

The rat HMGCS2 gene contains elements that control its multihormonal regulation and tissue specificity (11-14). To date, three main regulatory sequences have been described in the rat HMGCS2 promoter: (a) an insulin-responsive sequence, located at position −211 with respect to the transcriptional start site, which binds proteins from the forkhead family of transcription factors (12); (b) a peroxisome proliferator-activated receptor element, located at position −104, which binds retinoid X receptor-peroxisome proliferator-activated receptor heterodimers (13); and (c) a Sp1-binding site, located at position −56, which binds proteins from the Sp1 family of transcription factors (14). In the human promoter, the peroxisome proliferator-activated receptor element and the Sp1-binding site are also present in similar positions (14-16). Thus, transcriptional activation of the HMGCS2 promoter, mediated by a network of recruited transcription factors, can partly explain the induction of gene expression observed in liver under physiologic and pathologic conditions.

The HMGCS2 gene is also expressed in extrahepatic tissues. In rats, HMGCS2 is expressed in testis, ovary, intestine, kidney, brain, colon, and lymphocytes (6, 7, 17-19). In humans, expression has been reported in liver, skeletal muscle, heart, pancreas, testis, and colon (20). The role of HMGCS2 expression in several of these tissues remains to be fully elucidated. However, in tissues that oxidize fatty acids, it has been proposed that HMGCS2 expression prevents acetyl-CoA accumulation, which in turn impairs fatty acid oxidation rates (21). In resting lymphocytes, HMGCS2 expression may provide cell metabolism with a “branch point sensitivity” mechanism by which cells regulate the rate of precursor utilization (acetyl-CoA) during the cell cycle (19). Thus, a high flux through a pathway in which acetyl-CoA is an intermediate buffers the concentration of acetyl-CoA against changes in its rate of utilization in other pathways.

The capacity of the colon for ketogenesis has been correlated with HMGCS2 expression, which depends on the amount of butyrate produced by intestinal flora (18, 22). The effect of butyrate on HMGCS2 expression in vivo is exerted at the transcriptional level through a single Sp1-binding site present in the proximal promoter of the gene (14). Butyrate stimulates cell proliferation in healthy colonic epithelial cells. However, it reduces proliferation and induces cell differentiation and apoptosis in cancer cell lines in this tissue (23-25). One of the hypotheses proposed to explain the “butyrate paradox” is that healthy colonocytes metabolize butyrate efficiently, which would result in a reduction in intracellular concentration and therefore a decrease in capacity to inhibit growth (26). The capacity to metabolize butyrate could be lost in colon cancer cell lines (27). In this context, it has recently been described that a lack of induction of HMGCS2 leads to impaired β-oxidation (21).

In the colon, the β-catenin/TCF-4 complex forms the master switch that controls proliferation and differentiation in healthy and malignant intestinal epithelial cells. Thus, disruption of β-catenin/TCF-4 activity induces cell cycle arrest and differentiation (28). One of the target genes of TCF-4 is Myc, which encodes a transcription factor, c-Myc (herein called Myc). This factor can activate or repress gene expression. Myc regulates several major cellular functions, such as proliferation, adhesion, and size, which explains the diversity of Myc target genes found in a variety of metabolic pathways. In addition, Myc has been linked to energy metabolism via up-regulation of glycolysis (29), mitochondrial biogenesis (30), and abrogation of hepatic HMGCS2 expression and ketogenesis after the induction of diabetes (31). Myc represses gene transcription by at least two distinct mechanisms. The first is limited to the Inr sequence in the promoter start site and requires the inhibition of Miz-1-mediated transcription activation by the binding of Myc-Max heterodimers (32-34). The second mechanism is Inr independent and involves other promoter sequences in the recruitment of Myc-containing repressor complexes. In this mechanism, Sp1 has been reported to be one of several proteins with the capacity to recruit Myc (35, 36). However, other promoter sequences also participate in the recruitment of this transcription factor (37) or in Miz-1 transactivation (38).

To study the mechanism that regulates HMGCS2 expression, here we examined the expression of this gene in human colon. We observed a clear pattern of expression in differentiated cells of the colonic epithelium. Using chromatin immunoprecipitation assays, we also show that HMGCS2 is a direct target of Myc, which abrogates Miz-1-mediated transcriptional activation of HMGCS2 expression. Therefore, transcriptional repression by Myc could play a crucial role in regulating the expression of this gene. Our data led us to analyze the expression of HMGCS2 in several types of human intestinal cancer. A clear down-regulation of gene expression in colon, rectum, and small intestine was observed. In the first two tissues, this down-regulation correlated with Myc up-regulated tumorigenesis. Consistently with its expression in differentiated cells, HMGCS2 protein expression was down-regulated preferentially in moderately and poorly differentiated carcinomas.

In liver and colon, HMGCS2 is a histone deacetylase inhibitor target gene (14). Because histone acetylation is related to cell differentiation, we studied the expression pattern of a set of genes in proliferating and differentiated CaCo-2 cells. HMGCS2 expression was lower in proliferating cells (Fig. 1, compare lanes 1 and 3). Similar data were obtained when HMGCS2 expression was evaluated in terms of enzymatic activity (0.024 versus 0.25±0.02 mU/mg for proliferating and differentiated cells, respectively). As expected, differentiation induced the expression of intestinal alkaline phosphatase. Miz-1 expression also increased during differentiation, whereas Myc expression was clearly down-regulated (Fig. 1). As the expression of HMGCS2 correlated with acetylation of H4 promoter-associated histones (14), we studied the effect of trichostatin A on proliferating and differentiated cells. This drug induces mRNA expression of HMGCS2 only on the former (Fig. 1, compare lanes 1 and 2 with lanes 3 and 4), which indicates that histone acetylation is involved in cell differentiation-associated induction of HMGCS2. Trichostatin A did not affect mRNA levels of Miz-1 in proliferating or differentiated cells. However, trichostatin A down-regulated Myc expression in the former.

To study the expression pattern of HMGCS2 in human colon, we raised an antibody against a human HMGCS2 epitope. The antibody recognizes recombinant human HMGCS2 expressed in a stable cell line (Fig. 2A, lane 1). To ensure that antibody specificity recognized the human protein in the extracts, we used another stable cell line expressing rat recombinant HMGCS2 (39). The antibody recognized the rat HMGCS2 protein with a much less affinity (Fig. 2A, lane 2). HMGCS2 expression was localized mainly in differentiated cells of the colonic epithelium, whereas proliferating cells did not express the protein (Fig. 2B). In addition, no labeling was observed with a preimmune serum in either cell stage (Fig. 2B).

The HMGCS2 expression pattern is consistent with the Myc and Miz-1 expression gradients along the axis of the crypt (28). Moreover, expression of Myc in the liver of transgenic mice abolishes the ketosis associated with streptozotocin-induced diabetes and down-regulates HMGCS2 expression (31). Therefore, we studied the role of Myc in the regulation of HMGCS2 expression. Antibodies against Myc effectively precipitated the human HMGCS2 promoter in a chromatin immunoprecipitation assay (Fig. 3A). These data indicate that HMGCS2 is a direct target of Myc. In addition, Myc cotransfection abrogated the expression of a reporter construct driven by the HMGCS2 promoter, indicating a putative role for Myc in HMGCS2 down-regulation in proliferating cells (Fig. 3B). Moreover, Myc overexpression in CaCo-2 cells (Fig. 3D) down-regulated the endogenous HMGCS2 expression by 35% to 40% (Fig. 3C).

Repressor effects of Myc are produced by either Inr-dependent or Inr-independent mechanism. Therefore, we studied the effect of progressive deletions and Sp1-binding site mutation of the rat HMGCS2 promoter (Fig. 4A) on promoter activity and Miz-1-mediated effects (Fig. 4B). The previously characterized peroxisome proliferator-activated receptor element (15) and Sp1-binding sites (14) participated in sustaining transcriptional activity of the HMGCS2 promoter. Miz-1 transactivated all the constructs containing the Sp1-binding site assayed, whereas deletion or mutation of the Sp1 site abrogated Miz-1-mediated transcriptional activity (Fig. 4B). Conversely, Miz-1-mediated transactivation was not abolished by mutation of the transcription start site (data not shown). As expected, Miz-1-mediated transactivation was blocked by Myc expression (Fig. 4C). In addition, a Myc deletion mutant (pME18S-c-Myc D348-493) lacking the carboxyl end of the protein did not repress Miz-1 transactivation (Fig. 4C).

To assess whether the HMGCS2 Sp1-binding site recruited both transcription factors, we did DNA precipitation assays in which cellular extracts expressing tagged proteins were incubated with biotinylated Sp1 wild-type or mutant probes. Figure 4D shows the binding of Sp1, Myc, and Miz-1 to the Sp1 probe (Sp1-wt). This binding was specific, because a mutated Sp1 probe (Sp1-M) did not recruit any of these proteins.

These data indicate that the Sp1-binding site is involved in Miz-1-dependent transcriptional activity, which can explain, at least in part, Myc transrepression of the HMGCS2 promoter in colon-derived cell lines.

Myc has been implicated in the genesis of specific human tumors. In all cases, the relative amounts of Myc are increased in the tumor tissue relative to the surrounding healthy tissues, which indicates that elevated expression of Myc contributes to tumorigenesis (40). To examine the putative role of Myc-mediated down-regulation of HMGCS2 expression in distinct types of intestinal cancer, we studied the expression pattern of HMGCS2 (Fig. 5A) and Myc (Fig. 5B) in tumor and healthy matched tissues. HMGCS2 expression was higher in tissues, such as colon and small intestine, whereas the rectum and stomach showed lower levels of expression. In addition, expression differed between samples of healthy small intestine, colon, and rectum from the patient sample tested. Fluctuations between healthy tissues were much lower in the Myc expression pattern (Fig. 5B). This therefore hinders the establishment of a relationship between HMGCS2 and Myc expression in healthy tissues.

Colon HMGCS2 expression was down-regulated in 90% of the tumors (<0.7-fold expression in tumor relative to healthy tissue; Fig. 5A, right). In contrast, colon Myc expression was up-regulated in 80% of the tumors (>1.5-fold expression in tumor relative to healthy tissue; Fig. 5B, right). These findings indicate that, despite individual fluctuations in gene expression, HMGCS2 is down-regulated in Myc-dependent colon tumors. Similarly, an inverse correlation also occurred between the expressions of these two genes in rectum. In this tissue, HMGCS2 expression was down-regulated in 80% of the tumors (Fig. 5A, right), whereas Myc was up-regulated in 70% of tumors (Fig. 5A, right). However, this type of correlation was not detected in the small intestine or stomach. In the former, HMGCS2 down-regulation (84% of the tumors) did not correlate with an up-regulation of Myc (14% of the tumors), whereas, in stomach, no clear down-regulation of HMGCS2 gene expression was observed (40% of the tumors showed either down-regulation or up-regulation). These data indicate that HMGCS2 is down-regulated in small intestine, colon, and rectum, although only colon and rectum tumors were related to Myc overexpression. Therefore, HMGCS2 down-regulation in tumors may be a characteristic of extending Myc-dependent tumorigenesis.

To study correlations between differentiation and HMGCS2 down-regulation in colon cancer, immunohistochemistry was done on a tissue array (Colon Tumor Tissue Array from Lab Vision, Freemont, CA) containing samples from 64 colorectal adenocarcinoma specimens (Fig. 6). In 5 cases, no neoplastic tissue was observed. Two cases showed healthy, in addition to tumoral, tissue, whereas marginal adenoma was observed in another 2 cases. In these 4 cases, both the healthy crypts and the epithelium with features of adenoma showed the highest density of HMGCS2-positive cells in comparison with the malignant glands. The grade of differentiation of the adenocarcinomas was scored based on observations of the tumor areas shown in the tissue arrays. Nineteen carcinomata were regarded as well differentiated, 37 as moderately differentiated, and 3 as poorly differentiated. For the purpose of statistical analysis, the small sample of poorly differentiated tumors was grouped with moderately differentiated adenocarcinomas. Staining was categorized based on the intensity and location of staining as well as the percentage of tumor cells staining positive. The percentage of HMGCS2-positive cells was expressed as a ratio of positive cells to the total number of cells counted. The cutoff used was 35%, which was the mean of positive cells among the 59 cases with adenocarcinoma. Thirty-two carcinomas expressed low levels of HMGCS2, whereas the remaining 27 showed high expression. A good correlation was observed between the grade of tumor differentiation and the percentage of cells expressing HMGCS2 (P = 0.003; Table 1). Fourteen well-differentiated carcinomas expressed HMGCS2 in >35% of cells, whereas only 5 well-differentiated tumors showed low expression of this protein. In contrast to the well-differentiated tumors, the moderately and poorly differentiated carcinomas tended to show low expression of HMGCS2. Only 13 moderately and poorly differentiated tumors expressed HMGCS2 in >35% of cells.

The Myc transcription factor affects the expression of >10% of human genes and is a master switch that couples metabolism to cell growth and proliferation. Myc has been associated with energy metabolism through its up-regulation of glycolysis (29) or mitochondrial biogenesis (30). Here, we have shown that HMGCS2, the gene that controls the anabolic ketogenic pathway, is a Myc-target gene expressed in differentiated cells of the human colon. This expression pattern is consistent with the Myc and Miz-1 expression gradients along the axis of the crypt (28). The role of Myc in HMGCS2 down-regulation is supported by chromatin immunoprecipitation experiments, which showed occupancy of the HMGCS2 promoter by Myc. In agreement with our results, the overexpression of Myc in the liver of transgenic animals down-regulates HMGCS2 expression and ketogenesis (31). The expression pattern of HMGCS2 in differentiated cells of the colon and its down-regulation by Myc indicates that this gene may be under the control of the Wnt/APC/β-catenin signaling pathway (28, 34). In concordance with this hypothesis, the analysis of gene expression by microarrays shows that HMGCS2 is up-regulated by the expression of a dominant-negative form of TCF-4 in the colon cell line Ls147T (28).

Myc can directly repress gene expression through several mechanisms. The Inr-dependent mechanism implies abrogation of the Miz-1 transactivation effect by Myc/Max heterodimers (34). Alternatively, the Inr-independent mechanism involves Myc interaction with proteins, such as Sp1 (36). However, the Sp1-binding site has also been reported to participate in Miz-1-mediated transactivation (38). The HMGCS2 transcription start site has been established by primer extension, S1 nuclease protection, and rapid amplification of cDNA ends experiments in rat (11), pig (41), and human (15) genes and does not contain a canonical Inr sequence. Mutation of critical elements in the HMGCS2 proximal promoter diminished its transcriptional activity and therefore hinders the study of Myc transrepression. However, the Sp1-binding site, present in the proximal promoter of the gene, recruits Miz-1 and Myc transcription factors (Fig. 4D). This Sp1-binding site is involved in the transactivation mediated by Miz-1 (Fig. 4B), whereas Miz-1-mediated transactivation is blocked by Myc expression (Fig. 4C). Therefore, HMGCS2 expression seems to be under the control of a Inr-independent mechanism that still involves Miz-1 transactivation, which is blocked by Myc. Other redundant indirect mechanisms may also explain HMGCS2 down-regulation by Myc. Thus, the HMGCS2 gene is a target of HDAC1 (14) and HDAC2 (data not shown). Moreover, HDAC2 may be induced by Myc (42). Interestingly, the effect of histone deacetylase inhibitors is mediated by the same Sp1-binding site that participates in Miz-1 transactivation (14).

HMGCS2 is an inducible gene that controls ketone body production in the liver (1-3) under physiologic conditions, such as those occurring during suckling (4) or the fed-starved transition (5). HMGCS2 regulation occurs mainly at the transcriptional level. The proximal promoter of the gene contains elements that could account for its regulation by hormones, such as insulin (11), and metabolites, such as long-chain fatty acids or butyrate (13, 14). Insulin-mediated regulation of HMGCS2 involves forkhead transcription factors (12). Withdrawals of Foxo1 from the nucleus, as well as down-regulation by Myc, are critical steps in the regulation of cell proliferation. Thus, in addition to hormonal and metabolic signals, HMGCS2 seems to be down-regulated by signals that induce cell proliferation. Consistent with this hypothesis, the expression of this gene is down-regulated in a rat colon cancer model with the food-borne carcinogen, 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (43), or during liver regeneration after partial hepatectomy (44).

The molecular mechanism proposed to explain the colon cell-specific expression of HMGCS2 observed implies that the expression of this gene may be altered in cancer. The gene was down-regulated in 90% to 80% of colon, small intestine, and rectum tumors. However, only the colon and rectum tumors were Myc dependent. In addition, because of the high variability of HMGCS2 expression, we were unable to establish a straight correlation between Myc and HMGCS2 expression in healthy tissues. This variability in HMGCS2 expression was not related to levels of Myc. Therefore, although HMGCS2 is a direct target of Myc, the expression levels of these two genes depend on additional factors, which may be related to differences in nutritional status (4-6) or the intestinal germ flora (18, 22) of the patients. Down-regulation of HMGCS2 expression in proliferating cells was also observed at the protein level. The analysis of a tissue array-containing samples from 64 colorectal adenocarcinoma specimens indicates that HMGCS2 expression is preferentially down-regulated in moderately and poorly differentiated carcinomas when compared with well-differentiated tumors.

HMGCS2 expression has been associated with the ketogenic capacity of the colon (18). In addition, we have recently proposed that HMGCS2 forms part of a feed-forward mechanism in the colon to facilitate mitochondrial oxidation of butyrate (14). Nevertheless, we also speculate that ketone bodies are cell signaling molecules; therefore, the strict control of HMGCS2 expression is involved in a downstream cell signaling process. In this context, it has been shown that the ketone body acetoacetate, through increased oxidative stress, regulates gene expression (45), inhibits growth inhibition in human endothelial cells (46), and activates the extracellular signal-regulated kinase 1/2 and p38 mitogen-activated protein kinase signaling pathways in primary cultured rat hepatocytes (47).

Metabolism is an essential process for life and perturbations may have profound implications. Metabolism is tightly regulated by diet. Thus, there are different dietary manipulations capable to affect HMGCS2 expression (4-6, 18, 22). In this article, we show that HMGCS2 is down-regulated in poorly differentiated carcinomas. The relevance of HMGCS2 expression in tumor progression remains uncovered. However, we propose that some of these specific dietary manipulations can affect cancer incidence if ketone body production, or HMGCS2 expression, plays a role in tumor development/progression.

The human colon cancer cell line CaCo-2, obtained from the American Type Culture Collection (Rockville, MD), was maintained in DMEM supplemented with antibiotics and 10% fetal bovine serum at 37°C and 5% CO₂. CaCo-2 cells were cultured in medium alone for up to 5 and 21 days, with changes of medium every 2 days to obtain proliferating and differentiated cells for Northern blot experiments. Trichostatin A was obtained from Calbiochem (La Jolla, CA) and was used at a concentration of 0.3 μmol/L.

The rat HMGCS2 promoter-luciferase fusion plasmids −116pGl3, −62pGl3, and −34pGl3 were constructed by HindIII subcloning of the −116 to +28 fragment of pSMpCAT6A, the −62 to +28 fragment of pSMpCAT6D, and the −34 to +28 fragment of pSMpCAT7 into the pGL3Basic firefly luciferase reporter gene (Promega, Madison, WI). The plasmids pSMpCAT6A, pSMpCAT6D, and pSMpCAT7 were obtained as described previously (11, 13, 48). The −62 (M-Sp1) construct was generated by PCR amplification from −116pGl3 (M-Sp1) using the oligonucleotides 5′-aagcttTAAGTGGATCTAGAGCTTGCAGAGGCTG-3′ corresponding to coordinates −62 to −35 and 5′-aagcttCCAGCAGTCCACAGCCCCTGCG-3′ corresponding to coordinates +28 and +7. Nucleotides in lowercase were added to provide HindIII sites at the 5′ and 3′ ends. The PCR product was cloned into the p-GEM-T vector and subsequently subcloned into HindIII-cut pGL3Basic. The plasmid −116pGl3 (M-Sp1) was obtained as described previously (14). All construction sequences were confirmed by the dideoxynucleotide chain termination method using an automated fluorescence-based system (Perkin-Elmer, Wellesley, MA). Plasmid pcDNA3-hMS was constructed by insertion of the EcoRI fragment excised from pBS-MSH (20), corresponding to the cDNA of human HMGCS2, into the EcoRI site in the polylinker of the eukaryotic expression vector pcDNA3. The pUHD-Miz-1 and pcDNA3-Miz-1-Flag expression plasmids (32) were a gift from Dr. M. Eilers (Zentrum für Molekulare Biologie Heidelberg, Heidelberg, Germany). The pME18S, pME18S-c-Myc, and pME18S-c-Myc D348-493 (49) were kindly provided by Dr. Javier Leon (Universidad de Cantabria, Cantabria, Spain).

Hybridizations were done following standard methods (50) from extracted total RNA or cDNA immobilized on the Cancer Profiling Array II (Clontech, Palo Alto, CA). Total RNA was extracted from CaCo-2 cell lines with Trizol (Life Technologies, Gaithersburg, MD) and Northern blot analyses were done with 15 μg total RNA/lane using 2.2-kb human HMGCS2 cDNA, obtained by digesting pBS-MSH (20) with EcoRI, as a probe. Cancer Profiling Array II included normalized cDNA from tumor and corresponding healthy tissues from individual patients. This array was hybridized with a 3′ cDNA human HMGCS2 probe, a 789-bp fragment obtained by HindIII digestion of the pcDNA3-hMS plasmid. The probes for Miz-1 and Myc were the 2.4-kb fragment and the 1- to 4-kb fragment of human cDNA prepared by EcoRI/XbaI digestion of the pUHD-Miz-1 and pME18S-c-Myc plasmids, respectively. The alkaline phosphatase DNA probe corresponds to a 739-bp fragment of the human alkaline phosphatase cDNA and was obtained by reverse transcription-PCR as follows: 1 μg CaCo-2 RNA treated with butyrate was used in the reverse transcriptase reaction. Then, one fourth of the reverse transcriptase reaction was PCR amplified using the oligonucleotides 5′-tctagaCAGCTGAGGAGGAGAACCCGGC-3′ and 5′-ctcgagGTGCTTTGCCAGCCATTCCTGC-3. Nucleotides in lowercase were added to provide XbaI/XhoI sites at the 5′ and 3′ ends, respectively, for subsequent cloning.

CaCo-2 cells were seeded at a density of 2 × 10⁴/cm² 12 hours before transfection. Cells were transfected using the FuGene 6 reagent following the manufacturer's instructions (Roche, Indianapolis, IN) with 2 μg of the reporter 3-hydroxy-3-methylglutaryl-CoA synthase-luciferase gene construct and the amounts of the eukaryotic expression vector indicated in the figures or figure legends. In all cotransfection experiments, the amount of transfected DNA was kept constant by including salmon sperm DNA. After 48 hours, cell lysates were collected for luciferase assay. The pRL-CMV plasmid (Promega) was included in the transfection procedure to correct for variation in transfection efficiency. In the transfection experiments in which Myc was overexpressed, luciferase activity was corrected for the amount of protein assayed (51) because of the regulation of the pRL-CMV control plasmid expression by the expression of Myc.

Mev-SM cells, expressing rat HMGCS2, were obtained as described previously (39). Mev-SMh cells, expressing human HMGCS2, were obtained in a similar way by stable transfection of Mev-1 cells with the plasmid pcDNA3-SMh and were selected for their capacity to grow in the absence of mevalonate (39).

Recombinant adenoviruses expressing both green fluorescent protein and hemagglutinin-tagged c-Myc were purchased from Vector Biolabs (Philadelphia, PA). As a negative control, we generated adenoviruses expressing only green fluorescent protein. Recombinant adenoviruses were generated using the Ad-Easy system. The parent plasmids (pAdEasy-1 and pAdTrack-CMV) were a gift form B. Vogelstein (The Johns Hopkins Oncology Center). All adenovirus vectors were produced in HEK293 cells and purified as described previously (52). Differentiated CaCo-2 cells were infected with recombinant adenoviruses for 4 hours. Cells were left standing for 48 hours.

Cells were harvested using the passive lysis method (Promega) and luciferase assays were done using the Dual Luciferase Reporter Assay System (Promega). Luciferase activity was measured by means of a TD-20/20 luminometer (Turner Designs, Sunnyvale, CA).

Forty-eight hours after transfection, HeLa cells were lysed by sonication in HKMG [10 mmol/L HEPES (pH 7.9), 100 mmol/L KCl, 5 mmol/L MgCl₂, 10% glycerol, 1 mmol/L DTT, 0.5% NP40] containing protease inhibitors. Cellular debris was removed by centrifugation. Cell extracts were precleared with Tetralink Tetrameric Avidin Resin (Promega) for 1 hour and then incubated overnight with 1 μg biotinylated double-stranded oligonucleotides and 10 μg poly(deoxyinosinic-deoxycytidylic acid). DNA-bound proteins were collected with Tetralink Tetrameric Avidin Resin for 2 hours, washed four times in HKMG buffer, separated on a SDS-polyacrylamide gel, and identified by Western blotting. M2 anti-Flag monoclonal antibody (Sigma-Aldrich, St. Louis, MO) was used to detect Flag epitope-tagged proteins. Sp1 was detected using PEP2 antibody from Santa Cruz Biotechnology, Santa Cruz, CA. Biotinylated oligonucleotides comprising the Sp1 site of HMGCS2 (5′-GATCTGTGGAGGCGGGCAAGCT-3′: the consensus Sp1-binding site is shown in bold capital letters) and the mutated Sp1 site (5′-GATCTGTGGAtctaGaGCAAGCT-3: mutated nucleotides are shown in bold lowercase letters) were synthesized by Sigma.

Formaldehyde cross-linking and chromatin immunoprecipitation assays of CaCo-2 cells were done as described previously (14). Immunoprecipitations were done using 20 μg anti-c-Myc (Santa Cruz Biotechnology) or preimmune serum (Ig).

Polyclonal antibodies against human mitochondrial 3-hydroxy-3-methylglutaryl-CoA synthase were elicited against a peptide conjugated to keyhole limpet hemocyanin corresponding to the COOH terminus of the protein (sequence YHKVNFSPPGDTNS, amino acids 471-484; Sigma, Genosys). Immunoblotting was carried out using an enhanced chemiluminescence kit following the manufacturer's instructions.

Human colonic tissues embedded in paraffin were provided by the Banco de Tejidos y Tumores (Institut d'Investigació Biomèdica August Pi i Sunyer), Hospital Clínic de Barcelona (Barcelona, Spain). The Colon Tumor Tissue Array slides were purchased by Lab Vision. The sections for HMGCS2 analysis were antigen retrieved by preheating samples to 60°C for 30 minutes and then allowing them to cool for 5 minutes. Paraffin was removed from the slides using xylene and sections were then rehydrated by ethanol treatment followed by permeabilization in 0.1% Triton X-100, 20 mmol/L glycine, and 100 mmol/L PBS at room temperature for 10 minutes. Blocking was done for 20 minutes in 1% bovine serum albumin, 20 mmol/L glycine, and 100 mmol/L PBS and the sections were then incubated with the specific human HMGCS2 antibody at a dilution of 1:250 for 1 hour at 37°C. After washing the primary antibody with 20 mmol/L glycine and 100 mmol/L PBS (pH 7.4), the slides were incubated with a biotinylated goat anti-rabbit secondary antibody (Sigma) at a dilution of 1:250 for 30 minutes at 37°C. The sections were washed with 20 mmol/L glycine and 100 mmol/L PBS (pH 7.4) and then incubated with horseradish peroxidase-streptavidin (Zymed, San Francisco, CA) at 1:250 dilution for 30 minutes at 37°C. After washing the slides with 20 mmol/L glycine and 100 mmol/L PBS (pH 7.4) again, staining was detected using the AEC kit (Zymed) following the manufacturer's instructions. The slides were counterstained with hematoxylin (Sigma) and mounted with aqueous medium (Sigma).

Statistical correlation between HMGCS2 expression and tumor degree was determined using the two methods with SPSS statistical software (SPSS Worldwide Headquarters, Chicago, IL). Ps < 0.05 were considered statistically significant.