Expression of 15-Lipoxygenase-1 in Human Colorectal Cancer

Colorectal cancer is the second leading cause of cancer death in the United States. Recently, epidemiological studies demonstrated the association between long-term intake of nonsteroidal anti-inflammatory drugs and a reduced mortality from colorectal cancer (1, 2, 3). In as much as prostaglandin H synthase (Cox²) is the target of nonsteroidal anti-inflammatory drugs and high levels of prostaglandins have been detected in human colon tumors (4), suppression of prostaglandin production by the inhibition of Cox activity may inhibit colorectal carcinogenesis. Two isoforms of Cox have been identified (5). Cox-1 is constitutively expressed in most tissues; Cox-2 is not but is rapidly induced by a variety of stimuli. Several studies have demonstrated elevated expression of Cox-2, and not Cox-1, in human colorectal cancer tissue (6, 7, 8, 9), and expression of Cox-2 is considered to be an early event in colon carcinogenesis (10).

In addition to the Cox pathway, another arachidonate metabolic pathway, the lipoxygenase pathways are reported to participate in the development of cancer (11). Lipoxygenases are dioxygenases that recognize the 1,4-pentadiene structure of polyunsaturated fatty acids and incorporate single molecules of oxygen at specific carbon atoms of substrate fatty acids (12, 13). In human tissue, three major lipoxygenases have been characterized according to the oxygenation sites in the substrate arachidonic acid. These enzymes are named 5-, 12-, and 15-lipoxygenase. The 15-lipoxygenase forms 15(S)-hydroperoxy-5Z,8Z,11Z,13E-eicosatetraenoic acid, which is reduced to 15(S)-HETE. A previously known reticulocyte type 15-lipoxygenase has been termed 15-Lox-1, whereas a newly discovered 15-lipoxygenase (14) is called 15-Lox-2 (15). Unlike 5-lipoxygenase, which produces the leukotrienes that are biologically potent metabolites and mediate inflammation and asthma, the role of 15-lipoxygenases remains ambiguous (16). However, human 15-lipoxygenase is thought to be involved in inflammation (17, 18), reticulocyte differentiation (16, 19, 20), and atheroma formation (21, 22, 23, 24). Although no general concept for its biological role has been presented, there are two hypotheses that explain its mechanism of action (25). First, the bioactive metabolite is produced to mediate biological activities: (a) the metabolites may modify a host of cellular immune functions and may modulate other oxygenation enzymes (26); (b) 15-Lox-1 peroxidizes esterified polyunsaturated fatty acids in membranes, thus modifying the structure and function of lipid-protein complexes. For example, during maturation of reticulocytes to erythrocytes, peroxidation of mitochondria and other intracellular organelles (16) occurs, which destabilizes membrane structure and predisposes membrane to proteolysis during erythrocyte maturation. The 15-lipoxygenase appears to play a role in the induction of atherosclerosis. Oxidation of low density lipoprotein (21, 22, 23, 24) results in the formation of 13(S)-HODE from linoleic acid. This lipid metabolite is a ligand for the peroxisome proliferator-activated receptor γ, and the activation of gene expression by 13(S)-HODE appears to play an important role in foam cell development and pathogenesis of atherosclerosis (27).

Recently, we observed the induction of 15-Lox-1 in a sodium butyrate-stimulated human colon cancer cell line, Caco-2. In that study, the 15-Lox-1 pathway appears to modulate apoptosis or the differentiation of these colorectal carcinoma cells (28). Furthermore, incubation with indomethacin, a Cox-2 inhibitor, enhanced the expression of 15-Lox-1 during the process of apoptosis and cell differentiation. An important question is whether 15-Lox-1 is expressed in human colon tissue and whether its expression is increased in colorectal tumors. In this study, we report the expression of a 15-lipoxygenase in human colorectal tissue and its characterization as 15-Lox-1. A significant increase in the level of 15-Lox-1 expression was observed in most tumor tissue compared with the adjacent normal tissue. As reported in several previous studies (6, 7, 8, 9), Cox-2 expression was elevated, whereas the level of Cox-1 was frequently reduced in the tumor tissue. Furthermore, immunohistochemical studies indicated that 15-Lox-1 was present in the human colorectal epithelial cells. These results suggest that in addition to Cox-2, 15-Lox-1 may also play a role in colorectal carcinogenesis.

Anti-human recombinant 15-Lox-1 antibody (20, 29) was a generous gift from Dr. Elliott Sigal at the University of California at San Francisco. Anti-human Cox-1 monoclonal antibody, ovine Cox-1 purified protein, 13(S)-HODE, 15(S)-HETE, 12(S)-HETE, and 5(S)-hydroxy-6E,8Z,11Z,14Z-eicosatetraenoic acid were purchased from Cayman (Ann Arbor, MI). BCA protein assay kit was from Pierce (Rockford, IL). Murine recombinant Cox-2 protein was a generous gift from Monsanto (St. Louis, MO). Anti-human Cox-2 polyclonal antibody was purchased from Oxford Biomedical Research (Oxford, MI). Nonimmune rabbit serum was purchased from Jackson Immunoresearch Laboratories (West Grove, PA). Nonfat dry milk was purchased from Bio-Rad (Hercules, CA). Anti-rabbit immunoglobulin, horseradish peroxidase-linked whole antibody and Hybond ECL nitrocellulose membrane and ECL detection system were purchased from Amersham Life Science (Arlington Heights, IL). [1-¹⁴C]Arachidonic acid was purchased from NEN Life Science Products (Boston, MA). Vectastain rabbit ABC-Elite kit and normal goat serum was purchased from Vector Laboratories (Burlingame, CA). Taq DNA polymerase was purchased from Life Technologies (Gaithersburg, MD). QIAquick PCR purification kit was from QIAGEN (Valencia, CA). The restriction enzymes HindIII or PstI were from Boehringer-Mannheim (Indianapolis, IN). Automation buffer was purchased from Biomeda (Foster, CA).

Twenty-one surgically resected colorectal tumor samples and matched adjacent normal tissues were obtained from the University of North Carolina, Lineberger Comprehensive Cancer Center. The tissue was obtained with the approval of the University of North Carolina and National Institute of Environmental Health Sciences local boards governing research on human subjects. Of the 21 paired tissue samples, 12 were obtained from males, and 8 were from females with one sample number 10, for which the records were not available (Table 1). Also reported in Table 1 is the degree of differentiation, the age of the patient, the localization of the tumor, the clinical stage, and the histological evaluation. The tumors range from poorly to well differentiated and were mainly adenocarcinomas, except for case 2, which was dysplasia and cases 11 and 18 which were benign polyps. These samples were stored at −70°C. For Western analysis, ∼100–500 mg of each sample were crushed with a pestle and mortar in a liquid nitrogen bath and transferred to a plastic tube on ice. After the addition of five volumes of ice-cold 50 mm Tris-HCl (pH 7.4), containing 1 mm EDTA, the samples were homogenized by Polytron (Brinkmann Instrument, Westbury, NY) at 4°C. The homogenate was centrifuged at 10,000 × g for 10 min at 4°C, and the supernatants were used for Western blot analysis. Protein concentration was determined by the BCA protein assay with BSA as a standard. Homogenates from the nine paired samples of the most recently collected tissues were used to measure 15-lipoxygenase enzymatic activity by an HPLC assay for the conversion of arachidonic acid to 15-HETE.

Twenty μg of the samples were analyzed by SDS-PAGE using 8% acrylamide gels. Human tracheobronchial epithelial cells (NHTBE cells strain 2002; Clonetics Corp., San Diego, CA) were cultured with retinoic acid and incubated with IL-4 to induce expression of 15-Lox-1 for use as a standard (30). Murine recombinant Cox-2 protein or ovine Cox-1 purified protein was also loaded on the gels with the experimental samples. After electrophoretic transfer of the protein from the polyacrylamide gel to nitrocellulose membrane, nonspecific binding was blocked by incubation with 5% nonfat dry milk in TBST [10 mm Tris (pH 8.0), 150 mm NaCl, and 0.05% Tween 20] for 1 h at room temperature. After being washed three times in TBST, the membrane was probed with a polyclonal anti-human 15-Lox-1 antibody diluted 1:20,000 in TBST/1% dry milk or with polyclonal anti-human Cox-2 antibody diluted 1:4,000 in TBST/1% dry milk or with monoclonal anti-human Cox-1 antibody diluted 1:4000 in TBST/1% dry milk for overnight at 4°C. The membrane was then washed three times with TBST and incubated for 1 h at room temperature with an anti-rabbit (for 15-Lox-1 and Cox-2) or anti-mouse (for Cox-1) immunoglobulin peroxidase conjugated antibody diluted 1:5000 in TBST. The membrane was washed three more times with TBST and analyzed by the enhanced chemiluminescent detection system. The intensity of the positive signal was quantitated with a computing densitometer (Molecular Dynamics, Sunnyvale, CA).

Cell homogenates (total protein, 2–5 mg) were incubated with 25 μm [1-¹⁴C] arachidonic acid (1.0 × 10⁶ cpm; 15 nmol) in ethanol (2.5% of final volume) at 30°C for 15 min. The reaction mixture (600 μl) contained 50 mm Tris-HCl (pH 7.4) and 5 mm CaCl₂ with or without 20 μm indomethacin or 40 μm NDGA. After the incubation, 0.30 mg of sodium borohydride was added, and the mixture was kept on ice for 15 min and then was acidified to pH 3 with HCl. 13(S)-HODE was added as an internal standard. The sample was extracted with 2 ml of ethyl ether, the solvent was evaporated, and the dried material was dissolved in 50 μl of methanol:water (3:1) solvent. Arachidonic acid metabolites were analyzed by reverse-phase HPLC with a Beckman ODS Ultrashere column (5 μm, 4.6 × 250 mm). The solvent system was methanol:H₂O:acetic acid (75:25:0.01), and the column was washed at a flow rate of 1.0 ml/min. The radioactive eluent was monitored on-line with a Flo-One radioactivity detector (Packard, Downers Grove, IL) with EcoLume liquid scintillation cocktail (ICN Biochemicals, Costa Mesa, CA). Absorbance at 235 nm was monitored with a Waters 486 Tunable Absorbance Detector (Millipore, Milford, MA).

Total RNA was extracted from the frozen tumor samples by the use of TRI Reagent (Sigma, St. Louis, MO) according to the manufacturer’s protocol. First-strand cDNA was generated using 1 μg of total RNA as template, and a RT reaction was carried out to synthesize cDNA using Advantage RT-for-PCR kit (Clontech, Palo Alto, CA). The reaction solution was diluted to a final volume of 100 μl. Two μl of the cDNA solution were combined with the 5′ primer and 3′ primer, which were specific for 15-Lox-1 cDNA (EZ-Prime 15 LOX primers; Oxford Biomedical Research, Oxford, MI; Ref. 31). With the same cDNA, specific primers for 15-Lox-2 cDNA (5′-TG-CCT-CTC-GCC-ATC-CAG-CT-3′ and 5′-TG-TTC-CCC-TGG-GAT-TTA-GAT-GGA-3′; Ref. 14) were also used to detect the expression of 15-Lox-2 mRNA. The samples were then subjected to 35 cycles of denaturation (45 s at 94°C), annealing (45 s at 60°C), and extension (2 min at 72°C), followed by a final extension at 72°C for 7 min in the presence of 2 units of Taq DNA polymerase. The cDNA fragment obtained from primers of 15-Lox-1 was purified with a QIAquick PCR purification kit and digested with HindIII or PstI. The fragment was analyzed by electrophoresis on a 4% agarose gel.

Colorectal tumor and matched adjacent normal tissues were fixed in 10% neutral buffered formalin, processed routinely, and embedded in paraffin. Localization of 15-Lox-1 protein expression was investigated using the polyclonal rabbit anti-15-Lox-1 IgG antibody (1:200 dilution) on serial 6-μm sections. As a positive control tissue for 15-Lox-1 immunostaining, a section of human bronchus was immunostained (32). Slides were deparaffinezed in xylene and hydrated through a graded series of ethanols to 1× Automation buffer. Endogenous peroxidase activity was blocked with 3% H₂O₂ for 15 min. After rinsing in 1× Automation buffer, the sections were digested 15 min at 37°C with trypsin (0.25%), rinsed in distilled water, and blocked with 5% normal goat serum for 30 min. The primary antibody, 15-Lox-1, was then applied, and sections were incubated overnight at 4°C. Nonimmune rabbit IgG was used as the negative control at equivalent conditions in place of the primary antibody. The bound primary antibody was visualized by avidin-biotin-peroxidase detection using the Vectastain Rabbit Elite kit according to the manufacturer’s instructions and with 3,3′-diaminobenzidine as the color-developing reagent. Slides were counterstained with Harris hematoxylin, dehydrated through a graded series of ethanol to xylene, and coverslipped with Permount (Fisher, Springfield, NJ).

The results of immunoblot analysis were analyzed by the Wilcoxon sign-rank test to determine the significance of the difference between expression of 15-Lox-1, Cox-2, and Cox-1 in paired samples of adjacent normal and tumor tissue.

To examine the expression of 15-Lox-1, protein homogenates from paired tumor and adjacent normal colorectal tissues were examined by Western blotting (Table 1). Fig. 1 is a representative immunoblot obtained from six of the paired samples incubated with anti-human 15-Lox-1 antibody. A major band at M_r ∼70,000 was observed that comigrated with the 15-Lox-1 standard (30). The immunoblot results for 15-Lox-1 expression from all 21 paired samples were quantified by densitometric analysis, and the relative density was compared with the 15-Lox-1 standard loaded on each gel. The data (Table 2) are reported as the ratio of the experimental to the standard densitometry values. 15-Lox-1 was expressed in most of the samples in both adjacent normal and tumor tissue. The expression of 15-Lox-1 was at the limits of detection in adjacent normal and tumor tissue obtained from cases 7, 8, and 9 (Tables 1 and 2).

To further confirm the expression of 15-Lox-1 in colorectal tissue, we examined the metabolism of radiolabeled arachidonic acid to 15-HETE. The crude 10,000 × g supernatants of colorectal tissue samples, which had been examined previously by immunoblot and shown to express 15-Lox-1, was incubated with [1-¹⁴C]-labeled arachidonic acid. The borohydride reduced products were then separated and analyzed by reverse-phase HPLC. Fig. 2 is a representative HPLC profile obtained from one of the incubations from the nine paired tissue samples examined. A radiolabeled metabolite (retention time, 36.0 min), which absorbed at 235 nm, as expected of a conjugated diene, was the major species. This metabolite comigrated with authentic 15(S)-HETE. A minor metabolite (retention time, 40.6 min) was detected and coeluted with 12(S)-HETE. The ratio of 15-HETE:12-HETE was 6.5:1. Incubation with 20 μm indomethacin did not suppress the formation of this metabolite, but 40 μm NDGA inhibited formation (data not shown). These data support the conclusion that 15-Lox-1 is expressed in human colorectal tumor tissue.

To characterize further the 15-Lox-1 present in the colon samples, total RNA isolated from tumor samples was subjected to RT-PCR amplification with the primers specific for either human 15-Lox-1 or 15-Lox-2 cDNA. The amplified cDNA fragment (952 bp, Lane 1) from colon mRNA was detected with the primers of 15-Lox-1 and digested with HindIII (Lane 2) or PstI (Lane 3). The predicted fragments (648 and 304 bp by HindIII; 390, 291, and 271 bp by PstI) were detected (Fig. 3). In contrast, 15-Lox-2 mRNA was detectable by RT-PCR with primers specific for this isozyme as a very faint band (data not shown). We also sequenced the 952-bp PCR product using the same primers for 15-Lox-1, and the sequence (33) was identical to the human 15-Lox-1 sequence (data not shown). These results confirm the expression of 15-Lox-1 mRNA in the human colorectal tumor.

We also examined Cox-1 and Cox-2 expression in these paired colorectal samples by Western blotting. A protein of approximately M_r 68,000–72,000 was observed that comigrated with Cox-1 and Cox-2 standards. The Cox-2-specific antibody also reacted with proteins that migrated approximately as M_r 35,000 and M_r 38,000. These species were cleavage peptides of Cox-2, formed as a result of degradation, and were most apparent in the tumor samples. The immunoblot results for Cox-1 and Cox-2 expression (including the M_r 35,000 and M_r 38,000 bands) from all 21 paired samples were quantified by densitometric analysis, and the relative density was compared with the standard loaded on each gel. The expression of Cox-1 and Cox-2 was not detected in the adjacent normal and tumor tissue obtained from cases 18, 19, 20, and 21 for Cox-1 and cases 4 and 5 for Cox-2 (Table 2).

Inspection of the immunoblots obtained from the analysis of the paired samples and the relative expression on each blot compared with the authentic standards suggests that 15-Lox-1 and Cox-2 expression is higher in most tumor samples compared with the adjacent normal tissue, whereas the expression of Cox-1 is frequently lower in the tumor tissue (Table 2). Expression of 15-Lox-1 was noted in 18 of the 21 cases, and higher expression was observed in the tumor tissue in 14 of the 18 samples. As reported in several studies (6, 7, 8, 9), Cox-2 was expressed in most tumor samples (19 of 21 cases) and in those 19 cases, 17 pairs showed higher levels of expression in the tumor tissue than in adjacent normal tissue. Cox-1 was also expressed in both tumor and adjacent normal tissue in most samples (17 of 21 cases), and in 12 pairs of those 17 positive cases, Cox-1 levels were frequently lower in tumor tissue compared with adjacent normal tissue. Fig. 4 shows the distribution for the relative expression of the three enzymes in tumor and adjacent normal tissue from the 21 cases. One of the adjacent normal colon samples (case 2 in Table 1) highly expressed 15-Lox-1. This case was dysplasia, not an adenocarcinoma, which may account for the high expression. From this figure, it is clearly evident that both expression of Cox-2 and 15-Lox-1 is higher in the tumor tissue when compared with their paired adjacent normal tissue, whereas Cox-1 expression is lower in the tumor tissue. The paired differences between adjacent normal and tumor tissue was further analyzed by a Wilcoxon signed-rank test (Table 3). The expression of 15-Lox-1 was significantly higher in the tumor tissue compared with its paired adjacent normal tissue (P < 0.05). The expression of Cox-2 was also significantly higher in the tumor tissue compared with its paired adjacent normal tissue (P < 0.05). In contrast, Cox-1 expression was significantly lower in the tumor than in the paired adjacent normal tissue (P < 0.05). Neither age nor sex is associated with these significant differences. This analysis supports the conclusion that there is an increased expression of 15-Lox-1 and Cox-2 but a decrease in expression of Cox-1 in human colorectal tumors.

The localization of 15-Lox-1 was determined by immunohistochemistry in colorectal tissue. We stained formalin-fixed, paraffin-embedded tissues using a polyclonal anti-15-Lox-1 antibody. Anti-15-Lox-1 immunostaining was localized to the cytoplasm of the epithelial cells of both tumor and adjacent normal tissues. Fibroblasts located within the stromal tissue were also positive for 15-Lox-1 (Fig. 5,A). However, staining was not observed in the Peyer’s patch of the intestine. Rabbit nonimmune IgG produced negative staining in all sections (Fig. 5 B). We could not make a quantitation of expression between tumor and adjacent normal tissue by the immunohistochemical method.

Recently, we observed the induction of 15-Lox-1 in sodium butyrate-stimulated Caco-2 cells (28). In this report, we present evidence for the expression of 15-Lox-1 in human colorectal cells and report a significantly higher expression of 15-Lox-1 in the tumor sample compared with the adjacent normal tissues. Crude colorectal enzyme preparation metabolized arachidonic acid to 15-HETE, as determined by reverse-phase HPLC analysis. The profile showed the production of 15-HETE as well as a much lower level of 12-HETE (6.5:1). This finding supports the conclusion that 15-Lox-1, but not 15-Lox-2, is expressed in the colorectal tissue because 15-Lox-1 was reported to form 10–20% 12-HETE in addition to 15-HETE (34), and 15-Lox-2 forms only 15-HETE (14). In addition, we detected the expression of 15-Lox-1 mRNA by RT-PCR using specific primers for 15-Lox-1 cDNA. The sequence of this PCR fragment was identical to that of 15-Lox-1. Using the specific primers for 15-Lox-2, 15-Lox-2 mRNA was not detectable by RT-PCR. Several human tissues have been reported to contain both 15-Lox-1 and 15-Lox-2 (14), but our findings agree with the report (14) that 15-Lox-2 is not expressed in human colorectal tissues. The antibody we used for immunoblots was specific to the human recombinant 15-Lox-1 (20, 29) and shows little cross-reactivity to 15-Lox-2 because the amino acid sequence identity of 15-Lox-2 to 15-Lox-1 is only 40% (14). We screened 21 cases of surgically resected colorectal tissues with the antibody to 15-Lox-1 (Tables 1 and 2). Eighteen of 21 samples showed the expression of 15-Lox-1 in both tumor and adjacent normal tissue. Of the 18 15-Lox-1-positive samples, 14 tumors showed a higher expression level than the adjacent normal tissue. Thus, we conclude that the expression of 15-Lox-1 was increased in colon tumors. Furthermore, Cox-2 was expressed in 19 samples with higher expression observed in 17 of the tumor tissues. Cox-1 was observed in 17 samples with lower levels of expression observed in 12 of the tumor tissues. Our results on the expression of Cox-1 and Cox-2 are in agreement with previous reports in the literature (6, 7, 8, 9). Thus, we present evidence that both Cox-2 and 15-Lox-1 expression are higher in colorectal tumors, whereas Cox-1 expression is frequently lower in the adjacent normal tissue. Several studies suggested the presence of 15-lipoxygenase activity in human colon tissue (35, 36, 37), but in those studies, the enzyme was not characterized. To our knowledge, this is the first report that provides direct evidence for the expression of 15-Lox-1 in human colorectal tissue and, more importantly, elevated levels of expression in most of colorectal tumors examined.

Because Western analysis indicates the expression of 15-Lox-1 in human colorectal tissue, we determined the cellular localization of 15-Lox-1. Immunohistochemistry demonstrated the localization of 15-Lox-1 primarily in the epithelium. Contaminating blood cells, including eosinophils (20) and monocytes (31), also express this enzyme and could contribute to the detected expression of 15-Lox-1. However, we did not detect contaminating or resident blood cells. 15-Lox-1 is expressed primarily in epithelium, whereas some expression was observed in fibroblasts and stromal tissues.

The physiological importance of 15-Lox-1 expression in human colorectal tissue and increased expression in colorectal tumors is not clear. 15-HETE, the arachidonic acid metabolite formed by 15-Lox-1, is reported to have roles in both anti- and pro-inflammatory function, in the promotion of cell growth, in signal transduction events in mucous-secreting cells, and in the alteration of integrin expression and cell migration (20). In the epithelial barrier cells (in skin, gut, and airway), overexpression of 15-Lox-1 may catalyze membrane peroxidation that contributes to epithelial damage during inflammation (26). 15-Lox-1 appears to have a role in the regulation of apoptosis and cell differentiation (28) and cell proliferation in cell lines (38). Inhibitors of lipoxygenases can attenuate tumor growth (38), whereas other data suggest lipoxygenases block apoptosis by increasing the level of Bcl-2 (39) or by interacting with the interleukin 1β converting enzyme family cysteine proteases (40). Consideration must be given to the fact that linoleic acid is the preferred substrate for 15-Lox-1 and is metabolized to 13(S)-HODE. This lipid metabolite can modulate signaling pathways, thereby altering the tyrosine phosphorylation of signaling proteins (41), modulating cell growth and apoptosis. In this laboratory, we recently studied (28) sodium butyrate-induced apoptosis and cell differentiation in the human colorectal cell line, Caco-2, which highly expresses Cox-2 and mutant APC protein. The expression of 15-Lox-1 was only observed during apoptosis and cell differentiation. Inhibition of the lipoxygenase resulted in an enhancement of apoptosis, which supports the hypothesis that lipoxygenase metabolites act as inhibitors of apoptosis in this cell line. In the immunohistochemistry study, we observed the localization of 15-Lox-1 primarily in the epithelial cells. As the intestinal cells migrate up the crypt, they undergo differentiation; normal tissue homeostasis requires removal of the cells by the activation of apoptosis. Transformation of colorectal epithelial cells is associated with a progressive inhibition of apoptosis (42), and inhibition of apoptosis may be important in tumor development. The increased expression of Cox-2 observed in colorectal tumors also appears to have an important role in the attenuation of the apoptosis. Therefore, coordinated with Cox-2, 15-Lox-1 may act in colorectal tumor development by modulating apoptosis and/or cell differentiation in epithelial cells. We conclude that the expression of 15-Lox-1 in normal colorectal tissue and the increased expression observed in the tumors suggest a potential role for this enzyme in modulating apoptosis and cell differentiation of colorectal epithelium.

We thank Dr. Elliott Sigal for providing 15-Lox-1 antibody, Lynn Johnson and Evangeline Raynolds for collecting the samples, Drs. Natsuo Ueda and Yoshitaka Takahashi for helpful suggestions, and Mark Geller for technical assistance. We also thank Dr. Joseph Haseman for the analysis of the data by Wilcoxon signed-rank test.