Modulation of Intracellular β-Catenin Localization and Intestinal Tumorigenesis in Vivo and in Vitro by Sphingolipids¹

Many cancer susceptibility genes produce defects in the regulation of cell proliferation, differentiation, or apoptosis. One approach to cancer prevention and treatment would be to bypass these defects using naturally occurring or synthetic compounds. Sphingolipids are candidates for such intervention because their digestion products (sphingosine and ceramide) affect multiple intracellular targets that control cell growth and apoptosis, such as protein kinase C (1, 2), phosphatidylinositol 3-kinase (3, 4), other protein kinases and phosphoprotein phosphatases (5, 6), expression of inhibitors of cyclin-dependent protein kinases and other cell cycle-regulating proteins (7, 8, 9), induction of apoptosis via cytochrome c release from mitochondria (10), activation of effector caspases (11, 12), and inactivation of antiapoptosis factors (13), among other things.

In addition to inhibiting growth and inducing apoptosis in transformed cells in culture (14, 15, 16, 17), sphingolipids inhibit carcinogenesis in vivo. Topical application of sphingoid bases and ceramides to mouse skin treated with 7,12-dimethylbenz(a)anthracene reduces the number of carcinomas (18), and the feeding of complex sphingolipids to mice treated with DMH⁴ inhibits both early and late stages of colon carcinogenesis (19, 20, 21, 22). Chemotherapeutic uses of sphingolipids have also been explored by administration of sphingoid base analogues to nude mice injected with human tumor cells (23) and by a pilot clinical trial with safingol (l-threo-sphinganine) administered alone and in combination with doxorubicin (24).

These in vivo studies have been conducted with models that have multiple genetic mutations and provide little information about which specific genetic defects can be “bypassed” or “normalized” by sphingolipids. Of the multiple genetic mutations found in colon cancer, mutations in the APC gene and other defects in the APC/β-catenin pathway occur in a substantial portion of human colon cancers (25, 26, 27) and also as the result of an autosomal dominant human disease (familial adenomatous polyposis) that usually progresses to colon cancer. Modulation of the APC/β-catenin pathway can be studied in vivo using Min mice, which carry nonsense mutations in the APC gene and develop multiple intestinal neoplasia (28), as well as in vitro using human colon cancer cell lines. Therefore, this study used these models to discover that sphingolipids affect the biochemical alterations (i.e., reduce cytosolic accumulation of β-catenin) and tumorigenic consequences (i.e., reduce tumor number) caused by defects in the APC/β-catenin pathway.

Male Min mice (age, 32 days) were purchased from The Jackson Laboratory (Bar Harbor, ME), which maintained them on a regular diet before shipping. The mice were divided into experimental groups to achieve a similar initial mean weight, and housed five mice/cage in microisolator cages. They were monitored daily and weighed weekly for any signs of weight loss, distress, or sickness. The control group was fed an AIN 76A diet (Dyets, Bethlehem, PA), and the experimental groups were fed the AIN 76A supplemented with 0.1% (by weight) ceramide (N-palmitoylsphingosine; Avanti Polar Lipids, Alabaster, AL) or a mixture of complex sphingolipids in proportions similar to that in milk (sphingomyelin, 65%; glucosylceramide, 7.5%; lactosylceramide, 20%; and ganglioside G_D3, 7.5%; Matreya, Pleasant Gap, PA) or this mixture plus ceramide (60:40; sphingomyelin, 39%; ceramide, 40%; glucosylceramide, 4.5%; lactosylceramide, 12%; and ganglioside G_D3, 4.5%). The purity of the sphingolipids was confirmed by TLC and mass spectrometry (22). The diets were mixed fresh every 2 weeks and kept at 4°C in airtight containers. All protocols involving animals were approved by the Institutional Animal Care and Use Committee and conducted according to National Research Council Guidelines.

At 100 days of age (after 8 weeks of sphingolipid feeding), the mice were killed by CO₂ asphyxiation. Immediately afterward, blood was drawn by heart puncture, collected in heparin-treated Eppendorff tubes, and snap-frozen for later analyses. The intestines were excised, opened longitudinally, flushed with ice-cold PBS, and fixed flat overnight in 10% neutral buffered formalin. All sections of the intestine were examined by light microscopy, and tumor number and size were documented in a blinded manner. Randomly selected tumors were fixed, sectioned, and stained with H&E, and all were found to be adenomas.

For the immunohistochemical staining for β-catenin, intestinal sections from the Min mice and the genetic background C57BL/6J mice were embedded in paraffin and cut into 3–5-μm sections, which were deparaffinized with fresh xylene and rehydrated through graded alcohol (100, 95, 80, 50, 30, and 0%). After blocking with 3% BSA (Calbiochem) in PBS for 1 h at room temperature [for the C57BL/6J sections, 1% bovine serum, 1% BSA, 0.001% sodium azide, 0.2% nonfat milk, and 0.1% Tween 20 in 20 mm Tris-HCl, (pH 7.5) were added], the sections were incubated overnight at 4°C with the polyclonal antibody for β-catenin (Santa Cruz Biotechnology, Santa Cruz, CA). Serial sections were used as controls for antibody specificity (incubated with 100 ng of blocking peptide for the anti-β-catenin antibody from Santa Cruz Biotechnology; diluted 1:500), and background (incubated with nonspecific polyclonal goat antibody, raised against influenza HA protein, or blocking buffer only for determination of the background fluorescence of the secondary antibody). The resulting immunocomplexes were visualized with a FITC-conjugated secondary antigoat antibody (Sigma Chemical Co., St. Louis, MO), incubated for 60 min at room temperature, and then extensive rinses with PBS. Then the sections were covered with Mowiol (Calbiochem), an aqueous mounting solution supplemented with 0.25% diazobicyclooctane (an antifade reagent; Sigma Chemical Co.). After addition of a coverslip, the localization of the β-catenin was visualized using a Nikon E800 epifluorescence microscope equipped with DIC optics and FITC HYQ from Chroma (Brattleboro, VT). Images were processed using Adobe Photoshop software.

Min mice were killed by CO₂ asphyxiation, the abdominal cavity was opened, and the intestines removed, rinsed with PBS, and divided into the proximal, mid-, and distal small intestine (of approximately equal length) and colon. Each segment was opened longitudinally and added to a 3.0-ml incubation mixture that was prepared by mixing 500 μg each of phosphatidylcholine (from egg yolk; Sigma Chemical Co.), sodium cholate (Sigma Chemical Co.), and the sphingolipids (for the sphingolipid composition, see the legend to Fig. 3) in chloroform:methanol (1:1; v/v) and then evaporating the solvent under nitrogen, adding of 3 ml of potassium phosphate buffer (20 mm; pH 6.8), and sonicating until clear. After 0–6 h of incubation at 37°C, 100 μl aliquots were removed, the lipids were solubilized with 900 μl of methanol:chloroform (1:1; v/v), sonicated, centrifuged, and analyzed for the appearance of ceramide and free sphingoid bases by electrospray tandem mass spectrometry as described previously (22). The relative intensities of the metabolites are shown because the appropriate internal standards were not available at the time for quantitative analyses.

The lipids from blood drawn after CO₂ asphyxiation were extracted with methanol:chloroform (2:1; v/v) and free sphingoid bases were analyzed by the modified method of Merrill et al. (29) by reverse-phase high-performance liquid chromatography (Shimadzu, Columbia, MD) as described in Ref. 16. The resulting peaks were identified by comparison with known standards. Quantitation of the sphingolipid content was achieved by the use of C₂₀-sphinganine as an internal standard.

Cells were seeded in 100-mm dishes (1.5 × 10⁶), and the next day, the medium was removed and the new medium added with 2.5 μm sphingosine (Matreya) as the equimolar BSA complex, 2.5 μm C₂-ceramide (in ethanol), or with 2.5 μm long-chain ceramides (from bovine brain, in ethanol:dodecane, 98:2, by volume), or with only BSA, ethanol, or ethanol:dodecane as vehicle controls. After 4 or 6 h (SW480) or 24 h (T84), the cells were scraped off the plates and lysed with a hypotonic lysis buffer (1 mm NaCO₃, 0.2 mm PMSF, 1 μg/ml aprotinin, 1 μg/ml leupeptin, and 1 μg/ml pepstatin) for 30 min on ice. After passing through a 20-gauge needle (10 times), the lysates were centrifuged (2,000 × g for 5 min), the supernatants removed, and centrifuged at 4°C for 30 min at 15,000 × g (30, 31). Protein concentrations were determined by the Lowry assay (32) after protein precipitation with 5% trichloroacetic acid. Equal amounts of proteins [controlled by Ponceau S staining of the membranes and immunoblotting with antistriatin antibody, a cytosolic protein (a kind gift of David Pallas, Emory University) or anti-β-actin from Sigma Chemical Co.] were separated on a 12.5% SDS gel, transferred to nitrocellulose (Bio-Rad), and immunoblotted with monoclonal anti β-catenin antibody (clone 15B8, Sigma Chemical Co.), and horseradish peroxidase-conjugated secondary antibody (Sigma Chemical Co.). The bands were visualized by the enhanced chemiluminescence method (Amersham), recorded onto X-ray film, and identified by comparison to the size of commercially available protein standards.

SW480 cells (0.5 × 10⁵ cells/well) were seeded onto glass coverslips placed in 24-well dishes and incubated overnight. Then the media was removed, and fresh media containing 2.5 or 5 μm of sphingosine (as BSA complex; see above) or BSA as vehicle control were added for 24 or 48 h. The cells were then washed with PBS, and fixed in methanol for 4 min. After repeated washings with PBS, the coverslips were incubated for 20 min at room temperature with the monoclonal antibody for nuclear β-catenin (clone 8E4; Alexis Biochemicals, San Diego, CA), or with unspecific mouse IgG as a background control, and then incubated with a FITC-conjugated secondary antibody and 4′,6-diamidino-2-phenylindole dihydrochloride (to counter stain the nucleus) for 20 min, rinsed, and covered with Mowiol mounting solution and a coverslip. Images were acquired as described above at ×40 and processed with Adobe Photoshop software.

The colon carcinoma cell lines SW480 and T84 were from the American Type Culture Collection (Rockville, MD) and were, respectively, grown in Leibovitz medium and DMEM (Life Technologies, Inc.), in all cases with 10% FCS (Hyclone, Logan, UT, or Atlanta Biological, Atlanta, GA), and 61 mg/liter penicillin G and 100 mg/liter streptomycin (Sigma Chemical Co.). Cells were seeded into 96-well dishes (8000/well) and incubated overnight. The medium was changed, and the cells were incubated with increasing concentrations of sphingosine (as a 1:1 molar complex with fatty acid-free BSA; Ref. 33), C₂-ceramide (in ethanol; Matreya, Pleasant Gap, PA), or natural ceramides from brain (in ethanol:dodecane, 98:2, by volume; Matreya), or equal concentrations of BSA, ethanol, or ethanol:dodecane as vehicle controls. The number of viable cells was evaluated by the conversion of MTT to the formazan by viable cells (MTT assay; Ref. 34). The medium was removed after 24 h, and cells were incubated with 100 μl of MTT [Sigma Chemical Co.; 1 mg/ml in sterile PBS (pH 7.4)] at 37°C. After 2 h, 100 μl of lysis buffer [20% SDS in 50% N,N-dimethyl formamide in water (pH 4.7)] was added, and, after an overnight incubation at 37°C, the absorbance was measured with a microplate reader at a wavelength of 570 nm. Cell viability was expressed as a percentage of the appropriate control.

Min mice were fed four types of diets: (a) a control (AIN 76A) diet which contains <0.005% sphingolipid by weight (19); (b) the AIN 76A diet supplemented with ceramide at 0.1% by weight; (c) the AIN 76A diet supplemented with a mixture of sphingolipids isolated from milk, and in approximately the same proportion as in dairy products (35, 36), totaling 0.1% of the diet by weight; or, (d) a combination of this mixture of complex sphingolipids (60%) and ceramide (40%), maintaining the total sphingolipids at 0.1% of the diet by weight. These supplements were formulated to ensure that sphingolipid metabolites would reach all regions of the intestine, because Min mice develop tumors throughout the small and large intestine. Ceramide is digested relatively rapidly in the upper small intestine, whereas more complex sphingolipids such as sphingomyelin and glucosylceramide are hydrolyzed throughout the intestine (37, 38, 39).

Min mice fed the AIN 76A diet alone had 55.8 ± 4.6 tumors/mouse and supplementation with ceramide alone or the mixture of complex sphingolipids reduced the number of tumors by 40% (P < 0.05 and P < 0.01, respectively). The mixture of ceramide plus complex sphingolipids reduced the total number of tumors by 50% (P < 0.001 versus the control diet; Fig. 1,A). All three groups fed sphingolipids had significantly fewer tumors than the control group when specific regions of the intestine were analyzed (Fig. 1 B). The mixture of ceramide and the complex sphingolipids had the lowest number of tumors developed (P < 0.001, P < 0.001, P < 0.01, and P < 0.01 compared with the control; and P < 0.001 versus the ceramide or the milk mix alone in proximal, mid- and distal small intestine and colon, respectively).

If this reduction in tumor number results from “correction” of an abnormality in the APC/β-catenin pathway (i.e., a reduction in the cytosolic accumulation of β-catenin and its associated mitogenicity, see “Discussion”), this should be reflected in the intracellular localization of β-catenin. Therefore, intestinal sections of the genetic background mice C57BL/6J were stained with an anti-β-catenin antibody to establish the normal localization of β-catenin in these mice. In C57BL/6J mice, β-catenin is strictly localized throughout the small intestine in the lateral membranes (Fig. 2,a) along the full length of the villi. The cells lining the colonic lumen contain substantial amounts of β-catenin, which is also strictly associated with the membranes; very little expression of β-catenin was seen in lower parts of the colonic crypts (Fig. 2,b). Serial sections of the small intestine (data not shown) and the colon were coincubated with blocking peptide to test for antibody specificity (Fig. 2,c) or with an unspecific antibody to determine background fluorescence (Fig. 2 d) and showed essentially no fluorescence.

Intestinal epithelial cells from the small intestine of Min mice fed the sphingolipid-free control diet displayed a diffuse distribution of β-catenin throughout the cell, with much of the immunofluorescence in the cytoplasm (asterisks in Fig. 3,a). To a lesser degree, there is also an accumulation of cytosolic β-catenin in colonic epithelial cells lining the lumen, but there is also a substantial β-catenin staining in the membranes (Fig. 3,c). Min mice that displayed a low number of tumors after being fed a mixture of sphingolipids plus the ceramide had relatively less fluorescence in the cytoplasm of the epithelial cells of the small intestine (Fig. 3,b) and colon (Fig. 3,d). Most of the β-catenin was located at the lateral junctional complexes (arrows at the cell periphery in Fig. 3, b, c, and d), which is the normal distribution (compare with Fig. 2, a and b). Thus, sphingolipid feeding “normalized” the localization of β-catenin throughout the intestine, with the most pronounced changes in cells of the small intestine, where most of the tumors develop in Min mice (CF1). However, Min mice fed sphingolipids that had developed tumor numbers comparable with the controls did not show the drastic reduction of cytosolic β-catenin (data not shown), substantiating the critical role of β-catenin localization in tumor formation.

Although previous studies with rats and CF1 mice have shown that complex sphingolipids are digested to sphingosine (plus some ceramide) throughout the intestine (37, 38, 39) and have concluded that the sphingoid bases are the major inhibitors of colon carcinogenesis (21), this digestive capability has not been established for Min mice. Therefore, a mixture of these sphingolipids (ceramide, sphingomyelin, glucosylceramide, lactosylceramide, and ganglioside G_D3) was incubated with different intestinal segments, and the appearance of the hydrolysis products was monitored by electrospray tandem mass spectrometry. Complex sphingolipids were hydrolyzed to sphingosine and ceramide by segments from all regions of the small intestine and colon (Fig. 4); therefore, Min mice also have the ability to hydrolyze complex sphingolipids to these bioactive lipids.

Earlier studies have shown small increases in sphingolipids (mostly as free sphingoid bases) in lymph (37, 38, 39) or liver and kidney (37) after ingestion of a single dose of radiolabeled complex sphingolipids. To determine whether the sphingolipid supplements in these diets altered the circulatory sphingoid bases, the amounts in blood were determined immediately after the deaths of the mice. There was considerable sample-to-sample variability in some groups, but the nmol of free sphingoid bases/ml of whole blood did not differ significantly versus the control or among the groups: 0.57 ± 0.13, 0.51 ± 0.48, 0.49 ± 0.10, and 0.58 ± 0.09 (mean ± SE) for mice fed the control, ceramide, milk mix, and milk mix + ceramide diets, respectively (P = 0.65, by ANOVA).

To determine whether the product(s) of sphingolipid digestion can directly affect colon tumor cells (especially human colon cancer cells) with defects in the APC/β-catenin pathway, in vitro studies were conducted with SW480 and T84 cells. These colon cancer cell lines display a stable accumulation of β-catenin in the cytosol (Ref. 40; Fig. 5,A) caused by APC mutations. Treatment with 2.5 μm sphingosine for 4 h (SW480 cells) or 24 h (T84 cells) caused a substantial reduction in the amounts of cytosolic β-catenin (Fig. 5,A), suggesting that this effect, although time-dependent, is not restricted to sphingosine. In addition, SW480 cells were treated for 6 h with 2.5 μm of sphingosine, C₂-ceramide (in ethanol), or a natural long-chain ceramide (in ethanol:dodecane; 98:2, v/v), and the effect on cytosolic β-catenin was assessed. Treatment with sphingosine or bovine brain ceramide reduced cytosolic β-catenin, whereas C₂-ceramide (Fig. 5 B) and solvent vehicles (data not shown) had no effect. Under these conditions, the added sphingolipids were not toxic for either cell line, however, with longer incubation and/or higher concentrations, there was considerable toxicity (see below).

To determine whether this reduction in cytosolic β-catenin might also lead to reduced β-catenin in the nucleus, one of these experimental models (treatment of SW480 cells with sphingosine) was also examined using an anti-β-catenin antibody that recognizes nuclear (dephospho-) β-catenin (Fig. 6, a and b) as well as β-catenin outside the nucleus in a subpopulation of cells (Fig. 6, d and e). There was no obvious change in nuclear β-catenin after 6 h of incubation with sphingosine (data not shown), but a significant reduction in nuclear β-catenin was apparent 24 h after adding 5 μm sphingosine (Fig. 6, g and h) and after 48 h of incubation with 2.5 μm sphingosine (Fig. 6, k and l). The behavior of the cells in each Petri dish was not uniform, and at both concentrations and time points, a subpopulation of the cells did not respond with a visible reduction of nuclear β-catenin (data not shown).

By MTT assay, sphingosine was cytostatic and cytotoxic for both of these human colon cancer cell lines, with a 50% decrease in cell number after 24 h with 8 and 20 μm sphingosine for T84 and SW480 cells, respectively (Fig. 7). C₂-ceramide (N-acetylsphingosine) was less somewhat less potent, causing a 50% decrease in cell number at 50 and 30 μm, respectively (Fig. 7), and a natural (bovine brain) ceramide containing long-chain fatty acids was also less toxic than sphingosine (Fig. 7). Therefore, although both of the digestion products of complex sphingolipids (ceramides and sphingosine) are toxic for these cell lines, sphingosine has the greater potency, as we have seen also with HT29 cells and other colonic cell lines.⁵

Sphingolipids are provocative components of the diet because their backbone moieties (sphingosine and ceramide) are highly bioactive compounds that are thought to serve as intracellular regulators of growth and death (41, 42, 43), yet intestinal cells are regularly exposed to these compounds as a consequence of sphingolipid digestion and absorption (37, 38, 39). This raises the possibility that dietary sphingolipids might be injurious to intestinal epithelial cells; however, in our investigations (19, 20, 21, 22), as well as rodent feeding studies spanning two generations (44), no deleterious effects have been observed as yet. Instead, sphingolipid consumption has been found to inhibit early and late stages of colon carcinogenesis induced by DMH (19, 20, 21, 22) and, as shown by this study, to reduce tumorigenesis in Min mice. These findings are more consistent with the hypothesis that sphingolipids “bypass,” or normalize, a signaling defect in transformed cells rather than having nonselective effects on all of the cells of the intestine.

More studies are needed to elucidate the mechanism(s) for the tumor suppression in both Min mice and DMH-treated CF1 mice; however, they may share a common mechanism because APC and β-catenin mutations are found in this and closely related in vivo experimental models for colon cancer (45, 46). All intestinal cells of Min mice carry a nonsense mutation in one APC allele, but tumors develop only after the loss of the WT allele, suggesting that this heterozygous defect constitutes a predisposition for neoplastic lesions. Nonetheless, the truncated APC protein can associate with WT APC, and this appears to alter the behavior of the complex in a dominant-negative fashion in the cytosol (47). Furthermore, competitive binding of nuclear β-catenin by WT and mutated APC prevents the binding to and activation of transcription factors (48). This would account for β-catenin accumulation in the cytosol and nucleus even in the presence of the WT allele, as shown in Fig. 3 a. Although this has been seen by a number of other laboratories (40), some have reported that β-catenin is primarily membrane-associated in intestinal cells of Min mice (49). This variability is unexplained, as are the large variations in tumor numbers/mouse, which range between 20 and 80 for different experiments conducted under essentially identical conditions (as well as one can ascertain).

There are a number of mechanism(s) whereby sphingosine might induce the disappearance of β-catenin from the cytosol and the nucleus. Sphingosine is a potent inhibitor of protein kinase C, one isozyme of which (PKC_βII) is overexpressed in colon cancer (50, 51) and inhibits glycogen synthase kinase-3β (52), which phosphorylates β-catenin and triggers its ubiquitination and proteasomal degradation (53, 54). β-catenin is also degraded by activated caspase 3 in early apoptosis (55), and sphingosine and ceramide both induce this and other caspases (11, 12). Furthermore, induction of apoptosis in colon cancer cells by sulindac (a potent COX-2 inhibitor) involves the generation of ceramide by arachidonic acid-activated sphingomyelinases (56), suggesting that multiple mechanisms of cell regulation by sphingolipids are existent in this model. The bioactive sphingolipids affect a wide spectrum of intracellular signaling pathways, including phosphoprotein phosphatases (PP1 and PP2A; Ref. 57), a 14-3-3 kinase (58), cAMP (59), concentrations of intracellular calcium (60), and the regulation of cyclooxygenase(s) (61), among others, that might play a role in these complex events. By whatever mechanism, the removal of β-catenin from the cytosol and the nucleus may inhibit the association with Tcf/Lef transcription factors, and affect the transcription of proteins that are regulated by β-catenin and closely correlated with colon cancer (such as cyclin D1). These effects warrant additional investigation.

Defects in the APC/β-catenin pathway are common in human colon cancer and are thought to predispose cells to additional mutations that proceed to neoplasia. Intervention with compounds that suppress the abnormal phenotype of the cells before they acquire the later mutations have considerable promise as chemopreventive agents. The finding that sphingosine induces the disappearance of cytosolic and nuclear β-catenin in human cancer cell lines as well as being cytotoxic for these cells raises the possibility that consumption of sphingolipids might reduce the risk of human colon cancer. Diets rich in dairy products and/or soy, for example (36), could provide the amounts of sphingolipids that suppressed tumorigenesis in CF1 mice (19, 20, 21) and Min mice. Unfortunately, currently it is not possible to conduct epidemiological studies to test this association further, because there have been no systematic analyses of the amounts and types of sphingolipids in food, nor is much known about the potency of the sphingoid base backbones of other foods.⁶ From another perspective, previous epidemiological studies have not considered the sphingolipid content of the diets in evaluating associations between specific nutrients and colon cancer and this might account for some of the disagreement among these studies (62, 63, 64, 65). This should be borne in mind for future chemoprevention (and possibly chemotherapeutic) interventions, including studies of how diet might be used to delay the progression of colonic adenomas to adenocarcinomas in familial adenomatous polyposis, which usually necessitates removal of the colon.