Glycosphingolipids Are Essential for Intestinal Endocytic Function*

Background: The intestine contains high concentrations of glycosphingolipids, but their function remained unclear. Results: In newborn mice lacking glycosphingolipids, intestinal epithelia were indistinguishable from control littermates. However, a few days after birth, severe defects in epithelial differentiation occurred. Conclusion: Glycosphingolipid expression in the intestinal epithelium is quintessential for maintenance of resorptive function. Significance: Glycosphingolipids are essential for enterocyte function but not for brush border formation. Glycosphingolipids (GSLs) constitute major components of enterocytes and were hypothesized to be potentially important for intestinal epithelial polarization. The enzyme UDP-glucose ceramide glucosyltransferase (Ugcg) catalyzes the initial step of GSL biosynthesis. Newborn and adult mice with enterocyte-specific genetic deletion of the gene Ugcg were generated. In newborn mutants lacking GSLs at day P0, intestinal epithelia were indistinguishable from those in control littermates displaying an intact polarization with regular brush border. However, those mice were not consistently able to absorb nutritional lipids from milk. Between postnatal days 5 and 7, severe defects in intestinal epithelial differentiation occurred accompanied by impaired intestinal uptake of nutrients. Villi of mutant mice became stunted, and enterocytes lacked brush border. The defects observed in mutant mice caused diarrhea, malabsorption, and early death. In this study, we show that GSLs are essential for enterocyte resorptive function but are primarily not for polarization; GSLs are required for intracellular vesicular transport in resorption-active intestine.

The intestine is the principal organ for digestion and absorption of nutrients. In the small intestine, enterocytes (columnar epithelial cells) cover the surfaces of villi, which contain thousands of densely packed apical microvilli that further amplify the surface area for absorption of nutrients.
Glycosphingolipids (GSLs) 3 are major constituents of enterocytes (1). Because of their high enterocytic expression, GSLs might be involved in several cellular processes in polarized intestinal epithelia. It could be shown that GSLs are integral constituents of lipid microdomains that seem to play important roles in membrane organization (2) and in endocytosis (3) of nutrients. In addition, GSLs might influence the biosynthesis (4) and intracellular transport of proteins (5). GSLs have been described to influence membrane receptors such as the insulin and EGF receptor (6 -9).
The structural relevance of GSL synthesis in the intestine could be demonstrated in Caenorhabditis elegans mutants with significantly decreased glucosylceramide-based GSLs. Worms devoid of GSLs showed a pronounced growth arrest. The defect could be revoked by reexpression of glucosylceramide synthase in the digestive tract (10). This study highlighted the importance of intestinal GSLs, whereas GSL expression, different from the mouse, was dispensable for other tissues, including neuronal cells (10). In addition, a recent report suggested that GSLs were pivotal for epithelial polarization in C. elegans. With a marked reduction of GSL synthesis, worms developed multiple ectopic lumens in their intestine and showed degraded microvilli (11).
To dissect the role of GSLs for intestinal function in mammals, we used a novel mouse model in which GSL biosynthesis was genetically deleted in an enterocyte-specific manner by elimination of the gene Ugcg (glucosylceramide synthase), which encodes the key enzyme of the GSL biosynthesis pathway for all GSLs produced by enterocytes. To achieve specific deletion of GSLs in enterocytes, mice with the loxP-flanked Ugcg gene (12) were mated with mice expressing Cre recombinase under control of the villin promoter (13). As villin expression starts early with the developing intestine at embryonic day E10 and is then ubiquitously distributed throughout the small intestine at embryonal days E14 to E15 (14), our model was also especially suited to investigate the consequences of Ugcg deletion in newborn mice. A second model was employed in which GSLs were deleted specifically in enterocytes during adulthood, again under control of the villin promoter (15).
Complete genetic deletion of GSLs in the intestinal epithelium was achieved in both models. Although GSLs were absent in enterocytes of newborn mutant mice at postnatal day P0, the epithelial brush border had been formed regularly. However, the ability to absorb lipids from milk was already significantly reduced at P0. A few days after birth, GSL deletion in enterocytes caused severe structural defects of the epithelium accompanied by loss of the brush border. The lack of nutrient absorption led to the early death of mice.
This study has now demonstrated that glycosphingolipids are initially not required for intestinal epithelial polarization, but they are quintessential for proper intestinal function, namely endocytic uptake of nutrients, in particular lipids, when intestine starts its activity after birth.

EXPERIMENTAL PROCEDURES
Mice-Mice with loxP-flanked Ugcg gene (Ugcg f/f) were generated and genotyped by PCR and Southern blot analysis as described previously (12). Cell-specific Ugcg-deficient newborn mice were generated by crossing Ugcg f/f mice with animals expressing cre recombinase constitutively (VilCre, The Jackson Laboratory, Bar Harbor, ME) (13). To investigate the consequences of the Ugcg deletion during adulthood, an inducible model was developed by crossing Ugcg f/f mice with tamoxifeninducible VilCreERT2 animals (15). Mice were genotyped as described previously (13,15). Resulting heterozygous mice (Ugcg f/ϩ/VilCre or Ugcg f/ϩ/VilCreERT2) were bred again with Ugcg f/f mice, and 25% of the offspring, reflecting normal Mendelian inheritance, carried the homozygous genetic deletion of Ugcg (Ugcg f/f/VilCre and Ugcg f/f/VilCreERT2). Corresponding heterozygous (Ugcg f/ϩ/VilCre and Ugcg f/ϩ/VilCre-ERT2) as well as "wild type" litters (Ugcg f/ϩ/and Ugcg f/f) showed no obvious phenotype and were used as controls. Ugcg gene deletion in adult Ugcg f/f/VilCreERT2 mice was initiated by intraperitoneal application of 1 mg of tamoxifen in 100 l of sunflower seed oil (both, Sigma) for 3 consecutive days. Respective controls were treated in parallel with tamoxifen. All animal experiments were approved by federal law.
LacZ Staining to Evaluate Cre Expression-LacZ-Rosa26 reporter mice (kindly provided by G. Schütz, DKFZ-Heidelberg) bearing a ␤-galactosidase gene were crossed with mice expressing the VilCre transgene. In the resulting offspring bearing both LacZ and cre, ␤-galactosidase activity was measured as described previously (15).
Sphingolipid and Phospholipid Analysis-Freshly dissected tissue ϳ2 cm in length was frozen in liquid nitrogen and lyophilized. Tissues were powdered, and dry weight was determined. Lyophilized samples were extracted with 2 ml of CHCl 3 / CH 3 OH/H 2 O, 10:10:1, v/v, under sonication in a water bath at 50°C for 15 min. Supernatants were collected after centrifugation at 4000 rpm for 10 min. The pellets were again extracted with 2 ml of CHCl 3 /CH 3 OH/H 2 O, 10:10:1 (v/v), and thereafter with CHCl 3 /CH 3 OH/H 2 O, 30:60:8 (v/v), as described above. All supernatants were combined and dried under a stream of nitrogen. For the determination of phospholipids, aliquots from the crude intestinal extracts corresponding to 0.05 mg of dry tissue were loaded on thin layer chromatography (TLC) plates (Merck). TLC plates were developed in running solvent CHCl 3 /CH 3  For quantification of sphingomyelin, aliquots from the saponified samples corresponding to 0.05 mg of dry tissue were spotted on TLC plates. TLC plates were developed and stained as described above for phospholipids. Amounts corresponding to 0.25 mg of dry tissue were spotted on TLC plates for determination of ceramides. Running solvent was CHCl 3 /CH 3 OH/ HAc, 190:9:1, v/v. Visualization of ceramides with copper reagent was performed as described above. Glycosphingolipids were separated into neutral and acidic (sialic acid-containing) components by ion exchange chromatography. Glass pipettes were filled with 200 l of DEAE-Sephadex A25 (GE Healthcare), and columns were preconditioned with 2 ml of methanol. The samples were solved in 2 ml of methanol and applied to the columns. Reagent tubes were rinsed two times with 1 ml of methanol, and neutral GSLs were completely eluted from the columns with additional 2 ml of methanol. Acidic GSLs were then eluted with 2 ml of 0.5 M KAc in CH 3 OH. Acidic GSLs were dried and solved in 5 ml of H 2 O and desalted via preconditioned RP18 columns as described above. Neutral GSLs were thoroughly dried and kept for several hours in a vacuum desiccator. GSLs were solved in 200 l of dichloroethane (DCE) and acetylated by the addition of 50 l of acetic anhydride and 50 l of 0.1% dimethylaminopyridine in DCE at 37°C for 1 h. The solvents were evaporated under a N 2 stream by addition of 200 l of toluene, respectively, three times. Glass pipettes were filled with 400 l of Florisil (Merck) and washed with 4 ml of DCE/n-hexane, 4:1, v/v. The acetylated neutral GSLs were solved in the same solvent mixture and applied to the columns followed by washing steps with DCE/n-hexane, 4:1, and DCE, 4 ml each. Neutral GSLs were eluted from Florisil columns with 4 ml of DCE/acetone, 1:1 (v/v), and dried. To remove acetyl groups from sugar moieties, neutral GSLs were solved in 0.1 M KOH in methanol and incubated at 37°C for 2 h. After neutralization with acetic acid, the purified neutral GSLs were dried Quantification of Fatty Acids and Cholesterol in Stool-1 cm of the distal rectum of sacrificed mice (controls, n ϭ 8, and Ugcg f/f/VilCre, n ϭ 6) was resected and the feces content carefully removed with a small forceps. Isolated stools were freeze-dried, and the dry weight was determined and extracted essentially as described previously (16). Lipids were quantified using a Shimadzu CS-9310PC TLC scanner (Shimadzu Europe, Duisburg, Germany) at a wavelength of 580 nm.
Sphingolipids were detected in selected reaction monitoring mode with their respective transitions. The decision basis for the incorporation of individual sphingolipids into the final selected reaction monitoring list was based on nanospray direct infusion in precursor ion scan m/z ϩ264 mode identifying ceramides and hexosylceramides containing either a d18:1 or a t18:0 sphingoid base. For quantification, peak areas of individual lipid species were divided by the respective average internal standard area, multiplied by the amount of this internal standard, and finally divided by the tissue dry weight used. With the internal ceramide standards Cer(d18:1;31:0) and Cer(t18:0;31: 0), an average response factor of 12.2 Ϯ 1.8 for the sensitivity of sphingosine species over phytosphingosine species was determined and taken into account to quantify the phytosphingosine-containing ceramides and hexosylceramides.
Isolation of Milk Sugars from Feces and TLC-Feces were extracted essentially as described previously (16). An aliquot of the crude extract was directly loaded on a TLC plate and developed in running solvent pyridine/ethyl acetate/glacial acetic acid/water 5:5:1:3 by volume. Visualization with orcinol reagent was performed as described above.
Measurement of Intestinal Sialidase Activity-Homogenates of total small intestines of mutant and control (n ϭ 4 each) were   mixed with siallyllactose (Sigma,) and the digest was performed at 37°C either for 30 min or in a time-dependent manner essentially as described previously (17). The cleavage product lactose was quantified by densitometry. Fatty Acid/Glucose Uptake-Animals 3-4 days and 6 -7 days after birth (controls, n ϭ 12, and Ugcg f/f/VilCre, n ϭ 9) were anesthetized with isoflurane. The peritoneum was opened, and a 1-cm part of the intestinal jejunum of the same area was ligated. 100 l of a solution containing 10 M fluorescent-labeled stearic acid (NBD-stearic acid, Avanti Polar Lipids, Alabaster, AL) and 20 M fatty acid-free BSA (Sigma) in PBS was injected, and uptake was performed for 10 min in the dark. For investigation of glucose uptake studies, 60 M 2-NBD-glucose (Invitrogen) in PBS was used. The ligated part of the jejunum was resected, washed with pre-warmed PBS, and immediately frozen in liquid nitrogen. Cryosections of 5 M thickness were air-dried for 30 min in the dark, embedded in Fluoromount G (Southern Biotech, Birmingham, AL) for detection of NBD-stearic acid or Vitro-Clud (Langenbrink, Emmendingen, Germany) for NBD-glucose staining, and scanned with a confocal laser microscope (TCS-SL, Leica, Wetzlar, Germany).
Liposome Preparation and Uptake-To investigate whether GSL depletion influenced the absorption of fat in macroscale amounts, liposomes were prepared for this uptake approach. Each lipid, i.e. cholesterol, phosphatidylcholine, and triolein, was solved at a concentration of 10 mg/ml in CHCl 3 /CH 3 OH, 2:1 (v/v). Fluorescent NBD-stearic acid (Avanti Polar Lipids, Alabaster, AL) was solved in CHCl 3 at a concentration of 1 mg/ml. For the preparation of liposomes, 160 l of triolein, and 20 l of cholesterol, phosphatidylcholine, and NBD stearic acid each were mixed and dried under a stream of nitrogen. The lipid mixture was subsequently dried in a vacuum desiccator for 30 min. After addition of 500 l of PBS, liposomes were formed by incubation of the mixture in an ultrasound water bath at 50°C for 15 min. For uptake studies of liposomes, mice were anesthetized, and a 2-cm part of the jejunum ϳ4 cm caudally from the duodenum was ligated with clamps. 100 l of the liposome solution was injected (controls, n ϭ 8; Ugcg f/f/Vil-CreERT2, n ϭ 3). The liposome containing jejunum was excised after 15 min, rinsed with 5 ml of PBS, and immediately frozen on dry ice in O.C.T. Tissue-Tek embedding medium (Sukura Finetek, Alphen aan den Rijn, The Netherlands). Sections of 5 m were mounted with fluorescent mounting medium (DAKO, Hamburg, Germany) containing 1:2000 diluted DRAQ5 (Alexis, Shepshed, UK) for nuclear staining and scanned with a confocal laser microscope (TCS-SL, Leica, Wetzlar, Germany). Light Microscopy and Immunohistochemistry-Light microscopy at different time points (controls, n Ͼ40, and Ugcg f/f/ VilCre, n Ͼ30) and immunohistochemistry/immunofluorescence (controls, n ϭ 4, and Ugcg f/f/VilCre, n ϭ 4) were performed as described previously (12). Paraffin-embedded buffered paraformaldehyde (4%, pH 7.3)-fixed sections were stained by hematoxylin and eosin (H&E, Chroma, Köngen, Germany). Activity of alkaline phosphatase was determined by overlay of sections with alkaline phosphatase substrate solution. For immunohistochemistry of rat anti-mouse Ki67, 1:200 (DAKO, Hamburg, Germany), an alkaline phosphatase-antialkaline-phosphatase reaction was used (12). Goat anti-villin antibodies (Santa Cruz Biotechnology, Santa Cruz, CA) were diluted 1:20. A 1:300 diluted biotinylated anti-goat antibody was used as secondary antibody (Dianova, Hamburg, Germany), followed by a streptavidin-HRP complex (Vector SA-5004, Vector Laboratories, Burlington, CA) at a dilution of 1:200. All antibodies and dilutions used in immunofluorescence are listed in Table 2.
In Vivo DNA Labeling with EDU-Mice were injected intraperitoneally with 200 g of EDU (Invitrogen). After 24 h, mice were sacrificed, and intestinal parts were embedded in paraffin. Sections of 5 m thickness were taken, and EDU-labeled cells were stained by click reaction as described by Salic and Mitchison (18).
Immune Electron Microscopy-Intestinal sections were prepared for cryoimmune-EM according to Tokuyasu et al. (20) to ensure that GSLs neither were extracted nor delocalized within the tissue. Sections of ϳ2-3 mm in length were immersion-fixed with 4% paraformaldehyde in PBS for 2 h. Subsequently, sections were allowed to sediment overnight in 2.3 M sucrose and have thereafter been frozen until use. Frozen ultrathin sections were obtained using an ultracut (UC6, Leica, Nussloch, Germany) and were placed on carbon-coated EM grids. The sections were blocked and labeled with antibodies as described previously (20). Primary antibodies and gold-labeled secondary antibodies or protein A gold are listed in Table 3. All primary antibodies were incubated at 4°C overnight. Gold-labeled secondary antibodies or protein A were incubated at room temperature for 2 h. Images of labeled sections were taken using a transmission EM (Leo 910, Zeiss, Oberkochen, Germany).
Nile Red Staining for Lipids-For the staining of cryosections at P0, newborn mice were sacrificed after initial suckling ϳ2-4 h after birth, and one part of the intestinal jejunum with visible FIGURE 2. Sphingolipid analysis. A, GSLs as synthesis products of Ugcg were depleted in mutant tissue already in newborn mice at day P0. B, as revealed by mass spectrometry, the total GlcCer depletion rate in newborn Ugcg f/f/VilCre mice at P0 was Ͼ90% (controls (Contr.) and mutant, n ϭ 3, respectively). A minor GlcCer fraction with sphingosine and nonhydroxylated fatty acids in its ceramide anchor was not affected by the Ugcg-gene deletion and might likely be originated from cells in which the cre recombinase was not active. C, sphingomyelin in mutant intestine had a similar concentration as in control tissue, and ceramide content increased slightly in Ugcg f/f/VilCre intestine as compared with controls. D, phospholipid content was not altered in newborn Ugcg f/f/VilCre mice at P0 (n ϭ 4, respectively). E, intestinal phospholipid concentrations significantly decreased in mutant tissue at P5 to P7 (PE, phosphatidylethanolamine; PG, phosphatidylglycerol; PC, phosphatidylcholine), controls and Ugcg f/f/VilCre, n ϭ 6 each. nutrient content was shortly perfused with 2 ml of ice-cold PBS and immediately frozen in Tissue-Tek embedding medium. Similarly, also one part of the jejunum was prepared from newborns at P7. 0.05% Nile red was dissolved in acetone. 10 l of the stock solution was diluted 1:200 in glycerol/H 2 O, 3:1 (v/v). 5-m cryosections were covered with Nile red solution and scanned with a confocal laser microscope (TCS-SL, Leica, Wetzlar, Germany).
Western Blotting-Proteins were isolated from freshly prepared small intestine from the jejunum essentially as described elsewhere (21). Antibodies used are listed in Table 4.
mRNA Isolation and Quantitative RT-PCR-One ϳ1-cm piece of the intestinal jejunum was prepared from sacrificed animals. Adhering tissue was removed. Total mRNA was extracted as described previously (22). Quantitative real time RT-PCR was done using the LC-fast DNA Master SYBR Green  I kit PCR for the LightCycler (Roche Applied Science) as described previously (12), and mouse 18 S ribosomal RNA was used as reference (controls, n ϭ 8, and Ugcg f/f/VilCre, n ϭ 8).
Lipase Assay-Feces of newborn control (n ϭ 7) and Ugcg f/f/VilCre (n ϭ 6) mice were collected from rectums at day P6 and immediately frozen and stored at Ϫ80°C until analysis. Feces were weighed and diluted 1:20 (w/v) with distilled water and dispersed. The suspension was centrifuged at 12,000 rpm and at 4°C for 10 min. 5 l of the respective supernatants were used for lipase assay. The assay was performed according to Ma et al. (23).

Ugcg f/f/VilCre Mice Lack GSLs in the Intestine of Newborn
Mice-To investigate the role of GSLs for intestinal function, mice were generated in which the Ugcg gene, encoding the key enzyme Ugcg involved in the initial step of the glucosylceramide based GSL synthesis pathway, was deleted in a cellspecific manner in enterocytes by using a villin-cre recombinase transgene (Fig. 1A). Deletion of Ugcg can be expected to result in absence of GSLs in the epithelium of the intestine in newborn and adult mice (Fig. 1B). A product of 383 bp indicated the floxed Ugcg allele and a fragment of 216 bp was indicative for the villin-cre transgene (Fig. 1C). As shown by Southern blot analysis, Ugcg gene deletion was restricted to the intestine (Fig. 1D). The specificity of the Ugcg gene deletion for the complete intestinal tract but not for the stomach was indicated by a LacZ staining (Fig. 1E). Consequently, all investigated intestinal parts of mutant mice showed GSL depletion (Fig. 1F). In agreement with the Ugcg gene deletion, depletion of glucosylceramide (GlcCer) and its further synthesis products, e.g. ganglioside GM1 and GD1a (24), was seen in newborn mice at P0 and P7 ( Fig. 2A). The TLC data could be confirmed by LC-MS/MS. In our analysis from intestinal tissue at day P0, we could demonstrate that GlcCers with phytosphingosine-and ␣-hydroxylated fatty acid in their lipid anchor were by far the most prominent occurring species in the intestine (Fig. 2B). Those species of GlcCers were 40-fold reduced in mutant intestine of newborn mice at P0 (Fig. 2B). However, one minor GlcCer species with sphingosine and nonhydroxylated fatty acids, representing less than 5% of the total GlcCers, appeared unaltered in intestinal GSL extracts (Fig. 2B). These GlcCers may have originated from cells not affected by the Ugcg deletion such as muscle, endothelial, stroma, and goblet cells. Also, the possibility that the remnant was not GlcCer but GalCer cannot be excluded. The total sphingomyelin content in mutant intestine appeared almost unaltered (Fig. 2C), and ceramide increased ϳ20% in mutant tissue at P7 (Fig.  2C). The phospholipid concentration in intestinal extracts was not changed in newborn mice at P0 (Fig. 2D), but they significantly decreased in mutant intestines 5-7 days after birth (Fig. 2E). The intestinal cholesterol concentrations were slightly but not significantly reduced in mutant jejunum at P7 (i.e. Ugcg control, 22.5 Ϯ 1.1 nmol/mg dry weight; Ugcg f/f/VilCre, 20.0 Ϯ 1.9 nmol/mg dry weight).
Newborn Ugcg f/f/VilCre Mice at P0 Displayed Inconspicuous Morphology and Behavior-Although newborn mice lacked enterocytic GSLs at P0 (Fig. 2A), they and enterocytes were indistinguishable from their littermates in size and behavior. Expression of villin, required for correct alignment of the microvillar actin core bundle to anchor the brush border, was normal in mutants at P0, implying a regular polarization of the epithelium (Fig. 3, A and E, anti-villin immunofluorescence). The ultrastructure of intestinal epithelium was similar in mutants and controls (Fig. 3, B and F, electron micrographs). An increased proliferative property of the epithelium due to the absence of GSLs was not observed at P0 (Fig. 3, C and G). GSLs could be detected equally distributed in vesicular like dots throughout the cytosol of enterocytes in control mice, as shown by anti-GM1 immunofluorescence (Fig. 3D). In agreement with TLC GSL analysis, a staining for GM1 was negative in mutant enterocytes (Fig. 3H).
Ugcg f/f/VilCre Mice Demonstrated Loss of Fat Deposits-As a consequence of the intestinal GSL deficiency, mutant mice did not gain weight after birth with increasing age (Fig. 4A,  inset), although they sucked milk normally (Fig. 4C). The usual fat deposits in the abdomen and pelvis were not seen in mutants (Fig. 4E). Mutant mice regularly defecated attesting a not significantly altered gut motility. However, stools of mutant mice appeared soft and lipid-rich. Offsprings, depleted of intestinal GSLs, succumbed between P5 and P8.
Histology Showed Age-dependent Severe Structural Alterations in Mutant Intestine-In contrast to newborn mutant mice, which showed an inconspicuous morphology at P0, postnatally, between days 5 and 7, the intestinal epithelium demonstrated massively altered structure throughout both the small and large intestines (Fig. 5A). The morphology of mucosal villi in all parts of the small intestine was severely distorted in mutants 1 week after birth; they often appeared to be attached or "glued" to one another. In the colon, severe vacuolization was observed in particular in the uppermost layers of the enterocytes.
Cell proliferation in the intestinal epithelium was significantly increased in mutant tissue, with proliferating cells extending to the intestinal lumen. In control litters, proliferating cells were only located in the base of the intestinal crypts (Fig. 5B, Ki67). Alkaline phosphatase, usually stringently and continuously located at the apical surface of the intestinal brush border, was detected in a patchy pattern in mutant mice implying a disorganization of microvillar structure (Fig. 5B). In comparison with newborn Ugcg f/f/VilCre mice at P0, in which the microvillus protein villin was regularly distributed, anti-villin failed to stain the apical brush border region in GSL-deficient enterocytes at P7 (Fig. 5B, villin).  Ugcg f/f/VilCre Mice Displayed Severe Disturbance in Uptake of Nutrients-We thought that absence of GSLs in enterocytes would be associated with a decrease in the uptake of nutrients. Indeed, the lipid marker Nile red indicated a drastic reduction of intracellular lipid depots in Ugcg f/f/VilCre mutant mice at P0 (Fig. 6B) and P7 (Fig. 6E), although controls were strongly stained (Fig. 6, A and D, and C and F, quantification). A reduced uptake of NBD-labeled stearic acid further demonstrated the restricted ability of the mutant epithelium to absorb lipids (Fig.  6, H and I, quantification).
Ultrastructurally, microvilli of enterocytes in control intestines form a regular array of finger-like protrusions (Fig. 6J). Different from newborn mice with GSL depletion at P0, postnatally, 1 week after birth, mutant mice displayed only rudimentary stump-like microvilli on the apical surface of enterocytes. The enterocytes of GSL-deficient mice displayed fewer fat containing vacuoles (Fig. 6K, LD) corroborating the Nile red lipid staining (Fig. 6, B and E). Caveola-like invaginations as seen in enterocytes of control mice were almost absent in newborn mutants. In addition, their lysosome-like structures appeared smaller than in controls (Fig. 6K).
To investigate whether the observed reduced lipid uptake by mutant intestines led to increased fecal lipid excretion, feces of control and mutant mice were investigated for their lipid content. Fatty acids (Fig. 7, A and B) and cholesterol (Fig. 7, A and C) significantly accumulated in feces. Secretion of bile acids (Fig. 7D) and pancreatic lipases (Fig. 7E) into the gut lumen were unaltered indicating that their production and release into the gut lumen were not affected in mutant mice. A subcutaneous fat layer was observed in newborn controls as well as in mutant mice at P0 (Fig. 7, F and H). Fat deposits were completely used up in mutants until P7 (Fig. 7I), whereas controls increased significantly their subcutaneous fat layer (Fig. 7G). This showed that the mutant mice metabolized their preterm subcutaneous fat deposits in compensation of the reduced intestinal lipid uptake. In concordance, plasma triglyceride concentrations significantly decreased in mutants (Fig. 7J).
To verify whether absorption deficits in mutant intestines were restricted to lipophilic nutrients or also occurred in the uptake of hydrophilic components, the absorption of glucose was investigated. Although Ugcg-deficient mice were able to absorb glucose (Fig. 7L), the extent was less than in wild type animals (Fig. 7, K and M, quantification). The absorption of carbohydrates by the mutant intestine was maintained to an extent that lactose after digestion into galactose and glucose has been completely taken up (Fig. 7N). However, sialylated lactose accumulated in feces of Ugcg f/f/VilCre mice (Fig. 7N) as a consequence of reduced sialidase activity in the small intestine of mutant mice (Fig. 7, O, OЈ, P, and PЈ).
Quantitative RT-PCR was performed to investigate whether altered membrane protein expression negatively influenced the absorption of nutrients in mutant mice. The expression levels of the intestinal fatty acid-binding proteins Fabp2 and CD36, the cholesterol transporters Abcg5 and Abcg8, the triacylglycerol synthesis and transfer proteins Dagt1 and Mttp, the glucose transporters Sglt and Glut5, as well as the small peptide transporter Pept1 were slightly but not significantly altered (Table 5). Our data demonstrate that the functional defects observed were not primarily caused due to alterations of membrane protein expression with one exception, aquaporin 8 (Aqp8). As an indication for reduced water absorption in the small intestine, the mRNA levels of Aqp8 decreased.
Upon induction with tamoxifen of the specific Ugcg gene deletion in the intestine, mice lacked GlcCer and its higher small intestinal synthesis product GA1 almost completely after 4 days (Fig. 8A). In heterozygous control mice (Ugcg f/ϩ/Vil-CreERT2), intestinal GSL composition did not change. A reduction of GSL synthesis of 50% was already achieved at day 1 after the start of the tamoxifen induction (Fig. 8B). Reduction of 70% of intestinal GSL content accompanied by initial structural alterations was observed 2 days after the start of the tamoxifen application (Fig. 8B). An almost complete absence of GSL synthesis and severe structural alterations in the villous epithelia were seen 3 days after beginning tamoxifen treatment (Fig. 8, B and BЈ). Our data obtained from the inducible model suggested a critical GSL reduction of ϳ50 -70% to induce relevant structural and functional problems in the intestinal tract. During the time-dependent reduction of GSLs, a drastic decrease in body weight (Fig. 8C), food, and water consumption (Fig. 8, D and E) was observed in mutant mice.
To investigate whether endocytosis of large scale lipid was inhibited by GSL depletion similar to newborn mice, fluorescence-labeled liposomes were injected into the gut lumen of control-and Ugcg f/f/VilCreERT2 mice. In this approach we show that liposomes entered the apical membrane of control mice epithelium unimpeded (Fig. 8F); in contrast, mutant Ugcg f/f/VilCreERT2 mice were not able to absorb liposomes (Fig.  8F).
Morphologically, Ugcg f/f/VilCreERT2 mice resembled the phenotype of mutant newborns at P5 to P7 (Fig. 9). However, the extent of changes was stronger as in newborns. Adult control-and mutant mice have been injected with EDU on day 2 after the tamoxifen-triggered Ugcg gene deletion. One day later, the proliferative zone was seen to be distributed almost throughout the complete epithelial layer, whereas proliferation in tamoxifen-treated controls was restricted to the crypts of the villi (Fig. 9B, EDU). Ultrastructurally, similarly as seen in newborn mutant mice 1 week after birth, Ugcg f/f/VilCreERT2 mice showed loss of brush border 3-4 days after tamoxifen induc-tion (Fig. 10, B and D), decreased phospholipid concentrations (Fig. 10G), similar sphingomyelin levels as in controls (Fig.  10H), and increased ceramide concentrations (Fig. 10H).
Very similar to the GlcCer profile of newborns, GlcCers with a phytosphingosine-containing ceramide anchor were found by LC-MS/MS to be the most prominent species in intestine (Fig.  10I). In the intestine of adult mutant mice, GlcCers were drastically reduced (Fig. 10I). As in newborns, GlcCers with (d18:1; n:0)-ceramide anchors were unaltered. The precursor ceramides increased in mutant tissue significantly (Fig. 10, H, TLC; and J, LC-MS/MS). In addition, sphingomyelins (SM) containing phytosphingosine residues were elevated about 5-fold (data not shown). However, these SMs represent only 2% of the total SM content in control intestine. On the other hand, the most prominent SMs in intestine (i.e. d18:1;n:0) with sphingosine and nonhydroxylated fatty acid residues decreased roughly by 25% in the mutant tissue. In sum, the total sphingomyelin content in mutant intestine appeared unaltered (Fig. 10, H, TLC; and K, LC-MS/MS).
A detachment of enterocytes from the basal membrane was observed 4 days past tamoxifen induction. To investigate whether this phenomenon occurred due to a dislocation or reduction of adhesion molecules, expression of E-cadherin and ␤-catenin was investigated. Enterocytes of both control and mutant mice expressed E-cadherin in a basal and basolateral fashion throughout the whole epithelium (Fig. 11A). ␤-Catenin was stained strongly in basal and basolateral regions of control epithelium and in the lower compartments of villi of Ugcg f/f/ VilCreERT2 mice. However, only partial ␤-catenin expression or complete absence was seen at the tips of the villi in mutant mice (Fig. 11B). Immunofluorescence data could be confirmed by Western blot analysis (Fig. 11, C and D).
Stromal Immune Cells in Intestine Were Not Changed by the Ugcg Deletion-To investigate whether an infiltration of immune competent mononuclear cells would have triggered the hyperproliferation in mutant intestines, both newborn and adult mutant mice and respective controls were stained for B-cells, T-cells, and monocyte-derived cells. The numbers of immune cells infiltrating intestinal tissue of newborn control mice was very low in general but increased during adulthood (data not shown). However, none of the stainings of mutant newborn and adult intestine showed significant increased numbers of immune cells as compared with the respective control tissues.
Intestinal Phenotype in Ugcg f/f/VilCreERT2 Mice Could Not be Restored by Exogenous GlcCer Feeding-Mutant mice have been fed with chow diet, supplemented with or without 2 mg/g of glucosylceramide ad libitum, and both groups were injected with tamoxifen as described. Mutant mice with GlcCer supplementation showed an almost identical decrease in body weight as mutants that received chow diet (data not shown). Animals of both groups had to be sacrificed due to their severely compromised state of health 3 days after the initial tamoxifen induction. Morphologically, GlcCer treated and untreated Ugcg f/f/ VilCreERT2 mice were undistinguishable (data not shown) suggesting that exogenous GlcCer-application could not compensate for the Ugcg deletion.
Subcellular Localization of Enterocytic GSLs-To verify the role of GSLs for the observed phenotype in mutant mice, their intestinal localization and possible involvement in intracellular FIGURE 8. Ugcg f/f/VilCreERT2 mice showed severe structural alterations in the intestine shortly after induction of the Ugcg gene deletion. Ugcg control and mutant mice were treated for 3 consecutive days with 1 mg of tamoxifen (TAM), respectively. A, GSLs in the intestine were absent 4 days past the initial tamoxifen application (left and right TLCs represent neutral and acidic GSLs) and were remarkably reduced by ϳ50 and ϳ70% already 1 and 2 days after the initial induction (B). The remaining bands in extracts of mutant intestine migrating at the height of GM3 might be explained by admixtures of both intestinal muscle and stroma that were not affected by the Ugcg gene deletion. The degree of GSL depletion (B) correlated well with the dramatic alterations of the structure of the small intestine as shown by hematoxylin/eosin staining (BЈ). C-E, controls and Ugcg f/f/VilCreERT2, n ϭ 5; heterozygous mice, n ϭ 3). Shortly after induction, Ugcg f/f/VilCreERT2 mice showed reduction of bodyweight (C) as well as a lower food (D) and water consumption (E). Heterozygous mice developed similar to other controls. F, uptake of liposomes is impaired in Ugcg f/f/VilCreERT2 mice. Liposomes were rapidly endocytosed by control intestine. In mutant intestine the uptake of liposomes is drastically diminished or absent; scale bars, 100 m. Contr., control. transport were investigated. In this respect, stainings of GM1 together with the brush border protein villin as well as clathrin and Rab11 were performed in control and mutant mice postnatally at days 5-7. GM1, ubiquitously distributed in vesicle-like structures in the cytosol of control mice postnatally at day P0 (see Fig. 3D), concentrated apically below the brush border 1 week after birth (Fig. 12, A, C, and E). Only a small overlap could be seen between GM1 and villin in the microvillar region of the wild type intestine (Fig. 12A). GM1 stained negative in mutant tissue, and the brush border protein villin was detected only rudimentarily. A costaining of GM1 with clathrin, one marker of coated pits and vesicles (Fig. 12C) as well as GM1 with the recycling endosomal marker protein Rab11 (Fig. 12E) in control tissue, indicated close proximity apically below the brush border. GM1 similarly as at P0 was again not stained in mutant tissue (Fig. 12, D and F). The expression of clathrin and Rab11 was drastically down-regulated in mutant tissue 1 week after birth (Fig. 12, D and F). In addition, also the lysosomal marker Lamp1 decreased in Ugcg f/f/VilCre intestine (Fig. 12H).
The data obtained by immunofluorescence could be confirmed with Western blot analysis. The expression of proteins involved in endocytotic processes and in protein recycling such as clathrin, caveolin-1, and Rab11 was drastically down-regulated (Fig. 12, G  and J). However, the early endosomal marker proteins Rab4 and Rab5 were not significantly altered (Fig. 12K).
GSLs Cluster in Vesicles and Are Located Predominantly Below or Close to the Brush Border-Confocal immunofluorescence images have a limited resolution. To answer the question whether GSLs were involved in membrane events such as formation of caveolae and transport vesicles, immunogold EM stainings have been performed to elucidate the exact intracellular localization of GM1 in newborn (P5) as well as GA1 in adult mice. GM1 was found to be predominantly located in vesicle-like structures (Fig. 13A, arrows). These vesicles showed two opposite migration directions. The vesicular content, including GSLs, was shed into the intestinal lumen where GSLs have been found in the glycocalyx (Fig. 13, A and B) and also was released into lipid droplets (Fig. 13C). Absence of GM1 went along with an almost complete absence of lipid droplets in Ugcg f/f/VilCre mice (Fig. 13D).
A costaining of GM1 with clathrin demonstrated that coated pits formed at the apical membrane surface (Fig. 13E) mostly independent from GSLs; only a few gold particles that labeled clathrin were associated with GSL vesicles (Fig. 13E). Specific staining for GM1 and clathrin did not occur in mutant tissue. In GM1 and Rab11 coimmune EM, Rab11 was (with a few exceptions (Fig. 13G, inset)) not directly associated with GSL vesicles (Fig. 13G). Our immune electron microscopic data suggested that GSL vesicle formation essentially did not depend on clathrin and Rab11.
Similar to newborn mice, in adult animals GSL (GA1) could be visualized in vesicle-like structures (Fig. 13I, arrows). Those vesicles appeared to be released into the intestinal lumen (Fig.  13J) and again attached to lipid depots (Fig. 13K). Ugcg f/f/Vil-CreERT2 tissue showed only few dots of background staining (Fig. 13L).

DISCUSSION
This study reports new insights into the role of GSLs for intestinal epithelial function. GSLs have been surmised to be important for epithelial polarization (28) and differentiation (29). Indeed, in C. elegans mutants in whom GSL synthesis was inhibited by feeding of RNAi against glucosylceramide synthase (Ugcg), an impaired distribution of proteins involved in intestinal polarization such as ERM-1 and actin has been shown. A depletion rate of ϳ75% of glucosylceramide was sufficient to induce those dramatic defects (11).
Are these results also to be expected in mammals? Based on our data with GSL depletion in the liver in which bile canalicular function was maintained (16), we have postulated a different function of GSLs in mouse enterocytes but for epithelial polarity. Two mouse models, an inducible model in adult mice and a constitutive model in which GSLs were depleted during embryogenesis, were applied. The latter model provided the advantage to investigate the morphology of the intestine shortly after birth when its absorptive activity was not yet fully developed. Such a state of functional quiescence cannot be tested in C. elegans, and might be one reason that a similar phenotype has not been observed in worms. Surprisingly, the intestine of newborn mutants at P0 could neither macroscopically nor by microscopy be distinguished from that of control littermates, but the mutants lacked GSLs in the intestine. Also, epithelial proliferation did not differ from control intestine at P0, suggesting that GSL depletion had no immediate influence on the cell cycle of enterocytes. An epithelial-to-mesenchymal transition process in which GSLs were described to play a critical role in vitro (30,31) also did not occur as enterocyte morphology appeared normal at that stage and enterocytes seemed fully differentiated and polarized. However, newborn mutant offspring at P0 were not able to absorb milk fat in a similar amount as their control littermates. This was an initial hint that GSLs might be directly involved in lipid absorption and/or transport. The finding that villin, a marker of enterocytic polarization, was unchanged in GSL-depleted intestine at P0 and that the brush border was regularly formed suggested that GSLs were not immediately needed for microvillar generation and consequently for epithelial enterocytic polarization. This observation corroborated the results from mouse models with GSL depletion in liver (16) and in kidney. 4 In these organs, neither hepatocytic autoplasmic microvillus-like protrusions into canaliculi of hepatocytic epithelial cells nor microvilli of the proximal renal epithelium were altered upon GSL depletion. The latter findings implied the question of the cellular localization of GSLs in enterocytes. We surmised that GSL could be associated with proteins involved in intracellular transport. In control mice a few days after birth, we could demonstrate, by confocal microscopy and more clearly by immune electron microscopy,   12. GM1 to a major extent was found apically in vesicular structures in the cytosol and stained negative in mutant mouse tissue. All stainings shown were performed on intestinal jejunum sections collected 5-7 days after birth. Animals of the same age (littermates) were compared. A and B, costaining of GM1 together with villin. In contrast to newborn mice at day P0 in which GM1 was ubiquitously located in small dots in the cytosol of the epithelium, 1 week after birth GM1 located predominantly on the apical site of the enterocytes (A). The structural protein villin was stained in the brush border, but only a small overlap could be seen together with GM1. GM1 was stained to a major extent apically, apart from the brush border. B, in mutant tissue, as expected, GM1 could not be detected but also villin expression was very weak. C and D, costaining of GM1 and clathrin. In control tissue GM1 and clathrin were found in similar regions apically in the enterocytes (C). Clathrin expression was much less pronounced in mutant epithelium but was still present in the intestinal zona muscularis propria (D). E and F, costaining of GM1 together with the recycling endosomal marker protein Rab11. GM1 could be located apically in similar regions as Rab11 (E). Rab11 staining in mutant intestine was very weak (F). G, Lamp1-positive lysosomes were stained and were regularly distributed in the cytosol of the intestinal epithelium of control mice. H, in mutant tissue, Lamp1 staining appeared weaker, and lysosomes were irregularly distributed within the cytosol of the enterocytes. I-K, Western blot analysis revealed drastically decreased levels of clathrin and caveolin-1, proteins involved in endocytosis (I) as well as Rab11 as a marker for recycling endosomes important for protein recycling (J). The early endosomal marker proteins Rab4 and Rab5 showed similar expression in control and mutant tissue (K). Contr., control.
the existence of small GSL-containing vesicles at the apical membrane located below the brush border in the cytosol, similar to data reported earlier (3). These vesicles followed a transport route from the apical membrane to the enterocytic lipid depots and hence seemed to be involved in endocytosis and intracellular transport of nutritional lipids. The GSL-dependent vesicle formation, essentially devoid of clathrin and Rab11, which are also involved in endocytosis and recycling processes (32)(33)(34)(35), might be unique for the intestine due to its high concentration of glycosphingolipids. The question whether GSL vesicles require proteins for their formation still remains to be elucidated. Because of the disturbed lipid absorption, the intestine reacted with hyperproliferation; infiltration of immune cells that potentially may have triggered such a process could not be observed. As consequence of the epithelial hyperproliferation, a regular differentiation of the enterocytes was inhibited. Usually, in wild type animals, the intestinal epithelium exhibits a small proliferative area located in the crypts of villi. Cells differentiate into enterocytes and, while migrating to the apex of the villus, exert their absorptive function; they are then released into the gut lumen after 3-5 days (36). In our mutant mice, with fully developed phenotype, the proliferative zone was not restricted to the crypts of the villi, and it extended to half of the epithelium or even higher if associated with immature differentiation of the epithelial enterocytes 5-7 days after birth or 3-4 days after tamoxifen-induced Ugcg deletion in adult animals. The altered epithelial differentiation could also be depicted by the rudimentary formation of the brush border. It should be stressed, however, that GSL depletion did not primarily induce hyperproliferation, because in newborn mice with GSL deficiency, the proliferative rate of enterocytes was comparable with control mice.
Caveolin-dependent invaginations have been proposed to be involved in several biological processes such as cell signaling, membrane traffic (2,37), and endocytosis (3). It cannot be definitively decided whether a disturbed epithelial differentiation might have been accompanied with a reduced expression of caveolin or its intracellular recycling was affected. In our GSL-deficient mice, a significant reduction of caveolae formation was indicated by electron microscopy and Western blot.
Although the structural alterations in the intestine after Ugcg deletion appeared very prominent, the mRNA expression levels of proteins, actively thought to be involved in lipid binding and intracellular transport, were essentially unaltered. Nei- FIGURE 13. By immune electron microscopy, GSLs predominantly localize in vesicles close to the brush border. A-C, immunogold EM stainings of GM1 in control intestine of newborn mice at P5. GM1 was visualized associated in vesicles, below the brush border (mv) and in the region of the glycocalyx (gc) (A, arrows). Only rudimentary staining was seen in the microvilli. B, GSL-containing vesicles seemed to release their content at the apical membrane of the enterocytes into the intestinal lumen and intracellularly into lipid depots (LD) (A and C). D, only scarce unspecific gold particles were observed in enterocytes of Ugcg f/f/VilCre mice. E and F, immunogold EM costaining of GM1 (10 nm gold particles) with clathrin (5-nm gold particles) in newborn control (E) and mutant mice (F) at P5. Clathrin-coated pits (arrows) originate in wild type mice mostly independent from GSLs at the apical membrane of the enterocytes (E). Only few clathrin-positive dots colocalized with GM1 vesicles (arrowheads). In mutant mice GM1 and clathrin were almost completely absent (F). G, similar as clathrin, also Rab11 (10-nm gold particles, arrowheads) predominantly did not colocalize with GM1 vesicles (G, arrows, 5-nm gold particles) with a few exceptions (inset). H, only a background staining for GM1 and a reduced Rab11 staining was seen in mutant tissue. I-L, in adult animals, the GSL GA1, similar as GM1 in newborns, also localized in vesicles. GSL (GA1)-associated vesicles (I, arrows) are present in a layer close to the brush border (J) or in lipid droplets (LD) (K) similarly to the location of GSLs in newborn animals 1 week after birth. L, Ugcg f/f/VilCreERT2 intestine showed only a few unspecific gold particles; m, mitochondria; scale bars, 200 nm. Contr., control.
ther the intestinal fatty acid-binding proteins Fabp2 and CD36 nor enzymes important for triglyceride synthesis and transfer such as Dgat1 and Mttp were statistically significantly changed. Also the intestinal cholesterol transporters Abcg5 and Abcg8 as well as the short chain peptide transporter Pept1 and the glucose transporters Sglt1 and Glu5 were not significantly altered. However, the down-regulation of the membrane integral protein Aqp8, which has been described to function on the apical side of duodenum, jejunum, and colon (38), was accompanied by a lower intestinal water absorption associated with a significant increase of hematocrit levels in Ugcg f/f/VilCreERT2 mice (data not shown). Reduced expression and activity of Aqp8 water channels may have been one factor for the extreme soft stools observed in mutant mice.
Fat from nutrients, predominantly in the form of triglycerides, is primarily cleaved into monoacylglycerols and free fatty acids by secreted pancreatic lipases. The first step in lipid absorption, the movement of monoacylglycerols and fatty acids across the apical membrane of enterocytes, is not well defined on a molecular level; it very likely involves different mechanisms. Several studies describe protein-dependent mechanisms, although others have reported protein-independent transport via diffusion through the apical membrane of the enterocytes (39,40) for which in particular bile salts appear to be necessary as they form micelles with the fatty acids and facilitate thereby their absorption (41). Lipids also might be taken up via endocytosis. Ex vivo, in intestinal tissue sections it could also be shown that exposure of high lipid concentrations led to increased formation of clathrin-containing invaginations and vesicles that were then transported and released to intracellular lipid depots (42). This lipid uptake process appears to be particularly important when the lipid concentration in the intestine is extremely high such as in the situation of newborn suckling mice receiving high fat milk. Indeed, in our in vivo studies, we now could demonstrate that large scale fat from milk and in the form of liposomes can be endocytosed. However, this process depended on GSL expression. In support, caveolar endocytosis in ovary cells was significantly reduced after GSL depletion by inhibition of GSL synthesis (43).
Intestinal GlcCers predominantly express phytosphingosine with ␣-hydroxylated fatty acid in their ceramide anchor. In particular, in the adult intestinal model, most of this GlcCer accumulated in the form of its precursor ceramide, and only about 1/5 of it was shuttled into phytosphingosine-containing SM. This is in contrast to what we had found previously in brain and liver, where almost all of the ceramide that could not be converted to GlcCer was shifted into sphingomyelin (12,16). However, to a similar extent as phospholipids, the major sphingosine-containing SM fraction decreased, probably as a consequence of loss of the brush border accompanied with loss of membrane. Increased ceramide accumulations particularly occurring in adult Ugcg-deficient mutants may have triggered the pronounced autophagy (26,27) and apoptosis.
␤and ␣-catenin form complexes with E-cadherin that are important for anchoring epithelial cells to the lamina propria. In adult tamoxifen-induced mutant mice, a lowered expression of ␤-catenin has been observed, which may have facilitated the detachment of enterocytes. Moreover, it has been demon-strated that ␤-catenin binds GA1, which is the prominent GSL in adult enterocytes. GSL depletion may have indirectly also negatively influenced the epithelial binding to the basal lamina (31).
The phenotype observed in our mutant mice could not be alleviated by the uptake of exogenous GSL via milk or by adding GlcCer to the food. In this respect, it has been convincingly shown that exogenously applied GSL was almost completely degraded (44 -47).
GSL expression in the intestine very likely may not have been influenced by an uptake of shed liver GSL via the blood as their concentration in the blood is very low. In addition, the blood GSL pattern differs from the intestinal GSL, and in particular the main GlcCer bands in the neutral GSLs showed a different migration speed as GlcCers from blood and milk (data not shown), suggesting a different ceramide anchor as the ones mainly found in intestinal GSLs.
In summary, glycosphingolipid expression in the intestinal epithelium is quintessential for maintenance of a GSL signature of absorption-dependent intracellular transport and intestinal structure. However, GSLs are not primarily required for enterocyte-polarized brush border formation.