Aminopeptidase N (CD13) Is a Molecular Target of the Cholesterol Absorption Inhibitor Ezetimibe in the Enterocyte Brush Border Membrane*

Intestinal cholesterol absorption is an important regulator of serum cholesterol levels. Ezetimibe is a specific inhibitor of intestinal cholesterol absorption recently introduced into medical practice; its mechanism of action, however, is still unknown. Ezetimibe neither influences the release of cholesterol from mixed micelles in the gut lumen nor the transfer of cholesterol to the enterocyte brush border membrane. With membrane-impermeable Ezetimibe analogues we could demonstrate that binding of cholesterol absorption inhibitors to the brush border membrane of small intestinal enterocytes from the gut lumen is sufficient for inhibition of cholesterol absorption. A 145-kDa integral membrane protein was identified as the molecular target for cholesterol absorption inhibitors in the enterocyte brush border membrane by photoaffinity labeling with photoreactive Ezetimibe analogues (Kramer, W., Glombik, H., Petry, S., Heuer, H., Schäfer, H. L., Wendler, W., Corsiero, D., Girbig, F., and Weyland, C. (2000) FEBS Lett. 487, 293–297). The 145-kDa Ezetimibe-binding protein was purified by three different methods and sequencing revealed its identity with the membrane-bound ectoenzyme aminopeptidase N ((alanyl)aminopeptidase; EC 3.4.11.2; APN; leukemia antigen CD13). The enzymatic activity of APN was not influenced by Ezetimibe (analogues). The uptake of cholesterol delivered by mixed micelles by confluent CaCo-2 cells was partially inhibited by Ezetimibe and nonabsorbable Ezetimibe analogues. Preincubation of confluent CaCo-2 cells with Ezetimibe led to a strong decrease of fluorescent APN staining with a monoclonal antibody in the plasma membrane. Independent on its enzymatic activity, aminopeptidase N is involved in endocytotic processes like the uptake of viruses. Our findings suggest that binding of Ezetimibe to APN from the lumen of the small intestine blocks endocytosis of cholesterol-rich membrane microdomains, thereby limiting intestinal cholesterol absorption.

Intestinal cholesterol absorption is a main regulator of serum cholesterol homeostasis (1) involving digestion and hydrolysis of dietary lipids with formation of mixed micelles containing cholesterol, bile salts, fatty acids, and phospholipids (2). The molecular mechanisms being involved in cholesterol absorption are not understood but the findings of a strong species difference (2), sterol specificity (3), and the existence of specific cholesterol absorption inhibitors (4,5) strongly argue for a protein-mediated process. Several proteins have been suggested as candidates for the putative intestinal cholesterol transporter (6 -10) but for none has evidence as a cholesterol uptake system been presented. Investigations with CaCo2 cells have shown that cholesterol taken up from mixed cholesterol bile salt micelles is distributed into the brush border membrane and is moved to detergentresistant microdomains (rafts) followed by transport from these microdomains to the endoplasmic reticulum for esterification and further assembly into chylomicron particles that are secreted by the enterocyte (11). This suggests that cholesterol absorption occurs by a complex process involving several proteins rather than by a single cholesterol transporter. Consequently, to elucidate the molecular mechanisms involved in intestinal cholesterol absorption we attempted to identify the protein components of this machinery by photoaffinity labeling using photoreactive analogues of the cholesterol absorption inhibitor Ezetimibe (12, 13) ( Fig. 1) and of cholesterol (14). With photoreactive Ezetimibe derivatives we identified an integral 145-kDa membrane protein as the target protein for cholesterol absorption inhibitors in the enterocyte brush border membrane (12,13), whereas an integral 80-kDa membrane protein was identified as a specific cholesterol-binding protein (15). The 145-kDa Ezetimibe-binding protein showed an exclusive affinity for cholesterol absorption inhibitors, but did not bind cholesterol or phytosterols. Vice versa, the 80-kDa cholesterol-binding protein only bound cholesterol and plant sterols without showing affinity for cholesterol absorption inhibitors. Both binding proteins have an identical tissue distribution restricted to the anatomical site of cholesterol absorption, the small intestine (13,15). In the present article we localized the molecular mode of action of cholesterol absorption inhibitors to the luminal side of the small enterocyte brush border membrane and identified the 145-kDa target protein for Ezetimibe in the enterocyte brush border membrane as the ectoenzyme aminopeptidase N ((alanyl) aminopeptidase; EC 3.4.11.2; leukemia antigen CD13).

Materials
The cholesterol absorption inhibitors C-1, C-2, S 6053, S 6130, ezetimibel, ezetimibe glucuronide, and S 6504 were synthesized at Aventis * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Animals and Membrane Preparations
Male New Zealand White rabbits weighing 4 -5 kg (Harlan Winkelmann, Borchem, Germany) were kept on Altromin® standard diet C 2023 (Altromin®, Lage, Germany) ad libitum. Brush border membrane vesicles (BBMV) from rabbit stomach, duodenum, jejunum, ileum, cecum, colon, rectum, and kidney were prepared by the Mg 2ϩ precipitation method as described earlier (20). For preparation of BBMV from small intestinal segments, the small intestine from 2 rabbits was divided in 8 segments of equal length (numbered 1 to 8 from duodenum to ileum) and BBMV were prepared and characterized as described (20). Rat liver microsomes and rat adipocyte membranes were prepared as described elsewhere (21,22).

Cholesterol Absorption and Transport Measurements
Intestinal cholesterol absorption was determined by a modification of the Zilversmit/Hughes method (23) as described earlier (12). Cholesterol monomerization activity from mixed micelles was determined by measurement of the dequenching of fluoresterol-labeled cholesterol micelles using the assay described by Cai et al. (24). The uptake of [ 3 H]cholesterol from mixed micelles by rabbit small intestinal BBMV was determined as described by Compassi et al. (25) using mixed micelles prepared by mixing 10 M [ 3 H]cholesterol with a solution of 4 mM taurocholate and 0.6 mM sodium oleate in 50 mM Tris/HCl buffer (pH 7.4), 150 mM NaCl buffer, followed by sonification and evaporation of chloroform. Uptake was measured by mixing BBMV (20 g of protein) in 50 l of 10 mM Tris/Hepes buffer (pH 7.4), 100 mM NaCl, 100 mM mannitol in the absence or presence of 150 M of the respective Ezetimibe analogues with 50 l of the mixed micelle suspension, followed by centrifugation at 100,000 ϫ g for 120 s after the indicated incubation times and subsequently radioactivity in the pelleted material and the supernatant was determined by liquid scintillation counting.
Wheat Germ Lectin and Hydroxylapatite Chromatography After Photoaffinity Labeling with 3  H]C-2) at 254 nm the samples were collected and after washing three times membrane proteins were solubilized with solubilization buffer. The supernatant containing solubilized membrane proteins was added to 0.5 ml of wheat germ lectin-agarose gel. After 60 min at 20°C the beads were collected by centrifugation and washed 3 times with 2 ml of 10 mM PBS, 1% n-octyl glucoside. Adsorbed proteins were eluted with 4 portions of 1 ml of 10 mM PBS, 1% n-octyl glucoside, 300 mM N-acetyl-Dglucosamine each. The N-acetyl-D-glucosamine eluates were diluted to 10 ml with 10 mM sodium phosphate buffer (pH 7.4), 1% (w/v) n-octyl glucoside and applied to a hydroxylapatite column (self-filled, 10 cm height, 1 cm diameter; or Bio-Rad CHT-II columns) equilibrated with 10 mM sodium phosphate buffer (pH 7.4), 1% (w/v) n-octyl glucoside at a flow rate of 0.15 ml/min and collection of 1-ml fractions. Subsequently proteins were eluted as follows: 10 ml of 10 mM sodium phosphate buffer (pH 7.4), 1% (w/v) n-octyl glucoside followed by linear phosphate gradients in the above buffer. In each fraction the activity of aminopeptidase N, sucrase, and cholesterol monomerization as well as radioactivity was determined. From each fraction 100-l aliquots were analyzed by SDS-PAGE with subsequent determination of the distribution of radioactively labeled proteins, after precipitation of proteins with chloroform/methanol (26).
Streptavidin-biotin Chromatography after Photolabeling with C-4 or C-5-Ten samples of rabbit ileal BBMV (200 g of protein) were incubated with the biotin-tagged cholesterol absorption inhibitors C-4 or C-5 in 10 mM Tris/Hepes buffer (pH 7.4), 100 mM NaCl, 100 mM mannitol for 30 min in the dark at 20°C followed by irradiation at 254 nm for 30 s. After washing, proteins were solubilized and the clear supernatant was mixed with 0.5 ml of streptavidin-agarose beads and kept under stirring at 4°C for 2 h. After centrifugation, the beads were incubated with 2 ml of 10 mM Tris/Hepes buffer (pH 7.4), 300 mM mannitol, 1% n-octyl glucoside, 4 mM phenylmethylsulfonyl fluoride, 4 mM iodacetamide, 4 mM EDTA for 10 min at 4°C followed by centrifugation. After repeating this procedure twice, proteins were eluted from the streptavidin-agarose beads with 2 ml of the above buffer containing 6 mM biotin standing overnight at 4°C, repeated with 2 ml for 1 h at 4°C. Final purification was achieved by preparative SDS-gel electrophoresis (6% gel, 28 mm diameter, 5-cm gel length, Bio-Rad) at 500 V (40 mA, 6 W) and the eluates were fractionated into 0.6-ml fractions.
Ezetimibe Affinity Chromatography-An Ezetimibe affinity matrix was synthesized by coupling of 1-(4-aminomethyl-phenyl)-3-(3-hydroxy-3-phenyl-propyl)-4-(4-methoxy-phenyl)-azetidin-2-on (16) to a hydroxy succinimidyl-hexanoyl matrix (Hi-Trap column, Amersham Biosciences) according to the protocol of the manufacturer. N-Acetyl-Dglucosamine eluates from wheat germ agglutinin chromatography were applied to the Ezetimibe affinity column at a flow rate of 0.25 ml/min followed by elution with 10 ml of buffer and collection of 500-l fractions. Bound proteins were eluted with 5 ml of PBS (pH 7.4), 3% Triton X-100. From all fractions aliquots were analyzed for the enzymatic activity of aminopeptidase N and sucrase as well as protein composition by SDS-gel electrophoresis.

SDS-gel Electrophoresis
SDS-PAGE was carried out in vertical stab gels (20 ϫ 17 ϫ 0.15 cm) using an electrophoresis system LE 2/4 (Amersham Biosciences) with gel concentrations of 7-10.5% at a ratio of 97.2% acrylamide and 2.8% N,N-methylene bisacrylamide or in pre-casted NOVEX gels (4 -12, 12, or 15%, Invitrogen, Groningen, The Netherlands) using an electrophoresis system XCell II from Novex (27). After electrophoresis the gels were fixed in 12.5% trichloroacetic acid followed by staining with Serva Blue R 250. For determination of the distribution of radioactivity, individual gel lanes were cut into 2-mm pieces, protein was hydrolyzed with 250 l of tissue solubilizer Biolute S, and after addition of 4 ml of scintillator Quickszint 501 radioactivity was measured by liquid scintillation counting. Western blotting and immunostaining was performed as described earlier (27).
Confluent CaCo-2 cells cultivated for a further 11 or 20 days were used for the transport experiments. Cells were incubated either with culture medium (see above) (control), medium containing 150 M of the cholesterol absorption inhibitors Ezetimibe or S 6130 and kept at 37°C overnight. After 12 h of incubation 100 l of mixed micelles containing 100 M [ 3 H]cholesterol, 4 mM taurocholate, and 0.6 mM sodium oleate in 50 mM Tris/HCl buffer (pH 7.4), 150 mM NaCl were added and [ 3 H]cho-lesterol uptake was measured after 0, 1, 2, 3, 4, and 5 h of incubation. Medium was removed, and cells were washed twice with 10 mM Tris/ Hepes buffer (pH 7.5), 300 mM mannitol, Pefabloc. After transfer to scintillation vials, 400 l of tissue solubilizer Biolute S (Zinsser Analytic, Frankfurt, Germany) was added followed after 2 h by scintillator Picofluor 40 (PerkinElmer Life Sciences) and radioactivity was measured with a Wallac liquid scintillation counter.

Effect of Ezetimibe on the Binding of the WM-47 mAB to APN/ CD13 as Assessed by Confocal Laser Scanning Microscopy
For immunofluorescence staining of CD13, CaCo-2 cells (3 days after post-confluence) were cultured on Lab-Tek TM sodium borosilicate cov- erslips (Nunc, Wiesbaden, Germany). The cells were rinsed several times with Dulbecco's modified PBS (Biochrom, Berlin, Germany). Incubation either with vehicle (Dulbecco's PBS) or with 10 M Ezetimibe at room temperature for 15 min was carried out before or after immunofluorescence staining of the cells with FITC-CD13 mAB. Following extensive washing the cells were fixed with methanol/acetone (1:1) at Ϫ20°C for 10 min and processed for confocal laser scanning microscopy according to standard protocols. Fluorescence staining was recorded at 488 nm excitation with an inverted TCS 4D confocal laser scanning microscope (Leica Lasertechnik, Heidelberg, Germany), equipped with an Ar-Kr 75 mV mixed ion laser (Melles Griot Inc., Carlsbad, CA) and an acousto-optical transmission filter. Images in horizontal x-y sections were acquired by using the ScanWare instrument software. Image analysis was performed using the MetaMorph® software package (Universal Imaging Corp., Downingtown, PA).

Matrix-assisted Laser Desorption Ionization and ESI Mass Spectrometry
Coomassie Blue-stained protein bands were excised, destained (3 times for 30 min with 100 l of 50% acetonitrile, 25 mM NH 4 HCO 3 , pH 8.0), and dried by acetonitrile. Gel pieces were rehydrated in 15 l of trypsin solution (5 g/ml) (rec., proteomics grade, Roche Diagnostics) and incubated at 37°C overnight. Peptide extracts (50% acetonitrile, 5% trifluoroacetic acid) were pooled (3 ϫ 30 l), lyophilized, and reconstituted in 13 l 2% acetonitrile, 0.1% trifluoroacetic acid. Analysis of the peptide samples was performed on a nano-ESI-LC-MS/MS system (LC Packings, Amsterdam, coupled to a LCQ Deka XP mass spectrometer, Thermo Finnigan, San Jose, CA). 13 l of sample were desalted on a C18 precolumn (PepMap, inner diameter 300 m, 5-mm length); loading and washing of the sample with 2% acetonitrile, 0.1% trifluoroacetic acid was performed at a flow rate of 30 l/min for 10 min. A C18 nanocolumn (PepMap, inner diameter 75 m, 150 mm length, LC Packings) was used to separate the peptides at a flow rate of 200 nl/min (solvent A: 2% acetonitrile, 0.1% formic acid; solvent B: 98% acetonitrile, 0.1% formic acid; linear gradient from 5% B to 35% B in 45 min/100% B for 10 min). Nano-electrospray needles were laboratory pulled (Sutter Instruments Co., model P-2000) from fused silica capillaries (inner diameter 25 m, outer diameter 280 m, Grom) resulting in a needle orifice of ϳ3 m in diameter (liquid tight junction in MicroTee (Upchurch) via a platinum conductor; spray voltage 1.2 kV). Mass spectrometric analyses were controlled by XCalibur software (Thermo Finnigan, San Jose, CA) using data-dependent aquisition switching between MS and MS/MS in the positive ion mode. The transfer capillary temperature was constantly held at 180°C, the capillary voltage and the tube lens offset on 46 and 55 V, respectively. Peptides eluted from the column were detected in a first scan event in MS mode (m/z 500 -2000, 3 microscans, maximum injection time 50 ms), followed by three consecutive data-dependent MS/MS scan events (isolation width 3 Da, 4 microscans, maximum injection time 400 ms, activation time 30 ms) for the three most abundant ions (above 5 ϫ 10 5 counts) using a relative collision energy of 35% (corresponding to the XCalibur software settings). The dynamic exclusion parameters were set as follows: "repeat count" 2, "repeat duration" 0.5 min, "exclusion list size" 25, "exclusion mass width" Ϯ 1.5 Da, "exclusion duration" 1.50 min, no rejections.

Protein Identification
Mass data were processed using the search engine SpectrumMill (version 2.8, Millenium Pharmaceuticals Inc., Cambridge, MA) using the following parameters: no "cystein modification," "minimal sequence tag" Ͼ1, "scan range" 1-9999 (all), "[MϩH] ϩ " 500 -4000 Da, "parent charge assignment" find force 1 through 4/find max(z) 7/min MS S/N 25, "merge scans with the same parent" m/zϮ1 scan (no merging of spectra). Protein identifications were obtained by comparison of experimen- tal data to the latest SwissProt data base (taxonomy: mammals). Searches were done with matching tolerances of Ϯ1.5 Da and Ϯ0.7 Da for parent and fragment masses, respectively. A maximum number of 2 missed tryptic cleavages was allowed. A peptide sequence tag was regarded as reliable when the identification met the following parameters set in the validation filter: "protein score" Ͼ8, "peptide score" Ͼ8, "%SPI" Ͼ70; manual inspection/deselection of these spectra was additionally carried out.

RESULTS
The Luminal Side of the Enterocyte Brush Border Membrane Is the Site of Action for Ezetimibe-The release of cholesterol from luminal mixed micelles as a prerequisite for cholesterol absorption is a protein-mediated process as shown with homogenates from hamster small intestine (24). Rabbit small intestinal BBMV protein dependently catalyzed the monomerization of fluoresterol from mixed micelles (Fig. 2, a and b), whereas after preincubation of BBMV with proteinase K no cholesterol monomerization occurred (Fig. 2a). Incubation of BBMV or mixed micelles with Ezetimibe (analogues) had no influence on cholesterol monomerization (Fig. 2c) . 2d), but neither the simultaneous presence nor the preincubation of BBMV or mixed micelles with Ezetimibe (analogues) had any influence on the uptake of [ 3 H]cholesterol by the enterocyte brush border membrane (Fig.  2e). An essential structural feature for the pharmacological activity of Ezetimibe is the ␤-lactam ring structure and it may be speculated that this structural element is responsible for its pharmacological activity by covalent modification of the target protein(s) similar to the antibacterial activity of the ␤-lactam antibiotics. However, even after 24 h of incubation of BBMV with radioactively labeled Ezetimibe analogues neither covalently labeled proteins nor the carboxyl metabolite could be detected (data not shown), making a mode of action analogue to ␤-lactam antibiotics unlikely. These findings clearly indicate that Ezetimibe blocks intestinal cholesterol absorption at the level of the enterocyte brush border membrane or by influencing post-membrane processes.
Ezetimibe is rapidly absorbed in the upper small intestine and undergoes extensive first-pass metabolism in the intestinal wall to a glucouronide as an active metabolite (28). Therefore, an intracellular molecular target for Ezetimibe action cannot be ruled out a priori. Consequently, we have designed membrane-impermeable Ezetimibe analogues (17) to investigate whether binding to the enterocyte brush border membrane is sufficient for inhibition of cholesterol absorption. Photoaffinity labeling of rabbit small intestinal BBMV with the 3 Hlabeled photoprobes C-1 or C-2 in the presence of the pharmacologically active membrane-impermeable Ezetimibe-analogue S 6130 concentration dependently inhibited labeling of the 145-kDa Ezetimibe-binding protein (Fig. 3a). S 6504, the S 6130 analogue with an opened ␤-lactam ring, is pharmacologically inactive and did not inhibit photoaffinity labeling of the 145-kDa band (Fig. 3b). These investigations suggest that Ezetimibe and Ezetimibe analogues exert their pharmacological activity from the luminal side of the small intestine by binding to a 145-kDa integral membrane protein.
Purification of the 145-kDa Ezetimibe-binding Protein-Purification of the 145-kDa Ezetimibe-binding protein was significantly hampered by its resistance to solubilization with nondenaturating detergents (15); after numerous failures we finally found a protocol yielding 60 -80% solubilization of the 145-kDa protein by modifying the method of Mooseker (29) using 10 mM Tris/Hepes buffer (pH 7.4), 75 mM KCl, 5 mM MgCl 2 , 1 mM EGTA, 1 mM dithiothreitol, 1% Triton X-100, 1% n-octyl glucoside (solubilization buffer) or stepwise solubilization with 0.25% N-lauroylsarcosine followed by dissolution with solubilization buffer. For the purification of the 145-kDa Ezetimibe-binding protein we developed three different protocols. (a) A classical approach by chromatographic isolation of the radiolabeled 145 kDa protein after photoaffinity labeling of rabbit small intestinal BBMV with 3 H-labeled photoreactive Ezetimibe analogues. (b) Biotin-streptavidin affinity chromatography after photoaffinity labeling of rabbit small intestinal BBMV with biotin-tagged Ezetimibe photoaffinity probes (18). (c) Ligand affinity chromatography using an Ezetimibe affinity matrix. Biochemical investigations revealed the glycoprotein nature of the 145-kDa protein; deglycosylation with N-glycanase shifted the molecular mass from 145 to 115 kDa (Fig.  4A). Consequently, we investigated various lectins as matrices for a lectin affinity chromatography. With wheat germ lectinagarose a complete retardation of the radiolabeled 145-kDa protein was achieved (Fig. 5A) and with N-acetylglucosamine the entire amount of the 145-kDa Ezetimibe-binding protein could be eluted from wheat germ agglutinin-agarose (Fig. 5b). By a subsequent hydroxylapatite chromatography (Fig. 5c) a nearly complete separation of the radioactively labeled 145-kDa Ezetimibe-binding protein from other membrane proteins was achieved (Fig. 5d). Fractions were analyzed by SDS-PAGE and the distribution of radioactivity was determined by slicing of the gel lanes. Depending on the electrophoresis system the 145-kDa band retained by the streptavidin matrix was split into a double band of 145 and 150 kDa. Radioactivity was found in the bands of 145 and 205/210 kDa (Fig. 5e) suggesting that the 205/210-kDa band results from a dimerization of the 145-kDa protein. The estimated amount of purified 145-kDa protein was 1-3% of the material applied to the hydroxylapatite column indicating an enrichment factor 240 -300-fold from BBMV and 4800 -6000-fold with respect to enterocyte total protein. The radioactively labeled bands were excised, digested with trypsin, and peptides were submitted to sequence determination by electrospray mass spectrometry. In the 145-and 150-kDa bands, sequences for aminopeptidase N (APN), aminopeptidase A (APA), and sucrase/isomaltase (Table  I) were found; in the 205-and 210-kDa bands, also predominantly APN, APA, and sucrase/isomaltase were found indicating a hetero-or homodimerization of the labeled protein in the 145-150-kDa bands. Identical results were obtained with the C-1 and C-2-Ezetimibe photoprobes. Ezetimibe does not interact with the catalytic site of APN or APA because neither the enzymatic activity of APN and APA were inhibited by Ezetimibe (analogues) nor did inhibitors of APN and APA like bestatin show any effect on photoaffinity labeling of the 145-kDa Ezetimibe-binding protein or inhibit cholesterol absorption in vivo (data not shown).
To verify the identification of the Ezetimibe-binding protein of 145 kDa in the small enterocyte brush border membrane as APN, we have developed biotin-tagged photolabile Ezetimibe analogues C-4 and C-5 (16, 18) (Fig. 1, E and F). These membrane-impermeable Ezetimibe photoprobes inhibit intestinal cholesterol absorption in vivo (18) and specifically interact with the 145-kDa Ezetimibe-binding protein as is evident from competition labeling experiments (Fig. 6a). Incubation of rabbit small intestinal BBMV with C-4 or C-5 and subsequent ultraviolet irradiation led to a concentration-dependent incorporation of the Ezetimibe photoprobe predominantly into membrane proteins of 145 and 43 kDa as shown in Fig. 6b with jejunal vesicles. After photoaffinity labeling of rabbit small intestinal BBMV with C-4 or C-5 followed by solubilization, predominantly membrane proteins of 145 kDa were specifically retained by the streptavidin beads (Fig 7a, ϩh), whereas no proteins were detectable if ultraviolet irradiation was omitted (Fig. 7a, Ϫh). If photoaffinity labeling with C-4 was performed in the presence of other cholesterol absorption inhibitors, the amount of extractable 145-kDa protein was reduced indicating a direct competition between C-4 with cholesterol absorption inhibitors for binding to the 145-kDa protein (Fig. 7b). In contrast, substrates of other intestinal nutrient transporters for bile acids, fatty acids, glucose, oligopeptides, or amino acids did not influence the amount of extractable 145-kDa protein (data not shown). Labeling experiments with cell membranes from different organs revealed that only by labeling of BBMV from rabbit small intestine could a 145-kDa protein be extracted, whereas after labeling of membranes from stomach, cecum, colon, rectum, kidney, liver, or fat tissue no 145-kDa protein was retarded by streptavidin beads (Fig. 7b). Final purification was achieved by preparative SDS-electrophoresis leading to a complete separation of the 145-and 150-kDa proteins (Fig. 7c). Sequence analysis of the bands eluted from the preparative SDS gels revealed the identity of the 145-kDa band with APN, whereas the upper band (150 kDa) was a mixture of APN, APA, and sucrase/isomaltase (Table II); no other proteins could be identified in these fractions. To avoid retention of proteins by unspecific binding to the matrix or of protein complexes by protein-protein interactions, BBMV-membrane proteins were, subsequently to photoaffinity labeling with C-4, solubilized with 1% SDS in PBS or 0.25% N-lauroylsarcosine in PBS followed by dilution with solubilization buffer to a concentration of SDS or N-lauroylsarcosine of less than 0.1% and following application to the streptavidin beads. Under these conditions with biotin, mainly one protein band of 145 kDa was eluted that was identified as APN.
Because of the high affinity of Ezetimibe to its molecular target in the enterocyte brush border membrane (12), specific Ezetimibe-binding proteins should be retained by an Ezetimibe affinity column. Based on the structure-activity relationships for 2-azetidinone cholesterol absorption inhibitors (12,16,30) we have synthesized an Ezetimibe affinity matrix (Fig. 1G). The eluates from wheat germ agglutinin chromatography were applied to the Ezetimibe matrix and bound proteins were eluted with 3% Triton X-100 and finally purified by preparative SDS-gel electrophoresis. Sequence analysis of the affinity purified 145-kDa protein revealed its identity with APN; the only other sequence identified stemmed from keratin, probably by contamination during the purification process. Depending on the BBMV preparation the 145-kDa Ezetimibe-binding protein purified by either of the three methods contained varying amounts of a typical protein triplet of 42, 40, and 38 kDa (Fig.  7c), which were identified as ␤-actin (42 kDa), ␤-actin (40 kDa), and annexin II (38 kDa). Because of the resolution limits of the mass spectrometric methodology it cannot be excluded with absolute certainty that the final preparation of the 145-kDa Ezetimibe-binding protein contains a contaminant that may represent the Ezetimibe target; in that case, however, the putative contaminant would have to be present in less than 1% in the purified 145-kDa protein to allow escape of detection by electrospray-mass spectrometric sequence analysis. In none of the experiments using the three different purification protocols could we identify amino acid sequences belonging to the proposed candidates for an intestinal cholesterol transporter like the scavenger receptor type BI (SR-BI), the ABC transporters ABCA1, ABCG5, ABCG8, the Niemann-Pick proteins NPC1, NPC2, or Niemann-Pick C1-like protein 1 (NPC1L1) or the acyl-CoA:cholesterol acyltransferases ACAT1 or ACAT2.
Tissue Distribution of APN-It is generally assumed that  3 Fig. 5d were excised, digested with trypsin, and peptide fragments were analyzed by nano-LC-MS/MS spectrometry.

Fraction
Protein band Parentheses around plus signs indicate a weak signal for the respective proteins in these peaks.
intestinal cholesterol absorption primarily occurs in the duodenum and the jejunum with little or no contribution of the ileum (31)(32)(33). However, a careful evaluation of the literature clearly reveals that the entire small intestine has the capability to absorb cholesterol (33)(34)(35)(36). During gastrointestinal passage the concentration of micellar cholesterol decreases proximal to distal because of the absorption of cholesterol and fatty acids in the upper and mid-small intestine and of bile acids in the ileum. As a result, cholesterol is physiologically more efficiently absorbed in the upper small intestine, but the ileum has as well the capability to absorb cholesterol: based on tissue weight or on the length of intestinal segments, the transport capacities for cholesterol were similar in jejunum and ileum (35). Photoaffinity labeling of BBMV from rabbit ileum and rabbit jejunum with Ezetimibe photoprobes yielded similar labeling patterns with a slightly higher labeling in the ileum as we have reported earlier (12). For a more precise analysis we have prepared BBMV from 8 segments of rabbit small intestine, from duodenum to ileum, and analyzed the distribution of APN and other brush border enzymes and of the 145-kDa Ezetimibebinding protein along the gastrocolic axis. APN exerted an increase in the specific enzymatic activity (Fig. 8a) and protein expression (Fig. 8b) from duodenum to ileum; the peptidases APA and dipeptidylpeptidase IV showed a similar or even steeper gradient, whereas sucrase/isomaltase and ␥-glutamyltransferase did not show a concentration gradient along the gastrocolic axis. Photoaffinity labeling of BBMV from these intestinal segments with radiolabeled Ezetimibe analogues illustrate that the intensity of labeling of the 145-kDa Ezetimibe increased from duodenum to ileum concomitantly with APN (Fig. 8c). After photoaffinity labeling with the biotin-Ezetimibe probes the 145-kDa Ezetimibe-binding protein could be extracted from solubilized membrane proteins of all 8 intestinal segments with varying amounts of a 43-kDa protein (Fig. 9, a  and b); immunostaining with actin antibodies and sequencing revealed its identity with actin (Fig. 9c). Recently, it was suggested that a heteromeric protein complex of annexin II and caveolin 1 serves as the molecular target for Ezetimibe in the cytosol of enterocytes from Zebrafish and mouse (37). In the protein preparations extracted with streptavidin beads from solubilized BBMV of the 8 small intestinal segments after photoaffinity labeling with the biotin-Ezetimibe probe C-5 ( Fig.  9a) annexin II could be detected as a 36-kDa band with antiannexin II antibodies (Fig. 9d). However, no significant labeling of 36-(for annexin II) or 21-kDa bands (for caveolin1) was observable after photoaffinity labeling with the biotin-Ezetimibe probes (Fig. 9b) or with radiolabeled Ezetimibe probes C1 or C2 (data not shown), making a direct interaction of Ezetimibe with an annexin II-caveolin 1 complex as its mode of action unlikely.
APN/CD13 is expressed in the plasma membrane of many different cell types in different organs (38). Photoaffinity labeling of BBMV isolated from stomach, small intestine, colon, rectum, or kidney resulted only with BBMV from the small intestine in a significant labeling of the 145-kDa Ezetimibebinding protein (15) despite the presence of APN-like enzymatic activity in all membrane preparations. Fig. 10a reveals the presence of APN-protein in BBMV of rabbit jejunum and kidney but not in stomach or colon transversum. Whereas a prominent labeling of the 145-kDa Ezetimibe-binding protein occurred in BBMV from the small intestine (Figs. 6b and 9b), no labeling of the 145-kDa protein occurred in BBMV from stomach or colon transversum with only a faint labeling in renal BBMV (Fig. 10b). The labeling of the 145-kDa Ezetimibebinding protein in plasma membranes from different cell types obviously correlates with the presence of APN-protein as de-  PAGE (B). b, BBMV from rabbit ileum, kidney, or cecum (200 g of protein) were incubated with 200 M photoprobe C-4 either kept in the dark (Ϫh) or irradiated at 254 nm in the absence or presence of Ezetimibe analogues. After solubilization of the membranes, proteins were extracted by streptavidin-biotin chromatography and analyzed by SDS-PAGE and scanning of the intensity of the 145-kDa band. c, 10 samples of rabbit ileal BBMV as described in A were photolabeled with 200 M photoprobe C-4 and the 145-kDa protein was extracted by streptavidin-biotin affinity chromatography followed by preparative SDS-electrophoresis. After SDS-PAGE of the eluted fractions protein bands were excised and submitted to sequence analysis by LC-MS spectrometry after tryptic digestion (Table II). tected by immunoblotting but not with the enzymatic activity of APN. It remains to be elucidated whether proteases different from APN are responsible for the APN-like substrate specificity in the membranes of cell types like stomach or colon, where no APN-protein and no labeling of the 145-kDa Ezetimibe-binding protein can be demonstrated.
Influence of Ezetimibe on the Micellar Cholesterol Uptake and on the Binding of the WM-47 mAB to APN/CD13 in CaCo-2 Cells-Confluent CaCo-2 cells cultivated for a further 11 to 20 days exert partial sensitivity of cholesterol uptake to Ezetimibe. Fig. 11 shows that the uptake of [ 3 H]cholesterol from mixed micelles was partially inhibited by preincubation of the cells with 150 M Ezetimibe or the nonabsorbable Ezetimibe analogue S 6130.
Because of the surprising finding that APN is a specific target protein for Ezetimibe in the brush border membrane of small intestinal enterocytes and because APN is involved in endocytosis of membrane domains (39 -41) and internalization of certain viruses (42)(43)(44)(45)(46)(47), we investigated whether Ezetimibe has an influence on the binding properties of the WM-47 mAB to APN/CD13 on CaCo-2 cells. Staining of confluent CaCo-2 cells with FITC-conjugated WM-47 mAB (specific monoclonal antibody against human CD13) resulted in a bright fluorescence that was more intense at the cell surfaces in a granular pattern (Fig. 12A). In addition, a discrete cytoplasmic fluorescence staining was also visible, whereas nuclei were clearly free of immunofluorescence (Fig. 12, A and C). Incubation of CaCo-2 cells with Ezetimibe prior to staining with WM-47 mAB nearly completely abolished the immunofluorescent staining of the cell surface and only a weak punctual surface staining is left (Fig. 12B). In contrast, however, if the cells were preincubated with WM-47 mAB first, subsequent treatment with Ezetimibe did not significantly change the immunofluorescence staining of APN/CD13 and the staining pattern reflecting the gross cellular distribution of APN was very similar to untreated CaCo-2 cells (Fig. 12D). This finding makes a direct competition between Ezetimibe and the APN antibodies unlikely. On the other hand, Ezetimibe may induce a conformational change on APN/CD13 that the WM-47 mAB cannot recognize its binding site on the Ezetimibe-pretreated APN/CD13. DISCUSSION Whereas the mechanisms and pathways of cholesterol biosynthesis and catabolism to bile acids and steroid hormones are well known, the molecular mechanisms underlying intestinal cholesterol absorption still lie in the dark. The existence of specific cholesterol absorption inhibitors with profound structure-activity relationships (4,5) strongly argues for the involvement of specific proteins in the cholesterol absorption process. The assumption of a single cholesterol transporter (48,49), however, is a too simplified view from a number of observations.
(a) Cholesterol from mixed micelles is transferred to the enterocyte brush border and is moved to detergent-resistant microdomains from which subsequently transport to the endoplasmic reticulum occurs (11).
(b) Cholesterol absorption probably occurs by endocytotic processes as was visualized with the fluorescent cholesterol analogue fluoresterol (50). The scavenger receptor SR-BI is in the fasting state predominantly localized in the microvillar membrane and in apical invaginations between adjacent microvilli as well as in a subapical compartment and small cytoplasmic lipid droplets. Upon exposure of mucosal explants from pigs to a mixture of corn oil, cholesterol, bile, and pancreatin, the number of coated pits between adjacent microvilli strongly increased, being immunoreactive for SR-BI followed by internalization of SR-BI with a huge increase of SR-BI accumulation in cytoplasmic lipid droplets (39).
(c) Cholesterol absorption inhibitors like Ezetimibe and cholesterol or phytosterols bind to different membrane proteins of the enterocyte brush border membrane not showing any mutual interaction and competition to common binding sites (12,13,15). Consequently, a genomic approach with the assumption of a sterol-binding domain will fail to identify the molecular target for cholesterol absorption inhibitors. During preparation of this manuscript Altmann et al. (49) reported the identification of the NPC1L1 as an essential protein being involved in the Ezetimibe-sensitive cholesterol absorption pathway; NPC1L1, however, was neither able to bind Ezetimibe (analogues) nor could cholesterol transport be reconstituted by overexpression of NPC1L1 in non-enterocyte cells (49) making its role as the primary molecular target for cholesterol absorption inhibitors like Ezetimibe unlikely.
When we discovered that Ezetimibe and cholesterol absorption inhibitors specifically bind to a 145-kDa integral membrane protein of the enterocyte brush border membrane, we screened protein data bases for membrane proteins in the molecular range of 130 -160 kDa and considered the Niemann-Pick C1 protein as a putative candidate for the Ezetimibe target. An identity of the Ezetimibe-binding protein with NPC1 and possibly NPC1L1 is unlikely from the following findings: (i) NPC1 is transporting fatty acids and the cationic dye acriflavin (51); upon photoaffinity labeling with Ezetimibe photoprobes we never observed an inhibition of the labeling of the 145-kDa protein by the presence of fatty acids or acriflavin (data not shown). (ii) Photoaffinity labeling of the 145-kDa Ezetimibebinding protein is not inhibited by cholesterol and phytosterols and the 145-kDa Ezetimibe-binding protein is not labeled by photolabile cholesterol (15). In contrast, NPC-1 and NPC1L1 contain a sterol-sensing domain and are labeled by a photoactivable cholesterol analogue if the sterol sensing domain is intact (52). (iii) NPC1L1 is highly glycosylated migrating with an apparent molecular mass of 165 kDa with a shift to 140 kDa upon deglycosylation. In contrast, the Ezetimibe-binding protein from the enterocyte brush border membrane migrates at 145 kDa and shifts to 115 kDa after deglycosylation. (iv) We never obtained sequence fragments belonging to NPC1 or NPC1L1 when analyzing fractions and proteins from brush border membranes covalently cross-linked to Ezetimibe analogues or retained by an Ezetimibe affinity column.
We could identify the luminal side of the enterocyte brush border membrane as the site for pharmacological activity of  8. Localization of APN and the 145-kDa Ezetimibe-binding protein along the gastrocolic axis. a, distribution of the specific enzymatic activities of APN (E), APA (•), dipeptidylpeptidase IV (f), sucrase (Ⅺ), and ␥-glutamyl transferase (ϫ) in small intestinal segments 1-8. Activities are expressed as the ratio of the specific activities in segment X to that in segment 1 (duodenum). b, 5 g of BBMV protein from intestinal segments 1-8 was separated by SDS-PAGE on Novex 4 -12% gels followed by immunoblotting with APN antibodies (APN-3625). c, BBMV (150 g of protein) from small intestinal segments 1, 4, and 8 were photolabeled with 66 nM (0.3 Ci) C-1 followed by SDS-PAGE and determination of radioactivity.
FIG. 9. Photoaffinity labeling of small intestinal segments with biotin-Ezetimibe-photoaffinity probe C-5. BBMV (100 g of protein) of intestinal segments 1-8 were incubated with 9 M C-5 for 60 min at 20°C in the dark followed by irradiation for 30 s at 254 nM. After washing, BBMV proteins were solubilized and proteins covalently modified with the Ezetimibe photoprobe were extracted with streptavidin beads and were eluted after extensive washing of the beads with SDS sample buffer. Equal aliquots of each extract (for in total 8 gels) were used for the various detection methods: a, Coomassie staining; b, streptavidin staining; c, staining with actin antibodies; d, staining with annexin II antibodies. ited by Ezetimibe (analogues). Ezetimibe and its analogues exert their pharmacological activity by specific binding from the intestinal lumen to a 145-kDa membrane protein in the enterocyte brush border membrane. Binding of cholesterol absorption inhibitors to this 145-kDa Ezetimibe-binding protein as determined by differential photoaffinity labeling studies revealed identical structure-activity relationships as for in vivo inhibition of cholesterol absorption (12,13). Independently on the methodology used for purification, sequence analysis consistently revealed to our surprise the identity of the 145-kDa Ezetimibe-binding protein with the membrane-bound metalloprotease APN ((alanyl)aminopeptidase; leukemia antigen CD13). APN is an exopeptidase anchored to the cell membrane by a transmembrane helical region near the N terminus with only a small region (8 -10 amino acids) of the N terminus protruding into the cytoplasm (53). APN is widely distributed with a preferential localization in the brush border membrane of small intestinal enterocytes, kidney, liver, and placental cells as well as in the plasma membrane of monocytes, basophiles, eosinophiles, and neutrophiles (38). Independently from its enzymatic activity, APN shows a number of different functions from insects to mammals: (a) APN is the high affinity receptor for Bacillus thuringiensis Cry1Aa toxin in the midgut of many insect species (54) mediating insecticidal specificity by opening cation selective ion channels causing cell lysis and death of the insect. (b) APN plays a role in blood-brain barrier permeability (55) and can exert signal transduction independently on its catalytic activity (56). (c) A major function of APN is its role in the internalization of various viruses such coronaviruses (42)(43)(44)(45) or cytomegaloviruses (46); APN is crucial for the endocytosis of the coronavirus TGEV (porcine transmissible gastroenteritis virus) mediating receptor-mediated endocy- tosis of TGEV (47). The TGEV virus particle contains numerous APN receptor sites and during binding of TGEV to the cell membrane via APN further APN molecules are recruited leading to a local high density of APN molecules beneath the virus particle. The binding site for TGEV was localized outside the catalytic site to the C-terminal part of APN (47). The formation of the APN-TGEV complexes in the plasma membrane increases the endocytosis rate as indicated by the increase in the number of coated pits and apical vesicles. This clustering of APN molecules is independent on the cytoplasmic N-terminal tail and the virus particles probably cross-link the APN molecules thereby inducing endocytosis similar to antibody-mediated endocytosis. The internalized APN-TGEV complexes are then transported to endosomes where finally the TGEV membrane fusion and penetration occurs dependent on an acidic compartment. Cross-linking of APN molecules evenly distributed on the cell surface of human fibroblasts and labeled with CD13 antibodies with a secondary antibody led to a sequestration of APN to caveolin-1-positive patches (57). During this redistribution, the cross-linked APN molecules showed a linear distribution along longitudinal lines, probably being actin stress fibers prior to sequestration to caveolin 1-positive patches, which suggests that the cross-linking of APN induces binding to actin filaments. Coronovirus 229E, being a specific ligand for APN (42), induced a similar redistribution as induced by the secondary antibody to APN suggesting that the same mechanism is activated by binding to APN. Cholesterol depletion of membranes did not change binding of the viruses to the cell surface but led to an inhibition of sequestration to caveolin 1-positive patches and internalization. The findings that on the one hand various viruses like coronaviruses, SARS, or cytomegaloviruses use APN as the specific receptor for body internalization and that on the other hand APN is the molecular target for cholesterol absorption inhibitors suggest that cholesterol absorption inhibitors like Ezetimibe and Ezetimibe analogues (16 -18) or sterol glycosides may be effective to prevent and treat infections by viruses using APN as receptor.
Changes in the cellular localization of APN between the plasma membrane and intracellular compartments (the deep apical tubules) also occur in enterocytes dependent on their cholesterol availability. Deep apical tubules are a novel surfacecontaining structure in the terminal web region of the brush border functioning as a hub in membrane trafficking of the brush border, thereby connecting the apical surface to the network of intracellular trafficking routes, including the transcytotic pathway (40,41). APN is present in these deep apical tubules, which are assumed to act as an apical membrane reservoir allowing the enterocyte to adjust rapidly to the microvillar length and thus absorption capacity. The finding that the deep apical tubules disappear upon cholesterol depletion by methyl-␤-cyclodextrin may indicate that these compartments act as putative cholesterol reservoirs during lipid absorption (40,41). Because of the involvement of APN in receptor-mediated endocytosis and the dynamics of the deep apical tubule compartment we have investigated the influence of Ezetimibe (i) on cholesterol uptake and (ii) on binding of the WM-47 mAB on CD13 in confluent CaCo-2 cells. The uptake of cholesterol uptake by CaCo-2 cells could partially be inhibited by Ezetimibe and the nonabsorbable Ezetimibe analogue S 6130, indicating that cultivated CaCo-2 cells express to some extent the Ezetimibe-sensitive pathway of cholesterol absorption that is in differentiated small intestinal enterocytes responsible for more than 80% of cholesterol absorption (50). The intense fluorescence staining of CD13 on CaCo-2 cells shown with FITC-conjugated WM-47 mAB was nearly completely diminished when the cells were treated with Ezetimibe prior to APN staining. If Ezetimibe was applied after binding of the WM-47 mAB to APN no changes in the cellular fluorescence staining of APN occurred. Our results that Ezetimibe and nonabsorbable Ezetimibe analogues inhibit intestinal cholesterol absorption by specific binding to APN (and actin) strongly suggest that cholesterol absorption inhibitors interfere with the cellular trafficking of APN between the plasma membrane and intracellular compartments, thereby influencing endocytosis of membrane microdomains and the dynamics of deep apical tubules.