Liver Fatty Acid-binding Protein Initiates Budding of Pre-chylomicron Transport Vesicles from Intestinal Endoplasmic Reticulum*

The rate-limiting step in the transit of absorbed dietary fat across the enterocyte is the generation of the pre-chylomicron transport vesicle (PCTV) from the endoplasmic reticulum (ER). This vesicle does not require coatomer-II (COPII) proteins for budding from the ER membrane and contains vesicle-associated membrane protein 7, found in intestinal ER, which is a unique intracellular location for this SNARE protein. We wished to identify the protein(s) responsible for budding this vesicle from ER membranes in the absence of the requirement for COPII proteins. We chromatographed rat intestinal cytosol on Sephacryl S-100 and found that PCTV budding activity appeared in the low molecular weight fractions. Additional chromatographic steps produced a single major and several minor bands on SDS-PAGE. By tandem mass spectroscopy, the bands contained both liver and intestinal fatty acid-binding proteins (L- and I-FABP) as well as four other proteins. Recombinant proteins for each of the six proteins identified were tested for PCTV budding activity; only L-FABP and I-FABP (23% the activity of L-FABP) were active. The vesicles generated by L-FABP were sealed, contained apolipoproteins B48 and AIV, were of the same size as PCTV on Sepharose CL-6B, and by electron microscopy, excluded calnexin and calreticulin but did not fuse with cis-Golgi nor did L-FABP generate COPII-dependent vesicles. Gene-disrupted L-FABP mouse cytosol had 60% the activity of wild type mouse cytosol. We conclude that L-FABP can select cargo for and bud PCTV from intestinal ER membranes.

During intestinal lipolysis of a fat-containing meal, 2 mol of fatty acid (FA) 4 are produced for every mole of triacylglycerol (TAG) hydrolyzed. These FA and the remaining sn-2-monoacylglycerol are absorbed across the enterocyte apical plasma membrane. Absorption of the FA is thought to occur, at least in part, via specific membrane proteins. FATP4, which is highly expressed at the apical surface of the enterocyte, was shown to increase FA uptake following overexpression, and antisense oligonucleotide knockdown of FATP4 expression reduced FA uptake (1). FABPpm, which was first identified in jejunal microvillus membranes, may also be involved in FA uptake, as anti-FABPpm antibodies have been found to inhibit uptake (2)(3)(4). CD36 has been shown to be important for FA uptake into muscle and adipose tissues (5,6) and may also play a role in intestinal FA transport (7). 5 CD36 is also important for intestinal TAG synthesis and subsequent TAG export from the intestine (8). Kinetic experiments have shown that in addition to protein-mediated transport, a diffusional mechanism for FA uptake by the enterocyte exists as well (9).
Once inside the intestinal absorptive cell, the FA are largely bound to intracellular fatty acid-binding proteins (FABP), and the majority are rapidly converted into TAG (10) for subsequent export from the cell as the major component of intestinal lipoproteins, chylomicrons, and very low density lipoproteins.
The absorptive enterocyte contains high levels of two members of the FABP family, the so-called liver FABP (L-FABP) (11), which is expressed in both intestine and liver, and the intestinespecific FABP (I-FABP) (11). These proteins, first identified over 30 years ago (12,13), are abundantly expressed at 2-4% of enterocyte cytosolic protein levels (14). Although these related proteins both bind FA, L-FABP binds 2 mol of long chain FA and a variety of other lipids including sn-2-monoacylglycerol, lysophospholipids, and bile salts (15)(16)(17), whereas I-FABP appears to be more specific, binding only 1 mol of long chain FA (15). In addition to their distinct substrate specificities, there is other evidence suggesting unique functions for each of these proteins. Both FABPs can transfer long chain FA to phospholipid membranes; however, they appear to utilize distinct mechanisms, I-FABP via collisional interactions with membranes, and L-FABP by aqueous diffusion (18). A role for L-FABP in FA transport has been further suggested by fluorescence photobleaching studies, which indicated that the rate of FA movement within the cell was directly proportional to the level of FABP (19,20). Interestingly, the phenotype of L-FABP gene-disrupted mice was found to be minimally deviant from wild type controls on a chow diet (4% fat); however, fasting resulted in a reduced ability to generate TAG in the liver from the increased FA present in the plasma under fasting conditions, as well as a decreased rate of hepatic fatty acid oxidation (21). The decrease in liver FA oxidation was shown not to be due to a reduced oxidative capacity (22), further suggesting an involvement of L-FABP in intracellular FA transport. In addition to a potential role in intracellular FA trafficking, L-FABP has been suggested to play a role in cell division (23), to serve as a cytosolic reservoir for high levels of unesterified FA (18), and to interact with nuclear hormone receptors, perhaps delivering specific ligands (24). Finally, and of particular relevance to the present studies, it has been reported recently that mice null for L-FABP exhibit impaired intestinal TAG secretion (25).
The ER exit step is the rate-limiting step in the transit of TAG across the intestinal absorptive cell, from its ER synthetic site to its secretion across the basolateral membrane (26,27). For this reason, in previous work we focused our attention on the mechanism by which pre-chylomicrons exit the ER membrane and fuse with the cis-Golgi, a process that is mediated by a sealed vesicle, the pre-chylomicron transport vesicle (PCTV) (28 -30). In contrast to vesicles that transport proteins from ER to Golgi, we found that the coat protein complex-II (COPII) proteins were not necessary for budding of PCTV from the ER but were required for eventual fusion of the formed vesicle with the cis-Golgi (28). This study describes our efforts at identifying the protein(s) that are responsible for budding PCTV. By analogy to the COPII protein complex, we expected that separation of cytosol by size exclusion chromatography would result in the identification of a large molecular weight complex. However, we found that the activity was in the portion of the eluate that contains low molecular weight proteins. As detailed in this study, we have identified L-FABP as a protein required for budding of PCTV.  (22) were backcrossed to C57Bl6/J wild type mice six times prior to use in the present experiments.

EXPERIMENTAL PROCEDURES
Antibodies-Rabbit antibodies to purified rat L-FABP and I-FABP were generated by Affinity Bioreagents (Golden, CO) and used at 1:30,000 and 1:20,000 dilution for immunoblotting, respectively. Rabbit polyclonal anti-apolipoprotein B (apoB) antibodies were a gift of Dr. Larry Swift (Vanderbilt University, Nashville, TN). Rabbit polyclonal anti-apolipoprotein AIV (apo-AIV) antibodies were a gift of Dr. Patrick Tso (University of Cincinnati, Cincinnati, OH). Antibodies to Sar1 and Sec31 have been described previously (28). Antibodies to protein kinase C were obtained from Santa Cruz Biotechnology. Goat polyclonal antibodies to calnexin and calreticulin were purchased from Santa Cruz Biotechnology. Goat anti-rabbit IgG, and goat anti-rabbit IgG conjugated with horseradish peroxidase were purchased from Santa Cruz Biotechnology.
Recombinant Proteins-rI-and rL-FABP were generated in Escherichia coli without tags and were isolated and delipidated using biochemical techniques (31) (28). The cells were homogenized, and the ER was isolated using a sucrose step gradient that was repeated to purify the ER. The [ 3 H]-or [ 14 C]TAG-loaded ER preparation contained no Golgi, endosomes, or lysosomes (30). cis-Golgi was isolated from nonradiolabeled rat enterocytes (32) as was cytosol (28). The cis-Golgi preparation contained GOS28 but not calnexin nor RAB11, indicating no significant ER or lysosomal contamination (28,30). When lack of ATP activity was required, cytosol was dialyzed twice against buffer A (25 mM Hepes, 125 mM KCl, 2.5 mM MgCl 2 , 0.5 mM DTT, and protease inhibitors, pH 7.2); no ATP nor an ATP-regenerating system was added, and the cytosol was supplemented with ATP␥S (2.5 mM) and EDTA (2 mM).
In Vitro PCTV Formation-PCTV containing [ 3 H]TAG were formed from [ 3 H]TAG-loaded intestinal ER (28). In brief, ER was incubated at 35°C for 30 min with cytosol, an ATP-regenerating system, and Mg 2ϩ in the absence of Golgi acceptor. As indicated, when recombinant proteins were utilized, no cytosol was included in the reaction. The incubation mixture was resolved on a continuous sucrose gradient, and PCTV was isolated from the light portions of the gradient (30). PCTVs thus formed were concentrated to 5.0 mg/ml protein using a Centricon-10 filter (Millipore Corp., Bedford, MA).
In Vitro PCTV Fusion with the Golgi-When PCTV fusion with the cis-Golgi was sought, 3 H-protein and [ 14 C]TAGloaded PCTV (150 g of protein) (28), cis-Golgi membranes (300 g of protein), and native cytosol (1 mg of protein) were incubated for 30 min at 35°C with an ATP-regenerating system, 5 mM MgCl 2 , 0.25 M sucrose, 30 mM Hepes, pH 7.2, 30 mM KCl, 5 mM CaCl 2 , and 2 mM DTT with a total volume of 500 l (30). The cis-Golgi were separated from unreacted PCTV by a sucrose step gradient, and the Golgi TAG radioactivity was determined.
Isolation of the PCTV-generating Active Fraction from Cytosol-Preliminary studies showed that the active fraction was cytosolic (data not shown). Purification of the active fraction commenced with rat intestinal cytosol, 80 mg of protein (32), which was dialyzed overnight against buffer A at 4°C. The cytosol was concentrated 5-fold on an Amicon filter (YM-10 membrane) (Amicon, Beverly. MA) and further concentrated on a Centricon filter (10-kDa cutoff) to 20 mg/ml protein. The concentrated cytosol was passed over a heparin-agarose column (Sigma), previously equilibrated with 20 mM NaHPO 4 , 2 mM MgCl 2ϩ , 0.5 mM DTT, and protease inhibitors at pH 7.2. The active fraction (data not shown) was eluted with the equilibration buffer, concentrated to 5 mg of protein/ml (Amicon YM-10 membrane), dialyzed against BisTris buffer, 10 mM, pH 6.9, plus protease inhibitors, and loaded onto an anion exchange column (5-ml Econo column; Bio-Rad). The column was eluted with the equilibration buffer to which 0.1, 0.5, and 1.0 M NaCl in BisTris buffer were sequentially added. Each fraction was collected and evaluated for budding activity. The flowthrough had the maximum activity. It was concentrated and dialyzed against buffer B (25 mM Hepes, 1 mM DTT, 5 mM MgCl 2 , 50 mM KCl, pH 7.4 with protease inhibitors), loaded onto a pre-packed Sephacryl S-100 column (total volume 320 ml, 26 ϫ 60 cm (Pfizer, New York)), and eluted with buffer B. 0.5-ml fractions were collected and checked for activity (Fig.  1B). The active fractions were pooled, concentrated on an YM-10 membrane to 5 mg/ml, dialyzed against buffer C (0.5 mM DTT, Tris-HCl, 20 mM, pH 8.0, and protease inhibitors), and loaded onto an anion exchange column (5 ml) (Econo-pac; Bio-Rad). The column was eluted with buffer C and buffer C to which 0.1 M NaCl was added. The flow-through and 0.1 M NaCl eluates were collected, and budding activity was determined. The flow-through had 25-30 times the budding activity of native cytosol (data not shown). The active fraction was concentrated on an Amicon YM-10 membrane to 5 mg/ml protein. It had one major and several minor bands on SDS-PAGE (silver stain). The components of the active fraction were analyzed by tandem mass spectroscopy. 90% of the proteins were identified as L-FABP.
Mass Spectroscopy-Prior to mass spectroscopy analysis, the active fraction was separated into five fractions using reverse phase-high pressure liquid chromatography on a C-4 column (Thermo Separation Products, Thermo Electron Corp., Waltham, MA) to reduce the complexity of the protein sample and remove salts and buffers. The fractions were freeze-lyophilized, resuspended in 50 mM ammonium bicarbonate, pH 7.8, and digested with trypsin for 14 h at 37°C. The tryptic peptides were desalted by C-18 ZipTip (Millipore) using the manufac-turer's protocol. Tandem mass spectrometry analysis was performed by injecting the tryptic fragments onto a capillary C-18 LC column (our manufacture, New Objective Pico Frit column, 360-m outer diameter, 75-m inner diameter, 15-m tip, 10.4-cm length, and Magic C18AQ packing material, 5-m beads, 200-m pores) on line with a Finnigan LCQDECA (ThermoQuest, San Jose, CA) ion trap mass analyzer equipped with a nano-ESI source. The tryptic peptides were fractionated using 0.1% formic acid in water as solvent A and 90% methanol as solvent B. The sequence-specific data obtained were used in the SEQUEST search engine against a nonredundant data base (NCBInr) to identify proteins.

Collection of [ 3 H]Chylomicrons and VLDL-Chylomicrons
were obtained from rats infused intraduodenally with [ 3 H]glyceryltrioleate (10). In brief, chow fed rats were provided with a duodenal cannula and a mesenteric lymph duct fistula, placed in a restraining cage, and infused overnight with a glucose/ saline solution. The next day, the rats were infused with [ 3 H]glyceryltrioleate (2, 22.5, or 135 mol/h as indicated) through the duodenal cannula. Hourly collections of the lymph were made on ice and chylomicrons from 4 to 6 h of infusion (steady state chylomicron output) collected by centrifugation through saline (33). The chylomicrons were combined and used for column chromatography. VLDL were isolated by serial centrifugations (34).
SDS-PAGE and Immunoblots-Proteins, separated by SDS-PAGE, were transblotted onto nitrocellulose membranes (Bio-Rad) (28). After incubation with specific primary and then secondary antibodies, labeled proteins were detected by using ECL reagents and exposing the developed blots to Biomax film (Eastman Kodak Co.).
Measurement of TAG Radioactivity-TAG radioactivity was determined by liquid scintillation (32), and mass was quantitated chemically after specific TAG extraction using organic solvents (35).
Statistical Analysis-Comparisons between means were carried out using a statistical package supplied by GraphPad Software (Instat, GraphPad Software, Inc., San Diego) using a twotailed t test.

RESULTS
In considering potential proteins associated with PCTV budding activity from intestinal ER, we first identified as a paradigm the COPII protein complex (Ϸ576 kDa), which is required for protein vesicle budding. In line with this postulate, we thought that passage of intestinal cytosol over a sizing column would most efficiently lead to identification of the expected protein complex. However, when 30 mg of native cytosol protein was chromatographed on a Sephacryl S-100 column, the results showed that most of the PCTV-generating activity appeared just before the elution of cytochrome c (Ϸ12 kDa) (Fig. 1A), contrary to our expectations, indicating that either a small molecular weight protein or a complex of very small proteins provided PCTV budding activity. To improve the resolution of the column and to verify the results obtained in Fig. 1A, we passed native cytosol over a heparin affinity column and next over a DEAE column eluted with buffer, pH 6.9, with all the activity being found in the flow-through of both columns (data not shown). Fig. 1B shows the post-DEAE column flowthrough chromatographed on an S-100 sizing column. As indicated, budding activity was again found to emerge from the column just before cytochrome c. Control experiments showed that no significant budding occurred at 4°C, in the absence of ATP, or in the absence of cytosol (Fig. 1C).
The active fractions from the S-100 column were concentrated and the proteins separated by SDS-PAGE. A silver stain of the transblotted proteins showed one major band at 14 kDa and several minor bands (data not shown). Because a large (Ϸ8%) component of native intestinal cytosol is composed of I-and L-FABP (Ϸ14 kDa) and the active fraction contained predominantly a small molecular weight protein, we immunoblotted the active fraction for both I-and L-FABP using monospecific antibodies. Both L-and I-FABP were immuno-identified in approximately equal amounts (data not shown). However, when the concentrated active fractions from the S-100 column were passed over a DEAE column and eluted with phosphate-buffered saline, pH 8.0, the activity appeared in the flow-through which, on immunoblot, identified L-FABP but not I-FABP and on silver stain contained a major band at Ϸ14 kDa and several minor bands of slightly larger molecular weight (data not shown). Tandem mass spectroscopy was performed on the active DEAE, pH 8, flow-through fraction, to identify the protein(s) in the active fraction potentially accounting for the budding activity. The identified proteins were as follows: L-and I-FABP, acyl-CoA-binding protein, glutathione S-transferase, profilin A, and cyclophilin A. Recombinant proteins for each identified protein were obtained and used in an assay that measures the ability of proteins to generate PCTV or other TAG-containing structures from intestinal ER in the absence of acceptor cis-Golgi (28). Of all six proteins tested, L-FABP had the greatest activity, I-FABP had 23% the activity of L-FABP, and the rest had negligible activity (Fig. 2). The amount of budding activity displayed by 40 g of rL-FABP approximated that of 1 mg of native cytosol protein (Fig. 2). This level of L-FABP is comparable with that expected in 1 mg of rat intestinal cytosolic protein (14).
To determine the relationship between intestinal ER budding activity and L-FABP, increasing amounts of L-FABP were incubated with [ 3 H]TAG-loaded intestinal ER. The amount of budding activity was proportional to the amount of L-FABP added up to 150 g of L-FABP after which activity did not increase (Fig. 3A). To confirm that these [ 3 H]TAG data in response to L-FABP represented a true increase in chylomicron output as the amount of L-FABP was increased, immunoblots for apoB48 were performed on PCTV generated from each amount of L-FABP added (Fig. 3B). The apoB48 signal increased as the L-FABP increased up to 150 g and then plateaued in parallel with the [ 3 H]TAG data shown in Fig. 3A. The finding that activity reaches a maximum suggests that the ability of L-FABP to generate TAG-containing structures from ER membranes is limited and supports the thesis that the action of L-FABP on intestinal ER is associated with a specific set of reac- A, native cytosol (80 mg of protein) was concentrated to 20 mg/ml protein and passed over a Sephacryl S-100 column. 5-ml fractions were assayed for protein and budding activity. Molecular weight markers shown above the curves were cytochrome c (12 kDa), Sar1 (25 kDa, identified by immunoblot), protein kinase C (76 kDa, identified by immunoblot), Sec31 (150 kDa, identified by immunoblot), and V 0 (dextran blue). B, concentrated native cytosol was passed over a heparinagarose column. The active flow-through fraction was loaded onto an anion exchange column eluted at pH 6.9 using a stepwise NaCl gradient (see under "Experimental Procedures"). The active flow-through fraction was concentrated and passed over a Sephacryl S-100 column. 5-ml fractions were assayed for protein and budding activity. C, PCTV budding activity under various incubation conditions. Bar 1, PCTV budding activity (see "Experimental Procedures") using native cytosol incubated at 35°C; bar 2, native cytosol incubated at 4°C; bar 3, cytosol treated to be devoid of ATP activity; and bar 4, incubation at 35°C in the absence of cytosol. The percentage of ER [ 3 H]TAG incorporated into PCTV under conditions favorable for budding activity is 12-15% (28).
tions that enable budding to occur and is not the result of a nonspecific interaction of L-FABP with the ER. This conclusion is supported by the greatly reduced or virtually absent budding activity of the other five recombinant proteins that were assayed, each in the same amount as L-FABP, 40 g (Fig. 2).
The data thus far presented only suggest that L-FABP releases [ 3 H]TAG from intestinal ER membranes. We sought evidence that the released TAG was in enclosed, intact vesicles of a size consistent with PCTV, i.e. large enough to contain chylomicrons, and that the proposed vesicles contained the apolipoproteins identified with chylomicrons, apoB48 and apoAIV (28). Although the TAG-containing structures appeared to be vesicles by electron microscopy (Fig. 4B) and comparable in shape to vesicles generated using native cytosol (Fig. 4A), their size was variable. In order to adequately compare the size of these structures to PCTV generated from native cytosol, we chromatographed both over a Sepharose CL-6B sizing column (Fig. 5). PCTV generated using native cytosol (Fig.  5, NC-PCTV) eluted in a homogeneous peak in fractions 17-19, whereas PCTV generated using L-FABP (Fig. 5, L-FABP-PCTV) appeared in fractions 13-15 suggesting that they were modestly larger. To compare the elution pattern of these vesicles to that of rat lymph chylomicrons, we isolated chylomicrons from rats (36) infused with glyceryltrioleate at either 135 or 22 mol/h. Chylomicrons produced under both these conditions eluted earlier in the chromatogram as a homogeneous peak near the void volume of the column (Fig. 5, Chylomicrons). By contrast, when the input rate of glyceryltrioleate into the rat was reduced to 2 mol/h, the chylomicrons produced varied greatly in size and appeared in fractions 2-20 ( Fig. 5, Chylomicrons LD). Importantly, this elution volume encompassed that of PCTV generated from either native cytosol or L-FABP. The reduction in apparent size of these chylomicrons as a result of the lower TAG intraduodenal input rate is consistent with the data of Hayashi et al. (37) who found that as the intraduodenal input rate of lipid into rats was reduced, the size of the resulting chylomicrons was also reduced. Isolated rat VLDL, whose diameter is 66 nm (38), were run over the column and eluted in two peaks. The first and smaller peak was consistent with contaminating chylomicrons and eluted early in the chromatogram. The majority of the radiolabeled TAG (66%) eluted much later in fractions 40 -41. Thus the TAG-containing structures generated from ER membranes by either native cytosol or L-FABP were significantly greater in size than 66 nm VLDL. One explanation for the reduced apparent size of these structures compared with the larger chylomicrons could be that the ER from which they came could not be loaded with the same amount of TAG as ER from rats infused intraduodenally with TAG because the primary culture of enterocytes necessitates low concentrations of FA compared with normal digestive conditions (39).   We sought additional evidence that the structures generated by rL-FABP were compatible with PCTV. Characteristically, transport vesicles concentrate cargo and exclude nonspecific proteins; specifically PCTV concentrate apoB48 and apoAIV by comparison to intestinal ER and exclude the ER proteins calnexin and calreticulin (28). These characteristics are consistent with the TAG-containing structures generated by L-FABP, which also concentrate apoB48 and apoAIV and exclude calnexin and calreticulin (Fig. 6). Another characteristic of transport vesicles is that they are sealed as evidenced by their cargo resisting external protease digestion. Proteinase K treatment of the TAG-rich structures budded from ER membranes by L-FABP demonstrated an absence of protein degradation, consistent with this criterion (Fig. 7). As shown, the cargo protein, apoB48, resisted Proteinase K digestion both in PCTV formed in native cytosol and the TAG-rich structures formed by L-FABP. By contrast, if Triton X-100 disrupted the vesicle membranes, then in both cases the apoB48 signal was extinguished (Fig. 7). In sum, the data presented thus far show that the structures generated by L-FABP produce a sealed vesicle containing TAG, concentrate the chylomicron-associated apolipoproteins B48 and AIV, exclude ER luminal proteins, and are of a size similar to PCTV. In toto, these data are consistent with the thesis that the TAG-rich vesicle produced by rL-FABP is comparable with the PCTV produced by native cytosol.
To further examine the involvement of L-FABP in PCTV budding from ER, we evaluated cytosol from the intestine of L-FABP null mice, compared with cytosol from wild type mice of the same genetic background. Using [ 3 H]TAG-loaded rat ER, preliminary studies showed that C57Bl6/J mouse cytosol was active in budding PCTV from rat ER albeit at an Ϸ40% reduced rate compared with native rat cytosol (data not shown). Cytosols from L-FABP Ϫ/Ϫ and L-FABP ϩ/ϩ mice were tested for budding activity from rat ER pre-loaded with [ 3 H]TAG. The budding activity of L-FABP Ϫ/Ϫ mice was significantly (40%) reduced compared with wild type mice (Fig. 8). The residual activity in the gene-disrupted mice may be due in part to I-FABP, which has some budding activity (Fig. 2), or perhaps other compensatory mechanisms. Although I-FABP expression in intestine is not increased in L-FABP Ϫ/Ϫ mice (data not shown), large amounts of I-FABP are present in intestinal cytosol (14). At present, we do not know if the observed decrease in TG output in PCTV from LFABP-KO cytosol is   because of a decrease in the size of the PCTV or a decrease in the number of PCTV generated. Current studies are addressing this question. Budding activity of the L-FABP null mice was entirely restored when 40 g of r-LFABP was added (Fig. 8). Furthermore, when L-FABP was added to wild type mouse cytosol, a significant (20%) increase in budding activity was also found.
Although L-FABP, as shown, can bud vesicles that contain chylomicrons, we considered the possibility that it might also participate in budding vesicles that contain newly synthesized proteins from intestinal ER. This potential has been given recent support in yeast, where it was shown that an unknown small molecular weight protein is required for ER budding activity to be fully expressed (40). However, as shown in Fig. 9, we found that L-FABP did not produce vesicles carrying newly synthesized protein from the ER, in contrast to its robust ability to bud TAG-carrying vesicles.
Because L-FABP can select cargo and bud chylomicron-containing vesicles from intestinal ER membranes, we next asked if L-FABP-budded vesicles were functionally capable of fusing with intestinal cis-Golgi (30). [ 3 H]TAG-loaded PCTV generated from native cytosol were incubated with intestinal cis-Golgi, ATP, and Mg 2ϩ at 35°C for 30 min. As expected, the results showed (30) that considerable disintegrations/min of [ 3 H]TAG became isodense with Golgi postincubation (Fig. 10, Native cytosol). In marked contrast, L-FABP generated PCTV, similarly incubated with cis-Golgi, ATP, and Mg 2ϩ , either with or without native cytosol, did not result in [ 3 H]TAG delivery to the Golgi (Fig. 10, rL-FABP, rL-FABP ϩ Cytosol). These results indicate that L-FABP-generated PCTV cannot fuse with Golgi membranes. One possibility for the absence of Golgi fusion for the L-FABP ϩ cytosol condition could be the lack of the COPII proteins on PCTV that are required for fusion of PCTV with the Golgi (28), a potential we tested in the next series of experiments.
Because we previously found that vesicles generated from intestinal ER using cytosol depleted of Sar1 were also incapable of fusing with Golgi (28), we hypothesized that the L-FABPgenerated vesicles, which were formed in a cytosol-free medium, would likely contain low levels of the cytosolic COPII protein, Sar1, on their surface. Indeed, in contrast to PCTV generated in native cytosol (28), Sar1 immunoblotting demonstrated no detectable Sar1 in L-FABP-generated PCTV incubated without cytosol (Fig. 11, rL-FABP-Buffer) despite the presence of Sar1 on intestinal ER membranes (Fig. 11, ER) (28,30). Interestingly, even when the rL-FABP-generated PCTV were subsequently incubated with native cytosol, no Sar1 loading occurred (Fig. 11, L-FABP-Cytosol) despite the presence of Sar1 in native cytosol (Fig. 11, Cytosol). This is the expected result because Sar1 vesicle loading requires the GDP-GTP exchange protein, Sec12, located on the ER membrane.
It is possible that L-FABP or a binding partner needs to be present at the inception of PCTV for PCTV fusion with the Golgi to occur because L-FABP is present on PCTV (data not   shown). We could not answer this question directly by using PCTV produced using cytosol from L-FABP KO mice because of the reduced amount of PCTV produced by these mice and the amount of protein required for our assays. To approximate these conditions, however, we generated PCTV using native cytosol, isolated the PCTV, and then washed them free of L-FABP (as judged by immunoblot, data not shown) and possible attached proteins using 2 M KCl. These washed PCTV were then incubated with L-FABP KO cytosol and Golgi in our fusion assay ("Experimental Procedures"). The results were compared with similarly washed PCTV but incubated with native cytosol and Golgi. When the PCTV were allowed to fuse with the Golgi using L-FABP KO cytosol, 986 Ϯ 104 [ 3 H] TAG dpm were delivered to the Golgi as compared with 1044 Ϯ 117 [ 3 H] TAG dpm when the washed PCTV were incubated with native cytosol and Golgi (p Ͼ 0.05, n ϭ 4). These data suggest that L-FABP or a binding partner do not have to be present on PCTV for fusion with the Golgi to occur.
Because L-FABP-generated vesicles do not fuse with the cis-Golgi (Fig. 10), we wondered if in addition to the absence of Sar1 there was an absence of the v-SNARE, VAMP7. VAMP7 is present on PCTV formed in native cytosol (30) and is a component of the PCTV-Golgi SNARE complex (30). To address this question, we incubated L-FABP-generated PCTV with cytosol (which does not contain the integral membrane protein, VAMP7) and found that these PCTV also concentrate VAMP7 from intestinal ER (Fig. 11, PCTV-Cytosol). Therefore, the lack of fusion of these vesicles with cis-Golgi cannot be explained by the absence of VAMP7. These data also support the thesis that the vesicle formed by L-FABP is similar to that formed in native cytosol with the exception that it does not contain Sar1 and cannot fuse with the cis-Golgi.
To address the possibility that L-and I-FABP, which both bind long chain FA, might interact in the process of budding PCTV, we compared the effects of co-incubation of the two proteins on PCTV generation. As shown in Fig. 12A, increasing amounts of I-and L-FABP, up to 80 g each, resulted in a proportional increase in the amount of budding activity. Activity was clearly greater for L-FABP than I-FABP, consistent with prior results (Fig. 2). When both rI-and rL-FABP were added together equally so that in toto they equaled the micrograms listed on the x axis, the curve generated was very close to the predicted curve using the data points for each FABP assayed separately at the appropriate concentration (Fig. 12A). These data indicate that there is no interaction between the two rFABPs. Next we considered the possibility that because I-FABP had less budding activity than L-FABP, it could interfere with the budding activity of L-FABP present in native cytosol. To examine this question, we added progressively more rI-FABP to a set amount (1 mg of protein) of rat cytosol and found (Fig. 12B) that the budding activity of the resulting mixture was a straight line whose increase in activity corresponded to the amount of rI-FABP added. No indication of inhibition by I-FABP was found, suggesting that I-FABP does not interfere with the budding activity of native cytosol.
Because we have found that I-FABP delivers and receives its bound FA to model membrane surfaces by collision, and L-FABP by diffusion (18,41), and because I-FABP is expressed only in the intestine, we questioned why L-FABP was a more FIGURE 11. Immunoblots for Sar1 and VAMP7 in cytosol, ER, and PCTV. The proteins from cytosol, ER, or PCTV generated with recombinant L-FABP, which had been incubated with native cytosol for 30 min at 35°C (PCTV, Cytosol), or buffer (PCTV, Buffer) were separated by SDS-PAGE, and the proteins were transblotted to a nitrocellulose membrane and probed for Sar1 or VAMP7, as indicated. PCTV incubated in buffer were not probed for VAMP7. Detection was by ECL. potent initiator of budding than I-FABP. One potential explanation for our findings is that L-FABP but not I-FABP binds to intestinal ER membranes on incubation with native cytosol under conditions in which budding is expected (ATP, Mg 2ϩ , cytosol, and ER at 35°C for 30 min). Immunoblot analysis showed that a small amount of L-FABP is present on the ER prior to incubation, consistent with our prior observations (28). L-FABP binding greatly increases on the ER membrane on exposure to cytosol (Fig. 13, L-FABP). By contrast, although there is ample I-FABP in intestinal cytosol, none is detected on ER membranes either before or after incubation with cytosol ( Fig. 13, I-FABP). These data suggest that under appropriate conditions, L-FABP interacts with the ER membrane, potentially initiating a budding reaction.

DISCUSSION
Intracellular transport vesicles concentrate cargo for export to another intracellular organelle or membrane for processing, storage, or secretion. Proteins on their surface are required for correct targeting and fusion of the vesicles with the appropriate membrane. For newly synthesized proteins exiting the ER, a well defined system has been described that utilizes COPII proteins to bud vesicles from ER membranes whose destination is the cis-Golgi. The initiator of these events is GTP-activated Sar1 (42). Perhaps because of size constraints imposed by the COPII proteins, the diameter of vesicles carrying newly synthesized proteins is 60 -80 nm (43). However, cargoes different from newly synthesized proteins can engender different budding mechanisms. In particular, pre-chylomicrons in PCTV have been shown to bud from intestinal ER membranes in the absence of COPII proteins, although the COPII proteins are required for subsequent fusion of the PCTV with the cis-Golgi (28). The ability of the ER to bud PCTV despite the lack of COPII proteins may be due in part to the much larger size required by these transport vesicles, Ϸ250 nm, to enable them to enclose TAG-rich pre-chylomicrons. In addition to the lack of requirement of COPII machinery for budding, VAMP7, a v-SNARE usually localized to a post-Golgi compartment, in the intestine is uniquely localized to the ER and is concentrated on PCTV (30). This novel mechanism suggests that a protein(s), not considered previously to be part of the vesicle-forming machinery, could be the initiating agent(s) in generating PCTV at the ER surface. This alternative system of vesicle formation may be useful under conditions in which vesicle formation is tied to the intermittent dietary intake of lipid and not to the constant requirement for movement of newly synthesized proteins to the Golgi.
By contrast with the COPII protein complex (Ϸ576 kDa), which mediates the budding of protein-containing vesicles from the ER, we were surprised to find that a small molecular weight protein such as L-FABP could, apparently by itself, initiate budding. That a small molecular weight protein was the only budding factor in intestinal cytosol is suggested by the lack of activity elsewhere in the elution pattern of whole cytosol on size exclusion chromatography (Fig. 1A) as compared with the COPII components, Sec23/24 and Sec13/31, which have a much larger Stokes radius when similarly chromatographed (44). Although we cannot rule out a larger active complex of proteins that dissociates during the preparatory steps for generating cytosol, L-FABP alone effected the same degree of budding as found with intact cytosol. These data, using L-FABP at concentrations comparable with those found in normal cytosol, are strongly suggestive that the protein is sufficient for formation of PCTV from the ER.
The present in vitro studies are strongly supported by recently reported in vivo studies. Newberry et al. (25) found that L-FABP gene-disrupted mice fed a high fat diet retained more TAG in their intestines, relative to WT mice. Furthermore, L-FABP KO mice given the lipoprotein lipase inhibitor, tyloxapol, had a reduction in their serum TAG on being fed a lipid bolus, suggesting a reduced chylomicron output from the intestine as compared with normal mice. These results are entirely consistent with what would be predicted if L-FABP-initiated chylomicron exit from the ER were impaired. Thus, the present findings are likely to be of physiological significance for intestinal chylomicron assembly/secretion. L-FABP has not been considered previously as a protein that initiates vesicle budding; however, several known properties of L-FABP are consistent with the novel function proposed herein. For example, the correlation of L-FABP expression with lipid absorption has long been appreciated. Intestinal L-FABP expression is highest in the proximal intestine (45), and in both rats and humans it has been localized to the cells of the villus tips, the site of maximal lipid absorption (45,46). Furthermore, it has been shown in mice that lipid loading causes an increase in the expression of L-FABP (47), and in Caco2 cells, both Land I-FABP are responsive to incubation with FA (48). Although little has been reported about the effects of L-FABP gene disruption in the intestine, L-FABP null mice grow normally, suggesting normal intestinal absorption. Of note, however, mouse chow has only Ϸ4% fat (w/w) so that a potential intestinal defect in lipid absorption is not phenotypically expressed in these mice. There is also an effect of L-FABP ablation in liver. On fasting, these mice accumulate TAG and have a reduced uptake of FA (21) suggesting a role for L-FABP in hepatic lipid metabolism.
Using in vitro model systems at physiologic ionic strength and pH, we found that L-FABP transfers fatty acids to and from acceptor phospholipid membranes by a diffusion-based mechanism that does not involve L-FABP-membrane interaction  2 and 6) for 30 min at 35°C with an ATP-regenerating system. Intestinal ER preincubation (lanes 3 and 7) and intestinal cytosol (lanes 4 and 8) were also evaluated. The proteins from ER pre-or postincubation or cytosol were separated by SDS-PAGE, transblotted to a nitrocellulose membrane, and probed with anti-L-FABP (L-FABP) or anti-I-FABP (I-FABP) antibodies as indicated below the lane numbers. Detection was by ECL. (18,41). However, we postulated that in vivo, L-FABP could be interacting with specific membrane proteins and/or lipid domains, thereby acting not only as a cytosolic reservoir for long chain FA but also perhaps as a regulator of intracellular lipid trafficking as well (18). L-FABP binding to membranes was shown to occur at very low ionic strength, presumably because of an absence of surface charge shielding (18,49). In the present results, we find that L-FABP is detectable in freshly isolated intestinal ER fractions and that incubation of ER with intestinal cytosol results in a large increase in bound L-FABP but not I-FABP. Interestingly, although I-FABP interactions with membranes have been demonstrated in vitro (50,51), and subcellular fractionation of jejunal mucosa shows a small amount of I-FABP to remain membrane-associated (52), the specific conditions that promote PCTV generation do not appear to support the association of I-FABP with membranes.
Although it seems remarkable that a small molecular weight (Ϸ14 kDa) protein such as L-FABP is able to cause ER membrane deformation and fission, a similar property has been reported for the Ϸ21-kDa protein Sar1p. Recently, Lee et al. (53) reported that Sar1p could initiate cargo sorting, membrane deformation, and tubulation but not fission in a liposomal system. The membrane perturbation is postulated to occur by the insertion of the N-terminal ␣-helix of Sar1p into the liposome. In the GDP-bound state, the Sar1p ␣-helix is enclosed in a hydrophobic pocket that is altered on GTP exchange by Sec12, making the ␣-helix available for membrane insertion. An examination of the tertiary structure of L-FABP shows that it contains two short ␣-helical segments (54), and in vitro studies suggest that this domain is responsible for the FA transport properties of the protein (55). It is thus possible that protein-ER membrane/ER protein interactions can occur for L-FABP, potentially involving the helix domain. It is also possible that L-FABP causes membrane deformation via ligand delivery to the membrane. L-FABP is known to bind lysophospholipids (56), which have a so-called "inverted cone" shape, and hence to promote nonbilayer phases to form in membranes (57). Thus, perhaps L-FABP becomes bound to intestinal ER membranes by protein-protein interaction and delivers its lysophospholipid ligand to the ER membrane, resulting in sufficient deformation to initiate bud formation.
As judged by its TAG, apoB48, and apoAIV cargo, the vesicle formed using L-FABP contains chylomicrons. Moreover, the vesicle is sealed, as evidenced by the interior contents being resistant to proteinase K treatment. This is consistent with our prior data showing that COPII proteins are not required for budding PCTV (28). However, like PCTV formed in the absence of COPII proteins (28), the L-FABP-generated vesicle is not fully functional in that it is not able to fuse with the cis-Golgi. Our current data support the lack of a role for COPII proteins in the selection of pre-chylomicron cargo for inclusion in PCTV, another proposed function of both Sar1 (58) and Sec24 (59), in the formation of protein vesicles. The COPII proteins may, however, participate in the selection of other cargo such as p58 for PCTV (28). The lack of recruitment of p58 for PCTV in the absence of COPII proteins may be important for Golgi fusion because p58 recruits COPI proteins, required for fusion of ER vesicles with cis-Golgi (60).
It is not likely that L-FABP could be releasing PCTV in a nonspecific manner. In particular, when more than 150 g of L-FABP was added to ER membranes, there was no further increase in PCTV formation, indicating that the mechanism for PCTV generation initiated by L-FABP becomes saturated. Moreover, the amount of L-FABP used in these studies (typically 40 g) is in the physiologic range reported previously for this abundant enterocyte protein (14). Furthermore, when recombinant proteins other than L-FABP of near the same molecular weight were added, either none (acyl CoA-binding protein, etc.) or considerably less (I-FABP) PCTV budding activity was observed. Finally, PCTV generated by L-FABP, like those generated using native cytosol, not only concentrate their cargo lipoproteins, apoB48 and apoAIV, from their parent ER but also concentrate VAMP7, the unique v-SNARE for ER to Golgi transport employed by PCTV for fusion with the Golgi (30). These results are consistent with a specific action of L-FABP on intestinal ER membranes to generate a cargo-containing vesicle.
The amount of L-FABP used in these studies, 40 g of protein, is consistent with the ratio of L-FABP in cytosol to the amount of ER present in vivo in rat enterocytes. Normal cytosol contains 23 g of L-FABP per mg of cytosol protein (14), but only 18% of total enterocyte protein is present in the ER (61). This should be compared with the 33% of the incubation mixture used here represented by ER protein (500 g of ER protein, 1 mg of cytosol protein). Using the percentage of ER protein under the in vivo conditions compared with our in vitro conditions, the amount of L-FABP expected in the cytosol would be 42 g. These data may at least explain in part the equivalence in budding activity using 40 g of L-FABP versus 1 mg of native cytosol protein.
It is not yet known whether L-FABP requires bound FA in order to effectively generate PCTV. Although we used delipidated protein for exogenous addition to the ER fractions, it is still possible that ER-derived FA became bound to the added protein. Nevertheless, the fact that other FA-binding proteins, including I-FABP (Fig. 2) and albumin (data not shown), did not substitute for L-FABP, implies that the FA-binding properties of the protein may not be necessary for its function in PCTV budding.
In sum, we describe a novel physiological function for L-FABP, a small molecular weight protein that is abundantly expressed in both intestine and liver. Utilizing L-FABP and intestinal ER membranes, we have been able to generate a sealed vesicle in which the pre-chylomicron TAG and associated apolipoproteins have been selected and nonspecific ER luminal and membrane proteins excluded. This proposed role for L-FABP is the first physiological function to be directly demonstrated for this abundant intestinal cytosolic protein.
The vesicle generated by L-FABP is not able to fuse with intestinal cis-Golgi. We postulate that this is because of the inability of the vesicle to attract Sar1 to its surface and thus the remainder of the COPII proteins, leaving it lacking specific targeting information. It is likely that an as yet unidentified additional protein is required for the PCTV to acquire Sar1, making it fusion-competent with the Golgi.