The lipidation by hepatocytes of human apolipoprotein A-I occurs by both ABCA1-dependent and -independent pathways.

The pathways of hepatic intra- and peri-cellular lipidation of apolipoprotein A-I (apoA-I) were studied by infecting primary mouse hepatocytes from either apoA-I-deficient or ABCA1-deficient mice with a recombinant adenovirus expressing the human apoA-I (hapoA-I) cDNA (endo apoA-I) or incubating the hepatocytes with exogenously added hapoA-I (exo apoA-I) and examining the hapoA-I-containing lipoproteins formed. The cells, maintained in serum-free medium, were labeled with [(3)H]choline, and the cell medium was separated by fast protein liquid chromatography or immunoprecipitated to quantify labeled choline phospholipids specifically associated with hapoA-I. With the apoA-I-deficient hepatocytes, the high density lipoprotein fraction formed with endo apoA-I contained proportionally more phospholipids than that formed with exo apoA-I. However, the lipoprotein size and electrophoretic mobility and phospholipid profiles were similar for exo apoA-I and endo apoA-I. Taken together, these data demonstrate that a significant proportion of hapoA-I is secreted from hepatocytes in a phospholipidated state but that hapoA-I is also phospholipidated peri-cellularly. With primary hepatocytes from ABCA1-deficient mice, the expression and net secretion of adenoviral-generated endogenous apoA-I was unchanged compared with control mice, but (3)H-phospholipids associated with endo apoA-I and exo apoA-I decreased by 63 and 25%, respectively. The lipoprotein size and electrophoretic migration and their phospholipid profiles remained unchanged. In conclusion, we demonstrated that intracellular and peri-cellular lipidation of apoA-I represent distinct and additive pathways that may be regulated independently. Hepatocyte expression of ABCA1 is central to the lipidation of newly synthesized apoA-I but also contributes to the lipidation of exogenous apoA-I. However, a significant basal level of phospholipidation occurs in the absence of ABCA1.

The hepatic and intestinal origins of the major high density lipoprotein (HDL) 1 apolipoproteins, apolipoprotein (apo)A-I and apoA-II, are well defined (1,2). In contrast, HDL lipid constituents have complex and multiple origins that include secretion as nascent lipoproteins containing apoA-I (3,4), acquisition of lipids from remnant lipoproteins arising from lipolysis of triglyceride-rich lipoproteins (5)(6)(7), and from cellular lipid efflux (8). The relative contributions of the different pathways are not well understood, particularly the secretion of nascent lipoproteins and the contribution of the efflux pathway. The ATP-binding cassette transporter, ABCA1, was recently shown to control the efflux of cellular phospholipids and cholesterol (9 -12) and through this pathway to maintain HDL in the circulation. Impairment of ABCA1, as in Tangier disease, leads to extremely low levels of HDL (13)(14)(15)(16). The major tissues affected in this disease are rich in macrophages, which express high levels of ABCA1 (17)(18)(19). This was first interpreted as evidence that excess lipids accumulated in scavenger receptor-expressing cells were a major source of HDL lipids (20,21), but recent evidence shows that macrophage contribution to HDL-cholesterol concentrations is minor (22). ABCA1 is expressed in many tissues and at high levels in liver, brain, and small intestine, but also testis, lung, spleen, and kidney (17)(18)(19). This suggests that in both liver and intestine, where apoA-I synthesis is also high, ABCA1 may contribute to the lipidation of newly secreted or nascent lipoproteins. Previously, work with hepatocytes from chicken (23,24) or rat (25) had suggested that apoA-I was lipidated intracellularly. However, Hamilton et al. (26), using electron microscopy, failed to identify any lipidated apoA-I particles in hepatocytes, putting the intracellular lipidation hypothesis in dispute. Recently, Chisholm et al. (27) investigated the secretion and lipidation of apoA-I from HepG2 cells. They concluded that some apoA-I acquired lipid intracellularly and was then secreted along with lipid-poor apoA-I. Subsequently, the secreted apoA-I could acquire lipids extracellularly to form buoyant HDL particles. In those studies, which support the model of intracellular lipidation of apoA-I, HDL particles were obtained by carbonate extraction of cell homogenates. Despite careful quantitation and inclusion of controls, mixing of cell contents and artificial lipidation of apoA-I may occur.
Here we have characterized the hepatic lipidation of apoA-I by using adenoviral expression of human apoA-I (hapoA-I) (28,29) in primary hepatocytes of apoA-I-deficient and ABCA1deficient mice. We also characterized the nascent lipoproteins formed by primary hepatocytes cultured in lipoprotein-free medium compared with those formed by interaction of exogenous apoA-I with the same cells. In both conditions hepatocytes generate a lipidated pool of apoA-I-containing lipoproteins via a pathway dependent on ABCA1. However, the lipidation of apoA-I is reduced but not abolished in experiments with ABCA1-deficient hepatocytes, suggesting the existence of alternate lipidation pathways.
Cell Labeling-Six h following the initial plating, the cells were washed in William's medium without fetal bovine serum (2 ϫ 2 ml) and incubated with Hepatozyme medium (Invitrogen) containing 10 Ci/ well of [ 3 H]choline (PerkinElmer Life Sciences). The following day (24 h) the labeled medium was removed and the cells were infected for 1 h with either the recombinant hapoA-I encoding Ad5 adenovirus (AdAI) or luciferase adenovirus (AdLuc) at a multiplicity of infection of 75:1 plaque-forming units per cell in William's medium without fetal bovine serum (28,29). After the 1 h infection, the hepatocytes were incubated for an additional 24 h with fresh labeling medium as described above. The third day (18 -24 h after adenovirus infection), following 2 ϫ 2 ml washes in non-radioactive medium, the cells were incubated with unlabeled Hepatozyme medium (1 ml per well) in the absence or presence of 5 g of hapoA-I. The cells were returned to the 37°C incubator (5% CO 2 ) for 3.5 h, and the medium was subsequently collected and spun down to pellet any cell debris. The medium with newly secreted apoA-I (AdAI-infected cells) or with exogenously added hapoA-I (AdLuc-infected cells) was analyzed as described below. In some experiments, 10 M 9-cis-retinoic acid, a retinoid X receptor (RXR) ligand, was added to the hepatocytes 12 h prior to and during the 3.5 h incubation. Glyburide (100 M), a known inhibitor of ABCA1-mediated lipid efflux, was added only during the 3.5 h incubation.
Distribution of Secreted ApoA-I in the Various Lipoprotein Pools-The medium from four 6-well plates (24 wells) were pooled and concentrated down to 2 ml (12-fold concentrated with Amicon 10K filter units). The samples were immediately loaded on two calibrated Superdex 200 columns connected in a series similar to that described previously for isolation of lipoproteins from plasma samples (28). Very low density lipoprotein (VLDL)-and low density lipoprotein (LDL)-sized species elute in the void volume on these columns. HDL 2/3 particles and smaller very high density lipoprotein (VHDL) fractions containing albumin (Յ 7.1 nm diameter) were carefully separated. Aliquots (200 l) from each fraction were analyzed for apoA-I by Western blot analysis following transfer to nitrocellulose with a slot blot apparatus (BioRad Bio-Dot SF unit) as described previously (28). No background signal could be detected as indicated by analysis of medium collected from hepatocytes infected with the AdLuc. The relative distribution of apoA-I in the VLDL, HDL 2/3 , and VHDL pools was determined by densitometric scanning (BioRad software, Quantity One, version 4.11). For comparison, the relative distribution of murine apoB (apoB48 and apoB100) was also measured using a polyclonal anti-mouse apoB antibody (BIODESIGN International, Kennebunk, ME) and visualized by chemiluminescence (Pierce West Pico SuperSignal substrate, Pierce) after incubation with horseradish peroxidase-conjugated anti-rabbit IgG (Amersham Biosciences).
The Heterogeneity and Charge of ApoA-I Secreted from Primary Hepatocytes-The charge and size of apoA-I secreted from the primary hepatocytes was determined by agarose gel (Beckman Lipogel, Beckman Coulter, Fullerton, CA) and 4 -20% non-denaturing polyacryl-amide gradient gel electrophoresis (PAGGE) (Novex, Invitrogen), respectively, as described previously (28,29). Briefly, following transfer of proteins from the gels to nitrocellulose, the membranes were probed with biotinylated monoclonal antibodies directed against human apoA-I (a combination of 4H1 (against the extreme N terminus) and 5F6 (against the central region)) (33). The antibodies were biotinylated with Sulfo-NHS-Biotin (Pierce) and visualized by chemiluminescence following treatment with Streptavidin-conjugated horseradish peroxidase (Amersham Biosciences). The size of the apoA-I species were compared with biotinylated molecular weight markers of known hydrodynamic diameter, and the charge of secreted apoA-I was compared with lipidfree apoA-I and HDL both isolated from human plasma.
Immunoprecipitation of ApoA-I and Associated Choline-containing Phospholipids-ApoA-I secreted from hepatocytes was immunoprecipitated under native conditions either directly from the medium or from lipoprotein fractions isolated by FPLC as follows. The immunoprecipitations were carried out with a polyclonal anti-human apoA-I antiserum from sheep (Roche Molecular Biochemicals) and protein G-Sepharose (Amersham Biosciences). An equal volume of an antihuman apoB antiserum from sheep, which does not cross-react with murine apoB, was used as a control where indicated. The immunoprecipitates were collected following centrifugation (10 min at 3000 ϫ g) and washed three times with 10 ml of phosphate-buffered saline (no detergents) and resuspended in a final volume of 1 ml of phosphatebuffered saline. These immunoprecipitates were either subjected directly to scintillation counting or were further analyzed by Bligh and Dyer lipid extraction (34) and thin layer chromatography (TLC). TLC separation was performed on silica gel plates and a solvent system (chloroform/methanol/acetic acid/formic acid/water, 70:30:12:4:2) for separation of phosphatidylcholine and sphingomyelin. The TLC bands corresponding to phosphatidylcholine and sphingomyelin were excised and counted for radioactivity. Alternatively, cells were labeled with 32 P-phosphate (Amersham Biosciences) to label all cellular phospholipids. Cells were treated as with labeling with [ 3 H]choline, except that 32 P-phosphate in Hepatozyme medium was only added after adenoviral infection on the second day (not also on the first day as for [ 3 H]choline). On the third day, the hepatocytes were washed as before and incubated in fresh Hepatozyme in the absence or presence of apoA-I for 3.5 h. The hapoA-I-containing lipoproteins were immunoprecipitated as described above, and then phospholipids were extracted by the method of Bligh and Dyer (34). The phospholipids were separated by TLC in the solvent system of chloroform, methanol, acetic acid, formic acid, water (at a volume ratio of 70:30:12:4:2). The TLC plate was exposed to a phosphorimaging plate and the relative amounts of phospholipids were determined by densitometry scanning (BioRad software, Quantity One, version 4.11). Results are expressed as the average of at least three replicates to control for variable loading and extraction efficiency.

RESULTS
Lipidation of a newly synthesized apoA-I can occur both intracellularly during transport from the endoplasmic reticulum to the cell surface and peri-cellularly through lipid efflux. To document the relative contribution to the lipidation observed, we compared endogenously synthesized apoA-I and exogenously added apoA-I. Primary hepatocytes were isolated from 4 -6-month-old mice by liver collagenase perfusion, and isolated cells were cultured on fibronectin-coated plates in the presence of serum-free media (Hepatozyme). The following day, cells were infected with either an adenoviral construct encoding human apoA-I (AdAI) or luciferase (AdLuc) for 1 h, washed, and then returned to Hepatozyme media. The next day, the cells were washed and then incubated with fresh Hepatozyme media in the absence or presence of exogenous hapoA-I for 3.5 h. This time period was chosen to allow sufficient secretion of hapoA-I for analysis, and yet minimize peri-cellular interactions. We chose a concentration of exo apoA-I that approximated the amount of hapoA-I secreted during the same time period. The hapoA-I-containing lipoproteins in the media were then analyzed by a number of methods. The adenoviral vector was selected specifically to ensure apoA-I synthesis and secretion independent of experimental factors.
Electrophoretic Migration of hapoA-I-containing Lipoproteins-The electrophoretic migration on agarose gels of hapoA-I newly secreted from primary hepatocytes (hereafter referred to as endogenously synthesized apoA-I or "endo apoA-I") or exogenously added hapoA-I (referred to as "exo apoA-I") was assessed ( Fig. 1). Endo apoA-I and exo apoA-I from apoA-I-deficient mice were both found to have exclusively pre-␤ migration, and no ␣-migrating immunoreactive apoA-I band appeared even with prolonged exposure. This is in contrast with a previous study in monkey hepatocytes (4), where apoA-I-containing lipoproteins were found to segregate into two pre-␤-and one ␣-migrating fractions. Similarly, lipoproteins formed by hepatocytes from ABCA1Ϫ/Ϫ mice endo apoA-I or exo apoA-I ( Fig. 1) possessed only pre-␤ migration. This result demonstrates that all HDL formed, whatever the source, have similar pre-␤ electrophoretic mobility, and thereby lack a significant hydrophobic core.
ApoA-I Is Found in Different Lipoprotein Pools-To evaluate the lipoprotein size distribution of apoA-I secreted by the hepatocytes expressing hapoA-I, the medium was concentrated and immediately fractionated by FPLC and analyzed. The distribution of immunoreactive human apoA-I and murine apoB (both apoB100 and apoB48) in the FPLC fractions were analyzed by slot blot. Immunoreactive hapoA-I segregated into 3 well separated peaks ( Fig. 2A). The largest apoA-I-containing lipoproteins eluted at a position previously calibrated for VLDL (fractions 10 -14). This fraction also overlapped with the largest peak of murine apoB-containing lipoproteins (fractions 9 -13; data not shown). The second peak of immunoreactive apoA-Icontaining lipoproteins eluted at the position of HDL 2/3 (fractions 19 -23) and the third corresponded to lipid-poor VHDL and apoA-I (fractions [25][26][27][28][29]. The results presented are representative of three separate experiments. The distribution of immunoreactive apoA-I after FPLC separation of medium lipoproteins shows that endo apoA-I and exo apoA-I form lipoproteins of similar sizes ranging from VLDL/ LDL to VHDL (see Fig. 2, A and B, and the distribution obtained with control hepatocytes in Fig. 3, A and B). Furthermore, when exo apoA-I was added to the medium of hepatocytes infected with AdAI, the amount of label associated with apoA-I was additive (data not shown). This result clearly indicates that lipidation occurs both peri-cellularly and intracellularly.
ABCA1ϩ/ϩ control and ABCA1Ϫ/Ϫ mouse hepatocytes were also infected with AdAI or AdLuc and then analyzed by FPLC for size separation of the endo apoA-I-and exo apoA-I-containing lipoproteins. Comparing the FPLC profiles of hapoA-I-containing lipoproteins from control and ABCA1-deficient hepatocytes, a reduction can be seen in the proportion of hapoA-I found in the buoyant VLDL and HDL fractions for both endo apoA-I (Fig. 3A) and exo apoA-I (Fig. 3B). These results demonstrate the important contribution of ABCA1 to both intracellular and peri-cellular lipidation of apoA-I.

ApoA-I in the HDL 2/3 Pool Is Heterogeneous in Size-
The different apoA-I-containing lipoprotein populations generated by the primary hepatocytes and separated by FPLC (VLDL, HDL, and VHDL) were further analyzed by non-denaturing PAGGE and Western blot analysis (Fig. 4). Similar amounts of immunoreactive hapoA-I were loaded from each lipoprotein pool. The same lipoprotein pool for different hepatocyte samples was similarly concentrated and loaded. The endo apoA-I present in VLDL (lane 1), HDL 2/3 (lane 2), and VHDL fractions (lane 3) are well separated from one another. The HDL 2/3 and lipid-poor apoA-I yield distinct bands, which are compatible with the known formation of lipoproteins with varying numbers of apoA-I and with varying degrees of lipidation. A large amount of hapoA-I is secreted as HDL 2/3 -sized species ( Fig. 2A), with a significant size heterogeneity, which in this pool can reach 10.4 nm (Fig. 4, lane 2). The three lipoprotein fractions from the media of AdLuc-infected hepatocytes incubated with exo apoA-I ( Interestingly, in comparison to endo apoA-I, lipidation of exo apoA-I produced profiles of similarly as well as differently sized hapoA-I-containing lipoproteins (comparing lanes 2 and 5 and 3 and 6). This suggests differences in how endo apoA-I and exo apoA-I lipoproteins are speciated and lipidated; it also indicates that our experiments distinguish between lipidation associated with secretion and efflux. Furthermore, when hepatocytes infected with AdLuc and incubated with exo apoA-I were incubated with 9-cis-retinoic acid, a retinoid X receptor ligand, the resulting hapoA-I-containing lipoproteins, VLDL (lane 7), HDL (lane 8), and VHDL (lane 9), were similar to control exo apoA-I fractions. An increased amount of larger-sized HDL particles is evident, suggesting that 9-cis-retinoic acid can enhance lipidation of exogenous hapoA-I. The 9-cis-retinoic acid effect was not observed with endo apoA-I (data not shown).
Non-denaturing PAGGE and Western blot analysis were performed on the hapoA-I-containing lipoprotein fractions generated by ABCA1ϩ/ϩ control and ABCA1Ϫ/Ϫ hepatocytes ( We know that the quantity of hapoA-I found in VLDL and HDL fractions is significantly reduced in ABCA1Ϫ/Ϫ hepatocytes (Fig. 3, A and B), but, importantly, the nature of the lipoprotein particles formed is unchanged.
Distribution of apoA-I and Phospholipids in the Lipoprotein Fractions Separated by FPLC-From the apoA-I-deficient hepatocytes, the calculated relative distribution of endo apoA-I in the different lipoprotein fractions is shown in Fig. 6A. Interestingly, ϳ20% of the total endo apoA-I secreted was found in HDL 2/3 -sized fractions. As well, a smaller but significant percentage of secreted apoA-I was also found associated with the VLDL pool. This result is in good general agreement with previous results in monkey hepatocytes (4) and in HepG2 cells, although the latter do not secrete VLDL and therefore have no apoA-I-containing lipoproteins in this lipoprotein size (27,35).
The association of [ 3 H]choline phospholipids with hapoA-I in the three-lipoprotein pools was estimated by immunoprecipitation of hapoA-I under native conditions. Equal volumes of the pooled FPLC lipoprotein fractions (identified as VLDL, HDL 2/3 , and VHDL in Fig. 2) were immunoprecipitated with an anti-hapoA-I antibody raised in sheep. An anti-hapoB antibody also raised in sheep, which does not cross-react with murine apoB, was used and subtracted as nonspecific binding. The results show that although the majority of secreted hapoA-I is in the lipid-poor fraction (Fig. 2), a significant amount of the phospholipid associated with apoA-I (16.8%) is in the HDL 2/3 lipoprotein pool (Fig. 6B). Therefore, this demonstrates that apoA-I can be secreted with significant quantities of phospholipid, which is consistent with the size and heterogeneity of hapoA-I in the HDL 2/3 pool as determined by 4 -20% nondenaturing PAGGE (Fig. 4).
Lipoproteins formed by lipidation of exo apoA-I were analyzed in the same manner. Fractions corresponding to VLDL, HDL, and VHDL were pooled and immunoprecipitated with antibodies against hapoA-I (Fig. 6C) and the radioactivity associated with apoA-I measured as described above (Fig. 6D). A smaller proportion of apoA-I (6.7%) was found in the HDL fraction for exo apoA-I compared with endo apoA-I. A significant amount (11.9%) of the [ 3 H]choline-labeled phospholipids were associated with the HDL 2/3 pool. Therefore, this demonstrates that apoA-I can acquire significant quantities of phospholipid peri-cellularly, which is consistent with the size and heterogeneity of hapoA-I in the HDL 2/3 pool as determined by 4 -20% non-denaturing PAGGE (Fig. 4).
Levels of hapoA-I and [ 3 H]choline phospholipids found in the different lipoprotein fractions of ABCA1ϩ/ϩ control and ABCA1Ϫ/Ϫ hepatocytes were also quantified (Fig. 7). There was a significant decrease (82%; p Ͻ 0.05) in endo apoA-I associated with the HDL in ABCA1Ϫ/Ϫ hepatocytes compared with control hepatocytes (Fig. 7A). A decrease in HDL-associated 3 H-phospholipids was also demonstrated, although to a lesser extent (35%, p Ͻ 0.10; Fig. 7B). There was also a signif-icant decrease (63%, p Ͻ 0.05; data not shown) in total hapoA-I-associated 3 H-phospholipids released from the ABCA1Ϫ/Ϫ hepatocytes compared with control hepatocytes. There was a slight reduction of apoA-I and 3 H-phospholipid found in the VLDL fraction of ABCA1Ϫ/Ϫ hepatocytes compared with control hepatocytes, which did not reach statistical significance. For exo apoA-I, there was a decrease (65%, p Ͻ 0.10) in association of exo apoA-I with the HDL fraction for ABCA1Ϫ/Ϫ hepatocytes compared with control hepatocytes (Fig. 7C), with a smaller decrease (42%, p Ͻ 0.07; Fig. 7D) in HDL-associated 3 H-phospholipids. Total hapoA-I-associated 3 H-phospholipids released were also decreased (25%, p Ͻ 0.10; data not shown). Importantly, the total amount of hapoA-I secreted by AdAIinfected control and ABCA1Ϫ/Ϫ hepatocytes was the same, demonstrating that ABCA1 deficiency did not impair secretion of apoA-I (ABCA1 (ϩ/ϩ) control ϭ 1.33 Ϯ 0.22 g/h/well; ABCA1 (Ϫ/Ϫ) ϭ 1.33 Ϯ 0.11 g/h/well).
[ 3 H]choline-labeled phospholipids were extracted, separated by TLC, and quantified. For both endo apoA-I and exo apoA-I, over 90% of the [ 3 H]choline label was found in phosphatidylcholine species with the remainder in sphingomyelin species (data not shown). For a more complete evaluation of all phospholipid species, hepatocytes were incubated with 32 P-phosphate to label all phospholipids, as described under "Experimental Procedures." 32 P-labeled phospholipids sphingomyelin (SPM), phosphatidylcholine (PC), phosphatidylinositol (PI), phosphatidylserine (PS), and phosphatidylethanolamine (PE) were separated by TLC and quantified by densitometry scanning. PC constituted between 85-90% of the total phospholipid, whereas SPM, PE, PI, and PS contributed the additional 10 -15%. There were no significant differences between the ratio of the minor phospholipid species associated with endo apoA-I and exo apoA-I for apoA-I (Ϫ/Ϫ),  taining Lipoproteins-Hepatocytes infected with AdAI or Ad-Luc were labeled with [ 3 H]choline and release of 3 H-phospholipids into the medium after a 3.5 h incubation were measured. Hepatocytes were incubated with 9-cis-retinoic acid and glyburide to alter ABCA1 activity and evaluate the role of ABCA1 in the lipidation of endo apoA-I or exo apoA-I. The result shown is the mean (Ϯ S.D.) of four experiments, each performed in triplicate. In AdAI-infected hepatocytes, 9-cis-retinoic acid did not significantly alter the amount of 3 H-phospholipid (Fig. 8) associated with hapoA-I, but glyburide treatment resulted in a modest but consistent decrease in the lipidation of newly secreted apoA-I. In general, neither 9-cis-retinoic acid nor glyburide showed a remarkable effect on lipidation of endo apoA-I. AdLuc-infected cells were incubated for the 3.5-h period with 50, 15, 5, or 2.5 g/well of exogenously added hapoA-I and released 3 H-phospholipids were measured. Varying amounts of exo apoA-I were added to investigate the effect of increasing amounts of hapoA-I protein on lipidation. Addition of 9-cisretinoic acid resulted in a significant increase in lipidation of hapoA-I, whereas glyburide treatment resulted in a more modest but significant decrease in hapoA-I-associated lipidation (Fig. 8). Increasing amounts of exo apoA-I increased the total labeled phospholipid associated with hapoA-I, but did not change the ratio of VLDL/HDL/VHDL-associated lipids, nor the effects of 9-cis-retinoic acid and glyburide (data not shown). Results shown are the mean (Ϯ S.D.) of four independent experiments at 5 g of exo apoA-I/well and are typical of results with varying concentrations of exo apoA-I. DISCUSSION The goals of the present study were to characterize the contribution of secretion and efflux pathways to the lipidation of apoA-I by hepatocytes and determine the role of hepatocyte ABCA1 in these pathways. ABCA1 deficiency did not affect the net secretion of apoA-I expressed by the adenoviral construct. Therefore the concentrations of apoA-I, whether endogenously expressed or exogenously added, allow the direct comparison of the relative distribution of apoA-I and phospholipids in the different lipoprotein fractions. Overall, ABCA1 deficiency significantly reduced the proportion of apoA-I found in the HDL fraction, reduced the proportion of lipid associated with the HDL fraction, and significantly reduced the total 3 H-phospholipids released from hepatocytes, but not the amount of apoA-I secreted, compared with control hepatocytes. We found that HDL formed by endogenously synthesized apoA-I was more profoundly reduced by ABCA1 deficiency than that formed by exogenously added apoA-I. Therefore, ABCA1 contributes to lipidation of apoA-I not only by the well documented extracellular efflux mechanism (9 -12), but also, and more significantly, to the lipidation of newly secreted apoA-I. ABCA1 has been shown to be present in intracellular compartments, specifically early endosome and late endosome compartments (36). Our results suggest that intracellular ABCA1 may contribute to the lipidation of newly synthesized apoA-I, although current evidence only correlates lipid efflux with ABCA1 activity at the cell surface (36). In the liver, ABCA1 contributes significantly to lipidation of hepatic apoA-I, although we do not know exactly the physiological contribution of liver ABCA1 to circulating plasma HDL levels. Macrophages have been shown to contribute only a minor proportion of HDL (22) and it is likely that liver ABCA1 may provide a larger proportion. There are also tissue-specific transcripts of ABCA1 that may provide alternative regulation of ABCA1 in different tissues (18,19,22,37). Although there has been much speculation and indirect evidence for a major contribution of hepatic ABCA1 to circulating levels of HDL, the present study provides direct evidence for the role of ABCA1 in the liver.
Our study also highlighted significant differences between lipidation of apoA-I by secretion and efflux mechanisms. Comparison of HDL and VHDL fractions formed by exo apoA-I and endo apoA-I to each other in apoA-I-deficient (Fig. 4) and ABCAI-deficient hepatocytes (Fig. 5) showed distinct size profiles, as secretion and efflux pathways generated similar but non-identical HDL and VHDL species. Agarose gel electrophoresis and phospholipid analysis demonstrated that the charge and composition of HDL species formed by endo apoA-I and exo apoA-I were unchanged, despite the difference in the distribution profile. Thus, HDL formed by endogenous apoA-I synthesis and secretion and by exogenous apoA-I and efflux mechanisms are subtly different, and may involve different pathways and proteins necessary to lipidate apoA-I. This assertion is reinforced by the observation that exo apoA-I was sensitive to stimulation by 9-cis-retinoic acid (Figs. 4 and 8), whereas endo apoA-I was relatively insensitive. 9-cis-retinoic acid is a ligand for the RXR, and RXR can heterodimerize with peroxisome proliferator-activated receptor (PPAR)-␣, -␥, and -␦, liver X receptor-␣ and -␤, and farnesoid X receptor (and others) to regulated many genes involved in lipid metabolism. Liver X receptor (38 -40), PPAR␣ (41), PPAR␥ (41,42), and PPAR␦ (43) in complex with RXR have been documented to increase ABCA1 expression in macrophages but not hepatocytes. There are many targets of RXR regulation that may affect apoA-I lipidation (44). We do not wish to comment on the mode of RXR activation, but rather use 9-cis-retinoic acid as a tool to differentiate between lipidation of endo apoA-I and exo apoA-I. Our results show that ABCA1 deficiency most strongly affects lipidation of endo apoA-I (Fig. 7), whereas 9-cis-retinoic acid strongly affects only lipidation of exo apoA-I. One possible explanation may be that lipidation of endo apoA-I is maximal in the wild-type hepatocytes, i.e. limited by lipid availability (Fig. 7A), but lipidation of exo apoA-I is not (Fig. 7C), which allows the stimulatory effect of 9-cis-retinoic acid to be observed. There exist other possibilities, and we are currently investigating the mechanism of 9-cis-retinoic acid stimulation.
Glyburide has been documented to be an inhibitor of ATPbinding cassette transporters, including ABCA1, and can inhibit lipid efflux effectively in fibroblasts, endothelial cells and macrophages (11,12,45,46). Glyburide was surprisingly ineffective in hepatocytes.
Plasma HDL and apoA-I levels are reduced to practically undetectable levels in Tangier disease due to the hypercatabolism of poorly lipidated apoA-I. It has been documented that apoA-I secretion is not affected in Tangier disease subjects and in an animal model of Tangier disease (Wisconsin Hypo-Alpha Mutant (WHAM) chicken) (47)(48)(49)(50)(51)(52). In our model system, we also observed no decrease of adenoviral-mediated apoA-I expression in the ABCA1-deficient hepatocytes compared with control hepatocytes. Instead, ABCA1 deficiency significantly reduced the amount of HDL formed by impairing the lipidation of apoA-I. However, there is still a significant basal level of lipidation of apoA-I in the absence of ABCA1, for both endo apoA-I and exo apoA-I. If our model system is an accurate depiction of hepatic apoA-I lipidation, it is unclear if this degree of impairment of lipidation by the isolated hepatocyte would be sufficient to generate an HDL/apoA-I profile of Tangier disease. Possibly, defective HDL lipidation in other cells or within the circulation could contribute to the phenotype. The source of the residual lipidation remains unknown. It is possible that other ABC transporters could be playing a minor role in hepatocytes. A retroendocytotic mechanism may also be involved (53). We have discovered a novel apoA-I-binding site on the extracellular matrix of macrophages (54) and hepatocytes 2 involved in ABCA1-mediated efflux, but, as yet, we have not attributed any specific physiological role to this binding site. Hepatocytes also secrete apoE, but the mechanism of lipidation of newly secreted apoE remains unclear with evidence for and against an ABCA1-independent mechanism (55)(56)(57)(58). In any case, our study demonstrates the existence of an ABCA1-independent lipidation pathway.
The absence of ␣-migrating HDL assessed by agarose gel electrophoresis, reflects the absence of a significant hydrophobic core. Our results 3 show that free cholesterol associates with apoA-I-containing lipoproteins and lecithin:cholesterol acetyltransferase is known to be secreted by the hepatocytes (59,60). The question is raised as to why ␣-migrating HDL is not formed. Even extended incubations of apoA-I with hepatocytes (24 h, data not shown) were not sufficient to generate a hydrophobic core. Another factor necessary for HDL maturation may not be present on hepatocytes but possibly on other cell types (i.e. peripheral cells).
Hepatocytes (primary and transformed) and enterocytes are the physiological models used to study apoA-I secretion. Thrift et al. (61) studied lipoprotein secretion from transformed HepG2 (hepatocyte) cells into serum-free medium. Similar to the results found in this study, analysis of concentrated HepG2 cell medium by PAGGE revealed a broad immunoreactive-apoA-I band between 7.1-12.2 nm with the majority of apoA-I in the lipid-free or lipid-poor form (Ͻ 8 nm). However, HepG2 cells do not secrete normal VLDL-sized particles, and not surprisingly no apoA-I was detected in this lipoprotein pool (61). In our study, between 3-8% of total apoA-I and 0.8 -2.4% of total 3 H-phospholipids associated with the VLDL fraction. ABCA1 deficiency did reduce, but not significantly, endo apoA-I and phospholipid associated with VLDL, but not exo apoA-I. We intend to evaluate further the nature of apoA-I association with VLDL secreted from hepatocytes from control and ABCA1deficient mice. We will also analyze the role of ABCA1 in the secretion of VLDL and its associated apolipoproteins, in view of the increase in VLDL noted in patients with Tangier disease. Cynomolgus monkey hepatocytes in culture also secreted nascent apoA-I particles (4). Unlike the results presented here (Fig. 4), very little heterogeneity in secreted apoA-I was observed, even after 3 d in culture. The results from that study are, however, difficult to interpret due to the extended period of time that secreted apoA-I was in the medium and perhaps due to the lack of sensitivity in the assay required for detection of minor subpopulations of apoA-I. Similar to the role of microsomal triglyceride transfer protein in apoB secretion, hepatocytes and enterocytes might also express proteins that facilitate the lipidation of newly secreted apoA-I, such as ABCA1. Therefore, it is difficult to interpret studies from 3T3 cells (62), polarized Madin-Darby canine kidney cells (63), and mouse C127 cells (64), which have each been transfected with apoA-I cDNAs and found to secrete lipid-poor apoA-I-containing particles. Furthermore, most of these studies isolated apoA-I by ultracentrifugation rather than non-denaturing techniques that do not dissociate apolipoproteins. A recent study by Chisholm et al. (27) examined the lipidation state of newly secreted apoA-I from HepG2 cells. The authors concluded that ϳ20% of newly secreted apoA-I is lipidated intracellularly and another 30% is minimally lipidated shortly thereafter extracellularly. In our study, the most physiologically relevant cell model, primary hepatocyte, was used to address nascent HDL secretion, and care was taken to separate the secreted lipoproteins by nondenaturing techniques, such as FPLC on calibrated Superdex 200 columns and immunoprecipitation. With this approach, a significant amount (ϳ20% of total) of hapoA-I is found secreted as mature-sized HDL particles with pre-␤ migration, in accordance with the work of Chisholm et al. (27).
The discovery of a major role for ABCA1 in phospholipid and cholesterol efflux to nascent HDL or lipid-poor apoA-I has established a renewed interest in the processes by which apoA-I is lipidated. Most reports have studied ABCA1 in peripheral cells (i.e. fibroblasts and macrophages) and shown that this transporter is responsible for efflux of cholesterol and phospholipid to apoA-I. Neufeld et al. (65) demonstrated that an expressed ABCA1-GFP fusion protein was localized to the basolateral surface of WIF-B cells, a polarized hepatocyte cell line, and stimulated apoA-I-mediated cholesterol efflux. Overexpression of ABCA1 in transgenic mice led to increased circulating HDL levels, although expression was not targeted specifically to the liver (66). A major role of hepatocyte ABCA1 in apoA-I lipidation agrees with the reports of ABCA1 being most abundant in the liver (17)(18)(19). Recently, Basso et al. (67) demonstrated that injection of a liver-targeted adenoviral ABCA1 construct into C57Bl/6 mice resulted in increased apoA-I-me-diated cholesterol efflux and HDL-cholesterol concentration. Our study is the first to evaluate the role of hepatocyte ABCA1 in the process of apoA-I lipidation.
In conclusion, we demonstrated that intracellular and pericellular lipidation of apoA-I represent distinct and additive pathways that may be regulated independently. Hepatocyte expression of ABCA1 is central to the lipidation of newly synthesized apoA-I, but an ABCA1-independent pathway may also contribute to the lipidation of apoA-I. Our future work will focus on the lipidation of apoA-I by cholesterol and the potential regulation of lipidation of apoA-I by hepatic cholesterol levels.