Identification of a Form of Acyl-CoA:Cholesterol Acyltransferase Specific to Liver and Intestine in Nonhuman Primates*

The present study demonstrates that two different forms of the intracellular cholesterol esterification enzyme acyl-CoA:cholesterol acyltransferase (ACAT) are present in the nonhuman primate hepatocyte; one is similar to that originally cloned from human genomic DNA, here termed ACAT1, while a second gene product, termed ACAT2, is reported here. The primate ACAT2 gene product was cloned from an African green monkey liver cDNA library. Sequence analysis of an isolated, full-length clone of ACAT2 cDNA identified an open reading frame encoding a 526-amino acid protein with essentially no sequence similarity to the ACAT1 cDNA over the N-terminal 101 amino acids but with 57% identity predicted over the remaining 425 amino acids. Transfection of the cloned ACAT2 cDNA into two different mammalian cell types resulted in the production of abundant ACAT activity which was sensitive to ACAT inhibitors. Northern blot analysis showed that the ACAT2 mRNA was expressed primarily in liver and intestine in monkeys. In contrast, ACAT1 mRNA was expressed in almost all tissues examined. Topologic predictions from the amino acid sequence of ACAT2 indicates that it has seven trans-membrane domains in a configuration that places the putative active site of the enzyme in the lumen of the endoplasmic reticulum. This orientation of ACAT2 in the endoplasmic reticulum membrane, in addition to its expression only in liver and intestine, suggests that this enzyme may have as a primary function, the secretion of cholesteryl esters into apoB-containing lipoproteins.

The intracellular formation of cholesteryl esters catalyzed by the action of the enzyme acyl-CoA:cholesterol acyltransferase (ACAT; EC 2.3.1.26) 1 appears to be nearly ubiqitous in mammalian cells (1). Elucidation of the details of the structure and catalytic mechanism of ACAT and of the regulation of its activity have been stymied by the difficulty in isolating and purifying an active form of this membrane-associated enzyme. It has taken the isolation of a cDNA for ACAT from human genomic DNA, accomplished through functional complementation of mutant Chinese hamster ovary cells lacking ACAT activity, to initiate progress in understanding the biochemistry of ACAT function (2). The mRNA for this ACAT is expressed in most human tissues and cDNAs with nearly identical ACAT sequences have likewise been found in a variety of tissues from mouse, hamster, and rabbit (3)(4)(5).
Several functions can be attributed to cholesterol esterification by ACAT. The enzyme appears to modulate the potentially toxic effects of cholesterol in cell membranes. By attaching a fatty acid to the free hydroxyl group of cholesterol, physical properties of the cholesterol molecule are changed and the solubility of esterified cholesterol in the lipids of the cell membrane is limited. Cholesteryl esters accumulate in lipid droplets in the cytoplasm, and maintenance of a balance between the free and esterified forms of cholesterol in a cell is believed to be a component of regulation of cholesterol signaling pathways (6).
Evidence is also accumulating that ACAT activity is important for cholesterol transport in lipoproteins by the liver and intestine. The majority of cholesterol absorbed from the intestinal lumen by the mucosal cell is esterified by ACAT (7) and incorporated into chylomicron particles (8). Furthermore, the secretion of apoB-containing lipoproteins by the liver appears to depend to an extent on ACAT-catalyzed cholesterol esterification (9 -11). Formation of cholesteryl esters in the primate liver is promoted by oleic acid-enriched diets (12) and is catalyzed by ACAT (13). Recent work from our laboratory has shown that hepatic secretion of cholesteryl esters in apoBcontaining lipoproteins, as monitored during isolated liver perfusion, was strongly associated with the extent of coronary artery atherosclerosis in nonhuman primates fed atherogenic diets, (12), and was sensitive to inhibition by ACAT inhibitors (11). Interestingly, the extent of diet-induced cholesteryl ester accumulation in the liver, while correlated to the rate of hepatic cholesteryl ester secretion, was not as highly correlated to the extent of coronary artery atherosclerosis as was the secretion rate (12). One possible explanation was that the hepatic ACAT enzyme that catalyzed the formation of the cholesteryl esters of lipoproteins was different from the enzyme that catalyzed the formation of cholesteryl esters of intracellular lipid droplets. In any case, an important role for hepatic ACAT in the development of coronary artery atherosclerosis was indicated.
Conceptually, the presence of various forms of ACAT differentially distributed in tissues and conceivably present in separate cellular compartments is consistent with the increasing evidence for multiple roles of sterol esterification in cholesterol metabolism. The possibility that more than one ACAT enzyme is present in at least some tissues of mammalian species was highlighted by the mouse knockout experiment showing that disruption of the ACAT gene analog to the human ACAT gene cloned by Chang et al. (2), did not alter the pattern of intestinal cholesterol absorption or hepatic cholesteryl ester accumulation when cholesterol was fed (14). Furthermore, the work of Sturley and colleagues (15) and Yu et al. (16) demonstrated two different ACAT analogs in yeast, suggesting that two or more ACAT enzymes in mammalian species would not be unprecedented. Additional work in the Sturley (17) and Farese laboratories (18) identifying at least two ACAT-related gene products (ARGP) in the EST data bases for the human and mouse genomes supports the possibility that more than one ACAT enzyme exists.
In the work reported here, the complete nucleotide sequence has been determined for the cDNA of a second ACAT (ACAT2) isoform in primates which varies substantially in structure from the initially described human ACAT (2). ACAT2 is present in an organ-specific distribution (only liver and intestine) that differs from the nearly ubiquitous expression pattern seen for the original human ACAT, here termed ACAT1. The specific tissue location and topology prediction of this new ACAT2 isolate appears to adapt it to the particular function of providing cholesteryl esters for lipoprotein secretion from hepatocytes and intestinal mucosa. In this way ACAT2 may complement the ACAT1 analog, the latter being found widely distributed among tissues in the monkey apparently serving more general cell functions.

EXPERIMENTAL PROCEDURES
Materials-For earlier experimental protocols (19), organs from adult male African green monkeys (Cercopithecus aethiops) were collected at the time of necropsy, cut into 1-cm cubes, and snap frozen in liquid nitrogen. Samples were stored at Ϫ80°C until use. Tissues from two adult male cynomolgus monkeys (Macaca fascicularis) fed monkey chow were also collected at necropsy. All experiments with monkeys had been subjected to prior approval by our institutional animal care and use committee. AC29 cells, a Chinese hamster ovary cell line lacking ACAT activity (20) were provided by Dr. T. Y. Chang and were maintained in Ham's F-12 medium supplemented with 2 mM L-glutamine, 1% Eagle's vitamins, 100 IU/ml penicillin, 100 g/ml streptomycin, and 10% (v/v) fetal bovine serum. COS-1 cells were obtained from the American Type Tissue Collection and grown in Dulbecco's modified Eagle's medium containing 4500 mg/liter D-glucose, 10% (v/v) fetal calf serum, 100 IU/ml penicillin, and 100 g/ml streptomycin.
Library Screening-An African green monkey liver, oligo(dT)-primed cDNA library was constructed in the pcDNAII plasmid vector by Invitrogen (San Diego) using RNA purified in our laboratory. 9.5 ϫ 10 6 primary recombinants were obtained. Colonies were screened by transfer to duplicate nitrocellulose filters as described (21). Bacterial clones containing ACAT sequences were detected by hybridization with a 32 P-labeled antisense riboprobe transcribed from a pBlueScriptSK Ϫ plasmid template (Stratagene) containing a 355-base pair African green monkey liver cDNA fragment that was derived by reverse transcriptase-polymerase chain reaction (RT-PCR) of green monkey total liver RNA. The primers used in that process were patterned on sequences found in a partial human clone termed ARGP2 (ACAT-related gene product 2) and provided to us by Dr. Steven Sturley (15,17). Primer sequences were selected from the deduced carboxyl-terminal and 3Ј-noncoding region of the putative ACAT: argp2f, 5Ј-TTGAAT-TCATGATGCATGACCAGC-3Ј and argp2r, 5Ј-TTGAATTCTCCAGA-TCTCAGTCCTTGC-3Ј. Avian myeloblastosis virus reverse transcriptase (Promega) with an oligo(dT) primer was used to synthesize first strand cDNA from 0.5 g of poly(A) mRNA which had been isolated from African green monkey liver biopsy tissue as described below. All PCRs, unless specifically noted, were performed with Vent polymerase using standard conditions suggested by the supplier of the enzyme (New England Biolabs). The nucleotide sequences of the inserts in purified positive clones were determined in both directions with automated sequencing performed on a Perkin-Elmer ABI Prism 377.
The full-length of the message for ACAT2 was determined with analysis by 5Ј-rapid amplification of cDNA ends using monkey-specific primers whose sequence was determined from the new cDNA in conjunction with the reagents and protocols supplied in a commercial kit (Life Science Technologies Inc.). The primer in the first extension round was argp2r2, 5Ј-GCAGACGCAGACGTCTCCTG-3Ј; the second nested primer was argp2r3, 5Ј-GGTTTGTCTTGAGATGGGTA-3Ј. The nucleotide sequence obtained for African green monkey ACAT2 has been submitted to GenBank, accession number AF053234.
The ACAT1 coding region nucleotide sequence for the African green monkey is filed as GenBank number AF053336, and for the cynomolgus monkey, GenBank number AF053337. Only one amino acid difference between the predicted amino acid sequences for ACAT1 was found when comparing between these two old world primate species.
mRNA Analysis-Total cell RNA was isolated from snap-frozen monkey tissue biopsies using TRIzol reagent (Life Technologies). Poly(A) mRNA was purified by batch oligo(dT)-cellulose binding (Ambion). For Northern blots, unless specifically noted, 1 g of poly(A) RNA from each tissue was electrophoresed into 1.2% agarose gels containing 2.2 M formaldehyde (21) and subsequently transferred onto a Nytran membrane by positive pressure (Turboblotter; Schleicher & Schuell). To detect specific ACAT mRNA the membranes were hybridized with 32 P-labeled antisense riboprobes designed to contain sequences unique to each of the two ACAT enzymes. A probe for the newly isolated ACAT2 message was synthesized from the 3Ј-noncoding region of the isolated African green monkey hepatic cDNA clone pACAT2.23 (nucleotide 1634 to 1902 in Fig. 1). In addition, a riboprobe to the 5Ј-end coding region of ACAT2 (nucleotides 1 to 386 in Fig. 1) was also used. A riboprobe for the monkey equivalent of human ACAT1 was prepared with a SmaI-BamI restriction fragment encompassing the 5Ј-end of the coding sequences of the monkey cDNA representing the region equivalent to nucleotide 1381 to 1725 of the human ACAT sequence (2). A cyclophilin cDNA (Ambion) was used as a load control on Northern blots (22).
Expression of ACAT cDNA-The cDNA insert from ACAT2 and the cDNA of the African green monkey ACAT1 were ligated into pcDNA3 ϩ (Invitrogen) or pCMV5 plasmid vectors (25) for transfection into established tissue culture cell lines. The AC29 cells were transfected using the FuGene 6™ reagent with the protocol provided by the manufacturer (Boehringer Mannheim). Five g of recombinant ACAT cDNA plasmids mixed with 10 l of FuGene were applied to cell monolayers at 80% confluence in 100-mm dishes. COS-1 cells were transfected using a DEAE-dextran procedure (26) in which 6 g of recombinant plasmid was applied to cells that were 70% confluent in 100-mm culture dishes. Cells were harvested by scraping from the plates 48 h after transfection. Cells were disrupted with sonication and ACAT activity was determined in cell homogenates or in isolated microsomes with methods that have been described previously (13) but modified to use cholesterol suspended in cyclodextrin instead of in detergent for the preincubation step that saturates the cholesterol substrate pool.
In Vitro Translation-ACAT1 and ACAT2 cDNAs in plasmid pcDNA3 ϩ were linearized at their 3Ј-ends with ApaI and XbaI, respectively. The ileal bile acid transporter cDNA in plasmid pCMX (23) was linearized with NotI. Transcription by T7 RNA polymerase and translation in rabbit reticuloctye lysates (Promega) containing canine pancreas rough microsomes and [ 35 S]methionine was performed as described (24). For carbonate extraction, translation reactions were adjusted to 0.1 M Na 2 C0 3 , pH 11.5, 5 mM dithiothreitol and a final volume of 190 l. After incubation on ice for 15 min, samples were transferred to Airfuge tubes and underlayered with 60 l of 0.2 M sucrose, 0.1 M Na 2 C0 3 , pH 11.5, 2.5 mM Mg(OAc) 2 , 5 mM dithiothreitol. After centrifugation at 30 p.s.i. for 15 min (Airfuge Ultracentrifuge; Beckman Instruments), supernatant fractions were recovered by precipitation with 2 volumes of saturated ammonium sulfate; pellet fractions were dissolved in SDS-PAGE sample buffer. Samples were analyzed by 12.5% SDS-PAGE and autoradiography.
Liver Cell Isolation-The hepatocyte isolation procedure was patterned after the work of Lanford et al. (27,28). Hepatocytes from young adult male African green monkeys or cynomolgus monkeys fed chow diets were isolated in a fashion similar to the original collagenase two-step perfusion method of Berry and Friend (29) with modifications. Animals were anesthetized with ketamine hydrochloride (25-50 mg/kg body weight), the abdomen was surgically opened, and the portal vein and vena cava were then cannulated. The liver was then continuously flushed with a calcium, magnesium-free Krebs-Henseleit original Ringer bicarbonate solution containing 0.5 mM EGTA, pH 7.4, for approximately 10 min at a rate of 1 ml/min/g liver while the liver was excised from the abdominal cavity. The caudate lobe of the liver was then tied off, removed, and snap frozen. The liver was then suspended on a nylon screen in a closed chamber. Hank's balanced salt solution containing 26 mM Tricine buffer, pH 7.4, 1 mM CaCl 2, and 0.05% collagenase (Boehringer Mannheim) was then recirculated for 15-20 min (1 ml/min/g liver). All procedures were done under sterile conditions.
Throughout perfusion, solutions were kept at 37°C and oxygenated with 100% oxygen. After perfusion, the liver was transferred to a sterile container, Williams E medium containing 0.5% bovine serum albumin was added, and the tissue was teased apart. The resulting hepatocyte suspension was filtered through a 500-m nylon mesh and washed twice by gentle resuspension in fresh Williams E medium with 0.5% albumin followed by centrifugation at 50 ϫ g for 5 min, at 4°C. Hepatocyte yield was typically in excess of 1.5 ϫ 10 9 cells/liver, and the viability was above 90% as determined by trypan blue exclusion. RNA was extracted using the TRIzol protocol from about 50 ϫ 10 8 freshly isolated hepatocytes.
From the remaining cell suspension, Kupffer cells were collected by centrifugation at 550 ϫ g at 4°C. The Kupffer cell pellet was resuspended in 10 ml of Hank's medium containing 0.2% Pronase and incubated at 37°C for 1 h to lyse any remaining hepatocytes. The cells were then repelleted by centrifugation, resuspended, and washed with 10 ml of a solution of 0.15 mM NH 4 Cl, 0.01 mM KHCO 3 , 1 M EDTA to lyse red blood cells. Kupffer cells were then repelleted, rinsed, and the RNA of the isolated cell pellet was promptly extracted using the TRIzol protocol. The yield was approximately 15 ϫ 10 8 cells/liver. The degree of contamination with hepatocytes, checked microscopically, was estimated to be Յ10%.
Structure Prediction-Topology and secondary structure predictions from the deduced ACAT amino acid sequences were performed using the Predict Protein service maintained by the Protein Design Group, European Molecular Biology Laboratory, Heidelberg, Germany (30,31). The analytical methods used for the probability projections of transmembrane helices (32) are based on multiple sequence alignments with proteins whose crystal structures are recorded in the SWISS-PROT data base.

RESULTS
Cloning and Sequence Analysis-As the initial step in the cDNA cloning strategy, RT-PCR using monkey liver mRNA as template was performed with primer sequences whose design was based on a partial human ACAT-related gene sequence termed ARGP2 by Sturley et al. (17). These primers were used to generate a monkey ACAT-like cDNA fragment used subsequently to screen the monkey liver cDNA library. Aligning the nucleotide sequence similarities in ARGP2 and human ACAT1, a 355-nucleotide seqment was chosen that corresponded to the presumed COOH-terminal portion of the coding sequence and the adjacent 3Ј-noncoding region. The resulting PCR product displayed 94% identity in nucleotide sequence to the region in human ARGP2 but only about 50% identity to the cognate patch in human ACAT1. Screening of 7.5 ϫ 10 4 colonies from the green monkey liver plasmid cDNA library with this probe yielded 13 positive signals that appeared on duplicate filters. Five of these were purified to homogeneity. The nucleotide sequence of the longest insert, present in plasmid pACAT2.23, is shown in Fig. 1. A 5Ј rapid amplification of cDNA ends determination involving primer-mediated extension of total RNA from African green monkey liver was performed in triplicate and confirmed that the sequence shown represents the full extent of the 5Ј-end of this message (data not shown).
Examination of this full-length cDNA sequence indicates that the structure of the mRNA encoding this new African green monkey ACAT protein is clearly distinct from that of the human ACAT. Assuming that the initiation of translation is at the initial ATG codon at nucleotide 80 (Fig. 1), the 5Ј-untranslated region for the monkey ACAT-like message is only 79 nucleotides long in contrast to the 1396 nucleotides for human ACAT1 mRNA (2). The 3Ј-end untranslated region of the pACAT2.23 cDNA insert consists of 421 nucleotides that contain the conventional AATAAA polyadenylation signal 24 residues upstream from the run of A residues that terminates the cDNA. The position of this motif is 45 nucleotides downstream of the first of 2 polyadenylation sequences that appear in the 965 nucleotide 3Ј-untranslated region of the human macrophage ACAT message (2).
The amino acid sequence deduced from the cDNA insert in pACAT2.23 is shown in Fig. 1. Translation beginning at the first methionine codon gives rise to an open reading frame of 526 amino acids residues. This compares to 551 amino acids for human and monkey ACAT1. Consistent with the sequence of human ACAT1, the NH 2 -terminal amino acid sequence of the new enzyme does not appear to encode a signal peptide sequence (33).
The deduced AA sequence for the African green monkey ACAT-related protein appears completely unrelated to that of the published sequence of human ACAT for the initial 101 amino acids, but from that point to the end of the protein, the sequences are 57% identical. In Fig. 1 the identical residues are indicated by shading. In particular, the amino acid pattern in the region centered around the DWWN residues near position 384 are highly conserved in all of the reported ACAT and ACAT-like genes (15,34). The primary structure of the region surrounding the serine residue at position 249 in the monkey ACAT2 sequence, which represents the cognate of the essential serine at position 269 in the hamster ACAT (4), also closely resembles the corresponding region in human ACAT1 (34). Other primary structural elements in the monkey sequence include a leucine heptad motif between amino acids 342 and 363 and 2 potential N-glycosylation motifs at N329 and N387, only the second of which is present in human ACAT.
Prediction of the topology of the African green monkey ACAT2 protein from the deduced amino acid sequence indicates the existence of 7 potential transmembrane domains, including amino acid residues 124 -141, 162-179, 205-222, 306 -323, 344 -366, 418 -435, and 440 -461, with the aminoterminal 123 amino acids as the only long span predicted to be intracytoplasmic (Fig. 2). In this model, 26% of the amino acids in the protein are predicted to interact directly with the ER membrane as transmembrane helices. Most of the remainder of the peptide, including the essential serine at position 249, is predicted to reside within the ER lumen.
Comparison of African Green Monkey ACAT1 and ACAT2-To further characterize the structure of the newly identified ACAT2 and compare its properties to ACAT1, cellfree translations were performed. A cDNA representing the African green monkey form of human ACAT1 was produced by RT-PCR (see "Experimental Procedures"). The amino acid sequence deduced from this cDNA differs from the human sequence at only 11 positions, most of which are conservative. The mRNAs for ACAT1 and ACAT2 were transcribed in vitro and used to program rabbit reticulocyte lysate translation extracts. Translation products were resolved by SDS-PAGE and detected by autoradiography. As observed in Fig. 3, lanes 1 and 2, translation of ACAT1 (A1) and ACAT2 (A2) mRNA gave products with apparent molecular masses of ϳ50 and 47 kDa, respectively.
To confirm that the newly discovered ACAT2 is an integral membrane protein and to determine if any of the potential sites for N-linked glycosylation are utilized, translations were repeated in the absence (Ϫ) or presence (ϩ) of nuclease-treated rough microsomes (RM). Following translation, samples were adjusted to 0.1 M Na 2 C0 3 , pH 11.5, and separated into pellet (P) and supernatant (S) fractions by centrifugation. Most of the protein was recovered in the carbonate supernatant when ACAT1 and ACAT2 were translated in the absence of RM (compare lanes 7 and 8; 11 and 12). In contrast, when translations were performed in the presence of RM, a significant proportion of the protein was recovered in the carbonate pellet, indicating successful integration into the microsomal membranes. In neither case, however, was a mobility shift detected upon translation in the presence of RM suggesting that neither ACAT is modified by N-linked glycosylation (compare lanes 8 and 9; 12 and 13). To confirm that the microsomal membranes were glycosylation competent, mRNA for the ileal bile acid transporter protein was translated. The primary translation product of ileal bile acid transporter is ϳ33 kDa (Fig. 3, asterisk) and is recovered in the carbonate supernatant. Upon addition of RM, most of the protein underwent glycosylation at two sites resulting in a mobility shift corresponding to an apparent increase in size of ϳ5 kDa (2). The glycosylated product was recovered primarily in the carbonate pellet (lane 5).
Expression-AC29 cells have been shown to lack ACAT activity, protein, and mRNA (20), and we did not find significant activity in microsomes made from AC29 cells that had not been transfected ( Table I). Transfection of cultured AC29 cells with monkey ACAT2 cDNA in either the pcDNA3 ϩ vector or pCMV5 vector gave levels of ACAT activity near 2,000 pmol/min/mg microsomal protein. The levels of activity achieved were essentially the same for the two vectors and were comparable to the activity found in the rat liver microsomes run with each assay as a control. Microsomes from transfected AC29 cells were incubated with two different ACAT inhibitors, one from Parke-Davis, PD138412 at 3.0 M, and one from Pfizer, CP113818 at 0.2 M, both of which had been shown to be effective against ACAT activity in monkey liver microsomes (11). Both inhibitors decreased activity by 80 to 90%.
ACAT activity was also measured in COS-1 cells, which were transfected with the pcDNA3 ϩ bearing the monkey ACAT2 cDNA using the DEAE-dextran procedure. COS cells have endogenous ACAT activity, but after transfection, ACAT activity was over 10-fold higher than background indicating that the monkey ACAT2 cDNA conferred ACAT activity to these cells as well (Table I). As an additional control, AC29 cells were transfected with monkey ACAT1 cDNA. These transfections gave ACAT activities 100-fold over background and similar to those The deduced amino acid sequence (singleletter code) is shown immediately above the nucleotide sequence of the ACAT cDNA insert in the pACAT2.23 clone. The ACAT cDNA insert nucleotides are numbered on the left and amino acid residues are numbered on the right with the predicted inititator methionine denoted as 1.
Letters representing amino acid residues that are in identical positions to those found in the human ACAT1 are shaded. Doubly underlined nucleotides represent the polyadenylation signal.
found with the African green monkey ACAT2 cDNA. The cell transfection studies indicate that the monkey ACAT2 clone encodes an enzyme that confers ACAT activity to mammalian cells. It is therefore appropriate to name this enzyme ACAT2 to distinquish it from the human ACAT originally cloned by Chang and colleagues (2), which we refer to as ACAT1.
Tissue Distribution-The tissue distribution of expression of ACAT1 and ACAT2 was determined by analysis of poly(A) ϩ mRNA isolated from 18 different monkey tissues using Northern blots (Fig. 4A). These data are typical of studies done on selected tissues from 5 different African green monkeys and 2 different cynomolgus monkeys. The presence of ACAT2 was detected with a riboprobe sequence representing the 3Ј-noncod-ing portion of the ACAT2 message. A single band consistent with a 2,200-nucleotide long ACAT2 mRNA appears prominently in the lanes for liver and intestine (Fig. 4B). With the exception of a less pronounced, single band of 2600 bases that appears in the kidney lane, none of the other tissues tested shows an ACAT2 signal. When liver and intestine from cynomolgus monkeys were examined, the additional band at 2600 bases was typically seen. Probing this Northern blot with a riboprobe representing sequence in the unique amino-terminal portion of ACAT2 showed the same pattern of expression except that the larger band in the kidney lane was not present (data not shown), suggesting that hybridization to this larger band does not represent ACAT2 mRNA.
The smaller, 850-nucleotide band represents the signal arising from a labeled riboprobe for cyclophilin which was to serve as a load control. However, the lighter cyclophilin signals that are evident in the lanes for skeletal muscle, heart, pancreas, thyroid, and testes are apparently due to the fact that this particular message is less abundant in these tissues (22) and thus the cyclophilin signal does not quantitatively correlate with the amount of mRNA loaded in all lanes; the same low signals for these tissues were seen on 3 separate Northern analyses done for this study. Nevertheless, longer exposure of the autoradiographs still failed to show any ACAT2 signal in  any of these lanes. While skeletal muscle mRNA failed to show any appreciable signal with the cyclophilin load control, a prominent band was seen in this lane when a glycerol-3-phosphate dehydrogenase probe was substituted as the load control (data not shown). The ACAT1 probe, in contrast, hybridized to a characteristic, multi-band array in almost all tissues examined (Fig. 4A). While the highest signal appeared in mRNA from adrenal tissue (note that 5-fold less mRNA was loaded in the adrenal lane), prominent ACAT1 message bands were also visible in RNA from trachea, heart, liver, kidney, adipose tissue, and pancreas. Relative to the high signal in adrenal, the amount of ACAT1 message in liver and particularly in intestine seem disproportionately low in view of the high level of ACAT activity seen in microsomes made from these tissues. Skin mRNA showed an ACAT1 signal almost identical to that in the adjacent lane for aorta in other Northern blots (data not shown).
The sizes of the ACAT1 bands are consistent in all of the monkey tissues examined: 2.1, 2.5, 3.2, and 3.6 kilobase pairs. Except for the largest size, these are close to the values reported for rabbit adrenal tissue (5) but are significantly smaller than the reported sizes of the 4 major bands reported for human adrenal ACAT message.
In an effort to determine whether different ACAT enzymes are expressed in different cell types in monkey liver, a Northern blot comparing the specific ACAT messages in mRNA extracted from isolated cynomolgus monkey hepatocytes, Kupffer cells, and whole liver before cell isolation, is shown in Fig. 5. The banding pattern for ACAT1 is similar in hepatocytes and Kupffer cells, while the signal intensity is approximately double for Kupffer cells compared with hepatocytes. In contrast, the signal for ACAT2 in hepatocytes is significantly greater (ϳ6 times) than in Kupffer cells. Thus, the data suggest that both ACAT1 and ACAT2 are present in hepatocytes, while Kupffer cells contain primarily, if not solely, ACAT1, depending on the degree of hepatocyte contamination. The probes used for ACAT1 and ACAT2 were specific to sequences in each enzyme and the binding efficiency of either probe may be different. Therefore, it is not possible from these data to be able to compare the actual amounts of mRNA for these two enzymes.
When comparing ACAT2 RNA in either cell type to that extracted from a sample of the same liver from which the cells were isolated, the ACAT2 bands in either isolated cell type were lighter in intensity than for fresh whole liver. This may indicate that some loss of ACAT2 mRNA occurred during cell isolation procedures. This is not an unreasonable possibility since the cells remain at 37°C for some 30 min during collagenase digestion, and then at room temperature for another 60 -90 min during cell isolation and washing. The data suggest that ACAT2 mRNA may be more labile than ACAT1 in isolated hepatocytes.
The upper band seen for ACAT2 is the same size as the larger band seen for kidney in Fig. 4. This band was not seen in liver from African green monkeys, but is apparent in cynomolgus monkey liver. The blot of Fig. 5 was reprobed with the 5Ј-end probe for ACAT2 and this band was not seen (data not shown). DISCUSSION Evidence in nonhuman primate models has been obtained over the past two decades indicating that the accumulation of cholesteryl esters in low density lipoprotein, with the associated low density lipoprotein particle enlargement, is highly correlated with the extent of coronary artery atherosclerosis (35)(36)(37)(38). Most recently, we have shown that the secretion of cholesteryl esters from the liver, as monitored during isolated liver perfusion, is correlated with the enrichment of the low density lipoprotein particles with cholesteryl esters and is also predictive of the extent of coronary artery atherosclerosis (12). The formation of cholesteryl esters in primate liver is catalyzed by ACAT (13), and the secretion of cholesteryl esters and apoB by the perfused primate liver is decreased by ACAT inhibitors (11). When diets rich in monounsaturated fat are fed to African green monkeys, the livers accumulated excess cholesteryl ester in the form of cholesteryl oleate, the secretion of cholesteryl oleate was enhanced, and the extent of coronary artery atherosclerosis occurred out of proportion to the low density lipoprotein and high density lipoprotein cholesterol concentrations (12,39). All of this evidence points to a key role for hepatic cholesterol esterification by ACAT in promoting diet-induced coronary artery atherosclerosis in primates. Therefore, it was important for us to identify the ACAT enzymes in primate liver as a prelude to study of their regulation.
The evidence supporting the conclusion that the African green monkey cDNA clone we have characterized in this publication is an authentic ACAT enzyme includes, first and foremost, that transfection of mammalian cells with this clone results in high levels of ACAT activity (Table I). The activity in the cells was shown to be associated with the microsomal fraction. The levels of activity were as high as that found in rat liver microsomes and as high or higher as that seen when monkey ACAT1 cDNA was transfected into the cells. Furthermore, the activity was inhibitable by at least two ACAT inhibitors. The AC29 cells from Chang and colleagues (2,20) used for most of these transfection studies have no ACAT activity, protein, or mRNA, providing a rigorous system in which to characterize ACAT gene expression. A second mammalian cell line, COS-1, was also successfully transfected with ACAT2 to make sure that the outcome was not peculiar to AC29 cells.
The predicted primary structure of the monkey ACAT2 clone is highly analogous to that described for ACAT1. The fact that the first approximately 100 amino acids of each enzyme are of a completely different sequence suggests that this portion of the enzyme may confer separate functions for the two enzymes. However, over the remaining 425 amino acids, a 57% sequence identity occurs for human ACAT1 and monkey ACAT2 suggesting similar functions for this portion of the protein. Easily recognizable aspects of the ACAT2 sequence that may contribute to its function as an ACAT enzyme are: 1) the analog to serine residue 269, which when converted to a leucine residue in SRD-4 cells resulted in an inactive ACAT (4), 2) leucine heptad motifs that may facilitate protein-protein interactions important in the presumed oligomerization that may occur for active ACAT (34), and 3) multiple transmembrane helical domains that may serve to anchor the protein in the ER membrane.
Other evidence that the primary structure of the protein coded for by the monkey ACAT2 clone is as indicated in Fig. 1 was obtained during in vitro translation studies. The primary translation product for the African green monkey ACAT2 cDNA ran at about 47,000 M r compared with the primary product for monkey ACAT1, which ran at about 50,000 M r during SDS-polyacrylamide gel electrophoresis (Fig. 3). These apparent differences in size are consistent with the presumed differences in the size of the two proteins based on the deduced amino acid sequences. The fact that both enzymes appear to run at molecular weights that are smaller than the predicted sizes based on primary structure has been observed by others (18,20).
The distribution of ACAT mRNA among 18 different primate tissues is the most detailed comparison yet made available (Fig. 4) and was facilitated by the collection of multiple tissues from monkeys being necropsied for atherosclerosis evaluations in other experiments. The fact that ACAT1 mRNA is present in most tissues, with the possible exception of skeletal muscle, is compatible with the prevalent notion that most cells can esterify excess cholesterol when the need arises. The remarkably high level of ACAT1 mRNA in adrenal, compared with that in other tissues, is likely responsible for the significant amount of cholesteryl ester accumulation that normally occurs in this organ. High adrenal cholesteryl ester levels were dramatically reduced by the disruption of the ACAT1 gene in mice (14).
The presence of multiple bands for ACAT1 mRNA in various tissues has been noted before but the reason is unexplained (2,5). These data show that the banding pattern was quite similar for most tissues with the 5Ј-coding region probe used in these studies. We have also probed the same Northern blots with a probe specific to the 3Ј-end of the ACAT1 coding sequence (data not shown). This probe appeared to show some suggestive differences in band intensity among tissues that were not seen with the 5Ј-end riboprobe, but the same size range of bands was seen with both probes. Both the 5Ј-and 3Ј-end untranslated nucleotide sequences for ACAT1 mRNA are relatively long, especially compared with ACAT2, but any metabolic role for alternate splicing remains uncertain.
With both the 3Ј-end and 5Ј-end riboprobes to ACAT2 mRNA that were made to sequences dissimilar to any found in ACAT1, the lack of reactivity in tissues other than liver and jejunum was striking, especially considering the near ubiquitous expression of ACAT1. The presence of a larger band in the kidney is believed to represent cross-hybridization with the message for an unidentified protein since the use of a riboprobe specific to the sequences in the coding region near the 5Ј-end did not identify the larger band in the kidney. Clearly these data suggest that ACAT2 has a specific function related to this specific tissue distribution. While we cannot be sure what this function is, the secretion of apoB-containing lipoproteins is carried out almost exclusively by liver and intestine, and it is possible that ACAT2 is responsible for formation of the cholesteryl esters of newly secreted lipoprotein particles. We have provided suggestive evidence that some ACAT activity is required for apoB particle secretion (11).
In the context of tissue distribution and function, the modeling of the secondary structure of the monkey ACAT1 and ACAT2 enzymes that was done with the Predict Protein technology (30 -32) is of interest. The prediction for ACAT2, as shown in Fig. 2, is that there are seven transmembrane domains with the longer peptide loops between these helical transmembrane regions being situated mostly on the lumenal side of the ER membrane. In contrast, for the topology prediction for monkey ACAT1, the model shows eight transmembrane domains. Basically, the same seven domains as found in ACAT2 are also predicted for ACAT1, but one additional transmembrane helix is found at amino acid residues 247-264. This has the effect of moving the putative active site serine at amino acid 269 as well as the remainder of the peptide loops that likely contain other active site residues to the cytoplasmic side of the membrane. The NH 2 -terminal 100ϩ amino acids for both enzymes are predicted to be on the cytoplasmic side of the ER membrane, but since there is almost no sequence similarity in this region between ACAT1 and ACAT2, this portion of the protein probably does not contain active site residues. Thus, the model predictions would be consistent with the active site of the ACAT1 enzyme being on the cytoplasmic side of the membrane where it could possibly facilitate cholesteryl ester incorporation into cytoplasmic lipid droplets, as occurs extensively in the adrenal. In contrast, ACAT2 is predicted to have its active site on the lumenal side of the ER membrane (Fig. 2). This orientation may facilitate cholesteryl ester incorporation into apoB-containing lipoproteins, as lipoprotein particle assembly and secretion is known to occur in the ER lumen. FIG. 5. Autoradiograph of Northern blots of mRNAs from liver cell types and whole liver. One-g samples of poly(A) ϩ RNA that was purified from isolated cell populations of hepatocytes (H) and Kupffer cells (K) and from a whole liver biopsy (L) were used to prepare and analyze Northern blots as described in the legend to Fig. 3. The riboprobes used were as described in the legend to Fig. 4 for ACAT1 (top panel), ACAT2 (middle panel), and cyclophilin (bottom panel). Sizes of marker RNAs appear at the right.
Perhaps it was the inhibition of ACAT2 by ACAT inhibitors that resulted in the decreased rate of appearance in perfusate of apoB and cholesteryl esters in the monkey liver perfusion studies (11). The data in Table I show that ACAT2 is sensitive to two of these ACAT inhibitors. Further work is needed to test these hypotheses.