We have isolated the full-length cDNA for human ATP-binding cassette, sub-family A, member 2 transporter (ABCA2). The ORF of this cDNA encodes a protein consisting of 2436 amino acids with apparent molecular weight of Mr 270,000. Accordingly, ABCA2 is the largest known mammalian ABC transporter described thus far. Analysis of mRNA expression levels indicated that ABCA2 is highest in human brain and has a broad expression pattern in a panel of tumor cell lines. Using specific antibodies to ABCA2 and various organelle marker proteins, ABCA2 was found to colocalize with the lysosomal/endosomal marker LAMP1, forming discrete, punctate intracellular vesicles. In ABCA2-transfected cells, the transporter also colocalized with a fluorescently labeled steroid analogue, estramustine. The sequestration of the steroid into the lysosomal/endosomal compartment indicates a potential substrate specificity for ABCA2. Furthermore, the presence of a lipocalin signature motif in the ABCA2 sequence suggests a possible broad role for this protein in the transport of steroids, lipids, and related molecules.
ABC4 transporters are cellular membrane proteins responsible for unidirectional (outward or inward) transport of substrates, dependent upon ATP-hydrolysis in prokaryotes and eukaryotes (1). Members of the ABC protein superfamily are readily recognizable by the presence of an ABC consensus sequence. This is comprised of the highly conserved Walker A and B motifs separated by 90–120 amino acids, within which is a characteristic “signature” motif (2). The human ABC superfamily of proteins consists of at least seven sub-families: A (ABC1); B (MDR/TAP); C (CFTR/MRP); D (ALD); E (OABP); F (GCN20); and G (WHITE). Typically, proteins from this superfamily form dimers consisting of two ABCs and two membrane domains (3). These are formed either from separately encoded polypeptides with ABC-membrane domains or from tandemly duplicated ABC-membrane domains within a single molecule. Analysis of membrane domain topology ranges from 6–11 transmembrane helices (4), which function in substrate recognition, binding, and channeling, whereas ATP hydrolysis in the ABC domain provides the necessary energy for movement. As evidenced by IR and fluorescence spectroscopy data, substrate binding and ATP hydrolysis induce conformational changes in the transporter family (5). Thus, the interactions between the ABC and membrane domains are integral in executing substrate transport.
Many ABC transporters play critical roles in the cellular efflux of endogenous or xenobiotic substrates, and their biological dysfunction has been implicated in a number of clinical disorders. For example, impaired function of the cyclic AMP-activated ABCC7 (CFTR) chloride channel appears to be the basic defect in epithelial and nonepithelial cells derived from cystic fibrosis patients (6). Some mutations in the peroxisomal membrane half-transporter ABCD1 (ALDP), are associated with abnormal peroxisomal β-oxidation of saturated, very long chain fatty acids, and result in the neurodegenerative disorder X-linked adrenoleukodystrophy (7). The absence of functional transporter subunits associated with antigen processing (ABCB2 and ABCB3), has been linked to different forms of HLA class I deficiency syndrome (8). Mutations in the mitochondrial half-transporter, ABCB7, (an iron-transporter) are responsible for X-linked sideroblastic anemia and ataxia (9). More recently, ABCA4, the rod photoreceptor ABC transporter, has been implicated in a whole spectrum of vision disorders (10). Mutations in the ABCA4 gene interfere with transport functions that can lead to conditions associated with Stargardt disease, age-related macular degeneration, fundus flavimaculatus, cone-rod dystrophy, and retinitis pigmentosa. Characterization of ABCA4 knockout mice (11) suggested that ABCA4 probably functions as an outwardly directed flippase for N-retinylidene-phosphatidylethanolamine, protecting the retinal pigment epithelium from toxic adducts of photobleaching. Mutations in a closely related transporter, ABCA1, have been linked to Tangier disease and familial high-density cholesterol deficiency syndrome (12, 13, 14). Disrupted export of cholesterol and phospholipids from the cells plays a major role in the pathogenesis of these diseases (15).
ABC transporters have been implicated in cellular drug resistance (16). In particular, members of the MDR and MRP subfamilies of proteins have been linked to simultaneous resistance to multiple cytotoxic drugs in cancer cells. MDR1 confers resistance to a variety of hydrophobic, amphipathic natural product drugs (17) whereas members of the MRP-subfamily cause resistance to anionic and neutral drugs frequently conjugated to acidic ligands (18). Extending the drug-resistance paradigm, research from our laboratory revealed that an ovarian carcinoma cell line made resistant to the estradiol-based agent, estramustine, has a homogeneously staining region at chromosome 9q34 (19). Fluorescence in situ hybridization with a probe specific to a region of ABCA2 indicated that this gene was amplified at 9q34. Both gene copy number and mRNA levels of this transporter were increased in the resistant cell line. In addition, antisense treatment directed toward ABCA2 mRNA sensitized the resistant cells to estramustine. Together, these results suggested that ABCA2 is causally involved in estramustine resistance and implied a possible role for ABCA2 in steroid transport. Recently, Kikuno et al. (20) showed that a partial-sequence human cDNA clone, highly similar to mouse Abca2, was more prevalent in brain than in other tissues. Our present data provide the first characterization of the full-length human ABCA2 cDNA and its detailed expression pattern. Functionally, the protein sequesters estramustine into intracellular vesicles with lysosomal/endosomal characteristics, indicating a potential role in steroid transport and/or metabolism.
MATERIALS AND METHODS
Isolation and Assembly of ABCA2 cDNA.
We cloned the ABCA2 cDNA using a variety of library screening and PCR-based approaches. As reported previously (19), we first isolated 1.75 kb of ABCA2 cDNA by PCR using primers designed against the expressed sequence tag representing human ABCA2 (EST0600) and mouse Abca2 cDNA (GenBank accession no. X75927).
Our initial library screening experiments were performed using a ready-made random and oligo(dT)-primed human fetal brain cDNA library generated in λ ZAP II (Stratagene, La Jolla, CA). This library was chosen because Abca2 expression in the mouse is most pronounced in the brain (21). The library was probed with a 32P-labeled (Prime-it II Random Primer Labeling Kit; Stratagene) 490-bp PCR fragment that we amplified from the 5′ end of the 1.75-kb ABCA2 fragment. Screening was performed according to the manufacturer’s instructions. The four longest ABCA2 clones were designated as λ15B, λ17B, λ5A, and λ9A (Fig. 1).
5′ RACE using nested-PCR was performed using human brain Marathon-Ready adapter-ligated cDNA as a template (Clontech, Palo Alto, CA). The first round of PCR used AP1, 5′CCATCCTAATACGACTCACTATAGGGC3′ (forward) and ABCA2-specific 5′TGAGTTTGTCCACGCAGACAACCAGAG3′ (reverse). The second round used PCR AP2, 5′ACTCACTATAGGGCTCGAGCGGC3′ (forward) with ABCA2 specific 5′CCAGCTCCACTCCCAGGCTTCTG3′ (reverse). The products were ligated into the pT-Adv plasmid (Clontech), and ligation extraction products were used to transform TOPO10′ cells (Invitrogen, Carlsbad, CA). On the basis of an additional 5′ sequence obtained from this approach, (clone 35, Fig. 1), the following primers were designed for subsequent 5′ RACE: (a) AP1 (forward) and 5′TGAGTTTGTCCACGCAGACAACCAGAG3′ (reverse); and (b) AP2 (forward) and 5′CCAGCTCCACTCCCAGGCTTCTG3′ (reverse) for second-round PCR. The longest clone isolated, designated 208, was ∼900 bp (Fig. 1).
To obtain the entire 5′ end sequence, we constructed a 5′ ABCA2 cDNA library from high quality human brain poly(A)+RNA (Clontech) using a Marathon cDNA Amplification Kit (Clontech). Gene-specific primer, 5′CCACTGGGCAGCGAGAAGTTGTC3′, was used for first-strand synthesis. Second-strand synthesis and creation of blunt ends was followed by DNA ligation with adapters provided with the kit, according to the manufacturer’s protocol. The library was then PCR-amplified using AP1 and 5′CCACTGGGCAGCGAGAAGTTGTC3′ and then amplified with the nested ABCA2 primer, 5′GAAGCTGGAGTTCTGGCGGATCT3′ and the adapter primer, AP2. Reaction products were cloned into pCR-XL-TOPO plasmid (Invitrogen) and transformed into DH5α cells. Fig. 1 indicates the location of the two longest clones, designated 65 and 119.
For 3′ RACE, we used 5′CAGACCACACTGGACAATGTGTTCGTG3′ (forward) and AP1 (reverse), and 5′TCATCAGCTTCGAGGAGGAGCGG3′ (forward) and AP2 (reverse). Only short, 300-bp fragments were isolated (data not shown).
Full-length ABCA2 cDNA was assembled from four fragments (A–D). Fragments A and B were obtained by PCR from a human brain cDNA library using the following primers: (5′ATAAGCTTGCTGAGGCGGCGGAGCGTGGC3′ and 5′CCACTGGGCGAGAAGTTGTC3′ for fragment A; and 5′CCTCATTTTCCCCTACAACC3′ and 5′ACCTGCTCCATCTTGCTGCTGAACAC, for fragment B. Fragment C was directly obtained by restriction digestion of KIA1065 clone (kindly provided by Dr. Takahiro Nagase from Kazusa the DNA Research Institute, Chiba, Japan; Ref. 20). Fragment D was obtained by PCR from KIA1065 clone using 5′CAGCGGCGGCAACAAGCGGAA3′ and 5′GGTGAATTCGGCAGGCACTGGGGGACTTGT3′ primers. PCR products were initially cloned into the pCR-XL-TOPO cloning vector. Fragment A was excised by HindIII and SalI digestion and subcloned into pCR-XL-TOPO clone containing fragment B. Fragment D was excised by KpnI and EcoRI digestion and subcloned into corresponding sites of (A+B)pCR-XL-TOPO construct. Finally, fragment C was cloned into the KpnI site of (A+B+D) pCR-XL-TOPO clone.
The ABCA2 cDNA fragment was inserted into HindIII and EcoRI sites of pcDNA3.1 vector (Stratagene).
The pEGFP-ABCA2 clone was constructed in the following way. The start codon of ABCA2 was modified using PCR (primers used were 5′TAGTACTCCTTGGGCTTCCTGCACCAGC3′ and 5′CCAGGGCAGATGAGGGACCAAAGA3′), and the resulting clone was inserted into ScaI and EcoRI sites of pEGFP-C3 vector (Clontech).
All PCR products were verified by double-stranded DNA sequencing.
Mapping of Transcription Start Site.
5′ RACE was used to map the start site of the ABCA2 transcript. Reverse transcription of total brain RNA (Clontech) was performed using the antisense gene-specific primer 5′CATCCAGCAGGTCCCCCAGAAGC 3′ with subsequent RNase H treatment. First-strand synthesis product was subjected to dC tailing reactions with terminal deoxynucleotidyl transferase. The first round of PCR amplification was then performed using 5′ RACE anchor primer 5′GGCCACGCGTCGACTAGTACGGGIIGGGIIGGGIIG3′ and gene-specific antisense primer 5′AAACAGGTTGCCCTTCCTCCACCAC3′. A second round of PCR amplification was performed with universal amplification primer 5′GGCCACGCGTCGACTAGTAC3′ and gene-specific antisense primer 5′ACAGCGATTGCATGACAGGCAG3′. A single ∼300-bp product was obtained, and, after purification, it was cloned and sequenced.
Northern Blot Analysis.
Samples of the total RNA from a selection of the National Cancer Institute human tumor cell line panels were provided by Dr. Anne Monks (National Cancer Institute, Rockville, MD) (22). Ten-μg samples were separated on formaldehyde agarose gels, transferred to nylon membranes (GeneScreen; NEN Life Science Products, Inc., Boston, MA), and hybridized by standard protocols. 32P-radioactive labeling of a gel-purified 1.75-kb ABCA2 probe (position 3754–5498) to high specific activity was performed by random priming kit (Prime-It II Random Primer Labeling Kit; Stratagene).
Multiple Tissue Northern blot (MTN; Clontech) was hybridized with an 850-bp probe (position 4157–5005) that was PCR-amplified using the following primers: 5′AGGGAGCTGGCTACACCGACG3′ (forward) and 5′CGCCTGTGACCACCCGCATCT3′ (reverse). 32P-radioactive labeling was performed according to the random primer labeling method.
ABCA2 Sequence Analysis.
Nucleotide sequencing was performed with an ABI 377 DNA sequencer. The sequences were assembled in the Sequencher program (Gene Codes Corporation, Ann Arbor, MI). Protein computer analyses were performed with the Genetics Computer Group Package, version 9.1 (Madison, WI), and McVector (Oxford Molecular, Oxford, United Kingdom). N-linked glycosylation sites were obtained by ScanProsite analysis.5 TMDs were obtained with TopPred 2 analysis.
HEK293 cell lines (American Type Culture Collection, Rockville, MD) were cultured in DMEM supplemented with 50 μg/ml streptomycin, 50 units/ml penicillin, 2 mm glutamine, and 10% (v/v) fetal bovine serum. Transient transfections were performed using SuperFect (Qiagen, Valencia, CA) transfection reagent according to the manufacturer’s instructions. Cells (2 × 105/well) were seeded on LabTek II chambered coverglass (Nunc 155379) and transfected with 1.5 μg/well of pEGFP-C3 or pEGFP-ABCA2 vector. After 48 h, slides were prepared for confocal microscopy or lysed in 18 mm 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate buffer supplemented with protease inhibitors for immunoblot analysis.
The stably overexpressing HEK293 clones were created by electroporation of 2 × 106 HEK293 cells with 10 μg of either the pcDNA-ABCA2 or the parental pcDNA3.1 vector using Bio-Rad (Hercules, CA) gene Pulser apparatus. At 48 h after electroporation, the growth medium was changed to include 0.8 mg/ml G418. At ∼3 weeks, independent G418-resistant colonies were isolated using the cloning cylinder technique and expanded for confocal, drug sensitivity, or immunoblot analysis (same preparation conditions as above).
Polyclonal rabbit antisera were raised against synthetic peptides corresponding to residues 1499–1522 of the primary ABCA2 sequence. Protein levels in whole cell lysates were quantified by Bradford assay (Bio-Rad) and separated by 6% SDS-PAGE and transferred to nitrocellulose membranes (Bio-Rad). Membranes were blocked with 5% powder milk in Tris-buffered saline and 0.1% Tween 20 and probed for 1 h at room temperature with anti-ABCA2 antibody or mouse anti-GFP antibody (Clontech 8362-1). Each primary antibody was used at 1:500 dilution. Antirabbit-horseradish peroxide (Amersham Pharmacia Biotech NA934, Uppsala, Sweden) or antimouse-horseradish peroxide (Amersham Pharmacia Biotech NA931) was used as secondary antibodies at 1:1000 dilution. Bands were visualized using Renaissance Western Blot Chemiluminescence Reagent Plus (NEN Life Science Products, Inc.).
Cells were prepared for immunofluorescence by standard methods. Fixation was accomplished by −20°C methanol or by 4% paraformaldehyde with 0.1% Tween 20. Primary antibodies used were a polyclonal rabbit anti-ABCA2; a polyclonal goat anti-LAMP1 (C-20), sc8090 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA); a polyclonal rabbit anti-PMP70 (Zymed Laboratories, Inc., San Francisco, CA); a polyclonal rabbit anti-Human Endoplasmic Reticulum Associated Amyloid Beta Binding Protein (Chemicon International, Inc., Temecula, CA); a monoclonal mouse anti-Golgi Complex ab-1 (Clone 371-4; Neomarkers, Fremont, CA); and monoclonal mouse antihuman mitochondria (Chemicon International, Inc., Temecula, CA). Secondary antibodies used were species-specific for the primary antibodies, and all were conjugated with Rhodamine Red X and Cy2 (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA). A Mowiol antifade in glycerol was used in an aqueous mounting medium.
Immunofluorescently stained cells were imaged using a Bio-Rad MRC-600 laser scanning confocal microscope. Using the two photomultiplier setting on the laser scanning confocal microscope, dual acquisition of GFP or Cy2 and Rhodamine Red X was accomplished. Optical sections were acquired with ×60 and ×100 objectives using a 0.5-μm step size. The volumes were rendered using Voxel View (Vital Images, Plymouth, MN).
Dansylated Estramustine Imaging.
Transfected HEK293 cells were treated for 2 h with a dansyl chloride derivative of estramustine (synthesis described in Ref. 23) at a final concentration of 200 nm. The drug was removed, the were cells washed with PBS, and fresh media was added to the cells. Drug-treated cells were then imaged with a Quantix 12-bit cooled CCD camera (Roper, Inc., Tuscon, AZ). For transiently transfected HEK293 cells, dansylated estramustine, fluorescent GFP, and phase-contrast images were generated simultaneously using Isee software (Inovision Corp., Durham, NC) to drive the CCD, a Ludl filter wheel, and a shutter attached to a Nikon TE300 inverted microscope. For stably transfected HEK293, dansylated estramustine, anti-ABCA2 conjugated to Rhodamine Red X, and DIC images were taken sequentially. A forced-air incubator encompasses the microscope so that cells can be observed at near-physiological conditions.
Analysis of Drug Sensitivity.
Estramustine sensitivity was analyzed using a tetrazolium salt microtiter plate assay (Cell Titer 96 Cell Proliferation Assay; Promega, Madison, WI). Cells were seeded in triplicate at 8000/well in 96-well dishes in complete medium supplemented with 10% fetal bovine serum. The next day, estramustine at various dilutions was added to the growth medium. Assays were performed at 72 h of growth in the presence of the drug.
Cloning of Human ABCA2.
The ABCA2 transporter was cloned by initially screening an oligo(dT) and random-primed human fetal brain cDNA library probed with a PCR-amplified 1.75-kb fragment of ABCA2 (19). Of four isolated clones, the longest, λ9A, was 4.8 kb in length (Fig. 1). This fragment showed 86% nucleotide identity to the 4.7 kb of available mouse Abca2, with the least homology in the last 300 bp, and no predicted in-frame stop codon. Initial 5′ RACE yielded another ∼1.4-kb of sequence at the 5′ end of ABCA2 cDNA from the longest overlapping clones, 35 and 208 (Fig. 1). To obtain 5′ ABCA2 cDNA, we created a human brain ABCA2 cDNA library using gene-specific primers near the 5′ end of extracted ABCA2 cDNA sequence. The two longest amplified products, 65 and 119, both contained the same 5′-UTR and the ATG start in standard Kozak context (24). 5′ RACE was performed to map the transcriptional start site and yielded a single band. Thus, the major transcriptional start site is at a cytosine, 50 bp upstream of the translational start site.
For the 3′ RACE experiment, we used gene-specific primers upstream from the predicted stop codon, (based on the mouse ORF; Ref. 21) and generated a number of 300–400-bp fragments containing putative stop codons and an additional ∼300 bp of 3′-UTR (data not shown). In addition, an interim BLAST search (25) of the sequence of the λ9A clone (Fig. 1) returned a brain clone KIAA1062 (20) with high identity to the last ∼4.7 kb of this clone and provided an additional ∼400 bp of 3′-UTR (Fig. 1). The assembled total sequence was 8056 bp with 7308 bp corresponding to the ORF, 50 bp 5′-UTR and 698 bp 3′-UTR (Fig. 1).
Primary Structure of ABCA2.
Nucleotide analysis revealed that the ABCA2 ORF was comprised of 7308 bp and encoded a 2436-amino acid protein with a predicted molecular weight of Mr 270,000 (Fig. 2 A).
Comparison of the ABCA2 primary structure with several other human ABC transporters showed that ABCA2 shares a 45.4% amino acid identity with ABCA1 and slightly less with ABCA7 (40.5%), ABCA4 (38.5%), and ABCA3 (38.2%). Homology comparison with the partial mouse-Abca2 sequence (21) revealed 94.4% amino acid identity within the published region. The amino terminal of ABCA2 shares high identity with the amino terminals of ABCA1, ABCA3, ABCA4, and ABCA7 (55–70% in the first 20 amino acids). The significance of this region has yet to be resolved.
The ABCA2 protein is a full-size transporter that contains a tandem repeat of the recognizable hydrophobic domain with six transmembrane helices followed by highly conserved ABCs (Fig. 2,A). Structurally, the hydrophobic domain spacing is reminiscent of that found in ABCA4 (Fig. 2, B and C). The cytosolic NH2 terminus is followed immediately by the first transmembrane segment and a long extra-cytosolic loop. The long hydrophilic linker portion of the protein (∼700 amino acids) is separated into two halves by a HHD. This region of the protein, as suggested by Luciani et al. (21), may correspond to a putative regulatory domain similar to the one found in ABCC7. The potential sites of N-glycosylation are concentrated in 2 regions with the first 15 on the large extracellular loop and the second 6 between the putative HHD and the 7th transmembrane segment. Although there is evidence for these posttranslation modifications in the first region for ABCA4 (26), no data on glycosylation within the second region for ABC1-subfamily members is presently available. A number of potential phosphorylation sites were also apparent, including protein kinase C, casein kinase, tyrosine kinase and cAMP dependent-protein kinase. Many of them are located in the putative linker region of ABCA2. Of particular interest to the potential role of ABCA2 in the transport of steroids, a lipocalin signature (GQSRKLDGGWLKV) was identified at position 1424 within the putative regulatory domain and close to the HHD. This motif is characteristic of small lipocalin proteins that transport lipids, steroids, bilins, and retinoids (27).
ABCA2 Northern Blot Analysis.
The expression patterns of ABCA2 in a panel of normal adult human tissues was investigated by Northern blot analysis. As presented in Fig. 3, ABCA2 is expressed as a single transcript of ∼8 kb. Of the 12 tissues analyzed, ABCA2 expression was highest in the brain, whereas lower levels of expression were observed in kidney and liver. These results are similar to those found with the mouse (21). For the National Cancer Institute tumor cell line panel, there was a broad pattern of ABCA2 expression (Table 1).
ABCA2 Protein Expression and Cellular Localization.
To study ABCA2 localization and function, ABCA2 cDNA was subcloned into GFP-tagged mammalian expression vector and transiently transfected into HEK 293 cells. To explore concerns that GFP might impose improper ABCA2 subcellular localization and function, we electroporated pcDNA-ABCA2 or pcDNA3.1 vector into HEK 293 cells to create stably overexpressing cell lines. Increased untagged/native ABCA2 expression was detected, and one such colony where ABCA2 was highly expressed was selected for additional characterization and presented as the HEK/ABCA2 cell line. The protein expression in transfected cells (HEK293/GFP-ABCA2, HEK293/GFP, HEK293/ABCA2-pcDNA, and HEK293/pcDNA) was examined by immunoblot using polyclonal antisera against the peptide containing the 24 amino acids of the linker region of ABCA2 and anti-GFP antibody (Fig. 4). The band visualized on our Western blot with either antibody was diffuse, specific for cells transfected with ABCA2 constructs, and estimated to be Mr >250,000, which is in agreement with the predicted molecular weight of Mr 270,000. This diffuse appearance is characteristic of glycosylated proteins (21 potential glycosylation sites are in ABCA2). The band diminished upon incubation of anti-ABCA2 antibody with excess quantities of the 24-amino acid peptide. In several untransfected cell lines, including HEK293, HL60, SKOV3, and SNB-75, no ABCA2-specific band was detected (data not shown).
The cellular localization of ABCA2 protein was visualized by confocal microscopy (Fig. 5). HEK/GFP-ABCA2 cells exhibited punctate cytoplasmic GFP-specific staining (Fig. 5, panel B). The distribution of green fluorescence colocalized with the fluorescent staining pattern obtained with the polyclonal antibodies to ABCA2, indicating a vesicle organelle location for ABCA2 (Fig. 5, panels A–D). We observed the same distribution pattern under divergent conditions. e.g., when we used a different transfection agent, such as FuGENE6 (Roche; Indianapolis, IN), stably transfected HEK293 with native ABCA2 cDNA (Fig. 5, panel V), or transfected an ovarian carcinoma cell line, SKOV3 (data not shown). Additional analysis was performed using a panel of fluorescently labeled organelle markers (Fig. 5, panels E–T). Although there was some indication of partial colocalization of GFP-ABCA2 fluorescence with endoplasmic reticulum (Fig. 5, panels I–K), Golgi (Fig. 5, panels L–N), and, to a lesser extent, peroxisomes (Fig. 5, panels R–T); mitochondria (Fig. 5, panels O–Q) were negative. The most distinct and significant colocalization was evident with the lysosomal marker LAMP1 (Fig. 5, panels E–H). The distribution of the GFP-ABCA2 fusion protein in HEK/GFP-ABCA2 cells was clearly distinct from that of the GFP alone, which could be seen as intense green fluorescence throughout the transfected cell, including within the nucleus (data not shown). In addition we found that native ABCA2 in stably overexpressing HEK293 colocalizes best with LAMP1, supporting our findings further (Fig. 5, panels V–W).
Previously, we reported a correlation between ABCA2 expression and resistance to estramustine (19). In Fig. 6, a derivative of estramustine with a fluorescently labeled dansyl chloride tag attached to the 17β position of the steroid D ring (23) is partially colocalized with GFP-ABCA2 fluorescence. When GFP-ABCA2 was replaced with native ABCA2, we found essentially total overlap between the subcellular localization of dansyl estramustine and ABCA2. These data suggest that ABCA2 plays a role in the sequestration and eventual cellular efflux of estramustine.
To show the functionality of ABCA2 in stably overexpressing HEK293 cells, we studied the response of HEK/ABCA2 and vector-transfected cells to estramustine. We found increased resistance to estramustine of HEK/ABCA2 cells relative to control cells within drug concentration range from 0 to 10 μm (Fig. 7). The HEK/ABCA2 cells displayed an ∼2-fold resistance to estramustine at the lower drug concentrations. At higher drug concentrations (>7 μm), the difference in cytotoxicity was diminished, perhaps reflecting a saturation of the transport system.
In this study, we isolated human ABCA2 cDNA and subsequently performed molecular and primary structure analysis. This protein is closely related to members of the ABC1-subfamily of transporters; up to 12 members are listed on the web site on ABC transporters.6 Only five members of this subfamily (ABCA1 through -4 and ABCA7) have been completely cloned and characterized for expression and possible function. A partial cDNA (∼50%) for mouse Abca2 has been generated (21) and, premised on an extrapolation of the degree of amino acid identity, ABCA2 can be regarded as orthologous. The estimated topology of ABCA2 is reminiscent of that seen in ABCA4 (26, 28, 29). The large size of this protein (Mr ∼270,000) makes it the largest ABC transporter reported to date. The extracellular loop between the first two transmembrane segments, together with the regulatory domain, account for the size difference compared with other full-size transporters. ABCA2 has a large number of potential sites for both glycosylation and phosphorylation. A total number of 21 putative glycosylation sites were identified. This is the highest number thus far described for any ABC transporter. The phosphorylation of ABCA2 may play a functional regulatory role, because the activity of other ABC transporters such as ABCC7 (CFTR) is modulated by protein kinase A (30), implying a possible similar role for kinases and ABCA2.
The expression of ABCA2 in the lysosome/endosome compartment is unusual among those ABC1-subfamily members thus far characterized. For example, ABCA1 is suggested to localize to Golgi vesicles and plasma membrane (15), whereas ABCA4 is found in the disc membrane of the retinal rod outer segment (31). The only other described mammalian ABC transporter located in lysosomes is ABCA9 (32). Like most of the other intracellular organelle ABC transporters, ABCB9 is a half-transporter, which means that it contains one ABC and TMD and would require a partner to function as a dimer. Thus, ABCA2 is distinct in that it has a full transporter structure (2* ABC + 2*TMD). Furthermore, ABCA2 does not contain the recognizable lysosomal signal sequence, implying that it contains a novel lysosomal targeting sequence yet to be defined.
The merged images of ABCA2 and other cellular organelles indicated partial colocalization was also present in both the endoplasmic reticulum and Golgi. Presumably, these results can be explained by the synthesis and processing of the protein. In particular, the large number of posttranslational sites (particularly glycosylation) may influence its passage through the Golgi. Additionally, ABCA2 showed partial association with peroxisomes. Many of the individual peroxisomal vesicles did not costain with ABCA2, and the overall pattern of merged staining was less compelling than for the endosome/lysosome compartment. Analysis of large numbers of cells indicated that the colocalization of ABCA2 with peroxisomes was less consistent than with endosomes/lysosomes. We confirmed the lysosomal localization of ABCA2 in several cell lines and with different transfection protocols. Whether the transporter has a role in general lysosomal function or facilitates specialized lysosome transport remains to be determined.
The tissue distribution of ABCA2 shows prevalence of expression in the central nervous system, brain, and spinal cord. This is in agreement with the data of Luciani et al., who showed a similar tissue distribution for mouse Abca2 (21). Lower, but significant, levels of expression were also observed in the kidney and liver. Thus far, several ABC transporters have been identified at the blood-brain barrier (e.g., ABCB1 and ABCC1), and their role in the transport of xenobiotics is documented (33, 34). Although the high level of ABCA2 in neuronal tissue implies a functional association, its presence in other tissues also suggests a more general role for this transporter.
In some cases, ABC transporters have been shown to participate in the active transport of a variety of substrates (35, 36). The data from our earlier study on estramustine-resistance clearly linked ABCA2 expression with resistance to this drug in an ovarian carcinoma cell line (19). Other resistance has been shown in this selected cell line, including changes in expression of tubulin isotopes and τ phosphorylation (37). Because acquired resistance is frequently accompanied by multiple cellular adaptations, ABCA2 may be a contributory factor in cell response without providing the entire resistant phenotype. Estramustine is a synthetic nitrogen mustard derivative of estradiol with an unexpected antimitotic activity (38). Thus, although estramustine would not represent a natural substrate for ABCA2, its crystal structure has shown that it maintains a structural component that is identical to the steroid estradiol (39). Another possible clue to the function of ABCA2 is the presence of lipocalin signature motifs in the putative regulatory domain of the protein. Lipocalins are a family of proteins linked to the transport of retinoids, steroids (including cholesterol), lipids, and bilins (27). They are characterized by the presence of eight antiparallel, β-sheet peptide conformations (up to 200 amino acids in length) that form a binding site for hydrophobic substrates. The signature motif resides near the start of the first β-strand. In ABCA2, the lipocalin-binding site would hypothetically span partial regions of both the regulatory (hydrophilic) and membrane (hydrophobic) domains. Perhaps not coincidentally, the locus of some members of this family (Lipocalin 1 and 2) is on 9q34, where ABCA2 resides (19, 21, 40, 41). In addition, ABCA2-related proteins (ABCA1, ABCA4, and ABCA7) are suggested to play a role in lipid transport (42, 43). Thus, it is possible that lipid or lipid-steroid complexes bind to this lipocalin component of ABCA2, facilitating their transport and sequestration into endosomal/lysosomal vesicles. Such vesicular compartments could eventually provide metabolic or secretory pathways for these molecules. The high expression of the transporter in the central nervous system may also bring into consideration such substrates as neurotransmitters and/or bioactive amino acids or peptides as possible endogenous substrates.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported in part by NIH Grants CA06927 and RR05539, by NIH Grant CA53893 (to K. D. T.), and by an appropriation from the Commonwealth of Pennsylvania. Sequence data from this article have been deposited with GenBank and European Molecular Biology Laboratory Data Libraries under Accession No. AF178941.
The abbreviations used are: ABC, ATP-binding cassette; RACE, rapid amplification of cDNA ends; oligo(dT), oligodeoxythymidylic acid; AP, adaptor primer; UTR, untranslated region; ORF, open reading frame; HHD, highly hydrophobic domain; TMD, transmembrane domain; DIC, differential interference contrast.
Internet address: http://www.expasy.ch/tools/scnpsit1.html.
Internet address: http://www.med.rug.nl/mdl/humanabc.htm.
|Cell line .||ABCA2 .|
|Cell line .||ABCA2 .|
Data are in arbitrary units standardized to the housekeeping gene 36B4.
We thank Drs. Srdjan Askovic and Rebecca Raftogianis for helpful discussion throughout this work. We also thank Dr. Takahiro Nagase for providing the KIA1069 clone and Pat Kraus for her assistance in assembling this manuscript.