INTRODUCTION

Dopaminergic neurotransmission is terminated by the activity of presynaptic dopamine transporters (DATs)¹ that utilize the energy of electrochemical gradients to transport extracellular dopamine into the neuron (Horn, 1990). DAT is an 80-kDa integral membrane glycoprotein containing sialic acids and N-linked carbohydrates (Grigoriadis et al., 1989; Sallee et al., 1989; Berger et al., 1991; Lew et al., 1991). Molecular cloning of this protein from rat, bovine, and human brain reveals a predicted primary sequence of 619-620 amino acids with 12 hydrophobic stretches suitable for transmembrane-spanning regions, numerous consensus phosphorylation sites, and three to five consensus N-glycosylation sites (Shimada et al., 1991; Kilty et al., 1991; Giros et al., 1992; Vandenbergh et al., 1992; Usdin et al., 1991). The working model for the topological orientation of the protein places the N and C termini intracellularly, three or four of the glycosylation sites extracellularly, and several consensus phosphorylation sites intracellularly (Fig. 1).

The psychotropic drugs cocaine and amphetamine inhibit dopamine reuptake, and the binding of cocaine by DAT is presumed to underlie its reinforcing effects (Ritz et al., 1987). Numerous other structurally diverse compounds such as N-[1-(2-benzo(b)thiophenyl)cyclohexyl]-piperidine, mazindol, and GBR compounds are also high affinity dopamine uptake blockers and are competitors in DAT ligand binding assays. Although the binding properties of these compounds at DAT have been extensively characterized (Berger et al., 1990; Madras et al., 1989; Maurice et al., 1991; Pristupa et al., 1994; Reith et al., 1992), the molecular basis of their action as binding and transport antagonists is not understood. Elucidation of these properties at a molecular level could yield insight into the basis of DAT action and provide information useful for the development of therapeutic cocaine treatments.

Site-directed mutagenesis studies have identified specific amino acid residues important for binding of the cocaine analog [³H]CFT in TMD 4 (Wang et al., 1993), for dopamine transport and [³H]CFT binding in TMD 1 (Kitayama et al., 1992), and for transport of the neurotoxin 1-methyl-4-phenylpyridinium in TMD 11 (Kitayama et al., 1993). Functional studies performed using chimeras of dopamine and norepinephrine transporters suggest that the N-terminal region of the protein (TMDs 1-4) provides an essential transport activity, that the middle third of the protein (TMDs 5-8) functions in cocaine binding, and that more C-terminal regions (TMDs 11-12) are involved in 1-methyl-4-phenylpyridinium transport (Giros et al., 1994; Buck and Amara, 1995).

DAT functional domains have also been studied using photoaffinity ligands based on two of the aforementioned classes of uptake blockers, [¹²⁵I]DEEP, a GBR analog, and [¹²⁵I]RTI 82, a cocaine analog. Epitope-specific immunoprecipitation of proteolytic fragments generated from SDS-solubilized DATs showed that [¹²⁵I]DEEP was incorporated near the first two putative transmembrane domains, whereas [¹²⁵I]RTI 82 was localized to the C-terminal half of the protein (Vaughan, 1995).

The present report describes an extension of this earlier peptide mapping study modified to investigate the relationship of the photolabeled domains to transmembrane structures. This was done by performing the proteolytic treatment on membrane suspensions rather than solubilized DAT samples and separately analyzing the resulting membrane and supernatant fractions for photolabeled DAT fragments. Proteolytic fragments containing integral membrane structures will be retained by the membranes, whereas those lacking membrane structures will be rendered soluble and will be released into the supernatant (Herald et al., 1994). The data presented in this report demonstrate that proteolyzed DAT domains containing either [¹²⁵I]DEEP or [¹²⁵I]RTI 82 remain in the membrane and thus contain integral membrane structures. The small size of two of the fragments indicates that the photoaffinity ligands are incorporated in, or in close proximity to, these transmembrane structures and verifies the position of at least one transmembrane domain in each of these fragments in terms of its maximum possible distance from the incorporated ligand and antibody epitope. While the site of incorporation of the ligands within the primary sequence was consistent with the findings of the previous study, the methodological differences resulted in the production of two novel fragments that yielded additional information relative to the site of [¹²⁵I]RTI 82 incorporation and to sites of glycosylation. It was also found that the two larger DAT tryptic fragments retained the ability to bind photoaffinity ligands and pharmacological displacers.

EXPERIMENTAL PROCEDURES

Rat striatal membranes were homogenized in 10 m sodium phosphate buffer, pH 7.4, containing 0.32  sucrose, centrifuged at 20,000 × g for 12 min, and resuspended in the same buffer at a concentration of 6.7 mg/ml original wet weight as described previously (Vaughan, 1995). [¹²⁵I]DEEP, or [¹²⁵I]RTI 82 (specific activity 1100-1800 Ci/mmol) was added to a final concentration of 5 n, the suspensions were incubated on ice for 60 min, and then irradiated with ultraviolet light for 45 s. Membranes were washed twice with 50 m Tris-HCl, pH 8.0, and resuspended in this buffer at a final concentration of 50 mg/ml (original wet weight).

25 µl of the membrane suspension prepared as above was treated for 5 min at 22 °C with 25 µl of trypsin prepared in 50 m Tris-HCl, pH 8.0. Final trypsin concentrations ranged from 1 to 200 µg/ml as indicated specifically in the figure legends. At the end of the incubation, soybean trypsin inhibitor was added to equal the concentration of trypsin, and samples were centrifuged at 20,000 × g for 20 min (Figs. 2 and 3) or 10,000 × g for 5 min (remainder of experiments). Supernatants were removed for analysis or were discarded, and the membrane pellets were solubilized with 25 µl of SDS-PAGE sample buffer (60 m Tris-HCl, pH 6.8, 10% glycerol, 10 m dithiothreitol, and 0.5% SDS). Solubilized membranes were analyzed by electrophoresis and autoradiography, either directly or after immunoprecipitation.

For some experiments, unlabeled membranes were treated with trypsin followed by photolabeling. For these assays, membranes were treated with trypsin exactly as described above, and the supernatants were discarded. The trypsin-treated membranes were resuspended in the sucrose-phosphate buffer followed by the photolabeling and immunoprecipitation procedures.

The epitope-specific antisera used in this study were generated against the following regions of the DAT-deduced amino acid sequence: peptide 15, amino acids 6-30; peptide 16, amino acids 42-59; peptide 5, amino acids 225-236; peptide 12, amino acids 541-550; peptide 18, amino acids 580-608 (Fig. 1). The sera were coded with the same number as the immunizing peptide and were determined to be specific for DAT based on enzyme-linked immunosorbent assay, immunoblots, immunoprecipitations, and immunohistochemistry as described elsewhere (Vaughan et al., 1993; Freed et al., 1995; Vaughan, 1995).

For immunoprecipitations, SDS-solubilized membranes were diluted to a final SDS concentration of 0.1% using immunoprecipitation buffer (phosphate-buffered saline, pH 7.4, plus 0.05% SDS and 0.1% Triton X-100). To avoid loading large amounts of IgG onto the gels, the DAT antisera were covalently coupled to the protein A-Sepharose beads using the cross-linking reagent dimethyl pimelimidate prior to use in immunoprecipitations (Schneider et al., 1982). Each sample of diluted, solubilized membranes was incubated with 30 µl of precoupled protein A-Sepharose beads for 3 h at 4 °C; the beads were washed twice with immunoprecipitation buffer, and DAT samples were eluted from the beads with SDS-PAGE sample buffer. Samples were analyzed on 15% polyacrylamide gels unless otherwise indicated. For peptide blocking experiments precoupled beads were preincubated for 15 min with either no addition, the immunizing peptide, or an irrelevant peptide prior to adding the DAT sample. The irrelevant peptides used for those experiments were either peptide 4 (amino acids 162-172) or peptide 11 (amino acids 505-516) (Vaughan et al., 1993). High and low range Rainbow molecular weight markers from Amersham Corp. were standards on all gels. All immunoprecipitation results reported were produced in two or more independent experiments.

Photolabeled, trypsinized samples were immunoprecipitated as described through the step of washing the protein A-Sepharose beads. At that point the beads were incubated with 30 µl of 25 m Tris glycine, pH 6.8, 0.1% SDS, 0.7% Nonidet P-40, and 5 m EDTA, with or without the addition of 0.3 units of N-glycanase and/or 0.01 unit of neuraminidase. For the coincubation treatments, the neuraminidase treatment was done first for 60 min followed by the addition of the N-glycanase. Samples were incubated overnight at 22 °C, followed by elution of DAT samples from the beads with SDS-PAGE sample buffer and analysis by electrophoresis and autoradiography.

[¹²⁵I]DEEP and [¹²⁵I]RTI 82 were synthesized by Dr. F. I. Carrol, Research Triangle Park, NC, and radioiodinated by Dr. John Lever, Johns Hopkins University, as described previously (Grigoriadis et al., 1989; Patel et al., 1992). Trypsin was from Boehringer Mannheim; soybean trypsin inhibitor was from Sigma; protein A-Sepharose CL 4B was from Pharmacia Biotech Inc.; dimethyl pimelimidate was from Pierce, and N-glycanase and neuraminidase were from Genzyme. GBR 12935 was obtained from Research Biochemicals, and mazindol and ()-cocaine were obtained from NIDA (Baltimore, MD).

RESULTS

The effect of trypsin treatment on [¹²⁵I]DEEP-labeled membranes is shown in Fig. 2. Lanes 1-3 contain control or trypsin-treated membranes, and the corresponding supernatants from these samples are in lanes 4-6. Most of the radioactivity remained in the membrane fraction, and only a negligible amount of radioactivity was released from the membranes into the supernatants. Some proteolytic degradation and redistribution of radiolabel from higher to lower molecular weight forms can be observed.

Although the identity of photolabeled DAT in these preparations is known from pharmacological and immunological evidence (Grigoriadis et al., 1989; Patel et al., 1992; Vaughan et al., 1993), the presence of many nonspecifically photolabeled proteins made it necessary to immunoprecipitate the samples to identify the DAT proteolytic fragments. Immunoprecipitation of the membranes and supernatants in lanes 1-6 with DAT antiserum 16 is shown in lanes 7-12. Both unproteolyzed full-length DAT (lane 7) and 45- and 14-kDa DAT fragments (lanes 8 and 9) were quantitatively recovered from membranes, whereas no fragments were detected in the soluble fractions (lanes 10-12). The proteolytic pattern shown here was extremely reproducible. Treatment of membranes with 1 µg/ml trypsin generated approximately equal amounts of full-length DAT and the 45-kDa fragment (lane 8), whereas at 10 µg/ml trypsin, most of the full-length DAT was gone, and the radioactivity was distributed relatively equally between the 45- and 14-kDa fragments (lane 9; see also Fig. 4). Further characterization of these fragments is presented in Figs. 4 and 5.

Tryptic proteolysis of [¹²⁵I]RTI 82-labeled membranes (Fig. 3) produced a comparable result in that, while degradation of photolabeled proteins was apparent in the membrane samples (lanes 1-3), essentially no radioactivity was released into the supernatants (lanes 4-6). Immunoprecipitation of the samples in lanes 1-6 with DAT antiserum 5 (lanes 7-12) shows that DAT and DAT fragments were recovered exclusively from membranes (lanes 7-9). The fragments generated from [¹²⁵I]RTI 82-labeled DAT migrated at about 32 and 16 kDa and are characterized in Figs. 6 and 7.

Full trypsin dose-response curves and time course studies were performed for DATs labeled with each ligand, and only the optimal conditions are shown. At these trypsin concentrations proteolysis was complete in 5 min. The supernatants from the trypsin-treated membranes labeled with both ligands were analyzed with all five of the epitope-specific antisera, but no immunoprecipitable photolabeled fragments were found. Proteolysis of membrane preparations was also examined using V8 protease. While this was the optimal enzyme in the previous study for the SDS-solubilized DATs, it was substantially less effective than trypsin in the membrane suspensions.

Because the DAT fragments were found entirely in the membrane fraction, subsequent experiments used only membranes from which the tryptic supernatants had been removed. Although the [¹²⁵I]DEEP and [¹²⁵I]RTI 82 fragments generated by V8 protease from SDS-solubilized DATs were immunologically characterized in the previous study, this analysis was re-done for the present study. This was warranted because of the methodological differences between the studies, because the sizes of the proteolytic fragments differed somewhat between the studies, and because [¹²⁵I]RTI 82 fragments were generated that were smaller than in the previous study.

Immunological characterization of the [¹²⁵I]DEEP fragments generated at two concentrations of trypsin is shown in Fig. 4. The intact DATs remaining after proteolysis (that migrate at about 97 kDa on these high concentration gels) were immunoprecipitated by all five antisera (lanes 1-5 and 7-11) but not by preimmune serum (lanes 6 and 12). The 45-kDa fragment was precipitated with sera 15 and 16 (lanes 1, 2, 7, and 8) but not by the others, and the 14-kDa fragment was precipitated exclusively by serum 16 (lane 8). A lightly labeled fragment at about 32 kDa was precipitated by sera 5, 12, and 18 (lanes 3-5 and 9-11) but not by sera 15 and 16. Precipitation of all fragments with the indicated sera is specific because none of the fragments was recognized by preimmune serum (lanes 6 and 12), and because for all fragments, precipitation was blocked by the inclusion of the immunizing peptide (Fig. 5, center lane of each set) but not an irrelevant peptide (Fig. 5, right lane of each set). Occasionally other minor fragments (e.g. at about 26 kDa) were observed, but their appearance was not reproducible and their characterization was not pursued.

Analysis of [¹²⁵I]RTI 82 tryptic fragments is shown in Fig. 6. Again, the intact DATs precipitated with all immune but not preimmune sera. The fragment at 32 kDa precipitated with sera 5, 12, and 18 (lanes 3-5, and 9-11) but not with sera 15 and 16, whereas the 16-kDa fragment was recognized only by serum 5 (lanes 3 and 9). None of these fragments was precipitated with preimmune serum (lanes 6 and 12), and inclusion of the immunizing peptide (Fig. 7, center lane of each set) but not an irrelevant peptide (Fig. 7, right lane of each set) blocked precipitation of these fragments. Other minor fragments that were not characterized were also observed in these experiments.

The photolabeled fragments were analyzed for the presence of carbohydrates by treatment with N-glycanase and neuraminidase, which cleave asparagine-linked oligosaccharides and terminal sialic acids, respectively. [¹²⁵I]DEEP and [¹²⁵I]RTI 82-labeled DAT fragments were immunoprecipitated and then treated with these enzymes (Fig. 8). The full-length DATs in these samples (lanes 1 and 5) lose about 25 kDa after N-glycanase treatment (lanes 2 and 6) and about 5 kDa after treatment with neuraminidase (lanes 3 and 7). Use of both enzymes together produces a band that migrates at about 50 kDa (lanes 4 and 8). The 45-kDa [¹²⁵I]DEEP-labeled fragment underwent parallel reductions in molecular mass with these treatments (lane 1-4), resulting in a band that migrated at 25-30 kDa after deglycosylation with both enzymes. The electrophoretic mobilities of the 14-kDa [¹²⁵I]DEEP-labeled fragment and the 32-kDa [¹²⁵I]RTI 82-labeled fragment were unaffected.

Assessment of the functional properties of the proteolytic fragments was done by performing photoaffinity labeling after trypsin treatment of membranes (Fig. 9). Unlabeled membranes were treated with 10 µg/ml trypsin for subsequent [¹²⁵I]DEEP labeling or 20 µg/ml trypsin for subsequent [¹²⁵I]RTI 82 labeling. After removal of supernatants, the membranes were resuspended and subjected to photoaffinity labeling followed by immunoprecipitation with serum 16 for [¹²⁵I]DEEP-labeled samples or serum 5 for [¹²⁵I]RTI 82-labeled samples. Pre-photolabeled membranes were treated with trypsin, immunoprecipitated, and electrophoresed in parallel with the experimental samples.

DATs pre-labeled with [¹²⁵I]DEEP (lane 1) produced the characteristic 45- and 14-kDa fragments when treated with 10 µg/ml trypsin (lane 2). When unlabeled membranes received the same trypsin treatment followed by photolabeling, both full-length DAT and the 45-kDa fragment became labeled (lane 3). The 14-kDa fragment did not become photolabeled even though it was present as determined by the pre-photolabeled sample. Although the labeling of the 45-kDa fragment relative to the intact DAT form is modest in this experiment, in most other experiments it labeled more strongly (e.g. Fig. 10).

The parallel analysis using [¹²⁵I]RTI 82 to label proteolyzed DAT is shown in lanes 4-9. Membranes pre-photolabeled with [¹²⁵I]RTI 82 and then treated with trypsin produced the previously described 32- and 16-kDa fragments (lanes 4-6). Unlabeled membranes treated with trypsin and then subjected to photolabeling with [¹²⁵I]RTI 82 produced labeled full-length DAT and the 32-kDa fragment (lanes 7-9). The 16-kDa fragment was not observed to become labeled after proteolysis, even upon long exposures of the films.

Further analysis of the function of these fragments involved examination of their pharmacological properties (Fig. 10). Photoaffinity labeling of tryptic fragments with [¹²⁵I]DEEP (lanes 1-4) or [¹²⁵I]RTI 82 (lanes 5-8) was performed in the presence of no displacer (lanes 1 and 5), 30 µ ()-cocaine (lanes 2 and 6), 1 µ GBR 12935 (lanes 3 and 7), or 1 µ mazindol (lanes 4 and 8). These compounds are dopamine transport inhibitors and block photoincorporation of both [¹²⁵I]DEEP and [¹²⁵I]RTI 82 at DAT (Grigoriadis et al., 1989; Patel et al., 1992). In the control samples, the 45-kDa fragment became labeled with [¹²⁵I]DEEP (lane 1), and the 32-kDa fragment became labeled with [¹²⁵I]RTI 82 (lane 5). Inclusion of any of the displacers during binding blocked photolabeling of the remaining full-length DATs and both of the fragments (lanes 2-4 and 6-8).

DISCUSSION

This study identifies tryptic domains of dopamine transporters photoaffinity labeled with two different ligands, [¹²⁵I]DEEP, a GBR analog, and [¹²⁵I]RTI 82, a cocaine analog. [¹²⁵I]DEEP was found in two fragments of 45 and 14 kDa that were both recognized by serum 16 generated against an epitope near putative TMD1, whereas [¹²⁵I]RTI 82 was found in two fragments of 32 and 16 kDa that were both precipitated with serum 5 directed against an epitope near putative TMD4. N-Linked oligosaccharides were found exclusively on the 45-kDa [¹²⁵I]DEEP-labeled fragment. The location of these fragments relative to antibody epitopes and predicted transmembrane helices is shown schematically in Fig. 11.

While the N and C termini of the fragments cannot be unequivocally determined from these data, lysine and arginine residues (potential trypsin sites) are present in regions of the protein that could produce fragments of the appropriate sizes and characteristics. The 45-kDa N-terminal fragment and the 32-kDa C-terminal fragment are probably produced by cleavage of DAT at Arg-218 and/or Arg-227. These are the only potential trypsin sites present between the consensus glycosylation sites and epitope 5, which would be required to produce a glycosylated N-terminal fragment not recognized by serum 5 and a nonglycosylated C-terminal fragment recognized by serum 5. Arg-227 is just within the boundary of epitope 5, but cleavage here might still leave sufficient immunoreactivity in this epitope to permit precipitation of the C-terminal fragments by serum 5. High protease sensitivity at Arg-218 and/or Arg-227 indicates that under nondenaturing conditions these residues are particularly exposed at the surface of the protein. The immunoreactivity of the 45-kDa fragment to serum 15 and the 32-kDa fragment to serum 18 indicates that little or no proteolysis is occurring at the DAT N or C termini. Thus, the 45-kDa [¹²⁵I]DEEP-labeled fragment represents most or all of the N-terminal half of DAT, and the 32-kDa [¹²⁵I]RTI 82-labeled fragment represents most or all of the C-terminal half.

Reduction of the 45-kDa [¹²⁵I]DEEP-labeled fragment to the 14-kDa form could be occurring by N-terminal proteolysis at residues Lys-19, Arg-27, or Arg-35, which would remove the serum 15 epitope, and C-terminally at Arg-125, Lys-133, or Lys-139, which would remove the region containing the four consensus glycosylation sites between TMDs 3 and 4. Lys-139 is just N-terminal to putative TMD3, and as there are no other lysines or arginines between Lys-139 and the glycosylation sites, the 14-kDa fragment can contain at most only putative TMDs 1 and 2.

Since the 32- and 16-kDa [¹²⁵I]RTI 82-labeled fragments are both recognized by serum 5, the reduction of the larger to the smaller form does not involve proteolysis of the N terminus, which would remove the antibody epitope. Cleavage of the C-terminal end of the fragment could be occurring at any of the numerous residues that would remove epitopes 12 and 18, possibly at Lys-373 or Arg-379. Proteolysis here would generate fragments of about 16 kDa that would contain predicted TMDs 4-7. It was noted in this study as well as the previous peptide mapping study that [¹²⁵I]RTI 82-labeled DATs were more resistant to proteolysis than [¹²⁵I]DEEP-labeled DATs, requiring 20-200 times as much enzyme to produce fragmentation. This raises the possibility that DATs labeled with different photoaffinity compounds assume different conformations resulting in differential protease sensitivities or that the two ligands differentially protect protease sites.

A very small fraction of [¹²⁵I]DEEP became incorporated into the 32-kDa domain labeled by [¹²⁵I]RTI 82. This may be due to binding of a small fraction of [¹²⁵I]DEEP either in a totally or a partially nonoverlapping binding site compared with the [¹²⁵I]DEEP found in the N-terminal fragments, or perhaps it represents incorporation of [¹²⁵I]DEEP reaction intermediates formed during photoactivation. Although the explanation cannot be determined without further experimentation, the latter two scenarios are compatible with the interpretation that, during binding, the two different regions of DAT labeled with [¹²⁵I]DEEP are in close three-dimensional proximity.

The incorporation sites of the two ligands within the primary protein sequence are consistent with the results of the previous peptide mapping study (Vaughan, 1995), although there were some differences between the two studies in the sizes of the fragments produced. In the previous study [¹²⁵I]DEEP was found in 10- and 7-kDa fragments that immunoprecipitated with serum 16 and a 4-kDa nonprecipitable peptide presumed to be a degradation product of these fragments. These fragments are thus contained within the boundary of the 14-kDa [¹²⁵I]DEEP-labeled region identified in the present study. Conversely, the smallest analyzable [¹²⁵I]RTI 82-labeled fragment obtained previously was a 34-kDa fragment immunologically localized by sera 5, 12, and 18 to the C-terminal half of DAT. The generation of the 16-kDa [¹²⁵I]RTI 82-labeled fragment in the present study and its localization to the central region of the protein near TMD 4-7 thus delineates the [¹²⁵I]RTI 82 incorporation site much more precisely than was previously possible.

Multiple lines of direct and indirect evidence indicate that incorporation of the photoaffinity ligands is occurring in, or in close proximity to, integral membrane structures. First, both the 14-kDa [¹²⁵I]DEEP-labeled fragment and the 16-kDa [¹²⁵I]RTI 82-labeled fragment are retained in membranes and are not released as soluble forms, indicating that they possess integral membrane structure. The fragments were retained in membranes after as much as an hour of trypsin treatment involving multiple episodes of vortexing. However, brief exposure of trypsinized membranes to 1% CHAPS resulted in the release of the fragments into supernatants that could be separated from remaining CHAPS-insoluble material, demonstrating that the fragments possess the characteristics of integral membrane proteins that require detergents for solubilization.

The small size of the fragments in conjunction with their retention in the membrane suggests that incorporation of the ligands is occurring in or near integral membrane structures. Although the data presented here cannot distinguish between incorporation of the ligands in membrane structures or in adjacent hydrophilic regions, it is possible to calculate the maximum distance that could be occurring between ligand incorporation sites and transmembrane structures. Using an average amino acid mass of 110 Da, it can be estimated that fragments of 14-16 kDa would contain approximately 125-145 amino acids, of which 20-22 would be present in a transmembrane spanning helix. Thus, if a fragment contained a single TMD at its extreme end, ligand incorporation could be occurring as much as 100-120 residues away. However, if the TMD is positioned centrally and/or if more than one TMD is present in the fragment, then the photolabeled site and membrane domain must be much closer together. For [¹²⁵I]DEEP this is very likely to be the case if the predicted model of DAT is correct. In the previous study it was found that serum 16 immunoprecipitated a 7-kDa [¹²⁵I]DEEP-labeled fragment. Based on size this fragment would be predicted to contain TMDs 1 and 2, the six amino acids linking the TMDs, and part of the N-terminal region containing the antibody 16 epitope. Subsequent proteolysis of the 7-kDa fragment to the 4-kDa form results in the loss of the antibody epitope, producing a labeled fragment whose size is consistent with a region containing at most the two TMDs and the linker or one of the TMDs and some flanking structure. Thus, if the proposed model is correct, these data indicate that [¹²⁵I]DEEP incorporation is occurring in one of the TMDs or in the six-amino acid linker. Although most of the specifics of DAT structure in this region remain hypothetical, this interpretation is supported by the data produced in these two studies that show that the 7- and 4-kDa fragments as well as at least one TMD are present within the 14-kDa [¹²⁵I]DEEP-labeled membrane-bound fragment.

A final observation relevant to this point is that even very low molecular weight photolabeled fragments were not observed to be released from trypsin-treated membranes. It is possible that photolabeled fragments smaller than 14-16 kDa were produced in these experiments, but they could not be detected by immunoprecipitation due to proteolytic loss of the antibody epitopes. Such fragments not possessing any integral membrane structure would be released from the membrane as soluble forms and might be visible in the supernatants in Figs. 2 or 3 even if they could not be immunoprecipitated. The resolution of these gels down to 3.5 kDa or less indicates that proteolyzed fragments containing as few as 25-30 amino acids (2.8-3.3 kDa) might be visible. Fragments this small containing each photolabel were generated in the previous mapping study, and the DAT sequence contains many potential trypsin sites whose use could produce fragments of this size. The lack of fragment release into supernatants suggests that any nonimmunoprecipitable fragments smaller than 14-16 kDa were retained in membranes where they could not be visualized because of the high background noise, and further indicates that binding sites are located very close to transmembrane structures.

Enzymatic deglycosylation with neuraminidase and N-glycanase demonstrated that both sialic acids and N-linked oligosaccharides were present exclusively on the 45-kDa [¹²⁵I]DEEP-labeled fragment immunoprecipitated by serum 16. Based on apparent mass, this fragment would be expected to contain the consensus N-glycosylation sites at positions 182, 188, 196, and 204 between TMDs 3 and 4, as well as a consensus site at Asn-44 prior to TMD1. However, Asn-44 is not used for glycosylation, as this residue is within the serum 16 epitope, and the 14-kDa [¹²⁵I]DEEP fragment immunoprecipitated by serum 16 is not affected by N-glycanase. Deglycosylation of the 45-kDa but not the 14-kDa [¹²⁵I]DEEP-labeled fragment indicates that DAT is glycosylated in the region between TMDs 3 and 4, presumably at one or more of the consensus glycosylation sites, and shows that sialic acids and N-linked sugars are present in the same region of the protein. Use of analogous consensus sites for N-glycosylation has been demonstrated for heterologously expressed serotonin and glycine transporters (Tate and Blakely, 1994; Olivares et al., 1995). The ~25-kDa apparent molecular mass of the deglycosylated fragment is compatible with C-terminal cleavage of this fragment at Arg-218 or Arg-227 as previously discussed.

The proposed model for DAT structure and topology is based on theoretical considerations of hydrophobicity values, lack of signal sequence, and consensus glycosylation and phosphorylation sites (Shimada et al., 1991; Kilty et al., 1991). The data presented here experimentally verify several aspects of this model. First, this study demonstrates the presence of at least one integral membrane structure in two different regions of the protein predicted to contain TMDs and delineates the approximate maximum possible distance between the antibody epitope and transmembrane structures in each fragment. By using the same logic applied to the estimate of distance between TMDs and ligand incorporation sites (i.e. estimating 125-145 residues per fragment, 20-22 residues per TMD, and 15-18 residues per epitope), it can be calculated that a maximum of about 80-110 residues can be present between the end of the antibody epitope and the beginning of the closest transmembrane structure in each fragment. Again, this is a maximum allowable distance, and the TMD(s) could be situated more closely than this to the epitope and/or more than one TMD could be present in each fragment.

Second, the documentation of carbohydrates in the region between putative TMDs 3 and 4 indicates that this loop is oriented extracellularly as predicted and provides a marker for refinement of the topology of the 45-kDa [¹²⁵I]DEEP-labeled fragment. Light and electron microscopic histochemical evidence indicates that epitope 16 is present on the intracellular side of the plasma membrane (Nirenberg et al., 1996).² Since the glycosylation sites are almost certain to be oriented extracellularly, this precludes the possibility of an even number of TMDs in this fragment and indicates that the polypeptide sequence of this region must traverse the membrane an odd number of times, either once or, as predicted, three times.

Functional analysis of the fragments using photoaffinity labeling showed that while the 45-kDa fragment precipitated by serum 16 was capable of binding [¹²⁵I]DEEP, the 14-kDa fragment was devoid of this capacity. Similarly, of the two fragments recognized by serum 5, the 32-kDa fragment but not the 16-kDa fragment was capable of binding [¹²⁵I]RTI 82. These results are subject to at least two possible interpretations. One is that the larger fragments independently possess the capacity for binding and incorporation of their respective photolabels (i.e. the domains of protein lost due to proteolysis do not participate in these functions), and subsequent proteolysis of the larger fragments to the smaller forms eliminates this property. Alternatively, although proteolysis severs the DAT primary sequence, it is possible that the larger N- and C-terminal fragments remain in close proximity due to noncovalent forces, and the adjacent regions of the protein continue to contribute determinants or tertiary stability necessary for binding.

The latter scenario is compatible with the pharmacological properties of the fragments, which were indistinguishable from full-length DAT. If GBR compounds and [¹²⁵I]DEEP require binding determinants found only in the the 45-kDa N-terminal half of the protein, and cocaine and [¹²⁵I]RTI 82 require binding determinants found only in the the 32-kDa C-terminal half of the protein, then pharmacological displacement of the photolabels by GBR 12935 and cocaine would have been different for the fragments than for the full-length DAT, i.e. cocaine would not displace [¹²⁵I]DEEP binding to the 45-kDa fragment and GBR 12935 would not displace [¹²⁵I]RTI 82 binding to the 32-kDa fragment. The finding that all of the displacers examined inhibit photolabeling of both the N- and C-terminal fragments suggests the two halves of the protein remain in close association.

At least two mechanisms can be envisioned to explain this reciprocal effect. First, it is possible that ligand binding pockets are generated by the juxtaposition of amino acids present in widely separated regions throughout the primary sequence, and thus both halves of the protein are required for the binding of photolabels and displacers. Site-directed mutagenesis studies suggest this type of mechanism, as alterations in [³H]CFT binding are produced by point mutations of residues in TMDs 1 and 4 (Kitayama et al., 1993; Wang et al., 1993). A second possibility compatible with these data is that ligand binding, in fact, occurs only in a discrete region within one or the other fragment, but because the two halves of the protein remain associated, binding-induced conformational changes in one domain are conveyed to the adjacent region, resulting in reciprocal inhibition of photolabeling. Further analysis of this phenomenon might be achieved using alternative experimental approaches such as detergent solubilization of DAT fragments prior to photolabeling or expression and co-expression of DAT fragments.

These results extend our knowledge of DAT structure-function properties by demonstrating that binding of uptake-blocking compounds occurs in, or in close proximity to, integral membrane structures. While some integral membrane proteins such as glutamate receptors possess binding sites on extracellular domains well-separated from transmembrane stretches (Stern-Bach et al., 1994), localization of binding determinants to transmembrane spanning helices has been documented for other proteins such as 7TMD receptors (Strader et al., 1994). For example, in -adrenergic receptors, deletion of transmembrane domains results in abnormal ligand binding (Dixon et al., 1987a, 1987b), and several charged and polar residues within the TMDs serve as contact points for agonists and antagonists (Strader et al., 1987, 1988). In the multidrug resistance protein, MDR1, an ATP-dependent, 12-transmembrane domain drug transporter, transport-inhibiting photoaffinity ligands are incorporated in transmembrane domains or in closely positioned linker regions between the TMDs (Bruggemann et al., 1989; Greenberger, 1993; Zhang et al., 1995).

Identification of different sites of incorporation for [¹²⁵I]DEEP and [¹²⁵I]RTI 82 suggests that these compounds interact at DAT in at least partially non-overlapping binding sites and indicates that these two classes of uptake blockers may exert their effects by differing mechanisms. Nonidentical modes of action of various classes of uptake blockers have been suggested by numerous transport and binding studies (Berger et al., 1990; Madras et al., 1989; Maurice et al., 1991; Pristupa et al., 1994; Reith et al., 1992; Eshleman et al., 1994). The two general regions of DAT that incorporate photoaffinity ligands have been shown by mutational and chimeric approaches as being important for both dopamine transport and cocaine analog binding (Kitayama et al., 1992; Wang et al., 1993; Giros et al., 1994; Buck and Amara, 1995), and TMDs 1-2 and 4-8 also display the highest sequence conservation among Na⁺- and Cl-dependent neurotransmitter transporters, implicating by homology the functional importance of these regions (Amara and Kuhar, 1993). The finding that [¹²⁵I]DEEP is incorporated in both of these regions of the protein is compatible with the hypothesis that during ligand binding these domains are in close three-dimensional proximity, perhaps forming a ligand binding pocket. This type of interpretation has also been reached for the MDR1 protein in which photoaffinity ligand incorporation occurs in two widely separated regions of the primary sequence (Bruggemann et al., 1989; Greenberger 1993).

This report also experimentally verifies several aspects of DAT structure and topology predicted from theoretical considerations, including identification of at least one transmembrane domain in two different regions of the protein, and positive identification of the region containing N-linked oligosaccharides and sialic acids. The finding that DAT N- and C-terminal fragments maintain the ability to become photoaffinity labeled and that photolabeling of both fragments is blocked in a similar fashion by a variety of pharmacological displacers suggests that the domains remain associated after proteolysis and suggests testable hypotheses concerning the mode of interaction between the two domains. These results thus extend our knowledge of the molecular and biochemical basis of binding of DAT uptake blockers and provide additional framework for further structure-function analysis of DAT. Because of the high degree of structural and functional homology between DAT and other members of the Na⁺- and Cl-dependent transporter family, it is likely that many of the results presented here will be generally applicable to these other important neuronal proteins.