The Structural and Functional Units of Heteromeric Amino Acid Transporters

Heteromeric amino acid transporters are composed of a catalytic light subunit and a heavy subunit linked by a disulfide bridge. We analyzed the structural and functional units of systems b0,+ and x -C, formed by the heterodimers b0,+AT-rBAT and xCT-4F2hc, respectively. Blue Native gel electrophoresis, cross-linking, and fluorescence resonance energy transfer in vivo indicate that system b0,+ is a heterotetramer [b0,+AT-rBAT]2, whereas xCT-4F2hc seems not to stably or efficiently oligomerize. However, substitution of the heavy subunit 4F2hc for rBAT was sufficient to form a heterotetrameric [xCT-rBAT]2 structure. The functional expression of concatamers of two light subunits (which differ only in their sensitivity to inactivation by a sulfhydryl reagent) suggests that a single heterodimer is the functional unit of systems b0,+ and x -C.

band of 250 kDa (4). However, neither the native structure nor the functional unit of the HAT is known. Here we present evidence indicating that the single heterodimer is the functional unit of systems b 0,ϩ and x C Ϫ . 4F2hc-associated heterodimers seem not to form stable oligomers, whereas b 0,ϩ AT-rBAT is a heterotetramer in vivo. Finally, we demonstrate that rBAT promotes the oligomerization of HAT.

EXPERIMENTAL PROCEDURES
The construction of the distinct cDNAs is described under supplemental "Experimental Procedures." Reagents and Antibodies-Reagents were obtained from Sigma if not indicated otherwise. Antibodies against human and mouse b 0,ϩ AT and rBAT are described elsewhere (4,24,25). The anti-Xpress antibody was purchased from Invitrogen, and the anti-Myc 9E10 hybridoma was from ATCC. The rabbit polyclonal antibody against mouse LAT2 was produced at Research Genetics. The antigenic peptide was PIFKPTPVKD-PDSEEQP (the C-terminal 17 residues). Specificity of this antibody was tested by comparing the signal in Western blot with the preimmune antisera obtained from the same rabbit (data not shown).
cRNA Synthesis, Injection, and Maintenance of Xenopus Oocytes-The synthesis of human 4F2hc cRNA has been described elsewhere (26). In vitro synthesis of human xCT wild type, xCT C327S, and xCT concatamers was conducted with the NotI-linearized plasmid template using the in vitro transcription protocol from AMBION (mMESSAGE mMACHINE, Ambion, Austin, TX). Mixtures of cRNAs were prepared immediately before injection with a calibrated pipette. The amount of transcribed RNA was calculated by 260-nm absorbance measurement before microinjection into Xenopus oocytes. Each of the cRNA species was synthesized at least on two occasions. In mixing experiments, oocytes were injected with 4F2hc cRNA (5 ng) together with xCT Wt (5 ng), xCT C327S (5 ng), or with a mixture of xCT Wt and xCT C327S (total amount of 5 ng of xCT cRNAs). For the concatamers, 25 ng of the corresponding cRNA together with 5 ng of 4F2hc cRNA were injected per oocyte.
The procedures for injection and maintenance of oocytes have been described in detail elsewhere (27). Oocytes were injected with either 50 nl of water or 50 nl of water containing the cRNAs. Oocytes were incubated in modified Barth's solution, and the experiments were performed 2-4 days after injection.
Cell Culture and Transfection-Growth, maintenance, and calcium phosphate transient transfection was performed as described (25). The efficiency of transfection was above 70% in all experiments. For fluorescence resonance energy transfer (FRET) analysis, the cells were plated on a 6-well plate and transiently transfected using FuGENE-6 (Roche Applied Science). This reagent (6 l) was added to serum-free medium (100 l) at room temperature for 5 min. This medium was incubated with plasmids encoding the CFP and YFP fusion proteins (1 g each) for 30 min, and the mixture was added to the cells grown in culture medium. After 6 h of incubation, the cells were washed twice with phosphate-buffered saline, trypsinized, and reseeded on a 6-well plate containing one coverslip (22 mm, Electron Microscopy Science) per well.
Transport Assays and Transport Reconstitution-Influx rates of 100 M L-[ 3 H]glutamate (ARC) or 20 M L-[ 35 S]cystine (Amersham Biosciences) in Xenopus oocytes or transfected HeLa cells were performed as described (22,25). The effect of pCMB and MTS reagents (Toronto Research Chemicals, Inc.) was assayed as described (3,22). 1 mM pCMB and 2.5 mM MTSEA inhibited the activity of wild-type xCT (22) and wildtype b 0,ϩ AT, respectively. Reconstitution of wild type b 0,ϩ AT and the C321S mutant into liposomes and uptake measurements in the reconstituted system were performed as described (21).
Membrane Preparation and Protein Purification by Ni 2ϩ -NTA Chromatography-Kidney brush border membranes were obtained as described (4). For the preparation of kidney total membranes, the kidney was homogenized in 25 mM Hepes, pH 7.4, 4 mM EDTA, 250 mM sucrose, and 20 mM N-ethylmaleimide with the protease inhibitors aprotinin, leupeptin, phenylmethylsulfonyl fluoride, and pepstatin on a CPCU Polytron. The homogenate was centrifuged at 10,000 ϫ g for 10 min at 4°C, and the supernatant was further centrifuged at 200,000 ϫ g for 90 min at 4°C. A similar procedure was used to obtain total membranes from HeLa cells, but the cells (ϳ10 7 cells/1 ml homogenization buffer) were homogenized by 15 passages through a 25-gauge needle. For Blue Native PAGE or cross-linking, the membranes were resuspended directly in the appropriate buffers (see below).
For Ni 2ϩ -NTA chromatography, total membranes from HeLa cells were resuspended in 25 mM Tris-HCl, pH 7, 50 mM NaCl, 1% digitonin (final detergent/protein (w/w) ratio of 3.3), and solubilization proceeded for 30 min. Insoluble material was discarded by a 10,000 ϫ g centrifugation at 4°C for 10 min. The supernatant was diluted in the above buffer containing 15 mM imidazole and applied to the Ni 2ϩ -NTA beads (Qiagen). After 30 min of end-over-end mixing at room temperature, the beads were washed 4 times in a similar buffer containing 0.1% digitonin and 10 mM imidazole. Elution was performed by raising the imidazole concentration to 100 mM for 10 min at room temperature. The eluted material was then processed for Blue Native PAGE (see below).
Blue Native and SDS-PAGE-In preliminary experiments we tested a range of detergent/protein (w/w) ratios with a fixed Coomassie Blue G/detergent ratio (w/w) of 1/2.5 (28). With 0.5-1% digitonin extracts and a 3.3/1 detergent/protein ratio, we observed that rBAT and b 0,ϩ AT appeared as a single band of 535 Ϯ 18 kDa (Fig. 1A). Similar results were obtained with 0.25-0.5% dodecyl-␤-D-maltoside, although a smear appeared above the ϳ535-kDa band. N-Octylglucoside did not resolve any band, and Triton-X-100 increased the smearing until no distinct band could be seen (data not shown). We did not observe any additional bands. Digitonin/protein ratios below 3.3/1 increased smearing and decreased the intensity of the ϳ535-kDa band, and at a ratio of 0.25/1 no band was detected; ratios up to 10/1 were similar to the 3.3/1 ratio (data not shown).
Membranes were solubilized for 30 min at room temperature in 25 mM Tris-HCl, pH 7, 50 mM NaCl, 1% digitonin at a deter-gent/protein ratio (w/w) of 3.3 (see above). Further treatment with a range of urea concentrations, 100 mM DTT, or 2% SDS with or without 100 mM DTT was carried out at 37°C for 30 min. After solubilization, Blue Native buffer was added, and Blue Native PAGE was performed as described (29,30). Native molecular weight markers (Amersham Biosciences) were visualized by Coomassie staining. For control SDS-PAGE, SDS sample buffer (without DTT) was added to the solubilized samples, which were immediately loaded (without heating) in SDS-PAGE gels. Western blots were performed as described (31). After transference of Blue Native gels, the membrane was destained in methanol 50% and acetic acid 10% to eliminate excess Coomassie Blue G (Serva) and washed with bi-distilled water before blocking.
Cross-linking-Total membrane proteins from transfected HeLa cells were resuspended in phosphate-buffered saline and incubated at 2 mg/ml with the cross-linker dimethyl suberimidate (DMS, from Pierce) for 30 min at room temperature. Cross-linking was terminated by the addition of 100 mM Tris-HCl, pH 8.0, for 15 min at room temperature.
A Leica TCS SL laser scanning confocal spectral microscope (Leica Microsystems Heidelberg GmbH, Manheim, Germany) equipped with an argon laser, 63ϫ oil immersion objective lens and a double dichroic filter (458/514 nm) was used. We used CFP as the donor fluorochrome paired with YFP as the acceptor fluorochrome.
FRET measurements were based on the sensitized emission method previously described (32,33) with minor modifications for the confocal microscope. In some experiments in which the sensitized emission method was performed, FRET efficiency was also calculated using the acceptor photobleaching method (34,35). The same set of transfected cells (but distinct dishes) was used for both methods.
Sensitized Emission Method-The sensitized emission method is based on the increase of acceptor fluorescence caused by FRET during excitation. To measure FRET, three images were acquired in the same order in all experiments through 1) the CFP channel (absorbance 458 nm, emission 465-510 nm), 2) the FRET channel (absorbance 458 nm, emission 525-600 nm) 3) the YFP channel (absorbance 514 nm, emission 525-600 nm). Background was subtracted from images before performing FRET calculations. Control and experiment images were taken under the same conditions of photomultiplier gain, offset, and pinhole aperture.
The FRET image must be corrected for the cross-talk of the donor emission and the direct excitation of the acceptor by the donor excitation wavelength. The crossover of the donor and acceptor fluorescence through the FRET filter is a constant pro-portion between the fluorescence intensity levels of donor and acceptor and their bleed-through. To correct for this spectral bleed-through of the donor and acceptor through the FRET filter, images of cells expressing only CFP-b 0,ϩ AT/rBAT, CFP-xCT/4F2hc, or CFP-EGFR and cells expressing only YFPb 0,ϩ AT/rBAT, YFP-xCT/4F2hc, or YFP-EGFR were also taken under the same conditions as for the experiments.
Corrected FRET was calculated on a pixel-by-pixel basis for the entire image by using equation, where FRET, CFP, and YFP correspond to background-subtracted images of cells expressing CFP and YFP acquired through the FRET, CFP, and YFP channels, respectively. A (equal to 0.20) and B (equal to 0.21) correspond to the cross-talk coefficient of CFP and YFP through the FRET filter set, respectively. Images of FRET C intensity were renormalized using a look-up table in which the minimum and maximum values are displayed as blue and red, respectively. Mean FRET C values were calculated from mean fluorescence intensities for each selected region of interest following Equation 1, and normalized sensitized FRET (FRETN) values for selected sub-regions of the image were calculated on the basis of the equation, where FRET C , CFP, and YFP are the mean intensities of FRET C , CFP, and YFP fluorescence in the selected sub-region of the image.
The negative FRET C values obtained in some experiments are due to slight over-estimation of the spectral bleed-through coefficients. Areas with unusually high or low CFP/YFP ratios (i.e. outside the 1:1 to 1:4 stoichiometric range) were excluded from analysis. All calculations were performed using the Image Processing Leica Confocal Software and Microsoft Excel.
Acceptor Photobleaching Method-In the presence of FRET, bleaching of the acceptor (YFP) results in a significant increase in fluorescence of the donor (CFP). Half the cell was bleached in the YFP channel using the 514 argon laser line at 100% intensity. Before and after YFP bleaching, CFP and YFP images were collected to assess changes in donor and acceptor fluorescence. To minimize the effect of photobleaching caused by imaging, images were collected at low laser intensity. The gain of the photomultiplier tubes was adjusted to obtain the best possible dynamic range. FRET efficiency was calculated as, where I pre is the pre-bleach CFP intensity, and I post is the post-bleach CFP intensity in the bleached region. As an internal negative control, an unbleached region in the same cell was measured.

RESULTS
In Vitro Analysis of the Quaternary Structure of HAT-To analyze oligomerization of HAT, we used Blue Native PAGE (29,30) followed by Western blot. We initially tested different detergents, detergent to protein ratios, and Coomassie Blue G to detergent ratios to find good experimental conditions for detection of the transporters (see "Experimental Procedures" and Refs. 29 and 30). rBAT and b 0,ϩ AT were found as a single band of 535 Ϯ 18 kDa (n ϭ 12) (Fig. 1A). This band did not correspond to isolated rBAT or b 0,ϩ AT subunits because (i) it was detected in brush border membranes, where only the disulfide-linked heterodimer is found (Fig. 1A, lanes 1) and (ii) a functional, purified, His-tagged b 0,ϩ AT/rBAT concatamer (36) (where "/" indicates fusion of the two subunits) had the same mobility (Fig. 1A, lanes 4). As expected, no b 0,ϩ AT appeared in membranes from the b 0,ϩ AT KO mice (37) (Fig. 1A, ␣b 0,ϩ AT panel, lane 2). In contrast, rBAT was still detected as the 535-kDa band because of its binding to an as yet unidentified light subunit (4, 37) (Fig. 1A, ␣rBAT panel, lane 2).
Blue Native PAGE allows the determination of oligomerization stoichiometry with agents that partially dissociate complexes (28,38). Up to 5 M urea did not alter the mobility of the b 0,ϩ AT-rBAT complex (data not shown). In contrast, 2% SDS partially dissociated the complex in bands I and II, both in HeLa cells (Figs. 1, B and C) and in brush border membranes (data not shown). The relative amounts of the two bands varied slightly between experiments, but b 0,ϩ AT-rBAT never shifted completely to band II. The slower mobility of band I compared with the untreated sample could be due to partial unfolding of b 0,ϩ AT-rBAT. On the b 0,ϩ AT/rBAT concatamer, DTT may cause further unfolding that results in even a slower mobility of bands I and II (Fig. 1C). The addition of the reducing agent DTT in the presence of SDS dissociated the complex into its subunits (Fig. 1B). rBAT appeared as a single band of 162 Ϯ 11 kDa, and b 0,ϩ AT appeared as two bands of 180 Ϯ 14 and 84 Ϯ 9 kDa (n ϭ 7). SDS-PAGE followed by Western blot confirmed that DTT completely reduced the samples (data not shown). No dissociated subunits appeared with the b 0,ϩ AT/rBAT concatamer (Fig. 1C). According to a recently proposed empirical rule, the molecular weight of a polytopic transporter multiplied by a factor of 1.8 fits with its mobility in Blue Native gels (39). This applied to b 0,ϩ AT, confirming that the 84-and the 180-kDa bands were b 0,ϩ AT monomers and dimers, respectively (Fig.  1B). rBAT mobility was slower than expected (compare 162 Ϯ 11 kDa in Fig. 1B (Blue Native PAGE) with ϳ94 kDa (SDS-PAGE) on Fig. 3B), which could be due to partial unfolding. In conclusion, Blue Native PAGE strongly suggested that system b 0,ϩ has a [b 0,ϩ AT-rBAT] 2 structure since (i) only two bands (band I and band II, see Figs. 1, B and C, lanes 2 of the gels) were observed after treatment of the digitonin-solubilized sample with SDS, suggesting that b 0,ϩ AT-rBAT is a heterotetramer (two heterodimers; band I) that can be partially dissociated by SDS to the single heterodimer (band II), (ii) after DTT reduction, b 0,ϩ AT appeared as two bands compatible with a monomer and a dimer, and (iii) the size of the untreated sample (ϳ535 kDa) was double that of band II (260 Ϯ 19 kDa; n ϭ 7). In addition, the sum of the sizes of the monomeric rBAT (162 Ϯ 11 kDa) and the monomeric b 0,ϩ AT (84 Ϯ 9 kDa) matched the size of band II (260 Ϯ 19 kDa).
The His-xCT-4F2hc and Myc-LAT2-4F2hc heterodimers expressed in HeLa cells were detected on Blue Native gels as a single band of ϳ250 kDa (Fig. 2, A and B, respectively), similar to band II of b 0,ϩ AT-rBAT. No dissociation occurred even with 2% SDS. Further addition of DTT resulted in the detection of the His-xCT and myc-LAT2 subunits in a monomeric form (ϳ80 kDa). LAT2-4F2hc from kidney basolateral membranes behaved similarly (data not shown). A functional His-xCT/ 4F2hc concatamer (see supplemental Fig. C) ran with the same mobility as His-xCT-4F2hc ( Fig. 2A, lanes 4 -6) that was not modified by SDS or DTT, and a Myc-xCT/xCT dimeric concatamer co-transfected with 4F2hc (see below for functional data) ran at ϳ540 kDa, about twice the size of His-xCT-4F2hc heterodimer and the His-xCT/4F2hc concatamer ( Fig. 2A,  lanes 7-8). As expected, SDS did not dissociate this complex. It is worth mentioning that a faint band of similar size was observed with the His-xCT/4F2hc concatamer after long exposure ( Fig. 2A, asterisk). In summary, our results suggest that FIGURE 1. Blue Native PAGE analysis of b 0,؉ AT-rBAT. A, 1% digitonin extracts from distinct sources were supplemented with Blue Native sample buffer and loaded on 5-15% Blue Native gels for Western blot with antibodies against rBAT (␣rBAT) or b 0,ϩ AT (␣b 0,ϩ AT). 1, mouse kidney brush border membranes; 2, kidney brush border membranes from the null b 0,ϩ AT KO mouse; 3, total membranes from HeLa cells co-transfected with rBAT and b 0,ϩ AT; 4, the His-b 0,ϩ AT/rBAT concatamer expressed in HeLa cells was purified on a Ni 2ϩ -NTA chromatography column and eluted in imidazole buffer containing 0.1% digitonin. Preimmune antisera did not detect any band in the Blue Native gels (not shown). B, HeLa cells were co-transfected with rBAT, and b 0,ϩ AT and total membranes were isolated. The membranes were either solubilized with 1% digitonin (D), or after solubilization with 1% digitonin, 2% SDS was added for 30 min at 37°C (S) in the absence (Ϫ) or in the presence (ϩ) of DTT. The membranes were then supplemented with Blue Native sample buffer and loaded on 5-15% Blue Native gels for Western blot with antibodies against rBAT (␣rBAT ) or b 0,ϩ AT (␣b 0,ϩ AT ). C, the concatamer His-b 0,ϩ AT/rBAT was transfected in HeLa cells. Total membranes were isolated, solubilized with 1% digitonin, and applied to a Ni 2ϩ -NTA chromatographic column. The purified concatamer was eluted with imidazole buffer containing 0.1% digitonin and either left untreated (D) or treated with 2% SDS for 30 min at 37°C (S) in the absence (Ϫ) or presence (ϩ) of DTT. Blue Native sample buffer was added for loading on 5-15% Blue Native gels, and the Western blot was decorated with antibodies against the Xpress tag (␣Xpress) contiguous to the histidine tag of the concatamer. Longer periods in SDS or higher incubation temperatures did not modify the results. No other bands were detected in any of the gels, even with longer exposure times. Representative experiments from at least n ϭ 4 experiments for each type of sample are shown. Ab, antibody. I, band I; II, Band II.
LSHAT-4F2hc heterodimers do not stably assemble into higher order oligomers. We cannot rule out that an equilibrium between xCT-4F2hc and [xCT-4F2hc] 2 exists, which is likely shifted to the heterodimeric form.
Blue Native PAGE does not escape general criticisms to detergent-based methods. Therefore, we applied DMS a primary amine permeable cross-linker to total membranes from HeLa cells expressing rBAT and b 0,ϩ AT or 4F2hc and His-xCT. DMS shifted the b 0,ϩ AT-rBAT heterodimer from 130 to 250 kDa in a dose-dependent manner, indicating that two heterodimers are in close contact (Fig. 3A). Cross-linking efficiency was very high (10 mM DMS quantitatively shifted b 0,ϩ AT-rBAT to the 250 kDa band), revealing that both the intracellular and the plasma membrane heterodimer, the sites  1-3), the concatamer His-xCT/4F2hc (lanes 4 -6), or 4F2hc with the concatamer myc-xCT/ xCT (lanes 7 and 8) (each transfection corresponds to a distinct gel). Total membranes were isolated and either solubilized with 1% digitonin (D), or after solubilization with 1% digitonin, 2% SDS was added for 30 min at 37°C (S) in the absence (Ϫ) or in the presence (ϩ) of DTT. The samples were supplemented with Blue Native sample buffer and loaded on 5-15% Blue Native gels for Western blot with antibodies against the Xpress or the Myc tag (␣Xpress or ␣myc). The asterisk (*) indicates the high molecular weight band in lanes 4 -8. A representative experiment from n ϭ 4 is shown. B, HeLa cells were co-transfected with 4F2hc and myc-LAT2. Total membranes were isolated and treated as in A. The Western blot was decorated with ␣-Myc antibody. A representative experiment from n ϭ 3 is shown. Ab, antibody. where b 0,ϩ AT-rBAT is found in HeLa cells, 9 were in the same oligomeric state. Under reducing conditions the size of rBAT was not modified by the cross-linker, whereas a dose-dependent increase in the dimeric form of b 0,ϩ AT was observed (Fig.  3B). Therefore, two or more lysines on two b 0,ϩ AT subunits must be in close contact (no more than 11 Å). In addition, under reducing conditions b 0,ϩ AT ran as a monomer and as an SDSresistant dimer even in the absence of cross-linker, consistent with the two bands observed in non-reducing conditions (compare Figs. 3, A and B) (4). DMS had no effect on the xCT-4F2hc heterodimer (Fig. 3C). Under non-reducing conditions, the 125-kDa band represented most (Ͼ90%) of the total xCT-4F2hc, and the 240-kDa band did not increase after the addition of the cross-linker.
In Vivo Analysis of the Quaternary Structure of HAT-We applied the sensitized FRET emission method to study oligomerization in vivo (40) in HeLa cells transiently transfected with N-terminal (YFP or CFP)-tagged versions of the concatamers b 0,ϩ AT/ rBAT (see Fig. 4A) and xCT/4F2hc (see below). CFP-b 0,ϩ AT/rBAT and YFP-b 0,ϩ AT/rBAT completely co-localized at the plasma membrane and intracellular sites (Fig. 4B), and induced cystine transport activity (data not shown). Positive CFP3 YFP FRET C signals were detected at the plasma membrane when CFP-and YFP-b 0,ϩ AT/rBAT were co-expressed (Fig. 4B), suggesting close proximity of the N termini of the two fusion proteins (i.e. the N termini of the b 0,ϩ AT subunits) and that b 0,ϩ AT-rBAT forms dimers or higher order oligomers at the cell surface. As a control, we measured FRET in cells co-expressing YFPb 0,ϩ AT/rBAT and an unrelated integral membrane protein, the receptor for the epidermal growth factor tagged with CFP (CFP-EGFR) (40). Despite the co-localization of the two proteins at the plasma membrane (Fig. 4B), no FRET C signals were detected, indicating that they did not form complexes.
Because the efficiency of energy transfer depends on the amount of donor and acceptor capable of interaction, FRET C values were divided by the product of YFP and CFP intensities on a pixel-by-pixel basis to obtain normalized FRET (FRETN) at the plasma membrane. FRETN values in cells co-expressing CFP-and YFP-b 0,ϩ AT/rBAT (23.3 (E Ϫ 05) Ϯ 3.6 (E Ϫ 05)) were substantially higher than in control cells co-expressing YFPb 0,ϩ AT/rBAT and CFP-EGFR (3.5 (E Ϫ 05) Ϯ 1.2 (E Ϫ 05)) (Fig. 4C).
We also measured FRET in HeLa cells co-expressing the YFP-xCT/4F2hc and CFP-xCT/4F2hc concatamers. Localization of the fusion proteins was similar to that observed for YFPand CFP-b 0,ϩ AT/rBAT, and they also induced cystine transport activity (not shown). The FRETN value (7.8 (E Ϫ 05) Ϯ 2.6 (E Ϫ 05)) was significantly different from the values obtained when YFP-xCT/4F2hc was co-expressed with CFP-EGFR (which were zero) but lower than for b 0,ϩ AT/rBAT (see above).
Finally, we also determined FRET efficiency (E) with an independent approach; that is, the acceptor photobleaching FRET method (34,35). In YFP-and CFP-b 0,ϩ AT/rBAT co-expressing cells, E was 12.9 Ϯ 2.2%; in contrast, in YFP-and CFP-xCT/ 4F2hc co-expressing cells, E was not different from zero (2.9 Ϯ 2.9%). These results are consistent with the presence of b 0,ϩ AT-rBAT oligomers at the cell surface. Some xCT-4F2hc might be in an oligomeric state, which forms either less efficiently or is less stable than b 0,ϩ AT-rBAT oligomers.
The Transport Unit of Systems b 0,ϩ and x C Ϫ Is the Single Heterodimer-To study the b 0,ϩ AT-rBAT transport unit we first generated functional b 0,ϩ AT transporters either sensitive or insensitive to a sulfhydryl reagent (41)(42)(43). Among the reagents tested, only MTSEA completely abolished b 0,ϩ AT-rBAT-mediated uptake (Fig. 5A). Because b 0,ϩ AT alone is functional in a reconstituted system (21), we reasoned that the residues responsible for MTSEA inactivation may reside in b 0,ϩ AT. Therefore several b 0,ϩ AT Cys to Ser mutants were analyzed. The C321S b 0,ϩ AT mutant was insensitive to MTSEA (Fig. 5A), and in reconstitution experiments its activity was not significantly different to that shown by the wild type (Fig. 5B). Because the structural data indicated a heterotetrameric structure for b 0,ϩ AT-rBAT, we constructed dimeric concatamers between the wild-type and the C321S b 0,ϩ AT (Wt-CS and CS-Wt), so that when expressed the wild type and the C321S b 0,ϩ AT would be represented in equal proportions. Concatamers of two wildtype subunits (Wt-Wt) and two C321S subunits (CS-CS) were also constructed. The four concatamers were expressed together with rBAT in HeLa cells. They displayed the expected mobility in Western blots, and no single b 0,ϩ AT subunits were observed (data not shown). The transport function of the concatamers was 15-30% of the activity of b 0,ϩ AT-rBAT (supplemental Fig. A1), caused at least in part to their lower expression (less than 50%, data not shown) compared with b 0,ϩ AT-rBAT. The transport activity of the concatamers was measured after preincubation with MTSEA or vehicle (Fig. 5C). In a monomeric functional unit, MTSEA addition would result in a 50% loss of transport activity, whereas a total loss of function would be expected for a dimeric functional unit. After incubation with MTSEA, the wild-type and the C321S concatamers were completely inhibited or unaffected, respectively. In contrast, the activity of the combined concatamers after this treatment was close to 50% (Fig. 5C), suggesting that the transport unit of system b 0,ϩ is the heterodimer b 0,ϩ AT-rBAT.
xCT is completely inactivated by the sulfhydryl reagent pCMB, whereas the fully functional C327S xCT mutant is unaffected (22). We constructed dimeric concatamers between the wild-type and the C327S xCT in both orientations. Blue Native gels showed that these concatamers had the expected size, and no single xCT subunits were observed (see Fig. 2A). Western blots confirmed this finding (data not shown). The concatamers cRNAs were injected together with 4F2hc in Xenopus oocytes. The activity of the concatamers was 20 -30% compared with the wild-type or C327S xCT (supplemental Fig. A2). After exposure to pCMB, the wild-type and the C327S concatamers were completely inhibited or unaffected, respectively. In contrast, the combined concatamers showed an inhibition very close to the theoretical 50% (Fig. 6). In addition, when a 1:1 ratio of wild-type and C327S xCT cRNAs was injected together with 4F2hc, transport was inhibited by 47.9 Ϯ 2% after treatment with pCMB (supplemental Fig. B). Therefore, systems b 0,ϩ and x C Ϫ have a common functional unit, the single heterodimer.
rBAT Promotes Oligomerization-To uncover determinants of b 0,ϩ AT-rBAT oligomerization, we compared the mobility of b 0,ϩ AT/4F2hc, His-xCT/rBAT, b 0,ϩ AT/rBAT, and His-xCT/ 4F2hc concatamers on Blue Native gels (Fig. 7). The chimaeras showed 28 Ϯ 5% (b 0,ϩ AT/4F2hc) and 36 Ϯ 3% (His-xCT/ FIGURE 5. The functional unit of b 0,؉ AT-rBAT. A, HeLa cells were co-transfected with wild-type or C321S b 0,ϩ AT together with rBAT. 36 h later, the transport of 20 M L-cystine was assayed after preincubation in 2.5 mM MTSEA or vehicle for 5 min. The % of activity was obtained as the ratio of the transport activities induced (i.e. the difference from non-transfected cells treated similarly) in the absence (white bars) and the presence (black bars) of MTSEA. A representative experiment of n ϭ 4 is shown. B, HeLa cells were transfected with wild-type or C321S b 0,ϩ AT. After 3 days the cells were used for reconstitution into liposomes containing (black) or not (white) 2 mM leucine, and the transport rates of 0.5 M L-[ 3 H]arginine were measured. Data (mean Ϯ S.E.) correspond to a representative experiment performed in triplicate, corrected for the amount of reconstituted b 0,ϩ AT in arbitrary units (AU). Arbitrary units were obtained by densitometry of Western blots with ␣b 0,ϩ AT using two distinct b 0,ϩ AT-reconstituted liposome volumes. In a second experiment the activity of C321S b 0,ϩ AT was 92.5 Ϯ 4.5% that of the wild type. C, HeLa cells were transfected with rBAT, and the concatamers were indicated. 36 h later transport of 20 M L-cystine was assayed after preincubation for 5 min in 2.5 mM MTSEA or vehicle. The % of activity was calculated as the ratio of the transport activities induced (i.e. the difference from non-transfected cells treated similarly) in the absence (white bars) and in the presence (black bars) of MTSEA. CS, C321S mutant. The mean Ϯ S.E. of n ϭ 5 experiments is shown. The mean Ϯ S.E. of the Wt-CS and the CS-Wt groups are significantly different from 25% (p Ͻ 0.05 for Wt-CS (*) and p Ͻ 0.001 for CS-Wt (***)) but not significantly different from 50% (p ϭ 0.0852 for Wt-CS and p ϭ 0.7487 for CS-Wt) (one sample t test; the Wilcoxon signed-rank test was also applied with similar results).
rBAT) activity compared with their controls (supplemental Fig.  C). b 0,ϩ AT/4F2hc ran as two defined bands and a high molecular weight smear at the top of the gel, similar to His-xCT/ 4F2hc. However, more than 90% of His-xCT/4F2hc was in the heterodimeric form, whereas a 50% distribution between heterodimers and heterotetramers was observed for b 0,ϩ AT/ 4F2hc. The mobility of His-xCT/rBAT was identical to b 0,ϩ AT/ rBAT, suggesting that rBAT is the major determinant for the oligomerization.

DISCUSSION
Our study demonstrates that b 0,ϩ AT-rBAT forms an oligomer at the plasma membrane, with a minimal heterotetrameric structure. We used initially Blue Native PAGE because it seems to maintain the oligomeric structures of several membrane protein complexes solubilized in nonionic detergents such as dodecylmaltoside and digitonin (28 -30, 38, 44). On Blue Native gels the b 0,ϩ AT-rBAT complex was detected as a single 535-kDa band both in kidney membranes and transfected HeLa cells. More important, the complex was partially dissociated by SDS only to a smaller band of about half the size, strongly suggesting a heterotetrameric structure. Strikingly, urea did not dissociate the complex, indicating a very stable interaction between the two heterodimers. A similar property was reported for the mitochondrial 400-kDa general import pore (GIP) complex (44). Moreover, DTT in the absence of SDS did not dissociate the complex despite complete reduction of the intersubunit disulfide bond (data not shown), suggesting that the stability of the heterotetramer is independent from the integrity of that link. The interaction between the heterodimers may be of high affinity, as only the 535-kDa species was found on Blue Native gels and because the crosslinker DMS quantitatively linked two (and only two) b 0,ϩ AT-rBAT heterodimers on non-solubilized total membranes. DMS cross-linking indicated that the heterotetramer was the main structure in vivo. Additional evidence for oligomerization of the b 0,ϩ AT-rBAT on the cell surface of live cells came from FRET analysis, which showed significant association of CFP-and YFP-b 0,ϩ AT/rBAT. FRET did not occur by random association because it was negligible after co-expression with an unrelated, but co-localized, plasma membrane protein, the EGFR. Taken together, the evidence is consistent with a high affinity and stable heterotetrameric b 0,ϩ AT-rBAT on the cell surface in vivo. It is worth mentioning that the size of rBAT associated with an unidentified subunit in brush border membranes (4,37) was also consistent with a heterotetramer (Fig. 1A, ␣rBAT panel, lane 2).
Because the functional unit of HAT is the single heterodimer (see below for the discussion), it is hard to propose a functional role for the heterotetramer in system b 0,ϩ activity. An alternative possibility is that oligomerization provides a late quality control step in the biogenesis pathway of the transporter. Preliminary data obtained in our laboratory 10 is consistent with this hypothesis, because the oligomerization of b 0,ϩ AT-rBAT 10 P. Bartoccioni, M. Palacín, and J. Chillaró n, manuscript in preparation.  Fig. 1A for Blue Native gel electrophoresis followed by Western blot with antibodies against ␣b 0,ϩ AT or ␣Xpress (the Xpress tag contiguous to the His tag of xCT). The same results were obtained with antibodies against ␣rBAT (for the constructs which contained this subunit) (data not shown). A representative experiment of n ϭ 3 is shown. The high molecular weight smear in lanes 2 and 3 may represent aggregated material which remained at the top of the gel. Ab, antibody. FIGURE 6. The functional unit of xCT-4F2hc. B, Xenopus oocytes were injected with 5 ng of 4F2hc cRNA and 25 ng of the concatamer cRNAs as indicated. Three days after, transport of 100 M L-glutamate was assayed after preincubation in 1 mM pCMB or vehicle for 5 min. The % of activity is obtained as the ratio of the transport activities induced (i.e. the difference from noninjected oocytes treated similarly) in the absence (white bars) and in the presence (black bars) of pCMB. CS, C327S mutant. The mean Ϯ S.E. of n ϭ 5 experiments is shown. The mean Ϯ S.E. of the Wt-CS and the CS-Wt groups are significantly different from 25% (p Ͻ 0.0001 (***)) but not significantly different from 50% (p ϭ 0.0896 for Wt-CS and p ϭ 0.7588 for CS-Wt) (one sample t test; the Wilcoxon signed-rank test was also applied with similar results). seems to occur just before or concomitant with the exit from the endoplasmic reticulum.
In contrast to b 0,ϩ AT-rBAT, the results suggest that xCT-4F2hc and LAT2-4F2hc heterodimers do not efficiently oligomerize and/or that the formed oligomers are not as stable as the b 0,ϩ AT-rBAT heterotetramers. Actually, the faint bands of heterotetrameric xCT/4F2hc detected in Blue Native gels and the positive FRETN value for xCT/4F2hc (not confirmed by the acceptor photobleaching FRET method) are the only data supporting oligomerization of xCT-4F2hc. Moreover, despite the similar distribution of lysine residues in b 0,ϩ AT and xCT, xCT-4F2hc was not cross-linked by DMS, and the positive FRETN value was lower than for b 0,ϩ AT-rBAT. Anyway, caution must be taken concerning the oligomerization of LSHAT-4F2hc heterodimers. For instance, digitonin may dissociate oligomers of LSHAT-4F2hc. In addition, most tissues and cell lines express 4F2hc and one or more 4F2hc-associated LSHAT (1, 2), which may compete with the xCT-4F2hc FRET constructs and reduce the FRET signal. Other factors may decrease FRET signal, such as interactions of the xCT N termini with other proteins. As a working hypothesis we propose that 4F2hc-associated heterodimers do not stably oligomerize and/or that the heterodimer 7 oligomer equilibrium is strongly shifted to the heterodimeric form.
The two b 0,ϩ AT light subunits were cross-linked by DMS, suggesting that they are in close contact within the b 0,ϩ AT-rBAT oligomer (Fig. 3). However, the heavy subunit rBAT appeared to be the major determinant of the oligomerization, as its presence was sufficient to confer a heterotetrameric structure to the xCT/rBAT concatamer (Fig. 7). rBAT may facilitate or stabilize the interaction between the two light subunits either by exposing a dimerization surface on b 0,ϩ AT and xCT and/or by prior dimerization of rBAT itself. The appearance of the heterotetramer band on the b 0,ϩ AT/4F2hc concatamer (Fig. 7) suggests that b 0,ϩ AT may tend to oligomerize, and 4F2hc could prevent a high affinity interaction. Alternatively, 4F2hc-associated heterodimers may oligomerize depending on the context. In this regard it has been shown that 4F2hc interacts with integrins (19), and this interaction requires the clustering of 4F2hc (45).
Here we provide the first strong evidence indicating that the single heterodimer is the functional unit of HAT (where the light subunit is the transporter itself (21)). This is based on the distinct sensitivity to sulfhydryl reagents of the wildtype transporters and the Cys-to-Ser mutants, which otherwise are functionally very similar (Ref. 22, and see "Results"). We assume that if the activity of one subunit of an oligomeric functional unit is inhibited through chemical modification, the transport of the entire functional complex is blocked. Given this assumption, the experiments are compatible with a functional unit composed by a single heterodimer both for b 0,ϩ AT-rBAT and xCT-4F2hc. The ϳ50% loss of function observed (Figs. 5C and 6) is best explained by the presence of two functionally independent transport pathways per concatamer. The low expression level of the concatamers used for the functional studies precluded a more detailed kinetic transport analysis. Therefore, we cannot exclude, especially for b 0,ϩ AT-rBAT (which is a heterotetramer; see above), some degree of cooper-ativity between the two b 0,ϩ AT subunits. An alternative proposal is that the dimeric concatamers associate themselves into tetrameric structures (four light subunits) where the only functional dimers are constituted by the unlinked light subunits. If the association occurs at random, the reagents would inactivate 75% of the transport activity, which is excluded by the data. Moreover, DMS cross-linking of the b 0,ϩ AT/b 0,ϩ AT concatamers failed to show any cross-linked band (data not shown).
It has been shown that system b 0,ϩ from the chicken small intestine shows a sequential exchange mechanism compatible with the formation of a ternary complex (46). If this applies to the expressed b 0,ϩ AT-rBAT complex, then export and import pathways should co-exist simultaneously in the proposed functional unit (i.e. the heterodimer) and, therefore, in a single b 0,ϩ AT catalytic subunit. Interestingly, mitochondrial carriers have a similar transport mechanism, although each transport pathway resides in a single six-transmembrane domain subunit of the functional dimeric carrier (47,48). Clearly, studies with purified and reconstituted HAT will be necessary to integrate the transport mechanism, the functional unit, and the structure of these transporters in a unique model.