Proximity of two oppositely oriented reentrant loops in the glutamate transporter GLT-1 identified by paired cysteine mutagenesis.

Sodium- and potassium-coupled transporters clear the excitatory neurotransmitter glutamate from the synaptic cleft. Their function is essential for effective glutamatergic neurotransmission. Glutamate transporters have an unusual topology, containing eight membrane-spanning domains and two reentrant loops of opposite orientation. We have introduced pairwise cysteine substitutions in several structural elements of the GLT-1 transporter. A complete inhibition of transport by Cu(II)(1,10-phenanthroline)(3) is observed in the double mutants A412C/V427C and A364C/S440C, but not in the corresponding single mutants. No inhibition is observed in more then 20 other double cysteine mutants. The Cu(II)(1,10-phenanthroline)(3) inhibition can be partly prevented by the nontransportable glutamate analogue dihydrokainate. Treatment with dithiothreitol restores much of the transport activity. Moreover, micromolar concentrations of cadmium ions reversibly inhibit transport catalyzed by A412C/V427C and A364C/S440C double mutants, but not by the corresponding single mutants. Inhibition by Cu(II)(1,10-phenanthroline)(3) and by cadmium is only observed when the cysteine pairs are introduced in the same polypeptide. Therefore, in both cases the proximity appears to be intra- rather than intermolecular. Positions 364 and 440 are located on reentrant loop I and II, respectively. Our results suggest that these two loops, previously shown to be essential for glutamate transport, come in close proximity.

Glutamate, the predominant excitatory neurotransmitter in the brain, is removed from the synaptic cleft by glutamate transporters located in the plasma membrane of nerve and glial cells. Uptake of glutamate by the glutamate transporters maintains its extracellular concentrations below the neurotoxic level (1)(2)(3)(4)(5)(6). In some synapses, glutamate transporters play an important role in limiting the duration of synaptic excitation (7)(8)(9)(10). The uptake process is electrogenic (11)(12)(13), involving co-transport of three sodium ions, a proton, and a glutamate molecule and countertransport of a potassium ion (14 -16). In addition to the coupled flux, glutamate transporters mediate a thermodynamically uncoupled chloride flux, activated by two of the molecules they transport, sodium and glutamate (17,18).
The five known eukaryotic glutamate transporters, GLT-1 (19), , EAAC-1 (21), , and EAAT-5 (22), have an overall amino acid identity of ϳ50%. The homology is significantly higher in the carboxyl-terminal half of the transporters. Topology studies suggest that this region of the protein has an intriguing arrangement, containing two oppositely oriented reentrant loops, transmembrane domains (TM) 1 7 and 8, as well as an outward facing hydrophobic region (Refs. 23-25; see Fig. 1). Some of the features of the membrane topology remain under debate (26).
Several amino acid residues critical for the function of glutamate transporters are located in the carboxyl-terminal half. Two adjacent amino acid residues of GLT-1, Tyr-403 and Glu-404, are located in TM7 and are conserved in all other glutamate transporters. Both were implicated in the binding of potassium ions (16,27) and appear to be close to one of the sodium binding sites (27). A conserved arginine residue, Arg-477 in GLT-1, is located in TM8 and has been shown to play a pivotal role in the sequential interaction of the transporters with amino acid substrates and with the potassium ion (28). Accessibility studies have shown that cysteine residues introduced at positions 364 and 440 of GLT-1, located on reentrant loops I and II, respectively, react with the impermeant sulfhydryl reagent MTSET added from the extracellular side (24 -26, 29). Substrates and nontransportable analogues partially protected against the modification of cysteines introduced at these positions. Therefore, we have suggested that positions 364 and 440 may be close in the three-dimensional structure of the protein (25).
To verify this prediction and to obtain the first information regarding the tertiary structure of the carboxyl-terminal half of the glutamate transporters, we have set out to determine proximity relationships between the different structural elements in this region. In this study, we have used two types of functional assays to infer proximity of engineered cysteine pairs. We report here the identification of two cysteine pairs, A412C/ V427C and A364C/S440C, which behave as if they are close in space. The data provide evidence that the two oppositely oriented reentrant loops are spatially close to one another. 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.

EXPERIMENTAL PROCEDURES
Mutagenesis-Specific oligonucleotide primers containing the desired mutation and an engineered restriction site were synthesized. One or two primers were used to introduce single or double mutations on uracil-containing single-stranded DNA derived from wild type GLT-1 (WT-GLT-1) or cysteine-less GLT-1 (CL-GLT-1) constructs (30,31). Mutated clones were selected using the engineered restriction site(s) as a diagnostic tool. The fragment of GLT-1 containing the mutation(s) was then subcloned into WT-or CL-GLT-1 using unique restriction sites flanking the mutation(s), and the mutated constructs were sequenced in both directions between these unique sites.
Whole Cell Transport-Uptake assays using D- 3 [H]aspartic acid were carried out essentially as described (31). Briefly, HeLa cells were cultured in 24-well plates, transfected with the indicated GLT-1 construct, and infected with the vaccinia/T 7 virus vTF 7-3 (32) as described (31). After 16 -20 h, the cells were washed twice with choline solution (10 mM potassium phosphate, pH 7.5, 150 mM choline chloride, 0.5 mM magnesium sulfate, and 0.3 mM calcium chloride). Next, the cells were preincubated with the indicated solution for 5 min at room temperature. Following a brief wash with the choline solution, a sodium solution (10 mM potassium phosphate, pH 7.5, 150 mM sodium chloride, 0.5 mM magnesium sulfate, 0.3 mM calcium chloride) supplemented with 0.4 Ci of the radiolabeled D-aspartic acid (18 -21 Ci/mol, Amersham Biosciences, Inc. or PerkinElmer Life Sciences) was added and the uptake was carried out for 10 min at room temperature. The cells were washed twice with ice-cold sodium solution and lysed with 1% SDS solution for scintillation counting. For oxidation studies, the preincubation solution contained 100 M Cu(II)(1,10-phenanthroline) 3  In addition to the nine double mutants presented here, we have produced other combinations of cysteine residues, including A364C/G437C, Y403C/V427C, Y403C/G437C, Y403C/S440C, Y403C/V469C, A412C/G437C, and V427C/G437C. The relative activity of these double mutants ranges from 20 to 80%, and no significant decrease of activity was observed following the oxidizing treatment. solution containing freshly prepared 12 mM DTT for 5 min at room temperature. After another washing step, the uptake activity assay was performed as above.
Inhibition by Cadmium-HeLa cells transfected with the indicated construct were washed twice with choline solution and preincubated with the indicted concentration of cadmium chloride in sodium solution for 5 min at room temperature. The solution was aspirated, and the cells were incubated with the transport solution and the indicated concentration of cadmium chloride. After the uptake step, the cells were processed as above.

Activity-based Screen for Thiol Cross-linking of Double Cysteine Mutants-
The pioneering studies of Kaback and co-workers (33) to detect proximity relationships in lactose permease were based on creating, under oxidative conditions, a disulfide bond between two single cysteines, each located at a different TM domain or connecting loop. Cross-linking of two cysteines can be easily detected when each is located on a different part of a split transporter (33,34). We have made unsuccessful attempts to generate a functional split GLT-1 transporter (25). Molecular engineering of artificial protease-sensitive sites in GLT-1, with the aim to obtain physical evidence of cross-linking of engineered cysteines, led to inactivation of the transporter (data not shown). Therefore, we have employed a functional assay to study proximity relationships. Based on the well established observation that GLT-1 undergoes extensive conformational changes during transport (35), we reasoned that introduction of a covalent disulfide bridge may hamper these changes, leading to inhibition of transport. To increase the likelihood to detect proximal pairs of cysteines, we started our study with positions where introduced cysteines are accessible to the impermeant hydrophilic sulfhydryl reagent MTSET. Such positions are presumably facing a waterfilled cavity in the transporter, possibly the permeation pathway. Pairs of cysteines were introduced in many combinations in the WT-GLT-1, and the double mutants were expressed in HeLa cells. Transport of the nonmetabolizable substrate D-[ 3 H]aspartate was assayed with and without preincubation with the oxidizing agent CuPh. Some of the most significant results are depicted in Fig. 2. Although WT-GLT-1 contains nine endogenous cysteine residues (see Fig. 1), its activity is only slightly decreased upon exposure to 100 M CuPh. Two double cysteine mutants, A412C/V427C and A364C/S440C, completely lose their activity under the same conditions. Significantly, other combinations of cysteine pairs among these four positions do not lead to an increased sensitivity of the double mutants to oxidizing conditions (Fig. 2). Moreover, pairing of A412C, V427V, A364C, or S440C with a cysteine residue introduced at position 469 in TM8 results in CuPh-insensitive double mutants. Uptake activity of additional double cysteine mutants is also not affected by CuPh treatment (see legend to Fig. 2).
Although most of the double cysteine mutants retain at least 50% transport activity compared with WT-GLT-1, two (A364C/ S440C and A364C/V469C) show ϳ20% of WT activity (Fig. 2). Analysis of the kinetic properties of the A364C/S440C transporter shows that the decrease of activity is the result of a lowering of V max (1.2 Ϯ 0.6 compared with 4.7 Ϯ 0.5 nmol/mg of protein/min for WT-GLT-1) (n ϭ 3). The apparent K m is not significantly changed (34 Ϯ 6 M compared with 38 Ϯ 2 M for the WT-GLT-1). Thus, the conformation of the double mutant is not altered, at least regarding its substrate affinity.
Characterization of the Cysteine Pairs A412C/V427C and A364C/S440C-Because the double mutants were constructed in the WT-GLT-1 background, it is possible that the oxidation of A412C/V427C and A364C/S440C double mutants leads to the cross-linking of one of the newly introduced cysteines with an endogenous cysteine residue. To investigate this possibility, the two cysteine pairs were introduced in the background of CL-GLT-1, where all nine endogenous cysteines have been removed (23,36). However, we could not analyze the effect of CuPh on the double mutants, because they are completely devoid of activity (Fig. 4). The single cysteine mutants in this background are active (Fig. 4). Surface biotinylation experiments indicate that this is probably a result of the fact that these double mutant transporters do not reach the plasma membrane (data not shown). In view of this result, we have continued to characterize the structural determinants leading to the loss of activity caused by oxidative conditions in the wild type background. None of the single mutants A412C, V427C, A364C, or S440C in the WT-GLT-1 background are sensitive to the oxidizing conditions (Fig. 3, A and B). Moreover, transport activity of the 412C/ V427A and A412S/V427C mutants is not significantly inhibited by CuPh, as opposed to the A412C/V427C mutant (Fig.  3A). The same is shown for A364S/S440C and A364C/S440A mutants (Fig. 3B). These observations render highly unlikely the possibility that perturbation of one of the positions causes a conformational change bringing an endogenous cysteine close to one of the introduced ones.
The formation of a disulfide bond between proximal cysteines induced by CuPh would be expected to be reversed by a reducing agent such as DTT, provided it can reach the dithiol bond. Indeed, transport activity of both A412C/V427C and A364C/ S440C previously treated with CuPh was significantly restored upon exposure to 12 mM DTT (Fig. 5). This concentration of DTT is similar to that used to reverse the inhibition of serotonin transporter mutants by methanethiosulfonate reagents (37).
Recent studies indicate that glutamate transporters may form oligomers (38,39), although it is still unclear if the monomer or the oligomer is the functional unit. Therefore, the oxidative cross-linking inferred for the two cysteine pairs may be intermolecular rather than intramolecular. Our results do not support this possibility. When the single mutants A412C and V427C are coexpressed in HeLa cells, no inhibition of transport activity by CuPh is observed. On the other hand, when both mutations are present on the same cDNA molecule, a potent inhibition by CuPh is observed (Fig. 3A). The same is observed for A364C and S440C (Fig. 3B).
When the substrate analogue DHK is present during the oxidative treatment, 6-and 2.5-fold protection is observed with A412C/V427C and A364C/S440C, respectively (Fig. 6). The substrate L-glutamate has a 3-fold protective effect on A412C/ V427C, whereas GABA, which is not a substrate, does not protect. In A364C/S440C, the protective effect of L-glutamate is smaller but statistically significant (Fig. 6). Replacement of the other cosubstrate (sodium) by choline, has no effect on inhibition by CuPh (data not shown).
The Pairs A412C/V427C and A364C/S440C Create a High Affinity Binding Site for Cadmium (II) Ions-As a complementary approach to establish the proximity between cysteines introduced at positions 412 and 427 as well as between those at positions 364 and 440, we have examined the ability of the double mutants to form a high affinity Cd 2ϩ binding site. This divalent cation interacts with cysteinyl side chains (40,41), and the affinity of the interaction is dramatically increased if the Cd 2ϩ ion can be coordinated by two cysteines (42). Exposure of the WT-GLT-1 to up to 100 M Cd 2ϩ has no significant effect on D-[ 3 H]aspartate transport (Fig. 7A). The single cysteine mutants are also unaffected by this treatment (Fig. 7A). On the other hand, the activity of the A412C/V427C double mutant is inhibited to ϳ50% at 100 M Cd 2ϩ (Fig. 7A). The activity of the A364C/S440C double mutant is dramatically more sensitive to the divalent cation. Half-maximal inhibition is attained at ϳ0.5 M and total inhibition at ϳ10 M Cd 2ϩ . In both double mutants, a partial reversal of this inhibition is observed when the Cd 2ϩ ions are washed away (data not shown). It is of interest to note that the A364C/ S440C transporter is also more sensitive to CuPh as compared with A412C/V427C (Fig. 7B), although the difference is not as dramatic as that observed with Cd 2ϩ .
We have examined the impact of CuPh and Cd 2ϩ on the kinetics of the double mutant transporters. Transfected cells were treated with appropriate concentrations of these reagents to achieve 50 -70% inhibition of transport. Because of the relative low transport rate of A364C/S440C, no reliable data could be obtained for this double mutant. However untreated controls, whereas the K m remains essentially unchanged Ϫ37.7 Ϯ 0.7 M (n ϭ 2).

DISCUSSION
The results described in this paper provide the first information on distance constraints in the (Na ϩ ϩ K ϩ )-coupled glutamate transporters. We have identified two cysteine pairs, A412C/V427C and A364C/S440C, which appear to be close in space. We have used two functional assays to infer proximity relationships between the engineered cysteine pairs. One assay (Figs. 2 and 3) is based on the observation that extensive conformational changes take place during glutamate transport (35) and the idea that formation of disulfide bonds could hamper them. This idea is supported by our observations that CuPh impacts V max rather than K m of A412C/V427C, indicating that its effect is the result of the inactivation of modified transporters. CuPh treatment of other transporters, containing two engineered proximal cysteines, has also been shown to impair function (43,44). In some transporters and pumps, the crosslinking of the intact molecule has been shown to result in a gel-shift (45,46). We have attempted this experiment, but could not observe a change in mobility on SDS-PAGE upon treatment with CuPh (data not shown). This is probably a result of the fact that our observed functional impact represents inactivation of surface transporters. In the vaccinia/T 7 virus expression system employed, these are only a small fraction of the total protein pool made in the HeLa cells. Thus, the system used in our studies is not optimal for the detection of surface-directed cross-linking.
The second functional assay is based on the creation of a high affinity Cd 2ϩ binding site between vicinal cysteines (Fig. 7A). Engineering of artificial high affinity Zn 2ϩ binding sites has recently been used to define proximity relationships in the dopamine transporter (47,48). The inhibitory effect of Zn 2ϩ on the dopamine transporter was primarily the result of a decrease in V max (47). Similar effects of Zn 2ϩ have recently been observed with the GABA transporter GAT-1 (49). This is in agreement with our results on Cd 2ϩ inhibition of A412C/ V427C, where V max is selectively reduced. Although it has been reported that the affinity of the interaction with Cd 2ϩ is increased even when the binding site is coordinated by two cys- teines (42), three cysteines have been implicated for high affinity Cd 2ϩ binding in another system (50). Therefore, it is possible that the markedly higher Cd 2ϩ sensitivity of A364C/ S440C compared with A412C/V427C can result from the contribution of an additional (yet unknown) contact site within the A364C/S440C transporter, similar to the finding with the zinc sensitivity studies of the dopamine transporters (48). Another explanation for the higher Cd 2ϩ sensitivity of A364C/S440C relative to A412C/V427C is the possibility that the bond angles of the former are more conducive to Cd 2ϩ binding than the latter. Alternatively, positions 364 and 440 may be closer to each other than the other pair. Consistent with this option is the higher CuPh sensitivity of A364C/S440C compared with A412C/V427C (Fig. 7B).
The activity of the two double cysteine mutants compared with their single counterparts suggests that the impact of the individual mutations is not simply additive (Fig. 3). This is seen even more dramatically in the cysteine-less background (Fig. 4). In the latter case, it appears that the double cysteine transporters do not reach the plasma membrane, consistent with our inability to observe activity in these mutants after exposure of the cells (or reconstituted preparations derived of them) to DTT (data not shown). One possibility is that a disulfide bridge between the two cysteines is generated because of the oxidative environment of the endoplasmic reticulum, thus preventing their arrival or maintenance at the plasma membrane. The presence of one or more of the endogenous cysteines of GLT-1 somehow moderates this phenomenon (Figs. 3 and 4).
We have attempted to see whether the distance of the two pairs is changing as a function of the conformation of the transporter. For both pairs we have observed a significant protection of inactivation by CuPh by the nontransportable blocker DHK (Fig. 6). However, the interpretation of this result is not straightforward, because DHK may restrict the accessibility of the oxidizing agent required for disulfide formation, rather than affect the distance between the two cysteines. In fact DHK has been shown to protect the A364C and S440C transporters against inactivation by methanethiosulfonate reagents (25,29). This issue must await future experimentation to distinguish between the two possibilities.
The observation that position 412 at the top of TM7 ( Fig. 1; Refs. [23][24][25] is close to position 427 (Figs. 2, 3, and 7), located at the middle of the descending limb of reentrant loop II, enables us to refine the topological model of GLT-1 (Fig. 8). Position 427 is apparently located close to the top of TM7. Therefore, reentrant loop II needs to be drawn more extracellularly and residues 412-427 probably are all extracellular, and position 431 probably reaches only halfway through the membrane bilayer. Because a cysteine introduced at position 431 is accessible to the bulky biotinylated sulfhydryl reagent 3-N-maleimidyl(propionyl)biocytin from the inside (23), it appears that there is a wide vestibule from the cytoplasm, which reaches the apex of reentrant loop II. Engineering of additional cysteine pairs may help to further define the tertiary structure of GLT-1.
As schematically illustrated in Fig. 8, it is of special interest that positions 364 and 440, each located on an oppositely reentrant loop, appear to be close in space (Figs. 2, 3, and 7). Proximity of two reentrant loops of opposite orientation has been observed in water and glycerol channels (51,52). In the water channel aquaporin 1 and the glycerol facilitator glpF, these reentrant loops line the permeability pathway. Even though the precise molecular details appear to be somewhat different, our results suggest that this may be true also for glutamate transporters. Consistent with this idea are our observations that the nontransportable glutamate analogue DHK protects the single mutants A412C, V427C, A364C, and S440C against modification by MTSET (25,29). 2 Moreover, DHK also reduces the cross-linking mediated by CuPh (Fig. 6). Glutamate itself has a smaller protective effect than DHK (Fig. 6), but this is not unreasonable because it was transported and would therefore occupy the binding site for a shorter time. Defining the interfaces between the two reentrant loops by identification of other cysteine pairs may shed more light on the molecular mechanism of glutamate transport.
Glutamate transporters form multimers (38,39), possibly pentamers, and permeation of the glutamate anion does not follow the same pathway as that of the uncoupled chloride flux (53). Therefore, it has been suggested that chloride passes through the space between the monomers and glutamate is translocated through the monomer itself (39). Using the approach initiated here, it should be possible to test this idea in the future on glutamate transporters that have higher expression levels in oocytes than GLT-1. Cross-linking of yet unidentified cysteine pairs would be expected to selectively block the chloride conductance, whereas that of others might hamper coupled transport alone.