Structural Selectivity and Molecular Nature of L-GlutamateTransport in Cultured Human Fibroblasts

Benedict Cooper,*,† Mary Chebib,‡ Jie Shen,§ Nicholas J. C. King,§ Ivan G. Darvey,¶

Philip W. Kuchel,¶ Jeffrey D. Rothstein,\ and Vladimir J. Balcar*,†,1

*Department of Anatomy and Histology, ‡Department of Pharmacology, §Department of Pathology, ¶Department ofBiochemistry, and †Institute for Biomedical Research, The University of Sydney, Australia NSW 2006; and\Neuromuscular Division, Department of Neurology, The Johns Hopkins University, Baltimore, Maryland 21287

Received December 17, 1997

Uptake of L-[3H]glutamate by monolayers of fibroblastscultured from human embryonic skin has been studied inthe presence of several nonradioactive structural analogsof glutamate and aspartate. Results have suggested thatthe structural specificities of glutamate transporters incultured human fibroblasts are similar to those of gluta-mate transporters in the mammalian brain. Only subtledifferences have been detected: in the mammalian cere-bral cortex, enantiomers of threo-3-hydroxyaspartate arealmost equipotent as inhibitors of L-[3H]glutamate uptakewhile, in human fibroblasts, the D-isomer has been found tobe an order of magnitude less potent than the correspond-ing L-isomer. Kinetic analysis of a model in which sub-strates are recognized by the glutamate transporter bind-ing site(s) as both a- and b-amino acids indicated that sucha mechanism cannot explain the apparent negative coop-erativity characterizing the effects of D- and L-aspartate.Molecular modeling has been used to estimate the opti-mum conformation of L-glutamate as it interacts with thetransporter(s). Flow cytometry has indicated that all fibro-blasts in culture express at least moderate levels of fourglutamate transporters cloned from human brain. Smallsubpopulations (<3%) of cells, however, were strongly la-beled with antibodies against EAAT1 (GLAST) and EAAT2(GLT-1) transporters. We conclude that these two trans-porters—known to be strongly expressed in brain tissue—can be principally responsible for the ‘‘high affinity’’ trans-port of glutamate also in nonneural cells. © 1998 Academic Press

Key Words: transport of acidic amino acids; gluta-mate analogues; GLAST; GLT-1; EAAC1; EAAT4; molec-ular modeling; flow cytometry; skin fibroblasts.

Cultured fibroblasts have been shown to accumu-late L-[3H]glutamate by a ‘‘high affinity’’ transport(Km , 50 mM; 1–3; for a review, see 4). While thecharacteristics of L-[3H]glutamate transport by 3T3fibroblasts have been found to be variable (2, 5, 6),possibly depending on the proliferative status of thecells (7), those of L-[3H]glutamate transport by fibro-blasts cultured from human tissue appear to be rel-atively stable; the tissue of origin has had no detect-able influence on the transport characteristics (3)and there has been no evidence of major variationsfrom one laboratory to another (1, 3). The transportprocess is Na1-dependent, it is reduced in the pres-ence of 60 mM external K1 and shows a high degreeof substrate stereoselectivity with respect to gluta-mate, but not toward aspartate, enantiomers (3). Itis inhibited by close structural analogs of glutamate andaspartate (3) such as L- and D-threo-3-hydroxyaspartates,L-trans-pyrrolidine-2,4-dicarboxylate, L-cysteate, L-cys-teinesulphinate, and DL-2-methylaspartate, but not byDL-2-aminoadipate and L-homocysteate. The apparent‘‘inhibitors’’ may actually be substrates competing withL-[3H]glutamate for transport.

The sum of the characteristics of L-[3H]glutamatetransport by cultured human fibroblasts suggests thatthe process is mediated by transport systems very sim-ilar to those known to operate in the central nervoustissue (for a review, see 4). In the present study wehave estimated the values of IC50 (concentrations caus-ing 50% inhibition), nH (Hill coefficient), and Imax (max-imum attainable inhibition) of 11 inhibitors/substratesof the uptake of 1 mM L-[3H]glutamate in confluentmonolayers of fibroblasts cultured from human skin.We have then correlated the values of IC50 of severaltypical inhibitors and, using computer-assisted molec-ular modeling, we have determined the optimum con-

1 To whom correspondence should be addressed. Fax: (1612) 93512813. E-mail: vibar@anatomy.su.oz.au.

356 0003-9861/98 $25.00Copyright © 1998 by Academic Press

All rights of reproduction in any form reserved.

ARCHIVES OF BIOCHEMISTRY AND BIOPHYSICS

Vol. 353, No. 2, May 15, pp. 356–364, 1998Article No. BB980626

formation of L-glutamate, L-aspartate, and D-aspartatein which they bind to the transporter(s).

We have also carried out a kinetic analysis of a modelin which the transporter can interact with substratesin two different ways to examine a possibility that L-and D-aspartate can bind to the recognition site on thetransporter(s) in two different orientations (not only asa-amino acids but also as ‘‘b-amino acids’’).

Finally, we have used antibodies against four glutamatetransporters (8–10) to estimate, by flow cytometry, the leveland frequency of their expression in the fibroblasts.

MATERIALS AND METHODS

Cultured cells. The fibroblasts were from human embryos 16 to 17weeks postfertilization (3). The project was approved by the HumanExperimentation Ethics Committee of The University of Sydney and inall procedures the guidelines issued by The National Health and Med-ical Research Council of Australia were strictly followed. The deeperlayers of skin were minced and digested with 0.05% w/v trypsin in 1 mMEDTA (20 ml/g of tissue for 20 min at 24°C). The trypsinisation wasterminated by addition of Dulbecco’s modified Eagle’s medium(DMEM,2 Commonwealth Serum Laboratory) containing 10% of fetalbovine serum (FBS). The suspensions were centrifuged at 400g 3 10min at room temperature. The pellets were resuspended in DMEM/FBS, transferred into 175 cm2 flasks, and cultured in a humidifiedatmosphere of 95% air/5% CO2 at 37°C. When confluent, the cells weresubcultured and the cycle of growth, confluency, and subculturing wasrepeated at least twice. The cells were then frozen and stored in liquidN2. When required, the cells were thawed, seeded at about 1 3 105/wellinto 24-well plates, allowed to grow to confluency (4–6 days), and usedfor glutamate uptake experiments.

Glutamate uptake. L-[3H]glutamate uptake was studied by amodification of previously described techniques (2, 11). The culturemedium was replaced with 0.8 ml of the incubation medium (120 mMNaCl, 4.5 mM KCl, 1.8 mM CaCl2, 1.8 mM MgCl2, and 5.5 mMD-glucose, buffered at pH 7.4 by 10 mM sodium phosphate). At thestart of the preincubation period either 100 ml of inhibitor solution(in the incubation medium, neutralized, if necessary, by NaOH) or100 ml of incubation medium (controls and blanks) were added. The24-well plates were preincubated for 5 min at 25°C in a water bath,100 ml of L-[3H]glutamate (1 mM final concentration, 1 mCi/well) wasadded, and the incubation was continued for 8 min. The plates weregently shaken by hand during both preincubation and incubation. Atthe end of the incubation the cell monolayers were rapidly washedtwice with 1 ml of the incubation medium, free of L-[3H]glutamateand inhibitors, at room temperature (typically 20–24°C). The mono-layers were extracted with 0.5 ml of 0.125 M NaOH and both radio-activity and protein were determined by scintillation counting andLowry’s method (12), respectively.

The values of IC50 were determined by GraphPad Prism softwarefrom normalized data (control 5 100) using an equation (13) modifiedto include nH and the maximum theoretically attainable inhibition(Imax, cf. Fig. 1, Table I). The F test (14) was used to establishwhether the more complex forms of the equations (i.e., those includ-

ing a term for both Imax or nH) resulted in a significantly improved fit(thus indicating whether Imax , 100%, or nH Þ 1).

Flow cytometry. The culture medium was removed from the culturewells, the cells were washed twice in phosphate-buffered saline (PBS) andremoved from the floor of the wells using 0.025% (w/v) trypsin in CMF–HBSS for 3 min at room temperature. Trypsin was inactivated by imme-diate removal of the cell suspension into DMEM/FBS. Cells were thenwashed twice in DMEM/FBS and placed into separate siliconized conicalglass tubes on ice for labeling with the relevant antibody (GLT-1, GLAST,EAAC1, or EAAT-4), as well as preimmune rabbit Ig fraction as a controlfor nonspecific fluorescent labeling. At this point the temperature of the cellsuspensions were maintained strictly between 0 and 4°C until after flowcytometric analysis. The cells were first resuspended in 100 ml of primaryantibody diluted appropriately in DMEM/FBS to give saturation labelingand incubated for 60 min and afterward centrifuged through a bed of FBSto remove the residual primary antibody and then resuspended and incu-bated in a secondary sheep anti-mouse IgG antibody and conjugated tofluorescein isothiocyanate (SAMIgG-FITC) for 60 min. After a final cen-trifugation through an FBS bed to remove the residue of the secondaryantibody, cells were resuspended in 300 ml of DMEM/FBS in preparationfor flow cytometry.

The flow cytometry was carried out using a FACSan (Becton Dick-inson) equipped with an argon ion laser set at 488 nm for excitation.Emitted fluorescence between 515 and 545 nm was measured. For-ward and side-scatter measurements were within the same range forall populations. Live cells were selected based on scatter character-istics and propidium iodide exclusion. For each sample, 10,000 gatedevents were analyzed.

Molecular modeling. A computer-assisted study was carried outon D- and L-aspartate, D- and L-glutamate, L-trans-pyrrolidine-2,4-dicarboxylate (L-t-PDC), 3-aminoglutarate, 4-methyleneglutamate, andcis-1-aminocyclobutane-1,3-dicarboxylate (cis-ABDA) using Chem-X(Chemical Design Ltd., Oxford, UK, 1994) and Chem-3D (CambridgeScientific Computing, Cambridge, MA) to determine the conforma-tion of substrates/inhibitors at the transporter site. The three-di-mensional matrices of the compounds were optimized using themolecular mechanics optimization routines in Chem-X. The conform-ers of each compound were then subjected to conformational searchroutines about the bonds at which torsional rotations were possible.A search was subsequently undertaken to determine the low energyconformation of each compound, that fitted D- and L-aspartate andL-glutamate to the conformationally restricted glutamate analog L-t-PDC. These conformations are then proposed as the optimum con-formations of each respective compound as it is interacting with thebinding site(s) at the transporter(s).

Sources of chemicals. [3,4-3H]L-glutamic acid (TRK 445, B 50L01), 1.81 TBq (49.0 Ci)/mmol was purchased from Amersham In-ternational (UK). DL-2-Aminoadipic acid, L- and D-aspartic acids,L-glutamate monopotassium salt, D-glutamic acid, 3-aminoglutaric(b-glutamic) acid, and DL-4-methyleneglutamic acid were fromSigma Chemical Company) (St. Louis, MO) and dihydrokainic acid,L- and D-threo-3-hydroxyaspartic acids, L-trans-pyrrolidine-2,4,dicar-boxylic acid, L-serine-O-sulfate, L-cysteinesulfinic acid, L-cysteic acid,and D- and L-cysteine-S-sulfates came from (Tocris Cookson, Bristol,UK). cis-1-Aminocyclobutane-1,3-dicarboxylic acid was donated byDr. Rob Allan (Department of Pharmacology, The University of Syd-ney). All other chemicals were purchased from commercial suppliersand were of at least analytical grade.

RESULTS

Under the present experimental conditions, uptake ofL-[3H]glutamate remained linear with time for at least15–20 min. Only a fraction (maximum inhibition, Imax, cf.Fig. 1) of L-[3H]glutamate uptake was susceptible to theinhibition by the acidic amino acids listed in Table I.

2 Abbreviations used: DMEM, Dulbecco’s modified Eagle’s medium;FBS, fetal bovine serum; 3-AGr (b-Gm), 3-aminoglutarate (b-glutamate);DL-g-MGm, DL-4(g)-methylene-glutamate; ABDA, 1-aminocyclobutane-2,3-dicarboxylate; DL-D- and L-t-3OHA, D- and L-threo-3-hydroxyaspartate; L-CSA, L-cysteinesulphinate; L-t-PDC, L-trans-pyrrolidine-2,4-dicarboxylate;L-SOS, L-serine-O-sulfate; PBS, phosphate-buffered, CMF-HBSSCa21,Mg21-free Haub’s Balanced Salt Solution.

357TRANSPORT OF L-GLUTAMATE BY CULTURED SKIN FIBROBLASTS

Analysis of variance using parameters from Table Iindicated that the eight most potent inhibitors could bearranged into three groups: (i) L-threo-3-hydroxyas-partate (L-t-3OHA), L-cysteinesulfinate (L-CSA), and L-cysteate; (ii) L-aspartate and D-aspartate; (iii) L-trans-pyrrolidine-2,4-dicarboxylate (L-t-PDC), D-threo-3-hy-droxyaspartate (D-t-3OHA), and L-serine-O-sulfate (L-SOS). While neither (i) nor (iii) produced IC50significantly different (P . 0.05) from those of L-gluta-mate (Fig. 1) and D-aspartate, they differed signifi-cantly, in terms of IC50s, from each other (Table I). Thevalue of IC50 for L-aspartate was significantly lowerthan that for L-t-PDC, D-t-3OHA, and L-SOS, too, butboth L-aspartate and D-aspartate were set apart fromeither (i) or (iii) and were considered as a separategroup (ii) because both enantiomers were the only in-hibitors with nH , 1 (Table I).

The remaining glutamate/aspartate analogs were ei-ther much less potent or completely ineffective as in-hibitors of L-[3H]glutamate transport: 3-aminoglut-arate (IC50 . 50 mM, Table I), DL-4-methylenegluta-mate, and cis-1-aminocyclobutane-1,3-dicarboxylate(also referred to as trans-2,4-methanoglutamate in thepharmacological literature, IC50 . 100 mM, Table I).Dihydrokainate, D-cysteine-S-sulfate, and L-cysteine-S-sulfate were weak or ineffective as inhibitors (TableI). Studies with D-glutamate (no inhibition at 250 mM)and DL-2-aminoadipate (,33% inhibition at 250 mM)were reported previously (3).

Analysis of variance indicated no statistically signif-icant variations in the value of Imax from one inhibitorto another.

The profiles in Fig. 2 show that the expression ofGLT-1, EAAC1, EAAT-4, and GLAST was clearly de-tectable above background nonspecific fluorescence.Moreover, in the case of GLT-1 and GLAST, a smallsubpopulation of cells in each case expressed signifi-cantly higher concentrations (up to 100-fold) of thesemolecules than the majority of cells from the culture.

We have modeled the conformations of various glu-tamate analogs used in this study in an attempt todetermine a model that may explain why some ana-logues apparently inhibit L-[3H]glutamate uptake infibroblasts and why other similar analogs do not. Theglutamate analogs used in this study were D- and L-aspartate, D- and L-glutamate, L-t-PDC, 3-aminoglut-arate (b-glutamate), 4-methyleneglutamate, and cis-ABDA. Most of the analogs, with the exceptions ofL-t-PDC (15) and cis-ABDA (16), are flexible moleculesand can exist in a large number of low energy confor-mations. One of the conformationally restricted com-pounds—L-t-PDC—is, however, among the stronger in-hibitors of L-[3H]glutamate uptake (group III) and thisreduces considerably the number of conformations thatcan be considered as interacting with the substrate-recognition site(s) on the transporter(s). L-t-PDC is in apartially folded conformation and may define the pu-tative conformation of substrate bound by the trans-

FIG. 1. Inhibition of L-[3H]glutamate uptake by nonradioactive L-glutamate, L-aspartate, and D-aspartate in cultured human fibroblasts.The values were normalized (control 5 100) and the points are means 6 SD of three to eight determinations (cf. Table I). The mean valuesof controls were approximately 10 pmol/mg/min and the blanks (zero time incubations) accounted for about 10% of the control values. Thevalues, pooled from three experiments, were fitted to the equation Y 5 100-Imax/(1 1 10∧(nHp(log(IC50)2X))). The fit was significantly betterthan to an equation in which Imax 5 100 (P , 0.0002, by F test) and, in the case of D-aspartate and L-aspartate, better than to an equationin which nH 5 1 (P , 0.0005, F test, cf. Table I). Cheng–Prussoff transformation (40), using a previously published value of Km (3), ispermissible in the case of L-glutamate because nH is not significantly different from unity (P . 0.05, F test). It gives Ki 5 20.0/mM.

358 COOPER ET AL.

porter(s). In this study we have used the conformationof L-t-PDC as the basis of our modeling (Fig. 3A). Thiscompound gives the relative positions of the C2 aminogroup and the two carboxylate groups.

In the case of glutamate, only the L-enantiomer in-teracts with the binding site on the transporter and itssuperimposition onto the relative positions of thecharged groups given by the alignment with L-t-PDCshould produce a putative ‘‘bound’’ conformation. Thefit resulted in a partial folding of the carbon chain(Figs. 3B and 3E). While attempting to superimposeD-glutamate—which interacts poorly or not at all withthe binding site (2, 3)—two possibilities were consid-ered. The first, the carbon backbone of D-glutamatealigned against the ‘‘bound’’ conformation of L-gluta-mate resulted in the C2 nitrogens pointing away fromeach other and the pharmacophores of D-glutamatecould not, therefore, be superimposed on those of thebound L-glutamate (Fig. 4A). The second fit superim-posed the three pharmacophores but, this time, thecarbon chains of the two enantiomers could not bealigned. In other words, the carbon chain of D-gluta-mate buckled in an area very different from that occu-pied by the C2 and C3 carbons of L-glutamate, thuspossibly precluding the interaction with the bindingsite through steric hindrance.

With aspartate, both the D- and the L-enantiomers(Figs. 3C and 3D) inhibit L-[3H]glutamate uptake in

fibroblasts. Since D- and L-aspartate have a shortercarbon chain length than glutamate, the compoundsmay fit against the putative optimum (‘‘bound’’) confor-mation of glutamate in the extended conformation. Thelow energy conformations allow D- and L-aspartate tobe first, superimposed against the ‘‘bound’’ conforma-tion of L-glutamate (Figs. 3B and 3E) by fitting thethree pharmacophores, the C2 amino and carboxylicacid groups and the terminal carboxylic acid group(Fig. 3F). Second, as both D- and L-aspartate interactwith the transporter(s), then the C2 amino and carbox-ylic acid groups and the terminal carboxylic acid groupmust be able to align with each other. For D- andL-aspartate to align, they must fit as mirror images(Figs. 3C, 3D, and 3F, cf. Ref. 17, where a similar fitwas proposed for D- and L-enatiomers of t-3OHA).

3-Aminoglutarate (b-glutamate), 4-methylenegluta-mate, and cis-ABDA were weak blockers of L-[3H]glu-tamate uptake. These compounds were superimposedwith the partially folded or ‘‘bound’’ conformation ofL-glutamate (Fig. 4). Like L-tPDC, cis-ABDA is a con-formationally restricted analog of glutamate. The threepharmacophores, the amino and the two carboxylicacid groups of cis-ABDA, appear to superimpose withthe ‘‘bound’’ conformation of L-glutamate (Fig. 4B).This fit appears very reasonable (root mean squaredistance is 0.32), however, as cis-ABDA is not a potentblocker of uptake, its structure must map an area of

TABLE I

Inhibition of 1 mM L-[3H]-glutamate Uptake in Cultured Fibroblasts by Structural Analogs of Glutamate and Aspartate

Compound IC50 (mM) Imax (%) nH NC Range (mM) n

(a) ‘‘Strong’’ inhibitors

L-CSA 4.5 6 0.8** 76.4 6 4.0 0.89 6 0.14 15 0.5–256 60L-t-3OHA 5.3 6 0.7** 75.3 6 3.6 1.26 6 0.16 9 0.25–128 68L-Aspartate 9.9 6 3.0* 73.7 6 5.1 0.59 6 0.08*** 10 1–512 44L-Cysteate 10.7 6 1.8* 70.1 6 3.3 0.89 6 0.11 9 2–512 44D-Aspartate 17.6 6 4.0 68.6 6 3.9 0.72 6 0.07*** 10 1–512 44L-SOS 36.1 6 4.5 84.9 6 4.0 1.04 6 0.12 8 4–512 47L-t-PDC 36.8 6 9.6 74.8 6 7.7 0.99 6 0.14 16 2–256 63D-t-3OHA 38.6 6 6.4 68.4 6 3.3 1.09 6 0.14 9 2–512 45

(b) ‘‘Weak’’ inhibitors

3-AGr (b-Gm) 140.8 6 27.4 71.1 6 7.2 1.35 6 0.23 18 4–800 71DL-g-MGm 155.2 6 22.8 69.2 6 4.6 1.31 6 0.23 13 20–2000 55cis-ABDA 166.8 6 33.3 75.3 6 6.1 1.18 6 0.25 13 10–2000 52

Note. Compounds very weak or inactive as inhibitors of L-[3H]glutamate uptake in cultured fibroblasts: Dihydrokainate (IC50 . 0.5 mM),D-cysteine-S-sulfate (no effect at 1 mM), L-cysteine-S-sulfate (;50% inhibition at 1 mM), D-glutamate (no inhibition at 250 mM, see Ref. 3)and DL-2-aminoadipate (,33% inhibition at 250 mM see Ref. 3). Inhibitions were measured at several concentrations within the given rangeand the data for each inhibitor were analyzed as those in Fig. 1. Each set of IC50, Imax and nH (expressed as mean 6 SE) is based on datapooled from three to five experiments; NC is the number of concentrations and n stands for the total number of experimental points includedin the analysis. The values of IC50 for 3-aminoglutarate (3-AGr), cis-ABDA, and 4-methyleneglutamate are significantly different from thevalues of IC50s of the ‘‘strong’’ inhibitors (at P , 0.01, at least).

* Significantly different from IC50 of L-t-PDC and D-t-3OHA at P , 0.01 and from IC50 of L-SOS at P , 0.05.** Significantly different from IC50 values of L-t-PDC, D-t-3OHA, and L-SOS at P , 0.001.

*** nH , 1 (P , 0.0005, F test).

359TRANSPORT OF L-GLUTAMATE BY CULTURED SKIN FIBROBLASTS

possible steric interaction that could explain the re-duced potency of this compound. The C3 carbon asshown in Fig. 4B may interact in a sterically unfavor-able area. This carbon would be in a similar position aswould be the methyl group of L-2-methylglutamate.L-2-Methylglutamate has been shown to be inactive at100 mM as either a blocker or a substrate againsteither EAAT1 or EAAT2 transporters expressed in Xe-nopus laevis oocytes (17) and 250 mM DL-2-methylglu-tamate did not inhibit uptake of 1 mM L-[3H]glutamatein 3T3 fibroblasts (2). Hence compounds that havegroups that fall into this area are potentially inactiveas substrates for the glutamate uptake system.

Like with cis-ABDA, the three pharmacophores, theC3 amino and the two carboxylic acid groups of 3-ami-noglutarate appear to superimpose with the ‘‘bound’’conformation of L-glutamate (Fig. 4C). However, in ob-taining the pharmacophores to fit, the carbon backboneof 3-aminoglutarate is arranged in such a way that theC2 and C3 carbons may fall into sterically unfavorablepositions.

With 4-methyleneglutamate, the terminal carboxy-late and the C3–C4 bond are fixed in the same plane asthe double bond between C4 and the methylene carbon(C49) (Fig. 4D). The rotation around the C4–C5 bond isnot possible and hence the terminal carboxylate may

FIG. 2. Flow cytometry of cultured human fibroblasts labeled with antibodies to glutamate transporters. The solid lines represent flowcytometry histogram profiles of human fibroblasts labeled with antibodies to GLT-1, GLAST, CEAAC, or EAAT4 followed by SAMIg–FITC.Preimmune rabbit Ig fraction followed by SAMIg–FITC was used to control for nonspecific fluorescence (broken line). Live cells were selectedbased on scatter characteristics and propidium iodide exclusion. For each sample, 10,000 gated events were analyzed. The ordinaterepresents number of fluorescent events; the abscissa represents log10 fluorescence in arbitrary units.

360 COOPER ET AL.

not be able to assume the same configuration as in thebound conformation of L-glutamate.

Figure 4G shows the superimposition of L-tPDC andthe bound conformation of L-glutamate. It also showsthe mapped areas of steric interaction. In summary,the bound conformation of L-glutamate may be a par-tially folded conformation and unfavorable interac-tions occur around the C2 carbon of L-glutamate. It isproposed that the amino group in the position a- (2-),not in the position b- (3-), is important for binding tothe site (Figs. 4E and 4F). Furthermore, this amino

group seems to have strict requirements that no othergroup can be present in the area for maximal interac-tion with the protein. Subsequently, the terminal car-boxylic acid group needs to have some flexibility tointeract with the protein.

DISCUSSION

The average Imax (about 75%) agrees very well withthe previously estimated values for the inhibitor-resis-tant portion of L-glutamate uptake in cultured humanfibroblasts (3). Resistance of uptake at 1 mM L-[3H]glu-tamate to inhibition by micromolar concentrations ofglutamate analogs does not necessarily mean that 25%of uptake under the present conditions can be ex-plained by nonspecific diffusion across the cytoplasmicmembrane; low affinity transport(s) with Km in themillimolar region may have operated in the presentexperiments (for a review, see 19) and a componentwith low affinity but sufficiently high capacity may nothave shown significant inhibition if both substrate andinhibitor were at concentrations several orders of mag-nitude lower than Km.

Most of the dose–response relationships (Fig. 1; Ta-ble I) did not indicate presence of more than one type oftransporter or of multiple populations of independenttransporter sites, all with near identical characteris-tics (nH ; 1). The only exceptions were the inhibitionsby L- and D-aspartate which produced nH , 1 (P , 0.01,F test), raising a possibility of negative cooperativityamong multiple populations of linked sites.

Previously published data (20, 21) imply that, atleast in the central nervous tissue where the structuralrequirements of L-glutamate transport have been stud-ied most extensively, substrates can interact efficientlywith the transporter(s) only if they can carry one pos-itive and two negative charges in their ionized states(for a review, see 4). The L-glutamate transport systemin human fibroblasts seems to have a similar require-ment and, furthermore, the molecular modeling anal-ysis based on the present experimental data demon-strates that the conformations of substrates recognizedby the binding site(s) on the transporter(s) are confinedto a rather narrow range. This raises a particular prob-lem; when Christensen et al. (22) discussed the ‘‘anom-aly’’ in the stereoselectivity of glutamate transport sys-tem (both enantiomers of aspartate but only L-, and notD-, glutamate are transported; 4, 20, 23), they proposedthat D-aspartate is recognized by the substrate bindingsite as a b- rather than an a-amino acid, a configura-tion not possible with D-glutamate. Given the strictrequirements for the optimum conformation of boundsubstrate it appears unlikely that either L- or D-aspar-tate could interact with the binding site(s) equallystrongly in the ‘‘b-amino acid’’ as in the ‘‘a-amino acid’’orientation.

FIG. 3. L-tPDC (A) is a conformationally restricted analogue ofglutamate. It is an active blocker of L-[3H]glutamate uptake in fibro-blasts and reduces considerably the number of conformations thatmay act at the active site. L-tPDC is in a partially folded conforma-tion and may define the putative conformation that blocks L-[3H]glu-tamate uptake. In this study, we use the conformation of L-tPDC asthe basis of our modeling. This compound gives the relative positionsof the C2 amino group and the two carboxylate groups. L-glutamate(B), D-aspartate (C), and L-aspartate (D) are flexible compounds andcan exist in a large number of low energy conformations. WhenL-glutamate is superimposed against the charged groups, the nitro-gen of the pyrrolidine ring and the two carboxylate groups, of L-tPDC, the fit resulted in a partial folding of the carbon chain ofL-glutamate (3). This may represent the ‘‘bound’’ conformation ofL-glutamate to the active site. D- and L-aspartate are superimposedagainst the charged groups, the C2 amino and carboxylate and theterminal carboxylate groups, of L-glutamate, the compounds fitagainst the ‘‘bound’’ conformation of L-glutamate in an extendedconformation (F). This conformation allows both enantiomers to bindto the site as mirror images.

361TRANSPORT OF L-GLUTAMATE BY CULTURED SKIN FIBROBLASTS

In fact, a poor recognition of D- and L-aspartate asb-amino acids by the binding site(s) on the transport-er(s) is consistent with the data obtained using 3-ami-noglutarate (‘‘b-glutamate’’; see also 2): 3-aminoglu-tarate is a much less potent inhibitor of L-[3H]glutamateuptake than either L- or D-aspartate. It is, therefore,very improbable that the lack of stereoselectivity of thetransport system with respect to the enantiomers ofaspartate can be explained by the binding of D-aspar-tate in the b-orientation. Molecular modeling analysessupport this conclusion (Figs. 4E, 4F, and 4G).

It has been conjectured that a bimodal binding mech-anism with a modest component of binding in the b-ori-entation competing with the dominant form of bindingin a-orientation could manifest itself by ‘‘negative co-operativity’’ (nH , 1) under some circumstances (24).In fact, such a proposal would have to be further re-fined to account for the observed actions of L-cysteate,L-CSA, and L-t-3OHA by postulating that either thepresence of a 3-substituent or characteristics of thesulfate and sulfinate groups could preclude the bindingof L-cysteate, L-CSA, and L-t-3OHA in the b-configura-tion, allowing only a single mode—the more efficientone—of binding (a-orientation), resulting in low IC50sand nH ; 1. We have examined this hypothesis inkinetic terms by deriving a general rate equation forthe transport of a substrate that can bind to a carrier intwo different ways. The results of this analysis arepresented in the Appendix where it is shown that thesteady-state rate equation for the transport of a sub-strate or a ligand that can bind to a carrier in twodifferent orientations is not, in terms of observableparameters, distinguishable from the steady-state rateequation describing the conventional carrier modelwhich has the substrate binding to the carrier in onlyone orientation. Thus the particular characteristics ofL- and D-aspartate as substrates/inhibitors require analternative explanation.

It should be stressed that the structural require-ments of L-glutamate transport observed in the presentexperiments may not exactly match those of the mostcommonly expressed transporters in the central ner-vous tissue. D-t-3OHA was only marginally weakerthan L-t-3OHA as an inhibitor of [3H]L-glutamate up-take in the slices of rat cerebral cortex (17); the onlypart of the brain where the difference between D- andL-enantiomers of t-3OHA was reported to be about asgreat as that found in the present experiments, was the

FIG. 4. D-Glutamate (A), cis-ABDA (B), 3-aminoglutarate (b-gluta-mate) (C), and 4-methyleneglutamate (D) were inactive or weakblockers of L-[3H]glutamate uptake. These compounds were super-imposed with the ‘‘bound’’ conformation of L-glutamate. The fit ofD-glutamate with the bound conformation of L-glutamate (A) alignedthe carbon backbone of D-glutamate against L-glutamate. As a result,the C2 nitrogens pointed away from each other and the pharmacoph-ores of D-glutamate could not, therefore, be superimposed on those ofthe bound L-glutamate. The fit of cis-ABDA with the bound confor-mation of L-glutamate (B) aligned the three pharmacophores. cis-ABDA maps an area of possible steric interaction that could explainthe reduced potency of this compound. The C3 carbon may interact ina sterically unfavorable way with the protein. The fit of 3-aminoglu-tarate (b-glutamate) with the ‘‘bound’’ conformation of L-glutamate(C) aligned the three pharmacophores; however, the carbon back-bone of 3-aminoglutarate is arranged in such a way that the C2 andC3 carbons may fall into sterically unfavorable positions. The fit of4-methyleneglutamate with the bound conformation of L-glutamate(D) showed the C2 amine and carboxylate groups to align well butthe third pharmacaphor, the terminal carboxylate, to be poorlyaligned. This carboxylate group and the C3–C4 bond are fixed in thesame plane as the double bond between C4 and the methylenecarbon (C49). Rotation around the C4–C5 bond is not possible andhence the terminal carboxylate may not be able to assume the sameconfiguration as in the bound conformation of L-glutamate. The b-ori-entations of L- and D-aspartate are shown in (E) and (F), while (G)

shows the superimposition of L-tPDC and the bound conformation ofL-glutamate with the mapped areas of steric interaction around theC2 carbon of L-glutamate (spotted circles) derived from (A), (B), and(D). It is apparent that c2 and c3 carbon atoms of either L- orD-aspartate bound in the b-orientations may fall within the areaswhere they cause steric hindrance.

362 COOPER ET AL.

cerebellum (24, 25). Testing of D- and L-aspartate de-rivatives with 3-substituents bulkier than hydroxylgroups (26) may reveal more about the differences be-tween the structural requirements of glutamate trans-porters expressed, respectively, in the CNS and in non-neural tissues.

Finally, the assumption that the transport system incultured fibroblasts consists of a single set of transport-ers may be an oversimplification; immunocytochemicaldata using flow cytometry (Fig. 2) and antibodiesagainst peptide sequences characteristic to transport-ers cloned from the central nervous tissue (27) indi-cated the presence of four distinct transporter mole-cules (GLT-1, GLAST, EAAC1, and EAAT-4) in thecultured human fibroblasts. The differences observedcomparing the profiles of GLT-1 and GLAST with thoseof EAAC1 and EAAT-4 suggest that a minority sub-population of these cells express relatively very highconcentrations of GLT-1 and GLAST, compared to therespective main populations. The possibility exists,therefore, that much of the glutamate uptake is takingplace in those cells with very high transporter concen-trations.

Defects of glutamate transport implicated in the eti-ology of neurological disorders (9, 28, 29) have beenlinked to specific pathogenic factors in the cellular en-vironment (30–33) or to aberrant expression of individ-ual transporter molecules (9, 10, 34) rather than toerrors in the genome (35). However, little is knownabout the posttranslational processing of individualtransporter molecules and their normal arrangementwithin the healthy cytoplasmic membrane (36, 37).Furthermore, a recent study has employed detailedinvestigation of substrate and structural selectivity ofglutamate transport system to reveal subtle alter-ations specifically associated with the regions of thehuman cerebral cortex most affected by the loss ofneurons in Alzheimer dementia (29). Consequently,studies similar to the present one may provide impor-tant clues necessary for the understanding of themechanisms involved in neurodegenerative disorders.

APPENDIX

In order to allow for a transport protein to bind atransportable substrate in two different ways, we ex-tend the conventional carrier model,

E1 E2

ES1

S1 S2

ES2,

so that it includes two different pairs of complexedtransport protein species ES1 and ES2 for the substrate

binding in one specified way and ES91 and ES92 for thesubstrate binding the other way:

E1 E2

ES1

S1 S2

ES2.

ES1 99 ES2

S1 S2

As is customary, the transport protein can exist intwo different conformations, E1 and E2, with E1 bind-ing the transportable substrate, at concentration [S1],on Side 1 of the membrane and E2 binding it at theconcentration [S2] on the other side. We can choose amodel with fewer intermediates (38), since there are nodifferences in the steady state solution, in terms ofobservable parameters, between the above model andthe model with fewer intermediates,

E1 ,E2

k5

k10[S2]

k4[S2]k2

k7[S1]

k1[S1] k3

k9

k8

k6

ES9

ES

where k1, k2, . . . , k6 are rate constants for the stepsindicated in the above scheme.

If v132 is the unidirectional flux of transportablesubstrate from solution 1 to solution 2, then v132 canbe measured experimentally by adding radioactivelylabeled substrate, S*1, to solution 1 only and then mea-suring initial rates of appearance of radioactivity insolution 2. Under these conditions, for the simplifiedmodel, we have

v13 2 5 2 d@S*1#/dt 5 d@S*2#/dt 5 k1@E1#@S*1# 2 k2@ES*#

1 k7@E1S*1# 2 k8@ES'*# 5 k3@ES*# 1 k9@ES'*#.

Expressions for [ES*] and [ES9*] can be obtained whenwe apply steady-state conditions to the differentialequations describing the changes in concentrations ofspecies ES* and ES9*, namely

d@ES*#/dt 5 k1@E1#@S*1# 2 k2@ES*# 2 k3@ES*# 5 0

d@ES'*#/dt 5 k7@E1#@S*1# 2 k8@ES'*# 2 k9@ES'*# 5 0,

363TRANSPORT OF L-GLUTAMATE BY CULTURED SKIN FIBROBLASTS

which allow us to write

@ES*# 5 k1@E1#@S*1#/~k2 1 k3!

@ES'*# 5 k7@E1#@S*1#/~k8 1 k9!,

Substitution of these expressions for v132 gives

v13 2 5 ~k1k3/~k2 1 k3! 1 k7k9 /~k8 1 k9!!@E1#@S*1#.

An expression for [E1] in terms of the concentrations ofthe unlabeled substrate molecules on both sides of themembrane can be most easily found by applying apublished procedure (39) to the above simplified model.This allows us to write

v13 2 5 ~k1k3/~k2 1 k3! 1 k7k9/~k8 1 k9!!@S*1#~~k8k10~k2 1 k3!

1 k2k4~k8 1 k9!!@S2# 1 k5~k2 1 k3!~k8 1 k9!!n/¥,

where n is the total concentration of carriers and S is asum of terms involving constant terms alone, constantterms multiplied by [S1], constant terms multiplied by[S2], and constant terms multiplied by both [S1] and [S2].The form of the equation is such that when it is expressedin terms of observable parameters, it is not distinguish-able from the steady state equation describing the con-ventional carrier model which binds the substrate in onlyone way (38).

REFERENCES

1. Dall’Asta, V., Gazzola, G. C., Franchi-Gazzola, R., Busolatti, O.,Longo, N., and Guidotti, G. C. (1983) J. Biol. Chem. 258, 6371–6379.

2. Balcar, V. J. (1992) FEBS Lett. 300, 203–207.3. Balcar, V. J., Shen, J., Bao, S., and King, N. J. C. (1994) FEBS

Lett. 339, 50–54.4. Balcar, V. J., and Li, Y. (1992) Life Sci. 51, 1467–1478.5. Faivre-Baumann, A., Rossier, J., and Benda, P. (1974) Brain.

Res. 76, 371–375.6. Frame, M. C., Freshney, R. I., Vaughan, P. F. T., Graham, D. R.,

and Shaw, R. (1984) Br. J. Cancer 49, 269–280.7. Busolatti, O., Uggeri, J., Rotoli, B. M., Franchi-Gazzola, R., and

Gazzola, G. C. (1993) Biochim. Biophys. Acta 1151, 153–160.8. Rothstein, J. D., Martin, L., Levey, A. I., Dykes-Hoberg, M., Jin, L.,

Wu, D., Nash, N., and Kuncl, R. W. (1994) Neuron 13, 713–725.9. Rothstein, J. D., Van Kammen, M., Levey, A. I., Martin, L. J.,

and Kuncl, R. W. (1995) Ann. Neurol. 38, 73–84.10. Rothstein, J. D., Dykes-Hoberg, M., Pardo, C. A., Bristol, L. A.,

Jin, L., Kanai, Y., Hediger, M. A., Wang, Y., Schielke, J., andWelty, D. F. (1996) Neuron 16, 675–685.

11. Westergaard, N., Beart, P. M., and Schousboe, A. (1993) J. Neu-rochem. 61, 364–367.

12. Lowry, O. H., Rosebrough, N. J., Farr, A. L., and Randall, R. J.(1951) J. Biol. Chem. 193, 265–275.

13. McGonnigle, P., and. Molinoff, P. B. (1989) in Basic Neurochem-istry, 4th ed., pp. 183–202, Raven Press, New York.

14. Munson, P. J., and Rodbard, D. (1980) Anal. Biochem. 107,220–239.

15. Bridges, R. J., Stanley, M. S., Anderson, M. W., Cotman, C. W.,and Chamberlin, A. R. (1991) J. Med. Chem. 34, 717–725.

16. Allan, R. D., Hanrahan, J. R., Hambley, T. W., Johnston,G. A. R., Mewett, K. N., and Mitrovic, A. D. (1990) J. Med. Chem.33, 2905–2910.

17. Balcar, V. J., Johnston, G. A. R., and Twitchin, B. (1977) J. Neu-rochem. 28, 1145–1146.

18. Vandenberg, R. J., Mitrovic, A. D., Chebib, M., Balcar, V. J., andJohnston, G. A. R. (1997) Mol. Pharmacol. 51, 809–819.

19. Erecinska, M., and Silver, I. A. (1990) Prog. Neurobiol. 35, 245–296.20. Balcar, V. J., and Johnston, G. A. R. (1972) J. Neurochem. 19,

2657–2666.21. Robinson, M. B., Sinor, J. D., Dowd, L. A., and Kerwin, J. F., Jr.

(1993) J. Neurochem. 60, 167–179.22. Christensen, H. N., Greene, A. A., Kakuda, D. K., and MacLeod,

C. L., (1994) J. Exp. Biol. 196, 297–305.23. Gazzola, G. C., Dall’Asta, V., Bussolati, O., Makowske, M., and

Christensen, H. N. (1981) J. Biol. Chem. 256, 6054–6059.24. Killinger, S., Blume, G. L., Bohart, L., Bested, A., Dias, L. S.,

Cooper, B., Allan, R. D., and Balcar, V. J. (1996) Neurosci. Lett.216, 101–104.

25. Mitrovic, A. D., and Johnston, G. A. R. (1994) Neurochem. Int.24, 583–688.

26. Lebrun, B., Sakaitani, M., Shimamoto, K., Yasuda-Kamatami,Y., and Nakajima, T. (1997) J. Biol. Chem. 272, 20336–20339.

27. Gegelashvili, G., and Schousboe, A. (1997) Mol. Pharmacol. 52,6–15.

28. Dodd, P. R., Scott, H. L., and Westphalen, R. I. (1994) Neuro-chem. Int. 25, 203–219.

29. Scott, H. L., Tannenberg, A. E. G., and Dodd, P. R. (1994)J. Neurochem. 64, 2193–2202.

30. Fine, S. M., Angel, R. A., Perry, S. W., Epstein, L. G., Rothstein,J. D., Dewhurst, S., and Gelbard, H. A. (1996) J. Biol. Chem. 271,15303–15306.

31. Harris, M. E., Wang, Y., Pedigo, N. W., Jr., Hensley, K., Butter-field, D. A., and Carney, J. M. (1996) J. Neurochem. 67, 277–286.

32. Smith, M., Harris, P. L. R., Sayre, L. M., Beckman, J. S., andPerry, G. (1997) Widespread peroxynitrite-mediated damage inAlzheimer’s Disease. J. Neurosci. 17, 2653–2657.

33. Trotti, D., Rossi, D., Gjesdal, O., Levy, L. M., Racagni, G., Danbolt,N. C., and Volterra, A. (1996) J. Biol. Chem. 271, 5976–5979.

34. Lin, C. G., Bristol, L. A., and Rothstein, J. D. (1996) Soc. Neu-rosci. Abstr. 22, 2142.

35. Meyer, T., Lenk, U., Kuther, G., Weindl, A., Speer, A., andLudolph, A. (1995) Ann. Neurol. 37, 817–819.

36. Hucho, F., and Tsetlin, V. (1996) J. Neurochem. 66, 1781–1792.37. Haugeto. Ø., Ullensvanf, K., Levy, L. M., Chaudry, F. A., Honore,

T., Nielsen, M., Lehre, K. P., and Danbolt, N. (1996) J. Biol.Chem. 271, 27715–27722.

38. Lieb, W. R. (1982) in Red Cell Membranes—A MethodologicalApproach (Ellory, J. C., and Young, J. D., Eds.), pp. 135–164,Academic Press, London.

39. King, E. L., and Altman, C. (1956) J. Phys. Chem. 60, 1375–1378.40. Cheng, Y. C., and Prusoff, W. H. (1973) Biochem. Pharmacol. 22,

3099–3108.

364 COOPER ET AL.