Substrate specificity of the Escherichia coli 4-aminobutyrate carrier encoded by gabP. Uptake and counterflow of structurally diverse molecules.

Transport of 4-aminobutyrate into Escherichia coli is catalyzed by gab permease (GabP). Although published studies show that GabP is relatively specific, recognizing the common α-amino acids with low affinity, recent work from this laboratory indicates that a number of synthetic compounds are high affinity transport inhibitors (50% inhibition at 5-100 μM). Here we present evidence that many of these structurally heterogeneous compounds not only inhibit transport but also function as alternative GabP substrates (i.e. a set of observations inconsistent with the idea that the core of the GabP transport channel exhibits rigid structural specificity for the native substrate, 4-aminobutyrate).

The gab gene cluster is required for metabolism of 4-aminobutyrate in Escherichia coli. The cluster consists of a regulatory gene, gabC, two structural genes (gabD and gabT) encoding the metabolic enzymes, succinic semialdehyde dehydrogenase, and glutamate:succinic semialdehyde transaminase, and a third structural gene (gabP), encoding the 4-aminobutyrate transporter (gab permease or GabP). The GabP is a hydrophobic, 466-residue polypeptide (7) that is readily modeled as a transmembrane protein consisting of 12 transmembrane ␣-helical segments. The permease is active in whole cells as well as in rightside-out vesicles (3), and uptake of 4-aminobutyrate is stimulated by membrane potential and abolished by proton ionophores (7). Recently, we showed that a number of synthetic compounds are potent GabP inhibitors (8,9). An unanswered question is whether any of these inhibitors might in fact be transported substrates of GabP. Here, we provide evidence consistent with the hypothesis that GabP transports at least nine different substrate analogs.
It is shown directly with radiolabeled compounds that two bulky and structurally distinct heterocyclic molecules are efficiently transported by the 4-aminobutyrate transporter expressed under control of a lac promoter. Transport of 3-piperidine carboxylic acid or 3-hydroxy-5-aminomethylisoxazole was dependent upon either the presence of plasmid-borne gabP or upon induction of gene expression with isopropyl-␤-D-thiogalactopyranoside. Transport of either novel substrate was inhibited by the native substrate (4-aminobutyrate) over a range of concentrations consistent with the transport K m .

Materials-[
used to grow the Escherichia coli strains SK45 and SK55 for use in experiments with cloned gabP. Cells grown overnight (16 h) were diluted 100-fold into fresh medium (approximately Klett 10) containing 1 mM IPTG to induce high level gabP expression. Cells were grown for three to four doublings (Klett 130 -150 with number 42 blue filter). Transport-[ 3 H]GABA or [ 3 H]nipecotic acid transport was studied under conditions in which gabP was expressed from the lac-inducible plasmid, pSCK-472A, contained in E. coli strain SK55. Log phase cells were processed by washing twice with 100 mM potassium phosphate (pH 7.0). The cell pellet was resuspended in this buffer using 25-50% of the original culture volume such that the protein concentration was 1-2 mg/ml; such cells are referred to hereafter as "washed cells." Transport reactions were initiated by adding 80 l of washed cells with rapid vortex mixing to 20 l of solution containing the radiolabeled substrate and other additions (conditions indicated in the figure legends). A metronome was used to reproducibly time short uptake reactions (2-20 s). Uptake was rapidly quenched by adding 1 ml of a "stop solution" (buffer containing 20 mM HgCl 2 ) to the rapidly mixing reaction vessel. The "quenched" reaction was vacuum-filtered. The reaction vessel was then washed with 1 ml of 100 mM potassium phosphate (pH 7.0), 5 mM HgCl 2 , and this was applied to the same filter. Finally, the filter was washed with 4 ml of buffer containing 5 mM HgCl 2 . The filter was dissolved in scintillation mixture, and the radioactivity (disintegrations per minute) was calculated by a Beckman LS3801 scintillation spectrometer using stored quench curves and automatic quench compensation (H number determination).
Cell Water-Cytoplasmic water was taken as the difference between the total aqueous space, measured with [ 3 H]H 2 O, and the non-cytoplasmic [ 14 C]taurine space (1,2). Labeled cells were separated from the bulk aqueous medium by centrifugation through a silicone oil mixture, which was empirically adjusted to an appropriate specific gravity (i.e. we mixed 7 parts 550 fluid with 3 parts 510 fluid, and after phase separation the dense fraction was harvested and used in the experiments). Disintegrations per minute for each isotope were calculated as described above. Cells utilized in this study contained about 5 l of cytoplasmic water/mg of protein, such that the typical data point represents about 0.5-1.0 l of cytoplasmic water and 0.1-0.2 mg of protein.
The above measured values are roughly equivalent to 0.6 -1.2 mg of cells (wet weight).
Counterflow Studies-Twice washed (100 mM potassium phosphate, pH 7.0) cells (approximately 2000 Klett ϫ ml or 3 mg of protein) were resuspended at room temperature in 5 ml of the same buffer containing 30 mM sodium azide. After incubation for 5 min at room temperature, the cells were centrifuged at 4°C and resuspended in 100 l of buffer having either the same composition as the supernatant or with this buffer plus the preloaded substrate (10 mM). The cells were placed on ice until use. Counterflow was initiated by rapidly diluting 5 l of the cells with 1 ml of 100 mM potassium phosphate (pH 7.0) containing 30 mM sodium azide and [ 3 H]GABA as detailed in the figure legends. Counterflow was terminated by quenching with 1 ml of 100 mM HgCl 2 . The "quenched" reaction was vacuum-filtered. The reaction vessel was then washed with 1 ml of 100 mM potassium phosphate (pH 7.0), 5 mM HgCl 2 , and this was applied to the same filter. Finally, the filter was washed with 4 ml of buffer containing 5 mM HgCl 2 . The filter was dissolved in scintillation mixture, and the radioactivity (disintegrations per minute) was quantitated as described above.
Analysis-The counterflow results have been expressed as a "GABA uptake ratio" to avoid any implication that the intracellular concentration of [ 3 H]GABA is known in absolute terms. The GABA uptake ratio in poisoned cells is the [ 3 H]GABA uptake achieved by preloaded SK55 cells divided by the average "equilibration" level of [ 3 H]GABA uptake achieved by non-preloaded SK55 cells. Thus, the GABA uptake ratio for non-preloaded cells is theoretically unity (and in practice it is close to unity). The GABA uptake ratio is often less than unity for the SK45 strain since there is no basis (no GabP) for rapid equilibration in these cells. Values substantially greater than unity require coupling between efflux of the preloaded compound and influx of [ 3 H]GABA. 2

RESULTS AND DISCUSSION
Nitrogen-limiting conditions induce in E. coli the ability to utilize 4-aminobutyrate as a source of carbon and nitrogen (3)(4)(5)(6). This capability derives jointly from induction of the requisite metabolic enzymes as well as from induction of a permease that catalyzes accumulation of 4-aminobutyrate from the environment. Previous studies (3,7) agree that the GabP exhibits biological specificity, rejecting common constituents from the cellular milieu (e.g. the 20 common ␣-amino acids, ornithine, putrescine) that exhibit at least superficial structural similarity to 4-aminobutyrate.
On the other hand, studies (8,9) show that a number of synthetic compounds (e.g. Fig. 1) are potent GabP inhibitors. Although these molecules bear some structural resemblance to the native substrate, 4-aminobutyrate (compound 4), the degree of dissimilarity was sufficient to warrant speculation that some of these molecules (particularly the heterocycles) might act at an allosteric inhibitory site rather than within the GabP transport channel itself. The experimental evidence presented 2 Transmembrane passage or "transport" of the preloaded compound is suggested by both the rising and the falling phases of the counterflow time course. The rising phase reflects trans-stimulation of uptake mediated by rapid exchange of the preloaded substrate for extracellular [ 3 H]GABA. That the unlabeled compound can participate in the exchange reaction implies that it is a substrate capable of transmembrane passage via the same carrier (i.e. GabP) that transports [ 3 H]GABA (10 -14). The transient nature of intracellular [ 3 H]GABA accumulation suggests additionally that the carrier (i.e. GabP) has the capacity (i) to catalyze net efflux of the preloaded substrate (reflected in the falling phase) and (ii) to transduce energy, available in the transmembrane gradient of preloaded substrate, to drive concentrative uptake of [ 3 H]GABA (reflected in the peak "GABA uptake ratio" defined under "Experimental Procedures"). a This is the IPTG-inducible expression vector without the gabP insert in its polylinker. here fails to support this hypothesis for at least nine such inhibitors.
Several lines of evidence suggest that both [ 3 H]nipecotic acid (compound 7) and [ 3 H]muscimol (compound 3) are transported quite specifically by GabP (Figs. 2 and 3, respectively). Transport of these compounds depended strongly upon either the presence of plasmid-borne gabP (Fig. 2) or upon specific induction of gabP expression by IPTG (Fig. 3). Transport of [ 3 H]GABA, the native substrate, was likewise induced by IPTG in this expression system (8). Moreover, 10 M GABA was found to inhibit transport of either [ 3 H]nipecotic acid (Fig. 2, inset) or [ 3 H]muscimol (Fig. 3, inset) by about 50%, an effect consistent with the published K m (10 -12 M) of the GabP for GABA (3,7). The observed transport of [ 3 H]nipecotic acid and [ 3 H]muscimol supports the proposal that these heterocyclic compounds occupy the GabP transport channel (rather than an allosteric site) to inhibit [ 3 H]GABA uptake. Additionally, it will be noted (Fig. 4) that nipecotic acid (compound 7) contains within its structure the basic elements of two open-chain amino acids, 5-aminopentanoic acid (compound 4) and 3-aminopropanoic acid (compound 10). Thus, nipecotic acid has the potential to be recognized by GabP in the context of either compound 4 or compound 10 (or both). Indeed, compounds 4 and 10 were found to inhibit [ 3 H]nipecotic acid transport (Fig. 5)   are not available in labeled form, but transport has be inferred from the participation of these molecules in the counterflow reaction (Fig. 7). It will be noted that compound 2 is incorporated into the compound 8 ring structure. Likewise, compounds 4 and 10 are incorporated into the compound 7 ring structure. In addition to participating in counterflow, all of these molecules inhibit either [ 3 H]GABA or [ 3 H]nipecotic acid transport mediated by GabP. metabolically poisoned SK55 cells (Fig. 6). Then the capacity of many other nonradiolabeled transport inhibitors to support GabP-mediated counterflow was investigated (see Fig. 7 for summary).
Counterflow with Known Substrates-Preloading metabolically poisoned E. coli SK55 with either the native substrate, GABA (compound 2), or the alternative substrates, nipecotic acid (compound 7) and muscimol (compound 3), resulted in transient [ 3 H]GABA accumulation, which was not evident in SK55 cells that were not preloaded or in SK45 control cells lacking a functional GabP. 4-aminobutyrate is, of course, the best known GabP substrate (3,7,8), and transport of this molecule is clearly reflected in the counterflow data (Fig. 6). Likewise transport of compounds 3 and 7, shown here to be alternative GabP substrates (Figs. 2 and 3), is clearly reflected in the counterflow data (Fig. 6). This well known (10,(12)(13)(14) correlation between counterflow in metabolically poisoned cells and uptake in metabolically active cells can thus reasonably 3 be used to implicate the GabP in the transport of other com-pounds that are not readily available in radioactive form (9). The results suggest that effective inhibitors are not always effective counterflow substrates (Fig. 7).
Within this series of unsaturated test compounds, those that failed to support counterflow share a common structural element (Fig. 8). The carbonyl carbon and the amine moiety are in the trans configuration relative to the double bond that these analogs share in common. It will be noted that the native substrate, 4-aminobutyrate (compound 2), rotates freely about the C2-C3 bond so that the molecule can assume conformations similar to either the cis or the trans isomers of 4-aminobutenoic acid. The profound functional differences between these conformationally restricted cis and trans isomers may imply something about the conformation(s) of GABA that are preferred by GabP.
Indeed, with appropriate bond rotations, the electronegative atoms in 4-amino-cis-butenoic acid (compound 11) can be made nearly isosteric with those of another conformationally restricted GABA analog, cis-3-aminocyclohexyl carboxylic acid (Fig. 4, compound 8) so that either molecule might be imagined to dock (hydrogen bond) with the same complementary surface (e.g. GabP). 4 Like compound 11 (4-amino-cis-butenoic acid), compound 8 (cis-3-aminocyclohexyl carboxylic acid) is not only 3 Counterflow provides only an indirect indication that GabP can carry a particular preloaded substrate across the membrane. Under what circumstances might the presence or absence of a counterflow phenomenon be misleading? A genuine substrate having poor affinity for the inwardly oriented carrier might not significantly entrap [ 3 H]GABA entering from the outside, i.e. a small ␥ factor in the formalism of Wright (15). Alternatively, rapid degradation of a preloaded substrate could prevent observation of counterflow. However, it should be noted in this regard that GABA, clearly metabolized by GabT (16), nevertheless survives metabolic degradation long enough to allow experimental demonstration of counterflow (Fig. 6). Other investigators (3) using thin layer chromatography to analyze the state of intracellular GABA have also concluded that there is little degradation over the course of a transport experiment. A positive counterflow result (and the associated trans-stimulation) is more difficult to explain as an artifact. Possibly, a nonsubstrate inhibitor molecule might be metabolized to a substrate capable of trans-stimulating [ 3 H]GABA uptake. However, we doubt that such a fortuitous mechanism could occur frequently enough to make it an attractive artifactual explanation for the present results (Fig. 7) in which nine different compounds were shown to support counterflow. Certainly this artifactual mechanism is unattractive for compounds 2, 3, and 7 since direct methods have shown these to be GabP substrates. 4 The idea that GabP prefers a nonextended form of GABA may help to rationalize why compound 10 can be a substrate as well as why compounds 5 and 7 are preferred (8,9) over compounds 14 and 15 (Fig.  7) even though the latter pair are more analogous to the extended conformation of GABA. Model building in fact substantiates that after appropriate rotations, the hydrogen bonding elements of compounds 4, 5, 7, and 10 can be made nearly isosteric with those of 4-amino-cisbutenoic acid (compound 11), a result consistent with the experimental observation that all five of these molecules are GabP substrates. a GabP inhibitor but also a counterflow substrate (Fig. 7). When preloaded into metabolically poisoned E. coli SK55, compound 8 stimulated a transient accumulation of [ 3 H]GABA, which was not observed in non-preloaded cells. It will be noted (Fig. 4) that part of the compound 8 ring structure indeed mimics the native substrate, GABA (compound 2), which is a high affinity ligand (Fig. 2). The remainder of the compound 8 ring mimics 6-aminohexanoic acid (compound 12), a relatively nonpotent inhibitor (Fig. 7) of GabP. We found no evidence that compound 12 could drive [ 3 H]GABA counterflow.
Other nonradiolabeled compounds of interest are 5-aminopentanoic acid (compound 4) and 3-aminopropanoic acid (compound 10). As noted compound 4, a high-affinity GabP inhibitor (Fig. 5), can be recognized within the ring structure (Fig. 4) of FIG. 7. Summary of Inhibitory potency and counterflow of GabP ligands. Dose-response curves were generated to assess the ability of each numbered compound to inhibit GabP-mediated transport in E. coli SK55. The inhibitory potency of compound 2, the native substrate, was assessed using 100 M [ 3 H]nipecotic acid as the substrate (compound 7). The remaining dose-response curves were generated to assess the ability of different compounds to inhibit transport of 10 M [ 3 H]GABA. These data have been arranged according to the inhibitor IC 50 value, which is defined as the inhibitor concentration that reduces transport to 50% of the noninhibited control. Compounds represented by white bars behaved as counterflow substrates. Approximate peak values for counterflow GABA uptake ratio (defined under "Experimental Procedures") were as follows: compound 2, 14-fold; compound 3, 14-fold; compound 4, 30-fold; compound 5, 22-fold; compound 7, 10-fold; compound 8, 30-fold; compound 9, 15-fold; compound 10, 3-fold; compound 11, 16-fold. In similar studies, no counterflow evidence was obtained to support the hypothesis that the remaining compounds represented by dark bars are transportable by GabP. Chemical names of compounds are given in Footnote 1.
FIG. 8. Structure of unsaturated 4-aminobutyrate analogs. Compound 2 is 4-aminobutyrate (GABA), the native GabP substrate. Free rotation about the C2-C3 bond enables GABA to assume conformations similar to either the cis (compound 11, a counterflow substrate) or the trans (compound 18) isomers of 4-aminobutenoic acid. The structure of compound 18 (a nonsubstrate) is apparent (shaded) in the structures of several other molecules, compounds 1, 6, and 14 (9), which likewise failed to behave as counterflow substrates. Interestingly, two of these molecules, compounds 1 and 6, were rather potent inhibitors of [ 3 H]GABA transport (Fig. 7). Compound 3 is a relatively potent inhibitor and behaves as a substrate in both direct uptake assays (Fig. 3) and in counterflow (Fig. 6). the transported substrate, nipecotic acid (compound 7). The remainder of the nipecotic acid ring mimics compound 10, a moderately potent (Fig. 5 (Fig. 9). Thus, both 5-aminopentanoate and 3-aminopropanoate behave as though they are compounds that GabP can recognize, transport, and couple to a GABA counterflux.
As mentioned, 5-aminopentanoate (compound 4) and 4-aminobutyrate (compound 2) are more potent GabP inhibitors (Fig.  7) than either the longer (6-aminohexanoate, compound 12) or the shorter (3-aminopropanoate, compound 10) open-chain amino acids (8). The separation distance between the amino and carboxyl groups could be a possible basis for the different inhibitory potencies. However, the behavior of these compounds in counterflow suggests that the hydrocarbon skeleton is also significant.
Like 3-aminopropanoate (compound 10), 3-aminobutyrate (compound 9) is only a moderately potent GabP inhibitor (Fig.  7). On the other hand, the extra methyl group in compound 9 preserves the charge separation relative to compound 10 and improves counterflow substantially (Fig. 9). The observed transient accumulation of [ 3 H]GABA and trans-stimulation of [ 3 H]GABA uptake are consistent with the notion that both 3-aminobutyrate (compound 9) and 3-aminopropanoate (compound 10) are molecules that GabP can recognize, transport, and couple to a GABA counterflux. That GabP effectively recognizes compound 9 (3-aminobutyrate) is consistent with other evidence (see Fig. 8 and Footnote 4) that the permease prefers analogs that mimic a nonextended conformation of the native substrate, 4-aminobutyrate. 4 Irrespective of detailed mechanism, the data presented here support a structure/function model in which numerous molecules having different sizes and shapes are capable of interacting with GabP either as inhibitory ligands or as transported substrates. Perhaps most significantly, the counterflow assay has provided the first evidence that several potent to moderately potent inhibitory compounds, unavailable in radiolabeled form, are likely to be transported substrates capable of traversing the core of the GabP transport channel. On the other hand, the counterflow assay also suggested that some potent inhibitors (compounds 1 and 6) that mimic the extended conformation of 4-aminobutyrate do not appear to be counterflow substrates.
In short, a broad range of structurally and perhaps functionally distinct GabP ligands has been identified. The availability of these ligands may allow successful implementation of strategies to select substrate specificity mutants and/or strategies to permit the synthesis of successful active site probes (affinity labels). Thus, dual approaches (genetic and biochemical) to identifying amino acid residues important in the recognition and/or translocation of substrates now appear more feasible than would have been suggested by previous models, which indicated that GabP might be highly selective and unable to recognize alternative ligands.