Amino acid substitutions in the pore of rat glutamate receptors at sites influencing block by polyamines

Abstract

1. The effect on polyamine block of mutations at the Q/R site and the conserved negative charge +4 site in AMPA and kainate receptors was studied using the rat glutamate receptor GluR6 expressed in Xenopus oocytes and human embryonic kidney (HEK) cells.

2. Introduction of negative charge at the Q/R site increased the equilibrium dissociation constant at 0 mV (K_d(0)) for spermine from 1.3 to 4.0 μm (Q590E); the smaller side chains Q590D and Q590N had K_d(0) values of 47 and 20 μm. Reductions in spermine affinity were also obtained for the small hydrophobic residues Q590V and Q590A, with K_d(0) values of 3.6 and 8.8 μm. Positively charged side chains produced outward rectifying responses similar to those recorded for GluR6(Q) with polyamine-free conditions, suggesting a complete absence of voltage-dependent block by spermine.

3. Substitution of tryptophan at the Q/R site produced high-affinity block with a K_d(0) of 190 pm. In Xenopus oocytes no outward current was observed at potentials up to +200 mV. A much smaller increase in affinity was observed for Q590F and Q590Y, which had K_d(0) values of 0.28 and 0.83 μm, respectively.

4. The Q590H mutant gave weakly birectifying responses strikingly different from those for other mutants. When ionization of the His group was increased by raising the external hydrogen ion concentration, responses became outward rectifying. The ratios of the conductance at 100 mV over that at −100 mV for Q590H were 0.52 at pH 8.3 and 2.5 at pH 5.3.

5. Neutralization of charge or aromatic residues at the +4 site produced a large reduction of spermine affinity, with K_d(0) values for E594N, E594Q and E594W of 109, 1020 and 2150 μm, respectively. In the absence of polyamines, E594K and E594R produced strongly inward rectifying responses while E594Q, E594A and E594W were birectifying.

6. A model for permeant block allowed quantitative comparisons between mutants. Despite large changes in well depth and barrier heights, there was little change in the voltage dependence of block for both Q/R and +4 site mutants. We propose a model with a distributed binding site for polyamines in which the +4 site is located near the entrance to the channel.

Cytoplasmic polyamines block current flow through multiple families of cation channels; this was discovered first for inward rectifier potassium channels (Nichols & Lopatin, 1997), and subsequently for three families of ligand-gated cation channels, the glutamate (Dingledine et al. 1999), neuronal nicotinic (Haghighi & Cooper, 1998) and cyclic nucleotide-gated (CNG) channels (Lu & Ding, 1999). In addition, polyamines block bacterial porins (delaVega & Delcour, 1995) and ryanodine receptors (Uehara et al. 1996). Although we currently lack detailed information on the structure of the pore regions for neurotransmitter-gated channels, some general features have been revealed by a broad variety of experiments. In the case of glutamate receptors (GluRs) a major part of the pore is formed by a re-entrant loop which has an inverted topology to that found in potassium channels (Hollmann et al. 1994; Stern-Bach et al. 1994; Bennett & Dingledine, 1995). Within this loop, Cys substitution accessibility analysis reveals a region where multiple adjacent residues are reactive; preceding this is a region with periodic accessibility, suggesting that the pore is formed in part by an alpha-helix supporting a short sequence of extended or random coil structure (Kuner et al. 1996, 1997; Beck et al. 1999). Together, these results suggest that in contrast to potassium channels, where the selectivity filter is formed by main chain carbonyl oxygen atoms (Doyle et al. 1998), the narrow region of the pore of glutamate receptors may contain a region in which multiple adjacent amino acid side chains are exposed to the lumen.

A unique feature of glutamate receptors compared with other ion channels is the regulation of polyamine block by an RNA editing mechanism which introduces a positively charged Arg residue within the pore loop, at a position for which genomic sequences in the GluRB, GluR5 and GluR6 subunits encode a neutral Gln residue (Higuchi et al. 1993; Herb et al. 1996). Compared with genomically encoded receptors, the RNA edited forms of glutamate receptors show low Ca²⁺ permeability (Burnashev et al. 1995), increased permeability to anions (Burnashev et al. 1996), 40-fold lower single-channel conductance (Swanson et al. 1997; Traynelis & Wahl, 1997) and weak outward rectification in the presence of polyamines (Bowie & Mayer, 1995). Within the pore loop of the unedited forms of polyamine-sensitive AMPA and kainate subtype glutamate receptors there is a conserved negatively charged amino acid located four residues towards the C-terminus from the RNA editing site, the +4 site. Neutralization of the +4 site Asp residue in the AMPA receptor GluR3 mutant D616N produces birectifying responses with much weaker rectification than observed for wild-type GluR3 (Dingledine et al. 1992). Although these experiments were performed before the topology of the pore loop in GluRs was established, and before block by polyamines was discovered as the mechanism responsible for the birectifying response of the unedited forms of GluRs, it was proposed that in the edited form of GluRs the Arg side chain might disrupt an internal cation-binding site responsible for inward rectification (Dingledine et al. 1992). Conversely, for amino acid substitutions at the Q/R site, the disruption of polyamine block produced by the exchange Q612N in GluR3 (Dingledine et al. 1992) could reasonably be expected to result from a reduced external barrier height due to the smaller size of the Asn versus the Gln residue facilitating permeation of polyamines past a narrow region of the pore.

In additional experiments on GluR3, it was found that the introduction of negatively charged residues at the Q/R site, both Glu and Asp, as well as the aromatic residue Trp, produced non-functional receptors (Dingledine et al. 1992). The mechanism(s) by which these strikingly different mutations block GluR responses was not established. One likely explanation would be that Glu or Asp produced a high-affinity binding site for internal cations, such that channels remained fully blocked over the range of membrane potentials examined, or alternatively that the folding, assembly or cell surface expression of these mutants was disrupted. To address these issues, and to gain further insight into the interaction of polyamines with the pore of GluRs, we generated a comprehensive series of amino acid substitutions at the RNA editing and conserved negative charge +4 sites in the kainate receptor GluR6 and applied a quantitative model for analysis of polyamine block developed previously for wild-type GluR6(Q) (Bähring et al. 1997; Bowie et al. 1998). In addition to clarifying the mechanisms underlying polyamine block in GluRs, the results of our experiments provide new insights into pore structure in glutamate receptors.

METHODS

Molecular biology

When cloned into appropriate expression vectors the cDNA for rat GluR6 gave large currents in both human embryonic kidney (HEK) cells and Xenopus oocytes, thus allowing the analysis of responses for those mutants which expressed less efficiently than wild-type. To facilitate mutagenesis, sequencing and subcloning, silent NheI and BglII restriction sites were inserted into the cDNA of wild-type rat GluR6(Q) at base pairs (bp) 1704 and 2093, respectively (ATG = 1) to create a cassette of 389 nucleotides, which was then further divided by the introduction of a silent MluI site at bp 1897. The 389 bp NheI and BglII fragment was subcloned into pLIT29 (NEB, Beverly, MA, USA) and used for oligonucleotide-directed mutagenesis by amplification of parental plasmid in vitro using complimentary mutagenic oligonucleotides (Gibco BRL, Rockville, MD, USA) and Pfu polymerase (Stratagene, La Jolla, CA, USA). Selection of mutants was achieved by subsequent digestion of the parental plasmid with DpnI. Following transformation and purification, plasmids were selected by dideoxy dye terminator sequencing (ABI 310, Foster City, CA, USA) to confirm the desired nucleotide changes. DNA fragments bounded by the appropriate restriction enzyme sites in pLIT29 were then completely sequenced and subcloned back into either of two GluR6 expression plasmids. A modified version of pGEM/HE (Liman et al. 1992) was used to synthesize cRNA for expression in Xenopus oocytes using T7 polymerase (Ambion mMessage mMachine, Austin, TX, USA) and XbaI linearized template. Expression in HEK cells was achieved by transient transfection with a eukaryotic expression vector in which a cytomegalovirus promoter controls the transcription of cDNAs (Keinänen et al. 1990).

Oocyte and HEK cell preparation

Oocytes were surgically removed from Xenopus laevis (Xenopus One) after induction of anaesthesia by immersion in water containing 3 g l⁻¹ tricaine. The animals were killed by decapitation while under anaesthesia; up to six surgeries were performed per toad. These procedures were carried out under a protocol approved by the NICHD animal care and use committee and followed NIH animal welfare guidelines. The oocytes were defolliculated by incubation for 60-90 min with 1.5 mg ml⁻¹ collagenase dissolved in solution containing (mm): 83 NaCl, 2 KCl, 1 MgCl₂, 5 Hepes, pH 7.5; the preparation was then thoroughly rinsed with solution containing (mm): 88 NaCl, 2.5 NaHCO₃, 1.1 KCl, 0.4 CaCl₂, 0.3 Ca(NO₃)₂, 0.8 MgCl₂, 2.5 sodium pyruvate, 15 Hepes, pH 7.3, and 5 μg ml⁻¹ gentamicin. After 18-24 h Dumont stage V-VI oocytes were injected with between 40 pg and 50 ng of cRNA as required and then stored at 18°C for 2-8 days prior to recording.

HEK 293 cells (ATCC CRL 1573) were maintained at a confluency of 70-80% in minimal essential medium with Earle's salts, 2 mm glutamine and 10% fetal bovine serum. Twenty-four hours after plating at low density (2 × 10⁴ cells ml⁻¹) onto the centre of 35 mm Petri dishes, cells were transfected using the calcium phosphate technique; co-transfection with the cDNA for green fluorescent protein (S65T mutation) helped to identify transfected cells during experiments, as described previously (Bowie & Mayer, 1995). Cells were washed with phosphate-buffered saline (PBS) 12-18 h after transfection and used for electrophysiological recordings after another 24-48 h.

Immunohistochemistry

Four days after injection with 10 ng cRNA, oocytes were fixed overnight by immersion in 100 mm PBS (pH 7.4) containing 4% paraformaldehyde and 15% of a saturated solution of picric acid; uninjected oocytes from the same preparations were processed identically. The oocytes were then cryoprotected by overnight immersion in a solution containing 10 mm PBS, 10% glycerol, 0.008% NaN₃ and 25 g (100 ml)⁻¹ sucrose; 100 μm sections were cut on a freezing microtome, incubated with 2% Triton in PBS, blocked with 10% goat serum in 0.2% Triton for 2 h, and reacted overnight at 4°C with rabbit anti-GluR6 antibody, 1 μg ml⁻¹(Wenthold et al. 1994) in PBS containing 10% goat serum and 0.2% Triton. Sections were then incubated at room temperature for 60 min with FITC-conjugated goat anti-rabbit serum (1/200). A Zeiss confocal microscope with a krypton/argon laser was used for imaging.

Recording conditions and solutions

Our experiments were performed using the GluR6 subtype of kainate receptor for which we have previously developed models that describe the action of polyamines as permeant blockers (Bähring et al. 1997; Bowie et al. 1998). To ensure a high probability of occupancy of the open state, and thus favour open versus closed channel block by polyamines (Bowie et al. 1998; Rozov et al. 1998), responses were activated using 100 μm kainate for Xenopus oocytes and 50 μm domoate for HEK cell patches; for both preparations agonist concentrations were greater than 100 × EC₅₀. Two-electrode voltage clamp recording for Xenopus oocytes was performed using 3 M KCl-filled agarose cushion microelectrodes (Schreibmayer et al. 1994) and an Axoclamp-2B amplifier (Axon Instruments, Foster City, CA, USA) with an extracellular microelectrode used as the input for a virtual ground bath clamp. The recording chamber had a volume of 5 μl; solutions were applied at 250 μl min⁻¹. The extracellular solution contained (mm): 88 NaCl, 2.5 NaHCO₃, 1.1 KCl, 0.32 Ba(NO₃)₂, 0.4 BaCl₂, 0.8 MgCl₂, 15 Hepes, pH 7.3 with NaOH. Concanavalin A (Sigma type IV) at 0.3 mg ml⁻¹, was applied for 4-16 min to reduce desensitization (Partin et al. 1993; Everts et al. 1999). For solutions at pH 5.3, 2-(N-morpholino))ethanesulphonic acid (Mes) was substituted for Hepes; 2-(N-cyclohexylamino))ethanesulphonic acid (Ches) was used as the buffer at pH 9.3. Experiments to study the effects of pH on Q590H were performed using 5 μm tetrakis(2-pyridylmethyl)ethylenediamine (TPEN) added to the external solution. This was required to prevent use-dependent block by trace amounts of unidentified metal ions, most probably Zn²⁺ or Cd²⁺. Although we have not studied this effect in detail, we found that tricine-buffered solutions containing nanomolar free concentrations of Zn²⁺ strongly block Q590H but not wild-type GluR6(Q), indicating that the introduction of an imidazole group at the Q/R site allows the co-ordination of metal ions with high affinity. In a few experiments, where noted, we used a low-divalent solution containing 0.1 mm Ca²⁺ and 0.1 mm Mg²⁺; for these experiments oocytes were injected with 50 nl of a solution containing 20 mm K₄ BAPTA and 60 mm potassium aspartate, pH 7.2, 1-4 h prior to recording. Where required, divalent-free solution was made by substituting 0.5 mm Na₂ EDTA for the Ca²⁺, Mg²⁺ and Ba²⁺ salts in the external solution.

Experiments on outside-out patches from HEK cells were performed in an external solution containing (mm): 150 NaCl, 1.1 KCl, 0.7 BaCl₂, 0.8 MgCl₂, 5 Hepes, pH 7.3, osmolarity adjusted to 295 mosmol l⁻¹ with sucrose. GluR6 responses were evoked using 50 μm domoic acid, which produces partial desensitization, applied via a stepper motor-based fast perfusion system (Vyklicky et al. 1990). In some experiments, patches were treated with concanavalin A (0.3 mg ml⁻¹ for 1-1.5 min) to further reduce desensitization. The internal solution contained (mm):120 NaCl, 10 NaF, 5 Hepes, 0.5 CaCl₂, 5 Na₄ BAPTA to which spermine (Sigma) was added as required. In experiments with polyamine-free conditions 10 mm NaCl was replaced by 10 mm Na₂ ATP. Outside-out patches were excised from HEK cells using fire-polished, thin-walled borosilicate glass pipettes (2-5 MΩ) coated with dental wax to reduce electrical noise. Currents were recorded with an Axopatch 200 amplifier (Axon Instruments) and filtered (8-pole Bessel filter) at 1.5 kHz, sampled at 0.16 ms intervals and stored on a Power Macintosh G3 computer using a 16-bit A/D converter (Instrutech ITC-16) under control of the program Synapse (Synergy Research Incorporated, Silver Spring, MD, USA).

Data analysis

Current-voltage (I-V) plots were generated with ramp protocols (0.11-2.7 V s⁻¹). Procedures in the Igor program (Wavemetrics, Lake Oswego, OR, USA) were used to generate and analyse conductance-voltage (G-V) plots. First, the reversal potential for I-V plots was estimated using a 5th-order polynomial fit to either single (oocytes) or the average of five (HEK cells) leak-subtracted responses. G-V plots were then generated and fitted with the following Boltzmann function over the range -100 to 20 mV to obtain an initial estimate of the voltage dependence of block by polyamines (Fig. 3):

(1)

where G_max is the conductance at a sufficiently hyperpolarized potential to produce full relief from block by polyamines; V_m is the membrane potential; V_b the potential at which 50% block occurs; and k_b is a slope factor describing the voltage dependence of block. For wild-type GluR6(Q) with 100 mm[Na⁺]_o, an equilibrium dissociation constant for spermine at 0 mV (K_d(0)) of 1.2 μm was estimated by interpolation of K_d(0) values obtained in HEK cells with 44, 120 and 330 mm[Na⁺]_o (Bähring et al. 1997). Oocyte cytoplasmic polyamine concentrations were then calculated from the following relationship:

(2)

where [Spm] is the spermine concentration. In order to analyse polyamine block over a wider range of membrane potentials it was necessary to correct for rectification which occurs in polyamine-free conditions (Bähring et al. 1997; Bowie et al. 1998). To do this, pooled responses recorded in separate experiments in the presence and absence of polyamines were normalized to have a value of 1 at -100 or -150 mV. Corrected G-V plots were then generated by dividing responses recorded in the presence of polyamines by those recorded in polyamine-free conditions (Figs 8 and 9). The resulting plots were then fitted with the sum of two Boltzmann functions of opposite voltage dependence:

(3)

where G_max, V_m, V_b and k_b have the same meanings as described above; V_p is the potential at which the second Boltzmann function, which describes permeation of polyamines on strong depolarization, reaches 50%; and k_p is the slope factor describing the voltage dependence of relief from block. A voltage-independent dissociation constant for polyamine block, K_d(0), was calculated from the following relationship assuming a value of 20 μm for [Spm] (see Fig. 1 and below):

(4)

Parameters from fits of eqn (3) were also used to estimate well depth (WD):

(5)

the difference between external and internal barrier heights (EB - IB):

(6)

and electrical distances for block using a two barrier one site (2B1S) model of polyamine block (Bähring et al. 1997; Bowie et al. 1998).

Figure 3. Amino acid side-chain volume at the Q/R site regulates polyamine block in GluR6.

A, C and E show leak-subtracted I-V plots recorded using the protocol shown in Fig. 2 for three sets of mutations; responses for individual oocytes have been normalized to have similar amplitudes at -100 mV. B, D and F show the corresponding averaged G-V plots recorded from 5-17 oocytes as indicated, with the mean ±s.d. plotted at 20 mV intervals; dotted lines show fits of Boltzmann functions over the range -100 to 20 mV; to allow comparison between mutants, responses in B and D are normalized at -100 mV; in F responses are normalized at 0 mV. A and B, negative charge at the Q/R site does not increase polyamine block; instead Q590E shows similar voltage dependence to wild-type GluR6(Q). In contrast, reduction in side-chain volume, both for Q590N and Q590D, produces a large rightward shift in the G-V plots. For illustrative purposes the response for wild-type GluR6(Q) shown in Fig. 3B is a representative sample of data from 17 cells. C and D, for the series Q590F, Q590V, Q590A and Q590G, decreases in side-chain volume also reduce polyamine block; note that although the half-block potential shifts to progressively depolarized membrane potentials there is little change in slope. E and F reveals that the positively charged side chains Q590K and Q590R produce nearly identical outward rectifying responses while Q590H shows the weakest rectification observed for any of the amino acid substitutions introduced at the Q/R site.

Figure 8. Low-affinity block by polyamines for E594 charge neutralization mutants.

A, responses recorded from HEK cell outside-out patches with 10 mm Na₂ ATP to chelate endogenous polyamines reveal biphasic (E594Q, E594N and E594A) or inward (E594K) rectification similar to that observed in oocytes (Fig. 7); data points show mean ±s.d. for averaged responses normalized at -100 mV for 4-8 patches as indicated. In contrast, responses for wild-type GluR6(Q) show weak outward rectification. B, HEK cell outside-out patch responses for E594Q and E594N recorded with 1 mm spermine added to the internal solution (no ATP) were corrected for intrinsic rectification using the responses shown in A. Data points were plotted every 20 mV and show mean ±s.d.; lines through the data points are least-squares fits over the range -100 to +100 mV of a 2B1S model for permeant block by spermine.

Figure 9. Different affinities for permeant block by polyamines for Q/R site mutants.

A, G-V plots for GluR6 Q/R site mutants reveal weak outward rectification in the absence of polyamines. Responses were recorded from 3-5 HEK cell outside-out patches per mutant using 10 mm internal Na₂ ATP to chelate endogenous polyamines. Data points show mean ±s.e.m.. Note that changes in amino acid side-chain size or charge have little influence on rectification in the absence of cytoplasmic polyamines. B, G-V plots for responses recorded in Xenopus oocytes corrected for rectification recorded in the absence of polyamines using data shown in A. Note that compared to uncorrected responses shown in Fig. 3 the conductance at 100 mV no longer exceeds that at -100 mV except for Q590R. Data points are plotted every 20 mV and show mean ±s.d.; lines through the data points are least-squares fits over the range -100 to +100 mV of a 2B1S model for permeant block; the dashed line shows the fit for wild-type GluR6(Q) and indicates a 3-fold lower affinity for polyamine block compared with Q590F. Responses for Q590R are fitted with a 3rd-order polynomial and reveal no voltage-dependent block by polyamines over the range -100 to 50 mV; the small increase in conductance at potentials depolarized to 50 mV was also observed when responses in HEK cell patches with 100 μm internal spermine were compared with those with 10 mm internal Na₂ ATP.

Figure 1. Analysis of block by endogenous polyamines in Xenopus oocytes expressing wild-type GluR6(Q).

A, voltage ramps from -100 to +100 mV were evoked before and during 4 s applications of 100 μm kainate (Kai). B, effects of external divalent cations on GluR6(Q) responses. Conductance-voltage analysis of leak-subtracted responses to 100 μm kainate were recorded from a single oocyte in external solution containing 1.6 mm MgCl₂, 1.6 mm CaCl₂ or standard BaCl₂ solution as indicated; symbols are plotted every 20 mV for identification of individual responses; smooth lines show Boltzmann functions fitted from -100 to 20 mV for responses recorded with Ba²⁺ and Mg²⁺; the inset shows the response for standard BaCl₂ solution plotted at 20-fold higher gain which reveals a well-defined minimum conductance. C, box plots from Boltzmann analysis and estimated polyamine concentrations ([Spm]) for wild-type GluR6(Q) responses recorded with standard BaCl₂ solution from 107 oocytes. The line in the middle of the boxes shows the median; the top and bottom edges of the boxes show the 25th and 75th percentiles of the data; the whiskers coming out of the boxes show the 10th and 90th percentiles; the shaded area indicates the 95% confidence interval of the mean.

Because our equilibrium measurements do not provide information on rate constants, we assumed that δ1, the electrical distance from the cytoplasm to the internal barrier, has the same value (0.065) as previously established from the kinetic analysis of responses for wild-type GluR6(Q) (Bowie et al. 1998); we were then able to use the following relationships to estimate the remaining electrical distances from k_b and k_p:

(7)

where R, T and F have their usual meanings, δ2 is the electrical distance from the internal barrier to the well; δ3 the electrical distance from the well to the external barrier; and δ4 the electrical distance from the external barrier to the extracellular solution (Woodhull, 1973). The electrical distance of block measured from the cytoplasm (θ) is given by (δ1 +δ2). The equilibrium measurements performed in this study also did not permit assignment of absolute values to barrier heights, but did allow calculation of the difference between the heights of the internal and external barriers. Numerical values and error bars in graphs indicate means ±s.d. unless noted differently.

Estimation of cytoplasmic polyamine concentrations in Xenopus oocytes

The goal of our experiments was to obtain estimates of the voltage dependence for onset and relief from polyamine block over a range of membrane potentials sufficiently wide to allow the estimation of changes in barrier heights and well depths for a permeant blocker mechanism (Bähring et al. 1997; Bowie et al. 1998). Thus, it was important to establish that other ion channels present in the oocyte plasma membrane did not interfere with the measurement of GluR responses, and that the cytoplasmic concentration of polyamines did not vary from experiment to experiment. Since unedited and some mutant GluRs are permeable to Ca²⁺ it was first necessary to establish conditions which prevented activation of the oocyte Ca²⁺-dependent chloride conductance during GluR responses. This was achieved by substituting Ba²⁺ for Ca²⁺ in the external solution, and by limiting the duration of application of kainate by using a small chamber and rapid solution exchange (Fig. 1). With applications longer than about 10 s we found that even with Ba²⁺ as the external cation there was sufficient activation of the oocyte Ca²⁺-dependent chloride current to distort analysis of polyamine block at strongly depolarized membrane potentials.

To establish whether our standard recording conditions were indeed adequate for preventing activation of Ca²⁺-dependent chloride currents due to divalent cation flux through GluR6 channels we compared responses for wild-type GluR6(Q) recorded from the same oocytes using Ba²⁺, then Mg²⁺, and finally Ca²⁺ in the external solution (Fig. 1). These experiments were based on microinjection studies (Miledi & Parker, 1984) which showed that Mg²⁺ does not activate the Xenopus oocyte Ca²⁺-dependent chloride current. From a holding potential of -60 mV, voltage ramp responses from -100 to +100 mV with Ba²⁺ or Mg²⁺ in the external solution gave biphasic rectifying responses typical of those for GluR6(Q) recorded from HEK cell patches with micromolar concentrations of polyamines and the Ca²⁺ buffer BAPTA added to the internal solution (Bowie & Mayer, 1995). For 10 oocytes studied with all three solutions, Boltzmann functions fitted over the range -100 to 20 mV (Fig. 1) gave similar estimates of V_b for Ba²⁺(-49.7 ± 3.0 mV) and Mg²⁺(-44.6 ± 3.4 mV) with k_b= 15.6 ± 1.6 mV and 14.8 ± 2.2 mV, respectively; however, the amplitudes of responses at -60 mV with Ba²⁺ were only 0.34 ± 0.13 of those recorded with Mg²⁺. In contrast, when Ca²⁺ was used as the external divalent cation, GluR6(Q) responses showed complex G-V relationships. That these were due to simultaneous activation of GluR6(Q) and Ca²⁺-dependent chloride current was demonstrated by the observation that following injection of BAPTA, G-V plots with Ca²⁺ had similar voltage dependence to those with Ba²⁺ and Mg²⁺. The identical shape of the G-V plots recorded with Mg²⁺ and with the standard Ba²⁺ solution used for the rest of our experiments indicates that it is possible to measure accurately voltage-dependent polyamine block independent of oocyte Ca²⁺-dependent chloride currents.

To determine the extent to which the cytoplasmic concentrations of polyamines varied between individual oocytes, we recorded responses from wild-type GluR6(Q) using oocytes from the same batches for which the effects of pore mutations were studied (Fig. 1). Responses recorded from 107 oocytes expressing wild-type GluR6(Q) were fitted with a Boltzmann function to allow the subsequent estimation of internal polyamine concentrations (eqns (1) and (2)). Because the ratio of the free concentrations of spermine and spermidine in the oocyte cytoplasm are unknown, we simplified our analysis by assuming that all of the block was mediated by spermine. This is reasonable because although spermine and spermidine block GluR6(Q) responses with 5-fold different affinity, the block develops with almost identical voltage dependence (Bowie & Mayer, 1995; Bähring et al. 1997). Thus, even when the relative concentrations of spermine and spermidine are not known it is still possible to analyse responses as though all of the block was produced by a single species and thus establish the extent to which polyamine concentrations vary from oocyte to oocyte. The results of this experiment revealed surprisingly little variation between eggs. Mean values for V_b and k_b were -41.6 ± 4.1 and 15.3 ± 1.0 mV, respectively (n= 107). From these values we calculated a mean oocyte cytoplasmic spermine concentration of 20.3 ± 7.2 μm(eqn (2)), in good agreement with the observation of Fakler et al.(1995) that 10 μm spermine reasonably mimicked the high voltage dependence of rectification for IRK1. The relatively low values for cytoplasmic spermine concentration in oocytes compared with our prior estimates of 60 μm for HEK cells (Bowie & Mayer, 1995) possibly results from buffering of free polyamine concentrations by the large amounts of maternal mRNA present in Xenopus oocytes.

RESULTS

We introduced mutations into the pore region of GluR6 at two sites previously shown to influence rectification in the AMPA receptor GluR3 subunit (Dingledine et al. 1992), the Q/R site and E594, a conserved negatively charged amino acid in AMPA and kainate receptors located four residues towards the C-terminus from the Q/R site (Fig. 2). Substituted cysteine accessibility analysis of the AMPA receptor GluRD subunit (Kuner et al. 1997) suggests that the amino acid side chains of both residues face the lumen of the pore in polyamine-sensitive GluRs. We found that amino acid substitutions at both positions were well tolerated, with only two out of 27 mutants (Q590P and Q590W) apparently non-functional when assayed in Xenopus oocytes using voltage ramps over the range ±100 mV.

Figure 2. Amino acid sequence alignment in the pore region for five glutamate receptor gene families.

A shows a schematic diagram of a glutamate receptor subunit; the positions of three membrane-crossing segments and the hydrophobic segment believed to form a pore loop are indicated by black and grey bars, respectively. An amino acid sequence alignment of the pore loop region for AMPA, kainate (GluR5-7, KA-1 and KA-2) and NMDA (NR1 and NR2A-D) receptor subunits is shown below and reveals much greater conservation for AMPA and kainate receptors than for NMDA receptors. Arrows indicate positions at which mutations were introduced in GluR6; conserved residues between GluR gene families are indicated by vertical lines; gaps introduced by the ClustalW alignment algorithm are indicated by a dash; for positions at which amino acid variations occur within individual gene families the alternative residues are indicated above and below the line common to all subunits. Positions in the NR1, NR2C and GluRD subunits accessible by substituted cysteine accessibility analysis are shown in bold (Kuner et al. 1996, 1997). B shows the pore loop sequence for GluR6 and the series of mutations introduced at the Q/R site (590) and the conserved negative charge +4 site (E594); asterisks indicate substitutions which failed to give functional responses in Xenopus oocytes when assayed using the protocol shown in Fig. 1.

Effects of Q/R site amino acid side-chain size and charge

Prior studies on the AMPA receptor GluR3 subunit revealed reduced rectification for the Q/R site mutation Q612N, but no functional responses when the negatively charged side chains Q612E and Q612D were introduced at this site (Dingledine et al. 1992). We found that the corresponding kainate receptor GluR6 mutants Q590D and Q590E showed reduced amplitude responses compared with wild-type GluR6(Q), but that expression of Q590E and Q590D was still sufficiently high to allow accurate analysis of polyamine block when concanavalin A (0.3 mg ml⁻¹) was applied for 8-16 min to increase block of desensitization to near-maximum values. Current amplitudes at -60 mV were 0.34 ± 0.09 μA (mean ±s.e.m., n= 9) for Q590E and 0.33 ± 0.14 μA (n= 5) for Q590D, compared with 2.7 ± 0.34 μA (n= 17) for wild-type GluR6(Q) for which concanavalin A was applied for only 2-4 min. Surprisingly, introduction of negative charge at the Q/R site appeared to reduce polyamine block such that at strongly depolarized membrane potentials, I-V plots for Q590E showed increased outward current compared with responses for wild-type GluR6(Q) (Fig. 3A). G-V plots fitted with Boltzmann functions over the range -100 to 20 mV revealed an 11 mV depolarizing shift in V_b for Q590E with little change in slope (Fig. 3B); mean values for V_b were -41.6 ± 4.1 mV (n= 107) for wild-type GluR6(Q) and -30.6 ± 4.4 mV (n= 9) for Q590E. Much larger shifts in half-block potential were produced by a reduction in the size of the side chain at the Q/R site by one CH₂ group. This produced a 60 mV depolarizing shift in V_b when the side chain was neutral (Q590N: V_b= 18.8 ± 2.6 mV, n= 7) and a 58 mV depolarizing shift when the side chain was negatively charged (Q590D: V_b= 16.3 ± 1.4 mV, n= 5).

The apparent reduction in polyamine block on switching from Gln to Asn, or from Glu to Asp, was recapitulated by introducing non-polar residues of different size at the Q/R site (Fig. 3C and D). For the series Q590V, Q590A and Q590G, outward currents at depolarized potentials became progressively larger with reduction in size of the amino acid side chain. From Boltzmann analysis of G-V plots for these mutants we obtained V_b values of -26 ± 3.9 mV (n= 6), -7.2 ± 2.3 mV (n= 7) and 23 ± 2.9 mV (n= 5) for Val, Ala and Gly, respectively. Similar to the results for polar and negatively charged side chains (Fig. 3A and B), these large shifts in V_b were not accompanied by marked changes in slope such that the G-V plots were shifted in an approximately parallel manner (Fig. 3C and D). The aromatic amino acids Phe and Tyr produced small hyperpolarizing shifts in V_b with values of -53 ± 0.9 mV (n= 5) and -45 ± 2.3 mV (n= 5) for Q590F and Q590Y. In contrast, no responses were obtained for the larger aromatic substitution Q590W (n= 9).

Introduction of positively charged side chains at the Q/R site produced outward rectifying responses (Fig. 3E and F) similar to those previously recorded for GluR6(Q) using HEK cell outside-out patches with polyamine-free conditions (Bowie & Mayer, 1995; Bähring et al. 1997). This suggests that polyamines do not produce voltage-dependent block of GluR6 responses when the Q/R site contains a positively charged amino acid. Similar outward rectification for mutants with reduced sensitivity to polyamines most likely accounts for the larger conductance at 100 mV versus-100 mV observed for Q590D and Q590G for which block by polyamines was relatively weak (Fig. 3B and D). This is consistent with two separate mechanisms underlying rectification of Q/R site mutants, only one of which is mediated by polyamines. Thus, as described later, further analysis of changes in rectification produced by mutations at the Q/R site requires additional measurements of rectification in polyamine-free conditions (Bähring et al. 1997).

Introduction of an imidazole ring at the Q/R site produced a strikingly different response from those observed for other mutants (Fig. 3E and F). For Q590H, rectification was much weaker than for Q590K or Q590R, or for mutants with neutral or negatively charged side chains, such that over the range ±100 mV the I-V plot for Q590H was the closest to linear for any of the mutants studied (Fig. 3E). A similar result was previously reported for the corresponding AMPA receptor mutant GluR1 Q600H (Curutchet et al. 1992). Analysis of G-V plots for Q590H revealed weak biphasic rectification with a ratio of the conductance at 100 mV over that at -100 mV (G₁₀₀ /G_-100) of 0.90 ± 0.08 (n= 7) and a minimum at 22.7 ± 5.1 mV. Because His residues have pK_a (-log of the dissociation constant) values which allow partial ionization at physiological pH (Creighton, 1993), one explanation for the weak biphasic voltage dependence observed for Q590H could be titration of the imidazole ring by protons entering the channel from the intracellular and extracellular solutions. This would be expected to produce a birectifying response if hydrogen ions acted as permeant blockers (Woodhull, 1973). An alternative explanation could be that the weak rectification for Q590H results from low-affinity block by cytoplasmic polyamines.

Titration by external protons of an engineered His residue at the Q/R site

With outside-out patches from HEK cells and an internal solution containing 10 mm Na₂ ATP to chelate endogenous polyamines, conditions which generate weakly outward rectifying responses for wild-type GluR6 (Bähring et al. 1997; Cui et al. 1998), responses for Q590H remained birectifying (Fig. 4A). This result eliminated polyamine block as the major mechanism underlying the weak biphasic rectification recorded for Q590H in Xenopus oocytes. To determine if voltage-dependent block by protons contributed to rectification for Q590H, we studied the effect of acidification of the external solution. To control for the effects of pH on surface potential, ion activity and titration of other amino acids in addition to the engineered His residue in Q590H, any of which might contribute to the block of GluR responses by protons independent of the histidine residue in the pore loop (Traynelis & Cull-Candy, 1991), we compared the effects of changing [pH]_o for Q590H with those for Q590R, for which we expected the Arg residue to remain fully ionized over the pH range examined (Fig. 4). These experiments were performed using Xenopus oocytes and low divalent solution and showed that when ionization of the imidazole ring was increased by raising the external hydrogen ion concentration from 5 nm(pH 8.3) to 5 μm(pH 5.3), responses for Q590H changed from birectifying at pH 8.3, G₁₀₀ /G_-100= 0.52 ± 0.03, n= 14 (Fig. 4B) to strongly outward rectifying at pH 5.3, G₁₀₀ /G_-100= 2.49 ± 1.5, n= 5 (Fig. 4C). At pH 5.3 the similar shape of the G-V plots for Q590H and Q590R suggests that the imidazole ring is almost fully ionized over the membrane potential range examined such that the His residue now behaves functionally like the fully charged Arg side chain, which like the Q590K mutant produces outward rectifying responses (Fig. 3).

Figure 4. Titration of Q590H by external protons produces outward rectifying responses.

A, I-V plots for Q590H recorded from HEK cell outside-out patches (n= 4) using 10 mm internal Na₂ ATP to chelate endogenous polyamines; the dotted line indicates the nearly identical I-V plot for Q590H responses recorded in Xenopus oocytes (n= 6). B and C show averaged G-V plots for responses recorded from Xenopus oocytes at different external pH values and compare rectification for Q590R and Q590H; responses are normalized at 0 mV; data points show the mean ±s.e.m. Note that responses for Q590H show biphasic rectification at pH 8.3 (B) but outward rectification at pH 5.3 (C). In contrast, responses for Q590R show outward rectification at both pH values.

The birectifying responses observed for Q590H at pH 7.3 (Fig. 4A) most likely results from voltage-dependent changes in ionization of the imidazole ring due to protons entering the pore and acting as permeant blockers. Our attempts to estimate a pK_a value and an electrical distance for proton block for the imidazole ring in the Q590H mutant were complicated by several factors. First, protons produced a voltage-dependent block of responses for Q590R (Fig. 4B and C); thus, the degree of outward rectification at pH 5.3 (G₁₀₀ /G_-100= 4.07 ± 1.20, n= 5) was greater than at pH 8.3 (G₁₀₀ /G_-100= 2.1 ± 0.20, n= 11). Although comparison of the responses shown in Fig. 4B and C indicates that voltage-dependent block by protons was weaker for Q590R than for Q590H, these results suggest that protons are likely to block Q590H by multiple mechanisms in addition to titration of the imidazole ring, complicating any simple interpretation of the pH sensitivity of the Q590H mutant. A second difficulty was raised by the observation that at [pH]_o 8.3, responses for Q590H continue to show biphasic rectification, most likely as a result of block of outward current by either hydrogen ions in the oocyte cytoplasm, or perhaps by polyamines if the charge on the imidazole ring becomes sufficiently reduced. As [pH]_o was raised further, from 8.3 to 9.3, the birectifying conductance for Q590H increased in amplitude, approximately 1.5-fold at -100 mV, but showed little change in voltage dependence such that the ratio G₁₀₀ /G_-100 was 0.53 ± 0.08 (n= 14) at pH 8.3 and 0.49 ± 0.01 (n= 3) at pH 9.3. In contrast, for Q590R, although there was a similar increase in amplitude on reduction of hydrogen ion concentration, rectification continued to decrease such that the ratio G₁₀₀ /G_-100 was 2.1 ± 0.20 (n= 11) at pH 8.3 and 1.6 ± 0.15 (n= 6) at pH 9.3. Thus, although our results indicate that the imidazole group in Q590H senses changes in hydrogen ion concentration within the membrane electric field, the mechanisms underlying proton block are too complex to allow use of this effect to determine the location and affinity of the binding site, and hence estimate the position of the Q/R site within the electric field.

Tryptophan at the Q/R site creates a high-affinity binding site for polyamines

Of 18 mutations introduced at the Q/R site only two failed to give functional responses when expressed in Xenopus oocytes and analysed using the protocol shown in Fig. 3. These were Q590P (n= 10) and Q590W (n= 9). To address the possibility that these mutations interfered with GluR6 cell surface expression, oocytes were fixed, sectioned and stained with antibodies that recognized the GluR6 C-terminus (Wenthold et al. 1994); a similar approach has been used previously to assay inward rectifier potassium channel subunit expression in the oocyte plasma membrane (Stevens et al. 1997). These experiments were performed on a total of 28 oocytes (Fig. 5). The background level of cell surface immunofluorescence was similar for uninjected oocytes (n= 7) and Q590P (n= 7), suggesting that in GluR6 the introduction of proline at the Q/R site interferes with receptor folding, assembly, stability or targeting to the plasma membrane. In contrast, cell surface immunofluorescence was similar for Q590W (n= 7) and wild-type GluR6(Q) (n= 7), indicating that introduction of tryptophan at the Q/R site does not prevent insertion into the plasma membrane. This large aromatic residue must interfere with GluR6 responses in Xenopus oocytes via some other mechanism.

Figure 5. The non-functional mutants Q590P and Q590W show different cell surface expression.

Oocytes stained by indirect immunofluorescence with an antibody directed against the COOH terminus of GluR6 were viewed by confocal microscopy using a single z section with constant settings between oocytes for laser intensity, iris diaphragm and gain. The intense cell surface immunofluorescence observed for Q590W was similar to that observed for wild-type GluR6(Q). In contrast, for Q590P the level of staining at the oocyte surface was similar to that observed in uninjected eggs. Similar results were obtained in a total of 28 oocytes from 2 frogs.

To address the possibility that the lack of functional responses for Q590W in Xenopus oocytes resulted from the use of a heterologous expression system we analysed responses for the Q590W mutant expressed in HEK cells. In contrast to results obtained for Xenopus oocytes, agonist-activated responses were reliably measured with polyamine-free solutions and HEK cell outside-out patches (n= 4) over the range ±100 mV (Fig. 6A). Because these measurements were performed using the same concentrations of Ba²⁺ and Mg²⁺ in the extracellular solution as used for experiments on Xenopus oocytes, we can exclude high-affinity block by divalent cations as a mechanism responsible for lack of functional responses. In our experiments on HEK cells transfected with Q590W we noticed that in some patches, despite the use of 10 mm Na₂ ATP to chelate endogenous polyamines, there was a time-dependent reduction of biphasic rectification following formation of outside-out patches; this was similar to what we observed previously for wild-type GluR6(Q) when ATP was not added to the internal solution (Cui et al. 1998). The slow washout for Q590W in the presence of ATP suggests that polyamines might bind with much higher affinity than for wild-type GluR6(Q) and that in Xenopus oocytes the endogenous polyamine concentrations might be sufficient to fully block current flow over the range ±100 mV. To test this we recorded oocyte responses for Q590W using voltage ramps over an extended range of potentials, from 10 to -230 mV. This revealed robust expression of agonist-activated currents at strongly hyperpolarized potentials which, when fitted by a Boltzmann function, gave V_b= -191 ± 4.7 mV (n= 6), with a slope factor of 17.1 ± 1.3 mV (Fig. 6B). For these experiments we used ramps starting from a positive membrane potential since this minimized the extent of activation of a hyperpolarization-activated chloride current (Parker & Miledi, 1998) which otherwise interfered with measurement of small agonist-activated responses at potentials where polyamine block was approaching its maximum value. With depolarizing ramps from -100 to 200 mV, the leak and agonist-activated currents became indistinguishable at potentials depolarized to about -80 mV and there was no detectable increase in outward current when the membrane potential was depolarized up to 200 mV, indicating that polyamines may be unable to pass through the pore of the Q590W mutant. We conclude that Q590W generates functional channels which bind polyamines with such high affinity that at the endogenous concentrations present in Xenopus oocytes, polyamine block essentially abolishes functional responses over the membrane potential range used to study other GluR6 pore mutants (Fig. 3).

Figure 6. High-affinity polyamine block for Q590W underlies an apparent loss of functional responses in Xenopus oocytes.

A, mean I-V plots for HEK cell outside-out patches recorded with 10 mm internal Na₂ ATP used to chelate endogenous polyamines; weakly rectifying responses for Q590W were recorded over the range ±100 mV from each of 4 cells examined. B, leak-subtracted G-V plots for Q590W responses recorded from Xenopus oocytes (n= 6) show relief from polyamine block at extreme negative potentials. The dotted line shows a Boltzmann function fit to averaged data over the range -220 to -50 mV; for comparison the dashed line shows a Boltzmann function fit to wild-type GluR6(Q) responses over the range -100 to +20 mV (data from Fig. 3).

A conserved negative charge at E594 is required for high-affinity polyamine block

Similar to results obtained previously for the D616N mutation in the AMPA receptor GluR3 subunit (Dingledine et al. 1992), weak birectifying responses with large outward currents were recorded when the corresponding GluR6 mutant E594Q was expressed in Xenopus oocytes (Fig. 7A). In contrast, for E594D we observed strong biphasic rectification similar to that for wild-type GluR6(Q) (Fig. 10 and Table 1). We noted that the voltage dependence of rectification for E594Q and for E594N was less than for those Q/R site mutants such as Q590D or Q590N which produced comparably large outward currents (see Fig. 3) suggesting that different or additional mechanisms underlie rectification for E594 mutants. To address the possibility that, in addition to disruption of the action of polyamines, block by external divalent cations contributed to rectification for E594Q and E594N, we compared responses recorded with standard Ba²⁺ solution with those recorded in the presence of 0.5 mm EDTA (Fig. 7). The activation of oocyte hemi-gap-junction channels which occurs on prolonged exposure to divalent free solutions (Zhang et al. 1998), produces increases in conductance which can be as large as those evoked by GluR6. Thus, it was first necessary to establish protocols which allowed measurement of GluR6 responses in isolation of the response to EDTA. Because the inward current response of oocytes to divalent cation-free solution develops with slow kinetics (Zhang et al. 1998), it was possible to minimize the activation of hemi-gap-junction channels by using rapid exchange of the extracellular solution and only brief applications of EDTA. Responses to application of EDTA applied for 7 s were subtracted from those to kainate applied in the presence of EDTA, with computer-activated solenoid valves used to apply EDTA for identical times in the presence and absence of agonist. This approach gave highly reproducible responses to EDTA and limited the activation of hemi-gap-junction channel currents such that the amplitude of responses to EDTA were now typically less than 10% of those to kainate. In 11 oocytes voltage clamped at -60 mV the current activated by 0.5 mm EDTA applied for 7 s was 0.11 ± 0.04 μA while in the same cells responses to 100 μm kainate for various E594 mutants were 2.03 ± 1.1 μA.

Figure 7. Amino acid substitutions at a conserved negatively charged residue in the pore loop.

A, C and E show leak-subtracted I-V plots recorded from individual oocytes using the protocol shown in Fig. 1. Responses were recorded first with standard BaCl₂ solution and subsequently with divalent-free external solution containing 0.5 mm EDTA. The mutations test the effects of charge neutralization (A),introduction of small hydrophobic and aromatic amino acids (C) and charge reversal (E). Averaged G-V plots recorded from 5-10 oocytes in divalent-free external (DF) solution and normalized at -100 mV are shown in B, D and F. Note that the birectifying responses observed for the charge neutralization mutants E594Q and E594N (A and B) are similar to those for small hydrophobic and large aromatic amino acids (C and D). In contrast, strongly inward rectifying responses were observed for the charge reversal mutants E594K and E594R as well as for E594H (E and F).

Figure 10. Analysis of permeant block by polyamines with a 2 barrier 1 site model.

A, changes in K_d(0) for mutations at the Q/R site and E594, plotted as K_d(0) Mut/K_d(0) WT, reveal a 10⁷-fold range from the most (Q590W) to least (E594W) sensitive mutant examined. B, energy profiles for changes in well depth and barrier heights (external - internal) were estimated using a 2B1S model for polyamine block. C, electrical distance for polyamine block from the internal face of the membrane calculated as the sum δ1 +δ2 reveals only small changes between mutants.

Table 1.

Estimates of K_d(0) and barrier well profiles for polyamine block

Mutant	n	Kd(0) (μm)	WD(RT)	EB–IB(RT)	δ1 +δ2
Q590G	5	33.1 ± 4.0	−10.6 ± 0.03	2.0 ± 0.20	0.47 ± 0.01
Q590A	7	8.75 ± 0.34	−11.9 ± 0.04	2.6 ± 0.18	0.43 ± 0.01
Q590V	6	3.60 ± 0.10	−12.7 ± 0.05	3.8 ± 0.23	0.44 ± 0.01
Q590L	6	6.11 ± 0.17	−12.2 ± 0.03	4.4 ± 0.22	0.36 ± 0.01
Q590Y	5	0.83 ± 0.02	−14.3 ± 0.05	2.1 ± 0.22	0.42 ± 0.01
Q590F	5	0.28 ± 0.01	−15.2 ± 0.05	4.9 ± 0.27	0.54 ± 0.01
Q590W*	6	0.19 ± 0.03	−22.5 ± 0.04	>7.8 ± 0.04	0.42 ± 0.01
Q590N	7	20.4 ± 2.2	−11.0 ± 0.02	3.0 ± 0.14	0.33 ± 0.01
Q590D	5	46.6 ± 3.5	−10.2 ± 0.05	2.5 ± 0.11	0.43 ± 0.01
WT	17	1.25 ± 0.02	−13.8 ± 0.04	4.0 ± 0.2	0.44 ± 0.01
Q590E	9	4.03 ± 0.14	−12.7 ± 0.06	2.2 ± 0.25	0.39 ± 0.01
Q590S	6	39.5 ± 3.3	−10.4 ± 0.04	2.2 ± 0.12	0.38 ± 0.01
Q590C	6	7.23 ± 0.26	−12.0 ± 0.04	3.6 ± 0.14	0.46 ± 0.01
Q590T	6	8.80 ± 0.39	−11.8 ± 0.04	3.2 ± 0.21	0.41 ± 0.01
Q590M	6	1.16 ± 0.03	−13.9 ± 0.06	2.9 ± 0.62	0.53 ± 0.01
E594D	5	1.26 ± 0.02	−13.8 ± 0.04	4.2 ± 0.18	0.45 ± 0.01
E594N	5	109 ± 3.4	−9.4 ± 0.08	1.0 ± 0.2	0.40 ± 0.01
E594Q	5	1020 ± 391	−7.3 ± 0.16	0.77 ± 0.43	0.29 ± 0.03
E594W	7	2150 ± 327	−6.2 ± 0.12	1.31 ± 0.18	0.42 ± 0.03

Responses from 5–9 oocytes per mutant and 17 oocytes for wild-type GluR6(Q) (WT) were corrected for rectification in the absence of polyamines, then fitted with the sum of two Boltzmann functions to estimate K_d(0); well depth (WD) and the difference in height between the external and internal barriers (EB – IB); and the electrical distance from the cytoplasm to the polyamine binding site (δ1 +δ2); nIndicates the number of cells per mutant. Well depth and barrier heights are given in units of RT

^*Value of K_d(0) for Q590W is given in nm; since no relief from block was observed on depolarization to 200 mV the difference in barrier heights is at least 7.8 RT. For E594N, E594Q and E594W, values are from HEK cell patches with 1 mm spermine added to the internal solution.

Comparison of responses for E594Q and E594N recorded in standard Ba²⁺ solution with those recorded in 0.5 mm EDTA revealed that divalent cations did indeed act as permeant blockers for these mutants (Fig. 7). The evidence for this was that the block by divalent cations first increased and then decreased with hyperpolarization from 100 to -100 mV. However, even in the absence of external divalent cations, responses for E594Q and E594N remained birectifying, suggesting that polyamines might also act as permeant blockers albeit with lower affinity than for wild-type GluR6(Q) (Fig. 7B). With the exception of the charge reversal mutants E594R and E594K, which produced strongly inward rectifying responses, and the similar result obtained for E594H (Fig. 7E and F), we observed only small differences in the amount of rectification when other amino acid side chains were introduced at E594. Of particular interest were the G-V plots recorded in EDTA following introduction of small hydrophobic (E594A) and large aromatic (E594W) side chains (Fig. 7C and D), which were nearly identical to those for E594Q (Fig. 7A and B). These results are strikingly different from those at the Q/R site where Q590W produced much more high-affinity polyamine block than Q590A or Q590N. Also different were the effects of positively charged side chains (Fig. 7E and F), which produced strongly inward rectifying responses with conductance ratios at G₁₀₀ /G_-100 of 0.57 ± 0.06 (n= 5), 0.42 ± 0.03 (n= 5) and 0.52 ± 0.03 (n= 4) for E594R, E594K and E594H, respectively. In contrast, at the Q/R site, Q590R and Q590K produced outward rectifying responses (Fig. 3E and F). In addition, at E594 introduction of a His residue had similar effects to Arg and Lys, while at the Q/R site His produced birectifying responses most probably because at the Q/R site the imidazole ring is within the membrane electrical field and thus senses the transmembrane electrochemical gradient for protons. The extensive series of amino acid substitutions described here suggests that the Q/R site and the conserved ring of negative charge at E594 play distinct roles in ion permeation, and although separated by only three amino acids, probably sense substantially different fractions of the membrane electric field. In view of this it was instructive to examine the double mutant Q590E/E594Q in which the position of the negative charge was moved from E594 to the Q/R site. This construct produced weakly rectifying responses similar to those observed for E594Q (data not shown), indicating that high-affinity polyamine block requires a negative charge at E594 and that moving the position of the side chain by just four residues disrupts polyamine binding.

When E594 mutants were expressed in oocytes, the weak biphasic rectification observed in the absence of polyamines, and permeant block by extracellular divalent cations, made it difficult to determine whether there was any additional rectification produced by endogenous polyamines. To determine if the weak birectifying GluR6 responses observed in the absence of divalent cations following the introduction of Gln, Asn, Ala or Trp at E594 (Fig. 7)in part resulted from low-affinity block by cytoplasmic polyamines, we studied HEK cell outside-out patch responses for these mutants under two conditions. First, with 10 mm Na₂ ATP added to the internal solution to chelate endogenous polyamines, and second with 1 mm spermine (no ATP) added to the internal solution in an attempt to increase the extent of polyamine block, which at the concentrations of spermine and spermidine present in Xenopus oocytes was too weak to analyse with accuracy (Fig. 8). The first set of experiments revealed that in the absence of cytoplasmic polyamines, E594Q, E594N, E594A and E594W remained weakly birectifying, suggesting that polyamine block is not the mechanism underlying the weak rectification observed in Xenopus oocytes. The second set of experiments, with 1 mm spermine added to the internal solution, revealed an increase in biphasic rectification, especially for E594N (Fig. 8B), indicating that polyamines can produce low-affinity block even when the negative charge at E594 is absent. The stronger block by polyamines for E594N versus E594Q probably accounts in part for the slightly greater rectification observed for E594N expressed in Xenopus oocytes (Fig. 7B), and suggests that at position 594 Asn produces higher affinity block by spermine than Gln, while at the Q/R site the opposite is true for this pair of residues (Fig. 3). In contrast, responses for E594K were strongly inward rectifying in both HEK cell patches and Xenopus oocytes (Fig. 8A). Together, the results of these experiments suggest that the conserved negative charge at E594 is absolutely required for high-affinity block by polyamines, and in addition, indicate that the charge at this site influences permeation of monovalent cations, producing weak outward rectification independent of polyamine block when negative and strong inward rectification when positive.

Woodhull analysis of permeant block by polyamines

Prior analysis of polyamine block has shown that an asymmetrical two barrier one site (2B1S) model originally developed by Woodhull (1973) to analyse the action of permeant blockers on Na⁺ channels describes well the action of internal spermine on wild-type GluR6(Q) (Bowie et al. 1998). In the absence of polyamines, responses for wild-type GluR6(Q) show outward rectification (Bähring et al. 1997). Our prior analysis gave the best fits when we assumed that the underlying mechanisms causing rectification in the presence and absence of polyamines were independent and additive (Bähring et al. 1997; Bowie et al. 1998). The same approach was used here for analysis of GluR6 pore region mutants. When G-V plots for E594W, E594Q and E594N recorded with 1 mm internal spermine were corrected using responses recorded with 10 mm internal Na₂ ATP and no added polyamines, the Woodhull model gave good fits to the resulting data (Fig. 8B). The results of such analysis revealed K_d(0) values of 2150 ± 327 μm(n= 7) for E594W, 1020 ± 391 μm(n= 5) for E594Q and 109 ± 3.4 μm(n= 5) for E594N versus a value of 1.25 ± 0.02 μm for wild-type GluR6(Q). Surprisingly, for these E594 charge neutralization mutants, the voltage dependence of onset and relief from block was similar to that observed for wild-type GluR6(Q), and estimates of zθ, the electrical distance to the binding site from the cytoplasm, gave values of 1.68 ± 0.12 for E594W, 1.16 ± 0.16 for E594Q and 1.60 ± 0.04 for E594N. The zθ values for E594W and E594N are similar to the value of 1.76 for wild-type GluR6(Q), which agrees well with prior estimates in HEK cells (Bähring et al. 1997; Bowie et al. 1998); the lower value of zθ obtained for E594Q possibly reflects a relatively inaccurate measurement due to the extremely weak block produced by even 1 mm spermine (Fig. 8B).

We then used a similar approach to analyse changes in polyamine block produced by mutations at the Q/R site (Table 1 and Fig. 9). Because block by polyamines was much stronger for Q/R site mutants than for E594 mutants, it was possible to analyse G-V curves for responses in Xenopus oocytes using block by endogenous polyamines rather than having to use HEK cell patches and high concentrations of spermine added to the internal solution as was required for E594 mutants. Different from results obtained for E594 mutants, responses for Q/R site mutants recorded from HEK cell patches in the absence of polyamines showed weak outward rectification that did not vary substantially with either the charge or size of the amino acid side chain (Fig. 9A). Thus rectification, measured as the ratio of the conductance at 100/-100 mV, was similar for Q590D (2.19 ± 0.11) and Q590R (2.62 ± 0.53); the only exception to this was the previously described weakly birectifying response for Q590H (Fig. 4). Responses for Q/R site mutants recorded in Xenopus oocytes were then corrected using data recorded from HEK cell patches with 10 mm internal Na₂ ATP. The resulting G-V plots were well fitted by a Woodhull model for permeant block over the range ±100 mV and for the mutants examined revealed a 2.5 × 10⁵-fold range of estimates for K_d(0). Representative examples are shown in Fig. 9B. In contrast to the large shifts in K_d(0), the voltage dependence of polyamine block did not vary greatly for Q/R site mutants; for example, for Q590D and Q590W, the pair of mutants with the greatest difference in K_d(0), we obtained zθ values of 1.72 and 1.68, comparable to values obtained for the other mutants examined and similar to the value of 1.76 obtained for wild-type GluR6(Q).

Effects of amino acid substitutions on barrier heights and well depth

By fitting G-V plots over an extended range of membrane potential we were able to estimate changes in well depth and differences between external and internal barrier heights as a function of the amino acid side chains substituted at the Q/R site and E594 (Table 1). The results of this analysis are plotted in Fig. 10 together with the ratio K_d(0) Mut/K_d(0) WT and an estimate of the electrical distance of block measured from the inside of the cell. Several trends emerge. Introduction of small polar side chains attenuated polyamine block independent of whether the residues were neutral or negatively charged, as shown by the similar K_d(0) values for Q590D and Q590S, which were 46.6 ± 3.5 μm and 39.5 ± 3.3 μm, respectively. K_d(0) values for small side chains were 10- to 40-fold greater than those for larger residues such as Q590E, K_d(0) 4.03 ± 0.14 μm, and Q590M, K_d(0) 1.2 ± 0.03 μm. The reduction in polyamine block for Q590D and Q590S occurred due to decreases in both well depth, by 3.6 and 3.4 RT respectively, as well as reductions by 1.5 and 1.8 RT of the difference in height between the internal and external barriers.

We also noted that for side chains of closely related structure an increase in polarity tended to lower the difference in height between the external and internal barriers (Fig. 10B), thus facilitating relief from polyamine block with strong depolarization (e.g. Fig. 3). Thus for Q590F versus Q590Y the difference in barrier heights decreased from 4.9 ± 0.3 RT to 2.1 ± 0.2 RT; similar trends are evident for wild-type GluR6(Q) versus Q590E, 4.0 ± 0.2 RT and 2.2 ± 0.2 RT, respectively, and for Q590N versus Q590D, 3.0 ± 0.1 RT and 2.5 ± 0.1 RT, respectively.

When the hydrophobic series Ala, Val, Leu was introduced at the Q/R site the difference in external and internal barrier heights increased with side-chain size, from 2.6 ± 0.2 RT(Ala), 3.8 ± 0.2 RT(Val) to 4.4 ± 0.2 RT(Leu), indicating that the side chain at the Q/R site is likely to form a barrier limiting the permeation of polyamines through the pore. The greatest difference was observed for the pair Q590G versus Q590F, 2.0 ± 0.2 RT and 4.9 ± 0.3 RT, respectively. For Q590W we were unable to record any relief from block even when the membrane potential was depolarized to 200 mV; this sets a lower limit for V_p of 300 mV with a difference in external and internal barrier heights of at least 7.8 RT. It is remarkable that the effects of all of these mutations occur without marked changes in the voltage dependence of polyamine block (Fig. 10C).

Mutations at E594 produced such large reductions in polyamine affinity that even with a spermine concentration of 1 mm it was difficult to obtain accurate estimates of changes in well depth, barrier height and electrical distance. The most reliable data was obtained for E594N with a K_d(0) of 109 ± 3.4 μm(Fig. 8), 90-fold greater than wild-type GluR6(Q) and 5-fold greater than for Q590N. Comparison of values for zθ, the electrical distance of block for E594N (1.6 ± 0.04) with the value of 1.8 ± 0.04 for wild-type GluR6(Q) revealed little change in the voltage dependence of block despite the 90-fold increase in K_d(0). What distinguished the effect of mutations at E594 from those at the Q/R site was the much larger reduction in difference between external and internal barrier heights produced by mutations at E594. For example, the change was 3.0 ± 0.2 RT for E594N but only 1.0 ± 0.1 RT for Q590N. A similar large effect on relative barrier heights was observed for E594Q (Fig. 10B); in contrast, for E594D, barrier heights, well depth and electrical distance for block were all comparable with those for wild-type GluR6(Q) (Table 1 and Fig. 10), indicating that the different profile for E594N reflects the change in charge but not size of the amino acid side chain at position 594.

DISCUSSION

Our experiments focused on two amino acids in the pore loop of polyamine-sensitive AMPA and kainate receptors, the Q/R site and a conserved glutamate or aspartate residue located four amino acids towards the C-terminus which we refer to as the +4 site. Mutation of both positions was already known to attenuate the biphasic rectification characteristic of polyamine block (Curutchet et al. 1992; Dingledine et al. 1992) but little was known about the underlying mechanisms. Both positions are labelled by substituted cysteine accessibility analysis (Kuner et al. 1996, 1997) and thus their amino acid side chains would be expected to interact with polyamines as these enter the channel. The goal of our experiments was to gain insight into the role the amino acid side chains at these positions play in the normal function of wild-type AMPA and kainate receptors via an analysis of the effects of introducing mutations at the Q/R site and +4 site.

Does the Q/R site form a binding site for polyamines?

The substantial increase in outward current observed for Xenopus oocyte responses when the smaller Asn residue was substituted for Gln at the Q/R site in GluR3 (Dingledine et al. 1992) could have resulted from a reduction in height of an external barrier influencing the movement of polyamines through the pore. Simulations using a two barrier one site (2B1S) model which describes extremely well polyamine block for wild-type GluR6(Q) (Bowie et al. 1998) showed that varying the external barrier height produced the expected attenuation of block but did not produce the parallel rightward shifts in G-V plots that were characteristic of mutations at the Q/R site (Fig. 3). Instead, when fitted to our data a 2B1S model revealed changes in both external barrier height and well depth (Fig. 10). The change in well depth, which accounts for most of the attenuation of polyamine block at negative potentials, suggests that the Q/R site contributes to a structure which binds polyamines, but the nature of the interaction of the Gln side chain with polyamines remains to be determined. Surprisingly, introduction of Glu or Asp at the Q/R site lowered affinity for polyamines. The mechanism for this is also unknown but might reflect competition between permeant monovalent cations and polyamines with increased binding of monovalent cations at the carboxyl groups of the introduced sites.

Our experiments also suggest that the Q/R site contributes to the external barrier which regulates movement of cytoplasmic polyamines from their binding site(s) in the pore to the external solution. Thus, while introduction of Met and Tyr produced almost no change in polyamine affinity compared with wild-type GluR6(Q), the barrier profiles for this pair of mutants differed substantially (Fig. 10). For more closely matched pairs of residues such as Gln-Glu, Asn-Asp and Phe-Tyr, we observed a systematic reduction of the difference in height of the external and internal barriers when the Q/R site side-chain polarity was increased, suggesting that this allows polyamines to pass through the pore more easily; we interpret this as resulting from a reduction in external barrier height. Because we found little change in the voltage dependence of polyamine block produced by mutations at the Q/R site, even for the large aromatic residues Phe, Tyr and Trp, it would appear that the channel structure is not greatly perturbed by the introduction of a broad variety of residues at this site.

Cation-π and van der Waals interactions as mechanisms for high-affinity polyamine block

The 6000-fold increase in affinity for polyamine block observed for the Q590W mutant (Figs 6 and 10) was unexpected. The mechanism we propose combines the binding energy of both the charged ammonium groups and the hydrophobic methylene groups in spermine and spermidine, which would normally be expected to have different amino acid side-chain preferences. The π-electron orbitals in the planar rings of aromatic amino acids, especially tryptophan, strongly favour the binding of cations (Dougherty, 1996) and thus would be expected to interact with the NH₂ ⁺ and NH₃ ⁺ groups in polyamines as they pass through the pore. This mechanism for high-affinity polyamine binding in the Q590W mutant is reminiscent of that previously proposed for the external TEA binding site in potassium channels (Heginbotham & MacKinnon, 1992) and further reinforces our proposal that the tryptophan side chains in the Q590W mutant interact directly with polyamine molecules in a narrow region of the glutamate receptor pore. In addition, the CH₂ groups in spermine and spermidine could potentially interact with the tryptophan side chain via van der Waals interactions. The crystal structure of the bacterial spermidine-binding protein PotD shows an example of the latter mechanism with the ligand-binding site in PotD containing three tryptophan residues which make van der Waals contacts with spermidine CH₂ groups (Sugiyama et al. 1996). The much higher polyamine affinity observed for Q590W versus Q590F and Q590Y most likely reflects both the known stronger cation-π bonding for Trp versus Phe and Tyr (Dougherty, 1996), as well as the larger surface area in the indole ring of Trp for van der Waals contacts with polyamine CH₂ groups. It seems likely that Trp side chains from multiple subunits are required to make the high-affinity polyamine binding site in homomeric Q590W because, when co-expressed with wild-type GluR6(Q) at ratios of mutant to wild-type of 1:1 and 2:1, the half-block potential was shifted by only -16 and -64 mV, respectively, much less than the -150 mV shift for homomeric Q590W, indicating that it is unlikely that the binding energy for spermine is a linear function of the number of Trp residues in heteromeric channels. As a result, we propose that the tryptophan side chains from individual subunits are likely to be parallel to the central axis of the ion channel pore in order to create an aromatic cage which allows simultaneous interaction of multiple indole rings with polyamine ammonium and CH₂ groups.

Role of the negative charge at E594

The functional role of the +4 site conserved negative charge in AMPA and kainate receptors is less well established than that of the Q/R site but it is clear from the present study that the amino acid at this position plays a major role in generating high-affinity polyamine block. Thus, the 800-fold decrease in polyamine affinity for E594Q was much larger than the 50-fold reduction in K_d(0) observed for all Q/R site mutants except those with aromatic or positively charged side chains. In addition, we found that mutations at the +4 site had effects on rectification in the absence of polyamines suggesting that this position is likely to influence the permeation of monovalent cations and is most probably much closer to the entrance of the pore than the Q/R site. It is of interest that the KA-1 subunit has a threonine at the +4 site (Fig. 2) since we would expect that heteromeric receptors assembled from GluR6 and KA-2 would show weaker polyamine block than receptors generated by co-assembly of GluR6 with KA-1 or other members of the kainate receptor gene family.

Several lines of evidence suggest that E594 may be located near the internal face of the pore, in a region where the electric field has not yet begun to drop appreciably. First, one effect of neutralization of the charge at E594 is to reduce the difference between the internal and external barrier heights. Most likely this occurs because the negative charge generates a local electric field that serves to lower the internal barrier height, facilitating entry of polyamines and permeant ions into the pore, and perhaps also increasing the concentration of polyamines near the entrance of the pore. The double mutant Q590E/E594Q failed to show high-affinity polyamine block, even though K_d(0) for the Q/R site point mutant Q590E was similar to that of wild-type GluR6, most likely because in the double mutant the missing negative charge at the +4 site makes it difficult for polyamines to enter the pore. It is also possible that E594 contributes to the site which binds polyamines in the absence of agonist to produce closed channel block (Bowie et al. 1998; Rozov et al. 1998).

The strong inward rectification observed in the absence of polyamines when the charge at E594 is reversed most likely results from a large increase in the internal barrier height, which lowers the rate at which permeant cations enter the pore from the cytoplasm. An additional observation supporting a location of E594 in a position where the membrane electric field is weak comes from comparison of responses for the histidine mutants Q590H and E594H. The biphasic rectification observed for Q590H almost certainly reflects entry of protons into the electric field, such that the time-averaged state of ionization of the imidazole ring changes with both membrane potential and extracellular hydrogen ion concentration. In contrast, responses for E594H are inward rectifying and essentially identical to those for E594K and E594R. Thus, E594H seems not to sense voltage-dependent changes in proton concentration within the ion channel pore, most probably because the imidazole ring is located either at the entrance to the channel, or perhaps in a vestibule where the electric field has not yet dropped appreciably. The similar inward rectification for E594H and E594R suggests that even a partial charge at this site is sufficient to greatly alter the energy profile for permeation of monovalent cations, although we cannot exclude the possibility that the local environment shifts either the pK_a of the His residue or the surface H⁺ concentration to cause full ionization of the imidazole ring at physiological values of cytoplasmic pH.

Our proposal that E594 is a major determinant of polyamine binding, but does not sense a large fraction of the membrane electric field, raises the issue of why block by polyamines shows strong voltage dependence, equivalent to moving 1.8 charges across the membrane electric field. A clue comes from studies on inward rectifier K⁺ channels for which there is already substantial evidence that the movement of polyamines is tightly coupled to movement of potassium ions (Oliver et al. 1998; Pearson & Nichols, 1998; Spassova & Lu, 1998). As a result, the voltage dependence of polyamine block overestimates the fraction of the membrane electric field sensed by polyamine molecules due to the coupled movement of potassium ions deeper in the electric field with polyamines nearer the cytoplasmic entrance. We have already obtained evidence suggesting that a similar mechanism may operate in GluRs, based on our observation that the voltage dependence of block of small N-alkyl-substituted diamines increases with molecular size (Cui et al. 1998) and that the half-block potential for cytoplasmic polyamines shifts with [Na⁺]_o (Bähring et al. 1997). However, different from potassium channels (Oliver et al. 1998; Swanson & Heinemann, 1998), the effect of changing the molecular size of N-alkyl-substituted diamines occurred over a much smaller range, suggesting that the length of the narrowest part of the pore in GluRs in which polyamines interact with permeant ions may be considerably shorter than in K⁺ channels.

A third observation, consistent with a location of E594 either at the entrance to the pore or in a wide vestibule preceding the Q/R site and the region where most of the electric field drops, is that introduction of a tryptophan residue at E594 failed to generate a binding site with high affinity for polyamines. This suggests that in the E594W mutant the tryptophan residues are too far apart to form an aromatic cage like that found in polyamine-binding proteins (Sugiyama et al. 1996) and which most probably occurs in the Q590W mutant. However, we cannot exclude the possibility that the side-chain geometry for the E594W mutant is different from that for Q590W such that the indole rings for the E594W mutant are rotated away from, rather than parallel to the axis of the pore.

A distributed structure for polyamine binding

The distributed cationic and hydrophobic surfaces in spermine, the extended conformation of which spans 1.7 nm, makes it likely that there will be multiple pore-lining amino acids which contribute to polyamine binding during open channel block of GluRs. Consistent with this, our prior studies with polyamines of different structure indicated that with a symmetrical transmembrane [Na⁺] gradient, at 0 mV membrane potential, around 85% of the binding energy for spermine arose from hydrophobic interactions (Cui et al. 1998). Compared with other GluR channel blockers, for example block of NMDA receptors by external and internal Mg²⁺(Wollmuth et al. 1998a,b), the distributed cationic and hydrophobic surfaces in polyamines make it considerably more challenging to interpret the mechanisms by which they block AMPA and kainate receptors. We propose a scheme in Fig. 11 which combines the present results with prior studies on polyamine block in GluRs (Bähring et al. 1997; Bowie et al. 1998). We suggest that spermine molecules are anchored to the entrance of the channel, or perhaps the entrance of an inner vestibule, by an interaction of one or more of the polyamine ammonium groups with the conserved negative charge at position +4, which is held in place by a pore loop (Doyle et al. 1998). Additional binding energy comes from interactions of the remainder of the polyamine molecule with adjacent amino acids.

Figure 11. A distributed binding site for polyamines.

The +4 site conserved negative charge (E) in AMPA and kainate receptors and the Q/R site (Q) lie on a short sequence of extended or random coil structure supported by a pore loop helix formed by the preceding sequence (Kuner et al. 1996, 1997). When bound to the +4 site, polyamine molecules can reach up into the pore towards the Q/R site. The movement of Na⁺ ions entering the pore from the external solution is coupled to movement of polyamines, giving a high voltage dependence of block (Cui et al. 1998) and coupling of half-block potential to [Na⁺]_o (Bähring et al. 1997).

The distributed binding surface for spermine could extend over a distance of up to 3.4 nm away from the +4 site and thus involve many amino acid side chains. When the distal ammonium group in spermine is bound at the +4 site, we propose that the polyamine molecule reaches into the narrower region of the pore towards the Q/R site. Positively charged residues at the Q/R site repulse entry of polyamines deeper into the pore, and thus edited AMPA and kainate receptors do not show voltage-dependent polyamine block. Permeant cations entering the pore from the external solution also repulse those polyamine molecules, which though bound at the +4 site have pushed their remaining ammonium groups into the narrower region of the pore, thus coupling movement of polyamines to movement of Na⁺(Bähring et al. 1997) and generating a high voltage dependence of polyamine block. The mechanism we propose is one that is highly dynamic and ultimately involves multiple pore residues rather than a discrete locus. During permeation of polyamines it is likely that different sets of amino acid side chains and perhaps main chain carbonyl oxygen atoms are involved in binding spermine such that as the blocker moves deeper into the pore the role of the Q/R site and surrounding amino acids becomes increasingly important. Of particular note, even when the negative charge of the +4 site is neutralized, block by polyamines shows almost no change in voltage dependence for E594N and E594W. We propose that the location of the +4 site is shallow compared with the electric field and that a substantial fraction of the voltage dependence of polyamine block is indirect and results from coupling to movement of permeant ions. A major challenge for the future will be to determine the affinity and molecular identity of the binding site(s) for permeant monovalent cations in AMPA and kainate as well as NMDA subtype GluRs (Antonov et al. 1998).