The channel of the glutamate N-methyl-d-aspartate receptor (NMDAR) transports Ca2+ approximately four times more efficiently than that of Ca2+-permeable α-amino-3-hydroxy-5-methyl-4-isoxazolepropionate receptors (AMPAR). To investigate the basis of this difference in these glutamate receptors (GluRs), we measured the ratio of Cs+ efflux and Ca2+ influx in recombinant NMDAR and Ca2+-permeable AMPAR channels expressed in human embryonic kidney 293 (HEK 293) cells over a wide voltage range. At any one potential, this biionic flux ratio was measured by quantifying the total charge and the charge carried by Ca2+ using whole-cell currents and fluorometric techniques (dye overload) with Cs+ internally and Ca2+ externally (1.8 or 10 mM) as the only permeant ions. In AMPAR channels, composed of either GluR-A(Q) or GluR-B(Q) subunits, the biionic flux ratio had a biionic flux-ratio exponent of 1, consistent with the prediction of the Goldman-Hodgkin-Katz current equation. In contrast, for NMDAR channels composed of NR1 and NR2A subunits, the biionic flux-ratio exponent was ∼2, indicating a deviation from Goldman-Hodgkin-Katz. Consistent with these results, in NMDAR channels under biionic conditions with high external Ca2+ and Cs+ as the reference ions, Ca2+ permeability (PCa/PCs) was concentration dependent, being highest around physiological concentrations (1–1.8 mM; PCa/PCs ≈ 6.1) and reduced at both higher (110 mM; PCa/PCs ≈ 2.6) and lower (0.18 mM; PCa/PCs ≈ 2.2) concentrations. PCa/PCs in AMPAR channels was not concentration dependent, being around 1.65 in 0.3–110 mM Ca2+. In AMPAR and NMDAR channels, the Q/R/N site is a critical determinant of Ca2+ permeability. However, mutant AMPAR channels, which had an asparagine substituted at the Q/R site, also showed a biionic flux-ratio exponent of 1 and concentration-independent permeability ratios, indicating that the difference in Ca2+ transport is not due to the amino acid residue located at the Q/R/N site. We suggest that the difference in Ca2+ transport properties between the glutamate receptor subtypes reflects that the pore of NMDAR channels has multiple sites for Ca2+, whereas that of AMPAR channels only a single site.
N-methyl-d-aspartate receptors (NMDAR)1 and α-amino- 3-hydroxy-5-methyl-4-isoxazolepropionate receptors (AMPAR) mediate fast excitatory neurotransmission in the mammalian central nervous system. Post-synaptic influx of Ca2+ via glutamate receptors (GluR) is thought to be a critical step for the induction of long-term changes in synaptic strength and neurotoxicity (Choi, 1988; Bliss and Collingridge, 1993). Native NMDAR channels are heteromers composed of the constitutive NR1 subunit and one or more of four different NR2 subunits (A, B, C, D; for review, see Hollmann and Heinemann, 1994). All NMDAR subtypes are, with small quantitative differences, highly permeable to Ca2+ (Monyer et al., 1994; Burnashev et al., 1995; Schneggenburger, 1996). In contrast, AMPAR subtypes are more diverse in their ability to transport Ca2+ (Hollmann et al., 1991; Hume et al., 1991; Burnashev et al., 1992a; for review, see Burnashev, 1996). Four different AMPAR subunits have been cloned: GluR-A, -B, -C, and -D (alternatively, GluR1-4) (Hollmann and Heinemann, 1994). Ca2+-impermeable AMPAR channels contain the edited form of the GluR-B subunit, termed GluR-B(R), which contains an arginine (R) at the functionally critical Q/R site. AMPAR channels containing only the GluR-A, -C, or -D subunits, which contain a glutamine (Q) at this position, are Ca2+ permeable. In NMDAR channels, the homologous position to the Q/R site, the N site, is occupied by an asparagine (N) that also contributes to Ca2+ transport (Burnashev et al., 1992b).
NMDAR and Ca2+-permeable AMPAR channels show a high but not exclusive selection of Ca2+ over monovalent alkali cations. The inward current that the open channel mediates under physiological conditions is carried by a mixture of monovalent (Na+ and K+) and Ca2+ ions (MacDermott et al., 1986; Mayer and Westbrook, 1987). Recently, a method has been developed using Ca2+ photometry and high concentrations of intracellular fura-2 (dye overload), which allows the fraction of the total current carried by Ca2+ to be quantified over a wide voltage range (for review, see Neher, 1995). The NMDAR and Ca2+-permeable AMPARs show a quantitative difference in their fractional Ca2+ currents (Schneggenburger et al., 1993; Burnashev et al., 1995). For example, at −60 mV and in 1.8 mM Ca2+, recombinant NMDAR channels composed of the NR1–NR2A subunits carry a fractional Ca2+ current of ∼11%, whereas, for Ca2+-permeable AMPAR, it is ∼3% (Burnashev et al., 1995). NMDAR channels also have a higher Ca2+ permeability as assessed under biionic conditions than Ca2+-permeable AMPAR channels (e.g., Burnashev et al., 1995).
The mechanism of Ca2+ transport in GluR channels remains unclear. Since current amplitudes at any one potential are a balance between inward- and outward-directed currents, understanding mechanisms of ion permeation would be facilitated by defining the properties of the pores that control unidirectional currents over a wide voltage range. Such measurements are typically made only at the zero current or reversal potential, where the inward- and outward-directed currents are exactly balanced, yielding no net current. Although such reversal potential measurements can provide insights into permeation mechanisms (see Hille, 1992), they are limited in that they give information only at a single potential. Alternatively, measurements of unidirectional fluxes at potentials other than the reversal potential have been made using a radioactive tracer on one side of the membrane to distinguish the fluxes. This approach, based on the Ussing flux-ratio test (Ussing, 1949), has been used for voltage-gated K+ channels using 42K (Hodgkin and Keynes, 1955; Horowicz et al., 1968; Begenisich and De Weer, 1980; Vestergaard-Bogind et al., 1985; Stampe and Begenisich, 1996), and for voltage-gated and amiloride-sensitive Na+ channels using 22Na and/or 24Na (Begenisich and Busath, 1981; Benos et al., 1983). While this approach has the advantage of examining unidirectional fluxes over a wide potential range, it has the disadvantage of requiring the use of radioactive tracers and therefore has not been extensively used, especially with recombinant ion channels.
To compare the mechanism of Ca2+ transport in different GluR channels, we measured at a fixed potential total charge and the charge carried by Ca2+ using dye overload with Cs+ intracellularly and Ca2+ extracellularly as the only permeant ions. This allowed us to quantify unidirectional Cs+ and Ca2+ fluxes over a wide voltage range. This approach is comparable with, but not identical to, the Ussing flux-ratio test, which requires that the ion species on both sides of the membrane be the same (see materials and methods and discussion). We find that in NMDAR channels the biionic flux ratio of Cs+ efflux to Ca2+ influx shows a strong deviation from the prediction of the Goldman-Hodgkin-Katz (GHK) current equation with a biionic flux-ratio exponent of ∼2. In contrast, Ca2+-permeable AMPA channels show no such deviation having a biionic flux-ratio exponent of 1. This difference between GluR channels is not due to the amino acid residue at the Q/R/N site. Thus, the mechanism underlying Ca2+ transport in the two types of GluRs is different, and this may be due to differences in the amino acid composition of the external vestibule.
Materials And Methods
Heterologous Expression of GluR Channels
All experiments were performed with previously described expression constructs for wild-type NR1–NR2A NMDAR subunits (Wollmuth et al., 1996) and AMPAR subunits (Burnashev et al., 1992a). AMPAR subunits were identified following the nomenclature of Seeburg (1993), with the amino acid residue occupying the Q/R site indicated in parenthesis. Channels were expressed transiently [NR1-NR2A, GluR-B(N)] or permanently [GluR-A(Q), GluR-B(Q)] in human embryonic kidney 293 (HEK 293) cells. All AMPAR channels were of the flip-splice variant form.
The standard intracellular solution used to measure fractional Ca2+ currents consisted of (mM): 140 CsCl, 10 HEPES, and 2 mM K5·fura-2, pH adjusted to 7.2 with CsOH. The total intracellular monovalent concentration, [M]i, was 153.5 mM. In measuring Ca2+ reversal potentials, the solution was the same except that 10 mM BAPTA replaced the fura-2 ([M]i = 163.5). HEPES and EGTA were obtained from Carl Roth (Karlsruhe, Germany) and BAPTA and fura-2 from Molecular Probes (Eugene, OR).
Fractional Ca2+ currents were measured using an extracellular solution consisting of (mM): 1.8 CaCl2, 140 NaCl, and 10 HEPES, pH adjusted to 7.2 with NaOH. To measure fractional Ca2+ currents in “pure” extracellular Ca2+, 1.8 or 10 mM Ca2+ was added to the following N-methyl-d-glucamine (NMDG)–based solution (mM): 140 NMDG and 10 HEPES, pH adjusted to 7.2 with HCl. The high Cs+ solution used as a reference to quantify Ca2+ permeation consisted of (mM): 140 CsCl and 10 HEPES, pH adjusted to 7.2 with CsOH ([M]o = 143.5). The high Ca2+ solution consisted of (mM): 108 CaCl2, 2 Ca(OH)2, and 10 HEPES, with the final pH 7.2. For other Ca2+ concentrations, the relevant Ca2+ concentration (0.18–10 mM) was added to the NMDG-based solution. For AMPAR channels, we used 15 μM cyclothiazide to reduce channel desensitization. Two stock solutions of cyclothiazide were used: for all extracellular solutions containing monovalents, the stock solution was 10 mM cyclothiazide dissolved in 100 mM NaOH. Alternatively, for solutions containing Ca2+ and NMDG, 0.5 mM cyclothiazide was dissolved in (mM): 10 Ca(OH)2, 140 NMDG, and 10 HEPES. Appropriate amounts of 140 mM NMDG and 10 mM HEPES and CaCl2 were mixed, along with this stock solution, to yield the final Ca2+ concentration indicated in the text.
Current Recordings and Data Analysis
Currents were recorded at room temperature (20–23°C) using an EPC-9 amplifier with PULSE software (HEKA Elektronik, Lambrecht, Germany), low pass filtered at 500 Hz, and digitized at 2 kHz. Pipettes were pulled from borosilicate glass and had resistances of 0.8–3 MΩ when filled with the pipette solution and measured in the Cs+ external solution. External solutions were applied using a piezo-driven double-barrel application system. For NMDAR channels, one barrel contained the external solution plus 50 μM glycine, and the other barrel the same solution but with added 200 μM glutamate. For AMPAR channels, one barrel contained the external solution plus 15 μM cyclothiazide and the other barrel the same solution but with added 1 mM glutamate. The liquid-junction potential between the CsCl pipette solution and the 143.5 mM Cs+ and 143.5 mM Na+ external solutions was −0.9 and −1.6 mV, respectively (pipette negative). Ca2+-containing solutions generated junction potentials between the ground electrode and high CsCl solution of −10.1 (0.18–1.8 mM CaCl2 in NMDG), −10.5 mV (10 mM CaCl2 in NMDG), and −10.2 mV (110 mM CaCl2) (ground electrode 0 mV). All curve fitting was done using Igor Pro (WaveMetrics, Inc., Lake Oswego, OR). Results are reported in the text as mean ± SEM and shown graphically as mean ± 2*SEM.
Fura-2 (2 mM) was loaded into HEK 293 cells via the patch pipette to measure the fraction of the total current (monovalents and Ca2+) carried by Ca2+ (see Neher, 1995). Cells were illuminated alternatively at 365 and 385 nm (2–10 Hz) by a polychromatic illumination system (T.I.L.L. Photonics, München, Germany). Excitation light was coupled to the microscope via a light guide and was attenuated to 90–95% with neutral density filters. A 425-nm dichroic mirror and a 500–530-nm band-pass emission filter were included in the light path. Fluorescence signals were measured with a photomultiplier (Luigs and Neumann, Ratingen, Germany). Changes in [Ca2+]i were estimated from the fluorescence signals according to Grynkiewicz et al. (1985).
Fractional Ca2+ currents were measured according to Burnashev et al. (1995) and Neher (1995). In brief, the decrement in fura-2 fluorescence at 385-nm excitation (ΔF385) evoked by a Ca2+ influx depends on the relative Ca2+-binding ratio of exogenous (κ′B) and endogenous (κS) buffers:
In the case of overload, that is when the fura-2 Ca2+-binding ratio is much greater than the Ca2+-binding ratio of the endogenous buffer
The proportionality constant, fmax, also termed maximal F/Q ratio (Schneggenburger et al., 1993), between the charge carried by inward Ca2+ (QCa) and ΔF385 was determined as outlined below. The overload approach to quantifying Ca2+ influx requires that all of the incoming Ca2+ is captured by fura-2 rather than by endogenous buffers. The duration of the glutamate application at any one potential therefore was chosen such that [Ca2+]i remained below 200 nM. Also, in instances where Ca2+ was the only external permeant ion, we held the cells at −10 or 0 mV, leading to a low resting [Ca2+]i of ∼10–50 nM. In any case, even under extreme conditions beyond those used in this study (e.g., resting [Ca2+]i of 100 nM and a peak [Ca2+]i of 300 nM), the Ca2+-binding ratio, κ′B, of 2 mM fura-2 (2,660; derived using Eq. 28 in Neher, 1995) still far exceeds the endogenous Ca2+-binding ratio, κS, of HEK 293 cells (≈50; Burnashev et al., 1995). To verify this assumption, we also measured the derived parameter, abs(QCs/ QCa) (see Figs. 3–5), at a single voltage (−21 mV) over a range of QCa (Fig. 1 C ). This parameter was independent of the amount of Ca2+ influx, confirming the idea that dye depletion was not a significant problem even up to 140 pC.
Fractional Ca2+ currents (Pf) were quantified using the relationship:
To account for instabilities of the illumination intensity or the detection efficiency, ΔF385 was normalized to the fluorescence of beads (4.5-μm-diameter fluoresbrite BB beads, Lot 460565; Polysciences Inc., Warrington, PA) and expressed in “bead units” (BU; Schneggenburger et al., 1993). The bead unit was determined on each experimental day as the mean fluorescence of 8–15 beads at 385-nm excitation. Fluorescent measurements were made only on small cells (<6 pF) that were either attached or lifted. Green fluorescent protein was cotransfected to detect transfected cells, but this fluorescent marker has no effect on Pf measurements (Schneggenburger, 1996).
The proportionality constant, fmax, which needs to be measured under conditions where the total charge (QT) ≈ QCa was quantified in two ways. (a) For NMDAR channels, ΔF385 at −100 mV was measured with Cs+ intracellularly and, with 10 mM Ca2+, 140 mM NMDG extracellularly. The total amount of charge at this potential, which was assumed to be carried by Ca2+, was varied by altering the length of the glutamate application (typically 100–1,000 ms) (Fig. 1,A). The fmax was derived from the slope of a fitted line to a plot of ΔF385 against QT and was 1.22 ± 0.05 BU/ nC (n = 5) (Fig. 1,B). We also measured fmax using the same approach at −80 and −60 mV (Fig. 1 B); these values were not significantly different from that measured at −100 mV, confirming the idea that the total current at −100 mV is carried almost exclusively by Ca2+. (b) The second approach to quantify fmax followed that of Schneggenburger (1996). In brief, reversal potential shifts on adding Ca2+ (1.8 or 10 mM) to a Ca2+-free solution (143.5 mM NaCl) were used to estimate the monovalent reversal potential in Ca2+-containing solutions during Pf measurements. Assuming independence of ion movement, the f ratio (or F/Q ratio) at the monovalent reversal potential should solely be due to Ca2+ influx. For NR1–NR2A channels in 1.8 mM Ca2+, this approach yielded an fmax of 1.00 ± 0.02 BU/nC (n = 3), significantly less than that measured using approach a. Based on our measured f ratios for NR1–NR2A channels at −60 mV in 1.8 mM Ca2+, this fmax would yield a Pf of ∼16.5%, comparable to the value measured by Schneggenburger (1996). Approach b was also used for GluR-B(N) channels, which are mutant GluR-B channels containing an asparagine at the Q/R site. These channels were selected since they are Ca2+ permeable but are not blocked by intracellular polyamines, allowing currents to be readily measured around the reversal potential. For GluR-B(N) channels, fmax was 1.30 ± 0.02 BU/nC (n = 5) in 1.8 mM Ca2+ and 1.25 ± 0.03 BU/nC (n = 4) in 10 mM Ca2+. For NR1– NR2A channels, this fmax would yield a Pf at −60 mV (1.8 mM Ca2+) of ∼13%, comparable to values obtained elsewhere for recombinant (Burnashev et al., 1995) and native (Rogers and Dani, 1995) NMDAR channels. We used as an fmax value a grand mean (1.26 ± 0.02 BU/nC) of those values derived from approaches a and b for GluR-B(N) channels for the following reasons: first, they yielded comparable values and, second, independence of ionic movement is a fundamental assumption of approach b, but our results suggest that this assumption is violated in NMDAR but not in GluR-B(N) channels.
Biionic Flux Ratios
The Ussing (1949) flux-ratio equation relates the unidirectional fluxes, φefflux and φinflux, for passive, independent movement of a single ion species across the membrane:
where [S]i and [S]o are the concentration of ion species S on the intracellular and extracellular side of membrane, respectively, and z S its valence. R, T, and F have their normal meanings, and the quantity RT/F was 25.4 mV (21°C). This equation is valid when the intracellular and extracellular ions are the same. However, in our case, we are interested in determining the unidirectional flux ratios for ions having different valence, namely the efflux of a monovalent ion (Cs+) and the influx of a divalent ion (Ca2+). As a theoretical approach to describe these ratios, we started with the unidirectional current components of the GHK current equation (see Hille, 1992):
where QCs and QCa are the total charge carried by Cs+ and Ca2+ during a specific time interval. At any one potential, QT and QCa (= ΔF385/fmax, Eq. 1) were quantified over the same time interval. QCs was derived using the relationship, QT = QCa + QCs. The quantity, QCs/QCa, is not a proper flux ratio since the ions have a different valence. Therefore, we refer to it as a biionic flux ratio to distinguish it from an Ussing type flux ratio. To compare results under different ionic conditions, we expressed Eq. 3 relative to the zero current or reversal potential (Erev,Ca) for Ca2+. Using the Lewis equation (see Eq. 7), Eq. 3 was expressed relative to Erev,Ca:
For simplicity, we define the right hand part of Eq. 4 as Λ.
where Eq. 4 is raised to some power, n′, denoted the biionic flux-ratio exponent. Eq. 5 has a formal similarity to the Ussing type flux-ratio exponent, but the mechanistic interpretation of n′ is obscure. We use it as a short-hand notation to indicate deviations from Eq. 4. At any one potential, it was quantified by solving for n′:
This approach to quantifying the biionic flux ratio requires that the only permeant ions are Cs+ intracellularly and Ca2+ extracellularly. To generate a pure Ca2+-containing solution, we added various concentrations of Ca2+ to 140 mM NMDG. NMDG is impermeant in NMDAR channels (Villarroel et al., 1995). On the other hand, Ca2+-permeable AMPAR channels show a weak NMDG permeability. For GluR-A(Q) channels, PNMDG/ PCs is ∼0.02 (Burnashev et al., 1996), whereas it is ∼0.01 in GluR-B(Q) and essentially impermeant in GluR-B(N) (N. Burnashev, personal communication). To avoid any contamination by NMDG in GluR-A(Q) and GluR-B(Q) channels, we measured the biionic flux ratio only in 10 mM [Ca2+]o, where any contribution of an NMDG current component would be small.
Ca2+ Reversal Potentials
To quantify Ca2+ permeability from reversal potential measurements, we started with the Lewis equation (Lewis, 1979), which takes into account the presence of multiple permeant ions having different valences. With Cs+ as the only permeant ion intracellularly, this equation has the form:
where Pmono/PCs is the permeability ratio of any other permeant monovalent species, and P′Ca is PCa/[1 + exp(ErevF/RT)]. Ca2+ permeability was determined by measuring the change of reversal potential (ΔErev) for glutamate-activated currents on replacing 143.5 mM Cs+ with a pure Ca2+ solution (0.18–10 mM Ca2+ in NMDG or 110 mM Ca2+). Permeability ratios, PCa/PCs, were calculated according to the relation:
where Erev,Ca is the reversal potential in the Ca2+-containing solution, and Erev,Cs is the reversal potential in the reference Cs+-containing solution. For simplicity, we present throughout the manuscript only Ca2+ concentrations. We also calculated PCa/ PCs using activity coefficients (i.e., concentrations in Eq. 8 were multiplied by γ, the activity coefficient). Mean molal activity coefficients were found in the NIST Standard Reference Database 44 (U.S. Department of Commerce, Washington, DC); individual activity coefficients were, following the Guggenheim convention, γCa 0.26 (110 mM Ca2+), 0.30 (10 mM Ca2+) and 0.31 (0.18–1.8 Ca2+), and γCs 0.72 (143.5 mM Cs+). Because these activity coefficients were not greatly different (Ca2+ was present in high concentrations of NMDG), they increased the magnitude of PCa/PCs approximately equally over the entire concentration range (data not shown). In most instances, the control recording was an average of the control recording made before and after exposure to the Ca2+-containing solution.
Measuring flux ratios requires quantifying the unidirectional movement of ions in a channel at a fixed membrane potential. To do so in GluR channels, we simultaneously measured whole-cell currents by voltage clamp and Ca2+ influx with fluorescence under ionic conditions where Cs+ internally and Ca2+ externally were the only permeant ions. Under such conditions, the total charge during a defined time interval (QT) is carried by inward moving Ca2+ and outward moving Cs+ (i.e., QT = QCa + QCs). The unidirectional QCs component was derived as QT − QCa. We refer to the quantity QCs/QCa as a biionic flux ratio (see materials and methods).
Unidirectional Currents in NR1-NR2A NMDAR Channels
Fig. 2, A and B, shows whole-cell glutamate-activated currents in a HEK 293 cell expressing NR1-NR2A channels. The cell was bathed either in 143.5 mM Cs+ (A) or in 10 mM Ca2+, 140 mM NMDG (B) with the pipette solution always containing 163.5 mM Cs+. The current– voltage (I-V) relation of the corresponding peak currents is shown in Fig. 2 C. The currents in the external Cs+ solution (○) cross the voltage axis at −3.6 mV, close to the Nernst potential for Cs+ (−3.3 mV). When Ca2+ replaces Cs+ (•), the I-V relation is strongly outwardly rectifying. However, despite the nearly 14-fold reduction in concentration of a permeant species, the reversal potential is shifted only slightly leftwards, to ∼−12 mV. On average, replacing Cs+ with 10 mM Ca2+ produced a shift in the reversal potential of −8.7 ± 0.3 mV (n = 6), yielding a mean PCa/PCs of 4.2, consistent with Ca2+ being more permeant in NMDAR channels than monovalent alkali cations.
The use of fluorometry, in combination with whole-cell current recording, to quantify unidirectional fluxes is shown in Fig. 3. The recordings are from a HEK 293 cell measured in 10 mM Ca2+, 140 mM NMDG, like the cell in Fig. 2 C (•), except that 2 mM fura-2 was included in the pipette solution. At such high concentrations, fura-2 captures most of the Ca2+ entering the cell (see materials and methods). Changes in the fluorescence signal at 385 nm excitation are therefore proportional to the total Ca2+ influx (QCa), with the proportionality constant defined by fmax (≈1.26 BU/nC; see materials and methods) according to the relationship QCa = ∫ICadt = ΔF385/fmax.
Fig. 3 shows recordings from the same cell with the reversal potential (Erev,Ca) at −11 mV. At −31 mV (Fig. 3,A), which is −20 mV of Erev,Ca, the application of glutamate (solid bar) yields an inward directed current (trace I). As expected, there is also a corresponding decrement in the fluorescence signal at 385 nm excitation (ΔF385), showing that Ca2+ flows into the cell. As shown in the Q plot (bottom), the unidirectional Ca2+ influx (QCa), derived from the ΔF385 measurement, and the total charge (QT), derived from the time integral of the whole-cell current, were inwardly directed but not of equal magnitude. With Cs+ internally and Ca2+ externally as the only permeant ions, yielding a net flux less than the influx of Ca2+ requires that there be a unidirectional efflux of Cs+ (QCs). The ratio of QCs to QCa was essentially constant at each time point (data not shown). However, we quantified QCa and QCs at a single time point typically when F385 reached an initial minimum. At the time interval indicated (Fig. 3, arrow), QT was −66.9 pC and the Ca2+ influx was −76.9 pC, yielding an efflux of Cs+ of ∼10 pC. (Like current flows, we assign inward directed charge movements a negative value.)
At −11 mV (Fig. 3,B), the application of glutamate (solid bar) yields no net current, indicating that this potential represents the reversal potential. There is a large decrement in F385, indicating an influx of Ca2+, which must be exactly balanced by Cs+ efflux to yield a net zero current (Q plot). At +9 mV (Fig. 3 C), which is +20 mV positive to Erev,Ca, a longer glutamate application time (1 s) was required since a smaller portion of the total current was carried by Ca2+. At this potential, QT was outwardly directed. At the indicated time interval, QT was 303 pC. The corresponding Ca2+ influx during the same time interval was −38.9 pC, yielding a Cs+ efflux of 341.9 pC.
Using the methods illustrated in Fig. 3, we quantified in NMDAR channels the unidirectional fluxes of Cs+ and Ca2+ (Fig. 4,A), and the corresponding absolute ratio of the unidirectional fluxes, abs(QCs/QCa), over a wide voltage range. For ease of comparison, the values are plotted relative to the reversal potential (E − Erev,Ca). The voltage range over which the unidirectional fluxes was measured was limited at negative potentials since QCa approached QT (Fig. 4 A), making QCs small and unreliable. At positive potentials, it was limited since long pulses of glutamate were required to obtain a detectable ΔF385 with the baseline F385 having to be extrapolated over a long time.
Biionic Flux Ratios in NMDAR Channels Deviate from the Prediction of the GHK Equation
Fig. 5 summarizes the ratio of the unidirectional Cs+ efflux and Ca2+ influx measured either with 10 mM [Ca2+]o (Fig. 5,A) or 1.8 mM [Ca2+]o (Fig. 5,B) in NR1-NR2A channels. As a theoretical basis, we used the unidirectional current components of the GHK current equation to describe these results (Eq. 3). At both concentrations, the biionic flux ratios show a strong deviation from a biionic flux-ratio exponent of 1 (Eq. 3 and Fig. 5, solid lines). Indeed, in 10 mM [Ca2+]o (Fig. 5,A), the results were best described using an average biionic flux-ratio exponent of 2.02 ± 0.06 (Eq. 5). This biionic flux-ratio exponent showed no voltage dependence from −40 to +40 mV of Erev,Ca. In 1.8 mM [Ca2+]o (Fig. 5,B) and from −40 to +40 mV of Erev,Ca, the biionic flux-ratio exponent was 1.90 ± 0.04, somewhat smaller than that found in 10 mM [Ca2+]o. At very negative potentials, starting at ∼−40 mV of Erev,Ca, the biionic flux ratios started to show a weak voltage dependence with the biionic flux-ratio exponent getting smaller. However, at potentials so negative from the reversal potential, the biionic flux ratio was difficult to quantify accurately since QCa approached QT (see Fig. 4 A; a similar problem occurs in 10 mM [Ca2+]o as seen by the greater variability in biionic flux ratios with negative potentials). Nevertheless, the strong deviation from Eq. 3 at both Ca2+ concentrations suggests that Cs+ efflux and Ca2+ influx in NMDAR channels do not follow the GHK equation, and that at least ±40 mV of the reversal potential, this process is essentially voltage independent.
Biionic Flux Ratios in Ca2+-permeable AMPAR Channels Show a Biionic Flux-Ratio Exponent of 1
Fig. 6 illustrates recordings, comparable to those in Fig. 3, from Ca2+-permeable AMPAR channels assembled from GluR-A(Q) subunits. Fig. 6, A and B, shows I-V relations for glutamate-activated currents in the presence of cyclothiazide from a HEK 293 cell expressing GluR-A(Q) channels bathed either in CsCl or CaCl2. (Fig. 6,B is the same I-V relation as A, but with an expanded current scale.) In Cs+ (□), the I-V relation shows a strong double rectification due to a voltage-dependent block of Ca2+-permeable AMPAR channels by intracellular polyamines (Bowie and Mayer, 1995; Koh et al., 1995a). The currents in the Cs+ solution cross the voltage axis near −4 mV. [Because of the strong block by polyamines around 0 mV, we normally used voltage steps in 2-mV increments to quantify the reversal potentials in the CsCl solution, but use Fig. 6, A and B, to illustrate the overall shape of the I-V relation in GluR-A(Q) channels]. When Ca2+ replaces Cs+ (▪), the I-V relation remains strongly doubly rectifying with the reversal potential shifted leftward to ∼−30 mV. On average, replacing Cs+ with 10 mM [Ca2+]o in GluR-A(Q) channels produced a shift in the reversal potential of −26.6 ± 0.7 mV (n = 4), yielding a mean PCa/PCs of 1.60. This result is consistent with Ca2+ being more permeant in Ca2+-permeable AMPAR channels than monovalent alkali cations, although the relative permeability is less than that in NMDAR channels.
Fig. 6, C and D, shows simultaneous measurement of whole-cell currents and Ca2+ influx, with Cs+ internally and 10 mM Ca2+ externally, to quantify unidirectional fluxes in GluR-A(Q) channels. In this cell, Erev,Ca is ∼−30 mV (data not shown). At −50 mV (Fig. 6,C), the glutamate application in the presence of cyclothiazide elicited a large inwardly directed current. QT at the selected time interval (arrows) was ∼−88.4 pC. Based on the corresponding ΔF385 measurement, inward Ca2+ carried −133.4 pC of this charge, yielding a Cs+ efflux of 45 pC and an abs(QCs/QCa) of 0.38, a value larger than that at the same relative potential found in NMDAR channels (≈0.13, Fig. 3,A). At −10 mV (Fig. 6,D), the glutamate application resulted in a small outward-directed current. The total charge during the selected time interval was 53.2 pC with inward Ca2+ carrying −31.2 pC of this charge, yielding a Cs+ efflux of 84.4 pC. The abs(QCs/QCa) was therefore 2.71, considerably smaller than that for NMDAR channels at the same relative potential (≈8.8, Fig. 3 C).
Fig. 7 summarizes the biionic flux ratio for GluR-A(Q) channels (open symbols) as well as GluR-B(Q) channels (solid symbols) in 10 mM [Ca2+]o over a wide voltage range. Because of the strong block by intracellular polyamines of GluR-A(Q) and GluR-B(Q) channels, this ratio could only be measured until ∼30 mV positive to Erev,Ca. Nevertheless, in both Ca2+-permeable AMPAR channel types, the biionic flux-ratio exponent was 1, following the predictions of the GHK current equation (Fig. 7, solid line).
Ca2+ Permeability in NMDAR but Not in Ca2+-permeable AMPAR Channels Is Concentration Dependent
The results for the biionic flux-ratio experiments suggest that NMDAR but not AMPAR channels deviate from the prediction of GHK. Another indication of such a deviation is a concentration dependence of permeability ratios. To test this possibility, we quantified Ca2+ permeability over a wide concentration range of extracellular Ca2+ for NMDAR (Fig. 8, A and B) and Ca2+-permeable AMPAR (Fig. 8, C and D) channels.
Fig. 8,A summarizes the change in reversal potential on replacing high Cs+ with a wide range of different Ca2+ concentrations (0.18–110 mM) in NR1-NR2A channels. The smooth curve was derived using the GHK equation and assuming a PCa/PCs of 4.8. The curve does not describe the results well as they fall below it both at very low and high concentrations and above it at intermediate concentrations. The deviation from GHK is seen more clearly when PCa/PCs is quantified for individual Ca2+ concentrations (Fig. 8 B). Around physiological concentrations of Ca2+ (1 and 1.8 mM), PCa/PCs shows a peak of 6.2 and 6.1, respectively. On the other hand, PCa/PCs is reduced to 2.6 in 110 and to 2.2 in 0.18.
The decrease in PCa/PCs at low concentrations of Ca2+ was surprising. It is not due to differences in activity coefficients that varied little due to the presence of the positively charged NMDG (see materials and methods). An alternative explanation is that NMDG binds to Ca2+, reducing its effective concentration. However, PCa/PCs measured in 0.18 mM Ca2+ in 140 mM tetramethylammonium (TMA) (PCa/PCs = 2.2 ± 0.2, n = 6) was indistinguishable from that measured in 140 mM NMDG (PCa/PCs = 2.2 ± 0.1, n = 7), requiring that if NMDG binds to Ca2+, then TMA must do so to the same extent. It also cannot be due to NMDG permeating the channel or contaminating concentrations of Ca2+, both of which would lead to smaller changes in ΔErev (i.e., larger PCa/PCss). Hence, the decrease in PCa/ PCs at low concentrations in NMDAR channels appears to be a property of the channel.
Fig. 8, C and D, shows the comparable measurement of Ca2+ permeability over a wide concentration range in AMPAR channels composed of GluR-A(Q) (○) or GluR-B(Q) (□) subunits. In Fig. 8,C, the smooth curve, derived using the GHK equation and assuming a PCa/ PCs of 1.65, describes the changes in reversal potential quite well in 1.8–110 mM Ca2+. The change in the reversal potential in 0.3 mM Ca2+, however, was not as large as expected. Nevertheless, NMDG is weakly permeant in these channels (see materials and methods). Indeed, when corrected for this weak NMDG permeability, PCa/PCs shows essentially no concentration dependence over the entire concentration range (Fig. 8 D).
In summary, Ca2+ permeability ratios in NMDAR channels depend strongly on Ca2+ concentration, indicating a deviation from GHK. In contrast, Ca2+-permeable AMPAR channels do not show concentration-dependent permeability ratios. These results are consistent with the results of the biionic flux-ratio measurements.
Differences in Ca2+ Permeability between NMDAR and Ca2+-permeable AMPAR Are Not Due to the Q/R/N Site
A structural determinant of Ca2+ influx in GluR channels is the amino acid residue located at the functionally critical Q/R/N site in the pore-lining M2 segment (Hollmann and Heinemann, 1994). Ca2+-permeable AMPAR channels have a glutamine (Q) at this site, whereas NMDAR channels have an asparagine (N). To test whether the composition of the Q/R/N site underlies the difference between NMDAR and Ca2+-permeable AMPAR channels, we measured biionic flux ratios, the concentration dependence of reversal potentials, and fractional Ca2+ currents in GluR-B(N) channels (Fig. 9 and Table I). This mutant Ca2+-permeable AMPAR channel has an asparagine (N) substituted at the Q/R site (Burnashev et al., 1992a). We did not test the reverse mutant [glutamine (Q), substituted for the asparagine at the N site in NMDAR channels] since this substitution has profound effects on channel gating (Schneggenburger and Ascher, 1997; see discussion).
Fig. 9,A shows the biionic flux ratio for GluR-B(N) channels in 10 mM [Ca2+]o. Because these channels were not blocked by intracellular polyamines, the biionic flux ratio could be determined at potentials at least 60 mV positive to Erev,Ca. Like GluR-A(Q) and GluR-B(Q) channels, GluR-B(N) channels showed a biionic flux-ratio exponent of 1 between −60 and +60 mV of the reversal potential. Further, PCa/PCs in GluR-B(N) channels (Fig. 9 B, ○) also showed no concentration dependence. Hence, the biionic flux-ratio exponent of 1 and concentration-independent permeability ratios in GluR-B(N) suggest that the difference between NMDAR channels and Ca2+-permeable AMPAR channels is not due to the composition of the Q/R site.
Table I compares fractional Ca2+ currents (Pf) in NMDAR channels and various subtypes of Ca2+-permeable AMPAR channels, including GluR-B(N) at −60 mV. In comparison to GluR-B(Q) (Pf ≈ 4.0%), the fractional Ca2+ current in GluR-B(N) channels is somewhat higher, ∼4.7%. Nevertheless, this value remains considerably below that in NMDAR channels, which is ∼13.3%. Thus, an asparagine at the Q/R site does not yield AMPAR channels with fractional Ca2+ currents comparable with those in NMDAR channels.
Previously, flux ratios have been measured in K+ and Na+ channels using a radioactive tracer on one side of the membrane to distinguish between unidirectional fluxes (see introduction). Using a combination of whole-cell current recordings and Ca2+ fluorescence measurements with intracellular Cs+ and extracellular Ca2+ as the only permeant ions, we quantified unidirectional fluxes in NMDAR and Ca2+-permeable AMPAR channels over a wide voltage range. A similar approach to quantifying unidirectional fluxes could be used in other classes of channels that have a mixed Ca2+ and monovalent permeability, such as nicotinic AChR and cGMP-gated channels (Rogers and Dani, 1995; see Table II in Neher, 1995).
Biionic Flux Ratios in NMDAR but Not in Ca2+-permeable AMPAR Channels Deviate from the Prediction of GHK
The biionic flux ratio in NMDAR channels showed a clear deviation from the predictions of the GHK current equation, requiring a biionic flux-ratio exponent of ∼2 in 10 mM and 1.9 in 1.8 mM [Ca2+]o (Fig. 5). In contrast, AMPAR channels, composed either of GluR-A(Q), -B(Q), or -B(N) subunits, did not show such a deviation, having a biionic flux-ratio exponent of 1 over a wide potential range (Figs. 7 and 9). The most basic conclusion from these results is that the mechanism of Ca2+ transport in NMDAR and AMPAR channels is different, apparently being more complex in NMDAR channels. The results for AMPAR channels also act as an internal control, arguing against calibration or activity problems.
NMDAR channels had a biionic flux-ratio exponent of ∼2. For Ussing type flux-ratio exponents, such behavior has been termed flux coupling and has been interpreted to reflect that the movement of ions across the pore does not occur independently (Hodgkin and Keynes, 1955). Further, for a variety of permeation models, including those based on absolute reaction (Eyring) rate theory, the absolute value of the biionic flux-ratio exponent reflects the minimum number of ions that can occupy the permeation pathway (see references in Hille, 1992; Stampe and Begenisich, 1996). At present, the interpretation of the biionic flux-ratio exponent is ambiguous. In part, this ambiguity arises because we used different ions on the opposite sides of the membrane, and whether these ions interact with the same or different sites in the pore is unknown. Nevertheless, although the mechanistic interpretation of the biionic flux-ratio exponent remains limited, this approach to quantifying biionic flux ratios, in combination with mutagenesis of functional residues, will provide insights into how ions interact with the pore over a wide voltage range.
In AMPAR channels, which have a smaller fractional Ca2+ current (Table I), the biionic flux-ratio exponent was ∼1, whereas in NMDAR channels, which have a nearly fourfold higher fractional Ca2+ current, it was ∼2, suggesting that this biionic flux-ratio exponent in NMDAR channels reflects the same properties of the channel that underlie their higher fractional Ca2+ currents. Recently, analyzing the block of wild-type and mutant NMDAR channels by extracellular Ca2+, Premkumar and Auerbach (1996) identified a high affinity site for Ca2+ in the outer vestibule. The molecular identity of this external site remains as yet unidentified. Nevertheless, it senses little of the transmembrane electric field and is distinct from the N-site asparagines, which are positioned at or near the channel's narrow constriction (Wollmuth et al., 1996; see also Sharma and Stevens, 1996), a structure located ∼50–60% across the transmembrane electric field (Villarroel et al., 1995; Zarei and Dani, 1995). Thus, the deviation from GHK in NMDAR channels may reflect that Ca2+ interacts with multiple sites in the pore: the N-site asparagines positioned at the channel's narrow constriction and an additional site putatively positioned externally, probably <10% across the field. Identifying this putative external site will allow a direct test of the relationship between this external site, fractional Ca2+ currents, and biionic flux ratios.
The GHK current equation is derived assuming independence of ion movement and a constant transmembrane electric field. A biionic flux-ratio exponent of 1 is consistent with the predictions of GHK current equation, but does not necessarily indicate that the ions in the pore follow the assumptions of this test since, even in the simpler case of the Ussing flux-ratio exponent, it can also arise in single ion pores (Hille, 1992), when the transmembrane electric field is not constant (Chen and Eisenberg, 1993), or in multi-ion pores with distinct energetic profiles (Begenisich and Busath, 1981). Indeed, AMPAR channels do show deviations from GHK (Burnashev et al., 1995) and ion–ion interactions in the pore (Bahring et al., 1997). Nevertheless, a possible explanation for the biionic flux-ratio exponent of 1 in AMPAR channels is that these channels lack the putative external site present in NMDAR channels and that the mechanism of Ca2+ transport in them is dominated by a single site, possibly the amino acid side chains occupying the Q/R site.
Differences in Ca2+ Permeability between GluR Subtypes Are Not Due to the Q/R/N Site
The amino acid residue occupying the Q/R/N site is an important determinant of Ca2+ transport in both NMDAR and AMPAR channels. This site, however, is not responsible for the quantitative difference in Ca2+ transport between these GluR subtypes since fractional Ca2+ currents in GluR-B(N) channels were comparable with those in GluR-B(Q) rather than in NMDAR channels (Table I). The reverse substitution in NMDAR channels (glutamine at the N-site) does lead to reduced fractional Ca2+ currents (L.P. Wollmuth, unpublished data). However, the interpretation of this mutant channel in terms of its permeation properties is complex because of multiple subconductance states with different permeation properties (Schneggenburger and Ascher, 1997). Also, these mutant NMDAR channels are more strongly blocked by extracellular Ca2+ than wild-type NMDAR channels (Premkumar and Auerbach, 1996), an effect opposite that seen in AMPAR channels.
NMDAR Channels Have Concentration-dependent Permeability Ratios
In NMDAR channels, PCa/PCs showed a maximum between 1 and 1.8 mM Ca2+ and was reduced at both higher and lower concentrations (Fig. 8, A and B). These concentration-dependent permeability ratios are presumably a manifestation of the same properties of the channel that underlie the deviation of the biionic flux ratios from GHK. If this is true, then surface charges do not seem to play a prominent role in Ca2+ transport in NMDAR channels since reversal potentials showed a downward deflection at very low concentrations in contrast to an upward deflection expected for surface charges. This conclusion is consistent with that of Zarei and Dani (1994).
Previous measurements of the concentration dependence of reversal potentials in GluR channels show similarities and differences from our measurements. Reversal potentials in native NMDAR and Ca2+-permeable AMPAR channels have been measured using a pure Ca2+ extracellular solution similar to what we used (Iino et al., 1990; Koh et al., 1995b). For NMDAR channels, the reversal potentials were indistinguishable from our measurements and showed a similar deviation from GHK. The main difference from our work is that we tested a much wider concentration range of Ca2+, where the deviation from GHK was more pronounced. For native AMPAR channels, it is difficult to quantitatively compare the results since a small amount of edited GluR-B(R) may be present, shifting negative the reversal potentials. Nevertheless, the relationship between measured reversal potentials and the fitted GHK equation shows less deviation than in NMDAR channels. The deviation disappears when PCa/PCs is corrected for the small NMDG permeability present in AMPAR channels (Burnashev et al., 1996) (compare Fig. 8, C and D).
Another approach to test for concentration dependence of Ca2+ permeability has been to record changes in reversal potentials on switching from a high extracellular monovalent solution with no added Ca2+ to the same solution with added Ca2+ (Jahr and Stevens, 1993; Zarei and Dani, 1994; Schneggenburger, 1996; L.P. Wollmuth, unpublished data). In these instances, any deviation from GHK is negligible. In addition, many channel types that contain multiple permeant ions show a paradoxical behavior in mixtures of permeant ions, where the conductance or reversal potential of currents passes through a minimum rather than changing monotonically. This anomalous mole fraction dependence is a property of channel models with more than one ion at a time in the single-file region of the pore (Hille, 1992). Neither NMDAR (Zarei and Dani, 1994) nor kainate receptor (Gu and Huang, 1991) channels show this effect (but see Mayer and Westbrook, 1987). The result for kainate receptor channels is in agreement with our results for AMPAR channels. However, the basis for the lack of a clear anomalous mole fraction effect in NMDAR channels is unknown, but its absence suggests that, while there may be multiple sites in the pore of NMDAR channels for permeating ions, these sites are not multiply occupied.
Activation of NMDAR and Ca2+-permeable AMPAR channels mediates post-synaptic Ca2+ influx. Functionally, Ca2+ transport differs in these channels in that it is nearly fourfold higher in NMDAR than in Ca2+-permeable AMPAR channels. This difference arises because the mechanism of Ca2+ transport in these channels is different, possibly reflecting the presence of an additional external site for Ca2+ in NMDAR channels that is absent in AMPAR channels.
We thank Professor P.H. Seeburg for his generous support, Drs. G. Borst, E. von Kitzing, T. Kuner, and C. Beck for their comments on the manuscript, A. Roth for many helpful discussions, Ms. Spiegel and Ms. Dücker for secretarial assistance, and M. Kaiser, S. Grünewald, and U. Warncke for technical assistance.
This work was supported in part by an Alexander von Humboldt Fellowship (L.P. Wollmuth).
Abbreviations used in this paper
Lonnie P. Wollmuth's present address is Department of Neurobiology and Behavior, State University of New York at Stony Brook, Stony Brook, NY 11794-5230.
Address correspondence to Dr. Lonnie P. Wollmuth, Department of Neurobiology and Behavior, State University of New York at Stony Brook, Stony Brook, NY 11794-5230. Fax: 516-632-6661; E-mail: firstname.lastname@example.org