The selectivity filter of voltage-gated Ca2+ channels is in part composed of four Glu residues, termed the EEEE locus. Ion selectivity in Ca2+ channels is based on interactions between permeant ions and the EEEE locus: in a mixture of ions, all of which can pass through the pore when present alone, those ions that bind weakly are impermeant, those that bind more strongly are permeant, and those that bind more strongly yet act as pore blockers as a consequence of their low rate of unbinding from the EEEE locus. Thus, competition among ion species is a determining feature of selectivity filter function in Ca2+ channels. Previous work has shown that Asp and Ala substitutions in the EEEE locus reduce ion selectivity by weakening ion binding affinity. Here we describe for wild-type and EEEE locus mutants an analysis at the single channel level of competition between Cd2+, which binds very tightly within the EEEE locus, and Ba2+ or Li+, which bind less tightly and hence exhibit high flux rates: Cd2+ binds to the EEEE locus ∼104× more tightly than does Ba2+, and ∼108× more tightly than does Li+. For wild-type channels, Cd2+ entry into the EEEE locus was 400× faster when Li+ rather than Ba2+ was the current carrier, reflecting the large difference between Ba2+ and Li+ in affinity for the EEEE locus. For the substitution mutants, analysis of Cd2+ block kinetics shows that their weakened ion binding affinity can result from either a reduction in blocker on rate or an enhancement of blocker off rate. Which of these rate effects underlay weakened binding was not specified by the nature of the mutation (Asp vs. Ala), but was instead determined by the valence and affinity of the current-carrying ion (Ba2+ vs. Li+). The dependence of Cd2+ block kinetics upon properties of the current-carrying ion can be understood by considering the number of EEEE locus oxygen atoms available to interact with the different ion pairs.
Introduction
How an ion channel can be highly discriminating in its choice of permeant ion while also supporting very rapid flux of that same ion species is a question that has attracted scrutiny for more than three decades (Hille 1992). Understanding this achievement by K+ and Ca2+ channels, which exhibit error rates of only one ion per several thousand transported during a typical 1-ms opening, has been especially interesting because both of these classes of channels are commonly thought to achieve their unusually high degree of selectivity via binding of the preferred ion (Hodgkin and Keynes 1955; Hille and Schwarz 1978; Armstrong and Taylor 1980; Almers and McCleskey 1984; Hess and Tsien 1984; Hess et al. 1986; Lansman et al. 1986; Neyton and Miller 1988a,Neyton and Miller 1988b; Miller 1999). It has been known for many years that K+ channels are sized to fit bare K+ ions (2.66 Å radius) and cannot fit, in an energetic sense, smaller Na+ ions (1.90 Å) (Bezanilla and Armstrong 1972; Hille 1973). The recently obtained crystal structure of a bacterial K+ channel has beautifully confirmed this view (Doyle et al. 1998).
In contrast to most K+ channels (but see Armstrong and Miller 1990; Korn and Ikeda 1995), Ca2+ channels readily transport ions other than Ca2+ when the preferred ion, Ca2+ (1.98 Å diameter), is removed from the bathing medium. From steric considerations alone, this can occur because the Ca2+ channel pore has a relatively wide diameter of 6 Å (McCleskey and Almers 1985), which allows ions such as Na+ to pass through unhindered even in a partly hydrated state. Under physiological conditions of approximately millimolar extracellular Ca2+, Ca2+ channels selectively transport Ca2+ and not Na+ because Ca2+ binding within the single-file pore blocks Na+ flux. As noted by Bezanilla and Armstrong 1972, pore binding of a preferred permeant ion, though advantageous for selectivity, retards movement of the preferred ion through the pore. Ca2+ channels circumvent this problem by allowing interaction between ions within the pore, in such a manner that Na+ ions are ineffective at dislodging a bound Ca2+ ion (thus block and selectivity), but entry of another Ca2+ ion increases the exit rate of the bound Ca2+ ion 20,000-fold (thus high unitary current) (Almers and McCleskey 1984; Hess and Tsien 1984; Tsien et al. 1987; Yue and Marban 1990). These multi-ion phenomena were originally described using models that incorporated two high-affinity binding sites for Ca2+, but other models have been put forward as well, including models with a single high-affinity site (Armstrong and Neyton 1991; Dang and McCleskey 1998) or even no formal binding site (Nonner and Eisenberg 1998; Nonner et al. 1999). As presently constituted, this latter no-site model is unsuccessful in accounting for some key aspects of the experimentally observed behavior of Ca2+ channels, however (McCleskey 1999).
In its details, the original two-site model also does not provide an accurate description of selective ion transport in Ca2+ channels. Most significantly, site-directed mutagenesis work has revealed not two high-affinity Ca2+ binding sites, but instead a locus of four pore-lining glutamate (E) residues that collectively can bind a single Ca2+ ion with high affinity or multiple Ca2+ ions with low affinity (Kim et al. 1993; Mikala et al. 1993; Tang et al. 1993; Yang et al. 1993; Yatani et al. 1994). The effects on Ca2+ binding affinity of mutations in the glutamate locus (EEEE locus) have been investigated for Ca2+ entering the pore from either side of the membrane, and this work has shown that the EEEE locus is the only high-affinity Ca2+ binding structure in the pore of Ca2+ channels (Yang et al. 1993; Ellinor et al. 1995; Cibulsky and Sather 2000). Although each of the four glutamates interacts with Ca2+, they do so with dissimilar strengths of interaction (Mikala et al. 1993; Yang et al. 1993; Ellinor et al. 1995; Parent and Gopalakrishnan 1995). Thus, the degree to which Ca2+ binding was reduced depended upon which glutamate residue was replaced. An attractive rationale for this is that the different glutamates carry out distinct roles in interacting with permeant ions entering and exiting the EEEE locus, with neighboring, nonglutamate residues involved in tuning the affinity of individual EEEE locus residues. The structural origins of this functional asymmetry within the EEEE locus have only been incompletely identified (Williamson and Sather 1999).
In previous work, the pattern of EEEE locus interactions with mono- and divalent cations also depended upon the chemical nature (aspartate, glutamine, alanine) of the substitutions introduced and upon the specific pair of ions chosen for competition (Ca2+ vs. Li+, Cd2+ vs. Li+, or Cd2+ vs. Ba2+) (Ellinor et al. 1995). Here we describe the effects of amino acid substitutions in the EEEE locus upon the kinetics of Cd2+ block of unitary divalent (Ba2+) and monovalent (Li+) currents carried by α1C Ca2+ channels. The nature of the interaction between EEEE locus glutamate residues and permeant ions was investigated by analyzing the effects of the mutations upon rates of entry and exit for Cd2+, a pore-blocking ion that binds in the EEEE locus with very high affinity.
Based on the fact that aspartate and alanine substitutions in the EEEE locus reduced binding affinity for Cd2+ (Ellinor et al. 1995), one simple expectation for our single-channel block experiments was that Cd2+ off rate would increase in the mutants. A second simple expectation was that in mutants with reduced EEEE locus charge (alanine substitutions), Cd2+ on rate might also be decreased owing to a weaker electrostatic attraction. Neither of these expectations was fulfilled. Instead, we found that when probed with a monovalent current-carrying ion, EEEE locus substitution mutants exhibited retarded rates of blocker entry, but when probed with a divalent current carrier, these same mutants exhibited enhanced rates of blocker exit. These results can be understood by considering the numbers of EEEE locus oxygen atoms available to interact with blocking and current-carrying ions, as developed in a schematic model presented in the discussion.
Materials And Methods
Expression of α1C Ca2+ Channels in Xenopus Oocytes
L-type Ca2+ channels (α1C, cardiac splice variant) having a subunit composition of α1Cβ2bα2δ1a were heterologously expressed in Xenopus laevis oocytes, as described previously (Sather et al. 1993; Ellinor et al. 1995). In a recently proposed systematic nomenclature, this channel is referred to as α11.2a/β2b/α2δ1a (Ertel et al. 2000). In brief, Ca2+ channel cRNAs were synthesized by in vitro transcription using wild-type or mutant versions of the recombinant plasmids pCARDHE (rabbit α1C, gene bank accession number X15539; constructed by subcloning the α1C insert from pCARD3 into a modified version of the pGEM-3Z vector bearing the 5′- and 3′-untranslated regions of the Xenopus β globin gene; Liman et al. 1992; Cibulsky and Sather 2000), pBH17 (β2b, accession number X64298), and pCA1S (α2δ1a, the skeletal muscle subunit described by Mikami et al. 1989, derived from pSPCA1, 3′-truncated, and subcloned into pCDNA3). Oocytes were obtained from female donor frogs anesthetized by ∼30-min immersion in water containing 0.2% tricaine methanesulfonate. Ovarian tissue was removed via an abdominal incision ∼5-mm in length, and individual oocytes were isolated from the ovarian tissue by gentle agitation for 90 min in Ca2+-free OR-2 solution containing (mM): 82.5 NaCl, 2 KCl, 1 MgCl2, 5 HEPES, pH 7.5 with NaOH, containing 2 mg/ml collagenase B (Boehringer). Dissociated oocytes were rinsed in fresh OR-2 and the stage V-VI oocytes were selected by hand. These oocytes were injected with cRNAs encoding α1C (0.3 μg/μl), α2δ1a (0.3 μg/μl), and β2b (0.8 μg/μl) subunits in a 1:1:1 molar ratio, and were maintained in an incubator at 18°C in ND-96 solution containing (mM): 96 NaCl, 2 KCl, 1.8 CaCl2, 1 MgCl2, 5 HEPES, pH 7.6, supplemented with 2.5 mM sodium pyruvate, penicillin (100 U/ml; Sigma-Aldrich), and streptomycin (0.1 mg/ml; Sigma-Aldrich). Oocytes were incubated for 4–12 d before electrophysiological recording.
Single-Channel Recording
The vitelline membrane was manually removed from oocytes expressing α1C Ca2+ channels immediately before patch-clamp experiments, as previously described (Sather et al. 1993). For recording, oocytes were bathed in a high K+ solution to fix the membrane potential at ∼0 mV. The solution contained (mM): 100 KCl, 10 ethyleneglycol-O,O′-bis(2-aminoethyl)-N,N,N ′,N ′-tetraacetic acid (EGTA), 10 HEPES, pH 7.4 with KOH. Cell-attached patch recordings (Hamill et al. 1981) were obtained using borosilicate glass pipets containing either a 110 Ba2+ solution (mM): 110 BaCl2, 10 HEPES, pH 7.4 with tetraethylammonium (TEA)-OH, or a 100 Li+ solution (mM): 100 LiCl, 10 HEPES, 14 TEA-Cl, pH 7.4 with TEA-OH. For solutions used to study Cd2+ block of Li+ current, the Li+ solution was pretreated with Chelex® 100 (Bio-Rad Laboratories) to reduce contamination by divalent metal cations. After removal of the Chelex® 100 beads, solutions with the desired final concentrations of Cd2+ were produced by diluting a 10-mM stock solution of CdCl2 into the Li+ solution that was otherwise nominally free of divalent cations. For solutions used to study Ca2+ block of Li+ current, Li+ solutions were not treated with Chelex® 100 and instead contained 10 mM of the Ca2+ buffer N-(2-hydroxyethyl) ethylenediamine-N,N ′,N ′-triacetic acid (HEDTA). Li+ solutions with desired free [Ca2+] were obtained by adding calculated amounts of CaCl2 (10 mM stock solution) to the HEDTA-supplemented Li+ solution, and the pH was adjusted to 7.4 with TEA-OH. The amount of CaCl2 added was calculated using the Chelator program (Theo J.M. Schoenmakers, University of Nijmegen, Nijmegen, The Netherlands).
Single-channel currents were recorded with an Axopatch 200B amplifier (Axon Instruments, Inc.). For all recordings, except those of Ba2+ currents carried by the EIIID mutant, records were low-pass filtered at 2 kHz (−3 dB) with an eight-pole Bessel filter (Frequency Devices, Inc.) and sampled at 10 kHz using Pulse software (HEKA, distributed by Instrutech Corp.). For Ba2+ currents carried by the EIIID mutant, data were filtered at 5 kHz, sampled at 25 kHz, and digitally refiltered at 5 kHz. To promote long-lasting channel openings, all experiments were carried out in the presence of 2 μM FPL 64176 (included in the bath solution; RBI). Experiments were carried out at 21–23°C.
Data Analysis
Kinetic analysis of block was carried out using TAC software (Bruxton Corp.) and a 50% threshold criterion. Dwell-time histograms for open and shut states were constructed from idealized records, binned logarithmically (20 bins/decade), and plotted against a square root transformation of the ordinate (number of events/bin). Exponential functions were fit to dwell-time histograms using a maximum likelihood method, and the number of exponential components fit (usually one) was determined using a log likelihood ratio test (Colquhoun and Hawkes 1995). In this display format, peaks in the fitted function correspond to the time constants of the exponential components (Sigworth and Sine 1987). The rise time of our recording system (0.166 ms at 2 kHz filtering) limited resolution of kinetic events to those lasting >0.166 ms in all recordings except for Ba2+ currents carried by EIIID (>0.1 ms); shorter duration events were excluded from the analysis. Artifactual lengthening of the observed open time as a result of missed brief closures was corrected according to the equation τO = τO,obs exp(−Td/τC), where τO is the corrected open-state lifetime, τO,obs is the open-state lifetime obtained from histogram fitting, Td is the system dead time (0.09 ms), and τC is the lifetime of the shortest fitted component in the closed-state histogram (Colquhoun and Hawkes 1995). The complimentary problem of missed brief openings artificially lengthening the measured shut lifetime (τS) was minimized by the use of FPL 64176.
Kuo and Hess 1993a have previously analyzed the error in estimating τO and τS for Ca2+ channels when the effects of missed events are dealt with using the simple correction described above. In their work on divalent cation block of current carried by single L-type Ca2+ channels, the same type of experiment carried out here, simulated single channel currents were filtered, digitally sampled and analyzed as for real data, and the measured τO and τS values obtained in this way were compared with the values of τO and τS that were used to generate the simulated currents. Following this procedure, errors in estimating kon and koff were found to be in the range of 10–30% when using the missed events correction method that we have used (Kuo and Hess 1993a).
Owing to the action of FPL 64176, open times in the absence of blocker were often tens of milliseconds in duration (see control records in Fig. 1A and Fig. B, below). Channel closures in the absence of blocker were also generally of long duration, typically >10 ms. Inclusion of blocker (Cd2+ or Ca2+) in the pipet solution produced brief interruptions (shut times <1 ms) of the FPL 64176–dependent long-duration openings. Shut times longer than 3.0 ms were excluded from the analysis because these mostly represented closures between channel openings rather than block events. Under these conditions, both the open- and the shut-time histograms were in almost all cases well-fit by single exponential functions, with the single component of the shut time largely representing the blocked time, as described below.
The results were analyzed using a simple kinetic scheme (Scheme I) that incorporates gating transitions (k−C, kC) between closed (C) and open (O) states, and block (kon · [X2+]) and unblock (koff) transitions between open and blocked (B) states (Lansman et al. 1986).
The values of kon and koff for divalent cation block were determined from measurements of channel open and shut times over a range of blocking ion concentrations ([X2+]). The block and unblock rates were related to the open and shut time constants (τO, τS) by the following equations: kon · [X2+] + kC = 1/τO and koff + k−C = 1/τS.
The block rate, kon, was obtained as the slope of a linear regression fit to plots of τO−1 versus [Cd2+], and the [Cd2+]-independent closing rate (kC; given by the fit intercept on the ordinate) consequently did not significantly distort estimation of kon. Furthermore, in the absence of a blocking ion, pure closing rates (kC) were obtained and compared with τO−1 values when blocker was included in the pipet. Closing rates for wild-type and mutant channels ranged from 26 to 323 s−1 for Ba2+ currents, and from 122 to 357 s−1 for Li+ currents. When a blocking ion, such as Cd2+, was included in the pipet solution, a combination of channel closing and block transitions were recorded, and the frequency of these combined events is referred to here as the shutting rate. The closing rates listed in Table can be compared with shutting rates that ranged from hundreds to thousands per second, depending upon the concentration of blocker used (see Fig. 2,Fig. 3,Fig. 4,Fig. 5,Fig. 6,Fig. 7). Estimates of kC obtained from the intercept of the fit with the ordinate ([X2+] = 0) were comparable with values of kC obtained in the absence of blocker. Overall, estimated rates of block were little affected by contamination with channel closing events.
The unblock rate (koff) and the opening rate (k−C) were both concentration-independent, however, making unambiguous determination of koff more difficult in principle. Nonetheless, accurate estimation of koff was possible because with increasing [X2+] the number of unblock events increased relative to the [X2+]-independent number of opening events. Thus, at higher [X2+], τS−1 largely reflects koff. How well τS−1 approximated koff can be appreciated by comparing τS−1 values measured over a range of [X2+] (see Fig. 2,Fig. 3,Fig. 4,Fig. 5,Fig. 6,Fig. 7): τS−1 values did not differ as greatly over a range of [X2+] for a given channel construct (wild type or mutant) as they differed between channel constructs. As a quantitative test, we used simulations to estimate the extent to which potential differences in construct gating might contaminate our estimates of channel block kinetics. We used the simple kinetic scheme described above and computer simulations of worst-case scenarios (SIMU, QuB software suite, available online at http://www.qub.buffalo.edu). In one salient example, the very brief duration of Cd2+ block events for EIIID Ba2+ currents necessitated use of higher sampling and filtering frequencies, which had the consequence of increasing the density of resolved gating events as well (Table). We therefore simulated an extreme case wherein EIIID channels were assumed to be identical to WT in block/unblock kinetics, and differed from WT in gating kinetics alone. The simulated data were analyzed in a manner identical to that used for experimental data. The simulations showed that, when a blocker was present, fast gating in EIIID cannot account for the experimentally determined differences between EIIID and WT in open- or shut-time constants. Overall, the simulations indicated that construct-dependent differences in gating kinetics of the magnitude observed here do not significantly interfere with estimation of kon and koff for a blocking ion. This was not unexpected, given that the number of unblock events was much greater than the number of opening events over much of the blocker concentration range employed. A similar conclusion was reached in a previous study of L-type Ca2+ channels, in which it was found that differences in gating kinetics between subconductance states introduced little error into estimates of Ca2+ block kinetics (Cloues and Sather 2000).
In cases where block kinetics were close to the resolution limit of our recording system, we supplemented the half-amplitude threshold analysis described above with a β-distribution analysis, for which we assumed a reversible two-state block model with transition rates of kon · [blocker] and koff (Yellen 1984; Pietrobon et al. 1989; Moss and Moczydlowski 1996). Flickery data that had originally been filtered at 2 or 5 kHz were typically refiltered at 250 Hz using a digital Gaussian filter. The refiltered current amplitudes were normalized (0 ≤ i ≤ 1) by the open channel amplitude, which was measured for each patch as the difference between well-resolved open and closed levels. To account for instrumentation and other noise unrelated to block and unblock events, a Gaussian function fit to the closed-channel noise was convolved with the normalized β distribution, and the resulting noise-broadened convolution was then fit to the normalized distribution of current amplitudes. The normalized β distribution had the form f(i) = [ia−1 · (1 − i) b −1]/B(a,b), where the β function:
was used to normalize the distribution, a = koff · τ, b = kon · [blocker] · τ, τ = 0.228/fc, and fc was the calculated −3-dB frequency for the cascaded filters. The β distribution analysis was carried out using a custom-written MATLAB routine (The MathWorks). β distribution analysis was useful for data obtained at higher blocker concentration, where kon and koff were such that 2 ≲ a,b ≲ 20. This method of analysis was tested with simulated data of known kinetics (SIMU), and the rates extracted via the analysis were within 13% of the true rates. In most cases, rates obtained from β distribution analysis of experimental data were within 15% of values obtained from half-amplitude threshold analysis. β distribution analysis was, however, employed solely as an independent methodological test of the fidelity of the half-amplitude threshold analysis.
Nomenclature for Channel Constructs
In the single letter code, glutamate = E, alanine = A, and aspartate = D. The previously identified four glutamate residues that are essential for normal ion selectivity in wild-type Ca2+ channels are referred to, in ensemble, as the EEEE locus. Individually, they are designated EI (E393), EII (E736), EIII (E1145), and EIV (E1446), with the Roman numeral subscripts indicating the motif of origin of each glutamate residue. For brevity, mutant channels are symbolized, for example, as EIA for the E→A substitution in motif I.
All error bars represent SEM, and all mean values are reported together with their standard errors. Student's t test was used to determine statistical significance.
Results
Effect of the Permeant Ion Species on Entry and Exit Rates of a Pore Blocker: Cd2+
Under physiological conditions, Ca2+ channels selectively transport Ca2+ and not other common cations because Ca2+ binding within the channel's pore blocks their flux. We have exploited this basic phenomenon of block to study, at the unitary current level, interactions between permeant cations within the EEEE locus. For wild-type (WT) and mutant α1C Ca2+ channels, we measured the kinetics of Cd2+ block (on rate, kon) and unblock (off rate, koff) of unitary currents carried by either a divalent cation (Ba2+) or a monovalent cation (Li+). Because α1C Ca2+ channels exhibited multiple conductance levels that differ in their on and off rates for blocking ions (Cloues and Sather 2000), we restricted our analyses of WT and mutant α1C channels to their largest conductance level (Table). Owing to ion-dependent differences in reversal potential and channel gating, the kinetics of Cd2+ block of Ba2+ and Li+ unitary currents could not be resolved at the same voltage, and so current records were collected at 0 mV for Ba2+ and at −100 mV for Li+.
Fig. 1 shows examples of single-channel records carried through WT channels by Ba2+ (A) or Li+ (B). With 0 Cd2+ in the patch pipet solution, openings were often tens of milliseconds in duration. When Cd2+ was included in the Ba2+ or Li+ pipet solutions, discrete block events were detected as brief interruptions of current flow, and the number of block events increased with [Cd2+]. Half-amplitude threshold analysis of open and shut (blocked) time durations yielded rate constants for pore entry and exit by Cd2+. Shutting rates increased linearly with blocker concentration, in accord with the state diagram described in the materials and methods.
Entry of Cd2+ into the pore was much slower when Ba2+, as compared with Li+, was the current-carrying ion (Fig. 2, left). The on rate for Cd2+ block of Ba2+ current was 1.8 × 107 M−1 s−1 (0 mV), whereas the on rate for Cd2+ block of Li+ current was 8.2 × 109 M−1 s−1 (−100 mV), an ∼400-fold difference in kon (Table). On rates for pore blocking ions such as Cd2+ are essentially voltage insensitive (Lansman et al. 1986; Kuo and Hess 1993a,Kuo and Hess 1993b), as confirmed here by the fact that kon values for Cd2+ versus Ba2+ were of similar magnitude at 0 mV (1.8 × 107 M−1 s−1) and −60 mV (1.1 × 107 M−1 s−1; data not shown). In comparing the Ba2+ and Li+ solutions, membrane surface potentials also do not greatly differ (Kuo and Hess 1992). Thus, the large difference in Cd2+ on rate between Ba2+ and Li+ solutions is apparently attributable to differing interactions within the pore between Cd2+ and Ba2+ in one case, and between Cd2+ and Li+ in the second case.
Rates of Cd2+ exit from the pore and into the cytosol were also different between the Ba2+ and Li+ solutions (Fig. 2, right). The τS−1 values measured for the lowest concentrations of Cd2+ tested probably provide the best estimate of the true off rate, but because we were unsure of the scatter in our data, we estimated off rates from the horizontal lines fitted to all of the data. Using this procedure, the Cd2+ off rate was 1,309 s−1 with Ba2+ as the current-carrying ion (0 mV), and 3,591 s−1 with Li+ as the current-carrying ion (−100 mV). However, off rates for pore blockers such as Cd2+ are strongly affected by membrane potential (Lansman et al. 1986; Kuo and Hess 1993a,Kuo and Hess 1993b), so to compare Cd2+ off rates for Ba2+ and Li+ currents, rates were adjusted for the −100-mV difference in potential. Based on an e-fold/−25 mV increase in off rate (Kuo and Hess 1993a,Kuo and Hess 1993b), the Cd2+ off rate at −100 mV with Ba2+ as the current carrier is extrapolated to 71,469 s−1, ∼20× faster than the Cd2+ off rate measured at this potential with Li+ as the current carrier.
As a test for the reliability of the WT rate estimates, particularly for the very fast on rate of Cd2+ in the Li+ solution, an independent means of estimating block and unblock rates was applied. β distribution analysis of the kinetics of Cd2+ block of Li+ current yielded, for 300 nM Cd2+, a block rate of 3,140 ± 1,380 s−1 (n = 3) and an unblock rate of 4,759 ± 1,712 s−1 (n = 3), as compared with threshold analysis values of 1,788 ± 128 s−1 (n = 3) and 3,936 ± 811 s−1 (n = 3), respectively. The rough similarity of the rates obtained by the two different methods of analysis provides confirmation of the results of the half-amplitude threshold analysis.
Effect of EEEE Locus Mutations on Rates of Blocker Entry and Exit: E→D Mutants
Previous work at the whole-cell level showed that EEEE locus mutations reduce pore binding affinity for divalent metal ions such as Ca2+ and Cd2+ in a mutation-specific manner (Ellinor et al. 1995). Here we have investigated the effects of EEEE locus mutations upon rates of entry and exit for a blocking ion. We have also investigated how mutation-specific block kinetics depend upon the current-carrying ion species. The substitutions studied were E→D mutants, which shortened side-chain length by one methylene carbon but preserved carboxylate groups, and E→A mutants, which further shortened side chain length and also eliminated one of the four carboxylate groups.
Fig. 3 shows data obtained for Cd2+ block of unitary Ba2+ currents carried by the E→D mutants. Examples of current records (Fig. 3 A), distributions of open (B) and shut (C) times, and plots of τO−1 (▪) and τS−1 (○) versus [Cd2+] (D) are shown. The on rates for Cd2+ block were little different across the four E→D mutants, and ranged from ∼55% to ∼140% of the on-rate value obtained for WT under these conditions (Table). In contrast, off rates for Cd2+ block of Ba2+ current in E→D mutants were not all of the same general magnitude, and were in some cases distinctly different from WT. The Cd2+ off rates for EIID and EIIID were significantly speeded up by the mutations (WT: 1,309 s−1; EIID: 4,411 s−1; EIIID: 8,242 s−1; P < 0.01) but those for EID and EIVD were similar to WT (EID: 1,391 s−1; EIVD: 1,195 s−1; P > 0.05). The degree of increase in unbinding rate followed the order III > II > IV = I, which is almost identical to the order of the decreases in Cd2+ binding affinity previously reported for these mutants (III > II > I > IV) when carrying Ba2+ currents (Ellinor et al. 1995).
To test the reliability of the fastest estimated rate of Cd2+ unblock, that for EIIID, we again supplemented the half-amplitude threshold analysis with a β distribution analysis. β distribution analysis of the kinetics of Cd2+ block of Ba2+ current yielded, for 100 μM Cd2+, a block rate of 2,448 ± 52 s−1 (n = 3) and an unblock rate of 5,899 ± 142 s−1 (n = 3), as compared with threshold analysis values of 2,169 ± 108 s−1 (n = 3) and 6,404 ± 261 s−1 (n = 3), respectively. Thus, the β distribution analysis confirms the results obtained from the half-amplitude threshold analysis.
Fig. 4 presents data for Cd2+ block of unitary Li+ currents carried by the E→D mutant channels. The records in Fig. 4 A showed clear increases in the number of block events as [Cd2+] was raised, although the size of these increases depended upon the mutation. The most obvious effects of the mutations were profound reductions in Cd2+ on rate (Fig. 4 D; sloped lines, ▪). The EIIID mutant (1.1 × 108 M−1 s−1) was the most different from WT (7.5 ×109 M−1 s−1), exhibiting a 70-fold reduction in Cd2+ entry rate relative to WT. Decreases in Cd2+ entry rate for the other E→D mutants ranged from 10- to 20-fold, with all four mutants following the order III > IV > I > II (Table). Off rates (horizontal lines, ○) varied somewhat relative to WT, ranging from ∼35% to ∼155% of the WT value (Table). For Li+ currents, the general off-rate pattern across the E→D mutants was reminiscent of the off-rate pattern observed with Ba2+, but was less pronounced (illustrated in Fig. 8 A, below).
Effect of EEEE Locus Mutations on Rates of Blocker Entry and Exit: E→A Mutants
The on and off rates obtained for Cd2+ block of unitary Ba2+ and Li+ currents were also compared when individual glutamate residues were replaced with alanine (E→A mutants). As observed with the E→D mutants, the effect of the E→A mutations depended on the nature of the permeant ion. For the E→A mutants, examples of Cd2+ block of unitary Ba2+ current records, exponential fits to the distributions of open and shut times, and block and unblock rate plots are shown in Fig. 5. The amplitude of unitary Ba2+ current through the EIIIA mutant was so small at 0 mV (−0.25 pA) and Cd2+ unblock was so fast that block events could not be adequately resolved, and so Cd2+ block of Ba2+ current was not analyzed for this mutant. As shown in Fig. 5 C, shut-time distributions obtained for EIIA and EIVA were best-fit by two exponential components in some cases, with the small, longer-duration component most likely reflecting incomplete exclusion of long closures.
For the E→A mutations, the primary effects on the kinetics of Cd2+ block of Ba2+ current were increases in the speed of Cd2+ exit, leaving the Cd2+ entry rate little changed (kon increased to ∼140–190% of the WT value; Table). The Cd2+ exit rate was fastest for EIIA, followed by EIVA, and both of these rates were significantly different from WT (Table II; P < 0.01). The rate of Cd2+ exit from EIA was not significantly different from that of WT (Table II; P > 0.01). The fact that Cd2+ exit was unresolvable for EIIIA indicates that Cd2+ exited this mutant even more rapidly than it exited EIIA, an interpretation consistent with the fact that the EIIIA mutation was previously found to have the most severe effects on half-block values among all of the A and D mutants, and for all of the blocker/current-carrier pairs studied (Ellinor et al. 1995). The pattern of the effects of the E→A mutations on Cd2+ exit was similar to that of the E→D mutations, although the size of the increases in koff were relatively larger for the E→A mutations (see Fig. 8 A).
With Li+ as the current carrier, Cd2+ on rates were profoundly altered by E→A mutations and off rates were much less affected. EIIIA was the most severe mutation, with an on rate more than three orders of magnitude slower than WT (2.1 × 106 M−1 s−1 compared with 8.2 × 109 M−1 s−1) (Fig. 6). The reductions in kon followed the order III > II > I > IV (Table). In general, the E→A mutations exhibited greater reductions in kon than did cognate E→D mutations.
With the exception of EIVA, Cd2+ off rates with Li+ as the current carrier were not significantly affected by the E→A mutations (Student's t test, P > 0.05). EIVA, like EIVD, had a faster Cd2+ exit rate than did WT (Table).
Effect of the Permeant Ion Species on Entry and Exit Rates of Another Pore Blocker: Ca2+
The results of the Cd2+ block experiments seemed to indicate that EEEE locus point mutations caused the on rate for blocker to be reduced when a monovalent cation carried the current, but when a divalent cation carried the current, these same mutations caused the off rate for the blocker to be enhanced. To examine the validity of this generalization, we studied Ca2+ block of Li+ current to test whether this monovalent–divalent cation pair would exhibit behavior similar to that of Cd2+ block of Li+ current. We previously analyzed Ca2+ block of Li+ currents carried by the WT channel, obtaining an on rate for Ca2+ of 4.8 × 108 M−1 s−1 and an off rate of 3,582 s−1 at −60 mV (Cloues and Sather 2000). Because comparison of these WT rates to those of severe EEEE locus mutations provided the desired test, we examined Ca2+ block of Li+ current in the EIIIA channel. Discrete Ca2+ block events were not detectable for EIIIA, perhaps because the off rate for Ca2+ at −60 mV was too fast in this mutant, and perhaps also because the on rate for Ca2+ was too slow. Either of these possibilities is consistent with the observed effects of the EIIIA mutation on Cd2+ block of Li+ current (see Fig. 8). Seeking a more definitive answer, we studied the EIIID mutant, in which discrete Ca2+ block events were resolved.
Fig. 7 compares Ca2+ block of Li+ currents for WT and EIIID channels. Examination of the unitary current records in A and B shows that 3 μM Ca2+ produced many more resolved block events in WT channels than did 100 μM Ca2+ in EIIID channels. Fig. 7 C shows that the Ca2+ off rates were not significantly different between EIIID and WT channels (4,748 s−1 versus 3,582 s−1; P > 0.1), just as the Cd2+ off rates were not different between EIIID and WT. On the other hand, the Ca2+ on rate for EIIID was 55-fold slower than for WT (8.8 × 106 M−1 s−1 versus 4.8 × 108 M−1 s−1; −60 mV), a finding consonant with the 68-fold slowing in Cd2+ on rate for this same mutant relative to WT (−100 mV). β distribution analysis of the kinetics of Ca2+ block of EIIID yielded, for 1 mM Ca2+, kon = 5,621 ± 466 s−1 and koff = 5,603 ± 574 s−1 (n = 3), values close to those predicted by the linear fits to the results of the threshold analysis (kon = 6,000 s−1, koff = 4,748 s−1).
Discussion
In the present work on selective permeability in an L-type Ca2+ channel, we studied inter-ion competition for binding to the EEEE locus for ion pairs in which one ion interacted much more strongly with the EEEE locus than did the other ion. For the three pairings studied, relative WT EEEE locus affinities at 0 mV were 8.0 × 10−9 M for Cd2+ vs. Li+, 7.3 × 10−5 M for Cd2+ vs. Ba2+ and 6.8 × 10−7 M for Ca2+ vs. Li+. All of these ions entered the pore from the extracellular side of the channel, and all exited to the intracellular side (Lansman et al. 1986; Kuo and Hess 1993b). A given ion acted as a blocker because its exit rate was low, or it acted as a current carrier if both its exit rate and concentration were high.
The rate of extracellular metal ion entry into the pore of an L-type Ca2+ channel was measured as the rate of binding of a high-affinity divalent cation (Cd2+ or Ca2+) to the WT or mutated EEEE locus. In this single-file, multi-ion pore (Almers and McCleskey 1984; Hess and Tsien 1984), binding of a high-affinity cation obstructed flow of the dominant current-carrying ion (Ba2+or Li+), and the frequency of these discrete block events yielded the rate of blocking ion entry. Similarly, the rate of divalent metal cation departure from the pore and into the intracellular region was determined as the rate of unblock of unitary current. The two principal observations made in this work were (a) that the rate of metal ion entry into the pore of a WT Ca2+ channel depended upon the binding affinity of the ion already occupying the EEEE locus, and (b) that the way in which EEEE locus mutations reduced divalent cation binding affinity depended upon the valence of the current-carrying ion. Specifically, EEEE locus substitutions increased the permeability of Cd2+ when Ba2+ was the dominant current carrier, but the same substitutions decreased the permeability of Cd2+ when Li+ was the current carrier. The interpretation of these observations is considered below.
Influence of the Current-carrying Ion on Cd2+ Entry into WT Channels
In the experiments on WT channels, Cd2+ competed with either Li+ or Ba2+ for entry into the pore. Competing against Li+, the rate of association of Cd2+ with the EEEE locus was 8.2 × 109 M−1 s−1. This on rate is approximately fivefold greater than the estimated diffusion-controlled rate of association of a divalent metal ion with the EEEE locus (Kuo and Hess 1992, Kuo and Hess 1993a), and indicates that Cd2+ binds to the EEEE locus at, or slightly above, the rate calculated from simple diffusional interaction. Remarkably, Cd2+ can apparently displace Li+ from the EEEE locus as fast as Cd2+ can diffuse into the pore. The data do, however, hint at an alternative possibility that rapid Cd2+ displacement of Li+ relies upon more than a diffusive process: taken at face value, the (barely) super-diffusional Cd2+ on rate may be a clue that divalent cations are attracted into the EEEE locus at a rate that is enhanced by electrostatic interaction with the anionic EEEE locus carboxylates (Creighton 1993). This possibility is reminiscent of the case of proton block of L-type Ca2+ channels, in which the proton association rate significantly exceeds the diffusion limit (Prod'hom et al. 1987).
Competing against Ba2+, Cd2+ entered the pore at a rate of 1.8 × 107 M−1 s−1, which was 1/400th the Cd2+ entry rate when Cd2+ competed against Li+. Ba2+ and Li+ differ in charge, ionic radius, water substitution rate, and enthalpy of dehydration, and any of these differences could underlie the distinctly different interactions of these ions with Cd2+ in the EEEE locus. One possibility is that Ba2+ more completely screens EEEE locus charge, so that Cd2+ is not as strongly attracted into a Ba2+-occupied pore as it is attracted into a Li+-occupied pore. However, it is unlikely that the effects of current-carrying ions upon Cd2+ on rate can be accounted for by their charge alone, because the on rate for Ca2+ versus Li+ was also much lower (4.8 × 108 M−1 s−1; Cloues and Sather 2000) than that of Cd2+ versus Li+. The most specific description of these observations is that blocker on rate in WT channels is very strongly influenced by the EEEE locus binding affinity of the current-carrying ion.
Influence of the Current-carrying Ion on Cd2+ Block Kinetics in EEEE Locus Mutants
In accord with previously obtained results at the whole-cell level, the severity of the effect of EEEE locus substitutions on Cd2+ binding affinity (K′d), calculated from koff and kon (Table), depended on the nature and location of the amino acid substitution (Ellinor et al. 1995). Substitution at residue EIII generally had the most profound effect, followed by substitution at EII. Substitutions at residues EI and EIV had much smaller effects in general.
The pattern of substitution-specific effects upon kon and koff is illustrated for Ba2+ and Li+ currents in Fig. 8. This summary shows that blocking ion interaction with the mutated EEEE locus depended not only on the nature and position of the amino acid substitution, but very much on the nature of the current-carrying ion as well. With Ba2+ as the current-carrier, D or A mutations predominantly affected koff for Cd2+, and had little systematic effect on kon. In contrast, with Li+ as the current-carrier, D or A mutations predominantly affected kon for Cd2+, with smaller effects on koff. As for the koff/kon ratio, effects on kon and koff were position specific. For Cd2+ koff values measured versus Ba2+, for example (Fig. 8 A, left), koff for EID and EIVD were not statistically different from the WT value (P > 0.1), koff for EIID was larger than for WT (P < 0.001), EID (P < 0.005), and EIVD (P < 0.002), but smaller than for EIIID (P < 0.01). The effects also differed between the D and A substitutions at any given position, with the A substitutions generally producing larger deviations from WT performance. Thus, the way in which the EEEE locus residues interact with a blocking ion is dependent in part upon the kinds and numbers of interactions made between the EEEE locus and other, current-carrying ions within the pore.
One way to rationalize the effects of EEEE locus mutations on block and unblock kinetics is based on consideration of the number of EEEE locus oxygen (O) atoms available to bind Cd2+, Ba2+, or Li+, an approach that represents an extension of an earlier model of EEEE locus function (Yang et al. 1993). The results of investigation of proton block (Chen et al. 1996; Klockner et al. 1996) and of substituted-cysteine accessibility work (Cibulsky and Sather 2000; Wu et al. 2000) indicate that the oxygen atoms of the EEEE locus glutamates project into the lumen of the pore, where they are presumed to interact with metal ions. In the scheme developed here, individual O atoms bind and unbind metal ions according to thermal motions in the pore and electrostatic interactions with the permeating metal ions, and metal ions attempt to bind several O atoms so that unbinding of any single O does not generally result in complete unbinding of that ion. The approximate numbers of O atoms that could interact with the metal ions is suggested by their preferred ligand coordination numbers of six for Li+ and for Cd2+, seven for Ca2+ and eight for Ba2+ (Falke et al. 1994; Martel and Hancock 1996). We speculate that some O atom(s) must be available to bind and stabilize entering Cd2+, Ba2+, or Li+ in the EEEE locus. In some trials, the entering ion fails to out compete the bound ion(s) for O atoms, and in other trials it succeeds in acquiring several O atoms. The proportion of successful entry events depends upon the relative affinities of the competing ions, so that blocking ions succeed relatively frequently when competing against lower affinity current-carrying ions, whereas current-carrying ions succeed only infrequently in displacing high-affinity blocking ions.
Blocker on rate in WT channels.
Fig. 9 presents a specific depiction of this model of selectivity filter function. The depiction is intended to capture the essence of the model only, so that some of the specifics of the structure (e.g., the precise positioning of carboxylate groups) should be taken in a figurative sense. Based on earlier Ca2+ channel work (Kuo and Hess 1993a,Kuo and Hess 1993b; Yang et al. 1993; Ellinor et al. 1995; Cloues and Sather 2000), and by analogy with other ion channels (reviewed by Khakh and Lester 1999), the selectivity filter is envisioned as a flexible structure that changes conformation as different kinds and numbers of ions pass through it. To emphasize this idea, Fig. 9 illustrates, for Ba2+ versus Li+, differing orientation angles of the glutamate side chains, differing patterns of glutamate interactions with current-carrying ions, and differing numbers of interactions made with entering Cd2+.
For WT channels, the on rate for Cd2+ against Li+ is higher than is its on rate against Ba2+ because single Li+–O binding events are briefer in duration than are single Ba2+–O events. The overall unbinding rate of Li+ is thus faster than that of Ba2+, and this allows relatively faster entry of Cd2+ into the EEEE locus when Li+ is the current carrier. Turning to the difference in on rate between Cd2+ and Ca2+ that was observed when these ions were tested against Li+, again in WT channels, the comparatively slower Ca2+ on rate can be construed as follows: the Ca2+–O binding events are briefer in duration than are the Cd2+–O events, so that in competing against Li+ for binding to EEEE locus O atoms, the probability that a Ca2+ ion will bind enough O atoms to displace Li+ is lower than is the analogous probability for Cd2+. The net effect is that Ca2+ entry into the EEEE locus is slower than is Cd2+ entry.
Cd2+ vs. Ba2+: on rate.
How can kon for Cd2+ versus Ba2+ be essentially unaffected by EEEE locus substitutions? In cases where one of the EEEE locus residues has been replaced with D or A, there may not be enough effectively positioned O atoms remaining to properly bind two Ba2+ ions (Fig. 9, top row). In these cases, we propose that there is a greater probability that only one Ba2+ ion occupies the locus, rather than the two that more probably occupy the WT EEEE locus. With one Ba2+ ion in the locus of such mutants, for example EII substitutions (denoted ExEE), O atoms would be available to help stabilize entering Cd2+. This reasoning predicts that kon for Cd2+ block of Ba2+ flux would be similar between the EEEE channel and ExEE mutants, as observed.
It is interesting that the D and A mutants are very similar to one another in having little impact on kon despite the fact that the A mutants possess two fewer O atoms in the selectivity locus. This suggests that the number of O atoms is not the only critical factor, but rather that the number of O atoms in the proper spatial arrangement is important.
Cd2+ vs. Ba2+: off rate.
Pursuing the question of mutational effects on block kinetics further, how might substitution in the EEEE locus shorten the bound lifetime of a blocking ion? This could be accounted for based on the idea that Cd2+ unbinding involves facilitation of departure by competition with another ion in the EEEE locus (knock-off; Neyton and Miller 1988a,Neyton and Miller 1988b; Kuo and Hess 1993a,Kuo and Hess 1993b), and by making in addition the assumption that, analogous to the case for Cd2+ entry into a Ba2+-occupied locus, Ba2+ can enter the Cd2+-loaded mutant locus as readily as it can enter the Cd2+-loaded WT EEEE locus (Fig. 9, second row). Compared with the case for WT, Cd2+ is not as strongly stabilized in the substitution mutants and so its Ba2+-enhanced off rate is greater in the mutants than in WT.
In examining Fig. 8 A (left), two lesser points emerge. First, there is almost no effect of EI and EIV mutants on Cd2+ koff, indicating either that the loss of O atoms normally present at these positions does not much destabilize Cd2+ binding or that any reduction in Cd2+ binding affinity is compensated by a reduced Ba2+ binding affinity. Second, the effects of the A mutants appear to be slightly larger than those of the D mutants, as anticipated from the fact that A substitution results in loss of two carboxylate O atoms, whereas D substitution preserves carboxylate O atoms but alters the position of two of them.
Cd2+ vs. Li+: on-rate.
How might the effects of mutations on block kinetics be so fundamentally different when Cd2+ must compete with Li+ rather than Ba2+? Consider first Cd2+ on rates, which are much more strongly affected by EEEE locus substitutions when Li+, rather than Ba2+, is the current carrier (Fig. 9, third row). We speculate that the mutant locus, like the WT EEEE locus, is frequently doubly occupied by Li+ ions, because unlike the case for divalent Ba2+ ions, we suppose that there are sufficient O atoms to bind two monovalent Li+ ions even in the mutant channels. However, compared with WT channels, Li+-occupied mutant channels may lack the O atoms needed to stabilize Cd2+ as it attempts to enter the pore. Without this stabilization, Cd2+ returns to the external solution at such a high rate that the ultra-brief block events are not resolved in our recordings, kon is slowed.
Cd2+ vs. Li+: off-rate.
Lastly, Cd2+ off rates measured with Li+ currents were much less affected by mutations than were on rates, and the off-rate changes introduced by mutations were somewhat smaller for Li+ currents than for Ba2+ currents. To account for the near absence of effect of mutations on koff with Li+ currents, suppose that in the mutants there remain sufficient O atoms available to bind Cd2+ tightly, as depicted in Fig. 9 (bottom row). In contrast to the case for Ba2+, Li+ is proposed to have such a comparatively low affinity for the EEEE locus O atoms that it is does not strongly compete against Cd2+ in the mutants, and hence Cd2+ does not experience much change in its complement of liganding O atoms. The result for the mutants is that, as for WT, Li+ only weakly facilitates Cd2+ exit, and thus the rate of Cd2+ exit to the cytosol is little affected by the mutations.
Cd2+ Block Kinetics and Unitary Conductance
Is conduction of current-carrying ions (Ba2+ or Li+) related in any simple way to the on and off rates of a blocking ion (Cd2+)? Fig. 10 shows that in this case it is not: for a narrow range in Cd2+ on rate, unitary Ba2+ flux did not vary in any systematic way over an eightfold range in Cd2+ off rate, nor did unitary Li+ flux vary systematically over a 4.5-fold range in Cd2+ off rate and 3,600-fold range in Cd2+ on rate. The absence of a simple pattern in Fig. 10 indicates that the interactions between selectivity filter O atoms and current-carrying ions that occur during ion conduction must be different from those that occur between selectivity filter O atoms, current-carrying ions, and a blocking ion. In contrast to this conclusion for EEEE locus substitution mutants, we have previously found for subconductance states of the WT channel that the amplitudes of unitary Ca2+ currents carried by the various substates could be predicted from substate-specific on and off rates for Ca2+ block of Li+ current (Cloues and Sather 2000). These divergent conclusions might be reconciled by noting that the structure of the EEEE locus was significantly altered in the D and A substitution mutants, whereas for the WT subconductance states, EEEE locus glutamates were preserved.
The present work therefore provides a description of the way in which EEEE locus substitutions modify the flux of blocking ions, but not the flux of lower affinity, current-carrying ions. Competing against Li+, Cd2+ flux was reduced in EEEE locus mutants because kon for Cd2+ was reduced; competing against Ba2+ in the same mutants, Cd2+ flux was increased because koff for Cd2+ was increased. Development of a more quantitative and complete model describing flux, binding, and block in the EEEE locus of Ca2+ channels awaits the combination of atomic-scale structure information and additional kinetic measurements.
Acknowledgments
We thank Emily Liman for the gift of a vector bearing the 5′- and 3′-untranslated regions from the Xenopus β globin gene, and Tsutomu Tanabe, Veit Flockerzi, and Franz Hofmann for gifts of the α1C, α2δ1a, and β2b subunit cDNAs. We thank Xin-Sheng Wu for participation in some of the experiments, and Joyce Rohan and David J. Gross for help with β-distribution analysis. Software used to simulate single channel kinetics was generously provided by Feng Qin, Anthony Auerbach, and Fred Sachs (SIMU).
This work was supported by a fellowship from the American Heart Association of Colorado and Wyoming CWFW-14-97 (R.K. Cloues), a National Research Service Award from the National Institutes of Health (NIH) MH11717 (S.M. Cibulsky), and NIH grants NS35245 and AG04418 (W.A. Sather).
References
Abbreviation used in this paper: WT, wild type.