Differences in apparent pore sizes of low and high voltage-activated Ca2+ channels.

Pore size is of considerable interest in voltage-gated Ca(2+) channels because they exemplify a fundamental ability of certain ion channels: to display large pore diameter, but also great selectivity for their ion of choice. We determined the pore size of several voltage-dependent Ca(2+) channels of known molecular composition with large organic cations as probes. T-type channels supported by the Ca(V)3.1, Ca(V)3.2, and Ca(V)3.3 subunits; L-type channels encoded by the Ca(V)1.2, beta(1), and alpha(2)delta(1) subunits; and R-type channels encoded by the Ca(V)2.3 and beta(3) subunits were each studied using a Xenopus oocyte expression system. The weak permeabilities to organic cations were resolved by looking at inward tails generated upon repolarization after a large depolarizing pulse. Large inward NH(4)(+) currents and sizable methylammonium and dimethylammonium currents were observed in all of the channels tested, whereas trimethylammonium permeated only through L- and R-type channels, and tetramethylammonium currents were observed only in L-type channels. Thus, our experiments revealed an unexpected heterogeneity in pore size among different Ca(2+) channels, with L-type channels having the largest pore (effective diameter = 6.2 A), T-type channels having the tiniest pore (effective diameter = 5.1 A), and R-type channels having a pore size intermediate between these extremes. These findings ran counter to first-order expectations for these channels based simply on their degree of selectivity among inorganic cations or on the bulkiness of their acidic side chains at the locus of selectivity.

T-type Ca 2ϩ channels are found in many cell types and are important for cellular functions as diverse as cell proliferation, cardiac pacemaker activity, and rhythmic firing of neural networks (1,2). In contrast to the better studied high voltageactivated (HVA) 1 Ca 2ϩ channels (L-, N-, P/Q-, and R-type), T-type channels activate at relatively negative membrane po-tentials and are often referred to as "low voltage-activated" (LVA). The distinctive permeation properties of T-type Ca 2ϩ channels were critical for their original identification as distinct entities (3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16). In native preparations, T-type channels are equally permeable to Ca 2ϩ and Ba 2ϩ ions, unlike all known HVA channels, for which unitary Ba 2ϩ fluxes are larger than those supported by Ca 2ϩ (17). Despite general agreement about these differences (1), relatively little is known about the underlying basis of ion selectivity and permeation in T-type channels. New impetus for approaching such questions is provided by the molecular cloning of three different pore-forming Ca V subunits with biophysical properties that clearly identify them as T-type channels (18 -20), Ca V 3.1, Ca V 3.2, and Ca V 3.3 (␣ 1G , ␣ 1H , and ␣ 1I , respectively, in the former voltage-dependent calcium channel nomenclature). These subunits form a closely related family with only limited sequence homology to the subfamilies of HVA channels (ϳ28%). In the regions thought to be important for permeation, the Ca V 3 subunits and HVA channel are more similar (40%). The Ca V 3 subunits also have "P-loops," each containing an amino acid with an acidic side chain thought to be important for permeation. However, instead of a set of four glutamates as in HVA channels (EEEE locus), the corresponding amino acids in the three T-type channel ␣ 1 subunits are glutamates in domains I and II and aspartates in domains III and IV, forming a putative EEDD locus (21)(22)(23).
Knowledge about the molecular differences between the pore-forming regions of T-type channels and HVA channels has redirected attention to differences in their permeation properties. Because current models of ion permeation through voltagedependent Ca 2ϩ channels are based almost entirely on studies of L-type channels (24 -26), characterizing permeation and selectivity in various T-type channels offers broader perspectives about mechanisms. A promising start was provided by Talavera et al. (27), who showed that the aspartates at the EEDD locus of a T-type ␣ 1 subunit contribute to its distinctive selectivity properties (see also Ref. 28).
Here, we focus on pore size, a matter of fundamental importance to understanding interactions between the pore-forming groups and permeant ions. Even in the absence of direct evidence about channel structure from x-ray crystallography, valuable information on minimal pore size can be gathered by studies on the permeation of large organic cations (17, 29 -34). Early studies on the frog skeletal L-type Ca 2ϩ channel (35,36) showed that the pore is permeable to tetramethylammonium and thus has a minimal internal diameter of ϳ6 Å. There was also a brief report that T-type channels in neoplastic B-lymphocytes are impermeable to tetramethylammonium (37). By contrast, no systematic study of the pore size of Ca 2ϩ channels of known molecular composition has yet been reported.
Here, we used the Xenopus oocyte expression system to determine the apparent pore size of calcium channels encoded by cloned Ca V subunits. Ca V 1.2 (␣ 1C ), the subject of the vast majority of recent permeation experiments, was a logical starting point for comparison with the three T-type channel subunits, Ca V 3.1, Ca V 3.2, and Ca V 3.3. We also studied the pore size of channels encoded by the Ca V 2.3 subunit (␣ 1E ), which supports the R-type current (38), of interest because this channel resembles T-type channels in permeation and selectivity properties, although its amino acid sequence is much closer to L-than T-type ␣ 1 subunits (39,40).

MATERIALS AND METHODS
Molecular Biology-For RNA preparation, the human Ca V 3.2 (19) and rat Ca V 3.1 subunits (GenBank TM /EBI accession numbers NM_021098 and AF027984, respectively) subcloned in the pGEM-HEA vector were linearized with AflII (18,19), whereas the rat Ca v 3.3 subunit (NM_020084) subcloned in the pSP73 vector (28) and the human Ca V 2.3 (L27745) subunit subcloned in the pHBE vector (41) were digested with EcoRI and HindIII, respectively. All experiments using the Ca V 1.2 subunit (X15539) of L-type voltage-dependent calcium channels were performed using the CARD5 isoform (42), a deletion mutant of the CARD3 isoform subcloned in the pSP72 vector and digested with XbaI. The rat ␤ 1a (43) ␤ 3 (44) accessory subunits (M25817 and NM_012828) were linearized with XbaI and NotI, respectively, whereas the rabbit ␣ 2 ␦ 1a subunit (M21948) (42) was digested with XhoI. Linearized DNAs were cleaned with phenol/chloroform, and RNA was prepared in vitro with commercial kits (Ambion Inc., Austin, TX) using the SP6 polymerase for the ␤ 1a subunit or the T7 polymerase for all other channels.
Ca 2ϩ Channel Expression in Xenopus Oocytes and Electrophysiology-Xenopus laevis females (NASCO, Fort Atkinson, WI) were anesthetized by immersion in 0.2% ice-cold Tricaine solution for 30 -45 min, and then three to four ovarian lobes were removed by abdominal incision, minced into small fragments with sterile scissors, and digested for 1-1.3 h with 1 mg/ml type I collagenase (Invitrogen) in nominally "calcium-free" solution (82.5 mM NaCl, 2.5 mM KCl, 1 mM MgCl 2 , and 5 mM HEPES (pH 7.6) with NaOH). Stage V and VI oocytes were selected under a dissecting microscope and kept overnight in calcium-containing solution (ND-96: 96 mM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , 1 mM MgCl 2 , 5 mM HEPES, and 2.5 mM sodium pyruvate with the addition of 1000 IU of penicillin and 100 IU of streptomycin). The day after surgical isolation, selected oocytes were injected with ϳ50 nl of solution containing 0.1 mg/ml RNA in water using a hydraulic injector (WPI, Sarasota, FL). To reconstitute L-type Ca 2ϩ channels, the Ca V 1.2 subunit was coinjected in equimolar amounts with ␤ 1a and ␣ 2 ␦ 1 as accessory subunits. The human Ca V 2.3 subunit was co-injected with the ␤ 3 accessory subunit, whereas no accessory subunit was added to the T-type channel ␣ subunits, Ca V 3.1, Ca V 3.2, and Ca V 3.3. Injected oocytes were kept in ND-96 solution at 18°C for 2-6 days before being used for electrophysiological recordings.
Whole cell oocyte currents were recorded with the two-microelectrode technique using an oocyte OC-725-A amplifier (Warner Instruments Corp., Hamden, CT) and pClamp Version 6.1 software (Axon Instruments, Inc., Foster City, CA). Currents were elicited using the protocols described under "Results," filtered at 500 Hz, digitized every 250 s, and stored on a Pentium III computer for off-line analysis. All of the currents were leak-subtracted using a P/4 protocol. The experiments were performed at room temperature using a 150-l laminar flow chamber continuously superfused with a gravity-fed multiline perfusion system. Solution was continuously removed from the chamber using a soft paper bridge. Perfusion flow averaged 1.9 ml/min; and as established in test experiments, the entire chamber content was exchanged in ϳ1.3 min after switching from one solution to another.
Monovalent currents were recorded using nominally Ca 2ϩ -free solutions containing 10 mM HEDTA, 10 mM HEPES, 14 mM tetraethylammonium Cl, and the selected monovalent ion at 100 mM. The osmolarity was adjusted to 300 mosM with the addition of sucrose. To prevent the influx of monovalent ions through the nonspecific cationic channels present in the plasma membrane of Xenopus oocytes (45), the oocytes were preincubated for 10 min with the reversible blocker N-phenylanthranilic acid (500 M; RBI, Natick, MA), and this drug was added to all solutions used during the experiments. To minimize changes in the recording conditions, the duration of the experiments was kept short by using a high acquisition frequency (0.6 Hz). In control experiments, no significant decrease in current amplitude was observed in response to repetitive depolarizations at a fixed voltage using this frequency.
Drugs and Chemicals-All chemicals were of analytical grade and were purchased from Sigma. All cations used in this study were chloride salts, with the exceptions of lithium and tetramethylammonium, which were used as hydroxide salts. Statistical Analysis-Statistical analysis was performed using oneway analysis of variance, and the threshold for significance was set at p Ͻ 0.01.

RESULTS
To estimate the minimal dimensions of the permeation pore, we pursued the classical approach of recording the Ca 2ϩ channel currents supported by a series of monovalent cations of increasing size (17,35). The experiments were carried out in buffered Ca 2ϩ -free solutions to avoid the potent blocking effects of Ca 2ϩ ions on monovalent cation currents (35,46). When currents were evoked by a series of increasingly large depolarizing steps, inward currents first grew in size as the channels activated, but became progressively smaller and eventually reversed when the depolarization became strong enough to drive a net positive charge out of the oocytes, presumably supported by K ϩ efflux. Fig. 1A compares currents carried by Li ϩ ions in oocytes expressing the T-type channel subunits, Ca V 3.1, Ca V 3.2, and Ca V 3.3, or the HVA channel subunits, Ca V 1.2 (L-type) and Ca V 2.3 (R-type). As expected, each of these channels became highly permeable to monovalent ions once divalent cations were removed from the bath, and large inward currents were recorded with Li ϩ as the external charge carrier (Fig. 1). Despite obvious differences in activation and inactivation kinetics from one channel to the next, Li ϩ current reversed at around ϩ5 mV in both L-and R-type channels and in each of the T-type channel isoforms. This value is in good agreement with previous determinations of Li ϩ current reversal in oocytes expressing Ca V 1.2 (47) and leads to an estimated P Li /P K of 1.46 using the Hodgkin-Goldman-Katz formalism for bi-ionic permeation. In calculating this permeability ratio, we assumed that the intracellular potassium concentration in Xenopus oocytes is 120 mM (48) and does not change significantly upon recording with KCl-filled pipettes because of the large volume of the oocytes. Pooled results from many oocytes displayed no significant difference in reversal potential (E rev ) among the  Table I).
Ca V 3.2 and Ca V 1.2 Differ in Their Permeability to Organic Cations-To determine whether there is any difference in the pore size of LVA and HVA Ca 2ϩ channels, we compared the relative permeabilities of L-and T-type channels to organic cations (results with R-type channels are deferred to the end of "Results"). Ammonium and its di-, tri-, and tetramethyl substituents were chosen for two main reasons (49). First of all, given their pK a values, all of these compounds are Ͼ95% ionized at pH 7.4, allowing recordings at physiological pH without complications due to pH-induced changes in the channel itself. Second and more importantly, these compounds are expected to offer reliable reagents to determine channel pore size because the progressive addition of methyl groups produces a gradual increase in the size of these molecules without introducing gross modifications in their overall three-dimensional structure. For reference (49), the maximal diameters and the volumes of these ions, respectively, are as follows: NH 4 ϩ , 3.6 Å and 20.1 Å 3 ; methylammonium (MA), 3.8 Å and 35.1 Å 3 ; dimethylammonium (DMA), 4.6 Å and 49.7 Å 3 ; trimethylammonium (TriMA), 6.0 Å and 64.2 Å 3 ; and tetramethylammonium (TMA), 6.0 Å and 78.8 Å 3 . Interesting differences were found in Ca V 1.2and Ca V 3.2-expressing oocytes when these ions were used as charge carriers for inward currents evoked by depolarizing steps (Fig. 2). NH 4 ϩ and MA clearly permeated through both types of channels, but only Ca V 1.2-expressing oocytes displayed clear inward DMA currents. This provided an initial clue that channels supported by Ca V 3.2 and Ca V 1.2 differ in their permeability.
We looked more closely at the currents with TriMA and TMA as external cations to develop a more quantitative picture of the cutoff sizes of T-and L-type channels. The failure of these ions to support significant inward currents during step depolarizations must be cautiously interpreted because of the possibility that T-and L-type channels are permeable to these cations, but not detectably so at voltage levels sufficiently depolarized to activate the various Ca 2ϩ channels. As an indication that this was of concern for Ca V 1.2, both TriMA and TMA supported clear inward current tails after the depolarizing pulses upon repolarization to Ϫ80 mV, where the driving force was larger (Fig. 2, lower row). Accordingly, to improve the resolution of weak permeabilities, we carried out additional experiments to look systematically at inward current tails over a wide range of potentials. This approach is illustrated in Fig.  3, with tail currents recorded in the presence of 100 mM Li ϩ . Ca 2ϩ channels were strongly activated by application of a 5-ms depolarizing pulse to ϩ70 mV, and then the inward driving force was sharply increased by sudden repolarizations to a range of voltage levels to generate inward tail currents through the open channels. The large tail currents that resulted were outward at repolarization levels at or positive to ϩ10 mV, but increasingly negative at levels at or below 0 mV. The values of reversal potential determined with the tail protocol agreed with E rev values obtained with step depolarizations to within 2-3 mV (Table I). In no case were the differences significant, validating the tail current procedure.
Next, we proceeded to use the same tail current protocol to explore the permeation of organic cations through the various voltage-dependent Ca 2ϩ channels, focusing first on the Ca V 1.2 and Ca V 3.2 subunits. In oocytes expressing either of these subunits, very large inward NH 4 ϩ currents and sizable MA and DMA currents were observed ( Fig. 4 and Table II). The contrasts became more dramatic in considering even larger organic cations. In Ca V 3.2-expressing oocytes, no inward tails were detected with either external TriMA or TMA. In contrast, in Ca V 1.2-expressing oocytes, TriMA carried small but unmistakable inward currents that reversed around Ϫ30 mV, and TMA supported extremely small tails negative to Ϫ60 mV (Fig. 4).
Measurements of reversal potential provided a more quantitative assessment of the differences between channels. For NH 4 ϩ , no significant difference was observed in E rev between Ca V 1.2 and Ca V 3.2 (Table II), suggesting that pore size was not critical in allowing permeation of these small ions. In the case of MA and DMA, E rev was ϳ20 mV more negative for Ca V 3.2 than for Ca V 1.2. This difference corresponds to a lower relative permeability of the T-type channels for the intermediatesized ions.
Ca V 3.1 and Ca V 3.3 Show Organic Cation Permeability Similar to That of Ca V 3.2-All of the data reported in above point to the conclusion that the channel pore is smaller in Ca V 3.2 than in Ca V 1.2. To establish whether a smaller pore size is a general feature of T-type channels, we extended our studies to the two other members of the T-type channel family, Ca V 3.1 and Ca V 3.3, using the tail protocol to study their permeability to NH 4 ϩ and its methyl-substituted derivatives (Fig. 5). Despite clear differences in the kinetic properties of tail currents carried by these two channels, their permeability to organic cations was remarkably similar. Very large tails were observed when ammonium was the charge carrier, and these currents reversed at approximately ϩ20 mV, as in the case of Ca V 3.2 and Ca V 1.2. Clear tails were observed with MA and DMA, and, once again, the reversal potentials for these currents were  remarkably similar to that observed in Ca V 3.2-expressing oocytes, ranging at around Ϫ20 mV for MA and Ϫ60 mV for DMA (Table II). No clear inward tails were observed with TriMA and TMA as charge carriers. Indeed, with either of these external cations, outward currents flowed through the channel at potentials as negative as Ϫ80 mV. All of these data support the conclusion that T-type channels constitute a homogeneous family with regard to pore size, significantly different from L-type channels.
Unusual Permeation Properties of Ca V 2.3-As mentioned in the Introduction, a large body of experimental evidence suggests that Ca V 2.3-encoded channels may differ from the other members of the HVA channel subfamily in displaying certain permeation properties typical of T-type channels. However, no information is available on the pore size of channels supported by Ca V 2.3. To explore this, we used the same approach that allowed us to establish a small pore size in T-type channels (Fig. 6). In Ca V 2.3-expressing oocytes, large inward tail currents were observed when Li ϩ or NH 4 ϩ was present as the charge carrier. The tail currents were smaller with external MA and DMA and very small but clearly detectable with the larger cation TriMA. No inward tail currents whatsoever were observed with TMA as the external cation. The reversal potential for Li ϩ and NH 4 ϩ was not significantly different in Ca V 2.3and Ca V 1.2-expressing oocytes. By contrast, MA, DMA, and TriMA currents reversed at significantly more negative potentials in Ca V 2.3-expressing oocytes than in Ca V 1.2-expressing oocytes. Thus, there were quantitative differences in permeation properties between channels supported by Ca V 2.3 and the best studied exemplar of the HVA subfamily.
Quantitative Comparison of Permeabilities of the Various Ca V Subunits-To quantify and summarize differences in the pore size of various voltage-dependent Ca 2ϩ channels, we used the reversal potential measurements to calculate relative permeability ratios for each of the organic cations using the reversal potentials obtained in the presence of Li ϩ as a reference (see "Materials and Methods"). The implicit presumption here is that whichever cation is present in the extracellular bath, the composition of relevant charge carriers in the intracellular compartment is similar and constant, a very reasonable assumption because the volume of the oocytes is so large. Using the data shown in Figs. 3-6 and Table II, we calculated the permeability ratios for all of the cations of the NH 4 ϩ methylsubstituted series for each of the ␣ 1 subunits and plotted the values against the estimated cation mean diameter (Fig. 7).
There is close agreement between the permeability characteristics of each of the three T-type channel subunits. Using a conventional theoretical analysis (Fig. 7 legend), the cutoff diameters were 5.08 Å for Ca V 3.1, 5.10 Å for Ca V 3.2, and 5.05 Å for Ca V 3.3. Their consistent behavior stands in clear contrast to that of Ca V 1.2, as assessed by clear differences in the relative permeability to individual permeant ion species (MA or DMA)

FIG. 3. Tail currents carried by Li ؉ in HVA and LVA channels.
A, representative traces of tail currents recorded in oocytes expressing the L-type channel Ca V 1.2 or the T-type channels, Ca V 3.1, Ca V 3.2, and Ca V 3.3, using Li ϩ as a carrier ion and the voltage protocol depicted in the insets. In particular, tail currents were elicited by increasing the membrane potential from a holding potential of Ϫ80 mV to ϩ70 mV for 5 ms to maximally open the voltage-dependent calcium channels and by stepping down the potential to values increasing from Ϫ30 to ϩ30 mV in repetitive episodes (10-mV increments, 50-ms pulse duration, 0.6-Hz frequency). The horizontal bar is set at 15 ms, whereas the vertical bar indicates 0.5 A. B, current-voltage plot derived from the data obtained in Ca V 1.2 (E), Ca V 3.1(‚), Ca V 3.2 (ƒ), and Ca V 3.3 (Ⅺ) using this tail protocol. To make the comparison among different channel types easier, current measurements were normalized to the value obtained when the potential was stepped down to Ϫ20 mV. Each point represents the mean Ϯ S.E. of the data obtained in at least five different oocytes.

FIG. 4. Tail currents carried by NH 4 ؉ and its methyl-substituted derivatives in Ca V 1.2 and Ca V 3.2. A, representative traces of tail currents carried by NH 4
ϩ , MA, DMA, TriMA, and TMA as indicated in oocytes expressing the Ca V 3.2 LVA channels (upper row) and the Ca V 1.2 HVA channels (lower row). Tail currents were elicited by increasing the membrane potential from a holding potential of Ϫ80 mV to ϩ70 mV for 5 ms to maximally open the voltage-dependent calcium channels and by stepping down the potential to values increasing from Ϫ30 to ϩ30 mV in repetitive episodes (10-mV increments, 0.6-Hz frequency). B, current-voltage plots derived from the data obtained in Ca V 3.2-expressing (left) and Ca V 1.2-expressing (right) oocytes using this tail protocol. The currents were normalized as the percentage of the tail current recorded at Ϫ20 mV when each oocyte was switched to a 100 mM Li ϩ solution. Each point represents the mean Ϯ S.E. of the data obtained in at least five different oocytes. The vertical bar indicates 2 A in the case of Ca V 3.2 and 0.5 A in the case of Ca V 1.2. or by differences in the cutoff size: the estimated cutoff diameter for Ca V 1.2 was 6.16 A. The plot of relative permeability also shows that Ca V 2.3 occupies an intermediate position between L-and T-type channels, inasmuch as the relative permeability to DMA, TriMA, and TMA is larger in Ca V 2.3 than in T-type channels, but smaller than in L-type channels. Although the data for Ca V 2.3 are not readily fitted by the same theoretical function as used for the other channel types, the intermediate nature of the behavior is clear.

T-type Channels Differ from L-type Channels in Pore Size-
This study demonstrates a striking heterogeneity between various subfamilies of voltage-gated Ca 2ϩ channel with respect to their minimal pore size, estimated by passage of large organic cations in the absence of external Ca 2ϩ . Within the T-type channel subfamily, each of the pore-forming subunits (Ca V 3.1, Ca V 3.2, and Ca V 3.3) showed the same orderly behavior, a monotonically negative relationship between the size of organic cations and their relative permeability. The effective diameter of the permeation pathways of Ca V 3.1, Ca V 3.2, and Ca V 3.3 was estimated as 5.1 Å based on the ability of these subunits to support the permeation of DMA, but not TriMA. This stands in contrast with 6.2 Å, the estimated minimal pore size of Ca V 1.2, an L-type channel subunit originally derived from cardiac muscle. The very small amplitude of the currents carried by Ca V 1.2 when expressed without any accessory subunit precluded the analysis of the pore size of Ca V 1.2 in isolation, but our results for Ca V 1.2 coexpressed in oocytes with the ␤ 1b and ␣ 2 ␦ 1a accessory subunits are in good agreement with earlier estimates for L-type channels in situ both in cardiac myocytes (Ca V 1.2 in a native environment) (50) and in skeletal muscle fibers (Ca V 1.1) (35,36). Thus, although the properties within each of the subfamilies appear homogeneous, T-and L-type channel classes show consistent differences in pore dimensions.
It was unexpected to find that T-type channels are smaller in their minimal pore than L-type channels, given the known properties of T-type channels: they do not show strong selectivity for Ca 2ϩ relative to Ba 2ϩ ; their selectivity for divalent cations over monovalent cations is weaker than for HVA channels; and their selectivity filter includes two aspartate residues, whose side chains are less bulky than the corresponding glutamates in HVA channels. If anything, these features would lead to the hypothesis that T-type channels possess larger diameter pores than L-type channels. Likewise, the logical expectation for R-type channels was that their pore diameter would be larger than that of other HVA channels (because  3-6. Only three points comprising the closest to 0 nA and the ones immediately before and immediately after were used to calculate the fit. Permeability ratios (P r ) were calculated using equation (1) and the experimental approach described in the legend to Fig. 7

FIG. 5. Tail currents carried by NH 4 ؉ and its methyl-substituted derivatives in Ca V 3.1 and Ca V 3.3. A, representative traces of tail currents carried by NH 4
ϩ , MA, DMA, TriMA, and TMA in oocytes expressing Ca V 3.1 (upper row) or Ca V 3.3 (lower row). Tail currents were elicited using the protocol described in the legend to Fig. 4 and are depicted in the insets. Each set of traces was obtained in a different oocyte and is representative of the behavior of a group of at least six different cells. The vertical bar is set at 1 A for Ca V 3.1 and at 0.5 A for Ca V 3.3. B, current-voltage plot derived from the data obtained using this tail protocol in Ca V 3.1-and Ca V 3.2-expressing oocytes for ammonium and each of the cations of its methyl-substituted series. Using the same approach described in the legend to Fig. 4, the currents were normalized to Li ϩ currents measured at Ϫ20 mV. Each point represents the mean Ϯ S.E. of the data obtained in at least five different oocytes. they show little discrimination among divalent cations) or at least equal in size (because Ca V 2.3 is highly homologous to other HVA ␣ 1 subunits). Once again, our observations of a smaller cutoff size for organic cation permeation through R-type channels ran counter to what might have been expected a priori, although they do reinforce previous evidence that R-type channels differ significantly from other HVA channels in their selectivity properties but share some characteristics of T-type channels (39).
Possible Determinants of Pore Size-We have considered various explanations for the differences in pore size (see also Ref. 27). One possibility is that the minimal pore diameter is dependent on residues at the EEDD locus itself. This would follow the precedent of the selectivity filter of sodium channels, where alterations of individual amino acid side chains of the DEKA locus have a strong influence on pore size (49). At the EEEE locus of L-type calcium channels (51), like the DEKA locus of sodium channels (52)(53)(54), side chains protrude into the lumen of the pore. In the absence of divalent cations, negatively charged carboxylate groups would separate from each other because of electrostatic repulsion, possibly finding favorable interactions with other structural components in the lining of the pore. Because aspartate side chains are one methylene group shorter than glutamate side chains, it is conceivable that the aspartates might be less favorably stabilized in a configuration that allows passage of large organic cations.
A second possibility is that the minimal pore diameter might be determined by residues neighboring those at the EEEE or EEDD locus in each repeat (termed "position 0"). Based on analysis of Ca V 1.2, side chains at position Ϫ1 also appear to project into the pore lumen (51) and clearly participate in controlling permeation (55). If this is the case for the T-type channels, one may consider the involvement of a conserved lysine at position Ϫ1 in domain III. Once again, the Na ϩ channel provides precedent for narrowing of a pore by a protruding lysine side chain (49). This hypothesis is worth considering for Ca V 3.1, Ca V 3.2, and Ca V 3.3, but it would not provide an explanation for Ca V 2.3, in which the corresponding position is occupied by a glycine, hardly a bulky residue (41,56). Outside of the P-loops, one may consider other regions of the channel that help shape the permeation pathway, including the boundary between the P-loops and S6 (57) and the S6 segment itself (58,59).
A third alternative is that the smaller minimal diameter of T-type channels arises from a different tertiary structure, possibly reflecting an adaptation in the rest of the channel to compensate for the difference between an EEDD locus and an EEEE locus. The size of the ion-binding pocket is critical in determining the Ca 2ϩ selectivity of organic chelators and Ca 2ϩ -binding proteins (60). To maintain a similar configuration of carboxylate oxygen groups in T-and L-type channels and similar Ca 2ϩ -coordinating capabilities, the smaller chain lengths of the aspartates may have required a narrower FIG. 6. Tail currents carried by NH 4 ؉ and its methyl-substituted derivatives in Ca V 2.3. A, tail currents carried in Ca V 2.3 subunit-expressing oocytes by Li ϩ , NH 4 ϩ , and its methyl-substituted derivatives MA, DMA, TriMA, and TMA. Each set of traces was obtained in a different oocyte and is representative of the behavior of a group of at least five different oocytes. The vertical bar is set at 1 A, and the time scale is set at 15 ms. Tail currents were elicited using the same protocol described in the legend to Fig. 4. B, current-voltage plot obtained by normalizing the current to Li ϩ current measured in the same oocyte at Ϫ20 mV as described in the legend to Fig. 4. Each point represents the mean Ϯ S.E. of the data obtained in at least five different oocytes. C, the same current-voltage data as in B on an expanded scale.  (47). Oocytes were bathed with the test cation, and tail currents were obtained; then, the oocytes were washed with Li ϩ for 2 min, and another tail protocol was run. Permeability ratios (P r ) were calculated using E rev values and the following equation (35): spacing between the opposing pore walls. If this were the case, it would not be a coincidence that the ϳ1.2-Å difference in chain length between glutamate and aspartate agrees closely with the difference in pore diameters. Although this line of thinking is both simplistic and speculative, it illustrates how differences in pore size might have physiological significance for divalent cation selectivity in various types of Ca 2ϩ channel.
Structural Clues Set Limits on Possible Models of Ca 2ϩ Channel Permeation-Our results establish that a relatively large pore size is a general feature of voltage-gated Ca 2ϩ channels. Even including the T-type subfamily, the estimated minimal diameter of any of the Ca 2ϩ channels is still much greater than the 1.99-Å diameter of the dehydrated Ca 2ϩ ion. It is notable that this holds for T-type channels as well as L-type channels (35,36,50) because T-type Ca 2ϩ channels, like HVA channels, are highly selective for their ion of choice under physiological conditions. Taken together, these findings are relevant to current efforts at approaching Ca 2ϩ channel permeation with experiments (55,61,62) and theoretical calculations (63)(64)(65)(66)(67).
In the absence of crystal structures or any other direct structural information about Ca 2ϩ channel pores, it is natural to ask whether Ca 2ϩ selectivity and permeation could be explained by analogy to current thinking about K ϩ channels using, as a springboard, the elegant molecular structure of the bacterial KcsA K ϩ channel (68). That structure has directly supported earlier hypotheses that selectivity for K ϩ ions is generated by a narrow passageway lined with oxygen groups acting as ligands for the dehydrated ion (30,69,70). Turning back to Ca 2ϩ channels, the minimal diameter of the pore is much too large to allow a snug fit with a completely dehydrated Ca 2ϩ ion, even in the case of T-type channels. This precludes a strict extrapolation from K ϩ and KcsA channels to Ca 2ϩ and Ca 2ϩ channels. In considering how a rigid cage and a snug fit might hold for Ca 2ϩ channels, McCleskey and Almers (35) suggested that the discrepancy between the minimal pore diameter of an L-type channel and the diameter of an unhydrated Ca 2ϩ ion might be made up by a closely associated water molecule. This hypothesis was reminiscent of classical proposals that Na ϩ ions pass through the sodium channel selectivity filter along with a single H 2 O (17,29,49) and still remains worthy of consideration for L-type channels. However, the same combination of Ca 2ϩ ion ϩ water molecule would exceed the narrower minimal aperture of T-type channels. As an alternative to such scenarios, we suggest that, across the spectrum of Ca 2ϩ channels, Ca 2ϩ permeation can be accommodated within the prevailing model for L-type channels (21). A generalized model might propose the following: 1) that the ion permeation pathway in all Ca 2ϩ channels is generally much wider than the minimal pore diameter; 2) that acidic side chains, be they glutamates or aspartates, protrude into the pore and provide a flexible complex to closely coordinate one or more Ca 2ϩ ions when these are present; and 3) that, in the absence of permeant divalent ions, the acidic side chains swing partly out of the way, allowing passage of large organic monovalent cations to varying degrees.