In a 1976 reminiscence of his landmark work with Andrew Huxley that elucidated the ionic basis of the action potential, Alan Hodgkin made a puzzling admission:
We spent over two years analysing and writing up the results and I have often been asked why this took so long. The answer, as usual, is multiple . . . although we had obtained much new information, the overall conclusion was basically a disappointment. (Hodgkin, 1976)
Why were Hodgkin and Huxley so disappointed with their work? Through a brilliant train of thought that considered the linearity of the current-voltage curve and the relative amplitudes of ionic and gating currents (Hodgkin and Huxley, 1952), they concluded that the Na+ current which drives the action potential passes through a channel—defined as a pore that passes ions at high rates compared to enzyme or carrier turnover. It is often forgotten that this disproved their working hypothesis. Reasoning that the high selectivity of the molecule responsible for Na+ permeation was inconsistent with high flux, they had predicted a Na+-selective carrier, not a pore (Hodgkin et al., 1949; also, see Fig. 7 in Hodgkin, 1976, from an unpublished manuscript). We now take for granted that pores can bind their preferred ion and still pass millions per second, but, although it may be taken for granted, it is not understood. Bezanilla and Armstrong (1972) succinctly showed that a single binding site in a rigid pore cannot generate selective flux. The problem is clear: if a selectivity site binds ion A far more tightly than ion B, then A exits the pore far slower than B, and A's advantage in pore occupancy is negated. Always precise, Hodgkin and Huxley never once used the word “channel” in their set of papers that serve as the discovery site for Na+ and K+ channels. Concluding that they could not provide a plausible molecular mechanism for the Na+ current, they settled, in Hodgkin's words, “for the more pedestrian aim of finding a simple set of mathematical equations” that described their results. The problem that frustrated them—how ions can pass quickly through sticky pores—gets a new wrinkle in unexpected results reported by Polo-Parada and Korn in this issue of The Journal of General Physiology.
Polo-Parada and Korn address the interaction of Na+ and Ca2+ in voltage-gated Ca2+ channels. The choice between Na+ and Ca2+ made by Ca2+ channels plays an important role in the understanding of ion channel selectivity. The channel must selectively pass Ca2+ rather than Na+ despite the hundred-fold concentration advantage held by Na+. It does spectacularly well, choosing Ca2+ over Na+ at more than a 1,000:1 ratio (Hess et al., 1986); this success was a severe challenge to classic ideas of selectivity. Historically, ion channel selectivity was explained by molecular sieving mechanisms rather than by selective binding. A sieve provides a means of rapidly passing either the smaller or the less hydrated ion, thereby allowing both high flux and selectivity. But no sieve can choose Ca2+ over Na+. The naked ions are identical in size (Pauling, 1960) and, having twice the charge, Ca2+ is the more heavily hydrated. Instead, an intrapore, high affinity Ca2+ binding site underlies Ca2+ channel selectivity. The site has a dissociation constant for Ca2+ of about 1 μM (Kostyuk et al., 1983; Almers et al., 1984). The maximum off-rate from such a site is about 103 ions per second, yet Ca2+ channels pass about 106 ions per second. This paradox was first posed by results showing that micromolar Ca2+ blocked Na+ permeation through Ca2+ channels. Polo-Parada and Korn convincingly demonstrate the converse: that Na+ blocks Ca2+ permeation through Ca2+ channels. This observation was not addressed by the various theories that suggest solutions to the sticky-pore problem of Ca2+ channels. Along with other recent work on Ca2+ channels, the results of Polo-Parada and Korn invite a rethinking of present models of Ca2+ channel permeation, and permeation through highly selective ion channels in general.
The most widely known theory for Ca2+ channel permeation proposes the presence of two high affinity Ca2+ binding sites in the pore (Hess and Tsien, 1984; Almers and McCleskey, 1984). How can two strong sites solve the problem caused by one? Multiple sites allow simultaneous occupancy of the pore by multiple ions and this could allow ion–ion interaction. If the ions interact negatively—perhaps by electrostatic repulsion or by competition for binding ligands—the effective binding affinity diminishes. When the negative interaction creates a low affinity state, Ca2+ can readily pass out of the pore. In this model, the essence of Ca2+ selectivity and permeation is: (a) a Ca2+ ion blocks permeation of foreign ions by binding to a high affinity site in the pore; (b) entry of a second Ca2+ into the pore causes a negative interaction that promotes exit of one of the two; (c) since one Ca2+ ion remains, the pore remains impassable to foreign ions. The paradox of getting high flux of an ion through a pore that has high affinity is solved simply by making the pore have low affinity when it is occupied by more than one Ca2+ ion. The model quantitatively reproduced the currents seen in mixtures of Ca2+ with either Na+ or Ba2+ and it predicted the on and off rates for Ca2+ blockade of single channel currents carried by monovalent ions (Lansman et al., 1986). Still, little of the model is actually proven. Its intuitive nature and the absence of alternative suggestions may be the real reasons that ion–ion interaction is so widely accepted as the means of getting significant flux through channels that select their preferred ion by binding. There is, however, a little problem: the two-site model for Ca2+ channels is explicitly disproven.
There simply are not two high affinity Ca2+ binding sites in the pore of the Ca2+ channel. Proof that there is only one high affinity intrapore site began by comparing the amino acid sequences of different voltage-gated channels. The main subunit of Ca2+ channels, like that of Na+ channels, consists of about 2,000 amino acids arranged in 4 homologous domains. Each domain is itself homologous to voltage-gated K+ channels, which have 6 putative transmembrane helices ( Jan and Jan, 1989). A variety of mutations and substitutions on K+ channels that modified permeation and pore blockade defined an amino acid sequence between the 5th and 6th helices that contributes to the lining of the pore (Heginbotham et al., 1992). Comparison of the sequences of the four pore-lining regions in the four domains of Na+ and Ca2+ channels suggested a quartet of amino acids, each on a different domain and therefore hundreds of amino acids separate in the linear sequence, that contribute to an intrapore binding site. In Ca2+ channels, these amino acids are all negatively charged glutamate residues and are called the EEEE locus. In Na+ channels, the four are aspartate, glutamate, lysine, and arginine (the DEKA locus). Heinemann et al. (1992) demonstrated the significance of the DEKA locus by modifying the lysine and arginine to glutamates. Then, the Na+ channel took on the key property of the Ca2+ channel: current carried by Na+ could be blocked by low concentrations of Ca2+.
After solving technical problems inhibiting heterologous expression of Ca2+ channels, the labs of R.W. Tsien and Y. Mori mutated the EEEE locus. The two-site model makes a very specific prediction of such an experiment. A point mutation that destroys a single high affinity Ca2+ binding site in the pore might modify Ca2+ permeation in an unpredictable way, but it should not modify Ca2+ block of monovalent current through the pore; since another high affinity site should remain, Ca2+ block should be unchanged. Instead, changing any of the glutamates in the EEEE locus to an uncharged amino acid shifted the concentration range for Ca2+ block of monovalent current to higher values (Yang et al., 1993; Kim et al., 1993; Ellinor et al., 1995). Therefore, there is just a single high affinity site in the pore and the four glutamates form it.
Disproof of the existence of 2 high affinity intrapore sites does not disprove the more important concept of ion–ion interaction driving high flux off of the selectivity site. Indeed, Yang et al. (1993) proposed that a second Ca2+ ion entering the pore would compete with the first for the glutamates at the EEEE locus, thereby turning a single high affinity site into a pair of low affinity sites. The concept is similar to that of Armstrong and Neyton (1991), which was proposed without the benefit of clear knowledge of channel structure. Like in the original two-site model, these ideas imagine that the first Ca2+ into the pore binds with high affinity and the second Ca2+ acts to lower the affinity and promote flux. These newer models focus on competition for binding ligands as the source of ion–ion interaction whereas the older model focused on electrostatic repulsion. No results distinguish between these two possibilities, and it is perfectly likely that both forces could act in concert to diminish pore affinity during simultaneous occupancy by two Ca2+ ions.
Now arrives the work of Polo-Parada and Korn describing a function for Na+ in permeation of Ca2+ through the channel. In quantitative models, Na+ affinity for the intrapore binding sites was assigned an arbitrarily low number in agreement with its weak ability to compete with Ca2+. Polo-Parada and Korn tested this assumption by systematically varying Na+ concentration and observing its effect on flux carried by 2 mM Ca2+ or Ba2+. They find that Na+ binds with low affinity, but the affinity is interesting because it is physiologically relevant. The relevance arises because: (a) external Na+ blocks Ca2+ flux through the channel; and (b) it does so at physiological Na+ concentrations.
At physiological Ca2+ and Na+ concentrations, flux through the Ca2+ channel is as low as 50% its maximal value due to block by Na+ at a site with an IC50 ≈ 120 mM. Ba2+ competes with Na+ for the site with an IC50 ≈ 2 mM. The near identity of these IC50's to the physiological concentrations of monovalent and divalent ions is striking. Polo-Parada and Korn suggest that events that alter monovalent ion concentration—intense electrical activity, ischemia, or renal pathology, for examples—might modify Ca2+ channel permeation in vivo.
If one assumes that the pore consists of a linear sequence of binding sites, the observations suggest that there is a site to the outside of the high affinity Ca2+ selectivity site. Here, monovalent and divalent ions at physiological concentrations can effectively compete with each other for occupancy. Additional evidence for such a site, as well as a similar site on the intracellular side, comes from Kuo and Hess (1993) which showed that monovalent ions “trap” the Ca2+ ion at the high affinity site. Trapping means that extracellular monovalent ions inhibit exit of Ca2+ outward from the selectivity site and intracellular monovalent ions inhibit exit inward. If there are symmetrical internal and external entry sites, monovalent ions should still compete better for the internal site because of the low intracellular concentration of Ca2+. True to this idea, Susumu Hagiwara, the pioneer of Ca2+ channel studies, showed in his final, posthumous paper that variations in the internal Na+ concentration modify the current through Ca2+ channels far more effectively than the same variations externally (Yamashita et al., 1990). Together, these observations strongly suggest that there are one or more binding sites that flank the high affinity selectivity site of the Ca2+ channel pore. The pore seems to have a minimum of three binding sites in sequence and one of them, which cannot be either the outermost or innermost, has a high affinity for Ca2+ and serves as the key element for choosing Ca2+ over monovalent ions.
Although unanticipated discoveries such as this Na+ block may make our understanding of selective permeation appear embarrassingly rudimentary, a fundamental advance has been made since Hodgkin and Huxley feared to use the word “channel” to describe a molecule that selectively carried current by only one ion. All the results discussed above point to a pore consisting of a sequence of binding sites through which ions move in single file. The theoretical understanding of “single filing” was developed on K+ (Hille and Schwarz, 1978) and gramicidin (Andersen and Procopio, 1980) channels, and appreciation of its significance grows as other types of channels are considered. In such theories, an ion hops from a binding site to a neighboring unoccupied site. This mechanism is identical to the way ions move through semiconductor crystals (Shockley et al., 1952), so it is intuitively unsurprising that this provides a way to obtain very high fluxes through pores. Interestingly, the strongest evidence for single filing came from Hodgkin himself a few years after the Hodgkin and Huxley papers. Hodgkin and Keynes (1955) showed that an unexpectedly steep concentration dependence of unidirectional flux through K+ channels could be explained if there were 4 sequential sites in the pore. The final figure of their paper diagrammed a mechanical model used to count the transit of balls through a long, multi-ball pore and a short, single-ball pore. This toy, which effectively reproduced their flux studies and confirmed their theoretical formulas, should give us hope that there are simple explanations for unanticipated, and even unintuitive, results.