The Ca-activated Cl Channel and its Control in Rat Olfactory Receptor Neurons

Sensory transduction in vertebrate olfactory receptor neurons (ORNs) begins with the activation of specific receptor proteins on the plasma membrane of the cells' cilia by odorant molecules, leading via a G protein to stimulation of adenylyl cyclase to increase cAMP synthesis (for reviews see Schild and Restrepo, 1998; Gold, 1999). The rise in cAMP opens cyclic nucleotide–gated (CNG) channels that are permeable to both monovalent and divalent cations (Kurahashi and Shibuya, 1990; Zufall and Firestein, 1993; Dzeja et al., 1999). Consequently, Ca²⁺ in the ciliary cytoplasm increases (Leinders-Zufall et al., 1998; Reisert and Matthews, 2001b) and opens Ca-activated Cl channels. Because the Cl equilibrium potential across the ciliary membrane is much more positive than the resting membrane potential of ORNs (Zhainazarov and Ache, 1995; Reuter et al., 1998; Kaneko et al., 2001), a Cl efflux ensues, thus depolarizing the cell. This unusual, excitatory Cl current has been demonstrated in ORNs from amphibia (Kleene and Gesteland, 1991; Kurahashi and Yau, 1993), mammals (Lowe and Gold, 1993), and fish (Sato and Suzuki, 2000). Intracellular Ca²⁺ itself is regulated by a Na/Ca exchanger, which utilizes the inward-directed electrochemical Na gradient across the ciliary membrane to extrude Ca²⁺, thereby restoring the basal intracellular Ca²⁺ concentration within a few seconds after the stimulus (Reisert and Matthews, 1998). In amphibians, the Cl channel is thought to serve as a safeguard of olfactory transduction by compensating for any decrease in transduction efficacy caused by dilution of the nasal mucus by fresh water (Kurahashi and Yau, 1993). When it was found to be present also in rat ORNs, a land-based mammal, the idea was proposed that this channel has evolved to serve conveniently as an amplifier of olfactory transduction (Lowe and Gold, 1993). Some general properties of this channel, including its activation caused by Ca²⁺ influx through the CNG channel, have been characterized in a frog excised-cilium preparation (Kleene, 1993). However, little is known about this channel in mammalian ORNs, despite its critical role in olfactory transduction. In the first part of this paper, we describe general properties of this channel in rat ORNs.

In the second part of the paper, we use these obtained properties of the Cl channel to investigate the functional relation between the Cl channel, the CNG channel and the Na Ca exchanger. A wide variety of neuronal functions depend on the interplay between Ca²⁺ influx through Ca-permeable ion channels and proteins controlled by intracellular Ca²⁺. Based largely on studies of neuronal exocytosis (Neher, 1998; Meinrenken et al., 2003), it now appears that a network of sources and sinks of Ca²⁺ act together to establish a complex, dynamic profile of Ca²⁺ concentration on the cytoplasmic surface of the plasma membrane. Thus, the activation of Ca-gated ion channels can be understood from the spatial distribution of these channels within such a Ca profile. Decisive parameters for this functional coupling are the distances between Ca sources, Ca sinks, and the Ca-activated channels, as well as the speed of Ca²⁺ diffusion and the concentrations of Ca buffers. Since intracellular Ca²⁺ stores do not contribute to Ca²⁺ homeostasis in olfactory cilia (Zufall et al., 2000), and because the CNG channels, the Ca-activated Cl channel, and the Na Ca exchanger are all situated on the plasma membrane, it is possible to examine their functional interactions in a highly reduced system consisting of an inside-out membrane patch excised from rat ORNs. We introduce a method that can generally be applied to determine the distance between neighboring Ca-permeable channels and Ca-activated channels, and provides an estimate of channel densities in excised membrane patches. Recently, it has been suggested that olfactory transduction may take place in microdomains through supramolecular protein complexes (Schreiber et al., 2000; Goldstein et al., 2003). Our results, however, suggest that the Cl channel is not organized into a transduction complex with the CNG channel or the Na/Ca exchanger.

ORNs were isolated from female 4–10 wk old Wistar rats. The animal was decapitated after an overdose of isofluran in accordance with legally approved procedures and the head was split along the septum to expose the olfactory turbinates. Septal tissue and turbinates were transferred to extracellular solution and the tissue was separated from remaining cartilage. To aid dissociation, the tissue was incubated in extracellular solution (Table I) containing 0.2mg/ml trypsin (Sigma-Aldrich) for 30 min at 37°C and transferred to DNase solution (Table I). After gentle trituration, 0.5 ml of cell suspension was transferred to a recording chamber mounted on an inverted microscope with phase-contrast optics, and allowed to settle for 30 min before being continuously superfused with LiCl or NaCl solution.

Inside-out patches were excised from the dendritic knob of ORNs, which were easily recognized by their morphology. Cilia were often visible and were sometimes sucked into the tip of the recording pipette before a gigaohm seal was established. Therefore, the excised patches contained not only apical membrane of the dendritic knob, but also ciliary membrane. Pipettes were pulled from standard-wall borosilicate glass (Warner Instrument Corp.) and fire-polished to give an open-pipette resistance of 6–7 MΩ. Currents were recorded using an Axopatch-1D amplifier and acquired using a Digidata interface and pClamp software (Axon Instruments, Inc.). Junction potentials arising from dissimilar pipette and bath solutions were compensated. All data were sampled at 500 Hz and low-pass filtered (8-pole Bessel) at 100 Hz, with the exception of the noise recordings, which were digitized at 5 kHz, low-pass-filtered (8-pole Bessel) at 2.5 kHz, and subsequently digitally high-pass filtered at 2 Hz before variance calculations were made. Unless noted otherwise, the holding potential was −40 mV.

For current-voltage relations, care was taken not to introduce an artificial rectification due to the inactivation/desensitization of the Cl channel. For this purpose, patches were preexposed to 67 μM Ca²⁺ for 4 s to allow the current to partially inactivate. Thereafter, a double voltage ramp was used from −60 to +60 mV and then back to −60 mV at ±120 mV/s, and the two current-voltage relations were averaged. Leak currents in the respective Ca²⁺-free solution were subtracted. Dose–response relations for the Ca-activated Cl conductance were constructed by normalizing individual patches to their maximal current in order to average data across patches. Subsequently, Hill functions were fitted to the collected data using Origin software (Microcal Software Inc., Northamptom, MA USA). Relative halide permeabilities and theoretical reversal potentials were calculated using the Goldman-Hodgkin-Katz equation. In calculations, the 9.98 mM CaCl₂ added to achieve 67 μM free Ca²⁺ were taken into account. Averaged values are mean ± SEM if not noted otherwise.

All solutions contained 10 mM HEPES, with the pH adjusted to 7.2 with NMDG. DNase solution contained 0.2 or 0.4 mg/ml DNase 1 (Roche Diagnostics GmbH). The program Chelator (T.J.M. Schoenmakers, University of Nijmegen, Nijmegen, the Netherlands) or WinMAXC (C. Patton, Stanford University, Palo Alto, CA USA) was used to calculate free divalent concentrations in buffered solutions. The free Ca²⁺ concentrations in the HEDTA-buffered Ca solutions were determined using Ca-sensitive electrodes (Kennedy and Thomas, 1996). These solutions contained slightly different amounts of Cl⁻ because varied amounts of CaCl₂ were added to the buffered Ca solutions to obtain different free Ca²⁺ concentrations.

Whenever possible, Li⁺ was used instead of Na⁺ as the monovalent cation. Li⁺ is permeant through the rat CNG channel (Frings et al., 1992) but does not support Na/Ca exchange (Reuter and Seitz, 1968; Reisert and Matthews, 1998), a mechanism that can potentially change the free Ca²⁺ concentration even in excised patches (see Fig. 8).

Rapid solution exchanges were achieved by transferring the pipette across the interface between neighboring solution streams using a system based on the Perfusion Fast-Step SF-77B (Warner Instrument Corp.). Solution exchange was complete within 15 ms.

In the presence of excess HEDTA compared with Ca²⁺, the free Ca²⁺ concentration at distance r from a Ca channel is linearly proportional to the Ca²⁺ flux, Φ (in number of Ca²⁺/s), through this channel (Neher, 1986; Bauer, 2001).

where N_Av is the Avogadro number, D_HEDTA and D_Ca the diffusion coefficients for HEDTA and Ca²⁺, respectively, and

with the reciprocal reaction time, τ⁻¹, for binding of Ca²⁺ to HEDTA at the basal-free Ca²⁺ concentration [Ca²⁺]₀ being:

and the Ca²⁺-binding ratio, κ, being:

In the latter equations, K_diss and k_on signify the thermodynamic equilibrium constant and the rate constant for Ca²⁺ binding to HEDTA, respectively. In our experiment, [Ca²⁺]_o = ∼0. Owing to the linear relation between

where

where r₁ denotes the distance between a given Cl channel and the closest CNG channel (Fig. 7 B). We then plot the difference (Δ[Ca²⁺]) between

occur at the same r₁. Subscripts indicate HEDTA concentrations in mM. Note that Eqs. 4 and 5 are defined to have opposite signs and are, therefore, mirrored functions to facilitate accurate determinations of r₁ and D_Ca.

We studied the Ca-activated Cl current with patch-clamp recording from inside-out membrane patches excised from the dendritic knob/cilia of rat ORNs (see materials and methods). Although Cl channels and CNG channels are both present in the membrane, each channel population could be selectively activated by Ca²⁺ or cAMP in the bath solution. In most experiments, the pipette and bath solutions both contained LiCl instead of NaCl in order to disable the Na/Ca exchanger, which cannot be driven by Li⁺ (Reuter and Seitz, 1968; Reisert and Matthews, 1998), so that the Ca²⁺ concentration near the membrane patch could be properly controlled by the bath solution. At −40 mV and with symmetrical 140 mM LiCl base solutions (no Ca²⁺ or Mg²⁺; see Table I), application of 67 μM Ca²⁺ or 100 μM cAMP elicited a respective saturated current (Fig. 1 A), with the Ca-activated current initially being larger (83 ± 35 pA; 12 patches) than the cAMP-activated current (52 ± 18 pA; same 12 patches). It will be confirmed below that the Ca-activated current was indeed carried by Cl⁻. In the course of time, this Ca-activated Cl current declined in amplitude, whereas the cAMP-activated current stayed quite constant (Fig. 1 B). This irreversible run-down of the Cl current, not hitherto described by others (Kleene and Gesteland, 1991; Hallani et al., 1998), occurred in most patches (10 of 12) and could be substantial. On average, the I_Cl/I_CNG ratio decreased from 2.11 ± 0.35 to 1.01 ± 0.23 (Fig. 1 C; 12 patches). The time course of run-down varied greatly between patches, ranging from a few minutes to almost an hour.

During current run-down, the quantitative sensitivity of the Cl channel to Ca²⁺ did not change, in that the ratio, I_sub/I_sat, of the currents elicited by a sub-saturating (2.4 μM) and a saturating (67 μM) Ca²⁺ concentrations remained constant over time (Fig. 1, D–F; five patches in F). Thus, the run-down appeared to reflect a progressive loss of functional Cl channels. This conclusion was supported by noise measurements to be described below.

The run-down of the Cl current was often biphasic. In the first few minutes after excision of the patch, the current declined fairly quickly but thereafter the decline slowed down significantly (see, for example, Figs. 1 E and 5 B). Once the rapid phase of decline was over, experiments requiring relative current stability over an extended time period could be performed. All experiments described below were performed after the initial run-down.

To confirm that the Ca-activated current was indeed carried by Cl⁻, we maintained the pipette Cl⁻ concentration ([Cl⁻]_o) at 140 mM (Na salt) and progressively reduced the bath Cl⁻ concentration ([Cl⁻]_i) by equimolar substitution with methanesulfonate (Na salt), an anion that does not permeate well through most Cl channels (for review see Frings et al., 2000). Current was activated by 67 μM Ca²⁺ in the bath. As [Cl⁻]_i was reduced, the reversal potential of the current-voltage relation shifted to negative values (Fig. 2 A) in a manner predicted fairly well by the Nernst equation (Fig. 2 B, mean ± SD from 9 patches), indicating that the channel was anion selective. The nonperfect Nernstian behavior could be due to methanesulfonate being slightly permeant through the channel. With equimolar replacement of bath NaCl by NaBr or NaI (NaCl in the pipette), and corresponding activation by 67 μM Ca²⁺ (see materialsandmethods), the reversal potential shifted positively (Fig. 2 C), indicating that Br⁻ and I⁻ were more permeant than Cl⁻. For the anion F⁻, because CaF₂ is insoluble, the F⁻ permeability was tested with F⁻ in the pipette and Cl⁻ in the bath instead (inset in Fig. 2 C). From the reversal potential values and the Goldman-Hodgkin-Katz equation, the permeability sequence was I⁻ > Br⁻ > Cl⁻ > F⁻ and the permeability ratios, P_x/P_Cl, were 3.23:1.98:1:0.16 (10 patches for Br⁻ and I⁻, 6 patches for F⁻). This sequence indicates that the relative permeability is inversely related to the hydration energy of the halide (Fig. 2 D).

We also tested the ability of divalent cations other than Ca²⁺ to activate the Cl channel. At −40 mV and 1 mM of each tested divalent cation, Sr²⁺ was quite effective (current ratio I_Sr/I_Ca = 0.78), Ba²⁺ was very ineffective (I_Ba/I_Ca = 0.015) and Mg²⁺ was completely ineffective (Fig. 2 E, F). Additional six patches exposed to 2 mM Mg²⁺ at 40 mV also produced no current. Therefore, it can be safely assumed that Mg²⁺ does not play a physiological role in the activation of these channels. Unless indicated otherwise, all subsequent experiments were performed in the absence of bath Mg²⁺.

In addition to run-down, there was a reversible reduction of the current with a time course of seconds after activation by Ca²⁺ (see Fig. 1, 3, 4, 5). This reduction, which we refer to as inactivation/desensitization, also has not been described previously by others. Like the rundown, it did not show clear dependence on Ca²⁺ concentration or membrane potential (see Fig. 4). It occurred regardless of whether Cl⁻, Br⁻, or I⁻ was the charge carrier (data not shown). The time course of current decline did not fit a single-exponential decline, and was not analyzed further in this work. This current decline could reflect the development of an inactivated/desensitized state of the Cl channel after Ca²⁺ binding, or result from current flow causing local depletion/accumulation of Cl⁻ (Zimmerman et al., 1988; Lindemann, 2001). We ruled out the second possibility by momentarily curtailing current flow to near zero by holding the membrane potential at zero. In such an experiment, we found that the current amplitude at a fixed time after Ca²⁺ exposure stayed the same regardless of how long the Cl current had been flowing (Fig. 3 A; experiment was performed in eight patches). Thus, the inactivation/desensitization appeared to be an intrinsic property of the channel itself.

Because calmodulin (CaM) is known to mediate multiple Ca-induced mechanisms of feedback inhibition in ORNs, we asked whether this protein also affected the inactivation/desensitization. In an excised patch, as expected from previous work (Chen and Yau, 1994; Bradley et al., 2001), the CNG current activated by a subsaturating (10 μM) cAMP concentration was reversibly inhibited by Ca²⁺-CaM; 300 μM niflumic acid was present throughout in order to remove most of the Cl current (compare “-CaM” and “1 μM CaM” traces in Fig. 3 B). On the other hand, the Cl current activated by a subsaturating, buffered Ca²⁺ concentration (2.4 μM) was not affected by CaM (Fig. 3 C). At the same time, the drop in the current (i.e., inactivation/desensitization) was unchanged in the presence of CaM. The choice of a subsaturating Ca²⁺ concentration here was deliberate, in order not to overlook a mere shift in the Ca²⁺ concentration dependence of the current due to CaM. Similar results were obtained from seven patches. Thus, the inactivation/desensitization of the Cl current was neither caused by CaM nor sensitive to it. Possibly, it resulted directly from Ca²⁺ binding to the channel.

To quantify the functional interaction between the Ca-activated Cl channel and the CNG channel, it is necessary to characterize the full dose–response relation between the Cl current and Ca²⁺ concentration. These experiments were performed after the rapid phase of the run-down was complete—a justifiable protocol because, as pointed out earlier, the run-down did not affect the quantitative dependence of the Cl channel on Ca²⁺ concentration. Only patches showing subsequent current run-down of <10% in the course of measurements were included for analysis. The Ca dependence was measured in symmetrical LiCl solutions and at ±40 mV (Fig. 4, A and C), using a Ca²⁺-exposure time of 10 s. To test for any sensitivity change due to inactivation/desensitization, the current amplitudes at transient peak and at 10 s were both used for constructing dose–response relations (mean ± SD of normalized data shown in Fig. 4 B; −40 mV: six patches; +40 mV: seven patches). The smooth curves are Hill-equation fits. At –40 mV, the Ca²⁺ concentration for half-maximal activation, K_1/2, was 2.2 μM for peak current (black) and 2.1 μM for current at 10 s (green); the Hill coefficient, n, was 2.8 and 2.8, respectively. At +40 mV, the K_1/2 was 1.5 μM (red) and 1.4 μM (blue), and n was 3 and 3, respectively. Thus, like run-down, channel inactivation/desensitization was not associated with a quantitative change in the Ca sensitivity (see also Fig. 4 D). The slight decrease in K_1/2 at positive voltages could in principle reflect a voltage sensitivity in the Ca binding step or the gating step.

Monovalent current through CNG channels is sensitive to blockage by external divalent cations (Haynes et al., 1986; Zufall and Firestein, 1993; Frings et al., 1995; Dzeja et al., 1999). Thus, in order to compare the magnitudes of the CNG and Cl currents in the intact-cell situation, we included physiologically relevant concentrations of divalent cations (2 mM Ca²⁺, 1 mM Mg²⁺) in the pipette solution (Fig. 5 A). Under these conditions, the Cl current activated by a saturating concentration (67 μM) of Ca²⁺ shortly after patch excision remained large (136 ± 41 pA; 16 patches), but the CNG current elicited by a saturating concentration (100 μM) of cAMP was very small (4.4 ± 1.2 pA; 16 patches). As in the absence of divalent cations, the Cl current still showed run-down and inactivation/desensitization, whereas the cAMP-activated current remained constant throughout (Fig. 5 B). The mean initial current ratio I_Cl/I_CNG in this case was 33.34 ± 8.97 (indicated by ▪ in Fig. 5 C; 16 patches), instead of just twofold in the absence of divalent cations (see above). Even after run-down, this ratio stayed at 22.54 ± 7.43. Thus, in physiological ionic conditions, the maximum Ca-activated Cl current was ∼30-fold that of the cAMP-activated current.

Under the above experimental conditions, Ca²⁺ from the pipette will enter the patch through the CNG channel during application of cAMP. This could potentially activate the Cl channel and generate an additional current component. However, with the high HEDTA concentration of 10 mM used, this was unlikely to happen (see Fig. 7).

For the modeling to be described later, it is necessary to know the current through a single Cl channel and also the quantitative amount of Ca²⁺ influx through a single olfactory CNG channel. At least for the Cl channel, single-channel measurements are not feasible because the unitary current is too small. We therefore resorted to noise analysis. We took advantage of the fact that, in some patches, the membrane current developed slowly and decayed even more slowly despite rapid application/removal of 100 μM cAMP or 67 μM Ca²⁺, presumably due to a slow exchange of solutions (Fig. 6, A and B). In the case of the Ca-activated Cl current, we deliberately slowed the Ca²⁺ washout further by using a bath solution containing a low concentration of Ca buffer (0.1 mM HEDTA) between Ca²⁺ exposures. When high-pass filtered at 2 Hz (Fig. 6, A and B, traces labeled “hi-pass”), an increase in noise was evident during the falling phases of the CNG and Cl currents, taken to originate from the more frequent transitions between the open and closed states of the channels at submaximal open probabilities. Assuming, for simplicity, a single open state and a single closed state, the increase in current variance (σ²) is related to the unfiltered mean current (I) by the parabolic function σ² = iI − I²/N, where i is the single-channel current and N is the total number of channels on the patch (DeFelice, 1981). This equation fits the data quite well for both the CNG channel (Fig. 6 C) and the Cl channel (Fig. 6 D, different patch). In the experiments shown, i = 0.49 pA and N = 385 for CNG channels, and i = 0.07 pA and N = 999 for Cl channels. The ratio of maximum current (see figure legend) to N ^* i gave a maximal open probability, P_o, of 0.83 for the CNG channel and near 1.0 for the Cl channel. From multiple experiments, the mean single-channel current at −40 mV was −0.49 ± 0.04 pA for the CNG channel (nine patches) and −0.051 ± 0.003 pA for the Cl channel (seven patches). The mean P_o was 0.8 for the CNG channel and 0.97 for the Cl channel. The ratio of channel numbers (i.e., N_Cl/N_CNG) was 3.6 ± 2.2.

Because the P_o value for the Cl channel remained near unity even though the measurements were made after the rapid run-down of the current, the run-down itself could not be due to a decrease in P_o. Thus, as for other ion channels (for review see Becq, 1996), run-down was most likely due to a loss of functional channels, a phenomenon that was also described for Ca-activated Cl channels in aortic smooth muscle cells (Hirakawa et al., 1999).

A similar ratio of Cl channel to CNG channel numbers can be derived in another way. In the absence of external divalent cations, and before Cl current run-down, the average I_Cl/I_CNG was ∼2 in symmetrical LiCl solutions (Fig. 1 C). Because Li⁺ carried current half as well as Na⁺ through the olfactory CNG channel (Frings et al., 1992), the I_Cl/I_CNG ratio in symmetrical NaCl solutions would therefore have been near unity, i.e., I_Cl = I_CNG. The ratio of the respective channel numbers is

During the response to odorants, Cl channels are activated by Ca²⁺ that enters the ORN cilia through the CNG channels. Therefore, the distance between the Cl and CNG channels is key to Ca signaling in these neurons. Recently, it has been suggested that scaffolding proteins in olfactory cilia (Schreiber et al., 2000; Goldstein et al., 2003) organize components of olfactory transduction into a tranducisome complex as in Drosophila photoreceptors (for review see Tsunoda and Zuker, 1999). In a transducisome, intermolecular distances are in the range of <10 nm because the transduction molecules are immediate neighbors within the supramolecular complex (for review see Sheng and Sala, 2001). Ca²⁺ entering the cell via a CNG channel would travel this distance in such a short time that binding to a buffer is effectively impossible (Neher, 1998) and the secondary activation of the Cl channel should therefore be unaffected by changes in the Ca-buffer concentration. A test for proximity on the scale of a transducisome would be to examine whether the functional coupling between CNG channels and Cl channels can be intercepted with a Ca buffer (HEDTA). If coupling was not affected with HEDTA, the two channels would likely be very close together and possibly components of a transducisome. In the following, we show first that the coupling is indeed strongly affected by HEDTA, and second, that this finding can be used for estimating the mean distance between the two channel types on the membrane.

A patch was held in symmetrical CholCl solutions in order to suppress monovalent cation flux through the CNG channels (Balasubramanian et al., 1995). In addition, the pipette solution contained 1 mM Ca²⁺ to support Ca²⁺ flux through the CNG channels. In the experiment of Fig. 7 A, saturating (67 μM) Ca²⁺ in the bath elicited a peak current of −30 pA at −40 mV (black trace). With 100 μM cAMP and at high Ca-buffering capacity (1 mM HEDTA) in the bath solution, the Ca²⁺ flux from the pipette solution through the fully activated CNG channels induced only about −8 pA of current (Fig. 7 A, light-blue trace). With reduced bath Ca-buffer capacity (0.2 mM HEDTA), the current induced by cAMP increased to −21 pA (green trace). The cAMP-induced current was greatly suppressed by 300 μM niflumic acid (dark-blue trace), demonstrating that it was largely carried by Cl⁻, plus a small fraction being carried by Ca²⁺ through the CNG channels. When a solution containing 0.2 mM HEDTA and no cAMP was applied (red trace) a smaller current was observed that probably originated from spontaneous gating of the CNG channel (Kleene, 2000) or from residual cAMP still present in the immediate vicinity of the patch. As expected, when Ca²⁺ influx was largely minimized by clamping the membrane potential at 40 mV, no Cl current was observed (data not shown). The time course and extent of the Cl current decline (inactivation/desensitization) during the cAMP steps resembled those recorded during Ca pulses (compare Fig. 4), indicating that the local Ca profile sensed by the Cl channels already reached steady state during the cAMP step. The cAMP-induced Cl current thus illustrates a functional link between the CNG channels and the Cl channels. However, its sensitivity to a change in HEDTA concentration provides qualitative evidence against the notion that the two transduction channels are assembled within transducisomes. Previously, Kleene (1993) has likewise shown that the presence of BAPTA (a faster buffer than HEDTA) in an excised cilium could affect the Cl current activated by Ca²⁺ entering through the CNG channel.

How far are CNG channels and Cl channels separated? The average time required for Ca²⁺ to bind a buffer molecule depends on the binding rate and the buffer concentration. This time period determines the mean distance Ca²⁺ can freely diffuse from the CNG channel, which is, generally, 100–200 nm (Neher, 1998). Based on the theory of buffered Ca²⁺ diffusion (Neher, 1986; Bauer, 2001), we calculated the mean distance, r, between neighboring CNG and Cl channels using the above measurements. Each CNG channel constitutes a point source of Ca²⁺, surrounded by a radially symmetrical Ca²⁺ concentration ([Ca²⁺]_free) profile that peaks at the pore and decays with increasing distance from the channel. This concentration profile depends on the Ca²⁺ flux through a CNG channel, the binding to HEDTA, and the Ca²⁺ diffusion coefficient D_Ca. With more than one CNG channel around, the [Ca²⁺]_free profiles generated by individual channels in the presence of excess Ca buffer simply sum (see below and materials and methods). The amplitude of the current through a Cl channel depended on its location within this summed Ca profile, and therefore on the distances between the CNG channels and the Cl channels. From our estimates of the single CNG channel current (0.49 pA) and its P_o (0.8), together with the previous finding that the macroscopic monovalent current is reduced to 16% in the presence of 1 mM external Ca²⁺, of which the fractional current carried by Ca²⁺ is 0.4 (Dzeja et al., 1999), we calculated a Ca²⁺ current through a single CNG channel under our experimental conditions to be 0.025 pA, or 78,000 Ca²⁺/s. Because the efficiency of Ca buffering by HEDTA is also known (Thomas, 1982; Naraghi and Neher, 1997), the remaining parameters, D_Ca and r, can be solved by experimentally measuring

Because the CNG channels and Cl channels appeared not to be clustered in transducisomes, we assumed that they are arranged in a regularly-spaced matrix (Fig. 7 B), with a ratio of four functional Cl channels per CNG channel after the fast run-down of the Cl current. This assumption allows to derive an analytical solution of the Ca diffusion problem (see materials and methods). As mentioned above, the local [Ca²⁺]_free at a given Cl channel is equal to the sum of contributions from surrounding CNG channels, with decreasing contributions from more distant CNG channels (represented by concentric circles in Fig. 7 B). To quantify the contributions from distant channels, we fitted the measurements with the appropriate D_Ca and r values for the case of up to 15 CNG channels (circles 1–7 inclusive) contributing to the local [Ca²⁺]_free sensed by each Cl channel. We then repeated the calculations for decreasing numbers of circles and related the results to that obtained with 15 channels (=100% in Fig. 7 C) (it was not feasible to extend the model beyond 15 channels without taking into account the specific geometry of the patch; see below). The conclusion was that at least 11 channels (circles 1–6 in Fig. 7 B) had to be included in the model in order to account for >90% of

In the matrix shown in Fig. 7 B, the distance r₁ between a Cl channel and the nearest CNG channel is related to the space constant, d, of the matrix by

Excised from the dendritic knob, the membrane patches often contained one or more cilia. We have therefore recalculated r₁ and D_Ca using, instead of a planar model as in Fig. 7 B, a model consisting of a linear channel arrangement in order to simulate a long slender cilium. The derived r₁ value and channel densities remained similar to above, though D_Ca was reduced by a factor of 2 (data not shown). Nonetheless, the calculations indicated that, even in this case, each Cl channel still monitors Ca²⁺ originating from several CNG channels. Thus, regardless of patch geometry, the conclusion of a spatial integration of Ca signals from more than one CNG channel appears to hold.

While the CNG channel is responsible for an increase in ciliary Ca²⁺ upon activation by odorants, a Na-dependent Ca extrusion returns Ca²⁺ to the basal level (Reisert and Matthews, 1998). We likewise asked whether the Na/Ca exchanger and the Cl channel are organized into transducisome-like complexes. To address this question, we ran the Na/Ca exchange in reverse mode in order for it to serve as a Ca²⁺ source for activating Cl channels in an excised patch, and asked whether the resulting Cl current was sensitive to Ca-buffering in the bath solution. In this experiment, the pipette solution contained 140 mM CholCl plus 2 mM Ca²⁺. Reverse Na/Ca exchange was facilitated by holding the membrane potential at 40 mV, in order to take advantage of its electrogenicity (for review see Blaustein and Lederer, 1999). With initially 140 mM CholCl in the bath, maximal activation of Cl channels by saturating (67 μM) Ca²⁺ in the bath produced a current of 40 pA (Fig. 8, black trace). In the absence of bath Ca²⁺, replacing the bath CholCl with NaCl elicited a steady-state current of ∼20 pA at the low Ca-buffering capacity of 25 μM HEDTA (Fig. 8, red trace). This current was substantially reduced by 300 μM niflumic acid, thus identifying its Cl constituency (Fig. 8, green trace). The current was also absent when the internal HEDTA concentration was increased to 1 mM (dark-blue trace). The involvement of the Na Ca exchange in generating the Cl current was further supported by the fact that Li⁺ was unable to substitute for Na⁺ for generating the Cl current (Fig. 8; Na⁺, red trace; Li⁺, light-blue trace). Similar results were obtained from a total of six patches. In our experiments, reverse Na/Ca exchange was able to operate in the absence of K⁺. Thus, the exchanger on olfactory cilia apparently does not depend on K⁺ cotransport, unlike the Na/Ca,K-cotransporter in photoreceptors (Cervetto et al., 1989). This experiment suggests that the Cl channel and the Na/Ca exchanger are likewise situated in close proximity on the ciliary membrane, but far enough to allow interception of Ca²⁺ by Ca buffers. A quantitative estimate of the distances between the two, however, is not possible because there is currently no information about the density of the exchanger and its Ca-transport rate.

Previously, the olfactory Ca-activated Cl channel has been studied in frog (Kleene and Gesteland, 1991; Kleene, 1993, 1997). Generally speaking, the properties of the frog channel are similar to those of the rat channel. In both animal species, this channel shows inward rectification, a small single-channel conductance (0.5–1.5 pS), an activation K_1/2 of slightly over 2 μM Ca²⁺ at −40 mV, an open probability of the liganded channel near unity, and a lack of sensitivity to calmodulin (Kleene and Gesteland, 1991; Kleene, 1997, 1999). The dose–response relation displays pronounced cooperativity, suggesting that at least three Ca²⁺ ions must bind to open a channel, a mechanism that might contribute to the steep dependence of the odor-induced response on stimulus concentration (Firestein et al., 1993), especially in mammals (Reisert and Matthews, 2001a). The above lack of any major species-dependent differences we found is in contrast to the recent results of Hallani et al. (1998). This group also studied the rat channel on excised, inside-out membrane patches from the dendritic knob of ORN, but found a linear current-voltage relation and a K_1/2 value of 26 μM Ca²⁺ at an unspecified membrane potential. Because they used unbuffered Ca solutions, it is difficult to evaluate their results. Also, they found an unusual halide permeability sequence (Cl⁻ > F⁻ > I⁻ > Br⁻), whereas the sequence we obtained (I⁻ > Br⁻ > Cl⁻ > F⁻) is similar to those reported for Ca-activated Cl channels in other cell types (for review see Frings et al., 2000). Finally, they reported a strong current suppression at all voltages when Cl⁻ was replaced by other halides, which we did not observe.

The current run-down we observed was not described by Hallani et al. (1998) or in frog (Kleene and Gesteland, 1991; Kleene, 1993). In the latter case, the current run-down could have been missed because in these experiments Cl current activation occurred via diffusion of Ca²⁺ into an excised cilium, a relatively slow process. The run-down could arise from the elution of a soluble factor that is required for the gating of the channel by Ca²⁺. In other tissues, a variety of modulators of Ca-activated Cl channels have been described or postulated. Protein kinases (CaMK II and PKC), inositol polyphosphates, and cytoskeletal proteins can affect the time course of Ca-activated Cl currents in epithelial and smooth-muscle cells (Ismailov et al., 1996; Schlenker and Fitz, 1996; Wang and Kotlikoff, 1997; Xie et al., 1998), although the details of these modulations are not understood. Whatever its mechanism, this run-down of the olfactory Cl current may contribute to the progressive reduction of the odor-induced receptor current recorded in the whole-cell configuration (e.g., Zhainazarov and Ache, 1995).

In contrast to the run-down, which is observed only for excised patches or dialyzed ORNs, the intrinsic inactivation/desensitization is expected to affect the shape of the olfactory response in vivo. It will accelerate the termination of the odor-induced receptor current, and is consistent with the observation that the receptor current declines more rapidly than the ciliary Ca²⁺ signal (Reisert and Matthews, 2001b).

The distance between Ca-permeable channels and Ca-activated channels can be measured with biochemical or biophysical techniques (e.g., chemical cross-linking or FRET) if the channels are no more than ∼1–2 nm apart, or with immunofluorescence methods if the distance is within the resolution limit of light microscopy (∼1 μm). However, these methods require knowledge of the molecular identity of the proteins involved, which is absent for the olfactory Ca-activated Cl channel. Here we describe a method that fills the gap between these techniques and permits measurements of mean distances in the range of several hundred nanometers. Our quantitative analysis of channel densities is based on insights into buffered Ca²⁺ diffusion near Ca channels (Neher, 1986; Bauer, 2001). This approach is in principle applicable to any situation in which the effect of Ca buffers on Ca-dependent currents can be analyzed.

Application of the Ca-buffer method to the ciliary membrane of ORNs yielded several results: (a) evidence against the colocalization of CNG channels and Ca-activated Cl channels in transducisomes; (b) estimates for the densities of CNG channels and Cl channels and the distances between them; (c) an estimate for the lateral Ca²⁺-diffusion coefficient on the cytoplasmic side of the ciliary membrane; (d) evidence that a number of CNG channels contribute to the local [Ca²⁺]_free sensed by each Cl channel; and (e) evidence for functional coupling, but against direct contact, between Cl channels and Na/Ca exchanger sites.

Efficient functional coupling within a transduction cascade requires soluble messengers such as Ca²⁺ having to travel only short distances to reach their targets. In extreme cases, direct contact or organization into transducisome complexes between the transduction proteins can minimize the signal-transfer distance, as proposed for certain components of phototransduction in vertebrates (Korschen et al., 1999; Schwarzer et al., 2000; Poetsch et al., 2001; Bauer, 2002) and invertebrates (Zuker, 1996). Such an arrangement gives rise to a fast light response in Drosophila photoreceptors with latencies as short as 2 ms (Tsunoda and Zuker, 1999). Here, we show that the transduction channels in ORNs are not constituents of transducisomes but are positioned ∼120 nm apart. This finding indicates that olfactory signal transduction is not optimized for speed. On the other hand, ORNs are equipped with an unusual current amplification mechanism, in the form of a large Cl current triggered by a primary Ca current through CNG channels. In this respect, the olfactory CNG channels serve a somewhat different purpose from that of the CNG channels in vertebrate photoreceptors, which conduct the receptor current alone (for review see Kaupp and Seifert, 2002) and are expressed at ∼30-fold higher density (300–500 μm⁻² and 160 μm⁻² in bovine and salamander rod outer segments; Bauer and Drechsler, 1992; Karpen et al., 1992). Interestingly, the number of CNG channels contributing significantly to the local Ca²⁺ concentration in the olfactory cilia appears to be much smaller than, for example, the number of voltage-activated Ca channels (∼60) in the rat calyx of Held synapse necessary for the release of a single vesicle (Borst and Sakmann, 1996).

Although the Cl channel density on ORN ciliary membrane is comparable in both rat and frog, the CNG channel density is an order of magnitude lower in rat than previously found in frog (Larsson et al., 1997). This species difference may explain the considerably larger contribution of Cl current to the overall odor-induced current in mammals (>80%) than in amphibians (36–60%) (Kurahashi and Yau, 1993; Lowe and Gold, 1993; Zhainazarov and Ache, 1995). Kleene (1997) was the first to point out that, because of its small unitary current and high open probability, Ca-activated Cl channels can amplify the CNG current with little amplification of its current noise. The result is an increase in gain without an increase in noise. Our finding that each Cl channel monitors a rise in Ca²⁺ concentration contributed from multiple CNG channels means that some of the noise from individual CNG channels will be removed by spatial integration. In conclusion, the high Cl fraction of the rat olfactory receptor current is based on a roughly eightfold excess of Cl channels over CNG channels. The low-noise amplification results from smoothing of local [Ca²⁺] fluctuations by summation of Ca profiles that originate from several CNG channels within a few hundred nm distance of each Cl channel.

J. Reisert is indebted to Dr. R.C. Thomas of the Physiological Laboratory, Cambridge, UK for assistance in determining the free Ca²⁺ concentrations in the solutions used.

This work was supported by the Deutsche Forschungsgemeinschaft (SPP 1025).

Olaf S. Andersen served as editor.