Elevated levels of carbon dioxide increase lung ventilation in Helix aspersa. The hypercapnic response originates from a discrete respiratory chemosensory region in the dorsal subesophageal ganglia that contains CO2-sensitive neurons. We tested the hypothesis that pH-dependent inhibition of potassium channels in neurons in this region mediated the chemosensory response to CO2. Cells isolated from the dorsal subesophageal ganglia retained CO2 chemosensitivity and exhibited membrane depolarization and/or an increase in input resistance during an acid challenge. Isolated somata expressed two voltage-dependent potassium channels, an A-type and a delayed-rectifier-type channel (IKA and IKDR). Both conductances were inhibited during hypercapnia. The pattern of voltage dependence indicated that IKA was affected by extracellular or intracellular pH, but the activity of IKDR was modulated by extracellular pH only. Application of inhibitors of either channel mimicked many of the effects of acidification in isolated cells and neurons in situ. We also detected evidence of a pH-sensitive calcium-activated potassium channel (IKCa) in neurons in situ. The results of these studies support the hypothesis that IKA initiates the chemosensory response, and IKDR and IKCa prolong the period of activation of CO2-sensitive neurons. Thus multiple potassium channels are inhibited by acidosis, and the combined effect of pH-dependent inhibition of these channels enhances neuronal excitability and mediates CO2 chemosensory responses in H. aspersa. We did not find a single “chemosensory channel,” and the chemosensitive channels that we did find were not unique in any way that we could detect. The protein “machinery” of CO2 chemosensitivity is probably widespread among neurons, and the selection process whereby a neuron acts or does not act as a respiratory CO2 chemosensor probably depends on the resting membrane potential and synaptic connectivity.
- carbon dioxide
elevated co2 stimulates lung ventilation, and central CO2-chemosensitive neurons mediate, in part, the ventilatory response to hypercapnia. Respiration in Helix aspersa is also regulated as a function of Pco2, and the CO2 sensitivity of the snail approaches that of terrestrial mammals (24). Exposure of the intact snail to 3–5% CO2 increased opening of the pneumostome, a muscular aperture that controls access of air to the gas exchange surface within the mantle cavity. The effect of hypercapnia on pneumostomal opening was recapitulated by focal application of hypercapnic saline to a discrete region of the dorsal subesophageal ganglia in the central nervous system of the snail. This respiratory CO2-chemosensitive region contained several repetitively firing CO2-excited neurons that were synaptically coupled to the pneumostome (25). However, the electrical activity of only a fraction of CO2-excited neurons within the chemosensitive area influenced pneumostomal opening. Moreover, CO2-excited neurons were found outside the chemosensitive region, but these neurons were not synaptically coupled to the pneumostome. Thus a subset of CO2-chemosensitive cells in the chemoreceptor region appeared to fulfill the criteria of bona fide respiratory chemoreceptors. The excitatory response to hypercapnia was associated with an increase in input resistance (25), suggesting that neuronal CO2 chemotransduction involved inhibition of a resting conductance. In the foregoing studies, we developed an operational definition of a respiratory CO2-chemosensitive neuron that required that two criteria be fulfilled: first, chemosensory cells had to increase the frequency of action potential (AP) firing during hypercapnic exposure in the absence of synaptic input, and second, respiratory chemoreceptor activity had to be synaptically connected to motor neurons controlling pneumostomal activity. We found respiratory chemosensory neurons. We also found neurons, however, that were CO2 sensitive, but not respiratory chemoreceptors; their activity did not modulate pneumostomal activity (25). In studies of isolated cells, we require only that CO2-sensitive cells be activated or inhibited by exposure to CO2 in the absence of synaptic activity, and we consider any significant change in membrane potential (even without AP formation) as evidence of CO2 sensitivity.
The ventilatory response to CO2 of terrestrial pulmonate snails shares many features with vertebrates, and snails are, therefore, a useful model system in which to examine the cellular mechanisms of CO2 sensitivity. The chemosensory process in mammalian CO2-excited neurons seems to involve H+-dependent potassium channel inhibition (14, 15, 27, 39, 46). Potassium channel inhibition leads to membrane depolarization and/or increased electrical excitability of respiratory-related chemosensory neurons. Therefore, we examined the ionic mechanism of central CO2 chemoreception in CO2-excited neurons of H. aspersa. We believe that similarities between invertebrate and invertebrate chemosensory systems extend to the cellular level, and we tested the hypothesis that pH-dependent potassium channel inhibition also mediates neuronal CO2 chemosensitivity in H. aspersa. We used sharp electrodes to study CO2 sensitivity in neurons in situ in the subesophageal ganglia, and we used voltage- and current-clamp methods to study the activity of specific potassium channels in acutely isolated individual neurons.
MATERIAL AND METHODS
Snails and dissection.
Juvenile H. aspersa (1–2 g) were purchased from Pennsylvania Snail Farm (Meyersdale, PA) and housed for no longer than 4 wk at room temperature (22–24°C) in a humidified terrarium. Snails were fed lettuce, cucumber, and corn meal two to three times per week. Before dissection, snails were anesthetized by injection with 2 ml of the following solution (in mM): 30 MgCl2, 42 CaCl2, and 12 HEPES-free acid, pH 7.8. After the snail was anesthetized, the circumesophageal nerve ring was removed and pinned dorsal-side-up in a 60-mm petri dish containing “snail” saline with the following composition (in mM): 85 NaCl, 4 KCl, 7 CaCl2, 5 MgCl2, 10 glucose, and 10 HEPES, pH 7.8 or 7.3 (see below). The outer connective tissue layer covering the subesophageal ganglia was removed with fine-tipped forceps, microscissors, and tungsten microscalpels (13). The thin inner capsule was loosened by proteolytic digestion (2–5 mg/ml Pronase in snail saline for 15–30 min; Sigma, St. Louis, MO) and subsequently removed by manual dissection. All salts and reagents were purchased from Sigma and were reagent grade.
Electrophysiological measurements in the isolated subesophageal ganglia.
For in situ electrophysiological recordings, the ventral half of the subesophageal ganglia was removed to enhance visualization of nerve cell bodies on the dorsal surface. Isolated dorsal subesophageal ganglia were pinned in a 2-ml recording chamber dorsal-side-up and continuously superfused with snail saline at a rate of ∼10 ml/min. Dorsal nerve cell bodies were viewed with a Nikon E600FN microscope equipped with differential interference contrast optics through a ×10 Plan Fluor objective lens.
The transmembrane potential (Vm) difference and input resistance (Rin) were recorded with conventional sharp microelectrodes. Microelectrodes were pulled on a Flaming-Brown electrode puller (model P-87; Sutter Instrument, Novato, CA) from capillary glass containing a filament (cat. no. 602000; A-M Systems, Sequim, WA). Electrodes were back-filled with 3 M KCl and had resistances between 40 and 80 MΩ when immersed in snail saline. Vm was recorded in bridge mode with an Axoclamp 2B amplifier (HS-2A/0.1 LU head stage; Axon Instruments, Foster City, CA). Rin was recorded in discontinuous current clamp (duty cycle = 10 kHz) to avoid errors in measurement of Vm during current injection due to the voltage drop across the microelectrode tip. Rin was estimated using Ohm's law from the change in Vm resulting from hyperpolarizing current steps.
Whole cell patch-clamp recording from isolated neurons.
To isolate neuronal somata for patch-clamp recording, neurons from the right parietal, left parietal, and visceral ganglia were axotomized 50–200 μm below the cell body with microscissors and transferred directly to a ∼150-μl recording chamber (Warner Instruments, Hamden, CT) with a “cell sucker” (21). Viable somata adhered tightly to clean glass coverslips within 20 min, whereas nonviable somata did not attach and were washed from the chamber after starting bath perfusion (flow rate = 3–5 ml/min). Cells were allowed to recover for at least 1 h before initiating experiments.
Vm and whole cell currents were recorded using the whole cell configuration of the patch-clamp technique. Patch pipettes were fabricated on a two-stage puller (model BB-CH; Mecanex, Geneva, Switzerland) from borosilicate glass capillaries (Corning no. 7052 glass; WPI, Sarasota, FL). Patch electrodes had resistances between 1 and 10 MΩ when filled with the following solution (in mM): 5 NaCl, 70 potassium aspartate, 15 KCl, 1 CaCl2, 2 MgCl2, 5 EGTA, 5 Tris-ATP, and HEPES-free acid, pH 7.4; the osmolality was ∼180 mosmol/kgH2O. Access resistance, which was calculated from capacitance current transients, ranged from 3 to 15 MΩ (mean = 7.9 ± 0.4 MΩ; n = 67) and was compensated electronically by up to 70% before initiating voltage-clamp experiments.
Vm and Rin were recorded in bridge mode. Pipette tip potentials were zeroed with a bridge circuit before and after achieving the whole cell configuration to minimize errors in measurement of Vm during hyperpolarizing current steps. Only somata with resting membrane potentials below −25 mV were considered viable and used for experiments. Greater than 75% of isolated neurons met this criterion. Current signals were recorded in continuous single electrode voltage-clamp mode, filtered with an 8-pole Bessel low-pass filter (Frequency Devices, Haverhill, MA) at a cutoff frequency of 5 kHz, and sampled at 10 kHz with a Digidata 1320A A/D converter (Axon Instruments). All current measurements were normalized by the capacitance of the cell and expressed as picoamperes per picofarad. Before initiating voltage-clamp experiments, the voltage-clamp circuitry was tuned using a square-wave current input and adjusting the clamp multiplier gain, phase lag, and series resistance compensation until Vm also changed as a square wave.
We studied the soma of axotomized neurons to avoid space clamp limitations. Isolated neurons typically resorbed their axonal stub, forming a nearly spherical, isopotential compartment that was adequately voltage clamped with a single patch electrode. Hyperpolarizing current steps evoked monoexponential changes in Vm (see Fig. 4), indicating that isolated somata were under good spatial control during voltage-clamp experiments (6). Isolated neurons had large Rin (average Rin = 490.0 ± 36.8 MΩ, n = 75); therefore, leak currents were not subtracted from whole cell current records. Data were recorded to a computer for subsequent analysis using pCLAMP 8 software (Axon Instruments).
Separation of outward currents.
We established voltage-clamp routines to separate the potassium currents expressed in H. aspersa neurons. Clamping to test potentials from a holding potential of −100 mV elicited a mixed current from which three outward currents could be dissected. A transient A-type potassium current (IKA), which has been characterized previously (3, 17, 37, 44), was activated at test potentials above −50 mV and was the only time-dependent current activated between −40 and −10 mV. Therefore, IKA was studied in isolation from the other outward currents between these voltages. Rapid inactivation around the resting potential is characteristic of IKA (29), and by including a 200-ms inactivation pulse to −30 mV, we removed IKA from the outward current. With IKA inactivated, clamping to depolarized test potentials selectively activated two biophysically and pharmacologically distinct delayed outward currents: a slowly inactivating delayed-rectifier potassium current (IKDR) and a noninactivating current. This noninactivating current had biophysical and pharmacological properties that were indistinguishable from those of the proton current (IProton) described previously in snail neurons (8). The threshold of activation of IProton was approximately −10 mV, and IProton had a similar activation time course as IKDR, but smaller amplitude. Therefore, IProton was hidden by the comparatively larger IKDR. To separate IKDR and IProton, IProton was selectively activated from a holding potential of −30 mV, which was maintained long enough to inactivate both IKA and IKDR completely, and IProton was digitally subtracted from the mixed outward current trace containing both IKDR and IProton to yield pure IKDR.
Bath acidification (pH 7.3) was achieved with either CO2/HCO3−-buffered or fixed acid (i.e., CO2-free HEPES-buffered) snail saline. HCO3−-buffered solutions had the following composition (in mM): 85 NaCl, 4 KCl, 7 CaCl2, 5 MgCl2, and 20 NaHCO3. Normocapnic (∼2.5% CO2, pH 7.8) or hypercapnic (∼5% CO2, pH 7.3) saline was generated by bubbling bicarbonate-buffered snail saline with an air-CO2 mixture and titrated with HCl until the appropriate pH was achieved. These CO2 and pH changes encompass the normal range of variation in CO2 in snails (4). The effect of hypercapnia on potassium channel activity was measured 5 min after switching to hypercapnic saline to ensure that extracellular and intracellular pH (pHi) had reached a steady-state level. The effect of pharmacological inhibitors of potassium channels was measured in HEPES-buffered snail saline (pH 7.8), except where indicated otherwise. Tetraethylammonium chloride (TEA) was substituted for NaCl on an equimolar basis to prevent changes in solution osmolarity or ionic strength. 4-Aminopyridine (4-AP) was dissolved in water and added to the snail saline. The osmolarity of all solutions was checked routinely and adjusted to 225 ± 5 mosmol/kgH2O with dry sucrose.
Data are presented as means ± SE. Statistical significance was determined using Student's t-test for paired comparisons or an ANOVA as appropriate. When the ANOVA indicated that differences existed among treatment conditions, specific comparisons were made using the Bonferroni correction. P values ≤0.05 indicated statistical significance.
CO2-excited neurons are widely distributed in the subesophageal ganglia.
We previously identified respiratory-related CO2-excited neurons within the CO2-chemosensitive area of the subesophageal ganglia (25), but the distribution of CO2-sensitive neurons outside the chemosensitive region was not evaluated systematically. To determine whether CO2-excited neurons existed only in the anatomically restricted chemosensitive region, we mapped the location of CO2-excited, CO2-inhibited, and nonchemosensitive neurons in the subesophageal ganglia by recording spontaneous or evoked activity under normocapnic and hypercapnic conditions. CO2-inhibited neurons exhibited membrane hyperpolarization and/or a reduction in electrical activity during hypercapnia. Nonchemosensitive neurons exhibited no electrical response to hypercapnia. No attempt was made to block synaptic transmission during in situ recordings, since modified divalent cation synaptic blockade solutions can have unpredictable effects on synaptic transmission (30) and cellular excitability in H. aspersa (19).
The average resting Vm of neurons in the subesophageal ganglia was −55.7 ± 1.4 mV (n = 54). Greater than 90% of the neurons studied were electrically silent and required stimulation with depolarizing current to evoke all-or-none AP. Once the level of depolarizing current necessary to evoke AP was established in a neuron, this level of current injection was used consistently in all subsequent studies of the neuron. Three general types of spontaneous or evoked activity were observed in the control normocapnic condition (2% CO2, pH 7.8). Fifty-two percent (28 of 54 cells) of neurons exhibited sustained repetitive spiking (Fig. 1A), whereas 28% (15 of 54 cells) showed strong spike adaptation over a period of several seconds (Fig. 1B; note the rapid cessation of firing, adaptation, despite the persistent hypercapnic stimulus). The remaining 20% (11 of 54 cells) of neurons exhibited clusters of spikes separated by periodic hyperpolarization (Fig. 1C). These last two electrophysiological phenotypes are consistent with expression of a calcium-activated potassium current (IKCa), which dampens spike generation after an activity-induced increase in intracellular calcium (36).
Of the 54 neurons studied, 47 (87%) were excited by increased CO2. The excitatory response to hypercapnia varied with the type of electrical activity exhibited by a neuron during normocapnia. Figure 1A shows a representative recording from a repetitively spiking, unnamed neuron in the “chemosensitive region” of the subesophageal ganglia. The spiking frequency of this cell at pH 7.8 was ∼ 0.7 spikes/s. Decreasing bath pH to 7.3 increased the frequency of firing to ∼4 spikes/s. In strongly adapting neurons, hypercapnia delayed the onset of adaptation but never completely prevented it. Under normocapnic conditions, depolarization with 2 nA of positive current evoked a single spike followed by hyperpolarization and cessation of firing. However, at pH 7.3, neuron F77 fired three spikes before spike adaptation occurred (Fig. 1B). Intermittent spiking neurons always showed an increase in spike frequency, although the effect of hypercapnia on the pattern of bursts of spikes varied from cell to cell. For example, in one of six snails, sustained depolarization of neuron F76 evoked clusters of spikes separated by periodic hyperpolarization (Fig. 1C). In addition to increasing spike frequency in this intermittently spiking neuron, raising bath CO2 from ∼2.5 to 5% increased the number of spikes per cluster from 14.6 ± 0.6 to 24.9 ± 0.8 spikes/cluster and decreased the duration of the postcluster hyperpolarization from 19.4 ± 0.7 to 12.6 ± 0.4 s. The hyperpolarization between spikes has been ascribed to the operation of IKCa (35), and we attribute this effect of hypercapnia on the pattern of firing to pH-mediated inhibition of IKCa. In some intermittently spiking neurons, hypercapnia increased spike frequency in each cluster but failed to change the duration of hyperpolarization. Thus most neurons were stimulated by hypercapnia, although the excitatory response was strongly influenced by the apparent operation of IKCa. The pattern of CO2 responsiveness did not vary as a function of the neuronal location within the subesophageal ganglia; we saw all patterns of activity represented throughout the ganglia.
Only six neurons were found to be nonchemosensitive. This included the giant neuron F1 (Fig. 2), which exhibited spontaneous repetitive spikes and was completely insensitive to hypercapnia. We observed only one neuron that was inhibited during hypercapnia. This cell hyperpolarized during hypercapnia and lost the ability to fire AP even during strong depolarization. Figure 2 summarizes the approximate location of CO2-sensitive and -insensitive neurons in the subesophageal ganglia examined in this study. Many CO2-sensitive neurons where found outside the respiratory chemosensory area defined previously (24).
Effect of hypercapnia on AP waveform.
CO2-dependent changes in the AP waveform can provide insight into the ionic mechanism of neuronal CO2 chemoreception. Hypercapnia prolonged spike duration and decreased the amplitude of the peak afterhyperpolarization (AHP) following an AP in all CO2-excited neurons regardless of the AP firing pattern (Fig. 3), but not in neuron F1 or other nonchemosensitive cells. For example, hypercapnia broadened the average spike duration from 5.1 ± 0.6 to 7.3 ± 0.8 ms in chemosensitive neurons (P < 0.05; n = 7) but had no significant effect on nonchemosensitive neurons (n = 6). Spike broadening during hypercapnia did not result from high-frequency firing (1, 2, 34), since broadening was still observed in both slowly firing neurons and neurons in which the spike frequency was slowed by hyperpolarizing current injection. Similarly, increasing bath CO2 depolarized the peak AHP from −57.0 ± 2.8 to −48.8 ± 2.3 mV in chemosensitive neurons (P < 0.05; n = 7) but did not significantly affect the AHP in nonchemosensitive cells (n = 6). Finally, Rin increased in every CO2-sensitive cell studied, and the increase in Rin was similar regardless of the AP firing pattern. These data, obtained from sharp electrode recordings, are consistent with the hypothesis that inhibition of a voltage-dependent potassium conductance by hypercapnia mediated CO2 chemosensitivity and AP repolarization.
General observations of isolated neurons.
To examine the role of specific conductances in CO2 sensitivity, we performed voltage- and current-clamp experiments in neurons isolated from the right parietal, left parietal, and visceral ganglia. The effect of hypercapnic acidosis on Vm was measured only when neurons exhibited sustained repetitive spiking, and we frequently had to inject modest amounts of current to obtain repetitive spiking activity in isolated cells. The effects of acidosis on Rin and the characteristics of the individual conductances that we identified were studied in excitable and nonexcitable neurons, but these variables did not differ between excitable and nonexcitable neurons, and results from these populations of neurons were pooled.
The mean resting Vm of isolated neurons was −42.6 ± 1.3 mV (n = 98), which is more depolarized than neurons in situ. The average Rin and membrane capacitance were 490.0 ± 36.8 MΩ (n = 75) and 90.3 ± 6.2 pF (n = 72), respectively. Similar to neurons in situ, resting Vm was hyperpolarized below the threshold for inward currents, and depolarizing current injection was required to evoke activity. Approximately one-half of somata exhibited repetitive spikes during current injection. Intermittent or strongly adapting spike activity was never observed in isolated neurons, which suggests that IKCa was inactive in patch-clamped neurons. Chelation of intracellular calcium by EGTA in the pipette solution probably suppressed the activity of IKCa (40).
Hypercapnia increases excitability by inhibition of a pH-sensitive potassium current.
Hypercapnia significantly enhanced excitability in every excitable soma membrane studied (n = 26). Figure 4A shows a representative trace of the effect of hypercapnia on evoked activity in an isolated neuron, and Fig. 4B summarizes this effect. Hypercapnia increased AP frequency from 20 to 200% in these cells (P < 0.05). Hypercapnic saline typically caused a 2- to 10-mV depolarization of Vm, and the magnitude of depolarization increased as Vm was depolarized away from the reversal potential for potassium (calculated EK = −77 mV). This latter observation suggested that CO2 enhanced excitability by inhibiting a potassium-selective conductance open below spike threshold. Hypercapnia increased Rin by ∼48% (control Rin = 468 ± 86 MΩ, hypercapnic Rin = 694 ± 147 MΩ; n = 23, P < 0.05) and caused spike broadening, as had been observed in CO2-excited neurons studied in situ.
Nominally CO2-free acidic saline (pH 7.3) mimicked the effect of hypercapnia and increased the spike frequency, broadened the AP waveform, and increased Rin in the isolated cells (n = 21), suggesting that protons, and not CO2, per se, mediated the excitatory effect of hypercapnia. Surprisingly, the effect of extracellular acidification by CO2-free saline on spike frequency was more rapid than the response to hypercapnic saline. Changes in excitability occurred within 2–5 s of switching the superfusate, which we believe is too fast to be explained by a secondary intracellular acidification, whereas hypercapnia required ∼10 s to activate the cells (18). The effect of CO2-free acidic saline on pHi is relatively slow and requires minutes, not seconds, to change pHi in snail neurons (J. Pfeiffer and J. S. Denton, unpublished observations). Thus the effect of acidic HEPES-buffered saline on excitability is the result of acidification of the extracellular space and extracellular inhibition of a potassium conductance that contributes to the resting conductance (see below).
Voltage-dependent properties of IKA and IKDR.
We characterized the voltage dependence of channel activation and steady-state inactivation for IKA and IKDR in isolated soma membranes before analyzing the effect of hypercapnia on these variables. The characterization of IKA is summarized in Fig. 5, A and B. IKA activated around −50 mV and increased sharply as the membrane was depolarized. The half activation of IKA between −50 and 100 mV was 18.0 ± 1.4 mV (n = 13). The fraction of IKA available for activation depends critically on the membrane potential before activation and decreases as the membrane is depolarized. The voltage-clamp protocol used to evaluate the steady-state inactivation characteristics of IKA and the inactivation curve are also shown in Fig. 5, A and B. The average voltage at which 50% of IKA was available for activation (i.e., VH) was −62.5 ± 1.0 (n = 13), a value that agrees well with previously published results (3, 37) for IKA in H. aspersa. The activation and inactivation curves for IKA overlap between −50 and −10 mV. This window current, which encompasses Vm, represents a range of voltages over which IKA is tonically active. Therefore, IKA probably contributes to the resting potential of these cells (32).
The voltage dependence of channel gating for IKDR is shown in Fig. 5, C and D. IKDR activated around 0 mV, and current amplitude increased as the membrane potential was further depolarization. The half-activation voltage between −10 and 100 mV was 32.3 ± 1.3 mV (n = 12). The current-voltage relationship for IKDR does not exhibit a region of negative conductance at positive test potentials. An “N-shaped” current-voltage relationship has been attributed to the activity of IKCa in these cells (36), and our failure to identify such a current-voltage relationship provides further evidence that IKCa was absent or suppressed in the neurons we studied. The availability of IKDR for activation was also voltage dependent. A representative family of IKDR evoked at 60 mV from a 15-s holding potential at the voltages shown (inactivation was much slower in IKDR than in IKA) in Fig. 5C, and the inactivation curve is shown in Fig. 5D. The half-inactivation voltage for IKDR was −19.3 ± 0.5, which is significantly more depolarized than the half-inactivation voltage of IKA. Furthermore, the window current of IKDR was well above the resting Vm of these neuronal somata.
Hypercapnia inhibits IKA and IKDR in a voltage-dependent manner.
Based on the studies of neurons in situ in the subesophageal ganglia, we expected hypercapnia to inhibit one or more of the potassium currents expressed in these neurons. Therefore, we analyzed the effect of hypercapnia on IKA and IKDR using a repeated-measures ANOVA with two factors (pH levels and test potentials). Increasing bath CO2 from ∼2.5 to 5% inhibited IKA (Fig. 6, A and B). The slope of the current-voltage relationship was significantly less during hypercapnia (P < 0.001), and specific comparisons at each test potential indicated that peak IKA was reduced during hypercapnia compared with normocapnia at all test potentials more positive than −30 mV. The slope of the current-voltage relationship of IKDR was also significantly less during hypercapnia (Fig. 6, C and D; P < 0.001). Specific comparisons at each test potential indicated that the peak IKDR was reduced during hypercapnia compared with normocapnia at all test potentials >10 mV. Block of IKA and IKDR during hypercapnia was not specific to hypercapnic acidosis. Both IKA and IKDR were inhibited by acidification with CO2-free acidic saline (18). For example, peak IKA measured at 0 mV was reduced by 26.5 ± 4% during bath acidification with CO2-free HEPES-buffered saline (n = 4), and reducing the bath pH from 7.8 to 7.3 decreased peak IKDR at 50 mV by 33.5 ± 10% (n = 5). Thus H+ and not CO2, per se, appeared to mediate the inhibitory effect of hypercapnia on IKA and IKDR.
Mechanism of potassium channel block by hypercapnia.
We examined the voltage dependence of channel block during hypercapnia to explore the mechanism by which IKA and IKDR were inhibited during hypercapnic acidosis. If H+ traverse a fraction of the membrane potential to reach a blocking site within the ion conduction path, then the potency of channel block should be voltage dependent (29, 47). Inhibition of both IKA and IKDR was voltage dependent (Fig. 7). The fractional block of IKA was significantly reduced at depolarized potentials in 6 of 13 cells (Fig. 7A, top; P = 0.017 using a repeated-measures ANOVA). This latter observation is consistent with block of the outer pore of IKA by a positively charged ion, probably H+ (29, 47). In the remaining cells (n = 7), IKA was blocked with lower affinity, and, interestingly, the voltage dependence of block suggested that the proton block occurred at the inner side of the conduction pore in these cells (Fig. 7A, bottom; P < 0.002 using a repeated-measures ANOVA). The voltage dependence of block of IKDR was not a linear function of the test potential, but a repeated-measures ANOVA indicated that the percent block of IKDR diminished significantly as the test potential was increased (P < 0.05; n = 13), and the predicted site of proton block of IKDR was extracellular in all cells studied.
Hypercapnia may also decrease ion conduction through voltage-dependent channels by reducing the fraction of channels available for activation. Therefore, we examined the effect of hypercapnia on the steady-state inactivation curves for IKA and IKDR. Hypercapnia had no significant effect on the half-inactivation voltage for IKA. VH for IKA under normocapnic and hypercapnic conditions were −67.1 ± 0.6 mV and −66.4 ± 1.0 mV, respectively (n = 13). However, hypercapnia caused a significant leftward shift in the steady-state inactivation curve for IKDR from −20.5 ± 0.6 mV at pH 7.8 to −26.9 ± 0.9 mV at pH 7.3 (n = 13, P = 0.008). This indicates that hypercapnia reduced the fraction of IKDR available for activation.
Selective inhibition of IKA or IKDR increases excitability.
We examined the effect of 4-AP and TEA on IKA and IKDR. Unfortunately, these agents did not specifically inhibit either IKA or IKDR. Bath-applied 4-AP reversibly blocked IKA with the highest potency (IC50 = 0.98 mM; n = 6), but 1 mM 4-AP also inhibited IKDR and IProton by 20.8 ± 3.8 and 28.4 ± 9.0% (n = 6). At a concentration of 200 μM, 4-AP had no effect on Rin and inhibited IKA exclusively, but the inhibition was modest, only 14.4 ± 6.3% (n = 6). IKDR exhibited the greatest sensitivity to extracellular TEA (IC50 = 2.2 mM; n = 6) and was selectively blocked by 46.4 ± 4.3 and 71.0 ± 7.2% by 2 and 5 mM TEA, respectively (n = 6 and 14). TEA was also nonselective at higher concentrations (17, 38).
Selective, but partial, inhibition of IKA by 200 μM 4-AP increased spike frequency in isolated somata (3/5 cells) and neurons in situ (4/9 cells; Fig. 8A). AP frequency in control saline was 78 spikes/s in control saline, whereas it increased to 102 spikes/s in the presence of 200 μM 4-AP. This neuron exhibited periodic high-frequency clusters of spikes in the presence of 4-AP, which may reflect use-dependent block of IKA by 4-AP (12, 48). Similar high-frequency spindles were observed in one neuron in situ. 4-AP did not depolarize Vm, increase Rin, or significantly prolong AP duration (AP half-width in control and 200 μM 4-AP-containing saline was 5.4 ± 2.6 and 5.8 ± 2.6 ms, respectively; P = 0.09, n = 6). The lack of an effect of 4-AP on AP morphology may reflect the low concentration of blocker that we were required to use to avoid nonspecific effects of the drug.
Inhibition of IKDR by TEA caused significant membrane depolarization and increased activity in 12 of 13 neurons in situ. Figure 8B shows the excitatory effect of 5 mM TEA on neuron F1 (30). TEA caused an 8- to 14-mV membrane depolarization in 4/18 somata (2 mM TEA) and a 4- to 15-mV membrane depolarization in 10/18 somata (5 mM TEA) and increased excitability in these cells. TEA (1.5 mM), which inhibits ∼30% of IKDR, increased Rin by 16 ± 6.2% (n = 5). TEA also significantly prolonged AP duration (Fig. 8C). The average spike half-width in control saline was 2.6 ± 0.6 ms, whereas it increased to 4.5 ± 0.8 ms in 5 mM TEA (n = 6 cells; P < 0.01). Thus IKDR contributes to AP repolarization and the resting potassium conductance in H. aspersa neurons. Moreover, TEA treatment mimicked the effect of hypercapnia.
Effect of hypercapnia on other currents.
Although this study focused on the role of potassium channel inhibition in the excitatory response of H. aspersa neurons to hypercapnia, we also found that acidification inhibited IProton. For example, reducing bath pH from 7.8 to 7.3 with hypercapnic saline decreased IProton recorded at 50 mV from 22.3 ± 2.4 to 8.6 ± 1.6 pA/pF (P < 0.01). Unlike block of IKA and IKDR, inhibition of IProton during hypercapnia was voltage independent.
Bath acidification with HEPES-buffered saline from pH 7.8 to pH 7.3 reduced the peak inward current and frequently reduced peak AP amplitude by 5–10% (n > 10). The AP upstroke was predominantly calcium dependent (n = 5), which is consistent with results from a previous study (25). The reduction in spike amplitude suggests that acidification inhibited the calcium channels expressed in H. aspersa neurons. This is consistent with a considerable literature indicating that calcium channels from a variety of cell types, including snail neurons, are inhibited by H+ (9, 16, 31, 41–43).
We investigated the role of potassium channels in neuronal CO2 chemoreception in the pulmonate snail, H. aspersa. CO2-excited neurons were widely dispersed on the dorsal surface of the subesophageal ganglia, and, in all the neurons studied, the process whereby acidic stimuli were transduced into enhanced neuronal excitability involved H+-dependent inhibition of transient IKA and IKDR currents. The evidence for this conclusion is as follows. Hypercapnic or CO2-free acidic saline reduced the resting membrane conductance and depolarized Vm. The degree of membrane depolarization depended on the membrane potential and increased as Vm was depolarized away from EK (−77 mV). Hypercapnia also prolonged AP duration and decreased the amplitude of the postspike AHP. These observations are consistent with inhibition of potassium currents responsible for AP repolarization. Inhibition of IKA or IKDR individually with pharmacological blockers recapitulated some of the effects of acidification on excitability and other membrane properties, suggesting that both currents contribute to acid-sensitive spike timing in H. aspersa neurons. Furthermore, voltage-clamp studies of individual currents in isolated somata demonstrated voltage-dependent proton block of IKA and IKDR and enhanced inactivation of IKDR during hypercapnic exposure. We also found evidence that hypercapnia inhibited IKCa and prolonged the duration of bursts of AP. Hypercapnic inhibition of IKCa may not initiate the response to CO2, but it modulates the duration of pH-dependent activation once it is established. Thus inhibition of multiple potassium channels provides a plausible mechanism whereby changes in pH influence neuronal chemosensory activity.
Limitations of the methods.
In studies of isolated neurons, we examined hypercapnic responses only in neurons capable of sustaining regenerative spike activity. This criterion may have biased the results we obtained. We feel this is unlikely. First, CO2-sensitive neurons were widely distributed in the subesophageal ganglia so that a random sample of cells isolated from these ganglia still contained a significant number of CO2-sensitive cells. Second, when we studied Rin and activation and inactivation curves of IKA and IKDR in neurons that were capable of regenerative spike activity and those that were incapable of regenerative spiking, the results were identical. The failure of some neurons to display regenerative spiking may have reflected reduced calcium and/or sodium channel density or function as a result of the isolation procedure, but the function of potassium currents seemed to be similar in these two populations.
Once current injection elicited spiking activity in isolated somata, these neurons were more sensitive to acidic stimulation than neurons in situ. We believe that this results from the following factors. First, the Rin of isolated somata was up to 20 times higher than the Rin of neurons in situ, which ranged from 20 to 150 MΩ. In isolated neurons, therefore, a smaller change in potassium current was required to elicit a given change in excitability compared with neurons in situ. Second, isolated neurons were more depolarized than neurons in situ (Vm = −42.6 vs. −55.7 mV, respectively). Because the driving force for current flow through potassium channels increases in proportion to the difference, (Vm − EK), the effect of potassium current inhibition will increase as Vm is depolarized away from EK. Moreover, current flow through IKA will increase as Vm approaches the activation voltage for IKA (approximately −50 mV). Thus inhibition of IKA in depolarized neurons will have a greater effect on membrane current and, hence, membrane excitability than in hyperpolarized neurons. Third, approximately one-half of the neurons studied in situ exhibited functional evidence of IKCa expression, and the excitatory response to hypercapnia was antagonized by the operation of IKCa (Fig. 1, B and C), although hypercapnia also seemed to reduce this effect of IKCa. However, IKCa was suppressed in patch-clamped neurons, presumably because of strong intracellular calcium buffering by EGTA (40). In the absence of IKCa in the isolated somata, neurons exhibited only repetitive spiking without spike adaptation. Thus we believe that the mechanisms of CO2 sensitivity identified in isolated somata and the in situ preparation of the dorsal subesophageal ganglia accurately reflect the mechanisms active in intact animals.
The foregoing discussion of studies in isolated somata is relevant to the other major limitation of our study. We cannot prove that any particular isolated somata is actually a respiratory CO2 chemosensory cell; we can only demonstrate the mechanism(s) of its CO2 sensitivity; the same limitation exists in vertebrate preparations. We used a semi-intact preparation to demonstrate that respiratory CO2-sensitive neurons actually modified pneumostomal activity, and we demonstrated that these cells were intrinsically CO2 sensitive (they retained pH sensitivity when synaptic activity was blocked; see Ref. 25). The semi-intact preparation allowed us to establish the required proof of respiratory connectivity but precluded study of the ionic basis of CO2 sensitivity (we could not establish adequate space clamp in the semi-intact preparation). Nonetheless, the ionic currents present in isolated somata were remarkably consistent among cells, and the behavior of the ionic currents identified in the isolated somata was consistent with the pH responses of neurons in the semi-intact preparation. IKA and IKDR were present in virtually every soma that we studied, and IKCa was present in at least one-third of the cells that we studied in situ. Therefore, it seems likely that we studied the ionic mechanisms of respiratory CO2 chemosensitivity even if we could not identify each cell as a respiratory chemoreceptor.
What is the identity of the pH-sensitive resting potassium conductance?
We believe that much of the excitatory effect of acidification in snail neurons was mediated by inhibition of a resting potassium conductance, which in turn depolarized Vm and increased the rate at which Vm reached the threshold for AP. Inhibition of IKA is a likely candidate mechanism for increased neuronal excitability. The activation and steady-state inactivation curves for IKA overlap between −50 and −10 mV (Fig. 5), and this location of this “window current” with respect to Vm suggests that IKA contributes to the resting potassium conductance. The fraction of resting membrane conductance contributed by IKA can be estimated from measurements of activation/inactivation characteristics of IKA and the biophysical properties of each cell. For example, the mean input conductance (i.e., 1/Rin) of isolated somata in this study was ∼2.0 pA/mV. At a typical resting potential of −50 mV, where ∼25% of IKA are available for activation, the mean conductance through IKA was ∼0.8 pA/mV. This indicates that IKA contributes ∼40% of the resting conductance in H. aspersa neurons. Moreover, activation of IKA between spikes short circuits inward currents that would otherwise depolarize Vm to threshold, and, consequently, activation of IKA prolonged the interspike interval.
There are, however, two problems with this interpretation. First, we did not actually find a statistically significant effect of pH on IKA until Vm was greater than or equal to −30 mV, which is well above the resting membrane potential. Second, selective inhibition of IKA with 4-AP increased spike frequency in a subset of neurons (Fig. 8A), and the lack of effect of 4-AP on the remaining cells likely resulted from the low concentration of blocker used to inhibit IKA selectively. However, the nonspecific effects of higher concentrations of 4-AP made it difficult to examine more rigorously the role of IKA in shaping AP morphology. The rapid activation kinetics of IKA in H. aspersa neurons suggest that IKA probably activates fully during the AP upstroke, but it inactivates relatively slowly (>100 ms) compared with the brief duration of a typical AP, which lasts ∼5–10 ms. We believe that IKA contributed to spike repolarization, but the lack of a significant effect of low doses of 4-AP on spike duration does not support this idea. We tried to identify selective blockers of IKA, but the IKA in H. aspersa was resistant to α-, β-, and δ-dendrotoxins, margatoxin, and tityustoxin (Alomone Laboratories, Jerusalem, Israel). Thus the electrophysiological and pharmacological studies do not conclusively prove that IKA initiates the neuronal response to CO2. To overcome these problems, we developed a computer model of a CO2 chemosensory cell (11). Within the computer model, we were able to completely separate the contribution of each channel (IKA, IKDR, and IKCa) to chemosensitivity. The results of the model are consistent with results summarized above; the IKA was the dominant CO2-sensitive channel initiating the chemosensory response. Thus the activation and inactivation characteristics of IKA, the effect of low-dose 4-AP, the effect of hypercapnia on spike frequency, and previous studies of IKA indicate that IKA contributes significantly to CO2 sensitivity, but the electrophysiological and pharmacological evidence that IKA initiates the response to CO2 is supportive, but not conclusive.
The voltage dependence properties of IKDR (Fig. 5) do not predict that this current contributes to the resting potential; the activation voltage is well above the resting membrane potential. However, we observed that selective inhibition of IKDR with TEA depolarized Vm and increased Rin, suggesting that IKDR does actually contribute to the resting potassium conductance. The increase in Rin observed when IKDR was inhibited (∼15%) is significantly lower than that observed during hypercapnia (∼50% increase), which again suggests that acidification inhibits a second resting conductance as well as, probably, IKA. In other CO2-sensitive cells, a two-pore domain potassium channel (i.e., TASK), which constitutes a leak potassium current, has been identified (5). This current is sensitive to extracellular pH (pHe) and has a pKa centered around physiological pH (7, 22). However, we think it is unlikely that a TASK-like channel contributes to the resting conductance in H. aspersa neurons because TASK is insensitive to extracellular TEA (22), and we found no evidence in our voltage-clamp studies of any significant leak current.
IKDR also contributes to the AP shape and frequency in H. aspersa neurons. For example, inhibition of IKDR by ∼75% caused significant spike broadening (Fig. 8C). IKDR also seemed to contribute to the resting Vm even though the window current of IKDR was well above the resting Vm. For example, application of 2–5 mM TEA, a concentration that selectively inhibited IKDR, caused membrane depolarization and increased spike frequency (Fig. 8B), and, even below the window current, ion channels have a small but finite open probability. IKDR contribute to spike repolarization (29), and inhibition of IKDR in vertebrate neurons increases excitability (33). Thus we believe that IKDR also contributes modestly to CO2 sensitivity in H. aspersa neurons.
pH effects on other outward conductances.
In addition to IKA and IKDR, we also found evidence of IProton and IKCa. We do not believe that these conductances act as the CO2/pH sensor that initiates the electrophysiological response to CO2, but pH modulation of these channels may modify the ultimate output of any chemosensor (11). IProton activates only at membrane potentials above −10 mV; it activates slowly compared with the duration of the AP; it is a relatively small conductance; and it is inhibited by hypercapnia. However, we do not believe that it plays a significant role in determining the resting membrane potential or a role in CO2 chemosensitivity. We did not study IKCa intensively, but it seemed to be inhibited by acidosis (Fig. 1C). Activation of IKCa depends on calcium entry, which requires some period of depolarization. Therefore, it seems unlikely that IKCa contributes to the resting membrane potential or that it acts as the CO2 sensor. However, to the extent that activation of IKCa tends to limit spiking activity, pH-dependent inhibition of this channel will tend to sustain any chemosensory response. Thus pH-dependent inhibition of IKCa may contribute to the final pattern of spiking activity of chemosensory neurons, and, in that sense, the channel is chemosensory. Our analysis of IKCa was limited to inferring its role in CO2 sensitivity from sharp electrode studies of neurons in situ in the subesophageal ganglia; our patch-clamp pipette solution contained EGTA and suppressed the activity of IKCa. Evaluation of the role of IKCa in our computer model did, however, confirm that both IProton and IKCa may play a role in CO2 chemosensitivity, particularly IKCa (11). The role of IKCa in the final output of chemosensory neurons highlights an interesting aspect of our findings. The function of chemosensory neurons in H. aspersa depends not on one channel but on interactions within a single neuron of multiple channels (11), a finding supported in vertebrate studies as well (27).
Location of the stimulus.
The location of the chemosensory stimulus (i.e., intracellular vs. extracellular) has been a long-standing issue in the field of central CO2 chemosensitivity (23, 26, 28, 45). Decreasing bath pH with HEPES-buffered, CO2-free saline rapidly stimulated electrical activity in the present study. Solutions of fixed acid are generally thought to be membrane impermeant and, therefore, should be restricted to the extracellular space (10). The rapidity of the response to extracellular acidification suggests that changes in pHe are sufficient to modulate neuronal activity. Furthermore, the pattern of voltage-dependent block of IKDR and a subset of IKA supports the idea that extracellular protons may modulate potassium channel function. Block of IKDR was voltage dependent between 10 and 40 mV and was relieved at test potentials >40 mV. Relief of proton block at higher test potentials is consistent with block of the external side of the pore by a positively charged ion, presumably H+ (47). The pattern of voltage dependence of proton block of IKA was bimodal (Fig. 7A), and, in one group of neurons, high-affinity block of IKA was less as the test potential became more positive, indicating that H+ inhibited this current by blocking the external pore. Taken together, these results suggest that changes in pHe can increase the activity of chemosensory neurons.
We also identified pH-dependent block of IKA that actually increased in a subset of neurons as the test potentials increased (Fig. 7A). This implies that protons blocked IKA more effectively as the Vm became more positive, which is consistent with proton-mediated channel block at the intracellular side of IKA. Strong intracellular acidification also inhibited both IKA and IKDR in Lymnaea stagnalis neurons (9). Thus changes in excitability may also be mediated by changes in pHi. In previous studies of semi-intact snails, intracellular acidification caused greater pneumostome opening than superfusion of the nervous system with fixed acid saline (23), which we interpreted as evidence that pHi was the proximate site of central chemoreception in the snail. However, the current studies suggest that different neurons may respond to pHi and pHe.
In summary, the chemosensory process in H. aspersa originates from pH-dependent inhibition of potassium channels that participate in the maintenance of the resting membrane potential. Once the potassium channels are inhibited, the membrane potential depolarizes and reaches the threshold of activation for inward sodium and calcium currents that generate the AP. We did not find a single “chemosensory channel,” and the chemosensitive channels that we did find were not unique in any way that we could detect; they were typical representatives of IKA, IKDR, and IKCa. Furthermore, virtually every neuron we studied expressed both IKA and IKDR. These findings have significant implications for studies of chemosensory mechanisms in both vertebrates and invertebrates. The protein “machinery” of CO2 chemosensitivity is probably widespread among neurons, and this implies that the selection process whereby a neuron acts or does not act as a chemosensor depends to a large extent on the resting membrane potential and synaptic connectivity (25). Thus synaptic input and expression of other channels and modulators that may determine the membrane potential will determine in turn whether a particular neuron has chemosensory activity. The specific pattern of connectivity with the respiratory pattern generator will determine whether the chemosensor has a respiratory function. We infer from this that the selection process may be quite plastic: a neuron may be chemosensory in one state but insensitive in another depending on the effect of synaptic input on Vm and connectivity. Finally, air breathing has evolved at least four times in different phyla, but in every air-breathing animal in which chemosensory function has been studied, pH-dependent inhibition of potassium channels seems to be the fundamental sensory process. Thus convergent chemosensory mechanisms have evolved using similar basic ionic mechanisms: there has been no evolution of a unique chemosensory channel, but the evolutionary process has repeatedly used the ubiquitous effects of pH on common potassium channels to modulate the activity of individual neurons.
This work was supported by National Heart, Lung, and Blood Institute Grants HL-07449 and HL-51238.
Some of this work was performed at the Marine Biological Laboratory in Woods Hole, MA, where J. S. Denton was a Grass Fellow during the summer of 2001.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2007 the American Physiological Society