Astrocytes in the Retrotrapezoid Nucleus Sense H⁺ by Inhibition of a Kir4.1–Kir5.1-Like Current and May Contribute to Chemoreception by a Purinergic Mechanism

Abstract

Central chemoreception is the mechanism by which CO₂/pH sensors regulate breathing in response to tissue pH changes. There is compelling evidence that pH-sensitive neurons in the retrotrapezoid nucleus (RTN) are important chemoreceptors. Evidence also indicates that CO₂/H⁺-evoked adenosine 5′-triphosphate (ATP) release in the RTN, from pH-sensitive astrocytes, contributes to chemoreception. However, mechanism(s) by which RTN astrocytes sense pH is unknown and their contribution to chemoreception remains controversial. Here, we use the brain slice preparation and a combination of patch-clamp electrophysiology and immunohistochemistry to confirm that RTN astrocytes are pH sensitive and to determine mechanisms by which they sense pH. We show that pH-sensitive RTN glia are immunoreactive for aldehyde dehydrogenase 1L1, a marker of astrocytes. In HEPES buffer the pH-sensitive current expressed by RTN astrocytes reversed near E_K⁺ (the equilibrium potential for K⁺) and was inhibited by Ba²⁺ and desipramine (blocker of Kir4.1-containing channels), characteristics most consistent with heteromeric Kir4.1–Kir5.1 channels. In bicarbonate buffer, the sodium/bicarbonate cotransporter also contributed to the CO₂/H⁺-sensitive current in RTN astrocytes. To test the hypothesis that RTN astrocytes contribute to chemoreception by a purinergic mechanism, we used fluorocitrate to selectively depolarize astrocytes while measuring neuronal activity. We found that fluorocitrate increased baseline activity and pH sensitivity of RTN neurons by a P2-receptor–dependent mechanism, suggesting that astrocytes may release ATP to activate RTN chemoreceptors. We also found in bicarbonate but not HEPES buffer that P2-receptor antagonists decreased CO₂ sensitivity of RTN neurons. We conclude that RTN astrocytes sense CO₂/H⁺ in part by inhibition of a Kir4.1–Kir5.1-like current and may provide an excitatory purinergic drive to pH-sensitive neurons.

INTRODUCTION

A region of the caudal brain stem called the retrotrapezoid nucleus (RTN) contains a population of neurons that are highly pH sensitive in vivo and in vitro, are glutamatergic, and project to respiratory centers to influence breathing (Abbott et al. 2009; Mulkey et al. 2004). The mechanism by which these RTN chemoreceptors sense pH involves inhibition of an unidentified voltage-independent K⁺ conductance (Mulkey et al. 2007). The absence of an identified molecular pH sensor in RTN neurons has raised the possibility that multiple mechanisms contribute to chemoreception (Filosa and Putnam 2003; Pessia et al. 1996) and, despite the apparent intrinsic pH sensitivity of RTN neurons, it remains possible that local paracrine mechanisms contribute to chemoreception by the RTN.

There is increasing evidence that adenosine 5′-triphosphate (ATP), released by astrocytes, is an important mediator of chemoreception. Previous in vivo studies showed that hypercapnia evoked the discrete release of ATP within the RTN (Gourine et al. 2005). Further, application of ATP into the RTN stimulated respiratory output, whereas application of an ATP receptor antagonist (pyridoxal-phosphate-6-azophenyl-2′,4′-disulfonate [PPADS]) to the RTN lowered CO₂ respiratory responses (Gourine et al. 2005). Recently, it was shown that RTN astrocytes respond to bath acidification with Ca²⁺-dependent, exocytosis-mediated, ATP release (Gourine et al. 2010). In addition, photostimulation of RTN astrocytes expressing channelrhodopsin-2 (ChR2) resulted in a robust increase in breathing (Gourine et al. 2010), strongly suggesting that astrocytes contribute to respiratory drive. This study also provided evidence suggesting that ATP release from astrocytes is a requisite signal for neuronal pH sensitivity (Gourine et al. 2010). At the cellular level, we have shown that purinergic signaling can activate pH-sensitive RTN neurons via P2Y receptors (Mulkey et al. 2006). However, we also found that the pH sensitivity of RTN neurons in HCO₃⁻-free HEPES-buffered media was retained when ATP receptors were blocked with PPADS (100 μM) (Mulkey et al. 2004, 2006), indicating that ATP is not a requisite component of neuronal pH sensitivity. These results suggest that RTN neurons are intrinsically pH sensitive but that CO₂-evoked release of ATP can modulate the activity of pH-sensitive neurons and contribute to the integrated output of the RTN (i.e., respiratory drive) during hypercapnia.

The goals of this study were to 1) confirm that RTN astrocytes are pH sensitive, 2) identify mechanism(s) by which these cells sense changes in pH, and 3) determine the extent to which they contribute to chemoreception. To make these determinations, we used the brain slice preparation and a combination of patch-clamp electrophysiology and immunohistochemistry to identity pH-sensitive glia and to determine the mechanism by which they sense changes in pH. To explore the possibility that pH-sensitive RTN astrocytes stimulate activity of pH-sensitive neurons by a purinergic mechanism, we determined the effect of fluorocitrate-mediated astrocyte activation on pH-sensitive neurons. We found that pH-sensitive RTN glia are protoplasmic astrocytes that sense H⁺, in part, by inhibition of a Kir4.1–Kir5.1-like current and may contribute to respiratory drive by increasing activity of pH-sensitive RTN neurons by a purinergic-dependent mechanism.

METHODS

Brain stem slices

Slices containing the RTN were prepared as previously described (Jiang et al. 2001; Kubo et al. 2005; Matthias et al. 2003). Briefly, neonatal rats (7–12 days postnatal) were decapitated under ketamine/xylazine anesthesia and transverse brain stem slices (300 μm) were cut using a microslicer (DSK 1500E; Dosaka, Kyoto, Japan) in ice-cold substituted Ringer solution containing (in mM): 260 sucrose, 3 KCl, 5 MgCl₂, 1 CaCl₂, 1.25 NaH₂PO₄, 26 NaHCO₃, 10 glucose, and 1 kynurenic acid. Slices were incubated for about 30 min at 37°C and subsequently at room temperature in normal Ringer solution (in mM): 130 NaCl, 3 KCl, 2 MgCl₂, 2 CaCl₂, 1.25 NaH₂PO₄, 26 NaHCO₃, and 10 glucose. Both substituted and normal Ringer solutions were bubbled with 95% O₂-5% CO₂.

Electrophysiology

Individual slices were transferred to a recording chamber mounted on a fixed-stage microscope (Zeiss Axioskop FS) and perfused continuously (∼2 ml min⁻¹) with a bath solution composed of (in mM): 140 NaCl, 3 KCl, 2 MgCl₂, 2 CaCl₂, 10 N-2-hydroxyethylpiperazin-N′-2-ethanesulfonic acid (HEPES), 10 glucose; pH was adjusted between 6.9 and 7.5 by addition of HCl or NaOH. To test pH sensitivity of RTN astrocytes in bicarbonate-based buffer slices were perfused with normal Ringer solution bubbled with 95% O₂-5% CO₂ (bath pH = 7.45). The pH of the bicarbonate-based bath solution was decreased to a pH of about 6.8 by bubbling with 15% CO₂ as previously described (Filosa and Putnam 2003).

All recordings were made with an Axopatch 200B patch-clamp amplifier, digitized with a Digidata 1322A A/D converter, and recorded using pCLAMP 10.0 software (Molecular Devices). Recordings were obtained at room temperature with patch electrodes pulled from borosilicate glass capillaries (Warner Instruments) on a two-stage puller (P89; Sutter Instrument) to a DC resistance of 4–7 MΩ when filled with an internal solution containing the following (in mM): 120 KCH₃SO₃, 4 NaCl, 1 MgCl₂, 0.5 CaCl₂, 10 HEPES, 10 EGTA, 3 Mg-ATP, 0.2% biocytin, and 0.3 GTP-Tris (pH 7.2); electrode tips were coated with Sylgard 184 (Dow Corning). Voltage-clamp recordings from glia were made at a holding potential of −80 mV and in the presence of tetrodotoxin (TTX, 0.1 μM) to block neuronal activity. Holding current, conductance, and current–voltage (I–V) relationships were determined using voltage steps between −40 and −150 mV. Neuronal firing rate histograms were generated by integrating action potential discharge in 10-s bins and plotted using Spike 5.0 software. It is important to recognize that in a whole cell recording circuit, recording cells with a low membrane resistance (e.g., passive astrocytes) in series with a large access resistance (R_a) can result in voltage-clamping error. In our experiments, we limited R_a to about 10 MΩ by preparing tissue from young animals; capacitance and R_a compensation (70%) were used to minimize voltage errors as previously described (Matthias et al. 2003). Recordings were discarded if R_a varied >10% during an experiment. A liquid junction potential of 10 mV was corrected off-line. Data were reported as means ± SE and were analyzed by the paired t-test or ANOVA and Dunnett's multiple comparison test when appropriate (P < 0.05).

Immunohistochemistry

Recorded slices were fixed in 4% paraformaldehyde in 0.1 M sodium phosphate buffer (pH 7.4) for 2 h at 4°C and rinsed in phosphate-buffered saline (PBS). Slices were blocked and permeabilized in PBS containing 5% normal goat serum and 0.4% Triton X-100 for 1 h at room temperature. Biocytin was detected by incubation in Alexa 488–conjugated strepavidin (1:1,000; Invitrogen) for 1 h at room temperature. After three PBS washes, the slices were incubated in primary antibodies overnight at 4°C, washed, and then incubated in secondary antibodies for 1 h at room temperature. Slices were washed in PBS and mounted in Vectashield. Images were collected on a Leica TCS SP2 confocal microscope equipped with 488-, 543-, and 633-nm laser lines and tunable emission wavelength detection. For each labeled slice, a biocytin-positive cell was identified and confocal z-stacks were collected sequentially for the other two channels to detect glial antigens. A cell was scored as immunopositive if the immunoreactivity for one or more glial markers was detected in the biocytin-positive cell body and processes. To determine background fluorescence of the tissue and nonspecific binding of secondary antibodies, we tested secondary antibodies without preincubation in primary antibody. These methods have been optimized for 300-μm slices sequentially detecting biocytin and glial antigens.

Drugs

Tetrodotoxin was purchased from Alomone Labs (Bethlehem, Israel); all other chemicals were obtained from Sigma (St. Louis, MO). Fluorocitrate was prepared as previously described (Holleran et al. 2001). Briefly, the Ba²⁺ salt of fluorocitric acid was dissolved in 0.1 M HCl and the Ba²⁺ was precipitated by the addition of Na₂SO₄ (0.1 M). This solution was centrifuged at 800 g for 10 min and the supernatant removed and diluted in HEPES-buffered medium to a final concentration of 100 μM (pH 7.3 by addition of 1 N NaOH). To confirm that Ba²⁺ was adequately removed during the preparation of fluorocitrate, we also prepared a solution of equal Ba²⁺ concentration minus fluorocitric acid and subjected it to the same precipitation protocol described earlier.

RESULTS

This study consists of two series of experiments. First, we made whole cell recordings to identify pH-sensitive RTN glia and to characterize the mechanism by which they sense pH. We initially identified glia by their small size, hyperpolarized membrane potential, and inability to form action potentials during depolarizing current injection; cellular identity was later confirmed using immunohistochemistry. In a second series of experiments, we made loose-patch recordings from pH-sensitive RTN neurons to determine the effects of fluorocitrate-mediated astrocyte activation on baseline activity and pH sensitivity of RTN neurons and to test effects of P2-receptor antagonists on neuronal CO₂ sensitivity.

Astrocyte chemoreceptors

Previous sharp electrode recordings showed that ≤44% of RTN glia respond to extracellular acidification with membrane depolarization (Fukuda et al. 1978). In the whole cell configuration and HCO₃⁻-free HEPES-buffered solution we found that roughly 20% of RTN glia were pH sensitive; cells were considered pH sensitive if they showed >10 mV/pH unit change in membrane potential, ≥40 pA change in holding current, or 35% change in conductance when cycling bath pH between 7.5 and 6.9. Based on these criteria, we found a bimodal distribution of pH-sensitive and -insensitive glial cells (Fig. 1F). In current-clamp, pH-sensitive RTN glia have a resting membrane potential of −80 ± 1 mV (n = 8) and respond to acidification from pH 7.5 to 6.9, with a depolarization of 8.8 ± 1.4 mV (n = 8) that occurs within about 6 min (Fig. 1, A and B). We hypothesize that the slow kinetics of pH-sensitive astrocytes is reflective of their modulatory role and likely helps maintain stable chemoreceptor output despite fast breath-by-breath oscillations in tissue pH (Millhorn et al. 1984). In contrast, pH-insensitive glia have a resting membrane potential of −74 ± 3 mV (n = 12) that did not vary over the same pH range.

Fig. 1.

Characteristics of pH-sensitive retrotrapezoid nucleus (RTN) glia in HEPES buffer. A: traces of membrane potential and bath pH show that acidification from 7.3 to 6.9 causes a reversible depolarization. B: summary of H⁺ effects on membrane potential of pH-sensitive astrocytes (n = 5, *P < 0.05). C: traces of holding current (top) and conductance (bottom) during changes in bath pH show that alkalization increases and acidification decreases outward current and conductance. D and E: current responses to voltage steps from −80 mV to between −40 and −150 mV (E) and corresponding current–voltage (I–V) relationships under alkaline and acidic conditions (D). To isolate the pH-sensitive I–V (♣), we subtract I–V relationships obtained at pH 6.9 (▴) from those recorded at pH 7.5 (■). F: percentage change in conductance during acidification from pH 7.5 to 6.9 observed from all glial cells recorded in the RTN. G: average I–V relationship (n = 5) of pH-sensitive current reverses near the equilibrium potential for K⁺ (E_K⁺, arrow).

After recording, cells were filled with biocytin to characterize morphology and immunohistochemical phenotype. We found that 18 of 20 (90%) pH-sensitive glia were immunoreactive for aldehyde dehydrogenase 1L1 (Aldh1L1) (Fig. 2, A–C), an antigen that is expressed by protoplasmic astrocytes (Cahoy et al. 2008). In addition, pH-sensitive glia typically extended numerous fine radial projections near blood vessels (Fig. 2D) and to be dye-coupled (evidence of gap junctions) to other astrocytes (Fig. 2, A–C). These results support recent evidence that GFAP-expressing cells near the ventral surface are CO₂ sensitive (Gourine et al. 2010) and indicate that pH-sensitive glia are protoplasmic astrocytes.

Fig. 2.

pH-sensitive RTN glia are gap junction coupled protoplasmic astrocytes. A, left: a transverse slice between bregma −11.6 to −11.2 mm that contains the RTN. Right: map showing relative locations of 18 pH-sensitive RTN astrocytes (green circles correspond to the cells shown in the bottom panels). B1 and B2: biocytin filled pH-sensitive RTN glial cells (green, arrows) and dye-coupled cells (other green cells). C1 and C2: aldehyde dehydrogenase 1L1-labeled (Aldh1L1, red) astrocytes. D1 and D2: merged images of the same cells in B and C show double-labeling for biocytin (green) and Aldh1L1 (red); the pH-sensitive astrocytes (arrows) and dye-coupled partners colocalize with Aldh1L1 labeling. E1 and E2: merged images of the same cells in B and C show double-labeling for biocytin (green) and laminin (blood vessels, blue). Note that pH-sensitive RTN astrocytes have cell bodies positioned near blood vessels (E1) and extend processes to blood vessels (truncated arrows in E2). Each image is the average projection of a confocal stack; around 30 images taken every 0.26 μm in the Z plane.

To further characterize electrical properties of pH-sensitive astrocytes and establish their role as intrinsic pH sensors, we made voltage-clamp recordings of whole cell currents in HEPES buffer at control pH 7.3 and during exposure to alkaline and acidic conditions. Note that these experiments were performed in TTX (0.1 μM) to block neuronal activity. Under control conditions, pH-sensitive astrocytes display high K⁺ conductance and exhibit fairly time- and voltage-independent responses to voltage steps. These properties are consistent with those of passive astrocytes described in the hippocampus (Matthias et al. 2003; Zhou et al. 2006) and brain stem (Grass et al. 2004). Astrocytes in the RTN respond to acidification from pH 7.5 to 6.9, with a decrease in holding current of 91.8 ± 10 pA (n = 23) and a decrease in conductance of 6.8 ± 0.8 nS (n = 23) (Fig. 1C). This corresponds with a decrease in whole cell current of about 53% at +40 mV. By comparison, RTN astrocytes are more pH sensitive than astrocytes in a nonrespiratory region like the hippocampus; passive hippocampal astrocytes respond to acidification from 7.4 to 6.0 with a decrease in whole cell current at +20 mV of only about 28% (Zhou et al. 2009). To confirm that the pH sensitivity of RTN astrocytes is determined by intrinsic membrane properties, rather than intercellular current spread through gap junctions, we tested the effects of carbenoxolone (a nonspecific gap junction blocker) on astrocyte pH sensitivity. Carbenoxolone (100 μM) increased the pH-sensitive current by roughly 18% (Supplemental Fig. S1B), suggesting that RTN astrocytes are intrinsically pH sensitive.¹

Astrocytes are known to express high levels of the electrogenic Na⁺/HCO₃⁻ cotransporter (NBC), which can generate an inward depolarizing current (i.e., HCO₃⁻ efflux) during bath acidification (Deitmer 1991; Munsch and Deitmer 1994). To determine whether activity of the NBC contributes to the pH-sensitive current in RTN astrocytes, we characterized CO₂/H⁺ sensitivity of RTN astrocytes in normal Ringer solution (26 mM HCO₃⁻, 5% CO₂). In current-clamp, pH-sensitive RTN astrocytes had a membrane potential of −82 ± 1.2 mV (n = 3), which is about 2 mV hyperpolarized to resting potential of astrocytes in HCO₃⁻-free HEPES buffer, and responded to hypercapnia (15% CO₂, pHo = 6.8) with a membrane depolarization of 9 ± 2.3 mV (Fig. 3, A and B). In voltage-clamp, exposure to 15% CO₂ decreased holding current from 139.8 ± 20 to 54.9 ± 8.9 pA (n = 8), which corresponded with a change in holding current of 130 ± 17 pA/pH unit, whereas in HEPES buffer acidification from 7.3 to 6.9 changed in the holding current 62.7 ± 6 pA/pH unit (Fig. 3G). In addition, the CO₂/H⁺-sensitive current had a reversal potential of −105 mV (Fig. 3E), which is more negative than the reversal potential of the pH-sensitive current in HCO₃⁻-free HEPES buffer, suggesting that activity of one or more HCO₃⁻ transporters contributes to pH sensitivity of RTN astrocytes. The most likely candidate is the electrogenic NBC, which can generate a depolarizing inward current (i.e., HCO₃⁻ efflux) during bath acidification (Chesler 2003; Deitmer 1991). Therefore we used diisothiocyanatostilbene-2,2′-disulfonic acid (DIDS, an NBC blocker) to isolate the contribution of this transporter to pH sensitivity of RTN astrocytes. At control pH = 7.4, exposure to DIDS (0.5 mM) decreased holding current by 41.6 ± 8 pA (n = 3) (Fig. 3C). In the continued presence of DIDS, exposure to 15% CO₂ decreased holding current by 45.2 ± 13 pA (n = 3) (Fig. 3C), which corresponds with a change in holding current of 69.5 ± 20 pA/pH unit (Fig. 3G). The CO₂/H⁺-sensitive current in DIDS reversed at −94 mV (i.e., near the equilibrium potential for K⁺ [E_K⁺]) (Fig. 3, D and F), as expected for a current dominated by K⁺ conductance.

Fig. 3.

Characteristics of pH-sensitive RTN glia in bicarbonate buffer. A: traces of membrane potential shows that exposure to 15% CO₂ causes a reversible membrane depolarization. B: summary of the effects of 15% CO₂ on membrane potential pH-sensitive astrocytes (n = 3). C: trace of holding current shows that 15% CO₂ decreases outward current and conductance. Under control conditions (i.e., 5% CO₂, pHo = 7.45), exposure to diisothiocyanatostilbene-2,2′-disulfonic acid (DIDS, 0.5 mM) also decreased holding current close to 100 pA. In the continued presence of DIDS, a second exposure to 15% CO₂ decreased holding current also by about 100 pA. D and E: current responses to voltage steps from −80 mV to between −40 and −150 mV during 5% and 15% CO₂ exposure in bicarbonate buffer alone or in the presence of DIDS (D) and corresponding I–V relationships of the average CO₂/H⁺-sensitive current (i.e., difference current) (n = 3) in bicarbonate buffer alone (E) or bicarbonate buffer with DIDS (F). Note that under control conditions the reversal potential of the CO₂/H⁺-sensitive current is hyperpolarized to E_K⁺ (arrow in E), but when activity of the electrogenic Na⁺/HCO₃⁻ cotransporter (NBC) is blocked with DIDS, the CO₂/H⁺-sensitive current reverses near E_K⁺ (arrow in F). G: summary data comparing effects of CO₂/H⁺ on holding current (at a holding potential of −80 mV) in HEPES and bicarbonate-buffered media. Activity of the NBC significantly increased acid-induced changes in holding current compared with HCO₃⁻-free HEPES solution or in the presence of DIDS (*P < 0.05).

The DIDS-sensitive current exhibited a reversal potential of −104 mV, which is slightly hyperpolarized to the predicted reversal potential for the NBC (E_NBC). For example, the E_NBC was calculated with the following equation (Newman 1991): E_NBC = RT/F(n − 1) ln [Na⁺]_i[HCO₃⁻]_iⁿ/[Na⁺]_o[HCO₃⁻]_oⁿ, where n is the 2HCO₃⁻:1Na⁺ stoichiometry and R (gas constant), T (temperature), and F (Faraday constant) have their usual meanings. For a temperature of 22°C, [Na⁺]_o = 157 mM, [Na⁺]_i = 4 mM, [HCO₃⁻]_o = 26 mM, and [HCO₃⁻]_i was estimated to be 22 mM using the Henderson–Hasselbalch equation as previously described (Gross et al. 2001); we calculated E_NBC to be −99 mV. It is not clear why the DIDS-sensitive current reversed about 5 mV negative to E_NBC, although DIDS has been shown to have nonspecific effects on anion exchangers and Cl⁻ channels in glia (Munsch and Deitmer 1994), which may account for the discrepancy. To confirm that Cl⁻ channels do not contribute to CO₂ sensitivity of RTN astrocytes, we tested CO₂ sensitivity in low chloride (50 mM) buffer.

The low Cl⁻ solution was made isoosmotic by replacing 80 mM NaCl with an equal amount of Na-gluconate. We found the CO₂/H⁺-sensitive current was not affected by chloride reduction. For example, there was no difference between the CO₂/H⁺-induced change in holding current under control conditions (92.3 ± 38 pA; n = 3) and in low Cl⁻ buffer (91.3 ± 37 pA; n = 3) (not shown). Likewise, amplitude and reversal potential of the CO₂/H⁺-sensitive current were similar in control and low Cl⁻ conditions (not shown).

In the absence of NBC activity (HCO₃⁻-free HEPES buffer or Ringer solution with DIDS), the CO₂/H⁺-sensitive current expressed by RTN astrocytes reversed near the equilibrium potential for K⁺ (E_K⁺ = −97 mV), suggesting that one or more pH-sensitive K⁺ channels contribute to the pH sensitivity of RTN astrocytes. Astrocytes have been shown to express several types of pH-sensitive K⁺ channels, including background TASK channels and heteromeric Kir4.1–Kir5.1 channels. In particular, heteromeric Kir4.1–Kir5.1 channels are exquisitely sensitive to pH in the physiological range (Casamassima et al. 2003; Xu et al. 2000a), are preferentially expressed by astrocytes (Hibino et al. 2004; Tang et al. 2009), and in situ hybridization shows detectable levels of both Kir4.1 and Kir5.1 mRNA in the RTN (Wu et al. 2004). Therefore we consider heteromeric Kir4.1–Kir5.1 channels the most likely molecular pH sensor in RTN astrocytes.

Heteromeric Kir4.1–Kir5.1 channels confer pH sensitivity to RTN astrocytes

The possible contribution of Kir channels can easily be distinguished from background K⁺ channels based on Ba²⁺ sensitivity (Olsen and Sontheimer 2008); most Kir channels, but not background K⁺ channels, are fully inhibited by micromolar concentrations of Ba²⁺ (for review of background and Kir channel pharmacology see Goldstein et al. 2005; Kubo et al. 2005; Patel and Honore 2001). Therefore to determine whether heteromeric Kir4.1–Kir5.1 channels contribute to pH sensitivity in RTN astrocytes, we first determine effects of Kir channel inhibition by Ba²⁺ (100 μM) on membrane potential and the pH-sensitive current. In current-clamp, Ba²⁺ depolarized the membrane potential by 16.7 ± 3.0 mV (n = 4) (Fig. 4A). In voltage-clamp (at −80 mV, TTX), Ba²⁺ decreased outward current by 86.5 ± 5 pA (n = 4) (Fig. 4B) and blocked the pH-sensitive current by 96.5 ± 3.5% (n = 4) (Fig. 4, B, D, and F). Note that if a more Ba²⁺-sensitive background channel like TWIK (Ba²⁺ half-maximal inhibitory concentration [IC₅₀] = 100 μM; Patel and Honore 2001) or TASK-1 (Ba²⁺ IC₅₀ = 350 μM; Patel and Honore 2001) contributed to the pH-sensitive current in RTN astrocytes, we would expect that only a portion (at best ∼50%) of the pH-sensitive current would be blocked by 100 mM Ba²⁺. However, we show that this concentration of Ba²⁺ virtually eliminated the pH-sensitive current, suggesting that background K⁺ channels do not confer pH sensitivity to RTN astrocytes. Inhibition of other voltage-gated K⁺ channels with tetraethylammonium (TEA, 10 mM) or 4-aminopyridine (4-AP, 50 μM) had no effect on the pH-sensitive current in RTN astrocytes (Supplemental Fig. S1A). These results indicate that one or more Kir channels contribute to the pH-sensitive current in RTN astrocytes.

Fig. 4.

pH sensitivity of RTN glia involves heteromeric Kir4.1 channels. A1: membrane potential trace shows that inhibition of Kir channels mimics the effects of H⁺; Ba²⁺ (100 μM) and desipramine (des, 100 μM) depolarize membrane potential by about 18 and 9 mV, respectively. A2: average depolarization of RTN pH-sensitive astrocyte when exposed to either H⁺ (acidification from 7.3 to 6.9, n = 5), 100 μM Ba²⁺ (n = 4), or 100 μM desipramine (n = 5). B: trace of holding current shows that H⁺ decreases outward current. At pH 7.3, Ba²⁺ decreases holding current and blocks effects of pH on holding current. C: trace of holding current from a pH-sensitive astrocyte shows desipramine decreases holding current and blocks effects of pH on holding current. D: average (n = 4) I–V relationship of the pH-sensitive current under control conditions and in the presence of 100 μM Ba²⁺ shows negligible pH-sensitive current in Ba²⁺. E: average (n = 4) I–V relationship of the pH-sensitive current under control conditions and in the presence of desipramine shows negligible pH-sensitive current in desipramine. F: average pH-sensitive current amplitude (at −80 mV) recorded from RTN astrocytes during acidification from pH 7.5 to 6.9, exposure to Ba²⁺ (n = 3), or desipramine (n = 4) (*P < 0.05).

To differentiate the contribution of Kir4.1–Kir5.1 channels from other Kir channels, we inhibited Kir4.1-containing channels (i.e., homomeric and heteromeric channels) with desipramine (100 μM). At this concentration, desipramine inhibits channels containing Kir4.1 subunits, with minimal effects on other pH-sensitive Kir channels (Su et al. 2007; Tang et al. 2009). Our results indicate that desipramine mimics effects of acidification; in current-clamp exposure to desipramine depolarized membrane potential by 8.4 ± 3.0 mV (n = 4) (Fig. 4A) and in voltage-clamp it decreased holding current by 35.0 ± 13 pA (n = 4) (Fig. 4C) and the pH-sensitive current by 72.7 ± 21.3% (n = 4) (Fig. 4, C, E, and F). It should be noted that desipramine is also a tricyclic antidepressant and may have nonspecific effects on neuronal activity. However, RTN astrocytes are not responsive to norepinephrine or serotonin (DK Mulkey, unpublished observations), suggesting the effects of desipramine on these cells is mediated by inhibition of Kir4.1-containing channels. These results suggest that heteromeric Kir4.1–Kir5.1 channels confer pH sensitivity to RTN astrocytes.

RTN astrocytes activate pH-sensitive neurons by a purinergic mechanism

Recent evidence suggests that a subset of RTN astrocytes contribute to the mechanism of chemoreception by sensing CO₂/H⁺ and releasing ATP to activate pH-sensitive neurons and increase respiratory drive (Gourine et al. 2010). These results are supported by our evidence that roughly 20% of RTN astrocytes are intrinsically pH sensitive. To test the possibility that pH-sensitive astrocytes modulate activity of pH-sensitive neurons, we paralleled previous in vivo experiments by using fluorocitrate to selectively depolarize astrocytes (Erlichman et al. 1998; Holleran et al. 2001) while recording activity of pH-sensitive neurons. Fluorocitrate is a metabolic toxin that is selectively taken up by astrocytes, where it can initiate membrane depolarization (Erlichman and Leiter 2010), presumably by energy depletion and loss of cytoplasmic K⁺. At subtoxic doses, fluorocitrate has been shown to reversibly depolarize RTN glia and increase ventilatory sensitivity to CO₂ (Erlichman et al. 1998; Holleran et al. 2001), by yet unknown mechanisms. We confirmed that fluorocitrate (50 μM) can reversibly depolarize pH-sensitive astrocytes (Supplemental Fig. S2). We also confirmed the effects of fluorocitrate are not due to residual Ba²⁺ by determining that Ba²⁺ precipitation control solution, prepared in a manner similar to fluorocitrate, had no effect on neuronal or glial membrane potential. We found that fluorocitrate evoked a large and reversible neuronal activation (Fig. 5, A and B). Importantly, a second exposure to fluorocitrate, this time in the presence of an ATP receptor antagonist (PPADS, 100 μM), had no effect on neuronal activity (Fig. 5A). In addition, fluorocitrate also increased neuronal pH sensitivity; in the continued presence of fluorocitrate, responsiveness to acidification increased by 0.7 ± 0.2 Hz (n = 3), compared with acid alone (Fig. 5B). These results also confirm previous results that P2 receptors are not required for pH sensitivity of RTN neurons in slices incubated in HEPES buffer (Mulkey et al. 2004, 2006); pH sensitivity of RTN neurons was wholly retained in the presence of PPADS (Fig. 5B). It has been suggested that CO₂ and not H⁺ per se is the stimulus required for ATP release from RTN astrocytes (Spyer 2009). Therefore we tested the effect of PPADS and another P2-receptor antagonist (suramin, 100 μM) on CO₂/H⁺ sensitivity in bicarbonate-buffered medium. As shown in Fig. 6, exposure to 15% CO₂ alone increased neuronal activity 2.2 ± 0.1 Hz (n = 12). After washing to 5% CO₂ exposure to PPADS (100 μM) significantly decreased baseline activity from 0.13 ± 0.05 to 0.02 ± 0.01 Hz (n = 8). Likewise, suramin (100 μM) decreased baseline activity 0.50 ± 0.1 Hz to 0.3 ± 0.1 Hz (n = 4). These results suggest that in bicarbonate buffer (pH 7.3) RTN chemoreceptors receive tonic excitatory purinergic input. Prior to testing CO₂/H⁺ sensitivity in PPADS or suramin, a depolarizing current (∼1 nA) was delivered to the neuron to approximate the control level of activity. In the continued presence of PPADS (with baseline activity adjusted by DC current injection to near control levels), a second exposure to 15% CO₂ increased neuronal activity by 1.6 ± 0.1 Hz (n = 8); this response was significantly less than the CO₂ response under control conditions (initial response or in wash) (Fig. 6A). In the presence of suramin (with baseline activity adjusted to near control levels), a second exposure to 15% CO₂ increased neuronal activity by 1.4 ± 0.1 Hz; this response was significantly less than the CO₂ response under control conditions (Fig. 6B). These results indicate that P2-receptor antagonists PPADS and suramin decreased CO₂/H⁺ sensitivity. To the extent that fluorocitrate selectively affects astrocytes (Erlichman and Leiter 2010), these results suggest that astrocytes can release ATP and activate pH-sensitive neurons, presumably by a P₂Y-dependent mechanism (Mulkey et al. 2006), to increase chemoreceptor output. These results also lend support to the hypothesis that pH-sensitive astrocytes contribute to chemoreception by increasing activity of pH-sensitive neurons.

Fig. 5.

RTN astrocytes increase baseline activity and pH sensitivity of RTN chemoreceptors by a P2-receptor–dependent mechanism. A: traces of pH and firing rate (Hz) from a pH-sensitive neuron shows a characteristic response to H⁺: increasing pH to 7.5 inhibited activity, whereas acidification to pH 6.9 increased firing rate >2 Hz. At a control pH of 7.3, exposure to fluorocitrate (fluoro, 50 μM) evoked a large and reversible increase in neuronal activity. A second exposure to fluorocitrate, this time in the presence of a P2-receptor antagonist pyridoxal-phosphate-6-azophenyl-2′,4′-disulfonate (PPADS), 100 μM, blocked the effects of fluorocitrate. Fluorocitrate responsiveness returned in wash. B: traces of pH and firing rate from a pH-sensitive RTN neuron exposed to acidic and alkaline conditions in the presence and absence of fluorocitrate (fluoro, 50 μM) or PPADS (100 μM). Fluorocitrate alone stimulated neuronal activity. In the continued presence of fluorocitrate, acidification further increased firing rate by about 1 Hz, whereas alkalization decreased activity. Note that the effects of pH but not fluorocitrate were retained in the presence of PPADS. Bar graphs summarize the effects of fluorocitrate in the presence of PPADS (top, n = 5) and in acidic and alkaline conditions (bottom, n = 3) (*P < 0.05).

Fig. 6.

In bicarbonate-buffered solution astrocytes contribute to chemosensitivity by a purinergic mechanism. A: trace of firing rate shows the response of an RTN neuron to increases in CO₂ from 5 to 15% in bicarbonate buffer alone or in the presence of PPADS. An initial exposure to 15% CO₂ increased the firing rate about 2.8 Hz. After returning to 5% CO₂, exposure to PPADS (100 μM) decreased baseline firing rate, suggesting that in bicarbonate buffer RTN chemoreceptors receive tonic excitatory purinergic input. A depolarizing current (∼1 nA) was delivered to the neuron to increase firing rate to near control levels. In the continued presence of PPADS (with baseline activity adjusted by DC current injection to near control levels), a second exposure to 15% CO₂ increased the firing rate about 1.5 Hz. After washing, PPADS CO₂ sensitivity returned to initial levels. Bar graph on the right summarizes the firing rate response of RTN neurons (n = 9) to 15% CO₂ in control and PPADS (*P < 0.05). B: firing rate trace shows the response of a RTN chemoreceptor to 15% CO₂ in bicarbonate-buffered solution alone and in the presence of suramin (100 μM). An initial exposure to 15% CO₂ increased the firing rate about 2.7 Hz. As was observed with PPADS, exposure to suramin (100 μM) in 5% CO₂ decreased baseline firing rate. A depolarizing current (∼1 nA) was delivered to the neuron to increase firing rate to near control levels. In the continued presence of suramin (with baseline activity adjusted by DC current injection to near control levels), a second exposure to 15% CO₂ increased firing rate about 1.7 Hz. After washing, suramin CO₂ sensitivity returned to initial levels. The asterisk designates a 5-min break. Bar graph on the right summarizes the firing rate response of RTN neurons (n = 5) to 15% CO₂ in control and suramin (*P < 0.05).

DISCUSSION

Neuron–glia communication has emerged in recent years as a fundamental process in the brain that can be an important determinant of neuronal activity. However, by most accounts the activity of astrocytes is thought to be under the strict control of neuronal input (Fellin et al. 2006). Here we show that RTN astrocytes sense H⁺ in part by inhibition of heteromeric Kir4.1–Kir5.1 channels and stimulate activity of RTN chemoreceptors by a P2-receptor–dependent mechanism. Our evidence also indicates that purinergic input is not required for RTN chemoreceptor function. These results identify the most likely mechanism by which RTN astrocytes sense pH and build on recent evidence that astrocytes contribute to respiratory drive by a purinergic mechanism (Gourine et al. 2010).

Molecular mechanism underlying pH sensing by RTN astrocytes

Fukuda et al. (1978) showed >30 yr ago that the RTN contains a population of pH-sensitive nonspiking cells and more recent evidence shows these pH sensors express the NBC (Ritucci et al. 2005), thus narrowing the scope of candidates to either astrocytes or oligodendrocytes. Our data indicate that pH-sensitive RTN glia are protoplasmic astrocytes; specifically, these cells have a hyperpolarized membrane potential; high K⁺ conductance; exhibit an electrically passive profile; and express Aldh1L1, a marker of protoplasmic astrocytes (Cahoy et al. 2008).

Several pH-sensitive channels have been proposed to contribute to the mechanism of chemoreception, including Ca⁺²-activated K⁺ channels (K_Ca) (Wellner-Kienitz et al. 1998), voltage-sensitive K⁺ channels (e.g., transient and delayed rectifying K⁺ channels) (Filosa and Putnam 2003; Putnam et al. 2004), background TASK channels (Bayliss et al. 2001; Mulkey et al. 2004), and Kir channels (Schultz et al. 2003; Xu et al. 2000a; Zhu et al. 2000). All of these channels are expressed in various types of glial cells and conceivably contribute to the pH sensitivity of RTN astrocytes. Our results show that the pH-sensitive current in RTN astrocytes can be blocked with 100 μM Ba²⁺, which will fully block most Kir channels, with only modest effects on background K⁺ channels (e.g., TASK channels) (Goldstein et al. 2005; Kubo et al. 2005; Patel and Honore 2001). In addition, the pH-sensitive current in RTN astrocytes was not blocked by TEA (10 mM) or 4-AP (50 μM), which can inhibit other voltage- and Ca²⁺-dependent K⁺ channels, thus narrowing the scope of candidate molecular sensors to one or more Kir channels.

The Kir family of channels consists of 16 different channel subunits grouped into seven subfamilies (Kir1.x–Kir7.x) based on sequence homology (Olsen and Sontheimer 2008). All Kir channels are composed of four α-subunits assembled as homomeric or heteromeric channels that show varying degrees of rectification, single-channel conductance, and pharmacology. A common feature of most Kir channels is high sensitivity to low micromolar concentrations of Ba²⁺ (Kubo et al. 2005; Mulkey et al. 2004), a property that is frequently used to differentiate contributions of Kir channels from other K⁺ channels, including background K⁺ channels (Mulkey et al. 2004). Only certain Kir channels are pH sensitive, e.g., Kir1.1 (pK = 6.7) (Xu et al. 2000b), Kir2.3 (pK = 6.8) (Putnam et al. 2004), homomeric Kir4.1 (pK = 6.1) (Pessia et al. 2001), and the heterodimeric Kir4.1–Kir5.1 channels (pK = 7.4) (Casamassima et al. 2003; Jiang et al. 2001; Xu et al. 2000b). Note that Kir5.1 subunits do not form functional homomeric channels, but preferentially heteromerize with Kir4.1 subunits in astrocytes (Casamassima et al. 2003; Pessia et al. 1996). Of all pH-sensitive Kir channels, only heteromeric Kir4.1–Kir5.1 channels are exquisitely sensitive to physiologically relevant pH changes. Further, in situ hybridization found that the RTN expresses high levels of Kir4.1 and Kir5.1 mRNA (Wu et al. 2004). Here, we show that RTN astrocytes respond to physiologically relevant pH changes. The pH-sensitive current expressed by these cells appears relatively voltage independent between −150 and −30 mV and was almost completely blocked by micromolar concentrations of Ba²⁺ and desipramine (a blocker of Kir4.1-containing channels; Su et al. 2007; Tang et al. 2009). This combination of features (i.e., inhibition by physiologically relevant changes in pH and micromolar concentrations of Ba²⁺ and desipramine) is most consistent with heteromeric Kir4.1–Kir5.1 channels. Therefore we consider heteromeric Kir4.1–Kir5.1 channels to be excellent candidate pH sensors of RTN astrocytes.

Our results also indicate that the NBC contributes to the pH-sensitive current in RTN astrocytes. It is well established in various types of astrocytes that activity of the NBC generates a hyperpolarizing outward current (i.e., net HCO₃⁻ influx) at membrane potentials depolarized to E_NBC (Chesler 2003). In our experimental conditions we calculated the E_NBC to be about −99 mV. As expected, at a holding potential of −80 mV in bicarbonate-buffered solution we observed a DIDS-sensitive outward current. This outward current will likely increase during depolarization and so potentially blunt H⁺-mediated inhibition of Kir4.1–Kir5.1 and the corresponding membrane potential. However, the NBC is also modulated by changes in extracellular pH; bath acidification can reverse Na⁺-HCO₃⁻ transport to produce an inward depolarizing current (i.e., net HCO₃⁻ efflux) (Chesler 2003; Deitmer 1991). In voltage-clamp (at a holding potential of −80 mV), we show that inward current produced by the NBC increased amplitude of the CO₂/H⁺-sensitive current and caused a negative shift in its reversal potential. When activity of the NBC was blocked with DIDS, we observed a CO₂/H⁺-sensitive current that was similar in amplitude to that of the pH-sensitive current in HCO₃⁻-free HEPES solution and reversed near E_K⁺. A contribution of the NBC to central chemoreception has yet to be established and considering that H⁺-induced depolarization and extracellular acidification have offsetting effects on directionality of Na⁺/HCO₃⁻ transport by the NBC, further experiments are required to define potential roles of this transporter in RTN chemoreception.

Possible role of astrocytes in central chemoreception

There are two mechanisms by which astrocytes are thought to influence chemoreception. First, a well-known function of astrocytes is to help regulate extracellular K⁺ by taking up extracellular K⁺ ions during increased neuronal activity (Olsen and Sontheimer 2008). This process effectively buffers changes in extracellular K⁺; however, it can also depolarize membrane potential and increase Na⁺/HCO₃⁻ cotransport into glia (Chesler 2003), which would enhance the CO₂-induced fall in extracellular pH and potentiate activation of chemoreceptors. This possibility is supported by a series of in vivo experiments in which fluorocitrate microinjected into the RTN caused extracellular acidification followed by increased respiratory frequency and ventilatory sensitivity to CO₂ (Erlichman and Leiter 2010; Holleran et al. 2001). Note that fluorocitrate is a toxin that inhibits metabolic activity; at low concentrations it has been shown to be selectively taken up by astrocytes but not neurons, where it initiates membrane depolarization (Erlichman and Leiter 2010), presumably by loss of cytoplasmic K⁺. However, it is also possible that the fluorocitrate-mediated effects described in these experiments resulted from the release of excitatory gliotransmitters acting on pH-sensitive neurons. Our data support this latter possibility. For example, we found that exposure to fluorocitrate activated pH-sensitive neurons under conditions in which the NBC is minimally active (i.e., in HEPES buffer). In addition, we were able to block effects of fluorocitrate with a P2-receptor antagonist. Therefore the most likely mechanism by which astrocytes contribute to chemoreception involves release of ATP and activation of pH-sensitive neurons. Our data also indicate that fluorocitrate caused an upward parallel shift in the pH sensitivity of RTN neurons, suggesting that fluorocitrate-mediated ATP release from RTN astrocytes contributes to chemoreception by increasing excitability of RTN neurons. These results are consistent with evidence in other brain regions that astrocytes serve as a source of excitatory purinergic drive able to rapidly modulate neural activity (Gordon et al. 2009). However, it should be noted that these experiments are limited by a dearth of pharmacological tools available to selectively manipulate astrocytes and the possibility that fluorocitrate may have nonspecific effects on other cell types, including neurons. Therefore additional experiments that use genetic approaches to selectively manipulate astrocytes are required to definitively establish a role of pH-sensitive astrocytes in chemoreception. It is important to note that our results are entirely consistent with a recent study that used the adenoviral system to target expression of ChR2 to RTN astrocytes (Gourine et al. 2010). This study showed that photostimulation of RTN astrocytes increased activity of RTN neurons and stimulated breathing by a purinergic-dependent mechanism. These results are similar to the observed effects of fluorocitrate, thus adding to our confidence that fluorocitrate effects are mediated by astrocytes.

Based on the results described earlier, it is reasonable to expect P2-receptor antagonists to decrease pH sensitivity of RTN neurons. However, previous experiments performed in HEPES-buffered media (Mulkey et al. 2004, 2006) indicate that PPADS had no effect on pH sensitivity of RTN neurons. A possible explanation for this apparent discrepancy is that CO₂, not H⁺, is the effector responsible for hypercapnic-induced ATP release (Spyer 2009). Therefore in a parallel series of experiments, we tested the effects of two P2-receptor antagonists (PPADS and suramin) on CO₂/H⁺ sensitivity of RTN neurons in bicarbonate-buffered media. Interestingly, we found that under these conditions both P2-receptor antagonists blunted neuronal CO₂/H⁺ sensitivity. These results are consistent with the possibility that CO₂ is the required stimulus for ATP release in the RTN.

In summary, we demonstrate that pH-sensitive RTN glia are protoplasmic astrocytes that sense H⁺ by inhibition of heteromeric Kir4.1–Kir5.1-like conductance and contribute to respiratory drive by increasing activity of pH-sensitive neurons by a purinergic-dependent mechanism. Our results also suggest that the NBC may contribute to the pH-sensitive current in RTN astrocytes. Thus we conclude that RTN astrocytes contribute to the mechanism of chemoreception and influence integrated output of the RTN.

GRANTS

This work was supported by the National Institutes of Health Grants HL-104101 to D. K. Mulkey and NS-049267 to A. Nishiyama and a University of Connecticut Research Foundation grant to D. K. Mulkey.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.