K_ir5.1-Dependent CO₂/H⁺-Sensitive Currents Contribute to Astrocyte Heterogeneity Across Brain Regions

Abstract

Astrocyte heterogeneity is an emerging concept in which astrocytes within or between brain regions show variable morphological and/or gene expression profiles that presumably reflect different functional roles. Recent evidence indicates that retrotrapezoid nucleus (RTN) astrocytes sense changes in tissue CO_2/H⁺ to regulate respiratory activity; however, mechanism(s) by which they do so remain unclear. Alterations in inward K⁺ currents represent a potential mechanism by which CO₂/H⁺ signals may be conveyed to neurons. Here, we use slice electrophysiology in rats of either sex to show that RTN astrocytes intrinsically respond to CO₂/H⁺ by inhibition of an inward rectifying potassium (K_ir) conductance and depolarization of the membrane, while cortical astrocytes do not exhibit such CO₂/H⁺-sensitive properties. Application of Ba²⁺ mimics the effect of CO₂/H⁺ on RTN astrocytes as measured by reductions in astrocyte K_ir-like currents and increased RTN neuronal firing. These CO₂/H⁺-sensitive currents increase developmentally, in parallel to an increased expression in K_ir4.1 and K_ir5.1 in the brainstem. Finally, the involvement of K_ir5.1 in the CO₂/H⁺-sensitive current was verified using a Kir5.1 KO rat. These data suggest that K_ir inhibition by CO₂/H⁺ may govern the degree to which astrocytes mediate downstream chemoreceptive signaling events through cell-autonomous mechanisms. These results identify K_ir channels as potentially important regional CO₂/H⁺ sensors early in development, thus expanding our understanding of how astrocyte heterogeneity may uniquely support specific neural circuits and behaviors.

Introduction

Astrocytes play essential roles in brain physiology, from maintaining extracellular ion and transmitter concentrations to directly communicating with neurons to regulate their activity. An emerging concept in the research of astrocytes is their heterogeneity: while once considered a homogenous cell type across the CNS, recent research indicates that astrocytes vary in their morphology (Matyash and Kettenmann 2010), gene and protein expression profile (Chai et al. 2017; Morel et al. 2017; Rusnakova et al. 2013), ion channel and receptor expression (Bordey and Sontheimer 2000; Hoft et al. 2014; Zhou and Kimelberg 2000), physiology (Verkhratsky and Nedergaard 2018), response to stimulus (Cholewinski and Wilkin 1988; el-Etr et al. 1989), and interaction with neurons (Ben Haim and Rowitch 2017; Oberheim et al. 2012; Xin and Bonci 2018) across different brain regions and developmental stages (Chaboub and Deneen 2012; Farmer and Murai 2017; Zhang and Barres 2010). Thus, understanding how astrocytes differ between brain areas may provide insight regarding the difference in their roles in CNS physiology from region to region.

We are particularly interested in contributions of astrocytes to control of breathing. A brainstem region known as the retrotrapezoid nucleus (RTN) regulates depth and frequency of breathing in response to changes in CO2/H+ (Guyenet and Bayliss 2015), process known as central respiratory chemoreception. Chemosensitive RTN neurons intrinsically respond to reduced pH by increasing their firing rate to drive respiration and correct changes in blood gases. Disruption of RTN neurons genetically (Dubreuil et al. 2008), by targeted lesion (Takakura et al. 2008), or selective ablation of their pH sensing mechanisms (Kumar et al. 2015) results in a dramatically reduced ventilatory response to CO₂ or respiratory acidosis. This evidence that RTN neurons are intrinsically pH sensitive does not preclude the possibility that CO₂/H⁺-dependent output of this region is also subject to modulation by surrounding cells. Glia in the region have been proposed as functional pH detectors since their depolarization in response to acidification was first recognized over 40 years ago (Fukuda et al. 1978). Indeed, pharmacologic depolarization (Erlichman et al. 1998; Holleran et al. 2001; Sobrinho et al. 2017) or channel rhodopsin-mediated stimulation (Gourine et al. 2010) of astrocytes is sufficient to activate RTN neurons in vitro and increase respiration in vivo. These results establish RTN astrocytes as elements in RTN chemotransduction. However, specific upstream mechanism(s) by which CO₂/H⁺ “activates” astrocytes and leads to altered respiratory drive appear varied and remain unclear.

Recent evidence suggests that RTN astrocytes may sense lowered pH by inhibition of inwardly rectifying potassium currents (K_ir) (Wenker et al. 2010), which are produced by channels of the K_ir family (for review, see (Hibino et al. 2010)). These currents are characterized by larger inward than outward currents (inward rectification). K_ir channels are tetramers that can be composed of four identical subunits (homotetrameric, e.g. K_ir4.1) or pairs of different subunits (heterotetrameric, e.g. K_ir4.1/5.1). The most abundant glial K_ir channel is K_ir4.1 (Tang et al. 2009), and as such, is a major determinant of astrocyte function (for review, see, Nwaobi et al. 2016). Some K_ir channels are intrinsically H⁺ sensitive, including K_ir4.1 channels, which possess an intracellular H⁺ sensor, albeit in a non-physiological range (pK_a=6.7) (Casamassima et al. 2003; Xu et al. 2000). Interestingly, when heteromerized with K_ir5.1, K_ir4.1/5.1 channels are H⁺-sensitive in the physiological range (pK_a=7.45). Thus, differentially expressed K_ir channels and/or different combination of K_ir channel subunits may contribute to regional differences in astrocyte CO₂/H⁺ sensing. Additionally, robust developmental upregulation of K_ir4.1 (Nwaobi et al. 2014) coincides with enhancement of the ventilatory response to CO₂/H⁺ from the time of birth to adulthood (Davis et al. 2006; Putnam et al. 2005). However, to date, no differences in K_ir currents have been shown to contribute to developmental or regional specificity of astrocyte chemosensitivity.

In the present study, we used slice-patch electrophysiology to characterize functional heterogeneity between RTN and cortical astrocytes based on K_ir CO₂/H⁺-sensitive currents. We found that RTN but not cortical astrocytes sense changes in CO₂/H⁺ by inhibition of Kir-like currents beginning at postnatal day 2–3 (PND2–3; the earliest time point studied), and that this CO₂/H⁺-sensitive conductance increases over postnatal development. We also identified a requisite role of K_ir5.1 in RTN astrocyte CO₂/H⁺-sensitivity. These collective findings suggest that K_ir inhibition by CO₂/H⁺ is another form of astrocyte heterogeneity, which may have implications for the development of resilience to respiratory failure throughout early life.

Results

While a limited number of studies have evaluated the function of astrocytes in any brainstem region, none have comprehensively characterized their developmental electrophysiological properties as has been done in the forebrain (Zhong et al. 2016; Zhou et al. 2006). Only one previous study evaluated the current-voltage (I-V) relationships in a small number of RTN astrocytes in response to CO₂/H⁺ (Wenker et al. 2010). This limited number of electrophysiological studies is in contrast to a larger cohort of calcium imaging studies (Forsberg et al. 2017; Gourine et al. 2010; Kasymov et al. 2013; Sheikhbahaei et al. 2018), which show that cortical and brainstem astrocytes differ in their responses to pH changes. To gain mechanistic insight into the basis of differential astrocyte CO₂/H⁺ sensitivity, we characterized properties of CO₂/H⁺ sensitive currents in cortical and RTN astrocytes during the first two weeks of life.

Whole-cell current-clamp recordings from astrocytes in the RTN and in deep layer 1 and layer 2/3 of the motor cortex were made under control conditions (95% O₂, 5% CO₂, pH=7.30) and hypercapnic conditions (90% O₂, 10% CO₂, pH=7.00) in slices from Sprague Dawley (SD) rat pups at three time points: PND2–3, PND4–7, and PND9–13. Because we were interested in the intrinsic, cell-autonomous properties of astrocytes in these regions, all electrophysiological experiments were performed in tetrodotoxin (TTX) unless otherwise noted. Additionally, we utilized a reporter SD rat model that expresses eGFP under the promoter of the gene encoding S100β, a predominantly astrocytic Ca²⁺-binding protein, to assist in identifying RTN astrocytes. S100β has long been considered an astrocyte marker and co-expression with glial fibrillary acidic protein (GFAP) is considered a reliable means of distinguishing astrocytes from other glial cell populations, including NG2 cells (Grosche et al. 2013; Nishiyama et al. 1996; Ogata and Kosaka 2002; Raponi et al. 2007; Zhu et al. 2008). Our previously published work in cortex (Nwaobi et al. 2014) and data obtained from RTN (Supplementary Figure 1) indicate eGFP+ cells at the ventral surface of the RTN (S100β-eGFP) also have emanating processes that are GFAP+, suggesting the majority of eGFP+ cells are astrocytes. Thus, astrocytes were identified by eGFP expression, relatively round somas, no predominant outward rectification, the resting membrane potential was more negative than −75mV, and the cell recording remained stable through the experiment. While we recognize that NG2 cells and astrocytes display some similar electrophysiological characteristics, we excluded cells that had the “typical” appearance of NG2 cells (predominantly outwardly rectifying, and Na⁺ channel activation in experiments without TTX). However, we cannot exclude the possibility we inadvertently evaluated some fraction of NG2 cells, however, based on our exclusion criteria- NG2 cells would represent a small fraction of the cells we recorded from. Although we were easily able to detect cells using the above criteria that remained stable upon whole-cell breakthrough at PND2–3 in the RTN, we were unable to find and record from such cells in the cortex at this early time point.

Electrical properties of cortical and RTN Astrocytes

Astrocyte K_ir conductances were evaluated using a voltage step protocol that included a 0 mV pre-pulse to inactivate transient outward rectifying K⁺ channels (Figure 1A–E representative traces, protocol below Figure 1A) (Olsen 2012). Current responses to voltage steps from −180 mV to 120 mV were recorded; for our analysis, we focused on currents measured from −180 mV to −80 mV to avoid symmetry around ~−80 mV that obscures the main effect. Current density did not significantly differ between regions both at PND4–7 (Figure 1G main effect of region: F_(1,28)=3.8, p=0.06; RTN: n=18; cortex: n=12) and at PND9–13 (Figure 1H main effect of region: F_(1,54)=2.1, p=0.15; RTN: n=31; cortex: n=25). As noted above, we were unable to identify and record from astrocytes at PND2–3 in cortex.

Figure 1. Cortical and RTN astrocytes display similar electrophysiological phenotypes under control conditions.

A-E. Representative traces of baseline voltage step in the RTN (A-C) and cortex (D-E) across three developmental time points (2 for cortex). Voltage step protocol utilized to elicit currents is shown on the left.

F-H. Currents elicited in cortical and RTN astrocytes do not differ from one another under control conditions (RTN: PND2–3 n=10, PND4–7 n=18, PND 9–13 n=31; cortex: PND 4–7 n=12, PND 9–13 n=25). As noted in the text, we were unable to identify or record from astrocytes at PND2–3 in cortex.

I. Plot of resting membrane potential of RTN and cortical astrocytes as a function of age, obtained in current clamp ~1 minute after whole-cell breakthrough. Developmental but not regional differences were observed throughout early postnatal development (RTN: PND2–3 n=10, PND4–7 n=18, PND 9–13 n=31; cortex: PND 4–7 n=12, PND 9–13 n=25).

J. Developmental but not regional differences were observed in whole cell capacitance (RTN: PND2–3 n=10, PND4–7 n=18, PND 9–13 n=30; cortex: PND 4–7 n=12, PND 9–13 n=25).

K. Input resistance decreases throughout early postnatal development in cortical astrocytes (PND 4–7 n=12, PND 9–13 n=25) but not RTN astrocytes (PND2–3 n=10, PND4–7 n=18, PND 9–13 n=31).

The RTN astrocyte resting membrane potential (V_m) was slighly hyperpolarized compaired to cortical astrocytes at similar developmental time points (RTN: PND2–3 V_m=−81.1±0.8 mV, n=10; PND4–7 V_m=−82.6±0.7 mV, n=18; PND9–13 V_m=−82.5±0.5 mV, n=31; cortex: PND4–7 V_m=−81.2±0.9 mV, n=12; PND9–13 V_m=−80.9±0.7 mV, n=25; 2-way ANOVA performed only on PND4–7 and PND9–13; main effect of age: F_(1,82)=0.41, p=0.52; main effect of region: F_(1,82)=3.5, p=0.06; interaction: F_(1,82)=0.26, p=0.61; Figure 1I). Whole cell capacitance (C_m), a measure of cellular membrane area, increased in both regions across development, as expected with presumably increasing morphologic complexity paralleling maturation (RTN: PND2–3 C_m=6.7±0.7 pF, n=10; PND4–7 C_m=6.6±0.5 pF, n=18; PND9–13 C_m=10.9±0.7 pF, n=30; cortex: PND4–7 C_m=6.9±0.7 pF, n=12, PND9–13 C_m=11.6±0.8 pF, n=25; 2-way ANOVA performed only on PND4–7 and PND9–13; main effect of age: F_(1,81)=31.2, p<0.0001; main effect of region: F_(1,81)=0.3, p=0.56; interaction: F_(1,81)=0.6, p=0.83; Figure 1J). We also found that astrocytes showed an age dependent decrease in input resistance specifically in the cortex (2-way ANOVA performed only on PND4–7 and PND9–13; main effect of age: F_(1,82)=37.4, p<0.0001; interaction: F_(1,92)=24.6, p<0.0001; for RTN: t₍₈₂₎=0.9, p=0.94; for cortex: t₍₈₂₎=7.2, p<0.0001), and lower input resistance in the RTN than in cortex (main effect of region: F_(1,82)=14.9; p=0.0002), specifically at PND4–7 (t₍₈₂₎=5.4, p<0.0001) but not in PND9–13 (t₍₈₂₎=0.9, p=0.92) (cortical astrocyte input resistance: PND4–7, 174.0±29.4 MΩ; PND9–13, 40.9±5.8 MΩ; RTN astrocyte input resistance: PND2–3, 86.3±20.8 MΩ; PND4–7, 67.9±10.0 MΩ; PND9–13, 54.0±8.0 MΩ; n same as V_m; Figure 1K). Interestingly, there also appears to be regional heterogeneity in resting membrane potential and input resistance of RTN astrocytes at the early PND2–3 time point relative to neonatal hippocampal astrocytes per prior reports (Zhong et al. 2016).

RTN astrocytes display intrinsic CO₂/H⁺-sensitive current; cortical astrocytes do not

Based on previous work that showed by two weeks of age, RTN astrocytes sense CO₂/H⁺ by inhibition of a Kir-like conductance (Wenker et al. 2010), therefore, we next wanted to determine whether astrocytes in the cortex and RTN show different regional sensitivity to CO₂/H⁺ prior to and including this developmental window. Whole-cell currents were recorded under control conditions (pH_o ~7.30) and after an exposure to 10% CO₂ (pH_o ~7.00). Membrane conductance was monitored using a single voltage step to −120 mV every 20 seconds. We found that exposure to 10% CO₂ (5–10 min, selected recordings from one PND9–13) minimally effected whole cell conductance in cortical astrocytes (Figure 2A). Conversely, RTN astrocytes showed a strong reversible decrease in conductance in response to this same level of CO₂/H⁺ (Figure 2B, selected recordings from one PND9–13). This is also reflected in selected current tracings from one PND9–13 from each region in Figure 2C. As expected, bath application of Ba²⁺ at 100 μM, a concentration which inhibits astrocytic but not neuronal K_ir currents (Cui et al. 2018; Olsen and Sontheimer 2008; Ransom and Sontheimer 1995), decreased conductance in both cortical and RTN astrocytes (Figure 2A–B). Furthermore, the CO₂/H⁺-sensitive current (I_CO2/H+), which was determined by subtracting I-V relationships recorded under each condition, underlying observed changes in conductance in RTN astrocytes also increased over the developmental window studied. We also found that RTN astrocytes demonstrate a I_CO2/H+ at PND2–3 (Figure 2D). The currents were larger than cortical I_CO2/H+ at both PND 4–7 (main effect of region: F_(1,29)=7.9, p=0.009; RTN: n=18; cortex: n=13; Figure 2E) and PND9–13 (main effect of region: F_(1,55)=6.2, p=0.0157; RTN: n=32; cortex: n=25; Figure 2F). Analysis of the development of the CO₂/H⁺-sensitive current in the RTN reveals developmental increase of these currents (main effect of age: F_(2,57)=9. 1, p=0.0003; Supplementary Figure 2).

Figure 2. CO₂-sensitive current amplitude is greater in the RTN than the cortex.

A-B. Conductance of a representative cortical astrocyte (A, measured by a single −120 mV voltage step every 20 seconds) is inhibited by 100 μM Ba²⁺, and not by 10 % CO₂. Conductance of a representative RTN astrocyte is inhibited by both (B).

C. Selected recordings from one PND9–13 RTN (top) and cortical (bottom) astrocyte demonstrating CO₂-sensitive currents (right) yielded from subtraction of currents measured in 10% CO₂ (middle) from currents measured in 5% CO₂ (left).

D-F. CO₂-sensitive current in RTN astrocytes is present as early as PND2–3 (D, n=10) and is significantly greater than that observed in cortical astrocytes at PND4–7 (E, cortex n=12, RTN n=28) and PND9–13 (F, cortex n=27, RTN n=32).

We next tested whether non-cell autonomous mechanisms may play a role in potentiating the astrocyte response to hypercapnia in either the cortex or the RTN. Thus, in a different group of PND9–13 cells, we performed identical experiments to those outlined thus far in the absence of TTX. We found that the I-V relationship of I_CO2/H+ was similar under both experimental conditions in RTN astrocytes (main treatment effect: F_(1,41)=0.2, p=0.670, TTX n=32, no TTX n=11; Supplementary Figure 3A). However, the I-V relationship of I_CO2/H+ in cortical astrocytes changed when TTX was omitted from the bath solution (main treatment effect: F_(1,38)=4.1, p=0.05, TTX n=25, no TTX n=15; Supplementary Figure 3B). This result suggests that yet unknown neuronal mechanisms may enhance I_CO2/_H+ in cortical astrocytes.

The above evidence suggests RTN astrocytes are intrinsically CO₂/H⁺ sensitive by a mechanism involving inhibition of Kir channels. To further explore this possibility, we compared the density and voltage dependence of the CO₂/H⁺-sensitive current to the Ba²⁺-sensitive current (measured in 5% CO₂) in the same cells. If Kir channels mediate CO₂/H⁺-sensitivity of RTN astrocytes, we expect CO₂/H⁺- and Ba²⁺-sensitive currents to exhibit similar voltage-dependent properties. Consistent with this, we found that the I-V relationship of the CO₂/H⁺- and Ba²⁺ sensitive currents were similar across all time points (main effect of treatment: PND2–3 F_(1,5)=0.4, p=0.578, n=6; PND4–7 F_(1,10)=0.2, p=0.628, n=11; PND9–13 F_(1,13)=1.2, p=0.301, n=14; Figure 3A–C). Furthermore, amplitude of the Ba²⁺-sensitive current is positively correlated with amplitude of the CO₂/H⁺-sensitive current in all RTN astrocytes (Figure 3F). In striking contrast to the RTN, we found that the CO₂/H⁺-sensitive and Ba²⁺-sensitive I-V plots in cortical astrocytes were dissimilar at all time points considered. For example, CO₂/H⁺-sensitive current densities are smaller and the I-V relationship more linear as compared to the Ba²⁺ sensitive current (main effect of treatment: PND4–7 F_(1,8)=26.6, p=0.0009, n=9; PND9–13 F_(1,18)=11.0, p=0.004, n=19; Figure 3D–E). The voltage-independent profile of the cortical CO₂/H⁺-sensitive current is suggestive of a pH-sensitive leak-type K⁺ conductance reminiscent of two-pore domain K⁺ channel such as TASK1 (for review (Ryoo and Park 2016)), or TREK-1, which is expressed highly in cortical astrocytes (Zhang et al. 2014).

Figure 3. CO₂-sensitive currents resemble Ba²⁺-sensitive currents in the RTN.

A-C. RTN astrocyte currents in response to 10% CO₂ application (lighter squares) resemble those elicited by 100μM Ba²⁺ application (darker squares) in the same cells throughout early postnatal development (A, PND2–3 n=6; B, PND4–7 n=11; C, PND9–13, n=14).

D-E. Cortical astrocyte currents in response to 10% CO₂ application (lighter circles) do not closely resemble those elicited by 100uM Ba²⁺ application in the same cells (darker circles) (D, PND4–7, n=9; E, PND9–13, n=19).

F. The correlation between the Ba²⁺-sensitive current (I_Barium) and the CO₂-sensitive current (I_CO2) taken at a single voltage step (V_h=−160 mV) is significantly stronger in the RTN than in the cortex.

One of the possible mechanisms for inhibition of I_Kir is via hypercapnia-induced changes in extracellular K⁺ rather than a direct inhibition of the currents, as K_ir4.1 channels decrease conductance with lower extracellular potassium, [K⁺]_o, in a pH-independent manner (Edvinsson et al. 2011). However, this modulation generally does not occur under physiologic conditions in the brain. In order to evaluate this hypothesis, we analyzed the changes in the reversal potential (E_K) of RTN astrocytes after perfusion of CO₂, (Supplementary Figure 4) and found that the positive shifts in E_K correspond to an increase, not decrease, in [K⁺]_o per the Nernst equation. Thus, modulation of I_Kir by [K⁺]_o cannot explain the observed decrease in K_ir currents.

To further explore the possibility that CO₂/H⁺ and Ba²⁺ are inhibiting the same Kir conductance in RTN astrocytes, we next wanted to determine whether Ba²⁺ could occlude I_CO2/H+. Consistent with this possibility, in voltage clamp we found that in the continued presence of Ba²⁺ (100 uM) subsequent exposure to 10% CO₂ minimally affected whole cell conductance (Figure 4A) and the amplitude of currents (Figure 4B). These data collectively suggest that K_ir channels in RTN astrocytes are largely intrinsically CO₂/H⁺-sensitive and that K_ir constitutes the majority of CO₂/H⁺-sensitive currents in these cells.

Figure 4. Application of 10% CO₂ in the presence of 100 μM Ba²⁺ does not inhibit currents, but depolarizes RTN astrocytes.

A-B. Representative example (A) and summary (B, n=7) of application of 5% CO₂ (A, black) 10% CO₂ alone (A, grey), Ba²⁺ alone (A, red) or 10% CO₂ in the presence of Ba²⁺ (A, blue). Ba²⁺ significantly diminishes the CO₂-sensitive current (A-B).

C-D. Representative example of RTN astrocyte membrane voltage (C) and summary of both RTN and cortical astrocytes (D) shows that both RTN and cortical astrocytes are depolarized by Ba²⁺. However, only RTN astrocytes depolarize in response to 10% CO₂, both with and without Ba²⁺. Letter code next to data refers to significance in a Sidak post-hoc comparisons after 2Way ANOVA, relative to other groups: a – RTN control, b – RTN 10% CO₂, c – RTN Ba²⁺, d – RTN 10% CO₂ + Ba²⁺, e – cortex control, f – cortex 10% CO₂, g – cortex Ba²⁺, h – cortex 10% CO₂ + Ba²⁺. Capital letter denotes inter-regional difference, and small letter denote effect of solution within region.

RTN astrocytes display intrinsic hypercapnia-induced depolarization

Previous studies suggest RTN astrocytes undergo intrinsic H⁺ or hypercapnia-induced depolarization (Fukuda et al. 1978; Ritucci et al. 2005; Sobrinho et al. 2017; Wenker et al. 2012), and that this process may lead to numerous downstream effects, including the further acidification of the extracellular space, Ca²⁺ influx, and the release of signaling molecules that further propagate the chemosensory message (Mulkey and Wenker 2011). Given the importance of K_ir in setting the resting membrane potential of astrocytes, its inhibition by intracellular H⁺ has been suggested as a candidate mechanism of this depolarization (Wenker et al. 2010). To further explore this possibility, we wanted to determine whether a more modest CO₂ stimulus (10% CO₂) than utilized in previous in vitro studies can depolarize RTN or cortical astrocytes. In a subset of experiments separate from those described previously, we patched RTN and cortical astrocytes at PND9–13 and, in whole-cell current clamp mode, measured changes in resting membrane potential in response to 10% CO₂ or Ba²⁺ alone and in combination (Figure 4C–D). We found that bath application of Ba²⁺ (100 μM) consistently depolarized cortical and RTN astrocytes (RTN: Δ8.3±1.3 mV, t₍₁₅₎=6.3, p<.0001; cortex: Δ15.2±2.7 mV, t₍₆₎=5.649, p=.008). However, only RTN astrocytes showed a depolarizing voltage response to 10% CO₂ alone (RTN: Δ3.5±1.0 mV, t₍₂₁₎=3.5, p=0.014; cortex: Δ0.3±0.5 mV, t₍₆₎=0.6, p=0.998)). As above (Figure 4C–D), exposure to 10% CO2 also depolarized RTN astrocyte membrane potential in the presence of Ba²⁺ (Δ4.8±0.7 mV, t₍₁₄₎=7.0, p<0.0001). To rule out the possibility of incomplete Ba²⁺ block at the concentration utilized we performed a dose response in HEK cells for two combinations of glial inward rectifiers – K_ir4.1 and K_ir4.1/5.1. We confirmed that Ba²⁺ blocks K_ir4.1 and K_ir4.1/5.1 with an IC₅₀ of 30.14 μM and 25.5 μM, respectively (Supplementary Figure 5). These results suggest that mechanisms in addition to Kir channels contribute to RTN astrocyte CO₂/H⁺ sensitivity with regard to depolarization. Nevertheless, astrocytes in the cortex do not depolarize in response to CO₂/H⁺_, even in the presence of Ba²⁺ (Δ1.7±0.9 mV, t₍₆₎=1.931, p=.4746).

The relevance of depolarization in astrocytes is generally underappreciated since these cells do not generate action potentials and so are considered electrically non-excitable. Despite this, previous studies have shown that fluorocitrate-induced depolarization of RTN astrocytes is sufficient to increase RTN neuronal firing rate (Wenker et al. 2010) and ventilatory output in vivo (Sobrinho et al. 2017). However, since fluorocitrate is a metabolic poison with potential non-specific effects, we wanted to determine if 100 μM Ba²⁺, which strongly depolarizes RTN astrocytes (Figure 4D), is sufficient to activate chemosensitive RTN neurons. In cell-attached voltage-clamp mode, chemosensitive RTN neurons were identified by their characteristic firing response to CO₂/H⁺; exposure to 10% CO₂ increased activity by 0.98±0.06 Hz (q₍₅₎=20.52, p=0.0001, Figure 5A–B). After returning to control conditions, bath application of Ba²⁺ alone mimicked the effects of CO₂/H⁺ by increasing neural activity (0.51±.04 Hz, q₍₅₎=16.27, p=.0003, Figure 5A–B). In the continued presence of Ba²⁺ subsequent exposure to 10% CO₂ increased neural firing by 0.98±.05 Hz (q₍₅₎=26.67, p<0.0001). This CO₂/H⁺-mediated change was similar to the CO₂/H⁺ firing response in the absence of Ba²⁺ (t₍₅₎=0.08, p=.9393, Figure 5C). In order to verify that 100 μM Ba²⁺ does not inhibit neuronal K⁺ channels in the RTN as reported in the lateral habenula (Cui et al. 2018), we repeated the same protocol utilized by Cui et al. to measure K_ir currents in RTN neurons in response to 100 μM Ba²⁺ under TTX block. While neurons exhibited some I_Kir, it is minute in comparison to astrocytic I_Kir (Figure 5D). These results suggest astrocytic I_Kir inhibition augments the intrinsic neuronal CO₂/H⁺ response. These results are consistent with the possibility that RTN neurons are intrinsically CO₂/H⁺-sensitive (Guyenet and Bayliss 2015; Kumar et al. 2015; Wang et al. 2013) and add evidence that astrocyte depolarization by CO₂/H⁺- or Ba²⁺-induced I_Kir inhibition may contribute to downstream mechanisms activating RTN neurons.

Figure 5. RTN neuron firing rate is increased by both application of Ba²⁺ and CO₂.

A-B. RTN neurons increase their firing rate in response to hypercapnia and Ba²⁺ alone, and fire most rapidly under application of both 10% CO₂ and Ba²⁺ together (A, representative example, spikes are truncated; B, summary; n=6).

C. Change in firing rate with the application of 10% CO₂ is not significantly different between control and Ba²⁺ (n=6).

D. I-V curve (−120 mV to −40 mV, 10 mV steps) from whole-cell patch clamped RTN neurons from PND9–13 rats (in blue). Superimposed, the Ba²⁺-sensitive current measured in astrocytes at the same age from Figure 3C (in red).

Distinct developmental changes in Kir proteins in the brainstem and cortex

Given these results, we evaluated regional expression of the most abundant K_ir subunit in glia cells, K_ir4.1, during development. Previous studies have found a developmental increase in K_ir4.1 protein as well as an elevated expression in the brainstem relative to cortex (Nwaobi et al. 2014). Here we show by Western blot that expression of K_ir4.1 protein is detectable early in development and continues to increase through the first month of life, with higher expression in the brainstem relative to cortex (main effect of brain region - F_(1,36)=78.03, p<.0001; main effect of age - F_(5,36)=16.18, p<.0001; Figure 6A–B). However, the measured baseline electrophysiological currents do not differ as robustly as might be expected based on the difference in Kir4.1 expression between these regions. There remains a possibility that Kir4.1 channel are differentially sequestered in these regions, contributing to downstream functional heterogeneity. Furthermore, expression of Kir4.1 in other glial cell populations may differ between cortex and RTN.

Figure 6: K_ir4.1 and K_ir5.1 are upregulated during development with more K_ir4.1 and K_ir5.1 in brainstem than in cortex.

A-C. K_ir4.1 and K_ir5.1 (A, representative Western blot; B-C, summary of n=4 animals) protein expression increases during development, with higher expression of both proteins in brainstem than cortex (C, D). *middle panel in A, indicates non-specific binding as it shows up in the negative control or HEK cell lane.

As K_ir4.1 can form functional heteromeric channels with K_ir5.1 that are pH-sensitive in the physiological range (Pessia et al. 2001), we next evaluated K_ir5.1 expression in brainstem and cortex. Similar to the K_ir4.1 expression profile, we found that K_ir5.1 protein is more abundantly expressed in the brainstem relative to the cortex beginning early in development, and is developmentally upregulated in both regions (Figure 6A, C; main effect of brain region - F_(1,36)=39.97, p<.0001; main effect of age - F_(5,36)=3.99, p=0.0056). Sex has no effect on these results (Supplementary Figure 6). K_ir5.1 cannot form a functional homomers but readily forms functional pH sensitive heteromers with K_ir4.1 (Hibino et al. 2004; Konstas et al. 2003; Tanemoto et al. 2002). These data suggest that an increase in pH sensitive K_ir4.1/5.1 heteromers in the brainstem relative to the cortex could underlie RTN astrocyte chemosensitivity.

K_ir5.1 deficiency causes attenuation of I_Kir and chemosensitivity in RTN astrocytes

Again using heterologous expression of either Kir4.1 alone or Kir4.1/5.1 in HEK cells as described above we confirmed that Kir4.1 channels do not demonstrate a CO₂/H⁺ sensitive inhibition when exchanging the bathing solution from 5% to 10% CO₂/H⁺ (Supplementary Figure 7A), while Kir4.1/5.1 show robust CO₂/H⁺ sensitivity (Supplementary Figure 7B). Additionally, the slow time course of the CO₂/H⁺ block of RTN astrocyte currents was recapitulated when heterologous pH-sensitive channel K_ir4.1 and 5.1 channels were expressed in HEK cells (Supplementary Figure 7C).

To further support the possibility that heteromeric Kir4.1/5.1 channels confer CO₂/H⁺ sensitivity to RTN astrocytes, we also characterized I_CO2/H+ in RTN astrocytes in slices from Kcnj16 (the gene encoding K_ir5.1) knockout rats (Dahl/salt-sensitive background) recently developed by Palygin et al. (Palygin et al. 2017). Consistent with the possibility that Kir4.1/5.1 channels contribute to RTN astrocyte chemoreception, Kcnj16^−/− rats show a reduced ventilatory response to CO₂ (Puissant et al. 2019). Since this model uses a different genetic background than the SD rats used throughout this manuscript, we first characterized I_CO2/H+ in RTN astrocytes in slices from Dahl/SS control rat PND8–13 pups as described above. We found that exposure to either 10% CO₂ or Ba²⁺ decreased whole cell conductance in RTN astrocytes from control Dahl/SS rats by 58±4 % and 78±3 %, respectively (Figure 7A, C). Both treatments minimally affected conductance in RTN astrocytes in slices from PND8–13 Kcnj16^−/− rats (CO₂: 20±4 %, comparison to WT: t(24)=6.46, p<.0001; Ba²⁺: 50±8 %, comparison to WT: t(24)=3.27, p=.0033; Figure 7B–C). Likewise, the I_CO2/H+ was smaller in RTN astrocytes from Kcnj16^−/− rats compared to control (main effect of genotype: F_(1,24)=10.38, p=.0036, Figure 7E) and reduced Ba²⁺-sensitive currents (main effect of genotype: F_(1,24)=11.44, p=.0025, Figure 7F). Note that Ba²⁺ alone caused a reversible decrease in whole cell conductance in RTN astrocytes from both genotypes (Figure 7A–B), suggesting astrocytes in slices from Kcnj16^−/− rats are healthy and express other pH-insensitive Kir channels. These results provide strong evidence suggesting K_ir4.1/5.1 channels mediate I_CO2/H+ in RTN astrocytes.

Figure 7: Genetic deletion of Kcnj16 causes attenuation of I_Kir and chemosensitivity in RTN astrocytes.

A-C. Representative example (A, B) and summary of inhibition (C, WT n=12, Kcnj16^−/− n=14) of membrane conductance of WT and Kcnj16^−/− RTN astrocytes. In WT astrocytes (A), conductance is markedly inhibited by 10% CO₂ and 100 μM Ba²⁺. The inhibition by both CO₂ and Ba²⁺ inhibition is markedly decreased in Kcnj16^−/− astrocytes (B, C).

D. Membrane potential measurements concur with voltage clamp measurements, as Kcnj16^−/− RTN astrocytes are significantly depolarized compared to controls animals under control conditions (right after seal break, before application of 10% CO₂ and in wash after 10% CO₂). However, when inhibiting control astrocytes with 10% CO₂, Ba²⁺, or both, the two groups have the average membrane potentials that do not significantly differ (n varies from 4 to 14 at each time point).

E-F. I-V plots reveal a significant effect of genotype on both Ba²⁺-sensitive currents (E) and CO₂ sensitive currents (F, WT n=12, Kcnj16^−/− n=14 for both graphs).

In current clamp, RTN astrocytes in slices from Kcnj16^−/− rats showed a more depolarized resting membrane potential under control conditions as compared to RTN astrocytes from Dahl/SS control rats (−81.4±1.4 mV Dahl/SS control vs −72.9±1.7 mV Kcnj16^−/− rats; t(22.24)=4.20, p=.0025). Furthermore, exposure to 10% CO₂ or Ba²⁺ depolarized membrane potential of RTN astrocytes in slices of control rats (in 10% CO₂: V_m=−74.0±2.8 mV, in Ba²⁺: V_m=−68.8±2.8 mV) by an amount similar to RTN astrocytes in slices from SD rats (Figure 7D). Conversely, RTN astrocytes in slices from Kcnj16^−/− rats did not show a voltage response to 10% CO₂ or Ba²⁺ (in 10% CO₂: V_m=−70.7±2.2 mV, t(21.59)=.917, p=.9602 compared to control; in Ba²⁺: V_m=−63.2±2.7 mV, t(21.52)=1.58, p=.6191 compared to control; Figure 7D).

Discussion

Regional Astrocyte Heterogeneity

Astrocyte heterogeneity is the subject of extensive recent interest. Understanding the difference in the physiology of astrocytes in different neuronal circuits aids our understanding of their role within the circuit (Zhang and Barres 2010). Astrocytes exhibit heterogeneous gene expression profiles, giving rise to the notion that astrocytes as a population are comprised of functionally discrete subtypes (Oberheim et al. 2012; Schitine et al. 2015); however, few studies have correlated unique gene or protein expression with function. Here, we show that astrocytes in the RTN but not cortex intrinsically sense CO₂/H⁺ by inhibition of Kir-like currents, establishing Kir channels as key molecular determinants of this astrocyte population. These mechanisms may define populations of chemosensitive astrocytes more broadly.

Collectively, these findings shed light on possible mechanisms underlying functional heterogeneity of astrocytes in distinct brain regions. Potentiation of the CO₂/H⁺ signal and subsequent excitatory neural output is part of a critical regulatory feedback loop in the RTN. Astrocyte CO₂/H⁺-induced K_ir inhibition and NBC-mediated extracellular acidification (Erlichman et al. 1998; Holleran et al. 2001), transmitter release (Gourine et al. 2010; Kasymov et al. 2013; Sobrinho et al. 2017; Wenker et al. 2010; Wenker et al. 2012), and accumulating K⁺ (Largo et al. 1997) may favor an increase in RTN neural firing rate to correct respiratory acidosis. In contrast, astrocytes in the cortex and other areas unassociated with central chemoreception may express lower levels of CO₂/H⁺-sensitive K_ir and primarily serve homeostatic roles by dampening CO₂-mediated acidification through NBC-mediated extracellular alkalization (Theparambil et al. 2017).

Astrocyte heterogeneity across development

Consequent to the overall limited number of electrophysiological studies in RTN astrocytes, the CO₂/H⁺-sensitivity of these cells during the first postnatal week of life has also not previously been determined. This is of significant importance because the ventilatory response to CO₂ in rodents is present at birth and changes dramatically during the first week of life before maturing to a high level in adulthood (Davis et al. 2006; Putnam et al. 2005; Stunden et al. 2001; Wickstrom et al. 2002). The initially low CO₂-sensitivity is thought to coincide with a period of respiratory vulnerability in fetal and early postnatal human life (Darnall 2010). Therefore, to determine whether CO₂/H⁺-sensitivity of RTN astrocytes parallels upregulation of the drive to breathe, we characterized amplitude, kinetics, and voltage-dependent properties of CO₂/H⁺-sensitive currents in RTN astrocytes throughout the first two weeks of postnatal life. For direct comparison, and to evaluate if CO₂/H⁺-sensitivity is ubiquitous, we also characterized properties of CO₂/H⁺-sensitive currents in cortical astrocytes over the same period. Our results show that RTN astrocytes exhibit CO₂/H⁺-sensitive I_Kir (K⁺ current) as early as PND2–3 and that this current increases during early development. These electrophysiological results mimic the behavioral change in the hypercapnic response previously reported in the literature.

The purinergic component of the rodent hypercapnic chemoreceptive response reported by others seems to be absent at PND0–4 and relatively stable from PND7–12 through adulthood (Onimaru et al. 2012). Based on hippocampal astrocytes that do not possess mature I_Kir at that time point, it was suggested that intrinsic neuronal hypercapnic sensitivity may be responsible for in vivo ventilatory drive up until the point at which astrocytes appear in their mature form (PND4–7 at the earliest) (Wenker et al. 2012). In contrast, our data indicate some CO₂/H⁺-sensitive I_Kir in RTN astrocytes as early as PND2–3. This supports the possibility that RTN astrocytes may contribute to in vivo ventilatory drive at or near the time of birth, and heightens the need to consider temporal heterogeneity when investigating the roles of astrocytes in physiological processes.

The increasing chemoreceptor function during the first weeks of postnatal life suggests that upregulation of astrocyte chemosensitivity contributes to the maturation of respiratory control. Since reduced chemoreception during the early neonatal period is associated with life-threatening respiratory failure, these results identify astrocyte K_ir channels as potential therapeutic targets to improve respiratory drive and reduce vulnerability. However, these data do not rule out other pH-sensitive mechanisms in astrocytes.

Molecular underpinning of the CO₂/H⁺-sensitive current

The slow time course of the CO₂/H⁺ dependent block is significant, as the effect of pH on channel block is fast (Pessia et al. 2001). However, we used standard internal solution with HEPES, which may slow the kinetics of intracellular acidification. Indeed, when evaluated in HEK cells with the same conditions, CO₂/H⁺-dependent block of I_Kir is similarly slow. There also exists possible biological explanations for this finding, as previous study work has demonstrated that astrocyte intracellular acidification in response to CO₂/H⁺ occurs at the same rate observed here, and is dependent on bicarbonate transporters (Theparambil et al. 2017).

Ion channel expression by astrocytes varies across different CNS regions (Bordey and Sontheimer 2000), including heterogeneity in expression of the most common K_ir channel, K_ir4.1 (Poopalasundaram et al. 2000). K_ir4.1 is more highly expressed in the brainstem than in the cortex and is developmentally upregulated (Nwaobi et al. 2014), suggesting a potential role in various brainstem functions including chemosensitivity. However, K_ir4.1 is not pH-sensitive in the physiological range (pK_a=5.99 (Pessia et al. 2001)). The correlation between the expression of K_ir4.1, K_ir5.1, and CO₂/H⁺-sensitive current in the RTN suggests that K_ir4.1/5.1 heteromers are involved in astrocytic chemosensation. The pH sensitivity of this channel is in the physiological range (pK_a=7.35 (Pessia et al. 2001)), which together with its high expression in the brainstem makes it a prime candidate pH-sensor in chemosensitive astrocytes. Our research supports this notion, as Kcnj16^−/− rats have diminished CO₂/H⁺-sensitive currents compared to astrocytes from control tissue. Of note, according to the pH dose-response curve published in Pessia et al. 2001, the pH changes we observe here (pH=7.3 at 5% CO₂, pH=7.0 at 10%CO₂) should cause only a small (~10%) inhibition. However, there are several published dose-response curves for K_ir4.1/5.1, and a different one published by the same group (Tucker et al. 2000) shows a steeper curve, that under our conditions inhibit ~80% of I_Kir.

There are several other possible inward rectifying channels that may be involved in pH chemosensation: K_ir5.1 homomers, which can be expressed as a functional channel in the presence of PSD95 (Tanemoto et al. 2002) (though whether this form is pH-sensitive is not known), K_ir4.2 homomers (pK_a=7.1), or K_ir4.2/5.1 heteromers (pK_a=7.6 (Pessia et al. 2001)). However, RNA sequencing data indicates Kcnj15/Kir4.2 is not expressed in brainstem and shows little to no expression in the rest of the CNS (Human Protein Atlas, Uhlén, M. et al., Science 2015). Single cell RNA sequencing in RTN astrocytes and animals with astrocyte-specific conditional knockouts of various combinations are needed to pinpoint the exact molecular determinants of the CO₂/H⁺-sensitive current.

Another possibility is that pH has a separate and distinct sensor on the membrane, while a secondary messenger produced by the sensor affects I_Kir. An additional pH sensor may be an unknown ancillary subunit of Kir4.1 that modulates its pH sensitivity, similar to how the minK subunit modulates K_v7.1 pH sensitivity (Heitzmann et al. 2007) or the effect of β_2A on the pH sensitivity of the voltage gated calcium channel, Ca_v2.1 (Schuhmann et al. 1997).

Of note, there also appears to be a K_ir-independent effect on RTN astrocyte membrane voltage, as depolarization occurs under CO₂/H⁺ exposure when K_ir channels are already blocked by Ba²⁺. NBCe1A, the most common astrocytic bicarbonate transporter (Zhang et al. 2014), represents one possibility for this non-K_ir effect. Consistent with this, inhibition of the NBC decreased amplitude of the CO₂/H⁺-sensitive current in RTN astrocytes (Wenker et al. 2010), suggesting that the NBC contributes to astrocyte chemoreception. The NBC current in RTN astrocytes can be expected to produce a hyperpolarizing outward current (bicarbonate influx) at resting membrane potential. However, exposure to high CO₂ may transiently increase intracellular bicarbonate relative to extracellular bicarbonate due to the activity of intracellular carbonic anhydrase, and cause the NBC instead to extrude bicarbonate (inward depolarizing current). More research is needed to confirm this, and many other possibilities are viable, including other pH-sensitive sensitive cationic channels (like ASIC or TRP), or direct binding of Ba²⁺ to astrocytic proteins.

Physiological Importance

Astrocytic K_ir channels have been long implicated in maintenance of [K⁺]_o homeostasis; after neuronal firing, K⁺ ions are extruded into a relatively small extracellular volume, resulting in large increases in [K⁺]_o that are efficiently dissipated by astrocytic I_Kir (Kofuji and Newman 2004). I_Kir has also proved important for maintaining baseline levels of [K⁺]_o in the hippocampus (D’Ambrosio et al. 2002), and a decrease in this current in murine models of Rett Syndrome and Huntington disease leads to altered potassium dynamics (Kahanovitch et al. 2018; Tong et al. 2014). Thus, we propose that under normal circumstances, I_Kir in RTN astrocytes helps maintain a low level of [K⁺]_o. However, inhibition of Kir4.1/5.1 channels during exposure to high CO₂/H⁺ will conceivably diminish astrocyte K⁺ buffering capacity and potentially enhance activity of nearby RTN neurons to increase respiratory drive. To support the importance of K_ir5.1-dependent I_Kir on CO₂ homeostasis, rats that do not express K_ir5.1 show a reduction in in the ventilatory response to normoxic hypercapnia (Puissant et al. 2019; Trapp et al. 2011). However, this respiratory deficit may involve peripheral chemoreceptors since Kir5.1^−/− mice showed a normal ventilatory response to CO₂ when peripheral chemoreceptor drive was minimized under hyperoxic conditions or in the reduced brainstem-spinal cord preparation (Trapp et al. 2011).

Future Experimental Approaches and Translational Implications

In this study, room temperature was utilized, as higher physiologic temperatures resulted in frequent destabilization of the chamber pH likely due to spontaneous conversion of bicarbonate to CO₂, which has been previously reported (Theparambil et al. 2014). Since I_Kir amplitude and inhibition by CO₂/H⁺ and Ba²⁺ might have a temperature-sensitive component, a future experiment is needed to confirm the CO₂/H⁺ block of these currents in physiological temperatures. Such an experiment would have to stabilize the pH at these temperatures, e.g. by using pressurized chamber to prevent the bicarbonate to CO₂ conversion.

In the current study we utilized a rat reporter line which expresses eGFP under the S100β promoter to aid in the identification of cortical and RTN astrocytes for electrophysiological recordings. This decision was based on previous studies indicating S100β expression is a reliable marker for astrocytes (Grosche et al. 2013; Nishiyama et al. 1996; Ogata and Kosaka 2002; Raponi et al. 2007; Zhu et al. 2008). However, S100β has been shown to be expressed in some populations of NG2 cells in mice (Karram et al. 2008). Confounding this, there is overlap in the electrical properties of NG2 cells and immature astrocytes, the former being more likely to display voltage gated currents and have higher input resistances (Zhong et al., 2016) when compared to mature astrocytes and similar to what has been reported for NG2 cells (Larson et al. 2016, Chittajallu et al. 2004). For instance, using BAC-ALDH1L1-eGFP transgenic mice, Zhong et al., report the vast majority of eGFP⁺ astrocytes (also double labeled with SR101) display variably rectifying currents and have significantly higher input resistances than mature astrocytes. With this in mind, we cannot exclude the possibility that some fraction of cells in our study are S100+ NG2 glia, rather than immature astrocytes. Future studies using these animals will employ a double labeling approach, either using SR101 to positively identify eGFP⁺/SR101⁺ astrocytes (Zhong et al., 2016), or post-hoc immunohistochemistry using astrocyte markers in eGFP⁺ biocytin labeled cells.

These results broadly have interesting implications for neurodevelopmental disorders such as Rett Syndrome, in which genes encoding Kir channels may be abnormally epigenetically regulated during key periods of brainstem respiratory network maturation. Indeed, in Rett Syndrome, Kcnj10/Kcnj16 gene expression levels have been found to be reduced relative to wild-type mice (Kahanovitch et al. 2018; Pacheco et al. 2017). Not only do Rett animals demonstrate abnormal respiratory response to hypercapnia (Bissonnette et al. 2014; Garg et al. 2015; Ren et al. 2012; Voituron et al. 2009), evidence also suggests that astrocytes from Rett animals fail to respond to hypercapnic challenge with appropriate Ca²⁺ influx (Turovsky et al. 2015). Thus, Kir channels represent exciting molecular targets for further investigation of the ways in which astrocyte ion channels influence both normal and abnormal development of respiratory networks.

Materials and Methods

Subjects

All experimental protocols were in accordance with the NIH guidelines and were carried out with approval from the Animal Care and Use Committee of the University of Alabama at Birmingham (Animal Permit Number: 09409) and of Virginia Tech (IACUC: 16–241). Every effort was made to minimize pain and discomfort. Animals used in all experiments were littermates bred in our animal facility by crosses of Sprague Dawley females from Jackson Laboratories to Wistar:Crlj S100b eGFP male rats obtained from the National Bioresource Project Rat (Japan) (Itakura et al. 2007) and subsequently bred out to a Sprague Dawley background. Kcnj16^−/− generation was described before (Palygin et al. 2017). Dahl/SS control rats were obtained from Charles River Laboratories (Wilmington, MA). Animals were housed with water and food available ad libitum on a 12/12 light/dark cycle and all experiments were performed during the animals’ wake (dark) cycle.

Brainstem and Cortical Slices

Mixed gender neonatal rats (PND2-PND13) were rapidly decapitated, brain removed and sliced in half and embedded in 4% agarose in PBS. Coronal brainstem and cortical slices (250μm) were cut using a microtome in ice cold ACSF containing (in mM): 121 NaCl, 3 KCl, 5 MgCl2, 1 CaCl2, 26.2 NaHCO3, 11.1 glucose. Slices were incubated for 30 min at 37°C and subsequently at room temperature in ACSF containing (in mM): 121 NaCl, 3 KCl, 2 MgCl2, 2 CaCl2, 26.2 NaHCO3, 11.1 glucose. All solutions were bubbled with 95% O2–5% CO2 throughout. Coronal sections containing motor cortex analogous to those between approximately +2.52 and −1.32mm relative to bregma in the adult rat were harvested. RTN slices analogous to those located between approximately −10.80mm and −11.88mm relative to bregma in the adult rat were harvested (i.e. at the caudal end of the facial nucleus and the pyramids).

Whole-Cell and Cell-Attached Recordings

Slices containing the motor cortex or RTN were transferred to a recording chamber mounted on a fixed-stage microscope (Zeiss Axio Examiner .D1, Zeiss, Oberkochen, Germany) and perfused (~3–4mL/min⁻¹) with ACSF at room temperature via a pinch-valve syringe delivery system (VC-6, Warner Instruments, Hamden, CT). Room temperature was utilized, as higher physiologic temperatures resulted in frequent destabilization of the chamber pH likely due to spontaneous conversion of bicarbonate to CO₂, which has been previously reported (Theparambil et al. 2014). Solutions were bubbled with 95% O2–5% CO₂ (bath pH=7.3) or 90% O2–10% CO₂ (bath pH=7.0) as indicated throughout experiments.

Whole-cell, voltage-clamp and current-clamp recordings were made from astrocytes in deep layer 1 and 2/3 of the motor cortex and from RTN astrocytes within 5 cell bodies or ~50μm of the pial surface. All recordings were made with an Axopatch 200B patch-clamp amplifier, digitized with Axon Digidata 1550B, and recorded using pCLAMP 10.6 (Molecular Devices, San Jose, CA). Electrodes were pulled from premium thin wall borosilicate glass (Warner Instruments) to a resistance of 4–6 MΩ when filled with internal solution containing the following (in mM): 127mM KCH3SO3, 4 NaCl, 1 MgCl2, 0.5 CaCl2, 10 HEPES, 10 EGTA, 3 Mg-ATP, and 0.3 Na-GTP, .05 Alexa Fluor 568 dye (Thermo Fisher). Resting membrane potentials were measured directly from the amplifier ~1 minute after whole cell access was obtained. Whole cell capacitance and series resistance were measured directly from the amplifier and ~70% compensated to minimize voltage errors. Astrocytes were held at −80mV and series resistance and input resistance were monitored during voltage clamp experiments with a 5 or 10 mV hyperpolarizing voltage step. Voltage step protocol is described in Figure 1A. Full protocol includes voltage steps from −180 mV to 120 mV. Due to brevity, all voltages above 0 mV are not shown, and statistical analysis was performed from −180 mV to −80 mV (see Statistical Analysis section). Changes in extracellular pH were monitored throughout all recordings utilizing a combination pH electrode for micro wells (Warner Instruments).

Activity of RTN neurons was measured in the cell-attached voltage-clamp mode (Vholding=−60 mV) at room temperature (~22°C) with electrodes filled with the same pipette internal solution as above. Action potential discharge was integrated in 10-second bins to construct firing rate histograms using Spike 5.0 software (Cambridge Electronic Design, CED, Cambridge, U.K.). Whole cell voltage clamp recording in neurons (Figure 5D) was performed according to a protocol by (Cui et al. 2018). Neurons were kept at current clamp between voltage clamp protocols that consisted holding at −60 mV, and then 2.5 sec 10 mV voltage steps from −120 mV to −40 mV.

Immunohistochemistry

P12 wildtype males were anesthetized with an intraperitoneal injection of ketamine and xylazine (90 mg/kg and 10 mg/kg, respectively) and perfused with ice cold PBS for 5 minutes followed by 4% paraformaldehyde for 30 minutes. The brain was removed and stored in 4% paraformaldehyde for at least 3 days prior to slicing. After rinsing in PBS, 50 μM coronal slices were cut (Vibratome 3000 sectioning system). Sections were incubated in blocking buffer containing 10% goat serum and .3% Triton-X in PBS for 1 hour shaking at room temperature. Following blocking, slices were incubated shaking overnight at 4°C in the following primary antibodies made up in diluted in blocking buffer (1:3 solution with PBS): 1:1000 mouse anti-GFAP (MAB360, Millipore); 1:200 anti-S100β (Z0311, Dako). Slices were then washed 3×15 minutes in diluted blocking buffer and incubated in 1:750 goat polyclonal to rabbit Alexa Fluor 546 and to mouse Alexa Fluor 488 secondary antibody (Thermo Fisher) for 1 hour shaking at room temperature. Slices were then washed for 5 minutes in PBS and mounted on slides using ProLong diamond antifade mountant (Thermo Fisher). Confocal z-stack images were obtained on a confocal microscope (Zeiss LSM510) utilizing 543 nm and 488 nm lasers.

Protein Analysis by Western Blotting

Mixed sex rats PND3–28 were rapidly decapitated. Brains were removed and cortex and brainstem were dissected from whole tissue in ice cold PBS and sonicated for 14 s in lysis buffer containing 1M Tris pH 7.5, 10% SDS, and 1:10 protease and phosphatase inhibitors. Tissue homogenates were centrifuged for 5 min at 12,000 g at 4°C. Following BCA protein assay (Thermo Fisher), equal amounts of protein (10ug) were heated to 60°C for 15 min in an equal volume of 2X sample buffer (100 mM Tris, pH 6.8, 4% SDS, in Laemmli-sodium dodecyl sulfate, 600 mM B-mercaptoethanol, 200 mM Dithiothreitol (DTT), and 20% glycerol) and electrophoresed on 4–20% Criterion TGX precast gels (BioRad, Hercules, CA) for 1.5 hours at 150V at room temperature. Protein was transferred onto nitrocellulose membranes (Licor, Lincoln, NE) at 10V overnight spinning at 4° C using two layers of sponges and cold packs. Membranes were blocked in 1:1 TBS and TBS blocking buffer (Licor) shaking for 1 hour at room temperature. Membranes were incubated shaking, in abovementioned blocking buffer with .2% Tween added, with the following primary antibodies: chicken polyclonal to alpha tubulin at .25ug/mL (Abcam, Cambridge, UK) for 40 minutes at room temperature, 1:1000 rabbit polyclonal to Kir4.1 (Alomone, Jerusalem, Israel) for 1 hour at room temperature, 1:1000 rabbit polyclonal to Kir5.1 (Abcam) for 3 hours at room temperature. Membranes were then washed 3×5 min in TBST and incubated shaking with secondary antibodies, prepared in the same solution as primary antibodies: 1:20,000 goat anti rabbit IR 800 and donkey anti chicken IR 680 (Licor) for 40 minutes at room temperature. Membranes were then washed 3×5 minutes in TBST and for 5 minutes in TBS before being imaged with an Odyssey scanner. Bands were quantified using Licor ImageStudio Version 4.0 and normalized to alpha tubulin. To confirm that protein expression did not differ between sexes, sets of blots in which both male and female samples were run together were also analyzed and demonstrated no sex-dependent difference in Kir4.1 or Kir5.1 protein expression in brainstem or cortex (data not shown).

Drugs

All experiments included TTX (0.1μM, Tocris Bioscience, Bristol, UK) in the recording solution except where indicated (Supplementary Figure 4). Barium chloride (100μM Thermo Fisher) was also used where indicated. Drugs were stored in frozen stock solution and dissolved in recording solution prior to each experiment.

Statistical Analysis

All raw data were graphed and statistically analyzed in GraphPad Prism 7.00 for Windows (GraphPad Software, La Jolla California USA). Statistical tests used are detailed in Supplementary Table 1. Voltage step protocol was analyzed for the range −180 to −80 mV, to avoid symmetry around resting V_m that would dampen differences between regions/genotype. We chose this method over analyzing current at a single voltage step, as to not influence analysis with that choice. We also preferred not to analyze slope, as the I-V curve is linear only in some of the voltage steps. All data are reported as mean±standard error of the mean within the text, with results of statistical tests and p values. n, number of cells (electrophysiology) or animals (Western blotting) are indicated in figure legends. * indicates p<.05, ** indicates p<.01, *** indicates p<.001, **** indicates p<.0001.

Main Points:

Astrocytes in the RTN but not cortex intrinsically sense CO₂/H⁺ by inhibition of a K_ir conductance

Inhibition of astrocyte K_ir is sufficient to increase RTN neuronal firing

Chemosensitivity of RTN astrocytes is dependent on K_ir4.1/5.1 channels

Abbreviations:

Ba²⁺

barium

Ca²⁺

calcium

C_m

cell capacitance

CO₂

carbon dioxide

DTT

Dithiothreitol

eGFP

enhanced green fluorescent protein

E_K

potassium equilibrium/reversal potential

GFAP

glial fibrillary acidic protein

H⁺

hydrogen ions

I-V

current-voltage

I_CO2/H+

CO₂/H⁺-sensitive current

I_Kir

Kir channel current

Kir

inward rectifying potassium

[K⁺]_o

extracellular potassium

Kcnj10

gene encoding Kir4.1 protein

Kcnj16

gene encoding Kir5.1 protein

NG2

polydendrocyte cell

PND

postnatal day

RTN

retrotrapezoid nucleus

S100β

S100 calcium binding protein beta

SD

Sprague Dawley

TTX

tetrodotoxin

V_m

resting membrane potential

wt

wild-type