The astrocytic Na+‐HCO3− cotransporter, NBCe1, is dispensable for respiratory chemosensitivity

The interoceptive homeostatic mechanism that controls breathing, blood gases and acid‐base balance in response to changes in CO2/H+ is exquisitely sensitive, with convergent roles proposed for chemosensory brainstem neurons in the retrotrapezoid nucleus (RTN) and their supporting glial cells. For astrocytes, a central role for NBCe1, a Na+‐HCO3− cotransporter encoded by Slc4a4, has been envisaged in multiple mechanistic models (i.e. underlying enhanced CO2‐induced local extracellular acidification or purinergic signalling). We tested these NBCe1‐centric models by using conditional knockout mice in which Slc4a4 was deleted from astrocytes. In GFAP‐Cre;Slc4a4fl/fl mice we found diminished expression of Slc4a4 in RTN astrocytes by comparison to control littermates, and a concomitant reduction in NBCe1‐mediated current. Despite disrupted NBCe1 function in RTN‐adjacent astrocytes from these conditional knockout mice, CO2‐induced activation of RTN neurons or astrocytes in vitro and in vivo, and CO2‐stimulated breathing, were indistinguishable from NBCe1‐intact littermates; hypoxia‐stimulated breathing and sighs were likewise unaffected. We obtained a more widespread deletion of NBCe1 in brainstem astrocytes by using tamoxifen‐treated Aldh1l1‐Cre/ERT2;Slc4a4fl/fl mice. Again, there was no difference in effects of CO2 or hypoxia on breathing or on neuron/astrocyte activation in NBCe1‐deleted mice. These data indicate that astrocytic NBCe1 is not required for the respiratory responses to these chemoreceptor stimuli in mice, and that any physiologically relevant astrocytic contributions must involve NBCe1‐independent mechanisms.


Introduction
The RTN neurons in the ventral parafacial region of the rodent brainstem contribute to direct sensing of CO 2 /H + in service of homeostatic regulation of breathing, representing an anatomic substrate for the long-sought ventral medullary central chemoreceptors Guyenet et al., 2019;Mulkey et al., 2004). The central respiratory chemoreflex -increased breathing in response to elevated CO 2 -is exquisitely sensitive, and multiple mechanisms have been proposed to account for this sensitivity Guyenet et al., 2019). These include direct actions of intrinsically chemosensory RTN neurons, as well as indirect modulatory actions by other chemosensory neurons and glia that are largely intermediated by RTN neurons Guyenet et al., 2019;Hawkins et al., 2017;Hodges & Richerson, 2010;Souza et al., 2018). Irrespective of whether their excitation arises from direct or indirect CO 2 /H + sensing mechanisms, chemosensitive RTN neurons convey a facilitatory drive onward to the various respiratory centres that dictate the rate and depth of breathing (Del Negro et al., 2018;Guyenet et al., 2019).
For RTN neurons, the intrinsic pH sensing mechanism has been attributed to two molecular proton detectors -TASK-2 and GPR4 -that together contribute a major component of the overall chemosensitivity (Gestreau et al., 2010;Guyenet et al., 2019;Kumar et al., 2015;Wang et al., 2013). For the proposed glial contributions, the molecular mechanisms are less certain, but increased activity of an electrogenic Na + -HCO 3 − co-transporter, NBCe1 (encoded by Slc4a4), has been proposed to play a critical role in two different non-exclusive processes initiated by RTN astrocytes: (1) 'proton boosting' whereby local pH changes are amplified (Chesler, 2003;Erlichman et al., 2008;Guyenet et al., 2019); and (2) purinergic paracrine signalling, based on astrocytic ATP release Turovsky et al., 2016). A separate NBCe1-independent mechanism involving direct activation of astrocytic, ATP-releasing connexin hemichannels by CO 2 has also been suggested to contribute to local purinergic signalling Huckstepp, id Bihi et al., 2010;Meigh et al., 2013).
For the two proposed NBCe1-dependent astrocytic mechanisms, intracellular acidification and/or membrane depolarization induced by CO 2 increases NBCe1 activity, resulting in enhanced HCO 3 − transport into the astrocyte (Chesler, 2003;Grichtchenko & Chesler, 1994;Theparambil et al., 2017;Turovsky et al., 2016), effectively removing proton buffering equivalents from the extracellular space (Erlichman et al., 2008;Guyenet et al., 2019;Mulkey & Wenker, 2011;Wenker et al., 2010). Concurrently, the increased NBCe1 activity also promotes entry of Na + ions, driving a secondary Na + -Ca 2+ exchange mechanism that provides the Ca 2+ needed for ATP release (Turovsky et al., 2016). In the latter instance, ATP-mediated purinergic signalling either directly activates RTN neurons , or it causes a local vasoconstriction to reduce washout of metabolic byproducts (i.e. CO 2 ), further enhancing acidification Guyenet et al., 2019;Hawkins et al., 2017). Accordingly, because of its central position in these proposed mechanisms, deletion of NBCe1 from astrocytes is expected to eliminate its contributions to glial respiratory chemosensation and thereby reduce CO 2 -stimulated breathing.
To test these proposed mechanisms, we used two different Cre driver lines for conditional deletion in mouse astrocytes of Slc4a4, the gene encoding NBCe1. Despite suppressing Slc4a4 expression and NBCe1-associated currents in astrocytes, we found no effect of astrocytic Slc4a4 depletion on CO 2 -induced RTN neuronal activation or CO 2 -stimulated ventilation; hypoxia-activated breathing and sighs were also unaffected. Based on these results, we conclude that NBCe1-mediated astrocytic mechanisms are not required for respiratory chemosensitivity in mice.

Ethical approval
Experiments were performed following procedures adhering to National Institutes of Health Animal Care and Use Guidelines and approved by the Animal Care and Use Committee of the University of Virginia (Protocol no. 2454). The investigators understand the ethical principles under which the journal operates, and this work complies with the animal ethics checklist as outlined by the journal (Percie du Sert et al., 2020).

Animals
Experiments were performed on mice of either sex. Mice were housed in HEPA-ventilated racks and steam-sterilized caging (up to 5 per cage), with ad libitum access to food and water. Animals were exposed to 12 h light/dark cycles in a vivarium maintained at 22−24°C and ∼40-50% relative humidity. For electrophysiological recordings from RTN neurons and RTN-adjacent astrocytes, we used a Phox2b-GFP BAC transgenic mouse line (Jx99) in which GFP expression is driven by the Phox2b promoter; these mice were developed by the GENSAT project, maintained in-house, and characterized previously (Lazarenko et al., 2009). A conditional deletion-ready mouse line was obtained in which exon 12 of Slc4a4, the gene encoding NBCe1, was flanked by loxP sites (kindly provided by Dr Gary E. Shull, University of Cincinnati); the 'floxed' exon is included in all splice variants of NBCe1 (Vairamani et al., 2018). We knocked out NBCe1 in astrocytes by crossing, in-house, the 'floxed' Slc4a4 mice with either GFAP-Cre transgenic mice (RRID: IMSR_JAX:012886) (Garcia et al., 2004) or tamoxifen-inducible Aldh1l1-Cre/ERT2 BAC transgenic mice (RRID: IMSR_JAX:031008) (Winchenbach et al., 2016); after intercrossing Cre-positive offspring heterozygous at the Slc4a4 locus, we established breeding between Cre-positive female Slc4a4 fl/fl and Cre-negative male Slc4a4 fl/fl mice to obtain homozygous male and female Slc4a4 fl/fl littermates that were either Cre-negative or Cre-positive (GFAP line, N = male: Cre -, 24; Cre + , 30; N = female: Cre − , 40; Cre + , 36; Aldh1l1 line, N = male: Cre − , 53; Cre + , 39; N = female: Cre − , 20; Cre + , 35). To induce Cre expression in Aldh1l1-CreERT2 mice, adult mice (>70 days old) were treated for 10 consecutive days with tamoxifen (150 μl/mouse/day, i.p., from 20 mg/ml stock in peanut oil; Sigma-Aldrich, St Louis, MO, USA); the vehicle and tamoxifen-treated mice were studied after 3 weeks. All studies were performed and analysed by individuals blinded to genotype and experimental treatment.

Breathing measurements
Ventilatory variables were measured in conscious adult mice by whole body plethysmography, following the manufacturer's instructions (EMKA Technologies, Sterling, VA, USA) and as previously described (Kumar et al., 2015;Shi et al., 2016). A mass flow regulator J Physiol 601.16 provided quiet, constant and smooth flow through the animal chamber (0.5 l/min), and the pressure transducers were calibrated by injecting a known volume of air with a syringe into the chamber (10 ml, <2 s). Mice were habituated to plethysmography chambers maintained at 25−28°C for 4 h at least one day prior to testing. The protocol included three sequential incrementing CO 2 challenges (7 min exposures to 2%, 4%, 6%, 8% CO 2 , combined with 60% O 2 , balance N 2 ; each separated by 5 min of 60% O 2 /40% N 2 ). Hypercapnic exposure was performed in hyperoxia to minimize contributions of peripheral chemoreceptors to the hypercapnic ventilatory reflex (Basting et al., 2015) and to attribute ventilatory effects to central chemoreceptors. For hypoxia challenges, mice were exposed to 10% O 2 , balance N 2 for up to 5 min.

Analysis of ventilatory responses
Ventilatory flow signals were recorded, amplified, digitized and analysed using Iox 2.7 (EMKA Technologies) to determine ventilatory parameters over sequential 20 s epochs (∼50 breaths), during periods of behavioural quiescence and regular breathing. Minute ventilation (V E , ml/min) was calculated as the product of respiratory frequency (f R , breaths/min) and tidal volume (V T , ml/breath), and normalized to body weight. In short, for each breath, the Iox software determines inspiratory time (T i ) and expiratory time (T e ) from the zero-flow crossing time point as the flow trace alternates between negative (inspiration) and positive (expiration) values, and breathing frequency is calculated from the sum of T i and T e . The area under the flow curve during inspiration (i.e. below the zero-flow point) is taken as V T . For analysis of the hypercapnic ventilatory response, we sampled 10 consecutive epochs (200 s, representing ∼400-500 breaths at rest) that showed the least inter-breath irregularity during the steady-state plateau period after each CO 2 exposure, as determined by Poincaré analysis (Kumar et al., 2015;Loria et al., 2013). The response to hypoxia (10% O 2 ) was determined from the peakV E (20 s epoch) within the 5 min of exposure to the hypoxic gas mixture, and the incidence of hypoxia-induced sighs was determined throughout the last 3 min of hypoxic exposure (Kumar et al., 2015). There were no Cre-dependent and/or tamoxifen-dependent differences in breath-to-breath variability measured at any level of CO 2 or O 2 in the different mouse lines (i.e. Poincaré ellipse method; Loria et al., 2013).

Fos expression after exposure to hypercapnia in vivo
Activation of RTN neurons and astrocytes by elevated CO 2 or O 2 was assessed by Fos expression (Shi et al., 2016(Shi et al., , 2017. In short, adult mice were habituated to the plethysmography chamber for 1 h in hyperoxia (60% O 2 , balance N 2 ) before an additional 35 min exposure to hyperoxia (60% O 2 , balance N 2 ) or hypercapnia (12% CO 2 , 60% O 2 , balance N 2 ). Immediately following exposure to the test condition, mice were anaesthetized (ketamine, 75 mg/kg; xylazine, 5 mg/kg; i.p.), examined for absence of response to a firm toe pinch and perfused transcardially (as below). All experiments, including the histology, were run in pairs (i.e. mice exposed to hypercapnia and hyperoxia). Sections were prepared for multiplex fluorescence in situ hybridization, as described below.

Multiplex fluorescence in situ hybridization (FISH)
Mice were anaesthetized with ketamine/xylazine (75 mg/kg, 5 mg/kg; i.p.), examined for absence of response to a firm toe pinch and perfused transcardially with 4% PFA/0.1 m PB. Brains and kidneys were removed, immersed in the same fixative for 16−18 h at 4°C, cut in the transverse plane (30 μm) and placed in cryoprotectant (30% ethylene glycol, 20% glycerol, 50 mm sodium phosphate buffer, pH 7.4) at −20°C until further processing. Tissue preparation and staining procedure utilized the RNAscope Multiplex Fluorescent Assay (Advanced Cell Diagnostics, Hayward, CA, USA; RRID: SCR_012481), according to the manufacturer's instructions and as previously described (Shi et al., 2017). Catalogue probes were used for Aldh1l1, Nmb, Gfap, Slc4a5, and Fos, whereas a custom Slc4a4 probe was designed specifically to encompass the floxed exon 12 that was deleted by Cre-mediated excision (Cat. no. 533358). Following hybridization, sections were air-dried, counterstained with ProLong Gold antifade reagent containing DAPI, and cover slipped for further analysis.

Cell counts and analysis
Serial coronal sections (1:3 series) through the rostrocaudal extent of the RTN were mounted on glass slides, and images were acquired using an epifluorescence microscope (Zeiss Axioimager Z1) equipped with Neurolucida software (MBF Bioscience, Williston, VT, USA). One section from each mouse was used for cell counts and was chosen from the rostrocaudal level where maximal numbers of RTN neurons are located (6.48 mm caudal to bregma; Shi et al., 2017). In the two mouse lines, there were no Cre-dependent and/or tamoxifen-dependent differences in the average number of RTN neurons counted from these sections (44.2 ± 12.7 Nmb + neurons, range 21-76, from N = 42 mice). We counted FISH-labelled cells with DAPI-counterstained nuclei located in the RTN region within ∼100 μm of the ventral medullary surface. We considered cells with at least 3 Fos-associated puncta as Fos-positive. Using these criteria, CO 2 -activated RTN neurons were defined as Fos-expressing Nmb + , DAPI-stained neurons and CO 2 -activated astrocytes were defined as Fos-expressing Aldh1l1 + , DAPI-stained profiles. No stereological correction factor was applied. The investigator assessing and quantifying cellular profiles was blinded to treatment.

Blood gas analysis
Adult mice were habituated to a tail warmer and restraint apparatus (Braintree Scientific, Inc., Braintree, MA, USA) for 30 min one day before, and then for an additional 30 min immediately prior to blood sampling. For blood sampling, awake mice were gently restrained, and the tail was slightly warmed before arterial blood was collected from the ventral tail artery (∼100 μl) into heparinized capillary tubes for immediate analysis using a handheld blood gas analyser (iSTAT, with CG4+ cartridge, Abbott Point of Care Diagnostics, Princeton, NJ, USA) (Kumar et al., 2015).

Statistics
All statistical analyses were performed using GraphPad Prism (v. 9.5); details of specific tests are provided in the text or figure legends, and parametric tests were used when data were normally distributed (Shapiro-Wilk test). Data are presented in box and whiskers format (the median bisects a box bounded by the 25 percentile and 75 percentile, with whiskers depicting the range), or as mean ± standard deviation (SD). Statistical significance was set at P < 0.05.

Results
We generated conditional knockout mice to test the role of astrocytic NBCe1 in the central respiratory chemoreflex. To eliminate NBCe1 expression in astrocytes, we obtained mice in which loxP sites surrounded the 12 th exon of Slc4a4 (Vairamani et al., 2018), the NBCe1 gene, and crossed those animals with GFAP-Cre or Aldh1l1-Cre/ERT2 mice in which Cre recombinase is expressed primarily in astrocytes in the CNS (Garcia et al., 2004;Winchenbach et al., 2016). We validated deletion of Slc4a4 in astrocytes using histochemical, molecular and/or electrophysiological approaches, and then examined neuron/astrocyte activation (Fos) and ventilatory responses in vivo to raised CO 2 (hypercapnia) or lowered O 2 (hypoxia).

Astrocytic NBCe1 expression is reduced in GFAP-Cre conditional knockout mice
We used multiplexed fluorescence in situ hybridization (FISH, by RNAscope) to examine Slc4a4 expression in astrocytes (identified by their expression of Aldh1l1) within the RTN region of conditional knockout mice. As previously described, the RTN comprises a distinct cluster of neurons, defined in part by selective expression of Nmb, that reside between the facial motor nucleus and the ventral surface of the rostral medulla oblongata (Shi et al., 2017). In the region of Nmb-expressing RTN neurons from Cre-negative control animals (Slc4a4 f/f ), we found robust expression of Slc4a4 transcripts in Aldh1l1-expressing astrocytes (Fig. 1A). By contrast, in Cre-positive GFAP-Cre;Slc4a4 fl/fl mice, Slc4a4 expression was strongly reduced in Aldh1l1-expressing astrocytes (Fig. 1B); notably, Slc4a4 depletion was observed most prominently in astrocytes located near the ventral medullary surface, while more dorsally located Aldh1l1-expressing astrocytes retained Slc4a4 expression. This differential depletion of Slc4a4 in ventral medullary surface astrocytes in GFAP-Cre mice appears to reflect differential expression of Gfap, which is often a marker of activated astrocytes and can be variably expressed among astrocytic populations, including preferentially at greater levels in surface astrocytes in the RTN region (see Fig. 5A) (Angelova et al., 2015;SheikhBahaei et al., 2018;Walz & Lang, 1998). We should also note that Cre expression has been reported in cells other than astrocytes in this GFAP-Cre mouse line (JAX:012886). However, since Slc4a4 expression is largely localized to astrocytes in control mice (Theparambil et al., 2020), it is likely that any effects in these mice are due preferentially, if not exclusively, to deletion of Slc4a4 from those cells.
The related NBCe2 transporter is prominently expressed in epithelium of the choroid plexus (Christensen et al., 2018;Damkier et al., 2013), but minimal Slc4a5 expression is expected normally for astrocytes and neurons. Consistent with this, we found little evidence for Slc4a5 expression in any cells within the RTN region; this was true for both Cre-negative and Cre-positive Slc4a4 fl/fl mice, also suggesting no compensatory upregulation of NBCe2 expression after deletion of NBCe1 ( Fig. 1C and D). Likewise, the strong Slc4a5 expression observed in the choroid plexus was unaffected in Cre-positive mice ( Fig. 1E and F).
To quantify deletion of Slc4a4 in control and conditional knockout mice we also performed qRT-PCR using cDNA obtained from the RTN region that was micro dissected from acute brain slices (Fig. 1G); we again also looked for any compensatory changes in Slc4a5 expression. Consistent with the FISH results, levels of Slc4a4 transcripts were significantly reduced in GFAP-Cre;Slc4a4 fl/fl mice, by comparison to control  Slc4a4 fl/fl mice (by ∼74%; 2 −Ct = 0.0066 ± 0.0040 vs. 0.0018 ± 0.0009 for Cre − and Cre + , N = 7 and 5 mice; P = 0.0025, Mann-Whitney U test), whereas Slc4a5 expression was found at equally low levels in both control and conditional knockout mice (2 −Ct = 0.0003 ± 0.0003 vs. 0.0003 ± 0.0002 for Cre − and Cre + , N = 6 and 5 mice;

NBCe1 function is disrupted in astrocytes from GFAP-Cre conditional knockout mice
We assessed NBCe1 function by recording from astrocytes in brain slices from control (Slc4a4 fl/fl ) and conditional GFAP-Cre;Slc4a4 fl/fl knockout mice. Astrocytes were identified in the RTN region by their distinctive size and shape, and their uptake of SR101 dye ( Fig. 1H) (Kafitz et al., 2008;Nimmerjahn et al., 2004). In addition, by whole-cell voltage clamp recording, they presented with a characteristically linear current-voltage (I-V) relationship and zero-current potential near the predicted E K (Fig. 1I). In slices from Cre − control mice and Cre + knockout mice, the input resistances (R N ) for all astrocytes recorded under resting condition were, respectively, 31.2 ± 16.4 MΩ and 47.0 ± 29.7 MΩ, with a corresponding zero current potentials of −94.5 ± 3.3 mV and −93.7 ± 3.5 mV (n = 26 and 39 cells, N = 5 and 10 mice). It should be noted that SR101 can label other cell types, especially at concentrations higher than we used here (Hulsmann et al., 2017). However, by targeting cells with these additional morphological features and characteristic membrane properties, along with the NBCe1-like currents described below, we believe this approach has appropriately identified the recorded cells as astrocytes.

CO 2 -and hypoxia-stimulated breathing is unaffected by NBCe1 deletion in astrocytes from GFAP-Cre mice
We used whole animal plethysmography to test ventilatory responses to CO 2 and hypoxia in control mice with intact NBCe1 and in GFAP-Cre-expressing littermates with deletion of NBCe1 in RTN-adjacent astrocytes (Fig. 2). Conscious mice were exposed to increasing concentrations of CO 2 in the inspired air, and with constant hyperoxia (60% O 2 , balanced with N 2 ) to minimize effects of peripheral chemoreceptors and emphasize the centrally mediated respiratory chemoreflex (Basting et al., 2015). As evident in the exemplar respiratory flow traces and grouped data ( Fig. 2A and B), the stimulatory effect of different levels of CO 2 on minute ventilation (V E ; ml/min/g) was indistinguishable between Cre − and Cre + mice (F 4,110 = 0.1181, P = 0.9758 for interaction, two-way ANOVA, N = 11 and 13), with no difference at the highest CO 2 tested (V E in 8% CO 2 : 1.26 ± 0.27 ml/min/g vs. 1.35 ± 0.34 ml/min/g, N = 11 and 13; P = 0.5172, unpaired t test). This reflected essentially identical CO 2 -dependent effects on both respiratory frequency (f R , F 4,110 = 0.1654, P = 0.9555 for interaction, two-way ANOVA; f R at 8% CO 2 : 48.3 ± 24.6 breaths/min vs. 43.6 ± 29.6 breaths/min for Cre − and Cre + , N = 11 and 13, P = 0.6771, unpaired t test) and tidal volume (V T , F 4,110 = 0.01937, P = 0.9993 for interaction, two-way ANOVA; V T at 8% CO 2 : 4.6 ± 1.5 μl/breath/g vs. 4.9 ± 1.5 μl/breath/g for Cre − and Cre + , N = 11 and 13, P = 0.6781, unpaired t test). Thus, NBCe1 deletion from astrocytes had little overall effect on CO 2 -stimulated breathing.
We also tested whether NBCe1 is required for activation of RTN neurons and astrocytes by CO 2 in vitro. We performed cell-attached recordings of action potential firing from GFP-expressing RTN neurons in brainstem slices from Phox2b-GFP mice during exposure to 5% CO 2 (pH = 7.4) and 10% CO 2 (pH = 7.1), under control conditions and in the presence of S0859 (30 μM). As shown in Fig. 4A, we found no difference in firing rate under control conditions and in the presence of S0859 during exposure to either 5% CO 2 (1.9 ± 1.5 Hz vs. 1.7 ± 1.6 Hz) or 10% CO 2 (2.9 ± 1.8 Hz vs. 2.7 ± 1.6 Hz; n = 10, N = 6; F 1,18 = 0.09559 for control vs. S0859, P = 0.7607, two-way RM-ANOVA); under both conditions, CO 2 had essentially identical effects on RTN firing rate ( Hz = 1.1 ± 0.6 Hz and 1.0 ± 0.5 Hz, for control vs. S0859; n = 10, N = 6; P = 0.7730, paired t test). In addition, we performed whole cell recordings of CO 2 -induced current from RTN astrocytes in brainstem Figure 2. Effects of CO 2 and O 2 on minute ventilation and sigh frequency are unaffected by NBCe1 deletion from astrocytes in GFAP-Cre;Slc4a4 fl/fl mice A, whole body plethysmography recordings from unanaesthetized Cre-negative and Cre-positive GFAP-Cre;Slc4a4 fl/fl mice at increasing levels of inspired CO 2 (in 60% O 2 , balance N 2 ). Dotted line shows approximate inspiratory-expiratory transition. B, effects of increasing inspired CO 2 (from 0% to 8%; 60% O 2 , balance N 2 ) on minute ventilation (left; mean ± SD; F 4,110 = 0.1181, P = 0.9758 for interaction, by two-way ANOVA) and the change inV E evoked by 8% CO 2 (right; P = 0.5172, by unpaired t test) were not different in Cre-negative and Cre-positive GFAP-Cre;Slc4a4 fl/fl mice (N = 11 and 13 mice). C, effects of hyperoxia (60% O 2 , balance N 2 ) and hypoxia (10% O 2 , balance N 2 ) on minute ventilation (left; mean ± SD; F 2,63 = 0.3825, P = 0.6837 for interaction, by two-way ANOVA) and frequency of hypoxia-induced sighs (right; mean ± SD; F 1,42 = 0.9778, P = 0.3284 for interaction, by two-way ANOVA) in Cre-negative and Cre-positive GFAP-Cre;Slc4a4 fl/fl mice (N = 11 and 12 mice, respectively). [Colour figure can be viewed at wileyonlinelibrary.com] J Physiol 601.16 slices obtained from GFAP-Cre;Slc4a4 fl/fl mice. A change from 5% CO 2 to 10% CO 2 in the perfusate consistently evoked an inward current in astrocytes from both Cre − and Cre + mice (Fig. 4B), and although this CO 2 -induced current appeared to trend towards smaller amplitudes in NBCe1-deleted astrocytes (Fig. 4C), the values were not significantly different between genotypes (63.3 ± 40.6 pA and 40.0 ± 32.8 pA, n = 22 and 17, N = 5 and 6; P = 0.0673, Mann-Whitney U test). In addition, we also found no difference in effects of CO 2 on membrane potential calculated in those same cells from Cre − and Cre + mice ( Fig. 4C; 1.8 ± 1.6 mV and 1.9 ± 1.8 mV, P = 0.7314, Mann-Whitney U test).
Together with the plethysmography results, these in vivo and in vitro data indicate that elimination of NBCe1 from GFAP-expressing astrocytes has no significant effect on activation of RTN neurons or astrocytes by CO 2 and does not alter the central respiratory chemoreflex.

Chemoreflex-stimulated breathing is retained when
Slc4a4 is deleted broadly in astrocytes of Aldh1l1-Cre/ERT2 mice As described above, the floxed Slc4a4 gene was effectively excised in ventral medullary astrocytes near the RTN using the GFAP-Cre mouse line, but Slc4a4 expression was not eliminated in numerous other nearby astrocytes.
This probably reflects elevated expression of Gfap that is restricted mainly to activated astrocytes (Angelova et al., 2015;Walz & Lang, 1998). Indeed, we find very high levels of Gfap transcripts in astrocytes in the vicinity of Nmb-expressing RTN neurons near the ventral medullary surface  whereas Gfap was undetectable in many other astrocytes identified by Aldh1l1 expression, a more broadly expressed marker (Fig. 5A).
Therefore, we sought to obtain more widespread astrocytic depletion of NBCe1 by using a mouse line with Cre expression driven from the Aldh1l1 locus. In initial work, we were unable to obtain Cre + pups after crossing Aldh1l1-Cre mice with Slc4a4 fl/fl animals, suggesting that early deletion of NBCe1 from astrocytes was not compatible with survival. However, we found we could use a tamoxifen-inducible Aldh1l1-Cre/ERT2 line to obtain Cre − and Cre + littermates for successful deletion of Slc4a4 by treating those adult animals with tamoxifen.
We verified Cre-and tamoxifen-dependent astrocyte deletion of Slc4a4 in this mouse model by FISH, as shown in Fig. 5B-E. Specifically, Slc4a4 expression was strongly reduced in Aldh1l1-expressing astrocytes of Cre-positive Slc4a4 fl/fl mice that were treated with tamoxifen (Cre + :T), by comparison to the three different control groups: Cre-negative mice treated with either vehicle or tamoxifen (Cre − :V and Cre − :T), or Cre-positive mice treated with vehicle (Cre + :V). Unlike with the GFAP-Cre mice, depletion of Slc4a4 expression in the RTN region in these Cre + :T mice was not limited only to the astrocytes near the ventral medullary surface (cf. Figs 5E and 1B), and indeed was much more widespread throughout the brainstem (not shown).
We examined the effects on breathing of chemoreceptor stimuli, both elevated CO 2 and variable O 2 , in these inducible, conditional knockout mice. There was no difference in the peak stimulatory effect of CO 2 on breathing among the different experimental groups before or after vehicle or tamoxifen treatment ( Fig. 6A; pre: F 3,27 = 0.5080, P = 0.6801; post: F 3,27 = 2.069, P = 0.1279, ANOVA). Although there was no difference in the stimulatory effect of increasing levels of CO 2 onV E in the experimental groups post-treatment (Fig. 6B, inset: F 12,135 = 0.8097, P = 0.6401, for interaction by two-way ANOVA) we did note a significantly higher overallV E in the Cre-positive, tamoxifen-treated NBCe1 knockout mice across CO 2 levels (F 3,135 = 16.66, P < 0.0001 for genotype-treatment by two-way ANOVA); this was particularly apparent when the Cre + :T response was compared to the 95% confidence interval of the combineḋ V E response to CO 2 for the three control groups (Fig. 6B). The higherV E in Cre + :T mice was largely due to differences in V T . Thus, the V T for Cre + :T was elevated across the CO 2 range over the other three experimental groups, which were not different from each other (F 3,135 = 37.83, P < 0.0001 for genotype-treatment by two-way ANOVA; at least P < 0.05 for Cre + :T vs. each of the other groups at all CO 2 levels, Tukey's multiple comparison test); on the other hand, there were no differences in f R among the experimental groups at any CO 2 level (F 3,135 = 3.605, P < 0.0152 for genotype-treatment by two-way ANOVA; at least P > 0.10 for Cre + :T vs. each of the other groups at all CO 2 levels, Tukey's multiple comparison test). Likewise, there was no difference between any of the experimental groups in hypoxia-stimulated breathing before or after vehicle or tamoxifen treatment ( Fig. 6C; pre: F 3,24 = 1.905, P = 0.1557; post: F 3,21 = 2.104, P = 0.1302, ANOVA), and the higher overallV E response for the Cre + :T conditional knockouts was again apparent by comparison to the three control groups ( Fig. 6D; F 1,685 = 16.34, P = 0.0001, for genotype-treatment by two-way ANOVA). Finally, hypoxia-induced sighs were not different across any of the experimental groups, before or following vehicle/tamoxifen treatment ( Fig. 6E; pre: F 3,48 = 0.2274, P = 0.8769; post: F 3,42 = 1.038, P = 0.3857, two-way ANOVA). So, these ventilatory responses to CO 2 and hypoxia were unaltered by NBCe1 deletion even as Figure 4. Effects of CO 2 on RTN neurons and astrocytes after blocking NBCe1 or deleting Slc4a4 in astrocytes A, left: effect of CO 2 on action potential firing rate in an individual RTN neuron in a brainstem slice from a Phox2b::GFP mice, tested under control conditions and after blocking NBCe1 with S0859 (30 μM). A, right: summary data (n = 10 cells, N = 6 mice) shows that firing rate was increased by CO 2 in both conditions (F 1,18 = 58.75 for CO 2 , P < 0.0001, by two-way ANOVA), but there was no effect of S0859 on firing rate in either 5% CO 2 or 10% CO 2 (F 1,18 = 0.09559 for S0859, P = 0.7607). B, exemplar records of CO 2 -induced current in astrocytes from Cre-negative (upper) and Cre-positive (lower) GFAP-Cre;Slc4a4 fl/fl mice. C, summary data reveals no significant difference in CO 2 -induced current or membrane depolarization from control and NBCe1-deleted astrocytes (by Mann-Whitney U test; n = 22 and 17 cells, N = 5 and 6 mice). [Colour figure can be viewed at wileyonlinelibrary.com] J Physiol 601.16 the overallV E was shifted to higher levels in Cre + :T mice.
We again used Fos expression to examine the stimulatory effect of CO 2 on RTN neurons and surrounding astrocytes in mice after NBCe1 depletion from astrocytes. As illustrated in the exemplar photomicrographs from RNAscope FISH (Fig. 7A-D), CO 2 -induced Fos expression was equally prominent in Nmb-expressing RTN neurons and Aldh1l1-expressing astrocytes after NBCe1 depletion (i.e. in Cre + mice treated with tamoxifen) as in the various control groups. This was borne out in quantitative cell counts, which revealed no significant differences among these groups in percentage of Fos-labelled RTN neurons ( Fig. 7E; F 1,17 = 0.01331, P = 0.9095 for interaction, two-way ANOVA) or numbers of Fos-labelled astrocytes ( Fig. 7F; F 1,17 = 8.114, P = 0.0111, two-way ANOVA; all pairwise comparisons non-significant by Šídák's post hoc test).
Thus, data with inducible, conditional Slc4a4 knockout mice obtained with an Aldh1l1-Cre/ERT2 driver line indicate that even widespread depletion of astrocytic NBCe1 has no demonstrable effect on CO 2 -mediated activation of RTN neurons or nearby astrocytes, and also no effect on respiratory reflexes evoked by raised CO 2 or lowered O 2 . We nevertheless observed a higher overallV E in these mice across those experimental conditions.

Metabolic acidosis is associated with deletion of Slc4a4 in kidneys of tamoxifen-treated Aldh1l1-CreERT2 mice
Considering the elevatedV E observed in the Aldh1l1-Cre/ERT2 line of conditional NBCe1 knockout mice, we performed blood gas analysis from tail arteries of the control and Slc4a4-deleted littermates. These data suggested a partially compensated metabolic acidosis in the Cre + , tamoxifen-treated mice, by comparison to all the other control groups (Table 1). That is, we observed that arterial pH was markedly acidotic in the conditional NBCe1 knockouts, relative to all controls (pH = 7.21 ± 0.02, N = 10 vs. 7.41 ± 0.14, N = 32; P < 0.0001, unpaired t test), and this occurred in combination with reduced HCO 3 − (10.5 ± 0.7 mm, N = 10 vs. 19.9 ± 0.7 mm, N = 32; P < 0.0001, unpaired t test) and lower P CO 2 (25.7 ± 1.1 mmHg, N = 8 vs. 31.1 ± 0.7 mmHg, N = 30; P = 0.0016, unpaired t test). The respiratory compensation was observed in the elevatedV E post treatment for the NBCe1 knockouts (see Fig. 6B and D) and by comparingV E of Cre + :T mice to pooled data from the three control groups under normoxic conditions (1.2 ± 0.1 ml/min/g vs. 0.9 ± 0.1 ml/min/g, N = 6 and 17; P = 0.0164, unpaired t test).
NBCe1 is an important component of acid-base control by the kidney. So, given the marked metabolic acidosis in these conditional knockout mice, we examined the kidney to determine if there was an effect on NBCe1 expression. Indeed, as shown in Fig. 8A-D, we confirmed that Aldh1l1 Figure 6. Effects of CO 2 and O 2 on minute ventilation and sigh frequency are unaffected by NBCe1 deletion from astrocytes in Aldh1l1-Cre/ERT2;Slc4a4 fl/fl mice A, increase in minute ventilation (V E ) by 8% CO 2 in the indicated groups of Aldh1l1-Cre/ERT2;Slc4a4 fl/fl mice before and after treatment with vehicle or tamoxifen (ANOVA pre: F 3,27 = 0.5080, P = 0.6801; ANOVA post: F 3,27 = 2.069, P = 0.1279). Number of mice (N) for each group is indicated on each of the figure panels. B, effects of increasing inspired CO 2 (from 0% to 8%; 60% O 2 , balance N 2 ) onV E (mean ± SD) in each of the indicated groups of Aldh1l1-Cre/ERT2;Slc4a4 fl/fl mice (Inset; F 3,135 = 16.66, P < 0.0001, for genotype-treatment by two-way ANOVA; * * P < 0.005, * * * * P < 0.0001 for Cre − vs. Cre + :T; † P = 0.0121, for Cre + :V vs. Cre + :T) and comparing the conditional knockout mice (Cre + :T, mean ± SD) to pooled values from all control mice (shaded area represents the 95% confidence interval; F 1,145 = 41.60, P < 0.0001, for genotype-treatment by two-way ANOVA; * P = 0.0111, * * * * P < 0.0001 for controls vs. Cre + :T). C, increase inV E by hypoxia (10% O 2 ) in the indicated groups of Aldh1l1-Cre/ERT2;Slc4a4 fl/fl mice before and treatment with vehicle or tamoxifen (ANOVA pre: F 3,24 = 1.905, P = 0.1557; ANOVA post: F 3,21 = 1.767, P = 0.1843). D, effects of varied inspired O 2 (10% O 2 , 21% O 2 , 60% O 2 ; balance N 2 ) onV E comparing conditional knockout mice (Cre + :T, mean ± SD) to pooled values from all control mice (area represents the 95% confidence interval; F 1,68 = 21.17, P < 0.0001, for genotype-treatment by two-way ANOVA; * * P = 0.0025, * P = 0.491, for controls vs. Cre + :T). E, effects of hypoxia on sigh frequency (mean ± SD) in Cre-negative and Cre-positive Aldh1l1-Cre/ERT2;Slc4a4 fl/fl mice pre-and post-treatment with vehicle or tamoxifen. Hypoxia evoked sighs in all groups before and after treatment (pre: F 1,48 = 183.6, P < 0.0001; post: F 1,42 = 82.37, P < 0.0001, for hypoxia, by two-way ANOVA), but there was no difference among any of the groups (pre: is expressed in kidney (Krupenko, 2009) and, accordingly, we found that Slc4a4 expression was eliminated from kidney in a Cre-and tamoxifen-dependent manner. Thus, we expect that deletion of NBCe1 from the kidney in Cre + , tamoxifen-treated Slc4a4 fl/fl mice interferes with proximal tubular reabsorption of HCO 3 − and results in the observed metabolic acidosis (Lee et al., 2022). Note that Gfap is not expressed in kidney, Slc4a4 expression was preserved in kidneys of GFAP-Cre conditional knockout mice ( Fig. 8E and F), and those animals did not show any signs of metabolic acidosis (Table 2).

Discussion
Chemosensitive astrocytes that respond to CO 2 /H + have been implicated in control of breathing by brainstem RTN neurons, with various hypotheses advanced for mechanisms that couple astrocytic responses to enhanced RTN neuronal activity Guyenet et al., 2019). A contribution from astrocytic NBCe1 has been deemed critical in a number of these proposed mechanisms (Erlichman et al., 2008;Guyenet et al., 2019;Marina et al., 2018;Turovsky et al., 2016). Our results from two different lines of conditional knockout mice, both with astrocytic deletion of Slc4a4, indicate that NBCe1 is not required for CO 2 -mediated activation of either RTN neurons or ventral medullary astrocytes -and is also dispensable for CO 2 -stimulated breathing. These data indicate that if astrocytes contribute to respiratory chemosensitivity, the mechanism must be able to function independently of contributions from NBCe1.

Limitations and caveats
It is difficult to be completely assured of either the cell selectivity or the degree of gene knockout in studies of conditional knockout mice. In this study, we used GFAP-Cre and Aldh1l1-Cre/ERT2 mouse lines to eliminate the NBCe1 gene, Slc4a4, in astrocytes. The  2 . 9 ± 1.1 3.5 ± 1.4 * 3.9 ± 1.6 * * 2.0 ± 0.9 3.5 ± 1.4 * * Data are mean ± SD. Individual control groups were compared to the Cre-positive, tamoxifen-treated conditional knockouts (Cre + :T) by ANOVA with Dunnett's post hoc test; pooled control data was compared to Cre + :T conditional knockouts by unpaired t tests. * P < 0.05 * * P < 0.01 * * * P < 0.0005 * * * * P < 0.0001. Data are mean ± SD. Cre − and Cre + littermates were compared by unpaired t tests.
possible concerns with Cre expression by other brain cells in these mice are allayed by the nearly exclusive astrocytic expression of Slc4a4 in the brain (Theparambil et al., 2020). On the other hand, it is clear that astrocytes in different brain regions can be phenotypically distinct (Lee et al., 2022;SheikhBahaei et al., 2018;Theparambil et al., 2020;Walz & Lang, 1998) and we indeed found differential expression of Gfap and Aldh1l1 in the RTN (and other regions of the brainstem). Accordingly, the Cre-mediated recombination driven by the cognate GFAP and Aldh1l1 promoters yielded distinct patterns of Slc4a4 deletion in the two different mouse lines, with incomplete penetrance and chimeric residual expression. Nonetheless, we found that respiratory reflexes induced by CO 2 and hypoxia were unaffected in either line. In this respect, it is worth noting that that a recent study that used yet another inducible, astrocyte-selective NBCe1 knockout mouse line, based on GLAST-Cre/ERT2, also reported no deficits in CO 2 -stimulated breathing (see supplemental data in Hosford et al., 2022). As with our tamoxifen-inducible Aldh1l1-Cre/ERT2 line, Slc4a4 was also deleted in adult GLAST-Cre/ERT2 mice (Hosford et al., 2022), such that both lines mitigated potential issues associated with developmental compensation. Thus, across multiple lines of mice with variable NBCe1 knockout across different astrocytic populations, the common observation was that the respiratory chemoreflexes were unaffected. Nonetheless, it remains a formal possibility that complete elimination of NBCe1 across all astrocytic populations might reveal some alteration in CO 2 -or hypoxia-mediated breathing.

NBCe1 function in RTN astrocytes is dispensable for CO 2 -stimulated breathing
The evidence to suggest an important role for NBCe1 in proton sensing by astrocytes for CO 2 -mediated, astrocytic-dependent regulation of breathing has been reviewed previously (Gourine & Dale, 2022;Guyenet et al., 2019;Mulkey & Wenker, 2011). In short, NBCe1 activity is increased in ventral medullary astrocytes during CO 2 -induced acidification, probably reflecting a combination of intracellular acidification and membrane depolarization (Mulkey & Wenker, 2011;Turovsky et al., 2016;Wenker et al., 2010). Upon CO 2 entry into the cell, intracellular acidification promotes elevated NBCe1 activity to take up additional HCO 3 − , buffering those pH changes (Theparambil et al., 2014;Turovsky et al., 2016); in brainstem astrocytes, CO 2 -induced changes in intracellular Ca 2+ were strongly reduced by NBCe1 block with S0859 or in NBCe1 knockout mice (Turovsky et al., 2016). In addition, CO 2 can induce a membrane depolarization of astrocytes by inhibition of inwardly rectifying K + current with pharmacological properties suggestive of the Kir4.1-Kir5.1 heteromeric channels (Mulkey & Wenker, 2011;Wenker et al., 2010); in turn, astrocytic depolarization can increase HCO 3 − influx by virtue of the electrogenic nature of the NBCe1 transporter, and this so-called depolarization-induced alkalization J Physiol 601.16 is a well-described feature of brain acid-base control by astrocytes (Chesler, 2003;Grichtchenko & Chesler, 1994). Our experiments with conditional knockout mice reinforce the conclusion that Slc4a4 expression is indeed responsible for S0859-sensitive HCO 3 − -induced currents in astrocytes.
Despite the demonstrated depletion of NBCe1 in astrocytes in the conditional knockout mice examined here, RTN neuronal activation (i.e. Fos expression) and ventilatory stimulation by CO 2 were unaffected. As mentioned (see Introduction), multiple non-exclusive mechanisms by which NBCe1 might contribute to astrocytic chemosensation and respiratory stimulation by RTN neurons have been proposed Guyenet et al., 2019). In one conception, increased astrocytic NBCe1 activity accentuates extracellular acidification (Erlichman et al., 2008), and this serves to activate proton-sensing mechanisms on RTN neurons Guyenet et al., 2019). In another, the elevated NBCe1 activity evoked by CO 2 in astrocytes initiates additional ion exchange mechanisms to raise intracellular Ca 2+ and allow release of ATP (Turovsky et al., 2016); enhanced RTN neuronal activity follows either directly by actions of ATP on those neurons , or indirectly by ATP-mediated decreases in local blood flow to further raise extracellular CO 2 /H + (Hawkins et al., 2017). The present work suggests that the proposed NBCe1-dependent astrocytic mechanisms are dispensable for central respiratory chemosensitivity.
It is important to note that CO 2 -evoked Fos expression was retained in NBCe1-depleted astrocytes, and thus our observations do not preclude a separate, NBCe1-independent mechanism for CO 2 activation of astrocytes and for their proposed contribution to respiratory chemosensitivity. For example, Kir-dependent membrane depolarization may promote some alternative mechanism for intracellular Ca 2+ elevation to support vesicular release of ATP or other gliotransmitters (Mulkey & Wenker, 2011;Wenker et al., 2010). Alternatively, non-vesicular release mechanisms may also be relevant, such as CO 2 -dependent carbamylation and activation of ATP-permeating connexons, as suggested previously Huckstepp, id Bihi et al., 2010). Additional mechanistic work is warranted to elucidate the molecular/cellular basis for proposed effects of astrocytes on the central respiratory chemoreflex.

NBCe1 function in astrocytes is not required for hypoxic ventilatory response or hypoxia-induced sighs
Hypoxia may also directly activate astrocytes (Angelova et al., 2015;Gourine & Funk, 2017). Indeed, hypoxic inhibition of mitochondrial respiration, via elevated levels of reactive oxygen species, can drive calcium transients and calcium-dependent release of ATP (Angelova et al., 2015;Gourine & Funk, 2017); activated astrocytes can also release PGE 2 (Howarth et al., 2017), which enhances sighs (Koch et al., 2015;Viemari et al., 2013). However, unlike for CO 2 -induced activation of astrocytes (Turovsky et al., 2016), a role for NBCe1 in direct activation of astrocytes by hypoxia has not been proposed. In this respect, the direct hypoxia-induced activation of astrocytes and associated purinergic signalling has been suggested to contribute to a central hypoxia reflex that stimulates breathing (Angelova et al., 2015;Gourine & Funk, 2017). We find no effect of astrocytic NBCe1 deletion on hypoxia-stimulated breathing or sighs, suggesting that NBCe1 does not play a role in hypoxia-activation of astrocytes for breathing control; however, this could also reflect a predominance of peripheral chemoreceptor actions for hypoxia-stimulated breathing under our experimental conditions.

Aldh1l1-Cre/ERT2-mediated NBCe1 deletion in kidney leads to partially compensated metabolic acidosis
We found that mice with Aldh1l1-Cre/ERT2-mediated NBCe1 deletion displayed a partially compensated metabolic acidosis, with reduced P CO 2 and elevated overall ventilation. This appears to reflect loss of NBCe1 and reduced bicarbonate reabsorption from the kidney because we verified that Aldh1l1 is expressed in kidney (Krupenko, 2009) and showed that Slc4a4 expression was strongly reduced in Cre + , tamoxifen-treated Slc4a4 fl/fl mice. Despite this chronic metabolic acidosis, which might itself have blunted the hypercapnic ventilatory response, we found no difference in CO 2 -stimulated breathing in those mice. In addition, there was no concomitant loss of Slc4a4 expression in the kidneys of GFAP-Cre mice, which presented neither metabolic acidosis nor changes in baseline or CO 2 -stimulated ventilation. Thus, it seems unlikely that effects on kidney account for the lack of effect on CO 2 -stimulated breathing following NBCe1 knockout in these two different conditional knockout lines.
In conclusion, Slc4a4 expression and NBCe1 function in astrocytes do not contribute appreciably to CO 2 -stimulated (or hypoxia-evoked) breathing reflexes in mice. The mechanisms by which astrocytes contribute to the central respiratory chemoreflex remain to be elucidated, and our work indicates that they must be independent of previously proposed hypotheses that invoked local acidification or paracrine purinergic signalling dependent on astrocytic NBCe1 expression and function (Marina et al., 2018;Turovsky et al., 2016). J Physiol 601.16 E.C.G., R.L.S., C.B.B., G.Y., S.M.B. and D.A.B. analysed the data; and all authors edited and approved the final version of the manuscript and agree to be accountable for all aspects of the work. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed.

Funding
This study was funded by grants from the National Institutes of Health (HL074011, to P.G.G.; HL148004, to S.B.G.A.; HL061974 to G.E.S.; HL108609, to D.A.B.) and the Centre for Clinical and Translational Science and Training, and a Research Innovation Seed Grant from the University of Cincinnati (to G.E.S.). E.C.G. was supported by T32 GM007055 and F31 HL154660.