Somatostatin-expressing parafacial neurons are CO2/H+ sensitive and regulate baseline breathing

  1. Colin M Cleary
  2. Brenda M Milla
  3. Fu-Shan Kuo
  4. Shaun James
  5. William F Flynn
  6. Paul Robson
  7. Daniel K Mulkey  Is a corresponding author
  1. Department of Physiology and Neurobiology, University of Connecticut, United States
  2. The Jackson Laboratory for Genomic Medicine, United States
  3. Institute for Systems Genomics, University of Connecticut, United States

Abstract

Glutamatergic neurons in the retrotrapezoid nucleus (RTN) function as respiratory chemoreceptors by regulating breathing in response to tissue CO2/H+. The RTN and greater parafacial region may also function as a chemosensing network composed of CO2/H+-sensitive excitatory and inhibitory synaptic interactions. In the context of disease, we showed that loss of inhibitory neural activity in a mouse model of Dravet syndrome disinhibited RTN chemoreceptors and destabilized breathing (Kuo et al., 2019). Despite this, contributions of parafacial inhibitory neurons to control of breathing are unknown, and synaptic properties of RTN neurons have not been characterized. Here, we show the parafacial region contains a limited diversity of inhibitory neurons including somatostatin (Sst)-, parvalbumin (Pvalb)-, and cholecystokinin (Cck)-expressing neurons. Of these, Sst-expressing interneurons appear uniquely inhibited by CO2/H+. We also show RTN chemoreceptors receive inhibitory input that is withdrawn in a CO2/H+-dependent manner, and chemogenetic suppression of Sst+ parafacial neurons, but not Pvalb+ or Cck+ neurons, increases baseline breathing. These results suggest Sst-expressing parafacial neurons contribute to RTN chemoreception and respiratory activity.

Introduction

The retrotrapezoid nucleus (RTN) is an important respiratory control center located in the ventral parafacial region of the medulla oblongata (Guyenet and Bayliss, 2015; Guyenet et al., 2019). Of particular interest are RTN neurons that function as central respiratory chemoreceptors by regulating depth and frequency of breathing in response to changes in tissue CO2/H+ (Guyenet et al., 2019; Nattie and Li, 2012). Chemosensitive RTN neurons express the transcription factor Phox2b (Stornetta et al., 2006), are glutamatergic (Mulkey et al., 2004; Weston et al., 2004), and are intrinsically responsive to CO2/H+ by mechanisms involving acid inhibition of TWIK-related Acid-Sensitive K+ channel 2 (TASK-2; Wang et al., 2013a) and activation of G-protein-coupled receptors (GPR4; Kumar et al., 2015). Activity of RTN chemoreceptors is also subject to modulation by purinergic signaling (Mulkey et al., 2006; Wenker et al., 2012) most likely from local astrocytes (Gourine et al., 2010; Huckstepp et al., 2010) as well as excitatory (Lazarenko et al., 2011; Mulkey et al., 2007) and inhibitory (Guyenet et al., 2005; Moreira et al., 2007; Takakura et al., 2007) synaptic input. In addition, previous work showed using a Phox2b-eGFP reporter line that RTN chemoreceptors have numerous presumptive inhibitory (symmetric) and glutamatergic synapses (asymmetric) (Lazarenko et al., 2009). These results are supported by functional evidence showing in urethane-anesthetized rats that unilateral RTN injection of glutamate receptor blockers (AP5 or CNQX) decreased baseline breathing and the ventilatory response to CO2 (Nattie and Li, 1995), whereas unilateral RTN injection of the GABAA receptor blocker bicuculline stimulated baseline breathing and blunted CO2-stimulated respiratory activity (Nattie et al., 2001). This suggests glutamatergic and GABAergic synaptic drive to the RTN from distal and possibly local sources are required for maintenance of baseline breathing and CO2/H+ chemoreception. Furthermore, in the context of disease, suppression of parafacial inhibitory neural activity in a mouse model of Dravet syndrome caused a disordered breathing phenotype, and at the cellular level increased baseline activity and CO2/H+ dependent firing of RTN chemoreceptors (Kuo et al., 2019). These results suggest loss of inhibitory neural activity disinhibited RTN chemoreceptors and contributed to breathing problems. Despite this potential significance, contributions of parafacial inhibitory neurons to control of breathing are undetermined and synaptic properties of RTN neurons have not been characterized.

Based on in vivo multi-electrode array recordings in cats that showed the RTN contains CO2/H+-activated and -inhibited neurons that communicate through CO2/H+-dependent excitatory and inhibitory interactions (Ott et al., 2011), we hypothesize that ventral parafacial inhibitory neurons in the region of the RTN sense changes in CO2/H+ and regulate baseline breathing by a mechanism involving disinhibition. To address this, we first performed single cell RNA sequencing (scRNA-seq) to further characterize molecular signatures of chemosensitive RTN neurons and identify types of inhibitory neurons present in the parafacial region. We confirm the RTN is composed of two subsets of glutamatergic Phox2b- and Nmb-expressing neurons with similar levels of proton sensing machinery (Gpr4 and Kcnk5) but differ in galanin (Gal) expression. We also determined that parafacial inhibitory neurons are predominantly GABAergic and glycinergic and are composed of Sst, Pvalb, Ndnf, and Cck subtypes, including an Sst subset that is strongly inhibited by CO2/H+. At the network level, we show that the RTN functions as a CO2/H+-sensing network where excitatory and inhibitory neurons interact in a CO2/H+-dependent manner to augment respiratory drive. Also, at the whole animal level, we show that ventral parafacial inhibitory neurons, and more specifically Sst+ inhibitory neurons, regulate respiratory activity under baseline conditions but do not contribute to respiratory output under high CO2 conditions. The CO2/H+ response profile of Sst+ parafacial neurons, together with their preferential contribution to baseline breathing, suggest these cells are important determinants of resting respiratory drive.

Results

Molecular profiles of glutamatergic and GABA/glycinergic parafacial neurons

Single cells from the ventral parafacial region were isolated from 10-day-old wild type C57BL/6J (N = 16; 8 of each sex). Note that four male and four female mice included in this cohort received 4OH-tamoxifen (0.2 mg/daily for 3 days). However, since tamoxifen treatment did not affect the proportion of cells obtained or relative transcript profiles across all major cell types (Figure 1—figure supplement 1), these data sets were pooled. Single-cell RNA-seq was performed using the 10X Genomics Chromium Controller (Zheng et al., 2017) and 10X v2 chemistry targeting 16,000 barcoded cells. After quality control filtering and doublet removal (see Materials and methods), we analyzed 11,810 cells with a median of 2892 unique transcripts (UMIs) and 1472 genes per cell. We used the 2000 most highly variable genes, measured by dispersion, as input for dimensionality reduction using PCA, BBKNN, and UMAP, and found 20 distinct clusters using Leiden community detection (Figure 1—figure supplement 1).

Non-neuronal cells comprised roughly 90% of the dataset, so we employed a two-state mixture model parameterized on four general neuronal markers (Snap25, Tubb3, Elavl2, Syp) (Mickelsen et al., 2019) to segregate putative neurons from non-neuronal cells. A second mixture model was then used to classify the neurons into three groups: vesicular glutamate-transport type 2 (Vglut2+; Slc17a6) glutamatergic excitatory neurons (N = 197), vesicular GABA transporter (Vgat+; Slc32a1) inhibitory neurons (N = 445), and cholinergic (Chat) neurons (Figure 1A, Figure 1—figure supplement 1). The resulting neuronal populations had markedly higher molecular content, with a median of 4587 UMIs and 2317 genes per Slc17a6+ neuron and 4439 UMIs and 2299 genes per Slc32a1+ neuron.

Figure 1 with 3 supplements see all
Molecular signatures of ventral parafacial glutamatergic and inhibitory neurons.

(A) Through a normalized dispersion analysis for dimension reduction, a t-distributed stochastic neighbor embedding (t-SNE) was created of ventral parafacial single-cell transcriptome, with cells color coded by cluster. Neurons were differentiated from non-neurons (gray) based on expression of Snap25, Syp, Tubb3, and Elavl2. Cells expressing either Slc17a6 or Slc32a1 were used for sub-cluster analysis of glutamatergic and inhibitory neurons. A fairly large population of Chat expressing neurons was also detected but since neither glutamatergic (B) nor Vgat+ neurons express Chat, this population was not analyzed further. (B) UMAP plot depicting four sub-clusters of glutamatergic neurons and corresponding violin plots showing cluster-specific differential gene expression. Cluster number is noted on the x axis and gene expression (from 0 to 4 counts/cell) on the y axis. Slc17a6 clusters 1–2 are presumed to be subsets of RTN chemoreceptors based on expression of Phox2b, Nmb, Gpr4, and Kcnk5 that differ in expression of galanin. Slc17a6 cluster 3 differentially expressed Tac1 (gene encoding substance P), suggesting that these cells may be parapyramidal raphe neurons, whereas cluster four differentially expressed tyrosine hydroxylase (Th) indicative of adrenergic C1 pre-sympathetic neurons. (C) UMAP plot showing seven sub-clusters of inhibitory (Slc32a1) neurons and corresponding violin plots showing cluster-specific differential gene expression. Cluster number is noted on the x axis and gene expression (from 0 to 4 counts/cell) on the y axis. Discrete subtypes of inhibitory neurons were identified based on non-overlapping expression of cholecystokinin (Cck; Slc32a1 cluster 1), neuron-derived neurotrophic factor (Ndnf, Slc32a1 cluster 2) and parvalbumin (Pvalb; Slc32a1 cluster 5). Three somatostatin (Sst+) clusters could be differentiated based on expression of calretinin (Calb2; Slc32a1 cluster 3), reelin (Reln, Slc32a1 cluster 6) and neuronal nitric oxide synthase 1 (Nos1, Slc32a1 clusters 3, 7).

Two-dimensional embeddings and cluster assignments were generated for both the Slc17a6+ excitatory and Slc32a1+ inhibitory populations (Figure 1B–C) using a similar process described above (see Materials and methods). The population of Slc17a6+ cells does not overlap with Chat+ cells and includes two clusters (Slc17a6 clusters 1–2) of Phox2b-positive and Nmb expressing neurons with similar levels of Gpr4 (G-protein-coupled receptor 4) and Kcnk5 (TASK-2; K2P5) but differ in galanin (Gal) expression (Figure 1B). The molecular profiles of these clusters are largely consistent with that of RTN chemoreceptors (Shi et al., 2017). Sympathetic C1 catecholamine neurons (Slc17a6 cluster 4) are identified by expression of tyrosine hydroxylase (Th) and Phox2b and the absence of Nmb (Li et al., 2008). The remaining glutamatergic cluster (Slc17a6 cluster 3) was distinguished by expression of tachykinin 1 (Tac1, precursor for substance P), suggesting these cells are parapyramidal raphe neurons. All four glutamatergic clusters had detectable levels of stathmin 4 (Stmn4, encodes a microtubual binding protein [Holmfeldt et al., 2003]), suggesting these populations are not yet fully differentiated.

The population of parafacial inhibitory Slc32a1+ neurons was composed of 7 discrete clusters that could be distinguished by largely non-overlapping expression of somatostatin (Sst, Slc32a1 clusters 3, 6–7), parvalbumin (Pvalb, Slc32a1 cluster 5), cholecystokinin (Cck, Slc32a1 cluster 1) and the more recently identified interneuron marker (Abs et al., 2018) neuron-derived neurotrophic factor (Ndnf, Slc32a1 cluster 2) (Figure 1C). Of these, Sst+ interneurons were the most abundant cell type, but varied in expression of calretinin (Calb2, Slc32a1 cluster 3), reelin (Reln, Slc32a1 cluster 6) and neuronal nitric oxide synthase 1 (Nos1, Slc32a1 clusters 3, 7). Consistent with other brainstem regions, we found that all Slc32a1+ parafacial neurons also express lysosome-associated membrane protein 5 (Lamp5) (Koebis et al., 2019) (not shown) and most express the sodium- and chloride-dependent glycine transporter 2 (GlyT2; Slc6a5) (Hirrlinger et al., 2019; Figure 1C), suggesting they have the capacity to release GABA and glycine. However, in contrast to the cortex (Lee et al., 2010; Lim et al., 2018), the ventral parafacial region was devoid of interneurons that express the serotonin receptor 5HT3aR (Htr3a), vasoactive intestinal peptide (Vip), or neuropeptide Y (Npy) (data not shown). In summary, these results provide the first molecular characterization of inhibitory neurons in a chemoreceptor region, and in doing so establish a cellular framework for understanding roles of inhibitory neurons in respiratory chemoreception.

We confirmed that TdT labeled Slc32a1 cells are distributed throughout the ventral parafacial region including juxtaposed to Phox2b-immunoreactive RTN chemoreceptors (Figure 1—figure supplement 2). In-line with our scRNA-seq results, we found by fluorescent in situ hybridization that ~85% of Slc32a1+ cells (n = 57 cells) co-express glutamic acid decarboxylase (GAD67; Gad1) and GlyT2 (Slc6a5) (Figure 1—figure supplement 2). Also, consistent with our previous work showing that expression of a Dravet syndrome-associated Scn1a mutation disrupted RTN chemoreception and respiratory activity (Kuo et al., 2016), we found that Scn1a is expressed by all parafacial neurons, particularly inhibitory neurons that showed relatively higher levels of expression compared to glutamatergic neurons (Figure 1—figure supplement 3). Together, these results provide insight into the diversity of parafacial inhibitory neurons and identify unique molecular markers that may facilitate future assessment of cell type specific functions.

Sst+ parafacial neurons are CO2/H+ sensitive

The function of inhibitory neurons in this region of the brainstem has not been characterized. However, they are located within a region that is specialized to sense changes in CO2/H+, therefore, we wondered whether parafacial inhibitory neurons also contribute to RTN chemoreception. These experiments were performed in slices isolated from a Cre recombinase-dependent reporter line (Slc32a1Cre::TdT) which allows for selective targeting of inhibitory neurons in the region of interest (Figure 2—figure supplement 1). CO2/H+-sensitivity was characterized in cell-attached voltage-clamp mode; neurons that responded reversibly to 10% CO2 (pH 7.0) with ≥20% change in activity were considered CO2/H+-sensitive. By this criterion, we found that exposure to 10% CO2 decreased activity in 48 of 130 (37%) fluorescent parafacial neurons by an average of 2.0 ± 0.2 Hz (Figure 2Ai,D). The majority of parafacial inhibitory neurons (56%) did not respond to stimulus and so were considered CO2/H+-insensitive (Figure 2Bi,D), while a small subset of fluorescent parafacial neurons (7%) showed an excitatory response to CO2 (F2,130 = 91.23, p<0.001) (Figure 2C–D). We also found the inhibitory effect of CO2/H+ on chemosensitive RTN neurons was retained when purinergic signaling was blocked by bath application of pyridoxalphosphate-6-azophenyl-2',4'-disulfonic acid (PPADS; 100 µM) and 8-phenyltheophylline (8-PT; 10 µM) (T7 = 0.1515, p>0.05), or in the presence of ionotropic receptor blockers including CNQX (10 μM) to block AMPA/kainite receptors, gabazine (10 μM) to block GABAA receptors, and strychnine (2 μM) to block glycine receptors (T7 = 3.100, p=0.02) (Figure 2—figure supplement 2). These results suggest parafacial inhibitory neurons are intrinsically CO2/H+-sensitive.

Figure 2 with 2 supplements see all
Somatostatin neurons in the ventral parafacial region are CO2/H+-sensitive.

(A-C) Traces of firing rate and segments of holding current from ventral parafacial inhibitory neurons in slices from Slc32a1Cre::TdT mice show that exposure to 10% CO2 suppressed activity by 2.0 ± 0.2 Hz in 43% of neurons tested (Ai), whereas the majority of Slc32a1+ neurons in this region did not respond (Δ firing 0.11 ± 0.04 Hz) to this same level of CO2/H+ (Bi), and a small minority are activated by CO2/H+ (C). (Aii) Example of a CO2/H+-inhibited Lucifer Yellow filled parafacial neuron that was Sst-immunoreactive (IR). (Bii) Example of a CO2/H+-insensitive Lucifer Yellow filled parafacial neuron that was not Sst-IR. In sum, 5 of 5 CO2/H+-inhibited parafacial neurons were Sst-IR, whereas 0 of 5 CO2/H+-insensitive neurons in this region were Sst-IR. (D) Summary data (N = 130 neurons from 45 mice) plotted as number of cells vs mean firing response to 10% CO2. Note that all neurons included in this analysis showed a similar baseline level of activity in 5% CO2 (data not shown).

Figure 2—source data 1

Parafacial inhibitory neuron responses to CO2/H+.

https://cdn.elifesciences.org/articles/60317/elife-60317-fig2-data1-v2.xlsx

The next goal was to identify the molecular identity of CO2/H+-inhibited parafacial neurons. In other brain regions, subtypes of inhibitory neurons exhibit characteristic electrical properties and firing behavior (Ascoli et al., 2008). Although it is not clear whether such properties are discriminating for brainstem inhibitory neurons; nevertheless, in whole-cell current-clamp mode, we found CO2/H+-inhibited (662 MΩ) and -insensitive (703 MΩ) cells showed similar input resistances (p=0.91), baseline activity (p=0.59) and firing responses to depolarizing current injections (data not shown). These results suggest CO2/H+ sensitivity does not correlate with these electrical properties. Therefore, after gaining whole-cell access, we labeled cell types of interest with Lucifer Yellow (included in the pipette internal solution) for post hoc immunohistochemical identification using markers based on results our single cell RNAseq analysis (Figure 1C). We found that 5 of 5 CO2/H+-inhibited cells were Sst-immunoreactive (Figure 2Aii) and were not immunoreactive for Pvalb or Cck (data not shown), whereas 0 of 5 CO2/H+-insensitive cells expressed Sst (Figure 2Bii). These results suggest CO2/H+-inhibited parafacial neurons are one or more types (clusters 3, 6–7; Figure 1C) of Sst-expressing inhibitory neurons.

CO2/H+-synaptic properties of RTN chemoreceptors

Chemosensitive RTN neurons were identified in slices from Slc32a1Cre::TdT mice based on their lack of fluorescence and characteristic firing response to CO2/H+. As previously defined (Kuo et al., 2016), RTN neurons were considered chemosensitive if they show some level of spontaneous activity under control conditions and a robust firing rate response to 10% CO2 (Δ1.6 ± 0.36 Hz; N = 7 cells). Neurons that showed <1.0 Hz firing response to 10% CO2 were considered non-chemosensitive and excluded from this study. Once the cell type of interest has been identified, we obtained whole-cell access and in voltage-clamp, recorded spontaneous synaptic currents. sIPSCs were recorded in relative isolation by holding cells at the reversal potential for AMPA-mediated EPSCs (sEPSCs; Ihold = 0 mV). Under control conditions (5% CO2), chemosensitive RTN neurons showed sIPSCs with an average frequency of 0.22 Hz and amplitude of 13.7 pA (Figure 3A–C). The kinetics of sIPSCs recorded in RTN neurons are similar to what has been described in other brain regions (Ali et al., 2007; Banks and Pearce, 2000); average rise (10%–90%) and decay (90%–10%) times of 3.1 ms and 25.5 ms, respectively. Exposure to 10% CO2 decreases sIPSC frequency to 0.1 ± 0.06 Hz (F2,6=21.04; p<0.001; N = 7 cells total) (Figure 3A–B) which corresponded with an increase in the inter-event interval (Figure 3D) and decreased sIPSC amplitude (F2,6=6.58; p<0.05) (Figure 3C) but with no change in sIPSC rise (3.7 msl p=0.516) or decay (27.7 ms; p=0.268) times. The amplitude and frequency of sIPSCs returned to near control levels after washing back to 5% CO2 (Figure 3B–C). Subsequent bath application of bicuculline (10 µM) to block GABAA receptors and strychnine (2 µM) to block glycine receptors eliminated all sIPSCs, thus confirming they are mediated by GABA or glycinergic input (Figure 3A). Although we cannot exclude the possibility that CO2/H+ disrupts transmitter release from synaptic terminals in the RTN from distal inhibitory neurons, nevertheless, these results are consistent with the possibility that a subset of parafacial Slc32a1 neurons are inhibited by CO2/H+ and contribute to RTN chemoreception by a mechanism involving disinhibition.

CO2/H+-dependent suppression of inhibitory synaptic input to RTN chemoreceptors.

(A) Traces of holding current (Ihold = 0 mV) from an RTN chemoreceptor in a slice from a Slc32a1Cre::TdT mouse shows sIPSC events under control conditions and during exposure to 10% CO2 or bicuculline (10 µM) and strychnine (2 µM). (B–C) Summary data (N = 7) show the average effect of CO2/H+ on sIPSC freq (B) and amplitude (C). (D–E) Effect of CO2/H+ on sIPSC frequency and amplitude are also reflected in cumulative distribution plots of sIPSC inter-event interval (D; bin size 250 ms) and amplitude (E; bin size 5 pA). Data was analyzed by one-way RM ANOVA followed by Tukey multiple comparison test. *p<0.05, **p<0.01 ***p<0.001.

To further test this possibility, we characterized baseline activity and CO2/H+-sensitivity of RTN chemoreceptors under control conditions and when GABA and glycine transmission was blocked with bicuculine and strychnine. In cell-attached voltage-clamp mode, bath application of bicuculine (10 µM) and strychnine (2 µM) increased baseline activity of RTN chemoreceptors by 0.7 ± 0.2 Hz (T12 = 2.201, p=0.022) (data not shown). This finding suggests inhibitory input partly limits activity of RTN chemoreceptors under baseline conditions. We also found the firing response to 10% CO2 was similar under control conditions and in the presence of bicuculine and strychnine (Δ −0.3 ± 0.2 Hz; T12 = 1.246, p>0.05) (data not shown). These results suggest CO2/H+-dependent inhibitory transmission regulates activity of RTN chemoreceptors under baseline conditions but not during exposure to high CO2.

Since application of glutamate receptor blockers into the RTN blunted the ventilatory response to CO2 (47), we next characterized CO2/H+-dependent modulation of sEPSC’s in chemosensitive RTN neurons. Under control conditions (5% CO2) and at a holding potential of −60 mV chemosensitive RTN neurons exhibit sEPSCs with an average frequency of 0.33 Hz and amplitude of −12.6 pA (Figure 4A–C). The kinetics of RTN sEPSCs are similar to AMPA-mediated events described in other brain regions (Magee and Cook, 2000; Rodriguez-Molina et al., 2007; Selyanko et al., 1979); average rise (10%–90%) and decay (90%–10%) times of 1.2 ms and 4.9 ms, respectively. This was also confirmed pharmacologically by bath application of CNQX (10 µM) at the end of each experiment (Figure 4A). Exposure to 10% CO2 increased sEPSC frequency to 0.65 ± 0.08 Hz (F2,7 = 26.91; p<0.001) (Figure 4A–B) which corresponded with a decrease in the inter-event interval (Figure 4D) but with no change in sEPSC amplitude (F2,7 = 1.398; p>0.05) (Figure 4C) or kinetics (rise time 1.3 ms, p=0.5772; decay time 5.0 ms, p=0.9642). Furthermore, when both excitatory and inhibitory spontaneous synaptic events were characterized in the same cells (N = 6), we confirmed that CO2/H+ increases the sEPSC/sIPSC ratio from 1.5 to 8.9 (Figure 4F) (T5 = 3.70; p<0.05). The differential effects of CO2/H+ on excitatory and inhibitory synaptic currents also argues against potential non-specific effects of H+ on neurotransmission (Sinning and Hübner, 2013). Together, these results support the possibility that the RTN functions as a CO2/H+-sensing network where excitatory and inhibitory neurons interact in a CO2/H+-dependent manner to augment respiratory drive.

CO2/H+-dependent facilitation of excitatory glutamatergic input to RTN chemoreceptors.

(A) Traces of holding current (Ihold = −60 mV) from a chemosensitive RTN neuron in a slice from a Slc32a1Cre::TdT mouse shows sEPSC events under control conditions and during exposure to 10% CO2 or CNQX (10 µM). (B-C) summary data (N = 8) show the average effect of CO2/H+ on sEPSC freq (B) and amplitude (C). Note that two outlier data points were excluded from analysis. (D–E) Effect (or lack thereof) of CO2/H+ on sEPSC frequency and amplitude are also shown in cumulative distribution plots of sIPSC inter-event interval (D; bin size 100 ms) and amplitude (E; bin size 5 pA). (F) CO2/H+-induced suppression of sIPSC frequency in conjunction with increased sEPSC frequency resulted in enhancement of the sEPSC/sIPSC ratio. *p<0.05, **p<0.01 ***p<0.001.

Sst+ parafacial neurons regulate baseline breathing

Evidence suggests Sst-expressing parafacial neurons are inhibited by CO2/H+ (Figure 2) and contribute to basal activity of RTN chemoreceptors. To test this possibility in vivo, we used an AAV delivery system to express an inhibitory (Gi-coupled) DREADD receptor in a Cre-recombinase-dependent manner in Sst+ neurons. Specifically, we injected AAV2-hSyn-DIO-hM4D(Gi)-mCherry (10 nL/side, Addgene) bilaterally into the medial parafacial region of SstCre mice (JAX #: 013044) (Figure 5). The virus spread laterally to also include a lateral portion of the parafacial region associated with active expiration (Figure 5—figure supplement 1). After 2 weeks recovery, we characterized baseline breathing and the CO2 ventilatory response following sequential injections (I.P.) of saline followed ~2.5 hr later by clozapine (1 mg/kg; I.P.). These experiments began under room air conditions followed by exposure to 0, 3, 5, and 7% CO2 (balance O2 to limit peripheral chemoreceptor input). Animals were exposed to each condition for 10 min and each trial was limited to a total duration of 50 min to minimize the impact of clozapine clearance on receptor activation (Jendryka et al., 2019). Consistent with our cellular (Figure 2) and synaptic data (Figures 34), we found that chemogenetic suppression of Sst+ parafacial neurons increased respiratory frequency (F1,5=148.1, p<0.0001) (Figure 5A–B), tidal volume (F1,5 = 7.360, p=0.0421) (Figure 5C) and minute ventilation (F1,5 = 81.06, p=0.0003) under baseline conditions (Figure 5D). However, clozapine and saline treated mice showed similar ventilatory responses to CO2 (F1,5 = 0.9089, p>0.05) (Figure 5D). This is not surprising since CO2/H+-dependent suppression of inhibitory neural activity may preclude the effects of further chemogenetic inhibition on respiratory activity. Interestingly, parallel experiments performed in PvalbCre (JAX #: 008069) (F1,5 = 0.0139, p>0.05) and CckCre (JAX #: 012706) (F1,5 = 0.0660, p>0.05) minimally affected baseline breathing or the ventilatory response to CO2 (Figure 5E–F). It should be noted that cell type-specific expression of Cre recombinase in SstCre (Soumier and Sibille, 2014), PvalbCre (Liu et al., 2019) and CckCre (Matsuda et al., 2020) lines have been confirmed previously. Together with our cellular evidence, these results suggest Sst+ parafacial neurons are specialized to sense changes in CO2/H+ and contribute to respiratory activity.

Figure 5 with 2 supplements see all
Chemogenetic suppression of Sst+ but not Pvalb+ or Cck+ parafacial neurons increased baseline breathing.

(Ai) Computer-assisted plots show centers of bilateral AAV2-hSyn-DIO-hM4D-mCherry injections in SstCre animals. The number above each section indicates the relative position of each slice behind bregma (Paxinos and Franklin, 2013). (Aii) Traces of respiratory activity from SstCre mice that received bilateral parafacial injections of AAV2-hSyn-DIO-hM4D-mCherry following systemic (I.P.) injection of clozapine (1 mg/kg) or saline (control). Aiii, summary data (n = 6) shows effects of chemogenetic suppression Sst+ parafacial neurons with clozapine on minute ventilation under room air conditions and in 0–7% CO2 (balance O2). (B–C) Left side shows computer-assisted plots of AAV2-hSyn-DIO-hM4D-mCherry injection centers in PvalbCre (B) and CckCre (C) mice, and summary data to the right shows effects of chemogenetic suppression of parafacial Pvalb+ (B) and Cck+ (C) neurons on minute ventilation under room air conditions and in 0–7% CO2 (balance O2). N = 6 for each genotype (mixed sex). Slopes between 0% and 7% CO2 were compared by analysis of covariance (ANCOVA). #, different between genotypes (two-way ANOVA with Tukey’s multiple comparison test; one symbol p<0.05, two symbols p<0.01).

We performed similar experiments in Slc32a1Cre mice (JAX #: 016962) to suppress activity of all parafacial inhibitory neurons including clusters 2 and 4 from our scRNAseq dataset (Figure 1C), which were not targeted in experiments described above. We found that chemogenetic suppression of all parafacial inhibitory neurons mirrored the effects of targeted inhibition of just Sst+ neurons. For example, Slc32a1Cre mice responded to clozapine (1 mg/kg) with an increase in baseline respiratory frequency (F1,9=5.560, p=0.043) and tidal volume (F1,9 = 19.13, p=0.002), which culminated in an increase minute ventilation (F1,9=6.017, p=0.037) (Figure 5—figure supplement 1). Also, as observed in SstCre mice, we found that inhibition of all parafacial inhibitory neurons by clozapine administration (I.P.) in Slc32a1Cre mice minimally effected the CO2 chemoreflex (F1,9=0.0724, p>0.05) (Figure 5—figure supplement 1). We confirmed Cre recombinase is specific to inhibitory parafacial neurons in Slc32a1Cre mice (95% of TdT fluorescence co-localized with Gad67-immunoreactivity), which confirms previously reported data (Lowery-Gionta et al., 2018; Figure 5—figure supplement 1). It should also be noted that clozapine had minimal effect on respiratory activity in Slc32a1Cre mice that did not receive intracranial viral injections (F1,5=0.2399, p>0.05) (Figure 5—figure supplement 2). Together, these results identify Sst-expressing parafacial neurons as important determinants of baseline breathing.

Discussion

This study provides the first characterization of synaptic properties of RTN neurons and in doing so identifies a role of ventral parafacial inhibitory neurons in CO2/H+ dependent control of breathing. Specifically, we show that (1) the ventral parafacial region contains Sst, Ndnf, Pvalb, and Cck classes of interneurons of which only Sst-expressing neurons are inhibited by CO2/H+; (2) chemosensitive RTN neurons receive inhibitory input under control conditions that is withdrawn during exposure to high CO2; (3) chemogenetic inhibition of Sst+ but not Pvalb+ or Cck+ parafacial neurons increases baseline respiratory activity. These results suggest Sst-expressing parafacial neurons are important determinants of respiratory activity under baseline conditions. This is important because disruption of baseline breathing is the root cause of disordered breathing in various disease states.

Molecular profile of glutamatergic and GABA/glycinergic ventral parafacial neurons

Consistent with previous work (Dubreuil et al., 2009; Shi et al., 2017; Stornetta et al., 2006), we show that several types of ventral parafacial neurons express Phox2b including presumptive cholinergic neurons (Chat+), C1 pre-sympathetic neurons (Th+) and two subsets that express Nmb and similar levels of H+ sensing machinery (Gpr4 and Kcnk5) but differ in galanin expression. It is not clear whether both Nmb+ populations have similar or divergent roles in control of breathing. Based on Gpr4 and Kcnk5 expression both Nmb+ subtypes likely function as chemoreceptors, however, CO2/H+-activated Phox2b+ parafacial neurons also contribute to pre-inspiratory rhythmogenesis early in development (Onimaru et al., 2008) so perhaps differential expression of galanin denotes these functional differences. It is also not clear whether expiratory parafacial neurons express Phox2b and respond to CO2/H+ similar to RTN chemoreceptors or comprise a functionally discrete respiratory center. For example, selective activation (Souza et al., 2020) or inhibition (Marina et al., 2010) of ventral parafacial Phox2b-expressing neurons increased and decreased expiratory activity, respectively, suggesting expiratory parafacial neurons may be an extension of Phox2b-positive RTN chemoreceptors that differentially regulate inspiratory or expiratory activity depending on projection targets. However, others have shown that expiratory parafacial neurons are not Phox2b-immunoreactive (de Britto and Moraes, 2017), suggesting these neurons are distinct from RTN chemoreceptors. In any case, since expiratory parafacial neurons are putatively glutamatergic (Silva et al., 2016), this population is most likely included in clusters 1–3. Of these, cluster 3 is of particular interest because it lacks Gpr4 and Kcnk5 and so is not likely to function as an RTN chemoreceptor. It is also worth noting that Slc17a6 clusters 3–4 express high levels of Stmn4, an important regulator of neural differentiation (Lin and Lee, 2016), so perhaps these populations diverge at later developmental time points. In any case, we are unable to disentangle putative rhythmogenic or expiratory parafacial neurons from other Phox2b+ cell types at this time. Nonetheless, it is also important to recognize that Phox2b is expressed by multiple cell types in the parafacial region, thus confounding interpretation of previous work that relied primarily on Phox2b expression to target RTN chemoreceptors (Gourine et al., 2010).

We also determined that the ventral parafacial region contains a limited diversity of inhibitory neurons including one Cck+ cluster, one Pvalb+ cluster, one Ndnf+ cluster and three Sst+ clusters that differ in terms of calretinin, reelin and Nos1 expression. Although our cellular experiments suggest Sst+ cells are CO2/H+-sensitive (Figure 2), we were not able to identify which Sst+ cluster(s) function as chemoreceptors based on gene expression. We also found the ventral parafacial region did not include Htra3-, Vip-, or Npy-expressing inhibitory neurons. This is in marked contrast to the cortex where Htra3+ inhibitory neurons represent ~30% of the total interneuron population (Tremblay et al., 2016). By defining the diversity if inhibitory neurons in this brainstem region, these results provide a basis for understanding how disruption of inhibitory neural function in diseases like Dravet syndrome cause brainstem disfunction and mortality.

Sst-expressing parafacial neurons are intrinsically inhibited by CO2

The RTN is an important respiratory chemoreceptor region (Guyenet et al., 2019); therefore, it is reasonable to speculate that any neurons in this region that respond to CO2/H+ might do so in a manner tailored to support this reflex. Consistent with this, we found that a subset of inhibitory parafacial neurons (37%) are inhibited by 10% CO2. This response was retained when purinergic signaling or fast neurotransmission was blocked, suggesting these cells are intrinsically CO2/H+ sensitive. We also showed that CO2/H+-inhibited parafacial neurons express Sst while CO2/H+-insensitive cells did not, thus honing the list of inhibitory chemoreceptor candidates to clusters 3, 6, and 7 (Figure 1C). Since the majority of inhibitory neurons in this region did not respond to this level of CO2, it is tempting to speculate that Sst+ parafacial neurons are specialized to contribute to RTN chemoreceptors and respiratory drive.

The network basis of RTN chemoreception

Previous evidence based on multi-electrode extracellular recordings from cats suggests that RTN neurons including CO2/H+-activated (presumably chemoreceptors) and -inhibited (most likely Slc32a1+ based on evidence presented in this study) interact through paucisynaptic connections (Ott et al., 2011). Consistent with this, we show that exposure to high CO2/H+ increased the frequency of spontaneous glutamatergic input to RTN neurons while simultaneously decreasing frequency and amplitude of spontaneous inhibitory synaptic inputs.

Mechanisms contributing to CO2/H+-induced activation of the RTN network likely involves both pre- and post-synaptic mechanisms. At the presynaptic level, we consider other chemosensitive RTN neurons or medullary raphe neurons which may co-release glutamate (El Mestikawy et al., 2011) as most the likely sources of CO2/H+ dependent glutamatergic drive to RTN chemoreceptors. Also, since CO2/H+ inhibited the activity of some parafacial Slc32a1+ neurons, we consider these cells the most likely substrate responsible for CO2/H+-induced suppression of sIPSC frequency. It should be noted that few inhibitory parafacial neurons were activated by CO2/H+ (Figure 3C), suggesting RTN chemoreceptors do not project to and regulate activity of parafacial inhibitory neurons. The differential effects of CO2/H+ on EPSC and IPSC frequency also argue against potential non-specific inhibitory effects of CO2/H+ on voltage gated Ca2+ channels (Shah et al., 2001) and neurotransmission (Sinning and Hübner, 2013). However, the RTN (Guyenet et al., 2005; Takakura et al., 2007; Yang and Feldman, 2018) and parafacial region (Silva et al., 2020) receive inhibitory input from various elements of the respiratory circuit that serves to limit chemoreceptor function during times of high respiratory activity, and we cannot exclude potential effects of CO2/H+ on transmitter release from these inputs.

At the postsynaptic level, our finding that CO2/H+ minimally affected sEPSC amplitude is consistent with evidence that AMPA receptors are largely unaffected by acidification. Conversely, CO2/H+-dependent suppression of sIPSC amplitude may involve H+-dependent inhibition of GABA or glycine receptors on RTN neurons. For example, depending on the subunit composition, certain recombinant GABAA receptors (Huang et al., 2004; Huang and Dillon, 1999; Wang et al., 2005) and glycine α1 and α1β receptors (Chen et al., 2004) are reversibly inhibited by acidification near the physiological range. It should also be noted that CO2/H+ may elicit release of non-glutamatergic neuromodulators that alter postsynaptic conductance and potentially contribute to diminished IPSC amplitude; however, we consider this unlikely since this non-specific mechanism is expected to affect both excitatory and inhibitory synaptic currents. Together these results suggest the RTN functions as a CO2/H+-sensing network composed of both faciliatory and disinhibitory interactions.

Parafacial inhibitory neurons contribute to baseline breathing

The possibility that disinhibition contributes to chemoreception is not novel or unique to the RTN. For example, CO2/H+-inhibited cells have been found in several putative chemosensitive regions (Conrad et al., 2009; Nichols et al., 2009; Wang and Richerson, 1999) including GABAergic neurons in the medullary raphe (Iceman et al., 2014) and parafacial region (Kuribayashi et al., 2008). These results suggest disinhibition contributes to respiratory chemoreception. However, in the absence of evidence that inhibitory neurons in these regions actually influence chemoreceptor function or contribute to respiratory behavior, this possibility has remained largely speculative. We addressed these knowledge gaps by first showing that chemosensitive RTN neurons receive inhibitory synaptic input under control conditions that is withdrawn in a CO2/H+-dependent manner (Figure 4). We also show that chemogenetic suppression of Sst+ parafacial neurons increased minute ventilation (frequency and tidal volume) (Figure 5A). Therefore, we propose that Sst+ parafacial neurons contribute to the drive to breathe by regulating baseline activity of RTN chemoreceptors. When the activity of Sst+ parafacial neurons is diminished under high CO2 conditions, it is perhaps not surprising that further inhibition of this population by chemogenetic means had negligible effect on respiratory output. These results are consistent with pharmacological evidence from anesthetized rats showing that application of bicuculline (GABAA receptor blocker) to the ventrolateral medulla increased inspiratory activity under control conditions but not during exposure to high CO2 (Gourine and Spyer, 2001). Furthermore, disinhibition of neurons located more laterally in the ventral parafacial region contributed to emergence of active expiration during exposure to high CO2 (Huckstepp et al., 2015; Pagliardini et al., 2011). Recent work suggests GABAergic neurons in medullary raphe regions regulate expiratory output of the lateral parafacial region during high CO2 (60); however, the source of CO2/H+-dependent disinhibition of expiratory parafacial neurons remains unclear. Our evidence that Sst+ parafacial neurons show a CO2/H+ response consistent with disinhibition and their close proximity to the lateral parafacial region makes them ideal candidates for such a function. However, potential roles of these neurons in regulation of expiratory activity requires further investigation.

Potential physiological significance

The ability of respiratory chemoreceptors to sense and respond to tissue CO2/H+ is what maintains breathing during sleep. As CO2/H+ levels decrease so too does respiratory activity until the apneic threshold is reached and breathing ceases. Therefore, baseline CO2 levels and the apneic threshold are critical determinants of stable breathing during sleep (Nakayama et al., 2002). The ability of ventral parafacial Sst+ neurons to respond to CO2 in the low physiological range (minimally active under high CO2 conditions) and preferentially regulate baseline breathing suggests these cells are important determinants of baseline CO2 levels and the apneic threshold. Although we did not observe an apnea phenotype during chemogenetic suppression of parafacial inhibitory neurons, these experiments were performed in awake mice in which arousal-dependent mechanism may help stabilize breathing. Interestingly, a yet unidentified population of Slc32a1+ neurons in nearby and potentially overlapping medullary regions are known to regulate rapid eye movement (REM) sleep, where activation of these cells promotes a REM-like state and inhibition of these cells does the opposite (Weber et al., 2015). Considering that CO2/H+ also stimulates arousal (Guyenet and Bayliss, 2015), it is possible CO2/H+ inhibition of parafacial Slc32a1+ neurons suppresses REM and promotes arousal, thus directly coordinating chemoreceptor activity with sleep-wake state. However, these possibilities require further investigation. An additional caveat to note is that in addition to RTN chemoreceptors, Sst+ parafacial neurons likely regulate respiratory activity at other levels of the respiratory circuit. Therefore, an important future direction of this work will be to identify projection targets of Sst+ parafacial neurons.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Strain, strain background (M. musculus, Vgat-iris-Cre, mixed 129/SvJ and C57BL6/J background)Slc32a1tm2(cre)Lowl/JJackson LaboratoriesRRID:IMSR_JAX:016962
Strain, strain background (TdTomato reporter Ai14, C57BL6/J background)B6.Cg-Gt(ROSA)26Sortm9(CAG-tdTomato)Hze/JJackson LaboratoriesRRID:IMSR_JAX:007909
Strain, strain background (Phox2b Cre, C57Bl6/J background)B6(Cg)-Tg(Phox2b-cre)3Jke/JJackson LaboratoriesRRID:IMSR_JAX:016223
Strain, strain background (Sst Cre, C57Bl6/J background)Ssttm2.1(cre)Zjh/JJackson LaboratoriesRRID:IMSR_JAX:013044
Strain, strain background (Pvalb Cre, C57Bl6/J background)Pvalbtm1(cre)Arbr/JJackson LaboratoriesRRID:IMSR_JAX:017320
Strain, strain background (Cck-IRES Cre, C57Bl6/J background)Ccktm1.1(cre)Zjh/JJackson LaboratoriesRRID:IMSR_JAX:012706
Transfected construct (M. musculus)AAV2-hSyn-DIO-hM4D(Gi)-mCherryPMID:21364278RRID:Addgene_44362
Antibody(goat polyclonal) anti-mouse Phox2b antibodyR and D SystemsRRID:AB_10889846(1:100 dilution)
Antibody(rat monoclonal) anti-mouse GAD67 antibody,close 1G10.2Millipore SigmaRRID:AB_2278725(1:250 dilution)
Antibody(mouse monoclonal) anti-mouse Parvalbumin antibodyMillipore SigmaRRID:AB_2174013(1:250 dilution)
Antibody(rabbit polyclonal) anti-mouse CCK-AR antibodyR and D SystemsRRID:AB_2275486(1:250 dilution)
Antibody(mouse monoclonal) anti-Somatostatin antibodySanta Cruz BiotechnologyRRID:AB_831726(1:200 dilution)
Antibody(rabbit polyclonal) anti-lucifer yellow antibodyThermoFisherRRID:AB_2536190(1:500 dilution)
Antibody(donkey polyclonal) anti-rabbit AlexaFluor 488Jackson Immunoresearch711-545-152(1:500 dilution)
Antibody(donkey polyclonal) anti-mouse AlexaFluor 647Jackson Immunoresearch715-605-150(1:500 dilution)
Sequence-based reagentRNAScope Probe-Gad1ACDBio400951-C31:50
Sequence-based reagentRNAScope Probe-Slc6a5ACDBio425351-C21:50
Sequence-based reagentRNAScope Probe-Slc32a1ACDBio31919150:1
Commerical assay or kitRNAscope Fresh Frozen Multiplex Fluorescent KitACDBio320851
Commerical assay or kitChromium Single Cell 3’ Reagent Kit10X GenomicsPN 120237Version 2
Chemical Compound, drugLucifer YellowSigmaL02590.2%
Chemical compound, drugClozapineSigma11421071 mg/mL
Chemical compound, drugStrychnineSigmaS05322 µM
Chemical compound, drugGabazineTocris126210 µM
Chemical compound, drugCNQXTocris019010 µM
Chemical compound, drugPPADSTocris0625100 µM
Chemical compound, drug1,3-dimethyl-8-phenyl-xantine (8-PT)SigmaP227810 µM
Software, algorithmLoupe Browser10X GenomicsRRID:SCR_018555Version 5.0.0
Software, algorithmPonemahDSIRRID:SCR_01701Version 5.32
Software, algorithmSpikeCambridge Electronic DesignRRID:SCR_00903Version 5.0
Software, algorithmPrism 7GraphPadRRID:SCR_002798Version 7.03
Software, algorithmpCLAMP 10Molecular DevicesRRID:SCR_011323Version 10
Software, algorithmImageJNIHRRID:SCR_003070Version 2.0.0
Software, algorithmSynaptosoftMini Analysis ProgramRRID:SCR_002184

Animals

All procedures were performed in accordance with National Institutes of Health and University of Connecticut Animal Care and Use Guidelines. All animals were housed in a 12:12 light dark cycle with normal chow ad libitum if of weaning age. The single-cell RNA-seq experiments used mixed sex wild-type C57BL6/J animals. The Slc32a1Cre (JAX # 016962) and TdTomato (Ai14) reporter mice (JAX # 007914) were maintained on 129S1/SvlmJ and C57BL6/J backgrounds, respectively, and only F1 pups were used for electrophysiological experiments. Mouse lines used for chemogenetic experiments including SstCre (JAX # 013044), PvalbCre (JAX # 017320), and CckCre (JAX # 012706) lines were ordered from a congenic C57BL6/J background, directly from Jackson Laboratories. Phox2bCre (JAX # 016223) mice were maintained on a C57BL6/J background and used solely for antibody specificity confirmation.

Single-cell isolation

Request a detailed protocol

Animals were euthanized under ketamine/xylazine anesthesia and brainstem slices were prepared using a vibratome in ice cold, high-sucrose slicing solution containing (in mM): 87 NaCl, 75 sucrose, 25 glucose, 25 NaHCO3, 1.25 NaH2PO4, 2.5 KCl, 7.5 MgCl2, 0.5 mM CaCl2, and 5 L-ascorbic acid (equilibrated with 5% CO2-95% O2). Coronal brainstem slices (300 μm thick) were prepared and then immediately enzymatically treated at 34°C with protease XVIII (6 mg/mL, Sigma) for 6 min. After enzyme incubation, slices were washed three times in cold dissociation solution and then transferred to an enzyme inhibitor mix containing trypsin inhibitor (10 mg/mL, Sigma) and bovine serum albumin (BSA, 10 mg/mL, Sigma) in cold sucrose dissociation solution. Next, we isolated the parafacial region which included cells within ~100 µm from the ventral surface and extended ~600 µm medially from the border of the trigeminal nucleus and ~500 µm rostrally from the caudal end of the facial nucleus. Motor neurons, adrenergic C1 cells and raphe neurons were identified based on expression of cell type specific markers and excluded from analysis. These tissue chunks were then warmed to 34°C for 10 min before trituration. A single cell suspension was achieved by trituration using a 25 and 30 gauge needles sequentially, attached to a 2 mL syringe. Samples were triturated for an average of 5 min. Immediately after, the samples were placed back on ice and filtered through a 30-micron filter (Miltenyi Biotech) into sterile microcentrifuge tubes for cell viability assessment.

Single-cell RNA sequencing

Request a detailed protocol

Cell viability for each sample was assessed on a Countess II automated cell counter (ThermoFisher), and 12,000 cells were loaded for capture onto an individual lane of a Chromium Controller (10X Genomics). Single cell capture, barcoding and library preparation were performed using the 10X Chromium platform according to the manufacturer’s protocol (#CG00052) using version 2 (V2) chemistry. cDNA and libraries were checked for quality on Agilent 4200 Tapestation, quantified by KAPA qPCR. All libraries were sequenced on individual lanes of an Illumina HiSeq4000 targeting 6000 barcoded cells with an average sequencing depth of 50,000 reads per cell.

scRNA-seq data processing, quality control, and analysis

Request a detailed protocol

Illumina base call files for both libraries were converted to FASTQs using bcl2fastq v2.18.0.12 (Illumina) and FASTQ files were aligned to the mm10 (GRCh38.84, 10X Genomics mm10 reference 2.1.0) using the version 2.2.0 Cell Ranger count pipeline (10X Genomics), resulting in two gene-by-cell digital count matrices. Source code for FASTQ files is available on GitHub at: https://github.com/TheJacksonLaboratory/ventral-parafacial-neuron-scrnaseq or through GEO accession GSE153172. Downstream analysis was performed using Scanpy (v1.4.6) (Wolf et al., 2018). Individual libraries were subjected to quality control and filtering independently. Putative doublets were first removed using Scrublet on the raw matrixes (Wolock et al., 2019). Then, for each matrix, cells containing fewer than 800 genes, more than 50 hemoglobin transcripts, or more than 20% mtRNA content were excluded from downstream analyses. Genes present in five or fewer cells, with fewer than 10 total counts were also excluded. For the dataset depicted in this work, we used two lanes of 10X Chromium chip with the following conditions: P10 C57BL6/J pups and 4-OH Tamoxifen treated P10 C57BL6/J pups. We used both of these conditions to create a comprehensive control database to compare Cre dependent and inducible lines with a C57BL6/J background strain. The individual filtered matrices (containing 3345 and 9054 cells, respectively) were concatenated together resulting in an initial aggregated counts matrix of 12,399 cells by 15,923 genes. This aggregated counts matrix was normalized by the total number of counts per cell then multiplied by the median number of counts across all cells, log2 transformed, and lastly scaled to zero mean and unit variance column-wise.

The 2000 most highly variable genes computed using 'scanpy.pp.highly_variable_genes' with `flavor=”cell_ranger’` were selected for the computation of principal components (PCs). Genes related to cell cycle, stress response, the Y-chromosome, hemoglobin as well as ribosomal, mitochondrial, and the gene Xist were excluded from this list of highly variable genes prior to the computation of PCs. Briefly, the list of cell cycle genes was adapted from Giotti et al., 2019, using all annotated genes except those with labels ‘Other’, ‘Function known but’, and ‘Uncharacterized’, whereas the stress response genes were adapted from O'Flanagan et al., 2019. The first 25 PCs were computed and were used to create a k = 16 batch-balanced nearest-neighbor graph using BBKNN (Polański et al., 2020) measured by cosine distance. A 2D UMAP embedding (https://arxiv.org/pdf/1802.03426.pdf) was subsequently generated using this graph ('min_dist = 0.5'). Initial clusters were assigned via the Leiden community detection algorithm (Traag et al., 2019) at 0.7 resolution on this k-NN graph, resulting in 20 initial clusters (Figure 1—figure supplement 1).

To further analyze the neuronal populations, all cells were first classified as neuronal or non-neuronal using a simple two-state Gaussian mixture model (Mickelsen et al., 2019), and neuronal cells were further classified as inhibitory or excitatory using a three-state Gaussian mixture model. Briefly, the median expression of Snap25, Syp, Tubb3, and Elavl2 in each cluster was used to fit the first model and classify the clusters as neuronal (high expression) and non-neuronal (low expression). Median expression of Slc17a6 and Slc32a1 was then used to classify the neurons as Slc17a6high (excitatory), Slc32a1high (inhibitory), or other. Inhibitory and excitatory neurons were subsequently reanalyzed separately (Figure 1—figure supplement 1).

The raw expression matrices of the 210 excitatory and 464 inhibitory neurons were re-normalized, batch corrected, and embedded with UMAP independently as described above, using only 1000 highly variable genes. Clustering with Leiden community detection led to four excitatory and seven inhibitory neuron clusters. In both analyses, a small population of Slc17a6 high/Slc32a1 high cells emerged (n = 28); it is unclear whether these cells are doublets, but due to the limited number of such cells, this unique population was discarded from both analyses, resulting in 197 excitatory and 445 inhibitory neurons analyzed. Markers genes for each cluster in each analysis were computed using the Wilcoxon-ranked sum test in a one-versus-rest fashion using the `scanpy.tl.rank_genes_groups` function.

Immunohistochemistry

Request a detailed protocol

Mice (Phox2bCre::TdT, Slc32a1Cre::TdT, Slc32a1Cre infected with virus) were transcardially perfused with 20 mL of room temperature phosphate buffered saline (PBS, pH 7.4) followed by 20 mL of chilled 4% paraformaldehyde (pH 7.4) in 0.1 M phosphate buffered saline. The brainstem was then removed from the animal and post-hoc fixed for 24 hr. After, 150 µm slices were made using a Zeiss VT100S vibratome. In the case of live slices from electrophysiology recordings, individual slices were placed in chilled 4% paraformaldehyde for at least 24 hr before processing. Free floating slices were then incubated in a 0.5% Triton-X/PBS solution for 45 min to permeabilize the tissue. The slices remained in a 0.1% Triton-X/10% Fetal Bovine Serum (FBS, ThermoFisher)/PBS solution for a 12 hr primary antibody incubation of selected primary antibody (see Key Resource Table). The tissue was then washed three times in 0.1% Triton-X/10% FBS/PBS solution; the secondary antibody was incubated with the tissue after the third wash for 2 hr (see Key Resource Table). The tissue was then washed three times in PBS before mounting on precleaned glass slides with Prolong Diamond with DAPI (ThermoFisher). Imaging of brain slices was achieved with a Leica SP8 confocal microscope.

Fluorescent in situ hybridization

Request a detailed protocol

To prepare fresh frozen slice, two week old Slc32a1Cre::TdT mice were anesthetized with isoflurane, decapitated, and brainstem tissues were rapidly frozen with dry ice and embedded with OCT compound. Brainstem slices (14 µm thick) containing the RTN were cryosectioned and collected onto SuperFrost Plus microscope slides. Slices were fixed with 4% paraformaldehyde and dehydrated with ethanol. This tissue was processed with the instruction of RNAscope Multiplex Fluorescent Assay (ACD, 320850); the probes used in our study were designed and validated by ACD. Confocal images were obtained using a Leica Sp8 and confocal image files containing image stacks were loaded into ImageJ (version 2.0.0, NIH, RRID:SCR_003070).

Acute brainstem slice preparation and in vitro electrophysiology

Request a detailed protocol

Slices containing the RTN were prepared as previously described (Kuo et al., 2016). In short, mice were anesthetized by administration of ketamine (375 mg/kg, I.P.) and xylazine (25 mg/kg; I.P.) and rapidly decapitated; brainstems were removed and transverse brainstem slices (250–300 µm) were cut using a microslicer (DSK 1500E; Dosaka) in ice-cold substituted Ringer solution containing the following (in mM): 260 sucrose, 3 KCl, 5 MgCl2, 1 CaCl2, 1.25 NaH2PO4, 26 NaHCO3, 10 glucose, and 1 kynurenic acid. Slices were incubated for 30 min at 37°C and subsequently at room temperature in a normal Ringer’s solution containing (in mM): 130 NaCl, 3 KCl, 2 MgCl2, 2 CaCl2, 1.25 NaH2PO4, 26 NaHCO3, and 10 glucose. Both substituted and normal Ringer’s solutions were bubbled with 95% O2 and 5% CO2 (pH = 7.30).

Electrophysiological recordings

Request a detailed protocol

Individual slices containing the ventral parafacial region were transferred to a recording chamber mounted on a fixed-stage microscope (Olympus BX5.1WI) and perfused continuously (~2 ml/min) with a bath solution containing (in mM): 130 NaCl, 3 KCl, 2 MgCl2, 2 CaCl2, 1.25 NaH2PO4, 26 NaHCO3, and 10 glucose (equilibrated with 5% CO2; pH = 7.3). All recordings were made with an Axopatch 200B patch-clamp amplifier, digitized with a Digidata 1322A A/D converter and recorded using pCLAMP 10.0 software (RRID:SCR_011323). The firing response of chemosensitive RTN neurons to 10% CO2 (duration of ~5 min) was assessed at room temperature (~22°C) in the cell-attached voltage-clamp configuration (seal resistance >1 GΩ) with holding potential matched to resting membrane potential (Vhold = −60 mV) and with no current generated by the amplifier (Iamp = 0 pA). Patch electrodes had a resistance of 5–6 MΩ when coated with Sylgard 184 and filled with a pipette solution containing the following (in mM): 120 KCH3SO3, 4 NaCl, 1 MgCl2, 0.5 CaCl2, 10 HEPES, 10 EGTA, 3 Mg‐ATP and 0.3 GTP‐Tris, 0.2% Lucifer yellow (pH 7.30). Firing rate histograms were generated by integrating action potential discharge in 10 to 20 s bins using Spike 5.0 software (RRID:SCR_000903). In a subset of experiments, we obtained whole-cell access to characterize input resistance over voltages ranging from −20 to −80 mV by injecting current steps (1 s) of varying amplitudes (−100 pA to −20 pA). Also, in the whole cell configuration, we filled cell types of interest with Lucifer Yellow for post hoc immunohistochemical identification using cell type specific markers.

Spontaneous synaptic currents were characterized in the absence of TTX using a Cs-based pipette solution containing the following (in mM): 135 CsCH3SO3, 10 HEPES, 1 EGTA, 1 MgCl2, 3.2 TEA-Cl, 5 Na-phosphocreatine, 4 Mg-ATP, and 0.3 Na-GTP (pH 7.3 using CsOH). To record spontaneous IPSCs (sIPSCs), cells were held at the reversal potential for AMPA-mediated excitatory synaptic currents (sEPSCs; Ihold 0 mV) and confirmed with bath application of GABA and glycine blockers. To record EPSCs, cells were held at −60 mV. Although this voltage is positive to the Cl- reversal potential under our experimental conditions, potential contaminating IPSCs could easily be excluded from analysis based on the direction of synaptic currents; inward for EPSCs and outward for IPSCs. We also pharmacologically confirmed EPSCs as glutamatergic at the end of each experiment by bath application of 6-cyano-7- nitroquinoxaline-2,3-dione (CNQX). Spontaneous EPSCs and IPSCs were analyzed using the Mini Analysis Program (Synaptosoft) and detected events based on amplitude (minimum 5 pA) and characteristic kinetics (fast rising phase followed by a slow decay). Each automatically detected event was also visually inspected to exclude obvious false responses. All whole-cell recordings had an access resistance (Ra) <20 MOhm, recordings were discarded if Ra varied 10% during an experiment, and capacitance and Ra compensation (70%) were used to minimize voltage errors. A liquid junction potential of −10 mV (KCH3SO3) or +11 mV (CsCH3SO3) was corrected off-line.

RTN viral injections

Request a detailed protocol

Adult Slc32a1Cre, SstCre, PvalbCre, and CckCre mice (>20 g) were anesthetized with 3% isoflurane. The right cheek of the animal was shaved and an incision was made to expose the right marginal mandibular branch of the facial nerve. The animals were then placed in a stereotaxic frame and a bipolar stimulating electrode was placed directly adjacent to the nerve. Animals were maintained on 1.5% isoflurane for the remainder of the surgery. An incision was made to expose the skull and two 1.5 mm holes were drilled left and right of the posterior fontanelle, caudal of the lambdoidal suture. The facial nerve was stimulated using a bipolar stimulating electrode to evoke antidromic field potentials within the facial motor nucleus. In this way, the facial nucleus on the right side of the animal was mapped in the X, Y, and Z direction using a quartz recording electrode. The viral vector, AAV2-hSyn-DIO-hM4D(Gi)-mCherry (Addgene #44361, titer 5.3 × 1012 GC∕mL), was loaded into a 1.2 mm internal diameter borosilicate glass pipette on a Nanoject III system (Drummond Scientific). One 10 µL injection of virus per side was delivered at least −0.02 mm ventral to the Z coordinates of the facial nucleus, to ensure injection into the RTN. These same coordinates were used for the left side of the animal. In all mice, incisions were closed with nylon sutures and surgical cyanoacrylate adhesive. Mice were placed on a heated pad until consciousness was regained. Meloxicam (1.5 mg/kg) was administered 24 and 48 hr postoperatively. Plethysmography was performed 2 weeks after viral injection. The location of all injection sites were later confirmed by post hoc histological analysis (Figure 5—figure supplement 1).

Unrestrained whole-body plethysmography

Request a detailed protocol

Respiratory activity was measured using a whole-body plethysmograph system (Data Scientific International; DSI), utilizing a small animal chamber maintained at room temperature and ventilated with room air or carbogen mixtures at a constant flow rate of 1.16 L/min. Following recovery from surgery, Slc32a1Cre, SstCre, PvalbCre, and CckCre mice injected with AAV were individually placed into a chamber and allowed 1 hr to acclimate prior to the start of an experiment. Respiratory activity was recorded using Ponemah 5.32 software (DSI) for a period of 20 min in room air immediately after I.P. injection of saline or clozapine followed by exposures to 0, 3, 5, and 7% CO2 (balance O2) (10 min/condition). For the chemogenetic experiments, we sequentially tested saline followed by clozapine on the ventilatory response to CO2 in the same animals on the same day (~2.5 hr between trials). Plethysmography experiments began immediately after injection of either saline or clozapine. Parameters of interests including respiratory frequency (FR, breaths per minute), tidal volume (VT, measured in mL; normalized to body weight and corrected to account for chamber and animal temperature, humidity, and atmospheric pressure), and minute ventilation (VE, mL/min/g) were measured during a 20 s period of relative quiescence, confirmed with synchronous video monitoring, after 5 min of exposure to each condition. All experiments were performed between 9 a.m. and 6 p.m. to minimize potential circadian effects.

Statistics

Data are reported as mean ± SE. Power analysis was used to determine sample size, all data sets were tested for normality using Shapiro-Wilk test, and outlier data points were identified by the Grubbs test and excluded from analysis. Statistical comparisons were made using t-test, Wilcoxon-ranked sum test, or one-way or two-way simple or repeated measures ANOVA or ANCOVA followed by Tukey’s multiple comparison tests as appropriate. The specific test used for each comparison is reported in the figure legend and all relevant values used for statistical analysis are included in the results section.

Data availability

Raw and processed scRNA-seq data are available through the Gene Expression Omnibus (accession GSE153172) and analysis code is available on GitHub. Analysis of FISH, electrophysiology, and respiratory activity data was done using standard software and no custom code was written.

The following data sets were generated
    1. Cleary CM
    2. Kuo FS
    3. James S
    4. Flynn WF
    5. Robson P
    6. Mulkey DK
    (2020) NCBI Gene Expression Omnibus
    ID GSE153172. Ventral parafacial inhibitory neurons regulate baseline breathing by a mechanism involving disinhibition.

References

    1. Banks MI
    2. Pearce RA
    (2000)
    Kinetic differences between synaptic and extrasynaptic GABA(A) receptors in CA1 pyramidal cells
    The Journal of Neuroscience 20:937–948.
    1. Dubreuil V
    2. Barhanin J
    3. Goridis C
    4. Brunet JF
    (2009) Breathing with phox2b
    Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences 364:2477–2483.
    https://doi.org/10.1098/rstb.2009.0085

Decision letter

  1. Muriel Thoby-Brisson
    Reviewing Editor; CNRS Université de Bordeaux, France
  2. Ronald L Calabrese
    Senior Editor; Emory University, United States
  3. Clément Menuet
    Reviewer; Institut de Neurobiologie de la Méditerranée, France
  4. Natasha N Kumar
    Reviewer; University of New South Wales, Australia
  5. Patrice Guyenet
    Reviewer; University of Virginia, School of Medicine

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

A parafacial region of the medulla called the retrotrapezoid nucleus (RTN) is an important respiratory control center that responds vigorously to CO2 changes, contributes to arterial PCO2 homeostasis and regulates breathing during sleep. Within the RTN, glutamatergic neurons function as respiratory chemoreceptors by regulating breathing in response to changes in tissue CO2/H+. However, only very limited attention has been devoted so far to investigate the role of inhibitory neurons in RTN function. The present study by Mulkey's lab explores the possible contribution of local inhibitory neurons to the activity and acid-sensitivity of RTN neurons. The authors used multiple technologies (RNAscope, RNA seq, electrophysiology, chemogenetics) in vivo and in vitro to specifically target inhibitory interneurons located at the anatomical site of the RTN. They show that this region contains a limited diversity of inhibitory neurons. Among these, only Sst-expressing inhibitory interneurons appear inhibited by CO2/H+ and to send inhibitory inputs on the RTN chemosensitive neurons. By doing so these neurons finely tune resting respiratory drive according to CO2 changes in low physiological ranges. These new findings establish parafacial SST-expressing inhibitory neurons as important regulators of baseline breathing and contributor of RTN chemoception. Furthermore, this paper brings valuable mechanistic insights into how loss of inhibition within the RTN might disrupt breathing in pathological conditions such as in the Dravet syndrome.

Decision letter after peer review:

Thank you for submitting your article "Ventral parafacial inhibitory neurons regulate baseline breathing by a mechanism involving disinhibition" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Ronald Calabrese as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Clément Menuet (Reviewer #1); Natasha N Kumar (Reviewer #2); Patrice Guyenet (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

We would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). Specifically, when editors judge that a submitted work as a whole belongs in eLife but that some conclusions require a modest amount of additional new data, as they do with your paper, we are asking that the manuscript be revised to either limit claims to those supported by data in hand, or to explicitly state that the relevant conclusions require additional supporting data.

Our expectation is that the authors will eventually carry out the additional experiments and report on how they affect the relevant conclusions either in a preprint on bioRxiv or medRxiv, or if appropriate, as a Research Advance in eLife, either of which would be linked to the original paper.

Essential revisions:

More specifically, I would like to draw your attention to the fact that the three reviewers have carefully examined your manuscript and while they found a great interest for your results they also pointed out on several important aspects of the paper that require modifications. Their reviews are detailed below but I would like to emphasize on two major concerns that must be adequately addressed before suitability for publication in eLife could be considered:

First, they all noted a lack of precision in the description of the different methods used, in the size of the samples analyzed and illustrated in the figures, the parameters measured, etc… In addition, the experimental conditions in some cases are not always clear. More importantly, the authors must provide evidences for the specificity and selectivity of the transgenic lines used and the Dreadd and TdTomato expression and viral injections. In general, the methods need to be more precise and better explained throughout the text.

Second, the transcriptomic part, as it stands, does not really add much to the physiological question and remains a little apart from the rest of the study. For instance, it does not help in the identification of the neuronal cluster involved in the acid sensitivity of the RTN, and potential specific genes that could be indicative of a specific population of inhibitory neurons playing a role here are not used in the subsequent experiments presented after. Thus, either you put less emphasis in this part of the results (and even consider presenting these data at the end of the result section) or you provide a deeper analysis and a better link with the other presented data that would definitively bring a level of specificity and understanding that would greatly enhance the quality of the work and the interest of providing such genetic findings.

Note that acceptability will have to be reassessed upon submission of the revised version.

Reviewer #1:

The proposed study is a Research Advance paper aimed at investigating the role of inhibitory parafacial neurons in the control of breathing. The authors first performed single cell RNA sequencing of the RTN/parafacial region to identify clusters of excitatory and inhibitory neurons. Then they showed using whole-cell recordings in slices that about 25% of recorded inhibitory parafacial neurons are tonically active and inhibited by hypercapnia, and that chemosensitive RTN neurons receive tonic inhibitory inputs that are decreased by hypercapnia, and excitatory inputs that are increased by hypercapnia. Chemogenetic inhibition of inhibitory parafacial neurons increased normocapnic/normoxic and normocapnic/hyperoxic breathing but not hypercapnic breathing in mice in vivo.

To my knowledge, this is the first study investigating the specific role of inhibitory parafacial neurons in the control of breathing activity. The main finding, that inhibitory parafacial neurons enable fine tuning of resting respiratory drive according to variations of CO2 in the low physiological range, is very interesting to the field, and potentially to a broader audience. The study is well-designed, using state-of-the-art techniques, and the paper is well written.

1. The Results section starts with single cell RNA sequencing data of RTN/parafacial neurons, identifying clusters of glutamatergic and GABA/glycinergic neurons. In itself, this is very valuable data, which could justify a stand-alone paper with further analysis. However, starting the Results section with these data is a bit awkward, as they are poorly linked to the rest of the study. Indeed, the authors provide new subclasses of inhibitory neurons, but then focus the rest of the study on overall inhibitory neurons without attempting to link their functional results with the subclasses of neurons molecularly identified. The transition line 164, to show that TdT is expressed in overall inhibitory neurons, without using tools for further specificity as one would expect after reading the previous paragraph on scRNAseq, is somewhat frustrating.

The paragraph starting line 172 shows that a subset of parafacial inhibitory neurons are CO2/H+ sensitive, and finishes with a statement (lines 188-190) that the identity of this subset of neurons remains to be determined. Testing whether this subset corresponds to a specific cluster of inhibitory neurons identified in the scRNAseq data presented just before would seem the obvious next step for the present study. The authors could have done immunohistochemistry on the neurons patched, as several of the clusters identified have markers for which antibodies exist, or better (but harder) they could have aspirated the patched cytoplasm and performed qPCR to reach the same results. Such data would not only provide a natural link between the first two experimental sets performed, but also provide more detailed information on the inhibitory parafacial neurons that are CO2/H+ sensitive that would significantly enhance the impact of this study.

With these comments in mind, if no further experiment is done for this paper, it might be better to move the scRNAseq data to the end of the Results section, to open the work towards more specificity in future studies. On that line, the end of the sentence lines 168-170, which closes the scRNAseq data presentation, is more suited for the end of the Results section, rather than the beginning.

2. In a previous article by the same group (Kuo FS et al., eLife 2019, DOI: https://doi.org/10.7554/eLife.43387), mice expressing a Dravet syndrome-associated Scn1a missense mutation conditionally in inhibitory neurons presented breathing alterations including diminished chemosensitivity, with hypo-excitable inhibitory neurons and hyper-excitable excitatory neurons in the RTN. The present study is proposed as Research Advance of this previous article. To strengthen the continuity with the previous study, the authors could have analysed the expression of Snc1a in their scRNAseq data, to identify whether Scn1a is expressed in a cluster of inhibitory neurons, and link it with the cluster that is responsive to CO2/H+. This way, the link would be more evident, and the present study would really make use of the scRNAseq data. In the current state of this study, I somewhat fail to see its strong and direct continuity with the previous work justifying a Research Advance format, it would maybe be better suited as a traditional stand-alone Research Article.

3. Was synaptic isolation performed when testing the CO2/H+ sensitivity of parafacial inhibitory neurons? Did you also block the usual gliotransmitters? The methods section is not clear, please provide more details, as whether these neurons are intrinsically sensitive to CO2/H+ or not is critical for interpretation of these data.

Reviewer #2:

General assessment: Previous work from the Dan Mulkey group demonstrated that mice expressing a Scn1a mutation exhibit a Dravet syndrome associated phenotype, wherein inhibitory neurons of the RTN were less excitable, whereas glutamatergic chemosensitive neurons were hyperexcitable. In this manuscript, Cleary and colleagues used multiple technologies (RNAscope, RNA seq, electrophysiology, chemogenetics) to target inhibitory interneurons located at the anatomical site of the retrotrapezoid nucleus, which harbours glutamatergic respiratory chemosensory neurons. It is now established that inputs from other chemoreceptor neurons as well as astrocytes-derived paracrine signals, contributes to RTN chemoreceptor drive. Here they show that RTN neurons receive tonic inhibitory inputs from non-chemosensing VGAT neurons in the parafacial region. The authors provide careful control conditions. The manuscript figures are presented well experimental procedures and results are sound and support the conclusions. This work warrants publication in eLife in principle, the results are novel both technically and from a biological perspective, upon the condition that the authors adequately address the reviewers questions.

1. In the first paragraph of the discussion, the authors conclude that inhibitory inputs to chemosensing RTN neurons are withdrawn/inhibited during exposure to high CO2. The net result is a lessening of GABA restraint on chemosensing neurons and an increase in their neural drive and neurotransmitter release. Another valid interpretation could be that increased neurotransmission from other tonic inputs to the RTN (arising from chemoreceptors activated during high CO2 exposures) contributes to the ventilatory response? Ie the VGAT inhibitory input isn't withdrawn but rather contributes much less to RTN drive during exposure to high CO2. Please include this interpretation, and other alternatives in the discussion.

2. DREADD-mediated inhibition of inhibitory RTN neurons increased baseline minute ventilation but had no effect on ventilatory responses to CO2. Figure 6D-E – most columns do not include data from n=11 mice. Please explain why certain mice/replicates were excluded from the dataset. Figure 6 figure legend is incomplete – there is no description of the data shown for control Vgatcre (Figure 6F-H). Also, there are two Figure 6F.

3. CO2/H+ synaptic properties of RTN neurons studies (page 8, line 193):

"As previously defined (26), RTN neurons were considered chemosensitive if they show some level of spontaneous activity under control conditions and a robust firing rate response to 10% CO2 (1.6 {plus minus} 0.36 Hz; N=10)." – 10% CO2 FR response of 1.6 {plus minus} 0.36 Hz does not seem 'robust' when compared to other previous studies. For example, Wang S et al. 2013 Figure 2D (citation #64), demonstrates FR at pH 7 (~8% CO2) to be >4 Hz for both TypeI and TypeII dissociated RTN neurons. Wouldn't you expect FR to be >>4Hz in 10% CO2, for RTN neurons in an intact brain slice?

4. Figure 2 and Figure 2 legend:

Figure 2B – error bars appear to be missing for this n=4 quantified data.

Figure 2Aii is an enlarged version of the image at bregma -6.05mm. You could omit this, I don't think it adds anything to the paper. Alternatively, magnify further, use arrows to draw attention to cells of interest.

Figure 2 legend: include abbreviations (7N, b)

5. Figure 3C legend: It is not clear what the summary data corresponds to. Indicate this in the figure legend as follows: ' firing responses of CO2/H+ -inhibited (n=20, blue), -insensitive (n=53, grey) and -activated (n=7, red). What parameters do the authors use, to define inhibited versus activated neurons?

6. Figure 5 – please confirm that there are no outliers in these values. One replicate seems to have a much higher baseline sEPSC frequency than the others.

7. Supplementary Figure 1 figure legend states that the colors of distributions correspond to classes shown in Figure 1A however Figure 1A displays only 3 colours, whereas Supp Figure 1D displays 5 colours/classes. Please clarify.

8. Authors used RNA seq to identify differentially expressed and coregulated genes in RTN neurons in order to infer biological meaning for further studies on this preeminent central chemoreceptor population. Single cell isolation method and RNA-seq methodology was extremely detailed, valuable information regarding inclusion and exclusion criteria were included. The authors use tSNE to investigate segmentation, clustering of genes in the pFRG transcriptome. The data are convincing, since it is well established using other methods, that ~50% of glutamatergic RTN neurons express galanin. Could the authors clarify how they minimised batch effect, which can occur during the experiment, the RNA library preparation, or the sequencing run?

There are 20 global clusters (from 2000 genes) – the distribution of cells/cluster and genes/cluster is illustrated in Supp Figure 1C. The basis for the clustering was not clear – could the authors clarify this and make it clear in the text. It appears (from Supp Figure 1 B) that each cluster is defined by related genes that display a high level of expression within the cluster.

Reviewer #3:

RTN is a small cluster of lower brainstem neurons with a well-described molecular signature. These neurons respond vigorously to CO2 in vivo, contribute to arterial PCO2 homeostasis and maintain breathing automaticity during sleep. Their synaptic inputs are not well identified however and the way in which they encode brain PCO2 in vivo is complex and much debated. Several mechanisms have been proposed (intrinsic pH sensitivity of RTN neurons, astrocyte-mediated paracrine effect of acid, paradoxical acid-mediated contraction of microvessels and synaptic inputs from acid-activated neurons, e.g. serotonergic). The present study explores the possible contribution of local inhibitory neurons to the activity and acid-sensitivity of RTN neurons. Consistent with prior EM evidence, the results show that RTN neurons receive GABA/glycinergic inputs. The contribution of a subset of this input to the CO2 response of RTN neurons is suggested by the results of experiments in slices but the existence of this mechanism is not convincingly supported by the in vivo experiments.

The authors confirm the transcriptome of RTN neurons and describe that of several subsets of inhibitory neurons (likely GABA- and glycine-ergic) that reside in their vicinity. This work greatly extends existing observations regarding these local inhibitory neurons. Unfortunately this work did not help identify the particular inhibitory interneurons postulated to be CO2-responsive.

Next, the authors show that a subset of these inhibitory neurons are mildly inhibited by acidification in slices. The effects are small and variable from cell to cell but suggest that CO2-triggered disinhibition could perhaps contribute to the overall excitatory effect of CO2 on RTN neurons.

This hypothesis is then tested, by determining whether chemogenetic inhibition of these inhibitory interneurons enhances breathing and the respiratory chemoreflex in conscious mice. The authors show is that inhibition of the GABA/glycine interneurons (via a Gi-coupled DREAAD) in unanesthetized mice elevates breathing slightly at rest but seems to have no effect on the breathing stimulation elicited by CO2. The increase in resting breathing is consistent with prior evidence that breathing is increased when a GABA-receptor antagonist is microinjected into the region containing the RTN neurons (Nattie et al) but this fact does not demonstrate that the increase in breathing is caused by activation of RTN neurons. As the authors acknowledge in the discussion the increase in breathing produced by inhibiting GABA interneurons in this brain region could also result from arousal. The increase in breathing could also result from the disinhibition of neurons other than RTN, for example the C1 neurons.

Finally, the main interpretative difficulty is that inhibition of the GABA/glycine interneurons does not change the respiratory chemoreflex; this is difficult to reconcile with the results obtained in slices and makes one wonder whether the acid inhibition of the GABA interneurons observed in slices is an artifact or at least whether it is a phenomenon that has any relevance to the CO2 response of RTN neurons in vivo.

1. Re: results line 177. Although the authors are probably correct, they need to validate the assumption that their reporter line (Vgatcre::TdT) does indeed "allow for selective targeting of inhibitory neurons in the region of interest". If this has been demonstrated before please give the specific quote. If not, the authors must demonstrate that the vast majority of the tomato-red positive neurons in their region of interest express VGat (transcript or protein) at the relevant postnatal age.

2. Re: lines 240-241. "Specifically, we bilaterally injected AAV2-hSyn-DIO-hM4D(Gi)-mCherry (10 nL/side, Addgene) into the medial portion of the RTN in VGAT-cre mice". The authors need to also demonstrate here that VGat+-neurons were selectively transduced (i.e. that mCherry and VGat were colocalized). The selectivity of this type of approach is typically overrated (ectopic expression of Cre and Cre-independent effects of the DIO viral prep) and must be verified.

3. Re: results lines 177-183. The defining criteria for pH sensitivity is fine but the results are less than impressive: 20% of neurons were inhibited to some degree (-20% or more) when exposed to a fairly severe acidification (pH 7.0), 66% were unaffected and 9 % were judged to be excited. Figure 3C should be replaced with a distribution histogram summarizing the % response of all the VGAt neurons sampled (e.g. 10% effect binning). Small positive or negative effects of acidification on unit-activity in slices have frequently been reported in other brainstem areas such as the nucleus of the solitary tract. Cherry-picking neurons that are excited and those that are inhibited is not terribly convincing unless this grouping correlates with an identifiable biochemical or structural marker, which was not the case here.

4. Re: lines 248-249 and discussion. The increase breathing could have many causes and may not have nothing to do with CO2 chemoreception. For example, selective activation of the rostral C1 neurons produces a very strong increase in breathing in unanesthetized mice and these neurons are tonically inhibited by GABA-ergic input (the basis of the baroreflex). The increased breathing noted at baseline (no CO2 added) could also denote arousal from sleep, an effect produced by activation of the C1 neurons, the RTN itself and perhaps other neurons in the area.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Somatostatin-expressing parafacial neurons regulate baseline breathing" for further consideration by eLife. Your revised article has been reviewed by 3 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Ronald Calabrese as the Senior Editor.

This is an unusual decision letter. The situation is as follows. Two of the original reviewers declined to review the manuscript. We decided to solicit extra reviewers. The new review team judge that you have successfully addressed all of the concerns from the last round of review. We are therefore happy to publish the manuscript in its current form.

However, the two new reviewers have also raised some new points that they think will improve the manuscript. You therefore have the option to address these new comments if you think it would improve the manuscript. I emphasise this is entirely your decision. In accordance with eLife policy, because you have addressed the previous concerns, we are happy to publish the manuscript as it stands.

We apologize for the delay in providing this review. But two out of the three initial reviewers declined to re-review the paper and we had to find two other ones. Consequently, you will find some new concerns. However, be aware that your responses to previous comments have been judged adequate and that the additional experiments now included definitively improve the paper. We are happy to publish the paper in this form.

However, please take the time to read carefully all comments and address those that you choose to address. Basically, they highlight the following potential improvements: A better definition of your hypothesis; clarifying the functional connections between the different sets of data; stating unambiguously whether you consider the neurons of interest as part of the RTN/pFRG or not; providing a better description of the inhibitory neurons investigated in terms of distribution and number; and paying attention to the scales in different figures that might be wrong.

Reviewer #2:

The article is much improved and I think the authors have responded adequately to my comments and the comments of the other reviewer.

A comprehensive body of work investigating the neurochemical diversity of parafacial inhibitory neurons. In particular the large majority of scl32a1 neurons (GABA of Glycinergic) 37% are chemosensing and in fact exhibit reduced FR in response to extracellular acidosis, 56%were pH insensitive, 7% showed an excitatory response. The acid-inhibited neurons are intrinsically pH sensitive. 5/5 CO2 inhibited cells are SST immunoreactive.

Chemogenetic suppression of Sst+ parafacial neurons increased respiratory frequency and minute ventilation (F1,5=81.06, p = 0.0003) under baseline conditions

Having seen the original manuscript, the addition of the 2 sets of experiments linking the transcriptomic analysis with the electrophysiological in vitro respiratory experiments, and toning down the emphasis on the transcriptomic analysis, make the narrative of the paper clearer.

I agree with Reviewer 1, the transcriptomic analysis comes across as an add on to the data. There is some information that is not relevant to main narrative of this paper however, this additional information is likely of great interest to a broader audience of neuroscientists.

Whilst important to point out the additional experiments required to definitively demonstrate that the inhibitory RNAseq cluster corresponds with the H+ inhibited inhibitory neurons in the RTN, Reviewer 1 acknowledges the difficult nature of these experiments. The authors could include this as future directions in the Discussion section.

The Dan Mulkey group adequately addressed both reviewers suggestion to strengthen continuity with the Kuo et al. 2019 study, in order to justify this manuscript as a Research Advance of the previous study. The Kuo et al. paper is pivotal in that it links impairment of inhibitory parafacial neurons with a clinical syndrome.

The response to Reviewer 1 Major Concern 2 is good in that it reveals Scn1a expression is not restricted to inhibitory neurons.

The response to Major Concern 3 is sound; the knowledge that these inhibitory parafacial neurons are intrinsically chemosensitive (similar to phox2b+ parafacial neurons) is of great value to other researchers in the field.

I am satisfied with the authors responses to Reviewer 2 (my comments).

I am satisfied with the authors responses to Reviewers3

I'd like to commend the authors on their production of this impressive body of work

Reviewer #4:

It is a well-performed and very interesting study showing that the ventral medullary region of mice, which includes the RTN, contains a diversity of inhibitory neurons (SSt, CCK, and Pvalb). Some of the SSt-expressing neurons were inhibited by hypercapnia/acidosis and chemogenetic inhibition of SSt neurons, including the SSt neurons insensitive and activated by hypercapnia/acidosis, increases the baseline breathing. Besides, the RTN neurons receive inhibitory synaptic transmission at low frequency, which was reduced in response to hypercapnia/acidosis. Therefore, the authors conclude that the ventral medullary SSt neurons contribute to RTN chemoreception and respiratory activity. The topic of this manuscript is original, relevant, and worth studying. Several omissions cloud the interpretation, and I have several concerns that preclude the straightforward interpretation of the data, which have not been adequately considered in the manuscript. The main issue is in the interpretation, or overinterpretation, of the data.

1. It would be essential to define the hypothesis of the present study clearly. The authors "considered" the possibility that ventral medullary neurons in the region of the RTN sense changes in CO2/H+ and regulate RTN chemoreception by a mechanism involving disinhibition. However, the authors provided evidence for the ventral medullary inhibitory neurons' involvement in neither the RTN chemoreception nor RTN disinhibition. The authors then propose a novel mode of chemotransduction involving regulation of basal breathing by CO2/H+-dependent disinhibition. In fact, the experiments involving the chemogenetic inhibition of SSt neurons did not demonstrate that the RTN neuronal function was affected. The neurons in the ventral medullary region close to RTN project to the pre-Bötzinger Complex (PMID: 26855425), and these projections could be involved in the increases of baseline breathing following SSt neurons chemogenetic inhibition.

2. There are no clear functional connections between some of the results described. The authors described the presence of a diversity of inhibitory neurons in the ventral medullary region and that the inhibition of SSt neurons increased the baseline breathing without affecting the ventilatory responses to hypercapnia/acidosis. On the other hand, there are data showing that RTN neurons receive synaptic inhibition, which was reduced in response to hypercapnia/acidosis. Therefore, the authors need to demonstrate that: i) the three different populations (activated, insensitive, and inhibited) of SSt neurons are projecting to RTN chemosensitive neurons, and: ii) the chemogenetic inhibition of SSt neurons increases the RTN neurons firing frequency by reducing the inhibitory synaptic transmission. The SSt neurons recorded in vitro were labeled, and the projections to the RTN neurons can be revealed in the same slice. Besides, the effects of SSt neurons inhibition (chemogenetics) on the RTN firing frequency and inhibitory synaptic transmission can also be analyzed in in vitro experiments using the transfected animals. It is important to note that without new experiments to prove the role of the inhibitory ventral medullary SSt neurons in the RTN chemosensitivity, the authors should consider presenting the data in two (I – somatostatin-expressing RTN neurons regulate baseline breathing; II – the effects of hypercapnia/acidosis on the synaptic transmission to RTN neurons) different manuscripts.

3. The authors demonstrated that the inhibitory neurons are located close to the RTN chemoreceptors. In fact, it is possible to observe in the provided images that the Tdt cells of Slc32a1 animals are in the RTN. The anatomical nomenclature is highly confusing in the manuscript. There is no evidence that the described inhibitory neurons are outside of the RTN. Parafacial or even pFRG terminologies only add to the confusion regarding the cell types present in this region of the brain and their role. Therefore, the authors studied the inhibitory neurons of the RTN of mice.

4. Is the CO2/H+ sensitivity of RTN neurons different before and after the inhibitory synaptic transmission blockade? In other words, what is the contribution of the synaptic disinhibition to the RTN neuronal firing response to hypercapnia/acidosis? The RTN neurons receive inhibitory synaptic inputs with very low frequency (0.2 Hz) and amplitude (12 pA), and hypercapnia/acidosis reduces it by nearly half. Therefore, in which extension such rare and small events contribute to the increases in RTN neurons firing frequency and breathing after SSt neurons chemogenetic inhibition?

5. The authors should provide the number and the rostro-caudal distributions of transfected SSt, CCK and Pvalb neurons, not only the "center" of bilateral virus injections or percentage of neurons, in the chemogenetic experiments. It is an important issue to report considering data reproducibility by other groups, and the extension of the transfected cells in the ventral medulla can provide evidence for different clusters of inhibitory neurons in mice controlling different functions.

6. What about the anatomical distribution of insensitive, activated, and inhibited SSt neurons in the RTN? The authors could provide this information in Figure 2. Are the capacitance values different between these neurons? In this regard, the neurons seem bigger (~100 µm) than other neurons described in the RTN based on the scales provided in Figure 2, panels Aii and Bii. Are the scales right?

7. Please remove the word "tonic" during the description/discussion of RTN neurons' synaptic inhibition. The authors did not evaluate whether or not the inhibitory synaptic events are tonic because there are no respiratory oscillations in the applied in vitro preparation.

8. Is the scale for measured current of the Figure 4 panel A right? The grouped data in panel B show that the amplitude of sEPSCs is -10 pA, but the representative traces show that it is the level of the electrical noise. Besides, the VC traces in Figure 3 panel A are not representing the grouped data of sIPSCs amplitude in panel B. The events are ~ 50 pA, while the grouped data are ~12.5 pA.

9. Please check the Cl- reversal potential; it is unlikely to be – 60 mV using the described intracellular and extracellular solutions.

10. – Discussion section, line 330: Expiratory neurons in the lateral parafacial region do not express Phox2b in rats (PMID: 28004411). Besides, what evidence shows that the expiratory neurons in the lateral parafacial region are not sensitive to hypercapnia/acidosis?

11. Discussion section, line 413: There is no evidence that disinhibition of Phox2b neurons in the lateral parafacial region evokes forced expiration in rats. In fact, previous studies (references: 19 and 53 and PMID: 28004411) demonstrated that bilateral disinhibition of neurons of the lateral parafacial region, which are located more laterally to the Phox2b positive RTN neurons, induces forced expiration in rats.

Reviewer #5:

The most important finding of this study is that Sst-expressing neurons in the parafacial region are involved in regulation of baseline respiratory activity via inhibitory effects on the RTN chemoreception. Authors provided evidences that supported this hypothesis.

This manuscript is revised version that has been reviewed by three referees. Authors' responses to comments seemed to be adequate.

First, they showed that the parafacial region contained a limited diversity of inhibitory neurons including somatostatin (Sst)-, parvalbumin (Pvalb)- and cholecystokinin (Cck)-expressing neurons. They showed that a subset of inhibitory parafacial neurons were inhibited by 10% CO2 and the response was retained when purinergic signaling and fast neurotransmission was blocked, suggesting these cells were intrinsically CO2/H+ sensitive. Next, they showed that CO2/H+ -inhibited cells were Sst-immunoreactive and were not immunoreactive for Pvalb or Cck, whereas CO2/H+ -insensitive cells did not express Sst. They also showed that sIPSCs in the RTN chemoreceptor neurons were depressed by 10% CO2. These results suggest that Sst-expressing neurons of inhibitory parafacial neurons were inhibited by CO2/H+ and then contributed to hypercapnic response of the RTN chemoreceptor. This hypothesis was further supported by results from in vivo experiment in which chemogenetic suppression of Sst+ parafacial neurons increases baseline breathing.

Although I have no major concern about this paper, please see below for my comments.

1. Page 9, line 212-214, "We found that 5 of 5 CO2/H+ -inhibited cells were Sst-immunoreactive (Figure 2Aii) and were not immunoreactive for Pvalb or Cck (data not shown), whereas 0 of 5 CO2/H+ -insensitive cells expressed Sst (Figure 2Bii).": This part is most important parts because it was the results indicating that Sst neurons were indeed inhibited by hypercapnic stimulation. Before this part, they showed 48 of 130 (37%) cells that were thought to be inhibitory neurons were inhibited hypercapnic stimulation (page 8). To further clarify properties of these cells, authors investigated 5 cells with combination of whole-cell recordings and immunoreactive examination and showed that cells that were inhibited by hypercapnic stimulation were Sst-immunoreactive. However, this result did not simply mean that all of above 37% neurons were Sst-immunoreactive. Although in vivo experiments supported the idea, is this number (n=5) really enough for the verification? In addition, I suppose that examination for Pvalb or Cck immunoreactivity was performed in cells different from the Sst examination group. Please describe the number of cells tested for Pvalb and Cck. Relating this issue, it would be important to show an example of detailed distribution of Sst (and maybe Pvalb and Cck) immunoreactive cells in the parafacial region together with Phox2b-expressing cells.

2. Page 8, the response to 10% CO2 was retained when purinergic signaling and fast neurotransmission was blocked: I suppose that these experiments were performed separately but not in the presence of purinergic signaling blockers plus fast neurotransmission blockers. I concern that an uncertainty may remain to conclude that parafacial inhibitory neurons are intrinsically CO2/H+ -sensitive.

3. Page 10, frequency of sIPSCs: "0.22Hz" seems to be rather low. How such low frequency could affect activity of the RTN chemoreceptor neurons?

https://doi.org/10.7554/eLife.60317.sa1

Author response

Essential revisions:

I would like to draw your attention to the fact that the three reviewers have carefully examined your manuscript and while they found a great interest for your results they also pointed out on several important aspects of the paper that require modifications. Their reviews are detailed below but I would like to emphasize on two major concerns that must be adequately addressed before suitability for publication in eLife could be considered:

First, they all noted a lack of precision in the description of the different methods used, in the size of the samples analyzed and illustrated in the figures, the parameters measured, etc… In addition, the experimental conditions in some cases are not always clear. More importantly, the authors must provide evidences for the specificity and selectivity of the transgenic lines used and the Dreadd and TdTomato expression and viral injections. In general, the methods need to be more precise and better explained throughout the text.

Second, the transcriptomic part, as it stands, does not really add much to the physiological question and remains a little apart from the rest of the study. For instance, it does not help in the identification of the neuronal cluster involved in the acid sensitivity of the RTN, and potential specific genes that could be indicative of a specific population of inhibitory neurons playing a role here are not used in the subsequent experiments presented after. Thus, either you put less emphasis in this part of the results (and even consider presenting these data at the end of the result section) or you provide a deeper analysis and a better link with the other presented data that would definitively bring a level of specificity and understanding that would greatly enhance the quality of the work and the interest of providing such genetic findings.

We thank the editor and reviewers for their time and thoughtful suggestions. To address the first concerns, we have significantly updated our descriptions and explanations in the results and methods sections. We have also confirmed specificity and selectivity of materials used and those new results are shown in Figure 1—figure supplement 2, Figure 2—figure supplement 1, and Figure 5—figure supplement 1.

To address the second concerns, we performed two additional sets of experiments to identify CO2/H+-sensitive inhibitory neurons in vitro and their contribution to respiratory activity in vivo. The details of these new experiments are described below but in short we show that i) CO2/H+-inhibited Vgat+ neurons express Sst (new Figure 2A-B), ii) and chemogenetic inhibition of Sst expressing parafacial neurons, but not Pvalb+ or CcK+ populations, modulated baseline breathing in a manner consistent with a role of these cells in regulating baseline respiratory activity (new Figure 5 and Figure 5—figure supplement 1). These new results clearly link our transcriptomic analysis of subpopulations of Vgat+ parafacial neurons to cellular CO2/H+ sensitivity and respiratory behavior.

In short, we believe that we have address all noted concerns and hope that the editor and reviewers agree the manuscript is ready for publication. Please seen point-by-point responses below for details.

Note that acceptability will have to be reassessed upon submission of the revised version.

Reviewer #1:

1. The Results section starts with single cell RNA sequencing data of RTN/parafacial neurons, identifying clusters of glutamatergic and GABA/glycinergic neurons. In itself, this is very valuable data, which could justify a stand-alone paper with further analysis. However, starting the Results section with these data is a bit awkward, as they are poorly linked to the rest of the study. Indeed, the authors provide new subclasses of inhibitory neurons, but then focus the rest of the study on overall inhibitory neurons without attempting to link their functional results with the subclasses of neurons molecularly identified. The transition line 164, to show that TdT is expressed in overall inhibitory neurons, without using tools for further specificity as one would expect after reading the previous paragraph on scRNAseq, is somewhat frustrating.

The paragraph starting line 172 shows that a subset of parafacial inhibitory neurons are CO2/H+ sensitive, and finishes with a statement (lines 188-190) that the identity of this subset of neurons remains to be determined. Testing whether this subset corresponds to a specific cluster of inhibitory neurons identified in the scRNAseq data presented just before would seem the obvious next step for the present study. The authors could have done immunohistochemistry on the neurons patched, as several of the clusters identified have markers for which antibodies exist, or better (but harder) they could have aspirated the patched cytoplasm and performed qPCR to reach the same results. Such data would not only provide a natural link between the first two experimental sets performed, but also provide more detailed information on the inhibitory parafacial neurons that are CO2/H+ sensitive that would significantly enhance the impact of this study.

With these comments in mind, if no further experiment is done for this paper, it might be better to move the scRNAseq data to the end of the Results section, to open the work towards more specificity in future studies. On that line, the end of the sentence lines 168-170, which closes the scRNAseq data presentation, is more suited for the end of the Results section, rather than the beginning.

We agree and have invested considerable time and effort into addressing this shortcoming. To identify CO2/H+ sensitive Vgat+ neurons, we used slice-patch electrophysiology to characterize Vgat+ parafacial neurons as CO2/H+-inhibited or -insensitive based on their firing response to 10% CO2. Once the cell type of interest was identified, we gained whole cell access to fill the cell with lucifer yellow (included in the pipette solution) followed by cell type specific immunolabeling. We found that 5 of 5 CO2/H+-inhibited neurons were Sst+, whereas 0 of 5 CO2/H+-insensitive neurons were Sst+. These new results (illustrated as new Figures 2Aii, Bii) suggest that Sst+ inhibitory neurons are CO2/H+ inhibited.

Next, we used chemogenetics to systematically suppress the activity of each main inhibitory neural population in the parafacial region. Specifically, we made bilateral injections of AAV2-hSyn-DIO-hM4D(Gi)-mCherry into the medial parafacial region of Sst-Cre (JAX #: 013044), Pval-Cre (JAX #: 008069) or Cck-Cre (JAX #: 012706) mice. After two weeks recovery, we found that systemic administration of clozapine increased baseline breathing (room air and 100% O2) in Sst-Cre mice but not the Pval- or Cck-Cre lines. In fact, inhibition of just Sst+ parafacial neurons recapitulated our original observation using a Vgat-cre line (Figure 5—figure supplement 1), suggesting that of these populations of parafacial inhibitory neurons only the SST+ population contribute to respiratory activity. Note that all chemogenetic experiments were compared pairwise to saline control, and we confirmed that clozapine had negligible effect on breathing control animals that did not receive AAV injections (Figure 5—figure supplement 2).

Our RNAseq results show that Sst+ defines the largest inhibitory population that varies based on expression of calretinin (Calb2, Slc32a1), reelin (Reln, Slc32a1) and neuronal nitric oxide synthase 1 (Nos1, Slc32a1). We have not yet determined whether CO2/H+-sensitivity is specific to one subset of Sst+ neurons or is a common feature of all Sst+ neurons in this region. These new results are shown in a new figure 5 and the text has been modified accordingly. Note that we opted to keep the scRNAseq results as figure 1 as it sets the stage for identifying populations of CO2/H+-sensitive inhibitory neurons and determine their contribution to breathing. We hope the reviewers agrees that these new data make for a more complete and exciting story.

2. In a previous article by the same group (Kuo FS et al., eLife 2019, DOI: https://doi.org/10.7554/eLife.43387), mice expressing a Dravet syndrome-associated Scn1a missense mutation conditionally in inhibitory neurons presented breathing alterations including diminished chemosensitivity, with hypo-excitable inhibitory neurons and hyper-excitable excitatory neurons in the RTN. The present study is proposed as Research Advance of this previous article. To strengthen the continuity with the previous study, the authors could have analysed the expression of Snc1a in their scRNAseq data, to identify whether Scn1a is expressed in a cluster of inhibitory neurons, and link it with the cluster that is responsive to CO2/H+. This way, the link would be more evident, and the present study would really make use of the scRNAseq data. In the current state of this study, I somewhat fail to see its strong and direct continuity with the previous work justifying a Research Advance format, it would maybe be better suited as a traditional stand-alone Research Article.

The previous study noted by the reviewer showed that expression of a Dravet syndrome associated ion channel mutation in inhibitory neurons disrupted RTN chemoreceptor function and respiratory activity. These findings underscore the need to understand whether and how inhibitory neurons in the parafacial region contribute to breathing. Here, we meet this need by i) characterizing inhibitory neuron diversity within the parafacial region, ii) identifying which subsets of inhibitory neurons are CO2/H+ sensitive, iii) establishing that inhibitory synaptic activity regulate activity of RTN chemoreceptors in a CO2/H+ dependent manner, and iv) showing how specific subsets of inhibitory parafacial neurons contribute to respiratory activity. Therefore, we believe these results represent an important extension of our previous study and are entirely appropriate for the Research Advance format.

To further link the present result to our previous study, we agree that it would be useful to characterize expression of Scn1a in the parafacial region. We have included a supplemental figure (Figure 1—figure supplement 3) that shows expression of Scn1a transcript in parafacial cells from 10 day old mice. Consistent with our previous study noted above, we show that Scn1a transcript is widely expressed in the parafacial region including both glutamatergic and inhibitory neurons. We have modified the text to include this point.

3. Was synaptic isolation performed when testing the CO2/H+ sensitivity of parafacial inhibitory neurons? Did you also block the usual gliotransmitters? The methods section is not clear, please provide more details, as whether these neurons are intrinsically sensitive to CO2/H+ or not is critical for interpretation of these data.

Good point. We performed additional slice-patch experiments to determine whether Vgat+ neurons are intrinsically inhibited by CO2/H+. We found that exposure to 10% CO2 decreased activity by 1.7 ± 0.2 Hz under control conditions, 2.6 ± 0.3 Hz when glutamate and GABA/glycine signaling was blocked with CNQX (10 µM), strychnine (2 µM) and gabazine (10 µM), and by 1.1 ± 0.1 Hz when purinergic receptors were blocked with 8PT (10 µM) and PPADS (100 µM). Note that CO2/H+ evoked ATP release from astrocytes is the only form of glial communication implicated in RTN chemoreception (PMID: 20647426) and previous work showed that inhibitory input to RTN neurons is augmented by P2X signaling (PMID: 16822980). Therefore, for these experiments we targeted purinergic signaling as a means of probing for roles of glia in this response. These results are shown a new Figure 2—figure supplement 2.

Reviewer #2:

[…] 1. In the first paragraph of the discussion, the authors conclude that inhibitory inputs to chemosensing RTN neurons are withdrawn/inhibited during exposure to high CO2. The net result is a lessening of GABA restraint on chemosensing neurons and an increase in their neural drive and neurotransmitter release. Another valid interpretation could be that increased neurotransmission from other tonic inputs to the RTN (arising from chemoreceptors activated during high CO2 exposures) contributes to the ventilatory response? Ie the VGAT inhibitory input isn't withdrawn but rather contributes much less to RTN drive during exposure to high CO2. Please include this interpretation, and other alternatives in the discussion.

As suggested by the reviewer, we think CO2/H+ increases excitatory synaptic drive to RTN chemoreceptors while simultaneously decreasing inhibitory synaptic input. To support this, we show that exposure to high CO2/H+ increased the frequency of excitatory postsynaptic currents (EPSCs) (Figure 4) and decreased frequency of inhibitory postsynaptic currents (IPSCs) recorded from RTN chemoreceptors (Figure 3). Note that these experiments were performed using a Cs+ based internal to block all postsynaptic K+ channels (including TASK2 and leak K+ targeted by GPR4). We also study the effects of CO2/H+ on EPSCs and IPSCs in relative isolation by holding membrane potential at the reversal potential for chloride (-60 mV) or AMPA receptors (0 mV). For example, to study IPSCs we held membrane potential at 0 mV which eliminated AMPA mediated EPSCs. Under these conditions, the observed effect of CO2/H+ on IPSC frequency is not due (at least directly) to changes in excitatory input. However, to the reviewers point, CO2/H+ may elicit the release of non-glutamatergic neuromodulators that alter postsynaptic conductance to indirectly effect postsynaptic current amplitude. We cannot exclude involvement of this mechanism in the observed CO2/H+-induced decrease in IPSC amplitude. Therefore, we have added the following statement to the discussion “It should also be noted that CO2/H+ may elicit release of non-glutamatergic neuromodulators that alter postsynaptic conductance and potentially contribute to diminished IPSC amplitude; however, we consider this unlikely since such a non-specific mechanism is expected to have similar effects on both excitatory and inhibitory synaptic currents”.

2. DREADD-mediated inhibition of inhibitory RTN neurons increased baseline minute ventilation but had no effect on ventilatory responses to CO2. Figure 6D-E – most columns do not include data from n=11 mice. Please explain why certain mice/replicates were excluded from the dataset. Figure 6 figure legend is incomplete – there is no description of the data shown for control Vgatcre (Figure 6F-H). Also, there are two Figure 6F.

Thank you for your comment. Results from all 11 mice are shown in Figure 6C-I (now moved to supplemental figure 1). Some data points overlap making is somewhat difficult to see all individual data points. We have restructured this figure (now Figure 5—figure supplement 1) to omit any panel numbering redundancies.

3. CO2/H+ synaptic properties of RTN neurons studies (page 8, line 193):

"As previously defined (26), RTN neurons were considered chemosensitive if they show some level of spontaneous activity under control conditions and a robust firing rate response to 10% CO2 (1.6 {plus minus} 0.36 Hz; N=10)." – 10% CO2 FR response of 1.6 {plus minus} 0.36 Hz does not seem 'robust' when compared to other previous studies. For example, Wang S et al. 2013 Figure 2D (citation #64), demonstrates FR at pH 7 (~8% CO2) to be >4 Hz for both TypeI and TypeII dissociated RTN neurons. Wouldn't you expect FR to be >>4Hz in 10% CO2, for RTN neurons in an intact brain slice?

The CO2/H+ response of RTN chemoreceptors described here is consistent with what has been previously described for type I RTN chemoreceptors in both the brain slice and acute dissociated preparations. For example, we show here that increasing CO2 from 5% (pH 7.3) to 10% (pH 7.0) increased activity by 1.6+/- 0.36 Hz. The study by Wang S et al., noted by the reviewer showed that acutely dissociated RTN chemoreceptors fire at ~2 Hz under control conditions (pH 7.3) and increase their firing discharge to ~3.5 Hz in pH 7.0. Thus, 0.3 pH unit acidification increased activity by ~ 1.5 Hz. Likewise, RTN chemoreceptors in medullary slices showed a similar degree of CO2/H+ sensitivity (PMID: 15558061).

4. Figure 2 and Figure 2 legend:

Figure 2B – error bars appear to be missing for this n=4 quantified data.

This mistake has been corrected.

Figure 2Aii is an enlarged version of the image at bregma -6.05mm. You could omit this, I don't think it adds anything to the paper. Alternatively, magnify further, use arrows to draw attention to cells of interest.

We agree. The figure panel has been omitted.

Figure 2 legend: include abbreviations (7N, b).

Added.

5. Figure 3C legend: It is not clear what the summary data corresponds to. Indicate this in the figure legend as follows: ' firing responses of CO2/H+ -inhibited (n=20, blue), -insensitive (n=53, grey) and -activated (n=7, red). What parameters do the authors use, to define inhibited versus activated neurons?

All VGAT+ cells that responded reversibly to 10% CO2 with ≥ 20% change in firing were considered CO2/H+-sensitive. We have edited the text and figure to make this clear.

6. Figure 5 – please confirm that there are no outliers in these values. One replicate seems to have a much higher baseline sEPSC frequency than the others.

Thanks for bringing this to our attention, outlier analysis was used to identify two cells that were subsequently omitted from the sEPSC dataset, making the total N=8 cells. We have also noted this in the figure legend and transparent reporting forms.

7. Supplementary Figure 1 figure legend states that the colors of distributions correspond to classes shown in Figure 1A however Figure 1A displays only 3 colours, whereas Supp Figure 1D displays 5 colours/classes. Please clarify.

Figure 1—figure supplement 1 shows cut offs for denoting Slc17a6 and Slc32a1 cells. There are only three colors shown- blue, green, and red correspond with Slc32a1, Slc17a6, and Chat, respectively. The top expression graph shows low Slc32a1 and Chat in Slc17a6 cells (3 colors). The bottom expression graph shows low Slc17a6 and Chat in Slc32a1 cells; however, those transcript counts overlap (red and green) making it appear as additional color. We have modified the legend to make this more clear.

8. Authors used RNA seq to identify differentially expressed and coregulated genes in RTN neurons in order to infer biological meaning for further studies on this preeminent central chemoreceptor population. Single cell isolation method and RNA-seq methodology was extremely detailed, valuable information regarding inclusion and exclusion criteria were included. The authors use tSNE to investigate segmentation, clustering of genes in the pFRG transcriptome. The data are convincing, since it is well established using other methods, that ~50% of glutamatergic RTN neurons express galanin. Could the authors clarify how they minimised batch effect, which can occur during the experiment, the RNA library preparation, or the sequencing run?

During library preparation, prep kits from the same batch/lot were used and libraries were constructed in parallel. To limit bias from sequencing on different flow cell lanes, and to remove unwanted biological variation, the scRNA-seq data analysis attempted to minimize batch effects in several ways.

After quality control (removing low quality cells) and data normalization, the subsequent analysis is guided by the selection of a set of influential genes. Often called highly-variable genes, these genes are those with the largest variance in expression relative to their mean expression (dispersion) within different bins of mean expression. Due to their high variance, using these genes as the primary features for downstream analysis preserves the variability in the dataset while greatly reduces its dimensionality. From this set of highly-variable genes, we exclude genes which may represent sources of unwanted biological and technical variation so that they do not influence downstream analysis; mitochondrial, ribosomal, hemoglobin, cell-cycle, and stress-related genes are excluded as are sex-related genes such as Xist and genes on the Y-chromosome.

One of the downstream analysis steps involves computing a nearest-neighbor graph- which is a weighted network connecting neighborhoods of related cells together based on shared (influential) gene expression levels. It’s at this step where we utilize the fast and efficient batch correction tool BBKNN (10.1093/bioinformatics/btz625) which computes a batch-corrected nearest-neighbor graph. This graph is then used for downstream clustering and the ensuing cell-type identification and sub-clustering analysis.

There are 20 global clusters (from 2000 genes) – the distribution of cells/cluster and genes/cluster is illustrated in Supp Figure 1C. The basis for the clustering was not clear – could the authors clarify this and make it clear in the text. It appears (from Supp Figure 1 B) that each cluster is defined by related genes that display a high level of expression within the cluster.

Yes, that is roughly correct. As described above, the analysis is guided by the selection of roughly 2000 influential genes that are fed into principal component analysis and then the construction of a batch corrected nearest-neighbor graph. The clustering assignments come from the Leiden community detection algorithm (10.1038/s41598-019-41695-z) which operates on a graph to yield near-optimal partitioning/clustering at a specified resolution (effective partition size). In other words, cells are organized into communities similar to a social network based on their shared expression patterns of these 2000 influential genes, and cluster labels are given to groups of cells based on how similar cells within the group are to one another versus cells outside the group.

This clustering process is the current standard in the field and is integrated into the two most popular scRNA-seq analysis toolkits, ScanPy (used in this study) (10.1186/s13059-017-1382-0) and Seurat (10.1038/nbt.4096). Furthermore, the analysis of this data can be found in the GitHub repository (https://github.com/TheJacksonLaboratory/ventral-parafacial-neuron-scrnaseq).

Reviewer #3:

RTN is a small cluster of lower brainstem neurons with a well-described molecular signature. These neurons respond vigorously to CO2 in vivo, contribute to arterial PCO2 homeostasis and maintain breathing automaticity during sleep. Their synaptic inputs are not well identified however and the way in which they encode brain PCO2 in vivo is complex and much debated. Several mechanisms have been proposed (intrinsic pH sensitivity of RTN neurons, astrocyte-mediated paracrine effect of acid, paradoxical acid-mediated contraction of microvessels and synaptic inputs from acid-activated neurons, e.g. serotonergic). The present study explores the possible contribution of local inhibitory neurons to the activity and acid-sensitivity of RTN neurons. Consistent with prior EM evidence, the results show that RTN neurons receive GABA/glycinergic inputs. The contribution of a subset of this input to the CO2 response of RTN neurons is suggested by the results of experiments in slices but the existence of this mechanism is not convincingly supported by the in vivo experiments.

The authors confirm the transcriptome of RTN neurons and describe that of several subsets of inhibitory neurons (likely GABA- and glycine-ergic) that reside in their vicinity. This work greatly extends existing observations regarding these local inhibitory neurons. Unfortunately this work did not help identify the particular inhibitory interneurons postulated to be CO2-responsive.

Next, the authors show that a subset of these inhibitory neurons are mildly inhibited by acidification in slices. The effects are small and variable from cell to cell but suggest that CO2-triggered disinhibition could perhaps contribute to the overall excitatory effect of CO2 on RTN neurons.

This hypothesis is then tested, by determining whether chemogenetic inhibition of these inhibitory interneurons enhances breathing and the respiratory chemoreflex in conscious mice. The authors show is that inhibition of the GABA/glycine interneurons (via a Gi-coupled DREAAD) in unanesthetized mice elevates breathing slightly at rest but seems to have no effect on the breathing stimulation elicited by CO2. The increase in resting breathing is consistent with prior evidence that breathing is increased when a GABA-receptor antagonist is microinjected into the region containing the RTN neurons (Nattie et al) but this fact does not demonstrate that the increase in breathing is caused by activation of RTN neurons. As the authors acknowledge in the discussion the increase in breathing produced by inhibiting GABA interneurons in this brain region could also result from arousal. The increase in breathing could also result from the disinhibition of neurons other than RTN, for example the C1 neurons.

Finally, the main interpretative difficulty is that inhibition of the GABA/glycine interneurons does not change the respiratory chemoreflex; this is difficult to reconcile with the results obtained in slices and makes one wonder whether the acid inhibition of the GABA interneurons observed in slices is an artifact or at least whether it is a phenomenon that has any relevance to the CO2 response of RTN neurons in vivo.

We thank the reviewer for their constructive suggestions. We hope that we have satisfactorily address noticed concerns in our point by point responses below.

1. Re: results line 177. Although the authors are probably correct, they need to validate the assumption that their reporter line (Vgatcre::TdT) does indeed "allow for selective targeting of inhibitory neurons in the region of interest". If this has been demonstrated before please give the specific quote. If not, the authors must demonstrate that the vast majority of the tomato-red positive neurons in their region of interest express VGat (transcript or protein) at the relevant postnatal age.

Agreed. We confirmed that 81% and 83% of TdT labeling co-colocalize with Gad67 in P7 and P20 VgatCre::TdT mice, respectively (Figure 2—figure supplement 1). At both developmental timepoints we found 8% of cells labeled with TdT only and the remaining were Gad67 only. Furthermore, cell type-specific expression of Cre recombinase in Sst-Cre (PMID 24690741), Pvalb-Cre (PMID 31160332) and Cck-Cre (PMID 33173030) lines used here have been confirmed previously. Therefore, we are confident in our ability to selectively target these populations for in vivo functional assessment.

2. Re: lines 240-241. "Specifically, we bilaterally injected AAV2-hSyn-DIO-hM4D(Gi)-mCherry (10 nL/side, Addgene) into the medial portion of the RTN in VGAT-cre mice". The authors need to also demonstrate here that VGat+-neurons were selectively transduced (i.e. that mCherry and VGat were colocalized). The selectivity of this type of approach is typically overrated (ectopic expression of Cre and Cre-independent effects of the DIO viral prep) and must be verified.

Agreed. We confirmed that 99% of AAV mCherry labeled cells were Gad67-immunoreactive. We also confirm that 95% of Gad67-immunoreactive neurons in the parafacial region were co-labeled with mCherry (Figure 5—figure supplement 1). These results confirm the specificity and efficiency of our viral expression system.

3. Re: results lines 177-183. The defining criteria for pH sensitivity is fine but the results are less than impressive: 20% of neurons were inhibited to some degree (-20% or more) when exposed to a fairly severe acidification (pH 7.0), 66% were unaffected and 9 % were judged to be excited. Figure 3C should be replaced with a distribution histogram summarizing the % response of all the VGAt neurons sampled (e.g. 10% effect binning). Small positive or negative effects of acidification on unit-activity in slices have frequently been reported in other brainstem areas such as the nucleus of the solitary tract. Cherry-picking neurons that are excited and those that are inhibited is not terribly convincing unless this grouping correlates with an identifiable biochemical or structural marker, which was not the case here.

We performed additional experiments and found that 48 of 130 (37%) parafacial Vgat+ cells are robustly and reversibly inhibited by 10% CO2. The fact that most inhibitory neurons in this region do not respond to this level of CO2 is reassuring because it suggests that our stimulus is not a sledgehammer. It also suggests those cells that do respond are different and perhaps specialized sense and respond to CO2 and considering this is an important respiratory chemoreceptor region, it’s tempting to speculate these CO2/H+-inhibited parafacial neurons contribute to RTN chemoreception. As suggested, we included a histogram showing CO2/H+-induced change in firing for all Vgat neurons sampled (Figure 2D).

We have also determined that CO2/H+-inhibited parafacial neurons are SST-immunoreactive, whereas Vgat+ neurons that do not respond to CO2/H+ are also not SST-immunoreactive (Figure 2A-B). Based on this and guided by our RNAseq results that identify the main subsets of parafacial inhibitory neurons (Figure 1C), we went on to show that chemogenetic inhibition of Sst+ cells (Figure 5A) but not Pvalb+ (Figure 5E) or CcK+ (Figure 5F) parafacial neurons increased baseline breathing in manner consistent with disinhibition. These results also mimic the respiratory phenotype elicited by silencing all parafacial inhibitory neurons (Figure 5 Supplement 1), thus suggesting Sst+ parafacial neurons have a unique role in control of breathing. The text has been modified extensively to include these new results.

4. Re: lines 248-249 and discussion. The increase breathing could have many causes and may not have nothing to do with CO2 chemoreception. For example, selective activation of the rostral C1 neurons produces a very strong increase in breathing in unanesthetized mice and these neurons are tonically inhibited by GABA-ergic input (the basis of the baroreflex). The increased breathing noted at baseline (no CO2 added) could also denote arousal from sleep, an effect produced by activation of the C1 neurons, the RTN itself and perhaps other neurons in the area.

We agree and have modified the last paragraph of the discussion to make clear that parafacial inhibitory neurons (Sst+ in particular) likely regulate breathing at multiple levels of the respiratory circuit. It is worth noting that activation of adrenergic C1 has been shown to increase sigh activity, whereas CO2 or optogenetic activation of chemosensitive RTN neurons has been shown to increase respiratory output but with no change in sighing (PMID: 25325789). Also chemogenetic inhibition of Vgat-cre or Sst-cre parafacial neurons did not elicit obvious sigh activity. Therefore, we do not think C1 neurons have a dominate role in this response.

On the other hand, a yet unknown population of parafacial Vgat+ neurons influence sleep-wake states and so arousal may contribute to baseline breathing responses caused by inhibition of Vgat+ or Sst+ parafacial neurons. This interesting possibility will require additional experiments that are beyond the scope of this study.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

We apologize for the delay in providing this review. But two out of the three initial reviewers declined to re-review the paper and we had to find two other ones. Consequently, you will find some new concerns. However, be aware that your responses to previous comments have been judged adequate and that the additional experiments now included definitively improve the paper. We are happy to publish the paper in this form.

However, please take the time to read carefully all comments and address those that you choose to address. Basically, they highlight the following potential improvements: A better definition of your hypothesis; clarifying the functional connections between the different sets of data; stating unambiguously whether you consider the neurons of interest as part of the RTN/pFRG or not; providing a better description of the inhibitory neurons investigated in terms of distribution and number; and paying attention to the scales in different figures that might be wrong.

Thank you for your time and constructive feedback! In short, we have clarified our hypothesis, avoided the use of terminology that confused CO2/H+-sensitive inhibitory cells with RTN chemoreceptors, and made all suggested edits to the text and figures.

However, for the sake of time we chose not to include a detailed characterization of the distribution of parafacial inhibitory neurons or assessment of RTN to Sst+ parafacial connectivity. We hope you and the reviewers find the revised manuscript accepted for publication in eLife.

Reviewer #2:

[…] Whilst important to point out the additional experiments required to definitively demonstrate that the inhibitory RNAseq cluster corresponds with the H+ inhibited inhibitory neurons in the RTN, Reviewer 1 acknowledges the difficult nature of these experiments. The authors could include this as future directions in the Discussion section.

We thank you for this comment. We have included this as a future direction in the discussion.

Reviewer #4:

[…] 1. It would be essential to define the hypothesis of the present study clearly. The authors "considered" the possibility that ventral medullary neurons in the region of the RTN sense changes in CO2/H+ and regulate RTN chemoreception by a mechanism involving disinhibition. However, the authors provided evidence for the ventral medullary inhibitory neurons' involvement in neither the RTN chemoreception nor RTN disinhibition. The authors then propose a novel mode of chemotransduction involving regulation of basal breathing by CO2/H+-dependent disinhibition. In fact, the experiments involving the chemogenetic inhibition of SSt neurons did not demonstrate that the RTN neuronal function was affected. The neurons in the ventral medullary region close to RTN project to the pre-Bötzinger Complex (PMID: 26855425), and these projections could be involved in the increases of baseline breathing following SSt neurons chemogenetic inhibition.

Good point. We show that chemogenetic inhibition of Sst+ parafacial neurons increased baseline breathing, and our cellular data is consistent with CO2/H+-dependent disinhibition of RTN chemoreceptors; however, we do not show that CO2/H+-sensitive Sst+ neurons directly innervate RTN chemoreceptors or that Sst+ parafacial neurons regulate breathing via the RTN. Based on this, we broadly define our hypothesis as “ventral parafacial inhibitory neurons in the region of the RTN sense changes in CO2/H+ and regulate baseline breathing by a mechanism involving disinhibition”. We also noted in the discussion that it is not yet clear whether CO2/H+ inhibited parafacial neurons regulate breathing by disinhibition of RTN or other elements of respiratory control including the pre-BötC.

2. There are no clear functional connections between some of the results described. The authors described the presence of a diversity of inhibitory neurons in the ventral medullary region and that the inhibition of SSt neurons increased the baseline breathing without affecting the ventilatory responses to hypercapnia/acidosis. On the other hand, there are data showing that RTN neurons receive synaptic inhibition, which was reduced in response to hypercapnia/acidosis. Therefore, the authors need to demonstrate that: i) the three different populations (activated, insensitive, and inhibited) of SSt neurons are projecting to RTN chemosensitive neurons, and: ii) the chemogenetic inhibition of SSt neurons increases the RTN neurons firing frequency by reducing the inhibitory synaptic transmission. The SSt neurons recorded in vitro were labeled, and the projections to the RTN neurons can be revealed in the same slice. Besides, the effects of SSt neurons inhibition (chemogenetics) on the RTN firing frequency and inhibitory synaptic transmission can also be analyzed in in vitro experiments using the transfected animals. It is important to note that without new experiments to prove the role of the inhibitory ventral medullary SSt neurons in the RTN chemosensitivity, the authors should consider presenting the data in two (I – somatostatin-expressing RTN neurons regulate baseline breathing; II – the effects of hypercapnia/acidosis on the synaptic transmission to RTN neurons) different manuscripts.

We agree that our results fall short of definitively proving that Sst+ parafacial neurons directly regulate activity of RTN chemoreceptors. Specifically, we show that i) Sst+ parafacial neurons are inhibited by CO2/H+, ii) RTN chemoreceptors receive inhibitory input that is withdrawn during CO2/H+; iii) bath application of GABA and glycine receptor blockers increase baseline activity of RTN chemoreceptors (new results), and iv) chemogenetic inhibition of Sst+ but not Pvalb+ or CcK+ parafacial neurons increased baseline breathing. Based on this, we think it is reasonable to suggest Sst+ parafacial neurons regulate activity of RTN chemoreceptors. We also believe these complementary results should be published as one manuscript.

3. The authors demonstrated that the inhibitory neurons are located close to the RTN chemoreceptors. In fact, it is possible to observe in the provided images that the Tdt cells of Slc32a1 animals are in the RTN. The anatomical nomenclature is highly confusing in the manuscript. There is no evidence that the described inhibitory neurons are outside of the RTN. Parafacial or even pFRG terminologies only add to the confusion regarding the cell types present in this region of the brain and their role. Therefore, the authors studied the inhibitory neurons of the RTN of mice.

We appreciate the reviewers point. All cells included in this study were located in the ventral parafacial region that overlaps with the RTN. However, since the term RTN is typically used to define CO2/H+ activated glutamatergic neurons in this region (PMID: 31635852) (clusters 1-2 in Figure 1B), we referring to CO2/H+ sensitive inhibitory neurons as ‘inhibitory neurons of the RTN’ is more confusing then CO2/H+-inhibited parafacial neurons. We used this terminology throughout the manuscript.

4. Is the CO2/H+ sensitivity of RTN neurons different before and after the inhibitory synaptic transmission blockade? In other words, what is the contribution of the synaptic disinhibition to the RTN neuronal firing response to hypercapnia/acidosis? The RTN neurons receive inhibitory synaptic inputs with very low frequency (0.2 Hz) and amplitude (12 pA), and hypercapnia/acidosis reduces it by nearly half. Therefore, in which extension such rare and small events contribute to the increases in RTN neurons firing frequency and breathing after SSt neurons chemogenetic inhibition?

Thank you for bringing up this point. We added new in vitro results to show that bath application of bicuculine (10 µM) and strychnine (2 µM) increased activity of RTN chemoreceptors by 0.7 ± 0.2 Hz (T12=2.201, p=0.022). This finding suggests inhibitory input partly limits activity of RTN chemoreceptors under baseline conditions. We also found the firing response to 10% CO2 was similar under control conditions and in the presence of bicuculine and strychnine (Δ -0.3 ± 0.2 Hz; T12=1.246, p>0.05). These results are consistent with in vivo chemogenetic experiments showing that inhibition of parafacial Sst+ parafacial neurons increased baseline activity but with no change in the ventilatory response to CO2. We have added these results to the text.

5. The authors should provide the number and the rostro-caudal distributions of transfected SSt, CCK and Pvalb neurons, not only the "center" of bilateral virus injections or percentage of neurons, in the chemogenetic experiments. It is an important issue to report considering data reproducibility by other groups, and the extension of the transfected cells in the ventral medulla can provide evidence for different clusters of inhibitory neurons in mice controlling different functions.

We have not systematically characterized the rostro-caudal distribution of transfected cells for each cre line. However, as shown in Figure 5A—figure supplement 1, injections were centered in the RTN region and diffused ~600 µm in the rostro-caudal direction and spread laterally to include expiratory parafacial neurons

6. What about the anatomical distribution of insensitive, activated, and inhibited SSt neurons in the RTN? The authors could provide this information in Figure 2.

CO2/H+-insensitive and -inhibited neurons were found in both the medial and lateral parafacial regions. Unfortunately, only a limited number of recorded cells were filled with a fluorescent dye so we are not able to provide more information regarding locations of these cell types at this time.

Are the capacitance values different between these neurons? In this regard, the neurons seem bigger (~100 µm) than other neurons described in the RTN based on the scales provided in Figure 2, panels Aii and Bii. Are the scales right?

Thank you for this point. The scale bars in Figure 2 Aii and Bii should be 50 μm not 100 um. The figure has been corrected.

7. Please remove the word "tonic" during the description/discussion of RTN neurons' synaptic inhibition. The authors did not evaluate whether or not the inhibitory synaptic events are tonic because there are no respiratory oscillations in the applied in vitro preparation.

Thanks. We no longer use the word ‘tonic’ to refer to inhibitory input to RTN neurons under control conditions.

8. Is the scale for measured current of the Figure 4 panel A right? The grouped data in panel B show that the amplitude of sEPSCs is -10 pA, but the representative traces show that it is the level of the electrical noise. Besides, the VC traces in Figure 3 panel A are not representing the grouped data of sIPSCs amplitude in panel B. The events are ~ 50 pA, while the grouped data are ~12.5 pA.

Thanks. The scale bars in Figure 3A and 4A were incorrectly labeled, 3A should be 25 pA (not 50 pA) and 4A should be 10 pA rather than 20 pA. The figures have been corrected.

9. Please check the Cl- reversal potential; it is unlikely to be – 60 mV using the described intracellular and extracellular solutions.

Thank you for bringing this to our attention. The reviewer is correct, the Cl- reversal potential for our experimental conditions is approximately -84 mV. Therefore, the holding potential used for recording EPSCs was depolarized to the Cl- reversal potential. However, we consider this a minor issue because potential contaminating IPSCs could easily be excluded from EPSC analysis based on the direction of synaptic currents; inward for EPSCs and outward for IPSCs. We also pharmacologically confirmed EPSCs as glutamatergic at the end of each experiment by bath application of CNQX. We have clarified these points in the text.

10. Discussion section, line 330: Expiratory neurons in the lateral parafacial region do not express Phox2b in rats (PMID: 28004411). Besides, what evidence shows that the expiratory neurons in the lateral parafacial region are not sensitive to hypercapnia/acidosis?

It is not entirely clear whether (or not) expiratory neurons in the lateral parafacial region express Phox2b. The study noted by the reviewer (PMID: 28004411) showed in rats that expiratory parafacial neurons are not Phox2b-immunoreactive. They also showed that blocking glutamate receptors in the lateral parafacial region minimally effected CO2/H+-induced active expiration. These results suggest expiratory parafacial neurons are distinct from and not dependent on RTN chemoreceptors. Conversely, other work in rats showed selective activation (PMID: 32973046) or inhibition (PMID: 20844141) of ventral parafacial Phox2b+ neurons increased and decreased expiratory activity, respectively. These results suggest one population of Phox2b+ parafacial neurons regulate both inspiratory and expiratory activity. We have clarified this in the text.

It is also possible expiratory parafacial neurons are CO2/H+-sensitive so we have modified the text to make this clearer.

11. Discussion section, line 413: There is no evidence that disinhibition of Phox2b neurons in the lateral parafacial region evokes forced expiration in rats. In fact, previous studies (references: 19 and 53 and PMID: 28004411) demonstrated that bilateral disinhibition of neurons of the lateral parafacial region, which are located more laterally to the Phox2b positive RTN neurons, induces forced expiration in rats.

Sorry for this confusion. We now refer to these cells as lateral parafacial neurons.

Reviewer #5:

[…] Although I have no major concern about this paper, please see below for my comments.

1. Page 9, line 212-214, "We found that 5 of 5 CO2/H+ -inhibited cells were Sst-immunoreactive (Figure 2Aii) and were not immunoreactive for Pvalb or Cck (data not shown), whereas 0 of 5 CO2/H+ -insensitive cells expressed Sst (Figure 2Bii).": This part is most important parts because it was the results indicating that Sst neurons were indeed inhibited by hypercapnic stimulation. Before this part, they showed 48 of 130 (37%) cells that were thought to be inhibitory neurons were inhibited hypercapnic stimulation (page 8). To further clarify properties of these cells, authors investigated 5 cells with combination of whole-cell recordings and immunoreactive examination and showed that cells that were inhibited by hypercapnic stimulation were Sst-immunoreactive. However, this result did not simply mean that all of above 37% neurons were Sst-immunoreactive. Although in vivo experiments supported the idea, is this number (n=5) really enough for the verification? In addition, I suppose that examination for Pvalb or Cck immunoreactivity was performed in cells different from the Sst examination group. Please describe the number of cells tested for Pvalb and Cck. Relating this issue, it would be important to show an example of detailed distribution of Sst (and maybe Pvalb and Cck) immunoreactive cells in the parafacial region together with Phox2b-expressing cells.

We recognize your concern. However, please note that in addition to showing that 5 of 5 CO2/H+ inhibited cells were Sst+ we also showed that 0 of 5 CO2/H+ insensitive cells were Sst+. Based on this result we tentatively ruled out involvement of Pvalb and CCK neurons in CO2 sensing, and thus did not expand our IHC assessment of CO2/H+ sensitive cells to include markers of Pvalb or CCK. This decision was further justified by chemogenetic experiments showing selective inhibition of Sst+ neurons but not Pvalb+ or CcK+ neurons in the parafacial region disrupted baseline breathing.

Considerable time and effort will be required to characterize the distribution of Sst+ or other subsets of inhibitory neurons in this region. Therefore, to avoid further delays we chose to publish this work without that analysis.

2. Page 8, the response to 10% CO2 was retained when purinergic signaling and fast neurotransmission was blocked: I suppose that these experiments were performed separately but not in the presence of purinergic signaling blockers plus fast neurotransmission blockers. I concern that an uncertainty may remain to conclude that parafacial inhibitory neurons are intrinsically CO2/H+ -sensitive.

To confirm, these experiments were performed separately and so it remains possible that synaptic or paracrine mechanisms contribute to CO2/H+ sensing by inhibitory parafacial neurons. However, preliminary whole cell voltage clamp experiments (in the presence of TTX) suggest CO2/H+ inhibits Sst+ parafacial neurons directly by activation of an inward rectifying K+ conductance. These experiments are ongoing and will be explored in detail in future publications.

3. Page 10, frequency of sIPSCs: "0.22Hz" seems to be rather low. How such low frequency could affect activity of the RTN chemoreceptor neurons?

Good point. To address this concern, we included new data showing bath application of bicuculine (10 µM) and strychnine (2 µM) increased baseline activity of RTN chemoreceptors by 0.7 ± 0.2 Hz (T12=2.201, p=0.022). This finding suggests inhibitory input partly limits activity of RTN chemoreceptors under baseline conditions. We also found the firing response to 10% CO2 was similar under control conditions and in the presence of bicuculine and strychnine (Δ -0.3 ± 0.2 Hz; T12=1.246, p>0.05). These results are consistent with in vivo chemogenetic experiments showing that inhibition of parafacial Sst+ parafacial neurons increased baseline activity but with no change in the ventilatory response to CO2. We have added these results to the text.

https://doi.org/10.7554/eLife.60317.sa2

Article and author information

Author details

  1. Colin M Cleary

    Department of Physiology and Neurobiology, University of Connecticut, Storrs, United States
    Contribution
    Data curation, Formal analysis, Funding acquisition, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-0305-1324
  2. Brenda M Milla

    Department of Physiology and Neurobiology, University of Connecticut, Storrs, United States
    Contribution
    Data curation, Funding acquisition, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  3. Fu-Shan Kuo

    Department of Physiology and Neurobiology, University of Connecticut, Storrs, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Writing - review and editing
    Competing interests
    No competing interests declared
  4. Shaun James

    Department of Physiology and Neurobiology, University of Connecticut, Storrs, United States
    Contribution
    Data curation, Formal analysis, Writing - review and editing
    Competing interests
    No competing interests declared
  5. William F Flynn

    The Jackson Laboratory for Genomic Medicine, Farmington, United States
    Contribution
    Data curation, Formal analysis, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6533-0340
  6. Paul Robson

    1. The Jackson Laboratory for Genomic Medicine, Farmington, United States
    2. Institute for Systems Genomics, University of Connecticut, Farmington, United States
    Contribution
    Data curation, Formal analysis, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-0191-3958
  7. Daniel K Mulkey

    Department of Physiology and Neurobiology, University of Connecticut, Storrs, United States
    Contribution
    Conceptualization, Supervision, Funding acquisition, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    daniel.mulkey@uconn.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7040-3927

Funding

National Institutes of Health (HL104101)

  • Daniel K Mulkey

National Institutes of Health (HL137094)

  • Daniel K Mulkey

National Institutes of Health (NS099887)

  • Daniel K Mulkey

National Institutes of Health (HL142227)

  • Colin M Cleary

National Institutes of Health (F31NS120467)

  • Brenda M Milla

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This work was supported by funds from the National Institutes of Health Grants HL104101 (DKM), HL137094 (DKM), NS099887 (DKM), F31HL142227 (CMC) and F31NS120467 (BMM).

Ethics

Animal experimentation: All procedures were performed in accordance with National Institutes of Health and University of Connecticut Animal Care and Use Guidelines (protocols A19-048 and A20-016).

Senior Editor

  1. Ronald L Calabrese, Emory University, United States

Reviewing Editor

  1. Muriel Thoby-Brisson, CNRS Université de Bordeaux, France

Reviewers

  1. Clément Menuet, Institut de Neurobiologie de la Méditerranée, France
  2. Natasha N Kumar, University of New South Wales, Australia
  3. Patrice Guyenet, University of Virginia, School of Medicine

Version history

  1. Received: June 23, 2020
  2. Accepted: May 19, 2021
  3. Accepted Manuscript published: May 20, 2021 (version 1)
  4. Version of Record published: June 1, 2021 (version 2)

Copyright

© 2021, Cleary et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,329
    Page views
  • 170
    Downloads
  • 7
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Colin M Cleary
  2. Brenda M Milla
  3. Fu-Shan Kuo
  4. Shaun James
  5. William F Flynn
  6. Paul Robson
  7. Daniel K Mulkey
(2021)
Somatostatin-expressing parafacial neurons are CO2/H+ sensitive and regulate baseline breathing
eLife 10:e60317.
https://doi.org/10.7554/eLife.60317

Further reading

    1. Neuroscience
    Yang Xu, Tian-liang Cui ... Jin-Hui Wang
    Research Article

    The joint storage and reciprocal retrieval of learnt associated signals are presumably encoded by associative memory cells. In the accumulation and enrichment of memory contents in lifespan, a signal often becomes a core signal associatively shared for other signals. One specific group of associative memory neurons that encode this core signal likely interconnects multiple groups of associative memory neurons that encode these other signals for their joint storage and reciprocal retrieval. We have examined this hypothesis in a mouse model of associative learning by pairing the whisker tactile signal sequentially with the olfactory signal, the gustatory signal, and the tail-heating signal. Mice experienced this associative learning show the whisker fluctuation induced by olfactory, gustatory, and tail-heating signals, or the other way around, that is, memories to multi-modal associated signals featured by their reciprocal retrievals. Barrel cortical neurons in these mice become able to encode olfactory, gustatory, and tail-heating signals alongside the whisker signal. Barrel cortical neurons interconnect piriform, S1-Tr, and gustatory cortical neurons. With the barrel cortex as the hub, the indirect activation occurs among piriform, gustatory, and S1-Tr cortices for the second-order associative memory. These associative memory neurons recruited to encode multi-modal signals in the barrel cortex for associative memory are downregulated by neuroligin-3 knockdown. Thus, associative memory neurons can be recruited as the core cellular substrate to memorize multiple associated signals for the first-order and the second-order of associative memories by neuroligin-3-mediated synapse formation, which constitutes neuronal substrates of cognitive activities in the field of memoriology.

    1. Chromosomes and Gene Expression
    2. Neuroscience
    Alan E Murphy, Nurun Fancy, Nathan Skene
    Research Article

    Mathys et al. conducted the first single-nucleus RNA-seq (snRNA-seq) study of Alzheimer’s disease (AD) (Mathys et al., 2019). With bulk RNA-seq, changes in gene expression across cell types can be lost, potentially masking the differentially expressed genes (DEGs) across different cell types. Through the use of single-cell techniques, the authors benefitted from increased resolution with the potential to uncover cell type-specific DEGs in AD for the first time. However, there were limitations in both their data processing and quality control and their differential expression analysis. Here, we correct these issues and use best-practice approaches to snRNA-seq differential expression, resulting in 549 times fewer DEGs at a false discovery rate of 0.05. Thus, this study highlights the impact of quality control and differential analysis methods on the discovery of disease-associated genes and aims to refocus the AD research field away from spuriously identified genes.