Lateral parabrachial FoxP2 neurons regulate respiratory responses to hypercapnia

Abstract

About half of the neurons in the parabrachial nucleus (PB) that are activated by CO₂ are located in the external lateral (el) subnucleus, express calcitonin gene-related peptide (CGRP), and cause forebrain arousal. We report here, in male mice, that most of the remaining CO₂-responsive neurons in the adjacent central lateral (PBcl) and Kölliker-Fuse (KF) PB subnuclei express the transcription factor FoxP2 and many of these neurons project to respiratory sites in the medulla. PBcl^FoxP2 neurons show increased intracellular calcium during wakefulness and REM sleep and in response to elevated CO₂ during NREM sleep. Photo-activation of the PBcl^FoxP2 neurons increases respiration, whereas either photo-inhibition of PBcl^FoxP2 or genetic deletion of PB/KF^FoxP2 neurons reduces the respiratory response to CO₂ stimulation without preventing awakening. Thus, augmenting the PBcl/KF^FoxP2 response to CO₂ in patients with sleep apnea in combination with inhibition of the PBel^CGRP neurons may avoid hypoventilation and minimize EEG arousals.

Fig. 1. CO₂ activates cFos expression in PBcl^FoxP2 and KF^FoxP2 neurons.

Photomicrographs in columns a–d represent sections at the rostral (level 1, first row), middle (level 2, second row), and more caudal (level 3, third row) portions of the PB labeled immunohistochemically for both the immediate early gene cFos (green) and FoxP2 (red), from a mouse that was exposed to 2 h of either normocapnic room air (columns a, b) or 10% CO₂ (columns c, d). The insets in a, c demarcate the areas that are magnified in b, d. Double-labeling (yellow nuclei) was prominent in the KF and in the central lateral FoxP2 clusters in mice exposed to CO₂ but not those breathing normocapnic air. The arrowheads in b and d point to doubly labeled neurons, while the arrows mark neurons that were only labeled for cFos (green), a large cluster of which represent the CGRP neurons in the PBel (c₂, c₃). The bar graphs in e and f compare the percentage of the cFos cells (mean ± SEM) that also expressed FoxP2 (e), and the percentage of FoxP2 cells (mean ± SEM) that were also labeled for cFos (f) in the PBcl and KF areas, after exposure to 10% CO₂ (hypercapnia, n = 5) or room air (normocapnia, n = 3). All data points for each mouse are also shown in both e and f). The groups were analyzed using a one-way ANOVA, followed by Holms-Sidak method for multiple comparisons, where ***P < 0.001; **P < 0.01. Source data are provided as a source data file. Scale in c₃ = 200 µm; d₃ = 100 µm. KF Kölliker-Fuse PB subnucleus, PBcl central lateral PB subnucleus, PBel external lateral PB subnucleus, scp superior cerebellar peduncle, vsct ventral spinocerebellar tract.

Fig. 2. In vivo Ca_i imaging of PBcl^FoxP2 neurons by fiber photometry during exposure to CO₂.

Adeno-associated virus (AAV) expressing Cre-dependent GCaMP6s was injected into the lateral PB of FoxP2-Cre mice (a), resulting in eutopic GCaMP6s expression (green) in the PB^FoxP2 neurons (red) (b). The track of the implanted optical fiber just dorsal to GCaMP-expressing PBcl^FoxP2 neurons is outlined in (b) and allowed us to record their calcium activity as shown in the schematic in (c). A representative fiber photometry recording d shows the activity profile of the PBcl^FoxP2 neurons (ΔF/F), respiration (R) with simultaneous EEG/EMG signals, and the respiratory rate (RR) and minute ventilation (MV = RR × tidal volume, V_T), both of which increased significantly in each CO₂ trial. The control UV signal (isosbestic at 405 nm) was also recorded. Note that the GCaMP signal increased slowly during CO2 exposure, but more sharply as the animal woke up (abrupt change in EMG and EEG about 25 s after CO₂ onset, marked by red arrow). The ΔF/F from 59 trials from n = 5 mice is depicted in e as a heat map illustrating activity 15 s before and after the onset of CO₂ exposure. The graphs below show the mean ± SEM ΔF/F normalized to the pre-CO₂ values (f), %CO₂ in the plethysmograph (g), RR (h), and tidal volume (V_T, i) across the same trials. Two-way ANOVA compared the changes in ΔF/F, RR, and V_T during CO₂ exposure (f–i) compared to the pre-CO₂ baseline, followed by the Holms-Sidak method for multiple comparisons, where ^***P < 0.001; ^**P < 0.01, and ^*P < 0.05. Scale in b 200 µm (lower left) and 20 µm (lower right). Source data are provided as a source data file. Scale in b top right = 400 µm, bottom left = 200 µm; and bottom right = 40 µm. PBcl central lateral PB subnucleus, PBel external lateral PB subnucleus, scp superior cerebellar peduncle.

Fig. 3. In vivo activity of individual PB^FoxP2 neurons during CO₂ exposure.

GRIN lens was implanted above the injection sites in the PBcl of FoxP2-Cre mice injected with Cre-dependent AAV-GCaMP6s (a) and the calcium activity profiles (ΔF/F) of individual PBcl^Foxp2 neurons in response to the CO₂ were acquired. Representative Ca_i activity of three neurons (b, c), of which two neurons peaked (ΔF/F) at point d, early during the hypercapnia stimulus while activity of the neuron shown in (c) peaked in the later half of hypercapnia stimulus (shown as point e). Two cells (marked by green and yellow arrowheads whose activity profiles are also plotted in green and yellow) shown in (b) had fluorescence that peaked at about 17–19 s after exposure, before the maximal changes in respiration. A third cell (marked by a magenta arrow and activity profile) peaked roughly 50 s after the onset of the CO₂ trial (c), but CO₂ levels were still high. The ΔF/F from nine cells is also plotted during a 30 s trial of 8% CO₂ (with no awakening), shows an overall increase in Ca_i across the population, with most cells showing peaks at the time of maximal respiration (R), but with substantial variability across neurons (f). A heat map of the mean ΔF/F over 4–7 trials (during which animals exposed to 8% CO₂ did not awaken) is shown for all 28 cells in (g) (blue to red: shows low to high ΔF/F). Note that although the RR and MV summated over these trials (mean ± SEM) increased relatively smoothly (h), the activation of the PBcl^FoxP2 neurons occurred in waves, and of the 17 cells that showed three peaks, ten of them showed second and third peaks that were 17–19 s apart and were synchronous in time after CO₂ exposure, but that not every neuron participated in each wave of excitation. Data in (h) is acquired from n = 3 mice. Source data are provided as a source data file. Two-way ANOVA compared the changes in RR and MV post CO₂ to the Pre-CO₂ baseline, followed by the Holms-Sidak method for multiple comparisons, where ***P < 0.001.

Fig. 4. Effect of photoactivation of PB^FoxP2 neurons on respiration during NREM sleep.

FoxP2-Cre mice were injected bilaterally targeting the PBcl with AAV-Flex-ChR2-mCherry (red a, b) and sections were immunostained (green) for FoxP2. Implanted optical fibers targeted ChR2-expressing FoxP2 neurons (b_{1, 2}). Panel b₂ shows the right side of b₂ at higher magnification, b₃–b₅ shows the box in b₂, and the inset shows the box in b₅, Scale in b₁ = 500; b₂–b₅ = 300 µm; inset = 50 µm. In c a representative trial of 20 Hz stimulation in normocapnic air showed gradually increasing respiration (R), with cortical arousal at 4.5 s in this trial. In trials with 5 s stimulation, the animals on average awakened around 15 s after the stimulation stopped (d), suggesting that the awakening was not due to the stimulation itself, but may have been elicited by subsequent respiratory efforts. Trials with stimulation for 10 s usually caused EEG arousal either just before or after the termination of stimulation. Arousal latency was decreased dramatically (by 83%) by exposing the mice to continuous 2% CO₂ during optostimulation (d, shown in a yellow rectangle). A representative trial of stimulation at 20 Hz for 10 s with continuous 2% CO₂ is shown in (e). Graphs in f compare the RR, V_T, and MV during laser stimulation at 5 Hz, 10 Hz, or 20 Hz vs no laser (laser-OFF) for either 5 s or 10 s during normocapnia. Graphs in g compare the RR, V_T, and MV parameters in the mice subjected to laser stimulation at 20 Hz vs no laser (laser-OFF) for 10 s, in mice continuously exposed to 2% CO₂. Values are mean ± SEM for five breaths before the onset of stimulation, during stimulation but before cortical arousal, and in the post-stimulation period after the cortical arousal when stable breathing is attained. Two-way ANOVA was used for statistical comparison, followed by the Holm-sidak test for multiple comparisons, where **P < 0.01 (vs pre-stimulation/laser-OFF); ***P < 0.001 (vs pre-stimulation/laser-OFF), ^#P < 0.05 (vs laser-ON at normocapnia). Source data are provided as a source data file. PBel external lateral PB subnucleus, scp superior cerebellar peduncle.

Fig. 5. Descending projections of the PB^FoxP2 neurons.

Bilateral injections of Cre-dependent AAV-ChR2 (magenta) into the PBcl in (a) FoxP2-Cre mouse (a₁) are shown with immuno-labeled for FoxP2 (green); nearly all of the ChR2-labeled cell bodies are doubly labeled (white) in the PBcl area along the dorsal margin of the PBel (which is unlabeled) (a, magnified view of injection site on right). The bottom right inset a is a magnified view of the area marked by the dashed rectangle. Descending fibers and terminals in the medulla are shown in magenta in (b, c) on the right side of the brain. The magnified views (3×) of the areas in the dashed boxes in both (b, c) are shown as b₁–c₂ to demonstrate that ChR2-mCherry labels only the fibers and does not label any cell bodies in the area. Note that inferior olive neurons (green in b, d) also express FoxP2. Scale in a = 100 µm (inset = 50 µm), in c = 500 µm and c₂ = 150 µm. Data shown in micrographs a–c were consistently observed in n = 6 mice. The photomicrograph in d shows the injection site of the retrograde tracer CTb (0.2%, magenta) in the pre-Bötzinger area (PBZ) (n = 3; scale = 500 µm). This injection retrogradely labeled many FoxP2 neurons in the KF (e₁–e₄) and the PBcl and lateral crescent areas that surround the PBel subnucleus (f₁–g₃). The arrowheads in e₄, f₃, and g₃ mark the cells doubly labeled for FoxP2 and CTb (white), while the arrows mark cells labeled only with CTb (magenta). The insets in f₃ and g₃ are 2× magnified views of the areas encompassed by dashed rectangles, highlighting the double-labeled cells. Scale in e₁ = 100 µm, in e₃ = 60 µm, and in e₄ = 30 µm. Scale in g₁ and g₃ = 100 µm. CVLM caudal ventrolateral medulla, KF Kölliker-Fuse PB subnucleus, NTS nucleus tractus solitarii, PBZ pre-Bötzinger area, PBcl central lateral PB subnucleus, PBel external lateral PB subnucleus, RVLM rostral ventrolateral medulla, scp superior cerebellar peduncle, vsct ventral spinocerebellar tract, XIIn hypoglossal nucleus.

Fig. 6. Genetic ablation of the PB/KF^FoxP2neurons blocks hypercapnia-induced respiratory drive.

AAV-FLEX-DTA injections in PB of the FoxP2-Cre mice and wild-type littermates selectively ablated the FoxP2 neurons in the PBcl and the KF regions of the FoxP2-Cre mice. Photomicrographs (a, b) from a wild-type mouse (with no Cre expression, a WT-DTA) and a FoxP2-Cre mouse b injected with AAV-mCherry-Flex-DTA, immuno-labeled for FoxP2 (green) and mCherry (red), at three levels (rostal, middle, and caudal; a₁–b₃), with magnified merged panels of FoxP2 and mCherry for better view. The photomicrographs show that the spread of the AAV-DTA injection with mCherry labeling covered both KF and PB regions in (a, b). In the first group of mice a that lacked cre-recombinase, AAV-FLEX-DTA did not ablate the KF^FoxP2 and the PB^FoxP2 neurons (green); while in b where the virus injections were in the FoxP2-Cre mice, very few FoxP2 cells in the KF and the PBcl were spared (b PB + KF^FoxP2-DTA). However, both groups showed expression of mCherry in neurons in PBcl and PBel which lacked FoxP2. Scale in a₁–b₃ = 100 µm. KF Kölliker-Fuse PB subnucleus, PBel external lateral PB subnucleus, scp superior cerebellar peduncle, vsct ventral spinocerebellar tract. Genetic ablation of both PB^FoxP2 and KF^FoxP2 neurons significantly reduced the increases in V_T (d) and MV (e) but not RR (c) caused by CO₂ exposure (hypercapnia shown by a yellow rectangle). Values of RR, V_T, and MV are mean ± SEM (n = 8 WT-DTA; PB + KF^FoxP2-DTA = 5). Two-way ANOVA, with *P < 0.05; **P < 0.001, followed by Holms-Sidak method for multiple comparisons, for PB + KF^FoxP2-DTA compared to the WT-DTA. Source data are provided as a source data file.

Fig. 7. Acute optogenetic silencing of the PB^FoxP2neurons blocks hypercapnia-induced respiratory drive.

FoxP2-Cre mice were injected with AAV-FLEX-ArchT (a), and implanted for EEG/EMG and with bilateral optical fibers targeting PBcl^FoxP2 neurons (c₁, c₂). There was virtually complete co-expression of FoxP2 (red) and ArchT (green, c_3–5). The experimental strategy to test respiratory responses to 10% CO₂ given for 30 s every 300 s, with and without photo-inhibition of the PB^FoxP2 neurons is shown as a schematic in (b). Representative examples of a laser-ON and a laser-OFF trial are shown in (d, e), with magnified views of the bounding box from each trial shown in d₁ and e₁. For each stimulation, the laser was switched ON for 60 s beginning 20 s prior and extended for 10 s after the CO₂ stimulus (30 s, illustrated by the yellow box) as shown in (e). Photoinhibition of PB^FoxP2 neurons significantly reduced the increases in V_T (g) and MV (h) but not RR (f) caused by CO₂ exposure, although not as much as ablating both the PB^FoxP2 and KF^FoxP2 neurons. Values of RR (f), V_T (g), and MV (h) are mean ± SEM. Two-way ANOVA, followed by Holms-Sidak method for multiple comparisons, with ^*P < 0.05, compared to the laser-OFF and wild type injected with ArchT (controls). The photo-inhibition of PBcl^FoxP2 neurons did not change the latency to arousal (i mean ± SEM, laser-OFF (n = 5) vs laser-ON (n = 6)). For data shown in f–i, 'n' represents the number of mice. Scale in c₂ and c₅ = 200 µm. Source data are provided as a source data file.

Fig. 8. Proposed neural circuit for mediation of hypercapnia-induced increases in ventilation.

PB neurons receive information about CO₂ levels via relays from the carotid body through the NTS (nucleus of the solitary tract), and directly from chemosensory neurons including the RTN (retrotrapezoid nucleus)^,,,, medullary (Med R), and dorsal (DR) raphe nuclei^,,. The projections of the PB^FoxP2 neurons to the PBZ, rostral ventrolateral medulla (RVLM), and the CVLM may mediate the increase in ventilation by increasing RR^,,, and tidal volume (V_T) or directly by the projections from the KF FoxP2 neurons to the phrenic motor nucleus as previously shown by others^,. Mechanical sensory feedback by thoracic stretch receptors during increased ventilatory effort may activate PBel^CGRP neurons to cause cortical arousal, via activation of neurons in the basal forebrain (BF), central nucleus of the amygdala (CeA), and lateral hypothalamic area (LHA)^,. Selective activation of PB^FoxP2 neurons may be effective for augmenting breathing in apneas if combined with ways to suppress the EEG arousal.