Main

The brain supervises the physiology of our internal organs to ensure bodily homeostasis. Sensory signals from the body provide constant feedback to the brain about the ongoing performance of each major physiological system3. Within the cardiovascular system, specialized sensory neurons report on the pressure, volume and chemical composition of blood to ensure appropriate cardiac output and tissue perfusion1,4,5,6. Neuronal surveillance helps to enable adaptation to environmental stressors such as reduced oxygen during a high-altitude hike or the effect of gravity on blood circulation when standing up. Failure to regulate blood pressure, even for a moment, can lead to dizziness, fainting and injury7. Moreover, cardiovascular signals can drive some of our most basic emotions, as optogenetic induction of a racing heart can cause anxiety8. The heart–brain axis therefore provides essential control over physiology, behaviour and mood.

The vagal and glossopharyngeal nerves provide the dominant sensory innervation of the heart and nearby vasculature and mediate four major cardiovascular reflexes: the baroreceptor reflex, the hypoxic ventilatory response, the Bezold–Jarisch reflex and the Bainbridge reflex1,4,5,6. The baroreceptor reflex detects momentary changes in arterial blood pressure, providing feedback to stabilize cardiovascular output in real time2,9. Baroreceptor terminals are strategically located in the great arteries, at the carotid sinus and aortic arch, and detect stretch of the blood vessel wall with every heartbeat. Knockout of both Piezo1 and Piezo2 in baroreceptor neurons, but not either one alone, eliminates the baroreceptor reflex2. The hypoxic ventilatory response is mediated by arterial chemoreceptors located in the carotid body and aortic arch, which detect low oxygen levels through tissue-resident glomus cells, and trigger a neural arc that increases respiration10. Other receptors in the heart mediate the Bainbridge and Bezold–Jarisch reflexes. The Bezold–Jarisch reflex is a defensive reflex that may be evoked by cytokines, injury-associated chemicals and/or abnormal ventricular contraction, leading to bradycardia and hypotension11. The Bainbridge reflex is a reported mechanical reflex whereby rapid infusion of saline evokes a reflexive heart rate increase through vagal fibres in heart atria12. However, the Bainbridge reflex is weak or non-apparent in some species, including humans13,14, so the physiological roles of cardiac mechanoreceptors have remained obscure.

Recent single-cell RNA-sequencing studies have enabled genetic access to a variety of vagal and glossopharyngeal sensory neuron types15,16. For example, different genetically defined sensory neurons in the respiratory system detect airway stretch17, airway closure18, irritants that cause cough19, and pathogen- and inflammation-induced cytokines20. Other sensory neurons in the digestive system detect nutrients, osmolarity changes, toxins, cytokines, and stretch of the stomach or intestine15,21,22,23. Expression atlases and genetic approaches have also revealed orphan neuron types whose sensory properties and functions are not understood9,16. Neuronal responses to blood volume changes have been difficult to parse because the cardiovascular system is a closed loop and volume changes simultaneously impact sensors at different locations. Here we used genetic tools for selective loss of function to distinguish the roles of different vagal mechanoreceptors in the cardiovascular system. We first focused on sensation of mechanical signals during posture change, which can cause widespread effects on blood volume and blood pressure throughout circulation.

Neuronal compensation to posture change

When we stand up, gravity exerts powerful effects on the cardiovascular system, reducing venous return to the heart as well as blood flow to the head and upper body24. A prolonged reduction in carotid blood pressure upon standing up (called orthostatic hypotension) can cause dizziness and fainting. Several sensory pathways have been proposed to help compensate for the effects of gravity on the cardiovascular system during posture change25,26,27, but their relative importance is unclear.

Orthostatic hypotension is assessed clinically using a tilt-table test28, in which blood pressure and heart rate are measured while patients are rotated from a horizontal to an upright position. Here we adapted the tilt-table test for mice to assess posture-dependent blood pressure changes after genetic manipulation of different sensory pathways. Anaesthetized mice were placed securely onto a platform that allowed for 180 degrees of rotation from supine to upright to prone position. During rotation, arterial blood pressure (carotid and femoral) and heart rate were continuously monitored. Rotation to an upright position caused an immediate drop in arterial blood pressure through gravity that was rapidly corrected for and accompanied by a sustained heart rate increase (Fig. 1a,b and Extended Data Fig. 1a). Conversely, rotation to a supine or prone position caused an increase in arterial blood pressure that was also rapidly corrected for and accompanied by a sustained heart rate decrease. The compensation setpoint was primarily determined by the original carotid blood pressure, as overcompensation of femoral blood pressure was observed during recovery (Extended Data Fig. 1b). Aligning the systemic setpoint with carotid blood pressure presumably prioritizes the constancy of blood flow to the brain. Together, these data indicate that mice, like humans on a tilt table24, have a robust feedback control mechanism that compensates for the effects of gravity on the cardiovascular system.

Fig. 1: PIEZO2 mediates a vagal reflex to posture change.

a, Representative traces of blood pressure (mean carotid) and heart rate while mice are rotated through different postures on a tilt table. bpm, beats per minute. b, The initial and recovered changes in blood pressure after tilt to the indicated position. Initial change, maximal deviation within first 5 s; recovery BP, the average blood pressure from 1 min after tilt to 1 min and 5 s after tilt; baseline, the average blood pressure for the 5 s before tilt. The individual circles show the average response of one mouse over at least three trials. n = 7 mice. c, Cartoons depicting the anatomy of the vestibular system and vagus nerve. ADN, aortic depressor nerve; CSN, carotid sinus nerve. d, Representative traces of blood pressure (mean carotid) while mice are rotated through different postures on the tilt table. e, Representative blood pressure responses (average across at least three tilts) during upright tilt in the indicated mice. Baroreceptors cut, transection of the glossopharyngeal nerve and SLN. The control mice are littermates lacking Phox2b-cre. f, Quantification of blood pressure recovery after upright rotation. BP recovery % = (ΔI − ΔR)/ΔI, where ΔI is the initial blood pressure change from the baseline and ΔR is the recovery blood pressure change from the baseline. Data are mean ± s.e.m. From left to right, n = 7, 5, 5, 4, 4, 4, 5, 6 and 11 mice. KO, knockout; Subdiaph. vagotomy, subdiaphragmatic vagotomy. The individual circles show the average response of one mouse over at least three trials. Statistical analysis was performed using one-way ANOVA with Bonferroni correction for multiple comparisons, compared with the WT; ****P < 0.0001. Mouse diagrams in a created in BioRender. Liu, Z. (2025) https://BioRender.com/y44s465. Mouse diagram in c created in BioRender. Liu, Z. (2025) https://BioRender.com/z28w571. Heart diagram in c created in BioRender. Liu, Z. (2025) https://BioRender.com/a34i357.

Source Data

We next used various surgical and genetic manipulations to define sensory pathways for gravity compensation (Fig. 1c–f and Extended Data Fig. 1c). We first examined a candidate role for the vestibular system in the inner ear, which has an important role in maintaining balance and stabilizing the visual field during motion and position changes. However, mutant mice that lack vestibular sensation (through Tmie29 or Otop130 knockout) displayed normal haemodynamic responses on the tilt table. We next performed surgical nerve transections to examine the roles of different peripheral nerves. Complete cervical transection of both vagal and glossopharyngeal nerves proximal to the ganglia (cervical vagotomy), which eliminates all sensory and motor fibres, abolished compensation to posture change, resulting in sustained hypotension after upright rotation. Vagus nerve transection below the diaphragm (subdiaphragmatic vagotomy) had no effect, while partial responses were observed after transection of (1) both the glossopharyngeal nerve and superior laryngeal nerve (SLN), which eliminates arterial baroreceptors, or (2) the vagus nerve trunk below the SLN departure point (vagal trunk cut), which preserves arterial baroreceptors but eliminates fibres to the heart. The partial effect observed following baroreceptor transection suggested a second residual sensory pathway important for gravity compensation.

PIEZO ion channels and posture change

PIEZOs are mechanosensory ion channels that directly sense force, mediate our sense of gentle touch31, and contribute to internal organ sensations such as airway stretch17, airway closure18, bladder fullness32 and the baroreceptor reflex2,9,33. Global knockout of Piezo2 is lethal17, but targeted knockout of Piezo2 in PHOX2B cells (Phox2b-cre;loxP-Piezo2, hereafter Vagal-ΔPiezo2), which includes petrosal and nodose ganglia, is tolerated and mice can survive to adulthood. We note that Phox2b-cre does not drive detectable reporter expression in intrinsic cardiac neurons, dorsal root ganglia or other intrinsic cells of the heart, while reporter expression was observed in stellate ganglia and brainstem neurons that do not express PIEZO217,34 (Extended Data Fig. 2). These findings are consistent with previous work indicating that the intersection of Piezo2 expression and Phox2b-cre-based reporters is largely or exclusively confined to placode-derived cranial sensory neurons2,17,35,36.

Phox2b-cre-driven knockout of Piezo**2 eliminates the vagal sensation of airway stretch17 but not the baroreceptor reflex2, which instead is lost only after knockout of both Piezo1 and Piezo2 (Phox2b-cre;loxP-Piezo1;loxP-Piezo2; hereafter Vagal-ΔPiezo1/2). Here we confirmed that the baroreceptor reflex is normal in Vagal-ΔPiezo2 mice across a wide range of stimulus conditions (Extended Data Fig. 3). We next analysed both Vagal-ΔPiezo2 and Vagal-ΔPiezo1/2 mice using the tilt-table test. Notably, we observed a significant impairment in gravity compensation after knockout of Piezo2 alone (Fig. 1d–f), even though these mice have a normal baroreceptor reflex, as additionally evidenced by quantitative analysis of baroreflex sensitivity (Δheart rate/Δblood pressure or ΔHR/ΔBP) on the tilt table (Extended Data Fig. 1c). Knockout of both Piezo1 and Piezo2 also caused a response characteristic of orthostatic hypotension. Tilt-table responses were normal in mice lacking another candidate mechanoreceptor, GPR6837. Together, these findings indicate a collaborative role for at least two neuron types in gravity compensation: the baroreceptors and a separate neuron type dependent only on PIEZO2. Moreover, nerve transection experiments suggested that the residual PIEZO2-dependent pathway resides in the vagus nerve trunk below the SLN departure point, consistent with a role for sensory neurons that innervate the heart.

PIEZO2 neurons in the heart

We next examined whether vagal PIEZO2 neurons project to the heart. To map the anatomical projections of Cre-defined vagal sensory neurons, we injected vagal ganglia of cre knock-in mice with an adeno-associated virus (AAV) containing a Cre-dependent tdTomato reporter (AAV-flex-tdTomato)9,38,39. Initial experiments involved Vglut2-ires-cre mice to label all vagal sensory neurons, followed by whole-mount light-sheet imaging of the entire heart (Fig. 2a). We observed extensive innervation across all four cardiac compartments (left and right atria and ventricles), as well as dense innervation of the nearby aortic arch and carotid sinus. Major vagal nerve branches approach and innervate the dorsal side of the heart, with finer nerve fibres ramifying across the entire cardiac surface toward the ventral side. Injection of different reporters in the left and right vagal ganglia showed no obvious difference in heart innervation (Extended Data Fig. 4), although the left and right vagal ganglia accessed different arterial sites with a similar repertoire of terminal types9.

Fig. 2: Anatomy of vagal sensory neurons in the heart.

a, Representative whole-mount images of tdTomato immunofluorescence in the heart (left, dorsal view; right, ventral view) of a Vglut2-ires-cre mouse injected with AAV-flex-tdTomato in the vagal ganglia. Scale bar, 1 mm. LA, left atrium; RA, right atrium. b, Representative confocal images (maximum-intensity projection, two replicates) of a heart from a showing flower-spray terminals (top) and end-net endings (bottom). Scale bars, 200 μm. c, Uniform manifold approximation and projection (UMAP) plots depicting expression (natural log scale) of Piezo2 (top) and Npy2r (bottom) across a previously published vagal/glossopharyngeal cell atlas16. d, Representative confocal images (maximum-intensity projection) of a heart from a Piezo2-ires-cre mouse (top, from eight replicates) or Npy2r-ires-cre mouse (bottom, from six replicates) injected with AAV-flex-tdTomato (magenta) and a Cre-independent AAV-eGFP (green) in the vagal ganglia. Scale bars, 200 μm. e, Quantification of the relative distribution of terminal types observed in each Cre line. Data are mean ± s.e.m. n = 6 (NPY2R) and 9 (PIEZO2) mice. Each dot represents cumulative data from one mouse involving both intact atria and a 1 mm ventricle section. f, Piezo2-ires-cre and Npy2r-ires-cre mice were injected bilaterally in the vagal ganglia with DT (vagalABLATE) or PBS (control) and the BP recovery was quantified after upright rotation across conditions. Data are mean ± s.e.m. The individual circles show the average response of one mouse over at least three trials. From left to right, n = 8, 4, 5 and 6 mice. Statistical analysis was performed using a two-sided Mann–Whitney U-test; **P = 0.004.

Source Data

We observed two morphologically distinct types of vagal terminals within the heart: large arrays of interconnected end-net endings and complex arborizations called flower sprays, consistent with previous reports39,40,41 (Fig. 2b). It has been speculated that these endings may have different sensory functions, but the lack of selective genetic tools has hindered investigation.

Anatomical mapping experiments were similarly performed using Cre lines that mark smaller groups of vagal sensory neurons. We used Piezo2-ires-cre and Npy2r-ires-cre mice, which mark predominantly discrete populations of sensory neurons according to vagal cell atlases16,38 (Fig. 2c). PIEZO2 neurons densely innervated the heart, and exclusively formed end-net endings; PIEZO2-containing flower-spray terminals were not readily observed (Fig. 2d,e). In the atrium, PIEZO2 terminals were enriched near the junction with the vena cava (Extended Data Fig. 5). Vagal NPY2R neurons represent a substantial fraction of vagal sensory neurons16, and were previously shown to form a variety of terminals throughout the body, including free endings in the lung38 and intraganglionic laminar endings in the stomach and intestinal wall23. NPY2R neurons evoke rapid and shallow breathing, and also elicit cardiovascular responses underlying the Bezold–Jarisch reflex11,38. Here we observed that NPY2R neurons mark a large cohort of heart-innervating sensory neurons, including separate neurons that form end-net endings and flower sprays (Fig. 2d,e); we note that a previous report described some NPY2R terminals as varicose surface endings or ventricular intramuscular arrays39, and we have summarized them both as end-net endings due to structural differences between atrial and ventricular muscles. NPY2R-containing end-net terminals were 4.0-fold more abundant than end-net endings containing PIEZO2, and were more distributed across the heart atria and ventricles (Extended Data Fig. 5). These findings suggest that there are probably at least three types of vagal neurons in the heart: NPY2R-containing flower sprays, NPY2R-containing end-net endings and PIEZO2-containing end-net endings.

Next, we examined which terminal type might contribute to cardiovascular compensation during the tilt-table test. We ablated vagal PIEZO2 and NPY2R neurons using a previously established approach involving diphtheria toxin (DT)9,20. In brief, the DT receptor (DTR) was expressed in PIEZO2-expressing cells (Piezo2-ires-cre;lsl-DTR) or NPY2R-expressing cells (Npy2r-ires-cre;lsl-DTR), and vagal ganglia were then injected bilaterally with DT to achieve efficient and selective neuronal ablation; the resulting mice are termed PIEZO2-vagalABLATE and NPY2R-vagalABLATE mice. Note that the ablation of PIEZO2 neurons also eliminates baroreceptors, as reported previously9, presumably because PIEZO1 and PIEZO2 are co-expressed in baroreceptor neurons. Ablation of vagal PIEZO2 neurons reduced neuronal compensation during posture change, while ablation of NPY2R neurons had no effect (Fig. 2f and Extended Data Fig. 6). The different phenotypes of NPY2R-vagalABLATE and PIEZO2-vagalABLATE mice suggest that end-net endings marked by PIEZO2 and NPY2R are functionally different, with only PIEZO2 endings required for normal posture compensation.

Heartbeat-coupled vagal responses

We investigated the response properties of vagal PIEZO2 fibres in the heart to understand how their activity patterns might vary with posture. Classical studies have reported that some myelinated vagal sensory neurons fire at particular phases of the cardiac cycle42,43,44; we therefore performed electrophysiological recordings of the mouse vagus nerve while simultaneously recording cardiac activity by electrocardiogram. In some animals, we additionally measured right atrial and left ventricular pressure as well as left ventricular volume through sensors introduced by a cardiac catheter. We excluded contributions from arterial baroreceptors, which display heartbeat-coupled responses with each arterial pressure pulse42, by recording from the sensory end of the vagus nerve trunk below the SLN departure point (Fig. 3a). We also excluded contributions from airway stretch receptors by subtraction of breathing-coupled responses (Methods).

Fig. 3: Vagal mechanosensors fire during atrial and ventricular systole and report on blood volume.

a, Cartoon of electrophysiological recordings of the thoracic vagus nerve trunk. Heart diagram created in BioRender. Liu, Z. (2025) https://BioRender.com/a34i357. Scissors diagram created in BioRender. Liu, Z. (2025) https://BioRender.com/s08uo8h. b, Representative, simultaneously recorded electrocardiograms (ECG) and vagal activity from WT (top) and vagal-ΔPiezo2 (bottom) mice at the baseline (left, raw activity) and after intravenous injection (black arrow) of serotonin (right, integrated activity over a 5 s window). c, Representative spike raster of vagal activity (the black dashes show action potentials) across 159 cardiac cycles in a WT mouse. d, The average physiological measurements over 30 s (around 250 cardiac cycles) in four WT mice. RAP, right atrial pressure; LVP, left ventricular pressure; LVV, left ventricular volume. e, Representative spike histograms of average vagal activity patterns (30 s). f, Quantification of phase I (left) and phase II (right) responses from e. Data are mean ± s.e.m. The individual circles show the average response of one mouse over at least 100 heartbeats. From left to right, n = 12, 6, 6, 7, 6 and 5 mice. g, Vagal nerve responses recorded in WT mice during slow withdrawal of blood (500 μl over 3 min). Right, representative responses (averaged from 10 s) after depletion of the volumes indicated, or reinfusion to the baseline volume (the x axis shows the timeline normalized to cardiac cycle; Methods). Left, quantitative analysis of average phase I (red) and phase II (blue) responses across a range of depletion volumes. Data are mean (dark lines) ± s.e.m. (shading). n = 4. Mouse diagram created in BioRender. Liu, Z. (2025) https://BioRender.com/k0bf0ej. h, Vagal nerve spike counts (left) over 10 s after the volume change indicated, normalized to the serotonin response. Data are mean ± s.e.m. n = 5, 5 and 3 (blood withdrawal); and 4, 3 and 3 (saline infusion). The slope of vagus nerve activity to changing blood volume normalized to body weight (right). Data are mean ± s.e.m. n = 12, 6, 9, 6, 6 and 7 (blood withdrawal); 9, 5, 6, 4, 5 and 7 (saline infusion). For f and** h**, statistical analysis was performed using one-way ANOVA with Bonferroni correction for multiple comparisons, compared with the WT; from left to right, *P = 0.01, ***P = 0.0007, **P = 0.0015, ***P *= 0.0044, ***P *= 0.0029 and ***P *= 0.005 (f), and ****P < 0.0001, ****P *= 0.0006, ****P *= 0.0002, ***P *= 0.0026, **P *= 0.0104 and ***P *= 0.0014 (h).

Source Data

In wild-type (WT) mice, we observed that the vagus nerve trunk fired with every heartbeat at two specific phases of the cardiac cycle, just after the P wave and after the QRS complex in the electrocardiogram (Fig. 3b–f). The first peak (phase I) was synchronous with atrial systole, matching Paintal type A atrial receptors43,45, while the second peak (phase II) was synchronous with ventricular systole44. We did not observe prominent activity during the V wave of atrial pressure, which occurs during atrial filling and would have corresponded to Paintal type B fibres5,43.

We next examined whether and how PIEZO2 neurons might contribute to heartbeat-coupled vagal responses by measuring vagal responses across the cardiac cycle after targeted neuron ablation. Phase I and phase II responses were intact in NPY2R-vagalABLATE mice, but were lost in PIEZO2-vagalABLATE mice (Fig. 3e,f). Paired with anatomical data in Fig. 2, these findings demonstrate that vagal mechanoreceptors responsible for physiological responses during the cardiac cycle form end-net terminals and not flower sprays.

PIEZO2 marks about a third of vagal sensory neurons38, including different cell types that express P2ry1, Pvalb, Oxtr or Glp1r. We ablated subsets of vagal PIEZO2 neurons marked in P2ry1-ires-cre, Pvalb-t2a-cre, Oxtr-t2a-cre and Glp1r-ires-cre mice using genetic ablation approaches involving DT, but heartbeat-coupled responses remained partly (P2RY1) or completely (PVALB, OXTR, GLP1R) intact (Extended Data Fig. 7). These studies provide exclusionary data that help to refine which transcriptome-defined PIEZO2 neurons fire across the cardiac cycle.

To examine a role for PIEZO proteins themselves, we analysed heartbeat-coupled vagal responses in Vagal-ΔPiezo1 (Phox2b-cre;loxP-Piezo1), Vagal-ΔPiezo2 and Vagal-ΔPiezo1/2 mice (Fig. 3b–f). Heartbeat-coupled vagal trunk responses were normal after deletion of Piezo1 alone, but were lost after deletion of Piezo2 or Piezo2 together with Piezo1. These findings indicate a key difference between mechanoreceptors activated during cardiac contractions, which require only PIEZO2, and arterial baroreceptors, which require either PIEZO1 or PIEZO22.

PIEZO2 neurons measure blood volume

Baroreceptors fire with every heartbeat42, and the intensity of activity informs about changing blood pressure46. Given the key role for vagal PIEZO2 neurons in posture responses, we reasoned that the intensity of cardiac mechanoreceptor activity across the cardiac cycle might vary with posture. Electrophysiological recordings on the tilt table were not technically achievable; we therefore measured how key cardiovascular parameters changed with posture to examine how such parameters may influence phase I and phase II nerve responses.

We obtained a pressure–volume (PV) loop of WT mice on the tilt table. Rotation to an upright position caused a transient decrease in systolic pressure in the left ventricle that returned to the baseline, similar to changes in arterial pressure (Extended Data Fig. 8a). However, the stroke volume of the left ventricle was reduced to 70% and did not recover, similar to human clinical data24 and consistent with a sustained decrease in venous return due to gravitational pull. These findings raised the possibility that central blood volume might be a key variable that influences the firing of cardiac mechanoreceptors.

To test this idea, we measured vagal trunk responses during the cardiac cycle while changing circulating blood volume by withdrawing blood or adding saline across a broad physiological range (Fig. 3g). Integrated vagal nerve activity decreased as blood was withdrawn and increased as saline was introduced, and the magnitude of the nerve response varied with the extent of blood volume change. Smaller volumes of blood withdrawal impacted phase I nerve responses more substantially than phase II nerve responses, suggesting differential tuning of atrial and ventricular mechanoreceptors to blood loss. We note that, in some species, increasing blood volume evokes the Bainbridge reflex13, but we did not observe a heart rate increase after saline infusion (Extended Data Fig. 8b). Knockout of Piezo2 or ablation of PIEZO2 neurons caused a striking reduction in heartbeat-coupled responses, and this low activity level was not changed further by volume manipulations (Fig. 3h). Knockout of Piezo1 alone or ablation of NPY2R had no effect on the volume-dependent modulation of heartbeat-associated vagal responses. These findings indicate that the magnitude of PIEZO2 neuron activity during atrial and ventricular systole provides information on circulating blood volume.

We used vagal ganglion calcium imaging to understand the cellular relationship between blood volume sensors and other PIEZO2 neurons in the lung that respond to airway stretch. Real-time calcium transients were recorded after airway distension or intravenous saline infusion in individual vagal sensory neurons of Vglut2-cre;lsl-SALSA mice, which express a GCaMP6f–tdTomato fusion protein in all vagal sensory neurons, or Snap25-GCamp6s;Piezo2-ires-cre;lsl-tdTomato mice in which PIEZO2+ and PIEZO2− neurons can be imaged in parallel (Fig. 4a). We note that calcium imaging does not provide sufficient temporal resolution to distinguish the timing of responses across the cardiac cycle, but does resolve the responses of single cells. Small groups of vagal neurons responded (Fig. 4b,c) to either airway stretch (92 out of 1,842 neurons, 5.0%) or venous saline infusion (116 out of 1,842 neurons, 6.3%), and these populations were largely non-overlapping (10 out of 1,842 neurons (0.5%) responded to both). The majority of neurons responsive to each stimulus expressed PIEZO2 (Fig. 4d and Extended Data Fig. 9; airway stretch: 28 out of 37; venous saline infusion: 21 out of 28). Previous work showed that some PIEZO2− neurons receive mechanosensory signals from neuroepithelial bodies; these neurons are not relevant for the Hering–Breuer reflex or breathing-coupled responses, mediate responses to airway closure and display high-threshold and rapidly adapting responses to large distensions of the conducting airways18. In electrophysiological experiments, PIEZO2− neurons do not respond with every heartbeat, so the responses observed by calcium imaging in PIEZO2 negative neurons may represent another sensory pathway or be due to a secondary physiological change. Taken together, these findings indicate that separate cohorts of PIEZO2 neurons mediate the major responses to airway stretch and blood volume infusion.

Fig. 4: Different PIEZO2 neurons detect changes in blood and airway volume.

a, Cartoon depicting vagal ganglia imaging (left) and a two-photon image (right) of SALSA fluorescence in a representative vagal ganglion (of six replicates) from Vglut2-cre;lsl-SALSA mice. Scale bar, 50 µm. Heart image created in BioRender. Liu, Z. (2025) https://BioRender.com/a34i357. Microscope objective image created in BioRender. Liu, Z. (2025) https://BioRender.com/quy3reu. b, He