Missing pieces of the Piezo1/Piezo2 baroreceptor hypothesis: an autonomic perspective

Abstract

Baroreceptors play a pivotal role in the regulation of blood pressure through moment to moment sensing of arterial blood pressure and providing information to the central nervous system to make autonomic adjustments to maintain appropriate tissue perfusion. A recent publication by Zeng and colleagues (Zeng WZ, Marshall KL, Min S, Daou I, Chapleau MW, Abboud FM, Liberles SD, Science 362: 464–467, 2018) suggests the mechanosensitive ion channels Piezo1 and Piezo2 represent the cellular mechanism by which baroreceptor nerve endings sense changes in arterial blood pressure. However, before Piezo1 and Piezo2 are accepted as the sensor of baroreceptors, the question must be asked of what criteria are necessary to establish this and how well the report of Zeng and colleagues (Zeng WZ, Marshall KL, Min S, Daou I, Chapleau MW, Abboud FM, Liberles SD, Science 362: 464–467, 2018) satisfies these criteria. We briefly review baroreceptor function, outline criteria that a putative neuronal sensor of blood pressure must satisfy, and discuss whether the recent findings of Zeng and colleagues suitably meet these criteria. Despite the provocative hypothesis, there are significant concerns regarding the evidence supporting a role of Piezo1/Piezo2 in arterial baroreceptor function.

INTRODUCTION

The baroreceptor reflex plays a key role in cardiovascular regulation, allowing for stabilization of arterial blood pressure and therefore maintains appropriate tissue blood flow. Key questions in recent years have been how baroreceptors function and what is the nature of the sensory receptors monitoring arterial pressure. The recent discovery of Piezo channels (Murthy et al. 2017), mechanosensitive channels present in a variety of nerve endings, led to the hypothesis that Piezo channels underlie barosensation. Zeng et al. (2018) recently reported that deletion of both Piezo1 and Piezo2 in mice produces a complete absence of barosensation and baroreceptor reflex function. This report was met with considerable excitement (Allison 2019; Burke et al. 2019; Ehmke 2018; Miglis and Muppidi 2019). However, before Piezo1 and Piezo2 are accepted as the baroreceptor channel, we must establish the necessary criteria to evaluate putative baroreceptor mechanisms and then assess whether the evidence of Zeng et al. (2018) satisfies these criteria. Thus, despite the appeal of this hypothesis, it is important to carefully and fully evaluate the evidence provided to support this claim. As we detail below, the evidence is severely lacking, and it is therefore very premature to declare that these Piezo channels are responsible for barosensation.

BRIEF OVERVIEW OF ARTERIAL BARORECEPTOR REFLEX FUNCTION

Arterial baroreceptors are sensory neurons with cell bodies in cervical ganglia (nodose, jugular, and petrosal) and mechanosensitive sensory endings embedded in thoracic artery walls, chiefly in the aortic arch and the cervical artery walls including the carotid sinus (for reviews, see Andresen and Kunze 1994; Kumada et al. 1990). These neurons project to the hindbrain with contacts concentrated in the caudal portions of the solitary tract nucleus. Distension of the vessel wall stretches these baroreceptor endings to activate action potential volleys in proportion to each change in blood pressure. It is important to note that arterial pressure is changing throughout the cardiac cycle and therefore the action potential rate in baroreceptor afferent axons increases and decreases as blood pressure increases and decreases during the cardiac cycle, resulting in pulse-synchronous fluctuations in action potential frequency (Fig. 1). Each baroreceptor fiber has a characteristic pressure (threshold) below which the baroreceptors are inactive and a high pressure above which the action potential frequency reaches a maximum. Thus the encoding range of the population of individual arterial baroreceptors generally coincides with physiological range of pressures from 40 to 180 mmHg. Increases in baroreceptor afferent activity reduce total peripheral resistance through inhibition of sympathetic nerve activity (arteriolar vasodilation in many vascular beds) and reduction in cardiac output (through reducing both heart rate and stroke volume) via decreased sympathetic and increased parasympathetic input (Fig. 1); decreases in baroreceptor afferent firing promote vasoconstriction and increased cardiac output. These reflex responses initiated by carotid and aortic baroreceptors are rapid, tracking almost instantaneously to changes in arterial blood pressure.

Fig. 1.

A: raw trace of arterial blood pressure (ABP), mean ABP (gray line), heart rate, original and rectified/integrated aortic depressor nerve (ADN) activity, and original and rectified/integrated lumbar sympathetic nerve activity (SNA) of a urethane-anesthetized Male Sprague-Dawley rat (390 g) during intravenous infusion of sodium nitroprusside and phenylephrine. Note, ADN activity was directly related to the level of ABP and changed rapidly to any perturbation in ABP. Lumbar SNA was inversely related to ABP. The rat was artificially-ventilated to maintain end-tidal CO₂ at 3.5–4.5%, and body temperature was maintained at 37 ± 0.2°C (for detailed methods, see Simmonds et al. 2014). B: 0.6-s examples of ABP, ADN, and lumbar SNA during 3 different time points highlighted by gray bars: i: note, the pulse-synchronous ADN activity time locked to the upstroke of the ABP pulse; ii: absence of ADN activity at low ABP; and iii: robust ADN activity time locked to the ABP at high ABP.

Baroreceptor neurons are a heterogeneous population of myelinated and unmyelinated fibers (for reviews, see Andresen and Kunze 1994; Kumada et al. 1990; Schild and Kunze 2012). A-type baroreceptors are myelinated and have a lower action potential threshold than C fibers. A fibers exhibit changes in action potential frequency that are directly proportional to the blood pressure change over a wide range of pressures. Even within a cardiac cycle, A fibers display pulse-synchronous discharge. Thus A fibers display a regular discharge pattern and reliably encode changes in blood pressure. On the other hand, C fibers are unmyelinated and greatly outnumber A fibers (as much as 10:1). Higher pressures are required to activate C fibers versus A fibers. For the same blood pressure, the discharge frequency is lower for C versus A fibers. In addition, most C fibers exhibit one or two spikes that are loosely correlated to the arterial pressure waveform. Maximum discharge frequency is also lower for C fibers (~20–30 Hz) versus A fibers (100 Hz). More detailed electrophysiological analyses have revealed A and C fibers have different cellular properties and express different voltage-gated ion channels (i.e., tetrodotoxin resistant sodium current in C but not A fibers). Moreover, the thresholds and sensitivity of A and C fibers are modified by changes in vessel wall tone, ion concentrations, and various hormone and paracrine modulators.

Despite these differences in baroreceptor nerve types, the functional role of A versus C fibers in baroreceptor reflexes remains unclear. Selective activation of A fibers in the aortic depressor nerve (ADN) using low-voltage intensities requires stimulation frequencies >10 Hz to decrease arterial blood pressure (Fan and Andresen 1998). In contrast, activation of both A and C fibers in the ADN using high-voltage intensities decreases blood pressure at 2–10 Hz (Fan and Andresen 1998). However, the absence of techniques to selectively ablate one fiber type versus another represents a major barrier to attribute specific role of A versus C fibers in baroreceptor function or baroreceptor reflexes.

Surgical denervation of arterial baroreceptors by cutting the ADN and carotid sinus nerve (CSN) bilaterally in experimental animals results in an initial increase in arterial blood pressure, but chronically the most apparent cardiovascular perturbation is a marked increase in the lability of arterial pressure (Sved et al. 1997). This large increase in arterial blood pressure variability is often unaccompanied by overall changes in mean arterial pressure, although some studies find an increase in blood pressure as well as increased lability. It has been suggested that the complete arterial baroreceptor denervation results in chronically normal blood pressure whereas partial denervation results in an increase in blood pressure (Schreihofer and Sved 1994; Sved et al. 1997). Due to the heterogeneity of baroreceptor fiber types and lack of approaches to selectively manipulate baroreceptor fiber subtypes, the impact of selective denervation or removal of A versus C fibers (or vice versa) on blood pressure or lability remains unclear. Either way, the key finding is that baroreceptor denervation markedly increases arterial blood pressure variability.

It is noteworthy that that aortic and carotid sinus arterial baroreceptors are not the only mechanoreceptors associated with the cardiovascular system. In particular, other stretch receptive sensory neurons with endings in the cardiopulmonary system transduce stretch of the atria, ventricles, the veno-atrial junctions (vena cava and right atrium or pulmonary vein and left atrium). Note that stretch of many airways also evokes cardiovascular reflex changes. Activation of mechanoreceptors in other peripheral tissues (i.e., kidney, muscle) alters hemodynamics as well.

WHAT EVIDENCE IS NECESSARY TO IDENTIFY A BARORECEPTOR MECHANISM?

To evaluate a putative arterial baroreceptor mechanism, it would be useful to have a set of criteria that a candidate mechanism would need to fulfil. Here we suggest a list of criteria. We realize that most these criteria require knockout (KO) of the candidate mechanism in mammals, which raises the additional issue that most of what we know about what an animal lacking baroreceptors is based on studies of baroreceptor function in adult organisms. Developmental KOs, where baroreceptor function is absent throughout development, may present differently as a result of developmental adaptations. Thus some of these criteria may be better studied using conditional KOs.

Criterion 1

Direct anatomical verification requires that the stretch-sensitive channel is expressed in the appropriate neurons of the nodose and petrosal ganglia and be present in the sensory nerve endings in the aortic arch and carotid sinus. The transcript should be present in neurons anatomically labeled from the aortic arch or carotid sinus. The channel protein should also be localized to the sensory endings in the aortic arch and carotid sinus. Although additional experiments may localize the protein or channel to A-type myelinated versus C-type unmyelinated fibers, a specific role of one type versus another remains elusive. While such data may represent an advance for the field and allow for a detailed investigation, localization of the channel to a specific fiber type is not necessary.

Criterion 2

Functional verification requires that baroreceptive activity be modified in a manner that is consistent with known properties of arterial baroreceptors. In vivo whole nerve recordings of ADN and CSN activity (both A and C fibers) must show the absence (or reduction) of baroreceptor activity during experimental manipulations of blood pressure after pharmacological blockade, knockdown, or genetic deletion of the putative channel. Typical ADN or CSN in vivo recordings of control animals exhibit 1) pulse synchronous ADN or CSN activity that is time locked to the upstroke of the arterial blood pressure pulse at a resting arterial blood pressure (~90–110 mmHg, Fig. 1Bi), and 2) rapid and parallel changes in ADN or CSN activity to pharmacologically evoked decreases and increases in arterial blood pressure ranging from 60 to 160 mmHg (Fig. 1). Absolute blood pressure values must be reported to properly assess the level of ADN or CSN activity. Proportionate responses or baroreflex gain (change afferent nerve activity/change in blood pressure) should be attenuated or abolished in animals in which the channel is blocked, knocked down, and deleted. It is noteworthy that established properties of baroreceptors indicate the discharge rate of baroreceptor fibers will change in response to discrete changes in arterial blood pressure of <10 mmHg depending on fiber type (for reviews, see Andresen and Kunze 1994; Kumada et al. 1990) (Fig. 1). However, control experiments for mechanotransduction specificity must test whether the ADN and/or CSN remain viable by demonstrating action potential discharge via electrically evoked depolarization. Pragmatically, such variables are most often measured in anesthetized preparation.

Criterion 3

Baroreceptor reflex responses must be attenuated or absent if the putative channel is blocked, knocked down, or deleted. Typically, changes in heart rate and sympathetic nerve activity are measured in response to pharmacologically evoked increases (e.g., phenylephrine) or decreases (e.g., nitroprusside) in arterial blood pressure (Fazan et al. 2005; Ma et al. 2002, 2003; Rodrigues et al. 2011) (Fig. 1). Commonly, a modified Oxford technique (Smyth et al. 1969) is employed using similar pharmacological manipulations to produce ramp-like changes in blood pressure and assess reflex responses over the full operating range (Fig. 1). Baroreflex gain is calculated by the ratio of the reflex response to the change in blood pressure for both increases and decreases in blood pressure. Alternatively, blood pressure and reflex responses are binned per unit time and then used to generate baroreflex curves fit to a sigmoidal function (Fazan et al. 2005; Ma et al. 2002, 2003; McDowall and Dampney 2006; Rodrigues et al. 2011; Salman et al. 2014). The advantage of the latter approach is that baroreflex function or gain can be assessed over a range of blood pressure (70–140 mmHg) including what is likely most reflective of physiological values. Although sympathetic nerve recordings usually require anesthetized preparations, changes in heart rate (which reflect both sympathetic and parasympathetic drive) can be readily assessed in awake animals. An important control experiment is to directly activate the ADN to confirm that the sensory signal is absent rather than an alteration in the central processing of the baroreceptor afferent input.

Criterion 4

Increased blood pressure lability in awake animals in which the putative channel is blocked, knocked down, or deleted. In many studies across a variety of species, the hallmark cardiovascular phenotype of a baroreceptor denervated animal is marked lability of arterial pressure (Fazan et al. 2005; Rodrigues et al. 2011; Sved et al. 1997). Blood pressure should be measured in awake animals via telemetry or a chronically implanted catheter. Lability is typically calculated by the standard deviation of arterial blood pressure. Whether this is accompanied by increased arterial pressure is debatable and thus should not be used as a criterion.

DOES THE PRESENT EVIDENCE SATISFY THE ABOVE CRITERIA FOR PIEZO1/PIEZO2 AS THE BARORECEPTOR MECHANISM?

Criterion 1: Are Piezo1/Piezo2 Present in Baroreceptor Afferent Nerves?

Zeng et al. (2018) utilized RNAscope to assess Piezo1 and/or Piezo2 transcript expression in the nodose and petrosal ganglion. Although both transcripts were present, a very small number of neurons retrogradely labeled from the carotid sinus after injection of cholera toxin subunit-B were Piezo1 (6%, 6/95) or Piezo2 (8%, 8/95) positive. Cholera toxin subunit-B will preferentially label myelinated fibers. There are two general concerns. First, no method was employed to verify Piezo1 and/or Piezo2 expression at the terminal ending, but this may reflect concerns of antibody specificity for either channel. Second, given the small number of Piezo1- and Piezo2-positive neurons, additional experiments should have colocalized the transcripts with a marker for a barosensitive cell. An important confounder is that Piezo2-positive neurons in the nodose are associated with airway stretch receptors (Nonomura et al. 2017). Therefore, transcript expression in ganglion cells is not unique to baroreceptors. Thus the data partially satisfy criterion 1.

Criterion 2: Does Deletion of Piezo1/Piezo2 Attenuate Baroreceptor Activity?

Zeng et al. (2018) assessed ADN activity of wild-type and double-KO mice with deletion of both Piezo1 and Piezo2 during pharmacologic-evoked changes in blood pressure. Double-KO mice were used as deletion of either channel individually did not alter baroreceptor function. Piezo1^f/fPiezo2^f/f were crossed to Phox2bCre lines (a separate important issue discussed below). Original traces and summary data are presented in Fig. 2 of Zeng et al. (2018). Unfortunately, there are multiple serious concerns regarding the data and responses reported.

1. Figure 2 of Zeng et al. (2018) illustrates original ADN activity, rectified/integrated ADN activity, and/or blood pressure of wild-type and KO animals. First, baseline blood pressure is not provided for these experiments. Therefore, baseline ADN activity or baroreflex responses cannot be assessed in a meaningful way. For example, no ADN activity of wild-type mice at a resting blood pressure of 100 mmHg has a very different implication versus no ADN activity at a resting blood pressure of 60 mmHg (the former should have obvious pulse-synchronous activity, but the latter should have little if any). Second, the signal-to-noise ratio of the original ADN activity in Fig. 2G of Zeng et al. (2018) is quite low with no obvious activity at resting blood pressure time locked to the upstroke of the blood pressure pulse. Most importantly, the baseline activity of the ADN in Fig. 2G of Zeng et al. (2018) occurs at the peak systolic or shortly thereafter and, therefore, inappropriately relates to the cardiac cycle. It is quite possible that this “baseline activity” represents a mechanical artifact from the pulsations of the adjacent carotid artery. When blood pressure is increased with phenylephrine, the ADN recording shows two peaks, which is inconsistent known function or the example illustrated here (see Fig. 1). Lastly, the pulse pressure (systolic minus diastolic) presented in Fig. 2G of Zeng et al. (2018) is ~1–2 mmHg. This unphysiologically low pulse pressure is present in multiple panels and raises concern regarding the validity of the blood pressure measurements.

   The authors report that ADN activity of double KO mice was not present at baseline and did not change in response to phenylephrine-evoked increases in arterial blood pressure. However, no control experiment was performed to assess nerve viability, the same artifacts are present in the original ADN recording at the incorrect phase of the cardiac cycle, and resting blood pressure was not reported.

2. Figure 2E of Zeng et al. (2018) illustrates changes in ADN activity during increases in blood pressure. Again, blood pressure values were not reported. Although a bolus injection of phenylephrine was used to manipulate pressure, ADN activity did not change rapidly and parallel to the changes in blood pressure of wild-type mice; that is, small (a few mmHg) to robust changes in blood pressure (15–20 mmHg) did not rapidly change ADN activity. This directly contradicts current knowledge of functional baroreceptors and ADN activity (see Fig. 1). However, ADN activity did change after a significant delay, once blood pressure increased. It is noteworthy that a decrease in blood pressure from baseline values was not tested. Baroreflex sensitivity or gain was not calculated for the ADN responses over a functional range of baroreceptors (70–140 mmHg).

   A troublesome characteristic of Fig. 2E (Zeng et al. 2018) is the delay in ADN activity relative to the increase in blood pressure. Rather than ADN activity rapidly changing to the change in blood pressure, ADN activity did not increase until ~2 min after blood pressure rose >30 mmHg. ADN recordings performed in multiple species have documented ADN activity responds immediately to <10 mmHg (Fig. 1) (Andresen and Kunze 1994; Kumada et al. 1990; Ma et al. 2002, 2003; Salman et al. 2014). One plausible explanation for such responses is that the recording was performed from the vagus nerve and reflects increased parasympathetic drive to the heart after the increase in blood pressure. A second plausible explanation is that baseline blood pressure was very low (<50 mmHg). This might explain why blood pressure increased >50 mmHg before changes in ADN activity were observed. Unfortunately, ADN responses were not plotted as a function of arterial blood pressure nor were blood pressure values reported for such experiments.

   In summary, Zeng et al. (2018) designed experiments to address this criterion, but the execution (quality of recordings, low pulse pressure, ADN recording artifacts, delayed changes in ADN activity, and no reporting of blood pressure values) raise multiple doubts about the experimental evidence and fall short to satisfy criterion 2.

Criterion 3: Does Deletion of Piezo1/Piezo2 Disrupt Baroreflex Changes in Heart Rate and Sympathetic Nerve Activity?

Zeng et al. (2018) measured heart rate responses to phenylephrine-induced increases in blood pressure. Figures 2, A–D, of Zeng et al. (2018) illustrate a substantial reduction of the baroreflex bradycardia in response to large changes in blood pressure (>40 mmHg). It is noteworthy that baroreflex gain was not calculated within a physiological range of blood pressure (~15-mmHg changes from baseline levels) nor were heart rate responses tested to a decrease in blood pressure. Again, baseline blood pressure or heart rate values were not reported. This is critical as differences in resting blood pressure (or heart rate) will impact whether the animals are responsive (or could sense) the blood pressure perturbations.

A critical control experiment is missing from the data set in Fig. 2 of Zeng et al. (2018). As noted above, the ADN should be directly activated to ensure any deficit in baroreflex responses can be attributed to the channel at the terminal ending (versus central processing of ADN input). In summary, Zeng et al. (2018) did not satisfy criterion 3 to support a role for Piezo1/Piezo2 due to the extremely large increases in blood pressure to evoke reflex responses in control animals, the absence of experiments to test responses to decreases in blood pressure, the absence of direct ADN stimulation in KO mice to ensure deficits were not due to alterations in central processing of ADN input, and the absence of data on baseline values needed to evaluate the results.

Criterion 4: Does Deletion of Piezo1/Piezo2 Increase Blood Pressure Lability?

Zeng et al. (2018) measured blood pressure and lability of wild-type and KO mice using telemetry. Figure 3 of Zeng et al. (2018) reports a slightly elevated blood pressure and increased standard deviation of systolic blood pressure. However, a curious observation is the frequency histogram reported in Fig. 3D of Zeng et al. (2018). The histogram reveals a slight right-shifted distribution in the KO versus wild-type mice with a few blood pressure values of the KO mice at very low or high blood pressures. It is noteworthy that surgical denervation of baroreceptors produces a markedly different frequency histogram characterized by a much broader range of blood pressures and an absence of a defined peak in the frequency distribution at the resting blood pressure. Although the methods employed are appropriate to satisfy criterion 4, there are clear differences regarding the impact of Piezo1/Piezo2 deletion versus surgical denervation of baroreceptors. Unfortunately, such differences were not discussed.

The heterogeneity of baroreceptor fibers (both A and C fibers) provides an additional level of complexity if Piezo channels are selectively expressed in a subset of baroreceptor fibers. Such differences may explain the apparent differences between Piezo1/Piezo2 KO versus surgical denervation of baroreceptors.

ADDITIONAL CONSIDERATIONS FOR THE PIEZO1/PIEZO2 HYPOTHESIS

There are a few additional considerations regarding the findings of Zeng et al. (2018) that merit discussion. First, the KO mice were generated by crosses to a Phox2bCre line. Phox2bCre mice were originally generated for studies regarding respiratory disease, central chemoreceptive neurons in the retrotrapezoid nucleus, respiratory rhythmogenesis, as well as autonomic associated neurons more generally. In fact, a variety of visceral receptors in nodose will be knocked out in Phox2bCre;Piezo2 cKO including vagal airway mechanoreceptors as some of these authors showed previously (Nonomura et al. 2017), and yet this potential confound is not mentioned in the present work. This is an important consideration as the number of Piezo1 or Piezo2 neurons labeled from the carotid versus those present in the nodose or petrosal ganglia are considerably different as noted by the authors.

Second, a parallel set of experiments to the KO studies employed optogenetic activation of Piezo2-positive fibers. Mice were generated by crosses between Piezo2Cre- and Cre-dependent channel rhodopsin reporter mice. These experiments complement the other studies by showing that activation of Piezo2-positive neurons of the ADN reflexively decreased blood pressure and heart rate. However, there are also significant concerns. First, the frequency of stimulation was 50 Hz, which exceeds the normal frequency response range of baroreceptors operating at physiological blood pressure levels. Second, the authors report that administration of the β-blocker propranolol reduced the optogenetically evoked hypotension and bradycardia by ~50%. This is concerning for two reasons: 1) numerous studies have documented that the bradycardia due to baroreflex loading is largely vagally mediated (Ma et al. 2002, 2003) and to a much lesser extent attributed to cardiac sympathoinhibition, and 2) the hypotension should be mediated by sympathoinhibition via α-adrenergic receptors. Changes in heart rate of mice do not largely affect blood pressure due to the opposing changes in filling-time and end-diastolic volume.

SUMMARY AND CONCLUSIONS

The molecular identity of baroreflex transduction remains a controversial issue. Indeed, prior reports have proposed a number of ion channels may underlie this process including the epithelial sodium channel (Drummond et al. 1998) or acid-sensing ion channels (Lu et al. 2009). The recent publication by Zeng et al. (2018) provides provocative data regarding Piezo1/Piezo2 as the putative ion channel underlying the mechanosensitivity of baroreceptor neurons. However, the validity of any study depends on the ability of the data to provide critical evidence to convincingly support the conclusions. Given the available data, there are significant concerns regarding the evidence supporting a role of Piezo1/Piezo2 in arterial baroreceptor function.

GRANTS

We gratefully acknowledge National Heart, Lung, and Blood Institute Grants HL-128388 and HL-145875 (to S. D. Stocker) and HL-133505 (to M. C. Andresen).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

S.D.S. prepared figures; S.D.S., A.F.S., and M.C.A. drafted manuscript; S.D.S., A.F.S., and M.C.A. edited and revised manuscript; S.D.S., A.F.S., and M.C.A. approved final version of manuscript.