Central and peripheral chemoreflexes in humans with treated hypertension

Increased peripheral chemoreflex sensitivity is a pathogenic feature of human hypertension (HTN), while both central and peripheral chemoreflex sensitivities are reportedly augmented in animal models of HTN. Herein, we tested the hypothesis that both central and combined central and peripheral chemoreflex sensitivities are augmented in HTN. Fifteen HTN participants (68 ± 5 years; mean ± SD) and 13 normotensives (NT; 65 ± 6 years) performed two modified rebreathing protocols in which the partial pressure of end‐tidal carbon dioxide ( PETCO2${P_{{\rm{ETC}}{{\rm{O}}_2}}}$ ) progressively increased while the partial pressure of end‐tidal oxygen was clamped at either 150 mmHg (isoxic hyperoxia; central chemoreflex activation) or 50 mmHg (isoxic hypoxia; combined central and peripheral chemoreflex activation). Ventilation ( V̇E${\dot{V}}_{\rm{E}}$ ; pneumotachometer) and muscle sympathetic nerve activity (MSNA; microneurography) were recorded, and ventilatory ( V̇E${\dot{V}}_{\rm{E}}$ vs. PETCO2${P_{{\rm{ETC}}{{\rm{O}}_2}}}$ slope) and sympathetic (MSNA vs. PETCO2${P_{{\rm{ETC}}{{\rm{O}}_2}}}$ slope) chemoreflex sensitivities and recruitment thresholds (breakpoint) were calculated. Global cerebral blood flow (gCBF; duplex Doppler) was measured, and the association with chemoreflex responses was examined. Central ventilatory and sympathetic chemoreflex sensitivities were greater in HTN than NT (2.48 ± 1.33 vs. 1.58 ± 0.42 L min−1 mmHg−1, P = 0.030: 3.32 ± 1.90 vs. 1.77 ± 0.62 a.u. mmHg−1, P = 0.034, respectively), while recruitment thresholds were not different between groups. HTN and NT had similar combined central and peripheral ventilatory and sympathetic chemoreflex sensitivities and recruitment thresholds. A lower gCBF was associated with an earlier recruitment threshold for V̇E${\dot{V}}_{\rm{E}}$ (R2 = 0.666, P < 0.0001) and MSNA (R2 = 0.698, P = 0.004) during isoxic hyperoxic rebreathing. These findings indicate that central ventilatory and sympathetic chemoreflex sensitivities are augmented in human HTN and perhaps suggest that targeting the central chemoreflex may help some forms of HTN.


Introduction
Hypertension (HTN) is a powerful independent predisposing factor for the development of coronary heart disease, stroke and heart failure, and as such is a global public health problem (Ettehad et al., 2016). The pathogenesis of HTN is complex and multifactorial involving genetic, environmental and behavioural factors (Whelton et al., 2018). Among the putative pathogenic mechanisms, there is evidence that heightened sympathetic nerve activity plays a role in the initiation and development of HTN, and thus constitutes a potential target for arresting the progression of HTN and target organ damage (Fisher & Paton, 2012;Fisher et al., 2022). To develop effective countermeasures, the mechanisms driving sympathetic dysregulation in HTN must be identified; one of these mechanisms is the chemoreceptors.
Peripheral chemoreceptors located in the carotid and aortic bodies are responsive to decreases in the partial pressure of oxygen (P O 2 ) and increases in the partial pressure of carbon dioxide (P CO 2 ) and hydrogen ion concentration ([H + ]) (Comroe, 1939;Prabhakar, 2016;Zera et al., 2019), whereas the central chemoreceptors, located principally in the ventrolateral medulla oblongata (e.g. retrotrapezoid nucleus), respond to increases in local tissue P CO 2 /[H + ] (Guyenet, 2014;Guyenet et al., 2016). Independent activation of the central and peripheral chemoreceptors in humans (e.g. with hyperoxic hypercapnia and hypoxic normocapnia, respectively) causes a marked increase in minute ventilation (V E ) and muscle sympathetic nerve activity (MSNA) (Duffin, 2005;Guyenet, 2014;Keir et al., 2019). Notably,V E and MSNA responses to peripheral chemoreflex activation with hypoxia have been reported to be greater in borderline HTN (Somers et al., 1988) and mild HTN (Trzebski et al., 1982) compared to normotension (NT) controls. Conversely, deactivation of the peripheral chemoreceptors with hyperoxia has been reported to evoke more pronounced reductions in restingV E (Tafil-Klawe et al., 1985) and MSNA (Sinski et al., 2012) in HTN. These findings support the view that the peripheral chemoreflex is a promising therapeutic target for treating HTN (Narkiewicz et al., 2016;Pijacka et al., 2016) and other cardiovascular diseases (Marcus et al., 2014;Niewinski et al., 2013;Niewinski et al., 2017).
In contrast to the peripheral chemoreflex, limited work has considered the role of the central chemoreflex in hypertension. It has been observed that the spontaneously hypertensive rat (SHR) has an augmented central chemoreflex sensitivity that is associated with the development of raised blood pressure (BP) (Li et al., 2016). However, sympathetic recordings were not obtained and, to date, the influence of human HTN on the sympathetic vasoconstrictor responses to central chemoreflex activation remains unknown. Moreover, the nature of the integrative response to combined central and peripheral chemoreflex activation has been intensely debated (Duffin & Mateika, 2013;Teppema & Smith, 2013;Wilson & Day, 2013), with additive, hypoadditive and hyperadditive responses described. Importantly, in human HTN the cardiorespiratory and sympathetic responses provoked by concurrent hypercapnia (acidosis) and hypoxia (i.e. combined central and peripheral chemoreflex activation) have not been determined, despite their relevance for clinical conditions (e.g. chronic obstructive pulmonary disease, obstructive sleep apnoea).
Using magnetic resonance imaging, Warnert et al. (2016) observed reduced perfusion and increased cerebral vascular resistance (CVR) in treated patients with HTN. This has been implicated in the pathogenesis of HTN via the sympatho-excitatory Cushing mechanism (i.e. the selfish brain hypothesis) (Hart, 2016;McBryde et al., 2017). A further consequence of the attenuated perfusion in HTN may be reduced CO 2 /[H + ] washout from chemosensitive areas of the brain and subsequent heightened central chemoreflex sensitivity (Carr et al., 2021;Hoiland et al., 2015;Xie et al., 2006). However, recently identified regional heterogeneities in cerebrovascular CO 2 reactivity, specifically hypercapnia-induced vasoconstriction in the retrotrapezoid nucleus, might question this concept (Cleary et al., 2020;Hawkins et al., 2017). Nevertheless, it remains to be explored whether a lower cerebral blood flow is related to exaggerated central chemoreflex sensitivity and earlier recruitment in HTN (Warnert et al., 2016).
It has recently been highlighted that the magnitude of the MSNA response to central and peripheral chemoreflex activation is weakly associated with the correspondinġ V E response (Keir et al., 2019;Prasad et al., 2020). An important potential implication of this is that characterizing the appropriate end-organ response (V E , MSNA) is essential in order that those patients for whom therapeutic targeting of the peripheral and/or central chemoreflexes might be most fruitful are optimally identified. However, it is currently unknown whether thė V E and MSNA responses to chemoreflex activation are associated in humans with HTN.
Herein, we compared theV E and MSNA responses to a hypercapnic rebreathing protocol (central chemoreflex activation) with the end-tidal P O 2 (P ETO 2 ) maintained isoxic at hypoxic (peripheral chemoreflex activation) or hyperoxic (peripheral chemoreflex deactivation) levels in HTN and NT using an established method (Duffin, 2011). In addition, duplex Doppler ultrasound was used to determine resting global cerebral blood flow (gCBF).
The following novel hypotheses were tested: (1)V E and MSNA responses to central chemoreflex activation, and combined central and peripheral chemoreflex activation, are both augmented in human HTN; (2) lower gCBF is associated with earlier and increasedV E and MSNA responses to central chemoreflex activation; and (3) thė V E response to chemoreflex activation is weakly associated with the corresponding MSNA responses.
A. L. C. Sayegh and others J Physiol 601.12

Ethical approval
Ethical approval was granted by the Northern B Health and Disability Ethics Committee, Auckland, New Zealand (19/NTB/125), by the Auckland District Health Board Research Review Committee (A+ 8687), and was registered with the Australian New Zealand Clinical Trials Registry (ACTRN12619001767190). All volunteers were provided with a detailed verbal and written explanation of the study procedures before providing written consent to participate in the study. The study was conducted according to the standards outlined in the latest revision of the Declaration of Helsinki (2013).

Participants
Fifteen participants with essential HTN (Stage 2, treated controlled or uncontrolled, office systolic BP ≥140 mmHg or diastolic BP ≥90 mmHg) and 13 NT controls (office systolic BP <130 mmHg and diastolic BP <85 mmHg)  were enrolled in the study. All participants were non-obese [body mass index (BMI) <30 kg m -2 ] and reported being non-smokers, not users of recreational drugs and not misusers of alcohol. Exclusion criteria involved: history or symptoms of pulmonary, metabolic (e.g. diabetes) or neurological disease, and any acute or chronic disorders associated with alterations in cardiovascular structure or function. No participant was taking over-the-counter or prescription medication aside from anti-hypertensive therapy. No women were active users, or past users, of any kind of hormone replacement therapy. Participant characteristics and medication usage are provided in Table 1.

Experimental protocol
The experimental protocol involved a familiarization visit and an experimental visit to the laboratory conducted on separate days. Prior to the sessions, participants were requested to abstain from caffeinated drinks and physical activity for >12 h, alcohol for >24 h and food for >2 h. Patients were asked to withhold their morning medications until the end of the study. All experiments were performed in a temperature-controlled room (21-22°C). The initial familiarization visit involved a careful introduction to all study methods and protocols. Anthropometric (height and weight), demographic and clinical information data were collected. Following this, vertebral artery (VA) and internal carotid artery (ICA) blood flow velocity and diameter data were obtained bilaterally, while participants rested in the same semi-recumbent position adopted for chemoreflex assessment. The session concluded with participants being fully instrumented (aside from microneurography) and performing the isoxic hypoxia rebreathing (trial 2) protocol (described below). This meant that participants could practice the rebreathing protocol and experience the related sensations ahead of the experimental visit. At the experimental visit, upon arrival at the laboratory, participants laid on a medical exam table bed in a semi-recumbent position. After instrumentation, the Duffin modified rebreathing protocol (Duffin, 2011;Keir et al., 2019) was undertaken. In brief, this comprised: (1) a resting period of 5 min, (2) a 5 min period while participants increased the frequency and depth of breathing to lower P ETCO 2 to ∼25 mmHg (i.e. hyperventilation), (3) switching from room air, at the end of a full expiration, and taking three deep breaths from a rebreathing bag, and (4) an ∼10-15 min period during which participants freely breathed in and out from the rebreathing bag. The rebreathing bag contained a medical-grade gas mixture with a concentration of oxygen sufficient to produce a P ETO 2 of either 150 mmHg (trial 1; Fig. 1A) or 50 mmHg (trial 2; Fig. 1B). The rebreathing bag also contained 6% CO 2 (nitrogen balanced). The hyperventilation period was used to deplete CO 2 stores (Duffin, 2011) and allow the identification of a P ETCO 2 recruitment threshold at whichV E and MSNA increased. The three deep breaths rapidly equalize P CO 2 at the mouth, lung and arterial blood with that of the mixed-venous blood (Duffin, 2011). In accordance with previous studies, the observation of a plateau in P ETCO 2 at the onset of rebreathing was taken as evidence for this and implemented as a prerequisite for continuing the protocol (Keir et al., 2019). By design, during the rebreathing period P ETCO 2 increased progressively until tests were terminated either when P ETCO 2 reached ∼55 mmHg, when ventilation reached ∼100 L min −1 or at the participant's discretion (whichever happened first). During the rebreathing, the target P ETO 2 was maintained using a computerized gas sensor-controller as previously described (Duffin, 2011). Participants were blinded to the order of testing, but to avoid any potentially confounding after-effects of hypoxia on MSNA (Xie et al., 2001), isoxic hyperoxia (trial 1) preceded isoxic hypoxia (trial 2). Trials were separated by ∼15 min of recovery (or until the restoration of measured variables).

Experimental measurements
Heart rate (HR) was continuously monitored with a lead II electrocardiogram (BioAmp, FE231; ADInstruments, Bella Vista, NSW, Australia) and beat-to-beat BP measured via finger photoplethysmography (Human NIBP Nano interface, MLA382; ADInstruments). Finger BP values were validated against brachial artery BP measurements (Tango M2 BP monitor; SunTech, Morrisville, NC, USA). Arterial oxygen saturation (S pO 2 ) was measured using a finger pulse oximeter (MLT321 and ML320/F; ADInstruments). Multi-unit postganglionic MSNA was obtained from the right peroneal nerve using the microneurography technique (Fisher et al., 2015;Sayegh et al., 2022). Palpation and percutaneous electrical stimulation were used to locate the peroneal nerve posterior to the fibular head. A tungsten microelectrode was then inserted through the skin and into the peroneal nerve at the fibular head and a reference electrode inserted subcutaneously 1−3 cm distal. Adjustments to the recording electrode were made to acquire a signal displaying a pulse-synchronous pattern of spontaneous sympathetic bursting activity. Raw signals were amplified (×100,000), bandpass filtered (700-2000 Hz), rectified, and integrated (time constant 0.1 s) to obtain a mean voltage neurogram (Nerve Traffic Analyzer, model 662C-4; Bioengineering, University of Iowa, Iowa City, IA, USA).
Participants wore a mouthpiece (with nose-clip) connected in series to a low-resistance pulmonary filter (Water filter, MLA304; ADInstruments), a pneumotachometer (3830 Series, Heated Linear E Pneumotachometer; Hans Rudolph Inc., Kansas City, MO, USA), and a three-way valve (2100 Series; Hans Rudolph Inc.) that permitted switching from breathing room air (open circuit) to rebreathing from a 6 L bag (closed circuit).V E , tidal volume (V T ) and respiratory frequency (Rf) were measured breath-by-breath with the pneumotachometer. A sample line, connected to the mouthpiece, permitted continuous sampling of expired air and subsequent determination of the partial pressure of end-tidal oxygen (P ETO 2 ) and carbon dioxide (P ETCO 2 ) (Respiratory Gas Analyzer, ML206; ADInstruments). Perception of breathlessness was assessed using a 0−10 point Borg scale (Borg, 1982) ranging from 'Your breathing is causing you no difficulty at all' (0 points) to 'Your breathing difficulty is maximal' (10 points).
Bilateral assessments of internal carotid (ICA) and vertebral (VA) artery diameter and blood velocity were obtained using duplex Doppler ultrasound (Terason uSmart 3300; Teratech Corporation, Burlington, MA, USA) in accordance with standard guidelines (Thomas et al., 2015). In brief, with the participant semi-recumbent and the chin elevated and orientated towards the contralateral side, a 4−15 MHz multi-frequency linear-array transducer was used to insonate the target vessel with a constant 60 degree angle relative to the skin. The ICA was assessed at a site 1−1.5 cm distal to the carotid bifurcation, while the VA was assessed between the transverse process of the C3 vertebra and the subclavian artery. The B-mode was used for vessel localization and diameter measures, while the pulse-wave mode was used for velocity measures. Care was taken to ensure that the sample volume was positioned in the centre of the vessel and adjusted to cover the width of the vessel diameter. Ultrasound images were screen captured (DVI2USB3.0; Epiphan Video, Ottawa, ON, Canada) and stored as digital AVI files for offline analysis using an automated edge detection and wall-tracking software program (Cardiovascular Suite; Quipu, Italy). All video files were analysed by a single operator (A.L.C.S.).

Data analysis
Cardiorespiratory signals underwent analog-to-digital conversion at 1000 Hz (PowerLab 16/35 and LabChart version 8; ADInstruments). From the ECG, HR was calculated on a beat-to-beat basis. The arterial waveform was used to identify systolic (SBP) and diastolic (DBP) pressure on a beat-to-beat basis, and mean arterial pressure (MAP) was obtained by integrating the arterial waveform over the complete cardiac cycle. Breath-by-breath V T , Rf,V E , P ETO 2 , and P ETCO 2 were obtained and artefacts (e.g. produced by swallowing) were removed. A custom written Spike2 script (Cambridge Electronic Design, Cambridge, UK) was used by a single operator (A.L.C.S.) as an interactive means of identifying MSNA bursts. To account for peripheral sympathetic conduction delays and align sympathetic bursts with the appropriate cardiac cycle, neurograms were time-shifted backwards relative to all other signals by computing a burst latency for each participant (in seconds) according to their height (HTN 1.31 ± 0.04, NT 1.32 ± 0.05 s) (Fagius & Wallin, 1980). To 'normalize' an MSNA burst the highest spontaneously occurring sympathetic burst during each baseline period (i.e. before isoxic hyperoxic and isoxic hypoxic rebreathing trials) was first identified for each participant. These bursts were set as 100% and all other bursts were expressed relative to these highest bursts in that respective trial. That is, the highest spontaneously occurring sympathetic burst during the isoxic hyperoxic baseline was used to normalize all other bursts occurring during that trial, and the same approach was taken for the isoxic hypoxic rebreathing trial.
Burst frequency (bursts per minute), burst incidence (number of bursts per 100 heartbeats), burst amplitude (i.e. strength) and total MSNA (product of burst frequency and mean burst amplitude) were determined. In addition, bursts were identified with respect to the timing of their occurrence in the respiratory cycle, using the Spike2 script (Cambridge Electronic Design). This then allowed us to compute total MSNA only during the low-lung volume phase (i.e. last 75% of expiration, first 25% of inspiration (MSNA LLV ) on a breath-by-breath basis, and thus limiting the influence of lung volume on MSNA (Keir et al., 2019). Mean values for cardiovascular, respiratory and MSNA variables were calculated for baseline (5 min), hyperventilation (last 3 min preceding isoxic hyperoxic and isoxic hypoxic rebreathing) and rebreathing periods for both rebreathing trials. In addition, the average from the last 30 s of rebreathing was taken as the 'peak rebreathing' response.
Breath-by-breath P ETCO 2 from the rebreathing period was plotted againstV E and slopes fitted with a segmental linear regression (least squares fit) using a commercially available graphing program (Prism 8.0; GraphPad Software, San Diego, CA, USA). In brief, theV E vs. P ETCO 2 slope comprises two segments separated by a single breakpoint. The first segment was fitted with a straight line (i.e. slope parameter set as 0 L min -1 mmHg -1 ), and taken as a measure of basalV E . The breakpoint of P ETCO 2 was taken as the ventilatory recruitment threshold (in mmHg). Thereafter, the second segment, wherebyV E increases with increasing P ETCO 2 , was taken as the ventilatory chemoreflex (central or combined central and peripheral) sensitivity to hypercapnia (L min −1 mmHg −1 ). Breath-by-breath P ETCO 2 from the rebreathing period was then bin-averaged over the range 39-53 mmHg (2 mmHg bins) and plotted against MSNA LLV . This range was selected to maximize the number of participants included in this analysis (i.e. had a value for each P ETCO 2 bin).
Here, the breakpoint of MSNA LLV vs. P ETCO 2 was taken as the sympathetic recruitment threshold (in mmHg) and the second segment, whereby MSNA LLV increases with increasing P ETCO 2 , was taken as the sympathetic chemoreflex sensitivity to hypercapnia (a.u. mmHg −1 ). By using MSNA LLV for this analysis of the chemoreflex control of MSNA, the effect of changes in lung volume per se is minimized (Keir et al., 2019). The peripheral chemoreflex responsiveness was identified from the difference between the isoxic hyperoxic rebreathing (central chemoreflex activation) and isoxic hypoxic rebreathing (combined central and peripheral chemoreflex activation) tests (Keir et al., 2019), because an additive relationship has been reported between the central and peripheral chemoreflexes in humans (Duffin, 2007).
Bilateral ICA and VA blood flow velocity and diameter were obtained from a continuous 60 s recording from each participant.
Blood flow (Q) was calculated as: Mean blood flow velocity was divided by 2 because the automated edge detection and wall-tracking software program used assesses peak blood velocity (i.e. it tracks the uppermost edge of the velocity spectra), and it is assumed that half the peak velocity is representative of the intensity-weighted mean velocity (Li et al., 1993). Values for ICA Q and VA Q were taken as the mean of the respective left and right measures (Thomas et al., 2015).
gCBF was calculated as: CVR was calculated as:

Statistical analysis
Normality was assessed using the Shapiro-Wilk test. Participant characteristics (e.g. HTN vs. NT) were compared using Student's t tests and chi-square tests.
The main effects of group (HTN, NT), breathing period (eucapnia, hyperventilation, peak rebreathing) and their interaction, and the main effects of group, breathing trial (isoxic hyperoxic and isoxic hypoxic rebreathing) and their interaction, were examined using mixed model ANOVA with repeated measures. Post hoc analysis was undertaken using a t test with Bonferroni correction. Pearson correlation was used to evaluate the relationship between gCBF and CVR withV E and MSNA responses and between ventilatory and sympathetic recruitment thresholds and sensitivities during central, and combined central and peripheral chemoreflex activation. Values are presented as mean ± SD, unless otherwise stated. Statistical analyses were undertaken using SigmaPlot version 14.0 (Systat Software, Inc., San Jose, CA, USA) and P < 0.05 was considered significant.

Discussion
The novel findings of this study are that: (1) the central ventilatory and sympathetic chemoreflex sensitivities are augmented in HTN; (2) in contrast, ventilatory and sympathetic responses to combined central and peripheral chemoreflex activation were not different between HTN and NT; (3) ventilatory and sympathetic recruitment thresholds during central chemoreflex activation (isoxic hyperoxic rebreathing) were positively correlated with gCBF and negatively correlated with CVR; and (4) during central chemoreflex activation the ventilatory recruitment threshold was higher than the sympathetic recruitment threshold. Collectively, these findings implicate the central chemoreflex in the pathogenesis of human HTN and support the view that it represents a potential therapeutic target in some forms of HTN.

Central and peripheral chemoreflexes
Hyperoxic hypercapnia has been widely used to assess central chemoreflex responsiveness while suppressing the activity of the peripheral chemoreflex (Dejours, 1962;Duffin, 2011). In NT humans, hyperoxic hypercapnia is known to elevateV E , MSNA, vascular resistance and BP (Sayegh et al., 2022). Li et al. (2016) and Somers et al. (1989) reported that the SHR, a neurogenic model of HTN, has an augmented ventilatory response to hyperoxic hypercapnia compared to Wistar-Kyoto rats with normal blood pressure, indicating an increased central chemoreflex sensitivity in the SHR. In this study, we show for the first time that both the central ventilatory and the sympathetic chemoreflex sensitivities are augmented in HTN compared to NT. Li et al. (2016) reported that alterations in orexin neurons explained the increased central chemoreflex sensitivity in the SHR. These neurons are located in the retrotrapezoid nucleus and the hypothalamus, are stimulated with hypercapnia, and send projections to central sites involved in the regulation of breathing, sympathetic activity and blood pressure (e.g. medullary raphe, nucleus of the solitary tract, paraventricular nucleus and rostral ventrolateral medulla) Ventilatory recruitment threshold data are n = 15 HTN and n = 13 NT, and sympathetic recruitment threshold data are n = 14 HTN and n = 10 NT (due to an absence of MSNA recording). Values are expressed as the mean ± SD. The main effects of group (HTN, NT), trial (central, combined central and peripheral) and their interaction were examined using mixed model ANOVA with repeated measures. HTN, hypertensive; NT, normotensive. Bold type indicates significant values. (Barnett & Li, 2020). The SHR not only has more orexin neurons, but hypercapnia activates a greater proportion of orexin neurons in the SHR than the Wistar-Kyoto rat (Li et al., 2016). Moreover, antagonism of the orexin receptors with almorexant lowered BP (Li et al., 2013) and normalized the ventilatory and BP responses to hypercapnia in SHR (Li et al., 2016). Whether alterations in orexin neurons explain the greater central ventilatory and sympathetic chemoreflex sensitivities we report in HTN requires further investigation.
In several clinical scenarios, such as chronic obstructive pulmonary disease and obstructive sleep apnoea, hypoxia and hypercapnia (acidosis) coexist (Narkiewicz et al., 1999;West, 2011). Moreover, this coexistence is recognized as being an important mechanism driving elevated MSNA (Narkiewicz et al., 1999), while in chronic heart failure enhanced ventilatory chemoreflex sensitivity to both hypoxia and hypercapnia is associated with poor survival (Giannoni et al., 2009;Ponikowski et al., 2001). It is well established that combined central and peripheral chemoreflex activation by breathing a hypoxic and hypercapnic gas, delivered either via the steady-state open-circuit method [fixed fraction of inspired oxygen (F IO 2 ): 0.10 and fraction of inspired carbon dioxide (F ICO 2 ): 0.07] (Sayegh et al., 2022;Somers et al., 1989) or rebreathing (Keir et al., 2019), evokes robust increases iṅ V E and MSNA in young healthy individuals. Contrary to expectation, in the current study we observed that the magnitude of ventilatory and sympathetic responses to combined central and peripheral chemoreflex activation (isoxic hypoxic rebreathing) was not different in HTN and NT. Keir et al. (2019) demonstrated in young healthy individuals that the ventilatory (V E vs. P ETCO 2 slope) and sympathetic (MSNA LLV vs. P ETCO 2 slope) chemoreflex sensitivities were greater during combined central and peripheral chemoreflex activation (isoxic hypoxic rebreathing) than during central chemoreflex activation alone (isoxic hyperoxic rebreathing). This was also true for the NT group in the current study, where chemoreflex sensitivity was greater during combined central and peripheral chemoreflex activation than during central chemoreflex activation for bothV E and MSNA. In contrast, in the HTN group, both the ventilatory and the sympathetic sensitivities were not different for the combined central and peripheral chemoreflex activation and central chemoreflex trial. Duffin (2007) proposed that peripheral chemoreflex responsiveness may be identified from the difference between the isoxic hyperoxic rebreathing (central  Cardiorespiratory variables are n = 15 HTN and n = 13 NT, and sympathetic variables are n = 14 HTN and n = 10 NT. Values are expressed as the mean ± SD. The main effects of group, breathing period (baseline, hyperventilation, peak rebreathing), and their interaction were examined using mixed model ANOVA with repeated measures. HTN, hypertensive; NT, normotensive; HR, heart rate; SBP, systolic blood pressure; DBP, diastolic blood pressure, MAP, mean arterial pressure; S pO 2 , oxygen saturation; P ETO 2 , partial pressure of end-tidal oxygen; P ETCO 2 , partial pressure of end-tidal carbon dioxide;V E , minute ventilation; V T , tidal volume; Rf, breathing frequency; MSNA, muscle sympathetic nerve activity; MSNA LLV , total muscle sympathetic nerve activity at low lung volume. Where a significant main effect of trial, but no interaction, was observed, differences identified during post hoc analysis (t tests with Bonferroni correction) are shown as a P < 0.05 eucapnia vs. hyperventilation, b P < 0.05 eucapnia vs. peak rebreathing, and c P < 0.05 hyperventilation vs. peak rebreathing. Bold type indicates significant values. chemoreflex activation) and isoxic hypoxic rebreathing (combined central and peripheral chemoreflex activation) tests, because an additive relationship has been reported between the central and peripheral chemoreflexes in humans. Applying this approach to our data for NT, the calculated peripheral chemoreflex sensitivities foṙ V E and MSNA are 1.60 ± 0.81 L min −1 mmHg −1 and 1.89 ± 0.85 a.u. mmHg −1 , respectively. This equates to the contribution of the peripheral chemoreflex to theV E and MSNA responses to the isoxic hypoxic rebreathing being 48% and 50% in NT. However, for HTN, the calculated peripheral chemoreflex sensitivities forV E and MSNA are 0.26 ± 0.24 L min −1 mmHg −1 and 0.29 ± 0.21 a.u. mmHg −1 , respectively. This equates to a much smaller contribution of the peripheral chemoreflex to the isoxic hypoxic rebreathing responses of ∼11% for both theV E and MSNA LLV . This blunted peripheral chemoreflex responsiveness in HTN compared to NT was unexpected and in stark contrast to the evidence for heightened peripheral chemoreflex in human HTN (Pijacka et al., 2016;Sinski et al., 2012;Somers et al., 1988;Tafil-Klawe et al., 1985;Trzebski et al., 1982) and a variety of animal models (Abdala et al., 2012;McBryde et al., 2013;Pijacka et al., 2016). One potential explanation for these conflicting findings is the differing methods used to assess the peripheral chemoreflex, which include isocapnic hypoxia (Somers et al., 1988), progressive isocapnic hypoxic rebreathing (Trzebski et al., 1982) and hyperoxia (Sinski et al., 2012;Tafil-Klawe et al., 1985). In contrast, herein peripheral chemoreflex responsiveness was estimated from the difference between the isoxic hyperoxic rebreathing and isoxic hypoxic rebreathing tests, with the underlying assumption that there is an additive relationship between the central and peripheral chemoreflexes in humans (Duffin, 2007). However, the nature of this relationship is controversial, and experimental animal studies (e.g. in rats, cats, goats and dogs) have variously identified additive, hypo-additive and hyper-additive relationships (Blain et al., 2010;Cummings, 2014;Day & Wilson, 2009;Smith et al., 1984;Smith et al., 2015;Wilson & Teppema, 2016). Therefore, synaptic occlusion of inputs from central and peripheral chemoreceptors, rather than their algebraic (or synergistic) summation, may explain theV E and MSNA responses we observed in HTN. Another potential explanation is a difference in the characteristics and aetiology of the HTN patients studied. In general, our HTN group were older, more medicated, and contained both males and females unlike previous studies that recruited only men (Pijacka et al., 2016;Sinski et al., 2012;Somers et al., 1988;Tafil-Klawe et al., 1985;Trzebski et al., 1982). Xie et al. (2006) demonstrated in young healthy individuals that oral administration of the cyclooxygenase inhibitor indomethacin reduced middle cerebral artery blood flow velocity (a proxy of cerebral blood flow) by 77% and increased hypercapnic ventilatory responsiveness (V E /P ETCO 2 slope), assessed using a steady-state open circuit method, by 40-60%. Such findings support the concept that reduced cerebral perfusion attenuates CO 2 /H + washout from central chemoreceptor regions and increases central ventilatory chemoreflex sensitivity (Carr et al., 2021;Hoiland et al., 2015). However, this runs contrary to intriguing observations in murine studies where high CO 2 /H + causes a unique constriction of retrotrapezoid nucleus arterioles, a key site of central chemoreception (Cleary et al., 2020;Hawkins et al., 2017). Nevertheless, in the present study, gCBF was positively correlated with the ventilatory and sympathetic recruitment thresholds during isoxic hyperoxic rebreathing, such that individuals with a lower gCBF had an earlier breakpoint of P ETCO 2 at whichV E and MSNA LLV started to increase during rebreathing. This supports the idea that people with lower gCBF may exhibit earlier central chemoreflex responses. Of note, we did not observe an association between gCBF/CVR and either central ventilatory or sympathetic chemoreflex sensitivities. A potential explanation for this may be the method used to assess chemoreflex sensitivity. Of note, indomethacin-induced reductions in cerebral perfusion increase ventilatory CO 2 sensitivity when determined using a steady-state open circuit method (Fan et al., 2010;Xie et al., 2006), but not when assessed using rebreathing (Fan et al., 2010). The reason for this appears to relate to the P CO 2 gradient that exists across the body compartments (end-tidal, arterial, brain tissue, jugular venous) under normal conditions and during the steady state method, being equalized during rebreathing (Berkenbosch et al., 1989;Read, 1967). For example, an indomethacin-induced lowering of cerebral blood flow would be expected to mean that for a given inspired CO 2 there would be a greater rise in central CO 2 due to an impaired washout from central chemoreceptor regions, thereby widening the P CO 2 gradient between the end-tidal and central compartments (Fan et al., 2010). However, during rebreathing the P CO 2 gradient becomes equalized and therefore differences in cerebral blood flow do not appreciably affect the P CO 2 gradient between the arterial blood and brain (Berkenbosch et al., 1989;Read, 1967). As such, the use of rebreathing may have diminished the extent to which cerebral blood flow influences CO 2 /H + washout from central chemoreceptor regions in the current study (Carr et al., 2021;Hoiland et al., 2015) and accounts for the absence of any relationship being observed between gCBF and chemoreflex sensitivity.

Central chemoreflex and cerebral blood flow
In the current study, neither gCBF nor CVR were different in the HTN vs. NT groups, so a mechanistic role for gCBF and CVR in the greater central chemoreflex sensitivity observed in HTN is not supported. Magnetic resonance imaging (MRI) studies have identified increased CVR and hypoperfusion in human HTN (Warnert et al., 2016) and the discrepancy with the results of the current study is unexpected. A notable difference between the studies is the experimental methods employed; in the present study duplex Doppler was used to measure gCBF (in mL min −1 ) while the MRI approach used by Warnert et al. (2016) permitted the measurement of cerebral blood flow relative to brain tissue volume (in mL 100 mL -1 min -1 ) which may provide a more sensitive assessment of HTN-induced alterations in cerebral perfusion. Keir et al. (2019) observed that the concurrentV E and MSNA responses to either isoxic hyperoxic rebreathing or isoxic hypoxic rebreathing differed such that the sensitivity of the responses was not correlated. Similarly, in the current study, no correlation was observed between ventilatory and sympathetic chemoreflex sensitivities during either central and combined central plus peripheral chemoreflex activation. This observation is consistent with the view that not only is the carotid body a multi-modal sensor, but via a complex and highly organized process of signalling to and integration within the CNS, distinct patterns of efferent output and end-organ responses can be evoked (Zera et al., 2019). The aortic bodies are also responsive to hypoxia and hypercapnia (Prabhakar, 2016), and we cannot exclude their contribution to the responses observed. However, the finding that bilateral carotid body resection in heart failure patients virtually abolished theV E response to hypoxia but did not affect the HR response (Niewinski et al., 2014), points to the predominant effect of aortic body stimulation being cardiac (Prabhakar, 2016), although their effect on MSNA remains unknown. Aside from the physiological insights provided by the observations made in the present study, the applied significance may be that any diagnostic test of chemoreflex sensitivity should carefully select the appropriate end-organ response (e.g. MSNA, BP,V E ) on the basis of the therapeutic objective (e.g. sympatho-inhibition, BP lowering).

Ventilation versus sympathetic responses
Interestingly, we observed that the P ETCO 2 recruitment threshold for bothV E and MSNA were lower during hypoxic than hyperoxic rebreathing, in agreement with previous studies utilizing Duffin's modified rebreathing method (Keir et al., 2019). Longer term exposure to isocapnic hypoxia (e.g. 3 h) has also been shown to decrease the recruitment threshold at whichV E increases during rebreathing and attributed to enhanced activation of the peripheral chemoreflex (Mahamed et al., 2003). Irrespective of the underlying mechanisms, lowering of the P ETCO 2 recruitment threshold for bothV E and MSNA observed during hypoxic rebreathing in the present study may have implications forV E and MSNA regulation in chronic diseases in which hypoxia and hypercapnia coexist (i.e. obstructive sleep apnoea and heart failure) (Giannoni et al., 2009;Levy et al., 2015;West, 2011).

Experimental considerations
A variety of methods have been developed to assess chemoreflex sensitivity in humans, and their relative merits and limitations are debated (Duffin, 2007(Duffin, , 2011Keir et al., 2019;Prasad et al., 2020). Herein, rebreathing was used to produce progressive hypercapnia and central chemoreflex engagement under conditions of isoxic hyperoxia (peripheral chemoreflex silencing) and isoxic hypoxia (peripheral chemoreflex activation) following the method described by Duffin (2007), which modified the approach developed by Read (1967). An advantage of this approach is that both chemoreflex sensitivity and recruitment threshold can be identified forV E and MSNA (Keir et al., 2019). A potential concern with this approach is the use of hyperoxia (150 mmHg), and while it is assumed to selectively diminish the peripheral chemoreflex response to hypercapnia (Duffin, 2007;Keir et al., 2020), we cannot exclude the possibility that the peripheral chemoreflex was completely suppressed and that off-target effects occurred. With respects to the latter, these include an increase central oxidative stress and CVR that may augment brain tissue P CO 2 (Becker et al., 1996;Dean et al., 2004) and augment central chemoreflex activation.
The absence of a specific measure of the peripheral chemoreflex responsiveness is a limitation of this study. During pilot testing it was found that the inclusion of another chemoreflex test, and the associated lengthening of the protocol, was not well tolerated by the participants. In addition, the use of Duffin's modified rebreathing approach permits the characterization of peripheral chemoreflex sensitivity by subtraction of isoxic hyperoxic rebreathing responses from the isoxic hypoxic rebreathing response (Duffin, 2007). Finally, the aim of this study was to examine the impact of hypertension on the central chemoreflex and combined central and peripheral chemoreflex sensitivity, and peripheral chemoreflex sensitivity has previously been characterized in HTN. For these reasons a separate protocol to directly investigate the peripheral chemoreflex was not included.
Participants breathed freely during trials, and increaseḋ V E may have led to the engagement of pulmonary afferents with sympatho-inhibitory properties (Seals et al., 1990). Previous studies have shown that lung inflation has an independent effect on sympathetic outflow (Eckberg et al., 1985;Seals et al., 1993). In humans, MSNA decreases during inspiration, reaching its nadir at end inspiration/early expiration, and increases during expiration, reaching its peak at end expiration . To try and obviate the confounding effect of changes in lung volume on MSNA during chemoreflex testing, the method of Keir et al. (2019) was adopted whereby MSNA bursts were identified with respect to the timing of their occurrence in the respiratory cycle, and average values were calculated only during the low-lung volume phase (i.e. last 75% of expiration, first 25% of inspiration) on a breath-by-breath basis (i.e. MSNA LLV ).
It is acknowledged that the participants we studied are of an older age (∼65 years) and that the HTN group were treated and were reasonably well controlled, and resting MSNA was not different between the groups. The rationale for studying Stage 2 HTN patients was partly because the procedures used for assessing central and peripheral chemoreflex sensitivity raise BP (by ∼20 mmHg during hyperoxic rebreathing and ∼40 mmHg during hypoxic rebreathing), which could place more severely hypertensive individuals at undue risk. MSNA is reported to be higher in human HTN (Adlan et al., 2017;Anderson et al., 1989;Esler et al., 2010;Fisher & Paton, 2012;Grassi et al., 1998), but this is not the case in all studies (Morlin et al., 1983;Wallin & Sundlof, 1979), particularly where the HTN and control groups are of an older age (Delaney et al., 2010). Further studies are required to establish how central, and combined central and peripheral  chemoreflex, sensitivities are modified in younger and/or poorly controlled HTN. Unfortunately, cardiorespiratory fitness and physical activity were not assessed in the present study. Exercise training is well established to improve a broad array of health outcomes, including cardiovascular autonomic control (Pedersen & Saltin, 2015). Notably, exercise training in rats with heart failure has been observed to reduce oxidative stress in the retrotrapezoid nucleus and lower central chemoreflex drive (Diaz-Jara et al., 2021). While such findings raise the possibility that exercise training might have therapeutic potential for lowering central chemoreflex sensitivity in clinical populations (e.g. heart failure, hypertension), they also indicate that any between-group differences in exercise training in the present study might have influenced the chemoreflex sensitivities observed. We have previously reported that young women have an augmented increase in total MSNA, but a blunted increase inV E , during central chemoreflex activation (hypercapnic hyperoxia) and combined central and peripheral chemoreflex (hypercapnic hypoxia) (Sayegh et al., 2022). This raises the possibility that there may be differences in the central chemoreflex sensitivities (and combined central and peripheral chemoreflex responses) between older NT and HTN men and women. We do not believe that our findings in normotensive young men and women influenced the conclusions of the present study, as men and women were recruited in both HTN and NT groups. However, group sizes are insufficient to make meaningful sex-difference comparisons and this is something that needs to be investigated in future studies.

Conclusions
HTN is a major cardiovascular risk factor and, despite treatment, many patients have inadequate BP control (Ettehad et al., 2016;Whelton et al., 2018). Recently, peripheral chemoreceptors have emerged as a potential therapeutic target in HTN (Narkiewicz et al., 2016;Pijacka et al., 2016) and other cardiovascular conditions such as heart failure (Marcus et al., 2014;Niewinski et al., 2013;Niewinski et al., 2017). The results of the present study demonstrate that central chemoreflex sensitivity is also augmented in human HTN. Animal studies have shown that elevated central chemoreflex sensitivity causes end-organ damage (Toledo et al., 2017), sympatho-excitation and elevates BP (Li et al., 2013;Li et al., 2016). This supports the possibility that therapeutic targeting the central chemoreflex [e.g. with orexin antagonism (Li et al., 2013;Li et al., 2016) or exercise training (Andrade et al., 2018;Diaz-Jara et al., 2021)] may potentially help some forms of HTN in humans.