Measuring Peripheral Chemoreflex Hypersensitivity in Heart Failure

Heart failure with reduced ejection fraction (HFrEF) induces chronic sympathetic activation. This disturbance is a consequence of both compensatory reflex disinhibition in response to lower cardiac output and patient-specific activation of one or more excitatory stimuli. The result is the net adrenergic output that exceeds homeostatic need, which compromises cardiac, renal, and vascular function and foreshortens lifespan. One such sympatho-excitatory mechanism, evident in ~40–45% of those with HFrEF, is the augmentation of carotid (peripheral) chemoreflex ventilatory and sympathetic responsiveness to reductions in arterial oxygen tension and acidosis. Recognition of the contribution of increased chemoreflex gain to the pathophysiology of HFrEF and to patients’ prognosis has focused attention on targeting the carotid body to attenuate sympathetic drive, alleviate heart failure symptoms, and prolong life. The current challenge is to identify those patients most likely to benefit from such interventions. Two assumptions underlying contemporary test protocols are that the ventilatory response to acute hypoxic exposure quantifies accurately peripheral chemoreflex sensitivity and that the unmeasured sympathetic response mirrors the determined ventilatory response. This Perspective questions both assumptions, illustrates the limitations of conventional transient hypoxic tests for assessing peripheral chemoreflex sensitivity and demonstrates how a modified rebreathing test capable of comprehensively quantifying both the ventilatory and sympathoneural efferent responses to peripheral chemoreflex perturbation, including their sensitivities and recruitment thresholds, can better identify individuals most likely to benefit from carotid body intervention.

2 inter-individual differences in peripheral chemoreflex gain. The first of these reasons is that the proximity of PCO2 to the PCO2 at the peripheral chemoreflex VRT (variable 3) is not standardized.
It is often assumed that resting PETCO2 is equivalent to the PCO2 at the VRT for the peripheral chemoreflex, but this is not always the case (Mahamed et al., 2001). Thus, it is impossible to know whether the magnitude of the measured HVR reflects peripheral chemoreflex sensitivity (variable 2) or a combination of this and the amount by which isocapnic PCO2 exceed the VRT (variable 3).
Second, because the HVR varies with PCO2 above the VRT, at least two HVR tests at different levels of isocapnia (both above the PCO2 associated with the VRT) are needed to ascertain peripheral respiratory chemoreflex sensitivityone cannot measure sensitivity without constructing a slope! By applying two such tests, the change in HVR for the change in PCO2 provides a slope reflective of peripheral chemoreflex sensitivity (variable 2) while controlling for variables 1 and 3. But why measure the effect of PCO2 on the ventilatory response to PO2 with multiple tests and with unknown VRT when one can measure the effect of PO2 on the ventilatory response to PCO2 in a single protocol?
The modified rebreathing protocol does just that: the ventilatory response to the progressive rise in PCO2 via rebreathing is measured in standardized iso-oxic low and high PO2 conditions. The high PO2 condition provides the central chemoreflex response to PCO2 and the low PO2 condition gives the response of both central and peripheral respiratory chemoreflexes.
The difference in response slopes above the VRT between conditions provides the peripheral chemoreflex sensitivity to CO2. Importantly, this measure of peripheral chemoreflex sensitivity is derived under standardized PO2 conditions and is not affected by between-individual differences in VRTs. Initiating rebreathing from a hypocapnic state after a period of slow deep breathing also permits identification of the VRTs for the central (hyperoxic condition) and peripheral (hypoxic 3 condition) chemoreflexes (although caution should be taken when translating the PCO2 at the VRT from rebreathing to steady-state experiments (Mohan et al., 1999)).
The modified rebreathing test, therefore, characterizes the CO2 stimulus-response characteristics of both central and peripheral respiratory chemoreflexes and has the distinct advantage of providing a means by which they can be compared between individuals and within individuals (Jensen et al., 2010) before and after specific interventions (Fan et al., 2010). Since the rebreathing responses describe the central and peripheral chemoreflex characteristics, these can be incorporated into models of the control of breathing (Mahamed et al., 2001;Duffin, 2010).
Several assumptions underly the modified breathing protocol and the interpretation of the measured ventilatory responses. The subsequent section will list each assumption and, briefly, present evidence for their validity. In addition, common critiques of the protocol will be addressed.

Hyperoxia silences the peripheral chemoreflex response to CO2
The relationship between the ventilatory response to PCO2 and PaO2 is well described by a rectangular hyperbola with minimal rise until PaO2 falls below ~85 mmHg (Lloyd et al., 1957;Duffin et al., 2000). As isoxic PO2 increases much above 150 mmHg, further reduction in the CO2 ventilatory response slope becomes insignificant (Lloyd et al., 1957;Mohan and Duffin, 1997;Duffin et al., 2000). Both in vitro and in vivo experiments demonstrate that a PO2 of 150 mmHg nearly abolishes Type I cell excitation (Fitzgerald and Parks, 1971;Duchen and Biscoe, 1992;Buckler and Vaughan-jones, 1994;Dasso et al., 2000) and carotid sinus afferent discharge (Hornbein and Roos, 1963;Lahiri and Delaney, 1975;Lahiri et al., 1993;Vidruk et al., 2001). It is possible that, in humans, afferent output from the carotid bodies may still exist in hyperoxia (albeit with a discharge frequency that is largely diminished), but importantly, this afferent input 4 seems insufficient to generate a reflex response to CO2. Thus, in our experience, the ventilatory response to PCO2 during hyperoxic rebreathing reflects reflex responses initiated by the central chemoreceptors.

Ventilation is controlled by both central and peripheral chemoreflexes and their drives are additive
Anesthetized animal preparations that isolate and control perfusate at the carotid body and medulla suggest that the central and peripheral respiratory chemoreflexes interact hypo-additively in rats (Day and Wilson, 2007) and hyper-additively in canines (Blain et al., 2010;Smith et al., 2015), respectively. In humans the central and peripheral drives to breathe are additive.
In a conscious resting human, there is a ~10 mmHg difference between arterial (~40 mmHg) and brain (~50 mmHg) PCO2. Although brain PCO2 is, in part, determined by arterial PCO2, differences in local metabolism and blood flow can cause the arterial-brain PCO2 difference to widen or narrow dynamically in response to a given steady-state CO2 stimulus (note that these dynamics are irrelevant in modified rebreathing where the arterial-venous PCO2 difference is minimized and PCO2 rises similarly at the carotid body and medulla; see below). In humans, the speed of ventilatory responsiveness to step-changes in PO2 and PCO2 has been used to test central and peripheral chemoreflex interactions. In response to a step-increase in PETCO2, arterial PCO2 increases instantaneously whereas central PCO2 takes longer to adjust due to slow central tissue compartment kinetics (Farhi and Rahn, 1960). Therefore, the rapid response to a step-decrement in PO2 or increment in PCO2 is thought to be mediated almost entirely by the peripheral chemoreceptors.
Adopting this approach, the bulk of human evidence suggests that the central and peripheral chemoreflexes do not interact and their inputs are additive. For example, Cui et al., (2012) showed that the magnitude of the ventilatory response to a step-decrement in PO2 was the same in the presence of low (prior hyperventilation) and high (prior isocapnic hypercapnia) central PCO2.
Using a similar temporal separation technique, other studies have failed to provide evidence that peripheral chemoreflex ventilatory responses are modulated by central chemoreceptor stimulation (Clement et al., 1992(Clement et al., , 1995St Croix et al., 1996).
Most importantly, isoxic hypoxic rebreathing exposes both central and peripheral chemoreflexes to the same 'PCO2 ramp' stimulus. In response to such simultaneous stimulation, the net ventilatory response (reflecting contributions from both reflex arcs) is consistently linear (Duffin et al., 2000;Preston et al., 2008;Jensen et al., 2010;Keir et al., 2019), and incompatible with a hypo-or hyper-additive response (Duffin and Mateika, 2013) Two other studies are often cited as evidence for a central-peripheral respiratory chemoreflex interaction in humans: Dahan et al., (2007) reported that the "slow" ventilatory response to multifrequency binary sequence changes in PETCO2 (with gain attributable to the central chemoreflex) was reduced after the "fast" ventilatory response (with gain attributable to the peripheral chemoreflex) was abolished by bilateral carotid body resection in three individuals. However, a major limitation of the study is the use of end-tidal forcing to control central PCO2. With end-tidal forcing the central PCO2 is amenable to any cerebrovascular and cardiovascular changes that might alter the arterial-brain PCO2 gradient. Unfortunately, the authors did not track changes in cardiovascular or cerebrovascular dynamics and therefore, it is impossible to tell whether temporal reductions in the magnitude of the ventilatory response to the same step increment in PETCO2 was related to absence of a hyper-additive input from the peripheral chemoreceptors or because their absence facilitated 6 greater cerebral blood flow changes (perhaps by a reduction in sympathetic outflow) with stepincrements in PETCO2 such that central PCO2 increased to a lesser extent. Teppema et al., (2010)

Arterial and medullary chemoreceptor CO2 are equal during rebreathing
Experimental testing of this assumption is difficult to obtain in humans. The initial PCO2 in the rebreathing bag is approximately equal to that in venous circulation (~35 mmHg). The transition from hyperventilation to rebreathing simply stops excretion of CO2 at the lungs so that PCO2 in the arterial and venous circulation rise in unison at a rate determined by metabolic production of CO2 and body CO2 stores. Mathematical simulations based on compartment models (for example the original test model by Read and Leigh, (1967)) showed that, after the initial equilibration to mixed venous values, the PCO2 in all compartments rise together, mixed by the circulation. Any PCO2 difference between compartments is therefore due to circulatory time delays. Assuming a circulatory time delay of ~10 s from the central compartment to the lung compartment and a rate of rise of CO2 during rebreathing of about 0.07 mmHg/s, a difference of 0.7 mmHg may be expected but this would be reduced as CO2 increases cerebral blood flow.

Prior voluntary hyperventilation produces short-term potentiation that will alter ventilatory responses to CO2 independent of the chemoreflexes
It is important to note that the hyperventilation prior to rebreathing is of a slow, deep pattern akin to that used in meditation and therefore is more likely to induce relaxation rather than excitation. The effect of short-term potentiation has been studied. Chatha and Duffin, (1997) found no short-term potentiation during rebreathing and Mohan et al. (1999) demonstrated that the ventilatory response to hyperoxic rebreathing was the same with and without prior hyperventilation. Other studies support that most healthy individuals do not display hyperpnea after ~12 s post-hyperventilation (Mahamed et al., 2004) and even apnea has been reported for some subjects rather than short-term potentiation (Skatrud and Dempsey, 1983;Meah and Gardner, 1994;Mateika and Ellythy, 2003). In most individuals performing modified rebreathing, the onset of the linear rise in ventilation occurs ~1.5 minutes after rebreathing onset in hypoxia (the PCO2 threshold is further delayed in hyperoxia). Any hyperpnea, if present, is not included in the model parameter estimates either by excluding these data or by fitting the data below VRT an exponential decay model (e.g. (Jensen et al., 2010)). Therefore, even if a brief period of hyperpnea is present, it is unlikely to influence the baseline, VRT or slope parameters describing the respiratory chemoreflexes. Furthermore, to our knowledge, there is no evidence that short-term potentiation has any effect other than raising basal ventilation and thus, if present, it is unlikely that the PCO2 VRT or ventilatory responsiveness to PCO2 are affected. 8

Hyperoxia will stimulate ventilation independent of its effect on CO2 responsiveness
This statement suggests that the ventilatory response to iso-oxic hyperoxic rebreathing will reflect the central respiratory chemoreflex response plus a non-chemoreflex mediated response to high PO2. The findings of Becker et al., (1996) often are cited to support this assertion. With respect to the Becker data, it should be emphasized that the increase in ventilation with 30% O2 breathing likely was related to central chemoreceptor stimulation by PCO2. Those experiments were performed in steady-state isocapnic conditions, where central PCO2 would be expected to rise progressively (relative to arterial PCO2) with hyperoxia due to the Haldane effect (Eldridge and Kiley, 1987). When their experiment was repeated under poikilocapnic conditions the hyperoxic ventilatory response disappeared.
Unlike steady-state experiments, rebreathing uncouples hyperoxia from its affects on central PCO2 and thereby the central chemoreflex drive to ventilation is isolated. Arterial and medullary PCO2 are equivalent and rise progressively and synchronously with time (Read and Leigh, 1967).

PaCO2 does not accurately reflect the central chemoreceptor PCO2 across subjects who differ in their cerebral blood flow sensitivities to PO2 and/or PCO2
Individual-specific differences in cerebral blood flow sensitivities to PO2 and PCO2 are only relevant in steady-state experiments and are irrelevant during rebreathing. During rebreathing, arterial-venous PCO2 differences are minimal. Any rise in cerebral blood flow with CO2 (or hypoxia) will contribute to a further reduction in the arterial to central PCO2 difference.

Hypoxic ventilatory decline will affect the ventilatory response to PCO2 in hypoxia
The isocapnic hypoxic ventilatory response in steady-state conditions can arbitrarily divided into two phases with an immediate increase (0-5 minutes) followed by a slow "hypoxic ventilatory decline (HVD)" (5-20 minutes) (Steinback and Poulin, 2007). While HVD is an important consideration in steady-state experiments, it is not significant in modified rebreathing where isoxic hypoxic tests are completed in less than 4 minutes.

Summary
In a normal resting human, there is an ~10 mmHg difference between arterial (~40 mmHg) and brain (~50 mmHg) PCO2. Although brain PCO2 is, in part, determined by arterial PCO2 (and PO2) differences in local metabolism and blood flow can cause the arterial-brain PCO2 difference to widen or narrow dynamically in response to a given CO2 stimulus. By minimizing (or eliminating) this difference, modified rebreathing is the only method that is capable of separating central from peripheral respiratory chemoreflexes in humans to confidently quantify their independent reflex responsiveness and, from the clinical perspective, identify with greater precision those heart failure patients most likely to benefit from carotid body interventions.