Integration of respiratory chemoreflexes in humans - diversity helps leave the dogma behind.

In this issue of The Journal of Physiology, Guluzade et al. (2023) tackle a fundamental property of the respiratory system: how peripheral (both O2 and CO2 sensitive) and central (assumed only CO2 sensitive) chemoreflexes interact. Remarkably, the nature and importance of this interaction are unresolved. Data from attempts to surgically separate peripheral and central reflexes in animals are conflicting. Studies in awake dogs support a hyperadditive interaction between peripheral and central reflexes (e.g. Blain et al., 2010), whereas nearly all reduced and intact rodent experiments suggest hypoadditivity in at least one respiratory variable (i.e. frequency, tidal volume, minute ventilation ( V ̇ E ${\dot V_{\mathrm{E}}}$ ) or their neuronal equivalent; e.g. Day & Wilson, 2009). In humans, where physical separation of peripheral and central chemoreflexes has not been achieved, most data suggest an additive interaction in V ̇ E ${\dot V_{\mathrm{E}}}$ (e.g. Milloy et al., 2022). Resolving and explaining differences across species are of critical importance to understanding how we breathe. Guluzade et al. (2023) use Duffin's modified rebreathing technique to mathematically dissect out peripheral and central chemoreflexes in humans. Specifically, the authors expose participants to progressive CO2 ramps (via rebreathing) while measuring ventilation. The authors assume the CO2 ramps are sufficiently slow such that central and peripheral compartments are at equilibrium at any one moment in time. During each ramp, inspired oxygen is clamped at one of several pre-set levels, one being hyperoxia, which the authors assume specifically silences peripheral chemoreceptors without having central effects. The authors then quantify ventilation for a specific P ETC O 2 ${P_{{\mathrm{ETC}}{{\mathrm{O}}_{\mathrm{2}}}}}$ value across all the CO2 ramps (i.e. at different clamped levels of inspired oxygen). The ventilatory response attributed to peripheral oxygen sensing is then calculated by subtraction of the hyperoxic value. Plotting these values across different oxygen levels yields a line, the slope of which is taken as the ‘peripheral chemoreflex sensitivity to hypoxia’ (PChS) at a given P ETC O 2 ${P_{{\mathrm{ETC}}{{\mathrm{O}}_{\mathrm{2}}}}}$ . The authors then calculated and plot PChS values across all P ETC O 2 ${P_{{\mathrm{ETC}}{{\mathrm{O}}_{\mathrm{2}}}}}$ levels (i.e. at different points along the P ETC O 2 ${P_{{\mathrm{ETC}}{{\mathrm{O}}_{\mathrm{2}}}}}$ ramps). PChS increased with P ETC O 2 ${P_{{\mathrm{ETC}}{{\mathrm{O}}_{\mathrm{2}}}}}$ in a linear manner in at least 50% of participants, which the authors interpreted as demonstrating an additive interaction between peripheral and central chemoreflexes. In contrast, they propose that the flattening or steepening curvilinear relationships observed in other participants indicates hypoadditive or hyperadditive interactions, respectively. A major strength of Guluzade et al.’s study is their quantitative assessment of the relationship between PChS and P ETC O 2 ${P_{{\mathrm{ETC}}{{\mathrm{O}}_{\mathrm{2}}}}}$ for each participant in V ̇ E ${\dot V_{\mathrm{E}}}$ . To achieve this, the authors utilize an Akaike information criterion (AIC) approach to determine whether a linear (additive) or a curvilinear (hypoadditive or hyperadditive) model provides the most parsimonious explanation of their data. A ratio of AIC values yields the relative likelihood of one model over another, and subtracting AIC values (delta AIC) gives a ‘strength of evidence’ measure. Based on these measures, Guluzade et al. conclude that eight participants had ‘strong’ and two had ‘moderate’ evidence for an additive interaction; four – all females – had a ‘strong’ hypoadditive interaction; and one participant had a ‘strong’ hyperadditive interaction. The authors described the remaining five participants as having ‘weak’ evidence of additive interaction. These participants had delta AIC of <2 (see Table 3 in Guluzade et al. 2023). In our view, a ‘weak’ designation in some of these cases is overpresumptuous; others consider delta AIC values <2 as insufficient grounds to champion one model over another (e.g. Burnham & Anderson, 2004). Indeed, a qualitative inspection of the V ̇ E ${\dot V_{\mathrm{E}}}$ data in Guluzade et al. suggests that several of the participants with AIC ratios <2 demonstrate a flattening of the PChS– P ETC O 2 ${P_{{\mathrm{ETC}}{{\mathrm{O}}_{\mathrm{2}}}}}$ slope at higher values of P ETC O 2 ${P_{{\mathrm{ETC}}{{\mathrm{O}}_{\mathrm{2}}}}}$ , indicative of hypoadditivity (see participants 4, 12, 11 and 13 in Figure 5 in Guluzade et al. 2023). Guluzade et al. also provide PChS and P ETC O 2 ${P_{{\mathrm{ETC}}{{\mathrm{O}}_{\mathrm{2}}}}}$ plots for breathing frequency and tidal volume, but they chose not to employ AIC analysis to these respiratory variables. While V ̇ E ${\dot V_{\mathrm{E}}}$ is the most important respiratory variable for gas exchange, mathematically an additive interaction in V ̇ E ${\dot V_{\mathrm{E}}}$ must involve a non-linear increase in frequency and/or tidal volume with increasing V ̇ E ${\dot V_{\mathrm{E}}}$ . Such a non-linear increase in turn, likely necessitates a non-additive interaction in at least one of these variables. Indeed, nine participants appear to have a hypoadditive interaction in tidal volume (Figure 6 in Guluzade et al. 2023), and at least eight appear to have hypoadditive interactions in breathing frequency (Figure 7 in Guluzade et al. 2023). Thus, Guluzade et al.’s data appear to confirm the importance of additivity and hypoadditivity in chemoreflex integration. As such, Guluzade et al.’s study is important because it suggests that the current ‘one-size-fits-all’ hyperadditive dogma around chemoreceptor interaction is wrong, with hyperadditivity the exception, not the rule. Moreover, by showing diversity in chemoreceptor interactions within (and by extension) between species, Guluzade et al.’s study re-ignites the key question as to the contribution and importance of different forms of chemoreceptor interactions in respiratory control. As peripheral and central responses to inspired gases have time constants that differ by up to an order of magnitude, and ventilation is matched to metabolic rate not arterial blood gases per se, chemoreceptor interaction may be less important than respiratory physiologists assume (the ‘heresy hypothesis’; Wilson & Teppema, 2016). Alternatively, the form of interaction may be adapted to accommodate specific conditions, capabilities and/or uses of the respiratory system. For example, hyperadditive interactions in dogs may be required to accommodate panting, a special case use of the respiratory system to maintain high frequency upper airway gas flow for thermoregulation. Similarly, in humans, a predominantly additive interaction in V ̇ E ${\dot V_{\mathrm{E}}}$ might be beneficial for precise control of gas flow over the larynx required for speech. As hypoadditivity in V ̇ E ${\dot V_{\mathrm{E}}}$ was only present in females, Guluzade et al. may have provided a further clue as to the need for diversity: a hypoadditive interaction might best accommodate the lower absolute ventilatory capacity of many females. These sexually dimorphic data add to the overall importance of this outstanding study in our ongoing debate about chemoreflex interactions. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article. None. Both authors have read and approved the final version of this manuscript and agree to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. All persons designated as authors qualify for authorship, and all those who qualify for authorship are listed. This work is supported in part by a Canadian Institutes of Health Research Project Grant (RJAW, RN291964-366421) and a Natural Sciences and Engineering Research Council of Canada Discovery Grant (TAD, RGPIN-2016-04915). We wish to thank the editors and staff at The Journal of Physiology for inviting us to write this Perspective.