The ‘connexin’ between astrocytes, ATP and central respiratory chemoreception

Breathing in vertebrates is a remarkable behaviour. It begins in utero, and continues virtually uninterrupted from birth until death, automatically adjusting its activity over orders of magnitude to take in oxygen and expel CO₂ in proportion to metabolic rate. Homeostatic control of breathing is achieved by feedback loops involving chemical receptors in the peripheral circulation and the central nervous system (CNS) that regulate O₂, and CO₂ or pH. Central chemoreception refers specifically to the feedback process whereby changes in brain CO₂ (or pH) bring about adaptive (homeostatic) changes in breathing to maintain arterial CO₂ (or pH) near steady-state levels. The dominant theory of central respiratory chemoreception is that CO₂ works indirectly through its effect on pH, which stimulates breathing through the simultaneous activation of numerous types of acid-sensitive CNS neurons located at multiple sites including the retrotrapezoid nucleus (RTN), medullary raphe, nucleus tractus solitarii and locus coeruleus. Central chemoreceptors, in turn, detect pH via a cell-specific combination of several, as yet unidentified, acid-sensitive ion channels (Corcoran et al. 2009; Nattie & Li, 2009; Guyenet et al. 2010).

Despite the appeal of this general scheme, many uncertainties remain. In companion papers in a recent issue of The Journal of Physiology, Huckstepp et al. (2010a,b); provide surprising new information that will not only change ideas about mechanisms of central chemoreception and the regulation of breathing, but provide novel insight into the potential roles of ATP, connexin hemichannels and astrocytes in brain function. In 2005, this same group demonstrated that CO₂-evoked release of ATP from chemosensitive regions of the ventral medullary surface contributes to central respiratory chemoreception (Gourine et al. 2005). Despite numerous studies showing that respiratory networks and neurons are sensitive to ATP, this was the first to establish a physiological role for ATP in central respiratory control. The present study extends this work by showing that elevations of CO₂, independent of pH changes, evoke ATP release from the ventral medullary surface of 400 μm thick horizontal slices at sites overlying two putative chemoreceptive sites, the medullary raphe and RTN (Huckstepp et al. 2010b). The sensitivity of this release to a variety of connexin blockers (which lack perfect selectivity) and its lack of dependence on Ca²⁺ implicates connexin 26 (Cx26) hemichannels as the path of ATP release. Subsequent analysis of CO₂-dependent dye uptake in relation to genetically tagged Cx26-expressing cells, and specific markers for astrocytes and neurons, revealed that Cx26 is not expressed by neurons, but concentrated to the glia limitans of the ventral medullary surface and penetrating blood vessels. Most importantly, block of Cx26 hemichannel opening in vivo dramatically reduced ATP release and attenuated the hypercapnic ventilatory response by ∼25%.

These data are significant on three levels. First, they challenge the general view of central chemoreception outlined above (Corcoran et al. 2009; Nattie & Li, 2009; Guyenet et al. 2010) by showing (i) that astrocytes of the glia limitans are likely to be chemosensory cells (i.e. neurons are not the only sensory cells), (ii) that Cx26 hemichannels, rather than ion channels, can act as molecular sensors of brain acid–base status, and (iii) that CO₂, rather than pH, is a sufficient stimulus. Data, however, also emphasize that CO₂-sensitive, Cx26-mediated astrocytic ATP release is not the only mechanism underlying central chemoreception. Many acid-sensitive ion channels have been identified as candidate pH sensors but none has yet been defined as responsible for central chemosensitivity. There are also many candidate chemosensory neurons. The best characterized among these are the Phox2B RTN neurons. Raphe neurons have been extensively characterized in culture, but there is debate about their role in vivo (Corcoran et al. 2009; Guyenet et al. 2010). Another point of debate is whether central respiratory chemoreception derives from a diffuse but weak pH sensitivity of multiple sites that converge to regulate the activity of the central respiratory controller (Nattie & Li, 2009) or from one (or a few) specialized cell group, such as the RTN, that acts as an integration centre for relevant chemosensory information (Guyenet et al. 2010). Data presented here by Huckstepp et al. (2010b) do not address this issue and are consistent with both possibilities. The hypothesized mechanism (Fig. 1) is that CO₂ causes the release of ATP, which acts via P2 receptors to increase the excitability of local chemosensitive RTN neurons, which directly innervate the respiratory controller, or chemosensitive raphe neurons that activate the respiratory controller directly or indirectly through RTN neurons. The ionic mechanisms are not known, but chemosensitive RTN neurons are sensitive to exogenous ATP (Mulkey et al. 2006), and are excited by ATP released from channel rhodopsin 2-expressing astrocytes following their selective photostimulation (Gourine et al. 2010). The greater level of ATP release over the medullary raphe may suggest a greater contribution to the CO₂-evoked, Cx26-, ATP-mediated component of the respiratory chemosensory response, but this could as easily reflect greater ectonucleotidase activity (and ATP breakdown) in the RTN.

Figure 1. Schematic illustration of the hypothetical mechanism(s) underlying the contribution of astrocytes and CO₂-sensitive, Cx26 hemichannel-mediated ATP release to central respiratory chemoreception.

Elevations in CO₂ (intracellular or extracellular) in regions of the ventral medullary surface immediately ventral to the RTN and the caudal medullary raphe evoke release of ATP (Gourine et al. 2005; Huckstepp et al. 2010b) from astrocytes through Cx26 hemichannels that are gated by CO₂ (i.e. Cx26 hemichannels act as the CO₂ sensor; Huckstepp et al. 2010a). (Note, this is unlikely to be the only mechanism of ATP release from astrocytes; Gourine et al. 2010.) Within the RTN, ATP excites a population of what are most likely to be H⁺-sensitive neurons through a P2Y G-protein coupled receptor-dependent mechanism (Mulkey et al. 2006) that modulates either an unknown membrane conductance or the acid-sensitive ion channels directly (Guyenet et al. 2010). The H⁺ sensors in RTN neurons may be K⁺ channels that are directly sensitive to intra- or extracellular acidification. Increased output from the RTN to the ventral respiratory column (VRC) causes ventilation to increase (Guyenet et al. 2010). At the level of the medullary raphe, Cx26-mediated release of ATP from astrocytes is similarly hypothesized to excite H⁺-sensitive raphe neurons (Huckstepp et al. 2010b), but the ATP sensitivity of chemosensitive raphe neurons remains to be established, as does the identity of the H⁺ sensor. Increased output from raphe neurons is hypothesized to increase breathing either through direct connections with the VRC or indirectly through excitatory connections with RTN neurons (Guyenet et al. 2010; Corcoran et al. 2009). These ATP-dependent processes mediate ∼25% of the central chemosensory response (Huckstepp et al. 2010b).

Second, these data provide evidence of a physiological role for connexin hemichannels in signalling and modulation of network activity. Hemichannel gating was initially observed under non-physiological conditions and its relevance was questioned. Huckstepp et al. (2010a), however, heterologously express various connexins in HeLa cells to demonstrate via three independent methods (whole-cell recording; CO₂-dependent dye-uptake; inside-out and outside-out membrane patch recordings) that expression of Cx26 (and related β-connexins to a lesser degree) is sufficient to bestow on Hela cells CO₂ sensitivity and the capacity for CO₂-evoked ATP release. The implications of these data are very significant. They suggest that connexins can act as sensors monitoring the extracellular environment. Moreoever, they demonstrate that gating of Cx26 by CO₂ modulates respiratory network activity, which adds to an extremely small database supporting the physiological relevance of hemichannel gating. Future work will need to resolve the significance for central chemoreception of the competing effects of CO₂ and pH on Cx26 gating. The authors clearly demonstrate CO₂-evoked ATP release via Cx26 hemichannels. However, they also show that reductions in pH counteract this ATP release. It will also be important to resolve the relative roles of Cx26 mediated ATP release described here compared to a second, Ca²⁺-dependent, vesicular mechanism of astrocytic-ATP release (Gourine et al. 2010).

Third, data suggest a role for astrocytes and gliotransmission in modulating the activity of respiratory networks. The historical view that astrocytes provide only structural and nutritive support to neurons has changed dramatically over the last decade. There is now plenty of evidence in culture that astrocytes can signal to neurons and influence their activity, but the significance of these observations for functioning networks is less certain. The demonstration here in vivo that astrocytes on the ventral surface of the medulla can respond to a physiological stimulus (CO₂) with a Cx26 hemichannel-mediated release of ATP that contributes to a homeostatic increase in respiratory network activity is therefore very significant.

In summary (see Fig. 1), data suggest that central chemoreception derives in part from CO₂-sensitive Cx26 hemichannels that provide a conduit for release of the gliotransmitter ATP from astrocytes on the ventral medullary surface, i.e. data support a physiological role for connexin hemichannels, astrocytes and ATP in central chemoreception. The strength of these conclusions derives from the rigorous demonstration that Cx26 expression is limited to astrocytes, followed by careful analysis of Cx26 gating properties. Such delimited protein expression is rare. The fact that astrocytes and neurons express many of the same channels and receptors, combined with the lack of tools to selectively manipulate astrocytes, represents a major challenge to understanding the role of glia in information processing. Optogenetic approaches are extremely powerful and are informing how activation of astrocytes influences network function, but selective inhibition of astrocytes remains a challenge. Selective manipulation of connexin hemichannels is similarly required to further explore the exciting possibility raised here that they act as sensors, monitoring the extracellular environment.