AMPK facilitates the hypoxic ventilatory response through non-adrenergic mechanisms at the brainstem

We recently demonstrated that the hypoxic ventilatory response (HVR) is facilitated by the AMP-activated protein kinase (AMPK) in catecholaminergic neural networks that likely lie downstream of the carotid bodies within the caudal brainstem. Here, we further subcategorise the neurons involved, by cross-comparison of mice in which the genes encoding the AMPK-α1 (Prkaa1) and AMPK-α2 (Prkaa2) catalytic subunits were deleted in catecholaminergic (TH-Cre) or adrenergic (PNMT-Cre) neurons. As expected, the HVR was markedly attenuated in mice with AMPK-α1/α2 deletion in catecholaminergic neurons, but surprisingly was modestly augmented in mice with AMPK-α1/α2 deletion in adrenergic neurons when compared against a variety of controls (TH-Cre, PNMT-Cre, AMPK-α1/α2 floxed). Moreover, AMPK-α1/α2 deletion in catecholaminergic neurons precipitated marked hypoventilation and apnoea during poikilocapnic hypoxia, relative to controls, while mice with AMPK-α1/α2 deletion in adrenergic neurons entered relative hyperventilation with reduced apnoea frequency and duration. We conclude, therefore, that AMPK-dependent modulation of non-adrenergic networks may facilitate increases in ventilatory drive that shape the classical HVR, whereas AMPK-dependent modulation of adrenergic networks may provide some form of negative feedback or inhibitory input to moderate HVR, which could, for example, protect against hyperventilation-induced hypocapnia and respiratory alkalosis. Supplementary Information The online version contains supplementary material available at 10.1007/s00424-022-02713-8.


Introduction
We recently identified a role for the AMP-activated protein kinase (AMPK) in regulating breathing and thus oxygen supply [11,37] by facilitating the hypoxic ventilatory response (HVR) [38]. Classically, AMPK has been regarded as a cellular energy sensor that acts to maintain energy homeostasis in a cell-autonomous manner. Cells may express as many as twelve AMPK isoforms through the heterotrimeric associations of one from each of the 2 α-catalytic and 2 β-and 3 γ-regulatory subunits. AMPK is coupled to mitochondrial oxidative phosphorylation, and thus oxygen supply, by two discrete albeit cooperative pathways. Binding of AMP to the AMPK γ-subunit increases activity tenfold by allosteric action, while AMP or ADP binding delivers increases in LKB1-dependent phosphorylation and reductions in dephosphorylation of Thr172 on the α-subunit that confer 100-fold further activation. All of these effects are inhibited by ATP [13]. There is also an alternative calcium-dependent pathway to AMPK activation that is governed by the calmodulin-dependent protein kinase CaMKK2, which delivers increases in Thr172 phosphorylation and thus AMPK activation independent of changes in cellular AM(D)P:ATP ratios. Importantly, each AMPK isoform may hold different sensitivities to activation by increases in cellular AMP and ADP and differ in their capacity to directly phosphorylate and thus regulate downstream targets [27].
Our studies suggest that the role of AMPK is not limited to cellular metabolic homeostasis but extends to breathing and thus oxygen [34,36] and energy (ATP) supply to the body as a whole. However, in this respect, AMPK does not appear to act at the level of the carotid bodies as one would predict given their role as the primary arterial chemoreceptors [14,25,29,34], but downstream and most likely at the caudal brainstem [23]. Located within the caudal portions of the brainstem are the noradrenergic A1 and A2 cell groups, which receive peripheral carotid body afferent inputs and relay these signals to the respiratory central pattern generators (rCPG) [17,30,34]. The rostral portion of the brainstem on the other hand contains the adrenergic (phenylethanolamine N-methyltransferase (PNMT) expressing) C1, C2, and C3 cell groups, which also contribute to sympathoexcitatory responses and the regulation of breathing [15]. The dorsal C2 and C3 groups are located bilaterally and medially, respectively, in rostral regions of the NTS, while the C1 group is located bilaterally in the ventral medulla where the majority of brainstem PNMT-positive neurons are located [24]. Of these, it is evident that C1 neurons innervate, among other sites, the respiratory centres of the brainstem, sympathetic preganglionic neurons in the spinal cord, and the dorsal motor nucleus of the vagus (10 N) [16]. Furthermore, optogenetic stimulation of C1 cells has been shown to increase breathing frequencies in conscious mice [2] and rats [5], mostly through direct projections to the retrotrapezoid nucleus (RTN), which in turn modulates respiratory output through the rCPG [1,5,6].
Our previous investigations into the role of AMPK in facilitating the HVR relied upon a gene deletion strategy that targeted all catecholaminergic neurons through Cre recombinase expression via the tyrosine hydroxylase promoter [23,38]. To gain further insight, the present study compares outcomes for these mice with those for mice in which AMPK deletion was targeted to adrenergic neurons through Cre expression via the gene that encodes PNMT (Pnmt). We show here that while AMPK-α1/α2 deletion in catecholaminergic neurons precipitated marked hypoventilation and apnoea during poikilocapnic hypoxia, AMPK deletion in adrenergic neurons modestly augments the HVR and reduces apnoea frequency. In short, AMPK facilitates the HVR through non-adrenergic mechanisms.

Mouse models
Experiments were performed in accordance with the regulations of the United Kingdom Animals (Scientific Procedures) Act of 1986. All studies and breeding were approved by the University of Edinburgh and performed under UK Home Office project licenses. For the main study described here, both male and female mice were used. All were on a C57/Bl6 background, with the exception of a secondary control experiment on mice developed on a CD1 background . Numbers of mice (≥ 3 per measure) used are indicated for each experiment. Global, dual knockout of the genes encoding AMPK-α1 (Prkaa1) and AMPK-α2 (Prkaa2) is embryonic lethal. We therefore employed conditional deletion of the genes for the AMPK-α1 and AMPK-α2 subunits, using mice in which the sequence encoding the catalytic site of both α-subunits was flanked by loxP sequences [21]. To direct AMPK deletion to adrenergic cells, we crossed these 'floxed' mice with a transgenic mouse model in which the sequence encoding Cre recombinase was inserted into the 5′ UTR of the phenylethanolamine N-methyltransferase (Pnmt) gene (a kind gift from Prof Steven Ebert, University of Central Florida) [10]. Mice with AMPK deletion in catecholaminergic cells were obtained as described previously [38] by crossing floxed mice with mice in which Cre recombinase expression was driven via the tyrosine hydroxylase (TH) promoter [22].

Plethysmography
As described previously [34,36], we used a whole-body unrestrained plethysmograph, incorporating a Halcyon™ low-noise pneumotachograph (Buxco Research Systems, UK) coupled to FinePointe acquisition and analysis software (Data Science International, USA). Following acclimation and baseline measurements (awake but quiet, undisturbed periods of breathing) under normoxia (room air), mice were exposed to hypoxia (8% O 2 , with 0.05% CO 2 , balanced with N 2 ) for 10 min. The FinePointe software automatically calculated the respiratory parameters assessed after application of exclusion criteria due to non-ventilatory artefacts (movement, sniffing, etc.). Data were acquired as 2-s averages, and 2 to 4 data points of undisturbed breathing were selected for each time point of the HVR. Apnoeas were defined as a period of cessation of breathing that was greater than the average duration, including inter-breath interval, of 2 successive breaths (~ 600 ms) of control mice during normoxia with a detection threshold of 0.25 mmHg (SD of noise).

Confocal imaging
To identify neuronal networks in which dual AMPK-α1/α2 deletion had been induced, mice with PNMT-Cre-driven AMPK-α1/α2 deletion and TH-Cre-driven AMPK-α1/α2 deletion were crossed with mice engineered for Cre-dependent expression of Rosa (tdTomato). These mice were deeply anaesthetised using 2 g/kg Pentobarbital Sodium (Merial), transcardially perfused with ice-cold heparinised saline and fixed with 4% paraformaldehyde in 0.1 M phosphate buffer (PB; pH 7.4). Brains were extracted, post-fixed, and stored in 30% sucrose in 0.1 M PB at 4 °C. Thirty-micrometre sections of the brainstem were cut using a Frigomobil freezing microtome (Leica). Alternate sections were collected together and mounted onto glass slides, briefly air-dried, and coverslipped using Vision™ PermaFluor™ Aqueous Mounting Medium (Thermo Fisher).
Brainstem sections were imaged using a Nikon A1R + confocal system and tdTomato autofluorescence detected using an excitation wavelength of 554 nm and emission wavelength of 581 nm. Relevant regions of the caudal brainstem harbouring catecholaminergic neurons were identified using the mouse brain atlas [39].

Statistical analysis
Statistical comparison was completed using GraphPad Prism 6. For plethysmography, one-way ANOVA with Tukey post hoc test was used when comparing across one variable (e.g. genotype) and two-way ANOVA with Sidak post hoc tests when comparing across two variables (e.g. genotype and time). Statistical significance was assumed when p ≤ 0.05.

Results
The original cohort of PNMT-Cre mice we received was bred on a CD1 background, whereas the TH-Cre and AMPK-α1/α2 floxed mice were bred on a C57Bl6 background. Therefore, the CD1 PNMT-Cre mice were backcrossed for at least four generations onto a C57Bl6 background, to eliminate genetic differences between background strains that influence the hypoxic ventilatory response (HVR) (Supplementary Figs. 2-5; see also [3,19,32,33,35]). We then selected for AMPK-α1/α2 deletion in adrenergic cells [21] by heterozygous PNMT-Cre expression, which is sufficient for cell-specific excision of AMPK-α1/α2 [38], and thus avoided loss of PNMT expression and consequent adrenaline deficiency that is known to be associated with homozygous insertion of Cre at exon 1 of Pnmt [10].
Cell-specific gene deletion was confirmed previously for TH-positive cells by single-cell endpoint reverse transcription-polymerase chain reaction and whole-brain quantitative polymerase chain reaction, and restriction of Cre to TH-positive cells in the adult mouse was verified by viral transfection of a Cre-inducible vector carrying a reporter gene [38]. Here, we additionally crossed TH-AMPK-α1/α2 knockouts and PNMT-AMPK-α1/α2 knockouts with mice expressing the Cre-inducible reporter gene Rosa (tdTomato), the expression of which was assessed by confocal imaging to identify cells. In accordance with the discrete distribution of adrenergic C1, C2, and C3 neurons within the brainstem when compared to nor-adrenergic A1 and A2 neurons [15,17,20], qualitative comparison of thirty-micrometre brain sections revealed ( Fig. 1) marked regional variations in Rosa expression between PNMT-AMPK-α1/α2 knockouts and TH-AMPK-α1/α2 knockouts at, for example, the area postrema (AP) and across proximal regions of the dorsal nucleus tractus solitarius (NTS), as one would expect given the distribution of adrenergic and noradrenergic neurons at the brainstem [20].
When taken together, the aforementioned findings are in agreement with the view that the augmenting phase of the HVR is primarily mediated by carotid body afferent input responses [8,34,36], while direct modulation by hypoxia of brainstem respiratory networks aids maintenance of the HVR in the longer term [7,23,28,34,36].
It would appear, therefore, that AMPK-dependent signalling pathways of noradrenergic networks facilitate the HVR and thus oppose ventilatory instability, hypoventilation, and apnoea during hypoxia, and in a manner that receives some form of negative feedback/input consequent to AMPKdependent modulation of adrenergic networks.

Discussion
Our previous investigations demonstrated that AMPK deletion in catecholaminergic cells attenuated the HVR and precipitated severe hypoventilation and apnoea, rather than hyperventilation [38]. By contrast, we have shown here that deletion of AMPK-α1/α2 in adrenergic cells moderately Fig. 6 AMPK deletion in adrenergic cells reduces and AMPK deletion in catecholaminergic cells increases apnoeas during exposures to severe hypoxia. Means ± SEM of the a apnoeic index (min −1 ), b apnoea duration (ms), and c apnoea-duration index (frequency x duration) for PNMT-Cre mice (blue, n = 32 exposures from 8 mice), PNMT Cre-driven AMPK-α1/α2 double knockout mice (PNMT-AMPK-α1/α2 dKO, purple, n = 32 exposures from 8 mice), AMPK-α1/α2 floxed mice (black, n = 58 exposures from 21 mice), and TH-Cre-driven AMPK-α1/α2 double knockout mice (TH-AMPK-α1/α2 knockout, red, n = 49 exposures from 21 mice). * = p < 0.05, **** = p < 0.0001 augmented the HVR, further stabilised respiratory rhythm, and reduced apnoea frequency during 10-min exposures to severe hypoxia. This indicates differential roles for AMPKdependent modulation of adrenergic inputs and other catecholaminergic (i.e. dopaminergic and/or noradrenergic) inputs in regulating the HVR and thus whole-body oxygen and energy supply. It seems plausible, given reductions in apnoea frequency following AMPK deletion in adrenergic cells, that AMPK-dependent modulation of non-adrenergic catecholaminergic signalling pathways facilitates the HVR and protects against apnoea during hypoxia by enhancing ventilation, whereas AMPK-dependent modulation of adrenergic signalling pathways may, if anything, moderate the HVR in a manner that also contributes to the prolongation of apnoea, at least during the onset of hypoxia.
Neither AMPK-α1/α2 deletion in catecholaminergic nor adrenergic neurons markedly affected the augmenting phase of the HVR, which is consistent with the generally held view that this phase of the HVR is primarily mediated by carotid body afferent input responses [8,34,36]. Strikingly, however, the HVR of TH-AMPK-α1/α2 knockout mice was markedly attenuated throughout the sustained phase of exposures to severe poikilocapnic hypoxia, whereas the HVR of PNMT-AMPK-α1/α2 knockouts was moderately augmented. In each case, this was due to changes in breathing frequency, tidal volume, and thus minute ventilation relative to controls. When considered alongside our previous demonstration that AMPK deletion in catecholaminergic neurons attenuated the HVR where carotid body afferent inputs are normal, this finding adds further support to the view that direct modulation by hypoxia of brainstem respiratory networks aids coordination of the HVR in the longer term [7,23,28,34,36]. Nevertheless, that AMPK deletion in adrenergic neurons augmented the HVR was surprising given that the adrenergic ventrolateral C1 neurons have been proposed to form an integral part of the neural circuit that drives increases in breathing frequency during hypoxia in mice and rats [2,5]. Considering this, our findings may point to the provision of an AMPK-dependent inhibitory input from the C2 and/or C3 neurons to the rCPGs, either directly or indirectly via moderation of afferent NTS inputs. Alternatively, AMPK may act to limit activity of C1 neurons during hypoxia, which is in line with the finding that AMPK slows action potential firing frequency in hippocampal neurons by direct phosphorylation and activation of K V 2.1 [18].
Because AMPK deletion in all catecholaminergic cell types precipitated severe hypoventilation and thus revealed AMPK-dependent positive inputs, it therefore seems plausible that inhibitory adrenergic inputs may act to limit and/ or periodically attenuate respiration triggered by hypoxic hyperventilation facilitated by AMPK-dependent regulation of noradrenergic and/or dopaminergic inputs, perhaps to limit respiratory alkalosis and aid efficient management of physiological CO 2 levels [4,12]. It is evident that attenuation and thus moderation of the HVR can be conferred through direct silencing of RTN neurons by respiratory alkalosis leading to consequent reductions in breathing frequency and tidal volume [4], although it should be noted that optogenetic inhibition of the RTN during hypoxia did not lead to further reductions in breathing frequency unless the hypoxic stimulus was supplemented with 3% CO 2 or pharmacological agents that led to blood re-acidification [4]. Therefore, a contribution to those mechanisms that oppose respiratory alkalosis through AMPK-dependent regulation of adrenergic neurons is not beyond the bounds of possibility, given that the RTN can be inhibited through the release of inhibitory neuropeptides (such as enkephalin or neuropeptide Y) or signalling pathways involving GABAergic interneurons, that could be coordinated by the wider adrenergic network [26,31].
Critically, AMPK deletion in adrenergic neurons also decreased the variability of inter-breath intervals during the first 3 min of the hypoxic exposures, and reduced apnoea frequency. Conversely one would expect that AMPK-dependent regulation of adrenergic pathways during hypoxia might ordinarily act to oppose the HVR and facilitate apnoea, which would lead to periodic cessations of ventilation that have been proposed to be the most rapid and efficient way to restore central CO 2 levels during periods of intense respiratory effort [9]. Thus, during hypoxia, the balance of positive inspiratory drive from the carotid bodies and AMPKdependent activation of brainstem hypoxia-responsive networks of noradrenergic/dopaminergic neurons on the one hand and AMPK-dependent activation of inhibitory inputs from adrenergic neurons on the other hand may coordinate HVR in a manner that aids maintenance of both oxygen supply and physiological CO 2 levels.
In summary, AMPK facilitates the HVR through nonadrenergic catecholaminergic mechanisms. Whereas AMPKdependent modulation of noradrenergic and/or dopaminergic networks may facilitate increases in ventilatory drive that shape the classical HVR, AMPK-dependent modulation of adrenergic networks may provide some form of negative feedback or inhibitory input to moderate HVR, which may protect against hyperventilation-induced hypocapnia and respiratory alkalosis [4,12]. However, further studies will be required to determine whether this is indeed the case.
Funding This research was funded in whole, or in part, by the Wellcome Trust (WT081195MA). For the purpose of open access, the author has applied a CC BY public copyright licence to any authoraccepted manuscript version arising from this submission. SM was also supported by a University of Edinburgh PhD studentship and then by a BHF Programme Grant held by AME (RG/12/14/29885).
Data availability All available data are included in this manuscript.

Declarations
Ethics approval and consent to participate Ethical approval was given by the Ethics Committee, College of Medicine, University of Edinburgh.
Human and Animal Ethics All experimental procedures were covered by the UK Home Office Project Licence PBA4DCF9D.

Competing interests The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.