Adenosine signalling to astrocytes coordinates brain metabolism and function

Abstract

Brain computation performed by billions of nerve cells relies on a sufficient and uninterrupted nutrient and oxygen supply^1,2. Astrocytes, the ubiquitous glial neighbours of neurons, govern brain glucose uptake and metabolism^3,4, but the exact mechanisms of metabolic coupling between neurons and astrocytes that ensure on-demand support of neuronal energy needs are not fully understood^5,6. Here we show, using experimental in vitro and in vivo animal models, that neuronal activity-dependent metabolic activation of astrocytes is mediated by neuromodulator adenosine acting on astrocytic A2B receptors. Stimulation of A2B receptors recruits the canonical cyclic adenosine 3′,5′-monophosphate–protein kinase A signalling pathway, leading to rapid activation of astrocyte glucose metabolism and the release of lactate, which supplements the extracellular pool of readily available energy substrates. Experimental mouse models involving conditional deletion of the gene encoding A2B receptors in astrocytes showed that adenosine-mediated metabolic signalling is essential for maintaining synaptic function, especially under conditions of high energy demand or reduced energy supply. Knockdown of A2B receptor expression in astrocytes led to a major reprogramming of brain energy metabolism, prevented synaptic plasticity in the hippocampus, severely impaired recognition memory and disrupted sleep. These data identify the adenosine A2B receptor as an astrocytic sensor of neuronal activity and show that cAMP signalling in astrocytes tunes brain energy metabolism to support its fundamental functions such as sleep and memory.

Main

Brain neurons lack significant metabolic reserves and require a continuous supply of energy substrates. Astrocytes store chemical energy in the form of glycogen and respond to increases in the activity of neighbouring neurons with rapid activation of glucose metabolism^5,6. Evidence exists that metabolic coupling between neurons and astrocytes is crucial for supporting the function of neural circuits that control core behaviours^4,5,6,7,8,9. One of the defining features of metabolic activation of astrocytes is the increased production and release of lactate. Lactate supplements the extracellular pool of readily available energy substrates, and its local concentration rapidly increases in response to neuronal activity¹⁰. Significant experimental evidence suggests that the transfer of lactate from astrocytes to neurons is important for metabolic support of neuronal function^4,6,11,12. However, it is not entirely clear how exactly astrocytes monitor the metabolic needs of neighbouring neurons, and which extracellular and intracellular signalling pathways control astrocyte glucose metabolism and ensure uninterrupted supply of chemical energy to support neuronal activity.

In peripheral tissues such as the liver and muscle, increased energy expenditure rapidly recruits intracellular stores of glucose via the actions of hormones like glucagon and catecholamines, and activation of the canonical cyclic adenosine 3′,5′-monophosphate (cAMP)–protein kinase A (PKA) signalling pathway¹³. Here we show that, in the brain, the activity of the same cAMP–PKA signalling pathway in astrocytes is regulated by adenosine and plays a major role in coordinating brain energy metabolism and function.

cAMP signalling in astrocytes

Using the genetically encoded fluorescent cAMP sensor Epac-S^H187 (ref. ¹⁴) and the sensor of PKA activity AKAR4 (ref. ¹⁵) (Extended Data Fig. 1a,b), expressed under the control of glial fibrillary acidic protein (Gfap) promoter (Fig. 1a), we recorded robust increases in intracellular [cAMP] and PKA activity in astrocytes of the CA1 area of the rat hippocampus (acute and organotypic slice preparations) in response to stimulation of Schaffer collateral fibres (Fig. 1b and Extended Data Fig. 1c–e), supporting the results of previously published studies¹⁶. Blockade of glutamate receptors (10 μM CPP, 20 μM NBQX and 200 μM MCPG) did not prevent these cAMP and PKA responses (Fig. 1b and Extended Data Fig. 1e), suggesting that increased neuronal activity recruits the cAMP–PKA pathway in astrocytes via signals other than glutamate.

There is significant evidence that in the brain, increased neuronal activity is associated with the release of purines into the extracellular space^{17,18,19,20,21,22}. We next found that astrocytes in culture and brain slices respond to the purine nucleotides ATP (30 µM) and ADP (30 µM) (but not to glutamate (100 µM)) with elevations in intracellular [cAMP] and PKA activity (Fig. 1c–f and Extended Data Figs. 1a,b and 2a–e). As expected, strong [cAMP] increases in astrocytes were induced by the specific adenylyl cyclase activator forskolin (5 µM; Extended Data Fig. 2a,d). ATP-induced cAMP responses were significantly reduced by the adenylyl cyclase inhibitor SQ22536 (100 µM; Fig. 1e and Extended Data Fig. 2b,d), but were not affected by blockade of Ca²⁺ signalling (in zero Ca²⁺ and 1 µM thapsigargin), inhibition of ionotropic P2X (with PPADS (100 µM)) or metabotropic P2Y₁ (with MRS2179 (20 µM) or MRS2500 (2 µM)) receptors for ATP (Extended Data Fig. 2c,d). cAMP responses induced by a non-hydrolysable ATP analogue AMP-PNP (10 µM) were much smaller than the responses induced by ATP applied in the same concentration (Fig. 1e and Extended Data Fig. 3a). Moreover, the effects of ATP on astrocyte [cAMP] were significantly reduced or abolished by the ecto-5′-nucleotidase inhibitor α,β-methylene-ADP (200 µM), the adenosine receptor inhibitor caffeine (1 mM) or by the more specific adenosine A2 receptor antagonist ZM241385 (10 µM; Fig. 1c–f and Extended Data Fig. 3b).

Collectively, these data suggested that the effect of ATP on cAMP–PKA signalling in astrocytes is indirect, independent of intracellular Ca²⁺ signalling, and mediated by adenosine formed in the extracellular space following ATP breakdown by ectonucleotidase activity²⁰. Indeed, adenosine (1–10 µM) triggered robust increases in intracellular [cAMP] and PKA activity in astrocytes (often leading to saturation of the Epac-S^H187 and AKAR4 sensors when applied at higher concentrations) (Fig. 1c,e and Extended Data Fig. 3c,d). The A2 receptor antagonist ZM241385 (10 µM) completely blocked the effects of adenosine (Fig. 1c,e and Extended Data Fig. 3c). Furthermore, adenosine-induced PKA responses were prevented by the PKA inhibitor H89 (10 µM; Extended Data Fig. 3c). In slice experiments, we observed that application of ZM241385 led to a significant reduction of the Epac-S^H187 sensor signal (Fig. 1d,f), suggesting that the adenosine A2 receptor-mediated cAMP signalling pathway in astrocytes is constitutively active.

Adenosine A2B receptors in astrocytes

Vertebrates express two G_s protein-coupled adenosine receptors that stimulate adenylyl cyclase and activate cAMP signalling pathways. These are A2A and A2B receptors, encoded by two paralogous genes: Adora2a and Adora2b²³. Analysis of single-cell RNA sequencing (RNA-seq) data from the mouse brain²⁴ demonstrated strong and specific expression of Adora2b in astrocytes (Fig. 2a,b), consistent with the recently published evidence placing the A2B receptor among the most highly expressed G protein-coupled receptors in astrocytes from all regions of the brain²⁵. The RNA-seq data suggested that astrocytes do not express Adora2a (Fig. 2b). This conclusion was further supported by the results of the experiments showing that adenosine-induced cAMP responses in astrocytes were effectively blocked by the selective A2B antagonist PSB 603 (10 µM), but were unaffected by the A2A antagonist SCH442416 (10 µM; Fig. 2f and Extended Data Fig. 3d). Moreover, the effects of adenosine on intracellular [cAMP] in astrocytes were mimicked by the A2B receptor agonist BAY 60-6583 (0.1–1 µM), whereas the A2A agonists CGS21680 (1 µM) and PSB 0777 (1 µM) had no effect (Fig. 2f and Extended Data Fig. 3e). In astrocytes of floxed Adora2b mice (Adora2b^flox/flox)²⁶ transduced with AAV5-Gfap-iCre-mCherry vector to express the improved Cre (iCre) recombinase and delete A2B receptors (Fig. 2c), adenosine (up to 100 µM) had no effect on intracellular [cAMP] and PKA activity, whereas cAMP and PKA responses induced by direct adenylyl cyclase activation with forskolin (10 µM) were unaffected (Fig. 2d–g).

We next found that increases in intracellular [cAMP] and PKA activity in CA1 astrocytes evoked by stimulation of Schaffer collateral fibres were blocked by inhibition of A2 receptors with ZM241385 or upon genetic deletion of A2B receptors (Fig. 2h,i). Collectively, these results suggested that adenosine acting via A2B receptors mediates synaptic activity-dependent recruitment of cAMP–PKA signalling in astrocytes (Fig. 2j). We next sought to investigate the significance of this signalling pathway in the regulation of astrocyte glucose metabolism.

Adenosine signalling and brain metabolism

Using the genetically encoded fluorescent sensor of glucose FLIP¹²glu-700μΔ6 (ref. ²⁷) and the sensor of cytosolic NADH–NAD⁺ redox state Peredox²⁸, we recorded robust increases in glucose consumption and glycolytic rate in astrocytes in response to ATP and adenosine (Fig. 3a–d and Extended Data Fig. 4a). The A2B receptor agonist BAY 60-6583 (1 µM) increased glucose consumption and glycolytic rate in astrocytes and this effect was prevented by A2B receptor deletion (Fig. 3a and Extended Data Fig. 4b). Increases in astrocyte glucose consumption and the rate of glycolysis would be expected to facilitate the production of lactate and its release into the extracellular space. We next used enzymatic microelectrode biosensors to record the release of lactate in acute slices of the rat brain. Both ATP and adenosine triggered significant lactate release (Extended Data Fig. 5a,b). ATP-induced lactate release was not affected by the ATP receptor antagonist PPADS (100 µM), but was blocked by ZM241385 (10 µM; Extended Data Fig. 5b). The effect of adenosine on lactate release was mimicked by the A2B receptor agonist BAY 60-6583 (Extended Data Fig. 5f) and blocked by ZM241385 (10 µM), PKA inhibition with H89 (10 µM) or inhibition of glycogen metabolism with 1,4-dideoxy-1,4-imino-D-arabinitol (DAB; 200 µM; Extended Data Fig. 5b). The A1 adenosine receptor antagonist DPCPX (1 µM) had no effect on adenosine-induced lactate release (Extended Data Fig. 5b).