Stroke ranks among the leading causes of death and disability worldwide. With 7.6 million new cases in 2019, ischemic stroke accounted for 62% of all strokes, and 3.3 million patients died from this widespread disease [1]. At present, treatment options of acute ischemic stroke are limited to intravenous (i.v.) thrombolysis and mechanical thrombectomy both aiming at reperfusion of the clotted vessel. Due to strict inclusion criteria, these therapies are only available for a subset of patients, and many patients present with persistent neurological deficits despite successful recanalization. Therefore, additional treatment strategies to combat ischemic brain injury are urgently needed. Research over the last three decades identified the substantial contribution of sterile inflammation to secondary neuronal injury following ischemic stroke. Consequently, targeting secondary infarct growth mediated by infiltrating immune cells emerged as a promising therapeutic strategy. The inflammatory cascade is initiated by the excessive release of danger-/damage-associated molecular patterns (DAMPs) which activate pattern recognition receptors on microglia and astrocytes with subsequent cytokine and chemokine production and immune cell attraction from the periphery [2]. Among these DAMPs, the purine adenosine triphosphate (ATP) plays a pivotal role. Geoffrey Burnstock pioneered the concept that ATP released by cells acts as an extracellular signaling molecule and thus exerts key functions beyond energy supply [3].

Under physiological conditions, cells within a tissue, including the brain, release ATP and nicotinamide adenine dinucleotide (NAD+) into the interstitial space in an activity-dependent and regulated manner. Release mechanisms include exocytosis, connexin or pannexin hemichannels, ATP-binding cassette (ABC) transporters, calcium homeostasis modulator (CALHM) channels, a macropore after P2X7 receptor overstimulation, maxi-anion channels (MACs), and volume-regulated ion channels (VRACs) as comprehensively reviewed by Giuliani et al. [4]. At sites of inflammation or ischemia, ATP is further liberated in an uncontrolled fashion via cell damage and death [4]. Indeed, an increase in extracellular ATP concentrations within the brain parenchyma during in vivo ischemia was measured by microdialysis [5, 6]. To improve the accuracy of extracellular ATP sensing and to circumvent methodological challenges of purine measurements within intact tissues, the group of Francesco Di Virgilio engineered a chimeric plasma membrane luciferase (pmeLUC) [7]. PmeLUC emits a measurable bioluminescent signal upon D-luciferin administration in the presence of extracellular ATP. Using this in vivo ATP biosensor, ATP release could be visualized as early as 30 min and up to 24 h after transient middle cerebral artery occlusion (tMCAO) [8]. After being released, ATP can either signal through its ionotropic P2X and metabotropic P2Y receptors or experience degradation to adenosine by ectoenzymes. The present review will give an overview about the different players of purinergic signaling, their expression in mice and humans (Fig. 1), and their contribution to tissue damage and regeneration in ischemic stroke. Details on the cited experimental stroke studies using mice with global genetic deficiency for the different players of purinergic signaling can be found in STab. 1.

Gene expression data heatmap of 26 enzymes and receptors (y axis) involved in purinergic signaling in 37 conserved cell types (x axis) between humans (left) and mice (right). All enzymes and receptors are listed with gene names. For human single-nucleus RNA-sequencing data (SMART-SEQ) of the middle temporal gyrus (MTG), 15,928 nuclei from 8 human tissue donors were analyzed [10] (Allen Institute for Brain Science. Allen Brain Map, Cell Types Database, RNA-Seq Data. Available from: https://celltypes.brain-map.org/rnaseq/human/mtg). For the mouse cortex single-cell RNA-sequencing dataset (SMART-SEQ) 14,249 cells from the primary visual cortex (V1) and 9573 cells from the anterior lateral motor cortex (ALM) were profiled [11] (Allen Institute for Brain Science. Allen Brain Map, Cell Types Database, RNA-Seq Data. Available from: https://celltypes.brain-map.org/rnaseq/mouse/v1-alm). Expression levels were quantified as log2-transformed counts per million (CPM) of intronic plus exonic reads. Homologous cell types between the two species [10] can be divided into 5 non-neuronal cells (astrocytes, microglia and perivascular macrophages (PVM), oligodendrocytes (Oligo), oligodendrocyte precursor cells (OPC), and endothelial cells), 12 excitatory neuron (Exc) types named after the originating brain lamina (L) and their projection targets (intratelencephalic (IT), extratelencephalic-pyramidal tract (ET), near-projecting (NP), corticothalamic (CT)) and 20 inhibitory neuron cell types. Inhibitory neurons are further subdivided into lysosomal associated membrane protein family member 5 (Lamp5), somatostatin (Sst), parvalbumin (Pvalb), vasoactive intestinal polypeptide (Vip), and paired box 6 (Pax6) expressing cells and chandelier neurons

Extracellular ATP serves as ligand for all purinergic receptors of the ionotropic P2X family and for several P2Y receptors, while its metabolite ADP binds exclusively to P2Y receptors. The P2X family comprises seven members (P2X1-P2X7). Three P2X receptor subunits form an ion channel that upon ATP-mediated gating conducts the rapid influx of calcium and sodium ions and efflux of potassium ions. The expression of P2X receptors has been ascribed to various brain-resident cells as well as to cells of the immune system [9] with P2X1, P2X4, and P2X7 being associated with the outcome of cerebral ischemic events. In the central nervous system (CNS) of both mice and humans, microglia show the highest expression of these three receptors [10,11,12,13] (Fig. 1).

Of importance in the context of ischemic stroke, P2X1 is expressed on platelets. ATP (10 μM) increases platelet aggregation in vitro, which can be prevented by the P2X1 antagonist NF449 [14]. P2X1 is also expressed by neutrophil granulocytes of the mouse, where its activation promotes neutrophil chemotaxis [15]. Pharmacological blockade of P2X1 and genetic deletion of P2rx1 in mice diminished the activation and recruitment of neutrophils to inflamed arteriolar endothelial cells, accompanied by impaired platelet aggregation and fibrin generation [16]. Therefore, targeting of P2X1 could be a promising new approach to ameliorate neutrophil driven post-stroke inflammation as well as stroke-associated thromboinflammation [17]. Apart from immune cells, P2X1 is also expressed on vascular smooth muscle cells (VSMCs) of rat middle cerebral arteries [18]. It has been shown that P2X1 and P2X4 form heterotrimers in VSMCs, which, in contrast to P2X1 homotrimers, contribute to ATP-mediated vasoconstriction. The impact of P2X1 signaling on VSMCs in the context of stroke still needs to be investigated.

ATP released during cerebral ischemia triggers P2X4 opening on brain innate immune cells such as microglia or infiltrating monocytes/macrophages and sustained P2X4 activation contributes to the ischemic injury [19, 20]. Among immune cells, P2X4 is expressed on T cells promoting early T cell activation in concert with P2X7 [21] as well as chemotaxis [22]. Indeed, P2rx4 deficiency in mice led to smaller infarcts after tMCAO [23]. Interestingly, when specifically deleting P2rx4 in myeloid cells using LysM-Cre mice, only female mice showed a reduction in infarct size, suggesting a sex difference in the P2X4 response after stroke [23]. Another experimental stroke study by the same group verified the detrimental role of P2X4 in brain ischemia by using the P2X4 antagonist 5-BDBD: When applied 4 hours (h) after tMCAO, 5-BDBD reduced infarct size, number of infiltrating pro-inflammatory myeloid cells, and blood-brain-barrier (BBB) disruption [24]. 5-BDBD was also shown to attenuate brain injury in a mouse model of intracerebral hemorrhage, significantly reducing brain edema, BBB disruption, neuronal cell death, and neurological deficits [25]. Another study demonstrated that P2X4 is also expressed by pyramidal neurons of the rat hippocampal CA1 region [26]. In line with tMCAO data, intracerebroventricular (i.c.v.) 5-BDBD treatment immediately after cerebral ischemia (four-vessel occlusion model) and once 24 h later was also protective by blocking neuronal apoptosis, thereby promoting neuronal survival. Similar results were obtained with the P2X4 antagonist PSB-12062 [26]. Given the central role of P2X4 in inflammation, cell migration, and also pain, new P2X4 antagonists are continuously developed. The group of Kenneth A. Jacobson recently described the generation and evaluation of MRS4719 and MRS4596, two potent new P2X4 antagonists (IC50 0.503 and 1.38 μM for human P2X4, respectively) with neuroprotective effects in murine ischemic stroke (tMCAO): MRS4719 dose dependently reduced infarct size and brain atrophy 3 and 35 days post-stroke [27]. First P2X4-specific antagonistic monoclonal antibodies have been developed by AstraZeneca [28]. The affinity-optimized clone IgG#151-LO showed high selectivity for human P2X4 and induced potent and complete blockage of P2X4 currents. Inhibition of spinal P2X4 either by intrathecal delivery of anti-P2X4 antibodies or by systemic delivery of an anti-P2X4 bispecific antibody with enhanced blood–spinal cord barrier permeability achieved analgesia for 7 days in a mouse model of neuropathic pain [28]. Further, single domain antibodies (nanobodies) against P2X4 have been developed, but did not show P2X4-antagonizing properties [29]. In principle, evaluation of new biological-based P2X4-targeting drugs in the setting of brain ischemia would be desirable, as these usually exhibit good pharmacological properties in vivo and are well suited for cell-specific targeting of P2X4. Both factors might be important for the successful translation into treating stroke patients. Two studies highlight the need for a differentiated approach regarding timing and cell tropism: Sustained absence of P2X4 signaling, as occurring in P2X4ko mice, reduced infarct size after tMCAO, but myeloid cell-specific P2X4 deficiency led to more depressive behavior 30 days after stroke onset compared to recovering wild-type (WT) mice [23]. As another example of potentially detrimental effects of P2X4 blockade, brain endothelial cells acquire ischemic tolerance in a P2X4-dependent fashion as 5-BDBD treatment 15 min before ischemic preconditioning prevents this tolerization [30]. Therefore, biologics-based spatiotemporal targeting of P2X4 could be the key to future therapies focusing on P2X4 in stroke. However, crossing the BBB still remains a challenge for most antibody-based constructs, an obstacle that needs to be overcome by modern antibody engineering technology [31].

P2X7 is most likely the most extensively studied P2X receptor in brain ischemia. P2X7 activation is known as a potent trigger for the NLRP3 inflammasome formation in macrophages and microglia, leading to the release of pro-inflammatory interleukin-1β (IL-1β), a powerful driver of post-stroke inflammation [32,33,34]. Yet, deciphering the complex role of P2X7 during the interlinked processes of post-stroke inflammation, BBB breakdown and neuronal recovery remains challenging.

First studies by Nancy Rothwell’s group did not show a difference in stroke size when comparing P2rx7-deficient and WT mice [35], and i.c.v. treatment with P2X antagonists oxidized ATP and PPADS immediately before and again 30 minutes (min) after tMCAO did not affect lesion size after tMCAO either [36]. However, other studies demonstrated that intraperitoneal treatment with the broad P2 antagonist Reactive Blue 2 5 min after intraluminal permanent MCAO (pMCAO) in rats reduced brain damage and that MCAO led to an upregulation of P2X7 on microglia [37]. P2X7 upregulation had already been observed by earlier studies with hypertensive rats subjected to pMCAO by electrocoagulation [38]. Other studies suggested a protective role of P2X7 in stroke: I.c.v. injection of the P2X7-specific agonist BzATP into rats 1 h after tMCAO improved their motor functions when compared to mock-injected controls [39]. Studies using Brilliant Blue G (BBG) as P2X7 antagonist reported different outcomes regarding the role of P2X7 in rodent stroke models: The group of Carlos Matute reported that intraperitoneal BBG treatment 30 min after ischemia onset reduced the lesion size in rats subjected to tMCAO [40] and later reproduced their findings in mice [41]. The study by Caglayan and colleagues in 2017 again confirmed that i.c.v. injection of BBG reduced lesion size in mice when injected 30 min before tMCAO [42]. Similarly, intraperitoneal BBG administration also turned out to be protective in murine photothrombotic stroke when applied 1, 3, and 5 days after ischemia [43]. In contrast, a study, in which BBG was injected intraperitoneally (i.p.) daily for 3 days immediately after a short (15 min) tMCAO in mice, did not demonstrate a difference in lesion size between treatment and control group [44].

As for P2X7 targeting compounds, contrasting results have also been reported in studies using P2rx7-deficient mice. As mentioned above, first stroke experiments with P2X7ko mice did not reveal differences in stroke size [35]. Later in 2015, the group of Carlos Matute reported smaller stroke lesions (tMCAO) in P2X7ko mice when compared to WT [41]. In contrast, the group of Michael Schäfer reported reduced edema formation in P2X7ko mice 3 days after tMCAO, while stroke lesion size did not differ between P2X7ko and WT mice [45]. A recent study published by our group again suggested a detrimental role of P2X7 in stroke: Mice overexpressing P2X7 in cells that naturally express P2X7 [46] exhibited larger infarcts compared to WT controls. Further, i.c.v. injection of P2X7-blocking nanobodies [47] in mice directly before tMCAO significantly reduced lesion size compared to the control group [8]. This study represents the first approach to use highly specific biologicals to block P2X7 in the context of stroke. As mentioned above, the BBB constitutes a major obstacle for i.v. application of antibody-based drugs targeting brain-resident cells like microglia. Even for the small P2X7-specific nanobodies only high doses beyond 1 mg lead to detectable amounts of the nanobodies on brain microglia [48]. Interestingly, an alternative approach based on adeno-associated virus (AAV)-mediated muscle cell transduction followed by in vivo production of P2X7 antagonistic nanobodies led to the successful diffusion of the nanobodies into the brain where they could be detected on microglia for up to 120 days after transduction [48]. Since the role of P2X7 in long-term recovery after stroke still remains elusive, the AAV approach could be utilized to investigate this technically challenging question of high translational relevance.

When using antibody or nanobody-based therapeutics, one has to monitor for potential immune-related adverse events (irAE). Antibody-induced inflammation via complement activation or activation of natural killer (NK) cells and macrophages can be prevented by modifying the Fc-region of antibodies. Nanobody monomers lack Fc regions in principle but can be tailored as Fc-fusion proteins with desired or no Fc-related functions [49, 50]. To date, no adverse effects of P2X7 blockade by antibodies or nanobodies have been described. However, one has to keep in mind that P2X7 blockade can also potentially alter the susceptibility towards infections. Indeed, the P2X7 receptor antagonism can have both beneficial and deleterious effects depending on the type of pathogen, its virulence, and the severity of infection as reviewed in [51].

In general, though various studies suggest a role of P2X7 in stroke, many P2X7-related pathways and mechanisms that participate in post-stroke inflammation, thromboinflammation, BBB breakdown, neuroprotection and recovery remain to be revealed and connected. Contrasting results may be the consequence of (1) variations in the specificity of applied P2X7 antagonists, (2) different stroke models that do or do not allow reperfusion or that are based on photothrombosis, (3) differences related to administration routes or pre- versus post-treatment, or (4) the use of P2rx7-deficient mice of different origin. For the latter aspect, differences between the GlaxoSmithKline (GSK) and the Pfizer P2X7ko mice have already been demonstrated [52]: GSK P2X7ko mice lack P2X7 expression on innate immune cells but exhibit robust expression of P2X7 on T cells due to an escape splice variant. In contrast, the Pfizer P2X7ko mice lack P2X7 in both immune cell compartments. These differences might explain the opposite outcome of experimental autoimmune encephalomyelitis (EAE) studies: GSK P2X7ko mice showed suppressed EAE development compared to WT mice [53], whereas Pfizer P2X7ko mice exhibited an exacerbated disease course in EAE when compared to WT mice [54]. GSK P2X7ko mice have not been investigated in murine stroke models yet. Additionally, one has to keep in mind that the Pfizer P2X7ko mice are congenic and therefore might contain passenger mutations from 129 embryonic stem cells used for P2rx7 gene targeting. Indeed, our group could identify a P2rx7 passenger mutation in the commonly used P2X4ko mouse [55, 56]. Since P2rx4 and P2rx7 are direct neighboring genes, differentially expressed genes in congenic P2X4ko might also occur in congenic P2X7ko mice. If such passenger mutations also exist in the widely used Pfizer P2X7ko mouse [57] still needs to be investigated, in particular if they affect the outcome of stroke.

Of note, the ATP-gated P2X receptors expressed in the brain differ greatly in their sensitivity to ATP. As measured in transfected X. laevis oocytes or HEK293 cells by electrophysiology, the ATP concentration to induce the half-maximal response (K1/2) is about ten times lower for rat P2X1 (0.82 μM) than for rat P2X4 (11 μM). Rat P2X7, in turn, has a K1/2 of 130 μM and is accordingly about ten times less sensitive to ATP than P2X4 [58]. Thus, the activation of P2X receptors in the brain does not only depend on the expression pattern of the receptors, but also on the availability of extracellular ATP.

P2Y receptors are G-protein-coupled receptors activated by ATP, ADP, and other naturally occurring nucleoside phosphates. The most abundant P2Y receptors in the brain are P2Y12 and P2Y13, expressed at high levels in microglial cells (Fig. 1). The natural ligand for these two receptors is ADP, which upon engagement inhibits adenylate cyclase leading to a decrease in cyclic AMP (cAMP) and subsequent activation of microglia. P2Y12 receptor inhibitors, such as clopidogrel and ticagrelor, are commonly used for the secondary prevention of atherosclerotic ischemic stroke because of their effective reduction of P2Y12-mediated platelet aggregation. However, the beneficial effects of these inhibitors are not limited to this anti-thrombotic effect, since they are also neuroprotective in murine tMCAO (oral administration from 24 h before to 8 h after occlusion) [59], anti-atherosclerotic [60], and can reduce microglial activation and chemotaxis after pMCAO in rats (oral treatment 10 min, 22 and 36 h after occlusion) [61]. Interestingly, physiological concentrations of ADP facilitate the contact of microglia processes with neuronal cell bodies. After neuronal injury, higher concentrations of ADP triggered microglial-mediated neuroprotection in a P2Y12 receptor-dependent manner, sparing viable neurons from cell death and maintaining functional connectivity [62]. These data pose a warning to the treatment with P2Y12 inhibitors for stroke prevention and prompt the analysis of neuronal function in MCAO models under P2Y12 inhibitor treatment.

ADP can also activate P2Y1 receptors, expressed in astrocytes, and is a partial agonist for the P2Y6 receptor, expressed by microglial cells. While pre- and post-stroke blockade of P2Y1 by MRS2500 i.c.v. proved beneficial for stroke outcome in transient and permanent MCAO models [63, 64], the role of P2Y6 in stroke is still controversial: Pharmacological inhibition of P2Y6-mediated phagocytosis with MRS2578 on 3 consecutive days after tMCAO aggravated infarct size, brain atrophy, and neurological deficits [65]. However, in a model of brain ischemia induced by endothelin-1 injection, knockdown of the P2Y6 receptor had a beneficial effect by delaying phagocytosis of stressed neurons in the peri-infarct region [66]. P2Y2 and P2Y11 receptors are not expressed on brain-resident cells but are present on immune cells that can be attracted to the brain after ischemia-reperfusion injury. Indeed, ATP signaling through P2Y2 and P2Y11 promotes neutrophil and monocyte migration [67,68,69]. There