Introduction
Stroke induces sterile inflammation, which worsens the initial brain damage and neurological outcome [
1,
2]. Hypoxic brain tissue releases many molecules, which can activate cells such as microglia in the surrounding tissue and lead to infiltration of other immune cells such as neutrophils, amplifying the inflammatory cascade [
3]. These molecules include adenosine triphosphate (ATP) as well as nicotinamide adenine dinucleotide (NAD), heat shock protein (HSP), and high-mobility group box 1 protein (HMGB1). These factors can activate the inflammasome and induce the secretion of proinflammatory cytokines by innate immune cells [
4,
5]. These molecules activate several pathways, such as the ATP/P2X7 pathway or the nuclear factor kappa-light-chain-enhancer of activated B cells (NFκB) pathway [
6]. The P2X7 receptor is a homotrimeric, ligand-gated nonselective cation channel that is expressed in the central nervous system as well as on immune cells [
7]. The P2X7 receptor consists of three polypeptide subunits, each with two transmembrane domains [
8,
9]. After activation by extracellular ATP (eATP), these subunits form an ion-permeable channel, which induces Na
+ and Ca
2+ influx and K
+ efflux, resulting in plasma membrane depolarization and initiation of Ca
2+ signaling cascades. The K
+ efflux through the P2X7 receptor supports the formation of the Nod-like receptor protein 3 (NLRP3)-mediated inflammasome complex, which cleaves pro-caspase 1 and leads to a subsequent cleavage of pro-IL-1β and pro-IL-18 into their biologically active forms [
5,
10,
11]. The amount of accessible intracellular pro-IL-1β and pro-IL-18 also depends on another signal transmitted by receptors, such as Toll-like receptors (TLRs) or tumor necrosis factor (TNF)-receptors, and subsequent NFκB activation.
In the central nervous system (CNS), P2X7 has been found primarily on microglia, with less on astrocytes and oligodendrocytes [
12‐
17]. These findings were confirmed by data from the Allen Brain Atlas for mice [
18] and humans [
19]. There are some similarities between human and rodent P2X7 expression in the brain, such as high expression on microglia and low expression on astrocytes, but there are also some differences such as high expression of P2X7 on human oligodendrocytes and low expression on rodent oligodendrocytes.
Several studies have shown that the experimental stroke size in P2X7
−/− mice is smaller than that in wild-type mice [
20,
21]. In addition, blocking the P2X7 channel with brilliant blue G (BBG) attenuated ischemic damage [
20]. However, systemic BBG cannot be used in humans since it is nonspecific and toxic.
Nanobodies (nbs), named for their small size (2.5 nm diameter, 4 nm height, 12 kDa) [
22], are single-domain antibodies derived from camelid heavy chain antibodies. Compared to small molecule inhibitors, nbs have key advantages, such as low toxicity, high specificity, no off-target effects and, in the case of P2X7, a more potent inhibition [
10,
21,
23]. With their long complementarity determining region 3 (CDR3), these molecules can access cavities or clefts on membrane proteins that are often inaccessible to antibodies [
24,
25]. Other advantages of nbs over conventional antibodies include high stability, better solubility and rapid and targetable in vivo biodistribution. In addition, the ability to form nb multimers and the low costs and ease of production make them ideal candidates for treatment [
26]. Fusion of an nb (monomer or multimer) to the Fc domain of a conventional antibody yields a heavy chain antibody with reconstituted Fc-mediated effector functions, including binding to Fc receptors, extended half-life and complement activation. This phenomenon allows a much broader tailoring of nbs than of conventional antibodies to different pathophysiologies [
27].
In this proof-of-concept study, we used P2X7-specific nbs to treat mice directly before temporary middle cerebral artery occlusion (tMCAO) surgery. We found that these nbs need to be injected intracerebroventricularly to reach P2X7 receptor on brain resident cells and protect against ischemic stroke.
Methods
Animals
All animal experiments were approved by the local animal care committees (Behörde für Justiz und Veterinärwesen Hamburg, Nr 006/18) and conducted following the “Guide of the Care and Use of Laboratory Animals” published by the US National Institutes of Health (NIH Publication No. 83–123, revised 1996). All mice were kept at a constant temperature of 22 ± 2 °C with a 12-h light–dark cycle and ad libitum access to food and water. Only 12- to 18-week-old male mice were used for this study. C57BL/6J mice were purchased from Charles River (Bar Harbor, ME 04609, USA), whereas the generation of pmeLUC transgenic and P2X7-EGFP transgenic mice (line 17 in C57BL/6J) was described previously [
28,
29].
Production of P2X7 nbs
The P2X7-antagonizing nbs 1c81 and 13A7 were selected and cloned into the pCSE2.5 expression vector (kindly provided by Thomas Schirrmann, Braunschweig, Germany) [
30] as described previously [
10,
28]. Then, 13A7 was fused to the hinge, constant domain heavy chain (CH) 2, and CH3 domains of mouse immunoglobulin (Ig) G2c, resulting in a heavy chain format (nb A), whereas 1c81 was dimerized and fused to the albumin-specific nb Alb8 (mAb77) [
31], resulting in a bispecific heterotrimeric nb with an extended half-life (nb B) (Additional file
1: Fig. S1). Since dimers showed a higher potency than monomers [
10], we used nb B for intracerebroventricular (icv) injection. For icv injection, we needed to create a construct that could be highly concentrated without aggregation, so we modified our nb B (Additional file
1: Fig. S1). The modified nb B-mod was concentrated up to 15 μg/μl without aggregation. The exact sequences and further information on the various constructs can be found in patent WO/2013/178783.
HEK-6E cells were transfected with the constructs, and 6 days after transfection, the nbs were purified from the cell supernatant by affinity chromatography on a protein-G Sepharose column. The buffer was exchanged by gel filtration on a PD-10 column. The concentration and purity were monitored by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS-PAGE) and a BCA™ Protein Assay Kit (Pierce).
tMCAO surgery and stroke size analysis
tMCAO was performed as previously described [
32‐
34]. Mice were anesthetized with 1.5% isoflurane in 100% O
2 and an intraperitoneal injection of 0.05 mg/kg body weight buprenorphine in saline was used as analgesic. A midline skin incision in the neck was made before ligating the proximal common carotid artery (CCA) and the external carotid artery (ECA) without disrupting the venous vessels. Vital parameters were continuously monitored with PhysioSuite (Kent Scientific Corporation, USA). Occlusion was confirmed by a laser Doppler monitor (moorVMS-LDF; Moor Instruments, UK) and persisted for 40 min. Mice with an occlusion rate of less than 80% were excluded.
Stroke size was measured by triphenyl tetrazolium chloride (TTC) staining and magnetic resonance imaging (MRI). We used a 7-Tesla MR small animal imaging system (ClinScan, Bruker, Ettlingen, Germany). The imaging protocol comprised T2-weighted imaging MRI. Calculation of corrected stroke volumes was performed as described previously [
35].
The infarct volumes and total areas of the treated hemisphere were calculated using NIH ImageJ software.
Intravenous (iv) and icv injections of nbs
Different methods of nb administration were used. P2X7-specific nbs (nb A, 13A7-Fc) were directly injected (100 µg in 100 µl of phosphate-buffered saline [PBS]) intravenously, or P2X7-specific nbs (nb B/nb B-mod, 1c81-dim-HLE) or isotype nbs against human cluster-of-differentiation (CD) 38 were injected (30 µg in 2 µl of PBS containing 60 mg/ml trehalose and 0.4 mg/ml Tween-20) directly into the ventricles of the brain by using a stereotaxic apparatus. Mice were pain treated with 1 mg tramadol/kg body weight one day before surgery. Directly before the surgery, the mice were anesthetized with isoflurane (4% for induction, 2.5% for maintenance) in 100% oxygen. After placing the mice in a stereotactic frame (Stoelting, 51615), we made a 1-cm-long incision above the midline. A cranial burr hole (0.9 mm) was drilled 1.1 mm lateral and 0.5 mm posterior to the bregma. Nbs were drawn into a 10-μl Hamilton syringe (Hamilton, 1701RN) connected to a 26-gauge needle (Hamilton, 26G, Point Style 4, 12°) controlled by a motorized stereotaxic injector (Stoelting, integrated stereotaxic injector [ISI]).
The needle was slowly introduced 2.3 mm deep into the left ventricle (Additional file
1: Fig. S2). Following a period of 5 min to let the ventricular system re-expand, 2 μl of dissolved nbs at a concentration of 15 μg/μl was injected at 1 μl/min. This step was followed by another 10-min break and slow removal of the needle. Vital parameters were monitored by an animal support unit (Minerve, Esternay, France). Body temperature was maintained throughout the procedure at 37 °C using a feedback-controlled heating device.
In vivo ATP measurement after tMCAO using pmeLUC-TG
Three hours before tMCAO surgery, 150 mg/kg luciferin (Promega) was injected intraperitoneally. Luciferin was reinjected 1 day after tMCAO in prior of the measurement. In vivo ATP release was monitored by whole-body luminometry performed using the IVIS-Perkin Elmer in vivo imaging system. In vitro calibration was performed in brain homogenates from pmeLUC-tg mice.
Microglia and macrophage preparation and FACS
Animals were euthanized and perfused with PBS. Brains were dissected and digested in 1 mg/ml collagenase A (Roche) and 0.1 mg/ml DNase type I (Sigma). Separation from myelin and debris was performed by density centrifugation with Percoll (GE Healthcare). The following antibodies and detection systems were used: CD45-APC-Cy7 (1:100, 30-F11, #103,115 BioLegend), CD45-PerCP (1:100, 30-F11, #103,129, BioLegend), CD11b-APC (1:100, M1/70, #17–0112-82, eBioscience), Ly6C-PerCP/Cy5.5 (1:100, HK1.4, #128,011, BioLegend), anti-mIgG1-brilliant violet (BV) 421 (1:100, RMG1-1, #406,615, BioLegend), anti-mIgG2-BV421 (1:100, RM223, #31–1103-02, Dianova), steptavidin-BV421 (1:100, #405,226, BioLegend), Fc blocking anti-CD16/CD32 (1:100, 2.4G2, #BE0307, BioXcell), and mAb77 (1:100, Alb8-specific mouse monoclonal antibody kindly provided by Ablynx). Microglia were gated as mentioned in the supplementary materials (Additional file
1: Fig. S3). In the first step, cells were incubated (30 min on ice) with Fc blocking anti-CD16/32, where ex vivo samples were incubated with 0.5 μg of P2X7 specific nb in the presence of Fc blocking anti-CD16/32. For detection of cell-bound P2X7 nbs, cells were incubated either with biotinylated anti-mouse IgG2c-fused antibody followed by streptavidin BV421 conjugated (nb A) or with mAb77 (nb B-mod) followed by fluorochrome-conjugated antibodies in the presence of Fc blocking anti-CD16/CD32 (Additional file
1: Fig. S1). Calcium influx was measured by a Fam-fluorochrome-labeled inhibitor of caspase-1 (FLICA) detection system. DAPI uptake and IL-1β release were monitored by flow cytometry. IL-1β enzyme-linked immunoassays (ELISAs) were performed according to Invitrogen Thermo Fisher Scientific (#BMS6002).
Differentiation between brain resident microglia and brain infiltrating macrophages was performed by FACS, where infiltrating cells were labeled CD45
+CD11b
+Ly6C
high and microglia were labeled CD45
intCD11b
+ [
36].
For functional analysis, brain cells from icv injected brain cells were stimulated with 0.5 mM ATP in RPMI containing DAPI at 37 °C for 5 min. Cells were washed and analyzed by flow cytometry.
Immunostaining
Mice were deeply anesthetized, and brains were fixed with 4% paraformaldehyde (PFA) by transcardial perfusion. After fixation in 4% PFA overnight, 50 µm thick sections were prepared using a vibratome. Immunostaining was performed at 4 °C on free-floating sections using an anti-Iba1 antibody to detect microglia (Fujifilm Wako Pure Chemical Corporation) and an anti-neuronal nuclear protein (NeuN) antibody to detect neurons (Thermo Fisher Scientific). DAPI (Thermo Fisher Scientific) was used to counterstain nuclei. Images were obtained by confocal laser scanning microscopy (LSM 880, Zeiss, Oberkochen, Germany).
Discussion
Here, we show that eATP is present early after cerebral ischemia and that blocking the ATP receptor P2X7 with specific nbs diminishes the tissue damage caused by ischemia. However, the nbs need to be injected intracerebroventricularly to bypass the BBB and reach the P2X7 receptor on brain resident cells.
Mounting evidence indicates that stroke triggers a sterile inflammatory response. The injured tissue releases a myriad of molecules that can activate the surrounding or infiltrating immune cells. Potent activators of local immune responses are danger-associated molecular patterns (DAMPs). Some of these endogenous danger signals can induce activation of the inflammasome and the secretion of proinflammatory cytokines by innate immune cells [
4,
38]. Using transgenic mice that express luciferase on the outer layer of the cell membrane, we showed that similar to traumatic brain injury [
39], eATP is released very early during ischemic tissue damage. In addition, the signal is sustained over 24 h, clearly indicating an ongoing release of eATP in the ischemic tissue. Therefore, eATP and its cognate receptors likely play an important role in the initiation of the inflammatory reaction following stroke. eATP activates purinergic receptors. While the microglial P2Y12 receptor is important for microglial neuron interactions, the proinflammatory response by microglia is likely triggered by P2X7, which is highly expressed by microglia [
40‐
42].
P2X7 is expressed in the brain mainly on glial cells. Expression data from the Allen Brain Atlas for mice [
18] and humans [
19] show that the P2X7 receptor is highly expressed by microglia in humans and rodents (Table
1). In contrast, astrocytes show low levels of P2X7 RNA. Species-specific differences in P2X7 expression can be found for oligodendrocytes, where P2X7 is highly expressed by human oligodendrocytes but not by murine oligodendrocytes. Therefore, it is likely that in rodents, the main effect of blocking P2X7 is mediated through microglial cells.
After ischemic stroke, the expression of P2X7 is increased on microglia [
43,
44] and can induce cell death in ischemic microglia [
15]. This increase in P2X7 expression is not found in astrocytes after ischemic stroke [
17]. We and others have shown that experimental stroke in P2X7-/- mice results in smaller infarcts and that blockade of P2X7 with BBG reduces cerebral ischemic damage [
20,
45]. In addition, the inhibition of the NLRP3 inflammasome decreased the amount of damage after cerebral ischemia, but there was no additional benefit if P2X7 was also blocked [
20]. These data are still controversial [
12]. Yanagisawa and colleagues observed an exacerbation of ischemic brain damage when P2X7 was blocked. Similar findings were also reported by Kang et al. [
46], who observed an effect on ciliary neurotrophic factor (CNTF) production but no effect on lesion size. One explanation for these discrepancies is the use of BBG. Small molecule inhibitors are often semispecific and toxic. In particular, BBG is not specific for P2X7 [
47] and is known to have dose-dependent off-target effects. Therefore, we used nbs that we had recently developed and are currently in the process of being patented (see MM; WO/2013/178783) [
10]. We not only generated several different families of murine P2X7-specific nbs but also different human P2X7-specific nbs. Nbs, recombinant single domain antibodies derived from camelid heavy chain antibodies, are a promising new technology platform. The first nb-based reagents developed by Ablynx-Sanofi have entered clinical trials and have achieved FDA approval (targeting TNF-α, von Willebrand factor, receptor activator of nuclear factor κB [RANK]-ligand, and IL-6 receptor [
48,
49]).
The BBB is a major obstacle for the treatment of brain disease with biologicals. Under healthy conditions, the BBB is only permeable for lipophilic molecules of up to 400 kDa [
50]. In addition, the delivery of conventional antibodies to the brain is further hampered by Fc receptor-mediated efflux to the blood [
51]. Therefore, nbs lacking an Fc part may reach targets behind the BBB. However, under nonpathological conditions, monovalent nbs do not attain sufficient concentrations for in vivo brain imaging [
52] or therapeutic purposes [
53]. In stroke, biphasic BBB breakdown is caused by activated matrix metalloproteinase (MMP)-2, MMP-3 and MMP-9 [
54,
55]. The breakdown of the BBB is initially reversible but is further increased with the release of MMP-3 and MMP-9 [
56]. These findings suggest that antibodies or nbs would have easier access to the brain in ischemic stroke. However, as we can show here, only a minor portion of the intravenously injected nbs reached the brain. While macrophages from the bloodstream were quickly covered with intravenously injected nbs, when they reached the brain, microglia did not carry any nbs, and their function was unimpaired (Fig.
4). These findings are similar to observations in antibodies crossing the BBB, where a direct shuttle system such as the transferrin receptor is usually needed to enter the brain [
57]. Since this problem prevents noninvasive iv administration of the nbs, it is necessary to find strategies to facilitate the transport of nbs across the BBB. For this study, we chose to directly inject our nbs into the ventricular system of the brain, which is difficult in the mouse system because of the small volume that can be injected. We were able to modify our nbs so they could be highly concentrated without aggregating (Additional file
1: Fig. S1). In humans, nb delivery would be less of a problem since it could be accomplished by lumbar puncture and injection into the cerebral spinal fluid (CSF). Direct injection in the CSF of therapeutics is already used for other neurological diseases, such as neuronal ceroid lipofuscinosis [
58]. Other promising possibilities for nb delivery to the CNS include the fusion of nbs to ligands of brain-endothelial receptors such as ApoE-LDL-receptor or to nbs directed against cell transcytosis receptors on cerebral endothelial cells [
59‐
61].
In stroke, microglia are the first immune cells to respond, while macrophages enter the brain at later stages [
32]. Therefore, it is not surprising that there was no difference in ischemic lesion size after iv nb injection, where the nbs could not pass the BBB. In contrast, after an icv injection of P2X7-specific nbs, we could reach up to 95% of the microglia. This level of P2X7R blockade was sufficient to inhibit microglial activation and improve the outcome. Our study shows that inhibition of signaling by eATP is only effective if it is done early and reaches locally expressed P2X7 in the brain.
Limitations
Our study was a proof-of-concept study, which was not designed to simulate the clinical setting. Further studies are needed to determine whether P2X7-specific nbs improve outcomes after stroke, how they influence long-term outcomes, and if they are similarly effective in female, comorbid and old mice. Our results will have to be reproduced in other laboratories and other model systems before translation.
Conclusion
Here, we demonstrate the importance of locally produced eATP for the damage in ischemic stroke and the potential of intracerebroventricularly injected P2X7 nbs to reduce this damage.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.