Tissue-Resident Ecto-5′ Nucleotidase (CD73) Regulates Leukocyte Trafficking in the Ischemic Brain

Ectoenzymes expressed on the surface of vascular cells and leukocytes modulate the ambient nucleotide milieu. CD73 is an ecto-5′ nucleotidase that catalyzes the terminal phosphohydrolysis of AMP and resides in the brain on glial cells, cells of the choroid plexus, and leukocytes. Though CD73 tightens epithelial barriers, its role in the ischemic brain remains undefined. When subjected to photothrombotic arterial occlusion, CD73−/− mice exhibited significantly larger (49%) cerebral infarct volumes than wild-type mice, with concordant increases in local accumulation of leukocyte subsets (neutrophils, T lymphocytes, macrophages, and microglia). CD73−/− mice were rescued from ischemic neurologic injury by soluble 5′-nucleotidase. In situ, CD73−/− macrophages upregulated expression of costimulatory molecules far more than wild-type macrophages, with a sharp increase of the CD80/CD86 ratio. To define the CD73-bearing cells responsible for ischemic cerebroprotection, mice were subjected to irradiative myeloablation, marrow reconstitution, and then stroke following engraftment. Chimeric mice lacking CD73 in tissue had larger cerebral infarct volumes and more tissue leukosequestration than did mice lacking CD73 on circulating cells. These data show a cardinal role for CD73 in suppressing ischemic tissue leukosequestration. This underscores a critical role for CD73 as a modulator of brain inflammation and immune function.

C erebral ischemia elicits a strong inflammatory response (1) involving multiple cellular and humoral mediators. Little is known about humoral mediators for which catabolism in the extracellular intravascular milieu modulates cell-cell interactions that promulgate inflammation and ischemic tissue damage. Within the primary area of cerebral infarction, neurons and glial cells become damaged, resulting in extensive Wallerian and terminal degeneraton, loss of distal microvascular flow, and regional edema (2). These characteristic histopathological changes are accompanied or exacerbated by infiltration of lymphocytes, polymorphonuclear and mononuclear leukocytes, as well as by reactive astrocytosis, all of which can play a role in the development of secondary injury after acute brain infarction (3,4). Recruitment of inflammatory cells into infarcted tissue occurs by a stepwise process of homing, adhesion, and, ultimately, diapedesis (5). Cells migrate between the endothelial cells that line the inner surface of blood vessels and astroglial feet that comprise the neurovascular unit, ultimately reaching brain parenchyma (6). Recent work has shown that transcellular metabolism by endothelial-surface ENTPDase1 (CD39) of extracellular ATP and ADP released by activated platelets can mitigate explosive amplification of thrombotic nidus formation, thereby reducing damage in ischemic/reperfused stroke (6).
CD73/ecto-59 nucleotidase (59-NT) is a GPI-anchored cellsurface glycoprotein, an ectoenzyme that colocalizes near CD39 at the endothelial surface. CD73 catalyzes hydrolysis of AMP, the terminal product resulting from CD39 activity, to generate the anti-inflammatory and vasodilator nucleoside adenosine. In the extracellular milieu, the CD73-mediated generation of adenosine serves as a key regulator of mucosal and endothelial properties critical to homeostasis, including dynamic processes such as mucosal integrity, vascular flow, and leukocyte traffic (7). Under basal conditions, ATP is continuously released into the extracellular space through regulated transport. Furthermore, at sites of hypoxic and/or ischemic injury, additional large boluses of ATP are liberated from apoptotic and necrotic cells (5). On resting vascular endothelium, extracellular ATP and ADP bind to purinoreceptors of the P2X and P2Y families, thereby eliciting prothrombotic and proinflammatory cascades. CD73, which colocalizes with CD39, quickly degrades these nucleotide-mediated inflammatory stimuli to yield adenosine, which binds to the P1 type of purino-receptors. This sets off a secondary signaling cascade that promotes vascular flow via vasodilation and suppresses inflammation by inhibiting leukocyte extravasation (5).
To specifically evaluate the contribution of CD73 as an inflammatory modulator in the microenvironment of ischemic brain injury, experiments were performed using a modification of a recently described model of photothrombotic occlusion of the middle cerebral artery (MCA) (8). The photothrombosis stroke model was employed because of its propensity to create intravascular thrombus similar to that seen in human stroke. It also produces cortical infarcts that are highly reproducible in location and size, which is essential for quantification of cellular response. Genetic, pharmacologic, and cellular approaches were used to study the contribution of CD73 to leukocyte trafficking and neurologic outcomes in the setting of stroke.

Photothrombotic model of cerebral ischemia
Permanent occlusion of MCA (MCAO) was induced as previously described (8) by an operator blinded to the genotype of the experimental animal. Ten-week-old male mice were anesthetized with 2.5 mg ketamine and 0.25 mg xylazine given i.p. (Phoenix Pharmaceutical, St. Joseph, MO). Body temperature was maintained during surgery at 37˚C and for 45 min thereafter using a temperature-controlled circulating liquid heating pad. After opening a 2-to 3-mm diameter oval bony window using a dental drill (Foredom Electric, Bethel, CT), the distal part of the left MCA was exposed. A laser Doppler flow probe (Type N, 18-gauge; Transonic Systems, Ithaca, NY) was attached to the surface of the cerebral cortex 1.5 mm dorsal median to the bifurcation of the distal MCA. The probe was connected to a flow meter (Transonic model BLF21) and flow recorded with a continuous data acquisition program (Windaq; DATAQ Instruments, Akron, OH). Rose Bengal was diluted to 10 mg/ml in PBS and injected i.v. to achieve a final concentration of 40 mg/kg body weight. A 1.5 mW green neon laser (540 nm; CVI Melles Griot, Albuquerque, NM) was directed at the MCA from a distance of 6 cm and occlusion monitored by a cerebral blood flow probe. Occlusion was defined as a .80% reduction in blood flow for ∼10 min. After obtaining stable occlusion, the laser remained in place for 15 additional min. In a subset of experiments, CD73 2/2 and wild-type (WT) mice were injected i.p. with 7.5 U soluble 59-NT purified from Crotalus atrox venom given 30 min before induction of brain ischemia, whereas controls were injected with an equal volume of saline. In another subset of experiments, CD73 was inhibited using adenosine 59(a, b-methylene) diphosphate (AOPCP; 20 mg/kg) injected i.p. All drugs were given 30 min before the experimental procedure.

Magnetic resonance imaging
Infarct volumes were measured using magnetic resonance imaging (MRI) and performed by the University of Michigan Small Animal Imaging Resource 48 h after induction of brain ischemia. Throughout the MRI scanning procedure, mice were anesthetized with a 2% isoflurane/air mixture. Mice were positioned prone, head first, in a 7.0T Varian MR scanner (183 mm horizontal bore; Varian, Palo Alto, CA), with their body temperatures maintained at 37˚C using circulating heated air. A doubletuned volume radiofrequency coil was used to scan the head region of the mice. Axial T2-weighted images were acquired using a spin-echo sequence using the following parameters: repetition time/effective echo time, 4000/40 ms; field of view, 30 3 30 mm; matrix, 256 3 256; slice thickness, 0.5 mm; slice spacing, 0 mm; number of slices, 25; and number of scans, 1 (total scan time 8 min). Cerebral infarct volumes were quantified at 48 h using VOI-11 software by an observer blinded to experimental conditions.

Neurologic deficit scoring
Forty-eight hours after stroke, mice were assessed for neurologic deficit using a previously described five-tiered grading system (10), with measurements performed by an observer blinded to experimental conditions. A score of 1 was given if the animal demonstrated normal spontaneous movement; a score of 2 was given if the animal was circling clockwise when viewed from above while receiving a mildly noxious stimuli (tail pinch); a score of 3 was given if the animal was observed to spin clockwise on a longitudinal axis including the tail; a score of 4 was given if the animal fell down on the contralateral side; and a score of 5 was given if the animal was crouched on all four paws unresponsive to noxious stimuli.

Brain water content
A separate cohort of mice not undergoing infarct volume or leukocyte trafficking measurements were euthanized, and brains were removed rapidly and divided into ischemic and nonischemic hemispheres. The samples were weighed and then dried at 95˚C for 24 h to obtain the dry weight. The brain water content was calculated as: (wet weight 2 dry weight)/dry weight.

Immunohistochemistry
Whole brains of mice that underwent photothrombotic occlusion were harvested 48 h postsurgery, fixed with 4% paraformaldehyde, and paraffinembedded. Sections to be stained with CD73 then underwent deparaffinization, rehydration, and the Ag was retrieved. Endogenous peroxidase was blocked in each of the 5-mm sections followed by a rodent serum block. Serial sections were then stained for CD73 (Abcam, Cambridge, MA) with a rabbit polyclonal Ab at a 1:200 dilution or von Willebrand factor (vWF; Abcam) at a 1:500 dilution. Images were taken on an Eclipse TE2000-E microscope (Nikon Instruments).

Flow cytometric analysis of inflammatory cells
Forty-eight hours after surgery, mice were euthanized and the brains removed and divided into ischemic and nonischemic hemispheres. The cerebral hemispheres were then minced with a scalpel and the tissue pieces repeatedly drawn into a syringe through an 18-gauge needle to obtain single-cell suspensions. A Percoll (GE Healthcare, Piscataway, NJ) gradient was then used for separation of infiltrating cells, discarding the myelin fractions and residual debris. FACS Lysis Buffer (BD Biosceinces, Franklin Lakes, NJ) was used to lyse RBC contaminants. Prior to flow cytometric analyses, live cells were counted using a hemocytometer and distinguished from dead cells by the absence of trypan blue (Invitrogen, Carlsbad, CA) staining. Nonspecific Ab binding was blocked using Fc Block (BD Biosciences). Cell populations were purified and identified in three stages. First, the leukocyte/microglia population fraction of cells was isolated using an Ab to a common leukocyte Ag (CD45). Double-staining pairings of CD45/ CD4 and CD45/CD8 were used to identify CD4 + and CD8 + T cell subpopulations. Cells with high expression of CD45 (CD45 hi ) were further identified as neutrophils based on FITC-conjugated LY-6G (BD Biosciences) positivity (CD45 hi LY6-G hi ) or as infiltrating macrophages by FITC-conjugated F4/80 (Serotec, Raleigh, NC) positivity. Among CD45bearing cells, cells with low expression of CD45 and also positive for F4/80 Ab were resident microglia. Dead cells were excluded with a propidium iodide gate. All samples were acquired on an FACS Calibur Flow Cytometer, and data were analyzed using CellQuest software (BD Biosciences).

Bone marrow transplantation
For certain experiments, mice were myeloablated followed by bone marrow reconstitution (11). Ten-week-old male CD73 2/2 or WT (CD73 +/+ ) littermate control mice were irradiated using 12.5 Gy radiation exposure administered in three doses 3 h apart. Mice were repopulated with CD73 2/2 bone marrow or WT bone marrow using ∼4 3 10 6 bone marrow cells administered i.v. Mice were allowed to recover for 8 wk before induction of brain ischemia and phenotype examination. For these experiments, four types of chimeric animals were generated: CD73 2/2 mice repopulated with: 1) CD73 2/2 bone marrow to create global KO controls; 2) WT bone marrow to create tissue-only CD73-deficient mice; 3) WT mice reconstituted with WT bone marrow to create WT controls; or 4) CD73 2/2 bone marrow to create mice that lacked CD73 only on their leukocytes. These mice are denoted using the following abbreviation scheme: bone marrow donor → bone marrow recipient. Chimeras are referred to as knockout (KO) → KO; WT → KO; WT → WT; and KO → WT.

Real-time RT-PCR assay
Total RNA in brain tissues at 48 h after photothrombotic MCAO was extracted via the RNAzol B method (Tel-Test, Friendswood, TX). Total RNA was reverse-transcribed into cDNA by using random primers (Life Technologies, Rockville, MD). To detect cerebral levels of IL-10, IL-6, chemokine (C-X-C motif) ligand 1 (KC), TNF-a, IL-1b, and VCAM-1 mRNA, real-time RT-PCR was performed by means of an ABI PRISM 7700 sequence detection system with TaqMan Universal PCR Master Mix and Assays-on-Demand gene expression probes (Applied Biosystems, Foster City, CA). TaqMan Rodent 18S rRNA Control Regents VIC (Applied Biosystems) was used as an endogenous control gene. A standard curve for the serial dilution of murine brain cDNA was generated. The amplification cycle consisted of 2 min at 50˚C, 10 min at 95˚C, 15 s at 95˚C, and 1 min at 60˚C. Relative quantitative values of targets were normalized according to the endogenous 18S rRNA gene control.

Statistical analyses
Values are reported as mean 6 SEM, with the number of experiments performed provided in the figure. The significance of differences between groups with multiple comparisons was estimated by one-way ANOVA followed by Newman-Keuls test. Statistical significance was confirmed at p , 0.05.

Effect of CD73 gene absence on stroke outcome
Forty-eight hours after induction of MCAO, cerebral infarct volumes were assessed in both CD73-deficient and WT mice using T2-weighted cortical MRIs (Fig. 1A). Total infarct volumes were increased by 49% in CD73 2/2 mice compared with their WT littermates (71.2 6 0.8 mm 3 compared with 36.6 6 1.7 mm 3 , respectively; p , 0.001) (Fig. 1B). The larger infarct volumes in CD73 2/2 mice corresponded with a functional outcome after MCAO, as the CD73 2/2 mice had greater neurologic deficits when compared with WT mice (Fig. 1C). Given the known role for CD73 in maintaining epithelial and endothelial barrier properties (9), a comparative analysis of cerebral edema was performed 48 h after induction of brain ischemia. In comparison with WT controls, brain water content was significantly increased (∼30%) in the infarcted hemisphere of CD73 2/2 mice (4.53 6 0.01 ml/g dry wt tissue versus 6.5 6 0.07 ml/g dry wt tissue (Fig.  1D).
Because inflammation can worsen the outcome of brain ischemia, experiments were performed to assess whether CD73 modulates leukocyte trafficking into ischemic cerebral tissue. Analysis of brain tissue was conducted using multiparameter flow cytometry 48 h after induction of brain ischemia to quantify leukocyte populations in ischemic hemispheres. Ischemic hemispheres of CD73 2/2 animals had a .30% increase in the total numbers of infiltrating nucleated cells when compared with WT ischemic hemispheres (data not shown). Because the nonischemic hemi-spheres showed no significant differences in terms of infiltrating cell numbers between the two genotypes, we concluded that CD73 does not affect basal levels of cerebral inflammation (data not shown). To identify the mononuclear fraction more precisely, we considered CD45-positive cells that expressed the F4/80 mononuclear cell-surface marker (CD45 hi F4/80 + ), to be blood-derived macrophages. F4/80 surface marker was used because it is more specific for macrophages than CD11b, because CD11b is also found on some B and polymorphonuclear cells (12). The second cell population examined was CD45 low F4/80 + cells, which expressed 10 to 15 times less CD45 Ag than macrophages and hence were considered to be resident microglial cells. Two additional Abs against B7-1 (CD80) and B7-2 (CD86) Ags were used to examine the degree of activation of mononuclear cells within the inflamed brain.
In addition to examining absolute levels of recruited effector lymphocyte and leukocyte populations, experiments were performed to determine local production of inflammatory cytokines and adhesion molecules that could drive leukosequestration in the ischemic zone. To confirm the proinflammatory phenotype of ischemic brains of CD73 2/2 mice, expression of proinflammatory cytokines and adhesion molecules was analyzed 48 h after in-duction of brain ischemia by quantitative RT-PCR of RNA isolated from the brains of WT or CD73 2/2 mice.
As shown in Fig. 4A-D, the amounts of mRNAs encoding the cytokines IL-1b, IL-6, TNF-a, and KC were significantly greater in the ischemic hemispheres of CD73 2/2 animals compared with the amounts seen in WT mice. Although amounts of VCAM-1 mRNA were slightly increased in the nonischemic hemispheres of CD73 2/2 mice, significantly greater amounts of VCAM-1 mRNA was present in the ischemic hemispheres of CD73 2/2 mice following MCA occlusion (Fig. 4E). These experiments went further to examine the induction of anti-inflammatory cytokine IL-10, which is known to suppress TNF-a, IL-1, and IL-12, thereby contributing to both the limitation and resolution of inflammation (13). Though not statistically significant, there was a trend toward diminished IL-10 mRNA in ischemic brains of CD73 2/2 mice (Fig. 4F). The upregulation of proinflammatory and downregulation of anti-inflammatory cytokines as well as leukocyte recruitment demonstrated the enhanced inflammation in the ischemic cerebrum of CD73 2/2 mice.
Rescue of CD73 genotype-null mice from stroke sequelae with soluble 59-NT To fulfill Koch's postulates regarding a causal role for a pathway in disease, we not only performed experiments in which the pathway was deleted, but also experiments in which the deleted pathway was reconstituted. To prove the assertion that CD73 plays an important role in regulation of leukocyte trafficking in brain ischemia, CD73 2/2 and WT mice were each reconstituted with 7 U of soluble 59-NT purified from Crotalus Atrox venom, given 30 min before induction of brain ischemia. As in the earlier experiments, operators blinded to the identity of the experimental animals evaluated infarct volumes and neurologic scores. The mice were then euthanized, and ischemic and nonischemic hemispheres were separated for flow cytometric analyses of infiltrating leukocyte populations. As shown in Fig. 5A, 59-NT treatment of CD73 2/2 mice was associated with the complete reconstitution of a WT phenotype (cerebral infarct volumes in saline-treated CD73 2/2 mice were 68 6 2.6 mm 3 versus 30.4 6 2.9 mm 3 in CD73 2/2 treated with 59-NT). For comparison, 59-NT was also able to reduce infarct volumes in WT animals (infarct volumes in WT mice treated with saline were 34.8 6 0.75 mm 3 versus 22.7 6 2.1 mm 3 in WT animals treated with 59-NT). The reduction of infarct volumes in CD73 2/2 mice treated with 59-NT corresponded with improved neurologic score after MCA occlusions (Fig. 5B).
Leukocyte populations (neutrophils, microglia, and mononuclear fraction) were identified using the same combinations of Abs as before. We had previously observed (Fig. 2B) that CD73 deficiency primarily affects the mononuclear fraction of infiltrating cells (CD45 hi F4/80 hi macrophages) 48 h after induction of brain ischemia. In this next set of experiments, 59-NT was administered 30 min prior to the ischemic episode. 59-NT not only suppressed macrophage recruitment in ischemic WT mice (Fig. 5C), but also did so even more in ischemic mice lacking native CD73. 59-NT reduced total numbers of infiltrating macrophages by 48% in WT animals and 57% in CD73 2/2 mice. As a percentage of total infiltrating leukocytes, 59-NT caused the macrophage population of ischemic CD73 2/2 mice to decrease to the same level as that seen in vehicle-treated WT controls (9.94 6 0.5% CD73 2/2 treated with 59-NT versus 9.33 6 0. 3% in saline-treated WT mice; data not shown).
One other important facet of recruited leukocytes is related to their activation state, which can affect their immune functions. Treatment with soluble 59-NT in both CD73 2/2 and WT mice not only resulted in a markedly reduced absolute number of infiltrating macrophages, but those infiltrating macrophages exhibited a striking reduction in activation phenotype 48 h after induction of brain ischemia (Fig. 5D, 5E). As in previous experiments, these data were obtained by measuring expression of CD80 and CD86 costimulatory molecules. Although CD86 is constitutively expressed on a variety of APCs, the expression of both CD80 and CD86 is elevated upon activation (14). B7-1-positive infiltrating macrophages (CD45 hi F4/80 hi CD80 + ) isolated from ischemic hemispheres of CD73 2/2 mice treated with 59-NT were shown to express lower levels of B7-1 Ag (39%) compared with macrophages isolated from ischemic hemispheres of saline-treated CD73 2/2 mice (Fig. 5D). Note that CD73-deficient macrophages treated with saline demonstrated far greater activation then WT macrophages with respect to CD80 expression (CD73 2/2 MFI 100.4 6 8.9 versus 48.6 6 0.5 in WT; Fig. 5D). However, in CD73 2/2 mice, treatment with soluble 59-NT downregulated CD86 expression only 25%, resulting in a substantial decrease in the CD80/CD86 ratio compared with saline-treated mice (Fig. 6E). Treatment of WT mice with soluble 59-NT resulted in an additional reduction of both costimulatory molecules (Fig. 5D, 5E).

Stroke sequelae in CD73 chimeric mice
In an attempt to differentiate the contribution of CD73 on brainresident cells from that of CD73 on circulating leukocytes, a series of CD73 chimeric mice were generated. Four groups of chimeras were made by myeloablation and bone marrow reconstitution according to the following schema (donor → recipient); WT marrow into WT recipient; CD73 2/2 marrow into WT recipient; WT marrow into CD73 2/2 recipient; and CD73 2/2 marrow into CD73 2/2 recipient. The first and last chimeras (KO → KO and WT → WT) served as transplantation controls. KO → WT chimeras served as an experimental condition in which endothelium and other resident cells express CD73, but circulating leukocytes do not. WT → KO chimeras served as an experimental condition in which CD73 is expressed on circulating leukocytes, but it is absent from resident vascular cells. All experiments were performed 8-10 wk after reconstitution to allow for full bone marrow reconstitution.
Forty-eight hours after photothrombotic occlusion of the MCA, cerebral MRI scans were obtained to quantify infarction, neuro-logic deficit was scored by an operator blinded to experimental conditions, and leukocyte trafficking was assessed by flow cytometry. Cerebral infarct volumes in the KO → KO group of mice were markedly larger (48%) than those in the WT → WT group (68.8 6 0.7 mm 3 versus 35.8 6 1.2 mm 3 , respectively; p , 0.001, Fig. 5A). We next examined the effect of selective CD73 rescue by myeloablating naive mice and reconstituting them with marrow cells possessing or lacking the CD73 gene, after which MCAO was performed. Infarct volumetric analysis demonstrated that expression of CD73 only on brain-resident tissue (i.e., KO marrow → WT recipient) provided some protection to mice from (B) Neurological deficit was measured using a five-tiered grading system in the same animals. Cells from contralateral (C) and ischemic (I) hemispheres of these mice were subjected to flow cytometric analysis to determine the absolute number of macrophages (C), microglia (F), and neutrophils (G) (n = 6/group). Similarly, MFI was measured in macrophages isolated from ischemic (I) hemispheres of WT and CD73-null mice treated with soluble 59-NT or vehicle and labeled with Abs against CD80 (D) and CD86 (E) (n = 4/group). *p , 0.05, **p , 0.01, ***p , 0.001. stroke when compared with mice with global deficiency of CD73 (the KO → KO group). Quantitatively, this protection was measurable as a 40% reduction in infarct volumes (41.6 6 0.7 mm 3 versus 68.8 6 0.7 mm 3 in KO → WT versus KO → KO groups, respectively; p , 0.001). These data show that there is a considerable contribution from tissue-resident CD73 in cerebroprotection after ischemic brain injury. We further examined the effect of selective expression of CD73 on bone marrow-derived FIGURE 6. Selective inactivation of CD73 molecule on tissue only attenuates MCAO-induced brain ischemia. (A) Quantitative analyses of infarct volumes in marrow-reconstituted mice (n = 4/group). (B) Locomotor activity determined by neurologic deficit score shown for individual animals across the genotype 48 h after induction of brain injury (n = 4/group). The contribution of CD73 on brain-resident tissue and leukocytes to leukosequestration of microglia (C), neutrophils (D), and macrophages (E) was examined using bone marrow-reconstitution studies (n = 4/group). (F) and (G) show MFI of macrophages expressing CD80 and CD86 molecules isolated from ischemic hemispheres of bone marrow-reconstituted mice (n = 4/group). Thirty minutes prior to induction of brain ischemia, a different cohort of WT and CD73 2/2 mice were treated i.p. with vehicle alone or the CD73 inhibitor AOPCP (H), and 48 h later, infarct volumes were calculated (n = 5/group except n = 8 for WT AOPCP). *p , 0.05, **p , 0.01, ***p , 0.001. cells using a strategy of WT marrow implanted into CD73 2/2 recipients (WT → KO). This expression of CD73 on bone marrow cells only provided limited protection from cerebral ischemia (58.9 6 0.2 mm 3 ), which represent a 14% decrease in infarct volume in comparison with global lack of the CD73 molecule (KO → KO; p = NS). Consonant with these data, locomotor activity in the KO → WT group were substantially better than that in the KO → KO group, but there was no difference in basal locomotor activity between WT → WT and KO → WT mice (Fig. 6B).
We next evaluated the effect of site-selective CD73 expression on the trafficking of leukocytes to the ischemic brain. Total numbers of nucleated cells infiltrating ischemic hemispheres paralleled infarct size as well as neurologic deficit scores in each of the four groups of myeloablated and marrow-reconstituted mice under study (data not shown). By using a dual staining technique with anti-CD45 and anti-F4/80 Abs, an infiltrating mononuclear fraction could be easily identified and distinguished from the resident microglial population. Levels of CD45 and F4/80 expression distinguish between microglial (CD45 lo F4/80 + ) and CNSassociated macrophage populations (CD45 hi F4/80 hi ) (15). As in our previous experiments, the relative percentage of CD45 lo F4/ 80 + cells (microglia) among all leukocytes and also the relative percentage of CD45 hi LY6-G hi cells (neutrophils) did not change across genotypes. However, the total numbers of infiltrating cells of either population were significantly higher in KO → KO mice when compared with control (WT → WT) mice or chimeric animals that have tissue-resident CD73 (KO → WT; Fig. 6C). Similarly, the presence of CD73 either on circulating or resident cells reduced the accumulation of neutrophils in the ischemic brain (Fig. 6D).
After the induction of unilateral brain ischemia, the total number of infiltrating cells in the contralateral (nonischemic) hemisphere did not vary with respect to CD73 genotype or chimerism (data not shown). In myeloablated and reconstituted mice completely devoid of CD73, there was a .50% increase in the relative ratio of CD45 hi F4/80 hi infiltrating macrophages compared with myeloablated and reconstituted control mice (WT → WT). In chimeric animals in which CD73 was present in brain-resident tissue (KO → WT), macrophage infiltration was similar to control chimeras (WT → WT). In contrast, in WT → KO mice, macrophage infiltration was increased significantly (by 37%) in comparison with WT → WT strokes (data not shown). Similarly, the total number of infiltrating macrophages was significantly increased (up to 2.5-fold) in KO → KO and WT → KO mice (Fig. 6E). When CD73 was present on brain tissue but absent from leukocytes, there is little effect on leukocyte trafficking compared with WT → WT chimeras (Fig.  6E). These data together indicate that CD73 has an important native role that suppresses leukocyte accumulation in an ischemic zone, especially when that CD73 is expressed on brain-resident tissue.
As in our previous experiments, to define the contribution of activated leukocytes following injury, the next set of experiments showed that macrophages isolated from completely CD73 2/2 chimeric mice (KO → KO) express ∼60% more CD80 and ∼26% more CD86 on their surface when compared with WT → WT controls. The presence of CD73 on brain-resident tissue alone (KO → WT) significantly lowered expression of CD80 relative to CD86, thereby decreasing the CD80/CD86 ratio, reflecting attenuation of macrophage infiltration as well as overall inflammation in those mice when compared with animals with global lack of CD73 (CD80 MFI 50 6 5.7 in KO → WT mice versus 113 6 8.3 in KO → KO mice; CD86 MFI 48 6 1 in KO → WT versus 66.4 6 2.2 in KO → KO mice; Fig. 6F, 6G). In contrast, macrophages isolated from ischemic hemispheres of mice that have CD73 on leukocytes only (WT → KO) express significantly higher levels of CD80 when compared with macrophages isolated from control chimeric mice (CD80 MFI 85.7 6 3.8 in WT → KO versus 43 6 2 in WT → WT) and when compared with macrophages isolated from the chimeras when CD73 is present on the brain tissue only (KO → WT CD80 MFI 49.9 6 5.7, Fig. 6F, 6G). These data imply that CD73 from bone marrow cells provides less neuroprotection than CD73 derived from the brain-resident tissue. Immunohistochemical experiments staining WT and CD73-null mice for CD73 or vWF revealed CD73 expression on cells of the choroid plexus and an absence on endothelial cells of the brain (Fig. 7).

Effect of CD73 inhibition on stroke outcome
We next pursued the functional contribution of CD73 to cerebroprotection by treating mice with an i.p. injection of 20 mg/kg of the specific CD73 inhibitor AOPCP or vehicle control 30 min before induction of permanent brain ischemia. As shown in Fig.  6H, treatment of WT animals with AOPCP increases infarct volume 38% in comparison with saline-treated WT mice. Infarct volume observed in CD73 2/2 mice after the treatment with AOPCP was not statistically different from infarct volume in CD73 2/2 animals treated with saline, but both are still larger (∼19%) than infarct volumes of WT mice treated with AOPCP. This observation suggested to us that treatment with AOPCP partially blocked the CD73 activity, at least in the concentration used in this study.

Discussion
The ecto-59-NT CD73, which catalyzes the hydrolysis of AMP to adenosine, is considered the rate-limiting step in the generation of extracellular adenosine (16). In this work, we hypothesized that the generation of extracellular adenosine by CD73 is required to maintain the balance between pro-and anti-inflammatory signaling during cerebral ischemia. We used both genetic and pharmacologic approaches to test this hypothesis and found that CD73null mice were more susceptible to stroke injury, contained more infiltrating macrophages and associated proinflammatory cytokines in ischemic brain tissues, and these macrophages had a small increase in CD86 expression and a huge increase in CD80 expression. In addition, we found that injection of soluble 59-NT attenuated these effects in CD73 2/2 mice, effectively rescuing them from ischemic brain injury. We furthermore used bone marrow transplantation to demonstrate the relative importance of brain tissue-derived CD73 compared with the CD73 on circulating cells. Data from these experiments showed that CD73 on circulating cells is less critical than CD73 on tissue-resident cells in conferring ischemic cerebral protection. These results support our hypothesis and lead to the novel conclusion that brain tissuederived CD73 attenuates stroke injury by reducing the infiltration and activation of macrophages.

Effect of CD73 on stroke
Our studies revealed that MCAO resulted in significant increases in infarct volume (∼50%), neurologic deficit (∼56%), and brain edema (∼30%) in CD73 2/2 mice compared with WT controls. Rat brain MCAO has been shown to increase CD73 expression in infarcted tissue (4), and this and other evidence suggests an increase in the ability of brain tissue to hydrolyze extracellular nucleotides released as a consequence of severe tissue damage (4,17). With regard to this latter point, it was recently shown that hypoxia drives transcriptional increase of both surface ecto-nucleotidases CD39 and CD73 (18). Consequently, this amplifies the extracellular production of adenosine from adenine nucleotide precursors. In fact, our previous work showed that, following challenge by cerebral ischemia, CD39-deficient animals developed worse clinical outcomes than their corresponding controls (6,19). Hypoxia can also induce the expression of CD73 via binding of hypoxia-inducible factor 1-a to the CD73 promoter region (9,20). In contrast, activation of the A 2A adenosine receptor by adenosine has been shown to suppress immune responses by inhibiting activated immune cells (21). These observations suggested that dephosphorylation of AMP by CD73 could represent a major pathway of extracellular adenosine formation and inflammatory suppression during the oxygen deficit of brain ischemia. In addition, administration of exogenous AMP in rats was shown to increase infarct volume and cause hypotension, hyperglycemia, hypocalcemia, and consequent metabolic acidosis (22). This suggests that CD73 deficiency could increase stroke injury by both suppressing anti-inflammatory adenosine production and increasing extracellular AMP levels. The deleterious effects of AMP during ischemic stroke may be mediated by activation of AMP-activated protein kinase in the vascular endothelial cells, leading to increased NO production and vessel dilation and then to additional cerebral hypoperfusion and brain damage (23).

CD73 control of leukocyte trafficking
The extent of neuronal damage correlates with the degree of the innate immune response, with numerous studies demonstrating the critical role of cellular and humoral immune activity after ischemic brain injury (24,25). Forty-eight hours after induction of permanent brain ischemia, we found a near doubling of the total number of infiltrating neutrophils and microglial cells in the ischemic hemispheres of CD73 2/2 mice, a far larger increase than seen in WT controls. Consistent with the present findings, studies of hypoxia-associated lung inflammation and LPS-induced acute lung injury confirmed an anti-inflammatory role for both CD73 and CD39 in these models (26). Lennon et al. (27), examining the interaction of leukocytes, particularly neutrophils, at cell-cell junctions, showed that inhibition of CD73 using either AOPCP or the anti-CD73 mAb IE9 inhibited the resealing of endothelial and epithelial barriers by 85%, suggesting the necessity for purine nucleotide metabolism in this pathway. This observation fits nicely with our data demonstrating a loss of barrier function in the is-chemic brains of CD73 2/2 mice. In contrast, a recent study has shown that CD73 2/2 mice are protected from experimental autoimmune encephalomyelitis, suggesting that CD73-dependent adenosine production and signaling through the A 2A adenosine receptor are required for the efficient entry of lymphocytes into the CNS (28). These disparate findings may be explained by differences in the mechanism governing inflammatory cell trafficking in the CNS versus the lungs, differences between the subtypes of inflammatory cells (neutrophils versus lymphocytes), or differences in leukocyte egress mechanisms. It is also possible that the absence of CD73 in other brain tissues could be due to the interaction of leukocytes and the endothelium. Adenosine, in our case generated by CD73, has been shown to decrease leukocyte trafficking through the inhibition of cytokine release from endothelial cells (29). Additionally, catalytic activity of CD73 on the cell surface is inhibited by the adhesion of lymphoid cells (30). Although CD73 may not be present on the endothelial cells of the brain, it is observed on the leukocytes that accumulate at the vasculature interfaces and infiltrate into the ischemic tissue. The adenosine generated by CD73 has been shown to be more important than CD73 itself in regulating lymphocyte entry into the CNS in the case of experimental autoimmune encephalomyelitis (28).
One of the most interesting observations in our study was an increase in the relative ratio (44%) and total number of infiltrating macrophages in ischemic hemispheres of CD73 2/2 mice compared with WT controls. The most important cellular response to postischemic inflammation is by cells of the innate immune system, predominantly resident microglia/brain macrophages and bloodderived monocytes/macrophages (31). Macrophages in the ischemic zone are more highly activated cells and therefore could be primed to exacerbate local inflammation and promote incremental tissue damage (32). In the inflammatory milieu, macrophages have multiple functions, including acting as phagocytic cells to clear the area of cell debris to promote remodeling and repair of damaged tissue. Those functions are responsible for the dual role of macrophages in inflamed tissues: to promote the cell-mediated immunity and promote resolution of inflammation (33). In addition, both CD4 + and CD8 + T cell numbers were increased in the ischemic brains of CD73 2/2 mice. CD25 + Foxp3 + regulatory T cells and CD25-uncommitted primed precursor Th cells have been shown to express CD73. Through the generation of adenosine by these cells, the inflammatory response seems to be diminished. It is possible the adenosine generation by CD73 present on the infiltrating T cells of the WT mice resulted in protection and smaller infarct volumes (34). IFN-b is known to increase CD73 expression on brain endothelial cells. The adenosine generated by CD73 may mediate the inhibitory effects of IFN-b on the transmigration of CD4 + T lymphocytes (35). This is in concordance with the increased numbers of T lymphocytes seen in CD73-null mice. The binding of various ligands to adenosine receptors on monocytes and macrophages strongly suppressed TLR4-mediated LPS induction of the proinflammatory cytokines IL-12 and TNF-a (36,37), which may be one of the central mechanisms whereby adenosine receptor occupancy prevents inflammation-induced tissue injury. In addition, adenosine facilitates anti-inflammatory IL-10 production by stimulation of A 2A and A 2B receptors on murine peritoneal macrophages (13,38), which may also contribute to the anti-inflammatory and immunosuppressive action of adenosine. A recent report showed that IL-1b and TNF-a were produced by a subset of microglia and macrophages after induction of permanent brain ischemia (39). These data are in agreement with our observation that the mRNA expression of the proinflammatory cytokines IL-6, KC, TNF-a, and IL-1b are increased in ischemic hemispheres of CD73 2/2 animals compared with WT controls, a finding consistent with the greater inflammation and macrophage infiltration we observed in CD73 2/2 mice. Previous work in ischemia/reperfusion has shown protective effects of IL-10 due to its ability to suppress macrophage activation, downregulate proinflammatory cytokine production, and suppress leukocyte-endothelial cell interactions (40). Additionally, IL-10 has been shown as a central mediator of the regulatory T cell cerebroprotective effects (41). The diminished amount of IL-10 mRNA in ischemic hemispheres of CD73 2/2 mice could, therefore, be a contributing factor to increased cerebral expression of proinflammatory cytokines and increased cerebral infarct volumes.
Activation of innate immune mechanisms is often triggered or amplified by expression of a class of molecules on the surface of macrophages. These molecules, termed activation/costimulatory molecules, engage ligands external to the macrophages, which lead to an internal activation signal. CD80 and CD86 are two classic costimulatory molecules implicated in innate immune functions. Our data showed that macrophages from ischemic hemispheres of CD73 2/2 mice express higher levels of CD80 relative to CD86 when compared with macrophages from WT mice. This observation is consistent with the similar observations in SJL mice immunized for induction of experimental autoimmune encephalomyelitis (42). Temporal upregulation of CD80 surface expression relative to CD86 was found on splenic and CNS, but not on lymph node B cells, T cells, and macrophages, demonstrating CD80 predominance on cells infiltrating the CNS of mice with active disease (43). In that work, the changes in the expression patterns in both spleen-and CNS-infiltrating F4/80 + macrophages had important functional consequences, as CD80 becomes the predominant costimulatory ligand in T cell proliferation assays. Srinivasan et al. (14) showed that increased expression of CD80 on APCs, after induction of experimental autoimmune encephalomyelitis, correlated with greater inflammation and disease progression in the CNS. Increased infiltration of CD80-expressing macrophages was also observed in colonic inflamed mucosa in a murine model of inflammatory bowel disease (44). If CD80 and CD86 are able to direct and activate the adaptive immune response, the outcome will depend on the level of expression of the two molecules on APCs. On most APCs, CD86 is constitutively expressed, whereas CD80 is induced after activation (45). Despite interacting with the same receptors (CD28/CD152), differences in the nature of expression (constitutive versus induced) and kinetics of interaction of CD80 and CD86 have been attributed to their different functions in disease processes. This study demonstrates that CD73 presence in inflamed CNS milieu can regulate expression of costimulatory molecules such as CD80 (B7-1) and CD86 (B7-2) on infiltrating macrophages, thereby playing an important role in the regulation of the immune response in the setting of stroke.

Rescue of the CD73-null phenotype by 59-NT
We showed that i.p. injection of soluble 59-NT resulted in significant attenuation of infarct volume, improved functional outcome, and reduced leukocyte infiltration 48 h after induction of brain ischemia. This supports our hypothesis that the conversion of extracellular AMP to adenosine by the 59-NT, CD73, is a key control point for the regulation of vascular inflammation associated with brain ischemia. A significant increase in leukocyte adherence to the vascular endothelium has previously been shown in CD73 2/2 mice in a model of cremaster muscle ischemiareperfusion injury (46). CD73 and its active metabolite adenosine dose-dependently inhibited VCAM-1 but not ICAM-1 ex-pression. In a model of wire-induced vascular injury, CD73 deficiency increased monocyte adhesion to ex vivo-perfused carotid arteries that was mediated by VLA-4/VCAM-1 (46). These data are in agreement with our observation, as we found more VCAM-1 mRNA in ischemic hemispheres of CD73 2/2 mice compared with WT controls. Moreover, CD73-deficient mice had more severe vascular leakage and greater neutrophil infiltration during hypoxia than WT animals, which could be reversed by administration of soluble CD73 (9). These data suggest that CD73 can control inflammation by regulating the dynamics of leukocyte-endothelial interaction.

Stroke sequelae in CD73 chimeric mice
Because CD73 is expressed both on brain-resident cells and circulating leukocytes, a bone marrow transplantation model was used to create chimeric mice to distinguish between the relative contributions of resident and circulating CD73 in brain ischemia. We found that brain tissue-derived CD73 attenuated both macrophage influx and activation in the settings of stroke CD73, whereas CD73 on circulating cells inhibited these inflammatory parameters to a lesser degree. Our data cannot rule out the possibility that circulating cells may provide an additional source of adenosine available at the vascular endothelial interface.
In summary, our studies using CD73 2/2 mice demonstrated a significant protective role for CD73 during cerebral ischemia. This protection was primarily due to the attenuation of the number of infiltrating inflammatory cells as well as suppression of macrophage activation and was associated with attenuated expression of proinflammatory cytokines IL-6, KC, TNF-a, and IL-1b. These conclusions are supported by our finding that soluble 59-NT mitigated the inflammatory phenotype of both CD73 2/2 and WT animals. Based on these observations, we propose the novel conclusion that CD73 derived from brain-resident tissue has a critical role in controlling leukocyte extravasation and tissue injury after brain ischemia.