HPC is a phenomenon in which advanced exposure to mild hypoxia reduces the stroke volume produced by a subsequent ischemic attack 
. This phenomenon would be valuable in protecting the brain from stroke especially in patients with high risk of recurrent ischemic stroke. Although the beneficial effect of HPC has been established, the neuroprotection it provides does not last long. Commonly, the peak benefit occurs 1 to 3 days after HPC, with some variation among different protocols and the ischemic animal models employed 
. Because the neuroprotective effect is transient, we employed different hypoxic protocols trying to induce longer-term protective effects before ischemic stroke. It has been reported that 3 days after hypoxic exposure, the neuroprotective function of HPC reaches its peak, yet this effect is lost by the sixth day 
. Our E6d HPC protocol was designed to provide sufficient time for recovery between hypoxic episodes. As expected, E6d HPC provided neuroprotection from an ischemic attack 3 days after the last exposure; however, the protective effect did not last longer than 6 days. Subsequently, we investigated our E3d HPC protocol based on the hypothesis that the peak neuroprotection of the last hypoxic exposure would be reinforced by subsequent exposure to hypoxia, thus sustaining the neuroprotective effect beyond 6 days. To our surprise, E3d HPC resulted in the loss of the protective function of HPC. This provoked our interest in exploring the underlying molecular changes responsible for the loss of neuroprotection.
The HIFs (HIF-1 and HIF-2) are transcription factors that orchestrate the molecular response to hypoxia. During hypoxia, the degradation of HIFs is blocked, leading to accumulation of these factors and upregulated transcriptional activity. HIFs targets, including EPO and VEGF, have been shown to be neuroprotective against acute cerebral damage, and have also been implicated in mediating the neuroprotection afforded by HPC 
. For example, EPO, which is expressed in the central nervous system, has been shown to have potent neuroprotective properties both in vivo 
and in vitro 
Given the importance of EPO and VEGF in mediating the neuroprotective effects of HPC, we speculated that a decrease in the expression of theses HIF-regulated genes could underlie the loss of protection in our E3d HPC mice. Surprisingly, as in the single HPC and E6d HPC groups, we still observed elevated expression of HIF targets in E3d HPC mice. Specifically, the expression level of EPO was elevated (approximately 10-fold) after E3d HPC. This finding is consistent with previous reports of upregulated EPO mRNA and protein levels following single episode HPC 
. Although there is solid evidence that exogenously applied EPO is neuroprotective both in vitro 
and in vivo 
, the effect of endogenous EPO in neuroprotection is still debatable. For example, in transgenic mice overexpressing human EPO, infarct volumes tended to be smaller although this effect did not reach statistical significance 
. In the present study, we observed a discrepancy in HIF target upregulation and loss of neuroprotection in the E3d group. This finding supports the hypothesis that HPC may induce the production of other potentially neuroprotective agents within the brain. This is consistent with our observation of increased HIF target gene expression in the absence of neuroprotection in E3d HPC mice. Taken together, this suggests that HIF target gene upregulation is insufficient to explain HPC-induced neuroprotection, and that other neuroprotective mechanisms must be involved.
One such factor that may also be involved in HPC-induced neuroprotection is extracellular adenosine. Adenosine is a purine nucleoside which acts via four subtypes (A1, A2a, A2b, A3) of G-protein-coupled cell surface receptors to restore homeostasis by increasing blood supply and decreasing energy demand. Adenosine increases blood flow through vasodilation of pre-existing vasculature and by stimulating angiogenesis 
. Adenosine has also been reported to activate presynaptic-A1 receptors, reduce glutamate release and to reduce activation of NMDA receptors. These synaptic inhibitory actions of adenosine exert a powerful neuroprotective effect during hypoxic and ischemic events 
The extracellular concentration of adenosine is controlled by the balance of its production and degradation through enzymes and by transmembrane transport processes. Several of these control mechanisms have been shown to be involved in HPC-induced protection. For example, in a myocardial infarct mouse model, the loss of CD39 increased infarct sizes and abolished cardioprotection by ischemia preconditioning 
. Moreover, increased cerebral infarct volumes and reduced post-ischemic perfusion were also demonstrated in CD39 knock-out mice 
. Interestingly, CD73, which hydrolyzes AMP to adenosine, was also shown to be necessary for cardioprotection by ischemic preconditioning 
. Our data showed a failure to upregulate CD39 and CD73 in our E3d HPC mice, even though other HIF target genes were still highly upregulated. Although HIF-1 also acts as an important transcriptional regulator of CD73 and CD39 
, other factors such as CREB (cAMP response element-binding) are also involved. Interestingly, it is known that the activation of HIF can recruit CBP which binds to and coactivates CREB, thus it is plausible that, under our hypoxic conditions, the repetitive activation of HIF would recruit endogenous CBP, and the coactivation of CBP to CREB would be reduced.
ENT1 is thought to be present on all cell types of the brain, including barrier endothelial and epithelial cells, neurons and glia 
. ENT1 mediates cellular influx or efflux of adenosine, with the direction of movement dependent upon the relative intra- and extracellular concentrations of adenosine. A recent study showed that during ischemic events, neuronal ENT1 activity leads to increased cellular uptake of adenosine, this uptake then decreases adenosine A1 receptor signaling and decreases the neuroprotective effects of adenosine 
. Our data demonstrated that following E3d HPC, the expression of ENT1 was significantly increased.
The upregulated ENT1 expression together with the failure to elevate CD39 and CD73 may account for the decreased adenosine level in the E3d group. To test this hypothesis, we analyzed the influence of specific CD73 and ENT1 antagonists on extracellular adenosine levels in our different hypoxic models. Although the CD73 antagonist suppressed the upregulation of adenosine levels in the single HPC and E6d HPC groups, it did not change the decreased adenosine levels observed in the E3d group. While ENT1 inhibition totally reversed this pattern, thus suggesting an essential role for ENT1 in the decrease of extracellular adenosine levels by frequent hypoxic exposure. By using another ENT1 inhibitor PPF administered intraperitioneally, we went on to test the influence of ENT1 upregulation on stroke volume in the E3d group. As expected, this inhibitor mediated a dose-dependent increased adenosine levels and decreased stroke volume.
The adenosine A1 receptor (A1R) has the highest affinity for adenosine of all the adenosine receptors 
. A1 receptor is generally considered to be protective in the context of cerebral damage. For example, A1R antagonist significantly increased cell death in the CA1 region in a mouse model of global ischemia 
. Furthermore, the neuroprotective effects against stroke associated with delayed preconditioning are attenuated by the administration of the selective A1 antagonist 
. The A2a receptor is also widespread in the central nervous system and binds adenosine with high affinity. However, its role in cerebral ischemia is debatable. Despite evidence showing that the relatively specific A2a agonist, CGS 21680, reduces ischemic damage 
, several relatively specific A2a antagonists have been found to reduce ischemic damage in animal models of global or permanent ischemia 
. In particular, genomic knock-out of the A2a receptor showed attenuated stroke volume in a focal cerebral ischemic model 
. Since we observed a dramatic change in extracellular adenosine levels associated with single and E3d HPC, we went on to evaluate the effects of these two receptors on ischemic outcome under these different conditions.
In accordance with the protective role of the A1 receptor, the A1 antagonist not only increased stroke volume but also reversed HPC-induced neuroprotection, which suggests a critical role for extracellular adenosine in HPC. In contrast, our data support a detrimental role for the A2a receptor in cerebral ischemia although the HPC-induced neuroprotection still existed in the A2a antagonist treated animals. These results suggest that extracellular adenosine, which acts mainly through the A1 receptor, is essential for HPC-induced neuroprotection.
Adenosine is a potent endogenous vasodilator considered to be involved in local blood flow regulation to various tissues. It is also known that ischemic preconditioning before temporary MCAO causes a significant improvement in rCBF 
, although this effect is not thought to be associated with increased angiogenesis stimulated by the preconditioning stimuli. In our present study, rCBF was increased after either a single HPC or E6d HPC intervention, while in the E3d HPC group with the reduced extracellular adenosine levels, rCBF was significantly decreased during the ischemic and reperfusion stage. The ENT1 antagonist mediated a dose-dependent increase in extracellular adenosine levels and reversed the decrease in rCBF and enlarged stroke volume following E3d HPC exposure. We interpret this data to indicate that attenuated extracellular adenosine concentrations in the E3d HPC group led to the decrease in rCBF, which may contribute to the loss of neuroprotection.
In summary, the main findings of our study demonstrate that frequent hypoxic exposure leads to the loss of neuroprotection associated with HPC. By upregulating the expression of ENT1, E3d HPC results in a decrease in the extracellular adenosine levels and the decline in rCBF which are essential for the loss of neuroprotection.