Given that brain depends upon rapid and continuous delivery of oxygen to maintain normal function and survival, it may have developed unique molecular responses to hypoxia as compared to other organs and organisms. In spite of the crucial role of oxygen in normal brain function, there have been relatively few studies of brain hypoxia at the whole genome level. Previous studies using early-generation microarrays found changes of expression in whole brain of neonatal rat (34 genes up-regulated, 42 genes down regulated) and in cerebral cortex of adult mice (29 genes up regulated) [1, 2]. Thus, this study was undertaken in part to examine the brain hypoxia response at the whole genome level in vivo.

One of the major goals of this study, however, was to define the regional, whole genome changes of gene expression that could account for the phenomenon of "hypoxia preconditioning" (HP)[3–7]. In the first described brain HP model, exposure of neonatal rats to 3 hours of 8% oxygen almost completely protected against hypoxia-ischemia induced infarction [3]. In adult mouse brain, hypoxia preconditioning 24 to 48 hours in advance can reduce infarct volume up to 64% [6, 8]. Moreover, the protection provided by HP can last 5-8 weeks [9, 10]. Thus, HP may offer an important platform for the discovery of neuroprotective targets. Though extensively studied, the precise mechanisms accounting for HP remain unknown. Since HP requires new RNA and protein synthesis, we set out to study the whole genome transcriptional response to HP in the adult C57BL/6 mouse brain using Affymetrix Expression Arrays (> 39,000 transcripts) [7, 11, 12]. The dynamic gene expression changes were assessed over the following 24 hours after HP- a time by which the brain would be protected against focal ischemia [7]. Moreover, the expression changes were assessed on a region by region basis: the regional differences of the brain regarding its expression response to hypoxia/HP have not been studied systemically. It is known that certain brain regions and cells, such as hippocampal pyramidal neurons, striatal medium spiny neurons and cerebellar Purkinje neurons, are quite vulnerable to hypoxia/ischemia whereas others can be relatively resistant[13, 14]. It is still not clear whether hypoxia/HP responses could have predominantly engaged common or different mechanisms in different regions. One previous study, using a neonatal rat hypoxia-preconditioning model, showed differential molecular responses in neocortex, hippocampus, striatum and thalamus, even though HP produced similar protection in these regions [15]. Thus, our current study further examined the regional response to HP systemically at the genomic level to determine whether there were common as well as region-specific responses to hypoxia (8% O₂) in the brain.

Our study examined both time- and region-dependent transcriptional responses induced by HP in the adult mouse brain. Besides hypoxia inducible factor (HIF), various nuclear receptor transcription factors have been found to play important roles in regulating both the region-independent and region-specific gene expression responses to HP. Different brain region exhibited differential but coordinated responses: in the forebrain, down regulation of gene expression was predominant during peak response period while the hindbrain, especially the cerebellum, demonstrated predominant up-regulation responses to HP. Surprisingly, the cerebellum demonstrated the most profound gene expression response among all regions and may play a pivotal role in the protective effect of HP in vivo.

The results demonstrate both time- and region-dependent gene expression responses to hypoxia. Of the 2324 hypoxia-regulated transcripts, most are specifically regulated in just a few brain regions. Some of them are under the control of similar transcription factors across regions like HIF-1α, GR and VDR/RXR, while others are not. A unique finding is that a large number of cerebellum-specific hypoxia responsive genes appear to be uniquely regulated by the HNF4A transcription factor. The finding of more up-regulated than down-regulated genes in cerebellum and other hindbrain structures compared to more down-regulated genes in forebrain, points to possible sacrifice of forebrain cognitive functions for support of life-preserving hindbrain functions during periods of marked hypoxic stress.

Predominant down-regulation of genes over up-regulation in forebrain following hypoxia could conserve energy but compromise specific forebrain functions. However, there are still many up-regulated genes in forebrain following hypoxia, and these are more involved in regulation of cell survival and cell growth than the down-regulated genes. This selective up regulation of cell survival genes, moreover, was a common feature for all the other investigated brain regions. Specific pathways included those for HIF signaling, glucocorticoid signaling, and P53 signaling, PI3K/AKT and ERK/MAPK signaling. Though the degree of hypoxia used in this study does not produce cell death [3, 22, 23], it may produce diffuse single strand DNA breaks that require repair [24, 25]. This may account for the up-regulation of p53, PI3K/AKT and other pathways involved in DNA repair [26]

Among the hypoxia regulated genes, many are known HIF target genes [5, 27, 28]. Though HIF-1α plays a pivotal role in hypoxia sensing and signal transduction, its role in mediating hypoxia induced preconditioning effect is controversial, with several studies suggesting that HIF-1α might promote ischemic injury [29–32]. Microarray studies with brain specific HIF-1α knock out mouse even show that HIF-1α is dispensable for the hypoxia response[31]. If HIF related responses do not account for hypoxia preconditioning, then there must be other transcription factors and genes. Indeed, over half of the hypoxia regulated genes common to all brain regions lack HIF binding sites in their promoters but have sites for other hypoxia responsive transcription factors.

The glucocorticoid transcription regulation pathways are not commonly thought of as being associated with hypoxia and other types of preconditioning. However, these pathways might be considered since our promoter analyses show that genes with GRE (Glucocorticoid Responsive Elements) in their promoters are over-represented in those induced by one hour of hypoxia(Table 2), which is sufficient to produce hypoxia preconditioning [8]. Glucocorticoids are systemic stress hormones and can activate the glucocorticoid receptor (GR), a transcription factor that acts on GRE promoter elements [33]. Although the role of glucocorticoids in the context of HP has not been addressed, it is known that hypoxia (8% oxygen) increases glucocorticoids in the blood within 30-60 minutes [34–36]. Glucocorticoids can augment the expression of HIF1α-dependent genes via a direct interaction with HIF-1α [37]. Administration of synthetic glucocorticoids can decrease high altitude sickness, i.e. brain edema due to hypoxia [38]. Glucocorticoids applied 20 hours in advance can protect rat neurons from excitotoxin-induced apoptosis through PI3K/Akt-dependent phosphorylation of Cdkn1a [39]. Of note, our data show that hypoxia increased the expression of Cdkn1a (Fig. S2). According to our data, hypoxia induces many other glucocorticoid-inducible genes, including Tsc22d3, Sgk1, and Nfkb1a, as well as a member of the GR complex - FKBP5 (FK506 binding protein 5). Thus, glucocorticoid signaling could produce an alarm response to the hypoxia stress to augment the adaptive genomic response. Whether GR signaling plays a role in preconditioning has yet to be tested.

There is probably not a single pro-survival molecule or one pro-survival signaling pathway that accounts for hypoxia preconditioning. The redundant activation of multiple pro-survival mechanisms has been observed in many models, including hypoxia and ischemic preconditioning models [9, 17, 40]. Cellular protection may occur at many levels - and thus multiple pathways are needed to protect the cells as shown in Table 1.

The hypoxia preconditioning response appears to have little in common with an injury response. For example, very few transcripts were regulated by both hypoxia and ischemic stroke in the cerebral cortex (Additional file 2 table S5). A similar result has been found between an ischemic preconditioning and an ischemic stroke-induced gene expression response [17]. This result indicates that the protective effect induced by hypoxia is not achieved through a simple rehearsal of the ischemia event, but through a reprogramming of the transcriptional response [17].

Unlike the forebrain, there were many more up-regulated than down-regulated genes in cerebellum and other hindbrain structures including the pons and medulla. This could be due to the fact that brainstem and cerebellum play critical roles in maintaining cardiac and respiratory functions during periods of severe hypoxia and that higher order cognitive functions are shut down to conserve energy stores during hypoxia[41–43]. Indeed, hypoxic mice and other mammals are lethargic and inactive, but have marked increases of respiration, heart rate and autonomic responses during hypoxia. Our findings on the regional differences serve to emphasize that some aspects of "hypoxia preconditioning" in vivo may not be recapitulated by hypoxia or ischemia preconditioning in vitro.

The large number of up-regulated genes in cerebellum following hypoxia could suggest an important central role for cerebellum in the brain's preconditioning response. As noted above, cerebellum may play a role in coordinating respiratory, circulatory and other autonomic responses to hypoxia[41, 42]. Though possible, it is not clear why there would not be similar numbers of up-regulated genes in the pons and medulla where these functions are primarily controlled. An alternative role for cerebellum in hypoxia preconditioning might occur through "central neurogenic neuroprotection," which is produced by stimulation of the cerebellum fastigial nucleus (FN) [44]. Reis and coworkers found that one hour of electrical stimulation of the FN prior to injury protected the rat brain from subsequent focal ischemia 1-3 days later [45, 46]. The time course of this central neurogenic neuroprotection is similar to that of hypoxia preconditioning. The mechanism of neuroprotection is unknown, though FN stimulation pretreatment can decrease inflammation following ischemia and can modify the intrinsic sensitivity of forebrain neurons to apoptotic stimuli [47, 48]. The many up-regulated genes in cerebellum produced by hypoxia imply that the cerebellum was indeed "stimulated." It is tempting to speculate that changes of gene expression in cerebellum during the hypoxia may be responsible, at least in part, for the hypoxia-induced neuroprotective effect through the "central neurogenic protection" mechanism. Indeed, an excitotoxic lesion of FN neurons prior to preconditioning ischemia blocks the neuroprotective effect of ischemic preconditioning against global ischemia in the rat [49]. Even though the cerebellar FN-mediated neuroprotection effect has been well documented, the neuronal circuitry from cerebellum to forebrain responsible for the preconditioning still has not been identified.

Although the cerebellum exhibits a different response to hypoxia compared to the hippocampus, the response in the cerebellum is still mediated to an extent by well-known hypoxia responsive signaling pathways. Besides HIF-1α, a series of nuclear receptor transcription factors appear to have important roles in the cerebellum-specific up regulation response, including GR, NRF2 (NF-E2-related factor 2), VDR/RXR, and estrogen receptor. One of the major findings of this study, however, is that over half of the 1241 hypoxia-regulated genes unique to cerebellum have binding sites in their promoters for HNF4A, another member of the nuclear receptor family.

As a tissue-specific transcription factor, hepatocyte nuclear factor 4A (HNF4A) was originally identified as a key transcription factor responsible for expression of hepatic-specific genes and is involved in tissue-specific cell differentiation and energy metabolism [20, 50]. Mutations in this gene have been associated with monogenic autosomal dominant non-insulin-dependent diabetes mellitus type I, which stresses the important role of HNF4A in energy metabolism [51]. In addition, conditional HNF4A-knockout mice have down-regulation of a series of cell adhesion and junction molecules and severe failure of epithelial transformation [52], which is consistent with the potential HNF4A target genes regulated in the hypoxic cerebellum.

HNF4A is also known to be involved in the hypoxia response. HNF4A can interact directly with both HIF-1α and HIF-1β and is required for erythropoietin (EPO) transcription during hypoxia through interaction with HNF4A binding element in the EPO enhancer in the liver and kidney [53]. Mutation of the HNF4A binding site in the EPO promoter abolishes hypoxic induction of EPO [54]. According to our data, HNF4A appears to activate transcription for a large number of cerebellum-specific hypoxia responsive genes. Though a role for HNF4A in the hypoxia response in brain has yet to be proven, the current findings suggest that it plays a major role in the transcriptional response to hypoxia in the mammalian cerebellum.

To the best of our knowledge, our data uncovered so far the most complete HP-induced gene expression responses in the adult mouse brain. Our results confirm that HP elicits both time- and region-dependent transcriptional responses that are required for maturation of the delayed protection effect of HP. Both region-independent and region-dependent expression changes were observed. Developmentally closely related regions share more commonality than remotely related regions: roughly, the response in the forebrain regions is distinctive from that in the hindbrain regions. At the same time, from the whole brain view the region-dependent response in vivo appears to be a well coordinated one under limited energy sources, with hindbrain function being well supported somewhat at the expenses of the forebrain function. Nonetheless, selective up-regulation of cell survival related genes appears to be a common feature for all the brain regions including the forebrain. A large proportion of HP-regulated genes themselves are gene expression regulators. The data analysis suggested the complexity of such underlying transcriptional regulation mechanism: both universal transcriptional regulation mechanism, such as HIF and GR, and region-specific transcription factors, such as HNF4A, have been responsible for the region-specific gene expression changes which is related to cell survival. It is likely that a cascade of signaling pathways rather than any single pathway has mediated the HP response and protection mechanism. Our data also revealed novel transcriptional regulation mechanisms that may have been underappreciated in HP neuroprotection mechanism, and indicated the potentially important role of cerebellum and other hindbrain structures in HP. The exact mechanism and function of the region-dependent responses to HP remain to be fully elucidated.

The following additional data files are available with the online version of this paper: Additional file 1 contains Figure S1 to S5; Additional file 2 contains Table S1 to S5.