Introduction

Stroke is the third leading cause of death and the leading cause of long-term disability, with very few effective treatments (Benjamin et al., 2017). The discovery of ischemic tolerance may lay the groundwork for future therapeutic development (Li et al., 2017). In 1986, Murry et al. reported that several episodes of non-injurious ischemia to the dog heart led to a 25% reduction in infarct size produced by a subsequent sustained occlusion of the coronary artery. They named this phenomenon “preconditioning” (Murry et al., 1986). Up to now, preconditioning has been observed in multiple animal species and across various organs, including the brain (Kitagawa et al., 1990), liver (Salomao et al., 2012), kidney (Cao et al., 2010), spinal cord (Liang et al., 2012), and pancreas (Hogan et al., 2012), etc. It should be noted that non-lethal ischemia is not the only means that confers brain ischemic tolerance. Other methods include hypothermia, hypoxia, cortical spreading depression, inflammation, oxidative stress, and epilepsy, etc. (Stetler et al., 2014). This phenomenon is termed “cross-conditioning” or “cross-tolerance”, denoting that stimulus of one type could provide protection against subsequent injury of an entirely different type (Stetler et al., 2014).

Various experimental models of brain IPC have been applied in the preclinical studies. In rodents, cerebral ischemia can be primarily induced by either global (forebrain) ischemia or focal ischemia models. Two global ischemia models have been reported: permanent occlusion of the vertebral arteries combined with brief occlusion of bilateral common carotid arteries (4VO), and bilateral carotid artery occlusion with systemic hypotension (2VO). In the first study that described IPC in the brain in 1990, Kitagawa et al. reported that IPC rendered hippocampal neurons tolerant to subsequent lethal ischemia in gerbils. Interestingly, although they merely occluded bilateral carotid arteries, they actually induced a 4VO global ischemia model due to lack in the posterior communicating artery in gerbils (Kitagawa et al., 1990). Focal ischemia is typically induced by a temporary or permanent occlusion of the middle cerebral artery (MCA), because it represents a clinical course of prodromal transient ischemic attacks with a subsequent stroke. The intraluminal suture MCA occlusion (MCAO) model, also known as the so-called “filament model”, is induced by inserting a suture filament through the internal carotid artery up to the initial segment of the MCA, and removing the suture after a period of time, yielding local ischemia/reperfusion (I/R) injury (Glazier et al., 1994). The distal MCAO (Morancho et al., 2012) usually requires a craniectomy to expose the MCA, which can be occluded by electrocoagulation and additional transection, resulting in permanent occlusion, or alternatively by a microclip, a snare ligature, or a tungsten hook temporarily interrupting the blood flow of MCA (Shigeno et al., 1985; Buchan et al., 1992; Popa-Wagner et al., 1999). Notably, a novel laser-induced photochemical reaction technique enables us to make a pinpoint permanent occlusion of a vessel, leaving the dura and the skull intact (Dietrich et al., 1987).

Disruption of continuous oxygen and glucose supply can lead to neuronal death within a few minutes. The oxygen-glucose deprivation (OGD) in cell cultures or tissue slices is the most widely used model system mimicking ischemic injury in vitro. The OGD method was first established by Goldberg and Choi in mixed neocortical cultures (Goldberg and Choi, 1993), and Bruer et al., who further modeled in vitro IPC in a neuronal-enriched culture, observed that neuronal death was significantly reduced with sublethal OGD 48-72 h before lethal OGD (Bruer et al., 1997). The in vitro models provide a useful tool to study in the mechanisms of IPC, which may be generalized into the whole-animal model systems.

The two time-windows of protection in IPC are well-established and thoroughly reviewed (Stetler et al., 2009). The rapid window, occurring within minutes after IPC, provides only a short-lived (1–2 hours) protection against lethal ischemia (Schurr et al., 1986; Perez-Pinzon et al., 1997). Moreover, the rapid window seems to be less universal, as no protection was observed in the rapid window in a global ischemia model in gerbils (Kato et al., 1991). Compared to the rapid window of IPC, the delayed window confers long-lasting and more robust neuroprotection. The delayed window starts 24 hours after IPC, peaking at 48 hours to 72 hours, and lasting up to 1 week (Chen and Simon, 1997). It is broadly accepted that de novo protein synthesis is required for IPC-mediated long-term neuroprotection; however, the identification of proteins that are necessary, and their relative regulatory mechanisms are poorly understood (Barone et al., 1998; Dirnagl et al., 2003; Koch et al., 2014).

Phases of IPC-mediated gene expression

To understand the mechanisms underlying the delayed window of IPC, we need to map how brief ischemia leads to subsequent protein synthesis after IPC. Three distinct phases have been described in the process of protein synthesis: the triggering phase, the signal transduction phase, and the genomic phase (Della-Morte et al., 2012). The triggering phase involves release of receptor agonists that bind membranes receptors, mainly G protein-coupled receptors. Next, in the signal transduction phase, intracellular secondary messengers transduce the signals and activate transcription factors. And finally, the last phase, the genomic phase, refers to genetic regulation by transcription factors.

1. The triggering phase
The most important receptor agonist in the triggering phase is adenosine, which is a purine nucleoside (Heurteaux et al., 1995; Perez-Pinzon et al., 1996). It is released by cultured neurons and can be detected as soon as 60 min after OGD (Parkinson and Xiong, 2004). The rapid release of adenosine potentiates it to mediate neuroprotection in both rapid window and delayed window, via different mechanisms, though. Adenosine-mediated rapid protection is via a decrease in glutamate release and inhibition of calcium fluxes, mainly though adenosine A1 receptor (Heurteaux et al., 1995; Zhou et al., 2004; Shen et al., 2011). Blockade of A1 receptor, leading to abrogation of IPC-mediated rapid protection (Perez-Pinzon et al., 1996). On the other hand, adenosine-mediated delayed protection is through passing on the signals to secondary messengers and pushing the story forward to the signal transduction phase and the genomic phase.

In addition to adenosine, release of other receptor activators and redox signaling are also reported in the triggering phase. For example, opioid receptors and nicotinic acetylcholine receptors are activated in IPC (Peart et al., 2005; Rehni et al., 2008; van Rensburg et al., 2009), and bradykinin and erythropoietin are released and confer neuroprotection through activation of multiple signaling pathways (Baker, 2005; Ping et al., 2005; Noguchi et al., 2007).

2. The signal transduction phase
Protein kinase C (PKC) pathway is a central player in the signal transduction phase. First, PKC targets the adenosine triphosphate (ATP)-sensitive potassium channel, which prevents calcium overload in the mitochondrion and slows the tricarboxylic acid (TCA) cycle (Critz and Byrne, 1992; Domanska-Janik and Zablocka, 1993; Nishi et al., 1999; Ivannikov et al., 2010). Consequently, scarce energy resources can be conserved, and excessive reactive oxygen species (ROS) product can be eliminated. In addition, opening ATP-sensitive potassium channels could promote uptake of glutamate by astrocytes (Sun et al., 2008), which reduces excitotoxicity in ischemia. Second, PKC also upregulates the expression of Sirtuin 1 (SIRT1) through activating nicotinamide phosphoribosyltransferase (Morris-Blanco et al., 2014), which boosts robust neuroprotection via both epigenetics-dependent and -independent mechanisms (Koronowski et al., 2015).

Other signaling pathways also play a role in the signal transduction phase, such as glycogen synthase kinase (GSK) 3β, mitogen-activated protein kinase (MAPK), c-Jun N-terminal kinase (JNK), and Janus kinase/signal transducers and activators of transcription (JAK/STAT), etc. (Otani, 2008), and regulate a large variety of cellular responses, including growth and proliferation, development and differentiation, inflammation, and apoptosis.

3. The genomic phase
The genomic phase is initiated by nuclear translocation of transcription factors in response to intracellular signal transduction. Of note is that, transcription factors of the activator protein-1 family are activated that induce the translation of proto-oncogenes such as c-fos, jun B, c-jun and jun D (Truettner et al., 2002), and thereby play a critical role in determining the fate of cells after ischemic injury. Hundreds of genes are changed after IPC, by either upregulation or downregulation. This is termed “genomic reprogramming”, which expands the previous concept “new protein synthesis” in IPC by stating that downregulation of some genes should also be considered in this phase. The reprogramming profile following IPC is described by either genomic-based or proteomic-based studies and will be reviewed in Section 3.

In addition, the important role of epigenetics has recently been introduced to the stroke field, based on the finding that stroke is associated with increased DNA methylation (Endres et al., 2000) and histone acetylation (Meisel et al., 2006). However, it remains largely unexplored how epigenetics regulates IPC responses; thus, it might be a promising future direction for a better understanding of the IPC biology. The remodeling of epigenetic marks is discussed in Section 4.

Genomic reprogramming after IPC

1. General features of genomic reprogramming after IPC
As previously mentioned, there exists a “cross-conditioning” phenomenon in the brain. Although multiple stimuli lead to similar ischemic tolerant status, the genomic reprogramming patterns they trigger are not the same. For example, IPC is associated with downregulation in genes related to metabolism and channels, while lipopolysaccharide preconditioning altered inflammatory patterns along with upregulation in defense genes are associated with type I interferons (Stenzel-Poore et al., 2007). Thus, it is concluded that the nature of preconditioning stimulus determines the neuroprotective phenotype and genomic reprogramming pattern (Stenzel-Poore et al., 2003; Vartanian et al., 2015).

Both IPC and I/R are of ischemic basis, though with different degrees in severity. As a result, one may speculate that the genomic reprogramming processes triggered by IPC should cover similar categories of genes and have similar pattern with I/R. Surprisingly, using microarray analysis to investigate gene expression after IPC and I/R in mice, Stenzel-Poore et al. found no common overlap among the IPC, I/R and IPC followed by I/R (IPC+I/R) groups (Stenzel-Poore et al., 2003). Moreover, when comparing IPC+I/R with I/R alone, up to 83% genomic responses were unique to IPC+I/R (Stenzel-Poore et al., 2007). This number is much lower, about 39%, in primary neuronal cultures with 45 min OGD mimicking in vivo IPC (Prasad et al., 2012). The difference between in vivo and in vitro setting is fairly reasonable, since other than neurons, glial cells and microvessel cells are also actively engaged in the adaptation to ischemia and/or IPC, and take part in forming an integrated network, termed neurovascular unit (Zhang et al., 2012). Collectively, IPC is associated with a unique programmed phenotypic alteration and confers a so-called “pro-survival” phenotype (Kawahara et al., 2004), which results in “programmed cell survival” (Gidday, 2006).

Both upregulation and downregulation of genes participate in IPC-mediated genomic reprogramming. Multiple studies have revealed that I/R prefers upregulation than downregulation, and IPC + I/R induces more pronounced downregulation. For example, up to 86% genes regulated showed increased expression in I/R, while 77% of regulated genes showed decreased expression in IPC + I/R (Stenzel-Poore et al., 2003). IPC alone is associated with a slight trend preferring upregulation than downregulation (Stenzel-Poore et al., 2003; Stenzel-Poore et al., 2003; Feng et al., 2007).

In this section, we review the overall genomic reprogramming pattern mediated by IPC, particularly focusing on genomic-based studies and proteomic-based studies but not those studying an individual protein or pathway, and try to disclose the commonly shared mechanisms that are necessary for IPC (Table 1).

Table 1. List of genomic reprogramming studies in IPC

2. Categories of genes regulated by IPC

2.1 Transcription factors
Transcription factors are effectors of IPC that directly regulate gene expression and protein synthesis, and several groups of transcription factors have been involved in IPC. For example, silencer factor B and C-jun are regulated in IPC (Feng et al., 2007), and c-Fos, Jun B, and zinc finger Egr family members are also upregulated (Kawahara et al., 2004). In the study conducted using a MetaCore’s network algorithm mapping of pathway analysis, it is revealed that protein expression changes in IPC are dependent mainly on three proteins: androgen receptor, HIF-1, and NF-κB (Scornavacca et al., 2012). Additionally, preconditioning in rat hippocampal slices significantly increases in transcription activities of Fosl1, Jun D, Med13 and Nr4a1, peaking around 3 h post IPC (Benardete and Bergold, 2009).

2.2 Ion channels and transporters
Ion channels and other transporters are reported to be downregulated by IPC (Stenzel-Poore et al., 2003; Stenzel-Poore et al., 2004). IPC significantly decreased potassium channel Kv1.5 expression in I/R brain in vivo. This finding is further validated by electrophysiological studies in in vitro rat cortical neuronal cultures which showed decreased outward potassium currents and whole-cell conductance after IPC (30 min OGD) (Stenzel-Poore et al., 2003). Consistently, downregulation of voltage-dependent anion channel 1 with proteomic analysis in the CA1 region after global IPC was reported (Nakajima et al., 2015). In another study (Feng et al., 2007) applying global ischemia following global IPC, calcium signaling pathway is among the most significantly modulated pathways. Since a four-way pairwise comparison was performed, their findings are more likely to reveal pathways necessary for IPC-mediated protection. Regrettably, they did not identify whether and which genes in this pathway were upregulated or downregulated. In addition, it is worth noting that they were comparing it with tissues from IPC group vs IPC + MK801 group, in which MK801 is a non-competitive NMDA receptor, which could abolish the protection provided by IPC (Zhang et al., 2009). Similarly, Kawahare et al. reported downregulation in calcium signaling genes, IP3 kinase and IP3 receptor, mediated by a prior IPC (2 min global ischemia) in ischemic rat brain (6 min global ischemia) (Kawahara et al., 2004).

However, Dhodda et al.'s study disagrees with the aforementioned results (Dhodda et al., 2004). They performed a 10-min MCAO to induce IPC in spontaneously hypertensive rats, and total RNA from the MCA territory was subjected to GeneChip analysis. They reported upregulation in ion channels, including sodium channel scn6a, potassium channel KCNJ13, and calcium channel CNGA3 and P2X-associated ATP-gated channel. Consistently, Benardete and Bergold reported upregulated calcium signaling pathway genes after IPC in ex vivo hippocampal slices (Benardete and Bergold, 2009).

Acid-sensing ion channels, ASICs, are reported to be downregulated after IPC (Pignataro et al., 2011), although no whole-genomic studies have revealed a similar phenomenon.

2.3 Signaling transduction
Upregulation in genes related to MAPK signaling pathway is well-documented both in vivo (Dhodda et al., 2004; Kawahara et al., 2004; Feng et al., 2007) and ex vivo (Benardete and Bergold, 2009). An article also reported an upregulation of other genes related to signaling transduction, including SMAD-1 and -7, guanylyl cyclase, and retinoid acid receptor, etc. (Dhodda et al., 2004). JAK/STAT signaling is reported to increase after IPC ex vivo in hippocampal slice culture (Benardete and Bergold, 2009), while cyclic adenosine triphosphate signaling and protein kinase A signaling are decreased in neuronal cultures after IPC (Prasad et al., 2012).

2.4 Inflammation
Toll-like receptors (TLR) play a major role in innate immune response against various insults. A hierarchical analysis revealed significant downregulation in genes in TLR signaling pathway (Feng et al., 2007). This is further confirmed by Western blot, showing decreased expression of cyclooxygenase 2, a TLR4 downstream inflammatory factor. However, in an ex vivo study using rat hippocampal slice culture subjected to IPC (induced by 5 min OGD), increased gene expression of TLR signaling was reported, associated with increased inflammatory factors interleukin (IL)-1a, IL-1b, IL-6 and TNF as confirmed by polymerase chain reaction (RT-PCR) (Benardete and Bergold, 2009). Consistently, in primary neuronal cultures, increased inflammatory responses after IPC were observed (Prasad et al., 2012).

2.5 Neuroplasticity
In spontaneously hypertensive rats, IPC was associated with significant upregulation in neurotrophic factors transforming growth factor (TGF)-α, TGF-β1, and TGF-β receptor, among which the increase in TGF-β1 was confirmed by RT-PCR (Dhodda et al., 2004). Feng et al. (Feng et al., 2007) dug deeper by focusing on the expression of bone morphogenetic protein (BMP)-7, a member of TGF-β superfamily with unique neuroprotective and regenerative effects (Harvey et al., 2005; Heinonen et al., 2014). They confirmed upregulation in BMP-7 in IPC with RT-PCR, Western blot and immunohistochemistry in the dentate gyrus. Increase in TGF-β pathway genes was also observed ex vivo in hippocampal slice culture after IPC, together with genes related to other neurotrophic pathways, like Wnt, ErbB, and vascular endothelial growth factor signaling (Benardete and Bergold, 2009). In another study with in vitro neuronal culture, IPC increased cell cycle genes and neurotrophic genes (Prasad et al., 2012). Naylor et al. confirmed that IPC alone could increase in a panel of growth factors, including insulin-like growth factor, fibroblast growth factor, TGF-β, epithelial grow factor and platelet-derived grow factor-A (Naylor et al., 2005).

2.6 Metabolism and ribosome pathway
In Stenzel-Poore's study (Stenzel-Poore et al., 2003), the authors reported significant decrease in metabolic gene expression; interestingly, they further interpreted this feature as a stimulation to hibernation, in which a suppression of energy use contributes to neuroprotection (Drew et al., 2001). Their findings agree with a later study using proteomic analysis (Scornavacca et al., 2012). While Feng et al.'s study, using a four-way pairwise comparison, elicited opposite results. They reported a significant increase in genes related to ribosome pathway, oxidative phosphorylation and protein synthesis (Feng et al., 2007). They also suggested that activated metabolism and ribosome pathway could facilitate the recovery protein processing machinery after injurious ischemia. Consistently, an ex vivo study inducing IPC in hippocampal slice culture reported increased genes related to glycolysis and gluconeogenesis (Benardete and Bergold, 2009). The above studies seem controversial, but in fact, a possible explanation for this discrepancy is the difference in treatment groups. In the first two studies, the authors compared IPC + I/R group versus I/R alone group, while in the third and fourth studies, the comparison was between IPC alone versus control or IPC alone versus IPC + MK801 group, without injurious ischemia.

However, one article reported opposite effect of IPC using proteomic analysis with two-dimensional electrophoresis, showing decreased mitochondrial aconitase, an enzyme related to TCA cycle, in CA1 region after IPC induced by 3 min global ischemia (Nakajima et al., 2015). A shortcoming of this study is that there was no entire pathway-based analysis, and change in one single protein can hardly reflect the general trend of one pathway.

2.7 Heat shock proteins (HSPs)
HSPs are important stress sensors in the cellular system. Increased HSP expression by IPC is reported in multiples studies (Stenzel-Poore et al., 2003; Dhodda et al., 2004; Kawahara et al., 2004). To specify, after 15 min MCAO as IPC in mice, Hsp70 increased at 24 h, and Hspb2 at 72 h (Stenzel-Poore et al., 2003). IPC also contributed to an additional increase in Hsp70, 105 and 10 by I/R after 24 h (Stenzel-Poore et al., 2003). In a global IPC model, the only upregulated Hsp mRNA after IPC is Hsp70 (Kawahara et al., 2004). In spontaneously hypertensive rats, IPC (10 min MCAO) induced increase in Hsp70, 27, 90, 10 and 60 as confirmed by RT-PCR (Dhodda et al., 2004), and the increase in HSP70 was then validated by Western blot and immunohistochemistry in vivo. The upregulation of HSP70 after IPC was also reported by using proteomic analysis by two-dimensional electrophoresis coupled to liquid chromatography-tandem mass spectrometry (Scornavacca et al., 2012).

2.8 Redox signaling
Although it was mentioned that genes related to oxidative stress were regulated by IPC, Western blot reveals no significant changes in the expression of superoxide dismutase, an ROS scavenger (Scornavacca et al., 2012). More glutathione S-transferase expression was observed in CA1 than CA3/dentate gyrus region, but no significant changes between the sham and IPC groups (Nakajima et al., 2015). Although some sporadic studies did elicit increased antioxidants induced by IPC both in vivo and in vitro (Dozza et al., 2004; Holtzclaw et al., 2004; Danielisova et al., 2005; Shokeir et al., 2014), no proteomic or genomic-based studies reported the indispensable role of redox signaling.

Given that buffering oxidative stress abolished IPC-mediated protection (Puisieux et al., 2004; Narayanan et al., 2017), a couple groups have reported that the mild oxidative stress in preconditioning initiates redox signaling and activates nuclear factor erythroid 2-related factor 2 (Nrf2), a master transcription factor the upregulates anti-oxidative enzymes, in astrocytes (Bell et al., 2011; Narayanan et al., 2017). In support, our group also found an important role of Nrf2 after IPC. We found the nuclear translocation of Nrf2 after IPC and validated the indispensable role of Nrf2 for IPC (Yang et al., 2016a).

3. Limitations and future directions

Despite the progress in genomic-based and proteomic-based techniques in the past decade, all the above studies failed to make a breakthrough to reveal a common “tolerasome” or to provide clues in finding biomarkers for IPC (Meller and Simon, 2013; Koch et al., 2014). Besides different models and different comparison in different studies, most studies only applied correlation analysis, without further determination on the causal relationship. In addition, the sample sizes may be too small which lowered the power in these studies. A third limitation is lack of cellular specificity, since samples from most in vivo studies were extracted from whole tissue. Last but not least, gene regulation at an epigenetic level is largely ignored. Epigenetics regulates the transcriptional potential through modifying the accessibility of DNA to the transcriptional machinery. In the next section, we are going to review epigenetic reprogramming in IPC.

Remodeling of epigenetic marks after IPC

During the past decade, there has been an emerging role of epigenetics in regulating pathologic process and outcome of stroke. Nevertheless, fewer studies focus on the role of epigenetics in IPC. How IPC alters epigenetic reprogramming is still at a start-up stage and is attracting more and more attention. Mechanisms underlying epigenetic modulation on DNA expression include noncoding RNAs -- typically microRNAs (miRNAs) (Saugstad, 2010), global SUMOylation (SUMO: small ubiquitin-like modifier) (Lee et al., 2010b), direct DNA modification by methylation, and histone protein modifications (Schweizer et al., 2015).

1. miRNAs
In an elegant study conducted by Lusardi et al. (Lusardi et al., 2010), miRNA array was performed in mouse cortex tissues following IPC (15 min MCAO) or I/R (60 min MCAO). IPC + I/R upregulated ~200 miRNAs, and downregulated ~100 miRNAs. However, I/R alone only triggered ~100 miRNA downregulation, suggesting a unique and robust miRNA response to IPC rather than I/R. Further target RNA analysis on those downregulated miRNAs revealed a common target, methyl CpG binding protein 2 (MeCP2), which binds to methylated DNA and helps with gene silencing. They also confirmed that IPC increased MeCP2 expression, and MeCP2 knockout abolished IPC-mediated neuroprotection.

Targeting pathways were also explored, though roughly. Dharap and Vemuganti profiled miRNAs in the cerebral cortices from spontaneously hypertensive rats after IPC and reported a quickly reactive miRNAome (Dharap and Vemuganti, 2010). Among 51 miRNAs that were altered, 26 were upregulated and 25 were downregulated. They further performed KEGG pathway analysis on their target proteins and found that MAPK pathway and mTOR pathway were top the 2 upregulated proteins, and Wnt pathway and GnRH pathway were the top 2 downregulated proteins. Lee et al. performed miRNA profiling covering a total of 360 miRNAs, and found selective upregulation of two miRNA families, miR-200 and miR-182, 3 hours after IPC effects (Lee et al., 2010b). They then transfected some of them and revealed that prolyl hydroxylase 2 and HIF-1 pathways had the best neuroprotective effects.

2. Global SUMOylation
SUMOylation is correlated with transcriptional repression via modulating diverse chromatin enzymes, chromatin associated proteins, and promoter specific transcription factors (Ouyang and Gill, 2009). Elevated SUMOylation was protective against focal ischemia mouse brains (Lee et al.,2009). In the setting of IPC, elevated SUMO-1 conjugation levels observed. Primary cortical neurons were more resistant to OGD when transfected with SUMO-1 or SUMO-2, and SUMO-1-siRNA transfection attenuated IPC-mediated protection (Lee et al., 2009).

3. DNA methylation
In general, DNA methylation at the promoter region results in repression in gene expression, and DNA hypomethylation is associated with gene transcription. DNA methylation is mediated by DNA methyltransferases (DNMTs). Both I/R in vivo and OGD in vitro led to increased DNA methylation (Endres et al., 2000; Hu et al., 2006), and a lower level, but not absence, of DNMT1 is protective against ischemic injury, associated with decreased DNA methylation (Endres et al., 2000). By contrast, in a study describing dynamics of overall DNA methylation after I/R, Meller et al. reported significantly decreased methylated DNA by up to 80% compared to sham, and IPC decreased methylated DNA by~50% (Meller et al., 2015). A possible reason for this discrepancy may lie in their different methodologies. In Endres et al.’s study, DNA methylation was measured by incorporating [3H]-methyl groups into the genomic DNA, which reflects dynamic DNA methylation after I/R-induced DNA injury, while in Meller et al.’s study, methylated DNA was measured by a pull-down of the methyl group and agarose gel analysis, which provides a snapshot of present methylated DNA. Taken together, it can be speculated that I/R-induced DNA damage leads to decrease in methylated DNA which can be attenuated by IPC; DNA remethylation process occurs after I/R.

A number of genes can be regulated by DNA methylation in IPC. The increase of NKCC1, an Na+/K+ co-transporter after I/R correlated to decreased methylation of the promoter (Lee et al., 2010a). Decrease in estrogen receptor methylation led to increased estrogen receptor expression following ischemia (Westberry et al., 2008; Wilson and Westberry, 2009). Moreover, IPC induced methylation to the promoter of thrombospondin 1, an anti-angiogenic factor (Hu et al., 2006; Lawler and Lawler, 2012). Nevertheless, detailed information and the causal relationship between DNA methylation and IPC remain largely unexplored.

4. Histone modification
Histone proteins include histone 2A, 2B, 3 and 4. They participate in forming nucleosomes, which blocks the access of transcription factors to DNA. Histones can be modulated by acetylation, phosphorylation, methylation, and ubiquitination.

4.1 Histone acetylation
Acetylation of histones is the most extensively studied type of epigenetics, not in IPC settings, though. In general, acetylation of histones opens the chromatin configuration and allows transcription factors to bind DNA, and deacetylation of histones is associated with repression in gene expression. Histone acetylation is mediated by histone deacetylases (HDACs). Class I and II HDACs are Zn2+-dependent, and class III HDACs, otherwise known as Sirtuins, are nicotinamide adenine dinucleotide (NAD)⁺-dependent.

HDAC inhibitors are powerful tools to assess the role of HDACs. Class I and II HDAC inhibitors selectively inhibit the zinc hydrolase domain (Bradner et al., 2010). Multiples studies demonstrated neuroprotection against stroke by class I and II HDAC inhibitors via modulating oxidative stress, DNA damage, and inflammation (Qi et al., 2004; Ren et al., 2004; Faraco et al., 2006; Kim et al., 2007). The critical role of HDAC3, a member of HDAC class I, was also confirmed in IPC both in vivo and in vitro, possibly via induction of Hspa1a, Bcl2l1, and Prdx2 expression (Yang et al., 2016b).

NAD⁺ is a vital oxidizing agent of the glycolytic and TCA cycle. Being NAD+ sensors in the cellular system, Sirtuins, especially SIRT1, are required for IPC. In both in vivo and ex vivo studies, SIRT1 activities were increased following IPC, and blockade SIRT1 abrogated IPC-mediated neuroprotection (Raval et al., 2006; Della-Morte et al., 2009; Koronowski et al., 2015). However, one should be aware of SIRT1’s diverse targets, including both histone and non-histone proteins. For example, SIRT1 can either deacetylate MeCP2 to elicit neuroprotection (Zocchi and Sassone-Corsi, 2012), or function indirectly via transcription factors, transcriptional co-factors, or nuclear receptors (Zhang et al., 2011).

4.2 Histone phosphorylation
The best-known function of histone phosphorylation occurs when cells respond to DNA damage. Histone phosphorylation is associated with DNA repair, cell cycle and mitosis (Rossetto et al., 2012). Activation of the 5’-adenosine monophosphate-activated protein kinase signaling pathway, a master pathway regulating cellular energy homeostasis, is the major mechanism that phosphorylates histone. Bungard et al. reported that stress induced H2B phosphorylation at S36 residue; S36A decreased cell viability under stress conditions (Bungard et al., 2010). There is no report on histone phosphorylation in IPC so far.

4.3 Histone mono-ubiquitination
Mono-ubiquitination of histone 2A and 2B contributes to repression of genes that consume ATP; for example, voltage-gated potassium channels, which is a key factor during IPC (Stapels et al., 2010). Such mono-ubiquitination is mediated by polycomb group proteins (PcG). PcG decreased potassium currents in cultured neuronal cells, and PcG knockdown precluded the induction of IPC. Some researchers believe PcG is one of the central modulators of IPC given its diverse target genes including those involved in electron and glucose transport (Brand and Ratan, 2013).

Conclusion

Over the past several decades since the discovery of IPC, piles of evidence have demonstrated the generality of this phenomenon across multiple species and organs that can be induced by diverse in vivo and in vitro models. This has largely expanded the broadness of our understanding in IPC. Disappointingly, our understanding in the depth of IPC mechanisms, such as which pathways, proteins, molecules, and genetic modulation account for IPC-endowed neuroprotection, remains superficial. Besides the two time-windows and the requirement of genomic reprogramming for long-term neuroprotection, no breakthrough has been made during the past two decades, despite extensive work on genomic-based and proteomic-based studies. Epigenetics in the stroke field has gotten increasingly hot, while its role in IPC remains poorly understood. Since drugs involving epigenetic regulation have already been put into clinical trials (Hwang et al., 2013), we expect that exploration on epigenetic regulation in IPC may lead to a new era for stroke and IPC research. The effects of IPC might be an integrated consequence from a complicated network, whose components actively interact with one another. Apparently, we are currently at the very beginning in understanding the integrated mechanisms of the genetic regulation in IPC. We still have a long way to go before we fully understand the biology of IPC, and, what is more important, bring it to clinic to fight against stroke.

Acknowledgements

This work was supported by grants from the National Institutes of Health (NS092810 to FZ). We thank Pat Strickler for her administrative support.