Ischemic preconditioning regulates expression of microRNAs and a predicted target, MeCP2, in mouse cortex

Abstract

Preconditioning describes the ischemic stimulus that triggers an endogenous, neuroprotective response that protects the brain during a subsequent severe ischemic injury, a phenomenon known as ‘tolerance'. Ischemic tolerance requires new protein synthesis, leads to genomic reprogramming of the brain's response to subsequent ischemia, and is transient. MicroRNAs (miRNAs) regulate posttranscriptional gene expression by exerting direct effects on messenger RNA (mRNA) translation. We examined miRNA expression in mouse cortex in response to preconditioning, ischemic injury, and tolerance. The results of our microarray analysis revealed that miRNA expression is consistently altered within each group, but that preconditioning was the foremost regulator of miRNAs. Our bioinformatic analysis results predicted that preconditioning-regulated miRNAs most prominently target mRNAs that encode transcriptional regulators; methyl-CpG binding protein 2 (MeCP2) was the most prominent target. No studies have linked MeCP2 to preconditioning or tolerance, yet miR-132, which regulates MeCP2 expression, is decreased in preconditioned cortex. Downregulation of miR-132 is consistent with our finding that preconditioning ischemia induces a rapid increase in MeCP2 protein, but not mRNA, in mouse cortex. These studies reveal that ischemic preconditioning regulates expression of miRNAs and their predicted targets in mouse brain cortex, and further suggest that miRNAs and MeCP2 could serve as effectors of ischemic preconditioning-induced tolerance.

Introduction

Ischemic brain injuries are among the most common and important causes of disability and death worldwide. However, a sublethal duration of ischemia, ischemic preconditioning, triggers endogenous responses that protect the brain against a subsequent severe ischemic insult, a phenomenon known as ‘tolerance' (Dirnagl et al, 2009). The mechanisms of preconditioning-induced tolerance are not well known, but are characterized by three key features. Ischemic tolerance requires de novo protein synthesis (Barone et al, 1998), is correlated with repressed gene expression (Bowen et al, 2006; Koerner et al, 2007; Stenzel-Poore et al, 2003), and is transient (Chen et al, 1996; Perez-Pinzon et al, 1997).

MicroRNAs (miRNAs) regulate posttranscriptional gene expression in plants, animals, and viruses (Ambros, 2004; Bartel, 2004; Chen and Meister, 2005) and are integral components of RNA-induced silencing complexes, which repress translation by directly interacting with messenger RNAs (mRNAs). In animals, miRNAs regulate mRNA translation through an imperfect pairing with nucleotide sequences within the 3′ untranslated region (3′ UTR) of targets, and repressed translation is enhanced for those mRNAs targeted by multiple miRNAs (Doench and Sharp, 2004). MiRNAs repressing translation are sequestered and localized in processing bodies (Pillai et al, 2005; Sheth and Parker, 2003); release of miRNA-targeted mRNAs sequestered in processing bodies can occur without affecting mRNA stability (Liu et al, 2005; Pillai et al, 2005; Sen and Blau, 2005). Cellular stress can cause the release of mRNA from processing bodies, leading to recruitment of translating ribosomes and protein synthesis (Bhattacharyya et al, 2006). MiRNAs expressed within dendrites regulate translation of proteins mediating dendritic growth (Schratt et al, 2006). MiRNAs that regulate synaptic plasticity can target, and be targeted by, plasticity mediators such as cAMP response element binding protein (CREB), fragile X mental retardation protein, and MeCP2 (methyl-CpG binding protein 2) (Smalheiser and Lugli, 2009).

We propose that ischemic preconditioning could regulate miRNA expression and thus serve as novel effectors of altered protein expression that leads to ischemic tolerance. Accordingly, we show that ischemia does result in significant changes in miRNA expression in preconditioned, ischemic, and tolerant cortices, relative to sham. Target prediction analysis revealed MeCP2 as the most prominent target of miRNAs regulated in preconditioned cortex; thus, we further propose that the preconditioning-regulated miRNAs serve, at least in part, to regulate protein expression of transcriptional regulators required for ischemic tolerance.

MeCP2 has been considered as a global transcriptional repressor because of its methyl-DNA binding and transcription repression domains. For example, in neurons, MeCP2 bound to the brain-derived neurotrophic factor promoter is released on membrane depolarization, resulting in transcription of brain-derived neurotrophic factor mRNA (Chen et al, 2003). However, results of new studies indicate that MeCP2 is a global regulator of transcription; MeCP2 can repress transcription when complexed with histone deacetylase, or activate transcription when complexed with CREB1 (Chahrour et al, 2008). Further studies have established MeCP2 as a multifunctional nuclear protein with roles in chromatin architecture, regulation of RNA splicing, and transcriptional activation (Hite et al, 2009). Together, these studies show a complex role for MeCP2 in brain function and synaptic plasticity; as such, MeCP2 is associated with epigenetic regulation of the nervous system (MacDonald and Roskams, 2009).

The results of our studies show that preconditioning regulates miRNA expression, and target prediction of the preconditioning-regulated miRNAs identified novel proteins that could serve as effectors governing the epigenetic changes that mediate preconditioning-induced tolerance. Among these novel proteins, expression of MeCP2 protein increases in preconditioned cortex, with no corresponding change in MeCP2 mRNA expression, consistent with speculation that MeCP2 is posttranscriptionally regulated (Shahbazian et al, 2002). Accordingly, MeCP2 protein is directly regulated by miR-132 in neurons: decreased miR-132 increases MeCP2 expression (Klein et al, 2007). Although neither miRNAs nor MeCP2 have previously been linked to ischemic preconditioning, these studies support the concept that both could serve as novel effectors of the molecular mechanisms underlying ischemic preconditioning-induced tolerance.

Materials and methods

Transient Focal Ischemia

Adult male C57BL/6J mice (25 to 30 g; Charles River Laboratories, Wilmington, MA, USA) were maintained under diurnal conditions (12 h light/dark cycle) on an ad libitum lab chow and drinking water. Experiments were performed in accordance with the Association for Assessment and Accreditation of Laboratory Animal Care and approved by the Institutional Animal Care and Use Committee of Legacy Research. Transient focal ischemia was induced by suture occlusion of the middle cerebral artery (MCAO) in male mice anesthetized using 1.5% isoflurane, 70% N₂O, and 28.5% O₂ (Longa et al, 1989). Cerebral blood flow in the middle cerebral artery was monitored through a fiber optic probe, mice showing less than 70% cerebral blood flow reduction were excluded from the analysis. Mouse brains were removed 24 h after final MCAO and the ipsilateral and contralateral cortices were dissected and frozen at −80°C. Three brains from each group were stained with vital dye TTC (2,3,5-triphenyltetrazolium hydrochloride), sliced into six sections, and each section was scanned. The stained and unstained areas of each hemisphere were quantified with ImageJ 1.32j (NIH, Bethesda, MD, USA), and values were used to calculate infarct volume expressed as a percentage of the unlesioned hemisphere.

MicroRNA Microarray

RNA was isolated from each cortex using the mirVana miRNA Isolation Kit (Ambion, Austin, TX, USA). The Duke University Microarray Facility labeled RNAs from the ipsilateral (treated) cortex with Cy5 fluorophore, and RNAs from the contralateral (untreated) cortex with Cy3 fluorophore (Amersham Biosciences, Piscataway, NJ, USA). Labeled RNAs were hybridized with mirVana Probe Set V2 (Ambion) microarray slides including probes for human (328), mouse (115), and rat (46) miRNAs. The microarray slides were scanned on a GenePix 4000B (Axon Instruments, Union City, CA, USA). Total Cy3 signal was set to equal total Cy5 signal for each microarray, thus the ratio of total Cy3 to total Cy5 was 1.

MicroRNA Microarray Data Analysis

The miRNA microarray data were analyzed using a Web-based miRNA microarray analysis program created by Rob Lusardi (Slowdog Software, Portland, OR, USA). The ratio of the median intensities for each signal was calculated: 1 indicated equal quantities of target miRNA in the ipsilateral and contralateral cortices, <1 reduced quantities of target miRNA in the ipsilateral cortex, and >1 increased quantities of target miRNA in the ipsilateral cortex, relative to contralateral cortex. Each ratio was log₂-transformed to produce a normally distributed data set amenable to standard statistical analysis: average log ratio (ALR) of miRNA expression=log₂(Cy5/Cy3), where Cy5 and Cy3 were the probe intensities of a single miRNA in the ipsilateral and contralateral cortices, respectively, from the same mouse. A log ratio of 0 indicated no change between ipsilateral and contralateral cortices, positive values indicated increased miRNA expression in ipsilateral cortex, and negative values decreased expression in ipsilateral cortex, relative to contralateral cortex. The ALR of miRNA expression in preconditioned, ischemic, or tolerant cortex was compared with that of miRNAs in sham cortex, and Student's t-test was used to identify those miRNAs that were statistically different from sham.

Target prediction was limited to high-confidence miRNAs, defined as those present in all experimental replicates. Analysis was performed using miRanda (version 2005; www.microrna.org) (John et al, 2004) that allows queries of several miRNAs and reports the total number of gene targets (ENSG), mRNA transcripts (ENST), and the number of ‘hits' that are predicted sites for miRNA binding based on mRNA complementarity (miRNA/ENST pair). The mRNA binding sites are located in the 3′ UTR, numbered 5′ to 3′ from nucleotide 1 beginning just after the stop codon.

Quantitative Real-Time Polymerase Chain Reaction

We analyzed miRNA expression in mouse cortex by qRT-PCR using the miRCURY RNA miRNA PCR System with miR-132 and control U6 primer sets labeled with SYBR Green (Exiqon Inc., Woburn, MA, USA). The mRNA qRT-PCRs were performed using primer sets for MeCP2, SLC2A3, and 18S RNA with TaqMan FAM-labeled probe reagents on a 7500 Fast Real-Time PCR System (Applied Biosystems Inc., Foster City, CA, USA).

MeCP2 Knockout Mice

MeCP2 knockout (KO) mice (strain B6.129P2(C)-Mecp2^tm1.1Bird/J) were obtained from Jackson Laboratories (Bar Harbor, ME, USA). Heterozygous females (Mecp2^tm1.1Bird/J) crossed with wild-type (WT) males (C57BL/6J) yielded WT females (Mecp2⁺/Mecp2⁺), heterozygous females (Mecp2⁻/Mecp2⁺), WT males (Mecp2⁺/y), and KO males (Mecp2⁻/y) (Guy et al, 2001). Genotyping was confirmed by PCR using DNA isolated from tail biopsies, as per the Jackson Laboratories protocol.

Immunoblots

Nuclear fractionation of mouse brain cortices was performed using the CelLytic NuCLEAR Extraction Kit (Sigma, St Louis, MO, USA). Proteins separated by one-dimensional gel electrophoresis were transferred to polyvinylidene difluoride membrane (Millipore, Billerica, MA, USA). Protein blots were incubated in 1:1,000 dilution of mouse antihuman MeCP2 antibody (ab50005; Abcam, Cambridge, MA, USA) then incubated in 1:10,000 dilution of goat antimouse horseradish-peroxidase-conjugated secondary antibody (Bio-Rad Laboratories, Hercules, CA, USA). Protein bands were detected using enhanced chemiluminescence (Amersham Biosciences) and Kodak BioMax film (Eastman Kodak Co., Rochester, NY, USA). Blots were scanned and quantified using the Kodak Image Station 2000rt and Kodak 1D version 3.6 software. Protein bands were background-subtracted, normalized to α-tubulin III (46 kDa; Sigma T8660), and the normalized intensity was expressed as relative to the average control tissue. Statistical analysis was performed by analysis of variance (ANOVA), with a post hoc Dunnett test for significant differences.

Immunohistochemistry

Mouse brains were flash frozen and sectioned on a Cryostat CM3050S (Leica Microsystems Inc., Bannockburn, IL, USA) into 12-μm-thick sections. Whole-brain mounts were fixed, permeabilized, blocked, and incubated with rabbit antimouse MeCP2 antibody (07-013; Millipore/Upstate), then incubated with goat antirabbit Cy3-conjugated antibody (111-165-003; Jackson ImmunoResearch Laboratories, West Grove, PA, USA). Slides were mounted with Vectasheild (Vector Labs, Burlingame, CA, USA) containing DAPI (4′,6-diamidino-2-phenylindole) and images were captured on a Leica TCS SP2 microscope (Leica Microsystems Inc.). For Nissl stain, whole-brain mounts were incubated in cresyl violet, covered in permount, and coverslipped. Each slice was scanned, and stained and unstained areas of the treated hemisphere were quantified using ImageJ 1.32j (NIH) to calculate the infarct volume as a percentage of the unlesioned hemisphere.

Results

Ischemic Preconditioning Alters miRNA Expression in Adult Mouse Cortex

We generated preconditioned, ischemic, and tolerant mice using varying durations of MCAO (Figure 1A). Representative images of TTC-stained sections reveal the extent of injury in mouse brains (Figure 1B). Infarct volume was quantified using TTC-stained sections (n=3 mice per group). Sham (S) mice show no injury, whereas preconditioned (P) mice show injury in striatum, and ischemic (I) mice show significant injury in cortex and striatum. Prior preconditioning provides tolerance (T) against ischemic challenge in cortex evident by 24 h (T-24), more robust at 72 h (T-72), and gone by 10 days after preconditioning (T-240). Representative miRNA microarray heatmaps show consistent patterns of miRNA expression within sham (n=3) and preconditioned (n=3) mice (Figure 1C), and distinct patterns between sham and preconditioned treatments. Thus, we examined sham, preconditioned, ischemic, and tolerant mouse cortices by miRNA microarray analysis.

Figure 1.

Ischemic preconditioning and microRNA microarrays in adult mouse cortex. (A) Schematic of MCAOs used to generate preconditioned (15 mins), ischemic (60 mins), or tolerant (15 mins, reperfusion, 60 mins) mouse brains, all reperfused 24 h after final MCAO. (B) Images of TTC-stained mouse brains revealing injury in each group, and a graph showing percent infarct volume for each condition (n=3 per group). Sham (S) mice are uninjured, preconditioned (P) show little injury in ipsilateral cortex, ischemic (I) show significant injury in ipsilateral cortex and striatum, tolerant (T) show robust protection of ipsilateral cortex against ischemic injury that is apparent by 24 h (T-24), more robust 72 h (T-72) and gone 10 days (T-240) after preconditioning. (C) Schematic of ipsilateral cortex (CTX) and striatum (ST) targeted by MCAO. Representative heatmaps show miRNAs from preconditioned mice both increase (red) and decrease (blue) signal intensities, relative to sham mice.

Statistical Analysis Reveals Distinct Changes in miRNA Expression in Preconditioned, Ischemic, and Tolerant Mouse Cortices

We first evaluated the consistency of miRNA expression within each group to identify animal-to-animal and/or diurnal variations in miRNA expression. For each mouse, the ALR of miRNA expression from one animal (Replicate) was compared with the ALR of miRNA expression for all replicates in the same group (Total); each point in Figure 2A represents one unique miRNA. Representative graphs show a positive correlation between one individual replicate relative to the average of the total replicates sampled within each group for sham (Figure 2Ai), preconditioned (Figure 2Aii), ischemic (Figure 2Aiii), and tolerant (Figure 2Aiv) animals. These studies show that miRNA expression is consistent within a treatment group and that regulation of miRNAs is not random in individual mice.

Figure 2.

Distinct changes in microRNA expression in preconditioned, ischemic, and tolerant mouse cortices. (A) miRNA expression patterns are consistent within each group, and ALR of miRNA expression in each cortex was compared with average ALR of miRNA expression for all replicates in the same group (Total). Representative correlations are shown for (i) sham #3, (ii) preconditioned #6, (iii) ischemic #1, and (iv) tolerant #1. (B) Histograms of miRNA distribution for sham (i) mice show ALR values near zero, preconditioned (ii) and ischemic (iii) mice both show large populations of decreased miRNAs, whereas preconditioned mice also show a large number of increased miRNAs. Tolerant (iv) mice ALRs are similar to sham with a shift toward decreased expression in the ipsilateral cortex. (C) Differences in miRNA expression in each group, relative to sham. Average total response for each group is plotted against the average total response in sham; each point represents the average ALR for a unique miRNA. Sham versus sham (i) illustrates a positive correlation in unchanged miRNA expression. Preconditioned (ii) and ischemic (iii) reveal negative correlations, in contrast to a positive correlation in tolerant (iv), relevant to sham. (D) A Venn diagram illustrating 287 miRNAs significantly altered in preconditioned (P) cortex relative to sham: 157 are uniquely altered, 86 in common with ischemic (I), 13 in common with tolerant (T), and 31 in common in all three groups. (E) Graph of miRNA distribution in each group shows number of miRNAs significantly increased or decreased in preconditioned (P), ischemic (I), or tolerant (T) cortex for All (total miRNAs), Unique, or Common (two or more groups).

We then analyzed miRNA expression in the ipsilateral cortex of mice in each group, relative to contralateral cortex (Figure 2B). A histogram of the ALR for sham miRNAs (n=9) shows a normal distribution near zero, with a mode of −0.20, indicating that miRNA expression between the cortices is similar (Figure 2Bi). However, a histogram of the ALR for miRNAs in preconditioned cortex (n=6) shows a broadening of the distribution with a negative ‘tail' region, with a shift in the mode to +0.28, indicating that miRNA expression is both increased and decreased (P<0.05, t-test; P<0.001, F test; Figure 2Bii). In contrast, a histogram of the ALR for miRNAs in ischemic cortex (n=6) shows a broadening of the distribution with a negative ‘tail' region, but a shift in the mode to −0.14 is not large, indicating a bias toward decreased miRNA expression (P<0.001, t-test; P<0.001, F test; Figure 2Biii). However, a histogram of the ALR for miRNAs in tolerant cortex (n=3) shows a tighter, normal distribution and a negative shift in the mode to −0.38, similar to distribution in the sham group (P<0.001, t-test; P<0.001, F test; Figure 2Biv). These studies show that each treatment induced distinct changes in mouse cortical miRNA expression.

We then compared miRNA expression in the preconditioned, ischemic, and tolerant mice to those of the sham mice (Figure 2C). Graphs show the total ALR for each miRNA in a group plotted against the total ALR of miRNAs in sham. Sham versus sham is provided to illustrate the expected outcome of a positive correlation in unchanged miRNA expression (Figure 2Ci). The graphs show that the ALRs of miRNAs from preconditioned cortices are negatively correlated to sham values (Figure 2Cii), and a similar negative correlation is seen in the ALR of miRNAs from ischemic cortices (Figure 2Ciii). In contrast, there is a positive correlation between the ALR of miRNAs from tolerant cortices relative to sham (Figure 2Civ). These results show distinct changes in miRNAs in the ipsilateral cortex of preconditioned, ischemic, and tolerant cortices, relative to sham.

Individual miRNAs from preconditioned, ischemic, and tolerant cortices were then examined by t-test to identify statistically significant changes in expression, relative to sham. A Venn diagram (Figure 2D) illustrates that of 488 unique miRNAs in the microarray probe set, 273 miRNAs in preconditioned (P), 144 miRNAs in ischemic (I), and 50 miRNAs in tolerant (T) cortex were significantly different (P<0.05) from miRNAs in sham. The 31 miRNAs regulated in all treatment groups likely represent miRNAs altered in response to stress. MiRNAs in preconditioned cortex increased and decreased (Figure 2E, All P), whereas miRNAs in ischemic and tolerant cortices predominantly decreased (Figure 2E, All I, T). Of miRNAs uniquely regulated in preconditioned cortex, 153 increased and 4 decreased (Figure 2E, Unique P), and of miRNAs uniquely regulated in ischemic cortex, 2 increased and 24 decreased (Figure 2E, Unique I). Further, of miRNAs uniquely regulated in tolerant cortex, one increased and two decreased (Figure 2E, Unique T). Remaining data reflect miRNAs regulated among two or more groups (Figure 2E, Common). These results show that miRNA expression is significantly regulated in the ipsilateral cortex in preconditioned, ischemic, and tolerant mice. Supplementary Tables 1–3 list those miRNAs significantly regulated in preconditioned, ischemic, or tolerant mouse cortices.

Target Prediction of the miRNAs Regulated in Preconditioned Mouse Cortex

We used bioinformatic software tools to identify potential mRNA targets of the miRNAs significantly regulated in preconditioned mouse cortex (John et al, 2004), and restricted our analysis to high-confidence miRNAs, defined as those miRNAs detected in all microarray replicates (nine sham, six preconditioned, six ischemic, three tolerant). Given the cooperative nature of miRNAs and that target regulation is more potent when several miRNAs bind to a given mRNA (Doench and Sharp, 2004), the number of ‘hits,' or miRNA binding sites on an mRNA 3′ UTR is an important consideration in target prediction. Thus, we used miRanda (version 2005) for target prediction as it can query several miRNAs at one time to identify predicted mRNA targets of a cohort of miRNAs. Significantly regulated (P<0.05) high-confidence miRNAs included 205 miRNAs; of these, 152 miRNAs were present in miRanda. The queries revealed that the number of predicted targets exceeded 28,000 unique Ensemble gene ID mRNAs, and given that many mRNAs have multiple miRNA binding sites, also revealed over 130,000 potential miRNA to mRNA interactions (Supplementary Table 4). Sorting the Ensemble gene ID targets of miRNAs in preconditioned cortex by the number of potential hits revealed that a striking number of transcriptional regulators are heavily targeted by the miRNAs. Assessment of Gene Ontology of the preconditioning-regulated miRNA targets using Fatigo (Al-Shahrour et al, 2005; www.babelomics.org/, accessed on 7 December 2007) confirmed that the most prominent group of mRNA targets encode proteins that function as regulators of transcription (Table 1).

Table 1. Prominent messenger RNA targets of high-confidence preconditioning-regulated microRNAs (total number of unique miRNAs targeting the 3′ UTR/number of potential miRNA/mRNA binding sites).

Protein	Of all miRNAs represented on the microarrays	Preconditioned miRNAs
		⇑	⇓
MeCP2	61/96	18/35	31/44
HDAC4	30/41	9/11	16/26
CREB1	11/22	4/10	4/8
BDNF	27/136	8/52	16/72
MEF2C	18/57	7/24	7/21
DICER1	17/21	5/8	7/8
EIF2C2	15/15	1/1	11/11
CAMK2A	23/33	8/10	11/18

Abbreviations: MeCP2, methyl-CpG binding protein 2; HDAC4, histone deacetylase 4; CREB1, cAMP response element binding protein; BDNF, brain-derived neurotrophic factor; MEF2C, myocyte-specific enhancer factor 2C; DICER1, endoribonuclease dicer; EIF2C2, eukaryotic translation initiation factor 2C 2; CAMK2A, Ca²⁺/calmodulin-dependent protein kinase II α.

The mRNA predicted to be most heavily targeted by the preconditioning-regulated miRNAs encodes for MeCP2. DNA microarray studies show that MeCP2 mRNA is not significantly regulated by preconditioning (Stenzel-Poore et al, 2003), suggesting that protein expression is regulated at the posttranscriptional level. Given that MeCP2 had not previously been examined in the context of preconditioning, we focused further studies on this protein. We first analyzed the nature and distribution of miRNAs predicted to bind to MeCP2 mRNA. The MeCP2 mRNA transcripts (NM_001081979, NM_010788) are identical throughout their 8,596 nucleotide 3′ UTR, and of 76 miRNAs predicted to target 117 sites in the 3′ UTR (miRanda, version 2005), 58 miRNAs targeting 92 total sites are significantly regulated in preconditioned cortex (Table 2A). In contrast, glucose transporter 3 (SLC2A3) mRNA is significantly increased in preconditioned mouse cortex (Stenzel-Poore et al, 2003). SLC2A3 mRNA transcript (NM_011401) contains a 3,716 nucleotide 3′ UTR, and of 13 miRNAs predicted to target 13 sites in the 3′ UTR, 10 miRNAs targeting 11 total sites are significantly altered in preconditioned cortex (Table 2B). These predictions support that MeCP2 mRNA is a prominent target of the preconditioning-regulated miRNAs.

Table 2. Preconditioning-regulated microRNAs predicted to the target (A) MeCP2 mRNA 3′ UTR and (B) SLC2A3 mRNA 3′ UTR.

Decreased	mRNA target site(s)	Increased	mRNA target site(s)
A
miR-22	29	miR-30e-3p	77, 8373
miR-425	69	miR-302b-AS	193
miR-218	183	miR-378	195, 4850
miR-331	204, 4134, 4487, 6075	miR-19a	224, 961
miR-149	207, 4138, 4264, 7519	miR-301	226, 964
miR-328	209	miR-424	263, 264, 1161, 1162, 2159, 2160
miR-222	213	miR-199a-AS	267
miR-106a	217	miR-17-3p	268, 7989
miR-17-5p	217	miR-183	379, 3665
miR-370	828, 3113, 4113, 5964	miR-346	784, 3126, 5729
miR-339	879, 3195, 3807, 6068	miR-196b	1067
miR-145	1013, 6719	miR-296	1643, 6077, 6098
miR-103	2156	miR-147	1871
miR-15a	2157	miR-340	2101
miR-15b	2159	miR-337	2152
miR-16	2160	miR-373	2264
miR-138	3241	miR-302b	2267
miR-320	3348	miR-302d	2267, 8380
miR-29a	3359	miR-372	2268, 8381
let-7c	3428	miR-96	3666
let-7b	3428	miR-34b	4483
let-7a	3430, 4966	miR-214	4912
let-7d	3431	miR-200c	5271
let-7e	3431, 7959	miR-141	5272
miR-379	3480	miR-330	5351
miR-185	4400
let-7f	4477
miR-34a	4481
miR-24	4936, 6456, 6889
miR-133a	6074
miR-132	6870
miR-212	6875
miR-27b	7752
B
miR-103	67	miR-140	22
miR-107	73	miR-424	80, 81
miR-15b	76	miR-338	671
miR-16	79	miR-148b	1465
miR-15a	80
miR-195	80

Ischemic Preconditioning Decreases miR-132 and Increases MeCP2 Protein, but has no Effect on MeCP2 mRNA Levels

Those miRNAs predicted by miRanda to target the MeCP2 3′ UTR are depicted in Figure 3A; among the decreased miRNAs is miR-132. Klein et al (2007) have shown that MeCP2 expression is controlled by miR-132: decreased miR-132 leads to increased MeCP2 expression in neurons, whereas increased miR-132 leads to decreased MeCP2 expression. Results of our miRNA microarray data show that miR-132 expression is significantly decreased in preconditioned (P) cortex (Figure 3B, −2.26±0.485 ALR, n=6), relative to control (C). Consistent with this, we also show qRT-PCR data validating decreased expression of miR-132 at 24 h after preconditioning (P24) (Figure 3C, −0.75±0.243 ΔΔC_t, n=4), relative to control (C). These studies validate that miR-132, an miRNA known to regulate MeCP2 protein expression, is decreased in preconditioned mouse cortex.

Figure 3.

Decreased miR-132 expression correlates with increased MeCP2 protein, but no change in MeCP2 mRNA, in preconditioned mouse cortex. (A) Predicted MeCP2 3′ UTR target sites for increased (up arrow) and decreased (down arrow) miRNAs significantly regulated in preconditioned cortex. (B) MiR-132 expression by microarray shows decreased ALR of −2.23±0.485 (n=6) in preconditioned (P) cortex, relative to sham (S) (0.026±0.211 ALR, n=9). (C) MiR-132 expression by qRT-PCR validating decreased miR-132 expression (n=4) 24 h after preconditioning (P24, −0.75±0.34 ΔΔC_t), relative to control (C). (D) Immunoblot of mouse cortex showing MeCP2 protein (72 kDa, white diamond) significantly increased in preconditioned (P) cortex, relative to control (C). α-Tubulin III (T) served as a loading control (bottom panel). Quantification of control (C) and preconditioned (P) protein bands (n=3 each) validates a significant increase in MeCP2 protein (P<0.05). (E) MeCP2 mRNA expression at 8, 24, 48, and 72 h after ischemic preconditioning and in sham cortex expressed as ΔC_t (sham−preconditioned) shows no detectable changes in MeCP2 mRNA in preconditioned cortex at the time point (data analyzed by repeated measures ANOVA). (F) Schematic of predicted SLC2A3 target sites for increased (up arrow) and decreased (down arrow) miRNAs significantly regulated in preconditioned cortex. (G) SLC2A3 mRNA expression expressed as ΔC_t (sham-preconditioned) shows a significant increase in preconditioned cortex (P<0.05 by repeated measures ANOVA).

We then used immunoblot analysis to quantify MeCP2 protein expression in preconditioned mouse cortex. A representative immunoblot shows MeCP2 expression in control (C) and preconditioned (P) mouse cortical nuclear lysates (Figure 3C). The MeCP2 antibody (ab50005; Abcam) detects more than one protein band on the immunoblots, thus we also examined nuclear lysates from MeCP2 KO mice to confirm that the 72 kDa protein band (white diamond) is indeed MeCP2. α-Tubulin III (T), which served as a loading control, is shown in the bottom panel. Quantitative analysis of MeCP2 protein expression for control (C, n=3) and preconditioned (P, n=3) mice reveals a significant increase in MeCP2 protein in preconditioned cortex (P<0.05 by t-test) relative to control. These studies reveal the novel finding that ischemic preconditioning increases protein expression of MeCP2, the most prominent predicted target of the preconditioning-regulated miRNAs.

DNA microarray studies do not find MeCP2 mRNA regulated in preconditioned cortex (Stenzel-Poore et al, 2003). Thus, we quantified expression of MeCP2 mRNA by qRT-PCR at 8, 24, 48, and 72 h after preconditioning (n=3 mice per time point, including sham). The results show no significant change in MeCP2 mRNA any time after preconditioning relative to sham (Figure 3E; ΔC_t=sham−experiment), and repeated-measures ANOVA confirmed that preconditioning does not affect MeCP2 mRNA (P<0.68). DNA microarray shows increased SLC2A3 mRNA in preconditioned mouse cortex (Stenzel-Poore et al, 2003) and the SLC2A3 3′ UTR is not a prominent target of preconditioning-regulated miRNAs (Figure 3F). Accordingly, SLC2A3 mRNA levels are significantly increased in preconditioned cortex (Figure 3G, n=3 mice per time point) relative to sham (ΔC_t=sham−experiment) with P<0.05 by repeated-measures ANOVA. Fisher post hoc test revealed a significant difference between sham and 8 h preconditioning (P<0.05). These studies validate that ischemic preconditioning has no effect on MeCP2 mRNA but does induce a significant increase in SLC2A3 mRNA.

MeCP2 Knockout Mice Show Increased Susceptibility to Preconditioning Ischemia

We next examined the cellular distribution and temporal expression of MeCP2 protein in control, sham, and at 8, 24, 48, and 72 h after preconditioning by immunohistochemistry (n=3 mice per time point). Representative images show increased MeCP2 protein in mouse cortical cells by 8 h after preconditioning, persisting at 24 h but not apparent by 48 h after preconditioning (Figure 4A, Cortex). In contrast, MeCP2 protein does not change in the striatum, a region of the brain that is not protected by preconditioning (Figure 4A, Striatum). Given the restricted expression of MeCP2 in the protected cortex, we examined whether depletion of MeCP2 would affect the ability of mice to form ischemic tolerance. Thus, we examined the ability of MeCP2 KO mice to form ischemic preconditioning-induced tolerance. The KO mice had a high mortality rate in response to the ischemic challenge administered 72 h after preconditioning, and only one of three KO mice survived. However, Nissl stain from the surviving KO mouse shows extensive injury in the cortex, relative to the WT mice (Figure 4B, Tolerance). We then examined the effect of preconditioning ischemia alone on KO mouse brains. Representative Nissl-stained brain sections from preconditioned WT mice with a 72 h recovery (Figure 4B) show no injury in the cortex, and neuronal loss in the striatum (n=3). Further, there is no apparent injury in the cortex of the preconditioned KO mice (n=3). However, higher magnification of the brain sections revealed irregular, nonuniformly stained nuclei in the cortex of KO mice, in contrast to cells in the cortex of WT mice (n=3 each). Calculated infarct volumes of WT mice from the Nissl-stained sections are similar to the TTC-stained sections, showing the same result from both techniques (Figure 4C). The Nissl-stained sections show comparable infarct volumes in preconditioned (P) WT and KO mice (n=3 each). However, the tolerant (T) KO mouse that survived the ischemic challenge (one of three) had extensive injury in the cortex, relative to WT mice. These results show that KO mice have an increased susceptibility to ischemia, and that further studies to examine a role for MeCP2 and other transcriptional regulators as effectors of ischemic tolerance are warranted.

Figure 4.

Increased susceptibility to ischemia in MeCP2 KO mouse cortex. (A) MeCP2 protein expression increased in cortex by 8 h (P8) after preconditioning, maintained at 24 h (P24), and returned to control levels by 48 h (P48, P72) after preconditioning. MeCP2 protein is not increased in preconditioned striatum any time after preconditioning. Scale bar: 100 μm. (B) Nissl-stained WT and KO mouse brains subjected to tolerance or preconditioning (72 h recovery) show extensive cell death in the ipsilateral cortex of a tolerant KO mouse (n=1), but not in the cortex of WT mice (n=3). Preconditioned mice show no apparent injury in the cortex of preconditioned WT or KO mice (n=3 each). However, magnified images reveal injured cortical cells in preconditioned KO, but not WT, mice (n=3 each). Cells in the striatum of WT and KO mice are injured by preconditioning. Scale bar: 50 μm. (C) Percent infarct volume for TTC- and Nissl-stained sections from preconditioned (P) and tolerant (T) WT brains (n=3 each), and preconditioned and tolerant KO brains (n=3 and 1, respectively).

Discussion

Cerebral miRNA expression is regulated by transient ischemia (Dharap et al, 2009; Jeyaseelan et al, 2008), traumatic brain injury (Redell et al, 2009), and several other neurologic disorders (Kuss and Chen, 2008). Ischemic preconditioning-induced tolerance requires de novo protein synthesis (Barone et al, 1998), is correlated with genomic reprogramming of the brains response to ischemia (Bowen et al, 2006; Koerner et al, 2007; Stenzel-Poore et al, 2003), and tolerance induced by prior preconditioning is transient (Chen et al, 1996; Perez-Pinzon et al, 1997). Our studies herein focus on miRNAs as effectors of ischemic preconditioning-induced tolerance based on their role as regulators of posttranscriptional gene expression (Bartel, 2004; Chen and Meister, 2005; Smalheiser and Lugli, 2009). We used microarrays to analyze miRNA expression in preconditioned, ischemic, and tolerant mouse cortices. We show that miRNA expression is consistent among mice within a treatment group, miRNA distribution profiles reflect specific responses to each treatment, and miRNA expression profiles in preconditioned, ischemic, and tolerant mouse cortices are significantly different (P<0.05) from sham-operated mice. As preconditioning was the foremost regulator of miRNAs, we focused our bioinformatic studies on identifying potential mRNA targets of the preconditioning-regulated miRNAs. As repressed mRNA translation is enhanced for those mRNAs targeted by multiple miRNAs (Doench and Sharp, 2004), we used miRanda (version 2005) for target prediction as this program allows simultaneous queries of multiple miRNAs. Using t-tests to determine significant changes in treated mice, we identified hundreds of miRNAs as potentially significant, even though their fold changes were frequently small. The power of miRanda prediction software is that we were able to predict the cumulative effects of many small, yet significant, miRNA changes on protein expression. The prediction studies revealed that the most prominent targets of preconditioning-regulated miRNAs were transcriptional and translational regulators, consistent with reports that overrepresented groups of miRNA targets include transcription factors, components of the miRNA machinery, and proteins involved in translational regulation (John et al, 2004). As MeCP2 mRNA was the most prominent target of preconditioning-regulated miRNAs, we selected it for further study and show that ischemic preconditioning rapidly increased MeCP2 protein, but not MeCP2 mRNA, in mouse cortex. These results are consistent with the suggestion that MeCP2 translation might be posttranscriptionally regulated by tissue-specific factors (Shahbazian et al, 2002).

We examined MeCP2 protein because as a transcriptional repressor (Fuks et al, 2003), MeCP2 could be an effector of genomic reprogramming, and had not been examined in the context of preconditioning. Thus our finding that MeCP2 expression is rapidly increased in preconditioned cortex provided a potential link between two key, yet seemingly paradoxical, features of ischemic preconditioning-induced tolerance: the requirement for new protein synthesis (Barone et al, 1998) and genomic reprogramming of the response to ischemia that leads to a transient repression of gene expression (Stenzel-Poore et al, 2003). MeCP2 is a potent transcriptional repressor, yet recent studies show that MeCP2 is a complex regulator of transcription (Chahrour et al, 2008): MeCP2 as a transcriptional activator requires CREB1, and as a transcriptional repressor requires histone deacetylase. Since MeCP2 had not previously been examined in the context of preconditioning or tolerance, it also served as a novel protein to test the power of miRNA target prediction. These studies support the concept that, in addition to preconditioning-induced changes in gene transcription and increased mRNA levels, mechanisms of preconditioning-induced protein expression could include posttranscriptional regulation of mRNAs by miRNAs, consistent with studies showing miRNA regulation of protein expression independent of changes in mRNA levels (Baek et al, 2008; Selbach et al, 2008).

Functionally, miRNAs expressed within dendrites regulate translation of proteins mediating dendritic growth (Schratt et al, 2006). Further, miRNAs are important for regulating synaptic plasticity, and miRNAs target (and are targeted by) plasticity mediators such as CREB, fragile X mental retardation protein, and MeCP2 (Smalheiser and Lugli, 2009). A recent study has shown that miR-132 directly regulates MeCP2 protein expression in rat cortical neurons: increased miR-132 leads to decreased MeCP2 protein, whereas decreased miR-132 leads to increased MeCP2 protein (Klein et al, 2007). This finding is consistent with our data showing decreased miR-132 and increased MeCP2 protein in ischemic preconditioned mouse cortex. Our microarrays and qRT-PCR both showed miR-132 significantly decreased in preconditioned cortex. On the basis of our target analysis showing that many miRNAs, including miR-132, are predicted to bind to the 3′ UTR of MeCP2 mRNA, we used immunoblot and immunohistochemistry studies to evaluate MeCP2 protein expression and show, for the first time, that MeCP2 is rapidly increased in preconditioned cortex. As miR-132 expression is decreased and MeCP2 protein increased in preconditioned cortex, we suggest that MeCP2 mRNA is translationally repressed by miRNAs in control brain and that preconditioning leads to derepression of MeCP2 mRNA by miRNAs with resultant synthesis of MeCP2 protein. When we examined the effect of preconditioning and tolerance in MeCP2 KO mice, we found increased susceptibility to ischemia, consistent with studies showing increased cell death in cerebellar granule neurons in MeCP2 KO mice exposed to excitotoxicity and hypoxic-ischemia (Russell et al, 2007). Further, MeCP2 KO mice show increased susceptibility to hypoxia in telencephalic neuronal networks that involve disturbed potassium channel function, suggesting that hypoxia might contribute to the vulnerability of male Rett patients who are either not viable or severely disabled (Fischer et al, 2009). Although these initial studies in MeCP2 KO mice are not conclusive, they do provide further evidence that MeCP2 is necessary for the induction of ischemic preconditioning.

Given that the MCAO treatments are identical except for duration of occlusion, we expected miRNAs regulated by preconditioning also to be regulated by ischemia. We predict that these common miRNAs mount a response to subsequent injury, but that cell death pathways induced by the longer duration of ischemia overcome this response. Although we focused on miRNAs decreased in preconditioned cortex, many uniquely regulated miRNAs were increased in preconditioned cortex. These miRNAs could serve to repress translation of mRNAs not essential for neuroprotection as a mechanism of energy conservation; studies focused on the role of these miRNAs in tolerance are currently in progress. The studies presented herein set the stage to address additional questions such as which miRNAs specifically target MeCP2 to regulate protein expression. Our use of a low-stringency prediction program (miRanda, version 2005) allowed us to assess the cooperative potential across miRNA changes. Current studies examining specific miRNA/MeCP2 interactions and their therapeutic potential are focused on five preconditioning-decreased miRNAs predicted by three increasingly stringent bioinformatic prediction programs (miRanda, TargetScan, and PicTar) to target MeCP2 mRNA. In addition, recent studies show that MeCP2 is not restricted to neuronal cells as previously thought, but is also expressed in glia (Ballas et al, 2009; Maezawa et al, 2009). Given that miRNAs can activate translation in quiescent cells but repress translation in proliferating cells (Vasudevan et al, 2007), differential regulation of target proteins could occur in neurons and glia. We trust that these studies will contribute to our understanding of the mechanisms underlying ischemic preconditioning-induced tolerance, and have potential to translate into novel strategies for the treatment of ischemic brain injury: the induction of tolerance.