Profiling of MicroRNAs Involved in Retinal Degeneration Caused by Selective Müller Cell Ablation

Dysfunction of Müller cells has been implicated in the pathogenesis of several retinal diseases. In order to understand the potential contribution of Müller cells to retinal disease better, we have developed a transgenic model in which foci of Müller cell ablation can be selectively induced. MicroRNAs (miRNAs), small non-coding RNAs that are involved in post-transcriptional modulation, have critical functions in various biological processes. The aim of this study was to profile differential expression of miRNAs and to examine changes in their target genes 2 weeks after Müller cell ablation. We identified 20 miRNAs using the miScript HC PCR array. Data analysis using two target gene prediction databases (TargetScan and mirTarBase) revealed 78 overlapping target genes. DAVID and KEGG pathway analysis suggested that the target genes were generally involved in cell apoptosis, p53, neurotrophin, calcium, chemokine and Jak-STAT signalling pathways. Changes in seven target genes including Cyclin D2, Caspase 9, insulin-like growth factor 1, IL-1 receptor-associated kinase (IRAK), calmodulin (CALM) and Janus kinase 2 (Jak2), were validated with qRT-PCR and western blots. The cellular localisation of cleaved-caspase 9, Cyclin D2, Jak2 and CALM was examined by immunofluorescence studies. We found that the transcription of some miRNAs was positively, rather than negatively, correlated with their target genes. After confirming that overexpressed miR-133a-3p was localised to the outer nuclear layer in the damaged retina, we validated the correlation between miR-133a-3p and one of its predicted target genes, cyclin D2, with a luciferase assay in 661 photoreceptor cells. Results revealed by miRNA profiling, target gene analysis and validation were generally consistent with our previous findings that selective Müller cell ablation causes photoreceptor degeneration and neuroinflammation. Our data on alterations of miRNAs and their target gene expression after Müller cell ablation provide further insights into the potential role of Müller cell dysfunction in retinal disease.


Introduction miRNA Polymerase Chain Reaction (PCR) Array
Retinae were thawed then total RNA, including miRNAs, was extracted with a miRNeasy Mini Kit (Qiagen, 217004) according to manufacturer's instructions. The quantity and quality were assessed by an Experion Automated Electrophoresis System (Biorad, 701-7001). Six retinae from each group were reverse-transcribed with miScript II RT kit (Qiagen, 218160) and diluted with RNase free water. miRNA PCR array was performed with 384 well-plate miScript miRNA High Content PCR array (Qiagen, MIMM-3001Z) which contains miScript primers for 372 of the most well characterised miRNAs and duplicates of 6 internal reference miRNAs. The identities of 372 miRNAs are listed in Supporting Information (Table A in S1 File). The thermocycle of the PCR array consisted of denaturing at 94°for 15 seconds, annealing at 55°for 20 seconds and extension at 70°for 30 seconds. Relative quantification was performed by the ΔΔC T method recommended by the manufacturer (http://pcrdataanalysis.sabiosciences.com/ mirna/arrayanalysis.php?target = analysis). A p-value <0.05 was considered statistically significant and the fold change cut-off criteria was ± 2.

Functional Analysis of miRNAs
We started with functional analysis of each differentially expressed miRNAs by performing a PubMed literature search to identify recent publications, which would shed light on their related functions (http://www.ncbi.nlm.nih.gov/pubmed). In addition, we used 2 target gene prediction databases: TargetScan (http://www.targetscan.org) and miRTarbase (http://mirtarbase. mbc.nctu.edu.tw). Overlapping target genes from these 2 databases were identified for further analysis. Functional analysis of the target genes was performed by Database for Annotation, Visualisation Integrated Discovery (DAVID, http://david.abcc.ncifcrf.gov) and associated pathways were identified by KEGG pathway database (http://www.genome.jp/kegg/pathway.html). We also performed in silico study with mirTarBase to examine direct seed matching sequence alignments between the identified miRNAs and their target genes prior to further investigations (Table 1).

Quantitative RT-PCR (qRT-PCR)
qRT-PCR was performed as previously described [2]. In brief, an equal amount of total RNA was reverse-transcribed with SuperScript VILO cDNA Synthesis kit (Life Technologies, 11754050) according to manufacturer's instruction, and diluted with RNase free water and Express SYBR GreenER qPCR Supermix (Life Technologies, 11784-200). Primers we used in this study are listed in Table 2. The relative quantitative analysis was performed by Relative Expression Software Tool (REST) 2009 [24]. Expression values were normalised to the geometric mean of 3 reference genes including 18SrRNA, GAPDH, β-Tubulin.

Immunofluorescence
Immunofluorescence labelling was performed as previously described [2]. Briefly, eyes were enucleated and anterior segments were dissected immediately. The remaining eye cups were fixed with 4% paraformaldehyde in PBS for an hour at room temperature and cryo-protected with 20% sucrose in PBS for overnight at 4°C. The eyecups were embedded with Optimum Cutting Medium and sectioned at 12μm on Superfrost glass slides. Sections were washed with PBS 3 times for 5 minutes and blocked with normal goat serum for 1 hour at room temperature. After washing with PBS, sections were incubated with primary antibody for overnight at 4°C. The primary antibodies used in this study were cleaved caspase 9 (cell signalling, #9509S), cyclin D2 (cell signalling, #2924), JAK2 (abcam, ab39636) and calmodulin (abcam, ab45689). Secondary antibody incubation was followed on the next day for 2 hours at room temperature and nucleus staining was performed with Hoechst. Stained sections were examined by confocal microscopy as described previously [2,6].

In Situ Hybridisation
In situ hybridisation was performed to localise miRNA expression in the retina using Dig-labelled miR-133a-3p miRCURY LNA probe (Exiqon, Cat#39270-01) according to the manufacturer's instruction with minor modifications. Briefly, cryosections were thawed, air-dried and washed with PBS for 3 times. Sections were treated with proteinase K (100ng in 200ml water) in a slide wash container for 5mins followed by acetylation treatment (2.3ml of triethanolamine and 500μl acetic anhydride in 200ml water) for 10mins. A hybridisation buffer containing 25nM of probe was pre-warmed for 5min at 65°C and then applied onto sections. Sections were cover-slipped and incubated in a chamber humidified with 50% formamide in saline sodium citrate (SSC) solution overnight at 53°C. On the next day, sections were washed, blocked with 10% normal goat serum and incubated with an anti-Dig antibody (Roche, 11093274910) overnight at 4°C. NBT/BCIP alkaline phosphatase colour reaction (Vector laboratories, SK-5400) was performed to visualise the cellular localisation of miR-133a expression according to the manufacturer's instruction.

Luciferase Assay
Luciferase assay was performed to validate the direct correlation between miRNA and its predicted target gene expression. We chose to validate miR-133a-3p and one of its target genes, cyclin D2, since cyclin D2 was the most differentially expressed gene after Muller cell ablation. Briefly, 3'UTR of cyclin D2 was amplified by PCR, digested with XhoI and NotI restriction enzymes (New England Biolabs, R0146S and R3189S) and cloned into a multiple cloning site located downstream of the renilla translational stop codon within the psiCHECK TM -2 vector (Promega, C8021). 661w photoreceptor cells were seeded in 12-well plates (50,000 cells per well) and cultured for 2 days. Once they reached 70% confluence, the cells were transfected with the vector alone, or co-transfected with a miRNA-133a-3p mimic (Ambion, 4464067) or a control miRNA (Ambion; 4464058) using Lipofectamin 3000 (Life Technologies, L3000001) for 48 hrs. Luciferase assay was performed using the Dual-Glo Luciferase Assay System (Promega, E2920) and the luciferase activity was measured by Safire2TM Tecan plate reader (Tecan Austria, Safire-BASIC). Firefly luciferase activity was normalised to the renilla luciferase activity.

Statistic Analysis
Results are expressed as mean ± SEM. Data were analysed using unpaired Student t-test. A p value <0.05 was regarded as statistically significant.

Target Gene Analysis and qRT-PCR Validation
We used 2 different databases, TargetScan and mirTarBase, to analyse genes which could be potentially targeted by the differentially expressed miRNAs. We identified 78 overlapping genes that could be potentially targeted by the 20 differentially expressed miRNAs. Of these, 72 corresponded to upregulated miRNAs and 6 to downregulated miRNAs (Table B and Table C  in S1 File). DAVID and KEGG pathway analysis revealed that 41 target genes had significant biological functions (Table D in S1 File). Based on their potential roles in retinal pathology as previously reported [2,6], we chose 7 genes for qRT-PCR validation, including Cyclin D2, Caspase 9, IGF1, IRAK, CALM1, CALM2, and Jak2 (Table 3). Prior to further validation with qRT-PCR, we performed in silico study with mirTarBAse which revealed seed sequence alignments between miRNAs and their target genes (Table 1). Glutamine synthetase (GS) and GFAP were used as Müller cell markers to correlate changes in these 7 genes with Müller cell disruption. Consistent with our previous studies [6,7], Müller cell ablation resulted in downregulation of GS and upregulation of GFAP (Table 3). qRT-PCR analysis revealed that Cyclin D2, IGF1, CALM2 and Jak2 were significantly upregulated 2 weeks after Müller cell ablation (p<0.05) ( Table 4 and Fig. 2). The KEGG pathway analysis suggested that these genes were involved in   cell apoptosis, p53, neurotrophin, calcium, chemokine and Jak-STAT signalling pathways (Table 4).

Western Blot Analysis
Western blot analysis after Müller cell ablation found changes that were consistent with target gene analysis revealed by qRT-PCR. Genes involved in the p53 signalling pathway, cell apoptosis including cleaved-caspase 9 and cyclin D2 and calmodulin which plays a role in calcium signalling and the neurotrophin pathway, were all significantly upregulated after Müller cell ablation (Fig. 3). Jak2 tended to be upregulated but the change was not statistical significant (P = 0.2).

Immunofluorescence
We performed immunofluorescence with antibodies against four proteins encoded by target genes including cleaved-caspase 9, cyclin D2, Jak2 and calmodulin to examine their cellular specific expression (Fig. 4). We found that cleaved-caspase9 (Fig. 4A) and Jak2 (Fig. 4C) were mainly expressed in the outer nuclear layer (ONL) in areas corresponding with patches of Müller cell ablation. Upregulation of cyclin D2 was observed in not only the inner nuclear layer (INL) and the ONL but also in some cell bodies of photoreceptors near the outer limiting membrane after Müller cell ablation compared with controls (Fig. 4B). Calmodulin was mainly expressed in the INL and weakly expressed in photoreceptor inner segments in the normal retinae. It was markedly upregulated in the outer plexiform layer, ONL and photoreceptor inner segments in areas of Müller cell ablation (Fig. 4D).
Validation of the Correlation of miR-133a-3p with Its Target Gene Cyclin D2 Expression Using Luciferase Assay miRNA PCR array showed 2.15-fold increase in miR-133a-3p expression (Table 3) while qRT-PCR analysis showed 5-fold increase in cyclin D2 expression after Muller cell ablation ( Table 4). As cyclin D2 is one of the genes targeted by miR-133a-3p, our data indicate a positive correlation between miR-133a-3p and cyclin D2 expression. We performed in situ hybridisation and found that the increased miR-133a-3p expression was mainly localised to the outer nuclear layer in the retina after Muller cell ablation ( Fig. 5A-C). In order to validate the functional correlation between miR-133a-3p and cyclin D2 expression, we cloned the 3'UTR region of cyclin D2 gene into a vector and conducted luciferase assay in 661w photoreceptor cells. We found that co-transfection of 661w cells with a miR-133a-3p mimic along with the vector resulted in significant increase in the normalised luciferase activity compared with cells cotransfected with a control miRNA and the vector (P<0.05, Fig. 5D), thus confirming the positive correlation between miR-133a-3p and cyclin D2 expression.

Discussion
We have identified 20 miRNAs that were differentially expressed during the photoreceptor degeneration that occurred after selective Müller cell ablation. We then used 2 different databases, TargetScan and mirTarBase, to identify a number of genes that could be potentially targeted by these miRNAs. Of 72 target genes that we identified, 41 were found, by DAVID and KEGG pathway analysis, to be involved in several biological functions. Based on our previous studies, we chose 7 target genes which were involved in apoptosis, p53, chemokine, Jak-STAT, calcium and neurotrophin signalling pathways for further validation. Upregulation of the chosen target genes was confirmed with qRT-PCR and validated by western blot analysis.
Immunofluorescence studies revealed that the upregulation of these miRNA-target genes were mainly observed in the ONL where photoreceptor degeneration occurred after selective Müller cell ablation. We conducted in situ hybridisation for miR-133a-3p and found that the increased expression of miR-133a-3p was mainly localised in the ONL. Furthermore, we performed luciferase assay to validate the positive correlation between miR133a-3p and one of its predicted targets genes, cyclin D2 in 661 photoreceptor cells. The results were generally consistent with our previous findings that selective Müller cell ablation causes photoreceptor degeneration [2,6]. Neuronal damage in this transgenic model is characterized by photoreceptor degeneration caused by selective Müller cell ablation [6]. MiRNA profiling conducted 2 weeks after Müller cell ablation identified a number of differentially expressed miRNAs whose target genes were associated with cell apoptosis and neurodegeneration in the central nervous system including retinal degeneration [21,23,[25][26][27][28][29]. For instance, one of the genes targeted by miR-133a-3p is caspase-9, which can form an apoptosome with cytochrome-c released from mitochondria and lead to activation of other caspases and cell death [30]. Calmodulin, which is targeted by mir-1a-3, has been reported that calmodulin plays a role in neurotransmitter release and trafficking in synapsis [31,32]. However, overexpression of calmodulin has been reported to be involved in neuronal apoptosis by activating calcium/calmodulin-dependent protein kinase II [33,34]. Inhibition of calmodulin activity has been reported to prevent retinal neuronal apoptosis [35][36][37][38].
Our previous studies found that the microarray profiling in which we performed 1 week after Müller cell ablation, also revealed upregulation of genes involved in apoptosis [2], and the characterization study showed photoreceptor apoptosis was peaked at 2 weeks after Müller cell ablation [6]. We did not observe any significant differences in the number of TUNNEL positive cells the retina between 1 week and 2 weeks after Müller cell ablation, thus we believe the current study is generally supported by our previously observations. Interestingly, miR-133a-3p also targets an anti-apoptotic gene. Cyclin D2, which has been reported to play a role in anti-apoptosis by phosphorylating retinoblastoma protein, cell cycle arrest and neurogenesis [39][40][41][42], was significantly upregulated after Müller cell ablation. After confirmation that the increased miR-133a-3p expression was mainly localised to the outer nuclear layer after Muller cell ablation, we conducted luciferase assay in 661w photoreceptors to validate the positive correlation between miR-133a-3p and cyclin D2 expression. We found that co-transfection of 661w cells using the vector with a miR-133a mimic induced significant increase in luciferase activity compared with cells co-transfected using the vector with a control miRNA (Fig. 5D). A previous study reported a negative correlation between miR-133a and cyclin D2 expression in cardiac muscle cells and cos-1 cells [41]. The discrepancy between that study and the results we present here could be due to the diverse roles of miR-133a and cyclin D2 in different tissues and biological models. Cyclin D family has been reported to play a neuroprotective role in brain injury and retinal degeneration [41,43]. Moreover, cyclin D2 is involved in neurogenesis [40]. It is unclear whether the overexpressed miR-133a and cyclin D2 are due to a defensive response to the retinal stress caused by Muller cell ablation. Future studies are warranted to investigate the effects of intervention of miR-133a and cyclin D2 overexpression on photoreceptor degeneration in our transgenic mice. Neuroinflammation was also shown in the transgenic model of selective Müller cell ablation [2,6,7]. Mir-146a-5p, which has been reported to play a role in neuroinflammation and neurodegeneration [22,27], was upregulated after induced Müller cell ablation. It has been reported that its target genes, such as IRAK1, IRAK2, and TRAF6, are involved in cell survival and neuroprotection. IRAK recruits the p75 neurotrophin receptor and activates nuclear factor keppa B (NF-kB) [44,45]. We recently showed that the p75 neurotrophin receptor was profoundly upregulated in activated surviving Müller cells and that photoreceptor degeneration occurred concurrently with activation of microglial cells in this model. Linking the result of upregulation of mir-146a-5p with features of photoreceptor degeneration and neuroinflammation caused by Müller cell ablation indicates a potential positive correlation between mir-146a-5p expression and IRAK expression, although our qRT-PCR analysis of the change in IRAK transcription was not statistically significant between transgenic and control mice. Further analysis on the relationship between miR-146a-5p upregulation of changes in different isoforms of IRAKs and their downstream effectors is warranted to provide a better understanding of the role of miRNA-146 in photoreceptor degeneration and neuroinflammation caused by Müller cell dysfunction.
Interestingly, miR-375-3p was downregulated after Müller cell ablation, and Jak2, one of its target genes, was upregulated. The negative correlation between miR-375 and Jak2 is consistent with a recent report in which target mutation of a predicted miR-375-binding site abolished the activity of a luciferase reporter carrying the 3' untranslated region of Jak2 [46]. Jak2 has been reported to be involved in activation of immune response/chemokine signalling and the Jak/STAT pathways. Inflammation induced by lipopolysachrides in the rat retina caused activation of the Jak2/STAT3 pathway and neuronal gliosis [47]. Our recent studies found that selective Müller cell ablation resulted in profound activation of surviving Müller cells and microglia and overexpression of TNFα. It is plausible that the downregulation of miR-375-3p may trigger activation of the Jak2/STAT3 pathway and chemokine signalling, both of which can damage photoreceptors.
We conducted immunostaining to study the cellular localization of a number of target genes in normal and degenerating retinae. Cleaved-caspase 9, cyclin D2 and JAK2, were mainly expressed in the ONL in areas corresponding to Müller cell ablation, whereas upregulation of calmodulin was observed in the outer plexiform layer, ONL and photoreceptor segments after Müller cell ablation. Calmodulin is a highly conserved regulatory protein found in all eukaryotic organisms which mediates a variety of calcium ion-dependent signaling pathways [31]. A recent study showed that over-activation of calcium/calmodulin-activated protein kinase II is involved in photoreceptor degeneration in the rd1 mouse [48]. Our observation that calmodulin was upregulated after Müller cell ablation suggests that activation of the calcium/calmodulin signalling may contribute to photoreceptor degeneration. Interestingly, cyclin D2 was highly upregulated in the INL of retinae with Müller cell ablation compared with the normal retina in controls. As cyclin D2 is anti-apoptosis [39], this observation may partially explain the phenomenon that neuronal damage is not obvious in the INL following Müller cell ablation in this transgenic model [6].
One of the interesting findings we found in this study is that a number of differentially expressed miRNAs appear to be positively correlated with their target gene expression after selective Müller cell ablation. While the precise molecular mechanisms of which miRNA control gene expression are still largely obscure, there has been a growing number of reports of a positive correlation between miRNA and mRNA expression [17,[49][50][51]. Some hypothetical and in silico studies of the mechanism of which miRNA promotes gene expression have recently been reported [14][15][16][17]50].
We acknowledge that this profiling study only covered 372 miRNAs rather than an unbiased genome-wide screen. We cannot rule out the possibility that the screening of a limited number of miRNAs might have missed some miRNAs which make important contributions towards photoreceptor degeneration and neuroinflammation in retinal disease.
We report here that miRNAs and their target genes that are differentially expressed soon after selective Müller ablation are mainly involved in neuronal apoptosis, inflammation and neurodegenration. Müller cell dysfunction has been implicated in the pathogenesis of many retinal diseases but its precise contribution to the photoreceptor degeneration, which ultimately results in loss of vision, remains poorly understood. The results of miRNA profiling after Müller cell ablation that we present here provide further insights into the pathogenesis of photoreceptor degeneration associated with Müller cell dysfunction.
Supporting Information S1 File. Supporting tables. Table A. The list of 372 miRNA IDs used in Qiagen HC PCR array. Table B. Target gene analysis from the upregulated miRNAs. The list shows the overlapping target genes from TargetScan and miTarBase. Table C. Target gene analysis from the dwonregulated miRNAs. The list is the overlapping target genes from TargetScan and miTarBase.