AMPK‐regulated miRNA‐210‐3p is activated during ischaemic neuronal injury and modulates PI3K‐p70S6K signalling

Progressive neuronal injury following ischaemic stroke is associated with glutamate‐induced depolarization, energetic stress and activation of AMP‐activated protein kinase (AMPK). We here identify a molecular signature associated with neuronal AMPK activation, as a critical regulator of cellular response to energetic stress following ischaemia. We report a robust induction of microRNA miR‐210‐3p both in vitro in primary cortical neurons in response to acute AMPK activation and following ischaemic stroke in vivo. Bioinformatics and reverse phase protein array analysis of neuronal protein expression changes in vivo following administration of a miR‐210‐3p mimic revealed altered expression of phosphatase and tensin homolog (PTEN), 3‐phosphoinositide‐dependent protein kinase 1 (PDK1), ribosomal protein S6 kinase (p70S6K) and ribosomal protein S6 (RPS6) signalling in response to increasing miR‐210‐3p. In vivo, we observed a corresponding reduction in p70S6K activity following ischaemic stroke. Utilizing models of glutamate receptor over‐activation in primary neurons, we demonstrated that induction of miR‐210‐3p was accompanied by sustained suppression of p70S6K activity and that this effect was reversed by miR‐210‐3p inhibition. Collectively, these results provide new molecular insight into the regulation of cell signalling during ischaemic injury, and suggest a novel mechanism whereby AMPK regulates miR‐210‐3p to control p70S6K activity in ischaemic stroke and excitotoxic injury.


| INTRODUC TI ON
The ischaemic penumbra represents a window for therapeutic intervention to limit the progressive damage and death in the compromised tissue and any potentially detrimental effects of reperfusion contributing to ischaemic injury (Green & Shuaib, 2006;Kalogeris et al., 2012). To effectively utilize this opportunity in the treatment of acute brain injury, the development of neuroprotective agents targeting multiple pathobiochemical pathways may be required, thereby widening the therapeutic window, improving outcome and facilitating brain recovery (Green, 2004). Cerebral ischaemia triggers a complex series of physiological, biochemical and gene expression changes precipitating the onset of neuronal injury and cell death (Dirnagl et al., 1999). Despite an increasing wealth of knowledge, the heterogeneous nature of the biochemical consequences of ischaemic stroke has resulted in halting progress in the development and translation for neuroprotective agents into the clinical setting (Majid, 2014). To reduce the devastating impact of stroke, there is an urgent need for a greater understanding of the molecular mechanisms mediating neuroprotective and neurotoxic events during acute stroke for novel approaches in the field of neuroprotection and better success in translational efficacy (Dirnagl et al., 1999;Liou et al., 2003).
Progressive neuronal injury in response to ischaemia is associated with a series of biochemical cascades, many of which result from impaired energetics and the collapse of ion gradients, precipitating neuronal cell death (Dirnagl et al., 1999). Glutamate receptor over-activation, in particular N-methyl-D-aspartate (NMDA) receptors, is believed to be a central process in the spreading of neuronal injury during ischaemic stroke. Ischaemia-induced excitotoxic neuronal injury caused by excessive glutamate release and aberrant Ca 2+ levels have been suggested to play a key role in peri-infarct depolarization, cell death activation, inflammation and oedema formation, all of which may contribute to secondary decompensation within the penumbra (Belov Kirdajova et al., 2020;Dirnagl et al., 1999).
AMP-activated protein kinase (AMPK) is a critical energy sensor and central regulator of energy homeostasis, with the ability to sense compromised availability of cellular energy (Carling, 2004;Hardie et al., 1999;Kahn et al., 2005). Rapid energy depletion in acute stroke results in the activation of AMPK in response to impaired cellular bioenergetics, functioning to maintain energy balance within the cells by promoting processes leading to ATP production and restoration (Marsin et al., 2002;Wu & Wei, 2012), while simultaneously inhibiting anabolic growth-promoting processes, such as protein synthesis, proliferation and cell cycle progression (Carling, 2004;Hardie, 2011;Hardie et al., 2012;Jones et al., 2005;Li & McCullough, 2010). The direct effects of AMPK activation have conflicting implications in neuronal outcome however, paradoxically promoting decisions in cell survival and death signalling (Culmsee et al., 2001;Li & McCullough, 2010). The extent and duration of AMPK activation, in combination with the duration and nature of the metabolic stressor and extent of ATP and bioenergetic recovery, are pivotal in determining the downstream effects and cellular response to insult (Davila et al., 2012). Transient increases in AMPK activity in neurons prior to ischaemic-like challenge have been shown to enhance cell survival, promoting an adaptive, neuroprotective response that reduces the impact of subsequent more severe stimuli, whereas chronic sustained AMPK activation may adversely exacerbate injury (Anilkumar et al., 2013;Culmsee et al., 2001). To this end, elucidation of the molecular signatures associated with AMPK activation, determining cellular fate in the balance of survival signalling and susceptibility towards cell death stimuli, warrants consideration in the setting of ischaemia.
Given the complex nature of the ischaemic cascade, identification of clinically useful biochemical targets for intervention has been challenging and as such, a single target of neuroprotection may be ineffective (Maas & Furie, 2009;Sharp et al., 2011). Endogenous mi-croRNAs (miRNAs) are potent modulators of gene function, regulating the expression of multiple target genes at a post-transcriptional level, with crucial roles as regulators of signalling pathways involved in pathophysiology and progression of ischaemia-reperfusion injury (Dharap et al., 2009;Khoshnam et al., 2017). Furthermore, the rapid induction of an ischaemic miRNA profile can be detected before the induction of protein markers (Lee et al., 2010;Sharp et al., 2011). The dysregulation of miRNA profiles in disease, impacting on and contributing to pathology development and outcome, combined with their ability to regulate multiple genes in similar pathways, leave them uniquely poised as ideal biomarkers and therapeutic targets. Therefore, we aimed to explore AMPK-regulated miRNA as modulators of key downstream pathways in the response to progressive neuronal injury and provide mechanistic insights into the regulation of neuronal fate in ischaemic stroke.

| Animals
All animal experiments were carried out under license from the Department of Health and Children (Ireland) and in accordance with the European Communities Council Directive (86/609/EEC).
All procedures were reviewed and approved by the RCSI Research Ethics Committee and reported following the ARRIVE guidelines.
The study was not pre-registered. Wild type (WT) C57BL/6 mice (Charles River Laboratories; RRID:IMSR_CRL:27) were housed (minimum two, maximum five animals / cage) in a controlled environment with 12 hr light-dark cycles (7a.m.-7p.m.), with standard laboratory animal chow and water ad libitum. Animals were monitored daily as part of standard husbandry and each individual cage checked daily.
Mice were used for breeding from 6 to 8 weeks of age and were closely matched for age when paired for breeding. Additional bedding is made available to pregnant females 3-4 days before delivery. A total of 109 adult (7-8 weeks) WT C57BL/6 mice were used (14 for in vitro experiments, 95 for in vivo experiments). Male mice 7-to 8-weeks old (20-24 g) were arbitrarily assigned to sham/experimental groups; no exclusion criteria were pre-determined and no randomization was performed. All surgical procedures were carried out in the morning (9:00 ± 1 hr) and post-procedural animals were monitored carefully after surgery for signs of discomfort or adverse effects. All researchers were blinded to experimental groups and analysis was carried out by researchers independent of the experimenter, also blinded to the experimental groups. Timeline of experimental design and analysis points is illustrated in Figure 1.

| Reverse transcription and real-time quantitative PCR
Total RNA enriched with miRNAs was isolated using miRNeasy kits (Qiagen, 217004) according to manufacturer's instructions. was performed using TaqMan Universal PCR Master Mix according to manufacturer's instructions (Applied Biosystems, 4324018) and cycled using the StepOnePlus Real-Time PCR System (Applied Biosystems; RRID:SCR_015805). Cycling parameters were as follows: 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. All PCR amplification reactions were carried out in triplicate using specific TaqMan assays. A negative control was included for each assay.

| Digital PCR
Digital PCR (dPCR) amplifications were carried out to quantify miRNA copy number using the QuantStudio™ 3D Digital System (Applied Biosystems, A29154). Briefly, total RNA was reverse transcribed using TaqMan miRNA-specific RT primers as described for miRNA RT-qPCR. An input volume of 2.3 μl RT product diluted 1:10 was amplified in a total reaction volume of 14.5 μl containing QuantStudio™ 3D Digital PCR Master Mix (Applied Biosystems, A26358) and miRNA-specific TaqMan assay and loaded onto a 20,000 nanoscale reaction well QuantStudio™ 3D Digital PCR 20K Chip (Applied Biosystems, A26316). Chips were cycled on the ProFlex 2X Flat PCR System (Applied Biosystems, 4484078) as follows: 96°C for 10 min followed by 39 cycles of 60°C for 2 min and 98°C for 30 s, and 60°C for 2 min. Chips were read and analysed on the QuantStudio™ 3D Digital PCR instrument to obtain the number of wells with an amplified target (FAM positive) and number of empty wells (FAM negative) and raw data were analysed using QuantStudio™ 3D AnalysisSuite to determine miRNA copies/µl.

| Protein extraction and western blotting
Tissue and cell pellets were lysed in ice-cold lysis buffer (50 mM Tris HCl, 150 mM NaCl, 1% NP40, 0.5% sodium deoxycholate and 0.1% sodium dodecyl sulphate) to obtain whole lysates. Protein concentration was determined with micro BCA (bicinchoninic acid) assay (Thermo Scientific, 23225). Equal amounts of protein were supplemented with Laemmli buffer, denatured at 95°C for 5 min and separated on 10% SDS-polyacrylamide gels. Proteins were transferred to nitrocellulose membranes and blocked in 5% non-fat dry milk in
Isoflurane inhalation anaesthesia was employed because of its rapid induction and recovery time, greater safety, lesser cardiovascular depression and steady maintenance of anaesthetic depth.
Body temperature was maintained normothermic at 36.8-37.4°C via a feedback-controlled heat blanket. A silicone-coated monofilament (701956PK5Re; Doccol Corporation, USA) was introduced into the left internal carotid artery and advanced past the carotid artery bifurcation to occlude the middle cerebral artery. After 60 min the suture was removed to allow reperfusion. To control for effects of occlusion, sham-treated mice underwent the same surgical procedure, but the filament was not advanced to occlude the vessel. Effective occlusion and assessment of microcirculatory function was monitored by LDF (Laser Doppler Flowmetry) with a probe fixed to the exposed left parietal skull for continuous monitoring of regional cerebral blood flow (Perimed 5001 Master, Perimed) (Table S1). Mice were killed via cervical dislocation at various time points after reperfusion and brains processed for analysis.
Sections (750 μm intervals) were stained with Cresyl violet/Nissl, and infarct area and hemispherical swelling were evaluated quantitatively using an image analysis system (Axiovision 4.8, Zeiss).
Hemisphere and ischaemic infarct volumes (mm 3 ) were calculated by the total area multiplied by the distance between sections (mm) (Plesnila et al., 2004). The extent of hemisphere oedema was calculated by dividing the total volume of ipsilateral hemisphere by the total volume of contralateral hemisphere and expressed as the percentage of contralateral hemisphere to correct for differences in the brain size and oedema.
Tyramide Signal Amplification (TSA) Plus Fluorescence kit (Perkin Elmer, NEL741001KT) was used to amplify probe-target miRNA labelling and detect fluorescent signals.

| Imaging of whole-brain sections
Imaging was performed in a tile imaging mode using an inverted microscope equipped with a 20X 0.5NA objective and hardware autofocus (AxioObserver) with motorized stage (Prior, UK), a scientific CMOS camera (Orca Flash 4, Hamamatsu, UK) and controlled by MetaMorph software (Molecular Devices, UK). Images were taken with 10% overlap, background corrected and stitched using the Fiji/ ImageJ software (Wayne Rasband, NIH) with the stitching plugin written by Stephan Preibisch (Preibisch et al., 2009).
Protein was extracted and RPPA performed as previously described (Hennessy et al., 2010). Briefly, lysates were two-fold-serial diluted and arrayed on nitrocellulose-coated slides and each slide was probed with a single primary antibody using the Catalysed Signal Amplification (CSA) System (DAKO, USA, K1500) and visualized by DAB colorimetric reaction using biotin-conjugated secondary antibodies. Slides were scanned on a flatbed scanner to produce 16bit TIFF images. Spots from TIFF images were identified and the density was quantified using MicroVigene software (VigeneTech) to obtain spot signal intensities. All data points were normalized for protein loading and transformed to linear values. Repeated measures ANOVA was used to test whether mean protein levels independently differed by treatment. Protein levels were treated as continuous linear variables and treatment was treated as a categorical variable, with the following categories: Invivofectamine control; non-targeting negative control (50 pmol); miR-210-3p mimic (0.5, 5, 50 pmol). Statistical significance was determined using false discovery rate (FDR) p-values to correct for multiple comparisons based on p-values from repeated measures ANOVA F tests. Tukey's honest significant difference test was used as a post hoc test to examine whether there were differences in mean protein levels between treatment categories. Mean protein levels within each treatment category and 95% confidence intervals were computed using the Loftus and Masson (1994) method.

| Statistical analysis
Statistics were carried out on IBM SPSS 25.0 software (IBM). No statistical method was employed for exploratory in vivo dosing studies; sample size was determined based on previous experience of the group to minimize the number of animals required (Pfeiffer et al., 2015;Plesnila et al., 2001). No animals were excluded from analysis. No test for outliers was conducted. Normality was confirmed by the Shapiro-Wilk test for each individual dataset and normally distributed data were analysed by analysis of variance (ANOVA) and post hoc test or Student's t-test; please refer to figure legends for statistical tests used. p-values <0.05 were considered to be statistically significant; p-values were adjusted for multiple comparisons with Benjamini-Hochberg correction. Data are presented as mean ± SEM unless otherwise indicated.

| miR-210-3p is up-regulated in the cortex following transient focal ischaemia with reperfusion
Next, we examined the expression of differentially expressed miR-NAs in vivo in the cortex of WT mice following 60 min tMCAO with reperfusion. Levels of miR-124 were detectable by PCR following 60 min tMCAO with reperfusion but no significant difference was observed in expression levels at 3 hr (log2FC −0.19, p =.53) and 24 hr (log2(FC) 0.1, p =.98) compared with sham-treated controls ( Figure S1b). Quantitative real-time and digital PCR confirmed a robust induction of miR-210-3p following ischaemic stroke in vivo at 24 hr (2,597.4 ± 92.9 copies /µl) compared to 3 hr (1951.8 ± 106.9 copies /µl) reperfusion and sham control (1816.2 ± 62.6 copies /µl) ( Figure 3a). This substantial increase of miR-210-3p was observed 24 hr following 60 min tMCAO, at which time the establishment of ischaemic infarct is clearly evident (Figure 3b and c). In situ hybridization performed on sections following tMCAO confirmed induction of miR-210-3p in the cortex at 3 hr following focal cerebral ischaemia with reperfusion, with increased expression detected at 24 hr ( Figure 3d to h).

| Increased miR-210-3p expression alters p70S6K signalling in vivo
To further investigate the mechanistic involvement of miR-210-3p in these critical signalling pathways and validate MTI predictions, we employed RPPA to examine the effects of increased miR-210-3p
To further examine the effects of miR-210-3p modulation on p70S6K activity, primary cortical neurons were transfected with synthetic LNA miR-210-3p mimic (5 nM), power inhibitor (10 nM), nontargeting negative control (10 nM) or vehicle for 48 hr and subsequently exposed to NMDA-mediated excitotoxic injury or conditions of OGD, followed by 24 hr recovery. TaqMan miRNA assays confirmed signif- Interestingly, under conditions of OGD, we did not identify a difference in the levels of p70S6K (Thr389) and nuclear p85S6K (Thr412) in non-targeting control-treated cultures and neurons over-expressing miR-210-3p following OGD (Figure 6j).

| D ISCUSS I ON
Increased levels of neuronal AMPK activation are observed rapidly following ischaemic injury both in vitro and in vivo within the ischaemic penumbra and contralateral hemisphere (Concannon et al., 2010;McCullough et al., 2005;Venna et al., 2012;Weisova et al., 2009Weisova et al., , 2011. We identified differential down-regulation of miR-124 and up-regulation of miR-210-3p in response to neuronal AMPK activation, potentially targeting downstream molecular pathways associated with ischaemia. Our validation of miR-210-3p expression demonstrated robust increases of miR-210-3p following AMPK activation with AICAR in primary cortical neurons and in the cortex of WT mice following 60 min tMCAO. In situ signal readily confirmed induction of miR-210-3p in the cortex following ischaemia at the same time points, and we demonstrated a rapid and specific increase in expression in response to NMDA excitotoxicity, as a central mechanism of neuronal injury following ischaemic stroke (Lai et al., 2014). miR-210-3p has previously been shown to be expressed in ischaemic cells and tissues, playing a key role in cellular adaptation to low oxygen environments such as tumourigenesis and ischaemia (Chan & Loscalzo, 2010). However, the mechanistic involvement of miR-210-3p in mediating cellular response to cerebral ischaemia is yet to be understood and studies to date have focused on ischaemic stroke in the subacute phase (Ren et al., 2016;Vijayan & Reddy, 2016;Zeng et al., 2011).
To elucidate the potential involvement of miR-210-3p in the regulation of downstream pathways we carried out pathway enrichment analysis. We report significant over-representation of several  (Jiang et al., 2020;Molloy et al., 2011). We also identified significant enrichment of signalling by Notch1 and Notch3. Notch signalling is activated under conditions of cerebral ischaemia and hypoxia, and in addition to its role in proliferation and differentiation, suppression of Notch activation in vivo has been reported to attenuate the development of ischaemic infarct development and neurological deficits (Arumugam et al., 2006(Arumugam et al., , 2018. Furthermore, Notch1 mediates excitotoxic neuronal injury through negative regulation of PTEN and activation of PI3K/Akt signalling via direct interaction with PI3K catalytic subunit p110γ (PI3Kγ) (Marathe et al., 2015). Of particular relevance, we predicted targeting of both PTEN and PI3Kγ (PIK3CG) by miR-210-3p.
PI3Kγ is induced by excitotoxic NMDA receptor activation, and the strong interaction of Notch1 with this catalytic subunit, along with inhibition of PTEN, induces Notch-dependent activation of Akt (Brennan-Minnella et al., 2013;Kim et al., 2011;Marathe et al., 2015).
This reciprocal crosstalk between Notch and PI3K/Akt signalling in the promotion of cell survival and proliferation is well reported as a promising targeted therapeutic strategy in the treatment of cancer (Calzavara et al., 2008;Saito et al., 2019). We also predicted targeting of Class I PI3K regulatory subunit p85α (PIK3R1), which plays a critical regulatory role in PI3K/PTEN signalling through direct interaction with both PI3Kγ and PTEN, inhibiting catalytic PI3Kγ activity and positively regulating PTEN activity (Chagpar et al., 2010;Chen et al., 2018).
We identified PI3K downstream effector PDK1 (PDPK1) as both a predicted and validated target of miR-210-3p and confirmed altered expression of PTEN, along with PDK1, p70S6K and target RPS6 signalling in response to increased miR-210-3p in vivo, further supporting a multi-targeting role for miR-210-3p in PI3K/PTEN/PDK1 signalling. PDK1 has previously been reported as a direct target of miR-210-3p, regulating miR-210-3p-induced apoptosis via inhibition of PI3K/Akt signalling, which is reversed upon PDK1 over-expression . PDK1 functions as a constitutively active 'master kinase' and is a required activator of p70S6K, phosphorylating the activation loop at Thr389/412 (Thr389) and Thr229/252 (Thr229) in vitro and in vivo, both of which must be phosphorylated for substantial kinase activation (Alessi et al., 1998;Balendran et al., 1999;Mora et al., 2004;Templeton, 2001). Phosphorylation of p70S6K at Thr389 is required for the subsequent selective phosphorylation of Thr229 by PDK1; catalytically inactive PDK1 has been shown to prevent p70S6K activation and phosphorylation at both Thr389 and Thr229, whereas p70S6K was found to be completely inactive, with no detectable phosphorylation at Thr389, in PDK1-knockout cells (Williams et al., 2000). Furthermore, PDK1 over-expression induces increased phosphorylation at Thr389 in unstimulated cells (Alessi et al., 1998;Balendran et al., 1999). Our analysis found a significant reduction in PDK1 following low-dose miR-210-3p, in- to environmental conditions such as energetic stress and ischaemia (van Rooij et al., 2007;Wilczynska & Bushell, 2015). Others studies have shown that miR-210-3p has regulatory roles in cell cycle, angiogenesis, DNA damage repair and immune response Pasculli et al., 2019).
To further support the potential modulation of PI3K-p70S6K signalling by miR-210-3p in settings of ischaemic stroke, we report that induction of miR-210-3p is accompanied by a rapid and sustained decrease in p70S6K (Thr389) activity following tMCAO in vivo and following glutamate receptor (NMDA) over-activation in primary neurons. This effect was reversed by miR-210-3p inhibition, resulting in restoration of p70S6K activity. In contrast, no induction of miR-210-3p was observed following tunicamycin-and thapsigargin-induced ER stress, indicating miR-210-3p induction and modulation of p70S6K activity as a specific response to NMDA receptor-mediated excitotoxicity (Concannon et al., 2008).
Both PI3K/PTEN and MEK/ERK signalling pathways converge on p70S6K to regulate its activity as a central molecule in the control of cell proliferation and growth through active protein translation and cell cycle progression (Lehman & Gomez-Cambronero, 2002).
The high energy cost of protein translation is tightly controlled by p70S6K through its ability to regulate the multiple phosphorylation of major downstream target, RPS6 and decreases in p70S6K (Thr389) and RPS6 (Ser235/236) phosphorylation have been observed following MCAO injury in vivo (Biever et al., 2015;Koh, 2013).
The critical role of p70S6K in the regulation of RPS6 phosphorylation is reflected in its ability to phosphorylate RPS6 at all residues, modulating mRNA translation in response to multiple signalling pathways (Biever et al., 2015). We detected decreased phosphorylation of RPS6 on Ser235/236 and Ser240/244 in response to increased miR-210-3p and decreased p70S6K in vivo. Inhibition of PI3K signalling during reperfusion has been shown to induce ERK-p70S6K activation, demonstrating a well-characterized 'cross-talk' ensuring robust signalling response (Hausenloy et al., 2004). While Ser240/244 is only phosphorylated by p70S6K, RPS6 phosphorylation at S235/236 can be regulated independently of p70S6K downstream of ERK through p90S6 kinases RSK1 and RSK2, representing an additional regulatory mechanism for RPS6 activation ensuring signal execution (Roux et al., 2007;Steelman et al., 2011). We also identified RPS6KA3 (RSK2) as both a predicted and validated target of miR-210-3p; ERK cannot activate RSK2 in the absence of PDK1 activity, further supporting a role for miR-210-3p in the regulation of integrated signalling outcomes that occur between PI3K-p70S6K and MEK/ERK-p70S6K cascades following reperfusion (Hausenloy et al., 2004;Jensen et al., 1999). Finally, our analysis also highlighted the potential for indirect regulation of PTEN, p70S6K and RPS6 via TF-mediated signalling, presenting an integrated picture of the multifaceted dynamics of miRNA-mediated gene regulation.
Rapid phosphorylation of AMPK downstream targets results in the down-regulation of energy-consuming processes, such as protein translation via inhibition of p70S6K, allowing for the restoration of energy homeostasis. Catalytic subunit AMPKα1 is a direct target of miR-210-3p, raising the interesting possibility that miR-210-3p regulates mature miRNA levels, presenting a potential feedback F I G U R E 6 miR-210-3p modulates p70S6K activity in response to NMDA-mediated excitotoxicity. (a and b), Western blot and dot plots depicting mean normalized protein abundance demonstrating decreased p70S6K (Thr389) in (a) ipsilateral cortex of WT mice 3 and 24 hr following reperfusion; representative of n = 4 per group and (b) primary cortical neurons to 24 hr following glutamate excitotoxicity (NMDA receptor over-activation, 100 μM, 5 min), n = 3 replicates per group, pooled. Analysis of miR-210-3p expression by digital PCR (c-e) and p70S6K mRNA (Rps6kb1) expression (f-h) in models of (c and f) NMDA-induced excitotoxicity (100 μM, 5 min) and (d and g) tunicamycin-(3 μM) and (e and h) thapsigargin-(1 μM) induced ER stress. Data presented as mean ± SEM, n = 2-3 independent cell culture preparations, carried out in triplicate and pooled. *p <.05 (ANOVA, post hoc Tukey's test). (i and j), western blot and box plots depicting mean normalized protein abundance of p70S6K (Thr389) expression in primary cortical neurons transfected with synthetic miRCURY LNA miR-210-3p mimic (5 nM), inhibitor (10 nM) non-targeting negative control (ctrl, 10 nM) or vehicle (sham) for 48 hr and subsequently exposed to (i) NMDA -mediated excitotoxicity (100 μM, 5 min) or (i) OGD (45 min), followed by 24 hr recovery. Data from n = 2 independent cell culture preparations, each with n = 3 replicates, pooled. Box plots depict mean (min -max) of protein from densitometry normalized to β-actin or tubulin. *p <.05 for p70S6K (Thr389) (ANOVA, post hoc Tukey's test) mechanism in the regulation of AMPK activity which may in part explain our findings. Notably, in the context of our findings, activation of AMPKα has been shown to play a central role in mediating downregulation of PDK1 expression (Hann et al., 2014).
The dual nature of many biochemical and gene expression responses to ischaemia, promoting both susceptibility and resistance to neuronal injury, presents a complex challenge in approaches to the regulation of cell signalling during ischaemic injury, impacting on the development of neuroprotective strategies. While we reported significant effects of miR-210-3p modulation on p70S6K activity through the use of a miR-210-3p mimic and inhibitor following NMDA excitotoxicity, we did not observe the same effect in response to OGD. Conditions of OGD are known to affect multiple other signalling pathways, including ER stress response and hypoxiainducible factor 1α (HIF1α) induction, with integral roles in adaptation to low-oxygen and/or glucose conditions. While NMDA-induced excitotoxicity is not associated with significant modulation of ER stress signalling, conditions of OGD have been shown to increase the expression of ER stress response genes (Concannon et al., 2008).
HIF1α plays a crucial role in hypoxia-sensing and adaptation through F I G U R E 7 Effects of exogenous miR-210-3p pre-treatment on neurological outcomes in WT mice following tMCAO with reperfusion. WT mice were administered LNA miR-210-3p mimic (0.5, 1.5, 5 pmol) or non-targeting control (ctrl) i.c.v. 4 hr prior to tMCAO (60 min, 24 hr reperfusion). (a and e) Quantification of infarct area (mm 3 ) in continuous coronal brain slice sections; values are mean ± SEM, n = 9-10 mice per group. (b and f) Total infarct volume (mm 3 ) and (c and g) hemispheric swelling at 24 hr following 60 min ischaemia with reperfusion. Hemispheric swelling is expressed as a percentage of the volume of contralateral hemisphere. (d and h) Neurological deficit scores at 1 hr following MCAO and 24 hr following reperfusion. Data presented as mean ± SEM, n = 9-10 for each group, *p <.005 compared with nontargeting control (ANOVA, post hoc Tukey's test) transcription of hypoxia-responsive genes involved in angiogenesis, cell proliferation and energy metabolism. miR-210-3p is a direct transcriptional target of HIF1α; HIF-1α binds to the miR-210-3p promoter and up-regulates miR-210-3p in response to low-oxygen conditions, and HIF1α is itself an established target of miR-210-3p, showing a mutually regulated feedback mechanism between miR-210-3p and HIF1α (Dang & Myers, 2015;Lu et al., 2019).
While not further investigated in our study, miR-124 is well established as a highly abundant brain-specific miRNA and its dysregulation is implicated in many CNS disorders (Sun et al., 2015).
Suppression of neuronal miR-124 has been shown to regulate AMPK/mTOR signalling, significantly increasing p-AMPK activity (Gong et al., 2016). Although we did not confirm a significant decrease in miR-124, we observed a biological effect, decreasing following AMPK activation in vitro and 3 hr following ischaemia in vivo. Increased plasma levels of miR-124 have been reported in rat MCAO models 24 -48 hr following occlusion and circulating levels positively correlated with tissue damage (Vuokila et al., 2020;Weng et al., 2011); disruption of brain tissue and increased blood-brain barrier permeability allowing for increased release may contribute to circulating miRNA levels not reflected in neuronal tissue. While we initially identified a decrease in miR-124 expression following 1 hr AICAR treatment in primary neurons, neuronal miR-124 levels were not measured in vivo until 3 hr post-ischaemia, potentially highlighting the importance of the time point of miRNA measurement in an evolving profile.
Our findings highlight the concerted action of multi-targeting modulation in ensuring physiological regulation of signalling pathways converging on specific protein activity. Elucidation of triggers and mediators of ischaemic cell death within the penumbra is particularly worthy of investigation given the complexity of interlinking physiological events impacting on brain injury maturation and outcome following stroke. However, for these reasons, artificial in vivo manipulation of miRNA may be associated with difficult-topredict off-target effects. We assessed the impact of pre-treatment with exogenous miR-210-3p in an in vivo setting. While we report no significant impact on cerebral injury, we identified a reduction in neurological deficit in low dose-treated groups; however, low level of expression may not be sufficient to have a substantial effect on cerebral injury. A subgroup analysis at higher doses reveals a dichotomous effect that is likely attributable to toxicity associated with the use of miRNA mimics at higher therapeutic levels in vivo. Off-target effects associated with the multi-targeting actions of miRNAs, based on seed complement frequencies in the transcriptome, may cause potential toxicities, off-setting potential for therapeutic efficacy.
This study is limited by the small sample size of in vivo exploratory studies carried out; these effects may be more pronounced in pathway analyses carried out in larger validation cohorts. Furthermore, the complex multi-targeting profile of miR-210-3p carries implications for other intracellular cascades not identified in our RPPA profiling. While not the purpose of this mechanistic study, future studies should address the therapeutic relevance of these findings on injury and functional outcome; miRNA mimics are a well-established method for characterization of miRNA function, however, success of in vivo therapeutic applications is limited as there are considerable implications associated with dose-dependent efficacy and toxicity at therapeutic concentrations.
The secretion of miRNAs, their high specificity, stability and easy detection in the circulation also places miR-210-3p as a functionally relevant blood biomarker in the diagnosis and management of ischaemic stroke and associated outcomes. Such a robust, non-invasive diagnostic and prognostic biomarker would contribute valuable and timely information necessary for prompt patient management decisions in the acute setting, enabling more effective determination of appropriate therapeutic intervention and monitoring of treatment response. Collectively, these results provide new molecular insight into the multi-targeting regulation of cell signalling in response to ischaemia, and suggest a novel mechanism whereby AMPK regulates miR-210-3p to control p70S6K activity in response to excitotoxic neuronal injury.

ACK N OWLED G EM ENTS
The authors acknowledge the funding supported by SFI (13/IA/1881; 17/COEN/3474) and the Munich Cluster of Systems Neurology (SyNergy; Project ID EXC2145/ID390857198). E.J. was supported by an RCSI International StAR Ph.D. studentship.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.