Non-amyloidogenic effects of α2 adrenergic agonists: implications for brimonidine-mediated neuroprotection

The amyloid beta (Aβ) pathway is strongly implicated in neurodegenerative conditions such as Alzheimer's disease and more recently, glaucoma. Here, we identify the α2 adrenergic receptor agonists (α2ARA) used to lower intraocular pressure can prevent retinal ganglion cell (RGC) death via the non-amyloidogenic Aβ-pathway. Neuroprotective effects were confirmed in vivo and in vitro in different glaucoma-related models using α2ARAs brimonidine (BMD), clonidine (Clo) and dexmedetomidine. α2ARA treatment significantly reduced RGC apoptosis in experimental-glaucoma models by 97.7% and 92.8% (BMD, P<0.01) and 98% and 92.3% (Clo, P<0.01)) at 3 and 8 weeks, respectively. A reduction was seen in an experimental Aβ-induced neurotoxicity model (67% BMD and 88.6% Clo, both P<0.01, respectively), and in vitro, where α2ARAs significantly (P<0.05) prevented cell death, under both hypoxic (CoCl2) and stress (UV) conditions. In experimental-glaucoma, BMD induced ninefold and 25-fold and 36-fold and fourfold reductions in Aβ and amyloid precursor protein (APP) levels at 3 and 8 weeks, respectively, in the RGC layer, with similar results with Clo, and in vitro with all three α2ARAs. BMD significantly increased soluble APPα (sAPPα) levels at 3 and 8 weeks (2.1 and 1.6-fold) in vivo and in vitro with the CoCl2 and UV-light insults. Furthermore, treatment of UV-insulted cells with an sAPPα antibody significantly reduced cell viability compared with BMD-treated control (52%), co-treatment (33%) and untreated control (27%). Finally, we show that α2ARAs modulate levels of laminin and MMP-9 in RGCs, potentially linked to changes in Aβ through APP processing. Together, these results provide new evidence that α2ARAs are neuroprotective through their effects on the Aβ pathway and sAPPα, which to our knowledge, is the first description. Studies have identified the need for α-secretase activators and sAPPα-mimetics in neurodegeneration; α2ARAs, already clinically available, present a promising therapy, with applications not only to reducing RGC death in glaucoma but also other neurodegenerative processes involving Aβ.

Glaucoma is a major cause of visual impairment worldwide and is characterised by optic neuropathy and visual field loss. Retinal ganglion cell (RGCs) apoptosis is considered an early hallmark of glaucoma 1 and raised intraocular pressure (IOP) is presently the only modifiable risk factor. 2 As a proportion of glaucoma patients continue to lose vision despite effective IOP control, 3 IOP-independent risk factors are increasingly thought to have a role in glaucoma pathology.
Amyloid beta (Aβ), the major constituent of senile plaques in Alzheimer's disease (AD), has recently been implicated in glaucoma pathology. 4,5 Aβ is associated with abnormal processing of amyloid precursor protein (APP). APP can be cleaved either by α-secretase via the non-amyloidogenic pathway, producing soluble APPα (sAPPα), or β-secretase producing sAPPβ, and Aβ, via the amyloidogenic pathway. 6 Using rodent glaucoma models, the amyloidogenic pathway has recently been identified as a target for the development of novel neuroprotective glaucoma therapies. 4,5 Here, Aβ deposition was found to induce RGC apoptosis, a finding supported by a study on glaucoma patients reporting reduced Aβ concentrations in the vitreous. 7 Aβ may therefore be important in the stress-response to glaucomatous neurodegeneration and offers a novel therapeutic target. 4 Brimonidine (BMD), Clonidine (Clo) and Dexmedetomidine (Dex) are α2 adrenergic receptor agonists (α2ARAs). Apraclonidine, a para-amino derivative of Clo, is a topical α 2ARA 8 routinely used in the clinic to reduce IOP spikes induced by neodymium:YAG laser treatment for posterior capsule scarring after cataract surgery. 9 However, the reduced activity of Apraclonidine in controlling IOP with chronic usage 10 coupled with increased risk of follicular conjunctivitis, 11 renders it unsuitable for long-term glaucoma management. BMD was introduced as an IOP-lowering agent; however, increasing experimental evidence suggests it also has IOP-independent neuroprotective activity. 12 This was clinically demonstrated in a prospective, randomisedcontrolled study where BMD was reported to significantly preserve visual field in low-tension glaucoma patients compared with the beta-blocker timolol. 13 Both Clo and Dex are used as anaesthetics, 14 Clo is used to treat migraine, hypertension and menopausal flushing, 15 and Dex for sedation during intensive care. 16 In vivo studies have demonstrated Clo and BMD to have retinal neuroprotection [17][18][19] with functional benefits, 20,21 and Dex to have neuroprotection against cerebral ischaemia, 22 excitotoxicity 23 as well as in a model of traumatic brain injury. 24 Although α2A A and α2A B receptors have been identified in the RGC layer (RGCL) of the inner retina, 25 the mechanisms by which α2A agonists exert neuroprotection are not wellestablished. Various pathways have been proposed, including cyclic adenosine monophosphate (cAMP) reduction, 26 NMDA receptor neuromodulation, 21 increasing cell survival proteins p-Akt and bcl-2 (ref. 27) and neurotrophic factor expression. 22 The present study seeks to delineate the mechanisms of α2A-mediated neuroprotection using glaucoma-related in vivo and in vitro models, and investigate the involvement of the Aβ-pathway.

Results
α2ARAs are neuroprotective against retinal neuronal death in vitro and in vivo. Retinal neuronal cells (RNs) were pre-treated for 24 h with BMD, Dex or Clo before insulting with the hypoxia-mimic cobalt chloride (CoCl 2 ) 28 or UV light to induce neurotoxicity, based on previously determined IC 50 doses. 29 CoCl 2 induced a decrease in cell viability in both primary and immortalised cell types, which was significantly reduced with BMD at all concentrations assessed (0.1-100 μM, Po0.01 Figures 1a and c), whereas Clo was protective at 0.1, 1 and 100 μM (Po0.05), whereas Dex was effective at 0.01 μM (Po0.001, Figure 1a).
Effects of systemic α2A agonists BMD and Clo were examined in vivo using an ocular-hypertensive (OHT) rat model. 30 Comparison of the IOP profile between untreated OHT control and no-OHT groups showed OHT surgery produced a significant increase in IOP up to 3 weeks post surgery (Po0.05, Figure 2a). Systemic Clo treatment lowered IOP at 1, 2 and 3 weeks compared with the untreated OHT group (all Po0.01), whereas systemic BMD had no effect on IOP (Figure 2a), as expected. 31 Significantly, more apoptotic RGCs were observed at 3 and 8 weeks in the untreated OHT model compared with the no-OHT control eyes (Po0.01; Figure 2b), with peak RGC apoptosis occurring at 3 weeks, consistent with previous findings. 1,4 Both BMD and Clo significantly reduced RGC  Figure 1 α2 agonists are neuroprotective against RGC death in vitro. α2 agonists' effects on neuroprotection were investigated in vitro using the MTT assay in immortalised RNs against CoCl 2 (hypoxic) and UV (stress) insults; against 250 μM CoCl 2 BMD showed significant neuroprotection at all concentrations tested, Dex showed protection at 0.01 μM and Clo was neuroprotective at 0.1, 1 and 100 μM (a). Against the insult of UV, BMD was protective at 100 and 10 μM and Clo and Dex were both protective at 0.01 μM (b). The protective effects of BMD at 10 and 100 μM against CoCl 2 were reproduced in mouse-derived primary mixed retinal cultures (c). In immortalised RNs, BMD (10 μM) was found to be protective against an UV-light insult of 80 j/cm 2 (d  Figure 3h). Consistent with previous in vivo findings, 4 elevated levels of Aβ were observed in the RGCL 3 weeks post OHT model induction compared with no-OHT control (Po0.001, Figure 4a) associated with a significant increase in APP (Po0.001, Figure 4b).   Figure 3 α2A agonists directly reduce retinal neuronal Aβ and APP levels in different in vitrostress models. α2 agonists' ability to reduce Aβ levels was next investigated in vitro in immortalised RNs using CoCl 2 and UV insults, which significantly increased Aβ levels (a). Aβ levels were significantly reduced in CoCl 2 samples when co-treated with 10 and 100 μM BMD (b). All α2A treatments were associated with a reduction in Aβ levels at concentrations of 10 and 100 μM, although BMD was found to be most effective. Representative images from CoCl 2 insulted cells show BMD treatment (c) reduced Aβ (red) staining in RNs compared with CoCl 2 only treatment (d); (α-tubulin staining (green), DAPI staining (blue)). APP levels were significantly reduced with 10 μM BMD treatment against CoCl 2 insult (e). Representative images from CoCl 2 insulted RGCs show BMD treatment (f); reduced APP (green) staining in RGCs compared with CoCl 2 only treatment (g); (DAPI staining (blue)). α2A agonists directly affect the non-amyloidogenic pathway and elicit neuroprotective activity through sAPPα. Having established α2ARA treatments reduce Aβ-associated-RGC apoptosis, we investigated whether these effects were mediated through changes to APP processing. Levels of sAPPα (a product of the nonamyloidogenic pathway 6 ) were histologically assessed in the OHT model. sAPPα staining was not significantly different either at 3 or 8 weeks in the RGCL in untreated OHT and control eyes ( Figure 5a). BMD treatment significantly increased sAPPα levels at both 3 and 8 weeks (2.1-and 1.6-fold increase, respectively) compared with untreated OHT controls (Po0.05 and Po0.01, Figures 5a-e). Clo treatment had no effect.
The effect of BMD on the non-amyloidogenic pathway was confirmed in vitro using the hypoxia mimetic CoCl 2 to induce RN toxicity. A twofold significant increase in sAPPα levels was observed in response to 10 μM and 100 μM BMD treatment (Po0.01, Po0.05, respectively, Figures 5f-h).
Having previously established BMD treatment reduces Aβ levels in response to UV light induced toxicity, the effect of an sAPPα antibody on UV-insulted and BMD-treated cells was investigated. Although sAPPα antibody exposure caused no significant change in RGC viability after UV exposure, it significantly inhibited protection by BMD therapy (Figure 5i). This observation, coupled with data demonstrating that α2ARA treatment significantly reduced APP levels, suggests α2ARA mediated RGC neuroprotection is achieved through increased sAPPα through upregulation of non-amyloidogenic APP processing.
The neuroprotective effect of α2ARAs via Aβ-related pathways involves modulation of the ECM. Laminin and MMP-9 expression have been implicated in Aβ-related pathways. 32 The occurrence of RGC apoptosis in OHT is associated with a reduction in laminin in the RGCL at 3 months in the same OHT model. 33 We therefore investigated whether extracellular matrix (ECM) modulation had a role in α2ARAs-mediated neuroprotection using the OHT model.
A twofold reduction in laminin was observed in the RGCL at 3 and 8 weeks post OHT induction compared with the no-OHT control eyes (Po0.05, Figure 6a). BMD treatment restored laminin levels in the OHT model RGCL at both 3 and 8 weeks (Po0.05, Figures 6a-e). A significant (Po0.05) twofold reduction in MMP-9 expression was observed in the RGCL 3 weeks post OHT induction but not at 8 weeks (Figures 6f-j). BMD treatment induced a marked but not statistically significant increase in MMP-9 levels at 8 weeks (Figure 6f). Laminin and MMP-9 expression was also investigated in the Aβ-inducing apoptosis model. 4 BMD treatment was associated with a significant reduction in laminin deposition in the RGCL versus Aβ-treated controls, although there were no differences in comparison to the no-OHT control (unpaired t-test, Po0.05, Figure 6k).
To further investigate the effect of α2ARAs on MMP-9 expression, in vitro studies were performed on RNs insulted with CoCl 2 . Using zymography, CoCl 2 increased both pro-and active MMP-9 ( Figure 6h, lane 2), compared with untreated cells (lane 1). Treatment with BMD reduced both pro-and active MMP-9 ( Figure 6l). A marked but not statistically significant reduction in MMP-9 was observed histologically in the in vivo Aβ-induced apoptosis model, although in the Aβ-treated eyes, the MMP-9 staining was significantly lower than in the no-OHT controls (Po0.001, Figure 6m). These results suggest that α2ARA modulate laminin and MMP-9 expression in RGCs, which may be linked to changes in Aβ and APP processing.

Discussion
The present study confirms the neuroprotective actions of α2ARAs using in vivo and in vitro models of retinal neurodegeneration with a novel IOP-independent mechanism of action. This mechanism proposes that a reduction in RGC apoptosis is achieved through reduced Aβ production, and its precursor APP, via stimulation of the non-amyloidogenic pathway as evidenced by a significant increase in sAPPα, which leads to modification of ECM proteins laminin and MMP-9 ( Figure 7).
Compared with Clo, systemic BMD administration does not affect IOP, 31,34 suggesting that BMD's neuroprotective effect is IOP independent. This observation is supported by the LoGTs study, 13 where BMD was effective in preserving visual field independent of IOP reduction in adults with low-pressure glaucoma. Further investigations of the clinical use of BMD as a neuroprotectant are currently underway (clinicaltrials.gov NCT00658619, NCT00693485, NCT01229410). Furthermore, the ability of BMD to improve visual function is documented in rodent models of retinal disease. [35][36][37] This study suggests that different activity levels between α2ARAs could be attributed to differences in their selectivity for the α2A and the α2A A subtype. BMD and Dex are direct agonists α2 agonist neuroprotection against amyloid β Shereen Nizari et al of the α2A A receptor subtype, and are inhibited by the α2A A antagonist yohimbine in vivo 27 and in vitro. 38 In comparison, Clo has reduced selectivity for α2 receptors. 8,17,39 Neuroprotective effects α2ARAs are understood to be primarily mediated though α2A receptors, 11 supporting BMD's effects in this study. α2A receptors are G-protein-coupledreceptors whose activation leads to an inhibition of adenylate cyclase and a decrease in cAMP. 40 cAMP is implicated in Aβ modulation by stimulation of APP synthesis and processing, affecting the amyloidogenic and non-amyloidogenic pathways, 41,42 potentially through the cAMP/PKA/APP/Aβ pathway. 43 Several mechanisms have been proposed for α2Amediated neuroprotection, including increased expression of   Representative images show the marked but non-significant increase in general (pro and active forms) MMP-9 levels (green) in the RGCL of BMD-treated OHT models at 3 (h) and 8 (j) weeks compared with OHT control eyes (g, i), (DAPI, blue). (k) In comparison, a different pattern was seen in the Aβ neurotoxicity model, where BMD treatment was associated with a significant reduction in laminin in the RGCL. (l) Further assessment of MMP-9 activity was performed using zymography on immortalised RN exposed to CoCl 2 , where BMD treatment appeared to substantially reduce MMP-9 activity in both pro-and active forms, although this effect was reduced at a high concentration (100 μM).  27,47 which support this study's observed increase in P-Bad(Ser136). α2A agonists have also been reported to effect the glutamate pathway, 48 with reductions in intracellular cAMP production and cytosolic calcium shown in RGCs. 21 Recently, BMD was reported to reduce expression of NMDA subunits NR1 and NR2A in a model of ischaemia. 19 Similar activity in excitotoxicity models is reported for Clo 17,49 and Dex 50 in the CNS. Elevated APP and Aβ have been reported in glaucoma models, 4,51 potentially linking elevated IOP promotion of APP processing, and inhibition of APP anterograde transport from the RGCL to the optic nerve. 51,52 BMD is reported to preserve optic nerve axons and active transport throughout the visual pathway in a rodent OHT model, 18 which may explain the present study's finding where α2A agonists reduced levels of APP and Aβ in the RCGL of the OHT model. This hypothesis is supported by our in vitro data, where CoCl 2 and UV-induced elevation of APP and Aβ was inhibited by α2ARAs. A possible pathway by which this may occur is through α2ARAs' ability to inhibit cAMP production. 21 This study suggests for the first time that the neuroprotective effects of α2ARA BMD are directly associated with increased sAPPα, as the observed BMD-mediated neuroprotective effect was reversed using an sAPPα antibody in vitro. Furthermore, the increase in sAPPα appears to be specific to α2A A activity, as Clo did not produce a comparable effect. 11 Previously, sAPPα was reported to be neuroprotective in vitro. 53,29 More recently, Obregon et al. reported that sAPPα decreases Aβ generation directly by associating with the β-site APP-converting enzyme (BACE1) both in vivo and in vitro. Importantly, the authors proposed that the levels of sAPPα were so crucial that an imbalance could stimulate amyloidogenic APP processing, 54 concluding that sAPPα mimetics are a potential therapeutic target for the treatment of AD. This was further highlighted in a recent paper by Willem where unwanted side-effects on neuronal activity were implicated by treatment with BACE1. 55 The results of the present study suggest that α2A A agonists are sAPPα modulators, promoting RGC survival in a non-IOP-dependent manner.
A mechanism for the stimulation of sAPPα by α2A agonists may be through the activation of α-secretases, which can occur via protein kinase C (PKC), phosphatidylinositol 3kinase (PI3K) and mitogen-activated protein kinase-ERK. 56 Evidence suggests that APP cleavage by the α-secretase ADAM10 (a disintegrin and metalloproteinase and physiological α-secretase in neurons) constitutively produces sAPPα through 5-HT4 receptors in a cAMP-independent pathway. When stimulated by 5-HT4 receptors agonists, however, sAPPα secretion is mediated through cAMP/Epac (exchange protein activated by cAMP) signalling. 57 A possible explanation for this study's findings is that α2ARAs could directly activate α-secretase ADAM10 and sAPPα production through cAMP/Epac modulation.
ADAM10 and MMP-9 have been reported to stimulate sAPPα production and decrease Aβ, through α-secretase-like activity. 58,59 We present in vitro data suggesting that α2ARAs increase levels of pro-MMP-9 and sAPPα, whereas decreasing Aβ levels. This concurs with in vivo studies where elevated MMP-9 levels in a AβPP transgenic mouse model led to a reduced plaque burden and increased sAPPα levels. 60 As both ADAM10 and MMP-9 are matrix metalloproteases, their activity levels may be similar. 61,62 In the retina, NMDA activation is reported to increase MMP-9 activity; 63 α2A agonists could therefore reduce MMP-9 activity by inhibiting retinal excitotoxicity. 21 Furthermore, MMP-9 is reported to be influenced by Aβ, as intracerebroventricularly injected Aβ is reported to increase MMP-9 levels in vivo. 63 The reduction in Aβ levels by α2ARAs observed may be responsible for the reported reduction in active MMP-9 levels. The effects of α2ARAs on MMP-9 appear complex and closely related: α2A agonists may reduce active MMP-9 levels through modulation of NMDA and Aβ activity, but this reduction in MMP-9 may inturn decrease its α-secretase-like activity and the subsequent promotion of the non-amyloidogenic pathway.
MMP-9 activity in retinal neurodegeneration is linked with laminin, it was reported that RGC degeneration is associated with reduced laminin and stimulation of MMP-9. 33 We report similar changes in laminin in the RGCL of the OHT model, but at earlier time points than previously described. Laminin is suggested to be neuroprotective through a laminin-integrin signalling pathway, with evidence that laminin promotes RGC survival in vitro through β1 integrin-focal adhesion kinase signalling. 64 The reduction in laminin reported in this study was reversed by α2A agonist BMD treatment. BMD has been reported to induce laminin binding protein expression, laminininduced neuronal and axonal changes, essential for promoting axonal preservation and growth. 65 Furthermore, laminin is linked to modification of APP and Aβ processing and found in senile plaques in Alzheimer's Disease, 66 where it prevents Aβ fibril formation and reduces neurotoxicity. 32 This effect was confirmed in vivo, with laminin

A2 Agonists Positively effects
Non-Amyloidgenic Figure 7 Diagram representing α2A agonists' neuroprotective effect against Aβ toxicity. APP can be processed into Aβ or sAPPα through the amyloidogenic and nonamyloidogenic pathways, respectively. α2A receptor agonists can negatively affect the amyloidogenic pathway, preventing cell death, partly by modulating excitotoxicity caused by glutamate (GLU). α2A receptor agonists can also affect APP processing via the extracellular matrix, by modulating MMP-9 and laminin through LBP, preventing further toxic interactions with Aβ and by increasing processing of APP into sAPPα, promoting the non-amyloidogenic pathway. α2A receptor agonists can also increase levels of P-Bad, and therefore overall act to promote cell survival and neuroprotection α2 agonist neuroprotection against amyloid β Shereen Nizari et al depletion accelerating Aβ-induced neurotoxicity, altering the distribution of Aβ aggregates in Caenorhabditis elegans. 67 Intravitreal injection of Aβ [25][26][27][28][29][30][31][32][33][34][35] was found to induce increased laminin in the RGCL, which was reversed by BMD treatment, contradictory to the effect of BMD treatment observed in the OHT model. The increase in laminin observed in the Aβ 25-35 model may be a response to direct Aβ-induced neurotoxicity. 32,67 The reduction of Aβ-induced neurotoxicity with BMD treatment may result in an attenuation of the increased laminin levels in response to the acute presence of Aβ. This hypothesis is supported by the observation that BMD reduces levels of Aβ deposition in both the OHTand Aβ models in vivo, inhibiting Aβ neurotoxicity and RGC apoptosis.
In addition to modulating Aβ levels through sAPPα, this study confirms that α2A agonist BMD can elicit neuroprotection through P-Bad (Ser136) (Supplementary Figure1a-d). Aβ is reported to increase Bad de-phosphorylation and cell death by stimulating calmodulin-calcineurin activity. 68 BMDmediated reduction of Aβ may increase P-Bad, promoting cell survival pathways involving PI3K, known to be upregulated by BMD upstream of P-Bad. 27 PI3K activity can itself decrease Aβ levels, 69 providing further corroboration that BMD reduces Aβ. In addition, upstream of Bad, PKC activation simultaneously increases sAPPα and reduces Aβ levels, providing further support for non-amyloidogenic promotion of cell survival. 70 An alternative and complementary pathway by which α2A agonists reduce Aβ expression in OHT, may be via oxidative stress modulation. 26 Oxidative stress reduces expression of sAPPα and α-secretase activity 71 and is implicated in the pathogenesis of glaucoma, where an ischaemic element is well-recognised. 72 The observation that BMD increases sAPPα and reduces Aβ levels in RGCs when exposed to a hypoxic insult (CoCl 2 ) supports evidence for α2A agonist's neuroprotective activity under oxidative stress. Although Aβ 1-42 monomers may possess neuroprotective function, 6 Aβ oligomers and Aβ [25][26][27][28][29][30][31][32][33][34][35] used in this study are widely reported to induce oxidative stress. 4 The effects of α2A agonists against oxidative stress are supported by our observation that BMD and Clo are protective against Aβ 25-35 -induced RGC apoptosis in vivo.
Aβ neurotoxicity is associated with glutamate NMDA receptor activation and is the basis of the FDA-approved NMDA receptor antagonist Memantine for the treatment of Alzheimer's Disease. 73 This pathway has also been implicated in the development of Aβ induced-dendritic spine loss and tauassociated neurodegeneration. 74 Short-term NMDA receptor stimulation is reported to increase non-amyloidogenic α-secretase-mediated APP processing and sAPPα production, 75 whereas chronic stimulation is reported to have the opposite effect. 76 α2ARAs can also modulate NMDA receptors; Dong et al. report that BMD acts post-synaptically on NMDA receptors by reducing levels of intracellular cAMP; this indirect effect could also be responsible for BMD suppressing Aβ-induced excitotoxicity. 21 Clo has been shown to regulate GABAergic synaptic inputs in the CNS, which could explain its neuroprotective effect reported in this study in the OHT model. 49 In conclusion, we confirm that α2ARAs are neuroprotective of RGC death in vivo and in vitro, substantiating previous reports that this effect is non-IOP dependent. We suggest a new mechanism by which this occurs; we show through multiple modalities that α2ARAs modulate Aβ toxicity, decreasing levels of Aβ and APP in vivo and in vitro, and increasing sAPPα formation through the non-amyloidogenic pathway (Figure 7). We specifically demonstrate that neuroprotective effects of α2ARA BMD are mediated through sAPPα. Furthermore, BMD affects laminin, in association with Aβ, which influences RGC survival. Recent studies have identified the need for α-secretase activators and sAPPαmimetics in neurodegeneration. We propose that α2ARAs may be the most commonly available clinical sAPPα modulators, and being 'tried and tested' may offer an economical advantage as a pre-existing therapeutic for neuroprotection. The applications of α2ARAs may therefore not be limited to reducing RGC death in glaucoma but also to any neurodegenerative process where Aβ neurotoxicity is involved, such as AD. This work strongly advocates investigation of the therapeutic potential of α2ARAs in these disorders.

Materials and Methods
Ethics statement. All procedures were approved by the UK Home Office and the University College London Ethics Committee and were conducted in accordance with the Association for Research in Vision and Ophthalmology statement and ARRIVE guidelines.
Chemicals. BMD was purchased from Sigma Aldrich (Dorset, UK), or obtained from Allergan (Oregon, USA); Clo and Dex were purchased from Santa Cruz Biotechnology Inc (Santa Cruz, CA, USA).
Rat model of OHT. The OHT model was surgically induced in 30 rats using a well-established method. 30 IOP was elevated in the left eye of each animal by injection of 50 μl hypertonic saline (1.80 M) into the episcleral veins using a syringe pump (UMP2, World Precision Instruments, Sarasota, FL, USA). A propylene ring, with a 1 mm gap cut out of its circumference, was placed around the equator to prevent saline outflow from other aqueous veins, as described previously. 4,33 Contralateral, un-operated eyes served as controls. The IOP of both eyes was measured in mmHg at regular intervals using a Tonolab Tonometer (Tiolat Oy, Heisinki, Finland). Treatments of BMD (n = 10), Clo (n = 10) and PBS control (n = 10) were administered intraperitoneally at the time of surgery. Animals were imaged for RGC apoptosis by administrating Alexa-488-labelled annexin A5 (1.25 μg in 5 μl) intravitreally at 3 (n = 15) and 8 (n = 15) weeks, before being killed for histological assessment. 1,33 The ratio of IOP was calculated by dividing the IOP from the operated, OHT left eye (OS) by that of the un-operated, no-OHT right eye (OD). Analysts were masked to treatment group identity. Statistical analysis on IOP data was carried out using the non-parametric two-tailed Mann-Whitney test comparing treated OHT to untreated no-OHT control (n = 5).
Rat model of Aβ. The Aβ model was prepared in 15 rats to induce RGC apoptosis as described previously. 4 In brief, 25nmol Aβ 25-35 (Sigma Aldrich) was dissolved in sterile water and intravitreally administered unilaterally. BMD (n = 5) Clo (n = 5) or vehicle (PBS) (n = 5) was administered intraperitoneally at the time of Aβ administration. Rats were imaged for RGC apoptosis (as described above) at baseline and 3 days following treatments before sacrifice.
Histology and immunohistochemistry. RGC apoptosis counts were performed in vivo and histologically ex vivo, using whole-retina mounts. 4 For wholeretina mounts, eyes were enucleated and immediately fixed in 4% α2 agonist neuroprotection against amyloid β Shereen Nizari et al paraformaldehyde following termination. Eyes were dissected at the equator, and the cornea, lens and vitreous were removed. RGC apoptosis was identified by fluorescent annexin A5 staining. RGC apoptosis density counts were performed in a masked fashion by three observers, and were calculated as previously described. 4,77 Immunohistochemistry was performed using antibodies listed in Table 1. Dissected eyes were fixed in 10% formalin before processing in methanol solutions with increasing concentrations, and embedding in paraffin. 33 In total, 3 μm thick sections were then cut and incubated with antibodies diluted in TBTA (Table 1). Stained sections were analysed using fluorescent microscopy (Zeiss Axiovert S100) at x20 magnification, by a masked observer. The RGCL was segmented using a box of fixed proportions and image background noise was subtracted using a 25 pixel rolling ball radius, before recording the mean fluorescence intensity through the pixels' grey values using ImageJ software (NIH) as previously described. 78 Colocalisation analysis was carried out using the Pearson's Coefficient test with the ImageJ JaCoP plugin 79 on segmented double-labelled RGCL images. Cell viability studies. The neuroprotective effects of α2 agonists were assessed in vitro using either the MTT or the alamarblue cell viability assays. In brief, RN were seeded at a density of 3 × 10 3 cells/well in a 96-well plate and cultured for 24 h (37°C, 5% CO 2 ) before incubation with specified concentrations of BMD prepared from a 24 mM stock solution in DMSO, Clo and Dex for 24 h. Cells were insulted with predetermined (IC 50 ) concentrations of CoCl 2 or UV-B light for 24 h. Cell viability was assessed using MTT assay, where cells were treated with 0.5 mg/ml MTT in culture media for 2 h before dissolving the resulting formazan crystals using DMSO (0.1 ml/well) and measuring absorbance (Safire microplate reader) at 570 nm. For the alamarblue viability assay, 10 μl of alamarblue solution was added per 100 μl DMEM and incubated for 1.5 h measuring absorbance (Safire microplate reader) at 570nm. All experiments were carried out in triplicate.
Assessment of secreted MMP-9 activity by zymography. RNs were cultured to 70-80% confluence before pretreatment with BMD and insulting with 250 μM CoCl 2 for 24 h. Conditioned media was collected in the presence of protease cocktail inhibitors (Merck, Millipore, Nottingham, UK), and concentrated using centrifugal filters with a MWCO of 10kDa. Total protein concentration was determined using a BCA assay kit (Thermo Scientific, Rockford, IL, USA) according to manufacturer's instructions. Conditioned media containing 20 μg of protein per sample was run on a 10% SDS gel containing 0.1% gelatin (Sigma) at 180 V for 1 h rtp. SDS was removed by two 45 min washes in 2.5% PBS-Triton, followed by incubation in developing buffer (Invitrogen, Paisley, UK) overnight at 37°C. Gels were stained with Coomassie blue to reveal clear bands indicating the presence of secreted pro-and active MMP-9. 82 Immunocytochemistry. Immunocytochemistry was performed to elucidate mechanisms of α2 agonist-associated neuroprotection. RNs were seeded on glass coverslips before pre-treating with BMD, Clo or Dex for 24 h and insulted with 250 μM CoCl 2 or UV-B light (80 j/cm 2 ) for a further 24 h. Cells were fixed in 4% paraformaldehyde before permeabilizing with 0.1% PBS-Triton and blocking with 3% BSA. Cells were incubated overnight in either anti-P-Bad (Ser136), anti-sAPPα, anti-APP or anti-Aβ with anti-α Tubulin (Cell Signalling Technology, Beverly, MA, USA) to visualise cell structure. Antibodies were diluted 1:100 in 3% BSA, except anti-α Tubulin (1:50) ( Table 1). Fluorescence microscopy and analysis was performed as described previously. All experiments were carried out in triplicate.
Experimental design. Sample size estimation for the OHT study was based on the size of 'neuroprotective' effects of BMD in glaucoma patients in preserving visual field in the LoGTs study. 13 This effect was 90.9%, and based on a power of 0.85 and alpha of 5%, this provided a minimum number of four animals per group per time point. OHT rats were randomly assigned in blocks of three to treatment with either BMD or Clo (n = 5 at 3 weeks and n = 5 at 8 weeks), or vehicle (PBS, n = 10). For the Aβ model, samples size was based on our previous published results. 4 Ex vivo analyses were planned with five rats per group, but owing to one death and one atrophic retina, a final number of three to five rats per group were used, as sample material was limited. Animals were randomly assigned in blocks of three to treatment, as above. All in vitro experiments were carried out in triplicate.
Statistical analysis. Unless otherwise stated all statistical tests comprise a one-way ANOVA with either Dunnett's or Tukey's post-tests, with Po0.05 taken as statistical significance.

Conflict of Interest
Professor MF Cordeiro holds a patent regarding DARC technology. The remaining authors declare no conflict of interest.