A Brain-Derived Neurotrophic Factor Mimetic Is Sufficient to Restore Cone Photoreceptor Visual Function in an Inherited Blindness Model

Controversially, histone deacetylase inhibitors (HDACi) are in clinical trial for the treatment of inherited retinal degeneration. Utilizing the zebrafish dye ucd6 model, we determined if treatment with HDACi can rescue cone photoreceptor-mediated visual function. dye exhibit defective visual behaviour and retinal morphology including ciliary marginal zone (CMZ) cell death and decreased photoreceptor outer segment (OS) length, as well as gross morphological defects including hypopigmentation and pericardial oedema. HDACi treatment of dye results in significantly improved optokinetic (OKR) (~43 fold, p < 0.001) and visualmotor (VMR) (~3 fold, p < 0.05) responses. HDACi treatment rescued gross morphological defects and reduced CMZ cell death by 80%. Proteomic analysis of dye eye extracts suggested BDNF-TrkB and Akt signaling as mediators of HDACi rescue in our dataset. Co-treatment with the TrkB antagonist ANA-12 blocked HDACi rescue of visual function and associated Akt phosphorylation. Notably, sole treatment with a BDNF mimetic, 7,8-dihydroxyflavone hydrate, significantly rescued dye visual function (~58 fold increase in OKR, p < 0.001, ~3 fold increase in VMR, p < 0.05). In summary, HDACi and a BDNF mimetic are sufficient to rescue retinal cell death and visual function in a vertebrate model of inherited blindness.

There is an unmet clinical need to develop effective treatments for inherited vision loss. Inherited retinal degenerations (iRD) are clinically and genetically heterogeneous, presenting as diverse forms of blindness that include retinitis pigmentosa (RP), Leber congenital amaurosis (LCA), achromatopsia (ACHM), cone-rod dystrophy (CORD) and macular dystrophy (MD) 1 . iRDs are estimated to affect 1 in every 2000-3000 people 2 , and many are characterised by loss and/or dysfunction of photoreceptor cells. Potential interventions include drug-, gene-or cell-based therapy 3 . Cell therapy has potential to restore functional retinal cells irrespective of the genetic cause of iRD 4,5 . However, the integration of replacement progentitor photoreceptor cells which restored visual function in pre-clinical models 6 has been challenged by evidence demonstrating intracellular exchange between donor and host cells 7 . Gene therapy demonstrated initial success in clinical trials 8,9 . However, this approach is hampered by limited applicability due to genetic heterogeneity of iRD 10 . Additional concerns regarding long-term efficacy are exemplified by RPE65 gene replacement which fails to halt the progressive loss of photoreceptors 11 . An alternative to replacing defective cells or genes is to identify neuroprotective agents preventing loss of visual function 12 . Selection of neuroprotectants can be based on efficacy in other neurodegenerations or de novo discovery 13 . Figure 1. dye mutants display aberrant retinal morphology. Representative lateral and dorsal views in brightfield of sibling larvae (a,a′) and dye mutants (b,b′), displaying pigmentation and gross morphological defects present in the dye mutant, including swim bladder deflation compared to sibling (*in a) and pericardial oedema (black arrow in b). Transverse sections of the retina and ciliary marginal zone of sibling larvae (c,e) and dye mutants (d,f), pyknotic nuclei (red arrows in d,f) present in the ciliary marginal zone indicated dying cells. Transmission electron micrographs of photoreceptors in sibling larvae (g) and dye mutants (h) displaying shortened photoreceptor OS and aberrant morphology despite preserved ultrastructure (i,j). RPE cells in dye mutants are hypopigmented and contain large inclusion bodies compared to sibling larvae (k,l), and the RPE fails to interdigitate with photoreceptors in dye compared to siblings (N = 1 for sibling, N = 3 for dye mutants). Neuroprotectants can modulate cell survival or death pathways common to many iRD, offering potential widespread applicability, irrespective of the genetic cause.
Previously reported retinal neuroprotectants include rod cone viability factor (rdCVF) 14 , brain-derived neurotrophic factor (BDNF), ciliary neurotrophic factor (CNTF) 15 , nicotinic acetylcholine or serotonin receptor agonists 16,17 and histone deacetylase inhibitors (HDACi). HDACi are clinically administered as mood stabilisers or anti-epileptic agents 18 and are in clinical trials as anti-cancer 19 or anti-inflammatory 20 drugs. Notably, HDACi are under investigation to treat neurodegeneration in Alzheimer's and Huntington's disease 21 . Not suprisingly, HDACi recently emerged as potential drugs for iRD. A retrospective review of retinitis pigmentosa patients prescribed the HDACi valproic acid (VPA), showed improved visual field (VF) and visual acuity (VA) over an average 4 month treatment period 22 . The average optotype increased from 20/47 to 20/32. The study noted no negative effects on either VF or blood chemistry. Improvements in VF were clinically significant and indicates that VPA prevents progressive peripheral-central degeneration of retinal photoreceptor cells. However, multiple study criticisms include the small sample size, the short study period and the statistical methodology 23 . A similar retrospective review of a 9.8 month average treatment study suggests VPA has no beneficial effect on VF or VA readings 24 . Sisk et al. suggest HDACi treatment is appropriate for autosomal dominant RP acting to chaperone mutant rhodopsin, but may be harmful in autosomal recessive RP 25 . In summary, there is growing but conflicting evidence that HDACi are neuroprotective in a wide range of pre-clinical models of RP, retinal ischemia and retinal drug induced excitotoxicity [26][27][28] .
Zebrafish provide novel opportunities to efficiently investigate vertebrate vision and pharmacological interventions in vivo 29 . At 3 days post fertilisation (dpf), ocular morphogenesis is largely complete and a differentiated laminated retina prevails 30 . Vision in zebrafish larvae is cone-mediated, unlike nocturnal rodent models whose vision is largely rod-mediated 31 . The zebrafish retina is abundant in cone photoreceptors, arranged in mosaic patterns of red/green, blue, and UV subtypes in the photoreceptor cell layer 32 and zebrafish models of inherited blindness (e.g. zatoichi s125 , pde6c w59 and no optokinetic response f w21 ) demonstrate that genes essential for cone function (gc3, pde6 and gnat2) are functionally conserved between fish and humans [33][34][35] . Convenient methods to assess visual function in zebrafish include behavioural assays or direct retinal electrophysiology [36][37][38] .
Here, we demonstrate pharmacological rescue of cone-mediated vision in the dye ucd6 (https://zfin.org/ ZDB-ALT-170502-2) zebrafish model of inherited blindness. HDACi treatment results in a rescue of some gross and retina-specific morphological defects, and significant improvement in visual function. Interrogation of proteomic datasets from treated and control dye mutants identified a correlation between upregulation of BDNF-TrkB signaling and HDACi mediated rescue of vision. The functional significance of BDNF-TrkB signaling was validated by reversal of HDACi-mediated improved vision when co-treated with ANA-12, a pharmacological antagonist of the TrkB receptor. Most significantly, treatment with a BDNF mimetic, 7,8-dihydroxyflavone hydrate (7,, was sufficient to rescue visual function in dye larvae. In summary, characterisation of the mechanisms by which histone deacetylase inhibitors (HDACi) rescue visual impairment uncovered BDNF-TrkB signalling as necessary and sufficient to restore visual function in a model of inherited blindness.

Results
dye mutants display aberrant retinal morphology. The dying on edge (dye) mutant was uncovered during an F 3 N-ethyl-N-nitrosourea (ENU) mutagenesis screen in zebrafish. At 5 days post fertilization (dpf), dye larvae exhibit slightly decreased body length and eye diameter, hypopigmentation, pericardial oedema and deflated swim bladder (Fig. 1a,b). In ultrathin retinal sections, dye display increased cell death in the CMZ indicated by pyknotic nuclei (Fig. 1d,f). In comparison to wildtype siblings (Fig. 1c,e), dye photoreceptor outer segments (OS) are decreased in length, lack an orderly tiered layering and are not orientated perpendicular to the retinal pigment epithelium (RPE) (Fig. 1g,h). However, the OS ultrastructure presents with normal stacking of outer segment membranes (Fig. 1i,j). The RPE of dye is hypopigmented, and forms large intracellular inclusion bodies (Fig. 1k,l). Positional cloning localised dye to the atp6v0e1 gene and the causative mutation was identified as a 180 bp deletion ( Fig. 1m-o) that removes the stop codon, a 15 bp sequence of the 3′ UTR, and the donor splice site from the exon 3-intron 3 boundary that is required for correct splicing of the atp6v0e1 mRNA (Fig. 1o). In agreement, RT-PCR demonstrated a large reduction of atp6v0e1 mRNA expression in dye compared to wildtype larvae ( Supplementary  Fig. S1) and similar morphological defects are present in other vacuolar ATPase subunit mutants 39 . dye mutants have significantly impaired cone photoreceptor-mediated visual function. To determine if dye have reduced visual function, we conducted OKR and VMR assays (Fig. 2). The OKR measures the ability to track objects by quantifying the number of saccadian eye movements a larva exhibits in response to a rotating grated stimulus 36 . Notably, 5 dpf dye display a ~98% reduction in OKR, producing an average of 0.33 (±0.67) saccades per minute (sacc./min.) compared to an average of 21.6 (±1.6) sacc./min. in unaffected siblings +/−,+/+ (Fig. 2a). The VMR quantifies larval locomotor activity in response to sudden changes in lighting 37 . Rapidly changing the environment to illuminated (ON response) or darkened (OFF response), results in a 'startled' behavioural response wherein locomotor activity is acutely increased. 5 dpf dye (blue trace, Fig. 2b) show reduced VMR activity with significantly reduced ON and OFF peak responses and no elevated light seeking behaviour in dark conditions compared to siblings (green trace, Fig. 2b). dye exhibit a 79% reduction in in the supplementary information ( Supplementary Fig. S3a). (o) Sequencing of gel extracted PCR products and pairwise sequence alignment to the reference genome sequence (assembly GRCz11) revealed that the deletion removes the stop codon in, and the 3′ UTR encoded in exon 3 and of the GT splice donor site encoded at the exon 3-intron 3 border. the average MAX OFF VMR peak producing an average activity response of 0.055 (±0.013) ms/s compared to 0.259 (±0.049) ms/s in unaffected siblings (Fig. 2c). dye display an 85% reduction in the MAX ON VMR peak, responding with an average activity of 0.054 (±0.011) ms/s compared to 0.354 (±0.068) ms/s in unaffected siblings (Fig. 2d). dye also exhibit a defective electroretinogram (ERG), a more direct measurement of outer retinal function (Fig. 2e,f), with dye displaying reduced or absent b-waves in response to 20 millisecond flashes of increasing light intensity. In summary, visual behaviour assays and ERG measurements confirm dye have significantly impaired visual function. Notably, initiation of visual behaviour and retinal function in zebrafish larvae is predominantly mediated by cone and not rod photoreceptors 35 . HDACi treatment restores retinal morphology and visual function in dye. As previous studies report histone deacetylase inhibitors (HDACi) to preserve retinal morphology in mouse models of rod photoreceptor degeneration 26 , we investigated the effects of a panel of HDACi on retinal morphology and cone-related visual function in dye. Initial dose-range finding (0.25-4 µM) assays indicated that 1 µM trichostatin A (TSA) was the maximum tolerated dose for zebrafish larvae submersed in the drug diluted in embryo medium. Strikingly, treatment with TSA acutely reversed many morphological defects observed in dye (Fig. 3a-i). Homozygous dye larvae treated with 1 µM TSA from 3-5 dpf present with increased tissue pigmentation, eye size and body length; inflated swim bladders; and reduced pericardial oedema (Fig. 3a-c). The genotype of dye larvae was validated by Statistical analysis was carried out using a student's t-test with unequal variances, ***p < 0.001, **p < 0.01. PCR showing that homozygous dye mutant DNA, but not sibling DNA, was negative for a 528 bp PCR amplicon encompassing exon 3 of the atp6v0e1 gene. By contrast, a β-actin control amplicon was present in all samples (Fig. 3d). Treatment with TSA also increases acetylation of histone 3 in dye (Fig. 3e). In the retina, the number of prominent photoreceptor outer segments is markedly increased following HDACi treatment (Fig. 3f,g). Cell death in the CMZ and peripheral retina is significantly reduced as evidenced by a reduction in the number of pyknotic nuclei (Red boxes, Fig. 3h,i, quantified in l) present in serial transverse sections from the central retina. However, RPE hypopigmentation and the presence of inclusion bodies is not rescued (Fig. 3j,k).
Proteomic profiling, the HDACi molecular mechanism of rescue. To mechanistically understand how HDACi mediate rescue of visual function in dye, protein expression in the eye was profiled by mass spectrometry. 3 dpf dye larvae were treated with 1 µM TSA or vehicle control (0.1% DMSO) until 4 or 5 dpf. At 4 dpf, 1920 and 1905 proteins were identified from eyes of vehicle control or 1 µM TSA-treated dye, respectively ( Fig. 4a, Supplementary data S1). Of the 1690 overlapping proteins within these datasets, 147 show significant (Fisher exact test p < 0.05) differential expression; 32 were decreased and 115 increased (Fig. 4b, Table 1). At the 5 dpf end point, 1923 and 1939 total proteins were identified from eyes of vehicle control or 1 µM TSA treated groups, respectively ( Fig. 4a, Supplementary data S1). Of the 1511 overlapping proteins identified in both test groups, 131 proteins are significantly (Fisher exact test p < 0.05) differentially expressed; 51 decreased and 80 increased (Fig. 4c, Table 2). Cluster analysis of the differentially expressed proteins highlighted the concordance of the triplicate experiments for vehicle-or TSA-treated samples and identified sub-groups of ocular proteins exhibiting equivalent alterations in protein expression post HDACi treatment (Fig. 4b). To determine biological processes contributing to HDACi mediated rescue, enriched gene ontology terms were identified in DAVID-DB (Supplementary data S2) and PANTHER-DB (Fig. 4d,e) databases. For the significantly differentially expressed proteins at both time points, processes involved in initiation of transcription (nucleotide binding), protein translation (aminoacyl-tRNA ligase activity), visual perception and metabolism are over-represented in the dataset.
BDNF-TrkB signaling is central to HDACi mediated rescue in the dye. In order to delineate the signaling pathways mediating HDACi rescue of dye, the proteomic datasets were analysed using Ingenuity Pathway Analysis (IPA) software. Upstream regulators of the effects observed in the dataset are inferred from changes in expression in downstream proteins. Additionally, the activation state of these inferred transcriptional regulators (ITRs) was predicted based on the abundance of downstream effector proteins. Brain derived neurotrophic factor (number of replicates). (k) Summary of VMR activities, 1 µM TSA and 6 µM Scriptaid significantly improve the MAX ON activity, N = 12, n = 3. Statistical analyses were performed using a Kruskal-Wallis one-way analysis of variance and post-hoc Dunn's multiple comparison test, comparing each group to 0.1% DMSO treated dye mutants, error bars represent average SEM, ***p < 0.001, **p < 0.01, *p < 0.05. Full length agarose gels and western blots are included in the supplementary information ( Supplementary Fig. S3b-d). (BDNF), N-myc proto-oncogene (MYCN), myc proto-oncogene (MYC), tumor protein 53 (TP53) and RPTOR independent companion of MTOR, complex 2 (RICTOR) were identified as ITRs, as were several HDAC isoforms (Fig. 5a). In HDACi rescued dye, MYC, MYCN and TP53 signaling was predicted to be inhibited; in contrast, RICTOR and BDNF signaling was predicted to be activated (Fig. 5a). Consistent with this, BDNF-NTRK, PI3K and mTOR signaling were all differentially activated by our HDACi treatment (Fig. 5b). HDAC isoforms, HDAC1 and HDAC2 were predicted to be activated at the 4 dpf treatment endpoint, though this activation was reduced somewhat at the 5 dpf treatment endpoint. Conversely, HDAC3 is inhibited at both timepoints. As bdnf transcription is rapidly induced by HDACi treatment in cultured rat cortical neurons 40 and is upstream of PI3K/ AKT/RICTOR signaling we assessed the contribution of BDNF-TrkB signaling to in the rescue of vision in dye.  Table 2. Significantly differentially expressed eye proteins in TSA treated dye at 5 dpf. To validate the requirement of BDNF-TrkB signaling for the rescue of visual function, co-treatments were performed with ANA-12, a known pharmacological antagonist of the TrkB receptor. Visual response rescue following co-treatment with 1 µM TSA plus 100-500 nM ANA-12 was significantly reduced but not to untreated mutant levels, in a dose-dependent manner for both the OKR (69% maximum reduction) and MAX-ON response of the VMR (85% maximum reduction) compared to 1 µM TSA alone (Fig. 5c,d). ANA-12 co-treatment partially reversed the TSA-mediated rescue of cell death in the CMZ (Fig. 6i,j). To further validate the requirement of BDNF-TrkB, changes in protein expression in treated eye homogenates were examined (Fig. 5e,f). In dye larvae treated only with 1 µM TSA, a significant 31% increase in Bdnf expression is observed compared to vehicle controls (Fig. 5e,f). In contrast, Bdnf expression was not significantly increased in 1 µM TSA + 500 nM ANA-12 co-treated dye larvae compared to vehicle controls (Fig. 5e,f). TrkB receptor expression remained consistent across treatments (Fig. 5e,f). As the PI3K/AKT pathway is activated downstream of BDNF-TrkB signaling, the phosphorylation status of Akt (Thr308) was examined. Compared to vehicle controls, 1 µM TSA treatment significantly increased phosphorylation of Akt by ~57%, which was blocked by 1 µM TSA + 500 nM ANA-12 co-treatment (Fig. 5e,f). Expression or phosphorylation of these proteins was not significantly altered in equivalent treatments of wildtype or heterozygous sibling larvae (Fig. 5e,f). A STRING-DB protein-protein interaction network was generated by combining the predicted or confirmed protein activation status or abundance in HDACi treated dye (Fig. 5g). Changes in expression/activation of proteins in this network are in agreement with activation of BDNF-TrkB signaling wherein HDACi treatment increases expression of BDNF, subsequently activating Ntrk2/TrkB then activation of Akt and inhibition of MycN, P53, Myc and HDAC3.
A BDNF mimetic is sufficient to restore visual function in dye. Our data suggested HDACi mediated rescue was largely due to modulation of TrkB signaling via increased endogenous production of BDNF. Thus, dye were treated with a BDNF mimetic, 7,8-dihydroxyflavone hydrate (7,. Treatment with 10 µM 7,8-DHF did not rescue gross morphological defects present in dye to a significant degree (Fig. 6a,b), conversely, retinal morphology was rescued (Fig. 6c,d,f,g) to a similar degree to HDACi treated dye larvae, including an 89% reduction in the number of dying cells present in the CMZ (red boxes, Fig. 6f,g, bar chart, i), and a 59% reduction in the size of the CMZ (j). Co-treatment with 500 nM ANA-12 partially blocked rescue of CMZ cell death (Fig. 6i, 72% reduction) and reduction in CMZ size (Fig. 6j, 41% reduction). Significantly, treatment with 10 µM 7,8-DHF rescued visual function as assessed by OKR (Fig. 6k) and VMR (Fig. 6l), in a dose-dependent manner. The degree of rescue with 7,8-DHF (57.58 fold increase in OKR, 3.89 and 3.33 fold increase in VMR MAX ON and MAX OFF activities, respectively) was greater than that observed by the most effective HDACi. In addition, co-treatment with 10 µM 7,8-DHF and 500 nM ANA-12 abrogated rescue of visual behaviour (71% reduction in OKR, 48% reduction in MAX OFF and 58% reduction in MAX ON VMR) (Fig. 6k,l). A comprehensive analysis of the visual behaviour responses, including dose-dependent improvements in OKR following HDACi or BDNF mimetic treatment is presented in Table 3.
In summary, pathway analysis of dye proteomic datasets suggested BDNF-TrkB signaling as a mediator of HDACi rescue of visual function. The requirement and sufficiency of BDNF-TrkB signaling for rescue of visual function was confirmed experimentally by pharmacological intervention, biochemical analysis and visual behaviour assays.

Discussion
Currently, there are no approved pharmacological interventions to stabilise or improve visual function in patients with inherited blindness. Advantages of drug treatment include the ability to fine-tune effective and safe concentrations for each patient, to co-administer drug combinations, to incorporate sustained drug delivery devices and to stop treatment if ineffective or unsafe. However, identifying drugs that retain or restore visual function is challenging. Here, we demonstrate in a vertebrate model of inherited blindness that HDAC inhibitors and BDNF mimetics are sufficient to significantly improve cone-photoreceptor mediated visual function via a BDNF mediated signalling dependent mechanism of action.
Translating drug treatments for inherited blindness from research models to the clinic has been remarkably unsuccessful. In part, this reflects poor correlations between biochemical or morphological readouts of neuroprotection in research models and the more desirable clinical end-point of improved functional vision 41,42 Clinical setbacks can arise from trials on patients harbouring heterogenous genetic mutations that mask treatment efficacy, or poor ocular pharmacokinetics, resulting in optimal drug concentration of inefficient duration 43,44 .
To enable efficient identification of drugs overcoming visual impairment, we applied the zebrafish dye model of inherited blindness coupled with established visual behaviour assays. Although dye is an imperfect model of retinal degeneration, it was chosen because i) the ~25% homozygous affected dye larvae can be reproducibly and easily phenotyped at 3 dpf; ii) dye larvae at 5 dpf exhibit pronounced visual behaviour and retinal morphology ANA-12 co-treatment. Akt (Thr308) phosphorylation was increased in TSA treated dye, while levels remained unchanged in ANA-12 co-treated larvae, N = 25 (number of eyes per replicate), n = 3 (number of replicates). Images represent a typical blot, bar chart represents signal intensity measured by densitometry, Trkb and Bdnf signals normalised to Actb expression and pAKT (Thr308) signal normalised to total Akt. Statistical analysis was performed using a two-tailed unpaired t-test in comparison to dye control. Error bars indicate S.E.M, ****p < 0.0001 ***p < 0.001, **p < 0.01. Full length western blots are included in supplementary information ( Supplementary Fig. S3e-i). (g) Protein interaction map (generated by STRING-DB) of ITRs differentially expressed upon TSA treatment. Akt1, Akt2 and TrkB proteins were added to the network based on ITR pathway analysis and western blot data. Activated nodes are highlighted with red, downregulated nodes with green, grey highlighted nodes are predicted regulators but in which the direction is unknown.
ScienTific REPORTS | 7: 11320 | DOI:10.1038/s41598-017-11513-5 Visual function rescue was abrogated by ANA-12 co-treatment. Visual behaviour data was analysed by a Kruskal-Wallis one-way analysis of deficits, and iii) drug treatment of dye from 3-5 dpf followed by optokinetic or visualmotor response assays enable identification of small molecule drugs that prevent loss of visual function. The mutation in dye is caused by a 180 bp deletion in atp6v0e1, a vacuolar ATPase v0e1 subunit 45 , which affects splicing of atp6v0e1 transcripts into mature mRNA. Vacuolar ATPase complexes acidify organelles via transportation of H + ions across vesicle membranes 31 . In the retina, disruption of this complex leads to phagocytosis defects and accumulation of undigested photoreceptor outer segments in the RPE. Defective outer segment phagocytosis is a hallmark of retinal diseases including RP and Usher syndrome type IB [46][47][48] . N-retinylidene-N-retinylethanolamine (A2E), a major component of lipofuscin, is a potent inhibitor of v-ATPase activity and accumulation of lipofuscin is associated with age-related macular degeneration (AMD) 49 . In humans, mutations in V-ATPase a3 subunit cause osteopetrosis, wherein patients display visual impairment amongst other pathologies 50 . In the zebrafish eye, atp6v0d1 expression is restricted to the outer nuclear layer (ONL) and inner nuclear layer (INL) of the retina at 2 dpf 39 , and to the RPE from 3 dpf. atp6v0e1 shows a similar expression pattern from 3 dpf, where expression is mainly present in the RPE (Supplementary Fig. S1), Mutants of several other vacuolar ATPase subunits have similar retinal morphological defects to dye including RPE hypopigmentation, ciliary marginal zone cell death and abnormal photoreceptor morphology 39 . Defects in photoreceptor morphology are likely explained by aggregate accumulation in the RPE/photoreceptors, degradation of which is perturbed by an inability to acidify lysosomes. CMZ cell death affects a subpopulation of progenitor cells in the teleost retina 51, 52 though it is currently unknown how the dye mutation impacts survival of these cells. Significantly, 5 dpf dye larvae recapitulate clinical features of human blindness including significantly attenuated vision. In dye electroretinograms, significantly diminished b-wave amplitudes are observed, indicating defective photoreceptor-induced depolarisation of bipolar cells. This is consistent with ERG recordings from alternative zebrafish models of blindness including mutations in other vacuolar ATPase subunits 39 , the noir mutant arising from defective pyruvate dehydrogenase metabolism 53 and mutations of the SNARE complex, formation of which required for synaptic vesicle fusion 54 . Importantly, dye larvae exhibit impaired visual behaviour in optokinetic and visualmotor responses. These technically simpler and higher throughput assays are compatible with drug screening. In summary, the dye zebrafish model of inherited blindness is appropriate to incorporate into efficient phenotype-based screens to identify drugs that preserve or restore visual function.
In dye, HDACi treatment is sufficient to restore visual function. HDACi were chosen due to the contradictory reports in the literature regarding their efficacy and safety in clinical trials of patients with inherited sight loss and concerns over patients self-medicating with unapproved HDACi [22][23][24][25] . Significantly, HDACi rescues some features of retinal morphology, but robustly rescues visual responses. Previous studies report HDACi to preserve rod photoreceptor morphology ex vivo but this report demonstrates an improvement of visual function. In agreement, HDACi increased expression of phototransduction proteins including phosphodiesterase, opsin, transducin and peripherin isoforms. Despite the efficacy of HDACi in restoring visual function, translation to clinical intervention for patients with inherited sight loss is justifiably occluded due to ongoing safety and specificity concerns 22,23 . Thus, we sought to identify alternative key signaling nodes that are sufficient to recapitulate the HDACi rescue.
Bioinformatic analysis unmasked Bdnf, Myc, MycN, Tp53 and RICTOR signaling as transcriptional regulators predicted to mediate the rescue of visual function by HDACi. Of the identified signalling mechanisms, the most notable was the correlation between visual rescue and BDNF signaling. BDNF is a neurotrophic growth factor whose mature form binds the tropomyosin related kinase B (TrkB) receptor. BDNF-TrkB signaling has diverse physiological functions, including regulation of photoreceptor development and maintenance 10 . For instance, TrkB null mice develop shortened photoreceptor outer segments 55 . In the adult rat retina, BDNF and TrkB co-localise in green-red-sensitive cone photoreceptor outer segments 56 . In developing zebrafish, Bdnf localisation is observed in the outer nuclear layer (ONL), outer plexiform layer (OPL) and inner plexiform layer (IPL), while TrkB localisation is restricted to the IPL 57 , suggesting that ganglion or Müller cells receive the BDNF signal in the retina. We speculate that cell signalling pathways activated by agonism of the TrkB receptor on ganglion or glial cells culminates in release of trophic factors that support photoreceptors, a mechanism reported in models of light-induced photoreceptor degeneration 58 .
BDNF transcription is known to be induced by HDACi treatment in cultured rat cortical neurons mediated by the CREB transcription factor 40 , and in combination with ciliary neurotrophic factor (CNTF), BDNF reduces cell death in rd1 mouse retinal explants 15 . Furthermore, small molecule BDNF mimetics are neuroprotective in mammalian models of traumatic brain injury 59,60 . Here, pharmacological approaches demonstrated the necessity and sufficiency of BDNF-TrkB signaling for visual rescue in dye larvae. ANA-12, a TrkB receptor antagonist blocked HDACi mediated rescue by 69% and 85% in OKR and VMR assays respectively. Critically, 7,8-dihydroxyflavone hydrate, a TrkB receptor agonist or BDNF mimetic is sufficient to restore visual function and reduce CMZ cell death in the dye model of inherited sight loss. Co-treatment with 7,8-DHF and ANA-12 blocked rescue of visual function (71% and ~54% decrease in OKR and VMR respectively, compared to 10 µM 7,8-DHF only treated larvae), with partial rescue of dying cells in the CMZ. Treatment appears to result in a decrease in the size of the CMZ, which is partially blocked by ANA-12 co-treatment. A summary of the signaling networks involved in HDACi and BDNF mimetic mediated rescue is presented in Supplementary Fig. S2. Our findings support the conclusion that BDNF-mediated retinal neuroprotection is conserved in zebrafish and mammalian models, highlighting the applicability of zebrafish as a model for translational neurodegeneration research. variance and post-hoc Dunn's multiple comparison test, comparing each group to 7,8-DHF treated dye mutants, error bars represent average SEM. Comparison of pyknotic nuclei counts and comparison of CMZ area was performed using an ordinary one-way ANOVA, ***p < 0.001, **p < 0.01, *p < 0.05. BDNF mimetics offer an exciting alternative approach to treating sight loss. They offer a more selective approach than HDACi which have diverse target profiles and global effects on gene expression. BDNF was identified as a photoreceptor protectant over 20 years ago 12 but translational studies delivering BDNF in biological forms were hampered by pharmacokinetic issues [61][62][63] . Small molecule BDNF mimetics can circumvent these issues as they: i) protect vision in vivo, ii) cross the blood-tissue barriers iii) can be delivered directly to the eye and iv) are more stable than BDNF. BDNF mimetics are well tolerated in mammalian models, demonstrate efficacy in rodent models of Alzheimer's, Parkinson's and Huntington's disease and no adverse outcomes are reported for humans consuming 7,8 dihydroxyflavone as a nootropic supplement.

Materials and Methods
Ethics Statement. All experiments carried out on animals were performed according to ethical approval granted by the UCD Animal Research Ethics Committee. Zebrafish in the first 5 days of life are not capable of feeding independently. The UCD Policy on the use of Animals for Research & Teaching states that "Protected animals are those which have the capacity to experience pain, suffering, distress or lasting harm as a result of procedures which may be carried out in the course of research or teaching. " … "Larval stages of fish are judged to be capable of experiencing pain, suffering or distress once they are capable of feeding independently". As zebrafish larvae under 5 days post fertilization were used in all experiments, no protected animals were used and experiments were not subject to full ethical review.

N-nitroso-N-ethylurea (ENU) mutagenesis screen.
Male zebrafish of the AB genetic background were treated for 1 hour with 3 mM ENU (Sigma) diluted in 10 mM sodium phosphate buffer, pH 6.6. This treatment was repeated 4 times at 7-14 day intervals. Treated males were mated and embryos were grown up to generate F 1 founders outcrossed to wildtype fish generating F 2 families. F 3 offspring were screened for recessive mutations using the optokinetic response (OKR) assay. Carriers of mutations were crossed to wildtype Tubingen fish to generate hybrid carriers. These carriers were incrossed and their mutant and normal offspring were collected for phenotyping and genotyping.
Identification of dye mutation. Genomic DNA was isolated from larval zebrafish using a DNeasy Blood and tissue kit (Qiagen) according to manufacturer's instructions. DNA concentration was determined by spectrophotometry using a Nano-drop 2000 (Thermo Scientific). PCR utilised the Crimson-Taq DNA polymerase system (New England Biolabs). Extracted DNA was used for bulk segregation analysis using Z-markers as described in Sapetto-Rebow, et al. 64 . Within the genomic lesion, atp6v0e1 was selected for further interrogation. Primers used for genotyping include: atp6v0e1 A forward: AGAACCACTGCCAGAACC atp6v0e1 A reverse: Drug treatments. At 3 dpf dye larvae were separated from unaffected sibling larvae based on phenotype. Visual behaviour assays. At 5 dpf, HDACi-treated larvae were removed from drug solutions and transferred to embryo medium for behavioural analyses. In the optokinetic response (OKR) assay, individual larvae were transferred to a petri dish containing 9% methylcellulose for immobilisation, and placed inside a circular grated pattern comprising 18° black and white stripes. The pattern was rotated at 18 rpm for 30 seconds clockwise and 30 seconds anti-clockwise. The number of saccadian eye movements per minute was recorded manually. For the visual motor response (VMR) assay, treated larvae were transferred in 600 µL embryo medium to individual wells of a 96 well clear polystyrene plate (Whatman). The plate was placed in the Zebrabox ® recording chamber (Viewpoint Life Sciences) and locomotor activity quantified in response to changing light conditions by a motion detecting infrared camera. Detection parameters and analysis was performed as described previously by Deeti, et al. 66 . All statistical analyses were performed in Graphpad Prism V6. A Kruskal-Wallis one-way analysis of variance with post-hoc Dunn's multiple comparison test was performed for visual function data, comparing dye control larvae to all other groups unless otherwise stated 67 .
Electroretinography. Larvae were dark adapted for 30 minutes and paralysed with 0.5 mg/mL mivacurium chloride (Mivacron), the dark-adapted status was maintained throughout the handling and measurement processes. The reference electrode was placed in E3 medium, the recording electrode was filled with 0.9% saline solution and positioned on the center of the cornea using a micromanipulator and amplified with a P55 pre-amplifier (Grass Instruments). A 300 W halogen light source was used for light stimulation. The maximum light intensity stimulus was 2.8 × 10 3 µW/cm 2 with a 20 ms flash duration at all intensities (controlled by a S48 stimulator (Grass Instruments). Three optical density filters produced flash intensities at −3.0, −2.0, −1.0 and −0 log of the maximum. Recordings were analysed as previously described 68 . Raw data from dye mutants and siblings was compared using a 2-sample t-test with unequal variances.
Histological analysis. Samples were fixed overnight in 4% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M Sorenson phosphate buffer at pH 7.3. For bright-field imaging, samples were washed in PBS and imaged using an Olympus SZX10 microscope. For light microscopy, samples were post-fixed in 1% osmium tetraoxide and dehydrated in gradient ascending series of ethanol concentrations prior to Epon 812 resin embedding overnight. 1 µm sections were prepared using a Leica EM UC6 microtome and glass knife, mounted on glass slides and stained with toluidine blue. Prepared sections were imaged by a Leica DMLB bright field illumination microscope and Leica DFC 480 camera. The number of dying cells in the ciliary marginal zone (CMZ) was quantified by recording pyknotic nuclei present in sections 5 µm apart surrounding the optic nerve. The area of the CMZ was measured in central retinal sections using the polygonal selection tool in imagej, morphology data was analysed using an ordinary one-way ANOVA. The area of the CMZ was determined according to the criteria in Raymond, et al. 69 . For transmission electron microscopy (TEM), 0.1 µm sections were prepared using a Leica EM UC6 microtome and diamond knife, transferred to a support grid, contrasted with uranyl acetate and lead citrate, and analysed on a FEI-Tecnai 12 BioTwin transmission electron microscope (FEI Electron Optics).
Protein extraction and western blot. 4  Densitometry was performed in ImageJ and relative optical-density was normalised to β-actin or total Akt expression. Western blot data was analysed in Graphpad Prism V6 using a two-tailed paired t-test.
Proteomic analysis. Protein from 50 homogenised larval eyes was isolated with the addition of trichloroacetic acid (TCA, 20%), reduced using 200 mM dithiothreitol, alkylated with 200 mM iodoacetamide and digested overnight with trypsin (Sigma). Peptides were desalted with C18 STAGE tips 70 and resuspended in 0.1% TFA. For mass spectrometry peptide fractions were analyzed on a quadrupole Orbitrap (Q-Exactive, Thermo Scientific) mass spectrometer equipped with a reversed-phase NanoLC UltiMate 3000 HPLC system (Thermo Scientific). Peptide samples were loaded onto C18 reversed phase columns (5 cm length, 75 µm inner diameter). Raw data from the Orbitrap Q-Exactive was processed using MaxQuant version 1.  74 . Protein identifications were filtered to eliminate the identifications from the reverse database and common contaminants. Data was log transformed and t-test comparison of fractions carried out. For visualization using heat maps, missing values were imputed with values from a normal distribution and the dataset was normalized by z-score 75 . Gene ontology terms were identified and visualised by submitting identified gene lists to DAVID and PANTHER databases.
Pathway analysis. Ingenuity Pathway analysis (IPA) was used to identify inferred transcriptional regulators (ITRs) of differentially expressed genes in the datasets. Statistical algorithms match each gene symbol, fold change value and ANOVA p-value to corresponding objects in the curated IPA knowledge database, which is then used to identify shared regulators of genes in the dataset and are assigned a score based on relevance to input genes. This analysis also infers activation status of upstream regulators based on abundance of downstream transcriptional targets. Top ITRs were included in bar-charts with the predicted activation status score. ITRs and their predicted activation status and western blot data was collated and used to generate an interaction network using STRING database.