Involvement of P2X7 receptor in neuronal degeneration triggered by traumatic injury

Axonal injury is a common feature of central nervous system insults that culminates with the death of the affected neurons, and an irreversible loss of function. Inflammation is an important component of the neurodegenerative process, where the microglia plays an important role by releasing proinflammatory factors as well as clearing the death neurons by phagocytosis. Here we have identified the purinergic signaling through the P2X7 receptor as an important component for the neuronal death in a model of optic nerve axotomy. We have found that in P2X7 receptor deficient mice there is a delayed loss of retinal ganglion cells and a decrease of phagocytic microglia at early times points after axotomy. In contralateral to the axotomy retinas, P2X7 receptor controlled the numbers of phagocytic microglia, suggesting that extracellular ATP could act as a danger signal activating the P2X7 receptor in mediating the loss of neurons in contralateral retinas. Finally, we show that intravitreal administration of the selective P2X7 receptor antagonist A438079 also delays axotomy-induced retinal ganglion cell death in retinas from wild type mice. Thus, our work demonstrates that P2X7 receptor signaling is involved in neuronal cell death after axonal injury, being P2X7 receptor antagonism a potential therapeutic strategy.

Scientific RepoRts | 6:38499 | DOI: 10.1038/srep38499 Extracellular nucleotides released upon traumatic cell injury have an important role in the initiation and maintenance of the inflammatory response 15 . Extracellular nucleotides serve as an initial find-me signal for immune cell migration to the injury area and if the damage is controlled the nucleotide find-me signal promotes phagocytic clearance 16 . An increase of extracellular nucleotide concentration due to irreparable tissue damage and accumulation of other damage-associated molecular patterns leads to the activation of potent pro-inflammatory pathways by activating purinergic P2X7 receptors in innate immune cells 15,17 . P2X7 receptor is a cationic ion channel gated by mM concentrations of extracellular ATP, these high concentrations of ATP have been found in vivo associated to areas of tissue injury and inflammation 17,18 , and its prolonged activation leads to calcium flux, formation of large membrane pores and exposure of phosphatidylserine, ultimately resulting in cell death 19 . However, P2X7 receptor activation in the immune system can also couple to responses distinct from cell death including generation of reactive oxygen species, production of prostaglandins, release of proteases, antigen-driven T-lymphocyte proliferation, and intracellular pathogen killing 15,[20][21][22][23][24] . P2X7 receptor is also the most potent plasma membrane receptor responsible for NLRP3 inflammasome formation and activation of caspase-1, inducing the release of pro-inflammatory cytokines of the interleukin (IL)-1 family, such as IL-1β and IL-18 21,25 . IL-1β is one of the main mediators of damaging during sterile inflammation and is implicated in the aetiology of many major diseases, including damage to the CNS 26,27 .Therefore, P2X7 receptor is a promising therapeutic target in the management of tissue damage, inflammation and pain, as witnessed by the large number of selective P2X7 receptor antagonists developed by several drug companies and currently under clinical trials [28][29][30] . In particular, A438079 is a selective competitive antagonist for the human and rodent P2X7 receptor with good bioavailability and CNS penetration widely used in preclinical animal models of disease 31,32 . A438079 reduces inflammation in models of colitis, ischemic acute kidney injury, contact dermatitis, and liver injury and fibrosis [33][34][35][36][37] .
P2X7 receptor is expressed in RGCs [38][39][40] and several works suggest that it is involved in their death triggered by different insults. Thus, intravitreally injected P2X7 receptor agonists kill RGCs 41,42 while its antagonism protects them in models of ocular hypertension 43,44 or hypoxia 45 . Furthermore, in rats its expression increases after optic nerve axotomy 46 and acute ocular hypertension 44 .
Therefore, we purpose to study the effect of P2X7 receptor on the survival of RGCs and the appearance of phagocytic microglial cells (PMCs) in the injured and contralateral uninjured retinas after unilateral optic nerve crush (ONC) by using the P2rx7 −/− mice and by pharmacologically targeting P2X7 receptor in wild type mice.

Results
P2X7 receptor expression is regulated by axotomy. We first found that the P2X7 receptor was expressed in the inner plexiform layer and retinal ganglion cell layer (Fig. 1A). In intact retinas, there is a weak expression of P2X7 receptor in RGCs, in accordance with previous reports 40,44 . Interestingly, 5 days after ONC, there is an increase of P2X7 receptor immunostaining in the nerve fibre layer (RGC axons or astrocytes) and in the RGC somas (Fig. 1A). This signal was reduced in retinas from P2rx7 −/− animals (Fig. 1A). We also found a staining of blood vessels with the anti-P2X7 receptor antibody, but this signal was also observed in retinas from P2rx7 −/− mice, suggesting that is a non-specific signal (Fig. 1A).To confirm P2X7 receptor expression in retinas, we next performed western blotting detection in extracts from intact, injured and contralateral to the injury retinas. We found a ~75 kDa band corresponding to the size of P2X7 receptor that was also present in lysates from mouse bone marrow derived macrophages, a cell type expressing high levels of P2X7 receptor (Fig. 1B). P2X7 receptor expression in fellow retinas did not change compared to intact ones. However in the injured retinas, the expression of P2X7 receptor has a slightly increase at early time points after ONC and decreases with the time post-lesion being almost negligible at 9 days after ONC. This suggests that upon axonal injury, surviving RGCs over-express P2X7 receptor, but in the whole retina extracts its net expression decreases due to the loss of RGCs.
Deficiency of P2X7 receptor results in a delayed loss of RGCs after axotomy. We have analyzed in parallel the general RGC population, that transmits image-forming information to the brain and expresses the transcription factor Brn3a (Brn3a + RGCs) 6,47-49 and the subpopulation of intrinsically photosensitive RGCs, that convey non-image forming information and express the photopigment melanopsin 50,51 but do not express Brn3a (m + RGCs) 48,49 . The analysis of both types of RGCs is of interest because m + RGCs respond differently to injury and neuroprotection than the general RGC population [52][53][54] . In intact retinas, we found that P2rx7 −/− mice presented similar total number of the general RGC population identified by tracing or by Brn3a immunodetection ( Fig. 2A and Table 1), as well as similar total number of m + RGCs when compared to wild type mice (Table 1). These data together with the topographic maps of intact retinas of both strains (Figs 3A and 4A,B), showed that the lack of the P2X7 receptor does not have an effect on the retrograde axonal transport of RGCs (tracing) nor on the number and distribution of RGCs.
In both strains the loss of Brn3a + RGCs occurred homogenously across the retina (Figs 2B and 3A), was exponential (RGC loss vs. time: R 2 = 0.95 in wild type, R 2 = 0.96 in P2rx7 −/− . Table 2 and Fig. 3B) and significant 3 days after the lesion in the wild type strain, being delayed until day 5 in the P2X7 deficient mice (Fig. 3). Thus, in P2rx7 −/− animals the survival of RGCs is significantly higher than in wild type at 3 and 5 days after the lesion, equating at day 9. In fact, regression analysis showed that the daily loss of RGCs accounts for ~200 more RGCs in the wild type than in the deficient strain (slope of the linear regression, Fig. 3B). Topographically, we found more RGC survival throughout the retina in P2rx7 −/− mice at 3 and 5 day after injury, and the same than wild type by day 9 post-ONC (Fig. 3).
Deficiency of P2X7 receptor similarly affected to the loss of m + RGCs, meanwhile there was a significant lost at 3 days in wild type animals, no significant variation was observed in P2rx7 −/− retinas ( Table 2 and Fig. 4B,C). However, at 5 and 9 days post-ONC the percentage of m + RGC survival was the same in both strains. Topographically, their loss was diffuse in both strains, but stronger in the dorsal retina (Fig. 4B).
Scientific RepoRts | 6:38499 | DOI: 10.1038/srep38499 Comparing the general RGC population (Brn3a + ) and m + RGCs, we found that the percentage of survival was higher for the m + RGCs (Table 2), in agreement with previous reports showing in rats that m + RGCs are more resilient to axotomy 52 . P2X7 receptor antagonism in wild type animals also delays RGC loss after axotomy. In light of the above results which indicate that the P2X7 receptor might be a potential therapeutic target for RGC loss, the next experiment was designed to study whether its antagonism was effective delaying RGC loss in the retinas of wild type mice after ONC. First, we assessed the toxicity on RGCs of an intravitreal injection of 300 ng of the selective P2X7 receptor antagonist A438079. Treated and intact retinas were analyzed 9 days after the administration, to match the longest experimental time point. The number of Brn3a + RGCs and of m + RGCs in these retinas (37,050 ± 1178 and 1197 ± 54, respectively) did not differ to that found in intact retinas (see Table 1). Hence, the antagonist appears not to be toxic for RGCs. Therefore, we administered the same dose of antagonist to ONC-injured retinas and analyzed them at 3, 5 or 9 days after the lesion. The results show that RGC loss is significantly delayed in A438079-treated retinas (Fig. 5), similarly to the delay observed in the P2rx7 −/− strain. In agreement with the quantitative data, and as observed in the P2X7 receptor deficient mice, a higher RGC survival is observed across the retina in the antagonist-treated animals (not shown).
Phagocytic microglial cell appearance after neuronal damage. We analyzed the appearance of phagocytic microglial cells (PMCs) measured as microglial cells that become transcellularly labelled with the tracer (OHSt) when they phagocytose a dead and traced-RGC 4,55 (Table 3 and Fig. 6).

Figure 1. P2X7 receptor expression in the retina is regulated by optic nerve damage. (A)
Retinal crosssections stained for P2X7 receptor (green), Brn3a (red, detecting RGC nuclei), and nuclei (DAPI, blue), from intact or 5 days after injured (ONC) from wild type or P2rx7 −/− mice. The images show that the P2X7 receptor (green signal) in intact wild type retinas is found in the inner plexiform layer (asterisk), blood vessels (arrowheads) and weakly in the somas of RGCs (arrows, identified with Brn3a, red signal). Upon optic nerve crush, P2X7 receptor signal is stronger in the ganglion cell layer, mainly in the axons and somas of RGCs (arrows). In the P2rx7 −/− mice, some P2X7 receptor staining is observed in the inner plexiform layer and in blood vessels (arrowhead), but none in the GCL from intact or injured retinas. (B) Western blotting for P2X7 receptor or β -actin from cellular protein extract from mouse bone marrow derived macrophages (BMDM) or from retina pools (n = 3 retinas/pool) of intact retinas, injured retinas (ONC) or contralateral fellow retinas (F) at different days after ONC. Asterisk denotes unspecific protein signal found in retinas. The right panel shows densitometry quantification of two Western blots for P2X7 receptor signal normalized to β -actin and to the expression on intact retinas (dashed line). d: days. Full-length blots are presented in Supplementary Figure 1.
In both strains PMCs were found across the injured retina from day 3 post-lesion, and they were more abundant in the central-medial region, in the areas of higher RGC density and therefore higher RGC death (Fig. 6A,B).
As reported previously 4 , the increase of PMCs is exponential (Fig. 6C) and inversely proportional to the loss of RGCs (linear regression RGC loss vs. PMC appearance: R 2 = 0.95 in wild type and R 2 = 0.99 in the P2X7 deficient mice, regression graph not shown), i.e. their number increases as the loss of RGCs progresses (Table 3, compare left graphs in Fig. 3B -RGCs-and 5C-PMCs-). However, in P2rx7 −/− mice, there were significantly fewer PMCs 3 days after ONC when compared to injured wild type animals (Table 3), this could be a consequence to the slower RGC death found in P2X7 receptor deficient mice.
P2X7 receptor deficiency affects the contralateral response. In a previous work, we reported that unilateral axotomy in mice caused a slight decrease of RGCs in the contralateral retina, and the appearance of PMCs 4 . Thus, we next studied RGCs and PMCs in the contralateral uninjured retinas of both strains. In wild type animals, there was a small but significant loss of RGCs after 5 days of damage that was maintained at 9 days     (Table 4). In P2rx7 −/− mice, we found a delay on the loss of RGC and a significant decrease, similar to the one found in wild type animals at day 5, was observed at 9 days post-ONC (Table 4).
With respect to PMCs, in both strains and at all time points, their number in the contralateral retinas was significantly higher than in traced but otherwise intact retinas, and we found them across the retina although     Table 4). There were, however, differences among both strains. While in wild type animals the number of PMCs increased with time, in the P2X7 deficient mice not only there were significantly fewer PMCs at 3 days than in wild type, but also their number stabilized from 5 to 9 days. (Fig. 6C and Table 4).

Discussion
In this work we show that the lack of P2X7 receptor delays, in the injured and contralateral retina, the loss of neurons and the appearance of phagocytic microglial cells after unilateral axonal trauma. The neuroprotective effect of the specific P2X7 receptor antagonist A438079 on retinal neurons has been reported after an excitotoxic insult 56 . Here we show for the first time that also protects CNS neurons from a traumatic axonal injury. Purinergic receptors have an important role controlling neuronal information processing and regulation of retinal tissue homeostasis (reviewed in ref. 57). In particular, expression of the ionotropic P2X7 receptor has been found in most cell types of the retina, including neurons such as the RGCs, glia cells and vascular cells [57][58][59][60][61] . However, here we found that adult P2rx7 −/− mice have a normal number of RGCs and m + RGCs, as well as a competent axonal transport as shown by the tracing experiments. Thus, it seems that this receptor is not necessary for neuron development and survival.
In agreement with previous reports, we show that RGCs express the P2X7 receptor, and that axotomy induces its up-regulation in the wounded neurons 44 . However, we cannot rule out that the P2X7 staining found in P2rx7 −/− retinas, especially in vascular cells, could be due to alternative splicing variants of P2X7 receptor in the P2rx7 −/− mice used in this study, as it has been already reported in this strain, P2X7 receptor variants which are predicted to escape inactivation in the brain 62 .
Prolonged P2X7 receptor stimulation induces cell death, either directly affecting the mitochondria 19 or by inducing pyroptosis in myeloid cells after activation of the NLRP3 inflammasome and caspase-1 63 . It is reported that stimulation of P2X7 receptor can kill RGC both in vitro and in vivo by a mechanism involving a sustained increase on intracellular Ca 2+ , although the exact mechanism underlying the role of P2X7 receptor in neuronal death is to date unknown 42,64 . Recently, it was reported that the NLRP3 inflammasome was involved in the RGC loss after optic nerve crush injury 65 , our study suggest that P2X7 receptor, an upstream regulator of NLRP3, could be mediating the death of RGC by modulating the inflammasome.
The level of RGC protection found in the injured P2rx7 −/− retinas and in the wild type retinas treated with the selective P2X7 receptor A438079 is similar to that found when administering brain derived neurotrophic factor (BDNF), to date the best described neuroprotectant 4,66,67 . Here we also show that a higher percentage of m + RGCs than of Brn3a + RGCs survive, in accordance with previous works 52 . Interestingly, while neuroprotective treatments that rescue RGCs do not seem to have an effect on m + RGCs 53 , we observe here that P2X7 receptor deficiency or antagonism is also beneficial for this RGC subpopulation, so P2X7 receptor signaling could be relevant in mediating m + RGC death.
These data are of clinical significance because several small compounds antagonizing the P2X7 receptor are being developed by the pharma industry as potential novel anti-inflammatory drugs and some of them have reached clinical trials 68 . In particular, the specific P2X7 receptor antagonism used in this study, A437089, has been satisfactory used in different preclinical models of inflammation in the periphery [33][34][35][36][37] . A438079 has been found to be CNS permeable, and its treatment reduced seizure severity during status epilepticus, as well as dopamine depletion in a model of Parkinson's disease 69,70 . Furthermore, different non-specific P2X7 receptors antagonists (Brilliant Blue G and oxidized ATP) injected into the vitreous body of the eye were able to significantly preserved RGC after ONC  Table 4. Contralateral response: RGCs and phagocytic microglia. Mean number and standard deviation (SD) of RGCs and phagocytic microglial cells (PMCs) in the contralateral to the lesion retinas (right retinas) from wild type and P2rx7 −/− mice after axotomy to the left optic nerve. The percentage of RGC loss and PMC increase was calculated using as 100% the number of RGCs (Table 1) or PMCs (Table 4) in intact retinas within each strain. In wild type mice there is a significant decrease of RGCs in the right retinas 5 and 9 days after axotomizing the left optic nerve (Kruskal-Wallis One way ANOVA, Dunn's post-hoc test p < 0.05 compared to intact retinas). A similar diminution is observed in P2rx7 −/− retinas at 9 days, but it does not reach statistical significance. At all time points and in both strains the number of PMCs in the contralateral retinas is significantly higher than in intact retinas (One way ANOVA, Tukey test p < 0.001). However, their number was significantly lower in the knock out strain at 3 and 9 days (Two ways ANOVA -time and strain-Bonferroni t-test *p < 0.001). § In wild type mice, PMCs increase between 3 to 9 days while in P2rx7 −/− mice the significant increase occurs between 3 and 5 days (One way ANOVA, Tukey test p ≤ 0.001).
Scientific RepoRts | 6:38499 | DOI: 10.1038/srep38499 injury 46 . Our study shows that A438079 treatment reduces RGC death, suggesting that small molecules selectively targeting P2X7 receptor could be beneficial and a novel therapeutic approach to delay neuronal cell death. The contralateral response to unilateral injury is a well-documented effect 4,5,71 , but the underlying mechanism(s) are unknown. In adult rodents, both retinas are communicated by a very small number of RGCs that project from one retina to the other [72][73][74] . They are so few, that their loss would not account for the observed RGC loss, thus there must be additional signalling mechanisms underlying this process. Our work suggests that extracellular ATP could act as a humoral signal that reaches the contralateral retina, independent of the retino-retinal projection, however this might be difficult to explain due to the activity of ectonucleotidases in the brain that quickly degrade extracellular ATP 75 . Alternatively, P2X7 receptor signalling in the contralateral response could be related to the retino-retinal RGCs that release ATP as stress signal in the contralateral retina where their cell bodies lie upon injury of their axons that are part of the axotomized optic nerve. In line with this, the delayed loss of RGCs and PMC response in the contralateral P2rx7 −/− retina is very interesting. In wild type animals, neuroprotection of RGCs in the injured retina with BDNF does not affect the decrease of RGCs nor ameliorates the PMC response in the contralateral one 4 . Interestingly, our study shows that the deficiency of P2X7 receptor changes the contralateral response by delaying the loss of RGCs and decreasing the number of PMCs. This supports the above hypothesis stating that the contralateral response could be mediated by extracellular ATP acting as a danger signal activating the P2X7 receptor. Most probably, the slower response of microglial cells in the P2rx7 −/− retinas results from a slower rate of RGC death. However, there are also evidences showing that P2X7 receptor is required for the activation and proliferation of microglia, regulating the neuroinflammatory response and neuronal death 76,77 and activation of the NLRP3 inflammasome in microglia has been found important for RGC loss after optic nerve crush injury 65 . So, even though we did not find expression of P2X7 receptor in the retinal microglia, we could not rule out the possibility that in the P2X7 receptor deficient mice, microglia function could be also impaired, for example at NLRP3 inflammasome level, and thus being the reason of a higher RGC survival.
Overall, our study suggests that after axonal injury, P2X7 receptor signal is involved in neuronal cell death, either by directly inducing death of neurons and/or by exacerbating the inflammatory response associated to tissue injury, therefore, P2X7 receptor antagonism could be a promising therapy to delay neuronal loss after traumatic injury.

Material and Methods
Animal handling and ethics statement. Adult pigmented C57/BL6 were obtained from the University of Murcia breeding colony and P2X7 receptor-deficient mice in a C57/BL6 background (P2rx7 −/− ) were purchased from Jackson 66 . Mice of 20-25 g body weight were used in this study. Animal care and experimental procedures were performed in accordance to the Association for Research in Vision and Ophthalmology, European Union guidelines for the use of animals in research and were approved by the Ethical and Animal Studies Committee of the University of Murcia (Spain).

Surgery.
Tracing from the superior colliculi: hydroxystilbamidinemethanesulfonate (OHSt, Molecular Probes, Leiden, The Netherland) diluted at 10% was applied to both superior colliculi one week before surgery or processing (intact animals), as previously described 4,6 .
Unilateral optic nerve crush: the left ON was crushed at 1 mm from the optic disk using previously reported methods 4,6 . Animals were sacrificed at increasing times post-lesion. Both retinas were analyzed, injured and contralateral to the lesion ones. Retinas from naive (intact) animals were used as control.
Intravitreal injection was carried out following previously described methods 4 . The left eye was injected with the P2X7 receptor selective antagonist A438079 ((3-(5-(2,3-dichlorophenyl)-1H-tetrazol-1-yl)methyl pyridine; Tocris, Bristol, United Kingdom) at a concentration of 300 ng/eye, dissolved in PBS. This dose was chosen based on previous reports 78 . The antagonist was administered to intact retinas to assess its toxicity, and to injured retinas from wild type animals right after performing ONC. The same volume of vehicle (2.5 μ l) was administered in ONC-injured animals as control.
The number of retinas and the performed analyses are detailed in Table 5.
Western blot. Retinas were fresh dissected and immediately frozen in dry ice. Then, retinas were homogenized in lysis buffer (50 mM Tris-HCl pH8.0, 150 mM NaCl, 2% Triton X-100) supplemented with 100 μ l/ml of protease inhibitor mixture (Sigma-Aldrich, Madrid, Spain) for 2 min at 50 Hz using the mechanical homogenizer TissueLyser LT (Qiagen). Lysates were incubated 30 min on ice and centrifuged to remove particulate matter. Protein concentration was determined with the Bradford assay (Bio-Rad, Hercules, CA). A total of 80 μ g of protein were resolved in 12% SDS-PAGE and transferred to nitrocellulose membranes (Bio-Rad, Hercules, CA) by electroblotting. Blots were blocked for 1 h with 5% skim milk in PBS containing 0.5% Tween-20 (PBS-T, pH 7.4) and then were incubated overnight at 4 °C with anti-P2X7 receptor C-terminal rabbit affinity isolated polyclonal antibody (Alomone Labs) at 1:1000 dilution followed by incubation with HRP-conjugated secondary anti-rabbit (GE Healthcare) at 1:5000 dilution. As loading control β -actin was identified using anti-β -Actin-HRP mouse monoclonal antibody (clone C4, Santa Cruz Biotechnology, Heidelberg, Germany Blotting Detection Reagent (GE Healthcare) and the density of the bands of proteins were quantified using image analyzer Chemidoc TM XRS + (Bio-Rad, Hercules, CA) and the software (ImageLab TM 5.2.1).

Retinal dissection and immunofluorescence. Unless otherwise stated, all the reagents were from
Sigma-Aldrich (Madrid, Spain). Animals were perfused transcardially with 4% paraformaldehyde (PFA) in phosphate buffer 0.1 M after a saline rinse. Then, eyes were prepared for cryosection or retinas were dissected as flat-mounts and immunofluorescence was carried out as previously reported 49 .The general population of RGCs was detected using goat anti-Brn3a (1:750, C-20, Santa Cruz Biotechnologies, Heidelberg, Germany). Melanopsin + RGCs (m + RGCs, including both M1 and M2 subtypes) were detected using rabbit anti-melanopsin UF006 antibody (1:5000 AB-N38, Advance Target Systems, ThermoFisher, Madrid, Spain) that binds to the NH 2 terminal of the mouse melanopsin protein and thus identifies both melanopsin isoforms and detects 90% of m + RGCs in Opn4 Cre ;Z/EG reporter mice 79  Media Cybernetics, Silver Spring, MD, USA) as previously described 6,49 . Individual frames were tiled to reconstruct the whole-mounts (140 individual frames/retina). Retinal cross sections were imaged with a x20 objective.
Retinal analysis: quantification and spatial distribution. Traced-and/or Brn3a + RGCs were automatically quantified (image analysis software: Image-Pro Plus, IPP 5.1 for Windows; Media Cybernetics, Silver Spring, MD) using established routines by our group 3,4,6,49 and their topography was visualized using isodensity maps as previously described 3,4,6,49 . m + RGCs and PMCs were manually dotted on the retinal photomontage, and the dots automatically quantified (IPP software) as described 4,49 . Their distribution was assessed by the fixed-radius (0.165 mm) near neighbour algorithm as reported. All maps were performed using Sigmaplot (SigmaPlot ® 9.0 for Windows ® ; Systat Software, Inc., Richmond, CA, EEUU).

Statistics.
Comparison of two groups (t-test or Mann Whitney test), more than two groups (pairwise multiple comparison procedures, ANOVA or Krustall Wallis ANOVA, and Tukey's or Bonferroni's post-hoc tests), the regression analysis and associated graphs were done with GraphPad Prism v 6 software (GraphPad San Diego, USA). Differences were considered significant when p < 0.05 and tests are detailed in results.  Table 5. Number of retinas analyzed in this study. In addition, 2 intact eyes, and 2 eyes/strain processed 5 days after ONC were used for cross-sections. WB: western blotting. PMC: phagocytic microglial cells, RGCs: retinal ganglion cells.