Early removal of senescent cells protects retinal ganglion cells loss in experimental ocular hypertension

Abstract Experimental ocular hypertension induces senescence of retinal ganglion cells (RGCs) that mimics events occurring in human glaucoma. Senescence‐related chromatin remodeling leads to profound transcriptional changes including the upregulation of a subset of genes that encode multiple proteins collectively referred to as the senescence‐associated secretory phenotype (SASP). Emerging evidence suggests that the presence of these proinflammatory and matrix‐degrading molecules has deleterious effects in a variety of tissues. In the current study, we demonstrated in a transgenic mouse model that early removal of senescent cells induced upon elevated intraocular pressure (IOP) protects unaffected RGCs from senescence and apoptosis. Visual evoked potential (VEP) analysis demonstrated that remaining RGCs are functional and that the treatment protected visual functions. Finally, removal of endogenous senescent retinal cells after IOP elevation by a treatment with senolytic drug dasatinib prevented loss of retinal functions and cellular structure. Senolytic drugs may have the potential to mitigate the deleterious impact of elevated IOP on RGC survival in glaucoma and other optic neuropathies.


| INTRODUC TI ON
Glaucoma is comprised of progressive optic neuropathies characterized by degeneration of retinal ganglion cells (RGC) and resulting changes in the optic nerve. It is a complex disease where multiple genetic and environmental factors interact (Skowronska-Krawczyk et al., 2015;Weinreb, Aung, & Medeiros, 2014). Two of the leading risk factors, increased intraocular pressure (IOP) and age, are related to the extent and rate of RGC loss. Although lowering IOP is the only approved and effective treatment for slowing worsening of vision, many treated glaucoma patients continue to experience loss of vision and some eventually become blind. Several findings suggest that age-related physiological tissue changes contribute significantly to neurodegenerative defects that cause result in the loss of vision.
Mammalian aging is a complex process where distinct molecular processes contribute to age-related tissue dysfunction. It is notable that specific molecular processes underlying RGC damage in aging eyes are poorly understood. While no single defect defines aging, several lines of evidence suggest that activation of senescence is a vital contributor (He & Sharpless, 2017). In a mouse model of glaucoma/ischemic stress, we reported the effects of p16Ink4a on RGC death (Skowronska-Krawczyk et al., 2015). Upon increased IOP, the expression of p16Ink4a was elevated, and this led to enhanced senescence in RGCs and their death. Such changes most likely cause further RGC death and directly cause loss of vision. In addition, the analysis of p16KO mice suggested that lack of p16Ink4a gene protected RGCs from cell death caused by elevated IOP (Skowronska-Krawczyk et al., 2015). Importantly, elevated expression of p16INK4a and senescence were both detected in human glaucomatous eyes (Skowronska-Krawczyk et al., 2015). Therefore, for the first time, p16Ink4a was implicated as a downstream integrator of diverse signals causing RGC aging and death, both characteristics changes in the pathogenesis of glaucoma. Our findings were further supported by a subsequent report showing that p16Ink4a was upregulated by TANK binding kinase 1 (TBK1) a key regulator of neuroinflammation, immunity, and autophagy activity. TBK also caused RGC death in ischemic retina injury (Li, Zhao, & Zhang, 2017). Of particular note, a recent bioinformatic meta-analysis of a published set of genes associated with primary open-angle glaucoma (POAG) pointed at senescence and inflammation as key factors in RGC degeneration in glaucoma (Danford et al., 2017).
Glaucoma remains relatively asymptomatic until it is severe, and the number of affected individuals is much higher than the number diagnosed. Numerous clinical studies have shown that lowering IOP slows the disease progression (Boland et al., 2013;Sihota, Angmo, Ramaswamy, & Dada, 2018). However, RGC and optic nerve damage are not halted despite lowered IOP, and deterioration of vision progresses in most treated patients. This suggests the possibility that an independent damaging agent or process persists even after the original insult (elevated IOP) has been ameliorated.
We hypothesized that early removal of senescent RGCs that secrete senescent associated secretory proteins (SASP) could protect remaining RGCs from senescence and death induced by IOP elevation. To test this hypothesis, we used an established transgenic p16-3MR mouse model (Demaria et al., 2014) in which the systemic administration of the small molecule ganciclovir (GCV) selectively kills p16INK4a-expressing cells. We show that the early removal of p16Ink4 + cells has a strong protective effect on RGC survival and visual function. We confirm the efficiency of the method by showing the reduced level of p16INK4a expression and lower number F I G U R E 1 Removal of early senescent cells has a neuroprotective effect on RGCs. (a) Schematic representation of the p16-3MR transgene. Triple fusion of luciferase, the red fluorescent protein, and tyrosine kinase from HSV virus are under control of the regulatory region of p16Ink4a gene. (b) Plan of the experiment. After unilateral IOP elevation mice are daily injected with GCV (25 mg/kg) intraperitoneally. At day 5 VEP is measured, and tissue is collected for further experiments. (c) Representative images of retina flat-mount immunohistochemistry at day five with anti-Brn3a antibody specifically labeling ~80% of RGC cells. (d) Quantification of RGC number at day five after the treatment of WT animals. N ≥ 5 animals in each group (e) Quantification of RGC number at day five after the treatment of p163MR animals. N = 8 animals in each group. In d and e, statistical tests were performed using ANOVA with post hoc Tukey correction for multiple testing. *p < .05, **p < .01, ***p < .001, n.s., not significant of senescent β-galactosidase-positive cells after GCV treatment.
Finally, we show that treatment of p16-3MR mice with a known senolytic drug (dasatinib) has a similar protective effect on RGCs as compared to GCV treatment in p16-3MR mice.

| Animals
All animal experiments were approved by the UC San Diego

| Drug treatment
The p16-3MR transgenic model (Figure 1a), in which the mice carry a trimodal reporter protein (3MR) under the control of p16 regulatory region (Demaria et al., 2014)

| Visual evoked potential
VEP measurements were taken at five days post-IOP elevation.
This protocol was adapted from prior studies (Ridder & Nusinowitz, 2006). Mice were dark-adapted for at least 12 hr before the procedure. Animals were anesthetized with ketamine/xylazine and their eyes dilated as above. The top of the mouse's head was cleaned with an antiseptic solution. A scalpel was used to incise the scalp skin, and a metal electrode was inserted into the primary visual cortex through the skull, 0.8 mm deep from the cranial surface, 2.3 mm lateral to the lambda. A platinum subdermal needle (Grass Telefactor) was inserted through the animal's mouth as a reference and through the tail as ground. The measurements commenced when the baseline waveform became stable, 10-15 s after attaching the electrodes.
Flashes of light at 2 log cd.s/m 2 were delivered through a full-field Ganzfeld bowl at 2 Hz. Signal was amplified, digitally processed by the software (Veris Instruments), then exported, and peak-to-peak responses were analyzed in Excel (Microsoft). To isolate VEP of the measured eye from the crossed signal originating in the contralateral eye, a black aluminum foil eyepatch was placed over the eye not undergoing measurement. For each eye, peak-to-peak response amplitude of the major component P1-N1 in IOP eyes was compared to that of their contralateral non-IOP controls. Following the readings, the animals were euthanized, and their eyes collected and processed for immunohistological analysis.

| Immunohistochemistry
Following euthanasia, eyes were enucleated and fixed in 4% paraformaldehyde (PFA) in PBS (Affymetrix) for 1 hr and subsequently transferred to PBS. The eyes were then dissected, the retinas flatmounted on microscope slides, and immunostained using a standard sandwich assay with anti-Brn3a antibodies (Millipore, MAB1595) and secondary AlexaFluor 555 anti-mouse (Invitrogen, A32727).
Mounted samples (Fluoromount, Southern Biotech 0100-01) were imaged in the fluorescent microscope at 20x magnification (Biorevo BZ-X700, Keyence), focusing on the central retina surrounding the optic nerve. Overall damage and retina morphology were also taken into consideration for optimal assessment of the retina integrity. Micrographs were quantified using manufacturer software for Brn3a-positive cells in 6 independent 350 × 350 µm areas per flat mount.

| Real-time PCR
Total RNA extraction from mouse tissues, cDNA synthesis, and RT-qPCR experiments were performed as previously described (Skowronska-Krawczyk et al., 2015). Assays were performed in triplicate. Relative mRNA levels were calculated by normalizing results using GAPDH. The primers used for RT-qPCR are the same as in (Skowronska-Krawczyk et al., 2015). The differences in quantitative PCR data were analyzed with an independent two-sample t test.

| SA-βgal assay to test senescence on retinas mouse eyes
Senescence assays were performed using the Senescence b-Galactosidase Staining Kit (Cell Signaling) according to the manufacturer's protocol. Images were acquired using a Hamamatsu Nanozoomer 2.0HT Slide Scanner and quantified in independent images of 0.1 mm 2 covering the areas of interest using Keyence software.

| RE SULTS
Intraocular pressure was increased in one eye of transgenic mice bearing the p16-3MR construct ( Figure 1a). After IOP elevation, mice were intraperitoneally injected with GCV for five consecutive days ( Figure 1b) to specifically deplete p16Ink4a-positive (p16 + ) cells.
In parallel, wild-type animals were subjected to the same protocol, that is, underwent five daily GCV injections after unilateral IOP elevation. Retina flat-mount immunohistochemistry and RGC quantification were used to assess potential impact of drug treatment. We observed that five-day administration of GCV after IOP elevation Next, to test whether the protection of RGC numbers in GCVtreated retinas was accompanied by the protection of the visual circuit integrity on day five, the in vivo signal transmission from the retina to the primary visual cortex was assessed by measuring visual evoked potentials (VEP) (Figure 2a) (Bui & Fortune, 2004;Porciatti, 2015). In brief, the reading electrode was placed in the striate visual cortex, with the reference electrode in the animal's mouth and ground electrode in the tail. Flash stimuli were presented in a Ganzfeld bowl. Response amplitudes were quantified from the peakto-peak analysis of the first negative component N1. Using this approach, we have found that eyes subjected to IOP elevation showed decreased VEP P1-N1 amplitude (Figure 2b), compared to contralateral non-IOP control eyes. However, there was a marked rescue of VEP signals in transgenic animals treated with GCV ( Figure 2b).
Further quantification showed significant vision rescue upon GCV treatment only in p16-3MR and not WT animals ( Figure 2c,2), confirming the specificity of GCV treatment.
Our previous studies indicated that the increase in p16INK4a expression could be first observed as early as day three post-IOP elevation (Skowronska-Krawczyk et al., 2015). Therefore, we chose this time-point to analyze the effectiveness of GCV treatment on senescent cells in treated and control retinas of p16-3MR animals. RGC quantification showed that in animals not injected with GCV only ~15%-20% of cells disappeared at day 3 (compared to ~45%-50% on day 5).
To test whether GCV treatment indeed removed senescent cells in the retina, we used two approaches. First, we quantified the  (Figure 3a). Day 3 also corresponds to the highest F I G U R E 3 Senescence is lowered upon GCV treatment ~2 days before the effects on RGC numbers are observed. (a) At day 3 after IOP, only 20% of RGCs are lost compared to the non-treated eye. Similar numbers of cells are lost in GCV-treated eyes at this stage. N = 3 (non-GCV) and N = 5 (GCV), ANOVA, *p < .05, **p < .01, n.s. -not significant (b) p16Ink4a expression is significantly lower in affected retinas isolated from GCVtreated p16-3MR animals at day 3 after IOP elevation. t-test, **p < .01 (c) Number of SA-b-gal positive cells is lowered upon GCV treatment. Blue arrow -remaining senescent cell (d). Quantification of number of senescent cells upon IOP elevation in retinas isolated from mouse treated and non-treated with GCV. N = 4 (non-GCV), N = 6 (GCV); ANOVA, **p < .01 To inquire in an unbiased way about the differences in signaling pathways and cellular processes affected by IOP, GO analysis using PANTHER was performed (Mi, Muruganujan, Ebert, Huang, & Thomas, 2019). This approach revealed that processes of the immune system response, inflammation, and extracellular matrix composition and cell-matrix interaction were significantly changed in IOP samples (Table 1). We have also detected the significantly deregulated genes involved in apoptosis, microglial activation and interlukin-6 and interlukin-8 production and secretion. This analysis shows that many mechanisms are induced upon an acute IOP elevation, most probably causing additional transcriptional stress to cell.
Further analysis revealed that the genes involved in cellular senescence, extracellular matrix molecules and in factors involved in apoptosis (Table 2) (Pawlikowski et al., 2013) were significantly de regulated upon IOP elevation. Importantly, 3-day treatment to remove p16 + cells significantly mitigated this response (Figure 4c).
These data are in agreement with the loss of the senescence cells upon GCV treatment ( Figure 3b3) and lower detrimental impact of senescent cells on surrounding cells.
Additional GO analysis of the 617 genes which were significantly de regulated upon IOP elevation specifically in non treated retinas (i.e., genes where the effects of IOP were dampened by GCVmediated removal of senescent cells) (Figure 4d) identified a specific enrichment of a class of genes belonging to the ABL1 pathway and ABL1 downstream targets (Fig. S1). Prompted by this finding, we  explored whether dasatinib, a well-known senolytic drug and a Bcr-Abl and Src family threonine kinase inhibitor, could have a beneficial effect similar to GCV in p16-3MR mice. To this end, p16-3MR mice were treated with dasatinib (5 mg/kg) or vehicle for 5 days by intraperitoneal injection, similarly to the experimental procedure used for GCV ( Figure 1b). Performing this experiment in the transgenic mice allowed direct comparison of the efficiencies of both treatments in the same mouse strain. At day five after IOP elevation, VEP measurement was performed and retinas were immunostained to quantify RGC loss. We observed that dasatinib treatment prevented the loss of RGC (Figure 5c) similar to what was observed in GCV-treated animals ( Figure 1e). Most importantly, VEP analysis revealed that senolytic drug treatment successfully prevented vision loss upon IOP elevation ( Figure 5d).
Finally, we explored whether the protective impact of the drug is caused by the sustained inhibition of the cellular processes and whether it is maintained even after the drug is no longer present.
To do that, p16-3MR mice were treated with dasatinib (5 mg/kg) or vehicle for 5 days by intraperitoneal injection, similarly to the experimental procedure used for GCV (Figure 1b). After that, the mice were no longer treated with drug or PBS and at day twelve after IOP elevation, functional measurement was performed and RGCs were quantified. Also in this treatment regime, dasatinib prevented the loss of RGC (Figure 5c) similar to what was observed in GCV-treated animals ( Figure 1e). Additionally, VEP analysis revealed that senolytic drug treatment with seven days "chase" still successfully prevented vision loss upon IOP elevation (Figure 5d).

| D ISCUSS I ON
The collective findings of the current study strongly support the notion that removal of senescent cells provides beneficial protective Dasatinib is a selective tyrosine kinase receptor inhibitor that is commonly used in the therapy of chronic myelogenous leukemia (CML). Other studies have shown that treatment with dasatinib is effective in destroying senescent fat cell precursors .
Our RNA-seq data pointed to this senolytic drug as a potential candidate for in vivo treatment of retinal damage induced by IOP elevation. Notably, we found that the level of RGC protection resembles the one obtained with GCV treatment of p16-3MR transgenic line.
Based on these findings, we conclude that dasatinib treatment resulted in RGC protection through removal of senescent cells. It will be of interest to further investigate the possible therapeutic effects of other senolytic drugs in glaucoma and glaucoma models.
The gene encoding p16INK4a, CDKN2A, lies within the INK4/ ARF tumor suppressor locus on human chromosome 9p21; this is the most significant region to be identified as having an association with POAG in different population samples (Ng, Casson, Burdon, & Craig, 2014). Although the molecular mechanism of many of these associations is yet to be described, we have shown that one of them Akt-Bmi1 phosphorylation pathway (Li et al., 2017). Given the complexity of the 9p21 locus, we believe that there are more pathways involved in p16Ink4a regulation and further work is needed to understand the role of p16Ink4a as a integrator of these signals especially upon IOP elevation.
Several collaborative efforts identified numerous SNPs localized within the 9p21 locus to be highly associated with the risk of open-angle glaucoma including normal-tension glaucoma (NTG), a glaucomatous optic neuropathy not associated with elevated IOP (Killer & Pircher, 2018;Wiggs & Pasquale, 2017). Intriguingly, one of the top variants associated with the risk of NTG is located in the gene TBK1, a factor that has been recently shown to be implicated in upregulation of p16ink4a gene (Li et al., 2017). Finally, recent studies have also revealed that specific methylation patterns in the 9p21 locus are strongly associated with the risk of NTG glaucoma (Burdon, 2018). It is notable that the positions of most, if not all, of these SNPs and methylation markers overlap with active regulatory regions within the locus identified by ENCODE (Consortium, 2012).
Although regulation of the 9p21 locus in the context of many diseases and aging is under extensive investigation, it still remains to be explicitly addressed in relation to glaucoma.
Another major type of glaucomatous optic neuropathy is angle closure glaucoma (ACG), a condition characterized by blockage of the drainage angle of the eye. To date, there is no study reporting F I G U R E 5 Dasatinib protects retina degeneration. (a) Plan of the experiment. After unilateral IOP elevation, mice are daily injected with dasatinib (5 mg/kg) intraperitoneally. At day 5, VEP is measured and tissue is collected for further experiments. Immunohistochemistry of Brn3a and activated caspase show increase of apoptosis at day 3 after IOP treatment. (b) Retina flat-mount immunohistochemistry at day 5 with anti-Brn3a antibody specifically labeling ~80% of RGC cells. (c,d) Quantification of RGC number (c) or VEP responses (d) at day 5 (four conditions) or day 12 (additional 7 days of "recovery," two conditions) after the 5 days treatment of p16-3MR animals with dasatinib. N > 4 animals in each group. Statistical tests were performed using ANOVA with post hoc Tukey correction for multiple testing. *p < .05, ***p < .001, n.s. -not significant (e). Model. Top: Upon elevated IOP damaged cells become senescent and start to express SASP molecules. While disease progresses, the SASP molecule induces senescence or apoptosis in neighboring cells. Bottom: When senescent cells are removed using senolytic drug the neighboring cells are not exposed to detrimental SASPs and the disease progression is significantly slowed down. Remaining cells are healthy genetic variants or methylation markers in the 9p21 locus significantly associated with the risk of ACG despite several studies implicating various molecular mechanisms (Evangelho, Mogilevskaya, Losada-Barragan, & Vargas-Sanchez, 2019). Nevertheless, the fact that progressive vision loss is observed in PACG patients, even after lowering the IOP (Brubaker, 1996), raises the question whether an association could be observed between 9p21 markers and the progression rather than the risk of the disease. Further studies to unravel such associations are necessary.
Markers of cellular senescence such as expression of the p16Ink4a and SASP molecules dramatically increase during aging in both humans and mice. Several studies suggest that p16Ink4a + cells act to shorten healthy lifespan by promoting age-dependent changes that functionally impair tissues and organs Jeon et al., 2017;Krishnamurthy et al., 2006). Intriguingly, a recent explosion of studies has shown that removal of senescent cells using senolytic drugs in progeroid (accelerated aging phenotype) and healthy mice induces lifespan extension and improves the health of animals (Baker et al., 2011Scudellari, 2017;Xu et al., 2018). Our studies suggest a potential use of such therapy to reduce glaucoma associated blindness, either as a stand-alone treatment or together with IOP-lowering therapies.

ACK N OWLED G M ENTS
We thank Sherrina Patel for help with this project. This work was

CO N FLI C T O F I NTE R E S T
Nothing to declare.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available in GEO database (GSE141725) and in Dryad at https ://doi. org/10.6075/J0707ZTM (Rocha, 2019).