Hypoxia and mechanical stretch during episodes of ocular hypertension (OHT) induce neuroinflammation in the retina and optic nerve. Neuroinflammation is marked by activation of microglia and astrocytes, which is linked to the release of pro-inflammatory cytokines (Yang et al., 2011; Krizaj et al., 2014; Lim et al., 2016; Lu et al., 2017), other danger factors and deposition of C1q complement on neuronal dendrites (Williams et al., 2016; Liddelow et al., 2017). Neuroinflammatory damage that includes dendritic pruning, axonal injury and neurotoxicity was shown to facilitate injury and loss of retinal ganglion cells (RGCs) in the OHT-injured retina (Soto and Howell, 2014). Recent studies, performed in the glia-neuron co-cultures (Orellana et al., 2012; Liddelow et al., 2017) indicated that cytokines released by glia play a major role in such damage. Consistently, blockade of different pro-inflammatory signaling resulted in RGC protection (Dvoriantchikova et al., 2009; Yang et al., 2011; Howell et al., 2014; Krizaj et al., 2014; Madeira et al., 2015; Silverman et al., 2016; Williams et al., 2016). Pro-inflammatory cytokines, including the IL-1 family, are produced by glial, microglial and neuronal cells in the retina in response to ischemic injury and in glaucoma (Yoneda et al., 2001; Ivanov et al., 2006, 2008; Namekata et al., 2009; Qi et al., 2014). Their production is controlled by inflammasome activation and significantly increased after exposure to OHT (Barakat et al., 2012; Chi et al., 2014). Genetic ablation and/or pharmacological inhibition of caspase-1 or inflammasome regulators, including Panx1, IL-1 or P2X receptors was shown to protect retinal neurons in several injury paradigms (Yoneda et al., 2001; Zhang and Chintala, 2004; Arai et al., 2006; Pelegrin et al., 2008; Seki et al., 2009; Dvoriantchikova et al., 2012; Puyang et al., 2016).

Sterile injury-induced inflammasome activation has been demonstrated in neural cells of the central nervous system (CNS; de Rivero Vaccari et al., 2008, 2009; Abulafia et al., 2009). Activation of both canonical and non-canonical inflammasomes was documented in the inner retina and optic nerve of rodent eyes challenged with transient elevation of intraocular pressure (IOP; Dvoriantchikova et al., 2012; Chi et al., 2014; Qi et al., 2014; Albalawi et al., 2017). However, temporal and cellular expression patterns of inflammasome pathways in the retina exposed to the OHT injury remain poorly characterized. Canonical inflammasomes, assembled by the NLRP1, NLRP3 and Aim2 sensor proteins, proteolytically activate caspases 1 and/or 11 (caspase-4 in humans) that are essential for processing and release of IL-1β. This is typically marked by oligomerization of the apoptosis-associated speck-like (ASC) adaptor protein into a megacomplex with NLRPs/Aim2 sensors and caspase-1 to form ASC “specks.” The most recent research showed that, in addition to IL-1 precursors, caspases 1 or 11 also cleave the N-terminal pore-forming part of gasdermin D protein (GSDMD-NT). The GSDMD-NT fragments oligomerize to form membrane megapores that release IL-1 but could become cytolytic and cause pyroptotic cell death (Aglietti et al., 2016; Chen et al., 2016; Liu et al., 2016; Sborgi et al., 2016). Although inflammasome-induced pyroptosis have been shown to contribute to pathogenesis of age-related macular degeneration (Tseng et al., 2013; Brandstetter et al., 2016; Viringipurampeer et al., 2016; Kerur et al., 2018), potential role of pyroptosis in OHT injury has not been investigated so far.

Pannexin1 (Panx1) has been implicated in the pathophysiological cascade triggered by the OHT injury (Thompson et al., 2006; Bargiotas et al., 2009), as well as in inflammasome activation in the CNS and the retina (Dvoriantchikova et al., 2006, 2012; Silverman et al., 2009; Adamson and Leitinger, 2014; Beckel et al., 2014). In the retina, Panx1 is abundant in RGCs, where its expression level correlates with RGC sensitivity to ischemic and mechanical damage (Dvoriantchikova et al., 2018). Panx1 interactions with various membrane receptors underlie its responsiveness to a variety of common neurotoxic and danger signals (Krizaj et al., 2014; Makarenkova and Shestopalov, 2014). While TNF, TLR4 and IL-1 receptor-signaling regulate transcriptional “priming” of inflammasome components (Signal 1), the signaling via Panx1-P2X7 signalosome regulates their assembly (Signal 2), which is critically required for activation (Zhang and Chintala, 2004; Silverman et al., 2009; Yang et al., 2011; Krizaj et al., 2014; de Rivero Vaccari et al., 2014). The Panx1-P2X7 signalosome was shown to play a central role in the increase of extracellular ATP and dysregulation of intracellular Ca²⁺ and K⁺, the key inflammasome-triggers in the CNS and retinal injuries such as the mechanical/ischemic insult during the OHT injury (Krizaj et al., 2014; Makarenkova and Shestopalov, 2014).

What cell types are known to activate inflammasome in the post-ischemic or mechanically injured retina? Currently, a growing number of reports indicate that retinal pigment epithelium (RPE; Anderson et al., 2013; Brandstetter et al., 2015; Gelfand et al., 2015), astrocytes (Albalawi et al., 2017), Muller cells (Devi et al., 2012; Mohamed et al., 2014; Natoli et al., 2017) and microglia (Abulafia et al., 2009; Ystgaard et al., 2015) can activate NLRP3 inflammasome. Mixed retinal glia cultures responded robustly to ATP stimulation after priming by E. coli lipopolysaccharide (LPS; Murphy et al., 2012). Muller cells were shown to produce IL-1β under hyperglycemic conditions (Busik et al., 2008; Devi et al., 2012), photo-oxidative injury (Natoli et al., 2017) and in amyloid beta toxicity models (Dinet et al., 2012). Both astrocytes and Muller cells were reported to cause neurotoxicity after stimulation with activated microglia in various disease models (Natoli et al., 2017; Yun et al., 2018). However, relative activation of inflammasome and production of IL-1 cytokines is reportedly more robust in retinal astrocytes and Muller cells, rather than in microglia (Li et al., 2012; Ystgaard et al., 2015), suggesting that macroglial cells are the major drivers of neuroinflammatory damage in the inner retina.

In this work, we sought to determine spatial and temporal patterns of inflammasome activation after OHT injury. We detected that activation of three canonical inflammasomes, NLRP1, NLRP3 and Aim2 in retinal glia, microglia and RGCs is co-regulated by Panx1. To explore which danger signals produce the most robust activation, we applied defined inflammasome inducers to mouse retinas in vivo using intravitreal injection. Our results picture a complex pattern of neuron-glia interactions that underlie innate immune responses in OHT injury and facilitate the injury-induced neurotoxicity.

We induced unilateral acute OHT in eyes of WT, Panx1^−/− and Casp1^−/−Casp4(11)^del mice lacking both caspase-1 and caspase-11 inflammasomes and analyzed vitreo-retinal extracts from experimental and contralateral normotensive controls using ELISA. We detected release of IL-1β cytokines in WT eyes starting at 1.73 ± 0.19 pg/ml 6 h postinjury, peaking at 4.62 ± 1.28 pg/ml in 12 h and declining to 2.36 ± 0.56 pg/ml at 24 h after OHT injury (P < 0.05 at all time points; Figure 1A, red line). Normotensive WT control eyes showed statistically significant IL-1β release of 0.69 ± 0.11 pg/ml only at 6 h postinjury, which is consistent with the reported bilateral effect of unilateral ischemic injury (Sapienza et al., 2016). Relative to the WT eyes, the OHT-induced IL-1β release in Panx1^−/− mice was approximately 2-fold lower at all time points, reached the maximum of 2.8 ± 0.25 pg/ml at 12 h and 1.12 ± 0.11 pg/ml at 24 h postinjury (Figure 1A, blue line). No IL-1β release was detected in normotensive control eye of Panx1^−/− (green dotted line) or Casp1^−/−Casp4(11)^del mice (CaspDKO, Figure 1A, dotted yellow line). Consistent with these findings, Western blot analysis of retinal extracts from the OHT-injured eyes showed increased levels of the principle IL-1 convertase caspase-1 and the major pore-forming protein GSDMD as well as that NLRP3 sensor protein at 12 and 24 h postinjury (Figure 1B). Both Casp1 and GSDMD showed cleaved mature isoforms, the evidence of their proteolytic activation, at the same time points. To further validate the OHT-induced acute activation of GSDMD, we used RT-PCR to confirm gene activation. Our analysis detected robust increase of GSDMD expression, averaging 6.5- and 3.4-fold increase relative to normotensive control eyes (8.1- and 12.2-fold relative to naïve eyes) at 12 h and 24 h postinjury, respectively (Figure 1C). Noteworthy, the control eyes showed a significant 2.1- and 3.6-fold increase in GSDMD expression relative to naïve eyes at 6 h and 24 h postinjury.

Next, we asked the question whether inflammasome activation contributes to neuronal loss in OHT injury. Using a gene knockout approach, we showed that ablation of the Casp1 gene protected RGCs (3.4 ± 3.4% vs 26.4 ± 4.9% loss in WT OHT retinas, Figure 2A). A similar effect was achieved by genetic ablation of the Panx1 gene (4.5 ± 5.1% loss), the upstream regulator of the inflammasome assembly. Importantly, a similar level of protection was achieved with pharmacological blockade of the Panx1 channel by the specific blocker probenecid (Pbcd). Noteworthy, Casp1^−/− mouse line from the Jackson Labs repository has the Casp11-inactivating Casp4(11)^del passenger mutation, known to block pyroptosis induced by intracellular LPS (Yang et al., 2011; Aglietti et al., 2016). To test whether caspase-11 is also involved in OHT-induced neurotoxicity, we measured RGC survival in injured retinas of the Casp11^−/− knockout strains, as well as in the Panx1^−/−Casp1⁺ animals (obtained from Genentech, back-crossed to C57Bl6 background for seven generations) that have functional Casp4(11) gene. In these experiments, we applied isoflurane gas instead of ketamine/xylasine anesthesia to take advantage of an increased (from 26% to about 65%) post-IR injury RGC loss rate, which would allow us detecting even minor differences in protection. Casp11 KO mice showed a 72.7 ± 2.9% loss of NeuN-positive cells that was not statistically different from the loss in the WT (64.7 ± 5.2%) retina (Figure 2). The fact that similar levels of RGC protection were observed in both Panx1^−/− and Casp1^−/−Casp4(11)^del animals (4.5 ± 5.1% and 3.4 ± 3.4%, respectively), while the Casp11^−/− animals lacked (72.7 ± 2.9% vs 64.7 ± 5.2% loss) protection, helped us to rule out active Casp11 involvement in post-OHT degeneration (Figure 2B).

To map the activity of Casp1 interleukin convertase in retinal cells, we performed in vivo intravitreal injections of fluorescent Casp1 substrate FLICA 660-YVAD-FMK into both experimental (OHT) and normotensive control mouse eyes. The majority of Casp1-active cells localized within the inner nuclear layer (INL) and GCL layers of the inner retina (Figure 3A). As expected, Panx1 knockout or the treatment with the Panx1 blocker probenecid resulted in a dramatic suppression of Casp1 activity, particularly in the GCL, at 12 h post-OHT. Immunohistochemistry showed very low levels of Casp1 colocalizing to RGCs and their processes and retinal microvasculature in normotensive control eyes (NT Cntr, Figure 3B). In agreement with the activity mapping data, these levels were significantly increased in experimental OHT eyes of WT at 12 h and peaked at 24 h postinjury. Colocalization analysis showed the most intense specific staining for Casp1 protein in the GCL, colocalizing with RGCs (RBPMS⁺) and RBPMS⁻ neurons at 12 h and with astrocytes in the neurofiber layer at 24 h (Figure 3B).

Next, we investigated inflammasome sensor proteins potentially involved in proteolytic activation of Casp1 and release of IL-1β in OHT injured retinas. For this purpose, we examined expression of canonical NLRP1, NLRP3 and Aim2 using RNAscope mRNA in situ hybridization and immunostaining. RNAscope data analysis showed differential levels of expression of NLRP3 and Aim2 genes in the INL and GCL of OHT-challenged retinas relative to normotensive retinas at 12 h post-injury (Figure 4). In normotensive eyes, the expression of NLRP1 and NLRP3 genes (individual transcripts visualized as green and white dots, respectively) was observed in both GCL and INL layers. The NLRP3 and Aim2 gene transcripts were relatively abundant in the INL vs. GCL layers, while NLRP1 transcript showed similar abundance across the two layers (Figure 4). In the OHT-injured retinas, the NLRP3 and Aim2 transcripts showed an increase in abundance at 12 h after acute OHT injury compared to normotensive sham-treated control eyes. NLRP3 became more abundant in both GCL and INL (Figure 4A), Aim2 transcript had increased abundance only in the INL after OHT injury (Figure 4B), and NLRP1 remained unchanged.

To determine the cell types expressing distinct inflammasomes, we used multiplexing of the RNAscope labeling of transcripts with immunostaining for common glial cell markers: GFAP for astrocytes and glutamine synthetase (GlSyn) for Muller cells. In control normotensive retinas, the transcripts for NLRP1 colocalized predominantly with cells in the GCL, where astrocytes, microglia and two types of neurons, RGCs and displaced amacrine cells, are located. In both normotensive and experimental eyes 12 h after the OHT injury, the NLRP1 transcripts showed colocalization with cells in the INL and GCL, the retinal layer populated with astrocytes and RGCs (Supplementary Figure S1). Further analysis using co-immunostaining for GFAP marker protein showed no colocalization of NLRP1 transcripts with astrocytes (GFAP⁺ cells in the NFL later), in either normotensive control of OHT-challenged eyes (Supplementary Figure S1). The transcripts for gene encoding Aim2 sensor protein co-localized predominantly with GlSyn-positive Muller cells in the INL (Figure 4B).

Immunohistochemistry analysis showed labeling specific to the NLRP1 protein in the inner retina of normotensive eyes, where it predominantly colocalized with the RGC marker RBPMS (Figure 5A). The OHT challenge did not cause an increase in the NLRP1 labeling in these neurons and their axons in the NFL at 24 h post-insult (Figures 5A,D). NLRP1 labeling was also detected in axonal bundles in the NFL and in the RGC dendrites (Figures 5A,D). However, in contrast to NLRP1, the immunolabeling for NLRP3 protein has increased by 24 h post-OHT, colocalizing mostly with RGCs and RBPMS-negative neurons in the GCL and astrocytes in the NFL (Figures 5C,D). In both naïve an normotensive control retinas, NLRP3 labeling co-localized with GFAP⁺ astrocytes and, to a less degree, with CD11b⁺ microglial cells (Supplementary Figure S2A). No overlap was detected between NLRP3 and Aim2 labeling. Notably, in the outer retina, the highest NLRP3 labeling localized to the RPE (Supplementary Figure S2B). Immunohistochemistry showed strong colocalization of Aim2-specific labeling with the Muller glia marker GlSyn (Figure 5B). Combined, these data allow us to conclude that three canonical inflammasomes become activated early after OHT injury: NLRP1 and NLRP3 inflammasomes in the GCL, the NLRP3 inflammasome in astrocytes and the Aim2 inflammasome in Muller glia. Immunohistochemical evidence indicates expression of NLRP1 in RGCs of normotensive and naïve control retinas and expression of both NLRP1 and NLRP3 in these neurons after OHT challenge.

Next, we studied formation of the mature inflammasome complexes, referred to as ASC specks, which we visualized using a transgenic bioindicator model expressing fluorescent ASC-citrin fusion protein (Tzeng et al., 2016). ASC-citrin was shown to incorporate into oligomerizing inflammasome complexes to brightly label the ASC specks in vivo, thus providing a surrogate inflammasome activation marker in mouse tissues (Scheiblich et al., 2017; Venegas et al., 2017). Scarce small size ASC specks of incompletely assembled complexes colocalized primarily with glia-vascular interfaces in naïve retinas, a striking contrast to the large, mature ASC specks of active fully assembled complexes that were abundant in the GCL and INL of the OHT-challenged retinas (Figures 6A,B). Relative to naïve tissues, the normotensive control retinas had an increased speck abundance, with the most specks observed along blood vessels (Figure 6A), suggesting that the pro-inflammatory effect of the contralateral eye injury (Sapienza et al., 2016) is spread via systemic circulation. Immunostaining for cell markers in retinal sections showed occasional colocalization with astrocytes and RGCs in control normotensive eyes (Figures 6B,C), where specks aligned well with blood vessels ensheathed with astrocytes. In the post-OHT retina, a proportion of specks colocalized with astrocytes and CD11+ microglial/macrophage cells in NFL as well as with RBPMS-positive RGCs in the GCL (Figures 6B–D). The analysis of the INL showed ASC speck colocalization with GlSyn-positive Muller cells in retinal wholemounts, whose nuclei reside at approximately 40–45 μm deep below NFL layer (Figure 6E). The analysis of immunostaining for the ASC protein showed a robust induction at 24 h post-OHT but the pattern was different from ASC-citrin labeling, most likely because antibodies selectively recognized monomeric ASC. The labeling was observed in both RGCs, RBPMS-negative neurons in the GCL and in astrocytes in the NFL (Supplementary Figure S3).

To explore which inflammasome complexes are activated in the retina in response to distinct pro-inflammatory danger signals, we unilaterally injected cocktails of well-characterized inflammasome agonists into the vitreous body of the mouse eye without IOP elevation. The four cocktails, containing 100 ng/ml LPS mixed with 10 μg/ml nigericin (NLRP3 agonist), 2.5 mM ATP (NLRP1 and NLRP3 agonist), S. typhimurium flagellin 20 ng/ml (FLA-ST, an NLRC4 agonist) and Lipofectamine-conjugated poly dA:dT DNA (intracellular agonist of Aim2, 15 μM/eye), were unilaterally injected into experimental eyes with contralateral control eyes receiving vehicle only. The analysis of the ELISA data showed release of IL-1β in vivo in the retinas treated with these cocktails (Figure 7A). We observed the strongest release in the eyes treated with the LPS/bzATP cocktail. Importantly, this treatment triggered an early (6 h postinjection) induction of the NLRP1/NLRP3 inflammasome in RGCs and astrocytes as evidenced by the analysis of the ASC-citrin speck colocalization with RBPMS and GFAP markers, respectively (Figures 7B–D).

Our results, showing upregulation of three distinct canonical inflammasome complexes at 12–24 h post-OHT, raised the question whether RGC pathology in the inner retina is associated with pyroptotic cell death. The Western blot analysis of retinal extracts (Figure 1B) showed a time-dependent increase in expression and cleavage of gasdermin D (GSDMD), which is required for pyroptotic pore formation. Immunohistochemistry analysis confirmed an increase in GSDMD staining at 12–24 h postinjury (Figure 8). Importantly, RGCs were the first cell type in the GCL to become strongly GSDMD-positive at 12 h postinjury, while the labeling in astrocytes and microglia was very weak at this time point (Figure 8A). At 24 h, we detected a mild GSDMD-positive labeling of RGCs with patchy labeling of the plasma membrane that is consistent with pre-pyroptotic bubbling of the plasma membrane (Figures 8B–D). These cells were also weakly positive for Casp1 and RBPMS, and had shrunken nuclei or lacked one completely. Taken together, our results indicate that OHT injury-induced release of IL-1β occurred within hours after the insult and correlated with similarly rapid increases in gasdermin D and caspase-1 activation, first in RGCs and later in astroglia. Thus, it is reasonable to suggest that a retinal inflammasome caused Casp1-mediated pyroptotic pore formation that mediated IL-1β release have subsequently induced pyroptotic cell death in GSDMD-positive neurons.

In this study, we demonstrate that activation of the inflammasome and release of IL-1β in the retina occur within a few hours after OHT injury. Our evidence indicates that this acute inflammasome activation triggers pyroptotic death of RGCs via caspase-1-mediated mechanism.

We demonstrate that activation of the inflammasome and release of IL-1β in the retina occur within a few hours after OHT injury. We found three canonical inflammasome complexes in the retina: NLRP1, NLRP3 and Aim2, which had a distinct distribution in cell types (Figures 4, 5, Supplementary Figures S1–S3). Inflammasome activation was detected at the transcriptional level by RNAscope, at the protein level by immunohistochemistry, and at the functional level by the analysis of responses (i.e., IL-1β release) to the OHT injury or specific agonists in vivo. Our results are consistent with earlier studies that have detected NLRP1 and NLRP3 activation in RPE and astrocytes (Doyle et al., 2012; Tarallo et al., 2012; Chi et al., 2014; Qi et al., 2014; Albalawi et al., 2017). To the best of our knowledge, activation of NLRP1 and NLRP3 in retinal neurons are new findings. The neurotoxic activity of the Aim2 inflammasome has only been reported in the brain (Adamczak et al., 2014; Ge et al., 2018), but not in the retina. It remains to be determined, however, if activation of Aim2 which we detected in the Muller glia, plays any role in OHT injury-induced neurotoxicity. Activity of another type of inflammasome, NLRC4, was only detected by immunohistochemistry, so we did not pursue its further study.

Our immunohistochemistry data show that each type of inflammasome complex had a distinct cellular and temporal pattern of activation. In particular, the NLRP1 complex is constitutively active in RGCs, NLRP3 has low grade constitutive activity in astrocytes and vasculature but is OHT-inducible in GCL neurons, and the Aim2 complex is active in Muller glia. A certain level of constitutive activity of the three inflammasomes suggests that IL-1 signaling has a role in retinal homeostasis, as proposed in the literature (Namekata et al., 2009).

Analysis of events induced by OHT showed acute increase in inflammasome activity both in retinal neurons and glial cells in the GCL and INL layers. Which mechanism of danger signaling is the key for OHT injury-induction of inflammasomes? To this end, we show that the disruption of Panx1 signaling inhibits caspase-1 activation, suppresses and delays IL-1β release (Figures 1A, 2A). Most likely, Panx1 induces inflammasome via ATP release from mechanically stressed or ischemically injured cells. Purinergic signaling via a cell surface complex comprised from the Panx1 channel and P2X7 receptor has been implicated in inflammasome activation in the brain (Pelegrin et al., 2008; Silverman et al., 2009; de Rivero Vaccari et al., 2009) and retina (Kerur et al., 2013; Albalawi et al., 2017; Platania et al., 2017). It responds to extracellular ATP, facilitating inflammasome activation in paracrine and autocrine fashions (Reigada et al., 2008; Xia et al., 2012; Beckel et al., 2014). The fact that intravitreal injection of ATP in triggered the highest level of IL-1β release and ASC speck induction in RGCs and astrocytes (Figures 6, 7A), supports the importance of ATP-induced danger signaling for retinal inflammasome induction.

The current paradigm identifies microglia and macroglia as main drivers of neurotoxic inflammation in the retina (Hernandez, 2000; Bosco et al., 2012; Tezel et al., 2012; Abcouwer et al., 2013; Formichella et al., 2014; Mac Nair et al., 2016; Silverman et al., 2016). This model was based largely on the observed protection for RGCs against excitotoxicity and OHT-injury in mouse genetic models with suppression of glial (Ganesh and Chintala, 2011) or microglial (Silverman et al., 2016) activity or astroglial NF-kappaB signaling (Dvoriantchikova et al., 2009; Barakat et al., 2012). Contrary to this glia-centric model of neuroinflammation, our results argue that RGCs and other neurons in the GCL also directly participate in the innate immune response. Furthermore, we propose that neurons are the cell type that triggers injury response within a few hours; this occurs via activation of the neuronal NLRP1 and NLRP3 inflammasomes. The presence of basal NLRP1 and NLRP3 activity can prime the cells for quick response via the acute upregulation of Casp1 and cleavage of gasdermin D, which we detected in the inner retina after the OHT injury.

Inflammasome activation in various pathologies was shown to cause cell death via pyroptosis (Celkova et al., 2015; Viringipurampeer et al., 2016; Kerur et al., 2018). We have obtained evidence of caspase-1-mediated pyroptotic cell death during acute inflammasome activation and asked whether pyroptosis can contribute to neuronal death following IOP elevation? Together with the strong RGC protection by the knockout of pro-pyroptotic caspase-1, the induction of the gasdermin D pore protein in these cells supports such contribution (Figures 1C, 2C, 7B and 8). Interestingly, the strongest induction of caspase-1 and its substrate gasdermin D occurred in GCL neurons (Figures 3B, 7C, 8A). The dramatic morphological changes in RGCs spatially and temporally correlated with activation of caspase-1 and gasdermin D at 12–24 h postinjury (Figures 8B–D). These changes included nuclei shrinkage/loss in the RBPMS⁺cells with GSDMD-specific immunolabeling at the surface, which is consistent with pyroptotic death-induced membrane bubbling (Chen et al., 2016). Taken together, our results indicate that inflammasome activity-induced pyroptosis directly contributes to RGC death after transient IOP elevation.

Our discovery of pyroptosis as the early cell death type in OHT injury is highly significant because this is the most pro-inflammatory type of cell death. Pyroptotic release of ATP, ASC specks, HMGB1 and other danger signals from dying cells was shown to spread inflammation and induce secondary death of neural and blood endothelial cells (de Rivero Vaccari et al., 2014; Venegas et al., 2017; Ge et al., 2018). Retinal microenvironment changes, conditioned by pyroptotic release of these danger factors can significantly contribute the wave of secondary death both directly, via circulation or glial neuroinflammation. In fact, we found evidence that inflammasomal ASC specks can spread from the experimental eye into the contralateral control (normotensive) via blood capillaries (Figure 6A).

This study illuminates the role of neuronal inflammasome in triggering cell death. However, our data here and in earlier studies (Brambilla et al., 2009; Dvoriantchikova et al., 2009; Nikolskaya et al., 2009; Barakat et al., 2012) as well as other investigators (Tezel and Wax, 2000; Yuan and Neufeld, 2000; Ju et al., 2006; Bosco et al., 2008, 2012; Orellana et al., 2012) establish the involvement of glia and microglia in RGC pathology. It is also possible that active glia-microglia or glia-neuron signaling could facilitate neuronal pyroptosis, and indeed astrocytes and microglia are present in the vicinity of GSDMD-positive RGCs (Figures 7C, Supplementary Figure S3B). Alternatively, the timing of the observed inflammasome induction in RGCs may suggest that danger signaling, particularly ATP and ASC speck release from pyroptotic neurons can attract phagocytic microglia and astrocytes (Verderio and Matteoli, 2001; Choi et al., 2007; Brown and Neher, 2010; Chu et al., 2012; Venegas et al., 2017). An active role of such indirect signaling has been suggested in reports on hemichannel-mediated neurotoxic signaling (Bennett et al., 2012; Orellana et al., 2014) and microglial cytokine-mediated conversion of A1 astrocytes (Liddelow et al., 2017; Yun et al., 2018). These channels are particularly abundant in RGCs and were shown to facilitate OHT-induced ATP release from both glia and RGCs (Reigada et al., 2008; Xia et al., 2012; Beckel et al., 2014; Albalawi et al., 2017). Overactivation of Panx1 was shown to be essential to RGC death and retinal neuroinflammation following OHT injury (Dvoriantchikova et al., 2012, 2018). In this work, we also showed that genetic ablation or pharmacological blockade of Panx1 suppressed inflammasome activity and protected RGCs similarly to the Casp1^−/−Casp4(11)^del knockout (Figure 2). We hypothesize that the initial insult kills neurons via pyroptosis during acute injury phase, whereas danger signaling (e.g., ATP and K⁺ release from dying neurons) facilitates activation of the retinal macroglial and microglial cells to induce secondary degeneration.

The question remains whether acute neuronal pyroptosis plays a role in the pathology of OHT injuries, including glaucoma. The IOP spike-induced degeneration of RGCs has been demonstrated in a rodent model (Joos et al., 2010). Provided that the exposure to OHT episodes was observed during the head-down posture in aging people (Ventura et al., 2013; Porciatti et al., 2017), this mechanism could potentially contribute to glaucomatous-like degeneration of RGCs.

VIS developed the central hypothesis. VIS, AP and VZS contributed to experimental design, implementation of the methodology and writing the article. AP, VIS, DP, WA, GD, ZK, JQ, GR and AR performed the experiments and the data analyses. DP and GD assisted with the research design, data interpretation, manuscript writing and editing.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.