Valproic Acid Reduces Neuroinflammation to Provide Retinal Ganglion Cell Neuroprotection in the Retina Axotomy Model

Abstract

Neuroinflammation is a critical and targetable pathogenic component of neurodegenerative diseases, including glaucoma, the leading cause of irreversible blindness. Valproic acid has previously been demonstrated to reduce neuroinflammation and is neuroprotective in a number of experimental settings. To determine whether valproic acid can limit retinal neuroinflammation and protect retinal neurons we used an ex vivo retina explant (axotomy) model to isolate resident glial responses from blood-derived monocytes. Neuroinflammatory status was defined using high resolution confocal imaging with 3D morphological reconstruction and cytokine protein arrays. Valproic acid significantly reduced microglia and astrocyte morphological changes, consistent with a reduction in pro-inflammatory phenotypes. Cytokine profiling demonstrated that valproic acid significantly attenuated or prevented expression of pro-inflammatory cytokines in injured retina. This identifies that the retinal explant model as a useful tool to explore resident neuroinflammation in a rapid timescale whilst maintaining a complex system of cell interactions and valproic acid as a useful drug to further explore anti-neuroinflammatory strategies in retinal disease.

Keywords: astrocyte; glaucoma; microglia; neuroinflammation; retina.

FIGURE 1

Microglial morphological changes associated with pro-inflammatory phenotypes occur in all retinal microglia niches in retinal explants. Retinal explants were maintained for 1 day (D1) or 2 days (D2) ex vivo and compared to D0 controls (eyes fixed immediately after enucleation). IBA1 labeled microglia were imaged and individually reconstructed using Imaris software from the three retinal microglia niches (A) nerve fiber layer/ganglion cell layer (NFL/GCL) (B) inner plexiform layer/inner nuclear layer (IPL/INL), and (C) outer plexiform layer (OPL). (D) Across all microglial niches, at D1 there was a significant reduction in the number of microglia processes, total process length, field area, and normalized volume, demonstrating that microglia became less complex and more voluminous, as occurs with a shift towards pro-inflammatory phenotypes. The degree of change was no greater in any one layer, suggesting a global retinal inflammatory environment. There was no significant difference in microglial density within niches demonstrating that there is no active proliferation, migration, or apoptosis of microglia by D1. By D2, microglia in the NFL/GCL had significantly increased number of microglia processes, total process length, field area, and normalized volume, demonstrating an outgrowth of processes reminiscent of microglia in cell culture. This was similar in the IPL/INL, but not in the OPL. Iba1 = microglia specific marker, n = 4 retinas for all conditions, scale bars = 50 μm, * = p < 0.05, ** = p < 0.01, *** = p < 0.001, NS = non-significant.

FIGURE 2

Valproic acid reduces microglial morphological changes associated with pro-inflammatory phenotypes. (A) Retinal explants were exposed to LPS, VPA, or untreated and assessed at D1 in comparison to D0 controls. Individual IBA1 labeled microglia were reconstructed using Imaris software. (B) At D1 there was a significant reduction in the number of microglia processes, total process length, field area, and normalized volume, demonstrating that microglia became less complex and more voluminous, as occurs with a shift towards pro-inflammatory phenotypes. LPS treatment did not significantly enhance this morphological change. Addition of VPA significantly attenuated these morphological changes, as the number of processes remained higher, and volume was lower than in D1 controls. Total process length and field area were unchanged. There were no significant changes in microglial numbers in retinas exposed to VPA or LPS. These suggest a reduction in the magnitude of a pro-inflammatory shift and a retention of some morphological aspects associated with microglial support functions. n = 4 retinas for all conditions, scale bars = 50 μm, * = p < 0.05, ** = p < 0.01, *** = p < 0.001, NS = non-significant.

FIGURE 3

Valproic acid reduces astrocyte morphological changes associated with inflammation and gliosis. (A) Retinal explants were exposed to LPS, VPA, or untreated and assessed at D1 in comparison to D0 controls. Individual GFAP fibers were reconstructed to generate astrocyte networks using Imaris software. Areas enclosed by these fibers were generated in Imaris to measure space filling (individual areas are demonstrated by a unique color). (B) At D1 there was a significant increase in the number of filaments and the normalized volume, with no significant change in total filament length. This suggests a thickening of GFAP filaments and an increase in filament intersection. Supporting this, the number of enclosed areas was significantly increased, and these were significantly smaller (decreased enclosed area) suggesting an increase in filament volume and space filling by the astrocyte network, as is typical in gliosis. Addition of LPS to the media had no significant effect on astrocyte morphology compared to those in untreated D1 retinas. Addition of VPA significantly attenuated changes to the astrocyte network, with a reduction in the number of filaments, total filament length, and the normalized volume relative to untreated D1 retinas. The number of enclosed areas was significantly reduced, and the area of these was significantly higher than in untreated D1 retinas. There was no significant difference to D0 controls. Together, these suggest a protection against astrocyte morphological change and attenuation of pro-inflammatory and gliotic astrocyte changes. GFAP = astrocyte specific marker in NFL when labeled as flat mounts, n = 4 retinas for all conditions, scale bars = 20 μm, * = p < 0.05, ** = p < 0.01, *** = p < 0.001, NS = non-significant. Individual enclosed areas demonstrated by condition as violin plots.

FIGURE 4

Valproic acid attenuates pro-inflammatory cytokine responses from resident glia. Retinal explants were exposed to LPS, VPA, or untreated and assessed at D1 in comparison to D0 controls. Whole retinal homogenates were probed by cytokine array (111 cytokines and chemokines). (A) Hierarchical clustering (Pearson correlation) demonstrated that D0 and D1 retinas had discrete global cytokine profiles and that LPS treated retinas were most similar to D1 retinas. VPA treated retinas clustered with D0, demonstrating a protection against global cytokine changes, and a profile most similar to naïve retina. (B) Principle component (PC) analysis demonstrated a continuum of change from D0 retina, with D1 VPA least altered, followed by D1, and D1 LPS the most significantly changed. (C) This separation was predominantly driven by CCL5 and CXCL10 levels (PC 1) with small contributions from a number of cytokines across PC 2. (D) One-way ANOVA demonstrated 10 significantly altered cytokines between conditions (FDR <0.05). Volcano plots for individual comparisons demonstrate significant upregulation of five pro-inflammatory cytokines at D1, with a further increase in three pro-inflammatory cytokines under LPS treatment. VPA treatment significantly reduced six pro-inflammatory cytokines relative to untreated D1 retina (E) CCL5, CXCL10, CC122, and IL-1Α demonstrated discrete graded responses by condition where D1 > D0, LPS > D1, and VPA < D1 (and not significantly changed from D0). VPA also reduced ICAM1, but did not affect FGF1 or cystatin (C). LPS significantly enhanced IL12 over all conditions. Collectively, these demonstrate that LPS enhanced the pro-inflammatory response to injury in the retina even if this did not result in detectible morphological change over untreated D1 retinas. VPA demonstrated both a morphological and pro-inflammatory cytokine expression attenuation, but with a greater effect on cytokine profile. (F) To validate these findings against an in vivo model of RGC injury, significantly changed cytokines between D0 and D1 retina (ret, whole tissue, explant model) were compared to gene array data (ONH/Ret, whole tissue, DBA/2J) and RNA-sequencing data from sorted monocytes (ONH, monocytes, DBA/2J) in the DBA/2J mouse model of glaucoma. Common significant changes and the directionality of change are demonstrated as a heatmap. There was complete overlap with the optic nerve head (ONH) and 4/7 in the retina (ret) when compared to late-stage disease (stage 5 for ONH, four for retina). Monocytes do not differentially express any of these genes (monocytes 1 = enriched for pro-inflammatory changes, monocytes 2 = fewer pro-inflammatory changes), further supporting that the explant model can identify resident glia-specific changes. The % overlap in whole tissue was greatest for later disease stages, and greatest in the ONH, but there was overlap with early stages. This supports that the explant changes reflect real in vivo changes from early to late disease. n = 4 retinas for all conditions, FC = fold change, FDR = false discovery rate (q), * = p < 0.05, ** = p < 0.01, *** = p < 0.001, NS = non-significant. In B * = group centroid, and the cloud = 95% CI. In C * = 0% contribution to a principle component.