Estrogen receptor 1 activation suppresses microglia-derived Tnf-α signaling as a photoreceptor self-protection mechanism Authors: Yuting Li, Yadi Li, Jiarui Luo, Lan Wang, Qianlu Yang, Qianxi Yang, Cong Duan, Wenrong Xu, Yujie Dong, Lei Kong, Yan Li, Wenjia Zhang, Kangwei Jiao, Zhijian Zhao, Christina Schwarz, François Paquet-Durand, Junchuan Ye, Zhulin Hu, Jie Yan Published: 2025 ABSTRACT AbstractRetinitis pigmentosa (RP) is an inherited retinal degenerative disorder characterized by progressive photoreceptor loss and irreversible blindness. Increasing evidence implicates neuroinflammation as a contributor to photoreceptor degeneration extending beyond the initial genetic insult. Although estrogen has been reported to exert anti-inflammatory effects in the central nervous system, its role in RP remains controversial, with some studies suggesting a paradoxical exacerbation of retinal pathology. To address this discrepancy, we identify estrogen receptor alpha (Esr1) as a central immunoregulatory hub in RP. Transcriptomic analyses of rd1 and rd10 revealed upregulation of estrogen-responsive and inflammatory pathways, with Esr1 expression markedly elevated during degeneration. TUNEL assays demonstrated that systemic estradiol (E2) exerted divergent effects, protective in rd1 yet deleterious in rd10, whereas selective pharmacological activation of Esr1 with propyl pyrazole triol (PPT) consistently reduced photoreceptor death, preserved dark-adapted ERG responses, and downregulated inflammatory mediators including Tnf-α, Cx3cl1/Cx3cr1, Cd68, and Iba1. Mechanistically, Esr1 activation repressed microglial Tnf transcription and disrupted a self-sustaining Cx3cl1/Cx3cr1–Tnf-α signaling loop driving microglial recruitment, activation and neurotoxicity in the outer nuclear layer (ONL). Targeted interventions confirmed tumor necrosis factor receptor 1 (Tnfr1) as the principal mediator of Tnf-induced photoreceptor death: selective inhibition with R7050 conferred superior protection compared with broad-spectrum Tnf-α inhibitors (etanercept, infliximab). Cx3cr1 blockade likewise suppressed microglial activation and improved visual outcomes. Collectively, our findings establish Esr1 activation as not merely an external intervention but the amplification of an intrinsic self-protective program, positioning Esr1, Tnfr1, and Cx3cr1 as actionable therapeutic targets to suppress neuroinflammation and preserve vision in RP.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12974-025-03669-z. Introduction Inherited retinal diseases (IRDs) encompass a spectrum of genetic disorders that drive progressive photoreceptor loss and ultimately blindness [1]. Retinitis pigmentosa (RP), the most prevalent and currently untreatable IRDs, affects over two million individuals worldwide (≈1 in 4,000) [1]. RP typically presents with nyctalopia due to primary rod photoreceptor degeneration, followed by concentric constriction of the visual field (“tunnel vision”) driven by secondary cone degeneration, a consequence of rod death [1]. In the absence of disease-modifying therapies, delineating novel molecular targets is critical for halting or delaying RP progression. The rd1 and rd10 mouse strains represent well-established RP models, each harboring Pde6b mutations encoding the β-subunit of rod cyclic guanosine monophosphate (cGMP) phosphodiesterase (PDE6β) [2]. Although both models share the same affected gene, they represent distinct genotypes: rd1 carries a nonsense mutation in exon 7 (Tyr347Stop; C to A transversion) that completely abolishes PDE6β activity, resulting in rapid and early-onset photoreceptor degeneration, whereas rd10 harbors a missense mutation in exon 13 (Arg560Cys; C to T transition) that partially retains enzymatic function, leading to a slower, progressive course of retinal degeneration [3–5]. These two models are therefore complementary, reflecting different severities of Pde6b-associated RP. Loss of PDE6β function impairs cGMP hydrolysis, causing pathological cGMP accumulation [6], aberrant Ca2⁺influx [7], and activation of non-apoptotic photoreceptor death pathways, notably Parthanatos [8]. In rd1 mice, degeneration begins around postnatal day 8 (P8) and peaks near P13 [9], whereas in rd10 the onset is delayed (P16–P20) and the degeneration more protracted, peaking at P21-P25 [10]. Pharmacological PDE6 inhibition by zaprinast mimics Pde6b-driven pathology by selectively blocking cGMP hydrolysis [11], thereby serving as a chemical surrogate for modeling cGMP-induced photoreceptor death [12]. However, interventions targeting cGMP accumulation [13], Ca2+ overload [7], or Parthanatos [14] have proven insufficient to fully halt photoreceptor degeneration, suggesting the existence of additional, yet unidentified pathogenic pathways and therapeutic targets. Among these, microglia-driven neuroinflammation has emerged as a non-cell-autonomous contributor to retinal degeneration [15, 16]. Retinal microglia, the primary immune sentinels, transition from a homeostatic surveillant state to an activated phenotype, migrating to degenerating photoreceptors and releasing pro-inflammatory cytokines such as tumor necrosis factor alpha (TNF-α) [16, 17]. TNF-α engages its receptor TNFR1 to initiate neuroinflammatory signaling [18], and the resulting sustained inflammatory milieu further accelerates neuronal loss [16, 17]. Microglial homeostasis is tightly regulated by chemokine (C-X3-C motif) ligand 1 (CX3CL1), and its receptor CX3CR1, where neuron- and photoreceptor-derived CX3CL1 maintains microglial quiescence [19, 20]. Disruption of this axis during degeneration likely amplifies microglial activation and exacerbates photoreceptor death [20–22]. Given the role of neuroinflammation in retinal degeneration, hormonal modulation has attracted attention as a potential strategy to restrain microglia-mediated damage [23]. Estrogen, particularly 17β-estradiol (E2), is well recognized for its neuroprotective and anti-inflammatory properties within the central nervous system [23, 24]. In the retina, estrogen has been proposed as an endogenous self-protective mechanism engaged in response to stress [25]; however, the precise pathways by which it confers such protection remain poorly defined. Among the four natural isoforms (estrone, estradiol, estriol, estetrol), E2 is the most potent and biologically active, acting primarily through estrogen receptor α (ERα/ESR1), with additional contributions from ERβ (ESR2) and the G-protein-coupled estrogen receptor (GPER) [23, 26]. Despite this protective reputation, accumulating evidence indicates that estrogen’s actions in neurodegeneration are context-dependent, with some studies reporting exacerbation of pathology under certain conditions [27, 28], raising controversy over its therapeutic value in IRDs. To resolve this controversy, We integrated bulk and single-cell transcriptomic analyses across two Pde6b-associated retinal degeneration models, including bulk RNA-seq from rd1 (P13) and rd10 (P21) retinas, and single-cell RNA-seq from rd10 (P19). Bulk RNA-seq of rd1 and rd10 retinas revealed progressive activation of estrogen-responsive and inflammatory pathways, along with upregulation of microglial markers including Cx3cl1, Cx3cr1, Cd68 and Aif1. Single-cell profiling confirmed microglial expansion, localized Tnf expression, and loss of estrogen-responsive gene signatures in mutant photoreceptor that were otherwise enriched in wt counterparts. CellChat analysis further identified disease-specific microglia-rod interactions that were absent in wt but established in Pde6b mutants. Immunostaining demonstrated markedly enhanced Esr1 expression in degenerating retinas, and co-immunostaining demonstrated Esr1 expression within microglia, indicating that microglia can directly respond to Esr1 activation. Functionally, while E2 conferred neuroprotection in rd1, it aggravated degeneration in rd10, highlighting genotype-dependent effects. In contrast, selective Esr1 activation uniformly attenuated photoreceptor loss in both models by suppressing Tnf and downstream inflammatory cascades. Blockade of Tnfr1 provided more robust neuroprotection than pan-Tnf-α inhibition, and microglia-specific Cx3cr1 inhibition confirmed microglia as the principal source of Tnf-α. In vivo, intravitreal administration of Esr1 agonists, Tnf-α antagonists, or Cx3cr1 inhibitors preserved retinal structure and function in both genetic and chemically induced degeneration models. Collectively, these findings position Esr1 as an intrinsic protective regulator in photoreceptor degeneration and identify selective Esr1 activation as a potent, inflammation-targeted, genotype-independent therapeutic strategy that surpasses systemic estrogen in the treatment of IRDs. Materials and methods Animals Experiments were conducted using C3H Pde6brd1/rd1 (rd1), C57BL/6 J Pde6brd10/rd10 (rd10), congenic C3H Pde6b+/+ (C3H wild-type), and C57BL/6 J Pde6b+/+ (C57 wild-type) mice. Thus, rd1 mice and their congenic C3H wild-type counterparts share a C3H background, whereas rd10 mice and their corresponding C57 wild-type controls share a C57BL/6 J background. All procedures were performed in accordance with the guidelines of the Laboratory Animal Welfare and Ethics Committee of Yunnan University. Both male and female mice were included, and rd1 and rd10 animals were each compared with their respective congenic wild-type (wt) controls. Ex vivo retinal explant culture Retinal explants from rd1, C3H, rd10, and C57 mice were prepared and maintained under serum-free, hormone-controlled conditions and prepared as described [29]. Retinas were collected at P5 for rd1 and C3H, and at P12 for rd10 and C57, reflecting the earlier onset of photoreceptor degeneration in rd1 and the more mature cellular composition of rd10 retinas, in which photoreceptors and synaptic layers are fully developed [10, 30]. To minimize inter-animal variability, the two retinas from each animal were assigned in a parallel manner, one to the experimental group and the other to the control group, thereby allowing within-animal comparisons and maximizing independent observations. After 48 h in culture, explants were treated with 10 μM E2 [31] (HY-B0141; MedChemExpress, Sollentuna, Sweden), 10 μM E2 plus 10 nM AZD9496 [32] (HY-12870; MedChemExpress), 15 nM PPT [33] (HY-100689; MedChemExpress), 0.63 nM Etanercept [34] (ETN, HY-108847; MedChemExpress), 6.7 μM Infliximab [35] (IFX, HY-P9970; MedChemExpress), 65 μM R7050 [36] (HY-110203; MedChemExpress), or 4 μM AZD8797 [37] (HY-13848; MedChemExpress). Compound concentrations were based on retinal dose–response curves (Figure S1A-G), which defined the most effective and non-toxic doses subsequently used for in vivo validation. The same concentrations were also tested on wild-type retinal explants to assess potential drug toxicity (Figure S2A). Cultures were terminated at P11 (rd1, C3H) or P20 (rd10, C57) by fixation with 4% paraformaldehyde. Explants were embedded in Tissue-Tek (Sakura Finetek, Alphen aan den Rijn, The Netherlands) and cryosectioned (12 μm) using a CryoStar NX50 (Thermo Fisher Scientific, Runcorn, UK). In vivo intravitreal injections Single intravitreal injections were performed at P16 in rd10 and C57 mice as previously described [38]. Mice were anesthetized by intraperitoneal injection of tribromoethanol (83.3 mg/kg), and pupils were dilated with Tropicamide-Phenylephrine Eye Drops (Santen Pharmaceutical Co., Ltd., Osaka, Japan). Under a ZEISS Stemi 305 stereomicroscope (ZEISS, Oberkochen, Germany), A 33-gauge needle (Hamilton Bonaduz AG, Bonaduz, Switzerland) was inserted from the limbus with a 45° injection angle into the vitreous. The direction and location of the needle was monitored through the microscope (ZEISS). 0.1µL of solution was gently delivered into the vitreous cavity using a microsyringe (Hamilton Bonaduz AG). In rd10 mice, the left eye received 0.1μL of PBS (pH 7.4) containing each compound dissolved in DMSO, resulting in the following final intraocular DMSO concentrations derived solely from the solubilization of each drug: 0.002% DMSO for PPT, 0.00018% for ETN, 0.01% for IFX, 0.1% for R7050, and 0.008% for AZD8797. The right eye received 0.1μL of PBS containing the exact same DMSO concentration as the corresponding treatment eye, serving as the sham control; therefore, the DMSO content was matched between the two eyes for each compound, eliminating DMSO as a confounding factor. In C57 mice, 20 mM zaprinast [12] (ZAP, HY-B1816; MedChemExpress) was simultaneously co-administered with each compound in the left eye at a final intraocular concentration of 400 μM (0.1μL injection; dose selected based on Figure S1H). The corresponding right eye again received a sham injection containing 0.4% DMSO, which represents the final intraocular DMSO concentration needed to dissolve zaprinast. Final intraocular concentrations were calculated assuming a 5μL ocular volume in mice [39, 40], consistent with those used in retinal explant cultures. All animals were euthanized at P23 for histological analyses, corresponding to 7 days post-injection. Dark-adapted electroretinography (ERG) Dark-adapted ERGs were performed using the Celeris system (Diagnosys LLC, Lowell, MA, USA). Intravitreal injections with single dose were administered in rd10 and C57BL/6 J mice starting from P16 and ending at P21, coinciding with the critical window of functional decline before near-complete loss of retinal responses at ~ P23 [30, 39]. Histological analyses were extended to P23 to capture cumulative structural degeneration beyond the functional floor effect [30, 39]. For all experiments, rd10 and ZAP-treated C57 mice served as paired controls, with the right eye as control and the left as experimental. Mice were dark-adapted for ≥ 12 h before testing. Anesthesia was induced by intraperitoneal injection of tribromoethanol (83.3 mg/kg), and pupils were dilated with Tropicamide-Phenylephrine Eye Drops (Santen Pharmaceutical Co.). dark-adapted responses were elicited by 3 cd·s/m2 flash stimuli, and signals were recorded within a 300 ms window. A-wave amplitudes were measured from baseline to the negative trough, and b-wave amplitudes from the a-wave trough to the subsequent positive peak. For comparative consistency, all a-wave values were converted to absolute values, thereby representing negative amplitudes as positive numbers. Importantly, all ERG traces displayed in the figures represent the averaged waveform derived from all acquired traces for each condition, rather than representative individual traces. All recordings were performed under dim red light, and data acquisition and analysis were conducted using the integrated Celeris platform. Single-cell RNA-seq (scRNA-seq) data processing and analysis Raw sequencing data (eyeballs collected from P19 rd10 and C57) were processed using Cell Ranger (v9.0.0) with the Mus musculus GRCm39 reference genome, and UMI count matrices were analyzed in Seurat (v5.0.3). Cells with < 200 detected genes, < 1000 UMIs, log10 (Genes/UMI) < 0.7, > 30% mitochondrial gene content, and > 5% hemoglobin gene content were excluded. Doublets were removed using DoubletFinder. Data were normalized, integrated using Harmony for batch correction, scaled, and clustered using PCA and the Louvain algorithm, and visualized by t-SNE. Cell types were annotated with canonical markers. UCell scoring was applied to assess pathway activity, and t-SNE plots were generated to visualize differential expression across conditions. Cell–cell communication analysis was performed using CellChat (v1.6.1) with the mouse ligand-receptor database [40]. Bulk RNA sequencing (RNA-seq) Total RNA was isolated from pooled rd10 and congenic wt mouse eyeballs (both eyes per animal) or from PPT-treated rd1 and wt retinal explants. Total RNA was extracted using standard protocols and quality-checked with the RNA Nano 6000 Assay Kit (Agilent Technologies, USA) and NanoPhotometer® spectrophotometer (IMPLEN, USA). Sequencing libraries were constructed using the NEBNext® Ultra™ RNA Library Prep Kit for Illumina® (NEB, USA) and sequenced on an Illumina NovaSeq platform to generate 150 bp paired-end reads. The mRNA expression comparison between rd1 and wt from P7 to P21 employed datasets downloaded from the GSE62020 [41]. GSEA algorithm GSEA was performed using GSEA software (v.4.3.3) in conjunction with the Molecular Signatures Database (v.7.4). One thousand permutations were used. Hub-Gene identification Differential genes between rd1/rd10 and congenic wt retinas were called in R 4.5.1 (Bioconductor) with cut-offs fold change (FC) > 1.5 and p < 0.05. The resulting protein–protein interaction network was retrieved from STRING and analysed in Cytoscape 3.10. Six algorithms (Degree, MNC, MCC, Closeness, Stress, EPC) ranked nodes; genes present in the top 10 list of every metric were designated as hubs. TUNEL staining TUNEL staining was performed using a commercial kit (Beyotime, Jiangsu, China) to detect dying cells. Retinal explants and eye cup sections were dried, stored at − 20 °C, rehydrated in 0.1 M PBS, and permeabilized with 0.5% Triton X-100 for 5 min. Sections were washed in PBS and incubated with 100μL of TUNEL solution (enzyme:labeling solution, 1:9) in a light-protected humid chamber at 37 °C for 60 min. After additional PBS washes, sections were counterstained with DAPI (G1407; Servicebio, Wuhan, China) and imaged using confocal microscopy. Immunohistochemistry Sections were rehydrated in PBS for 15 min and blocked with 10% normal goat serum, 1% bovine serum albumin, and PBS with 0.3% Tween-20 for 1 h. Primary antibodies: rabbit anti-Esr1 (1:200; YT1634; Immunoway, California, USA), rabbit anti-Cd68 (1:200; BA3638; BOSTER, California, USA), rabbit anti-Iba1 (1:400; ym8165; Immunoway), mouse anti-Iba1 (1:200; ab283319; Abcam), rabbit anti-Cx3cl1 (1:200; YT5354; Immunoway), rabbit anti-Cx3cr1 (1:200; YT5112; Immunoway), and rabbit anti-Tnf-α (1:200; YM8306; Immunoway), rabbit anti-Tnfr1(1:50; YT4687; Immunoway), Mouse anti-Glutamine Synthetase (1:1000; MAB302; Abcam), were diluted in blocking solution and incubated overnight at 4 °C. For co-immunostaining, combinations of primary antibodies raised in different host species (e.g., rabbit + mouse) were mixed in the same blocking buffer and applied simultaneously to the sections to ensure comparable staining conditions. After primary incubation, sections were rinsed in PBS (3 × 10 min) and incubated with appropriately species-matched secondary antibodies, including goat anti-rabbit Alexa Fluor 488 (1:350; A11034; Molecular Probes, Oregon, USA) and goat anti-mouse Alexa Fluor 647 (1:500; A32728; Molecular Probes), for 1 h at room temperature in the dark. Secondary antibodies were selected to ensure minimal spectral overlap and to enable clear distinction of different target antigens. After an additional PBS rinse (3 × 10 min), sections were mounted with DAPI-containing mounting medium (Servicebio) for nuclear visualization. Negative controls were processed in parallel following the exact same staining and washing procedures but without the addition of primary antibodies, incubating sections only with the secondary antibody, goat anti-rabbit Alexa Fluor 488 (1:350; A11034; Molecular Probes). These controls were used to assess background fluorescence and to exclude non-specific or autofluorescent signals. Microscopy and image analysis Retinal explant and in vivo samples were imaged using a Zeiss LSM 900 confocal microscope (Carl Zeiss, Oberkochen, Germany). Z-stack and tile scans were acquired at 20 × magnification using identical imaging parameters across all experimental and control groups to ensure quantitative comparability. Specifically, laser power, gain, pinhole size, and exposure times were kept constant for all samples within each staining set. To minimize regional variability, all images were acquired from anatomically matched retinal regions using identical acquisition settings. Cryosections were captured across four optical planes, and images were processed uniformly using ZEN (Blue edition) software. For co-immunostaining experiments, high-resolution acquisition was performed using a 63 × oil-immersion objective. This allowed precise visualization of subcellular localization and colocalization of target proteins. The same laser intensities, detector gain, and channel settings were applied across all samples to avoid cross-channel interference and to ensure accurate assessment of colocalized signals. Sequential scanning was used when necessary to minimize spectral overlap between fluorophores. Quantification of ONL cells and positive signals To quantify ONL cell density, six randomly selected, non-overlapping rectangular regions within the ONL were analyzed for each section. These regions were not fixed in size but were chosen to adequately sample different ONL locations while avoiding overlapping fields. Within each region, the number of DAPI-stained nuclei was manually counted, and ONL cell density was calculated as: The mean ONL cell density derived from the six regions (≈0.05 cells/µm2) was then used to estimate the total number of ONL cells in each section by multiplying this density by the ONL area measured using ZEN software. Because these values are density-based estimates rather than raw counts, they are presented as non-integer values in the supplemental tables. For positive cell quantification, the number of marker-positive cells within the full ONL region was manually counted across three images per section and at least three biological replicates per group. The reported values represent the mean counts across replicates, and thus also appear as non-integer values. The percentage of positive cells was calculated as: For Cx3cr1 and Cx3cl1 quantification, fluorescence signals were correlated with DAPI-stained nuclei within the ONL, and positive counts were assigned only when a DAPI-positive nucleus was clearly covered by Cx3cr1 or Cx3cl1 signal, rather than based solely on signal morphology. Fluorescence intensities were quantified using Zeiss ZEN 3.0 software under identical acquisition and processing conditions for all samples to eliminate variation introduced by exposure time, laser power, or detector sensitivity. Statistical analysis and software use Two-way comparisons were analyzed using paired t-tests or student’s t-test. Multiple comparisons were assessed by one-way ANOVA with Tukey’s post hoc test. Analyses were performed using GraphPad Prism 8 (GraphPad Software, La Jolla, CA, USA), with p < 0.05 considered significant. Significance levels were denoted as follows: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), and p < 0.0001 (****). Data in Figs. 9 and S1 were normalized by linear scaling χ_scaled = (χ – χ_min)/(χ_max – χ_min) using SPSS Statistics 26 (IBM, Armonk, NY, USA). Spearman correlation analyses were conducted in R (v4.0.1). Figures were generated using Photoshop 2025 and Illustrator 2025 (Adobe, San Jose, CA, USA), and Fig. 10 was created with BioRender.com. Inflammatory marker dynamics and their correlation with photoreceptor degeneration across models. A-C Comparison of Inflammatory markers across different experimental treatments in ex vivo rd1 (A), in vivo rd10 (B), and wt subjected to ZAP-induced retinal degeneration (C). Normalized cell counts positive for TUNEL (magenta), Iba1 (orange), Cd68 (cyan), Cx3cr1 (red), Cx3cl1 (yellow), and Tnf-α (green) were shown across experimental groups. D Radar plots illustrating Spearman correlation between each inflammatory marker and TUNEL-positive cells in rd1 (red), rd10 (blue), and ZAP-treated wt (green) models. E Global Spearman correlation matrix among inflammatory markers and TUNEL across all degeneration models. Circle asterisks denote statistical significance; numeric values in squares indicate correlation coefficients (R) Schematic model of Esr1-mediated suppression of inflammatory signaling in Pde6b-associated retinal degeneration. In the healthy retina, microglia remain quiescent and are largely excluded from the outer nuclear layer (ONL). In Pde6b-mutant retinas, disease-associated stress induces upregulation of the chemokine Cx3cl1, which engages its receptor Cx3cr1 on microglia to drive their activation and migration into the ONL. Concomitantly, Esr1 expression is upregulated in degenerating retinas, suggesting a self-protective mechanism. Activated microglia transcriptionally upregulate Tnf-α and release it onto Müller cells, which in turn amplify microglial recruitment and activation. Simultaneously, Tnf-α engages Tnf receptors, particularly Tnfr1, on photoreceptors or Müller cells, initiating cell death cascades via recruitment of Tnfr1-associated death domain protein (TRADD) and subsequent assembly of a death-inducing signaling complex (DISC) comprising TRAF2, RIPK, and FADD. These signaling events promote direct and indirect photoreceptor degeneration while reinforcing a feed-forward loop that sustains Cx3cl1 expression and chronic neuroinflammation. Selective Esr1 activation by PPT interrupts this pathogenic cascade by repressing Tnf-α transcription in microglia, thereby disrupting microglia-driven neurotoxicity and preserving retinal integrity. The drugs used in this study and their targets are indicated. Red colour indicates destructive processes, while green labeled compounds promote photoreceptor survival Results Esr1 activation as an intrinsic self-protective mechanism to attenuate photoreceptor degeneration and preserves retinal function To assess the involvement of estrogen signaling in Pde6b-associated photoreceptor degeneration, we performed gene set enrichment analysis (GSEA) on RNA-seq datasets from rd1 (P13) and rd10 (P21) retinas. Both models exhibited significant enrichment of estrogen-responsive pathways (Fig. 1A, Figure S2B), implicating a potential role for estrogen signaling in disease progression. scRNA-seq profiling further demonstrated that estrogen-response gene set was enriched in wt photoreceptors but was diminished in rd10 photoreceptors, indicating a loss of photoreceptor-intrinsic estrogen responsiveness. Conversely, wt microglia exhibited minimal estrogen-response gene set, whereas rd10 microglia showed a clear induction of this pathway. Müller cells similarly retained strong estrogen-response signatures (Figure S2C), suggesting that estrogen signaling shifts away from photoreceptors and toward microglia during degeneration. Immunofluorescence staining corroborated bulk RNA-seq datasets, showing robust upregulation of Esr1 protein throughout the degenerating retina, with predominant localization in the inner segment (IS), inner nuclear layer (INL), and ganglion cell layer (GCL) in both rd1 and rd10 models compared to wt controls (Fig. 1B,C, Figure S2D,E, Table S1A). Co-immunostaining with Iba1 and glutamine synthetase (GS) further demonstrated that Esr1 expression extended beyond photoreceptors to reactive microglia and Müller glia (Fig. 1D,E, Figure S2F,G), highlighting a broader involvement of estrogen signaling within the glial compartment during retinal degeneration. Esr1 as an intrinsic self-protective mechanism. A GSEA of RNA-seq data from rd10 retinas at postnatal day 21 (P21) revealed significant positive enrichment of Estrogen-related pathways. B-C Immunofluorescence staining for Esr1 (green) demonstrated increased signal intensity in the entire retina of rd10 mice compared to congenic wt controls and negative staining controls. Esr1 was predominantly localized to IS, INL, and GCL. Negative control rd10 P21: n = 3; wt P21: 3; rd10 P21: 3. D-E Co-immunostaining of Iba1 (red) and Esr1 (green) further confirmed Esr1 expression in reactive microglia. Panel (D) shows 20 × confocal imaging, revealing widespread Iba1 microglia exhibiting overlapping Esr1 immunoreactivity. Panel (E) shows high-resolution 63 × imaging, demonstrating clear colocalization of Iba1 and Esr1 at the single-cell level, with both markers displaying shared positive signal within individual microglial somata and processes. F-G TUNEL staining (magenta) of ex vivo rd1 retinal explants showed that treatment with 10 µM E2, E2 + 10 nM Esr1 antagonist AZD9496, or 15 nM Esr1 agonist PPT markedly reduced ONL TUNEL positivity compared with untreated rd1 explants, consistent with the quantification in (H). Untr. wt: 3; Untr. rd1: 12; E2 rd1: 6; E2 + AZD9496 rd1: 4; PPT rd1: 4. H-I TUNEL staining and quantification of ex vivo rd10 retinal explants under identical treatment conditions revealed that E2 increased ONL TUNEL positivity, and E2 combined with the Esr1 antagonist AZD9496 further exacerbated photoreceptor cell death. In contrast, PPT treatment markedly reduced ONL cell death. J-K In vivo single intravitreal injection of PPT from P16-P23 significantly decreased ONL TUNEL positivity in rd10 mice. Untr. rd10: 13; PPT rd10: 13. L-M The same treatment improved dark-adapted ERG a-wave amplitudes, indicating partial functional rescue of photoreceptor responses. Scatter plots depict the percentage of positive cells in the ONL or the total fluorescence intensity; each dot corresponds to one independent retinal sample. Error bars: SD; significance levels: ns = p > 0.05; * = p < 0.05; ** = p < 0.01; *** = p < 0.001; **** = p < 0.0001. OS = outer segment, IS = inner segment, ONL = outer nuclear layer, INL = inner nuclear layer, IPL = inner plexiform layer, GCL = ganglion cell layer, NFL = nerve fiber layer; scale bar = 50 µm. Dashed lines indicate untreated mutant or wt baseline levels. DAPI (grey) was used as nuclear counterstain To functionally probe estrogen signaling, we employed ex vivo retinal explant cultures treated with 17β-estradiol (E2), E2 plus the Esr1 antagonist AZD9496, or the Esr1-selective agonist PPT. TUNEL assays confirmed robust outer nuclear layer (ONL) cell death in rd1 and rd10 explants relative to wt (Fig. 1F-I, Table S1B). E2 administration significantly attenuated ONL cell death in rd1 but, conversely, exacerbated degeneration in rd10. The addition of AZD9496 abrogated E2-mediated neuroprotection in rd1 and further aggravated photoreceptor loss in rd10, supporting a genotype-dependent and context-specific effect of endogenous estrogen signaling. In contrast, selective Esr1 activation via propyl pyrazole triol (PPT) consistently reduced ONL TUNEL positivity in both models, underscoring the genotype-independent neuroprotective role of direct Esr1 stimulation (Fig. 1F-I, Table S1B). Dose–response curves were generated to select appropriate treatment concentrations (Figures S1A-C). In vivo intravitreal PPT administration in rd10 mice, with contralateral eyes as controls, significantly reduced ONL TUNEL positivity (Fig. 1J,K, Table S1C) and improved dark-adapted ERG responses, with prominent recovery of a-wave amplitudes (Fig. 1L,M, Table S1D) and b-wave amplitudes (Figure S3A). Pharmacological induction of Pde6b-associated pathology by ZAP in wt mice increased ONL TUNEL positivity (Figure S2H,I, Table S1C) and suppressed ERG amplitudes relative to untreated eyes (Figure S2J,K, Table S1D, Figure S3B). A dose–response curve for the effects of ZAP was compiled (Figure S1H) and a concentration of 400 µM was chosen for further experiments. In vivo PPT co-administration attenuated ZAP-induced photoreceptor degeneration (Figure S2L,M, Table S1C) and restored ERG responses, particularly both a-wave (Figure S2N,O, Table S1D) and b-wave amplitudes (Figure S3C). Together, these data indicate that targeted activation of Esr1, rather than non-specific estrogen stimulation, confers robust neuroprotection and preserves retinal function across both genetic and pharmacological models of retinal degeneration. These findings further support the notion that Esr1 signaling serves as an intrinsic, protective mechanism in the degenerating retina and represents a promising therapeutic target for retinal dystrophies. Suppression of Tnf-α as the downstream effect of Esr1 activation To delineate the downstream pathways regulated by Esr1, we first analyzed rd1 retinal transcriptomes at P13. GSEA revealed positive enrichment of Tnf-associated and inflammatory pathways (Figure S4A,E), suggesting activation of proinflammatory signaling during degeneration. Immunostaining confirmed elevated Tnf-α expression in both rd1 and rd10 retinas compared with their respective wt controls (Figure S4B–D), supporting a conserved Tnf-α-driven neuroinflammatory component across Pde6b-mutant models. Furthermore, co-immunostaining with Tnf-α and GS showed that Tnf-α upregulation extended to Müller glia (Figure S4G,H), implicating Müller-glial inflammatory activation as a key component of the Esr1-linked downstream signaling landscape. Following ex vivo PPT treatment, GSEA demonstrated negative enrichment of Tnf-related and inflammatory pathways (Fig. 2A, S4F), indicating suppression of this inflammatory axis upon Esr1 activation. Consistently, CytoHubba analysis of PPT-treated rd1 transcriptomes identified Tnf as a shared hub gene across six algorithms (Fig. 2B), with significantly reduced Tnf expression (Fig. 2C, Table S2A). Esr1 activation suppresses Tnf-α signaling to protect photoreceptors. A GSEA of rd1 retinal RNA-seq from explants showed negative enrichment of Tnf associated pathways after PPT treatment. B CytoHubba analysis identified Tnf as a common hub gene post-PPT. C Tnf expression (FPKM) is significantly reduced by PPT. D-K Immunostaining for Tnf-α (green) reveals elevated signal in rd1 and rd10 versus wt; PPT reduces Tnf-α in ex vivo explant and in vivo models, Untr. wt: n = 4; Untr. rd1: 8; PPT rd1: 8; Untr. rd10: 9; PPT rd10: 9; D-G. ZAP elevated Tnf-α in wt, which was reversed by PPT, Untr. wt: 10; ZAP wt: 10; ZAP wt: 8; ZAP + PPT wt: 8; H–K. L-M TUNEL assays (magenta) with DAPI (grey) as nuclear counterstain in ex vivo rd1 explants showed decreased ONL cell death with Tnf-α blockade (ETN, IFX) and further protection by the Tnf receptor antagonist R7050. Untr. wt: 3; Untr. rd1: 10; ETN rd1: 4; IFX rd1: 4; R7050 rd1: 4. N–O In in vivo rd10 mice, intravitreal ETN injection significantly reduced ONL TUNEL positivity compared with untreated controls. Untr. rd10: 5; ETN rd10: 5. P-Q The same treatment restored dark-adapted ERG a-wave amplitudes, indicating functional recovery of photoreceptor responses. Scatter plots depict the percentage of positive cells in the ONL or the total fluorescence intensity; each dot corresponds to one independent retinal sample. Error bars: SD; significance levels: * = p < 0.05; ** = p < 0.01; *** = p < 0.001; **** = p < 0.0001. OS = outer segment, IS = inner segment, ONL = outer nuclear layer, INL = inner nuclear layer, IPL = inner plexiform layer, GCL = ganglion cell layer, NFL = nerve fiber layer; scale bar = 50 µm. Dashed lines indicate untreated levels Immunostaining corroborated these transcriptomic finding, showing low Tnf-α signal in wt, robust elevation in rd1 and rd10 retina, particularly Müller cells, and significant reductions following PPT treatment in both ex vivo and in vivo models (Fig. 2D-G, Table S2B). In wt, ZAP markedly increased Tnf-α expression, while co-treatment with PPT reversed this effect (Fig. 2H-K, Table S2B). To verify the functional role of Tnf-α, rd1 ex vivo explant cultures were treated with the Tnf-α inhibitor ETN, the monoclonal antibody IFX, or the Tnf receptor antagonist R7050. Dose–response curves were generated to select appropriate treatment concentrations (Figures S1D-F). TUNEL assays demonstrated significantly fewer ONL dying cells in wt versus rd1 (Fig. 2L,M, Table S2C). ETN and IFX both reduced photoreceptor death, whereas R7050 yielded the most pronounced protection, surpassing untreated rd1 and single-agent treatments (Fig. 2L,M, Table S2C). Taken together, these findings indicate that Esr1 activation mitigates photoreceptor degeneration in Pde6b mutant models by suppressing Tnf-α-mediated neuroinflammation. Inhibition of Tnf-α/Tnfr1 signaling protects photoreceptors and restores retinal function To validate the pathogenic role of Tnf-α signaling in vivo, rd10 and ZAP-induced wt mice were treated with either the soluble Tnf-α inhibitor etanercept (ETN), the monoclonal antibody infliximab (IFX), or the Tnfr1-selective antagonist R7050. In rd10 mice, ETN treatment markedly reduced TUNEL-positive cells in ONL (Fig. 2N,O, Table S2D) and significantly improved darked-adapted ERG responses, showing preservation of both a-wave and b-wave amplitudes (Fig. 2P,Q; Figure S3D; Table S2E), indicative of photoreceptor functional preservation. Similarly, ETN attenuated ZAP-induced photoreceptor loss in wt mice (Fig. 3A,B, Table S2D) and restored ERG amplitudes (Fig. 3C), with particularly pronounced recovery of both a-wave (Fig. 3D; Table S2E) and b-wave components (Figure S3E). In vivo Tnf-α and Tnfr blockade protect photoreceptors and restore retinal function. A-B ETN treatment significantly reduced ZAP-induced ONL TUNEL positivity in wt mice, ZAP wt: n = 6; ZAP + ETN wt: 6; indicating attenuated photoreceptor cell death. C-D The same treatment restored dark-adapted ERG responses and a-wave amplitudes, reflecting improved photoreceptor function in wt. E–F In rd10 mice, IFX treatment lowered the percentage of ONL TUNEL-positive cells compared with untreated controls, Untr. rd10: 6; IFX rd10: 6. G-H Correspondingly, IFX improved ERG responses, demonstrating functional rescue of retinal activity. I-J In ZAP-induced wt degeneration, IFX also mitigated ONL photoreceptor death, ZAP wt: 5; ZAP + IFX wt: 5. K-L The treatment rescued ERG function and preserved retinal responses in wt. M–N In rd10 mice, the Tnfr1 antagonist R7050 markedly decreased ONL TUNEL positivity, Untr. rd10: 6; R7050 rd10: 6; indicating reduced photoreceptor loss. O-P R7050 further enhanced dark-adapted ERG responses, suggesting robust functional recovery. Q-R In ZAP-treated wt retinas, R7050 further suppressed ONL degeneration, ZAP wt: 6; ZAP + R7050 wt: 6. S-T The same treatment restored dark-adapted ERG responses and a-wave amplitudes, confirming effective protection of photoreceptor function. Scatter plots depict the percentage of positive cells in the ONL; each dot corresponds to one independent retinal sample. Error bars: SD; significance levels: * = p < 0.05; ** = p < 0.01; **** = p < 0.0001. ONL = outer nuclear layer, INL = inner nuclear layer, GCL = ganglion cell layer, scale bar = 50 µm; TUNEL assays (magenta) with DAPI (grey) as nuclear counterstain IFX exerted comparable neuroprotective effects. In rd10 mice, IFX significantly reduced ONL cell death (Fig. 3E,F, Table S2D) and enhanced ERG a-wave amplitudes (Fig. 3G,H, Table S2E). In the ZAP-induced wt degeneration model, IFX preserved retinal structure and photoreceptor function, as evidenced by reduced TUNEL staining (Fig. 3I,J, Table S2D) and improved a-wave responses (Fig. 3K,L, Table S2E). However, IFX failed to restore b-wave amplitudes in both rd10 and ZAP-treated wt mice (Figure S3F,G), suggesting incomplete recovery of inner retinal signaling. Among the three agents, R7050 provided the most robust protection. R7050 profoundly suppressed photoreceptor cell death in rd10 (Fig. 3M,N, Table S2D), significantly restored both a- and b-wave ERG responses (Fig. 3O,P, Figure S3H, Table S2E), and attenuated ZAP-induced retinal degeneration in wt relative to ZAP-treated controls (Fig. 3Q-T, Figure S3I, Table S2D,E). scRNA-seq revealed Tnfrsf1a (Tnfr1) expression was nearly absent in rod photoreceptors of wt mice, while a minor fraction of rd10 rods exhibited low but detectable Tnfrsf1a transcripts (Figure S5A). Temporal analysis further confirmed Tnfrsf1a upregulation during retinal degeneration in both rd1 and rd10 models (Figure S5B,C). Immunohistochemical analysis demonstrated Tnfr1 protein distribution throughout the retina, with prominent localization in Müller glia and photoreceptor inner segments (Figure S5D). Notably, overall Tnfr1 protein levels did not differ between rd1, rd10, and their respective congenic wt controls (Figure S5E,F). Collectively, these in vivo findings, supported by transcriptomic and immunohistochemical data-identify Tnf-α/Tnfr1 signaling as a central driver of photoreceptor degeneration in Pde6b-associated retinal disease. Pharmacological disruption of this axis, particularly via selective Tnfr1 inhibition, effectively preserves retinal structure and visual function, offering a promising therapeutic strategy for inherited retinal degenerations. Microglia-derived Tnf-α as a putative driver of photoreceptor degeneration To identify the cellular source of Tnf-α in degenerating retinas, we performed scRNA-seq in wt and rd10 mice. Compared to wt retinas, the rd10 retina exhibited a substantial expansion of the microglial population (Fig. 4A). Transcriptomic mapping revealed that Tnf expression was predominantly restricted to microglia, with minimal expression detected in other retinal cell types (Fig. 4B). Cell–cell communication analysis using CellChat further demonstrated a lack of microglia-photoreceptor interactions in wt mice (Fig. 4C), whereas such crosstalk was prominently established in the degenerating rd10 retina (Fig. 4D), indicating disease-specific engagement. Microglia-derived Tnf-α mediates photoreceptor degeneration in Pde6b-mutant retinas. A Single-cell RNA sequencing (scRNA-seq) of whole eyeballs from rd10 and congenic wild-type (wt) mice revealed an expanded microglial population in rd10 retinas. B Feature plot showing Tnf transcripts predominantly localized to microglia rather than other retinal cell types. C, D Cell–cell communication analysis (CellChat) demonstrating absence of microglia-rod photoreceptor interactions in wt retinas (C), but emergence of such interactions in the Pde6b-mutant context (D). E–F CytoHubba analysis of P13 rd1 transcriptomes identified Cd68 as a central hub gene with markedly elevated expression, consistent with microglial overactivation during retinal degeneration. G–H Similarly, CytoHubba analysis of P21 rd10 transcriptomes also pinpointed Cd68 as a shared hub gene showing significant upregulation, reinforcing the conserved involvement of activated microglia in Pde6b-associated pathology. significance levels: * = p < 0.05; *** = p < 0.001 To elucidate microglial drivers of degeneration, we next applied CytoHubba analysis across transcriptomic datasets from rd1 and rd10 retinas at their respective peak phases of photoreceptor loss. Cd68 emerged as a convergent hub gene in both models, ranking top across six centrality algorithms and exhibiting significant upregulation in rd1 and rd10 mice (Fig. 4E-H, Table S3A). Transcriptomic analyses further confirmed that Cd68 expression was predominantly confined to microglia (Figure S5G), consistent with the robust microglial activation observed at the histological and transcriptomic levels. Importantly, co-immunostaining of Iba1 and Cd68 (Figure S6A,B) demonstrated substantial spatial overlap between these two markers, indicating that the majority of Cd68-positive cells in the degenerating retina are likely microglia. Immunostaining and RNA-seq further confirmed progressive increases in Iba1 (Aif1) and Cd68 expression from P9 to P21 in rd1, and from P13 to P30 in rd10, relative to their respective wt controls (Fig. 5A-F; Figure S6C-F; Figure S7A-D). Temporal gene and protein expression of Iba1 and Cd68 during retinal degeneration in rd1, rd10, and their congenic wt controls. A-C Immunostaining and gene expression analyses revealed a time-dependent upregulation of Iba1 (orange, Aif1) and Cd68 (cyan) in rd1 retinas (P9-P21) compared with age-matched wt controls, indicating progressive microglial activation throughout degeneration. D-F Immunostaining and gene expression analyses in rd10 retinas (P13-P30) similarly demonstrated gradual elevation of Iba1 and Cd68 expression relative to wt, paralleling disease progression and confirming sustained microglial activation in degeneration. Protein expression (solid lines) is plotted against the left Y-axis, while gene expression (dashed lines) corresponds to the right Y-axis in dual-axis plots.For each indicated time point, three biological replicates (n = 3) were analyzed. Error bars: SD; ONL = outer nuclear layer, INL = inner nuclear layer, GCL = ganglion cell layer; scale bar = 50 µm. DAPI (grey) was used as nuclear counterstain Collectively, these findings highlight microglia as the principal source of Tnf-α in the degenerating retina and implicate microglia-photoreceptor interactions as a possible pathological axis in Pde6b-associated retinal degeneration. Cx3cr1/Cx3cl1 signaling drives microglial activation and photoreceptor degeneration Among chemokine pathways, the Cx3cl1-Cx3cr1 axis emerged as a central regulator of microglial activation during retinal degeneration [42]. In both rd1 and rd10 models of RP, transcript and protein levels of Cx3cl1 and its receptor Cx3cr1 were significantly upregulated during the critical windows of photoreceptor loss (Fig. 6A-F; Figure S8A-D; Figure S9A-D), implicating this pathway in disease progression. Temporal gene and protein expression of Cx3cr1 and Cx3cl1 during retinal degeneration in rd1, rd10, and their congenic wt controls. A-C Immunostaining and gene expression analyses showed a time-dependent upregulation of Cx3cl1 (yellow) and Cx3cr1 (red) in rd1 retinas (P9-P21) compared with age-matched wt controls, indicating progressive activation of the Cx3cl1-Cx3cr1 axis during degeneration. D-F Similar analyses in rd10 retinas (P13-P30) revealed a gradual increase in both Cx3cl1 and Cx3cr1 expression, consistent with sustained microglial activation and chronic inflammatory signaling. Protein expression (solid lines) is plotted against the left Y-axis, while gene expression (dashed lines) corresponds to the right Y-axis in dual-axis plots.For each indicated time point, three biological replicates (n = 3) were analyzed. Error bars: SD; ONL = outer nuclear layer, INL = inner nuclear layer, GCL = ganglion cell layer, scale bar = 50 µm; DAPI (grey) was used as nuclear counterstain To assess whether microglial activation via this axis causally contributes to photoreceptor degeneration, we administered the Cx3cr1 antagonist AZD8797 was applied in ex vivo rd1 retinal explants (Figue 7A-B), where a dose response analysis established 4 µM as the optimal concentration (Figure S1G). The same compound was subsequently evaluated in in vivo rd10 mice and in the ZAP-induced photoreceptor injury model in wt retinas. In all contexts, AZD8797 treatment significantly reduced ONL TUNEL-positive cells (Fig. 7C-D, G-H, Table S3B,C), indicating attenuated photoreceptor cell death. Functional assessment via dark-adapted ERG revealed that AZD8797 robustly restored retinal responses, particularly improving a- and b- wave amplitudes (Fig. 7E-F, I-J, Table S3D, Figure S3J, K). Cx3cl1-Cx3cr1 signaling mediates microglia-driven photoreceptor degeneration in Pde6b-mutant models. A-B Cx3cr1 inhibitor AZD8797 reduced ONL TUNEL-positive cells in ex vivo rd1 explants. Dashed lines indicate untreated levels. Untr. wt: n = 3; Untr. rd1: 9; AZD8797 rd1: 15. C-D In rd10 in vivo, AZD8797 treatment markedly decreased ONL cell death relative to untreated rd10 controls, confirming its neuroprotective efficacy. Untr. rd10: 6; R7050 rd10: 6. E–F The same treatment restored dark-adapted ERG responses and enhanced a-wave amplitudes, indicating functional recovery of photoreceptor integrity in rd10 mice. G-H In ZAP-induced wt degeneration, in vivo AZD8797 significantly reduced ONL photoreceptor loss, demonstrating that pharmacological inhibition of Cx3cr1 also protects against chemically induced retinal injury. ZAP wt: 6; ZAP + AZD8797 wt: 6. I-J AZD8797 further improved ERG function and a-wave amplitudes in the ZAP-treated wt model, underscoring its broad therapeutic potential in both genetic and pharmacological degeneration paradigms. Scatter plots depict the percentage of positive cells in the ONL; each dot corresponds to one independent retinal sample. Error bars: SD; significance levels: * = p < 0.05; ** = p < 0.01; **** = p < 0.0001. ONL = outer nuclear layer, INL = inner nuclear layer, GCL = ganglion cell layer, scale bar = 50 µm; TUNEL assays (magenta) with DAPI (grey) as nuclear counterstain Together, these results establish Cx3cl1-Cx3cr1 signaling as a key mediator of microglia-driven photoreceptor degeneration in Pde6b-associated RP. Pharmacological blockade of this axis using AZD8797 confers significant neuroprotection, preserving both retinal structure and visual function across multiple degenerative contexts. Targeting Esr1, Tnf-α signaling, and Cx3cr1 suppresses microglial recruitment, Cx3cl1/Cx3cr1 signaling, and neuroinflammation We next examined the impact of Esr1 activation, Tnf-α inhibition, and Cx3cr1 blockade on microglial recruitment and neuroinflammatory signaling across rd1 explants, rd10 retinas, and ZAP-induced wt degeneration. Treatment with the Esr1 agonist PPT, Tnf-α inhibitors (ETN, IFX), the Tnfr1 antagonist R7050, or the Cx3cr1 inhibitor AZD8797 markedly suppressed microglial recruitment, as evidenced by reduced Iba1 expression within the ONL in all models (Fig. 8A,B, Table S4A, Figure S10A,E, Table S5A, Figure S11A,B, Table S6A). Cd68 expression followed a similar trend: minimal in untreated wt, robustly induced during degeneration, and significantly attenuated upon intervention (Fig. 8C,D, Table S4B, Figure S10B,F, Table S5B, Figure S11C,D, Table S6B). In vivo Esr1 activation, Tnf-α inhibition, and Cx3cr1 blockade suppress microglial recruitment and Cx3cl1/Cx3cr1 signaling in rd10. A-B Immunostaining of Iba1 revealed robust microglial recruitment into the ONL in rd10 in vivo, which was significantly reduced by Esr1 activation (PPT), Tnf-α blockade (ETN, IFX), Tnfr1 antagonism (R7050), and Cx3cr1 inhibition (AZD8797). Untr. wt: n = 3; Untr. rd10: 32; PPT rd10: 6; ETN rd10: 5; IFX rd10: 6; R7050 rd10: 6; AZD8797 rd10: 6. C-D Cd68 Immunostaining followed a similar trend: minimal in untreated wt, upregulated in rd10, and attenuated following intervention. Untr. wt: 3; Untr. rd10: 27; PPT rd10: 5; ETN rd10: 5; IFX rd10: 6; R7050 rd10: 6; AZD8797 rd10: 6. E-F Immunostaining for Cx3cr1 (red) demonstrated that its expression was barely detectable in wt retinas, markedly elevated in rd10, and suppressed by all interventions, indicating effective blockade of microglial receptor signaling. Untr. wt: 3; Untr. rd10: 29; PPT rd10: 6; ETN rd10: 5; IFX rd10: 6; R7050 rd10: 6; AZD8797 rd10: 6. G-H Immunostaining for Cx3cl1 (yellow) showed a similar expression pattern, with low basal levels in wt, strong upregulation in rd10, and consistent reduction following Esr1, Tnf-α, Tnfr1, or Cx3cr1-targeted treatments, indicating effective suppression of chemokine-mediated microglial activation. Untr. wt: 3; Untr. rd10: 29; PPT rd10: 6; ETN rd10: 5; IFX rd10: 6; R7050 rd10: 6; AZD8797 rd10: 6. Scatter plots depict the percentage of positive cells in the ONL; each dot corresponds to one independent retinal sample. Error bars: SD; significance levels: * = p < 0.05; ** = p < 0.01; *** = p < 0.001; **** = p < 0.0001. ONL = outer nuclear layer, INL = inner nuclear layer, GCL = ganglion cell layer, scale bar = 50 µm. Dashed lines indicate expression levels in untreated rd10 controls. DAPI (grey) was used as nuclear counterstain. Statistical testing: one-way ANOVA with Tukey’s multiple comparison post hoc test Consistent with these findings,Cx3cl1 and Cx3cr1, which were barely detectable in untreated wt retinas, became strongly upregulated during degeneration in rd1, rd10, and ZAP-treated models (Fig. 8E-H; Table S4C-D; Figure S10C,D,G,H; Table S5C-D; Figure S11E-H; Table S6C-D). All interventions, PPT, ETN, IFX, R7050, and AZD8797, consistently reduced Cx3cl1 and Cx3cr1 expression. In parallel, Tnf-α expression mirrored the dynamics of microglial and chemokine signaling. Across ex vivo rd1 explants, in vivo rd10, and ZAP-induced wt models, ETN, IFX, R7050, and AZD8797 each produced significant and comparable suppression of Tnf-α (Figure S12A-F, Table S7A-C). Taken together, these findings demonstrate that Esr1 activation, Tnf-α/Tnfr1 inhibition, and Cx3cr1 blockade converge on a shared mechanism that dampens microglial recruitment, restrains Cx3cl1/Cx3cr1 signaling, and attenuates Tnf-α-driven neuroinflammation across multiple degenerative contexts. This convergence not only highlights a common anti-inflammatory pathway linking Esr1 activation with cytokine and chemokine modulation but also indicates effective suppression of chemokine-mediated microglial activation as a key mechanism underlying the observed neuroprotection. Inflammatory signatures correlate with photoreceptor degeneration and are modulated by Esr1, Tnf-α, and Cx3cr1 inhibition To enable cross-model comparisons, expression data were normalized by linear scaling, setting the minimum and maximum values to zero and one, respectively. In untreated wt retinas, both TUNEL positivity and inflammatory marker expression were consistently low relative to rd1, rd10, and ZAP-treated wt models (Fig. 9A-C). Among inflammatory proteins, Tnf-α exhibited relatively elevated baseline expression even in wt retinas. Pharmacological interventions targeting Esr1 (PPT), Tnf-α signaling (ETN, IFX, R7050), and Cx3cr1 (AZD8797) significantly reduced photoreceptor cell death and suppressed expression of Iba1, Cd68, Cx3cr1, Cx3cl1, and Tnf-α across all models. Despite this, residual Tnf-α levels remained comparatively higher in rd10 and ZAP-treated wt retinas (Fig. 9B-C), suggesting persistent or model-specific inflammatory resistance. Radar plot based Spearman analyses revealed strong positive correlations between photoreceptor death (TUNEL positivity) and inflammatory markers (Iba1, Cd68, Cx3cr1, Cx3cl1, Tnf-α) in all models examined (Fig. 9D). A global Spearman analysis integrating data across rd1, rd10, and ZAP-treated wt further confirmed robust intercorrelations among these markers (Fig. 9E), underscoring their co-regulation and central involvement in neuroinflammatory signaling. Together, these findings implicate Tnf-α, Cx3cl1-Cx3cr1, and microglial markers (Cd68, Iba1) as tightly linked inflammatory mediators that collectively drive photoreceptor degeneration, and highlight their therapeutic modulation via Esr1, Tnf-α, or Cx3cr1 inhibition. Discussion Our study establishes Esr1 as a pivotal neuroprotective modulator in Pde6b-related retinal degeneration, acting through suppression of microglial activation and downstream pro-inflammatory signaling. We observed robust upregulation of Esr1 in degenerating retinas, and selective pharmacological activation of Esr1 consistently mitigated photoreceptor loss, preserved dark-adapted ERG responses, and downregulated inflammatory mediators, including Tnf-α, Cx3cl1/Cx3cr1, Cd68, and Iba1, across rd1, rd10, and pharmacologically induced models. Conversely, blockade of Esr1 abolished these protective effects, underscoring its central role in maintaining photoreceptor survival. Interestingly, systemic administration of E2 produced divergent outcomes: neuroprotective in rd1 yet deleterious in rd10, highlighting the genotype-dependent complexity of endogenous estrogen signaling. By contrast, selective Esr1 activation with PPT provided consistent, genotype-independent neuroprotection, offering a more predictable and clinically translatable strategy than global estrogen supplementation. Mechanistically, Esr1 activation transcriptionally repressed Tnf-α, curtailed Cx3cl1/Cx3cr1-driven microglial recruitment, and dampened microglial activation, thereby limiting retinal neuroinflammation and preserving photoreceptor integrity. Moreover, selective inhibition of Tnfr1 with R7050 significantly outperformed broad-spectrum Tnf-α inhibitors (ETN and IFX) in reducing ONL cell death and maintaining retinal function. Collectively, these findings identify Esr1 as an intrinsic protective mechanism in degenerating photoreceptors and nominate downstream effectors, particularly Tnfr1 and Cx3cr1, as actionable therapeutic targets for retinal neurodegeneration (Fig. 10). Diverse effects of E2 in rd1 and rd10 mouse Estrogen is widely recognized for its neuroprotective effects. In our study, E2 reduced photoreceptor death in rd1 explants but exacerbated degeneration in rd10 explants, consistent with prior observations that E2 can have either protective or detrimental actions depending on biological context [27, 28]. Although differences in oxidative stress levels between rd1 and rd10 retinas have been documented [43, 44], and E2 has been shown to exert redox-dependent antioxidant or pro-oxidant effects in other systems [45–49], these observations alone are insufficient to fully explain the divergent outcomes observed here. At present, the mechanistic basis underlying the opposite E2 responses in rd1 and rd10 remains unclear, and future work will be required to determine whether redox status or additional genotype-specific factors contribute to this difference. Esr1 as an intrinsic self-protective mechanism in degenerating retina RNA-seq profiling of rd1 and rd10 retinas revealed positive enrichment of estrogen-related pathways, consistent with previous reports implicating endogenous estrogen as neuroprotective mechanisms in retinal stress responses [25]. However, the precise mechanisms underlying this self-protective pathway remain poorly defined. Althought the protective effects of estrogen have been described as genotype-dependent [27, 28], our data suggest that Esr1 represents a conserved axis of intrinsic retinal defense. scRNA-seq provided additional context. In wt retinas, photoreceptors showed strong enrichment of estrogen-response gene signatures, whereas this enrichment appeared diminished in rd10 photoreceptors, suggesting a reduction in photoreceptor-intrinsic estrogen responsiveness during degeneration. Conversely, while wt microglia exhibited minimal activation of estrogen-response pathways, rd10 microglia displayed a noticeable induction of this signature. These observations may reflect a shift in estrogen-responsive signaling from photoreceptors toward microglia as degeneration progresses. Immunohistochemical analyses showing increased Esr1 expression in rd1 and rd10 retinas further support the idea that Esr1 upregulation may form part of a broader compensatory response to photoreceptor stress. Pharmacological modulation of this pathway provided functional validation. Co-administration of E2 with the Esr1 antagonist AZD9496 abolished the protective effect of E2 in rd1 and exacerbated degeneration in rd10, whereas treatment with the ERα-selective agonist PPT significantly delayed photoreceptor death. These bidirectional outcomes clearly indicate that Esr1 signaling mediates estrogen-dependent neuroprotection. Mechanistically, AZD9496 is a potent, non-steroidal, orally bioavailable selective estrogen receptor degrader and antagonist, displaying high potency for Esr1 (IC50 = 0.82 nM for binding, 0.14 nM for downregulation, and 0.28 nM for antagonism) and markedly weaker activity toward other nuclear receptors, including androgen receptor (AR, IC50 = 30 μM), glucocorticoid receptor (GR, IC50 = 9.2 μM), and progesterone receptor (PR, IC50 = 0.54 μM), confirming its high receptor selectivity [32, 50]. In contrast, PPT is an Esr1-preferring agonist with approximately 400-fold greater affinity for Esr1 than Esr2 [51, 52]. The opposing outcomes produced by these two pharmacological probes, an Esr1 antagonist versus an Esr1 agonist, provide receptor-specific evidence that activation of Esr1 confers protection, whereas its inhibition deprives the retina of estrogen-mediated resilience. Collectively, these findings suggest that Esr1 upregulation represents an endogenous self-protective attempt by photoreceptors, Müller glia, or microglia to mitigate inflammation and preserve retinal integrity. Pharmacological reinforcement of this pathway can therefore be viewed as amplifying an intrinsic survival program, transforming Esr1 signaling from a passive compensatory response into a therapeutically actionable neuroprotective axis. In this context, Esr1 activation should be considered not merely as an exogenous intervention but as the potentiation of a built-in, self-protective defense against photoreceptor degeneration. The rescue effect of Esr1 activation likely arises from targeting retinal microglia Unlike the variable effects observed with systemic E2, selective Esr1 blockade consistently exacerbated photoreceptor loss in both rd1 and rd10 explants, whereas Esr1 activation reproducibly attenuated cell death across all models, underscoring its robust neuroprotective effect. scRNA-seq analysis revealed that Tnf transcripts were confined to microglia, while immunostaining localized Tnf-α protein predominantly to Müller glia, supporting a paracrine model in which microglia-derived Tnf-α acts on Müller cells that express high levels of TNFR1 [53, 54]. Our immunohistochemical analyses confirmed Tnfr1 expression in Müller glia, consistent with their role as amplifiers of neuroinflammation through gliotic transformation, impaired glutamate clearance, and secondary release of chemokines such as Cx3cl1 [54, 55]. Although Esr1 staining was broadly distributed across the retina, with prominent enrichment in the IS, INL, and GCL, our co-immunostaining demonstrated that Esr1 is expressed in both microglia and Müller glia, indicating that both glial populations possess the molecular machinery to respond to Esr1 agonism. However, scRNA-seq analysis provided decisive cellular resolution: Tnf expression was restricted almost exclusively to microglia, with negligible expression detected in Müller glia or any other retinal cell type. This transcriptomic specificity, combined with the observation that PPT treatment resulted in a marked reduction in Tnf transcription, strongly suggests that the primary effector population of Esr1 agonism is microglia rather than Müller glia. In this context, Esr1 activation likely suppresses microglial inflammatory output at the transcriptional level, while any effects on Müller cells are secondary to reduced upstream Tnf-α signaling. Transcriptomic profiling following PPT treatment revealed significant repression of Tnf, consistent with prior work showing that estrogen-bound Esr1 can tether NF-κB p65 and recruit transcriptional corepressors to suppress Tnf transcription [56]. Correspondingly, PPT-mediated Esr1 activation markedly reduced Tnf-α protein levels in rd1, rd10, and ZAP-induced wt models, alongside parallel reductions in Cd68, Iba1, and Cx3cl1/Cx3cr1 expression. Taken together, these findings support a model in which microglia constitute the principal, and transcriptionally exclusive, source of Tnf-α in degenerating retinas, and Esr1 activation exerts its photoreceptor-protective effect primarily by directly targeting microglial Esr1 to suppress Tnf transcription. While Esr1 signaling in Müller cells may modulate downstream inflammatory amplification, the convergence of scRNA-seq specificity, immunohistochemical colocalization, and transcriptional repression following PPT treatment strongly indicates that microglial Esr1 represents the central therapeutic node for interrupting the Tnf-α, driven degenerative cascade. Inhibition of Tnf-α signaling, particularly Tnfr1, mitigates photoreceptor degeneration TNF-α is a central mediator of neuroinflammation [57], predominantly produced by activated macrophages and microglia [58, 59]. It is initially synthesized as a type II transmembrane protein (tmTNF) and subsequently cleaved by TNF-α-converting enzyme (TACE/ADAM17) into its soluble form (sTNF) [60]. While sTNF selectively activates TNFR1, tmTNF can engage both TNFR1 and TNFR2 [58]. TNFR1 is strongly linked to cell death through the recruitment of TNFR1-associated death domain protein (TRADD), which assembles with TNFR-associated factor 2 (TRAF2), receptor-interacting protein kinase (RIPK), and Fas-associated death domain protein (FADD) to form the Death-Inducing Signaling Complex (DISC), initiating programmed cell death [61]. Our data revealed that Tnf-α expression remained elevated compared to other inflammatory markers (including Iba1, Cd68, Cx3cr1, and Cx3cl1), even in untreated wt retinas. This preferential upregulation of Tnf-α may be attributed to the photoreceptor outer segment (OS) phagocytosis [62]. Photoreceptor OS discs are constantly synthesized at the base and shed at the distal tip, a dynamic process critical for sustaining phototransduction efficiency and removing oxidatively damaged or dysfunctional components [62–64]. However, this high rate of metabolic and structural turnover may inherently generate low-grade, non-pathological inflammatory or stress signals [65], thereby contributing to baseline Tnf-α expression, even in the absence of overt degeneration. In our rd1, rd10, and ZAP-induced degeneration models, Esr1 activation consistently suppressed Tnf-α expression, and pharmacological inhibition of Tnf-α signaling further confirmed its pathogenic role. ETN and IFX neutralized both sTNF and tmTNF [60], thereby broadly inhibiting TNFR1 and TNFR2 signaling [58]. In contrast, R7050, a selective TNFR1 antagonist [66], conferred the most pronounced protective effect on photoreceptor survival. This aligns with the notion that TNFR1, which contains a death domain, primarily mediates inflammatory and cytotoxic responses [58, 61]. Indeed, our immunostaining confirmed Tnfr1 expression also within photoreceptor inner segments, while scRNA-seq revealed that Tnfrsf1a is absent in wt rods but aberrantly induced in Pde6b-mutant rods, suggesting that Tnfr1 may directly trigger photoreceptor death. By contrast, TNFR2 could promote pro-survival signaling. Indeed, TNFR2-mediated Etk phosphorylation has been shown to partially activate VEGFR2 [67], contributing to cell survival and proliferation [67, 68]. Thus, the non-selective blockade of Tnfr2 by ETN and IFX may partially offset the neuroprotective benefits of Tnfr1 inhibition. Collectively, these results highlight Tnfr1 as the principal effector of Tnf-α-induced photoreceptor degeneration in Pde6b-deficient retinas. Microglial activation and Cx3cl1-Cx3cr1 signaling drive progressive photoreceptor degeneration In rd1, rd10, and ZAP-induced models of retinal degeneration, neuronal stress triggers a sustained inflammatory response characterized by microglial activation. Among the key mediators, the CX3CL1-CX3CR1 signaling axis emerged as a central regulator [42]. Stress-induced activation of NF-κB and AP-1 promotes CX3CL1 transcription and accumulation of its membrane-bound precursor [69]. As pathology progresses, proteolytic cleavage by Ca2+/ROS-activated ADAM10, MAPK/TNF-driven ADAM17 (TACE), and microglial cathepsin S releases soluble CX3CL1 [70–72]. Acting as a damage-associated molecular pattern (DAMP), soluble CX3CL1 engages microglial CX3CR1 to promote recruitment and activation within the photoreceptor [21]. While transient fractalkine signaling may facilitate debris clearance [73], sustained CX3CL1-CX3CR1 engagement reinforces microglial reactivity and drives chronic neuroinflammation [21, 22]. We observed persistent upregulation of microglial markers (Cd68, Iba1) and fractalkine axis components (Cx3cl1, Cx3cr1) in both rd1 and rd10 retinas, supporting the role of prolonged microglial activation in disease progression. Notably, pharmacological inhibition of Cx3cr1 with AZD8797 robustly reduced ONL cell death across models, confirming this pathway as a critical effector of microglia-mediated neurotoxicity and a viable target for therapeutic intervention. Cone degeneration may also benefit from targeting Esr1 signaling Given the overlap in inflammatory and degenerative pathways among photoreceptor subtypes, it is possible that cone cells may also benefit from Esr1-mediated modulation of neuroinflammation. The cpfl1 (cone photoreceptor function loss 1) mouse model represents a well-established system of cone-specific degeneration. The cpfl1 mutation affects Pde6c, encoding the cone-specific α′-subunit of cGMP phosphodiesterase (Pde6c), leading to enzymatic inactivation and pathologic cGMP accumulation within cones [74, 75]. Sustained elevation of cGMP leads to persistent activation of cGMP-gated channels, excessive Ca2+ influx, metabolic stress, and activation of downstream degenerative cascades involving PKG, HDAC, PARP, and calpain [76, 77], mechanisms that parallel those described in Pde6b-mutant rod degeneration [8]. Importantly, HDAC inhibition significantly preserves cone survival in cpfl1 mice [78], indicating shared molecular vulnerabilities between rod and cone photoreceptors. Although direct evidence is currently lacking, the mechanistic convergence suggests that Esr1 activation, through limiting Tnf expression and reducing microglial reactivity, could potentially provide benefit in cpfl1 or other cone-dominant degeneration models. Targeting the Esr1-Tnf-microglia axis may therefore represent a promising, but still exploratory, strategy for extending neuroprotection to cone degeneration. Study limitations and future directions It should be noted that both rd1 and rd10 mice carry mutations in Pde6b and may therefore share overlapping downstream death mechanisms during degeneration. Whether Esr1-targeted interventions can provide broader photoreceptor protection beyond PDE6-deficient retinopathies remains an open question. Furthermore, the present study did not include genetic manipulation of Esr1, microglia-specific ablation, or behavioral assessments of visual function, all of which would provide additional mechanistic clarity and strengthen causal inference. Future work using genetically unrelated models of inherited retinal degeneration, combined with targeted genetic approaches and microglia-specific interventions, as well as complementary vision-based assays, will be essential to determine both the generalizability and the functional significance of the protective effects observed here. Conclusion The role of estrogen signaling in IRDs has been debated due to its context-dependent and often opposing effects. Here, we identify Esr1 as a central immunoregulatory hub in Pde6b-associated retinal degeneration. Esr1 is robustly upregulated in degenerating retinas, likely representing an intrinsic self-protective response, and its selective pharmacological activation confers consistent, genotype-independent neuroprotection across rd1, rd10, and ZAP-induced degeneration models, whereas Esr1 inhibition abolishes these benefits. In contrast, systemic E2 treatment elicits divergent, genotype-specific outcomes, underscoring the complexity of global estrogen signaling and highlighting the therapeutic advantage of receptor-selective modulation. Mechanistically, Esr1 activation interrupts a self-sustaining Cx3cl1/Cx3cr1-Tnf-α inflammatory loop that drives microglial recruitment, activation, and ONL damage. Interventions at critical nodes, including Esr1 activation (PPT), Tnf-α neutralization (ETN, IFX), Tnfr1-specific inhibition (R7050), and Cx3cr1 blockade (AZD8797), consistently suppressed inflammatory mediators, dampened microglial toxicity, preserved retinal architecture, and improved visual function, with Tnfr1 inhibition yielding the strongest protection. These findings establish Esr1 activation not merely as an external intervention but as the amplification of an endogenous defense circuit in the degenerating retina, and support a rational therapeutic framework in which Esr1 and its downstream inflammatory effectors are targeted to suppress microglia-driven neurotoxicity and safeguard photoreceptor survival in Pde6b-associated IRDs. Supplementary Information Supplementary Material 1. Acknowledgements The authors would like to thank Huiqin Li (from Yunnan Infectious Disease Hospital) for excellent assistance. Authors’ contributions Conceptualization, J.Y., ZL.H., JC.Y. and F.P.-D.; methodology, J.Y., L.W., YT.L., YD.L. and JR.L.; software, QL.Y., X.Y., YJ.D. and J.Y.; validation, J.Y. and L.W.; formal analysis, QL.Y., Y.L., WJ.Z. and QX.Y.; investigation, J.Y., YT.L., YD.L., JR.L. and L.W; data curation, J.Y., ZL.H., KW.J. WR.X. and JC.Y.; writing-original draft preparation, J.Y. and QL.Y.; writing-review and editing, F.P-D., C.S., D.C. and ZZ.Z.; visualization, J.Y. and QL.Y.; supervision, J.Y., ZL.H. and JC.Y.; project administration, J.Y., ZL.H. and JC.Y.; funding acquisition, J.Y., ZL.H. and JC.Y. All authors have read and agreed to the published version of the manuscript. Funding This research was funded by the National Natural Science Foundation of China (No.82360604), the Medical Leading Talents Training Program of Yunnan Provincial Health Commission (No.L-2019029), the Yunnan Provincial Health Commission Clinical Medicine Center Research Project (No.2024YNLCYXZX0326, No.2024YNLCYXZX0339), the Yunnan Science and Technology Plan Project (No.202401AT070454), the Project of Yunnan Fundamental research Projects (No.202401AT070453), the Open Project of National Clinical key Specialty (No.ZKF2024046), the Yunnan Fundamental Research Kunming Medical University Projects (No.202501AY070001-217, No.202501AY070001-209), the Yunnan University Medical Research Foundation (YDYXJJ2025-0056), the Yunnan Clinical Research Center for Infectious Disease (AIDS) (No. 202405AJ310002). Data availability No datasets were generated or analysed during the current study. Declarations Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Yuting Li, Yadi Li and Jiarui Luo shared first authors who contributed equally to this work. Junchuan Ye, Zhulin Hu and Jie Yan shared corresponding authors who contributed equally to this work. Contributor Information Junchuan Ye, Email: yejunchuan@outlook.com. Zhulin Hu, Email: hzl7755@outlook.com. Jie Yan, Email: jieyanbiologist@outlook.com. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement No datasets were generated or analysed during the current study.