All animal procedures were approved by the Institutional Animal Care and Use Committee of the University of Murcia (Murcia, Spain) and performed according to our institutional guidelines (approved protocol A13210201) and ARRIVE guidelines.

Two months old C57BL/6J male and female mice were obtained from the breeding colony of the University of Murcia or purchased from Envigo (Barcelona, Spain). Animals were kept at the University of Murcia animal housing facilities in temperature and light controlled rooms (12 h light/dark cycles) with food (pellet 12 mm, Teklad Global Diet^®, Inōtiv, Mucedola, Milán, Italy) and water administered ad libitum.

Optical coherence tomography was conducted with general inhalational anesthesia utilizing 3% isoflurane (Abbott Laboratories, Abbott Park, IL) at a rate of 1.5 L/min oxygen, employing a calibrated precision vaporizer. For electroretinogram analyses, animals underwent anesthesia through an intraperitoneal injection of a combination of ketamine (60 mg/kg, Ketolar, Parke-Davies, S.L., Barcelona, Spain) and xylazine (10 mg/kg, Rompún, Bayer S.A., Barcelona, Spain). Following anesthesia, a protective ointment (Tobrex; Alcon S.A., Barcelona, Spain) was applied to the eyes to prevent corneal desiccation. The euthanasia process involved an intraperitoneal injection of a lethal dose of sodium pentobarbital (Dolethal, Vetoquinol; Especialidades Veterinarias, S.A., Alcobendas, Madrid, Spain).

Figure 1 summarises the experimental groups and analyses. In vivo functional and anatomical analyses (ERG and OCT) were performed longitudinally before (pre- or baseline) and after the procedures, and animals sacrificed at different time points for anatomical analyses on flat-mounts. New groups were done for cytometry and ELISA assays. Intact animals were used as controls for the total number of RGCs, morphology of resting microglia, flow cytometry, and cytokine profiling analyses. The number of samples per assay, sex and time point is shown in the scatter dot plots.

LPS [dissolved in saline. E. coli O111: B4, 437627 Millipore, Merck Life Science S.L.U. Madrid, Spain] and TNFR1 antagonist [12 mg/kg i.p. in 5% of DMSO-saline; R7050, Tocris Bioscience; Bio-Techne R&D Systems, Madrid, Spain], as previously published (43), were both injected intraperitoneally in a final volume of 200 µL.

P2X7R antagonist [ITH15004] was injected subcutaneously in a final volume of 100 µL of 10% DMSO/saline. ITH15004 (2-[6-chloro-9H-purin-9-yl)-1-(2,4-dichlorophenyl]ethan-1-one) is a non-nucleotide purine derivative that has shown selective P2X7 receptors blocking properties, and it has been synthesized accordingly to the method recently described (44).

For LPS and the P2X7R antagonist, we did a dose titration (see results) and the selected final concentrations were 5 mg/kg for LPS and 15 mg/kg for the P2X7R antagonist.

The full-field electroretinogram (ERG) was conducted following previously published methods (45–47). Briefly, mice underwent 12 hours of dark adaptation, followed by anesthesia, and dilation of both eyes with a topical mydriatic (Tropicamide 1%; Alcon-Cusí, S.A., Barcelona, Spain). Scotopic and photopic responses were simultaneously recorded using Burian-Allen corneal bipolar electrodes. Methylcellulose (Methocel^® 2%; Novartis Laboratories CIBA Vision, Annonay, France) was applied between the cornea and electrodes to enhance signal conductivity. The reference electrode was positioned in the mouth, and a needle at the tail base served as the ground electrode. Scotopic responses reflecting retinal ganglion cell (RGC)-mediated activity were recorded with light flashes ranging from -4.4 log cd·s/m², while rod-mediated responses were recorded at -2.5 log cd·s/m². Mixed responses (a- and b-waves) were recorded at 0.5 log cd·s/m². For cone-mediated responses, a flash of 0.5 log cd·s/m² was applied on a 30 cd/m² rod-saturated background. The electrical signals were digitized at 20 KHz using a Power Lab data acquisition board (AD Instruments, Chalgrove, UK), and standard ERG waves were analyzed in accordance with the International Society for Clinical Electrophysiology of Vision (ISCEV).

Retinas were longitudinally examined using SD-OCT (Spectralis; Heidelberg Engineering, Heidelberg, Germany) adapted with a commercially available 78-D double aspheric fundus lens (Volk Optical, Inc., Mentor, OH, USA) positioned in front of the camera unit, as outlined in a previous publication (48). Following anesthesia, tropicamide eye drops (Tropicamide 1%; Alcon-Cusí, S.A. Barcelona, Spain) were administered to induce mydriasis in both eyes. Imaging was conducted with the proprietary software package (Eye Explorer, version 3.2.1.0; Heidelberg Engineering). A raster scan comprising 25 equally spaced horizontal B-scans was used to capture retinal images. Manual measurements of total, inner, and outer retinal thickness were taken near the optic nerve head (0.4 mm) and at a 1-mm distance from it, always within central sections spanning the optic disc. Subsequently, the software calculated the volume of the central retina, with manual alignment of inner and outer retinal limits in the 25 sections acquired per retina. Hyperreflective puncta in the vitreous were manually quantified in 3 central OCT sections/animal.

Euthanized animals underwent transcardial perfusion with a 0.9% saline solution, followed by 4% paraformaldehyde in 0.1 M phosphate buffer solution. Flat-mounted retinas were prepared according to established procedures (49). Immunodetection procedures followed previous protocols (50). Flat-mounted retinas underwent triple immunodetection with mouse anti-Brn3a (1:500; MAB1585, Merck Millipore; Madrid, Spain) and rabbit anti-melanopsin (1:750; UF008 (AB-N39), Advanced Targeting Systems, Carlsbad, CA, USA) antibodies to quantify the total number of vision-forming and non-vision-forming retinal ganglion cells (RGCs), respectively. Additionally, guinea pig anti-Iba1 antibody was used to identify microglial cells or infiltrated macrophages (1:500; 234308, Synaptic Systems, Göttingen, Germany).

Secondary detection employed Alexa Fluor-labeled secondary antibodies (1:500; Molecular Probes; Thermo Fisher Scientific, Madrid, Spain). Retinal whole-mounts were mounted using anti-fading mounting media (H-1200, Vectashield^®, Vector Laboratories Inc., Burlingame, CA, USA).

Images were captured using a Leica DM6B epifluorescence microscope (Leica, Wetzlar, Germany). Retinal photomontages were constructed from individual square images of 500 µm² each. The total population of Brn3a⁺RGCs was automatically quantified following established procedures (49). m⁺RGCs were manually marked on the photomontage, and the markings were subsequently automatically quantified. The topographic distribution of Brn3a⁺RGCs and m⁺RGCs was evaluated through isodensity and neighborhood maps, respectively, using methods previously described (49, 51). Isodensity maps depict RGC density with a color scale ranging from 0-500 (purple) to ≥ 3,200 RGCs/mm² (red). Neighbour maps illustrate the number of neighboring cells around a given cell within a radius of 0.165 mm, with a color scale from 0-2 (purple) to ≥ 21 neighbors (dark red).

Retinas freshly dissected post-euthanasia were collected in neurobasal medium (Thermo Fisher Scientific), supplemented with 10% fetal bovine serum (Thermo Fisher Scientific), 2% B-27 (Thermo Fisher Scientific), and 1% L-glutamine (Merck Life Science). The dissected retinas were mechanically processed with a scalpel. Following gentle resuspension through pipetting for enhanced cell dissociation, 0.2% collagenase A (0.223 U/mg; Roche Diagnostics GmbH, Mannheim, Germany) in Dulbecco’s-modified Eagle’s medium (DMEM) was added and incubated for 30 minutes at 37 °C. Subsequently, cellular suspensions were filtered through a 70-μm CorningTM cell strainer (Thermo Fisher Scientific) and promptly centrifuged for 5 minutes at 600g. The resulting cell pellets were resuspended in complete DMEM medium, and primary fluorescence-labeled antibodies (1:250 anti-mouse CD11b-FITC and 1:500 anti-mouse CD45-PE; eBioscience, Thermo Fisher Scientific) were added. After a 30-minute incubation at 4 °C, two washing steps were performed, and the cells were ultimately analyzed using a FACS Canto flow cytometer (Becton Dickinson, Franklin Lakes, NJ, United States). Flow cytometry data were analyzed with FlowJo software (FlowJo LLC, Ashland, OR, United States) at the Tissue Culture Facility (ACTI, University of Murcia and IMIB).

Six, 24 or 72h after LPS administration, blood was collected from the heart of pentobarbital overdosed mice, mixed with citrate buffer (3.3% in double distilled water) and placed in ice for 15 min. Then, samples were centrifuged at 3,000 rpm for 10 min at 4°C, the plasma collected and immediately frozen at -80°C until analysis.

Right after blood extraction, retinas from the same animals were fresh dissected and immediately submerged in Pro-Prep™ (Intron Biotechnology, Seongram, South Korea) and dissociated with a hand shredder. After 1h, samples were centrifuged for 15 min at 13,000 rpm. Finally, supernatants were collected and stored at -80°C until analysis. Controls were samples from intact animals.

Each cytokine was measured individually using murine TNF, IFN-γ and IL1-β ELISA kits from Raybiotech (ELM-TNFa-CL-1, ELM-IFN-g-CL-1 and ELM-IL1-b-CL-1, respectively, Bionova Científica, Madrid, Spain), and LT-α from Cloud Clone (SEA134Mu-96T; Bionova Científica, Madrid Spain) following the manufacturer´s instructions. Absorbances were measured at 450 nm in a spectrophotometer and concentrations calculated from standard curves.

The data were analyzed and graphed using GraphPad Prism v.7 (GraphPad Software, San Diego, CA, USA), and the results are presented as mean ± standard deviation (SD). Significance was determined at p<0.05. Detailed information regarding the statistical tests employed can be found in the Results section.

The response to LPS is modulated by the sex, age and strain of the mouse, as well as environmental factors such as diet (18, 35, 52). Therefore, the concentration of LPS used varies widely between reports, even within the same mouse strain. We started by determining LPS dose in C57BL/6J male mice from our facility (CEIB, IMIB, Murcia Spain). In the literature it has been described that males are more sensitive to LPS administration, therefore, we decided to test the appropriate LPS dose in males (30, 36). At first, we looked at whether there was RGC loss 3 days after LPS administration. ( Supplementary Figures 1A, B ). The lowest concentrations, 2 to 4 mg/kg, resulted in a loss of 10% of RGCs, while 5 and 7 mg/kg concentrations caused a 20% loss. A 10 mg/kg dose was lethal for these mice. At 3 days, there were no differences on RGC survival between 5 and 7 mg/kg. Finally, we compared the effect of the two doses at day 7 to confirm that RGC viability was still the same ( Supplementary Figure 1C ). As there were no differences in RGC survival between the two doses, all experiments were performed at the concentration of 5 mg/kg, being less toxic to mice ( Supplementary Figure 1D ).

We then analysed the time course of the loss of Brn3a-expressing RGCs (vision-forming RGCs; Brn3a⁺RGCs) and melanopsin-expressing RGCs (non-vision-forming RGCs, M1 to M3 subtypes of intrinsically photosensitive RGCs; m⁺RGCs) (39) after LPS administration.

In both sexes, there was a significant and diffuse loss of Brn3a⁺RGCs on day 3. This loss continued until day 7, when it stabilised ( Figures 2A, B ). Females (blue bars) had significantly fewer Brn3a⁺RGCs than males (red bars). Therefore, to compare RGC loss between sexes, we calculated the percentage of RGC survival compared with intact retinas. As shown in Figure 2C , Brn3a⁺RGC loss was at all times proportionally higher in males than in females (from day 7, 27 ± 0.4% loss in males, 20 ± 0.6% in females).

The population of m⁺RGCs did not differ between males and females. In addition, unlike Brn3a⁺RGCs, they were resistant to LPS-induced systemic inflammation ( Figure 2D ).

Retinas of females and males were imaged longitudinally with SD-OCT before and after LPS administration.

In the baseline images, the vitreous was clear. However, after LPS administration, hyperreflective puncta, presumably infiltrated cells, were visible in the vitreous from day 1 to day 45 in both sexes. Their numbers were significantly higher than at baseline imaging at all time points and were significantly more abundant in males than females from day 4 onwards ( Figures 3A, B ).

The retinal volume increased early after LPS administration similarly in both sexes ( Figure 3C ). Next, we measured the inner, outer and total retinal thickness at 0.4 mm and 1 mm from the optic disc ( Figure 3D ) and observed a significant thickening of the total retina, mainly due to enlargement of the inner retina. These changes were more marked in females than in males except at day 2 in the inner retina at 1mm.

In flat-mounts, morphological activation of Iba1⁺microglia/macrophages was observed as soon as day 3 after LPS administration, and started to resume at 21 days, though still some microglial cells showed signs of activation, mostly around the optic nerve until day 45 ( Figure 4 , higher magnifications are shown in Supplementary Figure 1E ). Iba1⁺cells were hypertrophic, with swollen cell bodies and few ramifications. In addition, amoeboid cells were found around retinal vessels, probably microglial cells, perivascular macrophages or infiltrating monocytes.

Because retinal swelling could be related to cell infiltration in the retina as well as to microglial activation, we quantified the proportion of CD11b⁺ (myeloid cells: microglial cells or macrophages, irrespective of their state of activation) or CD45⁺ cells (broad leukocyte marker: expressed by activated microglia/macrophages and by monocytes or lymphocytes [reviewed in (53)] by flow cytometry. In both sexes, there was a significant increase in CD11b⁺ and CD45⁺ cells at 3 and 7 days ( Supplementary Figure 2 ) returning to basal levels on day 21, in agreement with the morphological activation. When cell populations were separated according to their marker combination, we observed an increase in steady-state microglia/macrophages (CD11b⁺CD45^-) at 3 and 7 days, which was significantly higher in females on day 3. In both sexes, there was a similar transient increase in infiltrating neutrophils, monocytes or lymphocytes (CD11b^-CD45⁺) at 3 days and a smaller but still significant increase in activated microglia/macrophages (CD11b⁺CD45⁺) ( Figure 5 ).

We measured pro-inflammatory cytokine levels by ELISA in matched retinal and plasma samples at early time points after LPS administration ( Figure 6A ). TNF levels increased significantly in the retinas of both sexes 6 and 24h after LPS exposure. In plasma, this increase was observed at 6h and progressively increased, especially at 72h. Lymphotoxin-α (LT-α), the non-canonical ligand of TNFR1, whose canonical ligand is TNF, was increased only in male retinas 6h after LPS administration. IL1-β, secreted by macrophages through inflammasome activation after LPS challenge (54), was significantly upregulated in the plasma of females and males at 24h, while this increase was observed in the retina at 24h in females and 72h in males.

IFN-γ is necessary for LPS-responsive gene induction and facilitates the production of several proinflammatory mediators (55). In our mice model, IFN-γ levels in plasma increased in both sexes after 24 h, without changes in the retina. Contrary to the other cytokines measured, IFN-γ basal levels were higher in plasma than in retina [4.6-fold]. Indeed, and surprisingly, basal retinal levels of TNF, IL-1β and LT-α were significantly higher in retina than in plasma, exceeding plasma levels by a factor of about 12, 4 and 1.4, respectively. Yet, the relative increment in TNF and IL-1β was much higher in plasma than in the retina ( Figure 6B ). By contrast, LT-α was not altered in plasma.

In view of the above, we decided to block, alone or in combination, the P2X7 receptor (P2X7R), which induces IL-1β secretion by inflammasome activation, and the TNF receptor 1 (TNFR1), which activates the extrinsic pathway of apoptosis. Antagonists (α) were injected intraperitoneally (αTNFR1) or subcutaneously (αP2X7R) on a daily basis from day 1 after LPS administration.

The dose of the αTNFR1 (R7050) was already established in mice (43). For the αP2X7R (ITH15004), we did a preliminary experiment on LPS-treated male mice using a dose of 15 mg/kg based on in vitro assays ( Supplementary Figure 3 ). This dose is comparable to those used for evaluating the effect of ITH15004 on the LPS-induced IL-1β release in ATP-stimulated murine peritoneal macrophages, as an eligible model of inflammation. In those experiments, the selective αP2X7R halved the IL-1β release from 1 µM (44). In our experiments, there was significant RGC rescue at this dose. To determine the effect of higher concentrations, we tried 30 and 60 mg/kg. These two higher doses did not improve RGC neuroprotection, so we performed all subsequent experiments at the lowest dose of 15 mg/kg.

Each antagonist was able to reverse LPS-induced Brn3a⁺RGC loss. Both antagonists worked better in females, with no differences compared to intact animals, while RGC neuroprotection was not complete in males. The combination of both treatments significantly improved neuroprotection in males compared with either treatment alone. ( Figures 7A, B ). Strikingly, the rescue of RGCs with the TNFR1 antagonist was significantly better in females than in males ( Figure 7C ).

Electroretinographic waves were recorded in all animals before (pre) and at 3 and 7 days after LPS administration, with or without pharmacological treatments ( Figure 8 ). LPS caused a transient decrease in all wave amplitudes at day 3 in both sexes, indicating functional impairment of the inner and outer retina. In females, all waves were completely recovered by day 7, whereas in males this recovery was incomplete except for the photopic wave. P2X7R antagonism completely restored all waves at day 3, except the b-wave in females, whereas TNFR1 antagonism restored only the photopic wave in males. The combined treatment was better than the TNFR1 antagonism alone, but not as good as the single P2X7R antagonism. There were some subtle differences between females and males, such as smaller recovery at day 3 of the a- and b- mixed waves in males than in females when treated with both antagonists, or of the rod response after P2X7R antagonism.

Finally, when we measured the implicit time, we observed that the response of the female retina was significantly slower than that of the male retina for all waves 3 days after LPS administration ( Supplementary Figure 4 ). In animals treated with the antagonists, where the implicit times of females were faster, this difference was not observed.