Because different pathogens cause experimental endophthalmitis at certain infectious doses (Gupta et al., 2019; Francis et al., 2020; Kumar et al., 2020; Livingston et al., 2021), we performed a dose-response study beginning with an intravitreal injection of 5,000 cfu of S. epidermidis, a dose required for S. aureus endophthalmitis (Singh et al., 2021). However, this inoculum was rapidly cleared from the mouse eyes resulting in transient inflammation (data not shown). Afterward, eyes were challenged with three higher doses (10⁵, 10⁶, and 10⁷ cfu) of S. epidermidis. Our data showed a dose-dependent increase in corneal haze, anterior chamber opacity, and hypopyon formation at 24 h post-infection ( Figure 1A ). The ERG analysis revealed that all S. epidermidis infected mouse eyes have reduced amplitudes of a and b waves, indicating significant loss of retinal function in comparison to uninfected control eyes ( Figures 1B, C ). Eye challenged with 10⁷ cfu retained the least ERG response with flat wavefront. As expected, PBS injected control eyes did not show any bacteria growth, whereas the intraocular bacterial burden correlated with an infectious dose, i.e., the higher the injected inoculum, the more bacteria recovered from the eye ( Figure 1D ). Interestingly, we noticed that out of six eyes, three eyes in 10⁵ cfu, two eyes in 10⁶ cfu, and one eye in 10⁷ cfu groups had lower bacterial burden than the injected inoculum. The assessment of inflammatory mediators showed that S. epidermidis induced the production of inflammatory cytokines (IL1-β, IL-6, TNF-α) and chemokines (CXCL-1 and CXCL-2) in mouse eyes ( Figure 1E ). However, significant elevation was detected in eyes infected with 10⁷ cfu followed by 10⁶ cfu but not by 10⁵ cfu. These results indicate that S. epidermidis dose is a key determinant in the pathogenesis of endophthalmitis.

Our dose-response study revealed that 10⁷ cfu of S. epidermidis causes a significant decline in retinal function and induces persistent inflammation in mouse eyes. Therefore, we decided to use a dose of 10⁷ cfu/eye to determine temporal changes during S. epidermidis endophthalmitis via assessment of the disease progression up to 72 h. Our data showed that infected eyes had visible opacity and corneal haze at 24 h, which subsequently reduced at 48- and 72-h time points ( Figure 2A ). In contrast, the histological analysis revealed time-dependent retinal tissue damage, with increased retinal folds and heavy cellular infiltrates in the vitreous cavity ( Figure 2B ). This coincided with a significant reduction in retinal function assessed by ERG analysis ( Figure 2C ). The amplitudes of both a and b waves were significantly decreased at all time points, i.e., 24, 48, and 72 h post-infection ( Figure 2D ). Relatively, the b wave amplitudes were lower than a wave, indicating impairment of photoreceptor functions. Interestingly, bacterial burden estimation revealed a time-dependent decrease in the viable bacterial count, indicating enhanced bacterial clearance ( Figure 2E ). To ensure whether the observed reduction in bacterial burden is not due to S. epidermidis growth defect, we performed an ex vivo growth curve using human vitreous. Our data showed that similar to enriched bacterial culture media, S. epidermidis exhibited classical log and lag phases of bacterial growth in the vitreous, indicating the unlikelihood of defective bacterial growth ( Figure S1 ).

One of the hallmarks of bacterial infection is the induction of inflammatory response (Kumar et al., 2016; Kumar et al., 2020; Francis et al., 2020). To determine innate immune response during S. epidermidis endophthalmitis, we assessed the expression of key inflammatory mediators in infected eyes. Our data showed that S. epidermidis induced the expressions of inflammatory cytokines (IL1-β, IL-6, and TNF-α) as well as chemokines (CXCL1 and CXCL2), at both mRNA ( Figure 3A ) and protein ( Figure 3B ) levels. The time-course study revealed that inflammation peaked at 24 h post-infection followed by a significant decline at 48- and 72-h time points. We also noticed that while inflammatory cytokines were drastically reduced at 48 and 72 h, protein levels of the chemokines CXCL1 and CXCL2 were still higher at these time points. Since chemokines play a key role in the recruitment of immune cells, we performed flow cytometry to measure neutrophil infiltration during S. epidermidis endophthalmitis. Our data showed a gradual, time-dependent increase in PMN infiltration in S. epidermidis infected eyes ( Figure 4A ), as evidenced by increased percentages of CD11b⁺ Ly6G⁺ at all time points ( Figure 4B ). Together, these results indicate the induction of an innate inflammatory response in S. epidermidis endophthalmitis.

Upon pathogen invasion, host cells engage TLRs for pathogen recognition and to initiate innate immune responses (Pandey et al., 2013). As our prior studies have implicated the essential role of TLRs in retinal innate defense during endophthalmitis (Kumar et al., 2010; Singh et al., 2014; Talreja et al., 2015), we evaluated their expression in this study. Our data showed that S. epidermidis induced the expression of Tlr2, Tlr4, Tlr6, and Tlr9, the main TLRs involved in bacterial recognition. Time-course analysis revealed the highest expression of all TLRs at 24 h post-infection with a subsequent decline at 48 and 72 h ( Figure 5 ). After pathogen recognition, TLRs activate an inflammatory signaling cascade involving NF-kB and MAPKs (e.g., ERK, P38) resulting in the production of inflammatory mediators. S. epidermidis infection resulted in IκB-α phosphorylation, which was detectable and peaked in mouse retinal lysate at 24 h, followed by a slow decline at 48 h. Accompanying the increase in IκB-α phosphorylation, IκB-α degradation was more prominent at 24 h. A similar time-dependent increase was observed in the phosphorylation of P38 and ERK in retinal tissue ( Figure 6A ). We also assessed the activation of these pathways in retinal Muller glia and RPE, which were shown to play roles in retinal innate defense during intraocular infection (Lenkowski and Raymond, 2014; Singh et al., 2014; Singh et al., 2017). Our data showed that S. epidermidis induced the activation of NF-kB, P38, and ERK signaling in Muller glia ( Figure 6B ) and RPE ( Figure 6C ) as evidenced by increased phosphorylation at both 30- and 90-min time points. Altogether, these results indicate that both retina and retinal cells directly respond to S. epidermidis and activate TLR-mediated inflammatory signaling pathways.

Next, we assessed the biological relevance of the induced signaling pathways; we determined the effect of S. epidermidis on the expression and secretion of inflammatory cytokines. We observed that human Muller glia challenged with S. epidermidis had increased expression of IL-1β and TNF-α both at the mRNA ( Figure 7A ) and protein ( Figure 7B ) levels. Moreover, the response was found to be time-dependent with a relative increase at the 6-h time point. Human RPE cells also exhibited a similar pattern of IL-1β and TNF-α mRNA ( Figure 7C ) and protein ( Figure 7D ) expression in response to S. epidermidis infection. PMNs and monocytes/macrophages are also major infiltrating innate immune cells during ocular infections (Singh et al., 2021); therefore, we decided to assess the inflammatory response of mouse BMDMs. To this end, our data showed that S. epidermidis induced robust Il-6 and Cxcl1 mRNA ( Figure 7E ) and protein expressions ( Figure 7F ) in cultured BMDMs. The induced expression of other inflammatory cytokines such as Il-1β and Tnf-α both at mRNA and proteins was also detected in BMDMs ( Figure S2 ). It should be noted that there was a differential response based on cell types, with BMDMs expressing much higher levels of inflammatory mediators compared to Muller glia and RPE. These results suggest that retinal cells possess the ability to respond to S. epidermidis infection by secreting inflammatory mediators.

The bacterial strain used in this study is a coagulase-negative strain, S. epidermidis, identifiable as ATCC12228. The strain was routinely cultured in tryptic soy medium (TSA or TSB; Sigma, St. Louis, MO) at 37°C for all the experiments. For in vitro and in vivo infection experiments, overnight-grown bacteria were rinsed and diluted in 1X PBS to achieve the desired colony forming units (cfu). For S. epidermidis growth curve analysis, 180 µl of homogenized human vitreous, Muller Hinton broth (MHB), or TSB was added to each well of a microtiter plate and 20 µl of bacterial suspension (5x10⁴ cfu/ml). After 6, 12, 24, and 48 h from each well, 20-µl aliquots were obtained and serially diluted to determine the bacterial count on tryptic soy agar (TSA).

The human retinal pigment epithelial cell line ARPE-19 was maintained in Dulbecco’s modified Eagle’s medium nutrient mixture F-12 (DMEM F-12), whereas the human retinal Muller glia cell line MIO-M1 was cultured in DMEM GlutaMAX, both with supplementation of 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin antibiotic solution at 37°C in 5% CO₂. However, prior to infection, the cells were cultured in serum and antibiotic-free media and infected with S. epidermidis (MOI, 10:1) for various time points. The cells were used for extraction of RNA or proteins, whereas cell-free culture media were collected for cytokine ELISA.

The isolation of bone marrow-derived macrophages (BMDMs) was done from C57BL/6 mice as described earlier (Singh et al., 2021). Briefly, 6–8 weeks old mice were euthanized, and bone marrows were extracted from their tibias and femurs with RPMI-1640 media containing 10% FBS and 0.2 mM EDTA, maintaining an aseptic condition inside a BSL-2 cabinet. Cells were centrifuged at 400 x g for 5 min at 4°C, followed by the lysis of RBCs in differential NaCl solutions. The cells were then rinsed with the media and transferred to a Petri dish containing an RPMI-1640 medium supplemented with 10% FBS, 1% antibiotic, and 10 ng/ml macrophage colony-stimulating factor (M-CSF) to allow macrophage differentiation. Cells were maintained in a 5% CO₂ incubator at 37°C. After 6 days post isolation, BMDMs were seeded in six-well plates and challenged with S. epidermidis (MOI, 10:1). The cells were used for the extraction of RNA or proteins, whereas culture media were collected for cytokine ELISA.

After infection, the culture supernatants from in vitro experiments were collected, and the levels of IL-1β, IL-6, TNFα, CXCL1, and CXCL2 were determined by ELISA using commercially available kits as described previously (Talreja et al., 2015). ELISA was performed as per the manufacturer’s instructions (R&D systems, Minneapolis, MN). For in vivo cytokine estimation, the whole eyes were enucleated, homogenized in 1X PBS by beating against stainless-steel beads in a Tissue lyser (Qiagen, Valencia, CA, USA), and centrifuged, and the lysates were subjected to ELISA as mentioned above. It is pertinent to note that, before performing ELISA, protein estimation was done using the BCA method to ensure that equal protein concentrations were used for each sample.

Total RNA was extracted from cultured cells or mouse retina using a TRIzol reagent as per the manufacturer’s protocol (Invitrogen, Carlsbad, CA). Next, cDNA was synthesized using 1 μg of the isolated RNA using a Maxima first-strand cDNA synthesis kit according to the manufacturer’s instructions (Thermo Scientific, Rockford, IL). The cDNA was then subjected to qRT-PCR on a StepOnePlus Real-Time PCR System (Applied Biosystems, Foster City, CA, USA) using gene-specific PCR primers synthesized from Integrated DNA Technologies (Coralville, IA, USA) with a PCR condition of initial denaturation at 94°C for 5 min, followed by 40 cycles of denaturation (94°C, 45 s), annealing (60°C, 1 min), and extension (72°C, 45 s), with a final extension at 72°C for 10 min. The data were analyzed as a comparative ΔΔC_T method and were presented corresponding to the fold-change differences in gene expression in test samples with respect to control.

Both male and female C57BL/6 (B6) mice (age, 6–8 weeks) were purchased from the Jackson Laboratory (Bar Harbor, ME, USA) and were housed in a restricted access DLAR facility at the Kresge Eye Institute, maintained in a 12:12 light/dark cycle, and fed with LabDiet rodent chow (Labdiet; Pico Laboratory, St. Louis, MO, USA) and water ad libitum. Both male and female mice, around 8 weeks of age, were used. Mice were treated in compliance with the Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Ophthalmic and Vision Research, and all procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Wayne State University under protocol # IACUC-19-03-1012.

Human vitreous samples were collected under sterile conditions from patients undergoing vitrectomy and had signed a preoperative informed consent to use the excised vitreous fluid for basic and clinical research. The protocol and study design were approved by the Wayne State University School of Medicine Institutional Review Board. Collected samples were stored in -80°C until further use.

Bacterial endophthalmitis was induced in B6 mice by intravitreal injection with specified doses of S. epidermidis. As per our IACUC approved protocol, only the eye of each mouse can be injected with either sterile PBS (serving as control) or bacteria. To obtain a 10⁷ inoculum, we concentrated 10 ml of bacterial culture equivalent to 1 O.D (~10⁸ cfu/ml) by pelleting and resuspending in 200 μl of PBS. Mice were anesthetized with ketamine and xylazine, and intravitreal injections of PBS or bacteria (2-μl volume) were performed using a 34-gauge needle under a microscope. This procedure is routinely performed in the lab and reported in our several studies (Kumar et al., 2010; Talreja et al., 2014; Kumar et al., 2016; Singh et al., 2020; Singh et al., 2021). Disease progression was monitored using slit-lamp examination and testing retinal function using electroretinogram (ERG). Following the desired time point post-infection, enucleated eyes were subjected to bacterial burden estimation, cytokine/chemokine ELISA, polymorphonuclear neutrophil (PMN) infiltration, and histopathology, as described in the following sections.

Bacterial densities in infected eyes of WT mice were assessed using the standard serial dilution and the bacterial plate count method. At the indicated time points, the eyes were enucleated and homogenized in sterile 1X PBS in a tissue lyser (Qiagen, Valencia, CA, USA), followed by serial dilution and plating on tryptic soy agar (TSA) plates. Results were expressed as mean ± SD number of colony-forming units (cfu)/eye.

Flow cytometry was performed to estimate the infiltration of neutrophils in infected eyes as described earlier (Talreja et al., 2015). In brief, the retinas from euthanized mice were isolated and digested with Accumax (Millipore) for 10 min at 37°C, with intermittent mixing using a 22-gauge needle and a syringe. Next, to obtain a single-cell suspension, the lysate was filtered through a 40-μm cell strainer (BD Falcon, San Jose, CA, USA). The cells were then incubated with Fc Block (BD Biosciences) for 30 min, followed by three times washing with PBS containing 0.5% bovine serum albumin (BSA). For staining the cells, phycoerythrin (PE)-Cy5-conjugated CD45, Ly6G-FITC, and CD11b-APC antibodies (BD Biosciences) were used to incubate the cells for 30 min in the dark. Following incubation, cells were washed and suspended in sheath fluid. The stained cells were acquired on the Accuri C6 flow cytometer (BD Biosciences) at the NEI P30 immunology core at the Kresge Eye Institute. Data were analyzed using the manufacturer’s software.

Scotopic electroretinography (ERG) was done to evaluate retinal function in S. epidermidis induced endophthalmitis as described previously (Francis et al., 2020). Briefly, following overnight dark adaptation, ERGs were recorded in control and infected mice eyes using the Celeris ERG system (Diagnosis LLC, Lowell, MA, USA) according to the manufacturer’s instructions. The ERG a-wave was measured as an amplitude between the ERG baseline and the first negative peak, and the ERG b-wave was measured as an amplitude between the first negative peak and the first positive peak. Data were analyzed with respect to placebo control eyes.

Mice were euthanized and eyes were enucleated, fixed in 4% formalin for histopathological analysis. The embedding, sectioning, and hematoxylin and eosin (H&E) staining of the tissues were done by Excalibur Pathology, Inc. (Oklahoma City, OK, USA). The slides were further scanned under the PathScan Enabler IV (Meyer Instruments, Inc., Houston, TX, USA) to obtain images.

Following infection, proteins were extracted from cultured cells after being washed with 1X PBS and lysed with radioimmunoprecipitation assay (RIPA) buffer, supplemented with protease and phosphatase inhibitor cocktails. Retinal tissues from two eyes were pooled in RIPA buffer and sonicated, and lysates were obtained after centrifugation. Total protein concentration was detected using a Micro BCA protein assay kit (Thermo Scientific, Rockford, IL). For western blot, the samples were run on SDS polyacrylamide gels and electrotransferred to 0.45-μm nitrocellulose membranes using a wet blot transfer. The membranes were then treated with 5% skim milk in TBST (20 mM Tris HCl [pH 7.6], 0.15 M sodium chloride, and 0.5% Tween 20) for 1 h at RT and further incubated with respective primary antibodies (Cell Signaling Technology, USA or Santa Cruz Biotechnology, USA) as per the manufacturer’s protocol for overnight on a rocker at 4°C. After washing thrice with TBST, the membranes were further treated with horseradish peroxidase (HRP)-conjugated appropriate secondary antibodies (anti-mouse or anti-rabbit Ig) for 2 h. Following three TBST washes, the blots were developed with a Super Signal West Femto chemiluminescent substrate kit (Thermo Scientific, Rockford, IL). To generate quantitative data, immunodetected protein band intensities were measured using the Image Studio software (LI-COR Biosciences, NE, USA).

All the assays were performed independently three times in biological triplicates, and graphs were plotted showing mean ± standard deviation. The data were analyzed using either Student’s t-tests or ANOVA with the help of GraphPad Prism version 8.1 (Graph Pad, CA, USA). A confidence interval of 95% was maintained for all experimental values. A p-value < 0.05 was considered statistically significant.