Integrative metabolomics and transcriptomics identifies itaconate as an adjunct therapy to treat ocular bacterial infection.

The eye is highly susceptible to inflammation-mediated tissue damage evoked during bacterial infection. However, mechanisms regulating inflammation to protect the eye remain elusive. Here, we used integrated metabolomics and transcriptomics to show that the immunomodulatory metabolite itaconate and immune-responsive gene 1 (Irg1) are induced in bacterial (Staphylococcus aureus)-infected mouse eyes, bone-marrow-derived macrophages (BMDMs), and Müller glia. Itaconate levels are also elevated in the vitreous of patients with bacterial endophthalmitis. Irg1 deficiency in mice led to increased ocular pathology. Conversely, intraocular administration of itaconate protects both Irg1 -/- and wild-type mice from bacterial endophthalmitis by reducing inflammation, bacterial burden, and preserving retinal architecture and visual function. Notably, itaconate exerts synergistic effects with antibiotics. The protective, anti-inflammatory effects of itaconate are mediated via activation of NRF2/HO-1 signaling and inhibition of NLRP3 inflammasome. Collectively, our study demonstrates the Irg1/itaconate axis is a regulator of intraocular inflammation and provides evidence for using itaconate, along with antibiotics, to treat bacterial infections.

Introduction

The incidence of intraocular infections, such as bacterial endophthalmitis, is on the rise because of a growing elderly population, which has increased the demand for eye surgeries to treat diseases such as cataracts, age-related macular degeneration (AMD), and glaucoma., Treatment for each of these conditions, along with the widespread use of intravitreal injections to manage neovascular AMD and diabetic retinopathy, predisposes individuals to endophthalmitis.3, 4, 5 The current treatment for bacterial endophthalmitis involves the intravitreal injection of antibiotics that eliminates the bacteria but fails to suppress inflammation-mediated ocular tissue damage and often results in partial or complete loss of vision., Thus, host-directed therapeutics are needed to complement pathogen-targeted approaches to treat ocular infections.

Transcriptomic and metabolomic analyses are powerful high-throughput technologies that have been increasingly used to uncover disease mechanisms and biomarkers, including several pathways involved in the pathobiology of ocular infections.10, 11, 12 These analyses have revealed a connection between energy metabolism and ocular innate immunity in bacterial and viral infections., Specifically, in bacterial endophthalmitis, the cellular metabolism of both residential (e.g., microglia) and infiltrating cells (e.g., neutrophils and macrophages) exhibited increased glycolysis. Inhibiting glycolysis in bacterial endophthalmitis attenuates intraocular inflammation, suggesting a mechanism that can be targeted for the treatment of ocular infections. Thus, these pre-clinical studies have identified potential mechanisms that enhance protective pathways or limit destructive host responses to attenuate inflammation from bacterial infection in the eye.

In our transcriptomic data from the mouse retina after Staphylococcus aureus infection, we observed marked upregulation of the immune responsive gene 1 (Irg1), which is also called Acod1. Under inflammatory conditions, Irg1 drives the production of itaconate, a product of the Krebs cycle intermediate cis-aconitate and has profound immunomodulatory properties. Expression of Irg1 and the subsequent production of itaconate are increased in response to lipopolysaccharide (LPS) challenge, or bacterial and viral infections., Irg1 is essential for neutrophil infiltration, limiting immune-mediated tissue damage in Mycobacterium tuberculosis (Mtb) infection, and promoting trained immunity. In contrast, Irg1 deficiency in immune cells results in reduced antimicrobial activities. The antibacterial properties of itaconate are due to its ability to inhibit isocitrate lyase, a bacterial glyoxylate shunt enzyme. Moreover, it inhibits bacterial growth at supraphysiological concentrations24, 25, 26 and is predicted to be an inflammatory marker., Thus, because of its broad immunomodulatory properties, the Irg1/itaconate axis has recently become an active area of investigation in the field of immunometabolism.

To determine alterations in gene expression and metabolic pathways that occur during bacterial endophthalmitis, we performed transcriptomic and metabolomic analyses of retinal tissue from infected mouse eyes and the vitreous of patients with bacterial endophthalmitis, which showed increased levels of itaconate. Based on these observations, we hypothesized that the production of itaconate is a protective host response in the eye during bacterial infections. Using Irg1 mice and itaconate analogs, we show that Irg1/itaconate protects the eye from bacterial endophthalmitis by resolving inflammation. In addition, activation of the Irg1/itaconate pathway enhanced the NRF2/HO1 antioxidant pathway, which abrogates inflammation and reduces retinal cell death. Moreover, itaconate synergized with antibiotics to limit bacterial growth and potentiate anti-inflammatory/antioxidant pathways to prevent tissue damage.

Results

Irg1/Itaconate pathway is upregulated in bacterial endophthalmitis

In the present study, we integrated metabolomic and transcriptomic approaches to identify key metabolites, genes, and pathways altered during bacterial endophthalmitis. Using retinal tissue from uninfected or S.-aureus-infected mouse eyes at different time points after infection, we performed temporal metabolomic and transcriptomic analyses. Principal component analysis (PCA) of untargeted metabolomics showed distinct clustering of various experimental groups, suggesting alterations in metabolite profiles among the different time points (Figure 1A). To identify differentially produced metabolites across various time points with respect to controls, we used a partial least-squares discriminant analysis (PLS-DA), in which each group was divided into a two-dimensional score plot with two principal components (variance of 46.71%). To show the differentially altered metabolites from the PCA analysis, we listed the top 25 metabolites using variable importance in projection (VIP) scores, with the top five metabolites being biopterin, N-acetyl-β-d-glucosamine, itaconate, N-acetylaspartic acid, and nicotinamide mononucleotide, respectively (Figure 1B). Because of the variability in the experimental groups, we reaffirmed the capability classification of these top altered metabolites by applying the data to a random forest classification model, which showed a prediction error rate of the top metabolites to be less than 8.2% (Figure 1C). Itaconate was the top altered metabolite identified by this model, confirming its classification as one of the metabolites most affected upon S. aureus infection (Figure 1D).

Figure 1

Metabolomic and transcriptomic profiling reveals upregulation of Irg1 and increased itaconate production in mouse eyes during bacterial endophthalmitis

(A–D) Endophthalmitis was induced in C57BL/6 (B6) WT mice (n = 6 per time point: 6, 12, 24, 48, and 72 h) by intravitreal inoculation of S. aureus RN6390 (5,000 CFUs/eye). Eyes injected with PBS were used as controls. At the indicated time points, the retinas were harvested and subjected to untargeted metabolomics analysis. (A) Principal component (PC) of untargeted metabolomics data at indicated time points. (B) Top 25 altered metabolites based on VIP scores. (C) Random forest classification model and error identification model to validate the top altered metabolites. (D) Top 15 metabolites that were differentially altered across the various time points, as identified by random forest classification.

(E and F) In another set of experiments, endophthalmitis was induced in B6 WT mice (n = 3 at each time point), and retinas were harvested at 6, 12, and 24 h after infection. Eyes with PBS injection were used as controls. Total RNA was extracted from control and infected retinas and subjected to microarray for transcriptomic profiling. (E) Venn diagram representing differentially expressed genes (DEGs). (F) Heatmap of 63 common genes.

See also Figure S1.

To identify the genes associated with altered metabolic pathways, we analyzed transcriptomic data obtained from the retinas of S. aureus-infected mice at 6, 12, and 24 h and retinas of the uninfected control eyes. Among the top 100 differentially expressed genes (DEGs), 63 were differentially expressed across all three time points (Figure 1E). A heatmap of these DEGs shows the differential changes in expression across the three time points, with Irg1 having the highest variability (Figure 1F). Given the relationship between Irg1 and itaconate, we correlated the top 63 DEGs and the top 10 altered metabolites and extracted a correlation score for “itaconate” and “Irg1,” which we plotted in a pairs plot with each of the three time points (6, 12, and 24 h; Figures S1A and S1B). This analysis revealed a positive correlation between itaconate and Irg1. Collectively, our integrated omics analysis revealed that itaconate and Irg1 have the highest concurrent variability across all time points in the context of S. aureus endophthalmitis.

Bacterial infection increases Irg1 and Itaconate levels in mouse and human eyes

Although our transcriptomic and metabolomic analyses showed upregulation of Irg1 and elevated levels of itaconate in infected eyes, the functional relevance of this pathway is unknown in bacterial endophthalmitis. Therefore, we performed a time-course study using a mouse model of S. aureus endophthalmitis and cultured cells to validate our multi-omics findings. Our data showed a time-dependent increase in the expression of Irg1 at both the mRNA (Figure 2A) and protein (Figure 2B) levels in S. aureus-infected mouse retinas. Macrophages are one of the major populations of innate immune cells infiltrating the eye during infection; therefore, we used bone marrow-derived macrophages (BMDMs) and observed a similar increase in Irg1 expression after S. aureus infection in cultured wild-type (WT) BMDMs (Figures 2C and 2D) and in the human retinal Müller glia cell line MIO-M1, which are the residential glial cells of the retina (Figures 2E and 2F). Because S. aureus produces several virulence factors that contribute toward the pathogenesis of endophthalmitis, we sought to determine their effects on Irg1 expression. Our results show that, in addition to live S. aureus, heat-killed S. aureus (HK), lipoteichoic acid (LTA), and peptidoglycan (PGN), but not the α-toxin or toxic shock syndrome toxin-1 (TSST-1), induced the Irg1 expression both at transcript (Figure 2G) and protein (Figure 2H) levels. Because Irg1 is responsible for the synthesis of itaconate during the Krebs cycle (Figure 2I), we used a targeted metabolomics approach to measure itaconate levels after S. aureus infection. Consistent with induced Irg1 expression, the levels of itaconate were noticeably increased in S. aureus-infected mouse vitreous/retinal tissue (Figure 2J) as well as in BMDMs (Figure 2K).

Figure 2

IRG1 expression and itaconate production are induced in mouse and human eyes during bacterial endophthalmitis

(A and B) Endophthalmitis was induced in the eyes of B6 WT mice (n = 6 per time point) by intravitreal inoculation of S. aureus (SA) RN6390 (5,000 CFUs/eye). PBS-injected eyes were used as controls. At the indicated time points post-infection, eyes were enucleated, and retinal tissue was used to quantitate Irg1 expression by qPCR (A) and western blot (B), normalized with β-actin and heat shock protein 90 (HSP90) as endogenous controls, respectively.

(C and D) BMDMs from WT mice (n = 4/condition) were challenged with S. aureus (multiplicity of infection [MOI] 10:1) for the indicated time points. Irg1 expression was measured by qPCR (C) and western blot (D), normalized with β-actin and HSP90 as endogenous controls, respectively.

(E and F) Human retinal Müller glia cell line, MIO-M1, (n = 4/condition) was infected with S. aureus (MOI 10:1) for the indicated time points. Irg1 expression was measured by qPCR (E) and western blot (F), normalized with β-actin and HSP90 as endogenous controls, respectively.

(G and H) BMDMs from WT mice (n = 4/condition) were challenged with S. aureus (SA), heat-killed SA (HK), staphylococcal lipoteichoic acid (LTA), peptidoglycan (PGN), α-toxin, and, toxic shock syndrome toxin 1 (TSST1) for 8 h. Irg1 expression was measured by qPCR (G) and western blot (H), normalized with β-actin and HSP90 as endogenous controls, respectively.

(I) Schematic showing the role of Irg1 in itaconate production.

(J–L) Itaconate estimation by liquid chromatography-mass spectrometry (LC-MS). (J) retinal tissue from WT mice 24 h after S. aureus infection (n = 3, six retinas pooled per sample) or PBS-injected control eyes (C; n = 3);

(K and L) BMDMs 8 h after S. aureus infection (K; n = 4); and (L) vitreous samples from patients with Gram-negative (Gram−; n = 10) and Gram-positive (Gram+; n = 12) bacterial endophthalmitis or healthy controls (HCs; n = 10).

The data represented are the culmination of two independent experiments and are shown as means ± SD. Statistical analysis was performed using one-way ANOVA (panels A, C, E, G, and L) or unpaired t test (J and K). ns; non-significant, ∗p < 0.05, ∗∗p < 0.001, ∗∗∗p < 0.0001. See also Table S1.

To determine whether itaconate is also produced during ocular bacterial infection in humans, we assessed itaconate levels in the vitreous samples from 22 patients who were clinically diagnosed with endophthalmitis, which was confirmed by a positive bacterial culture. The vitreous samples (n = 10) from patients who underwent vitrectomy for other retinal diseases (e.g., retinal detachment) were used as healthy controls (HC). Detailed patient demographics are provided in Table S1. Targeted metabolomics analyses revealed elevated levels of itaconate in the vitreous of patients with bacterial endophthalmitis compared with HC, although no significant difference was observed in the itaconate levels of those infected with Gram-positive (mean ± SD, 50.56 ± 13.04 nM) versus Gram-negative (mean ± SD, 63.23 ± 13.24 nM) bacteria (Figure 2L; Table S1). These results show that Irg1 expression is induced and itaconate levels are elevated in mouse and human eyes upon infection, indicating a potential role for the Irg1/itaconate signaling in the pathobiology of bacterial endophthalmitis.

IRG1 deficiency exacerbates bacterial endophthalmitis

To investigate the role of IRG1 and itaconate in ocular infection, we induced S. aureus endophthalmitis in WT and Irg1 mice. We previously showed that WT mouse eyes challenged with an infective dosage of 500 colony forming units (CFUs) of S. aureus resolve the infection within 48–72 h (resolving condition), whereas a dosage of 5,000 CFUs causes severe endophthalmitis (non-resolving condition) leading to blindness., Thus, the bacterial dosage is an important determinant in disease outcomes in endophthalmitis., Given the anti-inflammatory role of Irg1/itaconate, we hypothesized that IRG1 deficiency would delay the resolution of inflammation and worsen disease outcomes. As such, we decided to assess the effects of the 500 CFUs/eye dosage, which is typically resolved in WT mice. In comparison to WT mice, endophthalmitis was exacerbated in Irg1 mice, as demonstrated by increased corneal haze, opacity, and hypopyon (Figure 3A, top panel). Histopathological analysis showed increased retinal tissue damage, retinal folding, and heavy cellular infiltrate in the eyes of Irg1 mice (Figure 3A, bottom panel), and this was associated with reduced retinal function as determined by electroretinography (ERG) showing a decline in a- and b-wave amplitudes (Figure 3B). Moreover, the intraocular bacterial burden (Figure 3C) was significantly higher in Irg1 mice with respect to their WT counterparts, which coincided with elevated levels of the inflammatory mediators, interleukin 1 beta (IL-1β), IL-6, and tumor necrosis factor alpha (TNF-α) (Figure 3D). The anti-inflammatory role of Irg1 in response to S. aureus infection was also confirmed in BMDMs isolated from Irg1 mice, which showed increased levels of inflammatory mediators (Figure 3E). The effect of Irg1 deficiency on itaconate production was revealed by drastically low levels of itaconate in Irg1 BMDMs (Figure 3F). Together, these findings indicate that Irg1 exerts protective effects in the eye by reducing intraocular inflammation.

Figure 3

Irg1 mice exhibit exacerbated bacterial endophthalmitis

(A–D) Endophthalmitis was induced in the eyes of B6 WT and Irg1 mice (n = 6) by intravitreal inoculation of S. aureus (SA) RN6390 (500 CFUs/eye). Eyes in WT (WT C) and Irg1 (Irg1 C) mice with PBS injection were included as controls. (A, top panel) Representative slit-lamp micrograph showing corneal haze/opacity at 48 h post-infection (hpi). (A, bottom panel) H&E staining of eyes that were enucleated at 48 hpi for histopathological analysis. (B) The percentage of a- and b-wave amplitudes from scotopic ERG that were recorded from S. aureus-infected WT and Irg1−/− mice (n = 6 for each group) and compared with PBS-injected control eyes whose wave amplitudes were adjusted to 100%. (C) The intraocular bacterial burden in eye lysates from mice treated in (A) was determined by serial dilution and plate count method and represented as CFUs/eye (n = 6) (D) The indicated inflammatory cytokines in whole eye lysates were quantified by ELISA.

(E) WT and Irg1 BMDMs (n = 4/condition) were either left uninfected 'C' or infected with S. aureus RN6390 'SA' (MOI 10:1) for 8 h. The inflammatory cytokines (IL-1β, IL-6, and TNF-α) were quantified from the conditioned media by ELISA.

(F) Itaconate levels in WT and Irg1 BMDMs (n = 3) that were infected with S. aureus (MOI 10:1) for 8 h and quantified by LC-MS.

The data represented are the culmination of two to three independent experiments and are shown as means ± SD. Statistical analysis was performed using two-way ANOVA (panels B, D, E, and F) with Tukey’s multiple comparison test or by unpaired t test (panel C). ns; non-significant, ∗p < 0.05, ∗∗p < 0.001, ∗∗∗p < 0.0001. C, cornea; AC, anterior chamber; L, lens; VC, vitreous chamber; R, retina. See also Figure S2.

To determine whether the increased susceptibility of Irg1 mice to bacterial endophthalmitis at an even lower infectious dose of S. aureus was due to defective production of itaconate, we performed a rescue experiment; in which, Irg1−/− mice were supplemented with 4-octyl itaconate (OI), an itaconate derivative, via intravitreal injection 6 h after S. aureus infection (Figure S2A). Mice that received OI had reduced corneal haze/opacity and hypopyon, and histopathological analysis showed the retinal architecture was preserved (Figure S2B). ERG analysis showed that itaconate treatment significantly retained a- and b-wave amplitudes as compared with the untreated disease group (Figure S2C). Similarly, the bacterial burden (Figure S2D) and inflammatory cytokines (Figure S2E) were significantly reduced in itaconate-treated eyes. These results indicate that exogenous supplementation of itaconate protected Irg1 mice from bacterial endophthalmitis.

Itaconate treatment ameliorates bacterial endophthalmitis

The elevated levels of itaconate observed in the vitreous of patients with bacterial endophthalmitis, who are unlikely to have a deficiency of IRG1, suggest itaconate may harbor protective properties. To test that possibility, we used a non-resolving infection model (i.e., a higher infective dose of 5,000 CFUs/eye) to mimic the severe bacterial endophthalmitis that leads to blindness in humans and tested the therapeutic efficacy of two itaconate derivatives, dimethyl itaconate (DMI) and OI. Mice were treated with a single intravitreal injection of DMI or OI 6 h after infection with a high dose of S. aureus (Figure 4A). Disease progression was evaluated 24 h after itaconate treatment by performing ophthalmic exams (slit-lamp microscopy) and histopathological analysis, assessing retinal function, and estimating intraocular bacterial burden and inflammation. In mice infected with S. aureus, DMI or OI treatments markedly reduced the corneal haze/opacity and hypopyon as compared with those without treatment (Figure 4B, top panel). Histopathological analyses showed a striking reduction in retinal tissue damage and cellular infiltration in mice treated with the itaconate derivatives (Figure 4B, bottom panel). Additionally, both DMI and OI treatment significantly reduced bacterial burden in the eyes of infected mice, with DMI treatment having a greater anti-bacterial effect (Figure 4C).

Figure 4

The itaconate derivatives, DMI and OI, ameliorate bacterial endophthalmitis

(A–E) Endophthalmitis was induced in the eyes of B6 WT mice by intravitreal inoculation of S. aureus (SA) RN6390 (5000 CFUs/eye). After 6 h, eyes were treated with either DMI or OI (10 μg/eye) via intravitreal injection. (A) Schematic showing a timeline for induction of endophthalmitis, itaconate treatment, and assays used to monitor disease progression. (B, top panel) Representative slit-lamp micrograph showing corneal haze/opacity in eyes under the indicated conditions 24 h after itaconate treatment. (B, bottom panel) H&E staining of enucleated eyes (n = 6). (C) Quantitation of intraocular bacterial burden in whole-eye lysates by serial dilution and plate counting, which is represented as CFUs/eye (n = 6). (D) The inflammatory cytokines in the retinas of mice at the indicated groups (n = 6) were measured by qPCR and are represented as relative fold change by normalizing gene expression with that of endogenous β-actin. (E) Quantification of protein levels of inflammatory cytokines in whole-eye lysates as determined by ELISA (n = 6).

(F and G) BMDMs from B6 WT mice (n = 4) were pretreated with DMI or OI (125 μM) for 2 h, followed by infection with S. aureus (MOI 10:1) for 8 h. The inflammatory cytokines were measured by qPCR (F) and represented as relative fold change by normalizing gene expression with that of endogenous β-actin (G). Quantification of protein levels of the indicated inflammatory cytokines in conditioned media from BMDMs, as determined by ELISA (n = 4).

The data represented are the culmination of two to four independent experiments and are shown as means ± SD. Statistical analysis was performed using one-way ANOVA with Tukey’s multiple comparison test by comparing S.-aureus-infected samples with or without DMI/OI treatment. ns; non-significant, ∗p < 0.05, ∗∗p < 0.001, ∗∗∗p < 0.0001. C, cornea; AC, anterior chamber; L, lens; VC, vitreous chamber; R, retina. See also Figure S3.

In addition to direct retinal damage as a result of bacterial toxins, the pathogenesis of bacterial endophthalmitis leads to ocular injury from excessive activation of inflammatory pathways. Thus, we assessed intraocular inflammation and found that itaconate treatment (either DMI or OI) drastically reduced expression of the inflammatory cytokines IL-1β and IL-6 at the gene (Figure 4D) and protein (Figure 4E) levels. Surprisingly, the levels of TNF-α remained unchanged, irrespective of treatment, indicating differential effects of itaconate on inflammatory molecules. Similar to our in vivo observation, DMI/OI treatment significantly reduced inflammatory cytokines (IL-1β and IL-6) at the level of the transcript (Figure 4F) and protein (Figure 4G) in WT BMDMs. In addition, DMI/OI treatment reduced the generation of reactive oxygen species (ROS) in S. aureus-infected BMDMs (Figure S3A), indicating its ability to diminish oxidative stress. Collectively, these results indicate that itaconate exerts both antibacterial and anti-inflammatory properties during bacterial endophthalmitis.

As in similar studies, we used the cell-permeable derivatives of itaconate and demonstrated their anti-inflammatory role in the eye. However, some recent reports show that the itaconate derivatives (DMI and OI), exert different mechanisms compared with endogenous itaconate in regulating inflammation., Thus, we tested the effect of pH-buffered (pH 7.0) itaconic acid (ITA), along with OI, in our disease model. We found that ITA treatment also reduced corneal haze/opacity and hypopyon (Figure S3B) and bacterial burden (Figure S3C) in S. aureus-infected eyes. Similarly, ITA was found to attenuate the expression of the inflammatory cytokines at transcript (Figure S3D) and protein (Figure S3E) levels. However, the comparative analysis revealed that OI was superior to ITA in reducing intraocular inflammation, supporting its use to treat ocular infections.

Itaconate synergizes with antibiotics to reduce inflammation and disease severity

Intravitreal injection of antibiotics (vancomycin and ceftazidime) is currently the only standard treatment for managing bacterial endophthalmitis. However, emerging resistance to antibiotics among common ocular pathogens necessitates an increased and continuous evaluation of the best antibiotic treatment for this condition. The anti-inflammatory effects of itaconate suggest that it could be used as an adjunct therapeutic in combination with antibiotics to ameliorate bacterial endophthalmitis. To test that, we first determined the minimum inhibitory concentrations (MICs) of vancomycin (V), DMI, and OI against S. aureus strain RN6390 (Figure S4), which were found to be 2.5 μg/mL, 10 mg/mL, and 12.5 mg/mL, respectively. Next, sub-MIC levels of vancomycin were used either alone or in a combination with DMI or OI (e.g., V+DMI or V+OI) to treat S. aureus-infected eyes. As expected, treatment with vancomycin drastically reduced bacterial burden, yet the combined treatment of both V+DMI or V+OI further reduced viable bacterial counts (Figure 5A). Moreover, the combination therapy protected the retina from tissue damage (Figure 5B) and preserved retinal function with better retention of a- and b-wave amplitudes relative to treatment with V, DMI, or OI alone (Figure 5C). Similarly, the levels of inflammatory mediators were significantly diminished in eyes treated with combination therapy as compared with V, DMI, and OI treatments alone (Figure 5D). These results suggest the potential use of itaconate as an adjunct therapy to complement the antibiotic treatment of bacterial endophthalmitis and potentially lower the antibiotic dosage required for the treatment of this condition.

Figure 5

Itaconate synergizes with antibiotics to attenuate intraocular inflammation in bacterial endophthalmitis

Endophthalmitis was induced in the eyes of WT mice (n = 6/group). Six hours after infection, eyes were treated by intravitreal injection with vancomycin (V, 0.6 μg/eye) or DMI or OI (10 μg/eye) either alone or in the indicated combinations. Twenty-four hours after drug treatment, eyes were enucleated for analysis.

(A) Bacterial burden was estimated by serial dilution and plate count method.

(B) Representative H&E staining of enucleated eyes treated as indicated. C, cornea; AC, anterior chamber; L, lens; VC, vitreous chamber; R, retina.

(C) Percentage of a- and b-wave amplitudes retained in scotopic ERG recorded from WT mice eyes after S. aureus (SA) infection and itaconate/vancomycin treatment.

(D) Quantification of the indicated inflammatory cytokines in whole-eye lysates as determined by ELISA.

The data represented are the culmination of two to three independent experiments and are shown as means ± SD. Statistical analysis was performed using one-way ANOVA with Tukey’s multiple comparison test by comparing S. aureus-infected samples with or without treatment. ns; non-significant, ∗p < 0.05, ∗∗p < 0.001, ∗∗∗p < 0.0001. See also Figure S4.

Itaconate exerts its anti-inflammatory effect by potentiating NRF2/HO-1 signaling

To understand the protective mechanisms evoked by itaconate, we performed RNA sequencing (RNA-seq) analysis of WT BMDMs infected with S. aureus in the presence or absence of OI. Itaconate modulated the expression of a total 355 genes (212↑; and 143↓), including those which regulate antioxidant (e.g., Nrf2 and Ho-1), innate immunity, immune activation, and inflammasome pathways (Figure 6A). We analyzed the top 20 DEGs based on the proteins they encode to generate a protein-protein interaction map (Figure 6B). This revealed that itaconate may affect genes encoding proteins associated with the inflammasome (group 1) and that are involved in antioxidant pathways (group 2). Prior studies have shown that itaconate exerts its anti-inflammatory effects via Nrf2, which, along with its downstream target gene Ho-1, was differentially expressed in our analysis. Therefore, we set out to determine the role of Nrf2 and Ho-1 in our disease model (Figure S5A).

Figure 6

Itaconate potentiates NRF2/HO-1 signaling in S. aureus-infected mouse retina and BMDMs

(A and B) BMDMs from WT mice (n = 3/condition) were either left untreated or pre-treated with OI (125 μM) for 2 h, followed by infection with S. aureus for 8 h. RNA was extracted, and RNA sequencing was performed using the Illumina platform. (A) Heatmap showing DEGs in S. aureus (SA)-infected and OI-treated BMDMs. (B) The DEGs shown in (A) were analyzed using the String database (version 11.0) to create a protein-protein interaction network. Group-1: genes involved in the inflammasome pathway. Group-2: genes involved in the antioxidant pathway. Group-3: transcription factors and hypoxia-inducible factor. Thick lines denote strong evidence of interactions, and dotted lines represent possible interactions among the genes in these pathways.

(C–E) After induction of S. aureus (SA) endophthalmitis and itaconate (DMI or OI) treatment, mouse retinas (n = 6) were used to quantitate Nrf2 and Ho-1 transcripts by qPCR (C) and protein expression by western blot with densitometry (n = 4) (D and E), normalized with β-actin and HSP90 as endogenous controls, respectively.

(F–H) BMDMs from WT mice were pretreated with DMI or OI (125 μM) for 2 h, followed by infection with S. aureus (MOI 10:1) for 8 h (n = 4). Nrf2 and Ho-1 expression was assessed by qPCR (F) and western blot with densitometry (n = 4) (G and H), normalized with β-actin and HSP90 as endogenous controls, respectively.

(I–K) Endophthalmitis was induced in the eyes of WT and Nrf2 mice (n = 6). After 6 h, eyes were treated with OI (10 μg/eye) via intravitreal injection. (I) Representative slit-lamp micrograph showing corneal haze/opacity at 24 h after itaconate treatment. (J) Quantitation of intraocular bacterial burden in whole-eye lysates by serial dilution and plate counting method. (K) The inflammatory cytokines in whole-eye lysates were determined by ELISA (n = 6).

The data represented are the culmination of two to four independent experiments and are shown as means ± SD. Statistical analysis was performed using one-way (C, E, F, and H) or two-way (J and K) ANOVA with Tukey’s multiple comparison test by comparing S. aureus-infected samples with or without OI treatment. (# indicates comparison between WT versus Nrf2; J and K). ns; non-significant, ∗p < 0.05, # #∗∗p < 0.001, ∗∗∗p < 0.0001. See also Figures S5 and S6.

We found S. aureus infection induced the expression of Nrf2 and Ho-1 in mouse retina and DMI or OI treatment potentiated their expression both at the transcript (Figure 6C) and protein (Figures 6D and 6E) levels. The induction and potentiation of Nrf2 and Ho-1 at the transcript (Figure 6F) and protein (Figures 6G and 6H) levels was also observed in S. aureus-infected and OI-treated WT BMDMs. Like DMI/OI, itaconic acid (ITA) also induced the expression of Nrf2 and Ho-1 in S. aureus-infected mouse retina (Figure S5B), albeit at lower levels than OI induced, suggesting the involvement of Nrf2 in both endogenous and exogenously administered itaconate derivatives in the eye. We also assessed Nrf2/Ho-1 signaling in BMDMs from Irg1 mice and found no significant induction of Nrf2 and Ho-1 transcripts (Figure S5C) or protein (Figure S5D), in response to S. aureus infection alone. However, DMI/OI treatment markedly induced NRF2 and HO-1 expression. These results indicate that Irg1 deficiency (i.e., reduced endogenous itaconate) impaired NRF2/HO-1 antioxidant pathway in ocular infections.

We also assessed the effect of DMI and OI on Nrf2 and Ho-1 expression in the human retinal Müller glia cell line MIO-M1. Cells that were infected with S. aureus and treated with either DMI or OI showed increased expression of Nrf2 and Ho-1 at both the mRNA (data not shown) and protein levels (Figures S5E and S5F). Because DMI and OI were found to synergize with vancomycin in reducing intraocular inflammation (Figure 5), we determined whether the combination therapy also boosted antioxidant pathways. Our data show that in vivo, the combination of vancomycin and DMI or OI significantly increased expression of Nrf2 and Ho-1 transcripts as compared with vancomycin or itaconate alone (Figures S5G and S5H).

The ability of itaconate to induce NRF2 and HO-1 expression in both in vivo and in vitro models suggests that NRF2/HO-1 signaling regulates the inflammatory response in bacterial endophthalmitis. Therefore, to test that hypothesis, we used small hairpin RNAs (shRNAs) to silence their expression in WT BMDMs. Knockdown of NRF2 and HO-1 at the protein level was confirmed by western blot (Figure S6A) and that led to increased expression of S. aureus-induced inflammatory mediators (Figure S6B). To further validate the anti-inflammatory role of NRF2 signaling in our model, we used Nrf2 mice. As expected, OI treatment significantly reduced corneal haze, opacity, and hypopyon in WT mice; however, that effect was diminished in Nrf2 mice (Figure 6I). A similar trend of relatively lower antibacterial (Figure 6J) and anti-inflammatory (Figure 6K) properties of OI was observed in Nrf2 versus WT mice. ERG analysis also revealed that, although OI treatment significantly retained retinal function (i.e., amplitudes of a- and b-waves) in both WT and Nrf2 mice compared with the untreated eye, the percentage retention was more in the WT mice. (Figure S6C). In vitro studies showed significant induction of NRF2 and HO-1 in DMI or OI treated WT BMDMs but not in Nrf2 BMDMs (Figure S6D). Like in vivo observation, reduced anti-inflammatory effects of OI and DMI were observed in Nrf2 BMDMs compared with WT cells (Figure S6E). Collectively, these results indicate that OI/DMI treatment was more effective in protecting WT as compared with Nrf2 mouse eyes, indicating the role of NRF2/HO-1 signaling in regulating ocular inflammation in bacterial endophthalmitis.

Itaconate inhibits the expression of bacterial-induced NLRP3 and Caspase−1

Thus far, our studies showed that S. aureus induces the production of IL-1β in both the mouse retina and in cultured BMDMs and that itaconate attenuated that response. Because activation of the NLRP3 inflammasome is primarily responsible for the cleavage of pro-IL-1β into active IL-1β via activation of caspase 1, we sought to determine the effect of itaconate on NLRP3 activation. Using WT and Nrf2 BMDMs, our data showed that S. aureus induced the expression of Nlrp3 (Figure 7A) and caspase-1 (Figure 7B) transcripts, whereas DMI or OI treatment significantly inhibited their expression in both cell types, albeit more in WT BMDMs. This observation was confirmed by western blot for NLRP3 (Figures 7C and 7E) and cleaved caspase 1 P20 (Figures 7D and 7F), which showed inhibition of NLRP3 and Casp-1 in DMI- or OI-treated WT BMDMs. These findings indicate that itaconate attenuates the NLRP3-Casp-1-IL-1β cascade to downregulate ocular inflammation, and Nrf2 is partially involved in that phenomenon.

Figure 7

Itaconate inhibits S. aureus-induced activation of the NLRP3 inflammasome

(A and B) BMDMs from WT and Nrf2 mice (n = 4/condition) either were left untreated or were pre-treated with DMI or OI (125 μM) for 2 h, followed by infection with S. aureus (SA) (MOI 10:1) for 8 h. Nlrp3 (A) and Casp1 (B) expression were assessed by qPCR. The data are expressed as relative fold change by normalizing gene expression to that of the endogenous β-actin gene.

(C and D) Western blot detection of NLRP3 (C) and cleaved caspase-1 (D) proteins.

(E and F) Densitometry analysis was performed using ImageJ, and data are expressed as relative fold changes normalized to the loading control, HSP90.

The data represented are the culmination of two to four independent experiments and are shown as means ± SD. Statistical analysis was performed using one-way ANOVA with Tukey’s multiple comparison test by comparing S. aureus-infected samples with or without DMI or OI treatment. (# indicates comparison between WT versus Nrf2; A, B, E, and F). ns; non-significant, #p < 0.05, ∗∗p < 0.001, ∗∗∗p < 0.0001.

Discussion

The incidence of bacterial endophthalmitis is on the rise, and current treatments are often inadequate in preventing adverse outcomes, necessitating the discovery of alternative therapeutics that alleviate the underlying pathophysiology of this condition. However, the immune-privilege status of the eye poses significant challenges in treating ocular infections. Ophthalmologists have generally been reluctant to prescribe adjunctive anti-inflammatory drugs, such as corticosteroids, because of their immunosuppressive effects.40, 41, 42 Thus, there is significant interest in identifying biologic agents and immunomodulatory therapies that target ocular inflammation without major disruption to the normal immune response in the eye.,

Here, we used an integrated transcriptomic and untargeted metabolomics approach to identify genes and metabolites involved in experimental bacterial endophthalmitis. Using the vitreous of human patients and mouse eyes during bacterial endophthalmitis, we identified an increase in the expression of Irg1 and the concentration of its metabolite, itaconate, and that there was a strong pairwise correlation between the two upon infection with S. aureus. Given the anti-inflammatory role of itaconate, we hypothesized that activation of the Irg1/itaconate axis promotes resolution of inflammation during bacterial endophthalmitis. Using Irg1 mice and through shRNA-mediated gene silencing, we demonstrate that itaconate exerts protective effects by modulating NRF2/HO1 signaling. Most notably, intraocular administration of itaconate either alone or in combination with antibiotic treatment drastically improved disease outcomes. Collectively, our study reveals an essential role of IRG1 and itaconate in regulating protective responses during bacterial infection in the eye and suggests a model wherein intraocular administration of itaconate reduces inflammation and bacterial burden to ameliorate bacterial endophthalmitis. Thus, our study demonstrates considerable promise for the translation of an itaconate-based treatment strategy to mitigate ocular bacterial infections.

An increasing number of studies have shown that under inflammatory stimuli (e.g., LPS challenge), Irg1 is expressed, and consequently, itaconate is produced in macrophages.,, However, the role of Irg1/itaconate remains largely unknown in the pathobiology of eye diseases, especially in ocular bacterial infections. Here, we show that Irg1 is upregulated both in vivo (mouse retina) and in vitro (BMDMs and human retinal Müller glia) during S. aureus infection. Moreover, induced Irg1 expression coincided with an increased accumulation of itaconate in infected mouse eyes, which was also consistent with elevated levels observed in the vitreous of patients with culture-positive bacterial (both Gram-positive and Gram-negative) endophthalmitis. The increased itaconate levels in Gram-negative infection are likely due to LPS, which is known to induce Irg1 and is abundantly present in Gram-negative bacterial cell walls., Here, we show that among the S. aureus virulence factors, cell-wall components, but not the toxins, are the inducers of Irg1 expression, which is consistent with our prior study showing their inflammatory role in the eye., Together, these findings provide the evidence for the involvement of the Irg1/itaconate pathway in ocular infection.

In the absence of Irg1, we found mice are more susceptible to developing bacterial endophthalmitis, even when subjected to a lower infectious dose of bacteria that is usually resolved in WT mice. Irg1 mice had a higher bacterial burden, increased inflammatory mediators, and more tissue damage, culminating in the loss of retinal/visual function. This is in agreement with other studies showing increased infection and inflammation in Irg1 mice in response to Mtb, leading to greater mortality rates. As anticipated, BMDMs from Irg1 mice had reduced itaconate levels, and supplementing those mice with itaconate derivatives ameliorated bacterial endophthalmitis by reducing intraocular inflammation and preserving retinal tissue, suggesting itaconate could serve as a therapeutic option for humans with this condition. However, humans with this condition are likely not commonly deficient for IRG1. Therefore, we assessed the therapeutic efficacy of itaconate in WT mice with S. aureus endophthalmitis and observed profound protection. This suggests interventions that enhance Irg1 expression and/or itaconate levels within the eye could present potential therapeutic strategies for the treatment of ocular bacterial infections.

The clinical outcomes of bacterial endophthalmitis depend upon both the virulence of the infecting organism and the time at which appropriate therapy is initiated. Because of the poor penetration of systemic antibiotics, intravitreal injections of antibiotics remain the most widely accepted treatment option to manage the condition. Although these intraocular antibiotics eliminate the bacterial infection, they often liberate bacterial cell wall components,, contributing to bystander inflammatory damage to ocular tissue. Given the anti-inflammatory and antimicrobial role of itaconate,, and its elevated levels in patients with endophthalmitis, we tested its therapeutic potential when co-administered with antibiotics. Notably, co-administration of itaconate with sub-MIC concentrations of vancomycin significantly reduced ocular inflammation and improved disease outcomes. Thus, our study provides evidence that the combination of antibiotics and itaconate function synergistically to ameliorate endophthalmitis and reduce ocular inflammation. Moreover, including itaconate as an anti-inflammatory therapy could reduce the overall dosage of antibiotics required to treat bacterial endophthalmitis.

Macrophages and Müller glia, are known to have essential roles in orchestrating retinal innate responses in bacterial endophthalmitis, and our RNA-seq analysis in S. aureus-infected BMDMs indicated that NRF2/HO-1 and NLRP3 signaling are differentially modulated by itaconate treatment. Both itaconate derivatives, DMI and OI, potentiated the expression of NRF2 and its downstream signaling molecule HO1 in S. aureus-infected BMDMs and human retinal Müller glia. Moreover, levels of NRF2 and HO1 were increased in the mouse retina treated with itaconate. The reduction in inflammatory mediators upon itaconate treatment in S. aureus-infected cultured cells and its reversal by silencing Nrf2 indicate that itaconate likely exerts its anti-inflammatory effects in the eye by regulating NRF2/HO1 signaling. The mechanisms for anti-inflammatory effects of itaconate have been reported to involve the inhibition of succinate dehydrogenase,, blockade of IκBζ translation, and activation of NRF2. Although our data support the idea of itaconate-induced Nrf2/Ho-1 antioxidant signaling in exerting protective effects,,, recent studies report distinct mechanisms for endogenous itaconate versus its derivatives in regulating inflammation., The itaconate derivative OI reduces inflammation by triggering electrophilic stress, which is compensated by activation of the antioxidant Nrf2/Ho-1 axis, whereas endogenous itaconate evokes NRF2-independent mechanisms. However, a more recent study using 13C5-labeled OI showed that exogenously added OI can directly generate endogenous itaconate and reduce inflammation by carboxypropylation of NLRP3 and prohibiting its interaction with NEK7, which is essential for inflammasome activation. Therefore, those authors argued that OI could be used as a surrogate to study itaconate because it can be taken up and converted to intracellular itaconate by macrophages. We also found that OI treatment reduced NLRP3 coinciding with attenuated IL-1β secretion.

In view of recent studies and to clarify whether DMI- or OI-mediated protection is dependent on Nrf2 signaling, we used Nrf2 mice. Our data showed that (1) in comparison to WT mice, infected Nrf2 mouse eyes had higher inflammation and bacterial burden; and (2) the protective and anti-inflammatory effects of OI were reduced but not completely abolished in Nrf2 mice. Previous studies investigating the Irg1/itaconate axis have used LPS-stimulated macrophages. In contrast, we used live bacteria (which possess multiple virulence factors), and the observed differences in Nrf2 dependency could be attributed to the nature of the inflammatory stimuli. Nonetheless, we conclude that Nrf2 signaling, in part, contributes to OI-mediated protection, and Nrf2 deficiency exacerbates S. aureus endophthalmitis. Our data show that S. aureus induced NLRP3/Casp1 expression and IL-1β secretion in WT BMDMs, and Nrf2 silencing or knockout (Nrf2) potentiated IL-1β levels. However, in Nrf2 BMDMs, although, reduced by OI or DMI treatments, NLRP3/Casp1 levels did not change by S. aureus alone, indicating Nrf2-independent regulation of NLRP3 inflammasome. Some potential mechanisms by which S. aureus could activate NLRP3/Casp-1 in an Nrf2-independent manner include (1) induction of oxidative (i.e., ROS generation; Figure S3A) and endoplasmic reticulum (ER) stress, which can directly activate NLRP3/Casp1/IL-1β cascade;, and (2) the production of saturated fatty acids, such as palmitate, under inflammatory conditions in the retina,, which could also trigger NLRP3 activation, leading to IL-1β and IL-18 production. We acknowledge that further studies using Irg1 and Nrf2; Nlrp3 and Nrf2 double knockout mice and cells are needed to determine the role of endogenous itaconate and to dissect Irg1-Nrf2-Nlrp3 crosstalk in ocular infection.

Controlling intraocular inflammation is an effective approach to improve the outcome of ocular infections. In support of that, our results provide evidence of the effectiveness of itaconate as adjunctive therapy with antibiotics to treat bacterial endophthalmitis by reducing the bacterial burden and correcting the dysregulated inflammatory response, thus preventing ocular tissue damage.

Limitations of study

Despite the importance of these findings, there are a few limitations to our study that should be considered. First, elevated levels of itaconate were detected in both Gram-positive and Gram-negative bacterial endophthalmitis, whereas we only tested the effect of itaconate in a S. aureus (e.g., Gram-positive) infection model. In a recent study, itaconate was found to exacerbate Pseudomonas lung infection. Therefore, additional studies are needed to evaluate the effect of itaconate in Gram-negative ocular infections. Second, we only evaluated the therapeutic efficacy of itaconate at 6 h after infection. However, patients may not seek treatment until they experience pain and blurry vision, indicative of an advanced disease state. Hence, the timing (e.g., too early or very late) of treatment and optimal dose of anti-inflammatory therapy must be further studied because excessive suppression of inflammation could reduce the ability of the host to mount a sufficient antimicrobial response. Corresponding with that, the efficacy of itaconate as an adjunct therapeutic needs to be compared with ocular corticosteroids and evaluated with other types of antibiotics used to treat ocular infections. Although we observed reduced inflammation and protection in itaconic-acid-treated eyes, the effect was not as profound as that of OI. As reported recently, alteration of the itaconate structure in DMI or OI is responsible for extra effects that may not be attributed to itaconate. Therefore, itaconate derivatives used in this study may not fully recapitulate the endogenous itaconate mode of action and warrant further investigation. We also acknowledge that the amount of itaconate (4OI [∼41 mM] and DMI [∼63 mM]) administered was greater than the endogenous itaconate levels (∼55–120 pM at 24 h after infection) detected in the mouse eyes. Additional time kinetics studies are needed to determine the physiological range of itaconate in both human and mouse eyes.

STAR★Methods

Key resources table

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead contact, Ashok Kumar (akuma@med.wayne.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

The data generated from this study is deposited in the NIH Gene Expression Omnibus. The accession number for the raw and processed data is GSE168928.

Experimental model and subject details

Mice and ethics statement

Wild-type (WT) C57BL/6, Nrf2 and Irg1−/− mice (both male and female, aged 6-8 weeks) were purchased from the Jackson Laboratory and/or bred in-house in the Division of Laboratory Animal Resources facility at Kresge Eye Institute. Mice were maintained in a 12h light/dark cycle at 22°C and fed LabDiet rodent chow (Labdiet Pico lab Laboratory, St Louis, MO) and water ad libitum. All experimental procedures were performed in compliance with the Animals in Ophthalmic and Vision Research (ARVO) statement for the use of animals and were approved by the institutional animal care and use committee (IACUC) of Wayne State University.

Mouse model of bacterial endophthalmitis

Endophthalmitis was induced in WT, Nrf2 and Irg1−/− mice as described previously. Briefly, mice were anesthetized and intravitreally injected with S. aureus RN6390 (5000 CFUs/eye in 2 μL volume) using a 32-G needle under an ophthalmoscope. Eyes injected with PBS served as controls. For treatment groups, itaconic acid (Sigma-Aldrich) or itaconate derivatives dimethyl itaconate (DMI, Sigma-Aldrich) or 4-octyl itaconate (OI) were injected intravitreally (10 μg/eye in 1μL volume) at 6 h post-infection.

Patient vitreous collection

A total of 22 patients clinically diagnosed and treated for infectious endophthalmitis (12 Gram-positive and 10 Gram-negative), who presented to the L V Prasad Eye Institute, Hyderabad, India, retina clinic and underwent diagnostic vitreous biopsy/vitrectomy between November 2018 and February 2019 were recruited for the study. Patient’s demographical and clinical details are provided in the Table S1. In the control group, 10 patients with uninflamed eyes undergoing vitrectomy for non-infectious retinal disorders during the same period were included. In the test group, we excluded cases where the diagnosis was suspect or where the clinical characteristics were ambiguous. All patients diagnosed underwent complete ophthalmological examinations, including slit-lamp microscopy, visual acuity recordings, and later underwent pars plana vitrectomy. All cases were culture positive and had clinical features such as intense anterior chamber inflammation, extensive vitreous exudates, corneal infiltrates, or a lens abscess suggestive of infectious endophthalmitis. Written informed consent was obtained from all subjects and the study was approved by the Institutional Review Board, (LEC 09-18-125) L V Prasad Eye Institute, Hyderabad, India. The microbiological processing of the vitreous samples from the study group included direct microscopy and culture. A small portion of the vitreous sample (∼100μL) was immediately stored at −80°C for itaconate measurement.

Cell and culture conditions

The immortalized human Müller glia cell line MIO-M1 was cultured in DMEM at 37°C in 5% CO2 supplemented with 10% FBS, 1% penicillin-streptomycin and 10 μg/mL L-glutamine. Prior to infection cells were grown overnight in serum and antibiotic-free media and infected with S. aureus (MOI, 10:1) for various time points.

Method details

Bacterial burden estimation

Mouse eyes were enucleated following infection and 24h post-drug treatment. Whole-eye lysates were prepared in 250 μL of PBS by homogenization using stainless steel beads in a tissue lyser (QIAGEN, Valencia, CA). A small portion (50 μL) of the tissue homogenate was serially diluted and plated on Tryptic Soy Agar (TSA) to enumerate bacterial burden. The following day, colonies were counted, and results were expressed as the mean number of CFUs/eye ± SD.

Isolation of bone marrow-derived macrophages

Mouse bone marrow-derived macrophages (BMDMs) were isolated as described previously., Briefly, mice were euthanized, and bone marrow was flushed from tibias and femurs with RPMI media containing 10% FBS and 0.2 mM EDTA. Cells were pelleted by centrifugation at 400 x g for 5 minutes at 4°C. Red blood cells (RBCs) were lysed by adding 0.2% NaCl solution for 20-30 s, followed by the addition of 1.6% NaCl and centrifugation. Following RBC lysis, cell pellets were washed with RPMI media by centrifugation. Cells were re-suspended and seeded in RPMI media supplemented with 10% FBS, 100 U/ml penicillin, 100 mg/mL streptomycin and 10 ng/ml M-CSF for macrophage differentiation at 37°C in 5% CO2. Six days post differentiation 4 × 106 BMDMs/mL were seeded in 65 mm Petri-dishes for in vitro experiments. For drug treatments, BMDMs were pretreated with DMI or OI (125 μM) for 2 h followed by challenge with S. aureus (multiplicity of infection (MOI) 10:1), for 8h.

Determination of the minimum inhibitory concentration (MIC)

A micro broth-dilution method was used to determine the MIC for vancomycin, OI, and DMI as described previously. Briefly, bacterial cultures (105 CFU/well) were exposed to a two-fold serial dilution of the test compound in a 96-well plate. Following overnight incubation, the optical density (A600) of each microplate well was recorded using a spectrophotometer. MICs were determined based on the optical density of the growth in control and the lowest vancomycin, DMI, or OI concentrations that resulted in S. aureus growth inhibition compared with media alone.

RNA extraction and real-time PCR

Mouse neural retinas were removed and pooled (two retinas per sample) in Trizol for RNA isolation. Total RNA was extracted from the retina as per the manufacturer’s instructions (Thermo Scientific, Rockford, IL). RNA was reversed transcribed using a Maxima first-strand cDNA synthesis kit as per the manufacturer’s protocol (Thermo Scientific, Rockford, IL). Quantitative assessment of gene expression was carried out by quantitative real-time PCR (qPCR) using gene-specific primers on a StepOne Plus Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). The data were analyzed using the comparative ΔΔCT method as described previously. The gene expression in the test samples was normalized to endogenous β-actin controls.

Histology

Following infection and drug treatment, eyes were enucleated and fixed in 4% formalin for histopathological examination. The embedding, sectioning, and hematoxylin & eosin (H&E) staining was performed by Excalibur Pathology, Inc. (Oklahoma City, OK, USA). Pathscan Enabler IV (Meyer Instruments, Inc.) was used to scan H&E stained slides.

Cytokine ELISA

For in vivo samples, eyes were enucleated, and lysates were prepared by homogenization using a tissue lyser as described above. The tissue homogenates were centrifuged at 15,000 x g for 20 minutes at 4°C and the clear supernatants were used to estimate cytokine concentration. For in vitro studies, the conditioned media from S. aureus-infected and drug-treated groups along with vehicle control were used for cytokine measurements. ELISA for inflammatory cytokines IL-1β, IL-6, and TNF-α was performed as per the manufacturer’s instructions (R&D System, Minneapolis, USA).

Measurement of intracellular reactive oxygen species (ROS)

ROS production was measured by fluorescence microscopy using dichloro-dihydrofluorescein diacetate (DCFH-DA) as described previously. Briefly, BMDMs were grown in a 4-well chamber slide and infected with S. aureus (MOI, 10:1) for 4h. Following infection, cells were washed with PBS and incubated with 10 μM of DCFH-DA for 30min at 37°C. The cells were then washed twice with PBS and observed via an Eclipse 90i fluorescence microscope (Nikon).

shRNA mediated knockdown

The lentiviral particles containing scrambled control shRNA, NRF2 shRNA (sc-37049-V) were purchased from Santa Cruz Biotechnology (Dallas, TX). The shRNA knockdowns were performed as per the manufacturer’s protocol. The knockdown of target proteins was confirmed by western blotting. The conditioned media were preserved for cytokine ELISA.

Western blotting

Protein samples from BMDMs and Müller glia (MIO-M1) cells were prepared by direct lysis of cells in radioimmunoprecipitation assay (RIPA) buffer containing protease and phosphatase inhibitor cocktails. For in vivo samples, two neuroretina (without RPE and choroid) were pooled in RIPA buffer, and homogenates were prepared using sonication followed by centrifugation at 15,000 x g for 20min at 4°C. The total protein concentration of the cell and retinal lysates were determined using a Micro BCA protein assay kit (Thermo Scientific, Rockford, IL). Protein samples (30-40 μg) were resolved on SDS–PAGE (8%–10%) and transferred onto a nitrocellulose membrane (0.45 μm) using a wet transfer system. Membranes were blocked in 5% (w/v) dried milk in tris-buffered saline (TBS)-Tween (TBST) for 1h at room temperature. Membranes were incubated with primary antibodies: anti-NRF2, anti-Hem-oxygenase 1 (HO-1), anti-NLRP-3, anti-HSP-90, (Cell Signaling Technology), anti-IRG1 (Aviva Bio-system), anti-Caspase-1 (Santa Cruz Biotechnology), as per the manufacturer’s instructions and followed by the appropriate anti-mouse/rabbit horseradish peroxidase (HRP)-conjugated secondary antibody. To visualize the protein bands, the membranes were exposed to Super signal West Femto chemiluminescent substrate (Thermo-scientific, Rockford, IL). For semiquantitative analysis, immunodetected protein band intensities were measured using ImageJ software (Rasband, W.S., ImageJ, U.S. National Institutes of Health, Bethesda, Maryland, http://rsb.info.nih.gov/ij/, 1997–2009).

Liquid chromatography coupled with a tandem mass spectrometer (LC-MS/MS)

Itaconate levels in the mouse retina/vitreous and cultured murine BMDMs were measured by LC-MS/MS methods at the Pharmacology and Metabolomics Core, Karmanos Cancer Institute, Wayne State University. Briefly, cell pellets were re-suspended in 1 mL of 80% MeOH (cooled in an ice bath), sonicated with a Misonix XL-2000 probe sonicator, and centrifuged at 28,672 x g for 10min at 4°C. The supernatant was transferred into a new 2 mL centrifuge tube. The pellet was washed with 200 μL of 80% MeOH (ice bath) by vortexing thoroughly and centrifuged at 28,672 x g for 10min at 4°C. Supernatant from two rounds of extraction was combined and dried in a CentriVap Refrigerated Centrifugal Concentrator. The residue was reconstituted with 50 μL of ddH2O, and a series of 20-time dilution samples were prepared for paralleled quantification by Waters Acquity H-class UPLC system connected to a Xevo TQ-xS triple LC-MS/MS system. The itaconate levels in the patient vitreous were measured by LC-MS/MS as indicated above at the Indian Institute of Chemical Technology, Hyderabad, India.

Preprocessing and annotation of metabolomic and transcriptomic data

The metabolomics raw data were acquired from LC-QTOF/MS and were converted to a mzData format using Mass Hunter Qualitative Analysis Software (Agilent) and then imported to xcms R package for preprocessing. The default parameters in xcms software were used. The preprocessed data were obtained with three features of the dataset, including the retention time, the mass-to-charge ratio (m/z), and peak intensity. The annotation was done using the CAMERA R package to annotate isotope peaks, adducts, and fragments. Lastly, the data were normalized for each sample as described in the statistical analysis section.

For transcriptomic data analysis, the original Affymetrix data was pre-processed using an oligo R package and later normalized and log-transformed using a multi-array average (RMA) method. Further, the CDF package was used for probe annotation of Affymetrix data. The probes of the normalized data were successfully mapped to Entrez Gene IDs and Gene Symbols by annotation package (annotate) in R. Wherever multiple probes matched a single gene symbol, we calculated the median values of those probes as the expression value for that gene.

Metabolomic and transcriptomic data analysis

Multivariate data analysis of Metabolomics data was performed using MetaboAnalyst R package. Identification of clusters to locate metabolites across various time points was performed using a supervised partial least-squares discriminant analysis (PLS-DA) method. The metabolites were ranked for PC1 and PC2, and in an absence of internal validation, the PLS-DA method avoided overfitting of metabolites. The PLS-DA found differences in metabolites between the various time points versus controls. Variable importance was calculated as a coefficient for the selection of each variable. In addition to the supervised method, the nonparametric Kruskal–Wallis test was also performed to measure the importance of each metabolite. The R package “caret” was used for the random forest method to classify key altered metabolites. All default values of the caret package were used. The top metabolites were obtained based on the “Gini” score ranking. For Affymetrix gene expression data, all statistical analyses were performed using the multiple R Bioconductor packages (http://www.Bioconductor.org). For analysis, the P-values were adjusted for multiple testing with the Benjamini and Hochberg’s method to control for false discovery rate (FDR). Probe sets showing at least a ± 2.5-fold change and an FDR < 0.05 were considered significant for the analysis. ggplot2 was used to generate the plots, R package VennDigram was used to generate the Venn diagram and the heatmap.2 package was used for heatmap generation from fold changes value.

Quantification and statistical analysis

Statistical analyses were performed using Prism version 8.1 (Graph Pad, San Diego, CA). Data were expressed as mean ± SD unless indicated otherwise and statistical significance was determined using either unpaired Student’s t test or two-way ANOVA or one-way ANOVA with multiple comparisons as indicated in the figure legends. A confidence interval of 95% was used for all statistical tests. A p < 0.05 was considered statistically significant.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
IRG1 (D6H2Y) Rabbit mAb	CST	77510S; RRID:AB_2799901
Anti-IRG1-Mouse	AvivaSysBio	ARP91387_P050
NRF2 (D1Z9C) XP (R) Rabbit mAb	CST	12721S; RRID:AB_2715528
HO-1 Rabbit mAb	CST	70081S; RRID:AB_2799772
NLRP-3 (D4D8T) Rabbit mAb	CST	15101S; RRID:AB_2722591
Caspase-1	Invitrogen	14-9832-82; RRID:AB_2016691
Chemicals, peptides, and recombinant proteins
4-Octyl Itaconate	Gifted Dr. Luke O Neil	Trinity College Dublin, Scotland
Dimethyl itaconate	Sigma-Aldrich	109533-100G
Itaconic acid	Sigma-Aldrich	I29204-100G
Vancomycin	Cayman Chemicals	15327
TRIzol reagent	Life technologies	15596018
FBS (Fetal bovine serum)	CPS Serum	FBS-500-HI
Peptidoglycan (PGN)	Sigma-Aldrich	77140-10MG
Lipoteichoic acid (LTA)	Sigma-Aldrich	L2515-5MG
Alpha toxin (α-toxin)	Sigma-Aldrich	616385-250UG
Toxic shock syndrome toxin-1 (TSST1)	Sigma-Aldrich	T5662
Critical commercial assays
Mice IL-1β ELISA kit	R&D System	DY401
Mice IL-6 ELISA kit	R&D System	DY406
Mice TNF-α ELISA kit	R&D System	DY410
Micro BCA Protein Assay Kit	Thermo Fisher Scientific	23235
SuperSignal West Femto Maximum Sensitivity Substrate	Thermo Fisher Scientific	34096
Maxima First Strand cDNA Synthesis Kit for RT-qPCR	Thermo Fisher Scientific	K1641
Radiant Green HiROX qPCR Kit	Alkali scientific	QS2050
Deposited data
RNA sequencing data	NCBI	GEO: GSE168928
Experimental models: organisms/strains
Mouse: C57BL/6J	The Jackson Laboratory	https://www.jax.org/strain/000664
Mouse: C57BL/6N-Acod1 em1(IMPC)J/J (Irg1−/−)	The Jackson Laboratory	https://www.jax.org/strain/029340
Mouse: B6.129X1-Nfe2l2tm1Ywk/J	The Jackson Laboratory	https://www.jax.org/strain/017009
Software and algorithms
Prism 8	GraphPad	https://www.graphpad.com
R version 3.6.3	R project	https://www.r-project.org/
MetaboAnalystR 3.0	Metaboanalyst	https://www.metaboanalyst.ca/
Other
shRNA NRF-2	Santa Cruz	sc-37049-V