Proinflammatory Cytokines Trigger the Onset of Retinal Abnormalities and Metabolic Dysregulation in a Hyperglycemic Mouse Model

Purpose Recent evidence has shown that retinal inflammation is a key player in diabetic retinopathy (DR) pathogenesis. To further understand and validate the metabolic biomarkers of DR, we investigated the effect of intravitreal proinflammatory cytokines on the retinal structure, function, and metabolism in an in vivo hyperglycemic mouse model. Methods C57Bl/6 mice were rendered hyperglycemic within one week of administration of a single high-dose intraperitoneal injection of streptozotocin, while control mice received vehicle injection. After confirming hyperglycemia, the mice received an intravitreal injection of either proinflammatory cytokines (TNF-α and IL-1β) or vehicle. Similarly, control mice received an intravitreal injection of either proinflammatory cytokines or vehicle. The retinal structure was evaluated using fundus imaging and optical coherence tomography, and retinal function was assessed using a focal electroretinogram (ERG), two days after cytokine injection. Retinas were collected for biochemical analysis to determine key metabolite levels and enzymatic activities. Results Hyperglycemic mice intraocularly injected with cytokines developed visible retinal vascular damage and intravitreal and intraretinal hyper-reflective spots two days after the cytokines injection. These mice also developed a significant functional deficit with reduced a-wave and b-wave amplitudes of the ERG at high light intensities compared to control mice. Furthermore, metabolic disruption was evident in these mice, with significantly higher retinal glucose, lactate, ATP, and glutamine levels and a significant reduction in glutamate levels compared with control mice. Minimal or no metabolic changes were observed in hyperglycemic mice without intraocular cytokines or in control mice with intraocular cytokines at 2 days post hyperglycemia. Conclusions Proinflammatory cytokines accelerated the development of vascular damage in the eyes of hyperglycemic mice. Significant changes were observed in retinal structure, function, and metabolic homeostasis. These findings support the idea that with the onset of inflammation in DR, there is a deficit in metabolism. Therefore, early intervention to prevent inflammation-induced retinal changes in diabetic patients may improve the disease outcome.


Introduction
Diabetic retinopathy (DR) is a progressive sight-threatening complication of diabetes that is clinically classifed into early nonproliferative and late proliferative stages [1]. Te current treatments halt the vascular damage in the late stage of DR. However, to prevent disease progression before the onset of signifcant visual dysfunction, novel treatment strategies should target the earlier molecular events that lead to vascular damage [2,3].
Currently, the understanding is that not only hyperglycemia but also concomitant retinal infammation is the major driving factors in the development of vascular abnormalities in DR [4][5][6]. A recent study from our group found that proinfammatory cytokines promote the development of severe vascular and functional abnormalities including vessel dilation, vessel beading, increased vessel tortuosity, and retinal oedema in the nonobese diabetic mice [7]. Tis demonstrated that the retina is susceptible to both hyperglycemia and proinfammatory conditions, afecting various interlinked metabolic pathways with a direct detrimental efect on retinal homeostasis [8]. In fact, previous studies have reported that the hyperglycemic retina undergoes apoptosis [9,10], blood-retina barrier breakdown [11], activation of NF-κB [10], and increased proliferation of microglial cells [12] only after infammation becomes apparent.
Although current treatments such as panretinal photocoagulation, focal laser treatments, and intravitreal injection of anti-vascular endothelial growth factor (VEGF) target the late-stage sight-threatening vascular problems of DR, there is a high chance of reoccurrence of the disease [13][14][15]. Given that there has been signifcant advancement in an artifcial intelligence-based system to detect clinical referable DR [16], our research contribution is the evaluation of early-stage disease to assess the consequences of retinal infammation on retinal energy demand and on key metabolic pathways, which could be therapeutically modulated to delay the onset of vascular problems. Terefore, in this study, we investigated the morphological, functional, and biochemical outcomes of retinal infammation in early streptozotocin (STZ)-induced diabetic mice with infammation in the eye.

Animals.
All animal experiments were approved by the University of Auckland Animal Ethics Committee (AEC 2205) and were conducted in accordance with the Association for Research in Vision and Ophthalmology (ARVO) statement on the use of animals in research. Sixto seven-week-old male C57BL/6 mice from the Vernon Jansen Unit, the vertebrate containment facility at the University of Auckland, were used in this study. Mice were bred and housed in standard cages under normal lightdark cycle conditions (12-h light (174 lux) and 12-h dark (<62 lux)) and had access to normal food and water ad libitum.

Development of the STZ-Induced Hyperglycemic Mouse
Model. Baseline body weight and nonfasting blood glucose levels were measured before the induction of hyperglycemia. Body size and glucose measurements were obtained from eight mice that received a single 150 mg/kg body weight intraperitoneal injection of STZ prepared in 0.1 M sodium citrate as previously described [17]. Eight control mice received a sham intraperitoneal injection of 0.1 M sodium citrate. Nonfasting blood glucose levels were measured one week after administering STZ and repeated for three consecutive days to confrm hyperglycemia using a tail prick test. All blood glucose measurements were done consistently between 9.00 and 10.00 am. Mice with blood glucose levels higher than 15 mmol/L were considered hyperglycemic [18]. Blood glucose readings higher than the upper limit of detection (27.8 mmol/L) of the blood glucose meter (Freestyle Optium H Glucometer, UK) were recorded at 27.8 mmol/L for data analysis. Although measurement of glycated hemoglobin level (HbA1c) is routinely used as an indicator to assess long-term glycaemic control, previous studies have reported that STZinduced diabetic animals show a small, nonsignifcant increase in HbA1C levels after 1 week of STZ injection [19,20]. As the present study aims to investigate the early changes in the retina, it was unlikely that HbA1C levels would be signifcantly elevated after 1 week of STZ administration. Terefore, sustained hyperglycemic was confrmed by measuring blood glucose levels and HbA1C level was not evaluated in this study.

Intravitreal Injection of Proinfammatory Cytokines.
An intravitreal injection of a proinfammatory cytokine cocktail was performed, as previously described [7]. Te STZ or vehicle injection protocol was repeated to create control and hyperglycemic mice with or without intravitreal cytokines. Briefy, a group of hyperglycemic mice (n � 6) were intravitreally injected in both eyes with 500 ng/ml each of proinfammatory cytokine TNF-α (#RMTNFAI, Termo Fisher Scientifc, Waltham, MA), and 500 ng/ml IL-1β (#RMIL1BI; Termo Fisher Scientifc) in a total volume of 1 μl (hyperglycemic mice with intraocular cytokines), while another group of six hyperglycemic mice received a vehicle (0.1 M PBS) intravitreal injection into both eyes (hyperglycemic group). Control mice received an intravitreal injection of proinfammatory cytokines (n � 6 control mice with intraocular cytokines) or a vehicle injection (n � 6 control mice, 12 eyes). Mugisho et al. (2018) observed that nonobese NOD diabetic mice develop clinical signs of DR as early as two days after intravitreal cytokine injection. Terefore, in this study, we investigated the structural, functional, and metabolic changes in the STZ-induced hyperglycemic mouse retina after two days of cytokine injection.

Fundus Imaging and Spectral Domain Optical Coherence
Tomography (SD-OCT) Imaging. Fundus and SD-OCT imaging of the mouse retina was performed using the Micron IV imaging system as described previously [21,22] to identify the appearance of clinical signs of DR including vessel beading and vessel tortuosity. Briefy, mice were anaesthetised by an intraperitoneal injection of ketamine (75 mg/kg body weight; Parnell Technologies, New Zealand) and domitor (0.5 mg/kg body weight; Zoetis, New Zealand). Te pupils were dilated using 1% tropicamide (Bausch and Lomb New Zealand Ltd, New Zealand) and the cornea was maintained hydrated with 1% Poly Gel ® lubricating eye gel (Alcon, Switzerland). Te animals were then placed on a heating pad set at 37°C to maintain body temperature for the duration of anaesthesia. Eyes that developed cloudiness of the lens during anaesthesia were not possible to be analysed in the fundus and OCT studies. Te fundus camera 2 Journal of Ophthalmology was carefully advanced towards the cornea until it came into contact with the poly gel. Te Micron IV in-built StreamPix 6 software was used to visualise and record the retinal fundus. Te retinal image was centered on the optic nerve head and blood vessel appearance was assessed. Vessel tortuosity was identifed by the abnormal twists and turns of the blood vessels, and vessel beading was the alternating areas of blood vessel constriction that gave a "beaded appearance" to the blood vessels. For the SD-OCT imaging, an ultrabroadband (160 nm) superluminescent diode centered at 830 nm was used as the light source to capture an image of 1024 pixels per A-scan with 1.8 μm longitudinal resolutions. Te horizontal line Bscan had 2 μm axial resolutions. Each OCT scan consisted of twenty averaged B-scan images acquired at approximately one-disc diameter from the optic nerve head in the superior and inferior retinal quadrants. Images were analysed using ImageJ software (National Institute of Health, Maryland, USA) to evaluate retinal layer thickness.

Focal Electroretinography (ERG).
Retinal function was assessed using the image-guided focal ERG attachment of the Micron IV imaging system [23]. Briefy, mice were darkadapted overnight and were handled under dim red-light illumination (λ max � 650 nm). For the ERG recording, mice were anaesthetised by an intraperitoneal injection of ketamine (75 mg/kg body weight) and domitor (0.5 mg/kg body weight); the pupils were dilated using 1% tropicamide and the cornea was maintained hydrated with 1% Poly Gel ® lubricating eye gel. Te animal was positioned on the heating pad and a subdermal ground platinum electrode was inserted into the base of the tail and the reference platinum electrode was inserted under the scalp skin at the midpoint between the eyes. Te objective lens containing the recording gold electrode was advanced towards the cornea and the retinal image under red-light illumination guided the area used to record the ERG.
A recording area of 0.75 mm in diameter was consistently chosen within the central superior retina, approximately one-disc diameter away from the optic nerve head. Te ERG recordings were in response to white-light fashes from a light emitting diode (LED) source (5 millisecond duration) and were recorded using the LabScribeERG 3 software (Phoenix Research Labs). Te light intensity used to elicit the ERG response ranged from −0.40 to 3.20 log candela seconds per square meter (log cd s/m 2 ), with 20 sweeps and 10 second interval for −0.4 to 1.40 log cd s/m 2 , 10 sweeps and a 20 second interval for 2.00 log cd s/m 2 , 3 sweeps and a 60 second interval for 2.60 log cd s/m 2 , and 2 sweep and 2 second interval for 3.20 log cd s/m 2 . Multiple individual responses from each sweep were averaged to obtain an improved signal-to-noise ratio [24]. One eye (left or right) per animal was used in the study. Te ERG responses were recorded from the superior retina as this was the quadrant mostly associated with retinal thinning during DR [25][26][27][28]. Te a-wave, b-wave, and implicit time as well as summed OP response were measured using the Micron IV software.
2.6. Biochemical Assays. For these biochemical assays data was collected from one or both eyes per experimental group. Retinal metabolites were measured using commercially available kits: glucose (Glucose-Glo ™ Assay, #J6021, Promega, Madison, USA), lactate (Lactate-Glo ™ Assay, #J5021, Promega), glutamate-glutamine (Glutamine/Glutamate-Glo ™ Assay, #J8021, Promega) and adapted to analysis of the retina according to the manufacturer's protocol. Briefy, the retinal glutamate and glutamine levels were measured by incubating the sample with and without glutaminase for 40 minutes, followed by a 60-minute incubation with the supplier's glutamate detection reagent in 1 : 1 ratio. Te luminescence signal generated by the assay was recorded using the EnSpire ® Multimode Plate Reader (Perkin Elmer, Massachusetts, USA). A glutamate standard curve was used to determine the concentration of glutamate and glutamine in the retinal supernatant. Te retinal glucose and lactate levels were determined similarly, wherein the sample was incubated with the respective detection reagent and the luminescence signal was recorded. Glucose, lactate, glutamate, and glutamine levels are reported as nmol/mg protein.
ATP levels were measured using the Adenosine 5′triphosphate (ATP) Bioluminescent Assay Kit (#FLAA, Sigma-Aldrich, Missouri, USA) according to the manufacturer's protocol. Briefy, a freshly prepared ATP assay mix was added to the retinal supernatant and the luminescence signal was recorded immediately. Te ATP levels are reported as pmol/mg protein.

Enzymatic Activities in the Retina.
Enzymatic activity was evaluated using retinal samples from six eyes per animal group. Glutamine synthetase (GS) activity was evaluated using the method employed in a previous study [29]. Briefy, the reaction mixture was prepared with 50 mM Imidazole-HCl bufer (pH 7.1), 7.6 mM ATP, 1.0 mM phosphoenolpyruvate, 50 mM MgCl 2 , 10 mM KCl, 40 mM NH 4 Cl 2 , 0.35 mM NADH, 0.1 M monosodium glutamate, 25 μg of pyruvate kinase, and 50 μg of lactate dehydrogenase in a volume of 1.0 ml. To eliminate traces of ADP and pyruvate, the reaction mixture was equilibrated at 30°C for 10 minutes. Te retinal samples were added to the reaction mixture in a 1 : 1 ratio and the rate of change in NADH absorbance was measured at 340 nm, 37°C using the EnSpire ® Multimode Plate Reader for 10 minutes. Te specifc enzyme activity of GS was normalised to total retinal protein concentration and expressed as µmoles per minute per milligram protein.
Te glyceraldehyde 3-phosphate dehydrogenase (GAPDH) activity Assay kit (#ab204732, Abcam, Australia) was used to evaluate GAPDH enzyme activity. In this assay, GAPDH catalyses the conversion of glyceraldehyde-3phosphate to 1,3-bisphosphate glycerate resulting in a stoichiometric NADH generation, which further reacts with the developer to form a coloured product with an absorbance maximum at 450 nm. Te retina was homogenised in GAPDH assay bufer and centrifuged at 10,000 × g for 5 minutes at 4°C to remove any cellular debris. Te retinal supernatant was added to the reaction mix containing GAPDH assay bufer, GAPDH developer, and GAPDH Journal of Ophthalmology 3 substrate in a 1 : 1 ratio and the colorimetric change was measured kinetically every 5 minutes, for a total of 20 minutes at 450 nm using a plate reader. A NADH standard curve was generated to calculate the specifc GAPDH activity. GAPDH activity was expressed as nmoles per minute per milligram of protein.

Statistical
Analysis. Normality of biochemical data distribution was confrmed using the Shapiro-Wilk normality test and QQ plots (see appendix). Te statistical signifcance for focal ERGs was determined using two-way ANOVA and a post hoc Dunnett's multiple comparison tests. Statistical signifcance for the biochemical assays was determined using one-way ANOVA and a post hoc Dunnett's multiple comparison tests. Te added efect of cytokines over high glucose conditions alone was evaluated by an unpaired t-test. A p value of less than 0.05 was considered statistically signifcant. Te statistical analysis was conducted using six-eight eyes per animal group. Te "n" value in the fgure legend indicates individual retinas, with at least six per experimental group. All statistical analysis was performed using GraphPad Prism 8. As both eyes from the same animal were used for some experiments, sensitivity analysis using the generalised estimating equations was performed to confrm the outcome obtained through traditional statistics. Tis analysis was performed using IBM SPSS statistics software (version 29).

STZ-Induced Hyperglycemic Mice had Reduced Body Weight and Elevated Blood Glucose Levels within One
Week. Body weight and nonfasting blood glucose levels were determined before and after STZ injection ( Figure 1). One week after STZ administration, there was a signifcant 9% (p � 0.006) decrease in body weight in all STZ-injected animals and nonfasting blood glucose levels were signifcantly elevated by an average of 188% (p � 0.0002), ranging from 18.6 to 27.8 mmol/L in hyperglycemic mice compared to control mice (7.8 to 10.1 mmol/L).

Proinfammatory Cytokines-Induced Vascular Changes in the Hyperglycemic Mouse Retina.
Fundus examination was performed two days after the intravitreal injection, and the appearance of the retinal vasculature was assessed ( Figure 2). Control mice (Figure 2(a)), control mice with intraocular cytokines (Figure 2(b)) as well as hyperglycemic mice without intraocular cytokines (Figure 2(c)) did not show signs of blood vessel damage or changes in the retinal vasculature. However, intravitreal injection of cytokines to hyperglycemic mice resulted in blood vessel tortuosity in three out of ten eyes and blood vessel beading in fve out of ten eyes two days after the intravitreal injection (Figure 2(d)).

Proinfammatory Cytokines-Induced Retinal and Vitreous
Hyper-Refective Spots in Hyperglycemic Mice. SD-OCT imaging was performed to evaluate structural changes in the retina two days after intravitreal injection ( Figure 3). Te OCT scans showed that both hyperglycemic mice with and without intraocular cytokines developed distinct hyperrefective spots 2 days after the intravitreal injection. Two out of eight eyes (25%) of the hyperglycemic mice without intraocular cytokines and seven out of ten eyes (70%) of the hyperglycemic mice with intraocular cytokines developed small (less than 20 μm) hyper-refective spots, while six out of ten eyes (60%) of the hyperglycemic mice with intraocular cytokines developed large (greater than 50 μm) hyperrefective spots (Figure 3). Tese hyper-refective spots were located near the outer plexiform layer and within the inner nuclear layer. Furthermore, three out of ten eyes (30%) of the hyperglycemic mice with intraocular cytokines developed vitreal hyper-refective spots (Figure 3(g)), which was not observed in any of the hyperglycemic mice without intraocular cytokines. Intravitreal cytokine injection did not induce any retinal abnormalities in control mice. Retinal layer thickness measurements confrmed no retinal thinning in any of the groups compared to control mice (Supplementary Figure 2).

Proinfammatory Cytokines Caused Retinal Functional
Defcit in Hyperglycemic Mice. Focal ERG response was recorded to evaluate retinal function two days after intravitreal injection. Hyperglycemic mice with and without cytokines had a signifcant reduction in the ERG awave and b-wave amplitudes compared to control mice. Te a-wave amplitude of hyperglycemic mice without cytokines was not reduced signifcantly except at light intensity 3.2 log cd s/m 2 (two-way ANOVA, p = 0.033, Figure 4(a)) and the b-wave amplitude was not reduced except at light intensity 2.6 log cd s/m 2 (p = 0.030, Figures 5(c) and 4(b)). Whereas, intravitreal injection of proinfammatory cytokines to hyperglycemic mice significantly further reduced the a-wave amplitude at light intensities 2.6 (p = 0.001) and 3.2 log cd s/m 2 (p = 0.0007, Tere were no statistically signifcant diferences in the summed OP response in any experimental mouse group compared to control mice (Figure 4(e)), although the individual OP response appeared to be reduced in hyperglycemic mice injected with proinfammatory cytokines.
Hyperglycemic mice with intraocular cytokines showed signifcantly reduced GAPDH activity by 20% (p � 0.031), while GS activity remained unchanged compared to control mice. However, intravitreal injection of cytokines to hyperglycemic mice caused a signifcant decrease in GS activity by 24% (p � 0.004) compared to hyperglycemic mice without cytokines (Figure 7).

Discussion
In this study, it was confrmed that proinfammatory cytokines trigger the development of vascular and retinal abnormalities and metabolic dysregulation characteristic of DR in a STZ-induced hyperglycemic mouse model. Te STZ-induced diabetic mouse model is one of the most widely used diabetic models [30], and signifcant DR-related retinal and vascular signs were evident once exposed to a proinfammatory environment. In the STZ mouse model, proinfammatory cytokines were delivered as a single injection, and combined with hyperglycemia, this single dose of proinfammatory cytokines leads to ocular vascular damage within two days. Hyperglycemic mice did not develop any severe ocular vascular abnormalities for the duration of this study and confrmed that STZ-induced diabetic mice rarely develop serious vascular abnormalities until 6 to 12 month posthyperglycemia [31][32][33]. Te efects of STZinduced hyperglycemia are seen after 6 months, when retinal physiological and biochemical changes are observed [33]. However, the intravitreal administration of proinfammatory cytokines to hyperglycemic mice accelerated the onset of changes, and resulted in vessel beading, increased vessel tortuosity, and retinal and vitreal hyper-refective spots within two days. Tese vascular changes are considered early indicators of microvascular damage in DR [34] and were in line with previous fndings from Mugisho et al. (2018) [7], wherein intravitreal cytokine injection to nonobese diabetic (NOD) mice triggered the development of severe vascular abnormalities within one week. Te development of retinal and vitreal hyper-refective spots is in concurrence with previous fndings in patients with early DR [35,36] and in the nonobese NOD DR mouse model [7] and  Figure 1: Development of the hyperglycemic mouse model. Body weight and blood glucose levels were measured before and after the STZ injection. STZ administration caused a signifcant decrease in body weight (a) and a signifcant increase in blood glucose levels (b). Te unpaired t-test test was used to determine statistical signifcance. Values represent mean ± SEM (n � 8 mice per group). * * p < 0.01, * * * p < 0.001. Abbreviation-C: control mice, HG: hyperglycemic mice. DR rat model [37]. Tese previous studies have shown that intraretinal hyper-refective spots in the inner retina could be associated with microaneurysms and macroaneurysms based on their size [37,38], while other studies have demonstrated that the appearance of hyper-refective spots in the inner nuclear layer is a defnitive marker of vascular abnormality [35,38]. Although we did not collect histological sections for the hyper-refective spots in this model, hyper-refective spots were found to be associated with activated microglial cells that are responsible for mediating the early infammatory response in DR [35,[39][40][41]. Moreover, the precise location of these hyper-refective spots in the retina can be predictive of disease progression, as microglial cells tend to migrate from the inner retina towards the photoreceptors with time in the diabetic retina [35]. Although less frequent, vitreal hyper-refective spots are indicative of a more severe form of DR and are commonly reported in patients with severe proliferative DR [42]. Tese vitreal hyper-refective spots are considered infltrating macrophages that contribute to endothelial cell damage and eventually cause vascular breakdown in DR [43][44][45]. Hence, low-grade, subclinical infammation seems to be a critical factor for the development of vascular lesions in DR [7,43].  Journal of Ophthalmology In addition to vascular and retinal abnormalities, intravitreal injection of cytokines triggered retinal functional defcits in hyperglycemic mice. Our fndings were consistent with previous reports [46,47] that reduced ERG a-wave and b-wave amplitudes indicative of impairment of photoreceptor and inner retinal function were early events in DR pathology. Although photoreceptors are generally not considered afected in DR (most likely due to the substantial distance from the retinal vasculature), recent reports have found that they are important mediators of oxidative stress and infammation in DR [48,49]. In fact, it was observed that patients with retinitis pigmentosa, a rare genetic disorder causing photoreceptor degeneration, were less likely to develop DR associated retinal complications in diabetic patients [50,51]. Moreover, in addition to Müller cells and microglia, photoreceptors were found to be major contributors of oxidative stress and infammation in the retina during DR [48] and were found to be apoptotic in STZinduced diabetic rats [52]. Hence, early metabolic dysregulation in the photoreceptors and within the downstream inner retinal neurons [53] are the likely cause of the reduced a-wave and b-wave amplitudes in DR. However, in spite of the reduced a-wave and b-wave responses, the summed OPs remained unafected in these mice, implying that the inhibitory feedback pathway initiated in the inner retina by the amacrine cells remained unafected [54]. However, we hypothesize that signifcant change in OP amplitude becomes apparent with increasing DR duration.
Altered retinal function is often indicative of an underlying metabolic imbalance [55,56]. Our previous studies (e) Summed OP amplitude. Two days postintravitreal injection, there was a signifcant reduction in a-wave amplitude in the hyperglycemic mice at 3.2 log cd.s/m 2 (represented by the green asterisk) compared to control, while the hyperglycemic mice with intraocular cytokines showed further reduction of the a-wave amplitude at 2.6 and 3.2 log cd.s/m 2 (represented by the red asterisk) compared to control. Likewise, there was a signifcant reduction in bwave amplitude in the hyperglycemic mice at 2.6 log cd.s/m 2 (represented by the green asterisk) compared to control, while the hyperglycemic mice with intraocular cytokines showed further reduced b-wave amplitude at 2.0, 2.6, and 3.2 log cd.s/m 2 (represented by the red asterisk) compared to control. Tere were no signifcant changes in the OP amplitude in the experimental group compared to control mice. Statistical signifcance was determined using two-way ANOVA followed by post hoc Dunnett's multiple comparison tests. Values represent mean ± SEM (n � 6). * p < 0.05; * * p < 0.01; * * * p < 0.001.
have confrmed that co-exposure of mouse retinal explants to hyperglycemia and proinfammatory cytokines causes biochemical and neurochemical changes in retinal neurons and Müller cells, wherein the metabolite levels of glucose, lactate, ATP, and glutamate are altered, in addition to the redistribution of glutamate and glutamine within the inner retinal neurons [57,58]. As expected, hyperglycemic mice with intraocular cytokines showed altered glucose, lactate, ATP, glutamate, and glutamine levels, suggesting that cytokines can trigger metabolic dysregulation in hyperglycemic mice within two days. In addition, we also found that the activities of GAPDH and GS were altered in hyperglycemic mice with intraocular cytokines, and these enzymes are critical in regulating glucose and glutamate-regulating pathways [58]. Tis implies that infammation could be the causative factor of early functional defcits in DR. We observed the same outcome in an in vitro model of high glucose plus cytokines [58], where lactate accumulation may also be mediated by pyruvate recycling when GAPDH activity is reduced [59,60] or shunting through AGE pathways [61]. Tere is evidence of complete oxidative degradation of glutamate to form pyruvate and then lactate via a nonglycolytic route [62]. In support of this, previous studies have reported that metabolic pathways including glycolysis [63][64][65] and glutamate pathways [56], alter retinal function. Tese early functional defcits were also reported in diabetic patients with no apparent DR signs and without any signifcant visual dysfunction, suggesting that functional defcits may precede the manifestation of clinical signs of DR but are more likely to be a strong indicator of DR severity and the magnitude of functional loss [66]. Although retinal thinning is a clinical marker of neurodegeneration and is an early occurrence in DR [28,67], no changes in the retinal layer thickness at both time points were observed, suggesting that functional and metabolic changes in DR may precede the occurrence of neurodegeneration. Low glutamate and high glutamine levels observed in our model were not due to elevated glutamine synthetase activity. In such a scenario, glutamate hypersensitivity (a cause for retinal damage) is possible as overall decrease in glutamate refects a rapid decrease in its synthesis from carbon dioxide, rather than from glutamine as described by Gowda et al. [68] in diabetic conditions. Our previous studies showed this decrease in glutamate in bipolar cells [58] and   confrmed the decrease in glutamate and increase in glutamine. We have also shown that despite low glutamate levels, there are specifc retinal areas of glutamate accumulation in Müller endfeet, and this would be the areas of glutamate hypersensitivity in DR. Tus, we think that glutamine levels refect the state of the diabetic retina, but we do not have evidence that glutamine accumulation is damaging the retina. Rajagopal et al. (2016) have shown that development and progression of DR in fat-fed mice are also associated with hyperglycemia and infammation. Functional ocular defcits were characterised by electroretinographic dysfunction observed at beginning of 6 months due to glucose intolerance with microvascular disease appearing at 12 months. Interestingly, infammasome activation was reported at 3 months, before the development of systemic glucose intolerance, electroretinographic defects, or microvascular disease. Tese results reinforce our suggestion that disease in the diabetic environment may progress through infammatory stages long before the development of vascular lesions [69].
Terefore, the fndings from our study emphasise the importance of early intervention targeting mediators of infammation to slow down or prevent the progression of Although hyperglycemic mice without cytokines showed signifcantly high retinal glucose levels, intravitreal injection of cytokines to hyperglycemic mice further increased glucose, lactate, ATP, glutamine levels, and reduced glutamate levels compared to control mice (represented by the asterisk). Moreover, intravitreal injection of cytokines to hyperglycemic mice caused a signifcant change in glucose, lactate, ATP, and glutamine levels compared to hyperglycemic mice without intraocular cytokines (represented by the hashtag). One-way ANOVA followed by post hoc Dunnett's multiple comparison tests was used to determine statistical signifcance compared to controls. Te unpaired t-test was used to determine the added efect of cytokines on hyperglycemic mice. Values represent mean ± SEM (n � 6-8 eyes). * and # p < 0.05, * * and ## p < 0.01, * * * and ### p < 0.001, * * * * p < 0.0001. Abbreviations-C: control mice, Cyt: mice with cytokines, HG: hyperglycemic mice, HG + Cyt: hyperglycemic mice with cytokines.
DR. In support of this, recent studies have shown that metabolic inhibitors such as the polymethoxylated favone Niboletin [70] and a Chinese herbal formula, Shuangdan Mingmu capsule [71], promote the upregulation of GAPDH activity in DR, while resveratrol treatment [72] prevents GS downregulation in DR. Tese metabolic inhibitors could potentially prevent blood-retina barrier breakdown, oxidative stress-induced apoptosis of pericytes, and glutamate excitotoxicity in DR. Moreover, immunological therapy is fast gaining popularity to treat retinal infammation in DR [73,74], and drugs of particular interest are the Connexin-43 hemichannel inhibitors. Previous study from our group has shown that Connexin-43 hemichannel blocker mitigates retinal infammation by efectively blocking the NLRP3 infammasome pathway in an in vivo mouse model of DR [37,75,76].
In conclusion, this study is consistent with the hypothesis that proinfammatory cytokines aggravate the early morphological, functional, and metabolic imbalance in hyperglycemic mice and opens up the opportunity for a wider array of possible therapies concomitantly targeting infammation and early metabolic dysregulation in DR. Moreover, these fndings suggest that early visual dysfunction precedes retinal neurodegeneration and the appearance of severe vascular pathology in DR and are indicative of alerted retinal bioenergetics. Terefore, we have compiled evidence to suggest that it is crucial to control retinal infammation in diabetics to prevent or delay the rapid worsening of retinal metabolism, which may trigger functional changes and may progress with time if not treated early on.

Data Availability
Te data supporting the fndings of this study are included in the fgure and supplementary data.

Conflicts of Interest
Te authors declare that there are no conficts of interest regarding the publication of this paper.

Acknowledgments
Te authors would like to thank Mr. Zhenqiang Wu (School of Population Health, Section of Epidemiology and Biostatistics, Te University of Auckland) for his support with the statistical analysis. Gaganashree Shivashankar was supported by the Dean's International Doctoral Scholarship from the Faculty of Medical and Health Sciences, University of Auckland, New Zealand. Tis study was partially funded by Te New Zealand Optometric Vision Research Foundation (grant no. 3722031).

Supplementary Materials
Supplementary Figure 1. Normality tests, p > 0.05 on the Shapiro-Wilk test is indicative of a distribution of data that is not statistically signifcant from a normal distribution. No change in the retinal layer thickness was observed in experimental mice compared to control mice 2 days after intravitreal injection. One-way ANOVA followed by post hoc Dunnett's multiple comparison tests was used to determine signifcance. Values represent mean ± SEM (n = 6 eyes). Abbreviation-(C) control mice, Cyt: mice with intraocular cytokines, HG: hyperglycemic mice, HG + Cyt: Hyperglycemic mice with intraocular cytokines showed signifcantly reduced GAPDH activity compared to control mice. Although the GS activity in hyperglycemic mice with intraocular cytokines remained unchanged compared to control mice, intravitreal injection of cytokines to hyperglycemic mice caused a signifcant decrease in GS activity compared to hyperglycemic mice without intraocular cytokines. One-way ANOVA followed by post hoc Dunnett's multiple comparison tests was used to determine statistical signifcance. Te unpaired t-test was used to determine the added efect of cytokines on hyperglycemic mice. Values represent mean ± SEM (n � 6 eyes). * and # represent p < 0.05. Abbreviation-C: control mice, Cyt: mice with cytokines, HG: hyperglycemic mice, HG + Cyt: hyperglycemic mice with cytokines.