Multi-omics analysis of potential metabolic networks linking peripheral metabolic changes to inflammatory retinal conditions in STZ-induced early diabetic retinopathy

Abstract

Background

Diabetic retinopathy (DR), a leading cause of blindness among working-age adults, lacks targeted therapies besides glucose management. Early retinal lesions are linked to serum metabolites, but the underlying peripheral regulatory networks is unclear.

Methods

We first established a streptozotocin (STZ)-induced mouse model of early DR exhibiting retinal inflammation characteristics. This study employed an integrative approach, combining retinal and serum transcriptomic and metabolomic profiles with genome-wide association study (GWAS) data, to identify peripheral metabolites potentially linking early retinal lesions.

Results

STZ-induced mice exhibited retinal inflammation and metabolic dysregulation. Metabolites including glucose, sorbitol, and mannitol were altered in both serum and retina, implicating their potential involvement in retinal inflammation. Utilizing GWAS data of diabetic patients, we further explore the potential the upstream regulation of shared metabolites and their peripheral pathways potentially instigating early retinal inflammation through metabolite-related genes correlated with single nucleotide polymorphisms. Key enzyme genes including HK1, HKDC1, AKR1B1 in hyperglycemic pathway, CEL and HMGCR in cholesterol pathway, and ACSL1, PPT2 in palmitic acid pathway, may connect the metabolic network of hyperglycemia, hyperfructosemia and disrupted lipid metabolism to retinopathy.

Conclusion

This study elucidates the upstream regulatory network of peripheral serum metabolites associated with early retinal lesions. Specifically, the SNPs in key peripheral enzyme genes may exert remote effects on retinal inflammation in DR. This finding provides insights into the systemic metabolic management and offering peripheral precise early detection and treatment.

Graphical abstract

Highlights

• •Retinal metabolic network alterations during early retinal inflammation of diabetic retinopathy.
• •Peripheral metabolic rewiring links retinal metabolism shifts in initial phase of diabetic retinopathy.
• •Peripheral hyperglycemia and hyperfructosemia may initiate retinal inflammation.
• •GWAS identifies potential peripheral key regulatory factors for retinal inflammation.

1. Introduction

Diabetic retinopathy (DR) is the leading cause of blindness in the working-age population, and represents the most frequent etiology of vision impairment among individuals with diabetes [[1], [2], [3]]. Data from the 25-year follow-up of the type 1 diabetes mellitus (T1DM) cohort show that almost 97 % patients developed retinopathy over time, with a third to a half developing vision-threatening disease [4,5]. In 2020, the global number of adults with DR was estimated to be 103 million, and this number is projected to rise to approximately 160 million by 2045 [6]. Moreover, the escalating costs associated with managing DR patients will pose a significant challenge to public health system [7,8]. Hence, the prevention and treatment of DR require effective and economical strategies.

Currently, the most effective approach for reducing the incidence of DR remails early screening coupled with glucose management. Maintaining HbA1c<7 % significantly reduces DR progression risk [[9], [10], [11]], with intensive glycemic control lowering DR incidence and progression rates by approximately 50 % [11,12]. However, glucose-well controlled diabetic patients (GW-DR) lack early diagnostic and therapeutic strategies. Unique serum metabolomics profiles in GW-DR patients identify ethanolamine as a potential early biomarker and therapeutic target, supported by its anti-inflammatory effect in DR models [13]. Clinical metabolomics studies identify additional circulating candidates like 12-HETE, 2-piperidone [14,15], altered nucleosides (cytosine, cytidine, thymidine) [16], and amino acids (glutamine, glutamate, branched-chain amino acid) in DR patients [[17], [18], [19]]. These findings confirmed that non-glucose peripheral metabolites are linked to DR pathogenesis. However, the synergistic changes between peripheral circulation and in-situ retinal metabolites during early DR remain unclear. Elucidating this peripheral-retinal axis could enable glucose-independent early screening and precise interventions for DR.

Current research on the molecular pathophysiology of DR indicates that retinal inflammation is a hallmark pathological feature of its early-stage [20]. During this initial phase, pro-inflammation cytokines like interleukin-1β (IL-1β) and interleukin-6 (IL-6) are upregulated throughout the retinas of DR patients [[21], [22], [23], [24]]. These cytokines promote angiogenesis, increase vascular permeability, and facilitate retinal neovascularization [[25], [26], [27]]. Similarly, IL-1β and IL-6 contribute to the pathogenesis of DR in STZ-induced mice models [28,29]. Single-cell sequencing has identified active microglia as the primary source of IL-1β within the retinas of STZ mice [30]. The research has found that inflammation-related metabolic reprogramming plays a significant role in the course of DR [31]. Aberrant glucose metabolism and subsequent retinal inflammation lead to DR [31,32]. Peripheral metabolites may enter retinal cells via transporters or initiate downstream signaling pathways in retinal cells through receptors, which provides a possibility for uncovering peripheral metabolites potentially linked to retinal lesions. In addition, Genome-Wide Association Studies (GWAS) is a powerful tool to identify genetic loci conferring susceptibility to DR [[33], [34], [35]].

This study investigated coordinated metabolic changes and regulatory networks in peripheral metabolism associated with early retinal inflammation. Using metabolomics and transcriptomics, we analyzed metabolic alterations in the retina and peripheral serum of early DR mice. We focused on serum metabolites potentially triggering retinal inflammation via receptors or entering the retina through transporters. Integrated the analysis of serum metabolites and metabolic enzyme genes with single nucleotide polymorphisms (SNPs) reported in diabetic patients. Uncovering metabolic triggers and regulatory mechanisms beyond glycemic control may provide novel insights into DR pathogenesis and inform multifaceted precision prevention and treatment strategies.

2. Materials and methods

2.1. Animals

Murine work was carried out in accordance with The Animals (Scientific Procedures) Act 1986 (Amended 2012). Animals were maintained under 12:12 light-dark cycle with ad libitum access to standard chow and water, with daily welfare monitoring. All procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Shanghai General Hospital (Approval No. 2021AW039) and conducted in compliance with ARRIVE guidelines 2.0. No experimental procedures were performed on live mice and euthanasia was performed in accordance with ARRIVE guidelines.

2.2. STZ-induced diabetic mice

Previously published procedures were used for preparing diabetic model mice [36]. Male C57BL/6J mice (6-week-old, specific pathogen-free) were purchased from Shanghai Model Organisms Center, Inc. (https://www.modelorg.com/). Diabetes was induced by intraperitoneal administration of streptozotocin (STZ; Sigma, cat. V900890) dissolved in 0.1 M citrate buffer (pH 4.5). Following 12-h fasting with free access to water, mice received daily STZ injections (55 mg/kg body weight) for 5 consecutive days (n = 15/group). Age-matched control mice received equivalent citrate buffer injections (n = 10). Non-fasting blood glucose levels were measured weekly via tail vein sampling using Accu-Chek Active glucometer (Roche Diagnostics). Animals exhibiting sustained hyperglycemia (≥300 mg/dL) at 7 days post-injection were included in the diabetic group. The timeline of the animal experiments is shown in Fig. 1A.

Fig. 1.

Enhanced inflammatory cytokine expression in early STZ-induced diabetic retinopathy mouse retinas. (A) Flowchart illustrating the experimental design for a 2-month induction of diabetic retinopathy via streptozotocin (STZ) in 7-week-old mice. (B) Blood glucose levels of mice from two groups after 8 weeks of STZ-induction. Data are presented as mean ± SEM, *p-value <0.05, ** p-value <0.01, *** p-value <0.001, n = 5 in each group. (C) qPCR analysis of IL-6 and IL-1β gene expression in control mice versus STZ-induced diabetic mice (n = 4 in each group). Data are presented as mean ± SEM, * p-value <0.05, ** p-value <0.01, ***p-value <0.001.

2.3. Ocular tissue collection and retinal isolation

STZ-induced diabetic mice and control mice were anesthetized by intraperitoneal injection of tribromoethanol/tert-amyl alcohol solution (2.5 % w/v tribromoethanol in 2.5 % v/v tert-amyl alcohol). Retinas were gently detached from the retinal pigment epithelium using a stereomicroscope (Leica M80, Germany) and transferred to a 1.5 ml microcentrifuge tubes. All retinas were flash-frozen in liquid nitrogen within 2 min of dissection and stored at −80 °C in a cryogenic freezer.

2.4. Quantitative real-time PCR

Total RNA was isolated from mice retinas using the RNApure Micro Kit (JIANSHI Biotech, China, cat. TR50). RNA purity and concentration were quantified with the NanoDrop 2000c spectrophotometer (Thermo Fisher Scientific, USA) with demonstrating 260/280 nm absorbance ratios >2.0 selected for downstream processing. cDNA synthesis was performed using RT Master Mix (Vazyme, China; Cat# R212) according to the manufacturer's protocol. qRT-PCR analyses were conducted in triplicate using the ViiA™ 7 Real-Time PCR System (Applied Biosystems, USA). The thermal cycling protocol consisted of: initial denaturation at 95 °C for 30 s, followed by 40 cycles of 95 °C for 5 s, 60 °C for 5 s, and 72 °C for 34 s. The PCR protocol used was 40 cycles of 35 s at 95 °C, 5 s at 60 °C, and 34 s at 72 °C. Primers sequences used in this study are detailed in Table 1.

Table 1.

Primers sequences of qPCR.

Genes	Forward primer (5′-3′)	Reverse primer (5′-3′)
β-actin	CATTGCTGACAGGATGCAGAAGG	TGCTGGAAGGTGGACAGTGAGG
IL-1β	TGGACCTTCCAGGATGAGGACA	GTTCATCTCGGAGCCTGTAGTG
IL-6	TACCACTTCACAAGTCGGAGGC	CTGCAAGTGCATCATCGTTGTTC

2.5. Untargeted metabolomics by UHPLC-MS/MS

The samples of retina (n = 3) and serum (n = 5) from STZ-induced diabetic mice (n = 3) and control mice were separated. The metabolites were eluted and separated by linear gradient mode. The metabolites were ionized and mass spectrometry data were collected by QExactive quadrupole-orbital field mass spectrometer (QE) of Thermo Fisher Scientific Shier Company. The metabolites were ionized in positive ion mode (HESI+) and negative ion mode (HESI-) respectively by using a heated electrospray ionization (hesi) ion source. At the same time, the data-dependent identification (DDA) mode is used to collect the secondary mass spectrum fragment information of at most six parent ions in each scanning cycle. The collision energies of HCD are 15, 30 and 45 eV, respectively, and the mass spectrum resolution is 17,500, and the AGC threshold is 1 × 10⁵.

2.6. Untargeted metabolomics by GC-MS

The samples of retina (n = 3) and serum (n = 5) from STZ-induced diabetic mice (n = 3) and control mice were collected. Metabolomic profiling was performed using an Agilent 7890A/5975C GC-MS system (Agilent Technologies, USA) equipped with an OPTIMA® 5 MS Accent fused-silica capillary column (Macherey-Nagel, Germany). The samples were separated and analyzed by a gradient heating program. EI energy is set to 70 electron volts. The data with the mass-to-charge ratio (m/z) of 50–600 were collected by full scanning.

2.7. RNA sequencing analysis

Retinas tissues from STZ-induced diabetic mice (n = 3) and age-matched control mice (n = 3) were homogenized in Trizol Reagent (Thermo Fisher Scientific, Cat 15596026). Total RNA was isolated following the manufacturer's protocol. Equimolar amounts of RNA from three biological replicates per group were pooled to minimize individual variability. Transcriptome sequencing was performed by Genergy Biotechnology (Shanghai, China). The detected genes were screened by the fold change (FC). The genes with thresholds of FC > 5 or FC < 0.2 and p-value <0.05 were selected as differentially expressed genes (DEGs). Gene Ontology (GO) term enrichment was analyzed through DAVID Bioinformatics Resources (https://david.ncifcrf.gov/). Significant terms were selected based on EASE score p < 0.01 and false discovery rate (FDR) < 5 %. Visualization was performed using Bioinformatics.com.cn online platform (OmicShare Tools V8.2).

2.8. Protein–protein interactions analysis

The Search Tool for the Retrieval of Interacting Genes (STRING https://cn.string-db.org/) database was employed to establish protein interaction networks. Top 500 DEGs were selected for network analysis. The STRING output files (TSV format) were imported into Cytoscape v3.9.1 (National Institute of General Medical Sciences, USA) for visualization.

2.9. Genome-wide association study (GWAS) analysis

Single nucleotide polymorphism-trait association (risk site related to diabetes/diabetic retinopathy) was downloaded from the GWAS directory on December 23, 2024, and the keywords were “diabetes” and “diabetic retinopathy” respectively.

2.10. Statistics analysis

Data are shown as means ± SEM via the one sample t-test and visualized by GraphPad Prism 9.5. p-value <0.05 was considered statistically significant. DEGs of RNA-seq and differential metabolites with the criterion of FC > 1.2 or <0.83 and p-value <0.05. Principal Component Analysis (PCA), Orthogonal Partial least squares Discriminant Analysis (OPLS-DA) and KEGG pathways of metabolites between the STZ and control mice were analyzed with MetaboAnalyst (https://www.metaboanalyst.ca/).

3. Results

3.1. Retinal inflammation occurs in STZ-induced early diabetic retinopathy mice

The experimental timeline and schedule for streptozotocin (STZ)-induced early DR mouse model was shown in Fig. 1A. Blood glucose levels was significantly increased in STZ mice compared with control after 8 weeks. (Fig. 1B). To assess retinal inflammation, we measured the expression levels of inflammatory factor. qRT-PCR analysis revealed elevated IL-6 and IL-1β mRNA levels in the retinas of STZ-induced mice (Fig. 1C). These findings confirmed elevated inflammatory factor expression during STZ-induced early DR.

3.2. Transcriptomic alterations in the retinal inflammatory and metabolic network of STZ-induced early diabetic retinopathy mice

To explore the potential regulatory factors and inflammation in early DR, we performed RNA-seq analysis. A total of 444 differentially expressed genes (DEGs) were identified with the criterion of fold change (FC) > 5 and p-value <0.05 in the retinas of STZ mice compared to control. Gene Ontology (GO) analysis revealed pathways significantly associated with DR (Fig. 2A). Immune system process and inflammatory response were enriched in biological process (BP). Extracellular region was the most significant in the cellular component (CC), and protein binding, lipid transfer activity and cytokine receptor activity are enriched in the molecular function (MF).

Fig. 2.

RNA-sequencing reveals the metabolic and inflammatory signatures in early diabetic retinopathy mouse retinas. (A) 442 differential genes (FC > 5) were analyzed by Gene Ontology (GO), analysis showed the top 5 differential pathways were enriched in Biological Process (BP), Molecular Function (MF) and Cellular Component (CC) between STZ-induced group and control group respectively. (B) Protein-Protein Interaction (PPI) Networks diagram for the Top 200 up-regulated differentially expressed genes based on fold change values.

The protein-protein interaction was analyzed to further explore the molecular regulatory network in retina (Fig. 2B). Top 10 hub genes identified were IL-6, Alb, Ccll5, Serpinc1, Fgf8, Casp1, Apoa1, H3c8, Cyp2e1 and Cd163. Among these, IL-6, Ccl5, Casp1 and Cd163 were related to inflammation, while Alb, Cyp2e1 and Apoa1 were involved in metabolic process. These results indicated that inflammatory responses and dysregulation of metabolic processes occurred in early DR model, suggesting a close link between metabolism and inflammation.

3.3. Early metabolic alterations in the retinas of STZ-induced mice

To further investigate the metabolic alterations associated with the inflammatory process in DR, we conducted metabolomic profiling of retinas from STZ and control mice. PCA and OPLS-DA analysis revealed a clear separation between the two groups, indicating distinct metabolic differences (Fig. 3A and B). A total of 51 differential metabolites were identified between the STZ and control mice (FC > 1.20 or < 0.83, p-value <0.05). Among them, 40 metabolites including glucose, sorbitol, mannitol, fructose-6-phosphate, lactic acid and adenosine-5′-triphosphate were significantly upregulated in STZ mice, while 11 metabolites were significantly down-regulated, including PE38: 7, PE16: 1–22: 6, and aconitic acid. (Fig. 3C).

Fig. 3.

Metabolic profiling of the retina in early-stage diabetic retinopathy mice. (A) Two-dimensional score plot of the Principal Component Analysis (PCA) model comparing the metabolic profiles of the control (Red hollow circle) with STZ-induced diabetic (Green hollow circle) group based on ultra-high performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/MS) and gas chromatography-mass spectrometry (GC-MS) techniques. (B) Score plot of the orthogonal partial least squares-discriminant analysis (OPLS-DA) model based on UHPLC-MS/MS and GC-MS. (C) Volcano plot illustrating differential expression of metabolites between the two groups. Metabolites that are down-regulated and up-regulated metabolites are depicted in blue and red, respectively. Non-significant changed metabolites are represented by gray dots (p-value <0.05, FC > 1.2 or < 0.83). (D) Metabolome pathway analysis revealed that the differential metabolites between the STZ-induced diabetic group and the control group were associated with the enrichment of 25 metabolic pathways. The top 5 metabolic pathways with the most significant p-value are highlighted and labeled (p-value <0.01).

KEGG pathway enrichment analysis identified 25 significantly affected metabolic pathways, with the top 5 being fructose and mannose metabolism, pentose phosphate pathway, glycolysis/gluconeogenesis, galactose metabolism and arginine biosynthesis (Fig. 3D). These findings demonstrated distinct alterations in retinal metabolite profiles and associated metabolic pathway in DR, providing mechanistic insights into the pathophysiological processes underlying disease progression.

3.4. Early serum metabolomic characteristics in STZ-induced mice

Building on our previous finding of unique serum metabolomics in GW-DR patients, particularly the potential of ethanolamine as an early clinical warning [13], we investigated systemic metabolic changes concurrent with early DR retinal inflammation in STZ mice. Serum metabolomics profiling revealed clear separation between the STZ and control mice via PCA and OPLS-DA (Fig. 4A and B). A total of 79 differential metabolites were identified (FC > 1.20 or < 0.83, p-value <0.05) with 31 upregulated (e.g., glucose, fructose, mannitol and maltose) and 48 down-regulated (e.g., 1,5-anhydro-D-glucitol, 2,4-dihydroxybutanoic acid and palmitoleic acid) in STZ mice (Fig. 4C).

Fig. 4.

Serum metabolic characteristics in mice with early-onset diabetic retinopathy. (A) Two-dimensional score plot of the PCA model comparing the metabolic profiles of the control (Red hollow circle) with STZ-induced diabetic (Green hollow circle) group based on UHPLC-MS/MS and GC-MS techniques. (B) Score plot of the OPLS-DA model based on UHPLC-MS/MS and GC-MS. (C) Volcano plot illustrating differential expression of metabolites between the two groups. Metabolites that are down-regulated and up-regulated metabolites are depicted in blue and red, respectively. Non-significant changed metabolites are represented by gray dots (p-value <0.05, FC > 1.2 or < 0.83). (D) Metabolome pathway analysis revealed that the differential metabolites between the STZ-induced diabetic group and the control group were associated with the enrichment of 34 metabolic pathways. The top 6 metabolic pathways with the most significant p-value are highlighted and labeled (p-value <0.05).

KEGG analysis identified 34 altered pathways, including starch and sucrose metabolism, vitamin B6 metabolism, galactose metabolism, neomycin kanamycin and gentamicin biosynthesis and fructose and mannose metabolism (Fig. 4D). The metabolomic profiling of peripheral serum in early-stage STZ mice revealed systemic metabolic dysregulation, which may contribute to retinal pathway perturbations. These findings suggested that extraocular circulating metabolites—particularly non-canonical metabolic regulators distinct from glucose homeostasis—serve as key contributors to DR pathogenesis and disease advancement.

3.5. Correlation of serum and retinal metabolites in early diabetic retinopathy

To further investigate how peripheral blood metabolites influence the in-situ retina, we analyzed potential cross-metabolites and associations between systemic metabolic changes and the local retinal metabolism. Peripheral metabolites may interact with the retina via the bloodstream, either through transporter-mediated exchange of metabolites across the blood-retinal barrier or receptor-mediated signal transduction in the early DR.

Intersection analysis of the differential metabolites in serum and retina identified 6 overlapping metabolites (Fig. 5A). Among these, 1-Stearoylglycerol was decreased in serum but increased in retina. Glucose, sorbitol, pantothenic acid and mannitol were upregulated, while LPE18:1 was decreased in both serum and retina of STZ mice (Fig. 5B). These metabolites may enter the retina via specific transporters such as SGLT, GLUTs, SMVT, and SPNS1 respectively, suggesting their involvement in osmotic regulation or metabolic exchange.

Fig. 5.

Association of shared peripheral and retinal metabolic shifts in early diabetic retinopathy. (A) Venn plot illustrating the overlap of differential metabolites in serum and retina from diabetic retinopathy mice. (B) List of metabolites with synergistic changes in serum and retina, and their reported membrane permeability in the literature. (C) List of the top 6 metabolites with the most pronounced serum alterations and with documented receptor information in the literature.

Additionally, 73 metabolites were uniquely altered in serum (Fig. 5A). Of these, fructose, palmitoleic acid, N-acetylserotonin, palmitic acid, cholesterol and oleamide may affect retinal function via receptor-mediated pathways involving TAS1R, FFARs, MTs, CD36, LDLR, LXR, and 5-HT_2AR respectively (Fig. 5C). These interactions likely trigger downstream signaling cascades, modulating retinal function without direct metabolite accumulation.

3.6. Potential key players in the potential regulatory role of peripheral metabolites on retinal metabolism

To elucidate the genetic regulatory mechanisms underlying common differential metabolites in peripheral serum and the retina, and their association with early DR in diabetes. We integrated metabolomic and genomic data. Genome-wide association studies (GWAS) were employed to identify potential genetic variants, particularly single nucleotide polymorphisms (SNPs), which are key to understanding the genetic basis of complex trait [37]. A large GWAS dataset related to diabetes and its complications, with a focus on retinal lesions, was utilized. This integrated approach provides insight into molecular pathways linking systemic metabolic disturbances with early retinal inflammation in DR.

By analyzing the key enzyme genes of the common differential metabolites in peripheral serum and retinal tissue, we identified the genes containing SNPs, which associated with glucose, sorbitol, fructose, cholesterol, and palmitoleic acid, all related to increased diabetes risk. The diabetes-related enzyme genes with SNPs are marked in red within a green frame (Fig. 6). Among them, hexokinase 1 (HK1), glucokinase (GCK) and hexokinase domain containing 1 (HKDC1) mediate glucose phosphorylation and serve as critical regulators of glucose metabolism. HK1 and HKDC1 also participate in the phosphorylation and metabolism of fructose. Aldo-keto reductase family 1 member B1 (AKR1B1) a rate-limiting enzyme in the polyol pathway, mediates glucose-to-sorbitol conversion and contributes to metabolic disturbances in DR. In addition, 3-Hydroxy-3-Methylglutaryl-CoA Reductase (HMGCR) and carboxyl ester lipase (CEL) were involved in the cholesterol metabolism. Acyl-CoA synthetase long chain family member 1 (ACSL1) and palmitoyl-protein thioesterase 2 (PPT2) were linked to palmitic acid metabolism (Fig. 6). These results revealed an underlying metabolic network connecting peripheral serum and local retina. Metabolites and associated enzyme genes may influence the systemic environment of STZ-induced diabetic mice, potentially contributing to the development of DR.

Fig. 6.

GWAS indicates potential peripheral regulatory factors linking hyperglycemia, hyperfructosemia and disrupted lipid metabolism with retinopathy. Integration of peripheral metabolic pathways in early-onset diabetic retinopathy, highlighting the coincidence of elevated glucose and fructose (in red font) in peripheral serum and retina. The figure also illustrates the involvement of metabolizing enzymes with Single Nucleotide Polymorphisms (SNPs) reported in GWAS studies of diabetic patients (framed in green with red font).

4. Discussion

Diabetic retinopathy (DR) remains without targeted therapies beyond glycemic control. In this study, we integrated retinal transcriptomic and metabolomic date, serum metabolomics and genome-wide association study (GWAS) datasets to identify peripheral metabolites potentially linked to early retinal lesions in a streptozotocin (STZ)-induced DR mouse model. Our previous research demonstrated that 4-month-old STZ mice exhibited retinal inflammation, accompanied by microglial activation and visual function impairment [30]. To identify metabolites potentially responsible for initiating early retinal inflammation, we focused on earlier time points post-STZ induction. Although previous studies suggest that 2–4 weeks of hyperglycemia can induce systemic metabolic changes, they rarely result in robust retinal inflammation [36]. Consequently, we found that mice subjected to a 2-month high-glucose condition displayed significant retinal inflammation, evidenced by elevated IL-6 and IL-1β levels (Fig. 1C), along with concurrent metabolic-related genes alterations within the retina (Fig. 2).

Systemic metabolic dysregulation is a hallmark of diabetes and DR, with hyperglycemia being a primary factor driving retinopathy. Our metabolomics analysis revealed six differentially expressed metabolites - glucose, LPE18:1, sorbitol, pantothenic acid, 1-stearoylglycerol and mannitol - shared between the serum and retina of STZ mice (Fig. 5A and B). Dysregulated glucose metabolism is closely associated with chronic retinal inflammation, promoting elevated levels of pro-inflammatory cytokine [[38], [39], [40]]. Among the identified metabolites, sorbitol and mannitol are notable for their established roles in retinal inflammation of DR in addition to glucose. Sorbitol accumulatio has been observed in the retinas of both humans and rodent models under hyperglycemic conditions [41], in line with our findings. Inhibition of GLUT1, which mediates facilitated glucose transport, has been shown to reduce retinal sorbitol accumulation and mitigate oxidative stress and inflammation in diabetic mice [42,43]. Similarly, mannitol has been implicated in immune activation. IL-17A expression is significantly upregulated in myeloid cells from DR patients with T1DM following co-treatment with high glucose and D-mannitol [44]. Moreover, mannitol has been shown to increase CD11b expression in monocytes and neutrophils and to inhibit neutrophil apoptosis [45], although its direct contribution to retinal inflammation requires further investigation.

In addition to metabolites shared across the serum and retina, we also identified fructose and palmitoleic acid as uniquely altered in the serum of STZ-induced DR mice (Fig. 5C). These peripheral metabolites changes are also closely related to the increased inflammation in the retina. Elevated serum fructose levels [17] and decreased palmitoleic acid levels [46] have been reported in DR patients, consistent with our mouse model observations. In high-fructose-fed rats, alterations in cone photoreceptor responses were detected as early as day 8 [47]. Furthermore, these animals exhibited partial loss of rod sensitivity and reduced oscillatory potential amplitudes after six months of fructose feeding [48]. Palmitoleic acid, a lipokine with anti-inflammatory properties [49], has shown protective effects in the retina. It attenuated 4-hydroxynonenal-induced stress in retinal pigment epithelial cells and mitigates light-induced retinal degeneration in mice [50]. Additionally, lower circulating levels of palmitoleic acid have been associated with increased risk of metabolic disorders and diabetes [51,52]. These findings suggested that circulating metabolites, such as fructose and palmitoleic acid, may influence retinal inflammation and degeneration via systemic metabolic signaling. These metabolites may engage retinal cell receptors and activate downstream intracellular pathways that contribute to DR pathogenesis, representing potential targets for early intervention.

Given that genetic variants (SNPs) associated with metabolic pathways have been identified in patients with diabetes and DR, we further integrated metabolite-related genes with GWAS data to explore potential upstream regulatory mechanisms. This analysis identified key enzyme-coding genes associated with peripheral metabolites implicated in retinal inflammation (Fig. 6). These genes included HK1, HKDC1, AKR1B1 in the hyper-glycemic pathway; CEL and HMGCR in the cholesterol pathway; and ACSL1, PPT2 in the palmitic acid pathway. Among them, HKDC1, AKR1B1, HMGCR, and ACSL1 exhibit particularly strong links to DR pathogenesis.

HKDC1 affected glucose metabolism and polymorphisms such as rs10762264 and rs4746822 have been associated with gestational diabetes mellitus [53]. Hkdc1-deficient mice exhibited reduced scotopic electroretinogram responses and thinner outer nuclear layer, resembling phenotypes seen in retinitis pigmentosa [54]. The AKR1B1–106C > T polymorphism was significantly associate with DR [55], and another variant, rs759853, may confer protection against DR in T1DM patients [56]. Furthermore, genetic variants [57] and pharmacologic inhibitors [58] of HMGCR, a key enzyme in cholesterol biosynthesis, have also been linked to DR risk. ACSL1, which catalyzes the formation of acyl-CoAs from free fatty acids, is involved in diabetes-accelerated atherosclerosis, and SNPs within this gene (rs7681334, rs735949, and rs4862423) are associated with fasting glucose levels and diabetes status [59]. Collectively, our findings suggested that specific peripheral metabolites and their upstream genetic regulators are closely associated with early retinal inflammation in DR. These systemic metabolic disruptions may serve as mechanistic bridges between hyperglycemia and retinal pathology, offering promising avenues for early diagnosis and therapeutic intervention. Targeting these metabolite-gene networks could provide novel strategies to prevent or attenuate DR progression beyond current glucose-lowering treatments.

5. Limitations

Nevertheless, the functional validation of peripheral metabolites in inducing retinal inflammation, both in vitro and in vivo, as well as elucidating the mechanistic roles of key metabolic enzymes (such as AKR1B1, HMGCR), represents a future direction for research. Furthermore, upon functional validation of these potential regulatory factors, future research directions could be further expanded towards precise patient subtyping and the development of tailored, responsive modulation and intervention strategies. To achieve this, the integration of research approaches and methodologies-including the development of efficient delivery systems, the design of controllability and targeting specificity, the pursuit of stability and reusability, and the application of multimodal characterization techniques will pave new avenues for research on novel target modulation [60,61]. This integrated approach will not only offer innovative directions but also significantly enhance the potential for practical application and translational success.

6. Conclusions

In summary, this study employed a multi-omics approach to investigate the peripheral metabolic reprogramming associate with retinal metabolism shifts during the early inflammatory phase of DR. Our findings suggest that elevated peripheral levels of glucose, sorbitol, and mannitol may trigger retinal inflammation. Furthermore, key regulatory genes identified in the periphery, which are associated with the observed metabolites and their corresponding SNPs-including HK1, HKDC1, AKR1B1 in the hyperglycemic pathway, CEL and HMGCR in cholesterol pathway, and ACSL1, PPT2 in the palmitic acid pathway, represent promising targets for DR therapy beyond conventional glucose control. These insights enhance our understanding of peripheral metabolic regulation and its systemic implications in DR, thereby facilitating more precise early detection and therapeutic strategies for the disease.

CRediT authorship contribution statement

Xiaonan Zhang: Writing – review & editing, Writing – original draft, Software, Methodology, Data curation. Yan Liu: Methodology, Data curation. Mengxue Xia: Writing – review & editing, Software. Manwen Yang: Writing – review & editing. Yingjie Wu: Validation, Supervision, Resources, Project administration. Fang Zhang: Validation, Supervision, Resources, Project administration, Investigation, Funding acquisition, Conceptualization.

Funding

This research was supported by the Key R&D project of the National Ministry of Science and Technology (2023YFA1801100), and the National Natural Science Foundation of China (92357307, 32171177, 3247124).

Declaration of competing interest

The authors declare that they have no conflicts of interest.

Abbreviations

DR	Diabetic retinopathy
STZ	Streptozotocin
T1DM	Type 1 diabetes mellitus
VEGF	Linear dichroism vascular endothelial growth factor
HbA1c	Hemoglobin A1c
IL-1β	Interleukin-1β
IL-6	Interleukin-6
GW-DR	Glucose-well-controlled diabetic patients
SNPs	Single nucleotide polymorphisms
PCA	Principal Component Analysis
OPLS-DA	Orthogonal Partial least squares Discriminant Analysis
DEGs	Differentially expressed genes
GO	Gene Ontology
BP	Biological process
MF	Molecular function
CC	Cellular component
HK1	Hexokinase 1
HKDC1	Hexokinase domain containing 1
AKR1B1	Aldoketo reductase family 1 member B
GCK	Glucokinase
CEL	Carboxyl ester lipase
HMGCR	3-Hydroxy-3-Methylglutaryl-CoA Reductase
ACSL1	acyl-CoA synthetase long chain family member 1
PPT2	palmitoyl-protein thioesterase 2
AR	Aldose reductase
ChREBP	Carbohydrate-responsive element-binding protein

Data availability

Data will be made available on request.