Abstract Objectives Diabetic retinopathy (DR), an earnest complication of diabetes, is one of the most common causes of blindness worldwide. This study aimed to investigate the altered metabolites in the serum of non-DR (NDR) and DR including non-proliferative diabetic retinopathy (NPDR), and proliferative diabetic retinopathy (PDR) subjects. Methods In this study, the 1HNMR platform was applied to reveal the discriminating serum metabolites in three diabetic groups based on the status of their complications: T2D or NDR (n = 15), NPDR, (n = 15), and PDR (n = 15) groups. Multivariate analyses include principal component analysis (PCA) and Partial Least Structures-Discriminant Analysis (PLS-DA) analysis that were performed using R software. The main metabolic pathways were also revealed by KEGG pathway enrichment analysis. Results The results revealed the significantly different metabolites include 10 metabolites of the NPDR versus PDR group, 24 metabolites of the PDR versus NDR group, and 25 metabolites of the NPDR versus NDR group. The results showed that the significantly altered metabolites in DR compared with NDR serum samples mainly belonged to amino acids. The most important pathways between NPDR/PDR, and NDR/DR groups include ascorbate and aldarate metabolism, galactose metabolism, glutathione metabolism, and tryptophan metabolism, respectively. In addition, some metabolites were detected for the first time. Conclusions We created a metabolomics profile for NDR, PDR and NPDR groups. The impairment in the ascorbate/aldarate, galactose, and especially amino acids metabolism was identified as metabolic dysregulation associated with DR, which may provide new insights into potential pathogenesis pathways for DR. Graphical Abstract [31]graphic file with name 40200_2024_1462_Figa_HTML.jpg Keywords: Type 2 diabetes, Diabetic retinopathy, Proliferative diabetic retinopathy, Non-proliferative diabetic retinopathy, Metabolomics, 1HNMR Introduction Diabetes has become the third most common endocrine disorder in the world, and its prevalence is rising in developing and developed countries. Constantly high blood glucose damages the blood vessels of the eyes, heart, kidneys and nerves. In 2019, approximately 463 million diabetic patients were reported in the world [[32]1]. Diabetic retinopathy (DR) is a chronic retinal microvascular complication in patients with type 1 and 2 diabetes and the most common reason for blindness among people [[33]2]. Additionally, the DR also proposes a high risk of life-threatening systemic microvascular complications. Risk factors for DR include hyperglycemia, hypertension, dyslipidemia, continuity of diabetes, and genetic factors. DR consists of two stages including NPDR and PDR. NPDR is the first stage of DR in which micro-aneurysms and exudates form in the fundus. At this stage, patients may lose vision due to a micro-aneurysm laceration or macular edema [[34]3]. PDR is the end-stage of DR in which new blood vessels form with severe macular hypoxia. Newly formed blood vessels are very delicate and easily grow into the vitreous, resulting in significant vision loss [[35]4]. However, despite this clinical importance, studies indicate that DR has not been well-screened compared to other diabetes complications [[36]5]. Standard treatments that mainly target end-stage DR include laser photocoagulation, intravitreal injections of corticosteroids or anti-vascular endothelial growth factor (VEGF) drugs, and vitreoretinal surgery. The treatment methods mentioned above are quite expensive, all require the help of a vitreous specialist and cause several side effects. Irreversible damage to a person’s vision is also possible despite the initiation of treatment [[37]6]. Therefore, the identification of potential screening, treatment, and prognostic biomarkers is necessary for the management of DR. In the recent decade, “Omics-based” strategies including transcriptomics, proteomics, and metabolomics have been widely developed in the context of biosignature discovery. Among these, metabolomics, as an emerging field of “Omics”, comprehensively analyzes low molecular weight (< 1500 Da) metabolites in biological samples such as serum, plasma, urine, saliva, tissues, or cultured cells [[38]7–[39]10]. By identifying and quantifying metabolite profiles in a system, metabolomics provides a “snapshot” of metabolic changes associated with diseases [[40]11]. Due to the more dynamic nature of metabolomics rather than genomics and proteomics, this method can identify metabolic variations related to different physiological conditions in a shorter time. Recently, several metabolomics-based investigations have been reported for diabetic complications, especially DR, highlighting the importance of metabolomics-based studies. For instance, Zhu et al. reported plasma metabolomics profiling of PDR compared to NPDR [[41]12]. In another study, Tomita et al. investigated vitreous metabolomics profiling of PDR [[42]13]. The aim of this study was to find metabolites molecular signature associated with diabetic retinopathy. Therefore, we have conducted a metabolomics profiling of serum samples from T2D non-DR as control versus NPDR and PDR patients using an un-targeted 1HNMR approach. Materials and methods Study ethics approval This work was reviewed and approved by the Institutional Ethics Committee of the Zanjan University of Medical Sciences (IR.ZUMS.REC.1402.091). Written informed consent for the collection of clinical data and samples was obtained from all subjects. Study subjects and baseline characteristics The samples consisted of 45 T2D or NDR, NPDR, and PDR. All participants’ demographic and medical information including age, sex, height, weight, diabetes duration, drug use, blood pressure, smoking, alcohol consumption and body mass index (BMI). The routine biochemical and blood parameters including hemoglobin A1c (HbA1c), fasting plasma sugar (FBS), albumin, urea, total cholesterol, triglycerides (TG), High-density lipoprotein (HDL), Low-density lipoprotein (LDL), serum creatinine, and high-sensitivity C-reactive protein (hsCRP) were measured using standard automatic analyzers. Differences in baseline characteristics between groups were analyzed by one-way ANOVA test, and p-value < 0.05 was considered significant. All participants were diagnosed by a retinal specialist. The diagnosis of NPDR is based on the presence of bleeding points, micro-aneurysm, cotton wool points, or retinal microvascular disorders and the absence of evidence of active PDR or a history of treatment for PDR. A PDR diagnosis was based on the presence of neovascularization of the iris or retina or clinical imaging of vitreous hemorrhage or preretinal hemorrhage. Based on the retinal specialist examination and based on the Early Treatment Diabetic Retinopathy Study (ETDRS) trial, the patients were divided into three groups: group 1 (T2D or NDR, n = 15) as a control group, group 2 (T2D-NPDR, n = 15) -type 2 diabetes patients with non-proliferative diabetic retinopathy, and group 3 (T2D-PDR, n = 15)-type 2 diabetes patients with proliferative diabetic retinopathy. In addition, the patients of each group were matched based on age and gender. Inclusion and exclusion criteria The inclusion criteria were as follows: informed consent, type 2 diabetes, age over 30 years, diabetes over 5 years, diabetes with a body mass index below 40 kg/m2, does not use antioxidants, vitamin preparations and immunosuppressive drugs past three months, infectious diseases, liver diseases, absence of kidney and thyroid diseases, absence of other diabetic complications and non-smoking and alcohol consumption. Exclusion criteria included the following: the presence of other eye diseases in ophthalmology examination, history of eye surgery for reasons other than diabetic retinopathy, BMI > 40, smoking and alcohol consumption, infectious diseases, cancer, allergies and immune-based diseases, high blood pressure (systolic ≤ 180 or diastolic ≤ 110), severe heart and liver disease, and pregnancy. Sample collection and processing After confirming 8 h of overnight fasting, 6 mL venous blood was collected using vacutainer tubes. The serum was isolated by centrifugation at 3000 rpm for 10 min (4 °C), and then supernatant serum was transferred into a 1.5 ml sterile microtube and stored at − 80 °C. Just before doing the 1HNMR experiment, serum samples were thawed at 4 °C. Afterward, a 500µL serum sample was supplemented with a 200µL deuterium oxide (D2O). NMR spectroscopy NMR experiments were performed using Bruker-Avance 400 MHz, equipped with a 5 mm probe at 298 K (Zanjan, Iran). D2O added to serum samples leads to a rapid exchange of deuterium atoms for sensitive protons in -OH and -NH- groups. Since deuterium atoms do not show up in proton NMR spectra, the resonance for the -OH or -NH- groups disappears, confirming its identity. Spectroscopy of the samples was performed based on the One-dimensional CPMG (Carr-Purcell-Meiboom-Gill) protocol, which irradiates the residual water peak, with a relaxation delay of 2 sec and a mixing time of 100 ms. A total of 150 scans was collected over a spectral width of 8389.26 Hz with a 90° pulse width and a repetition time of 3 sec. 0.3 Hz line broadening was applied to the spectra before Fourier transformation and phase- and baseline-correcting. Spectra were referenced to the tetramethylsilane (TMS) standard at 0 ppm. TMS became the established internal reference compound for 1 H NMR because the addition of TMS does not usually interfere with other resonances. Additionally, TMS is highly volatile, so it can be easily removed if a sample needs to be recovered. Statistical analysis and metabolites identification NMR spectra were imported into ProMetab3 and normalized by the TSA method. The 4.5–5.5 ppm spectral range was also excluded to eliminate the variable effectiveness of water attenuation. Multivariate analysis by R V3.5.1 software including principal component analysis (PCA) and projections to latent structures-discriminant analysis (PLS-DA) was then performed to detect significantly altered metabolites between under-studied groups. The Human Metabolome Database, HMDB ([43]https://hmdb.ca/), was used to identify significant metabolites. Variables with variable influence on projection (VIP) values > 1 and p-value < 0.05 were used to identify significant discriminating metabolites in every comparison. Metabolic pathway enrichment analysis In our study, the significant metabolites between the compared groups were assessed for the pathway analysis and visualization by the Metabolite Set Enrichment Analysis (MSEA) method based on the biochemical pathways in the KEGG database using the Metaboanalyst v6.0 ([44]http://www.metaboanalyst.ca/). Results Demographic and clinical information Demographic and clinical information of all subjects in this study are presented in Table [45]1. No statistical difference was found between the groups in terms of sex (p = 0.914), age (p = 0.096), gender (0.914), systolic blood pressure (SBP) (p = 0.281), diastolic blood pressure (DBP) (p = 0.264), BMI (P = 0.205), TG (p = 0.227), TC (p = 0.525), LDL (p = 0.057), HDL (p = 0.318); FBS (p = 0.606), A1C (p = 0.344), Cr (p = 0.500), and Alb/Cr (p = 0.991). Table 1. Baseline demographics of subjects in this study T2D (NDR) NPDR PDR P-value n 15 15 15 - Gender (male/Female) 7/8 7/8 6/9 0.914 Age (years) 59.6 ± 9.61 59.6 ± 8.54 60.4 ± 7.65 0.096 Diabetes duration (years) 8 (6–10) 10 (5–15) 12 (8–16) 0.370 BMI (kg/m2) 26.5 ± 1.91 25.5 ± 4.11 24.4 ± 2.78 0.205 SBP (mmHg) 126 ± 12.2 131 ± 13.5 134 ± 16.1 0.281 DBP (mmHg) 80 (70–80) 80 (75–85) 80 (80–85) 0.264 FBS 132 ± 11.7 131 ± 28.6 141 ± 41.9 0.606 HbA1c 7.29 ± 0.606 7.54 ± 0.764 7.16 ± 0.76 0.344 Cr (umol/L) 0.97 ± 0.179 1.05 ± 0.165 0.996 ± 0.194 0.500 TG (mmol/L) 165 ± 73.3 162 ± 69.5 128 ± 47.3 0.227 TC (mmol/L) 150 ± 32.6 152 ± 24.2 162 ± 33.3 0.525 LDL (mmol/L) 66 (58–95) 84 (66–100) 98 (82–107) 0.057 HDL (mmol/L) 41.1 ± 9.97 40.1 ± 8.08 45.7 ± 13.2 0.318 Alb/Cr 12.2 ± 4.78 12.2 ± 5.16 12.4 ± 5.41 0.991 Tablet/Insulin/tablet + Insulin 15/0/0 12/1/2 6/5/4 - [46]Open in a new tab Multivariate analysis Multivariate analysis was used on the data matrix obtained from ProMetab. PCA analysis was performed to determine trends in the data and groups of observations and to find outliers. The result of the score plot of PCA analysis showed that the discrimination of the all study groups all studied groups was acceptable (Fig. [47]1). The value of the first three principal components for discriminating NPDR and PDR was 83.7, 6.3, and 3.1, respectively. We also compared T2D without DR (NDR) with NPDR and PDR groups. The values ​​of the first three principal components for differentiating T2D from NPDR were 76.4, 6.8, and 5.6%, respectively, whereas, these values were 79.7, 6.1, and 3.7 for the comparison of T2D and PDR subjects. Based on these results, PCA analysis was able to create a 90% separation between the studied groups in the first two dimensions. All samples were within the 95% confidence interval in the PCA score plots. PLS-DA analysis, as a supervised statistical analysis, was performed on the data matrix to find differential metabolites between groups. The score plot shows that the PDR and NPDR groups are well differentiated from each other. The cumulative value of Q2 is equal to 0.80 and the RMSE value is equal to 0.0101. Validation of the model was done by the 7-Fold Cross Validation method. The predictive power of the model was determined based on the area under the graph (AUC). The area under the ROC plot for this comparison was 1, which shows the high accuracy of the PLS-DA model (Fig. [48]2). The cumulative Q2 values ​​for comparing T2D with NPDR and PDR were 0.81 and 0.70, respectively, while the RMSE values ​​for T2D (NDR) versus NPDR and T2D (NDR) versus PDR were 0.0429 and 0.234, respectively (Fig. [49]2). Fig. 1. [50]Fig. 1 [51]Open in a new tab Scatter score plots of the PCA analysis for the differentiation between a) NPDR and PDR, b) T2D (NDR) and NPDR, c) T2D (NDR) and PDR Fig. 2. [52]Fig. 2 [53]Open in a new tab PLS-DA models and validation of model based on the area under the ROC curve for the separation of groups. a) NPDR versus PDR, b) T2D versus NPDR, c) T2D versus PDR Identification of differentiated metabolites VIP values obtained from the PLS-DA models were utilized to find the most significantly changed metabolites. VIP ranks the variables based on their contribution to the model. Metabolites with VIP > 1 and p < 0.05 were considered significant. The differentially expressed metabolites are listed in Tables [54]2 and [55]3. As shown, the most significant metabolites between the NPDR and PDR groups included up- and down-regulated metabolites. Table 2. Altered metabolite markers between NPDR and PDR groups extracted from the PLS-DA model. Those metabolites which had VIP values > 1, and p-value < 0.05 have been considered Metabolites HMDB ID Variable ID (ppm) VIP p-value Fold change (PDR/NPDR) Glycine HMDB0000123 3.595 3.302 0.0323 1.89↓ suberic acid HMDB0000893 1.295 2.762 0.0086 1.11↓ D- (+)-Maltose HMDB0000163 3.265 2.509 0.0168 1.19↑ L-homocitrulline HMDB0000679 1.395 2.491 0.044 1.37↑ N-Acetyl-L-Glutamine HMDB0006029 4.145 2.278 0.039 1.73↓ 3-Oxoglutaric acid HMDB0013701 3.325 2.13 0.0481 1.9↑ 3-Hydroxy-3-methylglutaric acid HMDB0000355 1.305 1.943 0.0421 1.11↓ 2-hydroxybutyric acid HMDB0000008 3.975 1.535 0.0248 1.63↑ myo-Inositol HMDB0000211 3.615 1.303 0.0337 1.49↑ Acetylcarnitine HMDB0000201 3.715 1.238 0.0478 1.52↑ [56]Open in a new tab VIP: Variable Importance to Projection; T2D: Type 2 diabetes; NPDR: Non-proliferative diabetic retinopathy; PDR: Proliferative diabetic retinopathy; ↑:up-regulation; ↓: downregulation Table 3. Altered metabolite markers between T2D & NPDR, and T2D & PDR groups extracted from PLS-DA model. Those metabolites which had VIP values > 1, and p-value < 0.05 have been considered Metabolites HMDB IDs T2D (NDR) vs. NPDR T2D (NDR) vs. PDR Variable ID VIP p-value Fold change (NPDR/T2D) Variable ID VIP p-value Fold change (PDR/T2D) L-homocitrulline HMDB0000679 1.525, 0.945 2.777 7E-07 3.42↑ 1.525, 0.945 2.973 4.09E-05 3.17↑ cholic acid HMDB0000619 1.485 2.277 7.9E-06 5.51↓ 1.485 2.859 5.11E-05 3.08↓ L-Arginine HMDB0000517 3.235 2.831 0.0 3.27↓ 3.235 2.583 0.0 3.28↓ serotonin HMDB0000259 3.305 2.453 2.4E-06 8.91↑ 3.305 2.313 1.7E-06 9.63↑ Alpha-Hydroxy isobutyric acid HMDB0000729 - - - - 1.375 2.163 2.19E-05 1.94↑ Taurine HMDB0000251 3.415, 3.225 2.414 1.01E-05 3.97↓ 3.415 2.113 1.82E-05 3.76↓ L-Threonine HMDB0000167 3.565 2.199 5.82E-05 3.19↓ 3.565 2.112 0.0003 2.95↓ L-Glutamic acid HMDB0000148 2.075, 2.085 2.496 1.3E-06 1.56↑ 3.755, 2.075, 2.085 2.111 0.000114 1.61↑ L-Tryptophan HMDB0000929 3.295 2.258 7.06E-05 4.25↑ 3.295 2.097 7.54E-05 4.36↑ D-Galacturonic acid HMDB0002545 5.275 2.207 0.000162 4.88↑ 5.275 2.042 0.000425 6.21↑ Sucrose HMDB0000258 3.665 1.964 0.000217 4.11↓ - - - - L-Methionine HMDB0000696 2.115 2.326 6.19E-05 1.88↑ 2.115 2.034 0.00084 2.04↑ Methylmalonic acid HMDB0000202 - - - - 1.215 2.015 0.027227 1.7↓ 1-Methyl-L-histidine HMDB0000001 3.955 2.193 6.17E-05 3.82↑ 3.955 2.004 0.000142 4.37↑ Indole-3-acetic acid HMDB0000197 3.655 2.069 7.61E-05 5.99↓ - - - - Pantothenate HMDB0000210 3.425 2.318 1.67E-05 2.84↓ 3.425 1.998 0.00018 2.97↓ Pantothenate HMDB0000201 - - - - 3.715 1.993 0.000384 3.68↓ Acetylcarnitine HMDB0000201 3.715, 3.845 2.124 0.000195 5.74↓ 3.715 1.993 0.000384 3.68↓ suberic acid HMDB0000893 1.295 1.918 0.001293 1.72↓ 1.295 1.977 0.000213 1.91↓ 5-Hydroxy-L-tryptophan HMDB0000472 - - - - 3.405, 3.225 1.965 4.41E-05 6.89↓ Glycerol HMDB0000131 3.585 2.124 0.000193 7.3↑ 3.585 1.964 0.000457 7.36↑ 2-hydroxyoctanoic acid HMDB0000711 0.845 2.301 2.15E-05 2.1↓ 0.845, 0.855, 1.285 1.962 0.000832 1.84↓ L-Leucine HMDB0000687 3.725 2.269 6.62E-05 4.26↓ 3.725 1.953 0.00012 3.5↓ L-Isoleucine HMDB0000172 0.935 2.185 6.44E-05 1.69↑ - - - - trans 4 Hydroxy-L-proline HMDB0000725 3.475 2.211 0.000058 2.26↓ 3.475 1.947 0.00056 2.31↓ glycocholic acid HMDB0000138 - - - - 1.235 1.932 0.000789 1.66↓ Choline HMDB0000097 - - - - 3.515 1.861 0.007775 3.95↑ D-(+)-Glucose HMDB0000122 - - - - 3.835 1.842 0.002348 2.6↓ Glycine HMDB0000123 3.535 2.313 1.12E-05 6.98↓ - - - - L(-)-Carnitine HMDB0000062 3.215 2.02 0.000422 18.09↓ - - - - N-alpha-Acetyl-L-lysine HMDB0000446 2.015, 2.025 2.01 0.000138 0.000138 - - - - [57]Open in a new tab VIP: Variable Importance to Projection; T2D: Type 2 diabetes; NPDR: Non-proliferative diabetic retinopathy; PDR: Proliferative diabetic retinopathy; ↑:up-regulation; ↓: downregulation Based on this, there was a significant increase in the metabolites of maltose, homocitrulline, 3-oxoglutaric acid, 2-hydroxybutyric acid, myoinositol, and acetylcarnitine, and a significant decrease in the metabolites of glycine, suberic acid, N-acetylglutamine, and 3-hydroxy-3- Methyl-glutaric acid was observed (Table [58]2). The altered metabolites comparing the T2D (NDR) group with NPDR and PDR groups are listed in Table [59]3. As shown in Table [60]3, the alpha-Hydroxyisobutyric acid, Methymaloric acid, 5-Hydroxy-L-tryptophan, glycocholic acid, choline, and D(+) -Glucose were exclusively identified in the comparison of T2D (NDR) and PDR groups, while Indole-3-acetic acid, L-Isoleucine, Glycine, L(-) -Carnitine, and N-alphaAcetyl-L-lysine were found only in comparison of T2D (NDR) with NPDR groups. Metabolic pathway enrichment analysis To clarify the pathways involved in diabetic retinopathy, we used KEGG pathway enrichment analysis, which combines pathway enrichment analysis and pathway topology results to find the most important metabolic pathways. The main pathways between NPDR and PDR groups include ascorbate and aldarate metabolism, galactose metabolism, glutathione metabolism, lipoic acid metabolism, inositol phosphate metabolism, porphyrin metabolism, glyoxylate and dicarboxylate metabolism, and glycine, serine and threonine metabolism (Fig. [61]3. a). Fig. 3. [62]Fig. 3 [63]Open in a new tab Metabolite sets enrichment analysis (MSEA) results for Diabetic retinopathy: a) PDR versus NPDR, b) T2D (NDR) compared to NPDR, and c) T2D (NDR) compared to PDR group The most important pathways that were significantly enriched in T2D/NPDR comparison included the amino acid tryptophan metabolism, bile acid biosynthesis, galactose metabolism and glycine, serine and threonine metabolism (Fig. [64]3. b). Enrichment analysis of metabolite sets for altered metabolites between T2D/PDR demonstrated that amino acid tryptophan metabolism, galactose metabolism, glycine, serine and threonine metabolism and bile acid biosynthesis pathway are among the most important pathways that are significantly different between T2D (NDR) and PDR groups. The results are shown in Fig. [65]3c. Discussion Diabetic retinopathy (DR) is the prevalent microvascular complication of diabetes that leads to blindness in many patients with diabetes. To improve the screening rate, it is necessary to present potential biomarkers that can diagnose and predict the DR development and progression. However, this study was designed to identify significant changes in metabolites that could be demonstrated as a potential biomarker of DR compared to T2D patients without retinopathy (NDR) as a control group. We experimentally identified metabolites with significant differential expression in serum samples of DR patients compared to controls by 1HNMR technology. By comparing NPDR and T2D groups, we established 10 significantly changed metabolites between serum samples from NPDR compared to T2D control group. Among these altered metabolites, four (including Glycine, suberic acid, N-Acetyl-L-Glutamine, 3-Hydroxy-3-methylglutaric acid) and six (including Maltose, L-homocitrulline, 3-Oxoglutaric acid, 2-hydroxybutyric acid, myo-Inositol, Acetylcarnitine) metabolites were downregulated and upregulated in PDR/NPDR, respectively. Glycine, a non-essential amino acid, was decreased in the PDR compared to the NPDR group in this study. It is involved in a wide range of biochemical pathways [[66]14]. Some of in-vivo and clinical studies reported antidiabetic outcomes of dietary glycine [[67]15]. In addition, there is some documentation that vessels and neurons of the retina tolerate inflammation involving both innate and adaptive immunity in the pathogenesis of DR [[68]16]. It is established that glycine administration confines immune responses, protecting organs from damage associated with proinflammatory cytokines in patients with T2D [[69]17]. In a study by Alvarado-Vásquez et al., on rats with induced diabetes, the vascular complications of diabetes were improved with glycine therapy [[70]18]. It is believed that glycine plays a role in the adjustment of hyperglycemia, hypercholesterolemia, and A1C levels [[71]19]. Gholami et al. found that glycine supplementation effectually weakens retinal neuronal damage in rats, therefore, it might be a potential nominate to protect retinal ultrastructure against diabetes [[72]20]. Additionally, Song et al., in 2022, showed the correlations between some amino acids such as glycine related to the pathogenesis of DR by metabolomics approach [[73]21]. A recent study reported that glycine is abnormally regulated in human retinal endothelial cells (HREC) and PDR [[74]22]. Our data finding indicated that other amino acids, including N-Acetyl-L-Glutamine, were decreased in PDR compared to NPDR. N-acetyl-L-Glutamine (Gln) is an acetylated form of the conditionally essential amino acid L-Gln and has been shown to have neuroprotective effects against metabolic disorders such as diabetes. In addition, it is a precursor to glutamate, an important neurotransmitter in the brain. It has been reported that serum Gln is linked with a lower risk of T2D, while Glu is related to a higher risk of T2D [[75]23]. Additionally, dysregulated Gln metabolism (lower concentration in serum) has been reported in patients with T2D [[76]24], and T1D [[77]25]. Diabetes is characterized by inflammation, and Gln exhibits anti-inflammatory and antioxidant effects [[78]26]. Additionally, according to previous metabolomic studies in DR, glutamine was a repeated metabolic biomarker related to DR in various biological samples including plasma, vitreous, and aqueous humor [[79]27]. These findings indicate amino acids (AAs) may be used as likely biomarkers in evaluating PDR development, which may apply to the field for the setting of targeted strategies for PDR patient therapies. This study also identified increased serum levels of acetylcarnitine in patients with PDR versus NPDR. Carnitine plays a critical role in the transporting of long-chain fatty acids into mitochondria organelle through acylcarnitine intermediates before β-oxidation. A review by Bene et al. reported reduced carnitine levels in T2D and diabetic complications. On the other hand, other studies described no association between carnitine and diabetes complications [[80]28, [81]29]. Our finding is consistent with the study by Paris et al., study, which found increased acylcarnitine levels in vitreous samples from patients with PDR [[82]30]. Dissimilarity in study design, clinical phenotype of subjects, and sample sizes involved in the inconsistencies of the study results. However, investigations in detail are needed to further discover the association between carnitine and DR. As shown in Fig. [83]3a, the altered metabolites between PDR and NPDR were mapped to several important pathways. The ascorbate and aldarate metabolism, galactose metabolism, glutathione metabolism, lipoic acid metabolism, inositol phosphate metabolism, porphyrin metabolism, glyoxylate and dicarboxylate metabolism, and glycine-serine-threonine metabolism were eight pathways significantly perturbed between NPDR and PDR groups (p < 0.05). Among these, the ascorbate and aldarate metabolism was a markedly disturbed pathway (p-value: 0.011). Ascorbic acid and threonic acid are known as core signaling hubs in the ascorbate-aldarate pathway. Ascorbic acid is also a cofactor for several hydroxylases such as proline hydroxylase and dopamine hydroxylase [[84]31], and plays main roles in neuropeptide synthesis. Furthermore, the ascorbic acid deficiency may play a main role in the primary neurodegeneration described in DR. It is shown that the Ascorbic acid inhibits angiogenesis, as a hub event in DR [[85]32]. Ascorbic acid metabolism perturbation was reported in diabetic patients with retinopathy [[86]33]. In addition, according to previous metabolomics studies based on GCTOF-MS, abnormality of ascorbate-aldarate metabolism and galactose metabolism have been introduced as pathways with high weight in the development of DR, which is in accordance with our results [[87]34]. Glutathione metabolism was another metabolic pathway enriched between PDR patients and patients with T2DM, which was also reported as a significantly altered pathway in DR in other metabolomics-based studies [[88]35]. In this regard, Wang et al. (2022) reported significantly enriched pathways between patients with PDR and nondiabetic control including glutathione metabolism [[89]36]. Lipoic acid metabolism was one of the most important altered pathways in the PDR compared of the NPDR group. Oxidative stress, as a high production of reactive oxygen species (ROS), has a main role in the development of several diseases, including diabetes and its complications. In addition, hyperglycemia-induced ROS production is a core factor in the pathogenesis of endothelial damage in diabetes. Meanwhile, antioxidants can avert unwanted oxidation via their reaction with ROS or oxidation intermediates [[90]37]. Lipoic acid is one of the endogenous thiol antioxidants that has been broadly investigated on inhibition of DR development. Lipoic acid treatment prevents early elevation of vascular permeability and angiogenic factors in the diabetic retina, an inflammatory hallmark of DR that promotes vascular endothelial cell damage and capillary loss [[91]38, [92]39]. Lipoic acid administration has also been found to affect DR and can prevent neurophysiological dysfunction and protect retinal vascular and neuronal health during experimental DR [[93]40–[94]42]. Notwithstanding these effects of lipoic acid on DR, the results of a study by Haritoglou et al. showed that daily consumption of lipoic acid is ineffective in preventing diabetic macular edema (DME) [[95]43]. From the comparison of T2DM as a control group and DR group (including NPDR, and PDR), we found 25 and 24 significant dysregulated metabolites between NPDR, and PDR versus the T2DM control, respectively. The 18 out of these altered metabolites were common in both comparisons. As shown in Table [96]3, several metabolites have been identified exclusively in group NPDR (n = 6), and PDR (n = 7). Several amino acids were identified in the comparison of T2D with NPDR, and PDR groups (Table [97]3). Many studies have reported the preservative effect of amino acid metabolism on microvascular endothelial cells. In addition, amino acids (AAs) and their metabolites play an important role in retinal health and function and also act as neurotransmitters in the retina. Many studies have reported changed amino acid levels in serum and retina in DR patients and rodent models of DR [[98]44, [99]45]. Abnormalities in amino acid metabolism lead to increased oxidative stress and inflammation and induce apoptosis, which induces neuroretinal injury in diabetes [[100]45], as described below. As shown in Table [101]3, among common metabolites with differences in their levels in NPDR and PDR groups, compared to NDR patients, arginine and threonine showed lower concentrations in DR group compared to T2D. This result is consistent with previous metabolomic studies that reported a decrease in these metabolites [[102]46]. The effect of threonine on DR may be due to serine/threonine kinase-activated non-protein-coding RNA (BRAF-activated non-coding RNA [BANCR]), which has been implicated in the development of DR through a role in regulating apoptosis [[103]47]. In addition, in a systematic review of metabolomics profiles in DR, Hou et al., reported threonine as a repeatedly identified metabolite, as well as, a potential biomarker of DR [[104]27]. Arginine, a central step in the ornithine cycle, is involved in arginine synthesis and the arginine and proline metabolism pathways. The enzyme arginase metabolizes arginine, which leads to the formation of proline, polyamines, and glutamate. The high activity of arginase reduces the concentration of arginine, which leads to the separation of nitric oxide synthase and the induction of polyamine oxidation, as well as the formation of glutamate. The generated superoxide and nitric oxide fastly react to produce the toxic oxidant pernitrite, leading to nerve vascular damage in the retinopathy process [[105]27, [106]44]. We identified branched-chain amino acids (BCAAs) including leucine, and isoleucine as altered metabolites in our study. These AAs are essential amino acids that are needed for the growth and development process and also act as nutrient marks, neurotransmitter synthesis and glutamate/glutamine cycling in the brain and retina [[107]48]. In addition, BCAAs play critical roles in the protein synthesis process, energy generation, inflammation, oxidative stress, and glucose metabolism pathway in the metabolic organs of diabetes [[108]49]. Currently, several studies have linked BCAA metabolism and retinopathy and frequently reported elevated plasma BCAA levels associated with obesity, insulin resistance (IR), and the development of diabetes in human and in-vivo studies [[109]50]. Besides, metabolomics-based studies have reported significant alterations in BCAA levels in the retina of diabetic models and patients with DR [[110]50, [111]51]. In addition, the drug gabapentin, which can suppress BCAT, an important enzyme in BCAA metabolism, successfully reduces the activity of caspase 3, which is a key factor of oxidative stress, and reduces the concentration of ROS in the diabetic retina. Thus, diabetes-induced dysregulated BCAA metabolism appears to be a possibility for neuroretinal damage [[112]51]. Generally, all forms of diabetes have tremendous effects on whole-body metabolism, and tight control of blood glucose is needed to reduce the risk of diabetes-related complications. Hyperglycemia has been broadly considered the main contributor to the progression of the retinal damage. Emerging evidence proposes that besides diabetes-induced alterations in glucose, elevated levels of lipid and amino acids might be an important contributor to the progression of early neurovascular retinal injury. According to literatures, plasma amino acids are important for glucose management in T2D, so that, the disturbance in plasma free amino acid has been found previously in DM, and is associated with IR. In addition, the focus of some investigations are specifically the alterations in amino acid metabolism due to impaired glucose metabolism by glycolysis [[113]52]. In this regard, a prospective cohort study concluded that a higher dietary intake of BCAA was associated with increased risk of T2D [[114]53]. In addition, using an un-targeted metabolomics approach, BCAAs were significantly increased in both individuals with impaired fasting glucose or T2D compared to healthy controls [[115]54]. Previous studies have demonstrated that affected lipid metabolism is correlated with diabetes complications. Since L-carnitine has a crucial role in lipid metabolism through its function in the β-oxidation of long-chain fatty acids and it has antioxidant arginase as well, it is likely to be a potential metabolite marker for diabetic complications such as retinopathy [[116]55]. On the other hand, L-carnitine has been reported to have anti-inflammatory and antioxidant properties and ameliorates insulin sensitivity. Our data showed that acetyl-carnitine and L-carnitine were decreased in DR compared with T2DM group (Table [117]3). There is a very limited number of investigations that investigate the role of L-carnitine levels in diabetes and the appearance of its later complications. Poorabbas et al., (2007) found that women with T2D along with complications had 25% lower free L-carnitine levels in serum samples than in patients without complications of diabetes [[118]28]. While, it was showed that free carnitines levels were lower in the patient involved with retinopathy compared to T2D patients without complications [[119]28]. The low level of carnitine in diabetic patients with complications such as retinopathy can advance research towards the use of carnitine as a therapeutic supplement [[120]56]. Despite the results showing the satisfactory effects of carnitine on the handling diabetic complications in animal models, studies are still in the beginning [[121]57]. In addition, the metabolites whose expression changed only in one group are more important and can be used as an indicator for early detection of the disease at stages of NPDR and/or PDR. One of the metabolites that downregulated only in PDR compared to T2D was 5-hydroxy L- tryptophan (5-HTP). (5-HTP), also called oxitriptan, is a natural happening amino acid and chemical precursor as well as a metabolic intermediate in the serotonin biosynthesis. Indeed, serotonin level can be considered a potential biomarker for early detection of diabetes mellitus complications. Besides that, serotonin was identified in beta cells, and researchers have been attempting to discover its mechanism of action ever since. The altered metabolites between DR groups and T2D (NDR) subjects were mapped to some significant pathways. The main pathways in both NPDR, and PDR compared to T2D (NDR) observed to be altered were tryptophan metabolism, galactose metabolism, and glycin, serine and threonine metabolism. In previous studies, it has been found that these pathways are also altered in retinopathy. However, this result suggests that the development and progression of DR are linked with the disturbance of tryptophan metabolism. Conclusion In conclusion, our findings showed a significant disturbance of metabolic pathways in diabetic patients with retinopathy. It was described that the metabolite profile of patients with DR was associated with reprogramming of amino acid metabolism, mainly tryptophan and glycine, serine, and threonine metabolism. We conclude that studies based on metabolomics for DR patients may be hopeful to identify novel and potential diagnostic and prognostic markers, and also for monitoring treatment. There are some limitations in our investigation: (a) small sample size; and larger sample sizes are needed to confirm the findings and distinguish true metabolic differences between various grades and stages of patients with DR; (b) the lack of specific metabolites associated with DR for analysis with blood samples. As serum sample possesses the bulk of metabolites from whole body parts, it seems hard to trace a special serum metabolic pattern for a disease. From finding biomarkers to understanding the mechanisms underlying DR, metabolomics is showing its power. This has also become possible when metabolomics is more broadly combined with other disciplines such as transcriptomics and proteomics. The application of metabolomics in DR could also be extended to precise therapeutic evaluation and monitoring. Acknowledgements