Selection of IVsIn total, 486 blood metabolites were analysed in this investigation and SNP details for each metabolite are provided in Supplementary file 1: Table S2. The SNPs generated by 486 blood metabolites ranged from 3 to 493, summing up to a total of 10,936 SNPs. The minimum F statistic value of blood metabolite genetic IVs was 17.64. This suggested that there was no instrument weak bias in this study, as all of these values were above the threshold of 10.Effect of genetically determined metabolites on DRThe IVW method preliminarily identified 41 metabolites significantly related to DR (detailed MR analysis results are provided in Supplementary file 1: Table S3). Of these, 12 metabolites had unknown chemical composition, and the remaining 29 metabolites could be divided into 7 categories according to their chemical composition (17 lipids, 5 amino acids, 3 xenobiotics, 1 peptide, 1 nucleotide, 1 energy, and 1 carbohydrate) (Fig. 2).Fig. 2 IVW, MR-Egger and WM methods for the causal relationship between blood metabolites and DR risk. DR diabetic retinopathy; IVW inverse-variance weighted; WM weighted median; OR odds ratio.The results were verified according to the MR-Egger and WM methods, and the metabolites with different directions of conclusions for the three models were deleted. A total of 10 metabolites were deleted (citrulline, acetylphosphate, 1-palmitoylglycerol (1-monopalmitin), X-11485, X-11497, X-12038,1-stearoylglycerophosphocholine, 1-stearoylglycerophosphoethanolamine, 4-ethylphenylsulfate, and 2-hydroxyglutarate). Additionally, the Cochran’s Q test identified the presence of heterogeneity. MR-Egger intercept analysis and MR-PRESSO were used to analyse the horizontal pleiotropy of the results. The leave-one-out method was used to detect the presence of outliers. Metabolites with heterogeneity, horizontal pleiotropy or outliers were deleted. A total of 5 metabolites were eliminated (kynurenine, X-05907, acetylcarnitine, 4-androsten-3beta, 17beta-diol disulfate 1,4-androsten-3beta, and 17beta-diol disulfate 2). The onset of DR may be linked to the remaining 26 metabolites. Among them, fifteen metabolites exhibit a positive causal relationship with the onset of DR, including 3 amino acids (isoleucine, pipecolate, and hydroxyisovaleroyl carnitine), 6 lipids (palmitoleate (16:1n7), 1-docosahexaenoylglycerophosphocholine*, 1-arachidonoylglycerophosphoethanolamine*, 1-oleoylglyceroethanolamine, hexadecanedioate, and n-butyl oleate), and 6 metabolites whose chemical composition is unknown (X-09706, X-11876, X-12063, X-12450, X-12844, and X-14189). Eleven metabolites were negatively correlated with DR. In addition to 2 metabolites of unknown chemical composition (X-02973 and X-12833), 1 carbohydrate (mannitol), 4 lipids (palmitate (16:0), heptanoate (7:0), decanoylcarnitine, cis-4-decenoyl, and antarnitine), 1 nucleotide (inosine), 1 peptide (pro-hydroxy-pro), and 2 xenobiotics (hippurate and hydroxyhippurate) were included. Detailed IVW results of the metabolites obtained are shown in Fig. 3.Fig. 3Forest plot of the IVW analyses of the associations between metabolite and risk of DR. DR diabetic retinopathy, CI confidence interval, IVW inverse-variance weighted, OR odds ratio.The results of the MR Steiger test did not show reverse causality between metabolites and DR. Please refer to Table 1 for specific findings.Table 1 Sensitivity analysis and direction detection of causal relationships between blood metabolites and DR.Confounding analysisWe conducted various sensitivity analyses to ensure the soundness of the analysis results. However, to further ensure that IVs are not related to the remaining common risk factors for DR (smoking, alcohol consumption, hypertension, obesity, diabetes, heart rate and other retinal diseases), we examined each of the 26 metabolite-associated SNPs using Phenoscanner. A total of 34 SNPs associated with confounding factors were screened (details are available in Supplementary file 1: Table S4). After deleting the corresponding SNPs, MR analysis and sensitivity analysis were performed again. The IVW results suggested that 8 metabolites lost significant association with DR (palmitate (16:0), heptanoate (7:0), X-02973, mannitol, X-12063, palmitoleate (16:1n7), decanoylcarnitine, and 1-oleoylglycerophosphoethanolamine). The estimated values of the remaining 18 metabolites were still significant after discarding the promiscuous SNPs, and the direction of conclusion was consistent with the primary analysis. Specifically, serum levels of isoleucine (OR 4.79, 95% CI: 1.55–14.83, P = 0.006), pipecolate (OR 1.7, 95% CI: 1.06–2.71, P = 0.027), hydroxyisovaleroyl carnitine (OR 2.08, 95% CI: 1.24–3.46, P = 0.005), 1-docosahexaenoylglycerophosphocholine* (OR 1.89, 95% CI: 1.04–3.46, P = 0.038), 1-arachidonoylglycerophosphoethanolamine* (OR 1.95, 95% CI: 1.21–3.15, P = 0.006), hexadecanedioate (OR 1.25, 95% CI: 1.04–1.49, P = 0.017), n-Butyl Oleate (OR 2.19, 95% CI: 1.2–3.98, P = 0.011), X-09706 (OR 1.97, 95% CI: 1.16–3.33, P = 0.012), X-11,876 (OR 1.39, 95% CI: 1.02–1.9, P = 0.036), X-12,450 (OR 1.46, 95% CI: 1.03–2.05, P = 0.031), X-12,844 (OR 2.12, 95% CI: 1.29–3.48, P = 0.029) and X-14,189 (OR 1.28, 95% CI: 1.05–1.57, P = 0.015) may increase the risk of developing DR. In addition, serum levels of cis-4-decenoyl carnitine (OR 0.65, 95% CI: 0.44–0.95, P = 0.027), inosine (OR 0.89, 95% CI: 0.81–0.98, P = 0.023), pro-hydroxy-pro (OR 0.53, 95% CI: 0.32–0.88, P = 0.014), X-12,833 (OR 0.92, 95% CI: 0.87–0.97, P = 0.001), hippurate (OR 0.75, 95% CI: 0.58–0.98, P = 0.003), and 4-hydroxyhippurate (OR 0.61, 95% CI: 0.4–0.94, P = 0.025) may help reduce the risk of developing DR. The detailed IVW results are shown in Fig. 4.Fig. 4Forest plot for the associations between metabolite and risk of DR after removal of confounders. DR diabetic retinopathy; CI confidence interval; IVW inverse-variance weighted; OR odds ratio.Effect of genetically determined metabolites on PDRThe IVW approach tentatively flagged 19 metabolites as being notably associated with PDR. The results were verified according to the MR-Egger and WM methods, and the metabolites with different directions of conclusions for the three models were deleted. Meanwhile, multiple sensitivity analyses were used to validate the results, and a total of six metabolites were deleted. Among them, eight metabolites exhibit a positive causal relationship with the onset of PDR, including 3 Lipids (linoleate (18:2n6), 1-arachidonoylglycerophosphoethanolamine*, and tetradecanedioate), 1 energy (alpha-ketoglutarate), 1 nucleotide (N2,N2-dimethylguanosine), and 3 metabolites whose chemical composition is unknown (X-11529, X-11538 and X-13431). Five metabolites were negatively correlated with PDR. In addition to 3 metabolites of unknown chemical composition (X-03094, X-06351 and X-11381), 1 enobiotics (mannitol) and 1 xenobiotics (salicylate) were included.Confounding factor screening was performed, and after removing confounding SNPs, MR analysis and sensitivity analysis were performed again.IVW results showed that 5 metabolites lost significant association with PDR (linoleate (18:2n6), tetradecanedioate mannitol, X-03094, X-13431). Estimates for the remaining 8 metabolites remained significant, and the direction of conclusions was consistent with the primary analysis. Specifically, genetic predisposition for elevated levels of alpha-ketoglutarate (OR 1.41, 95% CI: 1.01–1.97, P = 0.044), N2,N2-dimethylguanosine (OR 1.43, 95% CI: 1.03–1.98, P = 0.032 ), 1-arachidonoylglycerophosphoethanolamine* (OR 1.75, 95% CI: 1.19–2.58, P = 0.004), X-11,529 (OR 1.19, 95% CI: 1.10–1.20, P < 0.0001), and X-11,538(OR 1.26, 95% CI: 1.02–1.56, P = 0.034) was associated with increasing susceptibility to PDR. And genetic predisposition for elevated levels of salicylate (OR 0.63, 95% CI: 0.39–0.99, P = 0.049), and X-06351 (OR 0.82, 95% CI: 0.67–0.99, P = 0.045 ), X-11,381 (OR 0.54, 95% CI: 0.30–0.98, P = 0.045) was associated with reduced susceptibility to PDR. Detailed IVW results are shown in Fig. 5.Fig. 5Forest plot for the associations between metabolite and risk of PDR after removal of confounders. PDR proliferative diabetic retinopathy; CI confidence interval; IVW inverse-variance weighted; OR odds ratio.Metabolic pathway analysisWe analysed 16 metabolites with known chemical compositions in MetaboAnalyst 5.0 and found two metabolic pathways associated with DR (Table 2). Isoleucine was involved in valine, leucine and isoleucine biosynthesis (P = 0.03062), and hippurate was involved in phenylalanine metabolism (P = 0.03815). The abovementioned metabolic mechanisms may influence the pathogenesis and development of DR.Table 2 Significant metabolic pathways involved in the pathogenesis of DR.General condition of ratsCon group rats grew steadily and symmetrically and were responsive and docile. The hair was relatively clean, white and smooth. Water intake, food intake, and urine output values were not high. The bedding was changed every 3 days. There were no significant changes in ocular characteristics. The DR group showed some typical symptoms of diabetes, including polydipsia, polyphagia and polyuria, from approximately 3 days after the model was established. The bedding material was replaced once a day, and the excreta was smelly and viscous. After 2 weeks, the rats in the DR group exhibited weight loss, and the body shape gradually showed a difference from that of the Con group. After 3 weeks, the rats in the DR group showed abdominal distention, messy hair, yellow and dull hair colour, and an irritable mood. In the DR group, metabolic cataracts gradually appeared in some rats from the 7th week of modelling, but there were individual differences (Fig. 6A).Fig. 6Eight weeks after intraperitoneal injection, rats developed the typical pathological manifestations of DM and DR. (A) Comparison of appearance between the two groups of rats. The DR group rats exhibited a thin body shape, withered hair and diabetic cataracts. (B) Body weights; (C) random blood glucose levels. (D) Representative retinal morphology images of H&E staining. All data are presented as the mean ± SD; n = 6 per group. ****p < 0.0001, ***p < 0.001, **p < 0.005, *p < 0.05. Con control; DR diabetic retinopathy.The rats in the Con group gained weight steadily, averaging 37.25 ± 2.875 g/week over the 8 weeks following intraperitoneal injection. In contrast, the rats in the DR group had a slower average weight gain of 6.67 ± 2.08 g/week (Fig. 6B). Additionally, the random blood glucose level in the Con group remained consistently within the normal range. That of the DR group significantly increased one week after modelling but then slightly decreased, yet it remained consistently higher than 16.7 mM and tended to be stable(Fig. 6C).Rat retinal tissue was observed by H&E stainingThe inner limiting membrane of the retina in the control group was found to be smooth and intact by (H&E) staining. The ganglion cell layer showed an orderly and regular arrangement with good morphology and no apparent abnormalities in its monolayer structure. Additionally, the inner nuclear layer was neatly arranged, whereas the nuclei of the outer nuclear layer were darkly pigmented, orderly, and closely placed. The borders between the layers were well defined and sharply demarcated. This contrasted with the DR group, which showed thinning of the retinal tissue. The nerve fiber layer demonstrated an elevated thickness, with the presence of oedema and a disorganized and sparse distribution of cells. The ganglion cell layer showed cavitation. The outer nuclear layer was pale, while the inner and outer nuclear layers were disorganised. Additionally, the boundaries between layers were unclear (Fig. 6D). These observations align with the established features of rat retinopathy.Apoptosis of rat retinal cellsThe TUNEL signal (green) in the retina of the DR group was obvious and overlapped with the DAPI signal (blue), indicating that cells in multiple layers of the retina had undergone apoptosis (Fig. 7A). No obvious apoptotic signal was observed in the Con group. The optical density values of the staining results for the total retinal thickness of each group were quantified. The apoptosis signal of retinal tissue of rats in the DR group (n = 6) was significantly increased, and the optical density was 1.8 times that of the average optical density of rats in the Con group (Fig. 7B). According to the H&E and TUNEL results, the DR group showed early diabetic retinopathy-related lesions.Fig. 7Rat retinal cell apoptosis and blood metabolite level changes. A TUNEL immunoresponse staining showed that retinal cell apoptosis was significantly increased in the DR group. B Representation of the mean TUNEL-positive ratio in the retinas of rats. C, D,E Serum metabolites were detected by ELISA. The levels of isoleucine and 4-HPA were increased, and the levels of inosine were decreased in the DR group. All data are presented as the mean ± SD; n = 6 per group. ****p < 0.0001, ***p < 0.001, **p < 0.005, *p < 0.05. Con control; DR diabetic retinopathy.Changes in serum metabolite levels of rats in each groupSerum samples were collected for determination of isoleucine, 4-HPA and inosine levels. Isoleucine levels were 259.7 ± 26.4 ng/L and 303.6 ± 22.2 ng/L in the Con and DR groups, respectively, as determined by ELISA. The 4-HPA levels measured were 331.0 ± 31.8 ng/L and 371.2 ± 30.9 ng/L. The inosine levels were 1115.1 ± 165.9 ng/L and 856.3 ± 244.6 ng/L, respectively. Serum isoleucine (Fig. 7C) and 4-HPA (Fig. 7D) levels were markedly increased (P < 0.001, P = 0.01), whereas inosine levels were reduced (P = 0.031) among DR group compared with Con group rats (Fig. 7E).
