Lipidomic analysis of vitreous humor derived from patients with proliferative diabetic retinopathy

Abstract

Purpose

To compare the lipid profiles of vitreous humor (VH) obtained from patients with proliferative diabetic retinopathy (PDR) and healthy controls and to determine possible metabolic pathways associated with PDR.

Methods

VH samples from both PDR patients (PDR group) and healthy subjects (Control group) were harvested for lipid analysis using high-throughput liquid chromatography tandem mass spectrometry (LC-MS/MS). The differential expressions of lipids between the control and PDR groups were determined by both univariate and multivariate methods. The sensitivity and specificity of the differentially expressed lipids as the candidate biomarkers for DR were determined through the area under receiver operating characteristic (ROC) analysis. MetaboAnalyst 5.0 was used for metabolic pathway analysis.

Results

Lipids were abundant in the VH samples, and LC-MS/MS analysis identified 35 lipid classes, 849 lipid species from both groups. There were much higher total lipid levels in the VH derived from PDR patients than that in the controls and the levels of cholesteryl Ester (ChE), phosphatidylserine (PS), phosphatidylcholine (PC), phosphatidylinositol (PI), and sphingomyelin (SM) obviously increased in the PDR patients as compared with the controls (all p < 0.01). Of these lipid species, a total of 256 lipids (245 upregulated and 11 downregulated) were determined by combining univariate and multivariate statistical analyses. ROC analysis showed the selected differentially expressed lipids had an area under ROC greater than 0.8, exhibiting good diagnostic value in distinguishing PDR patients from the controls. Pathway analysis showed that these differentially expressed lipids were mainly enriched in sphingolipid and purine metabolic pathways.

Conclusion

Differentially expressed lipids in the VH of PDR patients were shown as ideal candidate metabolic biomarkers for clinical diagnosis and prognosis of DR, shedding new light on the metabolic mechanism of DR.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12886-025-04347-7.

Introduction

Diabetic retinopathy(DR), a prevalent microvascular complication associated with diabetes, continues to pose a significant global public health challenge. It is the primary cause of vision loss among individuals of working age [1]. The prevalence of diabetic retinopathy (DR) is anticipated to rise due to the increasing rates of obesity, particularly in China, the largest developing nation. Epidemiological studies have revealed that China accounts for 25% of individuals diagnosed with diabetes globally, with 11.6% of the adult population in the country living with diabetes and 50.1% of adults have prediabetes. DR has therefore become a serious health issue in China [2–4].

DR typically starts with asymptomatic retinal abnormalities and can advance to sight-threatening complications, such as proliferative diabetic retinopathy (PDR). A pressing need exists for a better understanding of the basic mechanisms underlying DR [5, 6]. An enlarged number of risk factors, like obesity, blood pressure, and blood levels of glucose, urea nitrogen, glycosylated hemoglobin, and creatinine are associated with the incidence of DR [7, 8]. Lipid metabolic abnormalities have also been identified as potential factors that may increase the risk of developing and worsening of DR.

Landmark studies have shown that high plasma lipid and lipoprotein levels are potential risk factors for DR and show a positive association with DR progression [9–11]. However, the lipid profiles in the vitreous humor (VH) of patients with PDR remain relatively unexplored, as only a few investigations have focused on PDR-related changes in vitreous lipids [12–14]. VH is crucial for maintaining ocular function and any structural or molecular changes in the vitreous related to DR would be expected to arise as consequences of metabolic and functional modifications of the retinal tissues during DR development [15]. This suggests that the analysis of the lipids in VH from patients with PDR could represent a new tool for acquiring preclinical and experimental evidence for the development of novel drug candidates for DR.

In recent decades, the accurate identification of chemical components in natural products has been attained by ultra-high performance liquid chromatography (UPLC) coupled with mass spectrometry (MS) [16, 17]. In the current study, a strategy based on a coupled UPLC-Orbitrap MS system was used for quick identification and precise measurement of different lipid elements in VH. The purpose of the present study was to compare the lipid profiles between healthy controls and PDR patients and to determine possible metabolic pathways associated with PDR. Gaining a more thorough comprehension of the specific lipid alterations that occur in the VH of PDR patients may offer fresh perspectives on the disease’s pathophysiology.

Methods

Patients and samples

The protocol for this study adhered to the ethical principles detailed in the Declaration of Helsinki. It also obtained formal approval from the Institutional Review Board of Qilu Hospital, Shandong University. Prior to enrollment, All participants provided written informed consent after receiving a detailed explanation of the study’s objectives, experimental procedures, and potential risks and benefits.

This case-control study prospectively recruited 30 patients with PDR complicating type 2 diabetes mellitus (T2DM) along with 30 age-matched nondiabetic controls through consecutive enrollment at the Ophthalmology Department of Qilu Hospital, Shandong University between January 2022 and September 2024. All subjects with PDR underwent primary pars plana vitrectomy (PPV) surgery as required (PDR Group). The control cohort comprised age-matched non-diabetic individuals undergoing pars plana vitrectomy (PPV) for surgically managed idiopathic macular pathologies, including epiretinal membrane and/or macular hole formation. Table 1 showcases the clinical characteristics of all participants included in this study.

Table 1.

Clinical characteristics and metabolic parameters of the study subjects

	PDR patients (PDR group, n = 30)	Control (Control group, n = 30)	P values
Age, y	61.19 ± 3.09	60.26 ± 2.89	0.815
Male/Female	14/16	12/18	0.602
Duration of DM, y	13.41 ± 3.26	-	-
TCho(mmol/L)	4.98 ± 0.35	5.01 ± 0.17	0.327
HDL-C(mmol/L)	1.21 ± 0.09	1.19 ± 0.16	0.411
LDL-C(mmol/L)	2.29 ± 0.20	2.77 ± 0.49	0.661
TG(mmol/L)	1.81 ± 0.21	1.79 ± 0.28	0.860
FBG(mmol/L)	6.90 ± 0.56	5.32 ± 0.11	0.000

Data were expressed as mean ± standard errors (SE)

TCho Total cholesterol, TG Triglycerides, LDL-C Low-density lipoprotein cholesterol, HDL-C High-density lipoprotein cholesterol, FBG Fasting blood glucose, DM Diabetes mellitus

PDR is determined by two different graders based on the early treatment of diabetic retinopathy study (ETDRS) scale. Previous medical information were gathered through a standard questionnaire as we used in our previous studies [18, 19]. Every participant went through a thorough ocular evaluation. Exclusions included history of previous ocular surgery and other ocular disease, like glaucoma, high myopia and retinal diseases. Participants who were taking lipid-lowering drugs were not included in this study.

Non-dilute and transparent VH samples were collected before the infusion and promptly transferred to dry ice and stored in Eppendorf tubes at −80 °C until further assays. Fasting venous blood specimens were obtained through standard venipuncture protocol within 2 h of morning awakening during the initial hospitalization phase. Subsequent biochemical analyses conducted at Qilu Hospital’s institutional laboratory quantitatively assessed serum concentrations of lipid profile parameters (total cholesterol (TCho), triglycerides (TG), low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C) and fasting plasma glucose (FPG) using standardized enzymatic assays.

Lipid extraction

Lipidomic profiling was conducted through mass spectrometry (MS)-based lipid detection, performed by Applied Protein Technology (Shanghai) Co., Ltd. using standardized extraction protocols and quality-controlled analytical workflows [20, 21].

Lipid isolation was achieved through methyl tert-butyl ether (MTBE)-based phase separation methodology. Interbatch analytical quality assurance was implemented by generating composite reference samples through equimolar pooling of biological replicates from each study cohort. Take a separate sample in each group and mix them equally together to create a pooled quality control (QC) sample. These standardized QC specimens were strategically embedded within the experimental workflow using randomized injection sequences to continuously validate system equilibration status and ensure measurement reproducibility across longitudinal analyses. QC measurements were conducted systematically after ten measurements as well as before and after measuring the samples with elevated organic content.

Liquid chromatography tandem mass spectrometry (LC-MS/MS) method for lipid analysis

Chromatographic separation and mass spectrometric detection were conducted using a high-resolution Q-Exactive plus mass spectrometer (Thermo Fisher Scientific) integrated with an UHPLC Nexera LC-30 A (Shimadzu) through heated electrospray ionization interface [20, 22]. The LC separation was performed with a Shimadzu 20AC-XR system equipped with a Waters 2.1 × 100 mm, 3.5 μm XSelect Peptide CSH C18 column (1.7 μm, 2.1 mm× 100 mm, Waters) and a Vanguard pre-column with the same packing material. The oven temperature was set to 75 °C. MS signals were acquired with the Q-Exactive Plus in positive and negative modes.

As we used before, the lipid components were analyzed with the UPLC-Orbitrap MS system and simultaneously identified using an automated search tool (Lipidsearch software v.4.1, Thermo Scientific). Among the extracted ion features, only variables demonstrating ≥ 50% detection frequency across biological replicates within any experimental cohort were retained.

Statistical and data analysis

Statistical analyses were executed utilizing SPSS statistical software 20.0 (SPSS Inc., Chicago, IL, USA). The lipids showing significant differences between the PDR patients and normal controls were analyzed thorough combining both univariate and multivariate methods. The univariate analysis comprised analysis of variable fold change and the two-tailed Student’s test. The SIMCA-P software 14.1(Umetrics, Umea, Sweden) was used for multivariate analysis. Discriminatory lipid species between the two groups were identified through the combination of variable influence on projection (VIP) values obtained from the orthogonal partial least-squares-discriminant analysis (OPLS-DA) model and a two-tailed Student’s t test (p value) performed on the raw data. Lipids, exhibiting VIP scores exceeding 1.0, coupled with p-values lower than 0.05, were deemed to display significant differences. Receiver operating characteristic (ROC) analysis was utilized to evaluate the diagnostic value of the significant differential lipids in PDR patients compared with the healthy subjects. Depending on the scenario, either the Student’s t-test or the Mann-Whitney test was employed to compare two independent sample groups. A p-value of less than 0.05 was considered statistically significant. All the data were presented as mean ± SEM.

Results

Basic clinical characteristics of the subjects

The final study population consisted of 30 patients with PDR and 30 nondiabetic healthy participants. Except for the higher levels of FBG in PDR patients, whereas other parameters, including sex, age, and serum lipid components (TCho, HDL-C, sdLDL, and TG), did not differ significantly between the two groups. An overview of basic characteristics and clinical parameters of all participants were shown in Table 1.

Altered composition of lipid classes in VH of PDR patients

The LC-MS/MS analysis revealed much higher total lipid levels in the VH of the PDR patients than in the controls (Fig. 1A). In total, 35 lipid classes, 849 lipid species were identified in both groups (Fig. 1B). The overall lipid compositions and distribution patterns were shown in Fig. 1C, D. The top five abundant lipid classes in the VH were almost identical in both groups and consisted of cholesteryl ester (ChE) > monogalactosyl diacylglycerol (MGDG) > Diacylglycerol (DG) > Zymosterol (ZyE) > others. To be mentioned, ChE was the most abundant lipid in VH, accounting for 62.01% and 60.55% of the total lipids in the PDR and control groups, respectively. In addition, the relative compositions (%) of several lipid classes showed significant differences between the two groups. The levels of ChE, phosphatidylcholine (PC), phosphatidylinositol (PI), Phosphatidylserine (PS), and sphingomyelin (SM) were much higher in the PDR patients than in the controls (all p < 0.01, Fig. 2).

Fig. 1.

Comparisons of lipid contents in vitreous humor between the PDR patients and controls. A The histogram indicated the total lipid in both the PDR patients and control subjects. B Numbers of reliably detected lipid species in both two groups. C The distribution of lipid classes in vitreous humor of control subjects. D The distribution of lipid classes in vitreous humor of PDR patients. ChE: cholesterol ester; MGDG: monogalacto syldiacylglycerol; DG: diglyceride; ZyE, Zymosterol; PG: phosphatidylglycerol; PC: phosphatidylcholine; PI: phosphatidylinositol; TG: triglyceride; WE: wax exters. SM: sphingomyelin; Cer: ceramides; CerP: ceramides phosphate; PA: phosphatidic acid; PE: phosphatidylethanolamine; CL: cardiolipin;PS: phosphatidylserine; PIP2: phosphatidylinositol-phosphate 2; PIP3: phosphatidylinositol-phosphate; GM1: gangliosides 1; GM2: gangliosides 2; LPC: lysophosphatidylcholine; LPE: lysophosphatidylethanolamine; DGDG: digalactosyldiacylglycerol; SQDG: Sulfoquinovosyldiacylglycerol; SQMG: Sulfoquinovosylmonoacylglycerol; LSM: lysosphingomyelin; MG: monoglyceride; OAHFA: (O-acyl)-1-hydroxy fatty acid; phSM: sphingomyelin (phytosphingosine); PIP: phosphatidylinositol-phosphate; ST: sulfatide; StE: stigmasterol ester;

Fig. 2.

Comparisons of lipid classes and fold changes in vitreous humor of the PDR patients and control subjects (A, B)

Identification of candidate metabolic biomarkers for DR diagnosis

To determine whether these differentially expressed lipid species could be used as candidate metabolic biomarkers for DR diagnosis, ROC analysis was performed to assess the sensitivity and specificity of these metabolite biomarkers. Five lipids derived from significantly upregulated lipids in PDR vitreous (Fig. 2, all p < 0.01) were randomly selected and subjected to ROC curve analysis. The results showed that these five metabolites had an AUC greater than 0.8, exhibiting good diagnostic value in distinguishing PDR patients from the controls(Fig. 3).

Fig. 3.

ROC analysis for the selected differently expressed lipid species in distinguishing PDR from control subjects

Multivariate and univariate analyses of differentially expressed lipid species

Multivariate analysis was used to evaluate the separation between the two groups. Both principal component analysis (PCA) and the OPLS-DA model displayed remarkable clusters of lipids in the controls and PDR patients (Fig. 4A, B). OPLS-DA possibly overfits data, therefore, the model was verified again by a permutation test which indicated no over-fitting in the OPLS-DA model (Fig. 4C).

Fig. 4.

The Scatter Graph showed the differentially expressed lipid species between the two groups based on univariate analysis. Red circles represent differential lipids, while non-significant lipids are shown in black

Univariate statistical analysis revealed a total of 256 lipids (245 upregulated and 11 downregulated, Supplementary File S1) that were significantly different between the PDR and control groups. The key variables (p < 0.05 and VIP > 1) are presented in the volcano plot (Fig. 5).

Fig. 5.

The score plot of principal component analysis (PCA) and the orthogonal partial least-squares-discriminant analysis (OPLS-DA) model indicated intrinsic differences between the two groups. A The PCA score plot of PDR-controls. B The OPLS-DA score plot of PDR-controls. C The OPLS-DA displacement test of PDR-controls

Metabolic pathway analysis

Eleven differential metabolites were incorporated into metabolic pathway analysis. Among them, 10 metabolites, including ChE, PC, ceramides (Cer), ceramides phosphate (CerP), SM and triglyceride (TG), were upregulated, and 1 metabolity, sulfoquinovosylmonoacylglycerol (SQMG), was downregulated in the PDR patients compared to controls. The results from the pathway analysis indicated that these differential metabolites were significantly enriched in the metabolism of sphingolipids and purines. The detailed results are shown in Fig. 6.

Fig. 6.

Summary of pathway analysis for the differentially expressed lipids with MetaboAnalyst 5.0. (a) Sphingolipid metabolic pathway. (b) Purine metabolic pathway

Discussion

In addition to the known effects of hyperglycemia and the duration of diabetes, lipids also play many essential roles in the pathogenesis of DR, and many plasma lipids and lipoproteins have been proposed as potential risk factors for the development of this condition [10, 23]. Plasma lipids, such as LDL, VLDL, HDL, and TCho, have been reported to show positive associations with DR in several clinical trials and epidemiological studies [11]. Elevated levels of lipids, including eicosanoids and sphingolipids, in the diabetic vitreous have also been closely associated with diabetic progression [13, 14, 24]. In the present study, we clearly demonstrated the differentially expressed lipids in the VH of PDR patients compared with normal nondiabetic controls. Our study emphasizes the significance of changes in lipid profiles within the diabetic VH and endorses the idea that disturbances in lipid metabolism contribute to the development of PDR.

We observed notable increases in the total lipid levels present in the VH of patients with PDR, when compared to non-diabetic counterparts. However, the cellular source of these VH lipids is unknown. Lipids are fundamental constituents of the retina tissues, and many unique lipids, including free cholesterol and phospholipid species, in the retina play fundamental roles in retinal function and disease [25, 26]. In addition, each of the different cell types in the retina, especially the Müller glia cells retinal pigment epithelium (RPE) and retinal ganglion cells (RGCs) as well as from infiltrated inflammatory cells, contain numerous lipid species, making the distributions of specific lipids difficult to define. Structural and molecular changes of the VH, as observed during DR progression, result from metabolic and functional alterations in retinal tissues, which are stimulated by the hyperglycemia, hypoxia, and oxidative, inflammatory, neurodegenerative, and leukostatic conditions that occur during diabetes. Accumulation of lipids in the VH is an indication of a distinct paracrine effect on retinal tissues. Except for the lipids with paracrine functions from the retina, part of vitreous lipids may originate from the systemic lipid leakage due to blood-retinal barrier breakdown or vitreous hemorrhage during the development of DR.

Further analysis of the composition of the lipid classes showed that ChE was the most abundant lipid, accounting for approximately 60% of the total lipids in the VH of both the PDR and control groups. This finding is in agreement with a similar report of the lipid composition of human aqueous humor (AH) [27]. Jahn et al. found that there was about 9.0 mg/dl for ChE in human AH (16.4 mg/dl for total lipids, 54.8%). ChE plays a crucial role in regulating the cellular reactions to oxidative stress and is usually increased in response to oxidative conditions. In our study, the ChE level in PDR patients was statistically significantly higher comparable to the controls, indicating a potential link between disrupted cholesterol metabolism and the development of the disease.

In the present study, we also found higher levels of TG in the VH of the patients with PDR than in the controls. Elevated TG levels are common in diabetic patients, and Serum TG levels are linked to the initial occurrence of cardiovascular events in high-risk diabetic patients who also have hypercholesterolemia and retinopathy [28]. However, the expression level of TG in the human VH is unknown. Our finding showed that the vitreous levels of TG were significantly elevated in patients with PDR than in the control subjects, suggesting that TG may possibly participate in the pathogenesis of PDR.

PS is an essential component in all human cells and has important roles in the regulation of cell apoptosis [29, 30], where it creates a scaffold for blood-clotting factors on activated platelets and participates in blood coagulation (29). PDR is distinguished by the development of neovascularization on the retinal surface, frequently leading to vitreous hemorrhage. Given these characteristics, a reasonable presumption is that a significantly elevated level of PS in the diabetic vitreous may partly explain the common observance of vitreous hemorrhage in PDR. Previous reports have indicated that supplementation with PS can significantly reverse the reduced survival in diabetes mellitus due to the consumption of a high-glucose diet and that PS species can serve as an immune therapy target for cancer [31, 32], suggesting that PS may represent a promising novel drug for mono- or polytherapy for PDR. This raises the intriguing question of whether supplementation with PS could reverse or slow the progression of DR. Preclinical research studies and clinical trials will be needed in the future to verify whether PS could be an alternate therapy for treating DR.

Through the screening of differentially expressed lipid species, we identified a total of 256 differential lipids between the PDR and control groups. These differential metabolites were mainly significantly enriched in sphingolipid and purine metabolic pathways. Irregularities of various sphingolipids, such as glycosphingolipids and sphingosine, have been linked to the development of diabetes mellitus, and changes in sphingolipid profiles have been identified in the retinas of diabetic individuals and contribute to retinal vascular pathology in animal models of diabetes [33, 34]. Dysregulation of sphingolipid metabolism, with upregulation of acid sphingomyelinase (ASM) and higher hexosyl ceramide production plays an important role in diabetic dyslipidemia-induced retinal damage. In retinal microglia, ASM activation promotes cytokine release, while ceramide acts as a second messenger to trigger endothelial cell apoptosis. In this study, levels of various sphingolipid species including Cer, CerP and SM in PDR patients were obviously higher than the controls, which was consistent with one previous study about the sphingolipid composition in human VH from diabetic individuals [13].

Consistent with prior reports, our findings demonstrate significant alterations in purine metabolic pathways within the vitreous humor of PDR patients. Purines interact with purinoceptors to mediate extracellular signal transduction, and their corresponding metabolites are pivotal drivers in the progression of various human diseases. Purine metabolism was identified as the most enriched pathway through metabolic pathway analysis, indicating strong relation with the development of diabetic microvascular complication. In order to determine whether altered metabolism metabolites could be used as candidate metabolic biomarkers for DR diagnosis, we assessed the sensitivity and specificity of these metabolite biomarkers using ROC analysis. Five randomly selected lipids derived from significantly upregulated lipids in PDR vitreous had an AUC greater than 0.8. Among them, ChE had the highest sensitivity and specificity for DR diagnosis, suggesting the strong association between cholesterol and the process of purine metabolism.

While this investigation provides novel insights, certain constraints merit discussion. Firstly, the participant cohort was relatively small and from a single institution, which may affect the extrapolation of findings to broader populations. Secondly, the enrolled PDR patients had significantly higher FBG than controls. It has been known that FBG is significantly associated with all lipid profile parameters. Therefore, higher FBG may be‌ the ‌underlying cause‌ of abnormal vitreous lipids in PDR patients. Thirdly, The control group consists of non-diabetic patients undergoing vitrectomy for ERM or macular hole, which may not represent true normal VH. Chronic traction or localized inflammation in these eyes might subtly alter lipid profiles, these confounding effects are likely minimal compared to the profound metabolic changes in diabetic VH.

In summary, our study provides direct biological evidence for a role for abnormal lipid metabolism in the pathogenesis of PDR. Understanding the implications of the major differentially expressed lipids may be attained in further studies on transgenic animal models and human studies with larger sample sizes and may provide novel therapeutic targets for PDR.

Authors’ contributions

All authors conceived of and designed the experimental protocol. Jiawei Wang: Writing – review & editing, Methodology, Formal analysis.Zheyi Shao: Writing – review & editing, Formal analysis. Wanna li: Methodology, Formal analysis, Writing – review & editing. Yihan Zhao: Methodology, Formal analysis. Writing – review & editing. Shiqiang li: Methodology. Chenzhao Ma: Methodology. Jianqiao Li: design of the work, Writing – review & editing. All authors read and approved the final manuscript.

Funding

The study has been supported by Natural Science Foundation of Shandong Province (ZR2023MH139) and Grant of Shandong University (6010220078).

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

This study received approval from the the Ethical Review Committee of Qilu Hospital of Shandong University. All research methods adhered to the principles outlined in the “Declaration of Helsinki”. Prior to conducting the study, the objectives and methods were presented to the patients to obtain informed consent, and signatures were obtained accordingly.

Consent for publication

Informed, Not applicable.

Competing interests

The authors declare no competing interests.