Metabolomic Profiling of Vitreous Humor Reveals Distinct Metabolic Signatures in Proliferative Vitreoretinopathy

Abstract

Purpose

Proliferative vitreoretinopathy (PVR) remains associated with a poor prognosis despite advances in vitreoretinal surgery. In the absence of effective adjunctive therapies, this study aimed to characterize vitreous metabolic alterations in PVR to better understand its pathophysiology.

Methods

Untargeted metabolomic profiling was conducted on vitreous humor samples from 38 eyes that underwent surgery: PVR (grade C, n = 10), rhegmatogenous retinal detachment (RRD; grades A and B, n = 18), and controls undergoing intraocular lens (IOL) scleral fixation (n = 10). Metabolites were analyzed using capillary electrophoresis–mass spectrometry. Metabolic profiles were compared pairwise among the three groups with multiple testing correction, and pathway enrichment analysis was performed.

Results

RRD demonstrated no disease-specific metabolic signature relative to controls after multiple testing correction. In contrast, PVR exhibited a distinct metabolic profile compared with both RRD and controls. Pathway enrichment analysis revealed upregulation of amino acid and nitrogen metabolism—including arginine and proline metabolism, glycine and serine metabolism, the urea cycle, ammonia recycling, alanine metabolism, and the malate–aspartate shuttle—consistent with increased biosynthetic demand. Pathways related to fibrosis and stress responses were also enriched.

Conclusions

This study shows that PVR exhibits a unique metabolic signature distinct from RRD, characterized by the activation of metabolic pathways that support increased energy demand, cellular proliferation, stress responses, and fibrotic remodeling. These findings provide insights into the metabolic mechanisms associated with the progression from RRD to PVR, suggesting potential therapeutic targets to prevent fibrosis and aberrant cell proliferation.

Rhegmatogenous retinal detachment (RRD) is a vision-threatening disorder in which a retinal tear permits fluid to separate the neurosensory retina from the underlying retinal pigment epithelium (RPE).¹ Without timely treatment, RRD may progress to severe visual impairment and even blindness.²

Proliferative vitreoretinopathy (PVR) is the greatest complication of RRD and, despite modern surgical techniques, often results in surgical failure and marked vision loss. PVR is a complex fibrocellular process characterized by the proliferation, migration, and contraction of RPE, glial, fibroblastic, and inflammatory cells within the vitreous chamber and over the retinal surface.³ Among these, RPE cells are a predominant component of the fibrotic membranes in PVR and are regarded as central drivers of the disease.⁴ Epithelial–mesenchymal transition (EMT) is the process by which epithelial cells lose epithelial traits and acquire a mesenchymal phenotype, and EMT in RPE cells is considered a fundamental mechanism underlying PVR membrane formation.^5,6 This cellular activity results in the formation of epiretinal membranes, subretinal strands, and vitreous bands, causing retinal distortion, traction, and ultimately, recurrent retinal detachment.^3,7–15

Despite the clear clinical need, effective pharmacological treatments for PVR remain lacking. Several pharmacological treatments, such as systemic prednisolone,¹⁶ infused dexamethasone,¹⁷ intravitreal triamcinolone,^18,19 and slow-release dexamethasone,²⁰ have been investigated for PVR, but none have yet been translated into practical clinical use. As noted above, PVR is driven not by inflammation alone but by a complex fibrocellular process involving proliferation, migration, and contraction of multiple cell types; therefore, steroid therapy targeting inflammation alone is unlikely to be sufficient. Therapeutic strategies that target non-inflammatory mechanisms of PVR are needed.

Metabolomics systematically profiles the small-molecule metabolites in a biological system, offering a moment-in-time picture of the metabolic state of cells, tissues, or whole organisms.^21,22 As the final outputs of cellular pathways, metabolite abundances capture the combined influence of gene expression, protein function, and environmental conditions. Metabolomics offers a powerful approach to understanding the complex biochemical alterations underlying disease pathogenesis.^21,23

Previous metabolomic studies have identified altered metabolic profiles in various retinal diseases, such as diabetic retinopathy (DR) and age-related macular degeneration (AMD).²⁴ In DR, metabolomic analysis of the vitreous humor has revealed elevated levels of glucose, sorbitol, and mannitol, as well as decreased levels of galactitol and ascorbic acid.²⁵ In AMD, altered lipid metabolism and oxidative stress have been identified as key metabolic features.²⁶

Although metabolomic studies have been conducted in other retinal disorders, only a few have focused specifically on RRD and PVR. One study compared the metabolomic profiles of vitreous samples from patients with RRD, PVR, and healthy controls, identifying 31 differential metabolites and dysregulation of pathways related to inflammation, proliferation, and energy consumption.²⁷ Another study investigated the vitreous humor profiles of RRD associated with choroidal detachment (RRDCD), identifying 24 metabolites that differed between RRDCD and RRD samples.²⁸ These studies suggest that metabolomics have the potential to provide valuable insights into the pathogenesis of RRD and PVR, but further research is needed to validate these findings and identify potential biomarkers.

Based on these prior observations, we formulated an a priori hypothesis that, compared with RRD, the vitreous in PVR would exhibit alterations in energy-producing pathways such as glycolysis and the tricarboxylic acid (TCA) cycle, together with pronounced changes in amino acid and nucleotide metabolism required for cellular proliferation and fibrotic remodeling.

This study builds upon the established knowledge of cellular and molecular mechanisms of PVR by using a comprehensive metabolomics approach to identify novel metabolic signatures associated with PVR. By comparing the vitreous humor metabolome of patients with RRD and PVR to that of controls, this research aims to provide a more complete understanding of the biochemical processes driving PVR pathogenesis.

Methods

Ethics Approval

The protocol was reviewed and approved by the Ethics Review Committee of the Chiba University Graduate School of Medicine (Reference No. M10537; Approval Date: January 11, 2023), in accordance with the tenets of the Declaration of Helsinki. Written informed consent was obtained from all patients prior to their participation in the study.

Patient Selection

This study included patients with RRD who underwent surgery at Chiba University Hospital between February 2023 and April 2024. RRD cases were classified into grade A, grade B, and grade C based on the updated Retina Society classification of PVR.²⁹ Patients undergoing intraocular lens (IOL) scleral fixation for IOL dislocation were enrolled as the control group. Exclusion criteria included a history of any intraocular surgery (including intravitreal anti-VEGF injections) within the preceding 6 months, the presence of significant vitreous hemorrhage, or active ocular lesions associated with other conditions such as inflammatory eye disease.

Vitreous Humor Sample Collection

At the start of vitreous surgery and before initiating intraocular perfusion, approximately 1.0 mL of undiluted vitreous humor was collected using a 5-mL syringe connected to the vitreous cutter. The collected samples were centrifuged and the supernatants were aliquoted into tubes and stored in a freezer at −80°C until analysis.

Metabolite Extraction

Fifty microliters of each sample were added to 200 µL of methanol containing internal standards (H3304-1002; Human Metabolome Technologies, Inc. [HMT], Tsuruoka, Yamagata, Japan) at 0°C to suppress enzymatic activity. The extract solution was thoroughly mixed with 150 µL of Milli-Q water. A 300-µL aliquot of this solution was then filtered by centrifugation (9100×g at 4°C) through a 5-kDa cutoff filter (ULTRAFREE MC PLHCC; HMT) to remove macromolecules. The filtrate was then evaporated to dryness under vacuum and reconstituted in 50 µL of Milli-Q water for metabolome analysis at HMT.

Metabolome Analysis

Metabolome analysis was conducted according to HMT's ω Scan package, using capillary electrophoresis–Fourier transform mass spectrometry (CE–FTMS), as detailed previously.³⁰ In brief, CE–FTMS analysis was performed on an Agilent 7100 CE capillary electrophoresis system equipped with a Q Exactive Plus mass spectrometer (Thermo Fisher Scientific Inc., Waltham, MA, USA), an Agilent 1260 isocratic HPLC pump, an Agilent G1603A CE-MS adapter kit, and an Agilent G1607A CE-ESI-MS sprayer kit (Agilent Technologies, Inc., Santa Clara, CA, USA). Instrument control was handled by MassHunter Workstation Data Acquisition (Agilent Technologies) and Xcalibur (Thermo Fisher Scientific). A fused silica capillary (50 µm i.d. × 80 cm total length) was used with commercial electrophoresis buffer (H3301-1001 for cations and I3302-1023 for anions; HMT) as the electrolyte. The spectrometer was scanned from m/z 60 to 900 in positive mode, and from m/z 70 to 1050 in negative mode.³⁰ Peak detection and integration were performed with MasterHands (Keio University, Tsuruoka, Yamagata, Japan), yielding m/z, peak area, and migration time (MT).³¹ Signal peaks corresponding to isotopomers, adduct ions, or other product ions of known metabolites were removed prior to annotation. The remaining peaks were annotated against HMT's metabolite database based on their m/z values and MTs. If multiple peaks were assigned to the same candidate metabolite, a branch index was appended to each candidate's name. The areas of annotated peaks were then normalized to the internal standards and adjusted for sample volume to obtain the relative levels of each metabolite. A total of 110 metabolites were absolutely quantified based on 1-point calibrations using their respective standard compounds.

Statistical Analyses

Data filtering, imputation, and normalization were performed using the R package MetaboAnalystR (version 4.0.0).³² Specifically, features with >50% missing values were removed, missing values were imputed using the minimum value method, and data were normalized by sum normalization, log transformation, and auto-scaling. Principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were also conducted using MetaboAnalystR. Differentially abundant metabolites were identified using the R software package limma (version 3.54.2),³³ with age, sex, intraocular pressure (IOP), and axial length included as covariates. All statistical comparisons were performed on the same set of detected metabolites across all samples, without applying comparison-specific feature selection. Multiple testing correction was performed using the Benjamini–Hochberg procedure to control the false discovery rate (FDR). Metabolites with Benjamini–Hochberg-adjusted P values < 0.05 were considered significantly differentially abundant. For comparisons of absolute metabolite concentrations between groups, zero values were treated as below the limit of detection and replaced with half of the minimum non-zero value for each metabolite prior to log transformation, followed by Welch's t-test. Finally, pathway enrichment analysis was performed on the differentially abundant metabolites using MetaboAnalystR with SMPDB as a metabolite set library.^34,35 Here, peaks with multiple metabolite annotations were excluded from the analysis. Pathways with nominal P values < 0.05 and containing more than one matched metabolite were visualized in the plots.

Results

Patient Characteristics and Global Metabolic Overview

The study included a total of 38 cases: 8 eyes with grade A (RRD), 10 eyes with grade B (RRD), 10 eyes with grade C (PVR), and 10 eyes with IOL dislocation that underwent IOL scleral fixation, which served as the control group (Fig. 1A). Detailed characteristics of the subjects are provided in Table 1 and Supplementary Table S1. The proportion of macula-off detachment and the duration of retinal detachment were comparable between grade A and grade B RRD. Anterior chamber cells were infrequent in controls and RRD groups, whereas they were more commonly observed in PVR. There were no marked differences in diabetes mellitus or renal dysfunction across groups.

Figure 1.

Study design and multivariate overview of vitreous metabolomics. (A) Study workflow. Vitreous samples were collected from 38 eyes across 4 groups: control (IOL dislocation undergoing scleral fixation), n = 10; grade A (RRD), n = 8; grade B (RRD), n = 10; and grade C (PVR), n = 10. Representative fundus photographs are shown below each schematic illustration; wide-field images are presented for the control and grade C eyes, whereas magnified views highlighting representative retinal breaks are shown for grade A and grade B eyes. The eye cartoons were adapted from an Adobe Stock image, for which permission for use and modification was obtained. (B) Principal component analysis (PCA) scores plot. This unsupervised analysis shows each eye’s metabolic profile. The first two principal components (PC1 and PC2) are orthogonal linear combinations of metabolites that capture the largest sources of variance in the data. Ellipses indicate the 95% confidence interval regions for each group. (C) Partial least squares-discriminant analysis (PLS-DA) scores plot. This supervised analysis emphasizes class separation. The components are latent variables that maximize covariance between metabolite profiles and group labels, thereby achieving this separation. Ellipses show 95% confidence interval regions for each group.

Table 1.

Baseline Demographic and Clinical Characteristics of the Study Eyes

	Control	Grade A (RRD)	Grade B (RRD)	Grade C (PVR)
Number of eyes	10	8	10	10
Age	72.8 ± 15.1	54.6 ± 12.9	64.4 ± 8.5	61.3 ± 17.9
Sex, M/F	7/3	7/1	8/2	7/3
Axial length, mm	24.4 ± 1.6	27.0 ± 2.3	25.2 ± 1.0	25.0 ± 2.3
IOP, mm Hg	15.0 ± 4.8	11.4 ± 2.6	11.1 ± 5.2	8.5 ± 5.1
Macular detachment, on/off	–	1/7	1/9	0/10
Anterior chamber cells, present/absent	0/10	0/8	1/9	6/3 (1 N/A)
Diabetes mellitus, present/absent	1/9	0/8	4/6	1/9
Renal dysfunction, present/absent	4/6	2/6	3/7	3/7
Duration of retinal detachment, d	–	8.6 ± 2.5	11.2 ± 7.7	256.1 ± 196.5

IOP, intraocular pressure; mm Hg, millimeters of mercury.

Data are presented as mean ± standard deviation. Presence of anterior chamber cells was evaluated preoperatively by slit-lamp microscopy examination. N/A indicates absence of documented assessment. Renal dysfunction was defined as an estimated glomerular filtration rate (eGFR) < 60 mL/min/1.73 m².

In total, 534 metabolite peaks were detected. To obtain an overview of the global metabolic trends, we first performed PCA and PLS-DA. PCA revealed that the grade C (PVR) group exhibited a distinct metabolic profile compared with the other groups (Fig. 1B). To further clarify intergroup differences, we conducted PLS-DA. Because no substantial differences were observed between grade A and grade B (Fig. 1C), these two groups were combined into a single RRD group for all subsequent analyses.

RRD Versus Control: Minimal Metabolic Changes Without a Clear Disease-Specific Signature

Comparison between the RRD and control groups revealed only minimal metabolic alterations. After Benjamini–Hochberg correction, no metabolite reached statistical significance (Fig. 2A). The top 10 upregulated and the top 10 downregulated peaks, along with their annotated metabolites, are summarized in Table 2. Although these metabolites did not remain significant after multiple testing correction, N²-methylguanine and glucose 1-phosphate showed the most prominent increases in the RRD group. Absolute quantification of glucose 1-phosphate and betaine showed trends consistent with relative quantification (Fig. 2B), supporting the robustness of these directional changes. The most downregulated peaks were annotated as phenylephrine and synephrine, which are not endogenous metabolites. Phenylephrine is presumed to derive from preoperative mydriatic drops. Its higher vitreous levels in the control group undergoing IOL scleral fixation may reflect zonular disruption, reducing the posterior capsule's barrier function, and permitting greater translocation of topical agents into the vitreous than in RRD eyes. Overall, RRD did not show significant metabolic changes after multiple testing correction, and observed differences consisted mainly of non-endogenous changes.

Figure 2.

Metabolomic comparison between RRD and control groups. (A) Volcano plot of metabolic changes. Each point represents a detected metabolite peak. Axes = x, log₂ fold change; y, statistical significance (−log₁₀ adjusted P value). Points are color-coded by significance; however, no peaks reached statistical significance in this comparison, and all points are shown in gray. The vertical dotted line indicates log₂ fold change = 0, and the horizontal dotted line indicates adjusted P = 0.05. (B) Absolute quantification of markedly altered metabolites. Log₁₀-transformed concentrations of glucose 1-phosphate and betaine are shown for each group. Statistical analysis was performed using Welch's t-test (ns, not significant; *P < 0.05, ***P < 0.001). Centerlines within box plots represent medians. Box limits indicate 25th and 75th percentiles, and the whiskers extend to 1.5 times interquartile range (IQR) of 25th and 75th percentiles.

Table 2.

Top 10 Upregulated and Top 10 Downregulated Peaks in RRD Relative to the Control Group.

Direction	Compound Name	logFC	P Value	Adjusted P Value
Upregulated	N2-Methylguanine	1.52	4.3 × 10−4	0.085
	Glucose 1-phosphate	1.51	4.9 × 10−4	0.085
	Glucuronolactone	1.34	0.0019	0.219
	3-Methoxy-4-hydroxyphenylethyleneglycol	1.28	0.0031	0.219
	N-Formylmethionine	1.22	0.0048	0.219
	O-Succinylhomoserine	1.21	0.0049	0.219
	N2,N2-Dimethylguanosine	1.21	0.0052	0.219
	3,4-Dihydroxyphenylglycol	1.20	0.0056	0.219
	Ethanolamine	1.15	0.0080	0.224
	N1-Methylguanosine	1.14	0.0081	0.224
Downregulated	Phenylephrine synephrine	−1.19	0.0057	0.219
	Betaine	−1.17	0.0069	0.224
	N,N-Dimethylglycine	−1.09	0.0115	0.231
	Ectoine	−1.09	0.0116	0.231
	Pipecolic acid	−1.08	0.0124	0.231
	S-Methylmethionine	−1.08	0.0126	0.231
	Imidazole-4-acetic acid	−1.05	0.0154	0.259
	3-Aminoisobutyric acid	−0.97	0.0249	0.284
	Acetoacetamide	−0.96	0.0255	0.284
	Cystine	−0.96	0.0260	0.284

The top 10 upregulated and top 10 downregulated peaks are shown with their annotated metabolites, log₂ fold changes (logFC), P values, and adjusted P values.

PVR Versus Control: Broad Metabolic Reprogramming Supporting a Proliferative State

Comparison of the PVR and control groups revealed marked metabolic differences. Overall, 54 peaks changed significantly, with 47 increased and 7 decreased in the PVR group (Fig. 3A). The top 10 upregulated and the top 10 downregulated peaks, along with their annotated metabolites, are summarized in Table 3. Among them, aspartate, glutamate, and glycolic acid were absolutely quantified and showed trends consistent with relative quantification (Fig. 3B). Metabolites increased in PVR included guanosine, N²,N²-dimethylguanosine, γ-Glu-Trp, aspartate, and glutamate. In addition, 2-oxoisovaleric acid and ethanolamine phosphate were elevated. In contrast, decreased metabolites included histamine metabolites, such as methylhistamine, taurine-related compounds including γ-Glu-taurine, and N-acetyltaurine, as well as nucleotide-related metabolites, such as 2′-O-methylcytidine. Decreases were also observed in organic acids such as isethionic acid and threonic acid.

Figure 3.

Metabolomic comparison between PVR and control groups. (A) Volcano plot of metabolic changes. Each point represents a detected metabolite peak. Axes = x, log₂ fold change; y, statistical significance (−log₁₀ adjusted P value). Points are color-coded by significance: red = peaks significantly increased in PVR; blue = peaks significantly decreased in PVR; and gray = peaks without significant change. The vertical dotted line indicates log₂ fold change = 0, and the horizontal dotted line indicates adjusted P = 0.05. (B) Absolute quantification of markedly altered metabolites. Log₁₀-transformed concentrations of aspartate (Asp), glutamate (Glu), and glycolic acid are shown for each group. Statistical analysis was performed using Welch's t-test (ns, not significant; **P < 0.01, ***P < 0.001, ****P < 0.0001). Centerlines within box plots represent medians. Box limits indicate 25th and 75th percentiles, and the whiskers extend to 1.5 times interquartile range (IQR) of 25th and 75th percentiles. (C) Pathway enrichment analysis of metabolites significantly altered in PVR versus the control group. The dot size corresponds to the enrichment ratio, and the x-axis indicates statistical significance (−log₁₀ P value). The pathway remaining significant after Benjamini–Hochberg correction (FDR < 0.05), purine metabolism, is highlighted in bold.

Table 3.

Top 10 Upregulated and Top 10 Downregulated Peaks in PVR Relative to the Control Group

Direction	Compound Name	logFC	P Value	Adjusted P Value
Upregulated	Guanosine	2.19	1.3 × 10−5	0.005
	2-Oxoisovaleric acid	2.00	6.9 × 10−5	0.012
	N2,N2-Dimethylguanosine	1.87	2.1 × 10−4	0.019
	γ-Glu-Trp	1.83	2.7 × 10−4	0.019
	Asp	1.80	3.4 × 10−4	0.019
	Ethanolamine phosphate	1.75	5.0 × 10−4	0.019
	4-Methyl-2-oxovaleric acid	1.74	5.3 × 10−4	0.019
	3-Methyl-2-oxovaleric acid
	2-Oxohexanoic acid
	Glucose 6-phosphate	1.74	5.5 × 10−4	0.019
	Glu	1.71	6.7 × 10−4	0.021
	3-(2-Hydroxyphenyl)propionic acid	1.70	7.4 × 10−4	0.021
	m-Ethoxybenzoic acid
	3-Phenyllactic acid
	p-Methoxyphenylacetic acid
Downregulated	Argininosuccinic acid	−1.85	2.3 × 10−4	0.019
	1-Methylhistamine	−1.75	5.2 × 10−4	0.019
	3-Methylhistamine
	Isethionic acid	−1.52	0.0025	0.028
	γ-Glu-Taurine	−1.44	0.0043	0.036
	N-Acetyltaurine	−1.43	0.0045	0.037
	2′-O-Methylcytidine	−1.42	0.0047	0.037
	Threonic acid	−1.38	0.0060	0.042
	Glycolic acid	−1.33	0.0080	0.051
	Pelargonic acid	−1.30	0.0098	0.058
	Ascorbic acid	−1.27	0.0114	0.066

The top 10 upregulated and top 10 downregulated peaks are shown with their annotated metabolites, log₂ fold changes (logFC), P values, and adjusted P values.

Pathway enrichment analysis identified purine metabolism as significantly upregulated in the PVR group after multiple testing correction (Fig. 3C). In addition, several pathways related to amino acid and nitrogen metabolism—including the urea cycle, ammonia recycling, arginine and proline metabolism, and glycine and serine metabolism—were consistently enriched at the nominal level.

PVR Versus RRD: Identification of a Distinct Progression-Associated Metabolic Signature

Direct comparison between the PVR and RRD groups revealed the greatest metabolic divergence among all group comparisons, supporting the presence of a distinct metabolic signature associated with PVR. Overall, 110 peaks changed significantly, with 74 increased and 36 decreased in PVR relative to RRD (Fig. 4A). The top 10 upregulated and the top 10 downregulated peaks, along with their annotated metabolites, are summarized in Table 4. Among these, inosine and cytidine were absolutely quantified and showed trends consistent with relative quantification (Fig. 4B). Metabolites increased in the PVR group included branched-chain amino acid derivatives, such as 2-oxoisovaleric acid, 4-methyl-2-oxovaleric acid, and 3-methyl-2-oxovaleric acid, as well as fatty acid-related metabolites including hydroxyoctanoic acid. In addition, aromatic metabolites such as 3-phenyllactic acid, the nucleoside inosine, the amino acids glutamate and aspartate, and the carnitine precursor γ-butyrobetaine were elevated. Conversely, decreased metabolites included fatty acid-related metabolites, such as decanoic acid and pelargonic acid, histamine-related metabolites including methylhistamine, and nucleic acid-related metabolites such as guanine and cytidine. Notably, 30 metabolites were commonly upregulated in both the PVR versus control and PVR versus RRD comparisons. In addition, all seven metabolites that were decreased in the PVR versus control comparison were also consistently decreased in the PVR versus RRD comparison.

Figure 4.

Metabolomic comparison between PVR and RRD groups. (A) Volcano plot of metabolic changes. Each point represents a detected metabolite peak. Axes = x, log₂ fold change; y, statistical significance (−log₁₀ adjusted P value). Points are color-coded by significance: red = peaks significantly increased in PVR; blue = peaks significantly decreased in PVR; and gray = peaks without significant change. The vertical dotted line indicates log₂ fold change = 0, and the horizontal dotted line indicates adjusted P = 0.05. (B) Absolute quantification of markedly altered metabolites. Log₁₀-transformed concentrations of inosine and cytidine are shown for each group. Statistical analysis was performed using Welch's t-test (ns, not significant; **P < 0.01, ***P < 0.001, ****P < 0.0001). Centerlines within box plots represent medians. Box limits indicate 25th and 75th percentiles, and the whiskers extend to 1.5 times interquartile range (IQR) of 25th and 75th percentiles. (C) Pathway enrichment analysis of metabolites significantly altered in PVR versus the RRD group. The dot size corresponds to the enrichment ratio, and the x-axis indicates statistical significance (−log₁₀ P value). Pathways remaining significant after Benjamini–Hochberg correction (FDR < 0.05) are highlighted in bold.

Table 4.

Top 10 Upregulated and Top 10 Downregulated Peaks in PVR Relative to the RRD Group

Direction	Compound Name	logFC	P Value	Adjusted P Value
Upregulated	3-(2-Hydroxyphenyl)propionic acid	1.70	8.4 × 10−6	7.3 × 10−4
	m-Ethoxybenzoic acid
	3-Phenyllactic acid
	p-Methoxyphenylacetic acid
	Inosine	1.67	1.3 × 10−5	8.8 × 10−4
	Hypotaurine	1.65	1.6 × 10−5	9.3 × 10−4
	2-Oxoisovaleric acid	1.61	2.6 × 10−5	0.0011
	Glu	1.60	2.9 × 10−5	0.0011
	Indole-3-lactic acid-2	1.57	4.1 × 10−5	0.0013
	5-Methoxyindoleacetic acid
	4-Methyl-2-oxovaleric acid	1.56	4.3 × 10−5	0.0013
	3-Methyl-2-oxovaleric acid
	2-Oxohexanoic acid
	γ-Butyrobetaine	1.53	6.3 × 10−5	0.0017
	Asp	1.50	9.1 × 10−5	0.0020
	8-Hydroxyoctanoic acid	1.49	9.9 × 10−5	0.0020
	3-Hydroxyoctanoic acid-1
Downregulated	Decanoic acid	−1.84	1.6 × 10−6	3.4 × 10−4
	Pelargonic acid	−1.82	1.9 × 10−6	3.4 × 10−4
	Phloroglucinol	−1.75	5.0 × 10−6	5.8 × 10−4
	1-Methylhistamine	−1.62	2.4 × 10−5	0.0011
	3-Methylhistamine
	Guanine	−1.56	4.5 × 10−5	0.0013
	Threonic acid	−1.52	6.9 × 10−5	0.0017
	Isethionic acid	−1.49	1.0 × 10−4	0.0020
	3-Methylcytidine	−1.48	1.1 × 10−4	0.0020
	Cytidine	−1.46	1.3 × 10−4	0.0022
	5-Methylcytidine	−1.44	1.7 × 10−4	0.0025

The top 10 upregulated and top 10 downregulated peaks are shown with their annotated metabolites, log₂ fold changes (logFC), P values, and adjusted P values.

Pathway enrichment analysis revealed six pathways associated with amino acid and nitrogen metabolism that were significantly upregulated in PVR compared with RRD (Fig. 4C), including arginine and proline metabolism, glycine and serine metabolism, the urea cycle, ammonia recycling, alanine metabolism, and the malate–aspartate shuttle. In addition, several pathways—including aspartate metabolism, carnitine synthesis, the glucose–alanine cycle, glutamate metabolism, beta-alanine metabolism, glutathione metabolism, and the Warburg effect—were enriched at the nominal level. By contrast, pathways related to fatty acid biosynthesis, beta-oxidation of very-long-chain fatty acids, and spermidine and spermine biosynthesis were downregulated at the nominal level.

Discussion

This study demonstrates that vitreous metabolomic profiling reveals a metabolic signature in PVR that is distinct from both RRD and control samples, whereas RRD showed only subtle changes relative to controls after multiple testing correction. In PVR, pathway enrichment analyses revealed prominent activation of amino acid, nitrogen, and nucleotide-related metabolic pathways, consistent with increased biosynthetic demand. Pathways related to fibrosis and stress responses were also enriched. Table 5 summarizes the altered pathways identified in PVR and their putative associated biological processes relevant to PVR pathophysiology.

Table 5.

Summary of Major Upregulated and Downregulated Pathways in PVR and Their Putative Biological Relevance

Direction	Pathway Name	Biological Function
Upregulated	Purine metabolism	P, F
	Urea cycle	P
	Ammonia recycling	P
	Arginine and proline metabolism	W, F
	Alanine metabolism	P
	Glycine and serine metabolism	P
	Carnitine synthesis	S
	Tyrosine metabolism	F
	Glutamate metabolism	P
	Malate–aspartate shuttle	E, S
	Glucose–alanine cycle	E
	Amino sugar metabolism	F
	Phosphatidylcholine biosynthesis	P, F
	Aspartate metabolism	P
	Beta-alanine metabolism	S
	Glutathione metabolism	S
	Warburg effect	P, E
	Tryptophan metabolism	S
	Methionine metabolism	S
	Cysteine metabolism	S
	Oxidation of branched chain fatty acids	S
Downregulated	Fatty acid biosynthesis	P, F
	Beta oxidation of very long chain fatty acids	E
	Spermidine and spermine biosynthesis	F

E, energy demand; F, fibrosis; P, proliferation; S, stress responses; W, wound healing.

Compared with controls, RRD showed no clear disease-specific metabolic signature. Any residual differences were subtle and may partly reflect non-endogenous drug-related peaks. Thus, RRD does not substantially alter the vitreous metabolome. RRD is typically an acute condition without sustained fibrocellular remodeling, which may explain the minimal metabolic alterations.

In contrast, the PVR versus control comparison showed broad metabolic alterations, allowing pathway-level analysis (see Fig. 3C). First, we observed upregulation of pathways associated with proliferative biosynthesis, including nucleotide synthesis, amino acid metabolism, and nitrogen utilization. Purine metabolism was significantly enriched, supporting nucleotide turnover and proliferative demand.³⁶ The urea cycle, ammonia recycling, alanine metabolism, glycine and serine metabolism, glutamate metabolism, and aspartate metabolism were concurrently upregulated, indicating increased nitrogen utilization required for biosynthetic processes.^37–39 Upregulation of the malate–aspartate shuttle supports oxidative glucose metabolism and cytosolic–mitochondrial redox balance, which protects retinal glutamate from oxidation.^40–45 Elevation of the glucose–alanine cycle is consistent with enhanced glycolysis associated with RPE EMT in PVR.⁴⁶ Second, pathways associated with fibrosis and extracellular matrix remodeling were enriched. Arginine and proline metabolism, required for collagen synthesis during wound healing and fibrosis,⁴⁷ was upregulated. Tyrosine metabolism was increased, which may relate to fibrotic activation via dopamine released from RPE cells under TGF-β stimulation.^48,49 In addition, amino sugar metabolism and phosphatidylcholine biosynthesis were elevated, supporting hyaluronan production and membrane biogenesis relevant to fibrosis and cellular proliferation.^50,51 Third, upregulation of carnitine metabolism suggested a metabolic stress response. Li et al. reported reduced carnitine in RRD and PVR, interpreting this as reduced anti-inflammatory capacity that may amplify inflammation.^27,52,53 In contrast, Yu et al. observed increased carnitine in RRD with choroidal detachment, suggesting a compensatory response to inflammation.²⁸ In our study, carnitine levels and carnitine synthesis were increased in PVR, consistent with the interpretation that carnitine elevation may reflect a compensatory response to inflammatory stress. Collectively, these pathway-level alterations indicate increased nucleotide turnover, energy demand, nitrogen utilization, and fibrotic remodeling in PVR, consistent with a proliferative and fibrotic disease state.

In the direct PVR versus RRD comparison, pathways elevated versus controls were generally higher again (see Fig. 4C), indicating further gains in proliferation, wound healing, fibrosis, energy demand, and stress responses. Additional pathway alterations not observed in the PVR versus control comparison also emerged. The Warburg effect, a hallmark of rapid proliferation, was upregulated, suggesting increased energy demand and enhanced proliferative activity.⁵⁴ Furthermore, β-alanine metabolism, glutathione metabolism, tryptophan metabolism, methionine metabolism, cysteine metabolism, and oxidation of branched-chain fatty acids were upregulated, consistent with heightened oxidative and metabolic stress responses.^55–59 Downregulated pathways further supported these interpretations, particularly those involved in lipid metabolism and polyamine synthesis. Fatty acid biosynthesis, which contributes to membrane lipid synthesis during proliferation,⁶⁰ was reduced. Perturbation of this pathway during TGF-β2–induced EMT in RPE cells has been reported,⁶¹ supporting its relevance to PVR pathogenesis. Reduced β-oxidation of very-long-chain fatty acids, which contributes to energy production,⁶² may be consistent with increased energy demand. In addition, Nicoletti et al. reported reduced putrescine and elevated spermidine in the vitreous humor of eyes with PVR,⁶³ and we similarly observed decreased putrescine and increased acetylspermidine, an acetylated derivative of spermidine, in PVR. Polyamines are known to participate in fibroblast activation and extracellular matrix remodeling,^64,65 and reduced spermidine levels have been reported in idiopathic pulmonary fibrosis.⁶⁶ In this context, excessive utilization of putrescine to support polyamine flux during fibrotic remodeling may reduce de novo polyamine synthesis, resulting in downregulation of the spermidine and spermine biosynthesis pathway. Taken together, these findings indicate that progression from RRD to PVR is accompanied by heightened oxidative and metabolic stress, along with relative suppression of polyamine and lipid biosynthetic pathways associated with fibrotic remodeling and altered lipid homeostasis.

EMT of RPE is a key mechanism underlying retinal fibrotic diseases, including PVR and subretinal fibrosis in AMD.⁶⁷ Our findings suggest that suppressing EMT may represent a rational pharmacologic strategy for PVR. In vitro, nintedanib suppresses glycolysis in TGF-β2–induced mesenchymal-like cells derived from RPE, with metabolic reprogramming initiating mesenchymal–epithelial transition.⁶⁸ Similarly, ZLN005, a small-molecule activator of the PGC-1α gene—a key regulator of mitochondrial biogenesis and metabolic function in RPE—can suppress TGF-β2–induced EMT and glycolytic upregulation.⁴⁶ Given that EMT-related metabolic pathways were upregulated in PVR, these agents may offer a potential therapeutic strategy.

This study has several limitations. First, the sample size is relatively small, and an independent validation cohort was not included. However, the acquisition of vitreous samples from patients with PVR is challenging, because PVR develops in only 5.1% to 11.7% of RRD cases,⁶⁹ limiting surgical availability during the study period. Second, the choice of control subjects requires careful consideration. We used patients undergoing IOL scleral fixation as controls because they required vitrectomy but did not have retinal disease or apparent intraocular inflammation expected to affect the vitreous metabolome. They do not represent a perfectly “healthy” reference, because underlying conditions (e.g., prior trauma or pseudoexfoliation) may introduce metabolic variability. Importantly, however, no metabolites differed significantly between the RRD and control groups after multiple testing correction, suggesting that such conditions were unlikely to affect major vitreous metabolic profiles in our dataset. Future studies with more rigorous control selection would be beneficial. Third, disease duration differed substantially between the RRD and PVR groups, with PVR showing a markedly longer interval from onset to surgery. Prolonged disease duration inherently involves chronic inflammation, sustained retinal ischemia, and long-term cellular stress, all of which could cause metabolic alterations. Therefore, it is challenging to distinguish the metabolic changes caused by the fibroproliferative PVR process itself from those caused by chronicity. Although we identified a distinct metabolic signature in PVR, longitudinal studies following patients with RRD who progress to PVR are crucial to define shifts associated with this transition. Finally, our metabolomic profiling captures only a single time-point snapshot. Whereas we identified strong associations between specific pathways and PVR, these cross-sectional data cannot establish causality. Further studies incorporating functional validation at the cellular level and interventional evaluations in animal models are warranted to clarify causality and assess the therapeutic potential of targeting these pathways.

Conclusions

This study shows that PVR exhibits a unique metabolic profile distinct from RRD and controls, with upregulation of pathways related to proliferation, wound healing, fibrosis, energy demand, and stress responses. These metabolic alterations may reflect key biological processes involved in PVR development. Taken together, our findings provide a metabolic framework for understanding PVR pathogenesis and suggest that these pathways may serve as potential therapeutic targets to restrain fibrosis and aberrant cellular proliferation. Further studies are warranted to functionally validate these findings and to develop novel treatments.