Elevated Very-Long-Chain Ceramides in the Vitreous Humor of Patients With Proliferative Diabetic Retinopathy

Abstract

Purpose

To quantitate the levels of various ceramide species in the vitreous of patients with proliferative diabetic retinopathy (PDR) and to investigate the role of vitreal ceramides in the pathogenesis of PDR.

Study Design

A case control study.

Methods

We collected vitreous samples from 25 type 2 diabetes patients with PDR and 25 age- and sex-matched nondiabetic controls undergoing vitrectomy. The levels of ceramide species (C16:0, 18:0, 20:0, 22:0, 24:1, and 24:0) were measured by ultra-high-performance liquid chromatography-tandem mass spectrometry with positive electrospray ionization mode. The correlation of baseline characteristics, blood test data, and clinical manifestation of PDR were analyzed with vitreal ceramides levels.

Results

The total level of ceramides was substantially higher in the PDR group than the control group (18.626 ± 19.264 versus 3.524 ± 2.456 pmol/mg protein; P < 0.001). Among ceramides of various acyl chain lengths, the increases of very-long-chain (VLC) ceramides (C22–C24) were more drastic than those of long-chain ceramides (C16–C20). In the PDR group, VLC ceramide species accounted for 76.1%, whereas in the control group, C16 ceramide predominated at 40.5%. Based on the multivariate linear regression analysis, diagnosis of diabetes (β = 14.5751; P = 0.0327) and lower body mass index (β = −2.1396; P = 0.0173) were significantly associated with higher level of VLC ceramides. Intravitreal injection of anti-VEGF leads to insignificant reduction of VLC ceramides (P = 0.068).

Conclusions

Vitreal ceramide levels were elevated in diabetic subjects, especially the VLC species, which may contribute to the pathogenesis of diabetic retinopathy.

Proliferative diabetic retinopathy (PDR) is a severe microvascular complication of type 2 diabetes (T2DM) and is the fifth global leading cause of blindness in people over 50 years of age in 2020.¹ A meta-analysis estimated the global prevalence of diabetic retinopathy at 22.27% and vision-threatening retinopathy at 6.17%.² The affected population is expected to rise with increasing prevalence of diabetes.^2–4 PDR pathogenesis involves chronic hyperglycemia, inflammation, oxidative stress, pathological angiogenesis, and blood–retinal barrier dysfunction, as well as genetic and epigenetic factors.⁵ Although good glycemic control decreases the risk for the development and progression of PDR, it is clear that aberrant lipid metabolism is also a risk factor.

Various clinical trials have demonstrated a clear association between the levels of circulating lipids and PDR,^6,7 and suggest that targeting dyslipidemia may decrease VEGF levels, decrease exudate formation, and prevent vision loss in PDR patients.^8,9 Among lipids that are elevated under diabetic conditions, ceramide is a sphingolipid that plays a crucial role in cell signaling pathways. Increased ceramide not only contributes to the development of insulin resistance,¹⁰ but has also been associated with the onset and progression of T2DM as well as diabetic complications.¹¹ In vitro studies and animal models have shown that increased ceramide causes retinal cell death, oxidative stress, and inflammation, all hallmarks of PDR.^12–16 Also, ceramide has been shown to impair the function of the endothelial cells in the retinal blood vessels, leading to increased vascular permeability and leakage, which is a major contributor to diabetic macular edema. Increasing evidence reveals that ceramides consist of various lengths of the fatty acyl side chains add show different biophysical property and bioactivity. For example, ultra-long-chain ceramides (>C26) are essential for skin barrier,¹⁷ whereas ceramides of C16 to C20 are proinflammatory and proapoptosis.¹⁸ Increased circulating ceramides have also been linked to (macrovascular) cardiovascular diseases, and C16:0, C18:0, and C24:1 ceramides can predict plaque instability and/or future fatality.^19,20 However, reports of ceramide levels in the context of clinical PDR are scarce.^21,22 The levels of various ceramide species and their mechanistic implications in PDR remains to be determined.

In addition, pathological vitreous has been shown to alter the characteristics of retinal pigment epithelial cells including viability, proliferation, and mesenchymal features,²³ suggesting vitreous contents participate in PDR pathogenesis. Here, we compared the levels of ceramide species of various acyl chain lengths in the vitreous of PDR patients with that of age- and sex-matched nondiabetic patients to provide a comprehensive view of the local concentration of ceramide that is in direct contact with the inner retina. Understanding the local level of ceramide in the vitreous could provide new insights into the pathogenesis of PDR and potentially lead to the development of novel therapies.

Methods

This study followed the tenets of the Declaration of Helsinki, and the protocol was approved by the Ethics Review Board of the National Taiwan University Hospital (REC ID: 202006201RINA). All of the participants provided written informed consent after receiving an explanation of the nature and possible consequences of the study.

Subjects and Sample Collection

This case-control study at National Taiwan University Hospital (March to December 2021) recruited consecutive patients (>20 years) who required vitrectomy. The study group included those with PDR, while controls were age- and sex-matched subjects who required vitrectomy owing to conditions irrelevant to diabetes, such as idiopathic macular pucker, macular hole, or myopic tractional maculopathy. The age difference between participants in the study and control groups was limited to 2 years or less. Patients who meet the following criteria were excluded: type 1 diabetes, prior intraocular surgery in the same eye, familial dyslipidemia, diseases affecting lipid metabolism, use of hormones, immunosuppressive therapy, drugs affecting sphingolipids, and pregnancy. Retina specialists (C.-M.Y., T.-C.H., Y.-T.H., and T.-T.L.) performed the vitrectomies, collecting up to 500 µL of undiluted vitreous humor per eye with 25G trocars after retrobulbar anesthesia. Samples were preserved at −80°C until analysis. No T2DM patients without DR or T2DM patients with nonproliferative DR were included in the present study because we did not encounter such patients requiring vitrectomy during the study period.

Collection of Baseline Characteristics

All participants underwent comprehensive ophthalmologic examinations before vitrectomy, including best-corrected visual acuity , intraocular pressure, slit-lamp biomicroscopy, dilated fundus exam, and optical coherence tomography (OCT) imaging. Visual acuity was measured with a Snellen chart and expressed as logMAR for further analysis. OCT was performed using RTVue XR Avanti (Optovue, Fremont, CA), RTVue RT100 (Optovue), or Cirrus HD-OCT (Carl Zeiss Meditec, Dublin, CA). Diagnoses and previous treatment histories including intravitreal injection (IVI) of anti-VEGF and panretinal photocoagulation (PRP) were recorded. No patient received steroid treatment in the present study. Baseline characteristics including age, gender, body mass index (BMI), diabetes duration, diastolic/systolic blood pressure, medications, and comorbidities were documented. The data of blood test measuring glycosylated hemoglobin (HbA1c), blood glucose, total cholesterol, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, and triglycerides, as well as liver and kidney function were collected.

Extraction and Analysis of Sphingolipids

The levels of ceramide species were measured by ultra-high-performance liquid chromatography-tandem mass spectrometry (UHPLC-MS/MS) at the Metabolomics Core Laboratory of National Taiwan University Centers of Genomic and Precision Medicine. All ceramide standards were purchased from Avanti Polar Lipids (Alabaster, AL, USA). Chloroform and acetonitrile were purchased from J.T. Baker (Avantor, Inc., Radnor, PA, USA). Formic acid and ammonium acetate were obtained from Sigma-Aldrich (St. Louis, MO, USA). Isopropanol and methanol were acquired from Merck KGaA (Darmstadt, Germany) and Scharlab, S.L. (Barcelona, Spain), respectively.

Sample processing and analysis were performed by the Metabolomics Core Laboratory of National Taiwan University Center of Genomic and Precision Medicine. Ceramides were extracted by the modified Folch method.²⁴ In brief, 100 µL of vitreous humor was combined with 1 µL of isotope-labeled Deuterated Ceramide LIPIDOMIX Mass Spec Standard (Avanti Research, cat# 330713), which contains 21.8 µg/mL C16:0 ceramide-d7 (d18:1-d7/16:0), 11.5 µg/mL C18:0 ceramide-d7 (d18:1-d7/18:0), 13.1 µg/mL C24:1 ceramide-d7 (d18:1-d7/24:1(15Z)), and 26.3 µg/mL C24:0 ceramide-d7 (d18:1-d7/24:0). After vortexing, the mixture was extracted with 500 µL of a mixture of chloroform:methanol (2:1, v/v) and homogenized at 1000 rpm for 5 minutes by a Geno/Grinder 2010 (SPEX SamplePrep., Metuchen, NJ, USA). Next, the sample was centrifuged by using Eppendorf Centrifuge 5810R at 15,000× g for 5 minutes at 4 °C, and the lower layer was transferred to a new tube. This procedure was repeated, and the combined phases were dried by nitrogen stream. The extract was reconstituted with 100 µL of methanol. After sonication for 15 minutes and centrifugation at 15,000× g for 5 minutes at 4 °C, the supernatant was filtered with 0.2 µm Ministart RC 4 filter (Sartorius, Goettingen, Germany) and subjected to UHPLC-MS/MS analysis.

Ceramides were analyzed by Agilent 1290 UHPLC coupled with an Agilent 6460 triple quadrupole mass spectrometer (Agilent Technologies, Santa Clara, CA) in multiple reaction monitoring mode. The separation was performed using an Agilent ZORBAX Eclipse Plus C18 column (2.1 × 100 mm, 1.8 µm, Agilent Technologies, Santa Clara, CA, USA), which was thermostated at 55°C. The mobile phase was composed of solvent A (10 mM ammonium acetate and 0.2% formic acid in methanol:water = 6:4, v/v) and solvent B (10 mM ammonium acetate and 0.2% formic acid in methanol:isopropanol = 6:4, v/v). A linear gradient elution was used: 0 to 2 minutes, 35% to 80% solvent B, 2 to 7 minutes, 80% to 100% solvent B, 7 to 14 minutes, 100% solvent B, and column re-equilibration with 35% solvent B for 2 minutes. The flow rate was 0.35 mL min⁻¹. The injection volume was 5 µL. The positive electrospray ionization mode was used with the following parameters: 325 °C for drying gas temperature, 7 L min⁻¹ for drying gas flow, 45 psi for nebulizer pressure, 325 °C for sheath gas temperature, 11 L min⁻¹ for sheath gas flow rate, and 3500 V for capillary voltage. Nozzle voltage was set at 500 V. All the peaks were integrated with MassHunter Quantitative Analysis software (Agilent Technologies). Concentrations of each ceramide in vitreous humor samples were determined from calibration curves using the peak area ratio of the analyte to its corresponding isotope-labeled internal standard. A ceramide standards mixture QC sample was designed and inserted to be analyzed between every 10 real samples during UHPLC-MS/MS analysis. The relative standard deviation of ceramides concentrations in QC sample were less than 3%.

Statistics

Categorical data were expressed as numbers and percentages, and continuous data were presented as either the mean ± standard deviation or the median with interquartile range depending on the distribution of the data. Laboratory data, ocular characteristics, and ceramide levels between the two groups were analyzed using paired-samples t-test, and categorical variables were compared with Fisher's exact test. Linear regression analyses were performed for the association of sphingolipids level in the vitreous with systemic characteristics. The Kruskal–Wallis test and the Mann–Whitney U test was used to determine statistically significant differences between the PDR subgroups with small samples. P values of <0.05 were considered statistically significant. All statistical analyses were performed using SPSS version 22 (SPSS, Chicago, IL, USA).

Results

Demographics and Baseline Characteristics

Vitreous samples were collected from 25 T2DM patients with PDR and 25 nondiabetic controls. The demographics, BMI, laboratory data, and ocular characteristics for all subjects are shown in Table 1. Patients in the two groups were matched based on age (range, 46–77 years) and sex (male = 44.0%, both). The PDR group had significantly higher BMI (P = 0.008). Subjects in the control group had never been diagnosed of T2DM, and their HbA1c levels were within the normal range. The HbA1c levels at the time of enrollment were not significantly different between the two groups (PDR group, 7.9 ± 3.0%; control group, 5.5 ± 0.2%; P = 0.157), which can be attributed to effective medical control of T2DM in the PDR group. The PDR group had significantly lower estimated glomerular filtration rate (eGFR) (P = 0.001) and higher creatinine level (P = 0.008) compared with the nondiabetic control group, suggesting moderate loss of kidney function.

Table 1.

Demographics, Laboratory Data, and Ocular Characteristics of the PDR and Control Groups

	Control (n = 25)	PDR (n = 25)	P Value*
Age (years) (median, IQR)	61.8 (55–70)	61.5 (57–65)	0.503
Gender (male:female)	11:14	11:14	1.000
BMI (kg/m2)	23.5 ± 2.5	25.8 ± 3.6	0.008
HbA1c (%)	5.5 ± 0.2	7.9 ± 3.0	0.157
Duration of diabetes (years)	—	9.6 ± 8.0
Total cholesterol (mg/dL)	161.5 ± 48.4	197.5 ± 29.9	0.242
Triglyceride (mg/dL)	100.0 ± 49.9	247.3 ± 220.3	0.092
LDL (mg/dL)	107.7 ± 31.2	75.5 ± 17.7	0.208
HDL (mg/dL)	54.5 ± 11.6	68.5 ± 24.1	0.256
ALT (U/L)	22.6 ± 17.1	11.2 ± 4.4	0.065
eGFR (mL/min/1.73m2)	97.9 ± 19.6	48.7 ± 41.3	0.001
Creatinine (mg/dL)	0.74 ± 0.13	3.0 ± 2.8	0.008
BCVA (logMAR)	0.630 ± 0.272	1.443 ± 0.724	<0.001
IOP (mm Hg)	15.9 ± 7.3	15.4 ± 2.9	0.735
Phakia: pseudophakia	17: 8	19: 6	0.754
OCT CST (µm)	428.7 ± 128.8	404.1 ± 253.4	0.666
IVI anti-VEGF	0	12	0.000
PRP treatment	0	19	0.000
Surgical indications: VH:FVP:TRD	0	8:11:6	0.000

ALT, alanine aminotransferase; BCVA, best-corrected visual acuity; CST, central subfield thickness; eGFR, estimated glomerular filtration rate; HDL, high-density lipoprotein cholesterol; IOP, intraocular pressure; LDL, low-density lipoprotein cholesterol; OCT, optical coherence tomography; TRD, traction retinal detachment.

Values are mean ± SD unless otherwise noted.

^*Paired-samples t test for continuous variable; Fisher's exact test for categorical variables.

P  values of <0.05 were considered statistically significant and were presented in bold.

Among ocular characteristics, the preoperative best-corrected visual acuity in the PDR group (logMAR 1.443 ± 0.724) was significantly worse than the control group (logMAR 0.630 ± 0.272; P < 0.001). Twelve subjects (48%) in the PDR group had received previous treatment of intravitreal anti-VEGF injection and 19 (76%) had undergone PRP, with 9 (36%) receiving both. Surgical indications for vitrectomy included uncleared vitreous hemorrhage (VH; 32%), fibrovascular proliferation (FVP) with or without VH (44%), and tractional retinal detachment (TRD)/tractional rhegmatogenous retinal detachment with or without VH (24%) in the PDR group. The surgical indications in the control group were epiretinal membrane (68%), macular hole (16%), myopic tractional maculopathy (8%), or others (8%).

Comparison of the Ceramide Level in the Vitreous of the PDR and Control Groups

Levels of overall ceramide and C16:0, C18:0, C20:0, C22:0, C24:1, and C24:0 ceramides were quantified by UHPLC-MS/MS analyses (Table 2). The vitreous from the control and PDR patients had similar protein concentrations (P = 0.2360). The overall level of ceramide was substantially higher in the PDR group compared with the control group (P = 0.006). The levels of all except for C18 ceramide species were significantly higher in the PDR group. Among ceramides of different acyl chain lengths, the increases of very-long-chain (VLC) ceramides (C22–C24) were more drastic than those of long-chain ceramides (C16–C20) (Fig. 1A). The vitreous of control and PDR eyes had distinct composition ratios of ceramide with different acyl chain lengths. In the PDR group, 76.1% of ceramides consist of VLC ceramide species (Fig. 1C), whereas C16 ceramide was the predominant species in the control vitreous, accounting for 40.5% of all ceramides (Fig. 1B).

Table 2.

Levels of Ceramide Species in the Vitreous of the Control and PDR Groups

	Control (n = 25)	PDR (n = 25)	P Value*
Protein concentration (mg/mL)	5.824 ± 1.591	6.736 ± 3.062	0.2360
Pooled ceramide concentration (pmol/mg protein)	3.524 ± 2.456	18.626 ± 19.264	0.0006
Cer (d18:1/16:0)	1.426 ± 0.883	3.211 ± 2.760	0.0034
Cer (d18:1/18:0)	0.372 ± 0.678	0.576 ± 0.747	0.2942
Cer (d18:1/20:0)	0.124 ± 0.035	0.664 ± 0.992	0.0122
Cer (d18:1/22:0)	0.307 ± 0.318	2.158 ± 2.365	0.0007
Cer (d18:1/24:1)	0.765 ± 0.653	5.703 ± 6.486	0.0006
Cer (d18:1/24:0)	0.530 ± 0.644	6.313 ± 7.729	0.0009

Cer, ceramide.

^*Paired-sample t test for continuous variable.

P Values of <0.05 were considered statistically significant and were presented in bold.

Values are mean ± SD.

Figure 1.

The relative differences and the proportional composition of ceramide species in the vitreous of the control and PDR groups. (A) The scattered box plot illustrates the levels of C16 to C24 ceramide species as well as the overall change between the control and PDR vitreous. (B, C) The pie charts show the proportional composition of C16 to C24 ceramide species in the control (B) and PDR (C) vitreous, respectively.

Systemic Factors and the Concentration of VLC Ceramide

Stepwise linear regression analysis was performed to identify potential systemic factors related to the levels of VLC ceramides including C22:0, C24:1, and C24:0 in the vitreous among all subjects (Table 3). In the univariable analysis, only diabetes (β = 13.8437; P = 0.0002), BMI (β = −1.0584; P = 0.0770), eGFR (β = −0.1426; P = 0.0484), and creatinine (β = 2.9810; P = 0.0131) were found to be associated with the level of VLC ceramide in the vitreous. No significant association was identified between age, sex, total ceramide levels, HbA1c, total cholesterol, triglyceride, low-density lipoprotein cholesterol, high-density lipoprotein cholesterol, or alanine aminotransferase. In the multivariable regression analysis, diabetes, BMI, and eGFR were included (Table 3). Creatinine was not included in the model owing to high multicollinearity with eGFR. The analysis showed that diagnosis of diabetes (β = 14.5751; P = 0.0327) and lower BMI (β = −2.1396; P = 0.0173) were significantly associated with higher level of VLC ceramide whereas eGFR was not (P  =  0.5762).

Table 3.

The Association Between Systemic Factors and the Concentration of VLC Ceramide in the Vitreous

	Univariate Analysis		Multivariate Analysis
	β (95% CI)	P Value	β (95% CI)	P Value
Diabetes (ref = control)	13.8437 (7.8997 to 20.6900)	0.0002	14.5751 (1.3202 to 27.8300)	0.0327
BMI, per 1 kg/m2 higher	−1.0584 (−2.2361 to 0.1192)	0.0770	−2.1396 (−3.8619 to −0.4172)	0.0173
eGFR, per 1 mL/min/1.73m2 higher	−0.1426 (−0.2841 to −0.0011)	0.0484	−0.0422 (−0.1969 to 0.1124)	0.5762
Creatinine, per 1 mg/dL higher	2.9810 (0.6873 to 5.2747)	0.0131	—	—
Age, per 1 year older	0.3383 (−0.1042 to 0.7809)	0.1308	—	—
Sex (ref = Female)	0.5365 (−7.4579 to 8.5309)	0.8932	—	—
HbA1c, per 1% higher	1.2450 (−3.8053 to 6.2953)	0.6084	—	—
Triglyceride, per 1 mg/dL higher	0.0209 (−0.0401,0.0818)	0.4635	—	—
T-cho, per 1 mg/dL higher	0.1526 (−0.1551 to 0.4603)	0.2951	—	—
LDL, per 1 mg/dL higher	−0.1740 (−0.7070 to 0.3590)	0.4790	—	—
HDL, per 1 mg/dL higher	0.2358 (−0.4771 to 0.9486)	0.4675	—	—
ALT, per 1 mg/dL higher	−0.1825 (−0.7352 to 0.3702)	0.4998	—	—

ALT, alanine aminotransferase; eGFR, estimated glomerular filtration rate; HDL, high-density lipoprotein cholesterol; LDL, low-density lipoprotein cholesterol; ref, reference; T-cho, total cholesterol.

The total level of VLC ceramides, including C22:0, C24:1, and C24:0, was used for the analysis.

Significant risk factors in univariate analysis were selected for multivariate analysis. Owing to collinearity among eGFR and creatinine, only eGFR was included in the linear regression model. β values are accompanied by the 95% CI in parentheses.

Manifestation of PDR and the Concentration of Ceramides

We further explored the association between vitreous ceramide levels and different vitreoretinal manifestations of PDR based on the indications for vitrectomy among the study subjects enrolled in the study. Figure 2 shows the levels of total ceramide in the PDR subgroups, including uncleared VH (8 subjects), FVP with or without VH (11 subjects), and TRD with or without VH (6 subjects). The average levels of total ceramide in the VH and FVP subgroups (VH, 21.169 ± 7.918 pmol/mg protein and FVP, 19.269 ± 5.612 pmol/mg protein, respectively) seem to be higher than the TRD subgroup (TRD: 14.055 ± 7.637 pmol/mg protein). However, the result did not reach statistical significance (P = 0.2335). The average level of total ceramide or VLC ceramide in the presence with and without VH showed no significant differences (P = 0.1410 and P = 0.1410, respectively).

Figure 2.

Comparison of total ceramide levels according to the manifestation of PDR. The scattered box plot demonstrates the total ceramide levels among different subgroups. (A) Total ceramide levels in patients with VH (8 subjects), FVP with or without VH (11 subjects), and tractional retinal detachment (TRD) with or without VH (6 subjects). (B) Total ceramide levels in patients with the presence VH (16 subjects) and without VH (no VH, 9 subjects). The levels of total ceramide in the VH and FVP subgroups appear higher than the TRD subgroup while the ceramide level in the subgroup with VH seem to be higher than that without VH. However, the differences did not reach statistical significance. ns: not significant.

Anti-VEGF and Panretinal Photocoagulation Treatments on the Concentration of Ceramides

To assess the effect of anti-VEGF IVI on ceramide levels, we divided patients into three groups: those who did not receive IVI, those who received a single IVI before vitrectomy, and those who received more than one IVI (Table 4). Patients who received a single IVI exhibited the highest vitreal ceramide levels compared with the control group and those receiving multiple IVIs. Although the differences were not statistically significant owing to the small sample size, our findings suggest that IVI may reduce ceramide levels in the vitreous. Among 25 PDR patients, 19 received PRP. The ceramide levels were similar between patients with or without PRP (Table 5). The lack of difference may be attributed to the long interval between the PRP treatments and sample collection. Our findings suggest that PRP has no lasting effect on ceramide levels in the vitreous.

Table 4.

Effect of IVI Anti-VEGF on Levels of Vitreal Ceramide in T2DM PDR Patients

pmol/mg Protein	No Treatment (n = 13)	Single IVI (n = 8)	Multiple IVI (n = 4)	P Value*
LC ceramide	3.337 ± 3.434	7.436 ± 4.315	3.753 ± 3.317	0.076
VLC ceramide	10.924 ± 14.165	25.873 ± 17.271	7.742 ± 5.950	0.068

^*By Kruskal-Wallis test.

Values are mean ± SD.

Single IVI, one IVI anti-VEFG before vitrectomy; multiple IVI, more than one IVI before vitrectomy; LC, long-chain.

Table 5.

Effect of PRP Treatment on Levels of Vitreal Ceramide in T2DM PDR Patients

pmol/mg Protein	No Treatment (n = 6)	PRP (n = 19)	P Value*
LC ceramide	4.308 ± 3.418	4.497 ± 3.452	0.726
VLC ceramide	12.795 ± 12.268	14.610 ± 14.854	0.976

LC, long-chain.

^*By Mann–Whitney U test.

Values are mean ± SD.

Discussion

In the present study, we collected vitreous samples from living PDR patients and compared the level of ceramide species with that of age- and sex-matched nondiabetic controls. We found that the increase of VLC ceramide dramatically changed the overall composition of vitreal ceramide in the PDR group. Multivariable analyses revealed that vitreal ceramide levels were associated with T2DM and BMI. These findings demonstrated a specific effect of T2DM on vitreal ceramide and hinted that ceramide may participate in PDR pathogenesis.

Although ceramide metabolism has been studied extensively in metabolic diseases, most analyses focused on measuring circulating ceramide in blood samples. Plasma ceramide accumulates in response to obesity and correlates with BMI,²⁵ and plasma ceramide levels are found to correlate with insulin resistance.²⁵ Further in vitro studies show that ceramide interferes with insulin signaling, likely by affecting receptor localization to ceramide-rich lipid microdomains.²⁶ These studies show that ceramide may be a link between hyperlipidemia and insulin resistance. However, several recent studies show that the sphingolipid profiles in the cerebrospinal fluid or other bodily fluids are different from the plasma levels. For example, recent studies revealed a lack of correlation between plasma ceramide levels and those of cerebrospinal fluid or urine.^27–29 Fox et al.³⁰ measured the levels of ceramide species in the plasma, heart, and liver in T2DM animal models. Compared with control animals, different tissues of diabetic animals showed different changes in ceramide profiles. Vitreous fills the posterior cavity of the eye and is in direct contact with the inner retina, ciliary body, and lens. Vitreous functions more than serving as a structural support and cushion, because it also contributes to the repository and diffusion of nutrients and wastes. Although the composition of vitreous is considered rather static, its components and consistency are known to vary by age, inflammation, and altered metabolic activities.³¹ Treating retinal pigment epithelial cells in vitro with pathological vitreous collected from patients with proliferative vitreoretinopathy induced the expression of EMT markers and produced significant contraction of collagen gels, whereas control cadaver vitreous treatment had no such effects.²³ Therefore, measuring ceramide level in the vitreous allows for a more accurate assessment of its possible effects at the site of pathogenesis and contributes to advancing the understanding of the disease mechanism.

Even though vitreous composition is highly relevant to the mechanism of PDR and other retinal diseases, reports of ceramide levels in the eyes of PDR patients are scarce. Priyadarsini et al.²¹ collected post mortem corneas from donors with T1DM, T2DM, and nondiabetic controls and analyzed three samples from each group. The results showed that, in patients with type 1 diabetes or T2DM, the levels of ceramide were elevated compared with nondiabetic controls. Wilmott et al.²² measured ceramide levels in post-mortem vitreous of four nondiabetic controls and nine patients with T2DM. They showed an overall increase in ceramide levels as well as a 1.74- to 2.08-fold increases in the levels of VLC ceramides, corresponding with 5% or lower changes in the composition ratio of individual ceramide species. The recent study by Dorweiler et al.³² (2024) measured vitreal ceramide levels in seven T2DM PDR patients and seven control patients with macular holes and reported increased in C16- and C24-ceramides in patients with PDR compared with nondiabetic controls. We analyzed the levels of C16 to C24 ceramide in larger numbers of living samples collected from diabetic and nondiabetic patients of matching age and sex. Similar to Wilmott et al., we also show increases in overall ceramides, especially VLC ceramides. Noticeably, our results showed more striking elevations of VLC ceramides, with a greater than 5-fold increase in ceramide levels and up to 18.9% change in the composition ratio. Because our samples were collected from living patients during surgery rather than post mortem, our findings likely better represent the differences in a clinical setting.

Dorweiler et al.³² showed that C26:0 ceramide was the most abundant ceramide species, accounting for 34% of total ceramides in nondiabetic control vitreous. In PDR patients, C26:0 ceramide greatly reduced to only 14% of total ceramides.³² Previous studies have shown that the retinal level of C26 ceramide was controlled by acid sphingomyelinase and the fatty acid elongase ELOVL4, and C26 ceramide plays an important role in preserving the barrier integrity of vasculature to prevent diabetic retinopathy. However, little C26 ceramides were detected in the post-mortem vitreous and cornea.^21,22 In the present study, we did not include C26 ceramide in the results because we could not detect C26 ceramides in most of our samples. Possible explanations for the discrepancy include differences in the study populations in terms of race, gender ratio, other clinical characteristics, and differences in the sphingolipidomic methodology.

PDR patients often suffer from macular edema and have newly formed blood vessels highly permeable to fluorescein.³³ Vascular leakage may contribute to the elevated ceramide level detected without hemorrhage in the vitreous of PDR patients. For example, Lim et al.³⁴ (2012) showed a significant correlation in homocysteine concentrations between the vitreous and plasma of PDR patients. Similarly, Wang et al.³⁵ (2022) compared the metabolic profiles between the plasma and vitreous and identified four of the five overlapping metabolites showing positive correlations between the plasma and vitreous. Indeed, our finding of elevated ceramide in the PDR vitreous is consistent with previous analyses of T2DM plasma and vitreous (summarized in Table 6).^{22,25,32,36–39} However, vascular leakage alone may not be the only determinant. Wu et al.⁴⁰ (2020) showed that vitreous and aqueous cytokines strongly correlated with each other in PDR patients, but found no significant correlation with plasma cytokines. These findings highlight the need for further studies to elucidate the factors regulating the compositions of plasma and vitreous.

Table 6.

Comparison of Ceramide Levels in the Plasma and the Vitreous of Diabetes With/Without PDR

		Total Ceramide Levels
Reference	Sample	Non-T2DM Control	T2DM	Ratio (T2DM/ Non-T2DM)
Haus et al. 200925	Plasma*	2370 ± 190 pmol/mL	3060 ± 260 pmol/mL	1.29
Chew et al. 201936	Plasma	3717.53 pmol/mL	4133.87 pmol/mL	1.11
Dugani et al. 202137	Plasma*	3714.87–3721.07 pmol/mL	Prevalent†: 3524.43 pmol/mL	0.95
			Incident†: 3763.03 pmol/mL	1.01
Denimal et al. 202338	Plasma	1882 (1472–6100) pmol/mL	3585 (2353–4650) pmol/mL	1.9
Mandal et al. 202439	Plasma*	FH−: 4273.9 ± 1297.4 pmol/mL	—	—
		FH+: 2712.77 ± 1238.3 pmol/mL
Wilmott et al. 201922	Cadavers vitreous	214.55 ± 8.20 pmol/mg protein	327.15 ± 46.70 pmol/mg protein	1.52
Dorweiler et al. 202432	Living PDR vitreous	71.62 ± 37.97 pmol/µl	203.13 ± 178.90 pmol/µl	2.84
This study	Living PDR vitreous	3.524 ± 2.456 pmol/mg protein	18.626 ± 19.264 pmol/mg protein	5.26

FH, family history of T2DM.

^*The ceramide levels have been converted to pmol/mL for plasma samples for easier comparison.

^†Prevalent: patients with pre-existing T2DM before the study; Incident: patients with newly diagnosed T2DM during the study.

Although the present study was underpowered for subgroup analysis, our data suggest a trend across patient groups with varying anti-VEGF treatment. In the present study, patients with severe disease received a single IVI before vitrectomy because anti-VEGF injections have been shown to reduce bleeding during vitrectomy.⁴¹ Consequently, a single IVI reflects a more severe disease status rather than a sustained anti-VEGF effect. These patients exhibited the highest ceramide levels, likely because the anti-VEGF effect was insufficient to reduce ceramide levels. This result aligns with previous findings that anti-VEGF therapy reduces VEGF levels more significantly over prolonged periods.⁴² Thus, multiple anti-VEGF IVIs may lower vitreous ceramide by reducing vascular leakage and promoting neovascular regression in PDR.

Ceramide is estimated to constitute approximately 1% of total retinal lipids or 15% of total sphingolipids.⁴³ Although ceramide is not present at a relatively high level, increased vitreal ceramide likely contributes to the hyperglycemic damage of the retina. Intra-vitreal injection of ceramide induced apoptosis of retinal ganglion cell layer and caused vision loss in rats.⁴⁴ Surgically induced retinal detachment caused ceramide accumulation and subsequent apoptosis in rabbit retina.⁴⁵ Mechanistically, ceramide colocalized with tight junction complexes, and the constitution of ceramide species affected the permeability of the retina.⁴⁶ Ceramide has been reported to inhibit tube formation in endothelial cells, block AKT/ERK signaling, and was implicated in mediating fatty acid-induced endothelial VEGF resistance, thereby impairing angiogenesis.⁴⁷ Ceramide is also a critical part of cellular stress responses.^48,49 In vitro and animal studies have established a role for ceramide in cell death under increased oxidative stress.^50–52 Ceramide can induce the expression of TXNIP,⁵³ a key regulator of cellular oxidation that is highly induced in the diabetic retina.^54,55 Also, high glucose increased reactive oxygen species production via the Rac1-Nox2 axis in streptozocin-induced diabetic mice.⁵⁶ Moreover, elevated ceramide can disrupt mitochondrial membrane potential to increase reactive oxygen species production, leading to the activation of the calpain-mediated apoptotic pathway, resulting in photoreceptor cell death.⁵² Inhibiting ceramide synthesis significantly reduced oxidative stress and prevented cell death in both cell and animal models,^16,57–61 demonstrating an essential role for ceramide in oxidative stress-induced cell death. Interestingly, VLC ceramide shows strong lipotoxicity, impairs mitochondrial function, and causes apoptotic cell death of cardiomyocytes, whereas LC ceramide has a much milder effect.⁶² Inhibition of ceramide, especially those with very long acyl side chains, may serve as a potential strategy to protect against diabetic retinopathy.

The present study has a few limitations. Owing to the scope of the study, the number of participants was insufficient to reveal the differences among PDR patients with different clinical presentations as shown in the subgroup analyses. Another weakness is that we were not able to recruit T2DM patients without DR or T2DM patients with nonproliferative DR for comparison. Also, we were not able to collect matching blood samples to assess the correlation between the plasma and vitreous ceramide levels. Moreover, the effects of vitreous ceramide levels on PDR progression could not be evaluated because of the lack of follow-up measurements. Last, most of the T2DM patients in the present study had normal HbA1c, potentially minimizing the differences in ceramide levels. The HbA1c values also indicated good glycemic control in our cohort, which may not be the case in other T2DM cohorts, thus limiting the clinical interpretation. Nonetheless, the present study provides a valuable collection of vitreal ceramide measurements between age- and sex-matched living patients with or without T2DM and PDR.

Conclusions

The present study compared the levels of vitreal ceramide in T2DM patients with PDR to those in age- and sex-matched nondiabetic patients who underwent vitrectomy. Our findings demonstrated an overall increase of ceramide in the vitreous, especially the VLC species. While the differences did not reach statistical significance, likely owing to the small sample size, our findings indicate a potential trend that anti-VEGF therapy may lower ceramide levels in the vitreous. The finding supports a potential role for ceramide in the pathogenesis of retinopathy that is associated with T2DM.