Altered High-Density Lipoprotein Expression Pattern in the Aqueous Humor From Eyes With Age-Related Macular Degeneration

Abstract

Purpose

Age-related macular degeneration (AMD) is associated with altered protein expression in the aqueous humor (AH). We aim to perform an unbiased proteomic analysis of the AH to identify proteomic signatures characterizing the disease's neovascular AMD (nAMD) and non-neovascular AMD (nnAMD) stages.

Methods

AH samples were collected from eyes with nAMD (n = 39), nnAMD (n = 30), and healthy control eyes (n = 36). Liquid chromatography-tandem mass spectrometry (LC-MS/MS) was used to analyze the AH proteome; differentially expressed proteins underwent functional analysis. Serum high-density lipoprotein cholesterol (HDL-C) levels were correlated with AH high-density lipoprotein (HDL) pathway proteins.

Results

Seventeen proteins were upregulated and five were downregulated in nAMD eyes compared with healthy controls. Enriched pathways include the fibrinogen complex, the complement alternate, the spherical HDL particle, and the serine protease inhibitor. Forty-six proteins were upregulated in nAMD versus nnAMD, whereas nine were downregulated. Enriched pathways included the spherical HDL particle, peptidase S1A, membrane attack complex, Sushi domain, and complement alternate pathway. No differentially expressed proteins were found between nnAMD and control eyes. Upregulated proteins associated with the spherical HDL particle pathway in nAMD, including apolipoprotein A-I, apolipoprotein A-II, and paraoxonase 1, exhibited a negative correlation with serum HDL-C levels (r = −0.43, P = 0.012; r = −0.37, P = 0.031; and r = −0.52, P = 0.002, respectively).

Conclusions

HDL-associated proteins exhibit increased expression in AH of nAMD eyes, independent of serum HDL-C levels. These findings suggest that serum HDL-C may not accurately reflect ocular HDL particle levels, highlighting the role of retinal lipid dysregulation in nAMD.

Proteomic analysis of involved tissues can provide important insights into complex multifactorial diseases, such as age-related macular degeneration (AMD). Recent advancements in proteomic technologies have made it possible to characterize tissue proteomes in an unbiased manner, and improvements in methodology and technology have made it possible to generate large datasets from minimal sample volumes. Aqueous humor (AH) proteomics reflects retinal metabolic and pathological processes. Several studies have identified alterations in AH protein composition associated with eye diseases, among them high myopia,¹ glaucoma_,² and diabetic retinopathy.³

Although AMD is localized to the macula, which constitutes less than 1% of the total fundus area, and the AH reflects the broader intraocular environment, AH sampling remains one of the least invasive methods available for investigating biochemical changes within the living human eye. Previous studies have demonstrated that localized retinal pathology, including AMD, can result in detectable alterations in the protein or cytokine composition of the AH.^4–6 Indeed, such alterations have been reported by our group and others.^7–11 In our previous study, we identified altered protein levels of several functional protein classes, among them complement protein dysregulation,¹² in neovascular AMD (nAMD) eyes versus controls. Coronado et al.¹³ identified notable changes in nAMD AH proteins associated with oxidative stress, inflammation, and angiogenesis, and Cao and colleagues observed APOB100 upregulation in nAMD eyes requiring fewer anti-vascular endothelial growth factor (VEGF) injections.¹⁴ Serum and plasma Omics studies have identified genetic, metabolic, and proteomic factors associated with AMD.^15–17 This suggests that the disease pathogenesis involves altered homeostasis of systemic pathways, among them, inflammation and lipid metabolism. These findings highlight the potential of proteomics for elucidating disease mechanisms and identifying biomarkers for improving prognosis.

The current study aimed to refine and extend previous studies by directly comparing the AH proteome between AMD subtypes. Whereas our prior study¹² utilized mass spectrometry on pooled nAMD AH samples, in the current study, we have analyzed individual samples to capture expression variation among eyes, and to compare nAMD with non-neovascular AMD (nnAMD).

Methods

Patients

The study protocol was approved by the Ethics Committee of the Hadassah Medical Center (#112-10 HMO), and each patient signed an informed consent form. Patients were recruited from the retina clinic of the Department of Ophthalmology at the Hadassah-Hebrew University Medical Center. Patients were diagnosed with AMD according to the age-related eye disease study (AREDS) criteria¹⁸ based on clinical examination combined with optical coherence tomography (OCT)-based assessment. Patients were excluded from the study if they had any other potential cause for macular neovascularization (MNV), such as high myopia (>6 diopters), trauma, or uveitis. Also excluded were patients with a major systemic illness, such as active cancer, autoimmune disease, congestive heart failure, or uncontrolled diabetes.

A sample of AH was collected during cataract surgery from 105 patients: 39 patients with nAMD (mean age = 79 ± 6.28 years, female/male = 20/19), 30 patients with nnAMD (mean age = 77.23 ± 7.72 years, female/male = 15/15), and 36 control patients (mean age = 76 ± 8.34 years, female/male = 21/15) at the Hadassah-Hebrew University Medical Center. Only one eye from each patient was included in the study. After topical anesthesia, the eye was prepared aseptically according to routine procedures for ophthalmic surgery. AH sample (50–150 µL) was collected using a 1-mL syringe and a 25-G cannula immediately following the generation of the first surgical wound. After collection, AH samples were rapidly transferred to an Eppendorf tube and immediately stored at −80°C for further analysis. Clinical information, including demographics, diagnoses, history of eye disease, medical history, visual acuity, and serum lipid profile (obtained after an overnight fast) within a 2-year window around the sampling date, was collected from the electronic medical records (EMRs).

Liquid Chromatography-Tandem Mass Spectrometry Analysis

Samples were prepared as follows: 30 µL AH sample was used, and 100 µL 8 M urea, 10 mM DTT, and 25 mM Tris-HCl pH 8.0 were added, and the samples were incubated for 30 minutes at 22°C. Next, 55 mM Iodoacetamide was added, and the sample was incubated for 30 minutes at 22°C, in the dark, followed by the addition of 20 mM DTT. The urea was diluted by the addition of 8 volumes of 25 mM Tris-HCl, pH 8.0. Trypsin was added (0.4 µg/sample), and the samples were incubated overnight at 37°C with gentle agitation. Trypsin-digested peptides were then desalted on C18 homemade Stage tips. Mass spectrometer (MS) analysis was performed using a Q Exactive Plus MS (Thermo Fisher Scientific, Waltham, MA, USA) coupled online to a nanoflow ultra-high-performance liquid chromatography (UHPLC) instrument, Ultimate 3000 Dionex (Thermo Fisher Scientific, Waltham, MA, USA). Peptides (0.45 µg, based on optical density [OD] of 280 nm estimation) were separated using a nonlinear gradient (0%–80% acetonitrile) run at a flow rate of 0.3 µL/min on a reverse phase 25-cm-long C18 column (75 µm ID, 2 µm, 100 Å, Thermo PepMap RSLC) for 105 minutes. The survey scans (380–2000 m/z, target value 3E6 charges, maximum ion injection times 50 ms) were acquired and followed by higher energy collisional dissociation (HCD) based fragmentation (normalized collision energy 25). A resolution of 70,000 was used for survey scans, and up to 15 dynamically chosen most abundant precursor ions, with a “peptide preferred” profile, were fragmented (isolation window 1.8 m/z). The tandem mass spectrometer (MS/MS) scans were acquired at a resolution of 17,500 (target value 5E4 charges, maximum ion injection times 57 ms). Dynamic exclusion was 60 seconds. Data were acquired using Xcalibur software (Thermo Fisher Scientific, Waltham, MA, USA). To avoid a carryover, the column was washed with 80% acetonitrile, 0.1% formic acid for 25 minutes between samples. Mass spectra data were processed using the MaxQuant computational platform, version 2.0.3.0. Peak lists were searched against translated coding sequences of the human proteome obtained from Uniprot. To control for false-positive identifications in this exploratory analysis, a target-decoy strategy was utilized. This approach uses a database built from reversed protein sequences (decoy) to empirically estimate the error rate, allowing for the enforcement of a strict false discovery rate (FDR). The search included cysteine carbamidomethylation as a fixed modification and oxidation of methionine as variable modifications and allowed up to two miscleavages. The match-between-runs option was used. Peptides with a length of at least seven amino acids were considered, and the required FDR was set to 1% at the peptide and protein levels. Relative protein quantification in MaxQuant was performed using the label-free quantification (LFQ) algorithm.

Statistical Analysis

Statistical analysis for the proteomic data was performed using the Perseus statistical package.¹⁹ To ensure statistical reliability, proteins with at least eight non-zero LFQ values in at least one sample group were included in the analysis. A two-tailed t-test and volcano plot were applied to screen differentially expressed proteins (DEPs) between the AMD and control groups. The default software parameters were used for all statistical computations, including FDR correction.

Although a formal a priori power calculation was not performed due to the study’s exploratory nature, a post hoc power analysis for a key finding, apolipoprotein A1, confirmed the adequacy of the sample size for these comparisons. The analysis was conducted using RStudio (version 2025.05.1, https://posit.co/) and demonstrated high statistical power (α = 0.05) for the nAMD versus control comparison (power = 0.99) and the nAMD versus nnAMD comparison (power = 0.92).

Pearson correlation analysis was performed using GraphPad Prism version 8 (GraphPad, San Diego, CA, USA) to assess the relationship between serum HDL-C levels and AH spherical HDL particle levels.

Proteome Functional Analysis

The biological functions of the DEPs were assessed by functional analysis via the DAVID website (https://david.ncifcrf.gov/tools.jsp). DAVID utilizes multiple data sources, for example, UniProt Keywords (KW), Sequence features, Gene Ontology (GO) Terms, Kyoto Encyclopedia of Genes and Genomes (KEGG), InterPro Protein interactions (IPR), and others. Significance was set at a Benjamini-Hochberg (BH) adjusted P value < 0.05 and an enrichment score (ES) >2. Significant clusters were also determined via χ2.

Area Under the Curve Analysis

To assess diagnostic accuracy, we performed receiver operating characteristic (ROC) curve analysis and calculated the area under the curve (AUC). This analysis compared the expression levels of spherical HDL particle proteins (ApoA-I, ApoA-II, and PON1) and key complement alternate pathway proteins (C3 and C5), as determined by liquid-chromatography tandem mass spectrometry (LC-MS/MS), among patients with nnAMD, nAMD, and control groups. The calculations were conducted using RStudio (version 2025.05.1, https://posit.co/). For the analysis, patients with AMD were coded as “1” and controls as “0.”

Results

A total of 726 proteins were identified in the AH samples through LC-MS/MS (Supplementary Table S1). Subsequently, statistical analysis was performed on 234 proteins, each possessing a minimum of 8 non-zero LFQ values in at least one sample group.

Differential AH Proteome Expression Between Patients With nAMD and Controls

Seventeen proteins were upregulated in nAMD (nAMD/Control > 1 and FDR < 0.05) compared with age-matched controls (Fig. 1A, Supplementary Table S2). Five proteins were downregulated in patients with nAMD compared with controls (nAMD/Control < 1 and FDR < 0.05). Functional analysis of the proteins upregulated in patients with nAMD demonstrated enrichment for four biologic groups including Serine protease inhibitor (enrichment score [ES]: 4.81 and KW-0722), spherical HDL particle (ES: 3.5 and GO:0034366), Complement alternate pathway (ES: 3.44 and KW-0179), and fibrinogen complex (ES: 2.3 and GO:0005577; Fig. 2A).

Figure 1.

Volcano plots illustrating differential protein expression in AH of patients with AMD. The x-axis represents the log2 fold difference in protein expression between the indicated groups, whereas the y-axis represents the negative log10 P value, indicating the statistical significance of the difference between the tested groups. Red dots represent the differentially expressed proteins (FDR ≤ 0.05). (A) Seventeen proteins were upregulated in patients with nAMD compared with controls, and five proteins were upregulated in the controls. (B) Forty-six proteins were upregulated in patients with nAMD compared to patients with nnAMD, whereas nine proteins were upregulated in nnAMD. (C) No differentially expressed proteins were identified between controls and patients with nnAMD.

Figure 2.

Histograms showing enrichment score levels for functional groups that were upregulated in patients with nAMD. Differentially expressed functional groups were identified using the DAVID algorithm. (A) Patients with nAMD versus controls and (B) patients with nAMD versus patients with nnAMD.

Differential AH Proteome Expression Between nAMD and nnAMD

There were 46 proteins upregulated in nAMD compared with nnAMD (nAMD/ nnAMD >1, FDR < 0.05; Supplementary Table S3, Fig. 1B). Nine proteins were upregulated in nnAMD compared with nAMD (nAMD/nnAMD < 1 and FDR < 0.05). Analysis was conducted to evaluate the biological functions of the nAMD upregulated proteins. The analysis demonstrated enrichment for five biological pathways upregulated in nAMD including complement alternate pathway (ES: 9.9 and GO:0006956), DOMAIN: Sushi (ES: 5.08 and KW-0768), membrane attack complex (ES: 4.34 and GO:0005579), Peptidase S1A, chymotrypsin-type (ES: 2.88 and IPR001314), and spherical HDL particle (ES: 2.66 and GO:0034366; Fig. 2B). The separate identification of these two complement-related pathways reflects the known biological sequence of the complement cascade, distinguishing between the early activation phase (alternate pathway) and the terminal, pore-forming effector phase (membrane attack complex).

There were no significant changes in protein expression between the control and nnAMD eyes (Figs. 1C).

Diagnostic Accuracy of Differentially Expressed Proteins

To assess the diagnostic accuracy of AH DEPs in identifying nAMD, we used the expression levels of ApoA1, ApoA2, and PON1, followed by ROC curve analysis. The AUC was calculated by comparing expression levels among the nnAMD, nAMD, and control groups. When comparing nAMD and control groups, ApoA1 exhibited an AUC of 0.74 (95% confidence interval [CI] = 0.63–0.86, P = 0.03), ApoA2 had an AUC of 0.75 (95% CI = 0.63–0.87, P = 0.04), and PON1 demonstrated an AUC of 0.71 (95% CI = 0.59–0.83, P = 0.08). For the comparison between the nnAMD and nAMD eyes, ApoA1 showed an AUC of 0.73 (95% CI = 0.61–0.85, P = 0.057), ApoA2 had an AUC of 0.74 (95% CI = 0.61–0.86, P = 0.051), and PON1 exhibited an AUC of 0.64 (95% CI = 0.50–0.77, P = 0.30; Fig. 3).

Figure 3.

ROC analysis. Receiver operating characteristic (ROC) analysis was performed to evaluate the discriminatory ability of ApoA1, ApoA2, and PON1, which were upregulated in patients with nAMD compared to patients with nnAMD and the control groups. For patients with nAMD versus controls: (A) ApoA1: AUC = 0.74, 95% CI = 0.63–0.86, P = 0.03. (B) ApoA2: AUC = 0.75, 95% CI = 0.63–0.87, P = 0.04. (C) PON1: AUC = 0.71, 95% CI = 0.59–0.83, P = 0.08. For patients with nAMD versus patients with nnAMD: (D) ApoA1: AUC = 0.73, 95% CI = 0.61–0.85, P = 0.057. (E) ApoA2: AUC = 0.74 95% CI = 0.61–0.86, P = 0.051. (F) PON1: AUC = 0.64, 95% CI = 0.50–0.77, P = 0.30.

To contextualize the diagnostic performance of these HDL markers, a comparative ROC analysis was performed using key proteins from the complement alternate pathway, which was also found to be significant. This analysis yielded comparable moderate AUC values. For C3: control versus nAMD (AUC = 0.746, 95% CI = 0.63–0.86, P < 0.001) and nnAMD versus nAMD (AUC = 0.738, 95% CI = 0.614–0.861, P < 0.001). For C5: control versus nAMD (AUC = 0.719, 95% CI = 0.6–0.83, P < 0.001) and nnAMD versus nAMD (AUC = 0.690, 95% CI = 0.556–0.825, P = 0.005). In contrast, proteins that were not found to be statistically differentially expressed lacked discriminatory power, with AUC values approximating 0.5. For example, APOE and CFD both yielded AUC values between 0.488 and 0.521, with all P values being nonsignificant (P > 0.75). Taken together, these analyses confirm that the moderate AUCs observed for the differentially expressed HDL proteins represent a true, albeit imperfect, disease signal, clearly distinct from the lack of signal in nonsignificant proteins.

Serum HDL-C Levels and AH Spherical HDL Particle Levels

To evaluate whether the upregulation of the spherical HDL particle proteins in AH of nAMD eyes was due to leakage from MNV tissue in nAMD eyes or if the HDL proteins were synthesized in the eye, we performed a correlation analysis between serum HDL-C levels and AH spherical HDL particle levels. We compared the levels of ApoA1, ApoA2, and PON1 in the AH with the serum HDL-C levels in patients who had serum HDL-C measurements available. Serum HDL-C data were available within 1 year of AH sample collection for most participants (n = 67). Five samples had a time interval between tests exceeding 1 year, with a maximum interval of 2 years. The median time interval between AH and serum HDL-C measurements was 0.26 years (range = 0–1.9 years). This analysis utilized a subgroup of the AH cohort, comprising 33 patients with nAMD (mean age = 78 ± 6 years, 16 female patients and 17 male patients), 21 patients with nnAMD (mean age = 75.7 ± 8.2 years, 11 female patients and 10 male patients), and 18 unaffected controls (mean age = 76.6 ± 9.2 years, 7 female patients and 11 male patients). In patients with nAMD, negative correlation was observed for ApoA1 (r = −0.43, P = 0.012), ApoA2 (r = −0.37, P = 0.031), and PON1 (r = −0.52, P = 0.002), suggesting that the higher AH level of the HDL particle proteins in this group is not due to leakage from blood vessels. There was no correlation between AH HDL particle protein levels and serum level of HDL-C in patients with nnAMD and control patients (Fig. 4).

Figure 4.

Correlation analysis among AH spherical HDL particle proteins (ApoA1, ApoA2, and PON1) levels and serum HDL-C in nAMD, nnAMD, and control eyes (Pearson correlation). Each scatter plot displays individual patient data points, with serum HDL-C levels plotted on the x-axis and AH protein levels on the y-axis. (A) In patients with nAMD, significant negative correlations were observed for ApoA1 (r = −0.43, P = 0.012), ApoA2 (r = −0.37, P = 0.031), and PON1 (r = −0.52, P = 0.002). (B) No correlations were found among patients with nnAMD (ApoA1: r = 0.35, P = 0.11; ApoA2: r = 0.34, P = 0.13; PON1: = 0.2, P = 0.37. (C) No correlations were found among controls (ApoA1: r = 0.28, P = 0.27; ApoA2: r = 0.36, P = 0.15; PON1: r = 0.23, P = 0.38).

Discussion

We conducted an unbiased proteomic analysis of AH samples from 105 individuals, including subjects with nAMD, nnAMD, and unaffected controls. Using an LFQ-based proteomics approach, we identified a total of 726 proteins across all AH samples, with 234 proteins selected for detailed analysis. This approach enabled the detection of 22 DEPs in patients with nAMD compared with controls (17 upregulated and 5 downregulated) and 55 DEPs in patients with nAMD compared to patients with nnAMD (46 upregulated and 9 downregulated). Functional analysis of the upregulated proteins in nAMD revealed enriched pathways, including the complement alternate pathway, fibrinogen complex, serine protease inhibitor, and spherical HDL particle.

We did not identify any DEPs between the nnAMD and control groups. This lack of significant findings is likely because the pathological changes in these earlier disease stages are subtle and located deep within the retina, producing a proteomic signal in the AH that is either too faint or composed of low-abundance proteins that fall below the detection limit of our MS platform. We found a negative correlation between serum HDL-C levels and the upregulated proteins involved in the spherical HDL particle pathway in patients with nAMD. This association was notably absent in both the nnAMD and control eyes. This finding suggests that lower systemic HDL-C levels may be linked to an increase in specific HDL-associated proteins within the ocular environment, potentially reflecting a compensatory or dysregulated lipid transport mechanism specific to nAMD. Notably, low systemic HDL-C or ApoA-I levels have been strongly and inversely associated with an increased risk of cardiovascular disease, metabolic syndrome, and type 2 diabetes, further highlighting the clinical significance of HDL-related dysfunction.^20,21 This negative correlation implies that the upregulated proteins in the AH are unlikely to be a result of leakage from newly formed blood vessels, as such a mechanism would be expected to yield a positive correlation. This interpretation aligns with the established view of the eye as a distinct biochemical compartment with limited proteomic correlation to serum.²² Further supporting this, the local retinal expression of key apolipoproteins like ApoA1, but not others such as ApoA2, suggests that HDL particles are remodeled or assembled intraocularly.^23,24 This local biogenesis would likely result in an ocular specific population of HDL particles differing in composition and size from their systemic counterparts. These results therefore demonstrate that serum HDL-C levels do not accurately reflect spherical HDL particle levels in the AH, and that this specific lipid dysregulation varies by AMD subtype.

In comparison to our previous study,¹² the current investigation benefited from an expanded sample size of 105 participants, up from 20 in the previous research and included an nnAMD patient group. Additionally, we performed individual analysis of the samples instead of analyzing pooled samples. In the comparison between the nAMD and control groups, we identified 17 DEPs, 15 of which were also found to be upregulated in our previous study, which reported 239 nominally upregulated proteins in nAMD compared with controls. Fewer DEPs were identified in this study, likely due to increased variability across individual samples, which could not be assessed by our previous pooling technique. Thus, the current study validated our previous results and expanded the findings. We again identified the complement alternative pathways as upregulated in nAMD versus controls. In addition, key proteins related to the spherical HDL particle, such as ApoA1, ApoA2, and PON1, were also found to be differentially expressed and increased ApoA1 levels in AMD AH was reported by others.^7,11,12 Additional differentially expressed pathways in our current study were previously implicated in retinal disease. For example, the serine protease inhibitors pathway was implicated in ocular pathologies through their anti-inflammatory, anti-angiogenic, antioxidant, and antifibrotic effects.^25,26

Plasma HDL biogenesis begins with the liver secreting lipid-poor ApoA1. This nascent particle acquires lipids from peripheral cells (via ABCA1) to form discoidal HDL. Through the action of LCAT, this complex is esterified and matures into the Spherical HDL, the fully lipidated form of high-density lipoproteins.²⁷ The proteins identified in this study, ApoA1, ApoA2, and PON1, are specifically associated with these fully formed spherical HDLs. ApoA1 and PON1 were previously reported to be expressed in the retina, and the presence of these proteins in the AH likely stem at least in part from local retinal production.²⁴

Our ROC analysis indicated that the individual spherical HDL proteins offer moderate diagnostic utility for nAMD. To determine the significance of this, we compared their performance with key markers from the complement pathway (C3 and C5) which is strongly associated with AMD, revealing comparable AUC values of the HDL lipoproteins and the complement proteins. It confirms that the moderate AUC values of the HDL proteins in the AH likely reflect the complex, multifactorial nature of nAMD, where no single protein serves as a perfect biomarker.^7,11,12

Our subsequent analysis focused on ApoA1, ApoA2, and PON1 as they were the differentially expressed proteins identified by our unbiased analysis within the enriched spherical HDL particle pathway. Although ApoA1, ApoA2, and PON1 are known components of spherical HDL particles, each of these proteins also exerts distinct functions beyond lipid transport, including roles in modulating inflammation, oxidative injury, and angiogenesis, all of which are key processes involved in the pathogenesis of nAMD. ApoA1, the primary structural protein of HDL, has been shown to reduce NF-κB activation in myeloid cells by interfering with inflammatory receptor signaling and inhibits angiogenesis.^28,29 ApoA2, the second most abundant HDL protein, regulates lipid and glucose metabolism and may confer proinflammatory properties to HDL.^30,31 ApoA2 modulates ApoA1 conformation, influencing its interaction with endothelial lipase.^32,33 PON1, the third upregulated HDL-associated protein in the AH of nAMD eyes, contributes to HDL’s anti-inflammatory and antioxidative functions. Overexpression of PON1 protects against LDL oxidation, whereas its knockout enhances oxLDL-induced inflammation.^34–36 The patients with nAMD exhibit lower serum PON1 activity,³⁷ aligning with our finding of a negative correlation between AH PON1 levels and plasma HDL-C. Furthermore, specific PON1 genetic polymorphisms that alter its activity have previously been associated with AMD risk.³⁸

Functions of these proteins are critical because cholesterol metabolism in the retina, like the central nervous system (CNS),³⁹ is highly localized and distinct from systemic plasma. In the retina, lipids are a major component, comprising approximately ⅓ of its total weight.⁴⁰ The retina receives cholesterol from both local syntheses, mainly by glial and retinal pigment epithelial (RPE) cells and from the circulation. In the outer retina, lipids are recycled through the shedding and reformation of lipid-rich photoreceptor outer segments serving as a potential fuel substrate for photoreceptor mitochondria.⁴¹ HDL-like particles are essential in this process, they facilitate the clearance of cholesterol and other lipids from photoreceptors to the RPE, preventing harmful lipid accumulation which can manifest is the development of drusen. Furthermore, HDL serves as the main carrier of lutein and zeaxanthin, these carotenoids are important for protecting the retina from oxidative stress and supporting visual function. Lutein and zeaxanthin are particularly important in the context of AMD, and supplementation of these carotenoids can reduce the risk for AMD progression.^42,43 Thus, HDL is crucial for maintaining macular health via both maintenance of lipid metabolism and by serving as a transporter for lutein and zeaxanthin.^44,45 The biological importance of this pathway in AMD is underscored by the significant genetic overlap between HDL-related risk genes and those involved in xanthophyll bioavailability.^46,47 Therefore, the altered HDL proteome we observed could potentially impair this critical transport, offering a plausible explanation for the macula-specific pathology of early AMD.^48,49

Multiple lines of evidence implicate lipid metabolism in AMD, most notably the lipid-enriched composition of drusen and genetic polymorphisms in 11 genes associated with the risk for developing AMD.^50,51 This link is further supported by observations of mitochondrial dysfunction in degenerating cones, the presence of lipids in drusens,⁵² and the known impact of dietary lipid intake on retinal lipid composition.^53–55

HDL is known for anti-inflammatory and antioxidant functions, in chronic disease states or aging, under specific circumstances HDL can transition to become pro-inflammatory.⁵⁶ This transition has been linked with alterations in the HDL proteome, such as an increased abundance of serum amyloid A (SAA) correlating with impaired cholesterol efflux, or changes in complement proteins, which impair its normal functions.^57–59 Multiple large studies, including the European Eye Epidemiology and EYE-RISK Consortium, and a major meta-analysis report that high systemic HDL-C is paradoxically associated with an increased risk for early and intermediate AMD.^60,61 In another EYE-RISK consortium study, metabolic analysis of the serum identified increased levels of large and extra-large HDL subclasses and decreased levels of very low-density lipoprotein, amino acids, and citrate in patients with AMD. Metabolites belonging to these large and extra-large HDL subclasses were significantly associated with AMD-linked genetic variants in or near key lipid metabolism genes.⁶²

The possibility that HDL may contribute to angiogenesis is supported by the finding of correlation between elevated HDL levels and improved prognosis in ischemic and inflammatory diseases where angiogenesis is a key factor.⁶³ HDL regulatory effects on angiogenesis appear to be context-dependent, promoting angiogenesis under hypoxic conditions while inhibiting inflammation-driven angiogenesis.^64,65 Both hypoxia and inflammation are relevant pathways in the context of AMD and additional studies are required to determine the role of HDL in nAMD.

Although we have identified HDL-associated proteins dysregulation in nAMD, others have reported on a potential role for HDL in nnAMD. Kelly et al. demonstrated a protective role of HDL in a mouse model for AMD. They found that an ApoA1 mimetic peptide, 5A, could mitigate the detrimental effects of a high-fat diet on the retina and reduce the accumulation of lipid- and protein-rich material in drusen-like deposits.⁶⁶ Similarly, Rudolf and colleagues used the L-4F ApoA1 mimetic to effectively remove lipids and improve Bruch’s membrane ultrastructure in aged nonhuman primates.⁶⁷

This study has several limitations. First, its retrospective design with a limited number of participants may affect the statistical power and generalizability of the findings. Second, we have used the HDL-C levels rather than the ApoA1, ApoA2, or PON1 levels for assessing the plasma levels of HDL particles. Yet, HDL-C is positively correlated with the plasma levels of these proteins, suggesting that assessing HDL-C is a relevant surrogate.^68,69 Additionally, the sensitivity limitations of MS may have prevented the detection of low-abundance proteins, such as cytokines and chemokines, which could further enrich our understanding of the disease processes involved.

In conclusion, the current study identified altered HDL particle proteins in the AH of nAMD eyes, and suggested that their levels do not positively correlate with the plasma HDL-C levels. Future studies should explore if altered HDL proteome in nAMD, is a driving factor in the disease or a consequence of its progression, and if HDL pathway is a potential therapeutic target in AMD.