thesis on macular degeneration

Log in using your username and password

• Search More Search for this keyword Advanced search
• Latest content
• Topic collections
• BMJ Journals

You are here

• Volume 6, Issue 1
• The 100 most-cited papers on age-related macular degeneration: a bibliographic perspective
• Article Text
• Article info
• Citation Tools
• Rapid Responses
• Article metrics

• Andrzej Grzybowski 1 , 2 ,
• Chen Shtayer 3 ,
• http://orcid.org/0000-0002-1441-9473 Stephen G Schwartz 4 ,
• Elad Moisseiev 3 , 5
• 1 Institute for Research in Ophthalmology , Poznan , Poland
• 2 Ophthalmology , University of Warmia and Mazury , Olsztyn , Poland
• 3 Ophthalmology , Meir Medical Center , Kfar Saba , Israel
• 4 Ophthalmology, Bascom Palmer Eye Institute , University of Miami Miller School of Medicine , Naples , Florida , USA
• 5 Ophthalmology , Sackler School of Medicine, Tel Aviv University , Tel Aviv , Israel
• Correspondence to Dr Stephen G Schwartz; sschwartz2{at}med.miami.edu

The 100 most-cited papers on age-related macular degeneration (AMD) were analysed using a bibliographic study. The bibliographic databases of the Institute for Scientific Information Web of Knowledge were searched, limited to research articles published between 1965 and 2020 in peer-reviewed journals. The papers were ranked in order of number of citations since publication. Five of the top 10 (and 3 of the top 4) papers reported randomised clinical trial results for either anti-vascular endothelial growth factor agents or nutritional supplements. Four of the top 10 papers reported genotype-phenotype associations between AMD and variants in Complement Factor H . This bibliographic study provides perspective and insight into many of the most influential contributions in the understanding and management of AMD and its evolution over time.

• degeneration
• neovascularisation
• treatment lasers

Data availability statement

All data relevant to the study are included in the article or uploaded as online supplemental information. All data relevant to the study are included in the article.

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:  http://creativecommons.org/licenses/by-nc/4.0/ .

https://doi.org/10.1136/bmjophth-2021-000823

Statistics from Altmetric.com

Request permissions.

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Age-related macular degeneration (AMD) remains the leading cause of irreversible visual loss among the elderly in industrialised nations. 1 Despite substantial progress in the diagnosis and treatment of AMD over the past 20 years, there are many unmet clinical needs. 2 AMD is the subject of extensive basic science, translational and clinical research efforts, resulting in thousands of peer-reviewed publications.

Historically, AMD was first described in the second half of the 19th century, after the invention and introduction of the ophthalmoscope. The disorder was known under different names, and these names might have varied between languages. In the first half of the 20th century, it was called among other names, disciform macular degeneration, retinal circinate degeneration (Fuchs’ circinate retinitis), external exudative retinitis, tumour-like tissue proliferation in the macula lutea, senile exudative macular retinitis, senile macular degeneration, central senile chorioretinitis, and serous and haemorrhagic disciform detachment of the macula. In the second half of the 20th century, most researchers used either senile macular degeneration or AMD, although other terms could be found, including senile exudative maculopathy, disciform detachment of the neuroepithelium, senile disciform macular detachment, senile macular choroidal degeneration, age-related maculopathy, age-dependent macular degeneration, age-related macular disease and ageing macular disease/disorder. 3–5 This evolution demonstrates the importance of using a unified terminology for knowledge dissemination and development.

Over the past 30 or so years, the primary diagnostic modality for AMD has evolved from fluorescein angiography to optical coherence tomography (OCT) and OCT angiography, and the primary therapeutic modality for patients with neovascular AMD has evolved from no therapy, to thermal photocoagulation, 6 to photodynamic therapy (PDT), 7 to antivascular endothelial growth factor (anti-VEGF) agents, including chronologically pegaptanib (Macugen), 8 bevacizumab (Avastin), 9 ranibizumab (Lucentis) 10 11 and Eylea (aflibercept) with different treatment protocols. 12 Most patients with at least intermediate AMD are offered nutritional supplementation per the Age-Related Eye Disease Study (AREDS). 13 And there is a growing understanding of the complex genetic risk factors affecting the pathogenesis of the disease. 14–17

One approach to evaluate the impact of an individual scientific article is by the number of subsequent citations. 18 19 Our group has used this approach to report the 100 most-cited papers on vitrectomy, 20 intravitreal injections 21 and retinal detachment. 22 Other investigators have used similar techniques in other areas of ophthalmology. 23–26

In this study, we identified the 100 most-cited articles on AMD over the past 55 years, in order to provide a bibliographic-historic perspective on the evolution of the understanding and management of this disease.

The bibliographic databases of the Institute for Scientific Information (ISI) Web of Knowledge databases were searched with the assistance of an expert medical librarian. The search was performed using the keyword combinations of ‘AMD’ and ‘senile macular degeneration’ that had to appear in the title of the manuscripts. The search included all publications in peer-reviewed journals from 1965 (the earliest year archived in the ISI Web of Knowledge databases) through the date of the search (31 December 2020). The search included all types of publications (original articles, reviews, meta-analyses, case reports, etc) and all available journals and sources, not only those specific to the field of ophthalmology.

The papers were then ranked by the number of total citations since publication. Each paper was reviewed and excluded if not relevant to the topic of AMD. After the list of the 100 most-cited papers was finalised, the following details were recorded for each paper: overall number of citations, mean citations per year since publication, journal name, year of publication, names of first and last authors, number of authors, country of origin (determined by the corresponding author), type of study, number of patients included, and the theme of its main topic. In some cases, the authors listed on the papers in the databases conflicted with the authors listed on the original publications; in these instances, the authors listed in the original publications were used for the present analysis.

Values of the results are presented as mean±SD. Correlations between year of publication and total number of citations and between the year of publication and mean number of citations per year were analysed using Pearson’s correlation coefficient, with a p value of 0.05 used to determine statistical significance. Data was analysed using SPSS for Windows V.20.

Overall, the search yielded 7755 articles. Most of these were published in ophthalmology journals (68.7%), and the rest in fields related to other major fields such as genetics, pharmacology, general medicine and biology. The most common countries of origin were the USA (35.5%), Germany (8.3%), United Kingdom (8.1%), Australia (6.1%), France (4.9%), China (4.7%) and Japan (4.5%).

Research interest in AMD has risen greatly in the past two decades. When looking at the total number of publications per year, an increase from about 200–300 to about 500 has occurred in 2004, and then again to over 1000 in 2013. This is presented in the figure 1 .

• Download figure
• Open in new tab
• Download powerpoint

Distribution of total publications related to AMD by year of publication. AMD, age-related macular degeneration.

The 100 most-cited papers on AMD, according to this methodology, are presented in the online supplemental table . The mean number of total citations was 754±585, with a median of 543 citations and a range of 343 to 3919. Sixty-one of the 100 most-cited papers on AMD were published in ophthalmology journals, and 39 were published in journals from other fields of research.

Supplemental material

The ophthalmology journals, in order of number of papers from the list, were Ophthalmology (n=17), Archives of Ophthalmology (n=11), American Journal of Ophthalmology (n=8), Survey of Ophthalmology (n=7), Investigative Ophthalmology and Visual Science (n=4), Retina (n=4), Progress in Retinal and Eye Research (n=3), Experimental Eye Research (n=3), BMC Ophthalmology (n=1), Molecular Vision (n=1), Eye (n=1) and Ophthalmic Surgery, Lasers and Imaging (n=1). More than half (36) of these papers were published in three leading journals in ophthalmology -Ophthalmology, Archives of Ophthalmology, and American Journal of Ophthalmology . Of note is that more than half of the papers published in other journals (22/38) were published in the most prestigious scientific journals- Science (n=6), New England Journal of Medicine (n=6), Lancet (n=4) and Nature/Nature Genetics (n=7). All papers in the top 100 were published in English.

The 100 most-cited papers on AMD included 75 original articles, 24 reviews and one case report. 9 Of the original articles, 27 (36%) were basic science or animal studies, and 48 (64%) were human studies. These included 30 prospective, 15 retrospective and 3 observational studies. Twenty-two of the articles reported the results of multicentre studies, corresponding with 29.3% of the original articles and 45.8% of the human clinical studies.

The top 100 papers were also analysed for the themes of their main topics. The most common topic was treatment of AMD (n=30), followed by genetics (n=23) and pathology and pathogenesis (n=21). Additional topics included AMD risk factors (n=8), epidemiology (n=7), classifications (n=4), general reviews (n=4), natural history (n=2) and imaging (n=1). Since treatment was the leading topic, the top 100 papers were also analysed for the various treatment methods mentioned. These included: ranibizumab (n=12), bevacizumab (n=5), aflibercept (n=2), pegaptanib (n=2), PDT (n=7), triamcinolone (n=1), and AREDS supplementation (n=5).

The 100 most-cited papers on AMD were published between 1983 and 2016. When further divided by decades, there were 18 papers published up to 1999, 61 papers published between 2000 and 2009, and 21 papers published after 2010. There was no correlation between year of publication and total number of citations (p=0.42), but a significant correlation was found between the year of publication and mean number of citations per year (p<0.001), with later publications on average had significantly higher citations per year.

The 100 most-cited papers in AMD illustrate the evolution of the diagnosis and treatment of the disease over the past three decades. In addition, they reflect the growing understanding of the role of genetics in the pathogenesis of the disease.

The #1, #3, #7, and #15 papers report, respectively, phase III randomised clinical trial (RCT) results for the anti-VEGF agents ranibizumab (Lucentis), 10 11 pegaptanib (Macugen), 8 and aflibercept (Eylea). 12 The #2, #5, #6, and #10 papers are the four original reports that variants in Complement Factor H ( CFH ) associate with clinical AMD. 14–17 The #4 paper reports the RCT for the original AREDS supplements. 13 The #8 paper is the Comparison of AMD Treatments Trials (CATT) report of an RCT comparing ranibizumab to bevacizumab (Avastin). 27 The #9, #11, and #14 papers are landmark epidemiological reports. 1 28 29 The #12 paper reviewed oxidative stress in disease pathogenesis. 30 The #13 paper is the initial RCT for PDT. 7 Collectively, these top 15 papers illustrate most of the major advances in AMD diagnosis, treatment, and understanding in the past 30 years. Interestingly, only 6 of these 15 papers (and only 2 of the top 10) were published in ophthalmology journals.

Bevacizumab, an off-label medication which never underwent a major phase III RCT, appears to be underrepresented in the top 15 papers. The first report of bevacizumab was in a single patient; this was the #29 paper and the only case report in the top 100 list. 9 Subsequent retrospective series, such as the #20 and #49 papers, 31 32 are also included on the list.

When viewed chronologically, there were about 200–300 papers published per year from 1997 through 2003. There is a modest but appreciable increase starting in 2004, the year pegaptanib received US Food and Drug Administration (FDA) approval. Similarly, there is a more substantial increase in 2013, the year aflibercept received US FDA approval.

The 100 most-cited AMD papers include nine papers authored by a study group with no individual authors named, plus two papers in which the study group is named as the first author, followed by individual investigators. These include the AREDS Research Group (#4 and #46), 13 33 the CATT Research Group (#8), 27 the Treatment of AMD With PDT Study Group (#13 and #18), 7 34 the Verteporfin in Photodynamic Therapy Study Group (#28), 35 the AREDS 2 Research Group (#51), 36 the Eye Disease Case-Control Study Group (#62), 37 and the Macular Photocoagulation Study Group (#68). 6 The two papers which listed the study group first, followed by individual investigators, were authored by the CATT Research Group (#16) 38 and the UK Inhibition of VEGF in Age-Related Choroidal Neovascularization Study Group (#55). 39

In terms of nomenclature, 98 of the top 100 papers exclusively used the term ‘AMD’. The only exceptions were the #66 paper (published in 1983, which used ‘senile macular degeneration’) 40 and the #11 paper (published in 1995, which used both ‘age-related maculopathy’ and ‘AMD’). 29

The limitations of this study are similar to the limitations of previous works using this methodology. The most-cited papers are not necessarily the most scientifically important or clinically relevant. For example, papers #1, 10 #3, 11 #4, 13 and #8 27 report RCT results for, respectively, ranibizumab, AREDS supplements, and bevacizumab; these interventions remain highly clinically relevant. Papers #2, 14 #5, 15 #6 16 and #10 17 report the genotype–phenotype association with variants in CFH ; this is scientifically important and has stimulated much further research but as yet has no clinical applicability. On the other hand, paper #7 8 reports the RCT results for pegaptanib, which is no longer available in the USA; and paper #9 28 reports prevalence rates from 2004 which are now outdated.

Regardless of these limitations, these 100 papers are among the most infuential in this field of study.

Ethics statements

Patient consent for publication.

Not required.

• Li X , et al
• Fleckenstein M ,
• Keenan TDL ,
• de Jong PTVM
• De Jong PTVM
• Macular Photocoagulation Study Group
• Treatment of Age-Related Macular Degeneration with Photodynamic Therapy (TAP) Study Group
• Gragoudas ES ,
• Adamis AP ,
• Cunningham ET , et al
• Rosenfeld PJ ,
• Moshfeghi AA ,
• Puliafito CA
• Heier JS , et al
• Kaiser PK ,
• Michels M , et al
• Chong V , et al
• Age-Related Eye Disease Study Research Group
• Chew EY , et al
• Edwards AO ,
• Abel KJ , et al
• Haines JL ,
• Hauser MA ,
• Schmidt S , et al
• Hageman GS ,
• Anderson DH ,
• Johnson LV , et al
• Grzybowski A
• Grzybowski A ,
• Krauthammer M ,
• Moisseiev E
• Shtayer C ,
• Schwartz SG , et al
• Isashiki Y , et al
• Sabanayagam C
• Nusinovici S , et al
• CATT Research Group ,
• Martin DF ,
• Maguire MG , et al
• Friedman DS ,
• O'Colmain BJ ,
• Muñoz B , et al
• Bressler NM ,
• Bressler SB , et al
• Phil M , et al
• Pieramici DJ ,
• Rabena MD , et al
• Spaide RF ,
• Fine HF , et al
• Verteporfin in Photodynamic Therapy Study Group
• Age-Related Eye Disease Study 2 Research Group
• The Eye Disease Case-Control Study Group
• Comparison of Age-related Macular Degeneration Treatments Trials (CATT) Research Group ,
• IVAN Study Investigators ,
• Chakravarthy U ,
• Harding SP , et al
• Lilienfeld AM ,
• Ferris FL , et al

Supplementary material

Supplementary data.

This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

• Data supplement 1

Funding Partially supported by NIH Centre Core Grant P30EY014801 and an Unrestricted Grant from Research to Prevent Blindness to the University of Miami.

Disclaimer The sponsor or funding organisation had no role in the design or conduct of this research.

Competing interests None declared.

Provenance and peer review Not commissioned; externally peer reviewed.

Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.

Read the full text or download the PDF:

Integrating Multi-omics to Identify Age-Related Macular Degeneration Subtypes and Biomarkers

• Open access
• Published: 07 August 2024
• Volume 74 , article number  74 , ( 2024 )

Cite this article

You have full access to this open access article

• Shenglai Zhang 1 ,
• Ying Yang 1 ,
• Jia Chen 1 ,
• Xiaowei Yang 1 &
• Aimin Sang 1  

149 Accesses

Explore all metrics

Age-related macular degeneration (AMD) is one of the most common causes of irreversible vision loss in the elderly. Its pathogenesis is likely multifactorial, involving a complex interaction of metabolic and environmental factors, and remains poorly understood. Previous studies have shown that mitochondrial dysfunction and oxidative stress play a crucial role in the development of AMD. Oxidative damage to the retinal pigment epithelium (RPE) has been identified as one of the major mediators in the pathogenesis of age-related macular degeneration (AMD). Therefore, this article combines transcriptome sequencing (RNA-seq) and single-cell sequencing (scRNA-seq) data to explore the role of mitochondria-related genes (MRGs) in AMD. Firstly, differential expression analysis was performed on the raw RNA-seq data. The intersection of differentially expressed genes (DEGs) and MRGs was performed. This paper proposes a deep subspace nonnegative matrix factorization (DS-NMF) algorithm to perform a multi-layer nonlinear transformation on the intersection of gene expression profiles corresponding to AMD samples. The age of AMD patients is used as prior information at the network’s top level to change the data distribution. The classification is based on reconstructed data with altered distribution. The types obtained significantly differ in scores of multiple immune-related pathways and immune cell infiltration abundance. Secondly, an optimal AMD diagnosis model was constructed using multiple machine learning algorithms for external and qRT-PCR verification. Finally, ten potential therapeutic drugs for AMD were identified based on cMAP analysis. The AMD subtypes identified in this article and the diagnostic model constructed can provide a reference for treating AMD and discovering new drug targets.

Similar content being viewed by others

The identification of signature genes and their relationship with immune cell infiltration in age-related macular degeneration

Single-cell transcriptomic atlas of the human retina identifies cell types associated with age-related macular degeneration

Identification of pathogenic genes and upstream regulators in age-related macular degeneration.

Avoid common mistakes on your manuscript.

Introduction

Age-related macular degeneration (AMD) is a chorioretinal disease closely related to age. Pathologically, the main manifestations are aging changes in the structure of the macular area and a decrease in the phagocytosis and digestion function of the retinal pigment epithelial cells on the outer disc membrane of the visual cells. Further features that increase the number and diameter of extracellular retinal deposits are called drusen (Lim et al. 2012 ). As the population ages, AMD is the leading cause of blindness in people over 50 years old worldwide (Newman et al. 2012 ). Genetic factors play an essential role in the pathogenesis of AMD, and multiple genetic variants have been associated with the risk of AMD. In a recent study, Qiao et al. identified multiple genetic susceptibility loci (including Lama5, Mtg2, Col9A3) through genome-wide association studies (GWAS) and whole-exome sequencing in older Asian people (Fan et al. 2023 ).

The RPE is particularly susceptible to oxidative damage because it is extremely metabolically active, highly oxidative, and exposed to photosensitizers such as the age pigment lipofuscin. This sensitivity leads to various age-related changes, ultimately leading to reduced RPE function and increased susceptibility to cell death. Oxidative stress is a recognized risk factor for AMD, in which changes in areas of focal loss of the RPE lead to photoreceptor degeneration and central vision loss (Jarrett and Boulton 2012 ). Increased mitochondrial damage and reactive oxygen species (ROS) production are associated with AMD, suggesting that damaged mitochondria and other oxidatively modified components are not efficiently removed by aging RPE cells (Karunadharma et al. 2010 ).

As a vital organ within cells, mitochondria are responsible for the energy cells, which are required and participate in biological processes such as cell metabolism and redox reactions (Kaarniranta and Salminen 2009 ). Therefore, mitochondria-related genes (MRGs) may play a vital role in the occurrence and development of AMD. Firstly, oxidative stress and mitochondrial dysfunction may accelerate the development of AMD (Beatty et al. 2000 ). Patients with AMD are often accompanied by increased oxidative stress, which may lead to oxidative damage to mitochondrial DNA, leading to mitochondrial dysfunction. Impaired mitochondrial function may further aggravate oxidative stress, forming a vicious cycle. Secondly, inflammation and immune response also impact the pathological process of AMD (Ambati et al. 2003 ). Abnormal mitochondrial function and oxidative stress can trigger immune system responses, leading to chronic inflammation in AMD. Finally, mitochondrial dysfunction may lead to insufficient intracellular energy supply, triggering apoptosis or necrosis (Khandhadia and Lotery 2010 ). These cell death processes can lead to cell loss in the macular area of AMD, further exacerbating chorioretinal damage.

Previous studies have utilized various machine learning methods to identify AMD diagnostic-related genes and construct AMD diagnostic models. Wang et al. constructed AMD diagnostic models based on DNA methylation and gene expression data using random forest models (Wang et al. 2021 ). Han et al. identified key modules and modular genes most relevant to AMD through weighted gene co-expression network analysis. They employed random forest, support vector machine, Xgboost, and GLM models to select predictive genes and build an AMD clinical prediction model (Han and He 2023 ). Additionally, Han et al. integrated weighted gene co-expression network analysis and differential expression to pinpoint genes intricately associated with Tfh cells. Using the MCODE function in Cytoscape software, they screened these genes and identified key diagnostic genes using the LASSO algorithm (Yang, et al. 2023 ).

Figure  1 shows the technical roadmap of this article. This paper aims to deeply explore the role of MRGs in AMD through bioinformatics analysis and experimental verification. We obtained MRGs from previous literature (Pei et al. 2023 ), and their expression levels were extracted from transcriptome data for differential expression analysis. AMD subtypes were then identified based on the deep subspace nonnegative matrix factorization algorithm (DS-NMF). This method can use the age of AMD patients as prior information, thereby changing the original data distribution so that patients of different age groups are distributed far away from each other. Furthermore, multiple machine learning algorithms were used to identify hub genes and construct a diagnostic model for AMD. Finally, the different patterns of multiple immune cells in trajectory analysis and cell communication were explored through AMD’s single-cell sequencing (scRNA-seq) data. This research is expected to provide a new theoretical basis for treating AMD and provide more critical insights into MRGs for future biomedical research.

The technical roadmap of the paper

Acquisition of Data Sets

This article downloaded two macular degeneration RNA-seq data sets (GSE29801 (Newman et al. 2012 ) and GSE135092 (Jones et al. 2023 ; Orozco et al. 2020 )) from the Gene Expression Omnibus (GEO) database. The GSE29801 data set is used as an internal data set containing 151 normal samples and 142 AMD samples. The GSE135092 data set is used as an external data set, which includes 50 normal samples and 50 AMD samples. scRNA-seq data sets of two AMD samples were collected from the GSE210543 data set.

Differential Expression Analysis and GO Enrichment Analysis

Differentially expressed genes (DEGs) between the control and diseased groups were identified based on the limma package (Ritchie et al. 2015 ). The parameters are set as follows: the absolute value of logFC is more than 0.25, and the p -value is less than 0.05. Gene Ontology (GO) enrichment analysis was performed on DEGs based on the R package “clusterProfiler” (Yu et al. 2012 ).

Lim et al. 2012 ) Ritchie et al. ( 2015 ). limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic acids research, 43(7), e47. https://doi.org/ https://doi.org/10.1093/nar/gkv007

Newman et al. 2012 ) Yu et al. ( 2012 ). clusterProfiler: an R package for comparing biological themes among gene clusters. Omics: a journal of integrative biology, 16(5), 284–287. https://doi.org/ https://doi.org/10.1089/omi.2011.0118

DS-NMF Algorithm

The DS-NMF algorithm consists of two parts: deep subspace reconstruction and NMF algorithm. Given data points \({\left\{{x}_{i}\right\}}_{i=1,\cdots ,N}\) extracted from multiple linear subspaces \({\left\{{R}_{i}\right\}}_{i=1,\cdots ,K}\) , specific points within a particular subspace can be represented as linear combinations of other points within the same subspace. This property is known as self-expression. For a data matrix \(X\) , the self-expression property can be formulated as \(X=\text{C}X\) , where \(C\) is the coefficient matrix capturing the linear relationships. Under certain permutation conditions, \(C\) should exhibit a block diagonal structure, with each block corresponding to samples from the same subspace (Fan et al. 2008 ).

This method differs notably from autoencoders. Self-expression networks focus on discovering subspace structures and clustering information in data through the self-expression property. They optimize the coefficient matrix WWW by leveraging linear combinations and sparse regularization to uncover relationships among data points. In contrast, autoencoders emphasize nonlinear dimensionality reduction and reconstruction of data. They achieve this through encoder and decoder networks for nonlinear mapping and reconstruction. In summary, while both self-expression networks and autoencoders are used for data representation and reconstruction, self-expression networks are more concerned with subspace structures and clustering in data, while autoencoders prioritize nonlinear dimensionality reduction and reconstruction performance.

In the deep subspace reconstruction part, the original input is reconstructed using the self-expression properties of the data. Define the gene expression matrix \(X\in {R}^{N*p}\) . \(N\) represents the number of genes. \(p\) represents the sample size. Let \(X=\left[{x}_{1},{x}_{2},\cdots ,{x}_{\text{Nz}}\right]\) , where \({x}_{1}\) is the first sample of the first label, and \(z\) represents the total number of categories. \({x}_{\text{Nz}}\) is the last sample of the \({z}_{\text{th}}\) label.

First, the original data is put into a multi-layer feedforward neural network, and the nonlinearly transformed matrix \({H}_{i}^{M}\) is output in the output layer. \(M\) indicates the total number of layers on the network. Iterative calculations are performed on the top \(M+1\) layer to achieve subspace reconstruction. Below is the definition of network parameters.

Among them, \(m=\text{1,2},,...,\) M indicates the number of network layers. \({W}^{(m)}\) and \({b}^{(m)}\) represent the weight and bias of the \(m-th\) layer network respectively. \({H}^{(M)}=\left[{h}_{1}^{(M)},{h}_{2}^{(M)},\cdots ,{h}_{{N}_{z}}^{\left(M\right)}\right]\) , where \({h}_{{N}_{z}}^{\left(M\right)}\) represents the \(N\) sample of the \({\text{z}}_{th}\) class after undergoing a nonlinear transformation by the multi-layer neural network. Finally, the objective function of the subspace reconstruction algorithm of the \({z}_{\text{th}}\) class samples is:

where \({c}_{l}\) represents the vector of self-expression coefficients of layer l and \({||\cdot ||}_{\text{F}}\) represents the Frobenius norm. \({h}_{l}^{(M)}\) is the output of the \(l-th\) feature of the top-level network after nonlinear transformation. \({H}^{(M)}\) is the output of the original data. The expression of \({h}_{l}^{(m)}\) is as follows:

where G(·) indicates the sigmoid activation function. The specific definitions are as follows.

The following equation can be obtained after calculating the partial derivatives of Eq.  2 and sorting it out (for the specific derivation process, see Supplementary material 1.1).

The self-expression coefficient matrix for all samples is defined as follows.

Finally, the reconstructed data can be expressed as the following formula.

The variable \(\widetilde{X}\) represents the data after deep subspace reconstruction. Leveraging neural networks to uncover nonlinear features in the data enables the incorporation of existing clinical information to reflect grouping information that reflects similar data structures. Incorporating prior information into the data as input to the NMF algorithm can enhance the algorithm’s performance to a certain extent. The NMF algorithm is a low-rank decomposition algorithm. This paper clusters the reconstructed data based on the NMF algorithm. The objective function of the NMF algorithm is given below. In Supplementary Material 1.2, we provide the parameters related to the neural network component in the DS-NMF algorithm. In Supplementary Material 1.3, the optimization and solving process of the algorithm is detailed.

Among them, \(W\in {R}^{S*K}\) and \(H\in {R}^{K*M}\) are the basis and coefficient matrices, respectively. \(W\) and \(H\) need to be guaranteed to be nonnegative. The sum needs to be guaranteed to be nonnegative. \(K\) is the number of clusters. Spectral clustering is performed on \(W\) to obtain the final clustering result.

Analysis of Gene Set Variation Between Subtypes and Immunoassays

This article implements gene set variation analysis of different subtypes based on the R package “GSVA” (Hänzelmann et al. 2013 ). Based on the “c2.cp.kegg.v7.5.1.symbols.gmt” reference gene set, multiple Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways with significantly different scores between subtypes were identified. The infiltration abundance of 22 immune cells in different subtypes was evaluated using the “CIBERSORT” algorithm (Chen et al. 2018 ). In addition, differences in the expression of immune checkpoints and HLA genes among different subtypes were evaluated. The above analyses all used the Wilcoxon method to compare differences.

Diagnostic Model Construction Methods

This paper implements the construction of diagnostic models through random forest (RF), support vector machine recursive feature elimination (SVM-RFE), K nearest neighbor (KNN), and adaptive boosting (Adaboost) algorithms. Python’s scikit-learn package (Pedregosa, et al. 2011 ) implements RF, KNN, and Adaboost. SVM-RFE is implemented by the R package “e1071.” After the internal data set was randomly divided into the training set and the test set at a ratio of 7:3, a tenfold cross-validation method was used on the training set to obtain the gene set with the highest accuracy in the test set by selecting the top features and perform AUC verification. In the RF algorithm, criteria are set to entropy, and n_estimators are set to 500. n_neighbors is set to 3 in the KNN algorithm. n_estimators is set to 500 in the Adaboost algorithm. The random seed of the SVM-RFE algorithm is 13,579. The remaining parameters of all algorithms involved are default parameters.

Nomogram Model Construction

This paper builds a nomogram model based on the R package “rms” and uses diagnostic genes. The construction effect of the nomogram model was evaluated through calibration curves. Decision curve analysis (DCA) is implemented based on the decision_curve function in the R package “rmda.” Clinical impact curves were also plotted to predict high-risk probability stratification for a population of 1000.

Analysis Methods of scRNA-seq Data

This article is based on the CreateSeuratObject function of the R package “Seurat” (Hao et al. 2024 ) to convert the original count matrix into a format readable by the Seurat package. In the quality control process, cells that meet the requirements of nFeature_RNA greater than 200 and less than 7000, nCount_RNA less than 5000, mitochondrial gene proportion less than 3%, and red blood cell proportion less than 0.2% are retained based on the subset function. The LogNormalize method based on the NormalizeData function implements standardization in the standardization process. The first 20 principal components are selected for clustering based on the RunPCA function in the dimensionality reduction and clustering process. After setting the resolution to 1, 21 cell clusters are obtained based on the FindClusters function. The removal of double cells is realized based on the R package “DoubletFinder” (Stoeckius et al. 2018 ). Subsequently, cell type annotation was implemented based on the R package “singleR” (Aran et al. 2019 ). Cell trajectory analysis is implemented based on the R package “monocle.” Cell communication analysis is implemented based on the “CellChat” (Jin et al. 2021 ) package.

Connectivity Map (cMAP) Analysis

The cMAP database ( https://clue.io/ ) can explore associations between diseases, genes, and small-molecule compounds based on gene expression. The dysregulated genes from the differential analysis were entered into the cMAP database to identify potential small-molecule drugs for AMD treatment. Potential small-molecule compounds are entered into the Pubchem database ( https://pubchem.ncbi.nlm.nih.gov/ ) to obtain the compound’s molecular structure.

Cell Cultures and Treatment

Human retinal pigment epithelial cells (ARPE-19) were purchased as frozen vials from Shanghai EK-Bioscience Biotechnology Co., Ltd. (Shanghai, China), and all cell experiments were performed between the third and fifth generations. The cells were cultured in DMEM/F-12 (supplemented with 10% FBS, 1% streptomycin/penicillin) at 37 °C in a humidified atmosphere containing 95% air and 5% CO 2 . Cells at 80–90% confluence were selected for subculture and subsequent experimentation. For H 2 O 2 -induced oxidative damage studies, the cells were treated with a serum-free medium containing various concentrations of H 2 O 2 (400 μM) for 24 h.

Real-Time Quantitative Polymerase Chain Reaction (qRT-PCR)

Total RNA was extracted from RPE cells using Trizol reagent (Invitrogen, USA). Total RNA was reverse transcribed into cDNA using HiScript II Q Select RT SuperMix for qPCR (Vazyme, China). qPCR was performed using SYBR green reagent (Vazyme, China) on Roche 96 (Roche, USA). The gene expression level was quantified using the 2 −ΔΔCt method. GAPDH (glyceraldehyde 3-phosphate dehydrogenase) was used as an internal control gene. Analysis of each sample was performed in triplicate. Primer sequences are listed in Supplement Table  1 . Statistical analysis was performed in GraphPad Prism software. Statistical difference between groups was assessed by Student’s t -test. p  < 0.05 was considered statistically significant.

Acquisition and Analysis of DEMRGs

First, this article conducts differential expression analysis on samples from the normal and diseased groups in the internal data set. A total of 528 DEGs were obtained. Figure  2 A shows the volcano plot obtained from differential expression analysis. The differentially expressed genes are in the gene diff.xls file in the Supplementary Material. Figure  3 C is a bar graph obtained by GO enrichment analysis of DEGs. We will discuss in part an analysis of the role of these pathways in the development of AMD. Subsequently, 31 intersection genes were obtained from the intersection of MRGs and DEGs collected from previous literature (Chang et al. 2023 ) (Fig.  2 B–E).

Results of differential expression analysis. A The volcano plot obtained by differential expression analysis. B A bar graph of GO enrichment analysis of DEGs. C The Venn diagram of the intersection of DEGs and mitochondria-related genes. D A box plot of intersection gene expression in normal and diseased groups. E A circle diagram of the chromosomal location of intersection genes

Clustering results based on DS-NMF algorithm. A , B PCA analysis scatter plots obtained before and after reconstruction using the deep subspace reconstruction algorithm. C A line chart of changes in cophenetic, dispersion, evar, residuals, rss, silhouette, and sparseness of the DS-NMF algorithm as the number of clusters increases. D The consensus matrix obtained by the DS-NMF algorithm when the number of clusters is 2. E A PCA analysis scatter plot obtained by clustering with the DS-NMF algorithm

Identification of Subtypes of Age-Related Macular Degeneration

To use age as prior information to change the distribution of the original data, this article first reconstructs AMD samples based on the DS-NMF algorithm. The input data are the expression profiles of 31 intersection genes and two group labels of AMD divided by the median age. This method performs multi-layer nonlinear mapping of gene expression profiles through multi-layer feedforward neural networks. It is then reconstructed on top of the network. During reconstruction, AMD samples were divided into two groups according to the median age. Figure  3 A and B show the distribution changes of the original data before and after reconstruction. Labels 1 and 2 represent samples younger and older than the median age of all AMD samples, respectively. The NMF algorithm was performed on the reconstructed data. The final number of clusters was set to 2 according to cophenetic (Fig.  3 C). Figure  3 D shows the consensus matrix with a cluster number of 2. Figure  3 E shows the final clustering results for AMD samples. The two subtypes were effectively distinguished. In addition, visualization results of different subtypes of samples based on t-SNE dimensionality reduction are presented in Supplementary materials Figure S3 .

To explore the differences between the two isoforms in terms of enrichment pathways and immune landscapes, differential expression analysis of the two isoforms was first performed (Fig.  4 A). Subsequent GSVA analysis identified multiple pathways with significantly different scores between the two subtypes (Fig.  4 B). We will analyze the biological significance of these pathways in the “Discussion” section. Furthermore, the CIBERSORT algorithm was used to evaluate the difference in infiltration abundance of 22 immune cells between the two subtypes. The infiltrating abundance of most immune cells in the two subtypes was significantly different (Fig.  4 C). Finally, immune checkpoints and HLA genes were collected from previous literature and found that the expression of most genes was significantly different in the two groups (Fig.  4 D, E ) (Liu et al. 2024 ). The results of enrichment and immunoassays confirmed the reliability of AMD subtype identification. Furthermore, in Supplementary Material Sect. 1.4, we present the AMD clustering results obtained using other clustering algorithms and highlight the differences in immune infiltration abundance across subtypes. Additionally, we included the GSVA results of the two baseline clustering algorithms in Supplementary Material 1.4, further confirming the biological significance of the subtypes identified by the DS-NMF algorithm.

Difference analysis, enrichment analysis, and immune analysis between subtypes. A The top 20 DEGs between the two subtypes. B The differential expression heat map of pathway scores obtained by GSVA analysis of the two subtypes. C A box plot of the difference in infiltration abundance of immune cells between different types based on CIBERSORT analysis. D , E Box plots of differential expression of immune checkpoints and HLA genes in different subtypes, respectively

Results of Correlation Analysis Between Diagnostic Genes and Immune Cells

In order to explore the diagnostic value of mRGs in the macula, this article constructed a diagnostic model based on the expression profiles of 31 intersection genes and using multiple machine learning algorithms. The diagnostic model can effectively classify AMD and control group. Specifically, this article selects different numbers of top features to build diagnostic models based on RF, Adaboost, KNN, and SVM-RFE algorithms. The RF, Adaboost, KNN, and SVM-RFE algorithms achieved the maximum accuracy in the internal test set when selecting the first 9, 26, 14, and 13 features, respectively (Fig.  5 A, C , E , and G ). Figure  5 B, D , F , and H are the AUCs of the four algorithms on the internal test set, respectively. Among them, the SVM-RFE algorithm reached the largest AUC (0.806). Therefore, we further tested the diagnostic model constructed on the external test set, and its AUC was 0.816 (F i g.  5 I). In addition, we give the maximum AUC of the deep neural network and the Rogeist regression algorithm in the supplementary material Figure S2 .

A diagnostic model based on multiple machine learning algorithms. A , C , E , G The accuracy change curves of the diagnostic model built based on RF, Adaboost, KNN, and SVM-RFE algorithms, respectively, after selecting different top features. B , D , F , H The ROC curves of the diagnostic models constructed by the four methods respectively. I The ROC curve used to verify the diagnostic model built by SVM on the external test set

Furthermore, the columnar line chart model is a graphical risk prediction tool that integrates multiple predictive factors into a single predictive model, providing intuitive and clinically accessible prediction outcomes. In our study, the columnar line chart model was used for AMD risk assessment and personalized prediction. We constructed the columnar line chart model based on the first 13 diagnostic genes (CYP27B1, FUS, FZD5, GLS2, GLYATL1, HSPA1A, NDUFA4L2, PDPN, SLC2A1, SNN, SPR, GSTZ1 and TRAF6) (Fig.  6 A). Figure  6 B gives the calibration curve of the nomogram model. The nomogram model based on 13 diagnostic genes agreed with the ideal model. DCA analysis showed that although both the nomogram model and individual diagnostic genes produced net benefits, the net usage of the nomogram model was significantly greater than that of individual diagnostic genes. This suggests that nomogram models may have more clinical value than individual diagnostic genes (Fig.  6 C). Clinical impact curve analysis showed that the nomogram model had high diagnostic ability (Fig.  6 D). In addition, to explore the correlation between diagnostic genes and immune cell infiltration abundance, this article screened the correlation results with p  < 0.0001 based on the Spearman correlation coefficient of the two. Figure  7 A–U show each gene and its two most strongly correlated immune cells. The remaining correlation analysis results are shown in Supplementary material Figures S4 - S6 .

Construction of nomogram model and DCA analysis results. A A nomogram model built based on diagnostic genes. B The calibration curve of the nomogram model. C DCA analysis. D The clinical decision curve

Scatter plot of correlation between diagnostic genes and immune cell infiltration abundance. A , B Correlation scatter plots between CYP27B1 and the two cells with the strongest correlation. C , D Correlation scatter plots between FUS and the two cells with the strongest correlation. E , F Correlation scatter plots between FZD5 and the two cells with the strongest correlation. G , H Correlation scatter plots between GLS2 and the two cells with the strongest correlation. I , J Correlation scatter plots between GLYATL1 and the two cells with the strongest correlation. K A scatter plot of the correlation between HSPA1A and macrophage M1. L A scatter plot of the correlation between NDUFA4L2 and T cells follicular helper. M , N The correlation scatter plot between PDPN and the two cells with the strongest correlation. O , P The correlation scatter plot between SLC2A1 and the two cells with the strongest correlation. Q , R A correlation scatter plot between SNN and the two cells with the strongest correlation. S , T A correlation scatter plot between SPR and the two cells with the strongest correlation. U A scatter plot of the correlation between TRAF6 and T cells follicular helper

Analysis Results of scRNA-seq Data

Immune infiltration analysis found that the infiltration abundance of various immune cells was significantly different between AMD and normal groups. This article further explores the interaction between various immune cells based on scRNA-seq data. Specifically, this article performs quality control, standardization, scaling, dimensionality reduction, clustering, and cell type identification on scRNA-seq of two AMD samples. Figure  8 A and B show violin plots of critical indicators before and after quality control. Figure  8 C shows the score heat map for identifying cell types based on the singleR algorithm. Figure  8 D offers two-dimensional plan views of different types of cells after nonlinear dimensionality reduction using uniform manifold approximation and projection (UMAP). A total of nine cell types (chondrocytes, CMP, endothelial cells, tissue stem cells, neurons, T cells, monocytes, NK cells, and fibroblasts) were identified in this article. Figure 9 A and B show the expression landscape of some hub genes in immune cells (T cells, NK cells, and monocytes). Among them, FUS is highly expressed in NK cells and T cells. HSPA1A and SNN are highly expressed in monocytes.

scRNA-seq data preprocessing and cell type annotation results. A , B Violin plots of nFeature_RNA, nCount_RNA, mitochondrial gene proportion (percent.mt), and red blood cell gene proportion (percent.HB) before and after quality control respectively. C A scoring heat map for cell type scoring of different cell clusters based on the singleR algorithm. D A two-dimensional plan view of cell types annotated after nonlinear dimensionality reduction of scRNA-seq based on the UMAP algorithm.

The expression landscape of some hub genes in immune cells (T cells, NK cells, and monocytes) is shown. Among them, FUS is highly expressed in NK cells and T cells. HSPA1A and SNN are highly expressed in monocytes

Furthermore, this article conducts a pseudo-chronological analysis of different types of cells after annotation (Fig. 10 A, B). Monocytes, T cells, and chondrocytes are in the early stages of differentiation. NK cells differentiate last. The remaining cells are located at multiple stages of differentiation. Cell communication analysis revealed that chondrocytes had stronger signaling than other cell types (Fig. 10 C, D). We found multiple significant pathways using each immune cell group (T cells, monocytes, and NK cells) as source and target, respectively. Then, we conducted communication analysis with other cell groups (Fig. 10 E, F). We will explore the role of these pathways in the development and progression of AMD in detail in the “Discussion” section.

Results of cell trajectory analysis and communication analysis between immune cells. A , B The results of grouping cells according to pseudo-chronological order and cell type respectively. C , D Network diagrams of the number and intensity of signaling pathways in the communication process between immune cells respectively. The size of the nodes in the graph reflects the number of cells of this type. The thickness of the line reflects the amount/strength of communication between cells. E , F Bubble diagrams of the pathways in which each immune cell acts as a source and target and communicates with other cells respectively

Potential Drug Identification

To identify potential small-molecule drugs that could treat AMD patients, we imported the top 150 upregulated DEGs and the top 150 downregulated DEGs into the cMAP database. The results showed that the top ten highest-scoring compounds included triamterene, LE-135, eflornithine, ellipticine, L-690330, IOX2, SAL-1, NF-449, miglitol, and SR-57227A, which are potential therapeutic candidates for AMD patients but have not yet been validated by existing literature (Fig. 11 A-J).

Screening of potential small-molecule compounds for AMD by cMAP analysis. A – J Eflornithine, ellipticine, IOX2, L-690330, LE-135, miglitol, NF-449, SAL-1, SR-57227A, and triamterene

Validation Results of the Hub Genes by qRT-PCR

To further verify the results of bioinformatics analysis, the mRNA levels of the 13 hub genes were determined with qRT-PCR. As illustrated in Fig .12 , the SNN, PDPN, GLYATL1, CYP27B1, GLS2, NDUFA4L2, FUS, and SLC2A1 were significantly downregulated in H 2 O 2 -treated ARPE‐19 cells compared to normal cells (all  p  < 0.05), while the HSPA1A and TRAF6 were significantly upregulated in H 2 O 2 -induced ARPE‐19 cells (all  p  < 0.05), as predicted by the bioinformatics analysis. Although the expression of FZD5, GSTZ1, and SPR showed no statistical difference between H 2 O 2 -treated cells and normal cells, their expression trend was consistent with bioinformatics analysis.

Quantitative reverse transcription-polymerase chain reaction (qRT-PCR) for the expression of the hub genes in ARPE-19 cells of oxidative damage and the controls. Expression of hub genes was normalized against GAPDH expression ( p  < 0.05). *** p  < 0.001, ** p  < 0.01, * p  < 0.05

From the bioinformatics perspective, this article explores the role of MRGs and the signaling pathways involved in the progression of AMD through RNA-seq, scRNA-seq, and other data. Firstly, this paper intersects the DEGs of the normal group and the AMD group. After conducting GO enrichment analysis on 31 intersection genes, it was found that some top pathways have been confirmed to play a critical role in AMD. Eszter Emri conducted a combined transcriptome, proteome, and secretome analysis from three genetically distinct human donors and found that AMD samples were involved in the unique pathway of the extracellular matrix (Emri, et al. 2020 ). Zhao et al. detected 44 and 53 significantly different metabolites in positive and negative ion modes in the AMD and control groups, respectively (Zhao et al. 2023 ). Retinal ganglion cell (RGC) death is the leading cause of AMD. The study by Zhong et al. found that K + channels, including ether-à-go-go (Eag), may contribute to dendritic repolarization during excitatory postsynaptic potentials and the attenuation of action potential backpropagation and protect RGCs (Zhong et al. 2013 ).

Secondly, this paper proposes a DS-NMF algorithm. This method reconstructs the input gene expression profile using the traditional NMF algorithm. Specifically, the original gene expression profile is passed through a multi-layer feedforward neural network during reconstruction. The sample age is then used as prior information at the network’s top level to change the data distribution. Finally, the reconstructed data is used as the input of the NMF algorithm. Based on the DS-NMF algorithm, this article identified two subtypes with apparent differences in enrichment pathways and immune cell infiltration. Most of the pathways with significant differences between the two subtypes have been confirmed to play a critical role in the progression of AMD. Research by Santos et al. revealed that complement and coagulation components and adhesion factors are differential biomarkers for vitreoretinal eye diseases, including AMD (Santos et al. 2023 ). Subramanian et al. found that OCT biomarkers were associated with visual impairment and vitreomacular adhesion in patients with diabetic macular edema (Subramanian et al. 2023 ). Chen et al. found that fenofibrate inhibited subretinal fibrosis by inhibiting TGF-β-Smad2/3 signaling and Wnt signaling in neovascular AMD (Chen et al. 2020 ).

This article builds a robust diagnostic model based on various machine learning algorithms to assist the clinical diagnosis of AMD. Among them, the AUC of the diagnostic model built using the SVM-RFE algorithm reached 0.806 and 0.816 in the internal and external test sets, respectively. We screened a total of 13 diagnostic genes (CYP27B1, FUS, FZD5, GLS2, GLYATL1, HSPA1A, NDUFA4L2, PDPN, SLC2A1, SNN, SPR, GSTZ1 and TRAF6). Some genes have been confirmed to be closely related to AMD. McKay and others found that two SNPs in the CYP27B1 gene are associated with early AMD (McKay et al. 2017 ). The findings of Choudhury et al. suggest that the interaction between HSPA1A and FHL-1 may impact AMD. This may mean that the expression and function of HSPA1A may be related to the onset and progression of AMD (Choudhury et al. 2021 ). In a multicenter cohort association study of SLC2A1 single nucleotide polymorphisms and AMD, Baas et al. found population-dependent genetic risk heterogeneity in AMD (Baas et al. 2012 ). Choroidal neovascularization (CNV) is a form of wet AMD. Ding et al. found that inhibiting TRAF6 can alleviate choroidal neovascularization in vivo (Ding et al. 2018 ).

Since various immune cells are significantly related to diagnostic genes, this article uses scRNA-seq data to analyze the differences in pseudo-chronology and communication of different immune cells. This article discovered the significant signaling pathways conducted through cell communication analysis when three immune cells communicate with other cells. The retinal pigment epithelium (RPE) performs many functions critical to retinal health and visual function and is implicated in the development of AMD. Studies by Jadeja et al. have shown that the loss of NAMPT in the aging RPE will promote cell senescence (Jadeja et al. 2018 ). Schlecht and others discovered that regulating the SPP1 pathway provides new opportunities for AMD therapeutic intervention by establishing a mouse model (Schlecht et al. 2020 ). Chandola et al. found that CD44 aptamer-mediated cargo delivery to retinal pigment epithelial cell lysosomes can prevent AMD (Chandola et al. 2019 ). Lee et al. found that COE and BP exert anti-angiogenic effects on retinal neovascularization by inhibiting the expression of AREG and other genes (Lee et al. 2016 ). Finally, qPCR validation was performed on all diagnostic genes. The expression trends of most genes were confirmed.

Aging changes in macular structure may cause different subtypes of AMD. This article proposes a DS-NMF algorithm to identify two subtypes of AMD. The two subtypes have significant differences in enriched pathways and immune infiltration. Based on the MRGs between subtypes, this paper constructed an AMD diagnosis model based on four machine learning methods. The diagnostic model constructed by the SVM-RFE algorithm can reasonably predict the occurrence of AMD. The communication patterns between immune cells and other cells in AMD samples were explored through scRNA-seq data sets. The subtypes and pathways identified in this article and the diagnostic model constructed in this article can provide new insights into the precise treatment of AMD.

Data Availability

The data used in the paper was downloaded from the GEO database ( https://www.ncbi.nlm.nih.gov/geo/ ).

Ambati J et al (2003) An animal model of age-related macular degeneration in senescent Ccl-2- or Ccr-2-deficient mice. Nat Med 9(11):1390–1397

Article   CAS   PubMed   Google Scholar  

Aran D et al (2019) Reference-based analysis of lung single-cell sequencing reveals a transitional profibrotic macrophage. Nat Immunol 20(2):163–172

Article   CAS   PubMed   PubMed Central   Google Scholar  

Baas DC et al (2012) Multicenter cohort association study of SLC2A1 single nucleotide polymorphisms and age-related macular degeneration. Mol vis 18:657–674

CAS   PubMed   PubMed Central   Google Scholar  

Beatty S et al (2000) The role of oxidative stress in the pathogenesis of age-related macular degeneration. Surv Ophthalmol 45(2):115–134

Chandola C et al (2019) CD44 aptamer mediated cargo delivery to lysosomes of retinal pigment epithelial cells to prevent age-related macular degeneration. Biochem Biophys Rep 18:100642

PubMed   PubMed Central   Google Scholar  

Chang J et al (2023) Constructing a novel mitochondrial-related gene signature for evaluating the tumor immune microenvironment and predicting survival in stomach adenocarcinoma. J Transl Med 21(1):191

Chen B et al (2018) Profiling tumor infiltrating immune cells with CIBERSORT. Methods Mol Biol 1711:243–259

Chen Q et al (2020) Fenofibrate inhibits subretinal fibrosis through suppressing TGF-β-Smad2/3 signaling and Wnt signaling in neovascular age-related macular degeneration. Front Pharmacol 11:580884

Choudhury R et al (2021) FHL-1 interacts with human RPE cells through the α5β1 integrin and confers protection against oxidative stress. Sci Rep 11(1):14175

Ding D et al (2018) Inhibition of TRAF6 alleviates choroidal neovascularization in vivo. Biochem Biophys Res Commun 503(4):2742–2748

Emri E et al (2020) A multi-omics approach identifies key regulatory pathways induced by long-term zinc supplementation in human primary retinal pigment epithelium. Nutrients 12(10).  https://doi.org/10.3390/nu12103051

Fan Y et al (2008) Spatial patterns of brain atrophy in MCI patients, identified via high-dimensional pattern classification, predict subsequent cognitive decline. Neuroimage 39(4):1731–1743

Article   PubMed   Google Scholar  

Fan Q et al (2023) Contribution of common and rare variants to Asian neovascular age-related macular degeneration subtypes. Nat Commun 14(1):5574

Han D, He X (2023) Screening for biomarkers in age-related macular degeneration. Heliyon 9(7):e16981

Hänzelmann S, Castelo R, Guinney J (2013) GSVA: gene set variation analysis for microarray and RNA-seq data. BMC Bioinformatics 14:7

Article   PubMed   PubMed Central   Google Scholar  

Hao Y et al (2024) Dictionary learning for integrative, multimodal and scalable single-cell analysis. Nat Biotechnol 42(2):293–304

Jadeja RN et al (2018) Loss of NAMPT in aging retinal pigment epithelium reduces NAD(+) availability and promotes cellular senescence. Aging (albany NY) 10(6):1306–1323

Jarrett SG, Boulton ME (2012) Consequences of oxidative stress in age-related macular degeneration. Mol Aspects Med 33(4):399–417

Jin S et al (2021) Inference and analysis of cell-cell communication using Cell Chat. Nat Commun 12(1):1088

Jones MK et al (2023) Integration of human stem cell-derived in vitro systems and mouse preclinical models identifies complex pathophysiologic mechanisms in retinal dystrophy. Front Cell Dev Biol 11:1252547

Kaarniranta K, Salminen A (2009) Age-related macular degeneration: activation of innate immunity system via pattern recognition receptors. J Mol Med (berl) 87(2):117–123

Karunadharma PP et al (2010) Mitochondrial DNA damage as a potential mechanism for age-related macular degeneration. Invest Ophthalmol vis Sci 51(11):5470–5479

Khandhadia S, Lotery A (2010) Oxidation and age-related macular degeneration: insights from molecular biology. Expert Rev Mol Med 12:e34

Lee YM et al (2016) Cnidium officinale extract and butylidenephthalide inhibits retinal neovascularization in vitro and in vivo. BMC Complement Altern Med 16:231

Lim LS et al (2012) Age-related macular degeneration. Lancet 379(9827):1728–1738

Liu Y et al (2024) The prognostic genes model of breast cancer drug resistance based on single-cell sequencing analysis and transcriptome analysis. Clin Exp Med 24(1):113

McKay GJ et al (2017) Associations between serum vitamin D and genetic variants in vitamin D pathways and age-related macular degeneration in the European Eye Study. Ophthalmology 124(1):90–96

Newman AM et al (2012) Systems-level analysis of age-related macular degeneration reveals global biomarkers and phenotype-specific functional networks. Genome Med 4(2):16

Orozco LD et al (2020) Integration of eQTL and a single-cell atlas in the human eye identifies causal genes for age-related macular degeneration. Cell Rep 30(4):1246-1259.e6

Pedregosa F et al (2011) Scikit-learn: machine learning in Python. J Mach Learn Res 12(null):2825–2830

Pei Y et al (2023) Construction and evaluation of Alzheimer’s disease diagnostic prediction model based on genes involved in mitophagy. Front Aging Neurosci 15:1146660

Ritchie ME et al (2015) limma powers differential expression analyses for RNA-sequencing and microarray studies. Nucleic Acids Res 43(7):e47

Santos FM et al (2023) Proteomics profiling of vitreous humor reveals complement and coagulation components, adhesion factors, and neurodegeneration markers as discriminatory biomarkers of vitreoretinal eye diseases. Front Immunol 14:1107295

Schlecht A et al (2020) Secreted phosphoprotein 1 expression in retinal mononuclear phagocytes links murine to human choroidal neovascularization. Front Cell Dev Biol 8:618598

Stoeckius M et al (2018) Cell Hashing with barcoded antibodies enables multiplexing and doublet detection for single cell genomics. Genome Biol 19(1):224

Subramanian B et al (2023) Association of OCT biomarkers and visual impairment in patients with diabetic macular oedema with vitreomacular adhesion. PLoS ONE 18(7):e0288879

Wang Z et al (2021) Integrated Analysis of DNA methylation and transcriptome profile to identify key features of age-related macular degeneration. Bioengineered 12(1):7061–7078

Yang Y et al (2023) Identification of the immune landscapes and follicular helper T cell-related genes for the diagnosis of age-related macular degeneration. Diagnostics (Basel) 13(17).  https://doi.org/10.3390/diagnostics13172732

Yu G et al (2012) clusterProfiler: an R package for comparing biological themes among gene clusters. OMICS 16(5):284–287

Zhao T et al (2023) Integrative metabolome and lipidome analyses of plasma in neovascular macular degeneration. Heliyon 9(10):e20329

Zhong YS et al (2013) Potassium ion channels in retinal ganglion cells (review). Mol Med Rep 8(2):311–319

Download references

This study was supported by the Jiangsu Graduate Practice and Innovation program (NO. 134422631103).

Author information

Authors and affiliations.

Eye Institute, Affiliated Hospital of Nantong University, Nantong, Jiangsu, China

Shenglai Zhang, Ying Yang, Jia Chen, Shu Su, Yu Cai, Xiaowei Yang & Aimin Sang

You can also search for this author in PubMed   Google Scholar

Contributions

Shenglai Zhang: Conceptualization, Methodology, Software, Visualization, Writing – Original Draft Preparation, Writing-review & editing.

Ying Yang: Conceptualization, Methodology, Software, Visualization,Writing-review and editing.

Jia Chen: Conceptualization, Methodology, Software, Writing-review and editing.

Shu Su: Data curation, Software, Writing-review & editing.

Yu Cai: Visualization, Writing-review and editing.

Xiaowei Yang: Data curation, Writing-review, Funding and editing.

All authors read and approved the manuscript.

Corresponding author

Correspondence to Aimin Sang .

Ethics declarations

Ethics approval.

Not applicable.

Consent to Participate

Consent for publication, competing interests.

The authors declare no competing interests.

Additional information

Publisher's note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary file1 (DOCX 8284 KB)

Rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/ .

Reprints and permissions

About this article

Zhang, S., Yang, Y., Chen, J. et al. Integrating Multi-omics to Identify Age-Related Macular Degeneration Subtypes and Biomarkers. J Mol Neurosci 74 , 74 (2024). https://doi.org/10.1007/s12031-024-02249-9

Download citation

Received : 03 April 2024

Accepted : 15 July 2024

Published : 07 August 2024

DOI : https://doi.org/10.1007/s12031-024-02249-9

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Age-related macular degeneration
• Immune landscape
• Mitochondria
• The first two authors should be regarded as joint first authors
• Find a journal
• Publish with us
• Track your research

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

• View all journals
• Explore content
• About the journal
• Publish with us
• Sign up for alerts
• Open access
• Published: 09 August 2024

PDK4-mediated metabolic reprogramming is a potential therapeutic target for neovascular age-related macular degeneration

• Juhee Kim 1 , 2   na1 ,
• Yujin Jeon 1 , 2   na1 ,
• Jinyoung Son 3 , 4 ,
• Haushabhau S. Pagire 5 , 6 ,
• Suvarna H. Pagire 5 , 6 ,
• Jin Hee Ahn 5 , 6 ,
• Akiyoshi Uemura 7 ,
• In-Kyu Lee   ORCID: orcid.org/0000-0002-1655-0985 8 ,
• Sungmi Park 1 &
• Dong Ho Park   ORCID: orcid.org/0000-0001-6911-3765 1 , 2 , 3 , 4  

Cell Death & Disease volume  15 , Article number:  582 ( 2024 ) Cite this article

Metrics details

• Diseases of the nervous system
• Metabolic disorders

Age-related macular degeneration (AMD) causes severe blindness in the elderly due to choroidal neovascularization (CNV), which results from the dysfunction of the retinal pigment epithelium (RPE). While normal RPE depends exclusively on mitochondrial oxidative phosphorylation for energy production, the inflammatory conditions associated with metabolic reprogramming of the RPE play a pivotal role in CNV. Although mitochondrial pyruvate dehydrogenase kinase (PDK) is a central node of energy metabolism, its role in the development of CNV in neovascular AMD has not been investigated. In the present study, we used a laser-induced CNV mouse model to evaluate the effects of Pdk4 gene ablation and treatment with pan-PDK or specific PDK4 inhibitors on fluorescein angiography and CNV lesion area. Among PDK isoforms, only PDK4 was upregulated in the RPE of laser-induced CNV mice, and Pdk4 gene ablation attenuated CNV. Next, we evaluated mitochondrial changes mediated by PDK1-4 inhibition using siRNA or PDK inhibitors in inflammatory cytokine mixture (ICM)-treated primary human RPE (hRPE) cells. PDK4 silencing only in ICM-treated hRPE cells restored mitochondrial respiration and reduced inflammatory cytokine secretion. Likewise, GM10395, a specific PDK4 inhibitor, restored oxidative phosphorylation and decreased ICM-induced upregulation of inflammatory cytokine secretion. In a laser-induced CNV mouse model, GM10395 significantly alleviated CNV. Taken together, we demonstrate that specific PDK4 inhibition could be a therapeutic strategy for neovascular AMD by preventing mitochondrial metabolic reprogramming in the RPE under inflammatory conditions.

Similar content being viewed by others

Oxidative stress-mediated TXNIP loss causes RPE dysfunction

The AKT2/SIRT5/TFEB pathway as a potential therapeutic target in non-neovascular AMD

TGF-β promotes pericyte-myofibroblast transition in subretinal fibrosis through the Smad2/3 and Akt/mTOR pathways

Introduction.

Age-related macular degeneration (AMD) is a leading cause of blindness among the elderly population [ 1 ]. Particularly, neovascular AMD exhibits characteristic choroidal neovascularization (CNV), in which abnormal new vessels penetrate through retinal pigment epithelium (RPE) into the outer retina, compromising vision [ 2 ]. Although intravitreal injections of anti-vascular endothelial growth factor (VEGF) antibodies are currently used to treat CNV, more than half of patients have an inadequate response to the treatment despite monthly injections [ 3 ]. Furthermore, the current therapeutics require repeated injections which can cause serious infections complications including endophthalmitis [ 4 ]. These limitations emphasize the critical need to develop alternative therapeutic targets for CNV.

Because the RPE phagocytoses and degrades shed outer segments throughout life, RPE cells require a significant energy supply that relies almost exclusively on mitochondria, primarily oxidative phosphorylation (OXPHOS) [ 5 , 6 , 7 ]. However, under hypoxic or inflammatory conditions, metabolic reprogramming towards aerobic glycolysis is mainly regulated by stabilization of hypoxia-inducible factor 1-alpha (HIF-1 α ), and the metabolic shift results in releasing proinflammatory cytokines [ 8 ]. A proinflammatory RPE environment promotes the development of AMD [ 9 ]. Furthermore, prior studies of human donors with AMD support that defective RPE mitochondrial function drives AMD pathology [ 10 ]. In primary cultures from AMD donors, ATP generation through OXPHOS is decreased, while ATP generation through glycolysis is increased [ 11 ].

The mitochondrial pyruvate dehydrogenase complex (PDC) is a central metabolic node mediating pyruvate oxidation, the critical step in OXPHOS. The PDC is regulated by post-translational subunit modifications, including phosphorylation of the E1α subunit of pyruvate dehydrogenase (PDHE1α) by pyruvate dehydrogenase kinase (PDK) 1–4 and de-phosphorylation regulated by pyruvate dehydrogenase phosphatase 1–2 [ 12 ]. Prior studies have reported that fursultiamine, a mitochondrial PDH cofactor, alleviated CNV by modulating the RPE metabolic and inflammatory response [ 13 ]. Similarly, PDK inhibition, which increases PDC activity, has been used as a therapeutic approach for inflammatory bowel disease [ 14 ]. However, the effects of PDK inhibition on metabolism processes in the RPE have not been investigated. Thus, in this study, we investigated whether PDK inhibition, particularly with small-molecule PDK4 inhibitors, alleviates CNV by regulating RPE metabolic and inflammatory reactions.

Materials and methods

Animal studies.

Animal experiments were conducted in accordance with the guidelines by the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research. Animal Care Committee of the Kyungpook National University (No. 2019-0104-1) approved the animal studies. Pdk4 knockout mice ( Pdk4 −/− ) were bred in an in-house animal facility [ 15 ].

Laser-induced CNV model and treatment

Animals were randomly assigned to one of experimental or control group (5 mice/group). Seven days before CNV induction, 7-week-old male C57BL/6J mice were treated with dichloroacetate (DCA 250 mg/kg in saline, Sigma-Aldrich, St. Louis, MO, USA) or GM10395 (1 mg/kg; 10 mg/mL in 1% DMSO), or normal saline by oral gavage daily for 14 days. A 532-nm OcuLight GLx Laser System (IRIDEX Corporation, Mountain View, CA, USA) was used to generate 4–10 CNV lesions in mice with the parameters as described previously [ 16 , 17 ]. Photocoagulation spots with hemorrhage or no bubble formation at the laser site were excluded. Seven days after CNV induction, pigmented RPE and choroid tissues (RPE/choroid) were dissected from the above transparent tissues (retina tissue; from nerve fiber layer to photoreceptor layer) and collected for choroidal flat mount and protein or RNA isolation.

Fluorescein angiography

Fundus fluorescein angiography images were acquired using Μicron IV Retinal Imaging Microscope (Phoenix Technology Group, Lakewood, CO, USA) as described previously [ 13 , 16 , 17 ]. Images were obtained at 3–5 min (early phase) and 7–10 min (late phase) after i.p. injection of 2% fluorescein sodium (Akorn, Inc., Lake Forest, IL, USA, H12099-0711). Lesion severity was graded as follows: lesions with patchy or faint fluorescence without any leakage were assigned a score of 0 (no leakage); lesions with hyperfluorescence without changes in intensity or size were assigned a score of 1 (mild leakage); lesions with hyperfluorescence of a consistent size but increasing intensity were assigned a score of 2A (moderate leakage); and lesions with hyperfluorescence of both increasing intensity and size were assigned a score of 2B (significant leakage). The classification was performed by two blinded examiners (JHK and YJJ), and disagreements were resolved by a third examiner (JYS).

Choroidal flat-mounts

On day 7 after laser-induced CNV induction, the eyes of the mice were removed and fixed in a solution containing 4% paraformaldehyde at RT for 30 min. The choroid and sclera were separated from the retinas to create a choroidal flat mount. To stain the eyecups, a solution of isolectin B4 conjugated to Alexa Fluor 488 (Invitrogen, Waltham, MA, USA, I21411) was incubated at 4 °C overnight. The eyecup was then mounted flat using PermaFluor aqueous mount (Thermo Fisher Scientific, Waltham, MA, USA). The flat-mounted eyecup was imaged using LSM 800 Airyscan confocal microscope (Carl Zeiss, Jena, Germany). The size of the CNV lesions was measured using ImageJ software as described in previous studies [ 13 , 16 , 17 ].

Immunohistochemical staining

Mouse eyes were enucleated and fixed in a solution containing 4% paraformaldehyde at RT for 1 h followed by the incubation in 30% sucrose solution at 4 °C overnight and then embedded in an OCT compound. Cryosections of 15 µm-thickness were incubated with a blocking buffer at RT for 1 h. After incubation with primary antibodies at 4 °C overnight, the sections were incubated with secondary antibodies at RT for 1 h. Antibody information is described in Table S1 .

Cell culture

Primary human RPE (hRPE) cells (Lonza, Walkersville, MD, USA, 00194987) were used between 5 and 6 passages. The cells were cultured in basal media with added supplements of RtEGM BulletKit (Lonza, 00195409) at 37 °C in a 5% CO 2 humidified incubator, according to the manufacturer’s instructions. The cells were seeded with 2% FBS/RtEGM media overnight and incubated with serum-free RtEGM media for 24 h. Confluent cells were treated in serum-free RtEGM containing 5 ng/mL of inflammatory cytokine mixture (ICM) including IL-1β, TNF-α, and IFN-γ from R&D system for 24 h mimicking the inflammatory state of AMD [ 18 ]. For the cell viability assay, the Cell Counting Kit 8 (Abcam, Waltham, MA, USA, ab228554) was used according to the manufacturer’s instructions.

siRNA transfection

Predesigned siRNAs targeting human PDK1 -4 and si Control (SN-1003) (Bioneer, Daejeon, Korea) were described in Table S2 . siRNA transfections were performed with Lipofectamine RNAiMAX and Opti-MEM according to the manufacturer’s instructions (Thermo Fisher Scientific). In brief, the cells were cultured in a 12-well plate and transfected at 60–70% confluency with siRNA at 500 pM. The next day, the cells were incubated with the media containing ICM for 24 h.

Quantitative real-time polymerase chain reaction (qPCR)

Total RNA was extracted from the tissues or cells using QIAzol lysis reagent (Qiagen, Hilden, Germany) and used for cDNA synthesis using a kit (Thermo Fisher Scientific). The qPCR analysis was performed with Luna Universal qPCR Master Mix (New England Biolabs, Ipswich, MA, USA) using a ViiA7 real-time PCR system (Applied Biosystems, Carlsbad, CA, USA). Specific primers used for real-time PCR are described in Table S3 .

Western blot analysis

Mouse tissues and the cells were lysed with RIPA buffer (Thermo Fisher Scientific) containing a phosphatase inhibitor cocktail (Sigma-Aldrich). The BCA protein assay (Thermo Fisher Scientific) measured protein concentrations. Nupage 4–12% Bis-Tris Mini Protein Gels (Thermo Fisher Scientific) separated the lysates. Antibody information is described in Table S1 .

Enzyme-linked immunosorbent assay (ELISA) for cytokines

After diluting 1:10 to 200 in cell culture medium, the secreted amounts of human IL6 (Thermo Fisher Scientific, 88-7066-76), IL8 (R&D system, Minneapolis, MN, USA, D8000C), and MCP1 (Thermofisher Scientific, 88-7399-76) were measured using ELISA kits, according to the manufacturer’s instructions.

Measurement of oxygen consumption rate (OCR) and extracellular acidification rate (ECAR)

OCR was measured using an XFe96 Extracellular Flux Analyzer (Seahorse Biosciences, Inc., Billerica, MA, USA). The cells were seeded in 6-well plates at a density of 1.4 × 10 5 cells/well. On day 3, the cells were washed and treated with ICM diluted in serum-free media with DCA or GM10395 for 24 h. Next, cells were transferred into the Xfe96 cell culture plate (Agilent Technologies, Santa Clara, CA, USA, 103794-100) with assay medium at a density of 1.0 × 10 4 cells/well, and the plate was incubated in 5% CO 2 at 37 °C for 5 h. The OCR assay medium consisted of XF DMEM medium pH 7.4 (Agilent Technologies, 102353) supplemented with 25 mM Glucose Solution (Sigma-Aldrich, G7528), 1 mM Pyruvate Solution (Sigma-Aldrich, S8636), and 2 mM Glutamine Solution (Thermo Fisher Scientific, 35050061). The ECAR assay medium consisted of XF DMEM medium pH 7.4 supplemented with 2 mM Glutamine Solution. Uncouplers and inhibitors were used at the indicated concentrations: oligomycin A (1 μM, Sigma-Aldrich, 75351), FCCP (carbonyl cyanide 4-(trifluoro-methoxy)phenylhydrazone, 1 μM, Sigma-Aldrich, C2920), rotenone (1 μM, Sigma-Aldrich, R8875), 2-DG (50 mM, Sigma-Aldrich, D6134) and Glucose (10 mM, Sigma-Aldrich, G7528). The nuclei were stained with DAPI solution (1 μg/mL, Thermo Fisher Scientific, 62248) and counted automatically using an ImageXpress Micro Confocal Microscope (Molecular Devices, San Jose, CA, USA) to normalize by cell number.

Cellular superoxide detection

A mitochondrial Superoxide Indicator (MitoSOX™, Thermo Fisher Scientific) was used to detect mitochondrial superoxide formation in the cells following ICM treatment for 24 h with drugs. The cells were seeded at a density of 2 × 10 4 cells/well in a black wall/clear bottom 96-well plate (Greiner Bio-One, Kremsmünster, Austria). At the end of the experiment, the cells were incubated for 10 min at 37 °C in PBS containing MitoSOX (5 μM) and NucBlue™ Live ReadyProbes™ Reagent solution (Thermo Fisher Scientific). Red-stained mitochondrial superoxide was image-captured using an ImageXpress Micro Confocal System. The mean fluorescence intensity from DAPI and TRITC in image planes was imaged and quantified automatically.

Immunofluorescence for mitochondrial morphology

Cells were seeded at a density of 6 × 10 3 cells/well in a black wall/clear bottom 96-well plate (Greiner Bio-One) and were treated with 4% paraformaldehyde for fixation, followed by permeabilization. The cells were then washed with PBS and incubated with primary antibody against TOM20 (Santa Cruz Biotechnology, Santa Cruz, CA, USA, sc-11415) at 4 °C overnight, followed by the incubation with Alexa Fluor 594-conjugated donkey anti-rabbit antibody (Thermo Fisher Scientific, A21207) at RT for 2 h. Nuclei were stained with DAPI solution (Thermo Fisher Scientific). Images were obtained from a 2.4 mm z-stack as described previously [ 13 ]. The sum of fluorescence pixels was calculated for each cell within the images. The total fluorescence intensity from each image was divided by the number of cells in that image to obtain the average fluorescence intensity per cell, as described previously [ 19 ].

All results were expressed as mean ± SEM. Datasets were analyzed using a two-sample t -test and one-way or two-way ANOVA followed by Tukey’s multiple comparison test. Data analyses were performed using Prism v. 6.0 (GraphPad Software, San Diego, CA, USA), and P  < 0.05 was considered statistically significant.

PDK4 activity is upregulated in the RPE of laser-induced CNV mice

To identify the specific mitochondrial PDK isoform involved in the development of neovascular AMD, we evaluated the expression of four PDK isoforms that suppress OXPHOS by PDK-mediated phosphorylation of PDHE1α (p-PDHE1α) in the laser-induced CNV mouse model. Total protein levels of PDK4 and p-PDHE1α significantly increased by 9.7 ± 0.7-fold and 2.5 ± 0.3-fold, respectively, in the RPE/choroid 1 day following CNV induction, while the levels of the other isoforms (PDK1–PDK3) remained unchanged (Fig. 1A, B ). Similarly, immunofluorescence for PDK4 and p-PDHE1α identified enhanced staining in the RPE/choroid, particularly following CNV induction (Fig. 1C, D ). These findings are consistent with a prior study demonstrating increased PDK4-mediated PDHE1α phosphorylation in the bowel tissues of inflammatory bowel disease patients, which was recapitulated in a mouse model [ 14 ]. By contrast, the retina tissue of the CNV model did not show changes in the levels of any PDK isoform (Fig. S1 ). These data suggest that increased PDK4 expression in the RPE/choroid could correlate with defective mitochondrial function during the early stages of CNV.

A , B Protein was isolated from the RPE/choroid of control and laser-induced CNV mice, and protein lysates were subjected to immunoblotting for PDK1–4 and phosphorylated pyruvate dehydrogenase E1-alpha subunit (p-PDHE1α). Only PDK4 was upregulated, and p-PDHE1α was increased in the RPE of CNV mice relative to the control. C , D Representative confocal images of retina and choroid from control and day 1 post-CNV induction animals. PDK4 + cells (green) and p-PDHE1α + cells (green) were present, especially in the RPE of CNV mice at higher magnifications, indicated by asterisks. Scale bar: 100 μm. Inset scale bar: 10 μm. Data are represented as mean ± SEM. * P  < 0.05; ** P  < 0.01; *** P  < 0.001 versus control ( n  = 5 mice/group). Two-tailed unpaired t -test. HSP90 heat shock protein 90, PDHE1α pyruvate dehydrogenase E1-alpha subunit, INL inner nuclear layer, OPL outer plexiform layer, ONL outer nuclear layer, PR photoreceptor.

Pdk4 knockout mice show reduced choroidal neovascularization

To investigate the effect of PDK4 deficiency in CNV, we evaluated the neovascular phenotypes of wild-type (Pdk4 +/+ ) and Pdk4 knockout ( Pdk4 −/− ) mice. Fluorescein angiography revealed abnormal vascular leakage from the new vessels on day 7 following CNV induction by laser photocoagulation. The proportion of grade 2B lesions with significant vascular leakage was lower in Pdk4 −/− mice than in the wild type (46% in wild-type and 5% in Pdk4 −/− mice) (Fig. 2A ). Similarly, Pdk4 −/− mice exhibited a significantly smaller CNV lesion area by 48.3 ± 5.6% compared to wild-type mice (21.9 ± 3.3 in wild-type and 11.3 ± 1.2 mm 2 × 10 −3 in Pdk4 −/− mice) based on an examination of flat-mounted choroids stained with isolectin B4 (Fig. 2B ). These findings demonstrated that PDK4 contributes to CNV, suggesting that it could be a potential therapeutic target for neovascular AMD. Inflammatory cytokines, including IL1B and MCP1, are highly expressed in the CNV mouse model and human CNV tissues [ 20 , 21 ]. Thus, we compared cytokine levels between wild-type and Pdk4 −/− mice. The mRNA levels of proinflammatory cytokines, such as Il1b and Il6 , were lower in the RPE/choroid of the CNV model of Pdk4 −/− mice than those in the CNV model of wild-type mice. (Fig. 2C ).

A Grading of CNV lesions (0, 1, 2A, and 2B) was conducted in fluorescein angiography images from wild-type (WT) and Pdk4 −/− ( Pdk4 KO) mice 7 days following CNV induction ( n  = 5 mice/group). B CNV lesion size was calculated in both groups. ** P  < 0.01; *** P  < 0.001 versus wild-type CNV ( n  = 34 lesions/group). Chi-square and two-tailed unpaired t -test. C mRNA levels of proinflammatory cytokines, including interleukin-1 beta (Il1b), interleukin-6 (Il6) , and monocyte chemoattractant protein-1 (Mcp1) , were measured. Data are represented as mean ± SEM. ## P  < 0.01; ### P  < 0.001 versus wild type; ** P < 0.01; *** P  < 0.001 versus wild-type CNV ( n  = 5 mice/group). Scale bar: 100 μm. One-way ANOVA with Tukey’s multiple comparisons test. NV neovascularization.

Only PDK4 gene silencing restored mitochondrial respiration in inflammation-challenged human RPE cells

To explore the specific roles of individual PDK isoforms on the inflammation-mediated decrease of mitochondrial respiration, we measured oxygen consumption rate (OCR) in primary hRPE cells, which were transiently transfected with either PDK1 , PDK2 , PDK3 , or PDK4 siRNA for 48 h, followed by incubation with vehicle (PBS) or ICM for an additional 24 h. Mitochondrial respiration was significantly decreased in ICM-treated human RPE cells, as evidenced by reductions in basal OCR, maximal OCR, and spare respiratory capacity. PDK4 gene silencing but not the silencing of any other PDK isozyme gene strongly reversed these reductions in respiratory parameters and ATP production (Fig. 3A ). The siRNA targeting inducible PDK4 under inflammatory conditions was confirmed at both mRNA and protein levels (Fig. 3B, C ). As inflammatory conditions are associated with impairment of energy metabolism of the RPE and contribute to the progression of AMD, restoring RPE oxidative metabolism could decrease inflammation. Thus, we investigated whether siRNA depletion of PDK4 expression affects mRNA levels of IL1B , IL6 , IL8 , or MCP1 in ICM-treated primary hRPE cells. Furthermore, it inhibited ICM-induced upregulation of IL1B , IL6 , IL8 , and MCP1 compared with the si Control (Fig. 3D ).

A Oxygen consumption rate (OCR) in primary hRPE that were transiently transfected with si PDK1-4 (small interfering RNA [siRNA] targeting PDK isoform 1, 2, 3, and 4 ) for 48 h followed by challenge with vehicle (PBS) or inflammatory cytokine mixture (ICM). Data are represented as mean ± SEM. # P  < 0.05; ## P  < 0.01; ### P  < 0.001 versus si Control ; *** P  < 0.001 versus si Control with ICM ( n  = 5/group). One-way ANOVA with Tukey’s multiple comparisons test. B , C PDK4 mRNA levels and PDK4 protein levels were measured in primary hRPE cells treated with ICM following transient knockdown of PDK4 . D mRNA levels of proinflammatory cytokines, including interleukin-1 beta (IL1B), interleukin-6 (IL6), interleukin-8 (IL8), and monocyte chemoattractant protein-1 (MCP1) , were measured in primary hRPE cells treated with ICM following transient PDK4 knockdown using siRNA (si PDK4 ) or si Control . Data are expressed as mean ± SEM. # P  < 0.05; ### P  < 0.001 versus si Control ; * P  < 0.05; *** P  < 0.001 versus si Control with ICM ( n  = 3/group). Two-way ANOVA with Tukey’s multiple comparisons test.

Small-molecule PDK4 inhibitors decrease inflammatory cytokine expression

Due to the lack of commercially available PDK4 selective inhibitors, pharmacological pan-PDK inhibitors, such as dichloroacetate (DCA), which target the ATP-binding pockets of all PDK isozymes, have been used to investigate metabolic diseases associated with mitochondrial dysfunction. Our previous studies showed that PDK4 inhibition by DCA protects against obesity-associated insulin resistance via regulation of adipose tissue inflammation [ 22 ], atherosclerosis [ 23 ], and diabetes [ 24 ]. Recently, structural modifications of hit anthraquinones have resulted in a new series of allosteric PDK4 inhibitors optimally fitted to the lipoamide binding site [ 25 ]. We developed a potent allosteric PDK4 inhibitor, GM10395, and used it in the current study (Fig. S2A ). ICM (5 and 10 ng/mL), DCA (1 and 10 µM), or GM10395 (0.5 and 1 µM) did not affect cell viability (Fig. S2B ). We assessed the effect of DCA and GM10395 on the inflammatory response and found that DCA significantly down-regulated inflammatory cytokine expression in primary hRPE cells (Fig. S2C ). These findings are consistent with those of a previous study reporting the effect of DCA on TGFβ-challenged hRPE cells [ 26 ].

PDK4 inhibition reverses the ICM-induced metabolic switch in primary hRPE cells

Next, we determined if DCA or GM10395 affected the level of p-PDHE1α in ICM-treated hRPE cells (Fig. 4A, B ). GM10395 at 0.5 µM was as effective in decreasing the level of p-PDHE1α as DCA, a pan-PDK inhibitor, at 10 µM. We previously reported that inflammation-induced metabolic stress of the RPE and decreased mitochondrial respiration exacerbate CNV [ 13 ]. To determine whether PDK regulates metabolic reprogramming in this context, we evaluated the effect of DCA and GM10395 on OCR and ECAR in ICM-treated primary hRPE cells. The maximal OCR, which was decreased by 36.4 ± 0.7% in ICM-treated cells, was significantly restored by 10 µM DCA or 1 µM GM10395, which was similar to the increase in the ATP production rate by PDK4 inhibitors (Fig. 4C ). Glycolysis, which was increased by 61.1 ± 0.7% in ICM-treated primary hRPE cells, was significantly decreased by 10 µM DCA and 0.5 and 1 µM GM10395 (Fig. S3 ).

A , B Protein was isolated from ICM-treated primary hRPE cells ± dichloroacetate (DCA; 1 and 10 μM) or GM10395 (0.5 and 1 μM) for 24 h. Lysates were subjected to immunoblotting for phosphorylated pyruvate dehydrogenase E1-alpha subunit (p-PDHE1α), and heat shock protein 90 (HSP90) ratios were determined by densitometry ( n  = 3 replicates/experiment). C Oxygen consumption rate (OCR) was measured in primary hRPE cells with ICM treatment for 24 h on day 3 ± DCA (1 and 10 μM) or GM10395 (0.5 and 1 μM). Basal respiration, maximal respiration, spare respiratory capacity, and ATP production rate were calculated based on the OCR response to specific inhibitors ( n  = 5/group). D Mitochondrial superoxide (MitoSOX) levels were measured in primary hRPE cells with ICM treatment for 24 h ± DCA (10 μM) or GM10395 (1 μM). Scale bar: 50 μm. Quantification of MitoSOX intensity ( n  = 27 microscopic fields with > 4 × 10 4 cells counted from three independent wells). E Primary hRPE cells treated with ICM ± DCA (10 μM) or GM10395 (1 μM) for 24 h were fixed and stained with translocase of outer membrane 20 (TOM20) antibody to visualize mitochondrial membranes. Scale bar: 20 μm. Inset scale bar: 5 μm. Quantification of mitochondrial area, fragmentation, and elongation of > 65 cells counted for each condition. Data are represented as mean ± SEM. # P  < 0.05; ## P  < 0.01; ### P  < 0.001 versus control; * P  < 0.05; ** P < 0.01; *** P  < 0.001 versus ICM group. One-way ANOVA with Tukey’s multiple comparisons test.

Next, we evaluated the effect of DCA and GM10395 on mitochondrial superoxide levels using MitoSOX staining. The ICM-treated primary hRPE cells exhibited a 19.5 ± 0.7% increase in superoxide levels, which was significantly reduced by DCA or GM10395 treatment (Fig. 4D ). Immunostaining of mitochondria with Tom20 showed that the mitochondrial area, which was reduced by 52.1 ± 1.6% in ICM-treated primary hRPE cells, was significantly restored by DCA (71.5 ± 2.8%) and GM10395 (82.2 ± 3.2%) (Fig. 4E ). Similarly, the increased mitochondrial fragmentation in ICM-treated hRPE cells was significantly reversed by DCA and GM10395 (Fig. 4E ).

To assess the anti-inflammatory effect of GM10395, we measured mRNA and protein levels of proinflammatory cytokines after DCA or GM10395 treatment of ICM-treated primary hRPE cells. Inhibition of PDK4 by DCA or GM10395 significantly attenuated the upregulation of inflammatory cytokines (Fig. S4 ). Previous studies have shown that mixed cytokine treatment in hRPE cells upregulated NFkB expression and increased cytokine secretion, mimicking the inflammatory CNV condition [ 18 , 27 , 28 ]. Furthermore, Kutty et al. showed the anti-inflammatory effect of an antioxidant occurs via regulation of the NFkB pathway in ICM-treated hRPE cells [ 29 ]. Our previous study showed that PDH activation by reversing the metabolic reprogramming of RPE cells under inflammatory conditions significantly decreased cytokine secretion via the regulation of NFkB pathway [ 13 ]. These findings demonstrate that small-molecule inhibition of PDK4 in RPE cells increases mitochondrial respiration, resulting in reduced cytokine levels under inflammatory conditions.

GM10395-mediated PDK4 inhibition alleviates laser-induced CNV

To evaluate the anti-angiogenic efficacy of DCA and GM10395, we orally administered the CNV model mice with 250 mg/kg DCA or 1 mg/kg GM10395. Both DCA and GM10395 significantly decreased the proportion of clinically significant Grade 2B CNV lesions compared with vehicle on day 7 following CNV induction (Fig. 5A ). Likewise, CNV size was 40.9 ± 6.7% lower in the DCA group and 35.4 ± 7.6% lower in the GM10395 group than in the vehicle group, as assessed using choroidal flat-mount preparations (Fig. 5B ). GM10395 decreased the p-PDHE1α levels in the RPE of CNV mice by 68.3 ± 4.5% compared to vehicle (Fig. 5C ), which is similar to the decrease in the p-PDHE1α levels by GM10395 in primary hRPE cells (Fig. 4B ). The above data indicate that oral administration of the PDK4 specific inhibitor GM10395 protected from inflammatory and metabolic reprogramming of RPE in the CNV model. Consistent with the in vitro data showing reduced proinflammatory cytokine levels (Fig. S4 ), mRNA levels of proinflammatory cytokines, such as Il1b , Il6 , and Mcp1 , were also decreased in the RPE/choroid of CNV animals (Fig. 5D ). Taken together, these findings demonstrate that PDK4 inhibition alleviates CNV and has anti-inflammatory effects.

A Grading of CNV lesions was conducted in fluorescein angiography images from the vehicle, dichloroacetate (DCA; 250 mg/kg), and GM10395 (1 mg/kg)-treated CNV mice 7 days following CNV induction ( n  = 5 mice/group). *** P  < 0.001 versus vehicle, Chi-square. B CNV lesion size was calculated using choroidal flat-mounts. ( n  = 34 lesions/group). Scale bar: 100 μm. Data are expressed as mean ± SEM. * P  < 0.05 versus vehicle. One-way ANOVA with Tukey’s multiple comparisons test. C Protein was isolated from retinal pigment epithelium (RPE)/choroid of the vehicle and GM10395 (1 mg/kg)-treated CNV mice 1 day after CNV induction. Lysates were subjected to immunoblotting for phosphorylated pyruvate dehydrogenase E1-alpha subunit (p-PDHE1α). The lanes were run on the same gel but were noncontiguous. Data are expressed as mean ± SEM. * P  < 0.05 versus vehicle ( n  = 3/group). Two-tailed unpaired t -test. D mRNA levels of the proinflammatory cytokines interleukin-1 beta ( Il1b ), interleukin-6 ( Il6 ), and monocyte chemoattractant protein-1 ( Mcp1 ) were measured in the RPE/choroid of control mice, and vehicle- and GM10395 (1 mg/kg)-treated CNV mice 3 days after CNV induction. ( n  = 3/group). Data are expressed as mean ± SEM. ### P  < 0.001 versus control; * P  < 0.05; ** P  < 0.01; *** P  < 0.001 versus vehicle. One-way ANOVA with Tukey’s multiple comparisons test. HSP90, heat shock protein 90; NV, neovascularization; PDHE1α, pyruvate dehydrogenase E1-alpha subunit.

The RPE relies almost exclusively on mitochondria OXPHOS for ATP production, which allows photoreceptors to utilize glucose transported from the choroid through the RPE [ 30 ]. This process is highly regulated by suppressing RPE glycolysis to preserve glucose for photoreceptors. However, AMD disrupts this metabolic ecosystem, causing the RPE to rely on glycolysis to fulfill its energetic needs during metabolic reprogramming [ 30 ].

Previous studies evaluated the link between PDK-mediated glycolytic metabolic shift and upregulation of proinflammatory cytokines in neural tissues such as dorsal root ganglion, which causes painful diabetic neuropathy [ 31 ]. Furthermore, pharmacologic inhibition of pan-PDK with DCA and its genetic deletion attenuates proinflammatory cytokine expression in this context, alleviating neuroinflammation in the spinal cord. Similarly, the metabolic shift in RPE mitochondria promotes a proinflammatory environment, exacerbating the neovascular phenotypes of AMD. However, the link between metabolic reprogramming in the human RPE and the neovascular phenotypes of CNV remains incompletely understood. Thus, in the present study, we evaluated the functional effects of PDK4-mediated RPE glycolytic reprogramming using genetic ablation and small-molecule targeting specific PDK4, identifying PDK4 as a modulator of mitochondrial dynamics in this context.

Among PDK isoforms, only PDK4 was upregulated in the RPE of CNV animals. Furthermore, Pdk4 ablation alleviated CNV. In vitro studies with primary hRPE cells demonstrated that only Pdk4 silencing restored ICM-induced downregulation of OCR and upregulation of proinflammatory cytokines. Thus, we proposed PDK4 as a potential therapeutic target for CNV.

DCA, a structural analog of pyruvate and an inhibitor for PDKs [ 32 ], has protective effects in various neurological disorders including neurodegenerative diseases [ 33 ]. However, DCA has severe systemic side effects, including peripheral neuropathy [ 34 , 35 ], necessitating the development of alternative PDK inhibitors, especially those targeting PDK4. Recently, we developed a new small-molecule PDK4 inhibitor, GM10395, and it showed a protective effect on mitochondrial dysfunction in ischemia-reperfusion kidney injury mouse models [ 36 ].

In the current study, GM10395 significantly increased maximal respiration and spare respiration. Likewise, it significantly decreased the glycolysis of hRPE cells. Furthermore, GM10395 alleviated inflammation-induced oxidative stress and mitochondrial fission. This indicates that enhancing mitochondrial activity renders the RPE more resistant to oxidative stress, further supporting the critical role of RPE mitochondria in CNV [ 37 ]. Together, our data suggest that inhibition of PDK4 protects the RPE from the inflammatory metabolic shift to aerobic glycolysis.

Inflammatory activation of the RPE changes mitochondrial dynamics and upregulates the expression of proinflammatory cytokines. A previous study reported that inhibiting LPS-induced metabolic reprogramming reduced releasing proinflammatory cytokines [ 38 ]. Similarly, GM10395 dose-dependently suppressed proinflammatory cytokine production. GM10395 attenuated the inflammatory response in the RPE by increasing mitochondrial metabolism via the inhibition of PDK4 expression, alleviating neovascularization.

As shown in Fig. 6 , we report that CNV is associated with the upregulation of PDK4 expression and increased levels of p-PDHE1α, causing a shift to glycolysis and inflammation. Genetic ablation of PDK4 and a small-molecule PDK4 inhibitor alleviated CNV, which was accompanied by decreased inflammation. These findings suggest that small-molecule PDK4 inhibitors administrated orally could be utilized to develop a new class of neovascular AMD therapeutics, which might also be applicable for the therapy of other spectrums of neovascular diseases.

PDK4 is upregulated in the retinal pigment epithelium (RPE) of laser-induced CNV mice. PDK4 inhibitors restore mitochondrial respiration and alleviate the CNV.

Data availability

All data generated or analyzed during this study are included in the main text and the supplementary information files.

Gehrs KM, Anderson DH, Johnson LV, Hageman GS. Age-related macular degeneration-emerging pathogenetic and therapeutic concepts. Ann Med. 2006;38:450–71.

Article   PubMed   Google Scholar  

Ambati J, Fowler BJ. Mechanisms of age-related macular degeneration. Neuron. 2012;75:26–39.

Article   PubMed   PubMed Central   CAS   Google Scholar  

Miller JW, Le Couter J, Strauss EC, Ferrara N. Vascular endothelial growth factor a in intraocular vascular disease. Ophthalmology. 2013;120:106–14.

Menchini F, Toneatto G, Miele A, Donati S, Lanzetta P, Virgili G. Antibiotic prophylaxis for preventing endophthalmitis after intravitreal injection: a systematic review. Eye. 2018;32:1423–31.

Jarrett SG, Lin H, Godley BF, Boulton ME. Mitochondrial DNA damage and its potential role in retinal degeneration. Prog Retin Eye Res. 2008;27:596–607.

Article   PubMed   CAS   Google Scholar  

Adijanto J, Philp NJ. Cultured primary human fetal retinal pigment epithelium (hfRPE) as a model for evaluating RPE metabolism. Exp Eye Res. 2014;126:77–84.

Miceli MV, Newsome DA, Schriver GW. Glucose uptake, hexose monophosphate shunt activity, and oxygen consumption in cultured human retinal pigment epithelial cells. Investig Ophthalmol Vis Sci. 1990;31:277–83.

CAS   Google Scholar  

Palsson-McDermott EM, Curtis AM, Goel G, Lauterbach MAR, Sheedy FJ, Gleeson LE, et al. Pyruvate Kinase M2 regulates Hif-1alpha activity and IL-1beta induction and is a critical determinant of the Warburg effect in LPS-activated macrophages. Cell Metab. 2015;21:347.

Nordgaard CL, Karunadharma PP, Feng X, Olsen TW, Ferrington DA. Mitochondrial proteomics of the retinal pigment epithelium at progressive stages of age-related macular degeneration. Investig Ophthalmol Vis Sci. 2008;49:2848–55.

Article   Google Scholar  

Fisher CR, Ferrington DA. Perspective on AMD pathobiology: a bioenergetic crisis in the RPE. Investig Ophthalmol Vis Sci. 2018;59:AMD41–AMD7.

Article   CAS   Google Scholar  

Golestaneh N, Chu Y, Xiao YY, Stoleru GL, Theos AC. Dysfunctional autophagy in RPE, a contributing factor in age-related macular degeneration. Cell Death Dis. 2017;8:e2537.

Park S, Jeon JH, Min BK, Ha CM, Thoudam T, Park BY, et al. Role of the pyruvate dehydrogenase complex in metabolic remodeling: differential pyruvate dehydrogenase complex functions in metabolism. Diabetes Metab J. 2018;42:270–81.

Article   PubMed   PubMed Central   Google Scholar  

Do JY, Kim J, Kim MJ, Lee JY, Park SY, Yanai R, et al. Fursultiamine alleviates choroidal neovascularization by suppressing inflammation and metabolic reprogramming. Investig Ophthalmol Vis Sci. 2020;61:24.

Lee H, Jeon JH, Lee YJ, Kim MJ, Kwon WH, Chanda D, et al. Inhibition of pyruvate dehydrogenase kinase 4 in CD4(+) T Cells Ameliorates Intestinal Inflammation. Cell Mol Gastroenterol Hepatol. 2023;15:439–61.

Jeoung NH, Wu P, Joshi MA, Jaskiewicz J, Bock CB, Depaoli-Roach AA, et al. Role of pyruvate dehydrogenase kinase isoenzyme 4 (PDHK4) in glucose homoeostasis during starvation. Biochem J. 2006;397:417–25.

Hasegawa E, Inafuku S, Mulki L, Okunuki Y, Yanai R, Smith KE, et al. Cytochrome P450 monooxygenase lipid metabolites are significant second messengers in the resolution of choroidal neovascularization. Proc Natl Acad Sci USA. 2017;114:E7545–E53.

Yanai R, Mulki L, Hasegawa E, Takeuchi K, Sweigard H, Suzuki J, et al. Cytochrome P450-generated metabolites derived from omega-3 fatty acids attenuate neovascularization. Proc Natl Acad Sci USA. 2014;111:9603–8.

Nagineni CN, Raju R, Nagineni KK, Kommineni VK, Cherukuri A, Kutty RK, et al. Resveratrol suppresses expression of VEGF by human retinal pigment epithelial cells: potential nutraceutical for age-related macular degeneration. Aging Dis. 2014;5:88–100.

PubMed   PubMed Central   Google Scholar  

Thoudam T, Chanda D, Sinam IS, Kim BG, Kim MJ, Oh CJ, et al. Noncanonical PDK4 action alters mitochondrial dynamics to affect the cellular respiratory status. Proc Natl Acad Sci USA. 2022;119:e2120157119.

Oh H, Takagi H, Takagi C, Suzuma K, Otani A, Ishida K, et al. The potential angiogenic role of macrophages in the formation of choroidal neovascular membranes. Investig Ophthalmol Vis Sci. 1999;40:1891–8.

Grossniklaus HE, Ling JX, Wallace TM, Dithmar S, Lawson DH, Cohen C, et al. Macrophage and retinal pigment epithelium expression of angiogenic cytokines in choroidal neovascularization. Mol Vis. 2002;8:119–26.

PubMed   CAS   Google Scholar  

Min BK, Park S, Kang HJ, Kim DW, Ham HJ, Ha CM, et al. Pyruvate dehydrogenase kinase is a metabolic checkpoint for polarization of macrophages to the M1 phenotype. Front Immunol. 2019;10:944.

Min BK, Oh CJ, Park S, Lee JM, Go Y, Park BY, et al. Therapeutic effect of dichloroacetate against atherosclerosis via hepatic FGF21 induction mediated by acute AMPK activation. Exp Mol Med. 2019;51:1–12.

Park BY, Jeon JH, Go Y, Ham HJ, Kim JE, Yoo EK, et al. PDK4 deficiency suppresses hepatic glucagon signaling by decreasing cAMP levels. Diabetes. 2018;67:2054–68.

Lee D, Pagire HS, Pagire SH, Bae EJ, Dighe M, Kim M, et al. Discovery of novel pyruvate dehydrogenase kinase 4 inhibitors for potential oral treatment of metabolic diseases. J Med Chem. 2019;62:575–88.

Shukal D, Bhadresha K, Shastri B, Mehta D, Vasavada A, Johar K Sr. Dichloroacetate prevents TGFbeta-induced epithelial-mesenchymal transition of retinal pigment epithelial cells. Exp Eye Res. 2020;197:108072.

Nagineni CN, Kommineni VK, William A, Detrick B, Hooks JJ. Regulation of VEGF expression in human retinal cells by cytokines: implications for the role of inflammation in age-related macular degeneration. J Cell Physiol. 2012;227:116–26.

Nagineni CN, Detrick B, Hooks JJ. Synergistic effects of gamma interferon on inflammatory mediators that induce interleukin-6 gene expression and secretion by human retinal pigment epithelial cells. Clin Diagn Lab Immunol. 1994;1:569–77.

Kutty RK, Samuel W, Abay R, Cherukuri A, Nagineni CN, Duncan T, et al. Resveratrol attenuates CXCL11 expression induced by proinflammatory cytokines in retinal pigment epithelial cells. Cytokine. 2015;74:335–8.

Kanow MA, Giarmarco MM, Jankowski CS, Tsantilas K, Engel AL, Du J, et al. Biochemical adaptations of the retina and retinal pigment epithelium support a metabolic ecosystem in the vertebrate eye. Elife. 2017;6:e28899.

Rahman MH, Jha MK, Kim JH, Nam Y, Lee MG, Go Y, et al. Pyruvate dehydrogenase kinase-mediated glycolytic metabolic shift in the dorsal root ganglion drives painful diabetic neuropathy. J Biol Chem. 2016;291:6011–25.

Stacpoole PW. Therapeutic targeting of the pyruvate dehydrogenase complex/pyruvate dehydrogenase kinase (PDC/PDK) axis in cancer. J Natl Cancer Inst. 2017;109:1–14.

Jha MK, Jeon S, Suk K. Pyruvate dehydrogenase kinases in the nervous system: their principal functions in neuronal-glial metabolic interaction and neuro-metabolic disorders. Curr Neuropharmacol. 2012;10:393–403.

Stacpoole PW, Henderson GN, Yan Z, Cornett R, James MO. Pharmacokinetics, metabolism and toxicology of dichloroacetate. Drug Metab Rev. 1998;30:499–539.

Spruijt L, Naviaux RK, McGowan KA, Nyhan WL, Sheean G, Haas RH, et al. Nerve conduction changes in patients with mitochondrial diseases treated with dichloroacetate. Muscle Nerve. 2001;24:916–24.

Oh CJ, Kim MJ, Lee JM, Kim DH, Kim IY, Park S, et al. Inhibition of pyruvate dehydrogenase kinase 4 ameliorates kidney ischemia-reperfusion injury by reducing succinate accumulation during ischemia and preserving mitochondrial function during reperfusion. Kidney Int. 2023;104:724–39.

Iacovelli J, Rowe GC, Khadka A, Diaz-Aguilar D, Spencer C, Arany Z, et al. PGC-1alpha induces human RPE oxidative metabolism and antioxidant capacity. Investig Ophthalmol Vis Sci. 2016;57:1038–51.

Palsson-McDermott EM, Curtis AM, Goel G, Lauterbach MA, Sheedy FJ, Gleeson LE, et al. Pyruvate kinase M2 regulates Hif-1alpha activity and IL-1beta induction and is a critical determinant of the Warburg effect in LPS-activated macrophages. Cell Metab. 2015;21:65–80.

Download references

Acknowledgements

DHP is financially supported by the Basic Science Research Program of the National Research Foundation of Korea (NRF), funded by the Ministry of Science and ICT (MSIT) (RS-2024-00334982, 2019R1A2C1084371); the Information Technology Research Center (ITRC) support program funded by MSIT and supervised by the Institute of Information and Communications Technology Planning & Evaluation (IITP) (IITP-2024-2020-0-01808); the Korea Drug Development Fund (KDDF) funded by the MSIT, Ministry of Trade, Industry, and Energy, and Ministry of Health and Welfare (MOHW) (RS-2021-DD120784 (HN21C0923000021)); the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), funded by the MOHW (HR22C1832, RS-2024-00437643); and the BK21 FOUR KNU Convergence Educational Program of Biomedical Sciences for Creative Future Talents.

Author information

These authors contributed equally: Juhee Kim, Yujin Jeon.

Authors and Affiliations

Department of Ophthalmology, School of Medicine, Kyungpook National University, Kyungpook National University Hospital, Daegu, Republic of Korea

Juhee Kim, Yujin Jeon, Sungmi Park & Dong Ho Park

Kyungpook National University Cell & Matrix Research Institute, Daegu, Republic of Korea

Juhee Kim, Yujin Jeon & Dong Ho Park

Department of Biomedical Science, The Graduate School, Kyungpook National University, Daegu, Republic of Korea

Jinyoung Son & Dong Ho Park

BK21 FOUR KNU Convergence Educational Program of Biomedical Sciences for Creative Future Talents, Daegu, Republic of Korea

Department of Chemistry, Gwangju Institute of Science and Technology, Gwangju, Republic of Korea

Haushabhau S. Pagire, Suvarna H. Pagire & Jin Hee Ahn

R&D center, JD Bioscience Inc, Gwangju, Republic of Korea

Department of Ophthalmology and Visual Science, Nagoya City University Graduate School of Medical Sciences, Nagoya, Japan

Akiyoshi Uemura

Department of Internal Medicine, School of Medicine, Kyungpook National University, Kyungpook National University Hospital, Daegu, Republic of Korea

You can also search for this author in PubMed   Google Scholar

Contributions

SP and DHP designed the study and wrote the manuscript. JK and YJ performed the most experiments. JK, YJ, JS, HSP, SHP, JHA, AU, IL, SP, and DHP helped with data collection and assembly. JK, YJ, JS, SP, and DHP performed data analysis and interpretation. SP and DHP wrote the original draft, review, and editing with the help of all authors. All authors reviewed the manuscript.

Corresponding authors

Correspondence to Sungmi Park or Dong Ho Park .

Ethics declarations

Competing interests.

The authors declare no competing interests.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Edited by Massimiliano Agostini

Supplementary information

Supplementary information_clean version, full and uncropped western blots, rights and permissions.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ .

Reprints and permissions

About this article

Cite this article.

Kim, J., Jeon, Y., Son, J. et al. PDK4-mediated metabolic reprogramming is a potential therapeutic target for neovascular age-related macular degeneration. Cell Death Dis 15 , 582 (2024). https://doi.org/10.1038/s41419-024-06968-0

Download citation

Received : 19 April 2024

Revised : 30 July 2024

Accepted : 01 August 2024

Published : 09 August 2024

DOI : https://doi.org/10.1038/s41419-024-06968-0

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

Quick links

• Explore articles by subject
• Guide to authors
• Editorial policies

Subscribe or Renew

Create an E-mail Alert for This Article

Age-related macular degeneration, permissions, information & authors, metrics & citations, continue reading this article.

Select an option below:

Create your account to get 2 free subscriber-only articles each month.

Already have an account, print subscriber, supplementary material, information, published in.

• Clinical Medicine General
• Geriatrics/Aging General
• Ophthalmology General

Affiliations

Export citation.

Select the format you want to export the citation of this publication.

• Hsin-Ying Lin,
• Willy Chou,
• Tsair-Wei Chien,
• Yu-Tsen Yeh,
• Shu-Chun Kuo,
• Sheng-Yao Hsu,
• Joseph B. Lin,
• Andrea Santeford,
• Darksha Usmani,
• Aaditya V. Shah,
• Philip A. Ruzycki,
• Rajendra S. Apte,
• Jason J. Colasanti,
• Tzu Jui Wang,
• Caixia Wang,
• Xuejing Li,
• Jialiang Duan,
• Qingli Shang,
• Joewee Boon,
• William Rojas-Carabali,
• Yusra Asad,
• Jonathan Tzien Yih Lim,
• Rajesh Rajagopalan,
• Rupesh Agrawal,
• Navid Manafi,
• Alireza Mahmoudi,
• Mehdi Emamverdi,
• Giulia Corradetti,
• Stephanie Trejo Corona,
• Charles C. Wykoff,
• SriniVas R. Sadda,
• Qiongtao Song,
• Wenyuan Jian,
• Yuanyuan Zhang,
• Fuwen Zhang,
• Junguo Duan,
• Dongchun Xie,
• Junjiu Huang,
• Congyan Liu,
• Wenting Su,
• Huangqin Zhang,
• Yuping Liu,
• Xinrong Xu,
• Ida Korva-Gurung,
• Anna-Maria Kubin,
• Pasi Ohtonen,
• Nina Hautala,

View Options

View options, content link.

Copying failed.

PREVIOUS ARTICLE

Next article, create your account for 2 free subscriber-only articles each month. get free access now..

• For Authors
• Editorial Board
• Journals Home

To View More...

Purchase this article with an account.

• Defining the Macula Lutea
• Polyak and the ETDRS Grid
• Human Photoreceptor and RPE Topography
• A Multilayer View of Aging Relevant to OCT
• A Small Area at Highest AMD Risk Aligns With the Macula Lutea
• How Feeding the Fovea Throughout Life Leads to High-Risk Drusen
• How RMDA Probes the High-Risk Area
• Cone Resilience and Rod Vulnerability, a Center-Surround Model
• Considerations for Visual Function Testing in Geographic Atrophy
• Strengths, Limitations, Future Directions, and Conclusions
• Acknowledgments
• Christine A. Curcio Department of Ophthalmology and Visual Sciences, University of Alabama at Birmingham Heersink School of Medicine, Birmingham, Alabama, United States
• Deepayan Kar Department of Ophthalmology and Visual Sciences, University of Alabama at Birmingham Heersink School of Medicine, Birmingham, Alabama, United States
• Cynthia Owsley Department of Ophthalmology and Visual Sciences, University of Alabama at Birmingham Heersink School of Medicine, Birmingham, Alabama, United States
• Kenneth R. Sloan Department of Ophthalmology and Visual Sciences, University of Alabama at Birmingham Heersink School of Medicine, Birmingham, Alabama, United States
• Thomas Ach Department of Ophthalmology, University Hospital Bonn, Bonn, Germany
• Correspondence: Christine A. Curcio, Department of Ophthalmology and Visual Sciences, EyeSight Foundation of Alabama Vision Research Laboratories, 1670 University Boulevard, Room 360, University of Alabama at Birmingham, Heersink School of Medicine, Birmingham, AL 35294-0019, USA; [email protected] . 
• Footnotes   Current affiliation: *DK, Apellis Pharmaceuticals, Inc., Waltham MA, USA.
• Full Article

This feature is available to authenticated users only.

Christine A. Curcio , Deepayan Kar , Cynthia Owsley , Kenneth R. Sloan , Thomas Ach; Age-Related Macular Degeneration, a Mathematically Tractable Disease. Invest. Ophthalmol. Vis. Sci. 2024;65(3):4. https://doi.org/10.1167/iovs.65.3.4 .

Download citation file:

• Ris (Zotero)
• Reference Manager

© ARVO (1962-2015); The Authors (2016-present)

• Get Permissions
• Supplements

A progression sequence for age-related macular degeneration onset may be determinable with consensus neuroanatomical nomenclature augmented by drusen biology and eye-tracked clinical imaging. This narrative review proposes to supplement the Early Treatment of Diabetic Retinopathy Study (sETDRS) grid with a ring to capture high rod densities. Published photoreceptor and retinal pigment epithelium (RPE) densities in flat mounted aged-normal donor eyes were recomputed for sETDRS rings including near-periphery rich in rods and cumulatively for circular fovea-centered regions. Literature was reviewed for tissue-level studies of aging outer retina, population-level epidemiology studies regionally assessing risk, vision studies regionally assessing rod-mediated dark adaptation (RMDA), and impact of atrophy on photopic visual acuity. The 3 mm-diameter xanthophyll-rich macula lutea is rod-dominant and loses rods in aging whereas cone and RPE numbers are relatively stable. Across layers, the largest aging effects are accumulation of lipids prominent in drusen, loss of choriocapillary coverage of Bruch's membrane, and loss of rods. Epidemiology shows maximal risk for drusen-related progression in the central subfield with only one third of this risk level in the inner ring. RMDA studies report greatest slowing at the perimeter of this high-risk area. Vision declines precipitously when the cone-rich central subfield is invaded by geographic atrophy. Lifelong sustenance of foveal cone vision within the macula lutea leads to vulnerability in late adulthood that especially impacts rods at its perimeter. Adherence to an sETDRS grid and outer retinal cell populations within it will help dissect mechanisms, prioritize research, and assist in selecting patients for emerging treatments.

Supplemental Links

Related articles, from other journals, related topics.

• Eye Anatomy and Disorders

This PDF is available to Subscribers Only

You must be signed into an individual account to use this feature.

Login to your account

If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password

If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password

• Submit Article

Access provided by

Download started.

• Academic & Personal: 24 hour online access
• Corporate R&D Professionals: 24 hour online access
• Add To Online Library Powered By Mendeley
• Add To My Reading List
• Export Citation
• Create Citation Alert

Age-related macular degeneration

Purchase one-time access:

• For academic or personal research use, select 'Academic and Personal'
• For corporate R&D use, select 'Corporate R&D Professionals'
• Google Scholar
• Scopus (2538)
• Full Text PDF
• Izekenova AK
• Abikulova AK
• UN Department of Economic and Social Affairs
• GBD 2019 Blindness and Vision Impairment Collaborators
• Scopus (632)
• Bressler NM
• Wittenborn JS
• Scopus (132)
• Ferris 3rd, FL
• Wilkinson CP
• Scopus (940)
• Hodgson LAB
• Scopus (33)
• Rosenfeld PJ
• Engelbert M
• Scopus (322)
• Scopus (262)
• Scopus (70)
• Dansingani K
• Scopus (338)
• de Oliveira Dias JR
• Scopus (123)
• Scopus (28)
• Shinojima A
• Scopus (543)
• Scopus (389)
• Whitmore SS
• Scopus (232)
• Scopus (53)
• Braunack-Mayer L
• Wickremasinghe SS
• Scopus (105)
• Holekamp NM
• Scopus (13)
• Scopus (10)
• Varadaraj V
• Scopus (16)
• Christie AH
• Scopus (14)
• Fleckenstein M
• Scopus (173)
• Feeney-Burns L
• Hilderbrand ES
• Anderson DH
• Scopus (519)
• Scopus (141)
• Heesterbeek TJ
• Lorés-Motta L
• Lechanteur YTE
• den Hollander AI
• Scopus (128)
• Scopus (839)
• Rudnicka AR
• Scopus (210)
• Harrison RA
• Scopus (398)
• Christen WG
• Hankinson SE
• Cougnard-Grégoire A
• Scopus (65)
• Woodside JV
• Age-related eye disease study research group
• Age-related eye disease study 2 research group
• Scopus (863)
• Mauschitz MM
• Schmitz M-T
• Verzijden T
• The eye disease case-control study group
• Scopus (36)
• Scopus (89)
• Eshtiaghi A
• Sothivannan A
• Scopus (3554)
• Ritter 3rd, R
• Panhuysen C
• Scopus (2091)
• Scopus (1707)
• Scopus (2081)
• Scopus (733)
• Fritsche LG
• Grassmann F
• Scopus (27)
• Scopus (360)
• Gharahkhani P
• MacGregor S
• Scopus (34)
• Gregory-Roberts E
• Scopus (280)
• Gomez-Caraballo M
• Liakopoulos S
• Scopus (196)
• Scopus (191)
• Scopus (524)
• Scopus (275)
• Scopus (199)
• Scopus (216)
• Scopus (54)
• Chakravarthy U
• Scopus (30)
• Waldstein SM
• Bogunovic H
• Sadeghipour A
• Schmidt-Erfurth U
• Scopus (68)
• Veerappan M
• El-Hage-Sleiman AM
• Scopus (49)
• Balaratnasingam C
• Cherepanoff S
• Dolz-Marco R
• Scopus (47)
• Messinger JD
• Medeiros NE
• Scopus (278)
• Schmitz-Valckenberg S
• Steinberg JS
• Scopus (149)
• Pumariega NM
• Scopus (129)
• Killingsworth M
• Scopus (108)
• Delyfer M-N
• Scopus (38)
• Fletcher EL
• Greferath U
• Scopus (26)
• Domalpally A
• Abdelfattah NS
• Scopus (72)
• Brassington K
• Scopus (42)
• Flamendorf J
• Scopus (45)
• Scopus (112)
• Perepelkina T
• Scopus (448)
• Krzystolik MG
• Scopus (4798)
• Scopus (1666)
• Scopus (388)
• Quezada Ruiz C
• Scopus (3016)
• Scopus (219)
• Scopus (204)
• Scopus (58)
• Scopus (152)
• Scopus (67)
• McAllister IL
• Scopus (177)
• Grunwald JE
• Scopus (395)
• Scopus (43)
• Gonzalez-Buendia L
• Delgado-Tirado S
• Sanabria MR
• Fernandez I
• Campochiaro PA
• Scopus (55)
• Nanegrungsunk O
• Cabral de Guimaraes TA
• Daich Varela M
• Michaelides M
• Scopus (46)
• Scopus (179)
• Scopus (139)
• Titchener SA
• Dafgård Kopp E
• Scopus (230)
• Georgiadis O
• Scopus (380)
• Boulanger-Scemama E
• Scopus (178)
• Brassington KH
• Scopus (51)
• Lindsley KB
• Kalter-Leibovici O
• Scopus (84)
• Johnston RL
• Scopus (40)
• Scopus (90)
• Jeganathan VS
• Ambrecht A-M

Uncited References

• Varsamidis M
• Scopus (56)

Article info

Publication history, identification.

DOI: https://doi.org/10.1016/S0140-6736(22)02609-5

ScienceDirect

• Download Hi-res image
• Download .PPT

• Open access
• Published: 15 March 2024

Artificial intelligence in age-related macular degeneration: state of the art and recent updates

• Emanuele Crincoli 1 ,
• Riccardo Sacconi 2 ,
• Lea Querques 2 &
• Giuseppe Querques 2  

BMC Ophthalmology volume  24 , Article number:  121 ( 2024 ) Cite this article

1538 Accesses

3 Citations

Metrics details

Age related macular degeneration (AMD) represents a leading cause of vision loss and it is expected to affect 288 million people by 2040. During the last decade, machine learning technologies have shown great potential to revolutionize clinical management of AMD and support research for a better understanding of the disease. The aim of this review is to provide a panoramic description of all the applications of AI to AMD management and screening that have been analyzed in recent past literature. Deep learning (DL) can be effectively used to diagnose AMD, to predict short term risk of exudation and need for injections within the next 2 years. Moreover, DL technology has the potential to customize anti-VEGF treatment choice with a higher accuracy than expert human experts. In addition, accurate prediction of VA response to treatment can be provided to the patients with the use of ML models, which could considerably increase patients’ compliance to treatment in favorable cases. Lastly, AI, especially in the form of DL, can effectively predict conversion to GA in 12 months and also suggest new biomarkers of conversion with an innovative reverse engineering approach.

Peer Review reports

Introduction

Artificial intelligence (AI) has already revolutionized our way of living, and it is destinated to induce even more profound changes in many sectors of modern society. Among them, healthcare, and especially imaging based subspecialties such as ophthalmology, have the highest potential to benefit from integration of AI in everyday clinical setting. Two main approaches can be distinguished: Feature learning (FL) and Deep Learning (DL). While in the first one classification is based on predetermined variables, the deep learning approach is based on the ability of neural networks to identify differences between cases. The unknown nature of this differences is at the origin of the so-called black box effect, consisting in the uncertainty over the reliability of the classification performed by the model due to the lack of information about guiding elements. To solve this problem, many strategies have been attempted to explain and visualize the decisional mechanism hidden within each model (explainability in AI). DL models are versatile software that can be used for different tasks. In imaging research, DL can be used for two main purposes: segmentation of structures or classification of cases (or a combination of both). Both FL and DL models need to be validated on a separate population after the training phase is completed, a process that goes by the name of testing (or external validation). The generalizability of a model defines the applicability of the model to the general population. Accurate models can support clinical management of eye diseases, especially high prevalence ones. In particular, the use of machine learning (ML) techniques has been tested in ophthalmology in the fields of screening, diagnosis, clinical decision-making and prediction of prognosis with promising results. Age related macular degeneration (AMD) is a multifactorial disorder representing a leading cause of vision loss and it is expected to affect 288 million people by 2040 [ 1 ]. The aim of this review is to provide a panoramic description of all the applications of AI to AMD management and screening that have been analyzed in recent past literature.

Diagnosis of AMD

Predicting incipient amd.

An effort towards a better understanding of the diseases on a genetic point of view using machine learning methods was done by Yan et [ 2 ]., that compared the performance of 4 different machine learning techniques (neural network, lasso regression, support vector machine, and random forest) in assessing the risk of AMD on a database of more than 32,000 caucasian individuals. The analysis was also meant to assess feasibility of prediction of AMD risk using genome analysis. All models reached around 0.80 area under the curve (AUC) when tested on data from the same biobank and an AUC of around 0.70 when tested on a different biobank.

An interesting study from Lee et al. [ 3 ] used a deep learning model trained on fovea crossing optical coherence tomography (OCT) images to identify OCT biomarkers of delayed rod-mediated dark adaptation (RMDA), which is a known functional biomarker for incipient AMD. The model identified hyporeflective outer retinal bands on macular spectral domain (SD) OCT associated with delayed RMDA with an acceptable mean absolute error (MAE).

AMD automatic diagnosis

Several algorithms have been trained for automatic detection of AMD on various imaging modalities. Many of them were based on the segmentation and counting of drusen and drusen-like deposits and were aimed to identification of the disease at its early stage. Yildirim et al. [ 4 ] trained and tested a U-Net deep learning (DL) segmenter to the identification of early AMD OCT biomarkers. The model obtained very good accuracy proving its potential in facilitating AMD screening with the contribution of automatic patient selection. Morelle et al. [ 5 ] reported the results of an OCT segmenter based on DL technology that was able to quantify drusen load with excellent accuracy based on layer positions, achieving an exceptional correlation between drusen volumes estimated with this method and two expert human readers, and increasing the Dice score compared to a previous state-of-the-art method [ 6 ]. Other authors [ 7 ] proposed a DL framework to automatically distinguish drusen from reticular pseudodrusen (RPD) that was meant to prompt further understanding of RPD as a separate entity from drusen in both research and clinical settings. The model achieved > 90% accuracy in classification and segmentation, which was similar to human experts’ performance. Accurate identification of RPD was also confirmed by different authors [ 8 ].

Saha et al. [ 9 ] tested for AMD diagnostic performance different DL algorithms pretrained for detection and classification of hyperreflective foci, hyporeflective foci within the drusen, and subretinal drusenoid deposits from OCT B-scans. An overall accuracy of 87% for identifying the presence of early AMD biomarkers was achieved.

Despite the good diagnostic results obtained with drusen identification, as highlighted by Thakoor et al. [ 10 ], the best diagnostic performance was obtained by DL models using multimodal imaging as input, in particular when OCT B scan and OCT angiography (OCTA) acquisitions were provided to the software. Other authors demonstrated good results with a combination of OCT B scan and color fundus imaging [ 11 ].

In a metanalysis from Leng et al. [ 12 ], the type of AMD and the architecture of the DL model appeared to be the main reasons for heterogeneity of the results obtained in AMD diagnostic performance. In particular ResNet architecture was identified as the most suitable DL design for optimization of the task. In alternative, architectures with < 10 layers might be preferable to overcomplicated models not addressing the problem of vanishing gradients (which is brilliantly managed in ResNet).

FDA recently approved iPredict AMD, a DL screening tool available on the market that can detect referrable AMD with 88% accuracy. This tool can also predict individual risk score for development of late AMD within 1 and 2 years [ 13 ].

Predicting progression to late stage AMD and identifying late stage biomarkers

Several studies have proven good performance of DL in segmentation and quantification of subretinal and intraretinal fluid in exudative AMD [ 14 , 15 , 16 , 17 , 18 ].

Identification of macular atrophy for automatic diagnosis of advanced AMD has also been tested. Wei et al. [ 19 ] demonstrated high performance of a DL model in identification of 6 imaging features associated to macular atrophy in AMD patients. The selected features were the presence of interrupted outer retina and interrupted retinal pigmented epithelium (RPE), the absence of outer retina and RPE, and the presence of hypertransmission < or > 250 μm.

Other authors [ 20 ] presented a highly performing fully automated algorithm segmenting Retinal Pigment Epithelial and Outer Retinal Atrophy (RORA) in dry AMD on macular OCT. The results of the segmenter turned out to be comparable to the ones of expert human graders.

Assessment of the risk of progression from an uncomplicated form of AMD to a late-stage AMD (either neovascular or atrophic) was also attempted.

Schmidt-Erfurth et al. [ 21 ] elaborated a ML model using a combination of demographic, and genetic input features as well as automated volumetric segmentation of outer neurosensory layers and retinal pigment epithelium, drusen, and hyperreflective foci by spectral domain-OCT image analysis with the aim of assessing the risk of conversion to advanced AMD. While the model obtained good results in prediction GA development (AUC 0.80), macular neovascularization (MNV) development was not as reliably predicted (AUC 0.68). Bhuiyan et al. [ 13 ] used color fundus photographs of the patients from the AREDS study to train a DL model for automatic recognition of the stage of the disease (early/none vs. intermediate/late), obtaining a 99.2% accuracy. They then used this information combined with sociodemographic data to train a feature learning model to assess the risk of conversion towards a neovascular AMD or geographic atrophy (GA) during the follow up. The prediction model for a 2-year incident late AMD (any) achieved 86.36% accuracy, with significantly lower performance when specific type of late-AMD (either wet or dry) was to be detected. Burlina et al. [ 22 ] also discussed how DL technology could not only classify AMD cases with the 9-step AREDS severity scale as accurately as expert human graders, but also provide reliable 5-years prediction of evolution to late-stage disease.

Neovascular AMD

The risk of conversion to the neovascular form of the disease and exudation has also been evaluated using ML technologies. Benerjee et al. [ 23 ] proposed a Deep sequence approach combining imaging features, demographic, and visual factors, with a recursive neural network (RNN) model in the same platform to predict the risk of exudation in non-exudative AMD eyes in the short term (within 3 months) and long term (within 21 months). In particular, results in short term prediction appeared to have high generalizability when tested on an external dataset.

Prediction of the burden of treatment

The first study to predict anti vascular endothelial growth factor (VEGF) treatment needs in AMD was published in 2017 by Bogunovic et al. [ 24 ]. The authors demonstrated high accuracy of a model integrating a combination of baseline, 1-month and 2-months OCT features, initial best corrected visual acuity (BCVA) and demographic characteristics in predicting the burden of intravitreal injections (IVIs) of ranibizumab needed within a 2 years follow up in a pro re nata (PRN) regimen (data from the HARBOR study). Classification of low (≤ 5) and high (≥ 16) treatment requirement subgroups demonstrated around 75% accuracy, with the best prediction obtained for values at 2 months. Subretinal fluid volume in the central 3 mm was identified as the most relevant feature for prediction. Recently, Chandra et al. [ 25 ] used data from the Comparison of AMD Treatments Trials (CATT) to investigate the performance of 3 different feature learning (ML) models in prediction of the number of IVI needed in a pro re nata (PRN) regimen after the loading phase in the first 2 years of treatment. The outcome was evaluated both as total number of injections in two year and in a categorial manner, identifying patients who received few (≤ 8) or many (≥ 19) injections within the same follow up time. According to their results, the best performing model was the SVM, with an area under the curve (AUC) of around 0.80 in binary prediction of few/many injections. Important features included fluid in optical coherence tomography (intraretinal, subretinal, or sub-RPE), lesion characteristics, and treatment trajectory in the first three months. Baseline lesion characteristics included macular neovascularization (MNV) lesion area, lesion location (subfoveal or non-subfoveal), lesion composition (considering lesions such as MNV, hemorrhage, blocked fluorescence, and serous retinal pigment epithelial detachment), and lesion type (occult only, minimally classic, or predominantly classic).

Pfau et al. [ 26 ] proposed a probabilistic forecasting of the number of injections needed in a real life setting with 1 year follow up, demonstrating a mean absolute error (MAE) in prediction of the burden of anti-VEGF treatment frequency of around 2.6 injections /year.with the proposed model.

As concerns treat and extend (TE) regimen, the potential of feature learning (in particular random forest architecture) to predict high (< 5 weeks interval) or low (> 10 weeks interval) treatment demand in AMD, retinal vein occlusion (RVO) and diabetic macular edema (DME) was analyzed by Gallardo et al. [ 27 ] The AMD-trained models yielded an AUCs around 0.80 for both low and high demand. Even more importantly, this study revealed that it is possible to predict low demand reasonably well at the first visit, before the first injection.

Deep learning technology was also tested in its ability to predict the need for treatment. Romo-Bucheli et al. [ 28 ] proposed a DL model including DenseNet [ 29 ] structure and a RNN (trainable end-to-end) architecture to predict IVIs burden during a PRN regimen. The model predicted number of received injections with a concordance index of 0.7 and demonstrated a 0.85 (0.81) AUC in detecting the patients with low vs. high treatment requirements.

Lastly, Hwang et al. [ 30 ] demonstrated how a DL algorithm trained on 35,000 OCT images could learn to provide correct treatment indications, which is particularly interesting in primary care and telemedicine settings.

Predicting the choice of treatment and treatment results

In a 2023 publication, Moon et al. [ 31 ] reported the results of a DL model conceived to guide the clinician in the choice of treatment (aflibercept vs. ranibizumab). The model was trained on OCT images and its architecture was based on an attention generative adversarial network (GAN) model. They highlighted how the AI model predicted anti-VEGF agent-specific short-term treatment outcomes with higher sensitivity than both highly and less experienced human examiners, thus proving the potential advantages of its use in everyday clinical practice.

Machine learning technology may also help predict the visual outcomes of anti VEGF treatment. The performance of 5 different feature learning algorithms to this task was tested, showing the Lasso protocol as the best performing [ 32 ]. This model obtained a 5-letters mean absolute error in 3 months prediction and 8 letters mean absolute error in 12 months prediction. The authors discussed how a similar tool might increase compliance to treatment, especially when 12 months results are prospected to the patient. Fu et al. [ 33 ] obtained even higher performance in post-treatment VA prediction using DL technology, particularly in the form of an OCT segmenter providing biomarkers quantification and changes registration during the course of the treatement.

Geographic atrophy

Quantification of GA is extremely important for disease monitoring, analysis of risk factors for progression and evaluation of clinical endpoints. Moreover, accurate, repeatable and easy methods for GA area calculation would also help investigating structure-function correlation and elucidating pathophysiological mechanisms of disease development and progression. Balaskas et al. [ 34 ] demonstrated feasibility of residual visual acuity prediction using a random forest model trained with DL-segmented GA biomarkers on OCT images. The status of the foveal region (46.5%) and RPE-loss (31.1%) had greatest predictive importance for VA. For low luminance VA, however, non-foveal regions (74.5%) and photoreceptors’ degeneration (38.9%) were most important. Other authors demonstrated accurate segmentation of GA on fundus autofluorescence imaging [ 35 ].

Conversion to GA and GA progression

With an interesting and innovative concept of AI use in ophthalmology [ 36 ], Wang et al. [ 37 ] proposed a different approach to biomarkers identification, which was based on reverse engineering technology. In fact, the model was intended to identify new potential biomarkers of GA with the help of explainability methods. The reconstructions consistently highlighted that large foveal drusen and drusen clusters with or without mixed hyper-reflective focus lesion on baseline OCT were often present in eyes experiencing conversion to GA after 12 months.

Gigon et al. [ 38 ] proposed a DL method for automatic retinal pigment epithelial and outer retinal atrophy (RORA) progression prediction. The proposed software was based on enface multiple reconstructions of the status of the outer retina and provided continuous-time output. It was used to compute atrophy risk maps, which indicate time-to-RORA-conversion, that represents a novel and clinically relevant way of representing disease progression.

New perspectives

Natural language processing models have been shown to provide satisfactory responses to medical queries posed by AMD patients. In a recent study, Johnson et al. showed how Chat-Generative Pre-Trained Transformer(Chat-GTP) generated responses that were judged with a mean score of “almost completely correct” and a mean score of “complete and comprehensive” as concerns respectively accuracy and completeness [ 39 ]. The use of Generative adversarial networks (GANs)(consisting in two competing types of deep neural networks, including a generator and a discriminator), although still in its early phases, is showing promising potential applications in ophthalmology as described in an interesting review from You et al. These include, conversion, artifact removal, denoising and database expansion, which could be applied to AMD imaging to aid diagnosis and interpretation [ 40 ].

Conclusions

During the last decade, machine learning technologies have shown great potential to revolutionize clinical management of AMD and support research for a better understanding of the disease.

DL based diagnosis of AMD is easier when multimodal imaging serves as input (OCTA and OCT B scan), even though the approach based on drusen identification only may lead to satisfactory results with lower economic burden. The use of ResNet architecture is advisable to optimize diagnostic performance. Accurate diagnosis of referrable AMD and prediction of risk of development of advanced form of the disease within 1 and 2 years can be provided by a commercially available software recently approved by FDA to this scope (iPredict AMD, iHealthScreen).

As concerns neovascular AMD, short term risk of exudation may be effectively predicted using a combination of imaging and demographic and clinical information. Machine learning can also help predict the need for injections within the next 2 years. This can be achieved with both feature learning methos (among which the SVM technology might be the most suitable method) and DL methods. Prediction of few IVIs needed is particularly proficient and can be accurately predicted very early during the treatment (ideally before the first injection). The most relevant feature appears to be subretinal fluid volume in the central 3 mm, even though in general the unsupervised approach used by the DL methods may obtain better results in this type of task. Moreover, DL technology has the potential to customize treatment choice with a higher accuracy than expert human graders. In addition, accurate prediction of VA response to treatment can be provided to the patients with the use of ML models, which could considerably increase patients’ compliance to treatment in favorable cases. Considering the positive results, there is a good chance that in the next future treatment interval and choice for wet AMD will be supported by AI technology. In order to make the best out of this additional tool, this revolution will certainly require economic evaluation and adjustments in the procedures for management of wet AMD patients in the real life.

Lastly, AI, especially in the form of DL, can effectively predict conversion to GA in 12 months and also suggest new biomarkers of conversion with an innovative reverse engeneering approach.

Data availability

No datasets were generated or analysed during the current study.

Abbreviations

artificial intelligence

age-related macular degeneration

area under the curve

best corrected visual acuity

deep learning

diabetic macular edema

geographic atrophy

intravitreal injection

Mean absolute error

machine learning

macular neovascularization

optical coherence tomography

optical coherence tomography angiography

pro re nata

random forest

rod-mediated dark adaptation

Retinal Pigment Epithelial and Outer Retinal Atrophy

reticular pseudodrusen

retinal pigmented epithelium

recursive neural network

retinal vein occlusion

spectral domain

support-vector machine

treat and extend

vascular endothelial growth factor

Wong WL, Su X, Li X, Cheung CMG, Klein R, Cheng CY, et al. Global prevalence of age-related macular degeneration and disease burden projection for 2020 and 2040: a systematic review and meta-analysis. Lancet Global Health. 2014;2(2):e106–16.

Article   PubMed   Google Scholar  

Yan Q, Jiang Y, Huang H, Swaroop A, Chew EY, Weeks DE, et al. Genome-Wide Association Studies-Based Machine Learning for prediction of age-related Macular Degeneration Risk. Transl Vis Sci Technol. 2021;10(2):29.

Article   PubMed   PubMed Central   Google Scholar  

Lee AY, Lee CS, Blazes MS, Owen JP, Bagdasarova Y, Wu Y, et al. Exploring a structural basis for delayed rod-mediated Dark Adaptation in Age-Related Macular Degeneration Via Deep Learning. Transl Vis Sci Technol. 2020;9(2):62.

Yildirim K, Al-Nawaiseh S, Ehlers S, Schießer L, Storck M, Brix T, et al. U-net-based segmentation of current imaging biomarkers in OCT-Scans of patients with Age Related Macular Degeneration. Stud Health Technol Inf. 2023;302:947–51.

Google Scholar  

Morelle O, Wintergerst MWM, Finger RP, Schultz T. Accurate drusen segmentation in optical coherence tomography via order-constrained regression of retinal layer heights. Sci Rep. 2023;13(1):8162.

Article   ADS   CAS   PubMed   PubMed Central   Google Scholar  

Pondorfer SG, Wintergerst MWM, Gorgi Zadeh S, Schultz T, Heinemann M, Holz FG, et al. Association of Visual Function Measures with Drusen volume in early stages of age-related Macular Degeneration. Invest Ophthalmol Vis Sci. 2020;61(3):55.

Schwartz R, Khalid H, Liakopoulos S, Ouyang Y, de Vente C, González-Gonzalo C, et al. A Deep Learning Framework for the detection and quantification of Reticular Pseudodrusen and Drusen on Optical Coherence Tomography. Transl Vis Sci Technol. 2022;11(12):3.

Chen Q, Keenan TDL, Allot A, Peng Y, Agrón E, Domalpally A, et al. Multimodal, Multitask, multiattention (M3) deep learning detection of reticular pseudodrusen: toward automated and accessible classification of age-related macular degeneration. J Am Med Inf Assoc. 2021;28(6):1135–48.

Article   CAS   Google Scholar  

Saha S, Nassisi M, Wang M, Lindenberg S, Kanagasingam Y, Sadda S, et al. Automated detection and classification of early AMD biomarkers using deep learning. Sci Rep. 2019;9(1):10990.

Article   ADS   PubMed   PubMed Central   Google Scholar  

Thakoor KA, Yao J, Bordbar D, Moussa O, Lin W, Sajda P, et al. A multimodal deep learning system to distinguish late stages of AMD and to compare expert vs. AI ocular biomarkers. Sci Rep. 2022;12(1):2585.

Yoo TK, Choi JY, Seo JG, Ramasubramanian B, Selvaperumal S, Kim DW. The possibility of the combination of OCT and fundus images for improving the diagnostic accuracy of deep learning for age-related macular degeneration: a preliminary experiment. Med Biol Eng Comput. 2019;57(3):677–87.

Leng X, Shi R, Wu Y, Zhu S, Cai X, Lu X, et al. Deep learning for detection of age-related macular degeneration: a systematic review and meta-analysis of diagnostic test accuracy studies. PLoS ONE. 2023;18(4):e0284060.

Article   CAS   PubMed   PubMed Central   Google Scholar  

Bhuiyan A, Wong TY, Ting DSW, Govindaiah A, Souied EH, Smith RT. Artificial Intelligence to stratify severity of age-related Macular Degeneration (AMD) and predict risk of progression to late AMD. Transl Vis Sci Technol. 2020;9(2):25.

Schlegl T, Waldstein SM, Bogunovic H, Endstraßer F, Sadeghipour A, Philip AM, et al. Fully Automated Detection and Quantification of Macular Fluid in OCT using deep learning. Ophthalmology. 2018;125(4):549–58.

Habra O, Gallardo M, Meyer Zu Westram T, De Zanet S, Jaggi D, Zinkernagel M, et al. Evaluation of an Artificial Intelligence-based detector of sub- and Intraretinal Fluid on a large set of optical coherence tomography volumes in age-related Macular Degeneration and Diabetic Macular Edema. Ophthalmologica. 2022;245(6):516–27.

Article   CAS   PubMed   Google Scholar  

Keenan TDL, Clemons TE, Domalpally A, Elman MJ, Havilio M, Agrón E, et al. Retinal specialist versus Artificial Intelligence Detection of Retinal Fluid from OCT: Age-Related Eye Disease Study 2: 10-Year Follow-On study. Ophthalmology. 2021;128(1):100–9.

López-Varela E, Vidal PL, Pascual NO, Novo J, Ortega M. Fully-automatic 3D intuitive visualization of age-related Macular Degeneration Fluid accumulations in OCT cubes. J Digit Imaging. 2022;35(5):1271–82.

Sappa LB, Okuwobi IP, Li M, Zhang Y, Xie S, Yuan S, et al. RetFluidNet: retinal fluid segmentation for SD-OCT images using convolutional neural network. J Digit Imaging. 2021;34(3):691–704.

Wei W, Southern J, Zhu K, Li Y, Cordeiro MF, Veselkov K. Deep learning to detect macular atrophy in wet age-related macular degeneration using optical coherence tomography. Sci Rep. 2023;13(1):8296.

Derradji Y, Mosinska A, Apostolopoulos S, Ciller C, De Zanet S, Mantel I. Fully-automated atrophy segmentation in dry age-related macular degeneration in optical coherence tomography. Sci Rep. 2021;11(1):21893.

Schmidt-Erfurth U, Waldstein SM, Klimscha S, Sadeghipour A, Hu X, Gerendas BS, et al. Prediction of Individual Disease Conversion in Early AMD using Artificial Intelligence. Invest Ophthalmol Vis Sci. 2018;59(8):3199–208.

Burlina PM, Joshi N, Pacheco KD, Freund DE, Kong J, Bressler NM. Use of Deep Learning for detailed severity characterization and estimation of 5-Year risk among patients with age-related Macular Degeneration. JAMA Ophthalmol. 2018;136(12):1359–66.

Banerjee I, de Sisternes L, Hallak JA, Leng T, Osborne A, Rosenfeld PJ, et al. Prediction of age-related macular degeneration disease using a sequential deep learning approach on longitudinal SD-OCT imaging biomarkers. Sci Rep. 2020;10(1):15434.

Bogunovic H, Waldstein SM, Schlegl T, Langs G, Sadeghipour A, Liu X, et al. Prediction of Anti-VEGF treatment requirements in Neovascular AMD using a machine learning Approach. Invest Ophthalmol Vis Sci. 2017;58(7):3240–8.

Chandra RS, Ying GS. Evaluation of multiple machine learning models for Predicting Number of Anti-VEGF injections in the comparison of AMD Treatment Trials (CATT). Transl Vis Sci Technol. 2023;12(1):18.

Pfau M, Sahu S, Rupnow RA, Romond K, Millet D, Holz FG, et al. Probabilistic forecasting of Anti-VEGF treatment frequency in Neovascular Age-Related Macular Degeneration. Transl Vis Sci Technol. 2021;10(7):30.

Gallardo M, Munk MR, Kurmann T, De Zanet S, Mosinska A, Karagoz IK, et al. Machine learning can predict Anti-VEGF treatment demand in a treat-and-extend regimen for patients with Neovascular AMD, DME, and RVO Associated Macular Edema. Ophthalmol Retina. 2021;5(7):604–24.

Romo-Bucheli D, Erfurth US, Bogunovic H. End-to-end Deep Learning Model for Predicting Treatment requirements in Neovascular AMD from longitudinal retinal OCT imaging. IEEE J Biomed Health Inf. 2020;24(12):3456–65.

Article   Google Scholar  

Sarkar A, Medium. 2020 [cited 2022 Oct 18]. Creating DenseNet 121 with TensorFlow. Available from: https://towardsdatascience.com/creating-densenet-121-with-tensorflow-edbc08a956d8

Hwang DK, Hsu CC, Chang KJ, Chao D, Sun CH, Jheng YC, et al. Artificial intelligence-based decision-making for age-related macular degeneration. Theranostics. 2019;9(1):232–45.

Moon S, Lee Y, Hwang J, Kim CG, Kim JW, Yoon WT, et al. Prediction of anti-vascular endothelial growth factor agent-specific treatment outcomes in neovascular age-related macular degeneration using a generative adversarial network. Sci Rep. 2023;13(1):5639.

Rohm M, Tresp V, Müller M, Kern C, Manakov I, Weiss M, et al. Predicting Visual Acuity by using machine learning in patients treated for Neovascular Age-Related Macular Degeneration. Ophthalmology. 2018;125(7):1028–36.

Fu DJ, Faes L, Wagner SK, Moraes G, Chopra R, Patel PJ, et al. Predicting Incremental and Future Visual Change in Neovascular Age-Related Macular Degeneration using deep learning. Ophthalmol Retina. 2021;5(11):1074–84.

Balaskas K, Glinton S, Keenan TDL, Faes L, Liefers B, Zhang G, et al. Prediction of visual function from automatically quantified optical coherence tomography biomarkers in patients with geographic atrophy using machine learning. Sci Rep. 2022;12(1):15565.

Treder M, Lauermann JL, Eter N. Deep learning-based detection and classification of geographic atrophy using a deep convolutional neural network classifier. Graefes Arch Clin Exp Ophthalmol. 2018;256(11):2053–60.

Crincoli E, Sacconi R, Querques G. Reshaping the use of artificial intelligence in Ophthalmology: sometimes you need to go backwards. RETINA. 2022. https://doi.org/10.1097/IAE.0000000000003878

Wang S, Wang Z, Vejalla S, Ganegoda A, Nittala MG, Sadda SR, et al. Reverse engineering for reconstructing baseline features of dry age-related macular degeneration in optical coherence tomography. Sci Rep. 2022;12:22620.

Gigon A, Mosinska A, Montesel A, Derradji Y, Apostolopoulos S, Ciller C, et al. Personalized atrophy risk mapping in Age-Related Macular Degeneration. Transl Vis Sci Technol. 2021;10(13):18.

Johnson D, Goodman R, Patrinely J, Stone C, Zimmerman E, Donald R et al. Assessing the Accuracy and Reliability of AI-Generated Medical Responses: An Evaluation of the Chat-GPT Model. Res Sq. 2023;rs.3.rs-2566942.

You A, Kim JK, Ryu IH, Yoo TK. Application of generative adversarial networks (GAN) for ophthalmology image domains: a survey. Eye Vis (Lond). 2022;9(1):6.

Download references

The authors did not receive support from any organization for the submitted work.

Author information

Authors and affiliations.

Ophthalmology Unit, “Fondazione Policlinico Universitario A. Gemelli IRCCS”, Rome, Italy

Emanuele Crincoli

Department of Ophthalmology, University Vita-Salute IRCCS San Raffaele Scientific Institute, Via Olgettina, 60, 20132, Milan, Italy

Riccardo Sacconi, Lea Querques & Giuseppe Querques

You can also search for this author in PubMed   Google Scholar

Contributions

Conceptualization, E.C., G.Q.; Methodology, E.C., R.S., L.Q. Validation, G.Q., R.S..Writing—original draft preparation, E.C., L.Q., R.S.; Writing—review and editing, E.C., G.Q., R.S., L.Q.; Supervision, G.Q.; Project administration G.Q.

Corresponding author

Correspondence to Giuseppe Querques .

Ethics declarations

Ethics approval and consent to participate.

Not applicable (no human subject involved in this research).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/ . The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/ ) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Cite this article.

Crincoli, E., Sacconi, R., Querques, L. et al. Artificial intelligence in age-related macular degeneration: state of the art and recent updates. BMC Ophthalmol 24 , 121 (2024). https://doi.org/10.1186/s12886-024-03381-1

Download citation

Received : 05 December 2023

Accepted : 06 March 2024

Published : 15 March 2024

DOI : https://doi.org/10.1186/s12886-024-03381-1

Share this article

Anyone you share the following link with will be able to read this content:

Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative

• Artificial intelligence
• Age-related macular degeneration
• Deep learning
• Support vector machine

BMC Ophthalmology

ISSN: 1471-2415

• General enquiries: [email protected]

Select type of work

Gw Work (Non-ETD)

Electronic Thesis/Dissertation

Journal Issue

Understanding the Pathogenesis of Age-related Macular Degeneration

Downloadable content.

Age related macular degeneration (AMD) is the leading cause of central vision loss in the elderly population. The disease affects the macula and it can occur in "dry" or "wet" form. The dry AMD is the most prevalent form and is characterized by accumulation of extracellular deposits named "drusen" between the retinal pigment epithelium (RPE) and Bruch's membrane while the wet AMD is less common form and is characterized with growth of new blood vessels that breach the RPE monolayer and invade the photoreceptors. AMD is a complex disease in which age, genetic variants, and environmental factors are all considered to play a role. Although aging is known as a major risk factor for AMD, it has been shown that more than 75 % of AMD patients harbor certain genetic variations. Three major genetic risk factors have been identified to date and these include variant in the complement factor H (CFH)gene, in the promoter region of the high temperature requirement protease 1 (HTRA1)gene and in the Age-related maculopathy susceptibility 2 (ARMS2) gene. However it is not clear how these apparently unrelated genes lead to the same pathological features of AMD (e.g. drusen formation and choroidal neovascularization). The goal of this dissertation is to understand the downstream consequences of these polymorphisms on RPE cells and define molecular mechanisms of AMD pathogenesis. We hypothesized that AMD risk genotypes influence homeostasis of RPE secretome leading to progressive drusen formation and macular degeneration. To test this hypothesis: (i) we established primary RPE cell cultures from human autopsy eyes of donors with and without AMD risk genotypes, (ii) compared their secretome profiles using stable isotope labeling by amino acid in cell culture (SILAC) strategy, and (iii) characterized key pathways linking the different risk genotypes to drusen formation and choroidal neovascularization. Our SILAC strategy readily facilitated high throughput screening for both detection and quantitation of RPE secreted proteins. In our first study of the RPE secretome, we found that RPE cells derived from AMD donors secrete elevated amounts of protein components found in drusen and these include complement components, amyloid, clusterin, fibronectin and TIPM-3 (Chapter 2). We also found that TNF-α, a pleiotropic cytokine, modulates secretion of specific proteins of RPE cells leading to dysregulation of the complement pathway and extracellular matrix remodeling (Chapter 3). Moreover, we found that RPE culture with HTRA1/ARMS2 risk genotype secrete elevated amounts of HTRA1 and that HTRA1 cleaves key proteins involved in regulation of the complement pathway (e.g. clusterin, vitronectin and fibromodulin) and in amyloid deposition (clusterin, alpha 2 macroglobulin and ADAM9). We propose models where alteration of RPE extracellular environment due to genetic variants and external stimuli favors amyloid accumulation and complement deposition leading to progressive drusen buildup and ocular cell degeneration.

• An, Eunkyung
• Bioinformatics
• Age-related Macular Degeneration
• Dissertation
• In Copyright
• Biomedical Sciences
• Hoffman, Eric P
• Hathout, Yetrib
• Colberg-Poley, Anamaris M
• Nagaraju, Kanneboyina
• Goldstein, Allan
• Rose, Mary C
•  https://scholarspace.library.gwu.edu/etd/b8515n42v

Notice to Authors

If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.

• Eye Diseases
• Ophthalmology

Age-related macular degeneration

• Nature Reviews Disease Primers 7(1):31

• University of Utah

• National Eye Institute
• This person is not on ResearchGate, or hasn't claimed this research yet.

• Queen's University Belfast

Discover the world's research

• 25+ million members
• 160+ million publication pages
• 2.3+ billion citations
• ARTIF INTELL REV

• Ibrahim Saleh
• Ayman El-Baz

• Robyn H Guymer
• Christian Quaia
• Richard J Krauzlis
• Lukas Schloesser
• Sara M Klose

• IMMUNOTHERAPY-UK
• Carl Danzig

• Anat Loewenstein
• OPHTHALMOLOGY

• PROG RETIN EYE RES

• Steffen Schmitz-Valckenberg

• Keith Westby
• Kourous A. Rezaei
• INT J MOL SCI
• Luis García-Onrubia
• Fco. Javier Valentín-Bravo

• James T. Handa

• Recruit researchers
• Join for free
• Login Email Tip: Most researchers use their institutional email address as their ResearchGate login Password Forgot password? Keep me logged in Log in or Continue with Google Welcome back! Please log in. Email · Hint Tip: Most researchers use their institutional email address as their ResearchGate login Password Forgot password? Keep me logged in Log in or Continue with Google No account? Sign up

• Consulting Editors
• For authors
• Publication ethics
• Publication alerts by email
• Advertising

GSK3 inhibition reduces ECM production and prevents age-related macular degeneration–like pathology

Malattia Leventinese/Doyne honeycomb retinal dystrophy (ML/DHRD) is an age-related macular degeneration–like (AMD-like) retinal dystrophy caused by an autosomal dominant R345W mutation in the secreted glycoprotein, fibulin-3 (F3). To identify new small molecules that reduce F3 production in retinal pigmented epithelium (RPE) cells, we knocked-in a luminescent peptide tag (HiBiT) into the endogenous F3 locus that enabled simple, sensitive, and high-throughput detection of the protein. The GSK3 inhibitor, CHIR99021 (CHIR), significantly reduced F3 burden (expression, secretion, and intracellular levels) in immortalized RPE and non-RPE cells. Low-level, long-term CHIR treatment promoted remodeling of the RPE extracellular matrix, reducing sub-RPE deposit-associated proteins (e.g., amelotin, complement component 3, collagen IV, and fibronectin), while increasing RPE differentiation factors (e.g., tyrosinase, and pigment epithelium-derived factor). In vivo, treatment of 8-month-old R345W+/+ knockin mice with CHIR (25 mg/kg i.p., 1 mo) was well tolerated and significantly reduced R345W F3-associated AMD-like basal laminar deposit number and size, thereby preventing the main pathological feature in these mice. This is an important demonstration of small molecule–based prevention of AMD-like pathology in ML/DHRD mice and may herald a rejuvenation of interest in GSK3 inhibition for the treatment of retinal degenerative diseases, including potentially AMD itself.

Sophia M. DiCesare, Antonio J. Ortega, Gracen E. Collier, Steffi Daniel, Krista N. Thompson, Melissa K. McCoy, Bruce A. Posner, John D. Hulleman

Full Text PDF | Download (11.81 MB)

Sign up for email alerts

An official website of the United States government

The .gov means it’s official. Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

The site is secure. The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

• Publications
• Account settings
• My Bibliography
• Collections
• Citation manager

Save citation to file

Email citation, add to collections.

• Create a new collection
• Add to an existing collection

Add to My Bibliography

Your saved search, create a file for external citation management software, your rss feed.

• Search in PubMed
• Search in NLM Catalog
• Add to Search

Juvenile Macular Degenerations

Affiliations.

• 1 Departamento de Oftalmología, Escuela de Medicina, Pontificia, Universidad Católica de Chile, Santiago, Chile. Electronic address: [email protected].
• 2 Department of Ophthalmology, Children׳s Hospital, Boston, MA; Department of Ophthalmology, Harvard Medical School, Boston, MA.
• 3 Department of Ophthalmology, Children׳s Hospital, Boston, MA.
• PMID: 28941524
• PMCID: PMC5709045
• DOI: 10.1016/j.spen.2017.05.005

In this article, we review the following 3 common juvenile macular degenerations: Stargardt disease, X-linked retinoschisis, and Best vitelliform macular dystrophy. These are inherited disorders that typically present during childhood, when vision is still developing. They are sufficiently common that they should be included in the differential diagnosis of visual loss in pediatric patients. Diagnosis is secured by a combination of clinical findings, optical coherence tomography imaging, and genetic testing. Early diagnosis promotes optimal management. Although there is currently no definitive cure for these conditions, therapeutic modalities under investigation include pharmacologic treatment, gene therapy, and stem cell transplantation.

Copyright © 2017 Elsevier Inc. All rights reserved.

PubMed Disclaimer

Fundus photos showing a normal…

Fundus photos showing a normal macula and examples of the typical macular findings…

Spectral domain ocular coherence tomography…

Spectral domain ocular coherence tomography (SD-OCT) scans showing a normal macula and examples…

Similar articles

• [Significance of ophthalmological imaging in common hereditary retinal diseases]. Kortüm K, Kernt M, Reznicek L. Kortüm K, et al. Klin Monbl Augenheilkd. 2013 Mar;230(3):223-31. doi: 10.1055/s-0032-1327906. Epub 2012 Dec 10. Klin Monbl Augenheilkd. 2013. PMID: 23229225 Review. German.
• Clinical and imaging findings of pattern dystrophy subtypes; Diagnostic errors and unnecessary treatment in clinical practice. Ozkaya A, Garip R, Nur Tarakcioglu H, Alkin Z, Taskapili M. Ozkaya A, et al. J Fr Ophtalmol. 2018 Jan;41(1):21-29. doi: 10.1016/j.jfo.2017.06.009. Epub 2017 Nov 28. J Fr Ophtalmol. 2018. PMID: 29195727
• Novel mutation in BEST1 associated with retinoschisis. Silva RA, Berrocal AM, Lam BL, Albini TA. Silva RA, et al. JAMA Ophthalmol. 2013 Jun;131(6):794-8. doi: 10.1001/jamaophthalmol.2013.2047. JAMA Ophthalmol. 2013. PMID: 23572118 No abstract available.
• Optical coherence tomography for multifocal vitelliform macular dystrophy. Xu H, Ying L, Lin P, Wu J. Xu H, et al. Optom Vis Sci. 2013 Jan;90(1):94-9. doi: 10.1097/OPX.0b013e318278fc5e. Optom Vis Sci. 2013. PMID: 23238259
• Novel therapeutics for Stargardt disease. Lu LJ, Liu J, Adelman RA. Lu LJ, et al. Graefes Arch Clin Exp Ophthalmol. 2017 Jun;255(6):1057-1062. doi: 10.1007/s00417-017-3619-8. Epub 2017 Mar 11. Graefes Arch Clin Exp Ophthalmol. 2017. PMID: 28285324 Review.
• The role of epigenetic methylation/demethylation in the regulation of retinal photoreceptors. Lu CF, Zhou YN, Zhang J, Su S, Liu Y, Peng GH, Zang W, Cao J. Lu CF, et al. Front Cell Dev Biol. 2023 May 26;11:1149132. doi: 10.3389/fcell.2023.1149132. eCollection 2023. Front Cell Dev Biol. 2023. PMID: 37305686 Free PMC article. Review.
• The Potential Role of Epigenetic Mechanisms in the Development of Retinitis Pigmentosa and Related Photoreceptor Dystrophies. Dvoriantchikova G, Lypka KR, Ivanov D. Dvoriantchikova G, et al. Front Genet. 2022 Mar 11;13:827274. doi: 10.3389/fgene.2022.827274. eCollection 2022. Front Genet. 2022. PMID: 35360866 Free PMC article.
• The Retina: A Window into the Brain. Ptito M, Bleau M, Bouskila J. Ptito M, et al. Cells. 2021 Nov 23;10(12):3269. doi: 10.3390/cells10123269. Cells. 2021. PMID: 34943777 Free PMC article.
• Behavioural Adaptation to Hereditary Macular Dystrophy: A Systematic Review on the Effect of Early Onset Central Field Loss on Peripheral Visual Abilities. Baig A, Buckley D, Codina C. Baig A, et al. Br Ir Orthopt J. 2021 Jun 16;17(1):104-118. doi: 10.22599/bioj.177. eCollection 2021. Br Ir Orthopt J. 2021. PMID: 34278226 Free PMC article. Review.
• Eye Movements in Macular Degeneration. Verghese P, Vullings C, Shanidze N. Verghese P, et al. Annu Rev Vis Sci. 2021 Sep 15;7:773-791. doi: 10.1146/annurev-vision-100119-125555. Epub 2021 May 26. Annu Rev Vis Sci. 2021. PMID: 34038144 Free PMC article. Review.
• Hendrickson AE, Yuodelis C. The morphological development of the human fovea. Ophthalmology. 1984;91:603–12. - PubMed
• Lee H, Purohit R, Patel A, Papageorgiou E, Sheth V, Maconachie G, et al. In Vivo Foveal Development Using Optical Coherence Tomography. Invest Ophthalmol Vis Sci. 2015;56:4537–45. - PubMed
• Yuodelis C, Hendrickson A. A qualitative and quantitative analysis of the human fovea during development. Vision Res. 1986;26:847–55. - PubMed
• Hendrickson A, Drucker D. The development of parafoveal and mid-peripheral human retina. Behav Brain Res. 1992;49:21–31. - PubMed
• Hendrickson A, Possin D, Vajzovic L, Toth CA. Histologic development of the human fovea from midgestation to maturity. Am J Ophthalmol. 2012;154:767–78. e2. - PMC - PubMed

Publication types

• Search in MeSH

Related information

Grants and funding.

• R43 EY018509/EY/NEI NIH HHS/United States
• R44 EY018509/EY/NEI NIH HHS/United States

LinkOut - more resources

Full text sources.

• Elsevier Science
• Europe PubMed Central
• PubMed Central

Other Literature Sources

• scite Smart Citations
• MedlinePlus Health Information
• Citation Manager

NCBI Literature Resources

MeSH PMC Bookshelf Disclaimer

The PubMed wordmark and PubMed logo are registered trademarks of the U.S. Department of Health and Human Services (HHS). Unauthorized use of these marks is strictly prohibited.

Clinical Trials in Age-Related Macular Degeneration Treatment

Publisher description.

This book is a compendium of the worldwide ocular stem cell, gene therapy, pharmaceutical, and other miscellaneous studies treating Age-Related Macular Degeneration registered with Clinicaltrials.gov. Clinicaltrials.gov is the largest website listing of registered clinical research studies in the world. The study and the clinical trial numbers are provided in order to make it easier for the reader to obtain further information. The book also includes an introduction on Age-Related Macular Degeneration as well as analysis of the clinical studies.   Clinical Trials in Age-Related Macular Degeneration Treatment is a valuable resource for ophthalmologists, optometrists, other physicians, and researchers.

More Books by Jeffrey N. Weiss