Retina Induced Expression of VEGFC, ANGPT, and EFNB2 and Their Receptors Characterizes Neovascularization in Proliferative Diabetic Retinopathy

Yaping Li,1,2 Dong Chen,3,4 Luguo Sun,1 Yannan Wu,1 Ying Zou,2 Chen Liang,1 Yongli Bao,1 Jingwen Yi,1 Yu Zhang,3 Jing Hou,3 Zhen Li,3 Fengyun Yu,4 Yanxin Huang,1 Chunlei Yu,1 Lei Liu,1 Zaoxia Liu,2 Yi Zhang,3,4 and Yuxin Li1 1National Engineering Laboratory for Druggable and Screening, Northeast Normal University, Changchun, China 2Department of Ophthalmology, Second Hospital, Jilin University, Changchun, China 3Center for Genome Analysis, ABLife, Inc., Wuhan, People’s Republic of China 4Laboratory for Genome Regulation and Human Health, ABLife, Inc., Wuhan, People’s Republic of China

Correspondence: Yuxin Li, National PURPOSE. To investigate whole transcriptional differences between proliferative diabetic Engineering Laboratory for Druggable retinopathy (PDR) neovascular membranes (NVMs) and retinas, and the regulatory Gene and Protein Screening, Northeast participating in retinal neovascularization in PDR. Normal University, 2555 Jingyue Ave- nue, Changchun 130117, China; METHODS. We used high-throughput sequencing technology to capture the whole-genome [email protected]. levels of all participants, including 23 patients with PDR or branch retinal Yi Zhang, Center for Genome Anal- vein occlusion (BRVO), 3 normal retinal samples, and 2 retinal samples from type II diabetic ysis, ABLife, Inc., 388 Gaoxin 2nd (T2D) eyes by donation, followed by analyses of expression patterns using bioinformatics Road, East Lake High-Tech Develop- methods, then validation of the data by in situ hybridization and Western blotting. ment Zone, Wuhan 430075, People’s Republic of China; RESULTS. We showed that transcriptional profiles of the NVMs were distinct from those of the [email protected]. retinas. growth factors VEGFC, ANGPT1, ANGPT2, and EFNB2, and their Zaoxia Liu, Department of Ophthal- receptors FLT4, TIE1, TIE2, and EPHB4, respectively, were overexpressed. Expression of mology, Second Hospital, Jilin Uni- VEGFA was highly upregulated in T2D retina, but low in the NVMs, while angiogenesis versity, Changchun 130021, China; transcription factors, including ETS1 and ERG, were coordinately upregulated in NVMs. [email protected]. LeiLiu,NationalEngineeringLabo- CONCLUSIONS. This study described a PDR neovascularization model in which pathological ratoryforDruggableGeneandPro- retina-secreted vascular endothelial A (VEGFA) enhanced the expression of a set tein Screening, Northeast Normal of angiogenesis transcription factors and growth factors, to cooperatively induce the retinal University, 2555 Jingyue Avenue, neovascularization. Based on these results, novel potential therapeutic targets and biomarkers Changchun 130117, China; for PDR treatment and diagnosis are suggested. [email protected]. Keywords: proliferative diabetic retinopathy, transcriptome profile, VEGF, ETS transcription Submitted: January 29, 2019 factors, neovascularization Accepted: August 16, 2019 Citation: Li Y, Chen D, Sun L, et al. Induced expression of VEGFC, ANGPT, and EFNB2 and their recep- tors characterizes neovascularization in proliferative diabetic retinopathy. Invest Ophthalmol Vis Sci. 2019;60:4084–4096. https://doi.org/ 10.1167/iovs.19-26767

iabetes mellitus is mainly caused by impaired in the 1980s,10,11 belonging to a gene family that includes D production or action, and does significant long-term VEGFB, VEGFC, VEGFD, and (PlGF).12 damage to many organs, including eyes.1,2 Diabetics have high These , as well as other vascular endothelium-specific risks of macrovascular and microvascular complications in- growth factors, play differential and cooperative roles in duced by oxidative stress.3 Diabetic retinopathy (DR) is one of promoting and regulating angiogenesis or neovascularization, the most common microvascular complications of diabetes.4,5 by interacting with their receptors and then stimulating the The advanced stage of DR, proliferative diabetic retinopathy following processes.12–17 (PDR) with retinal neovascularization, is closely related to During the pathogenesis of PDR, it has been widely reported vision loss.6–8 Investigations of the molecular mechanisms that VEGFA-VEGFR2 (VEGF receptor 2, also named KDR or Flk-1) involved in retinal neovascularization have shown that vascular mediates endothelial cell mitogenesis and vascular permeability, endothelial growth factor A (VEGFA), also known as VEGF, is a via VEGFR2-mediated signaling cascades.9,16 VEGFA could be master causative factor for retinal neovascularization.4,9 VEGF increased by altered glucose levels, insulin, and other growth was identified as a diffusible mitogenic -binding protein factors.18–20 The secreted VEGFA level from retinal pigment

Copyright 2019 The Authors iovs.arvojournals.org j ISSN: 1552-5783 4084

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License. Downloaded from iovs.arvojournals.org on 10/02/2021 Transcriptome Profile in PDR IOVS j October 2019 j Vol. 60 j No. 13 j 4085

epithelial cells was increased in the ocular fluid of DR patients (Epicentre, San Diego, CA, USA) before being used for and in the retinal ischemic monkey model,21–24 and its directional RNA-Seq library preparation. Purified RNA was expression level in DR patients has been correlated with disease fragmented, and then reverse transcribed with random primers progression.24 Consequently, anti-VEGFA therapy has been containing adaptor sequences and a randomized hexamer. The developed, with the first inhibitor approved in 2004.9,22 The cDNAs were then synthesized with terminal-tagging oligo- ocular anti-VEGFA therapies have shown clear efficacies in cDNA using the ScriptSeq v2 RNA-Seq Library Preparation Kit clinical trials.25,26 In addition to VEGFA, other proteins of the (Epicentre). The cDNAs were purified and amplified. PCR VEGF family also have important roles in angiogenesis. products corresponding to 300 to 500 (bp) were Competitive and collaborative relationships among VEGFs and purified, quantified, and stored at 808C until sequenced. their receptors constitute an important regulatory network The libraries were prepared following the manufacturer’s during angiogenesis.27–31 In human retinal pigment epithelial instructions and applied to an Illumina NextSeq 500 platform cells, VEGFA promotes the expression of VEGFC and VEGFR3.31 for 151 nucleotide pair-end sequencing (ABLife, Inc., Wuhan, Therefore, the roles of VEGF family members and China). The RNA-Seq data have been deposited in the GEO and family members,16,32 as well as their potential database with accession number GSE102485. cooperation during retinal neovascularization in humans, should 33 be further studied with advanced technologies. Data Processing and Differentially Expressed Gene An omics study on gene expression alterations of fibrovascular (DEG) Analysis membranes (FVMs) between PDR patients and normal retinal controls has revealed 91 genes that are preferentially expressed in Cutadapt35 was used to trim adaptors (with a 10% error) and FVMs, when compared with normal retinas.34 In the present low-quality bases (less than 16) of raw reads. Reads less than 16 study, we obtained qualified cDNA libraries from the small nucleotides in both ends were discarded. We used TopHat236 neovascular membranes (NVMs) of patients and retinas from software to align the filtered reads to the human reference normal or type II diabetic (T2D) individuals, and then sequenced genome (GRCH38) with four read mismatches using an end-to- the qualified cDNA libraries (RNA-Seq). The transcriptome end style. After alignment, total aligned reads for each gene were profiles of the NVMs were distinct from retinal membranes. We counted and saved as the initial expression level. Due to the also discovered that a number of genes belonging to the master nonuniform normalization of FPKM value,37 we used a tag-wise angiogenesis transcription factor (TF) ETS1 were systematically dispersion mode with the estimated size factors method38 to upregulated during retinal neovascularization. This expression normalize the gene transcriptional levels. After normalization, we network suggested distinct VEGF regulatory programs, involving selected the mRNA and long noncoding RNA (lncRNA) genes VEGFA acting as a trigger of the program switch. whose average transcriptional levels were no less than 0.1 as the expressed genes. We then calculated the Z-score for each gene across all MATERIALS AND METHODS samples. To retrieve genes showing diverse expression levels, we set the mean to Z-score 0.7 as the threshold to filter the Study Approval, Sample Collection, Treatment, genes with mediocre expression in all samples. DEGs were and RNA Extraction identified using the edgeR39 package from R software (R Project for Statistical Computing, Vienna, Austria). The false This study was approved by the Ethics Committee of the Second discovery rate was < 0.05 and jlog2 fold changej > 1). Hospital of Jilin University, and followed the tenets of the Declaration of Helsinki. Informed consent was obtained from the participants after explanation of the nature and possible Correlation Network Analysis consequences of the study. NVMs were surgically obtained from To obtain the expression modules and distinguish genes from a 23 eyes of patients with PDR or branch retinal vein occlusion union set by expression features, we used weighted gene (BRVO) during vitrectomies, which were performed at the coexpression network analysis (WGCNA).40 The normalized Second Hospital of Jilin University. Three normal and two T2D expression file of selected genes was the input. The output was retinas were from donated eyes. Surgical indications included the gene modules according to their expression patterns. For persistent vitreous hemorrhage, NVM formation, and NVM each gene module, eigengene was chosen to represent the traction retinal detachment. The surgical specimen for each case expression pattern. was frozen in liquid nitrogen immediately after removal. We used Pearson’s correlation coefficient (PCC) analysis to After homogenizing the sample with TRIzol reagent, the investigate the coexpressed genes with VEGFA.Weset samples were incubated at room temperature to permit absolute (PCC) ‡ 0.5 and P values 0.05 as the thresholds complete dissociation of the nucleoprotein complex. The to filter the coexpressed genes. We then obtained the human homogenate was then centrifuged at 48C, and the supernatant TFs from AnimalTFDB 2.0,41 and selected the TFs that were was collected and added to chloroform. The solution was coexpressed with VEGFA from the coexpressed genes. shaken vigorously and centrifuged at 48C to produce a clear upper aqueous layer (containing RNA), which was precipitated In Situ Hybridization (ISH) Experiments from the aqueous layer with isopropanol. The precipitated RNA was washed with 75% ethanol to remove impurities, and ISH was performed using a digoxigenin mRNA detection with then resuspended with diethyl pyrocarbonate (DEPC)-treated some modifications (Boster Biological Technology, Wuhan, water for use in downstream applications. China). Single-stranded sense and antisense digoxigenin-labeled corresponding mRNA probes were generated by in vitro cDNA RNA-Seq Library Construction transcription with T3 or T7 RNA polymerase. The tissues were fixed in 4% paraformaldehyde (containing 1/1000 DEPC) for 1 The quality, integrity, and quantity of the purified RNAs were hour, then dehydrated in ethanol gradients of 70%, 80%, 90%, determined using a 2100 Bioanalyzer system (Agilent Technol- 95%, and 100%, and incubated with xylene and embedded in ogies, Santa Clara, CA, USA). For each sample, 50 ng total RNA paraffin. Next, 4-mm-thick paraffin sections were baked face-up was used for the library preparation. Ribosomal RNAs were for 2 hours at 508C and deparaffinized. The tissue sections were depleted with a Ribo-Zero rRNA depletion kit for humans then treated with proteinase K (5 mg/mL; Sigma-Aldrich Corp.,

Downloaded from iovs.arvojournals.org on 10/02/2021 Transcriptome Profile in PDR IOVS j October 2019 j Vol. 60 j No. 13 j 4086

St. Louis, MO, USA) in 10 mM Tris-HCl (pH 8.0), 1 mM EDTA at types of retinal neovascular disorders. Qualified transcriptome 378C for 15 minutes before treatment with 3% H2O2 for 8 data were obtained from 23 NVM samples from patients, minutes, then rinsed with phosphate-buffered saline and including 17 from T2D patients, 3 from T1D patients, and 3 postfixed in 1% paraformaldehyde containing 1/1000 DEPC at from BRVO patients (Supplementary Table S1). T2D NVM room temperature for 10 minutes. Sections were then prehy- samples observed at seven different regions were collected (Fig. bridized for 4 hours at 408Cwith50lL prehybridization buffer. 1A; Supplementary Table S1). This sampling roughly reflected After the prehybridization, hybridization was performed in buffer the incidence of PDR in our hospital. Qualified transcriptome containing prehybridization fluid and 1.0 mgL1 either antisense data were also obtained from three retinal membranes from or sense digoxigenin-labeled mRNA probes at 408C overnight. normal individuals and two from T2D patients. A total of 1.43 Next, the sections were washed at 378Cfor30minutesina billion reads were sequenced with an average of 47.51 million buffer containing 50% formamide and 23 SSC, followed by reads per sample. After adaptor and low-quality read filtering, digestion with ribonuclease A (10 mg/mL; Sigma-Aldrich Corp.) approximately 40 million reads per sample were kept and at 378C for 30 minutes. After being washed in 23 SSC at 378Cfor aligned to the (Supplementary Table S2). 20 minutes, and twice in 0.23 SSC at 378Cfor30minutes, Similarly to a previous study,42 the sequencing results contained blocking was performed in blocking buffer containing 1% bovine a large fraction of reads (19%–50%) within intronic regions serum albumin for each section for 30 minutes at 378C. After (Supplementary Fig. S1A), and approximately 80% of the reads blocking buffer was removed, the sections were incubated with were not spliced (Supplementary Fig. S1B), indicating that a biotinylated mouse anti-digoxigenin antibody in a humidified large number of nascent transcripts were captured. We obtained chamber at 378C for 60 minutes, then postrinsed and incubated 25,920 expressed mRNA and lncRNA genes with normalized with Strept Avidin Biotin-peroxidase Complex (SABC) solution expressions ‡ 0.1 at least in one sample. for 30 minutes, and finally reacted with a solution containing To obtain a global understanding of transcriptional dynam- biotinylated peroxidase for 30 minutes at room temperature. ics from NVMs and retinas, we performed principal compo- After the sections were rinsed, a coloring reaction was nent analysis (PCA) based on the normalized expression level performed with a DAB Color Development Kit (Solarbio Science of all samples (Fig. 1B). NVMs and retinas could be clearly & Technology, Beijing, China). When the color reaction was separated by the second principal component (PC2), indicat- visible, it was stopped by washing in buffer solution, and the ing the presence of a large number of DEGs between NVMs sections were then counterstained with hematoxylin. Finally, the and retinas. NVMs originating from different regions or section was washed, dehydrated, and mounted. pathological states could not be clearly separated because of the large individual variation reflected by the PC1 direction. Western Blotting We found that six NVMs (P_I_W, two P_II_VAs, P_II_W, P_BRVO, and P_II_S) were separated from other samples Total protein was extracted by cell lysis. The extracts were (Fig. 1B). electroblotted onto a polyvinylidene difluoride (PVDF) mem- We also calculated PCCs for all samples. From the brane following separation by 12% SDS-PAGE. The PVDF hierarchical clustering results, samples were clustered into membrane was incubated with a blocking solution of 5% skim three major groups (Supplementary Fig. S1C, blue frames). milk for 1.5 hours at room temperature, followed by Group I (top left) contained 11 NVM samples and one retina incubation with a specific antibody (VEGF-C, 22601-1-AP; sample; group II (middle) contained two normal and two T2D Proteintech, Wuhan, China) overnight at 48C and then with retinal samples; and group III (bottom right) contained 12 horseradish peroxidase–conjugated secondary antibody NVM samples. Group II could be treated as the retina group. (SA00001-2; Proteintech) for 1 hour at room temperature. Six proliferative samples (P_BRVO3, P_II_S2, P_II_VA1, Finally, the blots were visualized using an ECL reagent P_II_W4, P_II_VA2, and P_I_W3) in group I showed low (Amersham, Buckinghamshire, UK). coefficients (R < 0.75) with most of the samples in group III (Supplementary Fig. S1C), consistent with the PCA results (Fig. Statistical Analysis 1B). The PCA and PCC analyses both showed that the transcriptional profile of NVMs was different from the retinal To describe the biologically functional significance of selected membranes, consistent with the fact that the NVM was gene sets, we used the hypergeometric test to calculate the essentially different from the retina.4,43 enrichment of (GO) terms and Kyoto We then compared the DEGs between each type of NVM Encyclopedia of Genes and Genomes (KEGG) pathways. Other and normal retinas. DEGs in NVMs from the surrounding area statistical analysis was performed using R software (R Project of the eye (P_II_S) were more distinct from other groups, and for Statistical Computing) unless otherwise stated. The those in the vascular arch and optic disc (P_II_VA_OD) significance of differences was evaluated with either Student’s displayed the largest DEG number (3322, Fig. 1C), among t-test, when only two groups were compared, or the hyper- which 911 (27.4%) were unique to this group and the rest of geometric test for a Venn diagram. No statistical methods were DEGs were shared by at least one other group. A total of 1387 used to predetermine sample size. *P < 0.05; **P < 0.01; ***P < DEGs (32.35%) were shared by at least four groups, and 1601 0.001. Hierarchical clustering was performed by Cluster 3.0 DEGs (37.35%) were identified only from one specific group. A (http://bonsai.hgc.jp/~mdehoon/software/cluster/software. total of 133 DEGs were common in all NVMs versus retina htm, in the public domain) or the heat map function in R. comparisons (Fig. 1C). Only 19 of the 87 DEGs from Ishikawa et al.34 were not included in our DEG set (Fig. 1C), implying the high sensitivity of the RNA-Seq method used in this study. RESULTS Transcriptome Profiles Distinguish NVMs From Expression of the VEGF Family and Two Other Retinal Membranes Vascular Endothelium-Specific Growth Factors in NVMs and Retinas In order to establish a link between the gene expression profiles in NVMs and PDR, we collected pathological tissues from retinal It has been reported that several proteins are major contrib- NVMs of patients that comprehensively included the major utors to angiogenesis and vascular development, including five

Downloaded from iovs.arvojournals.org on 10/02/2021 Transcriptome Profile in PDR IOVS j October 2019 j Vol. 60 j No. 13 j 4087

FIGURE 1. Transcriptome profiles of PDR samples were separated from nonproliferative samples. (A) A fundus photograph shows the right eye of a patient with 20/200 visual acuity, no signs of hypertension. Black arrows show the proliferative membranes (pale yellow) around the vascular arcade. The white arrow shows the blood vessel arcade. Other parts of the retina were labeled on the photograph that was related to Supplementary Table S1. (B) Principal component analysis of the sequenced samples revealed the separation between proliferative membrane and other samples. (C) An UpSet plot showing the overlapped differentially expressed genes between proliferative membrane and normal retina. Only groups with more than 10 genes are shown. DEGs from Ishikawa et al.34 were also included and are shown as a brown color.

proteins from the VEGF family, four proteins from the ephrin family, expression of EFNB2, EFNA1, and the receptor angiopoietin family, and at least one protein from the large EPHB4 was increased in the NVMs (Fig. 2A, right). ephrin family.15,16,44,45 By presenting the transcriptional levels To further validate the above observations, we included of all these growth factors and their receptors, we found a normal retina in the comparison group. We analyzed the significant decline of VEGFA and its major receptor KDR in the relative expression patterns of some known neovasculariza- NVM group (Fig. 2A). This expression pattern suggested that tion-related genes by sample-based normalized expression the VEGFA pathway was repressed in NVM samples. Expres- levels (Supplementary Fig. S2A). Notably, we found that sion of VEGFB did not show significant alterations between VEGFA, FLT1, KDR, and tumor neovascularization-related TFs NVMs and diabetic retinas. In contrast, VEGFC and its receptor HIF1A and PROX146,47 were highly expressed in normal gene, VEGFR-3 (FLT4), exhibited an increased expression retinas. In the diabetic state, expression of VEGFA increased pattern (Fig. 2A, left). Both angiopoietin genes and their dramatically in two T2D retinal samples, while expression of receptors showed a consistently higher expression level in the PROX1 was exclusively shut down. Remarkably, VEGFA NVMs than the diabetic retinas (Fig. 2A, middle). For the expression in all NVMs from T2D patients was drastically

Downloaded from iovs.arvojournals.org on 10/02/2021 Transcriptome Profile in PDR IOVS j October 2019 j Vol. 60 j No. 13 j 4088

FIGURE 2. Expression levels of three gene families were mainly elevated in neovascular membranes. (A) Box plot showing the gene expression level changes between the NVMs and the retinas. Three family genes of vascular growth factors (top) and their receptors (bottom) were presented. The brown box represents the NVM groups and the green box represents the retinal groups. (B) In situ hybridization validation for the expression and location of VEGF family genes. Upper: presentation for P_II_N sample; bottom: for P_II_VA_OD sample. Arrows in the figures represent the locations of target probe signal (dark brown regions). Control: treatment without target probes. (C) Western blotting results showing the protein level of VEGFC in different kinds of NVMs from patients.

decreased compared with the T2D retinas (Supplementary Fig. also validated the increased protein levels of VEGFC in T2D S2A). In contrast, we found that VEGFC was at a low level in NVMs (Fig. 2C). These results together showed that expression normal and T2D retinas, but at a high level in most of the NVM of VEGFC was highly induced in the NVMs, while VEGFA was samples (Supplementary Fig. S2A). The differentially regulated the most expressed in diabetic retinas. expression of VEGFA and VEGFC was also plotted by their gene-based normalized expression levels (Supplementary Fig. Genes Coexpressed With VEGFA and Their S2B). To validate the induced expression of VEGFC,we Network in the NVMs of PDR Patients performed RNA ISH in NVM samples (Fig. 2B). VEGF expression (dark brown) emerged at the margin or the inner We then characterized the VEGFA-related regulatory network membrane (arrows, Fig. 2B) of the tissues. The VEGFC signal by using a coexpression network approach. A correlation was quite strong, similar to that of VEGFA (Fig. 2B, upper). We matrix between VEGFA and the other 25,919 genes was

Downloaded from iovs.arvojournals.org on 10/02/2021 Transcriptome Profile in PDR IOVS j October 2019 j Vol. 60 j No. 13 j 4089

FIGURE 3. Functions of genes coexpressed with VEGFA and their expression patterns. (A, B) The dot plot represents the coexpression relationship between HMGA1 (A), ANGPT2 (B), and VEGFA.(C) The bar plot of enriched biological process terms by genes positively coexpressed with VEGFA. (D) The bar plot of enriched biological process terms by genes negatively coexpressed with VEGFA.(E) A heat map presentation of the genes belonging to the VEGF signaling pathway. Two group gene sets of the VEGF signaling pathway were labeled by blue and red colors.

generated by computing their PCCs. With criteria of P value subgroups, with one group (GI) showing a constitutively 0.01 and absolute PCCs ‡ 0.5, a total of 1615 genes were higher expression in retina but lower expression in NVMs (Fig. coexpressed with VEGFA, including 1421 positive and 194 3E). Genes of another group (GII) were expressed at lower negative pairs (Supplementary Table S3). Among these levels in normal retina and higher in T2D retina, but expressed coexpressed genes, we found HMGA1, a TF that controls at a much higher level in NVMs (Fig. 3E), suggesting that insulin receptor (INSR) gene transcription,48 was positively modulation of the VEGF signaling expression network could be coexpressed with VEGFA (Fig. 3A), coincident with the a key event for initiating neovascularization. previous finding that HMGA1 positively regulated the expres- 49 sion level of VEGFA. ANGPT2 was negatively coexpressed A Specific Group of ETS Transcription Factors with VEGFA (R ¼0.729; Fig. 3B), consistent with the Promote the Progression of PDR upregulation of ANGPT2 in NVMs. In a similar manner as the VEGFA expression patterns, genes positively coexpressed with To further explore VEGFA-mediated gene expression networks VEGFA were higher in diabetic and normal retinas than in in diabetic retinas and NVMs, we analyzed the TFs involved in NVMs (Supplementary Fig. S3A). In contrast, 194 genes that VEGFA regulation. We found that ETS1 was negatively negatively coexpressed with VEGFA showed a reciprocal coexpressed with VEGFA. ETS1 has been recently shown to expression pattern (Supplementary Fig. S3A). amplify the VEGF signal through acetylation to promote Based on the concept that coexpressed genes are function- angiogenesis.52 Two ETSs (ETS1 and ETS2) and two ERGs ally related,50 we further explored the VEGFA-regulated (ERG and FLI1) domain-containing genes showed a reverse networks by functional enrichment analysis. The genes expression pattern with VEGFA, that is, expressing at high positively coexpressed with VEGFA were mostly enriched in levels in NVMs (Fig. 4A). Three of these four genes (ETS1, ERG, the glycolysis and glucose metabolic processes (Fig. 3C; and FLI1) are known to promote angiogenesis by activating Supplementary Fig. S3B). Consistent with the role of hypoxia VEGFR and TIE1/TIE2.53,54 We also found that VEGFA was in increasing VEGFA expression,51 the HIF-1 signaling pathway negatively coexpressed with TIE1 (R ¼0.40) and TIE2 (also (KEGG) was also enriched (Supplementary Fig. S3B). The known as TEK; R ¼0.45). By adding VEGFC to this network, biological process terms, including blood coagulation and we observed the positive coexpression between VEGFC and platelet activation, as well as regulation of the apoptotic ETS family members (Fig. 4A). process, and vascular smooth muscle contraction were It is known that the ETS family contains both transcriptional enriched by genes negatively coexpressed with VEGFA (Fig. activators and repressors for angiogenesis.54 To define the 3D; Supplementary Fig. S3C). We also plotted the expression global function of ETS family genes in PDR, we searched for all profiles of genes in VEGF signaling pathways including 19 ETS family genes in humans and studied their expression genes. Expression of these genes was classified into two patterns. A clustering heat map of 28 expressed ETS family

Downloaded from iovs.arvojournals.org on 10/02/2021 Transcriptome Profile in PDR IOVS j October 2019 j Vol. 60 j No. 13 j 4090

FIGURE 4. ETS TFs promote the progression of neovascularization and PDR. (A) Correlation presentation of the four ETS TF genes with VEGFA and VEGFC. VEGFA was negatively coexpressed with these four genes, but VEGFC was positively coexpressed. (B) Hierarchical clustering heat map of the ETS TF family genes. Two distinct clusters emerged. (C) A Venn diagram shows the coexpression genes with the four ETS TF genes. (D) A bar plot of the enriched biological process terms for genes coexpressed with ETS1.(E) A Venn diagram showing the overlapped genes that were coexpressed with ETS1 and bound by ETS1. (F) Regulatory schema plot shows the coexpression relationships between ETS family genes and vascular growth family genes. The interactions in this network were generated by Reactom FI application from Cytoscape (https://cytoscape.org, in the public domain).

genes exhibited two distinct expression clusters (Fig. 4B). contrast, cluster II contained 12 genes, which showed a low Cluster I contained 16 genes that showed sporadic expression expression level in retina samples but globally overexpressed patterns. Among cluster I genes, ELF2 showed a similar levels in NVM samples (Fig. 4B), implying their potential role in overexpression pattern as VEGFA in the T2D retinas. In promoting PDR-related neovascularization.54,55

Downloaded from iovs.arvojournals.org on 10/02/2021 Transcriptome Profile in PDR IOVS j October 2019 j Vol. 60 j No. 13 j 4091

We then obtained a coexpressed gene network of ETS1/2, signaling pathways were among the top 10 enriched pathways EGR, and FLI1 to further explore their potential functions in (Supplementary Fig. S5D). NVMs. The coexpressed genes with these four TFs significantly T2D-repressed genes were mainly located in the green overlapped. For example, 732 genes were coexpressed with at module. GO enrichment analysis revealed that these genes least three of these four TFs (26.65%, P ¼ 2.30e-155; hyper- were highly enriched in synaptic transmission and neuronal geometric test; Fig. 4C). Consistently, functional analysis of functions, including visual perception, sensory perception of ETS-coexpressed genes showed significant enrichment of sound, and adult walking behavior, as well as response to angiogenesis and related biological processes (Fig. 4D; stimulus (Fig. 5B). KEGG analysis revealed that a large number Supplementary Figs. S4A–C). To assess whether these coex- of synaptic functions were selectively repressed under the T2D pressed genes resulted from transcriptional activation by TF state, including serotonergic synapse, cholinergic synapse, binding, we downloaded and analyzed the DNA binding profile GABAergic synapse, and glutamatergic synapse (Supplementa- data of ETS1 from human umbilical vein endothelial cells.52 We ry Fig. S5E). In contrast to the T2D-repressed genes, we found 305 coexpressed genes with ETS1, which were also noticed that the T2D-induced genes in the purple module were bound by ETS1 (P ¼ 1.24e-5, hypergeometric test; Fig. 4E), specifically enriched in glucose metabolic processes and small suggesting that a significant fraction of the coexpressed genes molecule metabolic processes, and responses to stress and with these TFs involved direct transcriptional regulation. protein targeting to mitochondrion (Fig. 5C; Supplementary We then characterized the coexpression relationship Fig. S5F). T2D-induced genes including VEGFA, ALDOA, ENO1, between these four TFs and vascular growth factors in Figure EGLN1, and LDHA were enriched in the HIF-1 signaling 2A. Most notably, besides VEGFC, ANGPT2, EFNB2,and pathway. These results suggested that the transcriptional EFNA1, their receptors TIE1/TEK, VEGFR1/3, and EPHB4 profiles in T2D retina had adapted to a stressful living were positively coexpressed with all these four TFs (Supple- condition containing high levels of glucose. mentary Table S4). This implied that the ETS and ERG domains To compare the differential expression of genes in the containing TFs could participate in PDR progression by retina modules and NVM modules, the top 25 most enriched promoting angiogenesis. Based on the coexpression results, BP terms and KEGG pathways in the four retina-specific we propose that ETS TF family genes, especially ETS1/2, ERG, modules were selected and plotted (Fig. 5E; Supplementary Fig. S5G), revealing that T2D-induced retina functions were and FLI1, participated in the formation and progression of PDR generally shared more by NVM groups than the T2D repressed. and angiogenesis via positively regulating the expression of TIE, VEGFR, and ephrin family genes (Fig. 4F). The Neovascularization-Specific Gene Expression Gene Expression in T2D Retina Is Reprogrammed Programs We identified gene sets associated with specific sample groups We then analyzed NVM-related modules, including eight using the WGCNA method. We first calculated the mean Z- modules exhibiting high levels in some or most NVM samples, score for each gene (see Materials and Methods and which were called NVM-related modules (7837 genes). The Supplementary Fig. S5A) to select genes with a mean Z-score turquoise (4306 genes) and brown (1073 genes) modules no less than 0.7 (12,293 genes), and then clustered these genes exhibited high expression in less than half of the NVMs using the WGCNA method40 to obtain a consensus coex- (Supplementary Figs. S6A–B). The other six modules (2459 genes), including red, black, yellow, pink, magenta and tan, pression network. We identified 12 modules with characteris- were prevalently high in NVM samples, showing similar or tic expression patterns (Supplementary Fig. S5B). These complementary expression patterns (Figs. 6A–F). For exam- modules could be further classified into two major groups ple, the red module showed elevated levels of gene based on clustering of eigengene values (Supplementary Fig. expression in most of P_I_W and P_II_W samples and some S5C): retinal group (four modules in the right branch), and other NVM samples (Fig. 6A). These genes were named NVM group (six modules in the left branch). The four retinal neovascularization-related genes, which could serve as modules had higher expression levels in R_norm and R_II potential PDR markers for further study. We provided detailed samples than others (Figs. 5A–D). Genes in the blue (2341 module information in the supplementary information sec- genes) module were specifically expressed in normal and T2D tion (Supplementary Table S5). retinas, and classified as the constitutive retina genes (Fig. 5A). We then plotted the top 25 BP and KEGG pathways in the Green (662 genes) and green–yellow modules (99 genes) had six neovascularization-specific modules (Fig. 6G; Supplemen- higher and specific expression levels in normal retina, which tary Fig. S6C). Most of these enriched terms were commonly we named T2D-repressed modules (Figs. 5B, 5D). The purple from multiple neovascularization-specific modules. Impressive- module (214 genes) exhibited consistently high levels in T2D ly, most NVM terms were shared among neovascularization- retina samples, which we named T2D-induced genes (Fig. 5C). specific modules, but were excluded from the retina groups. In summary, the T2D-repressed and -induced genes comprised The pink module included VEGFC, EPHB4, EFNB1, and other 29.4% of the four retina groups, indicating that the diabetic angiogenesis-related genes such as PDGFA, TGFBR1, FZD6, state profoundly altered the transcriptional profile of the ITGAV, ITGA5, TGFBI, FGFR1, and SHC1. Some angiogenesis retina. growth factors and receptors described in Figure 2 were We further explored the functions of genes in the four included in the magenta module, including ANGPT2, EFNA1, retina groups using GO and KEGG enrichment analysis. Among TIE2, FLT4, FLT1, and EPHB4 (Supplementary Table S5). the top 10 enriched GO biological process (BP) terms in the Furthermore, the pink (281 genes) and magenta (188 genes) blue module, seven of them were directly retinal related, modules were highly enriched in angiogenesis and extracellu- including phototransduction, retinal bipolar neuron differenti- lar matrix organization pathways (Fig. 6G). The expression ation, retina development in camera-type eye, and rhodopsin- profile of genes in angiogenesis and the related functional mediated signaling pathway (Fig. 5A). This enrichment pattern pathways had a global higher expression level in NVM samples was highly consistent with the constitutive expression of these (Supplementary Fig. S6D). In addition, several angiogenesis- genes in the retina, regardless of the pathological state. For related pathways such as blood coagulation,56,57 platelet enrichment of KEGG pathways, NOD-like receptor, phospha- activation,58 and leukocyte migration59 were also highly tidylinositol, and , as well as Fc epsilon RI enriched in NVM modules (Fig. 6G; Supplementary Fig. S6C).

Downloaded from iovs.arvojournals.org on 10/02/2021 Transcriptome Profile in PDR IOVS j October 2019 j Vol. 60 j No. 13 j 4092

FIGURE 5. WGCNA analysis of the varied genes revealed distinct expression pattern modules. (A–D) A bar plot showing the eigengene value (left) and the top 10 enriched biological processes (right) of retinal related modules. Samples were sorted by sample ID order. (E) A heat map presentation showing the top enriched biological process terms for genes in the retina-elated modules. Enrichment scores (log10 P value) were normalized by terms for the presentation.

Downloaded from iovs.arvojournals.org on 10/02/2021 Transcriptome Profile in PDR IOVS j October 2019 j Vol. 60 j No. 13 j 4093

FIGURE 6. Neovascularization-specific gene expression modules and their enriched functional pathways. (A–F) A bar plot showing the eigengene value of six epigene modules showing prevalent upregulated expression in neovascularization samples. Samples were sorted by sample ID order. (G) A heat map presentation of the top enriched biological process terms for genes in each module, choosing terms enriched in neovascularization- specific modules. Enrichment scores (log10 P value) were normalized by terms for the presentation.

DISCUSSION the retina, which provides an unprecedented opportunity to understand the mechanisms of retinal angiogenesis and Although PDR is one of the leading causes of preventable neovascularization, as well as to develop novel biomarkers blindness, a better understanding of its pathogenetic mecha- and therapies against PDR. nisms is needed to improve clinical treatments.8 In the present study, we have presented a new approach to study retinal Gene Expression in T2D Retinas Is Extensively angiogenesis and neovascularization associated with PDR, Reprogrammed to Adapt to High-Glucose Stress through robustly comparing the deep-sequenced transcrip- tomes of NVMs from T2D, T1D, and BRVO patients with those It has been reported that loss of neuroretinal adaptation was from T2D and normal retinas. To the best of our knowledge, detected before observable vascular pathologies,60 suggesting we are the first to characterize the global alterations of preceding transcriptome changes before DR diagnosis. Using transcriptome profiles between NVMs and retinas. The results the WGCNA method, we identified three modules of retina- showed a large transcriptional difference between NVMs and specific genes: constitutive retina genes (2341, 70.6%), T2D-

Downloaded from iovs.arvojournals.org on 10/02/2021 Transcriptome Profile in PDR IOVS j October 2019 j Vol. 60 j No. 13 j 4094

repressed retina genes (761, 22.9%), and T2D-induced retina VEGFA Expression Is Highly Regulated During genes (214, 6.5%). Although we obtained only five retinal Retina Pathogenesis and Neovascularization samples in this study, we could clearly separate the retinal samples from the NVMs to obtain three distinct retina-specific The essential function of VEGFA in promoting neovasculariza- modules. Functional clustering analysis showed that genes in tion in ischemic retinopathy has been proposed based on these modules were enriched in either expected or interesting multiple lines of evidence.9,25,33,67 Our results showed that functional pathways, which supported the validity of the VEGFA was expressed at high levels in normal retinas, diabetic results. We observed that the glucose metabolic process and retinas, and NVMs, while levels of VEGFA in diabetic retinas small molecule metabolic process pathways were enriched were significantly higher than in normal retinas and NVMs. The only in T2D-induced modules. The different pathways between finding that VEGFA expression is specifically increased in T2D T2D-induced and T2D-repressed modules suggested that the retina and reduced in NVMs was unexpected. However, it is transcriptional profile of retinal cells had been reprogrammed consistent with the finding that the onset of DR may be present to adapt to a stressful condition including high levels of at the time of diagnoses for T2D patients, and suggests a model glucose, and some retinal functions were compromised in T2D in which VEGFA plays an important role for the pathological patients. These findings are worthy of further investigation in transformation of retina from T2D to PDR.61 It also provides animal models and in the early clinical diagnosis of DR.61 mechanistic insight into clinical observations that earlier anti- VEGF therapy produces better results.68 Genes Overexpressed in NVMs as Potential Based on the results presented in this study, we propose a Biomarkers dynamic gene regulatory model that initiates and stabilizes retinal neovascularization and neovascularization. In this Approximately half of patients fail to respond to current anti- model, expression of VEGFA is high in diabetic retina and is VEGF treatment,8 necessitating the study of other factors sensitive to pathological states. Upon pathological induction, contributing to PDR to serve as more comprehensive VEGFA secretion into ocular fluid is increased. The increased prognostic and diagnostic biomarkers. The study of additional level of extracellular VEGFA stimulates the expression of ETS genes contributing to PDR progression is hindered by technical TFs in endothelial cells.52 Based on these coexpression difficulties in studying the very small amounts of patient relationships, ETS TFs may in turn regulate the expression of tissues. A microarray analysis of gene expression in six NVMs growth factors, VEGFC, angiopoietin, and ephrin, as well as from patients with PDR and three normal human retinas has their receptors54 (Figs. 2, 6E). These growth factors then only revealed 91 genes preferentially expressed in active coordinately drive the gene expression in endothelial cells to a neovascular patients, when compared with normal retinas, pattern toward the NVMs rather than retina cells. It is very excluding VEGFs and .34 In this study, we possible that the relatively high basal levels of VEGFA, and its reported the detection of 1387 overlapped DEGs among four main receptor, KDR, maintain the proper response of retinal diabetic proliferative membrane groups and the normal retina, endothelial cells to extracellular VEGFA, and this newly including reported active neovascular genes. WGCNA analysis established expression pattern collectively supports the identified six modules with prevalent higher expression in initiation and stabilization of retinal neovascularization. NVM samples, when compared with normal retinas, containing In summary, the transcriptome sequencing and bioinfor- 2459 genes. These genes were highly enriched in angiogenesis, matics analysis presented in this study have not only generated blood coagulation, platelet activation, leukocyte migration, and the most informative transcriptome data, a comprehensive list extracellular matrix organization, consistent with the current of candidate genes for developing PDR treatment, and 17,32 understanding of the mechanism of angiogenesis, and biomarkers, but have also provided mechanistic insights into suggested that a large number of angiogenesis-related genes the mechanism of retina angiogenesis and neovascularization. were highly expressed in NVMs. Genes in NVM modules The dynamic expression of VEGFA, the induced expression of therefore represented potential biomarkers and candidates for 62 endothelium-specific growth factors, and a group of ETS family developing PDR diagnostic and therapeutic methodologies. of TFs form the framework of an elaborate network that coordinates angiogenesis in PDR. Based on these observations, Expression of VEGFC, ANGPT, EFNB3, and Their further studies with higher-resolution methods in animal and Receptors Is Induced During Retina cell models are needed to further clarify, validate, and extend Neovascularization these regulatory mechanisms that occur during retinal neovascularization in PDR.69 The three major families, including VEGFs and the angiopoietin and ephrin protein families, play substantially differential and Acknowledgments cooperative roles in promoting and regulating angiogenesis and neovascularization.16,17 In this study, analysis of the The authors thank Jian Zhou for the discussion about data expression patterns of all genes in these three families showed processing and Hong Wu, ABLife, Inc., for language polishing. that expression of VEGFC, ANGPT1 and ANGPT2, EFNB3, and Supported by the National Natural Science Foundation of China their receptor genes were coordinately increased in NVMs (No. 81272242 and No. 81502284), National Key New Drug compared with those of retinas, and these observations were Creation and Manufacturing Program of Ministry of Science and confirmed by ISH and Western blot experiments. The essential Technology (No. 2013ZX09103003004), the Fundamental Re- role of VEGFC in angiogenesis has been reported in search Funds for the Central Universities and the Grant of Jilin embryonically developing mice and during cancer.63–66 The Province Science & Technology Committee, China (No. high expression of VEGFC in T2D-induced NVMs suggested its 20170414028GH, No. 20170520031JH, No. 201706230TC, No. potential role in maintaining the progression of NVM 20150204038YY, No. 20150101188JC, No. 20150204038YY, and development. These findings represent the first report of a No. 20150309003YY). This work was also supported by Appre- network of three major families of vascular endothelium- ciate the Beauty of Life, Inc. (No. 2014-11032, YZ). specific growth factors that may cooperate during retinal Disclosure: Y. Li, None; D. Chen, None; L. Sun, None; Y. Wu, neovascularization, which is consistent with recent re- None; Y. Zou, None; C. Liang, None; Y. Bao, None; J. Yi, None; ports.27,31 Y. Zhang, None; J. Hou, None; Z. Li, None; F. Yu, None; Y.

Downloaded from iovs.arvojournals.org on 10/02/2021 Transcriptome Profile in PDR IOVS j October 2019 j Vol. 60 j No. 13 j 4095

Huang, None; C. Yu, None; L. Liu, None; Z. Liu, None; Y. Zhang, retinopathy and other retinal disorders. NEnglJMed. 1994; None; Y. Li, None 331:1480–1487. 23. Adamis AP, Shima DT, Yeo KT, et al. Synthesis and secretion of References vascular permeability factor/vascular endothelial growth factor by human retinal pigment epithelial cells. Biochem 1. American Diabetes Association. Diagnosis and classification of Biophys Res Commun. 1993;193:631–638. diabetes mellitus. Diabetes Care. 2009;33(suppl 1):S62–S69. 24. Boulton M, Foreman D, Williams G, Mcleod D. VEGF 2. Saaddine JB, Honeycutt AA, Narayan KM, Zhang X, Klein R, localisation in diabetic retinopathy. Br J Ophthalmol. 1998; Boyle JP. Projection of diabetic retinopathy and other major 82:561–568. eye diseases among people with diabetes mellitus: United 25. Stewart MW. A review of for the treatment of States, 2005-2050. Arch Ophthalmol. 2008;126:1740–1747. diabetic retinopathy. Ophthalmol Ther. 2017;6:33–47. 3. Giacco F, Brownlee M. Oxidative stress and diabetic compli- 26. Cheung N, Wong IY, Wong TY. Ocular anti-VEGF therapy for cations. Circ Res. 2010;107:1058–1070. diabetic retinopathy: overview of clinical efficacy and 4. Antonetti DA, Klein R, Gardner TW. Diabetic retinopathy. N evolving applications. Diabetes Care. 2014;37:900–905. Engl J Med. 2012;366:1227–1239. 27. Joukov V, Pajusola K, Kaipainen A, et al. A novel vascular 5. Cheung N, Mitchell P, Wong TY. Diabetic retinopathy. Lancet. endothelial growth factor, VEGF-C, is a ligand for the Flt4 2010;376:124–136. (VEGFR-3) and KDR (VEGFR-2) receptor tyrosine kinases. EMBO J. 1996;15:290–298. 6. Tan GS, Cheung N, Simo´ R, Cheung GCM, Wong TY. Diabetic macular oedema. Lancet Diabetes Endocrinol. 2017;5:143– 28. Partanen TA, Arola J, Saaristo A, et al. VEGF-C and VEGF-D 155. expression in neuroendocrine cells and their receptor, VEGFR-3, in fenestrated blood vessels in human tissues. 7. Lang GE. Diabetic macular edema. Ophthalmologica. 2012; FASEB J. 2000;14:2087–2096. 227(suppl 1):21–29. 29. Tammela T, Zarkada G, Wallgard E, et al. Blocking VEGFR-3 8. Stitt AW, Curtis TM, Chen M, et al. The progress in suppresses angiogenic sprouting and vascular network understanding and treatment of diabetic retinopathy. Prog formation. Nature. 2008;454:656–660. Retin Eye Res. 2016;51:156–186. 30. Zarkada G, Heinolainen K, Makinen T, Kubota Y, Alitalo K. 9. Ferrara N, Adamis AP. Ten years of anti-vascular endothelial VEGFR3 does not sustain retinal angiogenesis without growth factor therapy. Nat Rev Drug Discov. 2016;15:385– VEGFR2. Proc Natl Acad Sci U S A. 2015;112:761–766. 403. 31. Zhao B, Ma A, Cai J, Boulton M. VEGF-A regulates the 10. Leung DW, Cachianes G, Kuang WJ, Goeddel DV, Ferrara N. expression of VEGF-C in human retinal pigment epithelial Vascular endothelial growth factor is a secreted angiogenic cells. Br J Ophthalmol. 2006;90:1052–1059. mitogen. Science. 1989;246:1306–1309. 32. Carmeliet P. Mechanisms of angiogenesis and arteriogenesis. 11. Senger DR, Galli SJ, Dvorak AM, Perruzzi CA, Harvey VS, Nat Med. 2000;6:389–395. Dvorak HF. Tumor cells secrete a vascular permeability factor 33. Apte RS, Chen DS, Ferrara N. VEGF in signaling and disease: that promotes accumulation of ascites fluid. Science. 1983; beyond discovery and development. . 2019;176:1248– 219:983–985. Cell 1264. 12. Neufeld G, Cohen T, Gengrinovitch S, Poltorak Z. Vascular 34. Ishikawa K, Yoshida S, Kobayashi Y, et al. Microarray analysis endothelial growth factor (VEGF) and its receptors. FASEB J. of gene expression in fibrovascular membranes excised from 1999;13:9–22. patients with proliferative diabetic retinopathy. Invest Oph- 13. Shweiki D, Itin A, Soffer D, Keshet E. Vascular endothelial thalmol Vis Sci. 2015;56:932–946. growth factor induced by hypoxia may mediate hypoxia- 35. Martin M. Cutadapt removes adapter sequences from high- initiated angiogenesis. Nature. 1992;359:843–845. throughput sequencing reads. EMBnet J. 2011;17:10–12. 14. Ferrara N, Gerber HP, LeCouter J. The biology of VEGF and its 36. Kim D, Pertea G, Trapnell C, Pimentel H, Kelley R, Salzberg receptors. Nat Med. 2003;9:669–676. SL. TopHat2: accurate alignment of transcriptomes in the 15. Shibuya M. Vascular endothelial growth factor and its presence of insertions, deletions and gene fusions. Genome receptor system: physiological functions in angiogenesis Biol. 2013;14:R36. and pathological roles in various diseases. JBiochem. 2013; 37. Dillies MA, Rau A, Aubert J, et al. A comprehensive evaluation 153:13–19. of normalization methods for Illumina high-throughput RNA 16. Yancopoulos GD, Davis S, Gale NW, Rudge JS, Wiegand SJ, sequencing data analysis. Brief Bioinform. 2013;14:671–683. Holash J. Vascular-specific growth factors and blood vessel 38. Anders S, Huber W. Differential expression analysis for formation. Nature. 2000;407:242–248. sequence count data. Genome Biol. 2010;11:R106. 17. Carmeliet P. Angiogenesis in health and disease. Nat Med. 39. Robinson MD, McCarthy DJ, Smyth GK. edgeR: a bioconduc- 2003;9:653–660. tor package for differential expression analysis of digital gene 18. Lu M, Kuroki M, Amano S, et al. Advanced glycation end expression data. Bioinformatics. 2010;26:139–140. products increase retinal vascular endothelial growth factor 40. Langfelder P, Horvath S. WGCNA: an R package for weighted expression. J Clin Invest. 1998;101:1219–1224. correlation network analysis. BMC Bioinformatics.2008;9: 19. Lu M, Amano S, Miyamoto K, et al. Insulin-induced vascular 559. endothelial growth factor expression in retina. Invest 41. Zhang HM, Liu T, Liu CJ, et al. AnimalTFDB 2.0: a resource for Ophthalmol Vis Sci. 1999;40:3281–3286. expression, prediction and functional study of animal 20. Ferrara N. Vascular endothelial growth factor: basic science transcription factors. Nucleic Acids Res. 2015;43:D76–D81. and clinical progress. Endocr Rev. 2004;25:581–611. 42. Ameur A, Zaghlool A, Halvardson J, et al. Total RNA 21. Miller JW, Adamis AP, Shima DT, et al. Vascular endothelial sequencing reveals nascent transcription and widespread growth factor/vascular permeability factor is temporally and co-transcriptional splicing in the human brain. Nat Struct Mol spatially correlated with ocular angiogenesis in a primate Biol. 2011;18:1435–1440. model. Am J Pathol. 1994;145:574–584. 43. Ljubimov AV, Burgeson RE, Butkowski RJ, et al. Basement 22. Aiello LP, Avery RL, Arrigg PG, et al. Vascular endothelial membrane abnormalities in human eyes with diabetic growth factor in ocular fluid of patients with diabetic retinopathy. J Histochem Cytochem. 1996;44:1469–1479.

Downloaded from iovs.arvojournals.org on 10/02/2021 Transcriptome Profile in PDR IOVS j October 2019 j Vol. 60 j No. 13 j 4096

44. Loukovaara S, Gucciardo E, Repo P, et al. Indications of 56. Belting M, Ahamed J, Ruf W. Signaling of the tissue factor lymphatic endothelial differentiation and endothelial progen- coagulation pathway in angiogenesis and cancer. Arterioscler itor cell activation in the pathology of proliferative diabetic Thromb Vasc Biol. 2005;25:1545–1550. retinopathy. Acta Ophthalmol. 2015;93:512–523. 57. Rothmeier AS, Liu E, Chakrabarty S, et al. Identification of the 45. Domigan CK, Ziyad S, Iruela-Arispe ML. Canonical and coagulation factor VIIa binding site required for pro- noncanonical vascular endothelial growth factor pathways: angiogenic PAR2 signaling. Blood. 2018;131:674–685. new developments in biology and signal transduction. 58. Janowska-Wieczorek A, Wysoczynski M, Kijowski J, et al. Arterioscler Thromb Vasc Biol. 2015;35:30–39. Microvesicles derived from activated platelets induce metas- 46. Hillen F, Griffioen AW. Tumour vascularization: sprouting tasis and angiogenesis in lung cancer. Int J Cancer. 2005;113: angiogenesis and beyond. Cancer Metastasis Rev. 2007;26: 752–760. 489–502. 59. Kreuger J, Phillipson M. Targeting vascular and leukocyte 47. Semenza GL. Targeting HIF-1 for cancer therapy. Nat Rev communication in angiogenesis, inflammation and fibrosis. . 2016;15:125–142. Cancer. 2003;3:721–732. Nat Rev Drug Discov 60. Abcouwer SF, Gardner TW. Diabetic retinopathy: loss of 48. Foti D, Iuliano R, Chiefari E, Brunetti A. A nucleoprotein neuroretinal adaptation to the diabetic metabolic environ- complex containing Sp1, C/EBP beta, and HMGI-Y controls ment. Ann N Y Acad Sci. 2014;1311:174–190. human insulin receptor gene transcription. Mol Cell Biol. 2003;23:2720–2732. 61. Ebneter A, Zinkernagel MS. Novelties in diabetic retinopathy. Endocr Dev. 2016;31:84–96. 49. Chiefari E, Ventura V, Capula C, et al. A polymorphism of HMGA1 protects against proliferative diabetic retinopathy by 62. MacLaren RE, Bennett J, Schwartz SD. Gene therapy and stem cell transplantation in retinal disease: the new frontier. impairing HMGA1-induced VEGFA expression. Sci Rep. 2016; Ophthalmology. 2016;123:S98–S106. 6:39429. 63. Bauer SM, Bauer RJ, Liu ZJ, Chen H, Goldstein L, Velazquez 50. Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster OC. Vascular endothelial growth factor-C promotes vasculo- analysis and display of genome-wide expression patterns. genesis, angiogenesis, and collagen constriction in three- Proc Natl Acad Sci U S A. 1998;95:14863–14868. dimensional collagen gels. JVascSurg. 2005;41:699–707. 51. Simo R, Sundstrom JM, Antonetti DA. Ocular anti-VEGF 64. Hamada K, Oike Y, Takakura N, et al. VEGF-C signaling therapy for diabetic retinopathy: the role of VEGF in the pathways through VEGFR-2 and VEGFR-3 in vasculoangio- pathogenesis of diabetic retinopathy. Diabetes Care. 2014;37: genesis and hematopoiesis. Blood. 2000;96:3793–3800. 893–899. 65. Bauer SM, Bauer RJ, Velazquez OC. Angiogenesis, vasculogen- 52. Chen J, Fu Y, Day DS, et al. VEGF amplifies transcription esis, and induction of healing in chronic wounds. Vasc through ETS1 acetylation to enable angiogenesis. Nat Endovasc Surg. 2005;39:293–306. Commun. 2017;8:383. 66. Valtola R, Salven P, Heikkila P, et al. VEGFR-3 and its ligand 53. Sharrocks AD. The ETS-domain transcription factor family. VEGF-C are associated with angiogenesis in breast cancer. Am Nat Rev Mol Cell Biol. 2001;2:827–837. J Pathol. 1999;154:1381–1390. 54. Lelievre E, Lionneton F, Soncin F, Vandenbunder B. The Ets 67. Miller JW. VEGF: from discovery to therapy: the Champali- family contains transcriptional activators and repressors maud Award Lecture. Trans Vis Sci Tech. 2016;5(2):9. involved in angiogenesis. Int J Biochem Cell Biol. 2001;33: 68. Brown DM, Nguyen QD, Marcus DM, et al. Long-term 391–407. outcomes of ranibizumab therapy for diabetic macular edema: 55. Watanabe D, Takagi H, Suzuma K, et al. Transcription factor the 36-month results from two phase III trials: RISE and RIDE. Ets-1 mediates ischemia- and vascular endothelial growth Ophthalmology. 2013;120:2013–2022. factor-dependent retinal neovascularization. Am J Pathol. 69. Gariano RF, Gardner TW. Retinal angiogenesis in development 2004;164:1827–1835. and disease. Nature. 2005;438:960–966.

Downloaded from iovs.arvojournals.org on 10/02/2021