Microarray Analysis of Gene Expression in Fibrovascular Membranes Excised From Patients With Proliferative Diabetic Retinopathy

Abstract

Purpose.: We determined the profile of genes expressed in fibrovascular membranes (FVMs).

Methods.: Six FVMs were surgically removed from patients with proliferative diabetic retinopathy (PDR) during pars plana vitrectomy with membrane peeling. The FVMs were classified into three active FVMs or three inactive FVMs according to the presence or absence of neovascularization (NV) in the membranes. Total RNA was isolated from the six FVMs and also from three normal human retinas. The DNA microarray analysis was performed to compare the genes expressed in the FVMs to those in normal human retinas, and also between active and inactive FVMs. Ingenuity pathway analysis (IPA) was used to determine the key biological networks related to the genes that were significantly altered. Quantitative RT-PCR and immunohistochemistry were performed to validate the microarray analyses.

Results.: There were 87 genes expressed at significantly higher levels in FVMs than in normal human retinas. Functional classification of these genes showed that the most clustered genes were those related to extracellular matrix formation. The top biological network generated by the IPA was cellular assembly and organization involving nodes of genes related to extracellular matrix formation. These networks included the collagen family and matricellular proteins, THBS2, POSTN, and TNC. There were 91 genes significantly upregulated in active FVMs, and the most clustered functional category was angiogenesis. In contrast, 89 genes were significantly upregulated in inactive FVMs, and the most clustered functional category was metabolism. The IPA revealed that the top biological network related to the genes that were significantly altered in this comparison was cell-to-cell signaling, and interactions involving the PDGF and TGFβ families. The results of quantitative RT-PCR analyses and immunohistochemistry for several selected molecules were in good agreement with the microarray data.

Conclusions.: Our data indicate that extracellular matrix-related molecules such as POSTN, TNC, TGFβ, and angiogenic factors have important roles in promoting the development of FVMs associated with PDR.

Introduction

Proliferative diabetic retinopathy (PDR) is characterized by neovascularization due to retinal ischemia, and subsequent formation of fibrovascular membranes (FVMs) on the surface of the neuronal retina. The FVMs can cause vitreous hemorrhages and tractional retinal detachments causing a severe reduction of vision. Despite recent advances in the treatment of PDR, such as vitreoretinal surgery with use of anti-VEGF drugs as an adjunct, PDR still remains a vision-threatening disease.¹ 

The FVMs consist of stromal tissues and cellular components, such as retinal glial cells, macrophages/monocytes/hyalocytes, fibroblasts, and vascular endothelial cells.^2,3 The pathophysiology of FVM formation is complex and involves proliferation, attachment, migration of the cells, and deposition of extracellular matrix followed by the cicatricial contraction. Earlier studies showed that several molecules, such as basic FGF (bFGF), platelet-derived growth factors (PDGFs), TGFβs, and VEGF, were involved in these cellular processes.⁴ Histologically, collagen type 2 and fibronectin are the main extracellular matrix components.⁵ However, the molecular mechanisms regulating the development and progression of FVMs have not been fully determined to our knowledge. 

To determine the underlying pathomechanism of proliferative vitreoretinal diseases, our group has applied two comprehensive gene expression analyses; expressed sequence tag (EST) analysis and DNA microarray analysis. These analyses were done on FVMs associated with PDR, fibrous membranes with proliferative vitreoretinopathy (PVR) from human patients and ischemic retinas from mouse model of oxygen-induced retinopathy (OIR). Our EST analyses identified the possible disease-causing genes that had not been identified earlier. These genes included TEM7, which is highly expressed in neovascular endothelial cells in the FVMs,^6,7 and soluble CD44 and soluble vascular cellular adhesion molecule-1 (VCAM-1) that are elevated in the vitreous of patients with PVR.⁸ In the DNA microarray analysis of fibrous membrane associated with PVR, the expression level of periostin, a novel PVR-associated matricellular protein, was significantly upregulated compared to that in normal human retinas. In addition, we demonstrated that periostin could have a pivotal role in the generation of fibrous membranes, and an inhibition of periostin could slow the progression of PVRs in vivo.⁹ Additionally, DNA microarray analysis of OIR in mice showed that MIP-1β, also known as chemokine CC motif ligand 4 (Ccl4), was upregulated in hypoxic retinas. This could then have an important role in macrophage-mediated angiogenesis.^10,11 

Thus, we hypothesized that determining the gene expression profiles of FVMs will reveal the pathomechanisms related to the development of PDRs. To test this hypothesis, we have performed two comparative analyses. First, we compared the gene expression of FVMs to that of normal retinas. Although the cellular composition is not comparable between FVMs and retina, this analysis would be useful to determine the molecules that are preferentially expressed in FVMs with the goal that blockade of such molecules might become beneficial in combating FVMs with minimal damage to the normal retina. It is notable that this analysis could identify periostin as a novel molecular target, blockade of which could favorably control fibroproliferative retinal diseases with lacking unfavorable side effects.^9,12 Second, we compared the gene expression profiles of active FVMs to that in nonactive FVMs to identify the factors regulating the progression and regression of FVMs. 

Materials and Methods

Study Approval and Clinical Specimens

This study was approved by the Ethics Committee of the Kyushu University Hospital and Fukuoka University Chikushi Hospital, and the surgical specimens were handled in accordance with the ethical standards of the 1989 Declaration of Helsinki. All patients gave informed consent for the surgery and the use of excised specimens before inclusion in the study. The entry criteria included a diagnosis of type 2 diabetes, age ≥ 20 years, and hemoglobin A1c (HbA1c) ≤ 13%. For the eye, the eligibility criteria included the presence of a tractional retinal detachment associated with the FVMs and repeated vitreous hemorrhages. 

Six FVMs were surgically removed from six eyes of patients with PDR during pars plana vitrectomy with internal limiting membrane peeling. Three of these specimens were classified as active FVMs (age 57.3 ± 2.1 years; men/women, 1:2; duration of diabetes, 15.7 ± 2.1 years; glycosylated hemoglobin, 8.3%) and three as inactive FVMs (age 48.0 ± 13 years; men/women, 1:2; duration of diabetes, 12.3 ± 2.5 years; glycosylated hemoglobin, 6.1%) according to a published classification.¹³ Briefly, FVMs were defined as active if there were perfused preretinal capillaries, and inactive if previously documented active proliferation had regressed fully or if only nonperfused, gliotic, or fibrotic vessels were present. 

Vitreous hemorrhage was seen in one patient in the inactive FVM group, and tractional retinal detachment was seen in three in the active group and two in the inactive FVM group. Anterior chamber neovascularization was not seen in either group. All six patients had previous photocoagulation, but no previous anti-VEGF treatment. The excised FVMs were processed for RNA isolation, microarray analyses, real-time quantitative RT-PCR analyses, and immunohistochemistry. 

RNA Isolation

To extract the RNAs, FVMs were homogenized with a Polytron homogenizer (Polytron, Luzen, Switzerland) in an extraction reagent (TRIzol; Invitrogen-Life Technologies, Carlsbad, CA, USA), then extracted with chloroform, and the aqueous phase was precipitated in isopropanol. The pellet was washed with 75% ethanol, dissolved in water, and frozen at −80°C. The concentration and quality of the RNA were assessed by spectroscopy (ND-1000 spectrophotometer; NanoDrop Technologies, Wilmington, DE, USA), and the integrity of the RNA was assessed by electrophoresis. The RNAs from human retinas were obtained from Clontech (Palo Alto, CA, USA). According to the manufacturer's data sheet, RNAs were extracted from normal retinas pooled from 99 male/female Caucasians, ages 15 to 80. 

Microarray Analyses

We performed microarray analysis as described in detail.¹⁰ Briefly, 250 ng of isolated total RNA was converted to biotinylated-cRNA and hybridized to human WG-6 V3 Expression BeadChips (Illumina, San Diego, CA, USA) according to the manufacturer's instructions. The raw gene expression data were summarized using the Illumina Bead Studio software, and the normalized values were calculated with GeneSpring GX 11.0 software (Agilent Technologies, Santa Clara, CA, USA). The datasets subsequently were subjected to Welch's t-tests coupled with multidimensional false-discovery control (FDR2D; OCplus package from the repository implemented on the R platform; available in the public domain at http://www.bioconductor.org). The FDR cutoff value between FVMs and retina was 0.5%, and that for active and inactive FVMs was 15% for each of the group comparisons. Functional annotations were performed by Gene Ontology (GO; available in the public domain at http://www.geneontology.org) classifications with GO category; biological process obtained through appropriate public databases, such as DAVID Bioinformatics Resources (available in the public domain at http://david.abcc.ncifcrf.gov/home.jsp). Biological networks were evaluated with the Ingenuity Pathway Analysis (IPA) software (available in the public domain at www.ingenuity.com). All expression array data have been deposited in the Gene Expression Omnibus (GEO) with a GEO accession number of GSE60436 (available in the public domain at http://www.ncbi.nlm.nih.gov/geo). All microarray data are Minimum Information About a Microarray Experiment (MIAME) compliant. The associated networks were generated through the use of IPA (Ingenuity Systems, available in the public domain at www.ingenuity.com). The IPA predicts the functional networks based on known protein–protein and functional interactions. To generate the functional networks that specifically interacted with the data sets, the selected genes were overlaid onto a global molecular meshwork developed from information contained in the Ingenuity's Knowledge Base. Networks of these genes then were algorithmically generated based on their connectivity. A network is a graphical representation of the molecular relationships between molecules. Molecules are represented as nodes, and the biological relationship between two nodes is represented as an edge (line). The intensity of the node color indicates the degree of up- (red) or down (green) regulation. Nodes are displayed using various shapes that represent the functional class of the gene product. 

Real-Time Quantitative RT-PCR

Selected genes from the microarray analyses were validated by real-time quantitative RT-PCR (qPCR). The primers used are shown in Table 1. Total RNA was reverse-transcribed with a first-strand cDNA synthesis kit for RT-PCR (Roche Diagnostics, Indianapolis, IN, USA) according to the manufacturer's instructions. The reverse-transcribed cDNAs then were subjected to qPCR (SYBR Premix Ex Taq; TaKaRa Bio, Inc., Otsu, Japan) and thermal cycling (LightCycler; Roche Diagnostics) as described.¹⁰ The level of mRNA expression was estimated from the fluorescence intensity relative to β-actin. 

Table 1

 

Primer Sequences Used for Real-Time RT-PCR

Gene Symbol	Refseq	5′-Forward Primer Sequence-3′	5′-Reverse Primer Sequence-3′
COL1A1	NM_000088.3	GACGCCATCAAGGTCTACTG	ACGGGAATCCATCGGTCA
THBS2	NM_003247.2	AGACTCCGCATCGCAAAGG	TCACCACGTTGTTGTCAAGGG
TNC	NM_002160.1	TCCCAGTGTTCGGTGGATCT	TTGATGCGATGTGTGAAGACA
POSTN	NM_006475.1	TGCCCAGCAGTTTTGCCCAT	CGTTGCTCTCCAAACCTCTA
APLN	NM_017413.3	GTCTCCTCCATAGATTGGTCTGC	GGAATCATCCAAACTACAGCCAG
ANGPT2	NM_001147.1	AACTTTCGGAAGAGCATGGAC	CGAGTCATCGTATTCGAGCGG
CRYAA	NM_000394.2	GCGAGGGCCTTTTTGAGTATG	GGTCGGATCGAACCTCAGAGA
GFAP	NM_002055.2	CTGCGGCTCGATCAACTCA	TCCAGCGACTCAATCTTCCTC
ACTB	NM_001101	CATGTACGTTGCTATCCAGGC	CTCCTTAATGTCACGCACGAT

Immunohistochemistry

Immunohistochemistry was performed as described previously.¹⁴ The FVMs were embedded in paraffin and cut at a 3-μm thickness. After removing the paraffin, the sections were rehydrated, blocked, and incubated with primary antibodies (Abs) overnight at 4°C and then incubated with secondary Abs. The bound antibody was made visible by a conventional avidin-biotin-peroxidase protocol with 3,3′-diaminobenzidine as the substrate. Hematoxylin and eosin (H&E) staining was performed for histology. The primary Abs were SEMA3C, SEMA3F (Sigma-Aldrich, St. Louis, MO, USA), and CRYAA (ENZO Life Sciences, Farmingdale, NY, USA). 

Results

Microarray Analyses

In a normalization method, 24,056 probe IDs of the 48,803 probes on the microarray chip were called present (detection P ≦ 0.05) in at least 3 or more of the 9 total samples, which consisted of 3 replicate samples in each group; human retina, active FVMs, and inactive FVMs. These probe IDs were processed for the group comparisons. To identify the genes that are expressed preferentially in FVMs compared to normal retina, we compared the mRNA expressions of all 6 FVMs to those of normal human retinas. To identify the genes specifically regulated in the processes involved in the progression and regression of FVMs, we compared the mRNA expression levels between active and inactive FVMs. 

In the comparisons of the mRNA expressions of FVMs to those of normal human retinas, 176 probe IDs were identified as differentially expressed at an FDR P value < 0.5%, and 87 of these genes were preferentially expressed in FVMs (Table 2). On the other hand, 89 genes were preferentially expressed in human retinas (Supplementary Table S1). In the analysis comparing the FVMs with different activities, 180 probe IDs were identified as differentially expressed at an FDR P value < 15%; 91 genes were upregulated and 89 genes were downregulated in active FVMs relative to inactive FVMs (Tables 3, 4). 

Table 2

 

Genes Preferentially Expressed in FVMs Compared to Retina

Description	Gene Symbol	Refseq	P Value	Fold Change
Extracellular matrix
Collagen, type VI, α3, transcript variant 1	COL6A3	NM_004369.2	5.40E-06	42.44
Thrombospondin 2	THBS2	NM_003247.2	2.49E-07	36.74
Collagen, type I, α1	COL1A1	NM_000088.3	2.23E-09	36.34
Collagen, type V, α2	COL5A2	NM_000393.3	4.37E-08	30.53
Serpin peptidase inhibitor, clade E (nexin, plasminogen activator inhibitor type 1), member 1	SERPINE1	NM_000602.1	3.60E-06	27.04
Collagen, type IV, α1	COL4A1	NM_001845.4	4.00E-08	20.89
Periostin	POSTN	NM_006475.1	2.39E-07	19.01
Versican	VCAN	NM_004385.2	7.22E-06	16.52
Collagen, type V, α1	COL5A1	NM_000093.3	5.84E-06	14.18
Collagen, type I, α2	COL1A2	NM_000089.3	3.16E-08	14.04
Collagen, type IV, α2	COL4A2	NM_001846.2	3.71E-06	13.96
Collagen, type III, α1 (Ehlers-Danlos syndrome type  IV, autosomal dominant)	COL3A1	NM_000090.3	4.01E-07	11.28
Matrix metallopeptidase 23B	MMP23B	NM_006983.1	4.30E-06	8.57
Plasminogen activator, urokinase	PLAU	NM_002658.2	2.71E-06	7.84
Tenascin C, hexabrachion	TNC	NM_002160.1	1.30E-06	5.05
Integrin, beta 1; fibronectin receptor, beta  polypeptide, antigen CD29 includes MDF2, MSK12	ITGB1	NM_002211.2	1.34E-05	4.17
Annexin A1	ANXA1	NM_000700.1	4.66E-06	3.97
Extracellular matrix protein 2, female organ and  adipocyte specific	ECM2	NM_001393.2	8.90E-06	3.65
Cadherin 2, type 1, N-cadherin, neuronal	CDH2	NM_001792.2	1.18E-05	3.38
Metabolism
fibroblast activation protein	FAP	NM_004460.2	8.16E-08	16.40
Protease, serine, 23	PRSS23	NM_007173.4	1.09E-06	6.59
T-box 15	TBX15	NM_152380.2	5.09E-06	6.06
Lymphoid enhancer-binding factor 1	LEF1	NM_016269.2	7.43E-06	5.71
Nuclear receptor subfamily 5, group A, member 2	NR5A2	NM_003822.3	1.54E-06	4.66
Nuclear factor (erythroid-derived 2)-like 3	NFE2L3	NM_004289.5	7.37E-06	4.35
CAMP responsive element binding protein 5	CREB5	NM_182898.2	4.85E-06	3.55
Thyroid hormone receptor interactor 11	TRIP11	NM_004239.1	4.79E-08	3.34
Ribonuclease P/MRP 40kDa subunit	RPP40	NM_006638.2	5.73E-06	3.29
Immune response
Runt-related transcription factor 1; acute myeloid  leukemia 1; aml1 oncogene	RUNX1	NM_001754.3	6.17E-06	8.03
C-type lectin domain family 5, member A	CLEC5A	NM_013252.2	6.33E-06	7.37
Lymphocyte antigen 96	LY96	NM_015364.2	2.94E-06	5.43
Dedicator of cytokinesis 11	DOCK11	NM_144658.3	5.15E-06	5.42
Endoplasmic reticulum aminopeptidase 2	ERAP2	NM_022350.2	1.27E-05	5.15
Interferon, gamma-inducible protein 16	IFI16	NM_005531.1	5.53E-06	4.03
Interferon regulatory factor 1	IRF1	NM_002198.1	4.82E-06	3.19
Cell adhesion
Integrin, α11	ITGA11	NM_012211.3	2.95E-06	6.08
Peptidylprolyl isomerase C, cyclophilin C	PPIC	NM_000943.4	6.07E-08	5.91
FAT tumor suppressor homolog 1, Drosophila	FAT	NM_005245.3	2.64E-06	5.05
Hyaluronan and proteoglycan link protein 3	HAPLN3	NM_178232.2	9.74E-06	3.38
FXYD domain containing ion transport regulator 5	FXYD5	NM_014164.4	1.97E-06	3.16
Cell differentiation
Lymphoid enhancer-binding factor 1	LEF1	NM_016269.2	8.56E-06	5.75
Cyclin-dependent kinase 6	CDK6	NM_001259.5	5.35E-06	4.08
Neuropilin 1	NRP1	NM_003873.4	1.29E-05	3.66
ST8 α-N-acetyl-neuraminide α-2,8-sialyltransferase 4	ST8SIA4	NM_005668.3	1.93E-07	3.37
Schlafen family member 5	SLFN5	NM_144975.3	6.39E-06	3.06
Apoptosis
Sulfatase 1	SULF1	NM_015170.1	2.68E-06	6.38
Cyclin-dependent kinase inhibitor 1A; p21, Cip1	CDKN1A	NM_000389.2	2.72E-06	4.85
XIAP associated factor 1	XAF1	NM_199139.1	3.87E-06	4.65
Tumor necrosis factor receptor superfamily, member  10b	TNFRSF10B	NM_003842.3	3.10E-06	3.51
Caspase 4, apoptosis-related cysteine peptidase	CASP4	NM_001225.3	1.52E-07	3.18
Cell signaling
G protein-coupled receptor 116	GPR116	NM_015234.4	3.00E-06	7.87
Rho GTPase activating protein 15	ARHGAP15	NM_018460.2	9.36E-06	4.58
RAB31, member RAS oncogene family	RAB31	NM_006868.2	5.11E-06	3.61
G protein-coupled receptor 65	GPR65	NM_003608.2	5.29E-07	3.41
RAB23, member RAS oncogene family	RAB23	NM_016277.3	1.03E-05	3.20
Cell proliferation
Sterile α motif domain containing 9	SAMD9	NM_017654.2	2.54E-08	7.01
Sterile α motif domain containing 9-like	SAMD9L	NM_152703.2	8.53E-06	5.04
S100 calcium binding protein A11	S100A11	NM_005620.1	7.03E-06	3.52
Nucleolar protein 8	NOL8	NM_017948.4	2.55E-06	3.06
Transport
Nicotinamide N-methyltransferase	NNMT	NM_006169.2	6.56E-06	5.94
Transient receptor potential cation channel, subfamily C, member 6	TRPC6	NM_004621.4	1.61E-06	4.40
Structural maintenance of chromosomes 4	SMC4	NM_001002800.1	6.23E-06	4.31
Cytoskeleton
Caldesmon 1	CALD1	NM_033138.2	1.35E-06	5.08
Retinoic acid induced 14	RAI14	NM_015577.1	7.35E-06	4.29
Stress response
RecQ protein-like; DNA helicase Q1-like	RECQL	NM_002907.2	5.51E-06	3.48
Zinc finger CCCH-type, antiviral 1	ZC3HAV1	NM_020119.3	3.35E-06	3.33
Transcription
Bromodomain adjacent to zinc finger domain, 1A	BAZ1A	NM_182648.1	2.82E-06	4.48
Ets variant gene 6, TEL oncogene	ETV6	NM_001987.4	1.04E-06	3.37
Cellular process
Tropomyosin 4	TPM4	NM_003290.1	4.30E-06	5.47
Cell cycle
Chromosome 14 open reading frame 106	C14orf106	NM_018353.3	4.11E-06	3.09
Cell migration
Collagen triple helix repeat containing 1	CTHRC1	NM_138455.2	1.20E-07	14.84
Coagulation
Coagulation factor II, thrombin receptor	F2R	NM_001992.2	1.66E-08	8.41
Sensory perception
Latexin	LXN	NM_020169.2	6.77E-07	3.42
Unknown
Mo sapiens hypothetical protein DKFZp761P0423	DKFZp761P0423	XM_291277.4	5.11E-06	6.22
Mo sapiens hypothetical protein MGC4677.	MGC4677	XM_939115.1	4.32E-06	4.89
Chromosome 4 open reading frame 18	C4orf18	NM_016613.5	6.33E-06	4.61
Chromosome 18 open reading frame 34	C18orf34	NM_198995.1	1.44E-07	4.35
Spermatogenesis associated 18 homolog (rat)	SPATA18	NM_145263.2	4.73E-06	3.87
Chromosome 12 open reading frame 35	C12orf35	NM_018169.2	4.42E-06	3.85
Leucine rich repeat containing 32	LRRC32	NM_005512.1	7.21E-06	3.82
4 jointed box 1, Drosophila	FJX1	NM_014344.2	1.14E-05	3.62
Outer dense fiber of sperm tails 2-like	ODF2L	NM_001007022.1	3.27E-07	3.61
Ankyrin repeat domain 50	ANKRD50	NM_020337.1	1.33E-05	3.43
Maternally expressed 3 on chromosome 14. XR_001346-XR_001372	MEG3	NR_002766.1	1.23E-05	3.25
Interferon stimulated exonuclease gene 20kDa-like 1.	ISG20L1	NM_022767.2	2.48E-06	3.15
CDNA clone IMAGE:5881642 3, sequence	EST	BM999001	1.80E-06	3.14
Fer-1-like 3, myoferlin, C. elegans	FER1L3	NM_133337.1	8.61E-06	3.14

Table 3

 

Genes Significantly Upregulated in Expression in Active FVMs Compared to Inactive FVMs

Description	Gene Symbol	Refseq	P Value	Fold Change
Angiogenesis
Apelin	APLN	NM_017413.3	0.0037	10.99
Matrix metallopeptidase 9; gelatinase B, 92kDa gelatinase,  92kDa type IV collagenase	MMP9	NM_004994.2	0.0046	6.11
Angiopoietin 2	ANGPT2	NM_001147.1	0.0052	4.75
Endothelial cell-specific molecule 1	ESM1	NM_007036.2	0.0004	3.44
CD34 antigen, transcript variant 2	CD34	NM_001773.1	0.0034	3.05
Nestin	NES	NM_006617.1	0.0067	2.66
Sema domain, immunoglobulin domain (Ig), short basic  domain, secreted, (semaphorin) 3F	SEMA3F	NM_004186.2	0.0037	2.62
Cadherin 5, type 2, VE-cadherin; vascular epithelium	CDH5	NM_001795.2	0.0076	2.43
Delta-like 4; Drosophila	DLL4	NM_019074.2	0.0025	2.43
Thy-1 cell surface antigen.	THY1	NM_006288.2	0.0072	2.27
Roundabout homolog 4, magic roundabout, Drosophila	ROBO4	NM_019055.4	0.0047	2.04
Annexin A2 pseudogene 1 on chromosome 4	ANXA2P1	NR_001562.1	0.0016	1.69
Stress response
Hydroxysteroid (11-β) dehydrogenase 2	HSD11B2	NM_000196.3	0.0004	3.03
Dysferlin, limb girdle muscular dystrophy 2B, autosomal  recessive	DYSF	NM_003494.2	0.0036	2.80
HIG1 domain family, member 1B	HIGD1B	NM_016438.2	0.0034	2.24
SH2 domain containing 3C	SH2D3C	NM_170600.1	0.0050	2.15
Secretory carrier membrane protein 5	SCAMP5	NM_138967.2	0.0006	1.98
NADPH oxidase 4	NOX4	NM_016931.2	0.0021	1.81
Protein phosphatase, EF-hand calcium binding domain 1	PPEF1	NM_006240.2	0.0028	1.80
Protein tyrosine phosphatase, receptor type, A	PTPRA	NM_002836.2	0.0001	1.35
ATG16 autophagy related 16-like 1, S. cerevisiae	ATG16L1	NM_017974.3	0.0001	1.31
Metabolism
Inhibin, beta B, activin AB beta polypeptide	INHBB	NM_002193.1	0.0077	4.89
Carbohydrate (keratan sulfate Gal-6) sulfotransferase 1	CHST1	NM_003654.3	0.0274	4.15
Serine/threonine kinase 32B	STK32B	NM_018401.1	0.0034	3.36
Sprouty homolog 4, Drosophila	SPRY4	NM_030964.2	0.0014	2.50
Protein tyrosine phosphatase type IVA, member 3	PTP4A3	NM_007079.2	0.0069	2.46
SRY (sex determining region Y)-box 7	SOX7	NM_031439.2	0.0058	2.09
Cholesteryl ester transfer protein, plasma	CETP	NM_000078.1	0.0004	1.84
Protein tyrosine phosphatase, receptor type, E	PTPRE	NM_130435.2	0.0030	1.67
Cell differentiation
Leucine rich repeat and Ig domain containing 1	LINGO1	NM_032808.5	0.0068	2.92
Transforming growth factor, β3	TGFB3	NM_003239.1	0.0007	2.41
Limb bud and heart development homolog, mouse	LBH	NM_030915.1	0.0081	2.32
Homo sapiens Notch homolog 3, Drosophila	NOTCH3	NM_000435.1	0.0082	2.29
H2.0-like homeobox	HLX	NM_021958.2	0.0064	2.26
Adenosine A2a receptor	ADORA2A	NM_000675.3	0.0061	1.90
Cysteine-rich protein 2	CRIP2	NM_001312.2	0.0029	1.78
Extracellular matrix
CD36 molecule, thrombospondin receptor	CD36	NM_000072.2	0.0029	3.55
Laminin, α4	LAMA4	NM_002290.2	0.0051	2.31
Platelet-derived growth factor beta polypeptide	PDGFB	NM_002608.1	0.0091	2.29
Integrin, α1	ITGA1	NM_181501.1	0.0015	2.21
Gglypican 1	GPC1	NM_002081.1	0.0056	1.91
Olfactomedin-like 3	OLFML3	NM_020190.2	0.0050	1.81
Cytoskeleton organization
Tubulin polymerization-promoting protein family member 3	TPPP3	NM_015964.2	0.0007	2.31
Transforming, acidic coiled-coil containing protein 2	TACC2	NM_206861.1	0.0015	1.99
Homo sapiens formin homology 2 domain containing 1	FHOD1	NM_013241.2	0.0069	1.96
Tubulin, β2C	TUBB2C	NM_006088.5	0.0020	1.90
ZW10 interactor	ZWINT	NM_001005413.1	0.0021	1.87
Ribosomal protein L29	RPL29	NM_000992.2	0.0046	1.76
FERM domain containing 4A	FRMD4A	NM_018027.3	0.0044	1.72
Apoptosis
Tribbles homolog 3, Drosophila	TRIB3	NM_021158.3	0.0074	2.56
Nonmetastatic cells 1	NME1	NM_000269.2	0.0018	1.95
Rho guanine nucleotide exchange factor (GEF) 17	ARHGEF17	NM_014786.2	0.0056	1.91
V-ets erythroblastosis virus E26 oncogene homolog 1, avian	ETS1	NM_005238.2	0.0015	1.72
Dedicator of cytokinesis 1	DOCK1	NM_001380.3	0.0025	1.68
Cell adhesion
Protocadherin 12	PCDH12	NM_016580.2	0.0058	2.82
Melanoma cell adhesion molecule	MCAM	NM_006500.2	0.0042	2.33
Parvin, β	PARVB	NM_013327.3	0.0055	2.21
Nidogen 2, osteonidogen	NID2	NM_007361.3	0.0029	1.84
Transport
Potassium inwardly-rectifying channel, subfamily J, member 2	KCNJ2	NM_000891.2	0.0008	3.99
Potassium voltage-gated channel, subfamily F, member 1	KCNF1	NM_002236.4	0.0020	2.63
GrpE-like 2, mitochondrial (E. coli), nuclear gene encoding  mitochondrial protein	GRPEL2	NM_152407.3	0.0076	2.33
Potassium intermediate/small conductance calcium-activated  channel, subfamily N, member 2	KCNN2	NM_021614.2	0.0035	1.99
Immune response
Major histocompatibility complex, class II, DR β5	HLA-DRB5	NM_002125.3	0.3859	6.11
Chemokine (C-X-C motif) receptor 7	CXCR7	NM_001047841.1	0.0043	3.70
Family with sequence similarity 19 (chemokine (C-C motif)-  like), member A3	FAM19A3	NM_182759.2	0.0017	2.05
Tumor necrosis factor, α-induced protein 6	TNFAIP6	NM_007115.2	0.0054	1.95
Cell cycle
Ras association (RalGDS/AF-6) domain family member 2	RASSF2	NM_170774.1	0.0075	2.52
Centromere protein N	CENPN	NM_018455.3	0.0025	2.26
Cyclin F	CCNF	NM_001761.1	0.0030	1.96
Signaling
Hairy/enhancer-of-split related with YRPW motif-like	HEYL	NM_014571.3	0.0036	2.60
Guanine nucleotide binding protein (G protein), γ11	GNG11	NM_004126.3	0.0006	2.49
FK506 binding protein 1A	FKBP1A	NM_000801.2	0.0033	1.77
Motor activity
Myosin IB	MYO1B	NM_012223.2	0.0036	2.09
Wound healing
Plasmalemma vesicle associated protein	PLVAP	NM_031310.1	0.0040	3.29
Unknown
PREDICTED: Similar to aggrecan 1 isoform 2 precursor	LOC649366	XM_938439.1	0.0372	7.56
Nitric oxide synthase 2A. inducible, hepatocytes	NOS2A	NM_153292.1	0.0041	2.68
Actin filament associated protein 1-like 1	AFAP1L1	NM_152406.1	0.0021	2.52
PREDICTED: Similar to Caspase-4 precursor (ICH-2 protease) (TX protease) (ICE(rel)-II) (LOC648470)	LOC648470	XM_937514.1	0.0085	2.41
Chromosome 3 open reading frame 54	C3orf54	NM_203370.1	0.0095	2.28
Family with sequence similarity 124B, transcript variant 2	FAM124B	NM_024785.2	0.0079	2.23
KIAA0672 gene product	KIAA0672	NM_014859.3	0.0028	2.14
AGENCOURT_8109349 Lupski_sympathetic_trunk cDNA clone IMAGE:6189406 5, sequence	EST	BQ717127	0.0014	2.08
Popeye domain containing 2	POPDC2	NM_022135.2	0.0030	1.98
PREDICTED: Similar to 6 transmembrane epithelial antigen  of prostate	MGC87042	XM_001128032.1	0.0003	1.96
WD repeat domain 51A	WDR51A	NM_015426.2	0.0032	1.90
Family with sequence similarity 101, member B	FAM101B	NM_182705.2	0.0053	1.89
Chromosome 1 open reading frame 54	C1orf54	NM_024579.2	0.0019	1.82
Chromosome 16 open reading frame 30	C16orf30	NM_024600.2	0.0025	1.77
CDNA FLJ41846 fis, clone NT2RI3003162	EST	AK123840	0.0014	1.77
PREDICTED: Similar to FK506-binding protein 1A	LOC642489	XM_925989.1	0.0018	1.77
PREDICTED: Similar to tubulin, β5	LOC647000	XM_929980.2	0.0020	1.65

Table 4

 

Genes Significantly Upregulated in Expression in Inactive FVMs Compared to Active FVMs

Description	Gene Symbol	Refseq	P Value	Fold Change
Metabolism
Serpin peptidase inhibitor, clade A (α-1 antiproteinase, antitrypsin), member 3	SERPINA3	NM_001085.4	0.0610	4.99
Chitinase 3-like 2	CHI3L2	NM_004000.2	0.0275	4.82
Lysozyme, renal amyloidosis	LYZ	NM_000239.1	0.0427	3.69
ATPase, Na+/K+ transporting, α2 (+) polypeptide	ATP1A2	NM_000702.2	0.0902	3.50
Sortilin-related VPS10 domain containing receptor 1	SORCS1	NM_001013031.1	0.0579	3.19
Decorin	DCN	NM_133503.2	0.0020	3.18
Nuclear receptor subfamily 1, group H, member 4	NR1H4	NM_005123.1	0.0092	2.93
UDP-N-acetyl-α-D-galactosamine:polypeptide N-acetylgalactosaminyltransferase-  like 1	GALNTL1	NM_020692.1	0.0322	2.90
Melanocortin 2 receptor accessory protein	MRAP	NM_178817.3	0.0012	2.83
Nuclear factor of κ light polypeptide gene enhancer in B-cells inhibitor, zeta	NFKBIZ	NM_001005474.1	0.0518	2.77
Microsomal glutathione S-transferase 1	MGST1	NM_145792.1	0.0270	2.72
CCAAT/enhancer binding protein (C/EBP), delta	CEBPD	NM_005195.3	0.0018	2.66
Mannosidase, α, class 1C, member 1	MAN1C1	NM_020379.2	0.0005	2.51
Glycogenin 2	GYG2	NM_003918.2	0.0022	1.69
Proteolipid protein 1; Pelizaeus-Merzbacher disease, spastic paraplegia 2,  uncomplicated	PLP1	NM_199478.1	0.0026	1.65
Immune response
PREDICTED: Similar to Ig κ chain V-I region HK102 precursor	LOC652493	XM_941953.1	0.0186	5.74
PREDICTED: Similar to Ig kappa chain V-I region HK101 precursor	LOC647450	XM_936518.1	0.0102	5.74
Duffy blood group, chemokine receptor, transcript variant 2.	DARC	NM_002036.2	0.1307	5.45
PREDICTED: Similar to Ig κ chain V-I region HK102 precursor	LOC652694	XM_942302.1	0.0343	4.08
PREDICTED: Similar to Ig γ-2 chain C region	LOC649923	XM_939003.1	0.0129	4.07
CD69 molecule	CD69	NM_001781.1	0.0743	3.54
Complement component 1, r subcomponent	C1R	NM_001733.4	0.0005	3.03
Interleukin 17 receptor B	IL17RB	NM_018725.3	0.0215	2.96
Complement factor I	CFI	NM_000204.2	0.0157	2.75
Major histocompatibility complex class I HLA-A29.1	HLA-A29.1	NM_001080840.1	0.6205	2.08
Complement component 1, s	C1S	NM_201442.1	0.0037	1.83
Major histocompatibility complex, class II, DR β1	HLA-DRB1	NM_002124.1	0.0092	1.17
Apoptosis
Crystallin, α A	CRYAA	NM_000394.2	0.0144	9.76
Paraneoplastic antigen MA3	PNMA3	NM_013364.4	0.0353	4.70
Aryl-hydrocarbon receptor nuclear translocator 2	ARNT2	NM_014862.3	0.0801	4.10
Cartilage oligomeric matrix protein	COMP	NM_000095.2	0.0938	3.87
Secreted frizzled-related protein 4	SFRP4	NM_003014.2	0.0119	3.85
Growth arrest and DNA-damage-inducible, β	GADD45B	NM_015675.2	0.0170	2.88
Secreted frizzled-related protein 1	SFRP1	NM_003012.3	0.0213	2.75
Mal, T-cell differentiation protein	MAL	NM_002371.2	0.0061	2.30
Growth arrest-specific 1	GAS1	NM_002048.1	0.0065	2.01
Signaling
Carboxypeptidase Z	CPZ	NM_001014447.1	0.0384	4.64
G protein-coupled receptor 37 (endothelin receptor type B-like)	GPR37	NM_005302.2	0.0960	3.91
RAS, dexamethasone-induced 1	RASD1	NM_016084.3	0.0290	3.37
Lysophosphatidic acid receptor 1	LPAR1	NM_057159.2	0.0031	2.91
GDNF family receptor α1	GFRA1	NM_005264.3	0.0002	2.03
Growth factor receptor-bound protein 14	GRB14	NM_004490.2	0.0044	1.98
Tudor and KH domain containing	TDRKH	NM_006862.3	0.0003	1.33
Transcription factor AP-2 gamma, activating enhancer binding protein 2 γ	TFAP2C	NM_003222.3	0.0002	1.31
A kinase (PRKA) anchor protein (gravin) 12	AKAP12	NM_005100.2	0.0002	1.29
Cell adhesion
Leucine rich repeat neuronal 2	LRRN2	NM_201630.1	0.0435	5.78
Microfibrillar-associated protein 4	MFAP4	NM_002404.1	0.0607	3.62
Osteomodulin	OMD	NM_005014.1	0.0089	3.47
Claudin 11, oligodendrocyte transmembrane protein	CLDN11	NM_005602.4	0.0250	3.39
Coxsackie virus and adenovirus receptor	CXADR	NM_001338.3	0.0293	3.19
Neuronal cell adhesion molecule	NRCAM	NM_005010.3	0.0436	2.73
CD48 molecule	CD48	NM_001778.2	0.0104	2.41
Ninjurin 2	NINJ2	NM_016533.4	0.0040	1.86
Cell growth
Retinoic acid receptor responder (tazarotene induced) 1	RARRES1	NM_206963.1	0.0157	4.98
Odd-skipped related 1, Drosophila	OSR1	NM_145260.2	0.0097	4.07
Transferrin	TF	NM_001063.2	0.0666	3.28
LIM homeobox 2	LHX2	NM_004789.3	0.0787	3.21
Fibromodulin	FMOD	NM_002023.3	0.0943	3.10
SRY (sex determining region Y)-box 2	SOX2	NM_003106.2	0.0801	3.09
Angiogenesis
Fibroblast growth factor receptor 3, achondroplasia, thanatophoric dwarfism	FGFR3	NM_022965.1	0.0296	2.81
Sema domain, immunoglobulin domain (Ig), short basic domain, secreted,  (semaphorin) 3C	SEMA3C	NM_006379.2	0.0003	2.64
Insulin-like growth factor 2 (somatomedin A)	IGF2	NM_001007139.3	0.0107	2.43
Bone morphogenetic protein 6	BMP6	NM_001718.4	0.0016	2.18
Carcinoembryonic antigen-related cell adhesion molecule 1, biliary glycoprotein)	CEACAM1	NM_001024912.1	0.0000	1.47
Oxydation
Flavin containing monooxygenase 2, nonfunctional	FMO2	NM_001460.2	0.0076	6.85
Alcohol dehydrogenase 1A (class I), α polypeptide	ADH1A	NM_000667.2	0.0280	3.30
Glutathione peroxidase 3 (plasma)	GPX3	NM_002084.3	0.0745	3.05
Monooxygenase, DBH-like 1	MOXD1	NM_015529.2	0.0024	2.56
Transport
FXYD domain containing ion transport regulator 1 (phospholemman)	FXYD1	NM_005031.3	0.0401	3.10
Calcium channel, voltage-dependent, T type, α1I subunit	CACNA1I	NM_001003406.1	0.0344	2.70
Solute carrier family 4, sodium bicarbonate cotransporter, member 4	SLC4A4	NM_003759.2	0.0243	2.33
Solute carrier family 2 (facilitated glucose transporter), member 9	SLC2A9	NM_001001290.1	0.0019	1.90
Extracellular matrix
Fibulin 1	FBLN1	NM_006486.2	0.0301	4.00
Collagen, type X, α 1(Schmid metaphyseal chondrodysplasia)	COL10A1	NM_000493.3	0.0469	2.75
Cartilage acidic protein 1	CRTAC1	NM_018058.4	0.0338	2.74
Cytoskeleton
Glial fibrillary acidic protein	GFAP	NM_002055.2	0.1059	5.07
Nebulette	NEBL	NM_006393.1	0.0268	2.66
Stress response
Ceruloplasmin, ferroxidase	CP	NM_000096.1	0.0345	2.75
Receptor tyrosine kinase-like orphan receptor 2	ROR2	NM_004560.2	0.0253	2.34
Unknown
LIM domain only 3, rhombotin-like 2	LMO3	NM_018640.3	0.0436	4.50
Complement component 1, q subcomponent-like 1	C1QL1	NM_006688.3	0.0366	3.44
MRNA; cDNA DKFZp686J0156, from clone DKFZp686J0156	EST	BX648964	0.0187	3.23
WD repeat domain 86	WDR86	NM_198285.1	0.0421	2.89
Chromosome 20 open reading frame 39	C20orf39	NM_024893.1	0.0080	2.85
FLJ31568 protein	FLJ31568	NM_152509.1	0.0201	2.50
Transmembrane protein 178	TMEM178	NM_152390.1	0.0027	2.14
Ribosomal protein S4, Y-linked 1	RPS4Y1	NM_001008.3	0.0807	1.49
F-box and leucine-rich repeat protein 5	FBXL5	NM_012161.2	0.0003	1.31
Chromosome 20 open reading frame 107	C20orf107	NM_001013646.2	0.0300	1.02

Microarray Validation by Quantitative Real-Time RT-PCR

To verify the outcome of the microarray analyses, qPCR analyses were performed on the same RNA samples used in the microarray studies. We chose COL1A1, THBS2, TNC, and POSTN from the genes preferentially expressed in FVMs compared to normal retina, and APLN and AGPT2 from the upregulated, and CRYAA and GFAP from the downregulated genes of the active and inactive FVMs. 

The changes in the genes analyzed were identical in terms of direction of variations and similar in the extent of the alterations (Fig. 1). These findings demonstrated the reliability of the microarray results.¹⁰ 

Figure 1

 

Real-time quantitative RT-PCR for microarray validation. Array validation was performed by quantitative RT-PCR for 8 selected genes; COL1A1, THBS2, TNC, and POSTN from Table 2; APLN and ANGPT2 from Table 3; and CRYAA and GFAP from Table 4. The β-Actin was used as an internal standard. Results are expressed as means ± SEMs. *P < 0.05, **P < 0.01; n = 6 in the group of FVMs and n = 3 in the groups of retina, active, and inactive FVMs.

Molecular and Biological Functions Associated With Genes Preferentially Expressed in FVMs Compared to Normal Retinas

There were 87 genes preferentially expressed in FVMs compared to that in normal retinas, and the maximum expression change was found for COL6A3 (42.4-fold, Table 2). These 87 genes were subdivided by functional subsets into those related to extracellular matrix, metabolism, immune response, cell adhesion, cell differentiation, and other functions according to the GO classification (Fig. 2A). The most clustered genes, ANXA1, CDH2, COL1A1, COL1A2, COL3A1, COL4A1, COL4A2, COL5A1, COL5A2, COL6A3, ECM2, MMP23B, POSTN, THBS2, TNC, ITGB1, PLAU, VCAN, and SERPINE1, were those related to extracellular matrix formation. 

Figure 2

 

Molecular and biological functions associated with genes significantly upregulated in FVMs compared to normal retinas. (A) Distribution of GO terms and (B) a top network with the highest score generated by IPA associated with genes significantly upregulated in FVMs compared to normal retina. The total number of analyzed genes was 87.

We used IPA to investigate the key biological functions related to the upregulated genes. This analysis generated 8 networks with scores ranging from 41 to 14 (Supplementary Table S2). The top network with the highest score was Connective Tissue Disorders, Cellular Assembly and Organization, and Cellular Function and Maintenance, and they involved nodes of genes related to extracellular matrix formation, such as collagen family and matricellular proteins, THBS2, POSTN, and TNC (Fig. 2B). In addition to these genes, TGFβ and PDGFBB emerged as key molecules closely linked to the genes in these networks. 

Next, IPA upstream regulator analysis was performed to identify the upstream transcriptional regulators that can explain the gene expression changes. This analysis identified 31 upstream regulators as significant (P < 1 × 10⁻⁹) including transcriptional regulators, growth factors chemical drugs, and others (Supplementary Table 4). Transcriptional factors, for example, MYC, SPDEF, and TP, and growth factors, for example TGF, FGF, and EGF, were predicted to regulate the genes preferentially expressed in FVMs. 

Molecular and Biological Functions Associated With Genes Significantly Changed in Active and Inactive FVMs

Among the 91 genes significantly upregulated in active FVMs compared to inactive FVMs, the maximum expression change was found for APLN (11.0-fold; Table 3). Those genes were subdivided into functional subsets that are related to angiogenesis, stress response, metabolism, cell differentiation, extracellular matrix, and other functions according to the GO classification (Fig. 3A). The most clustered was angiogenesis (ANGPT2, ANXA2P1, APLN, CD34, DLL4, MMP9, NES, ROBO4, SEMA3F, THY1, CDH5, and ESM1). Among the 89 genes significantly upregulated in inactive FVMs compared to active ones, the maximum expression change was found for CRYAA at 9.8-fold (Table 4). These genes were subdivided by functional subsets related to metabolism, immune response, apoptosis, signaling, cell adhesion, cell growth, and other functions according to the GO classification (Fig. 3B). 

Figure 3

 

Molecular and biological functions associated with the genes significantly changed between active and inactive FVMs. (A) Distribution of GO terms in genes significantly upregulated in active FVMs. The total number of the analyzed genes was 91. (B) Distribution of GO terms in genes significantly upregulated in inactive FVMs. The total number of the analyzed genes was 89. (C) Top network with the highest score generated by IPA associated with genes significantly altered between active and inactive FVMs. The total number of the analyzed genes was 180.

To identify key biological functions regulating the progression and regression of FVMs, 180 differentially expressed genes were identified by IPA. This analysis generated 10 networks with a score ranging from 48 to 13 (Supplementary Table S3). The top network with the highest score was cell-to-cell signaling and interaction, cellular compromise, cardiovascular system development, and function involving PDGF and TGFβ family, which are well-known growth factor families having critical roles in the pathogenesis of FVMs. Of note, Akt, a protein involved in multiple biological processes, including cell survival, proliferation, and growth, was shown to possess numerous interactions with the genes used in this network (Fig. 3C). 

The IPA upstream regulator analysis identified 25 upstream regulators as significant (P < 1 × 10⁻⁵), including transcriptional regulators, for example, Hif-1α and Notch, growth factors, for example, TGF-β and VEGFA, and chemical-endogenous, such as β-estradiol (Supplementary Table S5). 

Immunostaining of SEMA3C, SEMA3F, and CRYAA in FVMs

Next, to examine whether the molecules encoded by the selected genes were expressed at the protein level, we stained the sections of FVMs using Abs against SEMA3F, SEMA3C, and CRYAA. The H&E staining confirmed the presence of prominent neovascular vessels in active FVMs (Fig. 4A), whereas regression of neovascular vessels in inactive FVMs (Figs. 4B, 4C). Staining of SEMA3F can be seen in the neovascular endothelial cells in the active FVM (Fig. 4D). Staining of SEMA3C and CRYAA can be seen in the cells existing in the stromal tissue and the regressed neovascular vessels (Figs. 4E, 4F). Similar expression staining pattern can be seen in all the FVM specimens from the five PDR subjects examined. 

Figure 4

 

Localization of SEMA3F, SEMA3C, and CRYAA proteins in FVMs. (A–C) H&E staining of active FVMs (A) and inactive FVMs (B, C). Neovascular vessels are indicated by the arrows. (D–F) Immunostaining of SEMA3F in active FVMs (D) and SEMA3C (E) and CRYAA (F) in inactive FVMs. The SEMA3F staining can be seen in the neovascular endothelial cells in the active FVM. Staining of SEMA3C and CRYAA can be seen in the cells existing in the stromal tissue and the regressed neovascular vessels. Nuclei are stained blue. Scale bar: 100 μm.

Discussion

It is well known that the deposition of extracellular matrix is essential for developing neovascular vessel, and also for maintaining the structure of FVMs.^2,15,16 According to histological studies of FVMs, the structural components of extracellular matrices are collagen type II, a prominent component of stroma, and collagen type IV that is localized around neovascular vessels. In addition, proteins, such as fibronectin, laminin, and vitronectin, have been reported to be elements of FVMs.^2,17 Consistent with these observations, our comparative analysis between retina and FVMs showed that the most clustered function was extracellular matrix containing a variety of collagen family and matricellular proteins and their receptor, integrin. Among the genes, collagen type VI, II, IV, and V were highly expressed with the maximum expression difference in normal human retinas. On the other hand, the expression of collagen type IX was significantly higher in normal retinas indicating the existence of tissue-specific collagen types in the structure of FVMs and normal retina. This observation raises a concern that nonspecific vitreolytic enzymes, such as collagenase and dispase, that lyse all collagen types could have unfavorable effect on the normal structure of the retina.¹⁸ Thus, it may be important to develop reagents targeting specific subtypes of collagens that are highly expressed only in FVMs to minimize unfavorable side effects on the normal retina. 

A cluster of the genes at the top of the key functional networks revealed by IPA analysis was suggested to have a central role in FVMs development. They interact with each other and with several matricellular proteins, for example, THBS2, POSTN, and TNC in addition to structural ECM proteins. This agrees with our previous global gene expression profile using fibrous membranes associated with PVR.^8,9 

Matricellular proteins are suggested to interact with the extracellular environment, such as the matrix and cells. They also have biological functions through binding with receptors. Among the matricellular proteins, we focused on POSTN as a potential therapeutic target and showed an increased expression in the vitreous and FVMs obtained from PDR patients.⁷ In addition, POSTN has a significant role in cell proliferation, adhesion, migration, and collagen production during the generation of fibrous membranes in PVR.⁹ 

The TNC may be another potential therapeutic target, because it has been reported to be a major component of the ECM in FVMs.¹⁷ During tissue remodeling after cardiac injury, TNC has a role in releasing cardiomyocytes from their adherence to ECM through the upregulation of matrix metalloproteinases.¹⁹ Consistent with this, we also found that TNC was costained with αSMA-positive cells in FVMs, and the administration of recombinant TNC promoted tubular morphogenesis of human retinal endothelial cells (Ishikawa K, Yoshida S., manuscript in preparation). Interestingly, there was a significant correlation between the vitreous concentrations of TNC and POSTN in patients with PDR (Ishikawa K, Yoshida S, unpublished observation), and POSTN can promote the incorporation of TNC functioning as a bridge between ECMs resulting in the stabilization of the extracellular meshwork.²⁰ For a comprehensive understanding of the underlying mechanism of ECM formation in FVMs, further studies are necessary to determine the functional interaction of these matricellular proteins. 

The rationale for performing the comparative analysis between active and inactive FVMs was that it might obtain the functional gene sets between the presence of preretinal neovascularization, proliferative activity, and contractile property. The most clustered function, angiogenesis, contained well characterized molecules associated with hypoxia-induced neovascularization, such as, APLN,²¹ MMP9,²² ANGPT2,²³ ESM1,²⁴ CD34,²⁵ Nestin,²⁶ SEMA3F,²⁷ CDH5,²⁸ DLL4,²⁹ THY1,³⁰ ROBO4,³¹ and ANXA2.³² All of these molecules cause abnormal angiogenesis by interacting with VEGF, indicating the central role of VEGF in pathological neovascularization in PDR as proven by the clinical efficacy of bevacizumab with marked regression of neovascularization in FVMs.³³ 

Interestingly, we found that the gene expressions of two different types of class 3 semaphorin (SEMA3) had distinct patterns, upregulation of SEMA3F in active FVMs, and SEMA3C in inactive FVMs. The SEMAs are a family of secreted and membrane-bound proteins regulating multiple biological processes, including angiogenesis and cancer. The SEMA3s bind to plexins and neuropilins, also known as a receptor for VEGF, inducing pro- or anti-angiogenesis depending on diverse patterns of ligand-receptor interactions.³⁴ Subsets of the SEMA3 family are reported to regulate retinal angiogenesis, such as SEMA3E in physiological vascular regrowth in ischemic retinas,³⁵ and SEMA3A and SEMA3F in antiangiogenic effects in retinal neovascularization.^36,37 However, their roles have not been fully determined. Our data showing distinct gene expression patterns of the SEMA3 types could help in further studies on their diverse roles in hypoxia-induced retinal angiogenesis. The inflammation-related genes upregulated in active FVMs were chemokines, TNF and MHC. This can suggest the involvement of inflammatory cells, especially macrophages, in retinal neovascularization in PDR. Proangiogenic roles of macrophages have been well characterized.³⁸ Our previous work of gene expression profiling in the mouse model of OIR demonstrated the significance of chemokines and macrophages in hypoxia-induced retinal angiogenesis.^10,11 The present study confirmed the relevance of inflammation in retinal neovascularization. Taken together, further studies with use of mouse OIR model might help to facilitate the understanding of the functional role of these novel angiogenesis-related molecules. 

The gene with the highest change in expression in inactive FVMs was CRYAA, a member of the small heat shock proteins family, known to serve as a cell protection against apoptosis in retinal disorders, including diabetic retinopathy.^39,40 Studies have shown that apoptosis is involved in FVMs regression, and laser photocoagulation induces FVMs regression with an increase in the number of apoptotic cells in the FVMs.⁴¹ In addition, caspase-3 is highly expressed in the apoptotic cells.⁴² Because CRYAA exerts its antiapoptotic function through the inhibition of caspase-3 activation,⁴³ our observation of an upregulation of CRYAA in inactive FVMs could possibly reflect the cell survival response to the decreased expression of proliferation-inducible growth factors, such as VEGF. Although we showed the changes in the expression of apoptosis-related genes, further studies are necessary to identify the cell types expressing apoptosis-related molecules, such as endothelial cells or glial cells and the functional role of these molecules. 

The IPA analysis identified β-estradiol, a member of female sex hormone, as an upstream regulator with the lowest P value that can inhibit the genes altered in comparison between active and inactive FVMs. The β-estradiol is known to influence numerous metabolic processes⁴⁴ and its target genes, that is, INHBB, DCN, and SRPINA3, were clustered into metabolism. Interestingly, a recent report demonstrated that β-estradiol could inhibit TGFβ2-induced collagen contraction, ECM production, and α-SMA in the in vitro system of PVR. These suggest that β-estradiol might be an important player regulating fibrosis in the fibroproliferative retinal diseases. 

The top biological network inferred by IPA revealed that TGFβ and PDGF as prominent growth factors that characterize the activity of FVMs. It is well known that these growth factors have significant roles in promoting cell proliferation and transdifferentiation into fibroblasts, and contraction of cells in a collagen matrix in the development of fibrosis associated with proliferative retinal diseases.^45–47 Our comprehensive analysis could certainly identify them as central players determining proliferative activity and contractile property, one of the characteristics seen in active FVMs. Moreover, a key signaling pathway interacting with the genes in the network was the Akt pathway, which is known to mediate multiple biological processes, such as angiogenesis and fibrosis through activation of TGF and PDGF receptors in the pathogenesis of diabetic retinopathy.^48,49 In an earlier report, the Akt pathway was suggested as a potential therapeutic target against vascular leakage and angiogenesis through modulation of VEGF signaling, and inflammation through modulation of NF-κB.⁵⁰ 

In conclusion, we identified FVM-specific molecules by microarray analyses, especially matricellular proteins, such as POSTN and TNC. These molecules might be potential therapeutic targets to block the development and slow the progression of FVMs by blocking ECM-related proteins without disturbing normal retinal function. In addition, we identified genes that regulated angiogenesis and apoptosis in active FVMs. Thus, combining the molecular targeting reagents against those genes that determine the activity of FVMs with anti-POSTN/TNC reagents may offer more effective and individualized treatments based on the clinical appearance of the FVMs. 

Acknowledgments

The authors thank the staff of the Research Support Center (Graduate School of Medical Sciences, Kyushu University) for technical support, and Mari Imamura and Masayo Eto (Kyushu University) for their excellent technical assistance. 

Supported by a fellowship from The Japan Society for the Promotion of Science Postdoctoral Fellowships for Research Abroad (KI), and in part by JSPS KAKENHI Grants 24249083 26293374, 26670757, and Takeda Science Foundation. 

Disclosure: K. Ishikawa, None; S. Yoshida, None; Y. Kobayashi, None; Y. Zhou, None; T. Nakama, None; S. Nakao, None; Y. Sassa, None; Y. Oshima, None; H. Niiro, None; K. Akashi, None; T. Kono, None; T. Ishibashi, None