Gene Expression Profile of Hyperoxic and Hypoxic Retinas in a Mouse Model of Oxygen-Induced Retinopathy

Abstract

Purpose.: To determine a profile of gene expression in retinas of a murine model of oxygen-induced retinopathy (OIR).

Methods.: OIR was induced in C57BL/6N mice by exposing postnatal day (P)7 pups to 75% oxygen for 5 days and then returning them to room air at P12. Gene microarrays containing more than 47,000 transcripts were used to study the changes in gene expression in retinas isolated immediately (P12) and at 12 hours (P12.5) after exposure to hyperoxia. The retinas of P12 mice raised under normoxic conditions served as control subjects. Quantitative RT-PCR and multiplex ELISA were performed to validate the microarray analyses.

Results.: The expression of 83 gene transcripts was significantly altered in the hyperoxic P12 retinas. These genes were classified as cellular components or were associated with development, metabolism, transport, stress response, cell adhesion, inflammation, or vision. The genes related to retinal growth, such as Pdgfb and Robo4, which are associated with vascular development, were downregulated. In contrast, the expression levels of 95 genes were significantly altered in the hypoxic P12.5 retinas, which contained several known hypoxia-regulated genes including Vegfa and Hif1a. The differentially expressed genes were broadly clustered into the development, inflammation, metabolism, signaling, antiapoptosis, cellular component, transport, glycolysis, and vision groups. Those associated with organogenesis (e.g., Vegfa, Igfbp3, Tnfrsf12a, and Nestin) and to inflammation (e.g., Ccl3, Ccl4, and MHCs) were upregulated. The results of quantitative RT-PCR and multiplex ELISA were in agreement with the microarray data.

Conclusions.: These alterations in gene expression may determine the hyperoxic growth retardation, postischemic inflammation, neovascularization, and remodeling in retinas of murine OIR.

Retinal ischemia is an important intermediate step in the pathogenesis of many retinal neovascular diseases, such as diabetic retinopathy, retinopathy of prematurity, and retinal vein occlusion. The ischemia is followed by vascular dysfunction resulting in retinal neovascularization, which can lead to irreversible damage to visual function. The ischemia-induced mediators of angiogenesis have been determined in well-established murine models of oxygen-induced retinopathy (OIR). Using this model, researchers have shown that vascular endothelial growth factor (VEGF) is a major oxygen-regulated mediator of angiogenesis. ¹ Other angiogenic factors, such as angiopoietin-2, ² hypoxia inducible factor-1α, ³ and NAD(P)H oxidase, ⁴ have been reported to be also induced by retinal ischemia in this model. 

We have demonstrated that the transcription nuclear factor-κB, ^5,6 macrophage inflammatory protein (MIP)-1α, monocyte chemoattractant protein (MCP)-1, ⁷ and interleukin (IL)-8 ⁸ were induced in ischemic retinas, and these factors may be associated with retinal neovascularization. ^9–11  

Earlier conventional studies investigating the molecular effects of ischemia on the retina have focused mainly on one or a few molecules or pathways. Therefore, the molecular events taking place in ischemic retinas that may lead to retinal neovascularization remain undetermined. 

The advent of microarray technology, with its capacity to monitor the expression of thousands of genes simultaneously, has made it possible to determine a comprehensive pattern of gene expression in a specific tissue. ^12,13 Thus, Sato et al. ¹⁴ determined the gene expression profile in murine OIR by microarray-based gene profiling (GeneChip; Affymetrix, Santa Clara, CA). They focused only on a selected set of 94 genes related to inflammation and angiogenesis by performing RT-PCR with a low-density array (TLDA; TaqMan, Applied Biosystems, Inc. [ABI], Foster City, CA) selected from 1304 genes extracted by comparing single DNA microarray hybridization between retinas obtained from murine OIR and controls. 

Because their study focused only on genes related to inflammation and angiogenesis, we hypothesized that microarray analyses of the same murine model of OIR would reveal a new functional set of genes. To test this hypothesis, we applied microarray analyses in murine retinas subjected to OIR, with DNA microarray technology (Illumina, San Diego, CA). Plausible molecular signatures associated with retinal ischemia in murine OIR are discussed. 

Materials and Methods

Murine Model of OIR

All experimental procedures on the animals were performed according to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

OIR was induced in C57BL/6N mice by the protocol described by Smith et al. ¹⁵ Briefly, litters of 7-day-old (postnatal day [P]7) C57BL/6N pups along with their mothers were placed in a 75% ± 2% oxygen atmosphere for 5 days and then returned to room air at age P12. ⁶ The mice were killed by cervical dislocation, and the eyes were enucleated. Whole retinas free of vitreous were isolated from the mice immediately (P12), at 12 hours (P12.5), and at 1 day (P13) after exposure to hyperoxia. Pups raised under normoxic conditions were killed at P12 as control subjects. All samples were stored at −80°C until they were used. 

To isolate sufficient total RNA for labeling protocols, the two retinas from one mouse were pooled. For enzyme-linked immunosorbent assays, the four retinas from two mice raised under the same conditions were pooled. To minimize false positives due to biological variations, the results of three experiments for each condition (biological replicates) were used for the statistical analyses. 

RNA Isolation

To extract the RNA, retinas isolated from normal P12 and hyperoxia-exposed P12 and P12.5 mice were homogenized (Polytron; Kinematica GmbH, Luzern, Switzerland) in extraction reagent (TRIzol; Invitrogen-Life Technologies, Carlsbad, CA), 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), and the integrity of the RNA was assessed by electrophoresis. ^16,17  

Microarray Analyses

We used a mouse microarray (Sentrix BeadChip Array Mouse-6; Illumina), which allowed us to assay more than 47,000 transcripts, largely on the MEEBO (Mouse Exonic Evidence Based Oligonucleotide; http://waterdragon.stanford.edu/alizadehlab/credits.html) set and supplemented with more than 11,000 probes for additional targets from RIKEN FANTOM2 (http://fantom2.gsc.riken.jp/db/ provided in the public domain by Riken Bioresource Center) ¹⁸ and the National Center for Biotechnology Information (NCBI; Bethesda, MD) Reference Sequence (RefSeq) database. The cRNA preparation and array hybridization were performed with a microarray technique (Illumina). 

A total of 250 ng of isolated total RNA was converted to biotinylated-cRNA according to the manufacturer's procedures (Illumina TotalPrep RNA Amplification Kit; Ambion, Austin, TX). Briefly, reverse transcription to synthesize first-strand cDNA was performed for 2 hours at 42°C, primed with an oligo (dT) primer bearing a T7 promoter, and catalyzed by reverse transcriptase (ArrayScript; Ambion). Second strand cDNA was synthesized by adding DNA polymerase I, and RNase H, and incubated for 2 hours at 16°C. After cDNA purification, the eluted cDNA was used for in vitro transcription with T7 RNA polymerase. In vitro transcription was performed at 37°C for 14 hours, yielding multiple copies of biotinylated antisense RNA molecules from each mRNA in the sample. The measurement of the cRNA yield was determined spectrophotometer (ND 1000; NanoDrop). 

A total of 1500 ng of biotin-cRNA from each sample was loaded on an individual array spot on the mouse microarray (Mouse-WG6 BeadChip; Illumina), according to the hybridization protocol. The chips were hybridized at 58°C for 19 hours, washed, fluorescently labeled, and scanned in the array reader (BeadArray Reader; Illumina). Raw gene expression data were summarized using the accompanying software (Bead Studio; Illumina). The microarray datasets have been submitted to the GEO database under the identifier GSE19886 (www.ncbi.nlm.nih.gov/geo/ provided in the public domain by NCBI). 

Normalization (quantile method), calculation of signal intensities, and the change ratio analyses between groups were performed (Genespring ver., 7.3; Agilent Technologies, Santa Clara, CA). Three experimental replicates were hybridized independently resulting in three microarray replicates for each condition. The ratios of the average signal intensity were then calculated for the probesets (hyperoxic P12 retina relative to control normoxic P12 retina, hypoxic P12.5 retina relative to hyperoxic P12 retina, and hypoxic P12.5 retina relative to control normoxic P12 retina). If the signals were low, these probesets are reported as having a detection score of less than 0.99 (based on Bead Studio; Illumina) in all microarrays. For each condition, the probesets with the signals considered to be absent were removed. The filtered datasets were subsequently subjected to Welch's t-test ¹⁹ coupled with multidimensional false-discovery control (FDR2D) ²⁰ (OCplus package from the repository implemented on the R platform; http://www.bioconductor.org/ provided in the public domain by Bioconductor, open source and open development software project for the analysis and comprehension of genomic data). ²¹  

We compared the number of differentially expressed genes at various FDR levels and determined the cutoff values in order that the number in each group comparison was similar to one another. Consequently, the FDR cutoff values were 5%, 5%, and 7%, respectively, for each of the group comparisons between hyperoxic P12 and control normoxic P12 retinas; hypoxic P12.5 and hyperoxic P12 retinas; and hypoxic P12.5 and control normoxic P12 retinas. 

Functional annotation was performed by Gene Ontology (GO; http://www.geneontology.org/ developed by a consortium and provided in the public domain by The Gene Ontology Project) classifications with GO category obtained through appropriate public databases such as DAVID Bioinformatics Resources (http://david.abcc.ncifcrf.gov/home.jsp/ Database for Annotation, Visualization and Integrated Discovery, provided in the public domain by the National Institute of Allergy and Infectious Diseases). ^22,23  

Quantitative Real-Time Reverse Transcription–Polymerase Chain Reaction

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), according to the manufacturer's instructions. For all samples, a 1:50 dilution was used for the qPCR analyses. All samples were stored at −20°C until analyzed. The reverse-transcribed cDNAs were then subjected to qPCR (SYBR Premix Ex Taq; TaKaRa Bio Inc., Otsu, Japan) and thermal cycling (LightCycler; Roche Diagnostics). The reaction conditions were denaturing at 95°C for 10 seconds followed by 40 cycles of denaturing at 95°C for 5 seconds, and annealing and extending at 60°C for 20 seconds. 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 Name	Accession No.	5′-Forward Primer Sequence-3′	5′-Reverse Primer Sequence-3′
Vtn	NM_011707	TGTTTGAGCACTTTGCCTTG	GATAGCGCTTTCGGCTTCTA
Emcn	NM_016885	CCAAAAGTGACGTATCCCAAA	TGTTCTGGGAACCTGGTAGC
Cav1	NM_007616	GGCAAATACGTAGACTCCGAG	GACCAGGTCAATCTCCTTGG
Ktn	NM_008477	TGAAAGATCGGATTGGAACA	TTGCTTTCCAGTTGGTTTGT
Selenbp2	NM_009150	TGGATGACCGCTTCCTTTAC	CTCTTGGTCCTCCAGCACTT
Arr3	NM_133205	CAGCTTCTCCCAGACCTTTG	AGGACACCACCAAGTTGACTC
Peli3	NM_172835	GCATCCTGTCTTGTCCTGGT	ATGTGGTGCATGACATCTGG
ABCa8a	NM_153145	GGGACACAATGGAGCTGGTA	CCCCTTTATGGCTGCAAATA
Gstt1	NM_008185	TGCTCTACCTGGCACACAAG	CTGCCAGTGTTTCAGGAGGT
Ccl3	NM_011337	ATGAAGGTCTCCACCACTGC	CAAAGGCTGCTGGTTTCAA
Vegfa	NM_001025250	GGAGAGCAGAAGTCCCATGA	ACTCCAGGGCTTCATCGTTA
Gpi1	NM_008155	GCCAAAGTGAAAGAGTTTGGA	ATGGAAAGTCCAATGGCTGA
Igfbp3	NM_008343	CTCACTGCCCTCACTCTGCT	AACTTTGTAGCGCTGGCTGT
Nes	NM_016701	AGCAGGAGAAGCAGGGTCTA	CTGGGAACTTCTTCCAGGTG
Osmr	NM_011019	CGACATCAATGGCTCAGAGA	CAATGATGCTGAGCAAGACG
H2T23	NM_010398	ACCAACAGAGGGCATACCTG	GGTCTCCACAAGCTCCATGT
Cdkn1a	NM_007669	CGGTGGAACTTTGACTTCGT	TCTGCGCTTGGAGTGATAGA
Muc2	XM_133960	GGCATTGTGTGCCAACCAAAG	CCTTGGGCACACAGGAATAAACTG
Tnfrsf12	NM_013749	CAGATCCTCGTGTTGGGATT	GCAGAAGTCGCTGTGTGGT
Ccl4	NM_0113652	CTCTGACCCTCCCACTTCCT	CTCACTGGGGTTAGCACAGA
β-Actin	NM_007393	GATGACCCAGATCATGTTTGA	GGAGAGCATAGCCCTCGTAG

Multiplex Enzyme-Linked Immunosorbent Assay

The retinas were isolated from normal P12 and from hyperoxia-exposed P12- and P13 mice. P13 was chosen because it is likely to be the time point at which changes in protein synthesis resulting from transcriptional changes at P12.5 are evident. The retinas were individually immersed in 500 μL tissue protein extraction reagent (T-PER; Pierce, Rockford, IL) containing a proprietary detergent in 25 mM bicine and 150 mM sodium chloride (pH 7.6), supplemented with protease inhibitor cocktail (Halt; Pierce). The mixture was homogenized (Polytron; Kinematica AG) and clarified by centrifuging at 10,000g for 5 minutes. The clarified retinal lysates were then assayed by ELISA. The total protein was determined with a commercial assay (Bio-Rad protein assay reagent kit; Bio-Rad Laboratories, Inc, Hercules, CA). 

To measure the concentrations of KC (CCL8), MCP-1 (CCL2), MCP-5 (CCL12), MIP-1α (CCL3), MIP-2 (CXCL2), RANTES (CCL5), TARC (CCL17), EOTAXIN (CCL11), and SDF-1β (CXCL12), a mouse chemokine array (Searchlight; Pierce Biotechnology) was used, and for MIP-1β, an ELISA kit (R&D Systems, Minneapolis, MN) was used, both according to the manufacturers' instructions. The signals of the chemokine arrays were determined by CCD cameras (model LAS3000; Fujifilm, Tokyo, Japan) by chemiluminescence. The sensitivities of the assays for KC, MCP-1, MCP-5, MIP-1α, MIP-2, RANTES, TARC, EOTAXIN, SDF-1β, and MIP-1β were 6.0, 3.0, 1.6, 3.0, 6.0, 3.0, 6.0, 0.8, and 37.5 pg/mL, respectively. 

Results

Gene Expression Profile by Microarray Analysis

To obtain a differential gene expression profile of the retina to hypoxia, we determined the gene expression profiles in murine OIR to that of retinas obtained from mice raised under normoxic conditions. We extracted genes significantly altered by hyperoxia (hyperoxic P12 retina compared to normoxic P12 retina) and hypoxia (hypoxic P12.5 retinas compared to hyperoxic P12 retinas or to normoxic P12 retinas). We chose P12.5 pups to examine gene expression in the hypoxic retina because earlier studies demonstrated that the degrees of central avascular area and central vasoconstriction were at the maximum between P12 and P13. ¹⁴ Scatterplot analysis comparing gene expression levels in two different mice in each group showed an approximately linear distribution of expression levels along the 45° line (Fig. 1), indicating a high degree of reproducibility in the methods, measurements, and analyses. 

Figure 1.

 

Comparison of gene expression levels in two different mice in each group. The level of expression at each point in the scatterplot is determined by comparing the expression level at the x coordinate (one retina) and the expression level of the gene at the y coordinate (another retina). Analysis of expression levels in two independent experiments shows a tight linear distribution along the diagonal. Intragroup comparisons in P12 control (A), P12 hyperoxia (B), and P12.5 hypoxia (C).

A subsequent volcano plot demonstrated that the retinal hypoxia was associated with alterations of specific mRNAs (Fig. 2). Several known hypoxia-induced genes in the retina, including Vegfa, Hif1a, Cdkn1a, Igfbp3, Ccl3, Mt1, and Adm, were differentially expressed, suggesting that our microarray experiments were reliable (Table 2). 

Figure 2.

 

Analysis of gene expression comparing hyperoxic P12 to hypoxic P12.5 retinas. The vertical axis value is the negative log of the P-values from a Welch's t-test, and the horizontal axis value is the log₂ change between the two conditions. (•) The genes significantly regulated by comparison between hyperoxic P12 and hypoxic P12.5 retinas.

Table 2.

 

Genes Significantly Altered in Expression between Hyperoxic P12 Retina and Hypoxic P12.5 Retina

Description	Gene Symbol	RefSeq	Change Ratio	P
Inflammation
Chemokine (C-C motif) ligand 4	Ccl4	NM_013652	4.248	0.0002
Histocompatibility 2, class II antigen A, beta 1	H2-Ab1	NM_207105	3.139	0.0035
Histocompatibility 2, Q region locus 6	H2-Q6	NM_207648.1	2.431	0.0004
Histocompatibility 2, T region locus 23	H2-T23	NM_010398	2.356	0.0011
Histocompatibility 2, Q region locus 8	H2-Q8	NM_023124	2.151	0.0010
Histocompatibility 2, Q region locus 2	H2-D1	NM_010392	1.955	0.0001
Histocompatibility 2, K1, Kregion	H2-K1	NM_001001892.2	1.820	0.0031
Chemokine (C-C motif) ligand 3	Ccl3	NM_011337	1.714	0.0061
Histocompatibility 2, Q region locus 5	H2-Q5	NM_010393	1.638	0.0002
Linker for activation of T cells family, member 2 isoform a	Wbscr5	NM_022964	1.550	0.0007
Development
Tumor necrosis factor receptor superfamily, member 12a	Tnfrsf12a	NM_013749	3.540	0.0050
Annexin A2	Anxa2	NM_007585	2.726	0.0063
p8 protein	Nupr1	NM_019738	2.470	0.0188
Tubulin, beta 6	Tubb6	NM_026473	2.395	0.0055
Transgelin 2	Tagln2	NM_178598	2.116	0.0004
Epithelial membrane protein 1	Emp1	NM_010128	2.098	0.0098
EGL nine homolog 1	Egln1	NM_053207	2.073	0.0146
Nestin	Nes	NM_016701	2.046	0.0006
Insulin-like growth factor binding protein 3	Igfbp3	NM_008343	2.015	0.0063
Adrenomedullin	Adm	NM_009627	1.992	0.0044
RIKEN cDNA 1600014C23 gene	1600014C23Rik	XM_128667	1.962	0.0034
Spectrin beta 1	Spnb1	NM_013675	1.899	0.0025
Galectin 3	Lgals3	NM_010705	1.889	0.0064
Vascular endothelial growth factor A isoform 1	Vegfa	NM_009505	1.873	0.0200
Procollagen, type II, alpha 1	Col2a1	NM_031163	1.685	0.0036
Plexin D1	Plxnd1	XM_149784	1.598	0.0002
Ephrin A1	Efna1	NM_010107	1.577	0.0004
Endomucin	Emcn	NM_016885	1.509	0.0015
Neuronatin isoform alpha	Nnat	AK077465	1.361	0.0036
Calsenilin, presenilin-binding protein, EF hand transcription factor	Csen	NM_019789	−1.300	0.0003
Antiapoptosis
Bcl2-associated athanogene 3	Bag3	NM_013863	2.829	0.0003
Similar to Mucin-5AC	Muc5ac	XM_133960	2.794	0.0001
Cyclin-dependent kinase inhibitor 1A (P21)	Cdkn1a	NM_007669	2.250	0.0014
Unc-5 homolog B	Unc5b	NM_029770	1.872	0.0061
Glycolysis
Hexokinase 2	Hk2	NM_013820	2.273	0.0062
Phosphofructokinase, platelet	Pfkp	NM_019703	2.223	0.0202
Glucose phosphate isomerase 1	Gpi1	NM_008155	1.928	0.0001
Response to hypoxia
Hypoxia inducible factor 1, alpha subunit	Hif1a	NM_010431.1	2.219	0.0048
Vision
Retinitis pigmentosa 1 homolog (human)-like 1 isoform 1	Rp1hl1	XM_484387	−1.468	0.0003
Guanylate cyclase activator 1B	Guca1b	NM_146079	−2.020	0.0098
Arrestin 3, retinal	Arr3	NM_133205	−2.494	0.0006
Metabolism
Carboxyl terminal LIM domain protein 1	Pdlim1	NM_016861	3.059	0.0002
6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 3 isoform 1	Pfkfb3	NM_133232	1.922	0.0029
Ankyrin repeat and SOCS box-containing 16	Asb16	AK038411	1.905	0.0022
Glutaminase 2 (liver, mitochondrial)	Gls2	XM_125928	1.889	0.0041
Methylenetetrahydrofolate dehydrogenase (NADP+ dependent) 1–like	Fthfsdc1	NM_172308	1.888	0.0023
Plasminogen activator, tissue	Plat	NM_008872	1.674	0.0017
Serine hydroxymethyl transferase 2 (mitochondrial)	Shmt2	NM_028230	1.610	0.0006
PHD zinc finger transcription factor	Phf12	NM_174852	1.263	0.0003
Acyl-CoA thioesterase 7	Bach	NM_133348	−1.227	0.0001
Phospholipase A2, group IVB isoform 1	Pla2g4b	XM_355378	−1.279	0.0001
Hypothetical protein LOC194237	BC057371	NM_177572	−1.305	0.0007
Calcium/calmodulin-dependent protein kinase kinase 1, alpha	Camkk1	NM_018883	−1.321	0.0052
Mannose phosphate isomerase 1 isoform 1	Mpi1	XM_134931	−1.323	0.0037
RIKEN cDNA 9630033F20 gene	9630033F20Rik	NM_177003	−1.381	0.0006
Cofactor required for Sp1 transcriptional activation, subunit 2	Crsp2	NM_012005	−1.429	0.0010
Sushi domain containing 3	Susd3	NM_025491	−1.43	0.0003
Creatine kinase, mitochondrial 1, ubiquitous	Ckmt1	NM_009897	−1.475	0.0001
Thioesterase, adipose associated	Thea	NM_025590	−1.481	0.0002
Mitochondrial ribosomal protein L36	Mrpl36	NM_053163	−1.531	0.0006
Glutathione S-transferase, theta 1	Gstt1	NM_008185	−1.789	0.0087
Glutathione S-transferase, theta 3	Gstt3	NM_133994	−1.802	0.0008
Signaling
Synaptopodin	Synpo	NM_177340.2	2.475	0.0029
Metallothionein 1	Mt1	NM_013602	2.383	0.0002
Sterile alpha motif domain containing 7	Samd7	XM_130828	2.194	0.0019
Oncostatin M receptor	Osmr	NM_011019	2.161	0.0001
Spermatogenesis associated 13 isoform 1	Spata13	XM_147847	1.772	0.0072
Stanniocalcin 2	Stc2	NM_011491	1.747	0.0025
Nuclear factor of kappa light chain gene enhancer in B-cells inhibitor, beta	Nfkbib	NM_010908	1.585	0.006
Rhomboid family 1	Rhbdf1	NM_010117	1.387	0.0008
Similar to connector enhancer of kinase suppressor of Ras 1	Cnksr1	XM_110525	−1.629	0.0031
Opioid receptor, mu 1	Oprm1	NM_011013	−1.701	0.0012
Cellular component
Filamin C, gamma isoform 1	1110055E19Rik	XM_284175	2.532	0.0002
H1 histone family, member O, oocyte-specific	H1foo	NM_138311	1.626	0.0056
Dystrobrevin alpha isoform 2	Dtna	NM_010087	1.609	0.0006
Vacuolar protein sorting 45	Vps45	NM_013841	−1.282	0.0005
RAN guanine nucleotide release factor	Rangnrf	NM_021329	−1.608	0.0013
Transport
Sideroflexin 5	Sfxn5	NM_178639	−1.289	0.0001
ATP-binding cassette, sub-family B, member 9	Abcb9	NM_019875	−1.294	0.0001
ATP-binding cassette, sub-family A (ABC1), member 8a	Abca8a	NM_153145	−1.852	0.0024
Cell cycle
Cdk5 and Abl enzyme substrate 2	Cables2	NM_145851	−1.337	0.0008
Meiosis-specific nuclear structural protein 1	Mns1	NM_008613	−1.493	0.0019
Cell adhesion
Melanoma cell adhesion molecule	Mcam	NM_023061	2.017	0.0001
Cell part
RIKEN cDNA 9230117N10	9230117N10Rik	NM_133775	−1.969	0.0053
Unknown
Nitric oxide synthase 3 antisense	Nos3as	NM_001002897	1.871	0.0012
2010204K13Rik	0610010M13Rik	AK008437	1.783	0.0022
Hypothetical protein LOC320706	9830001H06Rik	XM_283804	1.758	0.0063
Complement component 1, q subcomponent, receptor 1	C1qr1	AK077882	1.742	0.0011
Arrestin domain containing 3	Arrdc3	NM_178917	1.403	0.0022
Epidermal growth factor receptor pathway substrate 15, related	Eps15-rs	AK036728	−1.462	0.0065
Hypothetical protein LOC231296	BC031901	NM_153568	−1.681	0.0011
RAD52 motif 1	Rad52b	NM_025654	−1.880	0.0021
Pellino 3	6030441F14Rik	NM_172835	−2.020	0.0086
Hypothetical protein LOC67304	3110070M22Rik	NM_026084	−2.075	0.0091
EST	EST	NM_001039244.2	−2.532	0.0176

 Positive values indicate upregulation, and negative values indicate downregulation in gene expression in hypoxic retina compared to hyperoxic retina. P-values were detected by Welch's t-tests.

Changes in Gene Expression between Hyperoxic and Normoxic Retinas

Eighty-three gene transcripts were significantly altered at the level of gene expression in hyperoxic P12 retinas compared with normoxic P12 retinas (Table 3). In hyperoxic P12 retinas, only eight genes (Selenbp1, Selenbp2, 1700060, C20, Rik, Mrpl15, Cks-1, and Ccl3) were upregulated. In contrast, 81 genes were found to be downregulated. These 75 genes were broadly classified into six categories: cellular components (Emp1, Dmn, Hist1h2af, E030006K04Rik, D11Ertd686e, Tubb6, Hist1h2bk, Slc40a1, Ktn1, Hist1h2bc, Hist1h2bp, Pecam1, Hist1h2bl, Hist1h2bj, C1qr1, Hist1h2bm, Ramp2, Hist1h2bn, Col4a1), development (Gja4, Vwf, Mglap, Cav1, Emcn, Igfbp7, Robo4, Gas7), metabolism (Pdlim1, Pfkfb3, Asb16, Gls2, Fthfsdc1, Plat, Shmt2, Phf12, Bach, Pla2g4b, BC057371, Camkk1, Mpi1, 9630033F20Rik, Crsp2, Susd3, Ckmt1, Thea, Mrpl36, Gstt1, Gstt3), transport (Zfp295, Pltp, Slco1c1, Hspg2, LOC381928, Slco2b1, Aqp1), cell adhesion (MGC68323, Vtn, Cldn5), and vision (Gnat1, Arr3; Table 3; Fig. 3). 

Table 3.

 

Genes Significantly Altered in Expression between Hyperoxic P12 Retina and Normoxic P12 Retina

Description	Gene Symbol	RefSeq	Change Ratio	P
Cellular component
Epithelial membrane protein 1	Emp1	NM_010128	−1.312	0.0059
Desmuslin (Dmn) transcript variant 1	Dmn	NM_207663	−1.538	0.0435
Histone 1 H2af	Hist1h2af	NM_175661	−1.715	0.0268
RIKEN cDNA E030006K04 gene	E030006K04Rik	NM_139206	−1.736	0.0289
DNA segment Chr 11 ERATO Doi 686 expressed	D11Ertd686e	XM_110968	−1.742	0.0023
RIKEN cDNA 2310057H16 gene	Tubb6	NM_026473	−1.848	0.0185
Histone 1 H2bk	Hist1h2bk	NM_175665	−1.912	0.0072
Histone 1 H2ag	Hist1h2ag	NM_178186	−1.938	0.0302
Solute carrier family 40 (iron-regulated transporter) member 1	Slc40a1	NM_016917	−1.938	0.0457
Kinectin 1 (Ktn1) mRNA.	Ktn1	NM_008477	−2.004	0.0486
Histone 1 H2bc	Hist1h2bc	NM_023422	−2.070	0.0180
Histone 1 H2bp	Hist1h2bp	NM_178202	−2.188	0.0315
Platelet/endothelial cell adhesion molecule 1	Pecam1	NM_008816	−2.227	0.0258
Histone 1 H2bl	Hist1h2bl	NM_178199	−2.358	0.0369
Histone 1 H2bj	Hist1h2bj	NM_178198	−2.375	0.0249
Complement component 1 q subcomponent receptor 1	Clqr1	NM_010740	−2.387	0.0040
Histone 1 H2bm	Hist1h2bm	NM_178200	−2.404	0.0445
Receptor (calcitonin) activity modifying protein 2	Ramp2	NM_019444	−2.457	0.0174
Histone 1 H2bn	Hist1h2bn	NM_178201	−2.551	0.0158
Procollagen type IV alpha 1	Col4a1	NM_009931	−2.618	0.0447
Development
Gap junction membrane channel protein alpha 4	Gja4	NM_008120	−1.681	0.0181
Von Willebrand factor homolog	Vwf	NM_011708	−2.053	0.0445
Matrix gamma-carboxyglutamate (gla) protein	Mglap	NM_008597	−2.128	0.0360
Caveolin	Cav1	NM_007616	−2.128	0.0376
Endomucin	Emcn	NM_016885	−2.336	0.0066
Insulin-like growth factor binding protein 7	Igfbp7	NM_008048	−2.740	0.0334
Roundabout homolog 4	Robo4	NM_028783	−2.809	0.0027
Growth arrest specific 7	Gas7	NM_008088	−3.597	0.0039
Metabolism
Mitochondrial ribosomal protein L15	Mrpl15	AK011775	1.494	0.0302
Low density lipoprotein-related protein 1B (deleted in tumors)	LRP-DIT	AK080989	−1.248	0.0046
Minichromosome maintenance deficient 6	Mcm6	NM_008567	−1.541	0.0107
Platelet derived growth factor B polypeptide	Pdgfb	NM_011057	−1.570	0.0166
Solute carrier family 17 (sodium phosphate) member 1	Slc17a1	NM_009198	−1.623	0.0103
E26 avian leukemia oncogene 2 3 domain	Ets2	NM_011809	−1.745	0.0119
EGF-like domain 7	Egfl7	NM_198724	−1.825	0.0010
Epoxide hydrolase 2 cytoplasmic	Ephx2	NM_007940	−1.890	0.0312
EGF latrophilin seven transmembrane domain containing 1	Eltd1	NM_133222	−1.984	0.0025
Zinc finger protein 306	Zfp307	NM_023685	−2.012	0.0404
RIKEN cDNA 2310003L22 gene	2310003L22Rik	NM_027093	−2.237	0.0374
Protein tyrosine phosphatase receptor type B	Ptprb	NM_029928	−2.278	0.0047
Dual specificity phosphatase 11 (RNA/RNP complex 1–interacting)	Dusp11	NM_028099	−2.639	0.0300
Transport
Zinc finger protein 295	Zfp295	NM_175428	−1.255	0.0013
Phospholipid transfer protein	Pltp	NM_011125	−1.764	0.0120
Solute carrier organic anion transporter family member 1c1	Slco1c1	NM_021471	−1.869	0.0090
Similar to deleted in malignant brain tumors 1 isoform c precursor	LOC381928	XM_355949	−1.980	0.0092
Perlecan (heparan sulfate proteoglycan 2)	Hspg2	NM_008305	−2.033	0.0266
Solute carrier organic anion transporter family member 2b1	Slco2b1	NM_175316	−2.053	0.0131
Aquaporin 1	Aqp1	NM_007472	−2.174	0.0165
Stress response
Selenium binding protein 1	Selenbp1	NM_009150	2.273	0.0166
Selenium binding protein 2	Selenbp2	NM_019414	2.243	0.0102
Chemokine-like factor super family 3	Cklfsf3	NM_024217	−1.524	0.0326
Heat shock 27kDa protein 8	Hspb8	NM_030704	−1.592	0.0389
Cell adhesion
Similar to glyceraldehyde-3-phosphate dehydrogenase	MGC68323	NM_199472	−1.961	0.0482
Vitronectin	Vtn	NM_011707	−2.146	0.0165
Claudin 5	Cldn5	NM_013805	−3.165	0.0154
Inflammation
Chemokine (C-C motif) ligand 3	Ccl3	NM_011337	1.290	0.0094
Chemokine (C-X-C motif) ligand 12	Cxcl12	NM_013655	−1.661	0.0174
Transforming growth factor beta receptor II	Tgfbr2	NM_009371	−1.701	0.0022
Vision
Arrestin 3 retinal	Arr3	NM_133205	−2.336	0.0042
Guanine nucleotide binding protein alpha transducing 1	Gnat1	NM_008140	−3.175	0.0140
Response to oxidative stress
RIKEN cDNA 2310016C16 gene	2310016C16Rik	NM_027127	−1.916	0.0149
Cell part
RIKEN cDNA 1700093K21 gene	1700093K21Rik	NM_026105	1.908	0.0244
Unknown
RIKEN cDNA 1700060C20 gene	1700060C20Rik	XM_149230	1.569	0.0074
Similar to Cyclin-dependent kinases regulatory subunit 1	Cks-1	XM_123604	1.439	0.0003
RIKEN cDNA 4930546H06 gene	4930546H06Rik	XM_283398	1.352	0.0002
Pleckstrin homology domain containing, family A member 6	Plekha6	NM_182930	−1.307	0.0091
RNA exonuclease 4 homolog	Rexo4	XM_130184	−1.330	0.0249
Thrombospondin type I domain 1	Thsd1	NM_019576	−1.370	0.0097
RIKEN cDNA 4931415C17 gene	2810048G06Rik	AK088235	−1.453	0.0130
Kelch-like 6	Klhl6	NM_183390	−1.466	0.0093
Hypothetical protein A930010102	A930010102	NM_177799	−1.471	0.0100
RIKEN cDNA 1700008G05 gene	1700008G05Rik	XM_485205	−1.597	0.0068
RIKEN cDNA 0610041G09 gene	Acta2	NM_183274	−1.658	0.0235
Hypothetical protein E130012K09	E130012K09	NM_174999	−1.894	0.0411
FERM and PDZ domain containing 1	Frmpd1	XM_204152	−1.949	0.0346
RIKEN cDNA 1700018O18 gene	Mfsd2	XM_131683	−2.033	0.0324
Solute carrier family 1 (glial high affinity glutamate transporter) member 2	Mfsd2	XM_131683	−2.033	0.0324
Similar to hypothetical protein	LOC381546	XM_355512	−2.208	0.0186
Arrestin domain containing 3	Arrdc3	NM_178917	−2.268	0.0215
RIKEN cDNA 1700025K23 gene	1700025K23Rik	NM_183254	−2.299	0.0199
RIKEN cDNA 2310056K19 gene	2310056K19Rik	NM_080846	−2.488	0.0108
Hypothetical AAA ATPase superfamily containing protein	4933439B08Rik	AK014074	−2.494	0.0431
Solute carrier family 38, member 5	Slc38a5	NM_172479	−3.401	0.0122

 Positive values indicate upregulation, and negative values indicate downregulation in gene expression in hypoxic retina compared to normoxic retina. P-values were detected by Welch's t-tests.

Figure 3.

 

Distribution of GO terms in genes regulated by hyperoxic retinas. Ontology of genes downregulated by hyperoxia. Seventy-five genes were analyzed.

The main functional category, cellular components, contained many subtypes of histone, the chief protein components of chromatin. The second main category, development, contained genes associated with vascular and neural development, such as Robo4, Egfl7, and Gas7. 

Changes in Gene Expression between Hypoxic and Hyperoxic Retinas

The expression levels of 95 genes were significantly altered in hypoxic P12.5 retinas compared with hyperoxic P12 retinas (Table 2). Sixty-two genes were found to be upregulated, and 33 transcripts were downregulated. The maximum expression change was found for Ccl4 (4.2-fold increase), and the most downregulated gene was an EST (2.5-fold decrease). 

The upregulated genes in the hypoxic retina were broadly clustered into six functional categories: development (Tnfrsf12a, Anxa2, Nupr1, Tubb6, Tagln2, Emp1, Egln1, Nes, Igfbp3, Adm, 1600014C23Rik, Spnb1, Lgals3, Vegfa, Col2a1, Plxnd1, Efna1, Emcn, and Nnat), inflammation (Ccl4, H2-Ab1, H2-Q6, H2-T23, H2-Q8, H2-D1, H2-K1, Ccl3, H2-Q5, and Wbscr5), metabolism (Pdlim1, Pfkfb3, Asb16, Gls2, Fthfsdc1, Plat, Shmt2, and Phf12), signaling (Synpo, Mt1, Samd7, Osmr, Spata13, Stc2, Nfkbib, and Rhbdf1), antiapoptosis (Bag3, Muc5ac, Cdkn1a, and Unc5b), and glycolysis (Hk2, Pfkp, and Gpi1) (Table 2; Fig. 4). The most clustered category, development, contained genes associated with vasculogenesis and neurogenesis (e.g., Vegfa, Igfbp3, Tnfrsf12a, and Nestin). The genes categorized into inflammation were made up of several kinds of antigen processing and CC chemokines (Ccl3 and Ccl4). In addition, the upregulated genes were categorized into the antiapoptosis and glycolysis groups. 

Figure 4.

 

Distribution of GO terms in genes regulated by hypoxic retinas. Ontology of genes upregulated by hypoxia. Sixty-two genes were analyzed.

The 32 transcripts that were downregulated included genes related to vision (Rp1hl1, Guca1b, Arr3), transport (Sfxn5, Abcb9, Abca8a), and cell cycle (Cables2, Mns1). 

Changes in Gene Expression between Hypoxic and Normoxic Retinas

The expression levels of 94 genes were significantly altered in hypoxic P12.5 retinas compared to normoxic P12 retinas (Table 4). Fifty-three genes were found to be upregulated, and 41 transcripts were downregulated. The maximum expression change was found for Selenbp1 (4.6-fold increase), and the most downregulated gene was an EST (4.1-fold decrease). 

Table 4.

 

Genes Significantly Altered in Expression between Hypoxic P12.5 Retina and Normoxic P12 Retina

Description	Gene Symbol	RefSeq	Change Ratio	P
Inflammation
Histocompatibility 2, class II antigen A, beta 1	H2-Ab1	NM_207105	3.410	0.0004
Chemokine (C-C motif) ligand 4	Ccl4	NM_013652	2.605	0.0009
Histocompatibility 2 T region locus 23	H2-T23	NM_010398	2.063	0.0006
Histocompatibility 2 Q region locus 8	H2-Q8	NM_023124	2.043	0.0002
Transcription factor E3	Tcfe3	NM_172472	1.263	0.0002
Chemokine (C-X-C motif) ligand 12	Cxcl12	NM_013655	−1.548	0.0001
Cellular component
RIKEN cDNA 1190002C06 gene	1190002C06Rik	NM_028447	1.554	0.0007
Fibrosin 1	Fbs1	XM_284344	−1.247	0.0001
RIKEN cDNA E030006K04 gene	E030006K04Rik	NM_139206	−1.408	0.0002
Platelet/endothelial cell adhesion molecule 1	Pecam1	NM_008816	−1.672	0.0001
Histone 1 H2bh	Hist1h2bh	NM_178197	−2.024	0.0003
Solute carrier family 40 (iron-regulated transporter) member 1	Slc40a1	NM_016917	−2.128	0.0001
Development
Nuclear protein 1	Nupr1	NM_019738	3.380	0.0004
Integrin alpha 3	Itga3	NM_013565	2.653	0.0001
Tumor necrosis factor receptor superfamily member 12a	Tnfrsf12a	NM_013749	2.589	0.0147
Cofilin 1 non-muscle	Cfl1	NM_007687	2.550	0.0005
Connective tissue growth factor	Ctgf	NM_010217	2.313	0.0013
Outer dense fiber of sperm tails 2	Odf2	NM_013615	1.558	0.0001
Growth hormone releasing hormone	Ghrh	NM_010285	1.505	0.0005
Microtubule-associated protein tau	Mapt	NM_010838	1.474	0.0023
Exocyst complex component 7	Exoc7	NM_016857	1.384	0.0002
Rnolase 3 beta muscle	Eno3	NM_007933	1.353	0.0001
TCF3 (E2A) fusion partner	Tfpt	NM_023524	1.326	0.0001
Calcium channel voltage-dependent alpha 1F subunit	Cacna1f	NM_019582	−1.546	0.0001
Growth arrest specific 7	Gas7	NM_008088	−2.513	0.0141
Antiapoptosis
Cyclin-dependent kinase inhibitor 1A (P21)	Cdkn1a	NM_007669	3.299	0.0001
similar to Mucin-5AC	Muc5ac	XM_133960	2.438	0.0001
Bcl2-associated athanogene 3	Bag3	NM_013863	2.302	0.0001
Apoptosis
Serine/threonine kinase 17b (apoptosis-inducing)	Stk17b	NM_133810	−1.247	0.0005
Glycolysis
Hexokinase 2	Hk2	NM_013820	1.607	0.0030
Metabolism
Formyltetrahydrofolate synthetase domain containing 1	Fthfsdc1	NM_172308	2.773	0.0008
Spectrin beta 1	Spnb1	NM_013675	2.738	0.0004
Glutaminase 2 (liver, mitochondrial)	Gls2	XM_125928	2.401	0.0108
Ubiquitin specific protease 2	Usp2	NM_198091	2.339	0.0008
PDZ and LIM domain 1 (elfin)	Pdlim1	NM_016861	2.318	0.0009
Zinc finger, RAN-binding domain containing 1	Zranb1	NM_207302	1.999	0.0031
Dimethylarginine dimethylaminohydrolase 1	Ddah1	NM_026993	1.509	0.0002
Intestinal cell kinase	Ick	NM_019987	1.416	0.0013
ADP-ribosyltransferase 1	Parp1	NM_007415	−1.292	0.0003
Inner mitochondrial membrane peptidase 2-like	Immp2l	NM_053122	−1.364	0.0001
Uroporphyrinogen III synthase	Uros	NM_009479	−1.464	0.0001
Phosphatidylinositol-4-phosphate 5-kinase type II alpha	Pip5k2a	NM_008845	−1.499	0.0004
Ectonucleoside triphosphate diphosphohydrolase 3	Entpd3	NM_178676	−1.543	0.0001
Alkaline phosphatase 2 liver	Akp2	NM_007431	−1.577	0.0029
RAB guanine nucleotide exchange factor (GEF) 1	Rabgef1	NM_019983	−1.832	0.0007
Solute carrier family 17 (sodium phosphate) member 1	Slc17a1	NM_009198	−1.927	0.0033
Trophoblast glycoprotein	Tpbg	NM_011627	−2.024	0.0018
Dual specificity phosphatase 11	Dusp11	NM_028099	−2.227	0.0001
Solute carrier organic anion transporter family member 1c1	Slco1c1	NM_021471	−2.370	0.0071
Signaling
Metallothionein 1	Mt1	NM_013602	3.472	0.0001
Sterile alpha motif domain containing 7	Samd7	XM_130828	2.789	0.0040
Calmodulin-like 4	Calml4	NM_138304	2.268	0.0140
Leucine rich repeat transmembrane neuronal 1	Lrrtm1	NM_028880	1.660	0.0005
Smoothelin-like 2	Smtnl2	NM_177776	1.473	0.0001
Inhibitory adapter molecule DOK3	Dok3	NM_013739	−1.499	0.0001
Transport
ATP synthase H+ transporting mitochondrial F1F0 complex subunit e	Atp5k	AK011342	2.894	0.0104
Solute carrier family 27 (fatty acid transporter) member 3	Slc27a3	XM_130954	1.695	0.0010
Solute carrier family 25 member 19	Slc25a19	NM_026071	−1.698	0.0001
RIKEN cDNA 1810012H11 gene	1810012H11Rik	NM_028048	−2.096	0.0095
Aquaporin 1	Aqp1	NM_007472	−2.128	0.0074
ATP-binding cassette sub-family A (ABC1) member 8a	Abca8a	NM_153145	−2.212	0.0001
Stress response
Selenium binding protein 1	Selenbp1	NM_009150	4.607	0.0036
Selenium binding protein 2	Selenbp2	NM_019414	4.356	0.0012
Cold inducible RNA binding protein	Cirbp	NM_007705	2.135	0.0012
Response to oxidative stress
RIKEN cDNA 2310016C16 gene	2310016C16Rik	NM_027127	−2.155	0.0011
Vision
Guanylate cyclase activator 1B	Guca1b	NM_146079	−2.217	0.0009
Guantne nucleotide binding protein alpha transducing 1	Gnat1	NM_008140	−2.451	0.0262
Arrestin 3 retinal	Arr3	NM_133205	−4.082	0.0001
Cell adhesion
Cerebral endothelial cell adhesion molecule 1	Ceecam1	NM_207298	2.040	0.0022
RIKEN cDNA 1700060C20 gene	1700060C20Rik	XM_149230	1.995	0.0009
Ependymin related protein 1	Epdr1	NM_134065	−1.239	0.0008
Vitronectin	Vtn	NM_011707	−2.404	0.0050
Claudin 5	Cldn5	NM_013805	−2.899	0.0001
Cell cycle
Anillin actin binding protein	Anln	NM_028390	1.441	0.0003
Septin 3	Sep3	NM_011889	1.338	0.0013
Unknown
Arrestin domain containing 4	Arrdc4	NM_025549	2.666	0.0014
Similar to rab interacting lysosomal protein	Rilp	XM_207763	2.252	0.0013
RIKEN cDNA 9830001H06 gene	9830001H06Rik	XM_283804	2.223	0.0001
RIKEN cDNA 1700093K21 gene	1700093K21Rik	NM_026105	2.141	0.0001
EST	EST	XM_484892	2.071	0.0001
RIKEN cDNA 9130213B05 gene	9130213B05Rik	NM_145562	1.935	0.0003
TRIO and F-actin binding protein	Tara	NM_138579	1.914	0.0001
RIKEN cDNA 1110038B12 gene	1110038B12Rik	XM_359415	1.661	0.0008
Alport syndrome mental retardation midface hypoplasia and elliptocytosis chromosomal region gene 1 homolog	Ammecr1	NM_019496	1.336	0.0015
RIKEN cDNA 1100001A21 gene	1100001A21Rik	NM_025651	1.274	0.0003
Hypothetical protein C130086A10	C130086A10	NM_173746	−1.399	0.0001
HD domain containing 3	Hddc3	NM_026812	−1.558	0.0002
cDNA sequence BC024537	BC024537	NM_146237	−1.773	0.0003
Similar to hypothetical protein MGC45441	LOC381546	XM_355512	−2.088	0.0001
WD repeat domain 78	Wdr78	NM_146254	−2.294	0.0005
RIKEN cDNA 6030441F14 gene	6030441F14Rik	NM_172835	−2.331	0.0001
Hypothetical AAA ATPase superfamily containing protein	4933439B08Rik	AK014074	−2.695	0.0029
Solute carrier family 38, member 5	Slc38a5	NM_172479	−3.497	0.0015
EST	EST	NM_001039244.2	−4.115	0.0028

 Positive values indicate upregulation, and negative values indicate downregulation in gene expression in hypoxic retina compared with normoxic retina. P-values were detected by Welch's t-test. The genes in bold are also significantly altered in expression by comparison between hyperoxic P12 retina and normoxic P12 retina (Table 3). The genes in italic are also significantly altered in expression by comparison between hyperoxic P12 retina and hypoxic P12.5 retina (Table 2).

More than 40% of the gene transcripts were overlapped by genes differentially expressed in hyperoxic P12 and normoxic P12 retina (18 genes) or in hyperoxic P12 and hypoxic P12.5 retina (20 genes; Table 4). In addition, the genes significantly altered were broadly clustered into cellular component, development, inflammation, metabolism, transport, stress response, signaling, antiapoptosis, glycolysis, and vision. 

These functional categories were in parallel with the functional categories obtained by comparing hyperoxic P12 to normoxic P12 retinas (Table 2) or hyperoxic P12 to hypoxic P12.5 retinas (Table 4). 

Microarray Validation by Quantitative Real-Time RT-PCR

To validate the outcome of our microarray analyses, we performed qPCR assays. Relative quantifications of the mRNA expression by qPCR were calculated for 5 genes selected from the gene significantly altered by comparing normoxic and hyperoxic P12 retinas (Table 3), and for 15 genes by comparing hyperoxic P12 and hypoxic P12 retinas (Table 2; Fig. 5). We found excellent agreement between microarray and qPCR. In all samples, the expression changes were identical in direction of variation and were very similar in the extent of the alteration. These findings further demonstrated the high reliability of our array results. 

Figure 5.

 

Real-time quantitative RT-PCR for microarray validation. Array validation was performed by quantitative RT-PCR for 20 selected genes (5 and 15 genes selected from gene lists in Table 3 and Table 2, respectively). β-Actin was used as an internal standard. Relative change was the mean differences in gene expression between the groups. Dashed black line at change of onefold indicates no change.

Chemokine Protein Levels in Hypoxic Retinas by ELISA

Because genes related to inflammation were upregulated in hypoxic retinas, which may reflect postischemic inflammation, we next asked whether molecules encoded by these genes were upregulated at the protein level. We performed multiplex ELISA to determine the protein levels of KC (CCL8), MCP-1 (CCL2), MCP-5 (CCL12), MIP-1α (CCL3), MIP-2 (CXCL2), RANTES (CCL5), TARC (CCL17), EOTAXIN (CCL11), and SDF-1β (CXCL12) (Fig. 6). 

Figure 6.

 

Protein levels of the chemokines MIP-1α, MIP-1β, MCP-1, SDF-1β, MIP-2, MCP-5, KC, TARC, RANTES, and EOTAXIN in normoxic P12, hyperoxic P12, and hypoxic P13 retinas. Retinal lysates were prepared and were individually assayed by multiplex ELISA. The bars show the mean ± SEM results in three independent experiments per group. *Statistically significant differences (P < 0.01) compared with control subjects.

In the P13 retinas at 1 day after hypoxia, the KC, MCP-1, MIP-1α, MIP-1β, and SDF-1β protein levels were significantly increased (25.2, 36.2, 14.5, 4.02, and 1596.0 pg/mL, respectively) compared with control P12 and hyperoxic P12 retinas. In contrast, the MCP-5, MIP-2, RANTES, and EOTAXIN protein levels were not significantly increased in the hypoxic retinas. KC and MIP-1β protein in normoxic P12 and hyperoxic P12 retinas were below the level of detection. The results of ELISA were in agreement with the microarray data. Moreover, the assay revealed that retinal hypoxia markedly altered the protein levels of specific chemokines attracting granulocyte-macrophage lineages (e.g., KC, MCP-1, MIP-1α, MIP-1β, and SDF-1β), but it did not alter the other chemokines mainly related to eosinophils, mast cells, and lymphocytes. 

Discussion

Our analyses showed that retinal hyperoxia/hypoxia were associated with specific changes in the patterns of gene expression as determined by comprehensive DNA microarray analyses. These alterations may determine the hyperoxic growth retardation, postischemic inflammation, and subsequent neural and vascular remodeling, and pathologic neovascularization in retinas of murine OIR. 

In an earlier study, Sato et al. ¹⁴ examined only a single microarray hybridization for each condition. We performed multiple hybridizations, and we were able to extract the differentially expressed genes by applying Welch's t-test coupled with multidimensional false-discovery control (FDR2D) ²⁰ as the cutoff criteria. FDR2D can guard against false-positive results from transcripts whose variance is underestimated by chance, whereas their change ratios were small. Consequently, our method allowed us to identify differentially expressed genes reliably. 

As a result, we were able to extract well-known ischemia-regulated genes such as Vegfa and Hif1a (Table 2). Moreover, the results of our microarray analyses were in good agreement with independent qPCR assays for all the 20 genes selected (Fig. 5). These results demonstrated the high reliability of our microarray experimental procedures and subsequent data mining, and the extracted genes are likely to represent those that are regulated by hyperoxia/hypoxia in retinas of murine OIR. 

Hyperoxic Exposure to Neonatal Mice Resulted in Retinal Growth Retardation

Only eight genes were upregulated in the hyperoxic retinas compared with the normoxic control P12 retinas. Two of these, Selenbp1 and Selenbp2, may play a central role in protecting cells from oxidative damage. ²⁴ Therefore, these two genes were enhanced in response to the hyperoxia to protect the retina from oxidative injury. 

The 75 genes that were downregulated in response to hyperoxia were broadly clustered into cellular components, development, metabolism, transport, cell adhesion, vision, inflammation, and stress response. About one-half of the classified genes were categorized into cellular components and development (Table 3; Fig. 3). In the mouse, much of the retinal development takes place during the first 3 weeks after birth, and the expression of different genes should be essential for neural differentiation and guidance. ²⁵  

An interesting set of the genes showing decreased expression primarily in the hyperoxic retina were the histones: Hist1h2af, Hist1h2bk, Hist1h2bc, Hist1h2bp, Hist1h2bl, Hist1h2bj, Hist1h2bm, and Hist1h2bn. The decrease indicates that the chromosome biogenesis required for retinal development is largely suppressed. Moreover, several genes related to development such as cell structure (Emp1, Dmn, E030006K04Rik, D11Ertd686e, Tubb6, Slc40a1, Ktn1, Pecam1, C1qr1, Ramp2, and Col4a1), cell adhesion (MGC68323, Vtn, and Cldn5), and organogenesis (Gja4, Vwf, Mglap, Cav1, Emcn, Igfbp7, Robo4, and Gas7) were also downregulated, suggesting hyperoxia-exposed retinas from P7 to P12 are in a state of growth retardation. ²⁶  

One of the key physiological processes that occur during postnatal mouse retinal development is vasculogenesis. Retinal hyperoxia was accompanied by a downregulation of genes related to vascular development (Tubb6, Egfl7, Pecam1, Emcn, Pdgfb, and Robo4). This change may correspond to the marked regression of the superficial network of vessels that had already formed in the central retina and delayed the development of the deep plexus observed in this model. ²⁷  

We compared our findings on the genes altered by hyperoxic exposure to those in a similar study by Natoli et al., ²⁸ who reported that the alteration of gene expression was mainly associated with the stress response and with apoptotic cell death in the adult C57BL6/J mouse retinas exposed to hyperoxia. However, there was only one overlapping gene (Vwf). We postulate that this discrepancy reflects the differences in the age of the mice used or the use of different statistical tests. Our study revealed a new major category of genes suppressed by hyperoxia, growth and development, in addition to the categories found by Natoli et al. 

Gene Expression Induced by Retinal Hypoxia: Inflammation, Development, Antiapoptosis, and Glycolysis

As shown, exposing the developing retina to hyperoxic conditions significantly altered the expression of several genes related to development and growth. These genes were also present in the group comparison between hypoxic P12.5 retinas and normoxic P12 retinas (Table 4). This finding indicates that the hypoxic P12.5 retinas were still in a state of retinal growth suppression by the 5-day hyperoxic exposure compared with the normal retinas (Table 3). However, even in this state of growth suppression, we observed that similar genes were detected when hypoxic P12.5 retinas were compared to normoxic or hyperoxic P12 retinas (Table 2). This finding suggests that certain genes that should be altered in hypoxia are also altered, even though retinal growth is still suppressed by the preceding hyperoxia, in murine OIR (Table 4). 

Despite the considerable differences in the extent and duration of the hypoxic stimulus, the age, and the number of animals used in the different studies, several genes reported to be involved in retinal hypoxia, such as Vegfa, Igfbp3, ²⁹ Hif1a, ³ Cdkn1a, Adm, Mt1, ^30,31 and Ccl3, ⁷ were also found to be differentially expressed in the hypoxic retina in our study. These results further confirm that our microarray analyses were reliable. This reproducibility also indicates that the gene expression changes associated with retinal ischemia are related to specific pathways. 

The genes altered in the hypoxic retina were broadly clustered into six categories; development, inflammation, metabolism, signaling, antiapoptosis, and glycolysis (Fig. 4). Several inflammation-related genes (Ccl4, H2-Ab1, H2-Q6, H2-T23, H2-Q8, H2-D1, H2-K1, Ccl3, H2-Q5, and Wbscr5) were upregulated, indicating that hypoxic retinas are in an elevated state of immune responsiveness. ^32,33 The inflammation-related genes were subdivided into two main categories: CC chemokines and MHCs. We have demonstrated that Ccl3 is a CC chemokine involved in retinal neovascularization in murine OIR. ⁷ Our microarray analyses confirmed the upregulation of this gene. In addition, the change in the expression of Ccl4, another CC chemokine, was the highest among the 62 genes that were upregulated in the hypoxic retina (a 4.3-fold increase, Table 2). Because Ccl4 is not known to be a hypoxia-responsive gene in the retina, it may play more roles in hypoxic retinas than previously believed. 

Subsequent multiplex ELISA of the protein levels of several chemokines in the hypoxic retinas revealed that the protein levels of monocyte lineage–recruiting CC chemokines, such as MCP-1, MIP-1α, and MIP-1β, were specifically and significantly increased. In contrast, the protein levels of other chemokines that recruit other types of cell lineages, including eosinophils, lymphocytes, and mast cells, were not significantly higher (Fig. 6), indicating the involvement of Ccl3 and Ccl4, like Vegfa, ³⁴ in recruiting monocyte lineage cells in response to ischemia. ^35,36  

Concordant upregulation of genes related to MHCs may represent activation of resident and/or recruited macrophages/microglia. ³⁷ Such activated macrophages/microglia may play crucial roles in postischemic inflammation. Moreover, these cells may participate in eliminating apoptotic cells, facilitating the subsequent pathologic neovascularization ^7,34 and promoting vascular remodeling in this model. ^27,38  

Several genes related to cell development (Tnfrsf12a, Anxa2, Nupr1, Tubb6, Tagln2, Emp1, Egln1, Nes, Igfbp3, Adm, 1600014C23Rik, Spnb1, Lgals3, Vegfa, Col2a1, Plxnd1, Efna1, Emcn, and Nnat) were also upregulated, indicating that 12 hours after hyperoxic exposure, retinas regain the ability to undergo retinal vascular and neural development. Among these, the genes related to vasculogenesis or neovascularization (Vegfa, Hif1a, Anxa2, Tnfrsf12a, Adm, and Igfbp3) were also increased. ^{1,3,29,39–41} These results indicate that the changes in the gene expression in the P12.5 retina may contain what is required for the subsequent pathologic and physiological retinal neovascularization. ^1,29,42,43  

The observation that genes related to inflammation and angiogenesis were upregulated in hypoxic retina is consistent with the previous microarray study on gene expression in the same model. ¹⁴ However, our comprehensive microarray analysis revealed alterations of a distinct set of genes belonging to functions other than inflammation and angiogenesis. An interesting set of upregulated genes (Tnfrsf12a, Igfbp3, Bag3, Cdkn1a, Unc5b, Vegfa, and Hif1a), are related to antiapoptosis. This may reflect the cells' adaptive response to retinal ischemia. Bag3, a modulator of chaperone activity, has antiapoptotic activity and increases the neuroprotective function of Bcl-2 induced by various stimuli. ⁴⁴ Cdkn1a and Cdkn2b are known as cyclin-dependent kinase inhibitors (CDKIs) and regulators of cell cycle progression at G₁ and have antiapoptotic properties. ⁴⁵ Vegfa is also a prosurvival factor protecting retinal neurons against ischemic injury. In addition, Adm is associated with the HIF-mediated cell protection mechanism. These genes may play important roles in neuroprotection of the retina from ischemic injury. Activation of a diverse array of transcripts related to antiapoptosis is consistent with earlier ideas that fully protecting the retina may require several different antiapoptotic factors. ³¹  

Other upregulated genes in ischemic retinas were those related to glycolysis (Hk2, Pfkp, and Gpi1). This finding is in good agreement with those in earlier studies demonstrating increased glycolysis after experimental retinal ischemia ^46,47 and further supports the idea that the techniques used to extract and identify the genes in this study are reliable. 

We also found that the photoreceptor-specific genes Rp1hl1, Guca1b, and Arr3 were downregulated in the hypoxic retina, probably because developing photoreceptors are known to be sensitive to oxygen, with increased apoptosis occurring after an ischemic insult. ⁴⁸ In support of this notion, it has been reported that Arr3 acts as a scaffold by directly interacting with the neuronal JNK3 isoform and also by recruiting MKK4 and apoptosis signal-regulating kinase (ASK1) to form the complexes of ASK1-MKK4-JNK3. ^49,50 The alterations of retina-specific genes in response to ischemia also suggest the existence of retina-specific pathways in ischemic retinas of murine OIR. 

Conclusions

Our comprehensive microarray studies on the retinas of murine OIR provide molecular insights on hyperoxic growth retardation, postischemic inflammation, neural and vascular development, and subsequent pathologic neovascularization. Recently, several trials of anti-VEGF therapy have been performed on patients with neovascular diseases, and the results have revolutionized the treatment of intraocular neovascular diseases. ^51,52 However, finding additional or earlier targets that can be treated less invasively is certainly still a goal. Our comprehensive study clearly demonstrates that there are factors other than VEGF that could be additional molecular targets for such antiangiogenesis therapy. Our comprehensive microarray analyses in combination with multiplex ELISA support the idea that the macrophage recruitment mediated by CC chemokines may play a major role in the postischemic inflammation that contributes to the pathogenesis of ischemia-induced retinal neovascularization. Therefore, CC chemokines, especially Ccl4, which was most upregulated among the differentially expressed genes in ischemic retinas, could be therapeutic targets for inhibiting subsequent retinal neovascularization. Moreover, the comprehensive analysis also revealed that retinal ischemia accompanies enhancement of a diverse set of genes associated with antiapoptosis. This result raises the possibility that simultaneous molecular targeting therapies augmenting antiapoptotic pathways along with antiangiogenesis therapy enable effective maintenance of visual functions, while inhibiting pathologic retinal neovascularization. 

Footnotes

 Supported by grants from the Ministry of Education, Science, Sports and Culture, Japan (TI, SY).

Footnotes

 Disclosure: K. Ishikawa, None; S. Yoshida, None; K. Kadota, None; T. Nakamura, None; H. Niiro, None; S. Arakawa, None; A. Yoshida, None; K. Akashi, None; T. Ishibashi, None

The authors thank the staff of the Research Support Center (Graduate School of Medical Sciences, Kyushu University) for technical support and Naomi Higuchi (Kyushu University) for excellent technical help.