Subconjunctival Gene Delivery of the Transcription Factor GA-Binding Protein Delays Corneal Neovascularization in a Mouse Model

Home , GABPA

Gene Therapy (2009) 16, 973–981 & 2009 Macmillan Publishers Limited All rights reserved 0969-7128/09 $32.00 www.nature.com/gt ORIGINAL ARTICLE Subconjunctival gene delivery of the transcription factor GA-binding protein delays corneal neovascularization in a mouse model

KC Yoon1, JA Bae2, HJ Park1,SKIm1,HJOh1, XH Lin1, MY Kim2, JH Lee2, SE Lee2, KY Ahn2 and KK Kim2 1Department of Ophthalmology, Chonnam National University Medical School, Kwangju, South Korea and 2Medical Research Center for Gene Regulation and the Brain Korea 21 Project, Center for Biomedical Human Resources at Chonnam National University, Chonnam National University Medical School, Kwangju, South Korea

Corneal neovascularization can reduce visual acuity. GA- intervals. The third group served as an experimental control. binding protein (GABP) is a transcription factor that regulates In vivo delivery of human GABPa/b into mouse neovascular- the expression of target genes including vascular endothelial ized cornea reduced VEGF and Robo4 gene expression. growth factor (VEGF) and roundabout4 (Robo4), which Biomicroscopic examination showed that, at 1 week after participate in pathologic angiogenesis. We assessed one or two injections, GABPa/b-treated eyes had signiﬁcantly whether intraocular injection of the GABP gene affects the less neovascularized corneal area than did eyes treated with growth of new corneal blood vessels in a mouse ocular the empty vector. Histologic examination showed signiﬁ- neovascularization model. Transfection of human GABPa cantly less vascularized area and fewer blood vessels in the and GABPb gene (GABPa/b) into human conjunctival GABP-treated group at 1 week after injections. However, epithelial cells resulted in decreased VEGF and Robo4 these angiosuppressive effects were weakened at 2 weeks expression. Three groups of mice underwent chemical and after injections. Our results indicate that subconjunctival mechanical denudation of the corneal epithelium. Subse- GABP gene delivery delays corneal neovascularization for quently, two groups were administered subconjunctival up to 2 weeks in a mouse model of deliberate corneal injury. injection of lipoplexes carrying plasmid DNA encoding for Gene Therapy (2009) 16, 973–981; doi:10.1038/gt.2009.50; human GABPa/b or an empty plasmid DNA at 1-week published online 7 May 2009

Keywords: corneal neovascularization; GA-binding protein; VEGF; Robo4

Introduction vascularized corneas of rats and humans.6 Thus, reducing VEGF levels is the most common approach for Corneal neovascularization is characterized by the inhibiting corneal neovascularization in corneal gene ingrowth of new vessels from the limbus caused by the therapy studies.7,8 loss of the limbal stem cell barrier and frequently results The imbalance between angiogenic and antiangiogenic in vision impairment or blindness. Corneal neovascular- factors responsible for pathologic angiogenesis may also ization is usually associated with inflammatory or result from downregulation of inhibitors of blood vessel infectious disorders of the ocular surface and is often growth. Recognition of the potential benefits of control- accompanied by stromal edema, scarring, lipid deposits ling angiogenesis has led to a search for inhibitors of and keratitis; corneal neovascularization can also occur angiogenesis, and there has been considerable recent following keratoplasty.1 Angiogenesis is initiated when progress in identifying molecular signals that participate the balance between angiogenic and antiangiogenic in ocular neovascularization. Reestablishing the balance factors is shifted toward the angiogenic factors.2–4 between angiogenic and antiangiogenic factors by ocular Various angiogenic factors mediate corneal neovascular- gene transfer to block stimulators or increase the ization: vascular endothelial growth factor (VEGF), basic expression of endogenous inhibitors is a promising fibroblast growth factor, matrix metalloproteinase and therapeutic approach.7 Recent studies have shown that insulin-like growth factor-1. Among these, VEGF is angiogenesis inhibitors, such as endostatin, angiostatin, reported to be the primary mediator of neovasculariza- somatostatin, tissue inhibitor of metalloproteinase-3, tion in the eye5 and is elevated in inflamed and peptide derived from thrombospondin-1, prolactin and brain-specific angiogenesis inhibitor 1 (BAI1), can inhibit corneal neovascularization and maintain corneal Correspondence: Professor KK Kim, Department of Pharmacology, avascularity.9–16 Chonnam National University Medical School, Hak-Dong 5, Dong-Ku, GA-binding protein (GABP) is a ubiquitously ex- Kwangju 501-190, South Korea. E-mail: [email protected] pressed nuclear transcription factor that is involved in a Received 26 January 2008; revised 31 January 2009; accepted 10 broad range of cellular processes and that regulates February 2009; published online 7 May 2009 target gene expression in the cell cycle and cellular GABP gene delivery delays corneal neovascularization KC Yoon et al 974 metabolism.17 GABP is composed of two subunits: Results GABPa contains the erythroblast transformation-specific DNA-binding domain, whereas the substitutable sub- Decreased VEGF and Robo4 expression after GABPa/ unit, GABPb or GABPg, contains ankyrin repeats and the b gene transfer in conjunctival epithelial cells transcriptional activation domain. Both GABPb and GABPa/b was reported to act both as a transcriptional GABPg contain a nuclear localization signal; however, repressor of VEGF expression in SH-SY5Y and CV-1 GABPa does not, and its entry into the nucleus cells21 and as an activator with SP1 of the human Robo4 absolutely requires the concomitant expression of either promoter in primary endothelial cells.23 We chose human GABPb or GABPg. Therefore, GABPb or GABPg should Chang conjunctival epithelial cells to test the expression bind GABPa in the cytoplasm to assist GABPa transport of the GABPa/b gene in the conjunctiva and cornea by into the nucleus.18–20 GABP is now also recognized as a transiently transfecting these cells with increasing doses key transcriptional regulator of dynamically regulated, of lipoplexes carrying a plasmid encoding for the human lineage-restricted genes, especially in myeloid cells and GABPa and GABPb (GABPa/b) genes. at the neuromuscular junction, and acts as an integrator Expression of VEGF mRNA or protein in conjunctival of the cellular signaling pathway by regulating key epithelial cells was decreased after transfection with hormones and transmembrane receptors.17 GABPa/b gene (Figure 1a), indicating that GABPa/b also Recently, we found that the heteromeric complex of acts as a transcriptional repressor of VEGF in conjuncti- GABP (GABPa/b) acts as a transcriptional repressor and val epithelial cells. Unexpectedly, Robo4 mRNA, an thereby inhibits VEGF expression in SH-SY5Y neuro- endothelial-specific gene, was expressed in conjunctival blastoma and green monkey kidney epithelial (CV-1) epithelial cells and was markedly decreased after cells.21 This suggested testing the possibility of using GABPa/b transfection (Figure 1a). This suggests that GABP to control neovascularization of the ocular surface. GABP can act as a transcriptional repressor, rather than Ectopic expression of GABP has been reported to induce as an activator, of Robo4 expression in conjunctival quiescent cells to reenter the cell cycle, which suggests epithelial cells. that GABP is necessary and sufficient for reentry into the cell cycle.22 Moreover, SP1 transcription factor and GABP bind to the human roundabout4 (Robo4) promoter in Exogenous GABPa expression in mouse cornea after primary endothelial cells from human coronary artery subconjunctival GABPa/b gene injection and induce promoter activation.23 Robo4 is expressed To determine whether exogenous human GABP would exclusively in endothelial cells and has been implicated be expressed in the mouse cornea after local gene in endothelial cell migration and angiogenesis,24,25 and transfer in the corneal angiogenesis model, we used also in stabilizing the vasculature.26 lipoplexes that carry a plasmid DNA encoding for the Therefore, in the present study, we investigated human GABPa/b gene. At 3 and 5 days after subcon- whether in vivo subconjunctival gene delivery of GABP junctival injection of the human GABPa/b plasmid, transcription factor affects gene expression pattern of human GABPa mRNA expression was observed in the Robo4 and VEGF in the corneal neovascularization mouse cornea (Figure 1b), thus indicating that the induced by chemical and mechanical injuries. Both genes exogenous human gene was expressed. Expression was are important in pathologic angiogenesis and modulate observed until 7 days after subconjunctival gene injection experimental corneal neovascularization process in a (data not shown). VEGF and Robo4 mRNA, which are mouse model. expressed primarily in the endothelium,26 were ex-

Figure 1 Expression of vascular endothelial growth factor (VEGF) and roundabout4 (Robo4) in cultured conjunctival epithelial cells and in mouse cornea. (a) Expression of human VEGF mRNA (hVEGF, upper panel) and protein (lower panel, 26 and 23 kDa) and of human Robo4 mRNA (hRobo4, upper panel) in conjunctival epithelial cells was decreased after transfection with lipoplexes carrying human GABPa/b plasmid DNA. M, molecular size marker. (b) Expression of human GABPa mRNA (hGABPa) was observed at 3 (G3) and 5 (G5) days after single subconjunctival injection of lipoplexes carrying human GABPa/b plasmid DNA. There was no expression of hGABPa after injection of lipoplexes carrying empty plasmid DNA (V3, V5). Also, expression of exogenous human GABPa/b in mouse cornea markedly decreased mouse VEGF (mVEGF) and mouse Robo4 mRNA (mRobo4) expression at 3 days (G3) after injection. Excised corneas from ﬁve mice per group were pooled for RNA extraction. C3, cornea harvested 3 days after corneal denudation without gene injection.

Gene Therapy GABP gene delivery delays corneal neovascularization KC Yoon et al 975 pressed in the cornea that was denudated but not treated Quantitative biomicroscopic analysis shows less (Figure 1b). This suggests that corneal neovascularization neovascularized corneal area after GABP gene may occur after corneal injury. Also, as in the conjuncti- delivery val epithelial cells, the expression of VEGF and Robo4 We tested whether the injected GABPa/b gene could mRNA was markedly decreased in the mouse cornea at 3 inhibit corneal neovascularization in a mouse model by days after the subconjunctival injection of the human biomicroscopic examination. The lipoplexes carrying the a b GABP / plasmid, but gradually recovered by 5 days GABPa/b plasmid DNA or the empty plasmid DNA a after injection. Thus, the lipoplexes carrying the GABP / were applied subconjunctivally to each eye at 1 week b plasmid DNA or empty plasmid DNA were injected at after debridement. Mean values of neovascularized 1-week intervals to maintain transgene expression, just 16 corneal area in GABPa/b- or empty-vector-treated eyes as in our previous study. were measured at 3, 7 and 14 days after one or two injections and were compared with experimental control Decreased in vivo VEGF and Robo4 expression in the eyes at corresponding periods after corneal denudation. cornea after GABPa/b gene injection Three days after the one-injection treatment of Next, to assess whether subconjunctival injection of the GABPa/b or empty plasmid DNA, the GABPa/b-treated human GABPa/b gene actually induces changes in the in group showed less neovascular growth (Figure 3c) than vivo expression of VEGF and Robo4 in the cornea, we did the empty-vector-treated group (Figure 3d) or the examined the expression of VEGF and Robo4 by experimental control group (data not shown). The mean immunohistologic examination of corneas excised 3 days percentages of neovascularized corneal area to the total after injection of lipoplexes carrying the GABPa/b corneal area as calculated by morphometric analyses of plasmid DNA or the empty plasmid DNA. eNOS the biomicroscopic images in GABPa/b-treated, empty- (endothelial nitric oxide synthase), which was used as a vector-treated and experimental control eyes are de- positive marker for blood vessels, was decreased in picted in Figure 4. Three days after one-injection the GABP-injected group (Figure 2b) compared with treatment, GABPa/b-treated eyes had signiﬁcantly less the empty-vector-treated group (Figure 2a). As for the neovascular corneal area than did empty-vector-treated mRNA expression after subconjunctival injection of the (25.4%, Po0.05) or experimental control eyes (26.7%, human GABPa/b plasmid, the staining of VEGF (Figures Po0.05) (Figure 4a). 2d versus c) and Robo4 (Figures 2f versus e) in the mouse Seven days after one-injection treatment, the GABPa/ cornea decreased at 7 days after GABP delivery in b-treated group showed less neovascular growth (Figure comparison with empty vector delivery. This result 3e) than did the empty-vector-treated group (Figure 3f) indicates that GABPa/b, when injected subconjunctivally or the experimental control group (Figure 3a). Thus, 7 as lipoplexes, acts as a transcriptional repressor of VEGF days after one-injection treatment, GABPa/b-treated and Robo4 expression in this in vivo corneal injury model. eyes had signiﬁcantly less neovascular corneal area than

Figure 2 Expression of vascular endothelial growth factor (VEGF) and roundabout4 (Robo4) in the cornea. Representative light micrographs of stained corneas at 3 days after injection of lipoplexes carrying the empty plasmid DNA or the GABPa/b plasmid DNA are shown. The expression of VEGF and Robo4 was examined by immunohistochemistry (dark brown, indicated by arrows) in serially sectioned mouse corneas. A high-power view (a–f, inset) marked by the square (red) in (a–f) shows decreased endothelial nitric oxide synthase (eNOS) (b), VEGF (d) and Robo4 (f) expression in the cornea after GABPa/b gene delivery compared with empty vector delivery. eNOS was used as a marker for blood vessels. Scale bar ¼ 20 mmin(a–f, inset).

Gene Therapy GABP gene delivery delays corneal neovascularization KC Yoon et al 976

Figure 3 Biomicroscopic findings of mouse corneas after subconjunctival delivery of lipoplexes carrying the GABPa/b gene or empty plasmid DNA. Representative pictures of treated mouse corneas and experimental control corneas are shown. The corneas were examined at 2(a) and 3 (b) weeks after corneal denudation in experimental control eyes and at 3 days after the first injection of GABPa/b (c) or empty vector (d). The cornea at 7 days after the first injection (2 weeks after corneal denudation) of GABPa/b (e) or empty vector (f). The cornea at 1 week after two injections (3 weeks after corneal denudation) of GABPa/b (g) or empty vector (h). The GABPa/b-treated eyes showed significantly less neovascular growth than did the empty-vector-treated eyes or experimental control eyes.

did empty-vector-treated (34.6%, Po0.01) or experimen- after two-injection treatment, however, the GABPa/b- tal control eyes (33.5%, Po0.01). Fourteen days after one- treated eyes had less neovascular corneal area, injection treatment, GABPa/b-treated eyes had less but the difference was not significant compared neovascular corneal area but the difference was not with empty-vector-treated or experimental control eyes significant compared with empty-vector-treated or ex- (Figure 4b). perimental control eyes (Figure 4a). Overall, there were no practical or significant differ- To evaluate whether the frequency of gene delivery ences in the magnitude of the antiangiogenic effect of affected the antiangiogenic results of GABPa/b on GABPa/b on the neovascularized cornea between mice neovascularized cornea, we gave two injections, 1 week treated with one injection compared to two injections. apart, of lipoplexes that carry the GABPa/b plasmid or empty plasmid DNA. Three days after this two-injection treatment, GABPa/b-treated eyes had significantly less Histologic analysis shows less vascularized area and neovascular corneal area than did empty-vector-treated fewer vessels in mouse cornea treated by (25.5%, Po0.01) or experimental control eyes (24.4%, subconjunctival GABP gene delivery Po0.01) (Figure 4b). Seven days after the second gene Suppression of experimental corneal angiogenesis by injection, the GABPa/b-treated group showed less GABPa/b gene delivery was also studied by histologic neovascular growth (Figure 3g) than did the empty- examination of corneas excised from empty-vector- and vector-treated group (Figure 3h) or experimental control GABPa/b-injected mice. We used toluidine blue staining group (Figure 3b), and GABPa/b-treated eyes had as a molecular method to detect neovascularization.27,28 significantly less neovascular corneal area than did After injury, the experimental control group showed empty-vector-treated (20.3%, Po0.01) or experimental more neovascular growth (Figures 5b and d) than did the control eyes (20.3%, Po0.01) (Figure 4b). Fourteen days untreated absolute control group without injury (Figures

Gene Therapy GABP gene delivery delays corneal neovascularization KC Yoon et al 977 Discussion The specialized anatomy of the eye is such that highly vascularized and avascular tissues exist side by side.29 This requires tight regulation of the vascular growth process in avascular corneal tissue, which is maintained by the expression of soluble VEGF receptor-1 (sFlt-1).30 Under certain inflammatory conditions, however, avascular cornea may be subject to unwanted new blood vessel growth from the limbus caused by the loss of the limbal stem cell barrier. Corneal neovascularization is a serious vision-threa- tening complication associated with corneal infection, chemical injury or corneal trauma.2,7,8 Angiogenesis is initiated when the balance between angiogenic and antiangiogenic factors is shifted toward the angiogenic factors. Accordingly, neovascularization requires not only the upregulation of angiogenic factors, but also the downregulation of antiangiogenic factors.2,4 More- over, the imbalance accountable for pathologic angiogenesis may also be caused by downregulation of inhibitors of angiogenesis, and we recently reported that the BAI1 gene reduces corneal neovascularization in a rabbit model.16 In the present study, we examined the possibility that subconjunctival delivery of the GABP transcription factor to neovascularized cornea could affect the expression Figure 4 Comparison of the neovascular area among GABPa/b- treated eyes, empty-vector-treated eyes and experimental control patterns of VEGF and Robo4, which were reported to be eyes. After one (a) or two (b) injections, mean fractions of transcriptionally regulated by GABP and to be critical neovascularized areas were measured at 3, 7 and 14 days in regulatory proteins in pathologic angiogenesis.21,23,25,26 GABPa/b-treated eyes, empty-vector-treated eyes and untreated Using this approach, we observed for the first time that eyes. The asterisk indicates significant differences between the GABP could delay new vessel formation in a mouse corresponding GABPa/b-treated and experimental control eyes neovascularized cornea for up to 2 weeks after subcon- (*Po0.05, **Po0.01), or between the corresponding GABPa/b- and w ww a b empty-vector-treated groups ( Po0.05, Po0.01). Values are junctival injection of lipoplexes carrying the GABP / mean±s.e.m. plasmid DNA adjacent to the corneal neovascular area. We found that gene delivery of the GABPa/b plasmid DNA reduced VEGF and Robo4 expression in conjunctival 5a and c). The GABPa/b-treated eyes showed less epithelial cells and in mouse neovascularized cornea, neovascularized area and fewer vessels than did the and that GABPa/b-injected cornea had significantly less empty-vector-treated eyes at 7 days after one (Figures 5e neovascular area and fewer vessels than did the empty- versus f) or two (Figures 5i versus j) injections. The mean vector-treated or experimental control groups. Thus, we values of neovascular area and number of vessels from confirmed that the transcriptionally active GABPa/b three representative sections in the GABPa/b-treated heteromeric complex acts as a repressor of in vivo VEGF eye, empty-vector-treated eyes and experimental control and Robo4 expression, and thereby leads to a reduction of eyes are summarized in Table 1. corneal neovascularization in a mouse model. Seven days after one-injection treatment of GABPa/b Although neovascularization was significantly sup- or empty plasmid DNA, the mean neovascular area and pressed at 7 days after treatment, the suppressive effect the mean number of vessels in three corneal sections of decreased at 2 weeks after gene injection. This finding each mouse in GABP-treated eyes were significantly less indicates that the duration of repressed neovasculariza- than those in empty-vector-treated eyes (Po0.01 and tion by GABP delivery is somewhat short compared with o0.05, respectively) and experimental control eyes that of direct VEGF suppression in rat cornea by (Po0.01 and o0.05, respectively). Also, 1 week after adenoviral antisense RNA delivery, in which reduced two-injection treatment of GABPa/b or empty plasmid neovascularization is seen 14 days after injury,6 but DNA, the mean neovascular area and the mean number equivalent to that of direct VEGF knockdown in mouse of vessels in three corneal sections of each mouse were cornea by siRNA delivery, in which reduced neovascu- significantly less than those in empty-vector-treated eyes larization is seen 7 days after implantation of CpG- (Po0.05 and o0.05, respectively) (see Table 1 ). containing oligonucleotides.31 In addition, subconjuncti- However, 2 weeks after one- or two-injection treat- val injection of bevacizumab, an anti-VEGF antibody, ment, the mean neovascular area and number of vessels was reported to induce regression of experimental in GABP-treated eyes were not significantly different corneal neovascularization up to 2 weeks in animal than in empty-vector-treated eyes (Table 1). Thus, models,32,33 and partial regression of corneal neovascu- subconjunctival GABPa/b gene delivery significantly larization was reported in short-term studies in reduced the neovascular area and the number of vessels humans.34 These reports suggest that the vascular at 1 week, but not 2 weeks, after treatments with either regression induced in the cornea by blocking VEGF one or two injections. action alone is partial and transient.

Gene Therapy GABP gene delivery delays corneal neovascularization KC Yoon et al 978

Figure 5 Light micrographs of corneas in an experimental control eye after corneal denudation and the effects of GABPa/b gene delivery on corneal angiogenesis. Representative micrographs of toluidine blue-stained corneas, given lipoplexes carrying the empty plasmid DNA or the GABPa/b gene are shown. The untreated absolute control group at 2 (a) and 3 (c) weeks showed no neovascularization, whereas the experimental control group at 2 (b) and 3 (d) weeks showed neovascular formation. Vascularity was examined 1 week after 1 injection of GABPa/b (e) or empty vector (f), at 2 weeks after one injection of GABPa/b (g) or vector (h), at 1 week after two injections of GABPa/b (i)or vector (j), and at 2 weeks after two injections of GABPa/b (k) or vector (l). GABPa/b-treated corneas showed signiﬁcantly fewer vessels and smaller areas of neovascularization than did empty vector-treated corneas. Scale bar ¼ 20 mm.

GABP was reported to act as a transcriptional activator vascular disease, Slit2 inhibited angiogenesis and vas- with SP1 of human Robo4,23 which has been implicated in cular leaking, whereas deletion of the Robo4 gene endothelial cell migration, angiogenesis and vascular enhanced these pathologic processes.26 Those ﬁndings development in vertebrates.24,25 Also, a GABP-binding suggest that Robo4 is involved in stabilizing the motif is necessary for mediating Robo4 expression in the vasculature and that activating the Robo4 receptor may intact endothelium.35 By contrast, the Robo4 receptor is have therapeutic application in diseases characterized essential for Slit2-dependent inhibition of VEGF-165- by excessive pathologic angiogenesis and vascular induced pulmonary endothelial cell migration, tube leaking.26 formation, and permeability in vitro and for inhibiting In this study, however, we observed that forced VEGF-165-driven vascular leaking in vivo by blocking Src expression of GABPa/b in the mouse cornea and in family kinase activation. Moreover, in mouse models of cultured human conjunctival epithelial cells suppressed oxygen-induced retinal and laser-induced choroidal Robo4 mRNA expression, thus suggesting the presence

Gene Therapy GABP gene delivery delays corneal neovascularization KC Yoon et al 979 Table 1 The neovascular area of each vessel and the number of vessels in sectioned corneas of mice (n ¼ 3, mean±s.e.m.)

Groups Experimental control Empty vector treated GABP treated Interval (area/vessel number)

1 week after 1 injection 1.38±0.32/7.2±2.9 1.36±0.56/6.8±2.3 0.39±0.18**ww/3.6±1.9*w 2 weeks after 1 injection 1.69±0.38/11.9±2.2 1.60±0.56/12.1±3.8 1.10±0.52/8.8±2.9 1 week after 2 injections 1.79±0.47/11.5±4.0 0.76±0.10w/6.9±1.1w 2 weeks after 2 injections 1.98±0.32/13.8±2.6 1.71±0.28/10.1±2.5

Abbreviation: GABP, GA-binding protein. *Significantly different from the corresponding experimental control group (*Po0.05, **Po0.01). wSignificantly different from the corresponding empty-vector-treated group (wPo0.05, wwPo0.01). of regional variation in the effect of GABP on Robo4 virus promoter, was used for GABPa and GABPb expression in the eye, especially in the avascular cornea. expression. Two microliters of empty pFLAG vector The molecular mechanisms of Robo4 expression in (2 mg) or pFLAG-GABP plasmid (2 mg; 1 mg GABPa and neovascularized cornea are not entirely elucidated at 1 mg of GABPb plasmid DNA) was mixed with 1 mlof present. However, because VEGF and Robo4 are not FuGENE 6 (Roche, Indianapolis, IN, USA), to which 17 ml expressed in normal avascular cornea but are expressed of serum-free medium was added to make a total volume primarily in neovascularized corneal endothelium in the of 20 ml. mouse corneal injury model used in this study (Figure 2), Human Chang conjunctival epithelial cells (clone 1-5c-4) we hypothesize that suppression of Robo4 and VEGF by were obtained from the Korean Cell Line Bank and were GABP gene delivery could be beneficial for reducing cultured in six-well plates with RPMI 1640 containing corneal neovascularization. 10% fetal bovine serum and antibiotic/antimycotic In addition, we cannot rule out the possibility that the solution. Cells were then transfected with pFLAG-tagged GABP expression driven by the injected lipoplexes was GABPa and GABPb or empty pFLAG vector using simply not maintained at a sufficient level to reduce the FuGENE 6 according to the manufacturer’s protocol. angiogenic stimulus. Considering all our results together, Cell lysates were prepared, resolved by SDS-polyacryla- we hypothesize that after subconjunctival GABPa/b mide gel electrophoresis, transferred and blotted injection, overexpressed GABPa/b in the injected area with anti-VEGF (NeoMarkers, Fremont, CA, USA) as may induce conjunctival vascular endothelial cells to described.21 reenter the cell cycle,22 while at the same time suppress Total RNAs were prepared from cultured cells or or delay corneal angiogenesis by transcriptional repres- excised pooled corneas. Total RNA (400 ng) was reverse- sion of VEGF and Robo4 in the neovascularized cornea. transcribed, as described elsewhere,16 with random Thus, in concert with these potential underlying primers and MMLV by the use of RT-PCR kit (Invitrogen, mechanisms of GABP, we observed partial recovery of Carlsbad, CA, USA). PCR consisted of 30–35 cycles of the angiogenesis 2 weeks after GABP injection (Table 1). following conditions: denaturing for 50 s at 94 1C, Therefore, it seems that GABP gene delivery into the annealing for 50 s and extension for 1 min at 72 1C. The neovascularized cornea may result in a relatively short- amplification products were analyzed on agarose gels term suppression of angiogenesis, in which the balance and visualized. The annealing temperature was 60 1C for between angiogenic and antiangiogenic processes is hVEGF, b-actin, and GAPDH and 58 1C for hRobo4. PCR reestablished around 2 weeks after gene injection. This primers for hVEGF: S1, 50-GCAGACCAAAG-AAAGAT suggests that the application of GABP gene therapy in AGAGCAAG-30,andAS1,50-CGCCTCGGCTTGTCACAT-30; treating corneal neovascularization may require further hRobo4: S1, 50-CCAGTAGACCTG-TCTGCCTCCTTT-30, experimental and preclinical studies. and AS1, 50-GGAGCCATAAAAAGTGCTGGTGTC-30.

Mouse model of corneal neovascularization Materials and methods General anesthesia was induced with an intramuscular injection of 50 mg kgÀ1 ketamine hydrochloride and Animals À1 The study was performed with C57BL/6 mice weighing 10 mg kg xylazine. After instillation of topical propar- between 20 and 25 g. Prior approval of the experimental acaine, the entire corneal surface was contacted with 2 ml protocol was obtained from the Chonnam National of 0.15 mM NaOH for 5 s. The corneal surface was then University Medical School Research Institutional Animal rinsed with sterile saline, and the corneal and limbal Care and Use Committee. Animal maintenance and all epithelia were removed with a corneal knife in a rotary in vivo experiments were performed in accordance with motion parallel to the limbus. Levoﬂoxacin eyedrops institutional guidelines and the Association for Research (Cravit, Santen, Osaka, Japan) were instilled immediately in Vision and Ophthalmology Statement for the Use of after epithelial debridement. Animals in Ophthalmic and Vision Research. Experimental design Plasmids, cell culture and RT-PCR One week after epithelial debridement, lipoplexes carry- The pFLAG mammalian expression vector (Eastman ing the GABPa/b plasmid DNA or empty plasmid DNA Kodak, Rochester, NY, USA), which has a cytomegalo- were injected into the bulbar conjunctiva. For identiﬁca-

Gene Therapy GABP gene delivery delays corneal neovascularization KC Yoon et al 980 tion of gene expression, mice were treated by subcon- innermost vessel of the limbal arcade as the border. junctival injection (20 ml) of lipoplexes prepared as The area of corneal vasculature was outlined with the described above into the right eye. In one group computer mouse, the total neovascularization area was (n ¼ 15), the lipoplexes carried 1 mg of GABPa and 1 mg then normalized to the total corneal area, and the of GABPb, whereas in the other group (n ¼ 15), the percentage of the cornea covered by vessels was lipoplexes carried 2 mg of an empty plasmid DNA. For calculated by morphometric analysis. each group, five eyes were used for corneal RNA preparation after 3 days, five eyes were used for Analysis of corneal neovascularization by histologic immunohistochemistry after 3 days and five eyes were examination used for corneal RNA extraction after 5 days. For RNA Mice (n ¼ 6) were anesthetized with 50 mg kgÀ1 of preparation, excised corneas were pooled per each time ketamine hydrochloride and 10 mg kgÀ1 of xylazine on point. day 7, and the common carotid artery was exposed. This For biomicroscopic and histologic examinations, mice artery was then flushed through a catheter with were treated by subconjunctival injection (20 ml) of phosphate-buffered saline, followed by 4% paraformal- lipoplexes. In one group (n ¼ 12), the lipoplexes carried dehyde. The eyes were enucleated and immediately 1 mg of GABPa and 1 mg of GABPb, whereas in the other placed into the same fixative. After fixation, the corneas group (n ¼ 12), the lipoplexes carried 2 mg of an empty were excised and immersed in 4% paraformaldehyde plasmid DNA. For each group, six mice were examined 1 fixative at 4 1C overnight. The tissue blocks were week after injection, and the remaining six mice were washed, dehydrated, embedded in paraffin, cut in evaluated after 2 weeks. The right eyes of 12 mice were 6-mm sections and mounted. The tissue sections were not injected after epithelial debridement and served as deparaffinized, rehydrated, rinsed, and stained with experimental controls. Four eyes that received neither eNOS and VEGF (Santa Cruz Biotech, Santa Cruz, CA, treatment nor epithelial debridement served as untreated USA) or Robo4 antibody (Abcam, Cambridge, UK) and absolute controls. examined as described.16 In a separate examination, 1 week after debridement, To determine the degree of induced angiogenesis, we subconjunctival injections of lipoplexes carrying the stained sections with toluidine blue.27,28 Three represen- GABPa/b plasmid DNA (n ¼ 12) or empty vector tative sections obtained from each cornea of experimen- (n ¼ 12) were performed twice at 1-week intervals into tal control, empty-vector-treated or GABPa/b-treated the right eye, as described above. Also, experimental groups were analyzed with a microscope (Olympus, controls (n ¼ 12) and untreated absolute controls (n ¼ 4) Tokyo, Japan) as described.16 The images were captured were used. with a Spot digital camera (Media Cybernetics) and morphometric analyses were performed using Image- Examination of GABP, VEGF and Robo4 expression Pro Plus software. The whole field at a magnification of in gene-injected cornea Â 400 was examined for each cross section. The sum and The expression of exogenous human GABPa/b, mouse mean of the vascularized area of each vessel and the VEGF and mouse Robo4 in the mouse cornea was number of vessels per section were determined. confirmed via reverse transcription (RT)-PCR and im- Statistical analysis munohistochemical analyses. Primer sets were designed Results are expressed as mean±s.e.m. The statistical to recognize the FLAG-specific portions of the pFLAG- significance of differences among the experimental GABP gene (FLAG sense, 50-GACTACAAAGAC control and empty-vector- and GABPa/b-treated groups GATGACGAC-30, and GABPa antisense, 50-ATATCTTG was determined by the Kruskal–Wallis test with SPSS CAGACAAATTT-30). Specific PCR primers for mouse 15.0 software (SPSS Inc., Chicago, IL, USA). The VEGF and Robo4 were designed (mVEGF: sense, statistical significance of differences between the one- 50-CCGAGAAGAGAGGGGAGGAA-30, and antisense, and two-injection treatment groups was determined by 50-GTTCATGGTTTCGGAGGCCC-30; mRobo4: sense, using the Mann–Whitney U-test. 50-TCCAGGGACGGCCACAAGAT-30, and antisense, 50-GACCAGGGGTTTCCCGTCTT-30). Acknowledgements Analysis of corneal neovascularization by biomicroscopic examination We thank Hong-Jae Chae (Chonnam National University Hospital) for assisting with the statistical analysis. This All mice were photographed at a standard magnification work was supported by the Korea Science and Engineer- 1 day before injection and on days 3, 7 and 14 after ing Foundation through the Medical Research Center for injection. After anesthesia with 50 mg kgÀ1 of ketamine Gene Regulation (R13-2002-013-04001-0) at Chonnam hydrochloride and 10 mg kgÀ1 of xylazine, images of National University. corneal neovascularization were captured with a camera (F-801s; Nikon, Miyagi, Japan) attached to a slit-lamp biomicroscope (FS-3; Nikon, Kanagawa, Japan). The References images were analyzed by using Image-Pro Plus software (Media Cybernetics, Silver Spring, MD, USA) as de- 1 Kvanta A, Sarman S, Fagerholm P, Seregard S, Steen B. 16 scribed. The images were resolved at 624 Â 480 pixels Expression of matrix metalloproteinase-2 and vascular endo- and converted to a tagged information file format. thelial growth factor in inflammation-associated corneal Neovascularization in the entire cornea was quantitated neovascularization. Exp Eye Res 2000; 70: 419–428. in a blinded manner to minimize sampling bias. The total 2 Chang JH, Gabison EE, Kato T, Azar DT. Corneal neovascualri- surface area of the cornea was outlined, using the zation. Curr Opin Ophthalmol 2001; 12: 242–249.

Gene Therapy GABP gene delivery delays corneal neovascularization KC Yoon et al 981 3 Gabison E, Chang JH, Hernandez-Quintela E, Javier J, Lu PC, ETS-related transcription factor GABPg1. Mol Cell Biol 2002; 22: Ye H et al. Anti-angiogenic role of angiostatin during corneal 2147–2158. wound healing. Exp Eye Res 2004; 78: 579–589. 20 O’Leary DA, Koleski D, Kola I, Hertzog PJ, Ristevski S. 4 Folkman J. Angiogenesis in cancer, vascular, rheumatoid and Identification and expression analysis of alternative transcripts other disease. Nat Med 1995; 1: 27–31. of the mouse GA-binding protein (Gabp) subunits alpha and 5 Lai YK, Shen WY, Brankov M, Lai CM, Constable IJ, Rakoczy PE. beta1. Gene 2005; 344: 79–92. Potential long-term inhibition of ocular neovascularisation by 21 Jeong BC, Kim MY, Lee JH, Kee HJ, Kho DH, Han KE et al. recombinant adeno-associated virus-mediated secretion gene Brain-specific angiogenesis inhibitor 2 regulates VEGF through therapy. Gene Therapy 2002; 9: 804–813. GABP that acts as a transcriptional repressor. FEBS Lett 2006; 6 Lai CM, Spilsbury K, Brankov M, Zaknich T, Rakoczy PE. 580: 669–676. Inhibition of corneal neovascularization by recombinant 22 Yang ZF, Mott S, Rosmarin AG. The Ets transcription factor adenovirus mediated antisense VEGF RNA. Exp Eye Res 2002; GABP is required for cell-cycle progression. Nat Cell Biol 2007; 9: 75: 625–634. 339–346. 7 Campochiaro PA. Gene therapy for ocular neovascularization. 23 Okada Y, Yano K, Jin E, Funahashi N, Kitayama M, Doi T et al. A Curr Gene Ther 2007; 7: 25–33. three-kilobase fragment of the human Robo4 promoter directs 8 Klausner EA, Peer D, Chapman RL, Multack RF, Andurkar SV. cell type-specific expression in endothelium. Circ Res 2007; 100: Corneal gene therapy. J Control Release 2007; 124: 107–133. 1712–1722. 9Va´zquez F, Hastings G, Ortega MA, Lane TF, Oikemus S, 24 Bedell VM, Yeo SY, Park KW, Chung J, Seth P, Shivalingappa V Lombardo M et al. METH-1, a human ortholog of ADAMTS-1, et al. Roundabout4 is essential for angiogenesis in vivo. Proc Natl and METH-2 are members of a new family of proteins with Acad Sci USA 2005; 102: 6373–6378. angio-inhibitory activity. J Biol Chem 1999; 274: 23349–23357. 25 Suchting S, Heal P, Tahtis K, Stewart LM, Bicknell R. Soluble 10 Ambati BK, Joussen AM, Ambati J, Moromizato Y, Guha C, Robo4 receptor inhibits in vivo angiogenesis and endothelial cell Javaherian K et al. Angiostatin inhibits and regresses corneal migration. FASEB J 2005; 19: 121–123. neovascularization. Arch Ophthalmol 2002; 120: 1063–1068. 26 Jones CA, London NR, Chen H, Park KW, Sauvaget D, Stockton 11 Wu PC, Liu CC, Chen CH, Kou HK, Shen SC, Lu CY et al. RA et al. Robo4 stabilizes the vascular network by inhibiting Inhibition of experimental angiogenesis of cornea by somatos- pathologic angiogenesis and endothelial hyperpermeability. tatin. Graefes Arch Clin Exp Ophthalmol 2003; 241: 63–69. Nature Med 2008; 14: 448–453. 12 Ma DH, Zhang F, Shi W, Yao JY, Hsiao CH, Wu HC et al. 27 Kenyon BM, Voest EE, Chen CC, Flynn E, Folkman J, D’Amato Expression of tissue inhibitor of metalloproteinase-4 in normal RJ. A model of angiogenesis in the mouse cornea. Invest human corneal cells and experimental corneal neovasculariza- Ophthalmol Vis Sci 1996; 37: 1625–1632. tion. Ophthalmic Res 2003; 35: 199–207. 28 Demir T, Celiker UO, Ku¨ kner A, Mogulkoç R, Celebi S, Celiker 13 Shafiee A, Penn JS, Krutzsch HC, Inman JK, Roberts DD, Blake H. Effect of Octreotide on experimental corneal neovasculariza- DA. Inhibition of retinal angiogenesis by peptides derived from tion. Acta Ophthalmol Scand 1999; 77: 386–390. thrombospondin-1. Invest Ophthalmol Vis Sci 2000; 41: 2378–2388. 29 Knop E, Knop N. Anatomy and immunology of the ocular 14 Shao C, Sima J, Zhang SX, Jin J, Reinach P, Wang Z et al. surface. Chem Immunol Allergy 2007; 92: 36–49. Suppression of corneal neovascularization by PEDF release from 30 Ambati BK, Nozaki M, Singh N, Takeda A, Jani PD, Suthar T human amniotic membranes. Invest Ophthalmol Vis Sci 2004; 45: et al. Corneal avascularity is due to soluble VEGF receptor-1. 1758–1762. Nature 2006; 443: 993–997. 15 Duenãs Z, Torner L, Corbacho AM, Ochoa A, Gutie´rrez-Ospina 31 Kim B, Tang Q, Biswas PS, Xu J, Schiffelers RM, Xie FY et al. G, Lo´pez-Barrera F et al. Inhibition of rat corneal angiogenesis by Inhibition of ocular angiogenesis by siRNA targeting vascular 16-kDa prolactin and by endogenous prolactin-like molecules. endothelial growth factor pathway genes: therapeutic strategy Invest Ophthalmol Vis Sci 1999; 40: 2498–2505. for herpetic stromal keratitis. Am J Pathol 2004; 165: 2177–2185. 16 Yoon KC, Ahn KY, Lee JH, Chun BJ, Park SW, Seo MS et al. Lipid- 32 Hurmeric V, Mumcuoglu T, Erdurman C, Kurt B, Dagli O, mediated delivery of brain-specific angiogenesis inhibitor 1 gene Durukan AH. Effect of subconjunctival bevacizumab (Avastin) reduces corneal neovascularization in an in vivo rabbit model. on experimental corneal neovascularization in guinea pigs. Gene Therapy 2005; 12: 617–624. Cornea 2008; 27: 357–362. 17 Rosmarin AG, Resendes KK, Yang Z, Mc Millan JN, Fleming SL. 33 Kim TI, Kim SW, Kim S, Kim T, Kim EK. Inhibition of GA-binding protein transcription factor: a review of GABP as an experimental corneal neovascularization by using subconjuncti- integrator of intracellular signaling and protein–protein interac- val injection of bevacizumab (Avastin). Cornea 2008; 27: 349–352. tions. Blood Cells Mol Dis 2004; 32: 143–154. 34 You IC, Kang IS, Lee SH, Yoon KC. Therapeutic effect of 18 Batchelor AH, Piper DE, de la Brousse FC, Mcknight SL, subconjunctival injection of bevacizumab in the treatment Wolberger C. The structure of GABPalpha/beta: an ETS domain- of corneal noevascularization. Acta Ophthalmol 2009; 87 ankyrin repeat heterodimer bound to DNA. Science 1998; 279: (in press). 1037–1041. 35 Okada Y, Jin E, Nikolova-Krstevski V, Yano K, Liu J, Beeler D 19 Hauck L, Kaba RG, Lipp M, Dietz R, von Harsdorf R. Regulation et al. A GABP-binding element in the Robo4 promoter is necessary of E2F1-dependent gene transcription and apoptosis by the for endothelial expression in vivo. Blood 2008; 112: 2336–2339.

Gene Therapy