TAK1: a potential therapeutic target for retinal neovascularization

Retinal neovascularization, or pathological angiogenesis in the retina, is a leading cause of blindness in developed countries. Pathological angiogenesis occurs through the complex activation of pro-inflammatory and pro-angiogenic pathways. Transforming growth factor-β-activated kinase 1 (TAK1) is a mitogen-activated protein kinase kinase kinase (MAPKKK) activated by TGF-β1 and other pro-inflammatory cytokines such as tumor necrosis factor α (TNFα) and interleukin-1 (IL-1). TAK1 is a key mediator of inflammation, innate immune responses, apoptosis and tissue homeostasis and plays an important role in physiological angiogenesis. Its role in pathological angiogenesis, particularly in retinal neovascularization, remains unclear. We investigated the regulatory role of TAK1 in pathological angiogenesis in the retina. Transcriptome analysis of human retina featuring retinal neovascularization revealed enrichment of known TAK1-mediated signaling pathways. Genetic or pharmacological inhibition of TAK1 activation in human endothelial cells induced by TNFα or IL-1 stimulation led to inhibition of phosphorylation of major kinases, including nuclear factor kappa B (NFκB) and mitogen-activated protein kinase (MAPK) signaling pathways. Suppression of this signaling, in turn, decreased expression of downstream genes associated with inflammation and angiogenesis, suggesting that TAK1 is required for these pathological processes. Transcriptome analysis of endothelial cells revealed that TAK1 knockout prevented inflammatory and immune responses induced by TNFα stimulation, mainly via cytokine and chemokine activity. Selective inhibition of TAK1 by 5Z-7-oxozeaenol in vitro ameliorated pro-angiogenic activity, including endothelial cell proliferation, migration and tube formation. Moreover, 5Z-7-oxozeaenol attenuated aberrant retinal angiogenesis in rats following oxygen-induced retinopathy. Our finding shows that TAK1 plays a key role in pathological angiogenesis in the retina. Selective inhibition of TAK1 prevents pathological angiogenesis, suggesting that TAK1 could be a potential therapeutic target for retinal neovascularization. GRAPHICAL ABSTRACT

signaling, in turn, decreased expression of downstream genes associated with inflammation and 48 angiogenesis, suggesting that TAK1 is required for these pathological processes. Transcriptome 49 analysis of endothelial cells revealed that TAK1 knockout prevented inflammatory and immune 50 responses induced by TNFα stimulation, mainly via cytokine and chemokine activity. Selective 51 inhibition of TAK1 by 5Z-7-oxozeaenol in vitro ameliorated pro-angiogenic activity, including 52 endothelial cell proliferation, migration and tube formation. Moreover, 5Z-7-oxozeaenol attenuated 53 aberrant retinal angiogenesis in rats following oxygen-induced retinopathy. Our finding shows that 54 TAK1 plays a key role in pathological angiogenesis in the retina. Selective inhibition of TAK1 55 prevents pathological angiogenesis, suggesting that TAK1 could be a potential therapeutic target for 56 retinal neovascularization. is increasing evidence that some patients do not respond well to anti-VEGF treatments (3). Therefore, 69 it is necessary to seek other therapeutic targets. 70 The common pathways shared between inflammation and angiogenesis play critical roles in 71 retinal neovascularization (4). Several studies have reported increased expression of various pro-72 inflammatory cytokines, chemokines, pro-angiogenic factors and adhesion molecules in patients with 73 retinal neovascularization, such as tumor necrosis factor α (TNFα), interleukin-1 (IL-1), chemokine 74 (C-X-C motif) ligand 8 (CXCL8; known as IL-8), vascular endothelial growth factor (VEGF), as well 75 as intercellular adhesion molecule 1 (ICAM-1) (5-9). TNFα was found to suppress the expression of 76 tight junction proteins and is required for VEGF-mediated hyperpermeability in endothelial cells and 77 the breakdown of the blood-retinal barrier in patients with diabetic retinopathy (DR) (10, 11). 78 Interleukins are also primary mediators of inflammation that can promote angiogenesis via induction 79 of proangiogenic factors such as VEGF (4). Moreover, TNFα and IL-1 stimulates expression of ICAM-80 1, CXCL8 and VEGF in endothelial cells and microglia cells, both of which promote progression of 81 retinal neovascularization (12, 13). A key role for inflammatory mediators is highlighted by their 82 growing use as prognostic markers of the severity of proliferative diabetic retinopathy (PDR) (4). 83 Transforming growth factor-β-activated kinase 1 (TAK1), a member of the mitogen-activated 84 protein kinase (MAPK) family, is a critical serine/threonine kinase in several cellular signaling 85 pathways. Such pathways can be activated by diverse pro-inflammatory stimuli including TNFα, IL-86 1, transforming growth factor beta (TGF-β), or toll-like receptor (TLR) ligands. The engagement of 87 TAK1 in turn activates downstream signaling pathways, including nuclear factor kappa B (NFκB) and 88 MAPK p38, extracellular signal-regulated kinase (ERK), and c-Jun kinase (JNK) signaling, 89 modulating inflammatory responses and cell survival (14-16). TAK1 inhibition has been shown to 90 reduce TNFα and CXCL8 secretion in leukocytes (17) and to suppress prostaglandin-endoperoxide 91 synthase 2 (PTGS2; known as cyclooxygenase-2, COX-2), ICAM-1, and VEGF expression in cancer 92 cells through NFκB signaling pathway (18). Expression of TAK1 provides protection for endothelial 93 cells against TNFα-induced apoptosis under inflammatory conditions (19). Indeed, TAK1 deletion 94 causes embryonic lethality as the capacity to prevent TNFα-induced endothelial cell death and vessel 95 regression is lost. Moreover, TAK1 is also important for supporting TNFα-independent vascular 96 formation and endothelial migration. These broad roles for TAK1 suggests that endothelial TAK1 97 inhibition may be a useful alternate anti-angiogenesis target in retinal disease (19,20). 98 We hypothesized that TAK1 plays a crucial role in pathological angiogenesis and TAK1 inhibition 99 can suppress retinal neovascularization. Our findings indicate that TAK1 is an important regulator of 100 inflammatory and angiogenesis-related pathways that contribute to retinal neovascularization. 101 Pharmacological inhibition of TAK1 alleviated aberrant retinal angiogenesis by modulating pro-102 inflammatory and pro-angiogenic signaling in a rat model of ischemia-induced retinopathy. 103 104 105

TAK1 is expressed in human retina and is positively associated with retinal neovascularization 107
To understand TAK1 expression profiles in the human retina, we first analyzed published of single 108 nuclei RNAseq (snRNAseq, GSE135133) data to examine TAK1 expression in different retinal cells. 109 Analysis of 121,799 individual cells from human retina revealed that both TAK1 (MAP3K7) and its 110 binding protein, TAB1-3, are expressed moderately in most neural and non-neural retinalcells (Fig. 1A  111 and S1). Notably, among the retinal cells examined, TAK1 is highly enriched in endothelial cells. 112 Immunostaining suggests that TAK1 is expressed in various layers of the human retina in both neural 113 and non-neural cells, such as mural cells including smooth muscle cells (Fig. 1B) in larger vessels 114 (diameter of artery or vein ranging from 0.6 -16 mm) (21). We also found that co-localization of 115 TAK1 and endothelial cell marker CD31 was more likely to be found in the microvasculature, 116 including arterioles or venules (20 -25 μm) and capillaries (9 μm). 117 To identify the key molecular signaling or biological pathways relevant to the angiogenic process 118 in patients with retinal neovascularization, a transcriptome dataset from humans with proliferative 119 diabetic retinopathy (PDR) was interrogated by taking advantage of Gene Set Enrichment Analysis 120 (GSEA), a knowledge-based approach for interpreting genome-wide expression profiles (22). 121 Compared with controls those with retinal neovascularization showed enrichment of known TAK1 122 upstream signaling pathways, including IL1, TNF, TGF-β, TLR, T-cell receptor (TCR) and B-cell 123 receptor (BCR) (Fig. 1C, S2 and S3A). Pathways related directly to TAK1 were enriched, such as 124 "TAK1 activation of NFkB by phosphorylation and activation of inhibitor of κB (IκB) kinase (IKK) 125 complex", and "activated TAK1 mediated p38 MAPK activation" (Fig. 1C and S3B). There was 126 general upregulation of genes associated with the TAK1 activates NFkB by phosphorylation pathway 127 and activation of IKK complex pathway in patients with retinal neovascularization compared with 128 controls (Fig. 1D and 1E). There was negative enrichment of genes associated with several biological 129 pathways in relation to visual phototransduction and neuronal systems (Fig 1C) in the endothelial response to inflammation, we generated TAK1-knockout (KO) human telomerase-138 immortalized microvascular endothelial (TIME) cells (TAK1-KO) using CRISPR/Cas9 ( Fig. 2A-2D). 139 To explore the role of TAK1 in endothelial cells, we generated a transcriptome through RNA 140 sequencing (RNA-seq) of wild type TIME cells (WT) and TAK1-KO TIME cells (TAK1-KO). We 141 found that 2178 genes were differentially expressed [false discovery rate (FDR), <0.05] following 142 TAK1-KO ( Fig. S4A and S4B). GSEA results suggested that TAK1 knockout negatively affected 143 pathways related to immune responses and inflammation, such as interferon signaling (Fig. S4C). 144 Since TNFα is strongly correlated with progression of retinal neovascularization and it is a major 145 activator of TAK1 signaling (10), we used TNFα as the stimulus for in vitro studies. Treatment with 146 TNFα significantly increased TAK1 phosphorylation in wild-type cells, an effect not observed in 147 TAK1-KO cells ( Fig. 2E and 2F). IκB degradation and NFκB p65 phosphorylation are key events in 148 NFκB signaling (24). TNFα significantly induced NFκB p65 phosphorylation and IκB degradation in 149 wild-type cells, while no such changes occurred in TAK1-KO cells ( Fig. 2E and 2G). Likewise, 150 phosphorylation of JNK, p38, and ERK in MAPK signaling was not increased in TNFα-treated TAK1-151 KO cells compared to TNFα-treated wild-type cells ( Fig. 2E and 2H). Collectively, these data indicate 152 that TAK1 is required for phosphorylation of the key mediators of NFκB and MAPK signaling in 153 human microvascular endothelial cells. 154 To explore the effects of TAK1-KO in TIME cells upon TNFα stimulation, we generated two 155 more transcriptomes by RNA sequencing TNFα-treated wild type (WT) TIME cells and TNFα-treated 156 TAK1-KO TIME cells. We found that 462 and 263 genes were differentially expressed (FDR<0.05 157 and Log2FoldChange >1 or <-1) in wild type cells upon TNFα treatment (WT vs WT TNFα; hereafter 158 referred to as the WT TNFα set) and TAK1-KO cells upon TNFα treatment (TAK1-KO vs TAK1-KO  159 TNFα; hereafter referred to as the TAK1-KO TNFα set), respectively. The principal components 160 analysis of our dataset revealed a complete separation of four groups of samples (n=4 in each group, 161 In keeping with the notion that TNFα induces TAK1-related signaling, we next sought 162 genes that are differentially expressed between WT TNFα and TAK1-KO TNFα sets. We observed 163 that 246 and 110 genes were exclusively up-regulated and down-regulated in WT TNFα set and not 164 the TAK1-KO TNFα set, suggesting these dysregulated genes were not induced by TNFα due to 165 TAK1-KO ( Fig. 3A and 3B). Gene Ontology (GO) enrichment analysis of the 246 exclusively up-166 regulated genes in the WT TNFα set suggested that TNFα stimulated inflammatory and immune 167 responses (Fig. S6A), mainly via cytokine and chemokine activity (Fig. S6B) while such TNF 168 responses did not occur in TAK1-KO TIME cells. 169 To further explore the biological significance of our differential expression findings, we 170 undertook GSEA of WT TNFα and TAK1-KO TNFα sets. This demonstrated a marked inflammatory 171 response via the NFkB pathway in cells stimulated by TNFα; however, activation of the NFkB 172 pathway was significantly reduced in TAK1-KO cells (Fig. 3C), confirming TNFα-induced 173 inflammation is mainly dependent on NFkB activation through TAK1. In keeping with the notion that 174 NFkB is the main regulator of immune responses and inflammation, and is also involved in cellular 175 apoptosis and cell cycle regulation (25), we selected five major biological functions regulated by NFkB 176 from a NFkB target gene database (26) to investigate the differential expression of included genes in 177 each function category between WT TNFα set and TAK1-KO TNFα set (Table 1) were negatively affected in cells due to TAK1-KO. Expression of the most significant genes (those 183 also shown to be involved in retinal neovascularization) from each functional category in each group 184 was also validated using qRT-PCR (Fig. 3D). Upon TNFα stimulation, the expression level of ICAM-185 1, PTGS2, CXCL8 and VEGF-A were substantially elevated in wild-type cells, while no TNFα induced 186 change was observed in TAK-1 KO cells (Fig. 3D). Similar results were found in IL1b-treated wild-187 type cells but not in IL1b-treated, TAK1-KO cells (Fig. S12). Likewise, TAK1 knockdown using 188 TAK1 siRNA (siTAK1) resulted in a significant anti-inflammatory effect in both TIMEs (Fig. S13A, 189 S13B and S13E) and human retinal microvascular endothelial cells (HRMECs) (Fig. S13C and S13F). Pharmacological inhibition of TAK1 suppresses inflammation-mediated angiogenic signaling 196 and pro-angiogenic activities in vitro 197 5Z-7-oxozeaenol, a resorcylic acid lactone derived from a fungus, is a selective inhibitor of TAK1 198 (27). We tested the effects of 5Z-7-oxozeaenol in human microvascular endothelial cells stimulated 199 with TNFα. Similar to TAK1-KO, 5Z-7-oxozeaenol significantly suppressed TNFα induced 200 phosphorylation of NFκB p65 ( Fig. 4A and 4B), JNK, p38, and ERK (Fig. S14), indicating that 5Z-201 7-oxozeaenol inhibits NFκB and MAPKs signaling in a dose-dependent manner. Consistent with the 202 anti-inflammatory effects observed in TAK1 knockout endothelial cells, 5Z-7-oxozeaenol inhibited 203 TNFα-induced expression of pro-inflammatory or pro-angiogenic genes, PTGS2, CXCL8 and ICAM-204 1 as well as VEGF-A expression (Fig. 4C). This inhibitory effect of 5Z-7-oxozeaenol was also found 205 in HRMECs stimulated with TNFα (Fig. 4D). We also observed similar inhibitory effects of 5Z-7-206 oxozeaenol stimulated with IL1β in HRMECs (Fig. S15). 207 In addition to examining the pro-angiogenic signals mediated by TAK1 in response to 208 inflammatory insult, we further investigated the involvement of TAK1 in various endothelial functions 209 related to angiogenesis through ex vivo and in vitro assays. We first used an ex vivo aortic ring 210 sprouting assay to test the anti-angiogenic effects of pharmacological TAK1 inhibition by 5Z-7-211 oxozeaenol. We found that vascular sprouting was significantly suppressed 3 days (45.9%) or 9 days 212 (45.8%) after 5Z-7-oxozeaenol treatment compared with vehicle treatment (Fig. 5A). We next 213 performed a range of in vitro angiogenesis assays including cell viability, scratch migration, and tube 214 formation assay to test anti-angiogenic effects of 5Z-7-oxozeaenol in TIMEs and HRMECs. Since 215 TAK1 was found to protect endothelial cells from TNFα-induced apoptosis under inflammatory 216 conditions (19), we also evaluated the effect of 5Z-7-oxozeaenol on in vitro angiogenesis activity in 217 the presence or absence of TNFα. MTT assays revealed that TNFα alone did not affect the cell viability. 218 However, a profound reduction of cell viability was observed with cells treated with 5Z-7-oxozeaenol 219 or 5Z-7-oxozeaenol in the presence of TNFα stimulation compared with the control groups in TIMEs 220 and HRMECs, respectively. (Fig. 5B and 5C). Moreover, TNFα significantly reduced endothelial 221 migration by 20% (Fig. 5D). Cells treated with 5Z-7-oxozeaenol or TNFα/5Z-7-oxozeaenol showed 222 greater reductions in TIME migration of around 50% (Fig. 5D). Interestingly, TNFα alone did not 223 affect endothelial migration in HRMECs (Fig. 5E). 5Z-7-oxozeaenol resulted in a 20% reduction in 224 HRMEC migration both in the presence or absence of TNFα stimulation (Fig. 5E). Furthermore, 5Z-225 7-oxozeaenol was effective at reducing TNFα-driven tube formation activity in TIMEs (Fig. 5F) and 226 HRMECs (Fig. 5G). No changes in tube formation activity were observed in 5Z-7-oxozeaenol-treated 227 TIMEs and HRMECs. Collectively, our data reveal that pharmacological inhibition of TAK1 228 significantly suppresses inflammation-mediated angiogenic signaling and reduces endothelial cell 229 angiogenic activity. 230 231

Inhibition of TAK1 attenuates retinal neovascularization in vivo 232
To investigate the role of TAK1 in pathological neovascularization in the retina, we used a rat model 233 of oxygen-induced retinopathy (OIR) which yields ischemic avascular zones and preretinal 234 neovascularization similar to that observed clinically in PDR and retinopathy of prematurity. Neonatal 235 rats were subjected to daily cycles of 80% oxygen for 21 hours and ambient air for 3 hours from 236 postnatal day 0 (P0) to P14 to induce vaso-obliteration. At P14 rats were returned to room air, where 237 maximal preretinal neovascularization occured by P18, which was followed by a phase of vascular 238 regression (Fig. 6A). A significant increase in TAK1 expression at both RNA and protein levels was 239 observed in OIR retinae at P18 compared with controls ( Fig. 6B and 6C). Examination of the retinal 240 distribution of TAK1 expression revealed that TAK1 expression was uniquely expressed in areas of 241 active neovascularization (Fig. 6D). To gain further insight into the role of TAK1 activation at the 242 peak of retinal neovascularization (P18), we then performed transcriptome analyses using bulk-RNA 243 sequencing. GSEA-based analysis revealed a positive enrichment of several TAK1 upstream pathways, 244 including IL1, TNF, TCR, BCR and TLR ( Fig. 6E and S16), which is consistent with our findings in 245 patients with retinal neovascularization due to PDR. Similar to findings in humans, there was a 246 negative enrichment of genes associated with phototransduction and neuronal systems at P18 in the 247 rats. Of note, certain canonical pathways and biological processes related to angiogenesis were 248 positively enriched in the OIR model, such as VEGF-A/VEGFR2, PDGF pathway, the cellular 249 response to hypoxia, angiogenesis and inflammation ( Fig. 6E and S16). Overall, our data implicates 250 active involvement of TAK1 in retinal neovascularization. 251 To assess whether intraocular application of a TAK1 inhibitor could be used to reduce retinal 252 neovascularization, 5Z-7-oxozeaenol was administered intravitreally to the OIR rats at P14 (Fig. 7A). 253 Four days after drug administration, retinal neovascularization was significantly suppressed by 51.3% 254 and 57.1% in OIR rats that had received low (18 ng) and high doses (90 ng) of 5Z-7-oxozeaenol, 255 respectively, compared to OIR rats that had vehicle only ( Fig. 7B and S17). No significant difference 256 in retinal vaso-obliteration was found among these groups (Fig. 7C). We also examined the 257 development of deep capillary beds to assess the effects of TAK1 inhibition in physiological vessel 258 growth (28). Normal control rats developed a complete deep layer vasculature, while such 259 development was severely disrupted in OIR rats (Fig. 7D). We found that 5Z-7-oxozeaenol treatment 260 did not improve or damage the formation of deep capillary beds in OIR retina, which is likely attributed 261 to suppression of proliferative vessels by TAK1 inhibition. In addition to ICAM-1-related leukocyte 262 infiltration, microglia, the central nervous system resident immune cells, are also emerging as key 263 mediators or modulators of retinal inflammatory responses (29). Co-immunostaining of IB4 and Iba1 264 (marker of microglia) illustrated that 5Z-7-oxozeaenol treatment at both low and high doses 265 significantly attenuated the microglial adhesion to the vascular surface ( Fig. 8A and 8B). Our data 266 also confirmed that TAK1 inhibition attenuated the expression of inflammatory and angiogenic signals, 267 TNFa, VEGFA and ICAM-1, in OIR retina (Fig. 8C). Here, we provide the first evidence that both genetic and pharmacological inhibition of TAK1 313 activity in human endothelial cells prevents inflammatory stimuli from inducing sequential 314 phosphorylation of upstream kinases in both NFκB (p65) and MAPK (JNK, p38 and ERK) pathways 315 and suppressed expression of downstream genes that drive inflammation and angiogenesis, including 316 ICAM-1, PTGS2, CXCL8 and VEGF-A. GSEA results confirmed that TAK1 knockout prevented TNFα 317 induced chemokine and cytokine activation predominantly through the NFκB pathway, which 318 attenuated inflammation and immune responses. Of note, GSEA did not identify significant changes 319 in the MAPK pathway in either wild type or TAK1-KO cells treated with TNFα, potentially because 320 the MAPK pathway activates more genes than the NFκB pathway (40) and fewer genes in MAPK 321 pathway showed significant changes compared to NFκB pathway. In addition, TAK1 inhibition by 5Z-322 7-oxozeaenol significantly reduced endothelial cell angiogenic activities including migration, 323 proliferation and tube formation. These data together suggest a significant role for TAK1 in mediating 324 crosstalk between inflammatory and angiogenic pathways, both of which are crucial in pathological 325 angiogenesis. 326 TAK1 is thus a central regulator of cell survival and death (16). 5Z-7-oxozeaenol, a resorcylic 327 acid lactone derived from a fungus, acts as a potent and selective inhibitor of TAK1, which displays 328 more than 33-and 62-fold selectivity for TAK1 over binding of MEKK1 and MEKK4, respectively 329 (27). It forms a covalent complex with TAK1 and impedes both the kinase and ATPase activity of 330 TAK1 following bi-phase kinetics (41). We demonstrated that TAK1 inhibition by 5Z-7-oxozeaenol 331 directly reduced endothelial cell viability. This effect is more likely attributable to the suppression of 332 cell proliferation (such as cell cycle arrest, data not shown) rather than inducing cell apoptosis for we 333 did not observe any obvious morphologic changes associated with apoptosis in endothelial cells, such 334 as a decline in cell volume or aberrant cell nuclei (42). Intravitreal injection of 5Z-7-oxozeaenol 335 suppressed retinal neovascularization, and modulated elevated TNFα expression in the retina of OIR 336 rats, without causing degeneration of the normal vasculature. These data suggest that TAK1 inhibition 337 by 5Z-7-oxozeaenol may more specifically attenuate active angiogenesis with less risk of damage to 338 normal endothelial cells. However, further studies are required to understand the effects of repeated 339 5Z-7-oxozeaenol treatment. Moreover, TAK1 Inhibition by 5Z-7-oxozeaenol is known to block pro-   creating an incision-like gap. After removing debris-containing medium, cells were washed with 100 481 µl PBS. Cells were then exposed to 5Z-7-oxozeaenol (1 µM) with or without TNF-α (10 ng/ml) in 100 482 µl serum/growth factor-free medium containing VEGF (20ng/ml) (293-VE-010; R&D Systems, 483 Minneapolis, MN). The scratch area was photographed using an Incucyte® Live-Cell Imaging and 484 Analysis System (Incucyte) immediately after wounding (0 hour) and 16 hours post-wounding. The 485 scratch area was quantified using ImageJ version 1.48 software (http://imagej.nih.gov/ij/). Cell 486 migration was expressed as a percentage of closure of the scratch area: (scratch area at 0 hour -scratch 487 area at 16 hour) / scratch area at 0 hour × 100%. 488

489
Endothelial tube formation assay 490 Endothelial tube formation was quantified as previously described (58). In brief, TIMEs or HRMECs 491 were seeded at a density of 2 × 10 5 /well in a 6-well plate. Cells were then pre-treated with 5Z-7-492 oxozeaenol (1 µM) for 0.5 hours followed by TNF-α (10 ng/ml) treatment for 24 hours. Tasmania (A0017311). Rats were randomly allocated to treatment groups such that littermates were 511 distributed equally between groups. 512 513

Rat model of oxygen-induced retinopathy and intravitreal injection of 5Z-7-oxozeaenol 514
The oxygen-induced retinopathy model was induced in rats as previously described (59). Briefly, 515 Sprague-Dawley litters (within 12 hours of birth; P0) and their nursing mothers were exposed to daily 516 cycles of 80% O2 for 21 hours and room air for 3 hours in a custom-built and humidity-controlled (< 517 80%) chamber until P14. At P14 animals were returned to room air until P16 or P18, when they were 518 sacrificed to harvest retinae. An ProOx 110 oxygen controller (BioSpherix; Parish, NY) was used to 519 regulate and monitor the oxygen level. At P14, intravitreal injections were performed to administer 520 5Z-7-oxozeaenol visualizing with a microsurgical microscope. Under surface anesthesia, a puncture 521 in the superior temporal quadrant of the limbus was made by a 30-gauge needle. information. Other original datasets will also be available to the research community upon reasonable 563 request. 564

ACKNOWLEDGEMENTS 565
The authors thank UTAS CFF animal technicians, Karen Shiels, Keri Smith, Heather Howard, Lisa 566 Harding and Danielle Eastley for their assistance with rat chamber operation. We also thank Dr 567 Jacqueline Y.K. Leung for assistance with intravitreal injection, and Zheng He for assistance with 568 retinal imaging. This work was supported by grants from the National Health and Medical Research 569 Council of Australia (1061912, 1185600 and 1123329   (D) After exposure to 5Z-7-oxozeaenol (1 µM) with or without TNFα (10 ng/ml), a wound healing 674 assay was performed to evaluate migration activity in TIMEs. Representative images were taken after 675 wounding (or scraping cells, 0 hr) and at 16 hours (16 hr) post-wounding. Quantitative analysis of cell 676 migration assay was characterized by wound closure (n = 5). (E) With same pre-treatment as Fig. 5D, 677 migration assay was performed in HRMECs (n = 4). (F) After exposing to 5Z-7-oxozeaenol (1 µM) 678 with or without TNFα (10 ng/ml) for 24 hours, tube formation assay was performed (n = 5). 679 Representative images were taken 6 hours post tube formation. Quantitative analysis was characterized 680 by mesh number. (G) With same pre-treatment as Fig. 5F, tube formation assay was performed in 681 HRMECs (n = 5). Group data are shown as means ± SEM. Statistical analysis was undertaken with 682 one-way ANOVA and Tukey's multiple comparison test; *P < 0.05, **P < 0.01, ***P < 0.001, ****P 683 Neonatal rats were exposed to 80% of oxygen for 21 hours each day from postnatal day 0 (P0) to P14. 690 Rats were returned to room air from P14 to P18. Vessel loss was induced during oxygen exposure and 691 retinal angiogenesis could be induced after rats returned to room air. (B) qPCR analysis of TAK1 692 expression in OIR at P18 as compared with the normoxia group (n = 6-7). (C) Western blot analysis 693 of TAK1 expression in OIR at P18 compared with normoxia group (n = 10-12). Representative 694 immunoblotting from 4 independent samples were shown for normoxia (lanes 1-4) and OIR group 695 and ICAM-1) and pro-angiogenesis (VEGFA) gene expression at P16 (n = 5-7). Group data are shown 727 as means ± SEM. Statistical analysis was undertaken with one-way ANOVA and Tukey's multiple 728 comparison test; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.