Epoxydocosapentaenoic acids are anti-angiogenic in vitro: all the chemically stable EDP regioisomers inhibited VEGF-induced angiogenesis in a Matrigel plug assay (Zhang et al., 2013). Of note, 19,20-EDP, which is the least efficient substrate for sEH, therefore most abundant isomer (Zhang et al., 2014), had no effect on endothelial cell proliferation but strongly inhibited human umbilical vein endothelial cell tubular network formation and migration and weakly inhibited matrix metalloproteinase 2 (MMP-2) activity via a VEGF receptor 2 dependent manner (Zhang et al., 2013), although the exact mechanism through which EDPs crosstalk with VEGF signaling remains to be clarified.

ω-3 EpFAs also have anti-inflammatory and anti-angiogenic effects in animal models of ocular angiogenesis. Dietary intake of ω-3 PUFAs but not ω-6 PUFAs reduces murine laser-induced choroidal neovascularization (L-CNV) (Yanai et al., 2014), a widely used model in which laser ruptures Bruch’s membrane, resulting in angiogenesis from the choroid into the subretinal space. This recapitulates key features of wet AMD and serves as a model in which to test anti-angiogenic therapies (Lambert et al., 2013).

Dietary intake of ω-3 PUFAs in mice substantially enhanced levels of 17,18-epoxyeicosatetraenoic acid (EEQ) and 19,20-EDP in the serum lipid profile. However, it did not increase levels of EDPs in the retinal lipid profile. Interestingly, dietary intake of 17,18-EEQ or 19,20-EDP also suppressed CNV, suggesting that the protective effect of ω-3 PUFAs against CNV could be mediated by its downstream epoxy metabolites that are generated by CYP. In addition, dietary ω-3 PUFAs interfered with leukocyte invasion into the CNV lesions, while ω-6 PUFA did not (Yanai et al., 2014). These effects were associated with anti-inflammatory properties of EpFAs. Specifically, they modulated leukocyte rolling velocity by changing the expression of adhesion molecules on the surfaces of leukocytes and in the CNV lesions (Hasegawa et al., 2017). In transgenic mice overexpressing CYP2C8 in endothelial cells, the dietary intake of ω-3 PUFAs, which increased production of 17,18-EEQ and 19,20-EDP in serum, reduced CNV lesions (Hasegawa et al., 2017). Likewise, dietary ω-3 PUFAs reduced CNV lesions in Ephx2^-/- mice. These mice also had increased plasma levels of 17,18-EEQ and 19,20-EDP, since sEH-dependent degradation of these epoxides into corresponding diols was blocked. In contrast, dietary ω-3 PUFAs did not confer inhibitory effects on CNV in mice overexpressing sEH.

The relevance of sEH to CNV was further supported by our recent study that showed an increase in the expression of sEH in the eyes of L-CNV mice and human wet AMD patients (Sulaiman et al., 2018). Interestingly, sEH was upregulated in the photoreceptors, and ocular enzymatic activity of sEH was increased upon L-CNV induction in adult mice. The lipid profile of retina/choroidal tissue of the mice revealed that the ratio of 19,20-EDP to 19,20-DHDP was significantly reduced in L-CNV, suggesting enhanced sEH activity (Sulaiman et al., 2018). In the developing mouse retina, sEH is highly expressed in Müller glia, and DHDP produced by sEH contributes to retinal angiogenesis (Hu et al., 2014). Müller glia span the entire retina radially, providing structural and metabolic support for retinal neurons (Reichenbach and Bringmann, 2013). The Müller cell specific knockout of sEH or systemic deletion of sEH significantly impaired developmental retinal angiogenesis and altered the retinal lipid profile. The level of 19,20-DHDP was significantly reduced in the retina of Ephx2^-/- mice. Intravitreal injection of 19,20-DHDP rescued impaired retinal angiogenesis in Ephx2^-/- mice. 19,20-DHDP was also found to be a signaling molecule, downregulating the endothelial Notch signaling pathway by inhibiting presenilin-1 dependent γ-secretase activity, which is required for release of the Notch intracellular domain (Hu et al., 2014). Interestingly, the crosstalk between Notch and VEGF pathways in angiogenesis has been reported in numerous studies, where activation of Notch signaling modulates VEGF signaling (Hellström et al., 2007; Li and Harris, 2009). Given this, inhibition of sEH not only stabilizes the anti-angiogenic and anti-inflammatory ω-3 EpFAs, but also inhibits production of pro-angiogenic DHDP.

Targeting sEH with small molecule inhibitors effectively reduces ocular angiogenesis (Table 1). SH-11037, a synthetic homoisoflavonoid that we developed in cell-based assays and subsequently identified as an sEH inhibitor (Sulaiman et al., 2018), effectively blocked key angiogenic properties of human retinal endothelial cells (HRECs) – proliferation, migration and tube formation – without inducing cell death (Basavarajappa et al., 2015). As well, SH-11037 reduced angiogenesis in an ex vivo choroidal sprouting assay and inhibited developmental ocular angiogenesis in zebrafish larvae (Sulaiman et al., 2016). Local application of SH-11037 (1 μM) into the eye via intravitreal injection significantly suppressed CNV lesions (Sulaiman et al., 2016; Figure 1C,D) and was also effective in reducing retinal neovascularization in the oxygen-induced retinopathy (OIR) model (Basavarajappa et al., 2015), in which neonatal mouse pups are subjected to hyperoxia during their developmental retinal vascularization, causing ischemia-induced angiogenesis on return to normoxia (Scott and Fruttiger, 2009; Kim et al., 2016). Structural, morphological and vascular examination of retina and electroretinography showed that up to 100 μM intravitreal SH-11037 does not exert ocular toxicity (Sulaiman et al., 2016). Excitingly, SH-11037 also synergized with anti-VEGF therapy to reduce L-CNV (Sulaiman et al., 2016).

Intravitreal injection of other known sEH inhibitors t-AUCB (1–10 μM) and “compound 7” (10–30 μM) also suppressed L-CNV lesions (Sulaiman et al., 2018; Table 1). Moreover, intravitreal 10 μM SH-11037 or t-AUCB treatment effectively normalized CNV-induced sEH enzymatic activity and increased the ratio of 19,20-EDP to 19,20-DHDP, indicating that local pharmacological inhibition of sEH can alter the lipid metabolism in the eye (Sulaiman et al., 2018). These studies used 6–15 mice per treatment, plus vehicle injected controls (Basavarajappa et al., 2015; Sulaiman et al., 2016, 2018), which should be sufficient to avoid any confounding effects of inflammation, which is a concern with this delivery route (Chiu et al., 2007). An inhibitory effect on ocular angiogenesis was also reported for other routes of administration of sEH inhibitors. Oral administration of TPPU (1 mg/kg/day) not only reduced CNV lesions and vascular leakage, but also its coadministration with i.p. injection of 17,18-EEQ or 19,20-EDP (50 μg/kg/day) potentiated anti-angiogenic effects on CNV (Hasegawa et al., 2017). In addition, normal neonatal mice that received i.p. injection of t-AUCB (2 mg/kg, twice/day) from 1 to 4 days had significantly reduced retinal vascularization (Hu et al., 2014). Together, these studies support the therapeutic potential of sEH inhibitors for ocular neovascularization.

However, some studies do not support the finding that sEH inhibition reduces ocular angiogenesis (Gong et al., 2017). Shao et al. (2014) found that EDP is pro-angiogenic and inhibition of CYP epoxygenase rather than sEH reduced retinal neovascularization in OIR. Systemic overexpression of the CYP epoxygenase CYP2C8 and downregulation of sEH expression in the retina of OIR mice were reflected in the increased and decreased retinal EDP to DHDP ratio, respectively. Dietary ω-3 PUFAs enhanced OIR-induced retinal neovascularization in CYP2C8 overexpressing mice and reduced retinal neovascularization in sEH overexpressing mice. In the aortic ring assay, DHA had an anti-angiogenic effect which was abolished by CYP2C8 overexpression, whereas 19,20-EDP alone had no effect but 19,20-EDP + sEH overexpression reduced aortic sprouting (Shao et al., 2014). However, the macrovessels of the aortic ring do not fully recapitulate microvascular features of the choroid capillaries (Shao et al., 2013).

Panigrahy and colleagues showed that CYP2CJ and CYP2C8 overexpressing mice and Ephx2^-/- mice had enhanced corneal and neonatal retinal vascularization (Panigrahy et al., 2013) and enhanced tumor dependent corneal angiogenesis (Panigrahy et al., 2012). CYP2C8 inhibition with montelukast further inhibited OIR and L-CNV in mice fed with ω-3 and ω-6 PUFAs, as opposed to sEH inhibition with i.p. injection of UC1770 (0.3 mg/kg) which enhanced OIR and L-CNV (Gong et al., 2016a). Also, UC1770 promoted aortic ring and choroidal sprouting, while montelukast enhanced the anti-angiogenic effects of DHA on aortic and choroidal sprouting, and HREC proliferation, which were all rescued by 19,20-EDP (Gong et al., 2016a). CYP2C8 inhibition with fenofibrate also further reduced OIR and L-CNV in animals fed with dietary ω-3 PUFAs (Gong et al., 2016b). Again, 19,20-EDP reversed the anti-angiogenic effects of fenofibrate in aortic and choroidal sprouting, HREC migration, and tube formation, in contrast to DHA which enhanced fenofibrate’s anti-angiogenic effects (Gong et al., 2016b). Thus, these studies present conflicting results regarding sEH as a therapeutic target for ocular angiogenesis.

There are important factors to consider in attempting to integrate the conflicting experimental data: stability of EpFAs, routes of administration of CYP or sEH inhibitors and systemic or local modulation of sEH/CYP expression, modulation of other lipid mediator pathways, and experimental design considerations.

First, we cannot conclude that 19,20-EDP promotes ocular angiogenesis when studies added 19,20-EDP into growth media without modulating sEH activity in the aortic ring and choroidal sprouting assay (Gong et al., 2016a,b). Although 19,20-EDP is a fairly stable isomer, it can be converted into its corresponding diol 19,20-DHDP by sEH. Co-treatment with an sEH inhibitor is required to stabilize 19,20-EDP to ascertain that a change in phenotype is due to 19,20-EDP and to minimize effects from 19,20-DHDP that is generated by sEH activity. Indeed, Hasegawa et al. (2017) demonstrated greater inhibitory effects on L-CNV upon oral administration of sEH inhibitor TPPU + 17,18-EEQ or 19,20-EDP compared to 17,18-EEQ or 19,20-EDP alone. Additionally, aortic rings from Ephx2^-/- mice treated with 19,20-EDP or 17,18-EEQ did not show any effect on angiogenic sprouting. One explanation for this lack of effect is that the lipid concentration used (1 μM) (Shao et al., 2014) may be below the inhibitory threshold; further characterization of dose dependent effects of these compounds on angiogenesis may be necessary.

Second, the studies suggesting that inhibition of sEH promotes ocular angiogenesis employed sEH blockade via constitutive Ephx2^-/- mice (Panigrahy et al., 2013) or oral administration (Gong et al., 2016b) or systemic i.p. injection of sEH inhibitors (Gong et al., 2016a). It is difficult to compare such results to those from tissue specific local targeting of sEH, considering the potential confounding effect of the blood-retina barrier (BRB) and the unique retinal lipid profile. To validate sEH as a key player in ocular angiogenesis, local targeting of sEH (e.g., tissue-specific knockout, intraocular injection, or topical therapy) is necessary. Moreover, sEH inhibition can have opposite effects on angiogenesis depending on tissue levels of fatty acids that are parent to EpFAs. EETs and EDPs are the most abundant epoxy substrates of sEH in ω-6 ARA and ω-3 DHA rich tissues, respectively. EETs usually have proangiogenic effects (Pozzi et al., 2005; Michaelis et al., 2008; Xu et al., 2013) while EDPs have antiangiogenic effects (Zhang et al., 2013; Capozzi et al., 2014; Hasegawa et al., 2017; Hu et al., 2017). Systemic targeting of sEH could accumulate EETs and their effects on angiogenesis could outweigh those of EDPs. Conversely, local targeting of sEH in the eye would predominantly accumulate EDPs since ω-3 DHA is most enriched in retina (Querques et al., 2011). Likewise, levels of EDPs are substantially greater than EETs in retina (Hu et al., 2014).

Third, the studies that showed that DHA delivery and CYP inhibition are antiangiogenic do not necessarily lead to the conclusion that downstream EpFA metabolites are proangiogenic, since CYP inhibition and the resulting accumulation of DHA could lead to accumulation of metabolites from the COX and LOX pathway. The dietary intake of DHA reduced retinal (Connor et al., 2007; Stahl et al., 2010b; Shao et al., 2014; Gong et al., 2016b) and choroidal neovascularization (Moghaddam-Taaheri et al., 2011; Gong et al., 2016a,b), but the inhibition of CYP epoxygenase potentiated the beneficial effect of DHA against ocular angiogenesis, suggesting that the resulting accumulation of DHA or reduced generation of EDP or both are playing a role in ocular angiogenesis. In addition, the results can also be partially explained by studies that show DHA derived metabolites of LOX and COX such as resolvins and neuroprotectin are anti-inflammatory and inhibit retinal angiogenesis (Connor et al., 2007). It is likely that the beneficial effects of exogenous DHA plus CYP blockade are mediated by increasing free DHA that can efficiently compete with ARA for the two other major metabolic pathways, LOX and COX, which produce ARA-derived pro-inflammatory metabolites such as prostaglandins and leukotrienes (Calder, 2010). In addition, certain COX metabolites from EETs were shown to be pro-angiogenic (Rand et al., 2017). Even though inhibition of CYP would reduce the formation of EDP, the effect of increased DHA leading to reduction in COX/LOX-dependent ARA derived pro-inflammatory and pro-angiogenic metabolites, and increase in COX/LOX dependent DHA derived anti-inflammatory metabolites could outweigh the loss of EDP.

Finally, experimental design factors should be considered. The ω-3/ω-6 lipid composition of mouse chow can vary between facilities, potentially influencing findings. Sex differences in animal models matter, too, given the estrogen dependent suppression of sEH expression (Yang et al., 2018). This could possibly reduce the response to sEH loss/inhibition in female mice whereas response to sEH loss/inhibition could be more apparent in male mice. Among the studies discussed above, many did not specify the sex (Connor et al., 2007; Stahl et al., 2010b; Hu et al., 2014; Shao et al., 2014; Basavarajappa et al., 2015; Gong et al., 2016a,b) while some studies reported using male mice (Panigrahy et al., 2012, 2013; Yanai et al., 2014; Hasegawa et al., 2017) or female mice (Sulaiman et al., 2016, 2018), thus posing a challenge in assessing potential variability in results due to sex differences. In addition, sex differences and age are critical factors in the L-CNV model, as aged female mice (>9 months) develop more severe CNV lesions than age-matched male mice whereas sex differences are not significant in younger mice (Gong et al., 2015). Reassuringly, mice at 6–8 weeks of age were used in all relevant studies (Yanai et al., 2014; Gong et al., 2016a; Sulaiman et al., 2016, 2018), which is an ideal age range for the L-CNV model (Gong et al., 2015). Likewise, all studies showed rigor in use of littermate controls for the OIR model (Panigrahy et al., 2012; Shao et al., 2014; Basavarajappa et al., 2015; Gong et al., 2016a,b), which is important as mice from larger litters with poor postnatal weight gain develop more severe OIR (Stahl et al., 2010a; Kim et al., 2016).