Inhibition of Soluble Epoxide Hydrolase 2 Ameliorates Diabetic Keratopathy and Impaired Wound Healing in Mouse Corneas Free

With the recent rapid increase in the prevalence of diabetes, the associated ocular complications such as retinopathy, cataract, uveitis, and neurophthalmic disorders have made diabetes a leading cause of blindness throughout the world (1). In addition to the aforementioned complications, various types of corneal disorders are also relatively common in patients with diabetes (2,3). Abnormalities of the cornea, termed diabetic keratopathy (DK), are resistant to conventional treatment regimens (for a comprehensive review, see Ljubimov [3]). Unlike diabetic retinopathy or cataracts, DK patients usually do not have detectable symptoms; however, once the cornea is injured, delayed epithelial wound healing is often observed (4) and may be associated with sight-threatening complications such as stromal opacification, surface irregularity, and microbial keratitis (5). Chronic low-grade inflammation and persistent oxidative stress are thought to be two major contributing pathogenic factors for the development of diabetic complications (3). However, the molecules and signaling pathways leading to these pathogenic events remain incompletely understood.

Corneal epithelium debridement is an ideal model to study re-epithelialization, delayed wound healing, and ulceration in the cornea (6). Using this model, we performed a genome-wide cDNA array analysis and observed that diabetes caused a general decline, at the transcriptional level, in both uninjured and healing corneal epithelial cells (CECs). In unwounded cells, 31 genes (loci) were upregulated and 72 genes were downregulated (7). Among these genes, EPHX2, encoding soluble epoxide hydrolase (sEH), was unique; its expression was not significantly altered in response to wounding and yet increased >14.8- and 11.9-fold in diabetic unwounded and healing epithelia when compared with that of normal corneas, respectively.

Arachidonic acid is a polyunsaturated omega-6 fatty acid and is the precursor that is metabolized to a wide range of biologically and clinically important eicosanoid molecules by various enzymes, including enzymes of the COX, lipoxygenase, and cytochrome P450 (CYP) monooxygenase pathways (8). The epoxygenase CYP enzymes generate four bioactive epoxyeicosatrienoic acids (EETs) by metabolizing arachidonic acid: 5,6-, 8,9-, 11,12-, and 14,15-EET. EETs contribute to the regulation of vascular tone, cardiovascular homeostasis, nociception, inflammatory response, angiogenesis, and cell proliferation (9–11). All EETs are then further metabolized by sEH (10,12), which is encoded by the gene EPHX2, and are converted into inactive or less active 1,2-diols, dihydroxyeicosatrienoic acids (DHETs) (13). A decrease in EET availability, due to an increased degradation by sEH, has been found to be a deleterious mechanism associated with various disease states such as cardiac hypertrophy, atherosclerosis, hypertension, pain, and diabetes (14–17). Accordingly, inhibition of sEH exerts beneficial actions in controlling or ameliorating these human diseases and pathologies (13,18,19). The role of EPHX2 has also been explored in the pathogenesis of diabetic complications, particularly in nephropathy; inhibition of sEH activity by gene deletion and by pharmacological inhibitor of EPHX2 reduced renal inflammation and injury in diabetic mice in an NF-κB–related manner (20). Moreover, the gain-of-function 55Arg polymorphism variant is found to be associated with acute kidney injury after cardiac surgery in patients without preexisting chronic kidney disease (21). Our cDNA array data, showing that hyperglycemia caused marked upregulation of EPHX2 in both unwounded and healing CECs (7), suggest that EPHX2 may contribute to the pathogenesis of DK (22).

In this study, we first confirmed the diabetes-associated expression of EPHX2 in CECs. We assessed its role during diabetes-impaired epithelium wound healing and sensory nerve innervation and regeneration. Moreover, we evaluated the therapeutic potential of the local application of sEH inhibitors and its downstream gene, heme oxygenase-1 (HO-1), to treat delayed diabetic wound healing.

All investigations using animals conformed to the regulations of the Association for Research in Vision and Ophthalmology (ARVO) Statement for the Use of Animals in Ophthalmic and Vision Research, the National Institutes of Health, and the guidelines of the Animal Investigation Committee of Wayne State University (WSU). Human autopsy corneas with or without diabetes were obtained from Michigan Eye Bank, without any personal information except age, sex, and cause of death.

Six-week-old C57BL/6 mice, both males and females purchased from The Jackson Laboratory, were induced to develop diabetes with streptozotocin (STZ) as described previously (23,24). Mice were considered as diabetic with blood glucose levels >300 mg/dL within 4 weeks postinjection and thereafter (24). Ephx2^−/− mice on a B6 background were a gift from Joan Graves (National Institutes of Health [NIH]/National Institute of Environmental Health Sciences [NIEHS]) and were induced to develop diabetes in the same manner as diabetic mice.

Tear secretion was determined with phenol red–impregnated cotton threads (Zone-Quick, Tokyo, Japan). The threads were placed in the medial canthus for 1 min and the length of the wetted part, turning red on soaking tears, was photographed and measured.

Diabetic and age-matched normal mice were anesthetized by an intraperitoneal injection of xylazine (7 mg/kg) and ketamine (70 mg/kg) plus topical proparacaine, and a 1.5-mm circular wound was first demarcated with a trephine in the central cornea, followed by the removal of epithelial cells within the circle with a blunt scalpel blade under a dissecting microscope (Zeiss). Two corneas were pooled in one tube, stored at −80°C. The collected cells were marked as unwounded (0 h). The progress of wound healing was monitored by fluorescence staining for epithelial defects and photographed with a slit lamp microscope. At the end of healing, the corneas were either snap frozen in optimal cutting temperature compound for cryostat sectioning or marked with the same size trephine for CEC collection, marked as healing CECs.

Western blotting of CECs was performed as follows. Cell lysates with an equal amount of proteins (20 μg) were separated with 5–15% gradient SDS-PAGE and transferred to pore size 0.2 μmol/L nitrocellulose membrane. The membranes were stained with EPHX2 (ab133173; Abcam), HO-1 (ab13243; Abcam), pSTAT3 (T705; Cell Signaling Technology), pSTAT3 (S727; Cell Signaling Technology), STAT3 (Cell Signaling Technology), or nonmuscle β-actin (A1978; Sigma-Aldrich), followed by incubation with horseradish peroxidase–conjugated donkey secondary antibodies (1:5,000 dilution; Jackson ImmunoResearch Laboratories). The bands were visualized with enhanced chemiluminescence (SuperSignal), and the images were acquired using Kodak Image Station 4000R Pro. Band intensity was analyzed using Carestream Molecular Imaging Software. Actin was used as loading control.

Mouse eyes were enucleated, embedded in Tissue-Tek optimal cutting temperature compound, and frozen in liquid nitrogen. Human 6-μm-thick sections were cut and mounted to poly-l-lysine–coated glass slides, fixed in 4% paraformaldehyde, blocked with 10 mmol/L PBS containing 2% BSA for 1 h at room temperature, and incubated with rabbit primary EPHX2 or HO-1 antibodies. This was followed by a secondary antibody, fluorescein isothiocyanate anti-rabbit (1:100; Jackson ImmunoResearch Laboratories). Slides were mounted with Vectashield mounting medium containing DAPI mounting media. Controls were similarly treated with rat or rabbit IgG, as well as using the depletion of primary antibodies with mouse recombinant EPHX2 (10:1). The sections were examined under a Nikon ECLIPSE 90i microscope. The center of unwounded or the leading edge of healing corneas was photographed.

Whole-mount immunostaining and quantitation of innervation of B6 mouse corneas were performed as described in our previous study (25). In brief, the enucleated eyes were fixed and the corneas were isolated and further fixed for an additional 10 min. The corneas were cut radially into six standardized sections and were incubated at 37°C in 20 mmol/L EDTA for 30 min, followed by 2-day incubation in 0.025% hyaluronidase and 0.1% EDTA in PBS. The tissues were blocked at room temperature for 2 h in PBS–Triton X-100 containing 2% BSA, followed by incubation overnight at 4°C with antibody against β-tubulin III. After secondary antibody, the tissues were mounted and examined under a confocal microscope (TCS SP2; Leica, Heidelberg, Germany). Innervation in a region was calculated as the percent area positive for β-tubulin III staining by ImageJ.

sEH enzymatic activity was measured using Cayman Chemical Cell-Based Assay Kit (600090) with 50 µg of cell lysates. For EET/DHET measurement, collected CECs from normal and diabetic mice were processed and epoxy/dihydroxy fatty acids were analyzed by the Lipidomics Core Facility at WSU using standardized liquid chromatography–mass spectrometry methods as described previously (26).

The statistical analyses were performed using the software GraphPad Prism 6. Data were presented as means ± SD. Experiments with two treatments and/or conditions were analyzed for statistical significance using two-tailed unpaired Student t test. Experiments with two groups were analyzed using one-way ANOVA (Fig. 1C), and more than two groups were analyzed with two-way ANOVA to determine overall differences. A Bonferroni posttest was performed to determine statistically significant differences. Significance was accepted at P < 0.05. Experiments were repeated at least twice to ensure reproducibility.

To determine whether EPHX2 was also expressed at the protein level, we first performed immunohistochemistry and observed that EPHX2 immunoreactivity was primarily found in the epithelium layer; its staining intensity increased with time post-STZ treatment (Fig. 1A). The presence of recombinant mouse EPHX2 in the first antibody incubation abolished immunoreactivity (Fig. 1A, insets, 8 weeks,). Since EPHX2 was mostly expressed in the epithelium, we performed Western blotting of CECs isolated at different times after STZ injection. Figure 1B revealed that EPHX2 was expressed in normal CECs; elevated expressions were detected at different time points post-STZ; relative intensities, normalized to actin, were 24.2 pre-STZ treatment, and increasing to 34.8, 49.1, and 42.8 at 2, 4, and 8 weeks, respectively (Fig. 1B). EPHX2 encodes a phosphatase at the N terminus and an epoxide hydrolase at the COOH terminus (9). sEH activity was assessed in diabetic CECs (Fig. 1C). Basal activity of sEH was detected in nondiabetic (NL) CECs, and STZ treatment resulted in 2.95-, 4.68-, and 5.21-fold increases at 2, 4, and 8 weeks, respectively (Fig. 1C). Figure 1D shows the immunostaining of human corneas, one from a 66-year-old male (NL) and the other from a 56-year-old female patient with severe diabetic retinopathy; stronger EPHX2 staining was seen in diabetic human corneal epithelium compared with that of a patient without diabetes. Taken together, these results indicate that sEH activity increases in diabetic mouse CECs.

To determine the amounts of EETs and their metabolite DHETs, we pooled epithelial cells from eight corneas from the control and STZ diabetic mice, two each, and assessed the levels of EETs. Among four EETs, 8.9-EET was detected with 0.30 and 0.43 ng per sample for NL and 0.17 and 0.24 ng per sample for diabetic CECs, respectively. 11,12-EET was barely detectable only in normal CECs (0.01). Tissue DHET levels were below the limit of quantification.

Ephx2^−/− mice have been used as models for hypertension and cardiovascular diseases (8,27,28). We obtained Ephx2-knockout (Ephx2-KO) mice on B6 background from Dr. Darryl C. Zeldin (NIH/NIEHS). An early study using intraperitoneal injection of 50 mg/kg/day STZ for 3 days reported that Ephx2 deletion prevented the development of diabetes (29). We used 50 mg/kg/day STZ injected intraperitoneally for 5 days. As shown in Fig. 2, Ephx2^−/− mice had similar average random blood glucose levels and became hyperglycemic in a similar time fashion with the control B6 mice (Fig. 2A); no differences in their body weight were observed as well (Fig. 2B). Hence, Ephx2 depletion had minimal effects on the development of diabetes induced by STZ. Our results were similar to those reported by Elmarakby et al. (20). Using Western blotting, we showed that diabetes induced EPHX2 expression in wild-type (WT), but not in Ephx2^−/−, mice (Fig. 2C).

One of the characteristics of DK in patients is decreased tear secretion. Whereas no difference was observed in NL WT and Ephx2^−/− mice, Schirmer test revealed that STZ treatment significantly decreased tear secretion in WT (65% of NL mice), but not Ephx2^−/− mice (Fig. 2D and E), suggesting an overall protective effect of Ephx2 depletion.

Having shown that EPHX1 expression and sEH activities were elevated in diabetic corneas and that diabetes can be induced in Ephx2^−/− mice, we next investigated the effects of Ephx2 depletion on epithelial wound closure in the corneas (Fig. 3). Ten weeks after induction of diabetes, the Ephx2^−/− diabetic and age-matched WT diabetic mice were wounded with a 1.5-mm diameter epithelium debridement procedure. The wounds were allowed to heal for 24 h, and the progress of wound healing was monitored by corneal fluorescence staining for epithelial defects and photographed with a slit lamp microscope (Fig. 3A). The remaining wound area (RWA) at 24 h postwounding was calculated (Fig. 3B). For NL corneas, Ephx2 depletion exhibited no detectable effects on the rate of epithelium wound closure, 15.5 ± 6.5% vs. 17.25 ± 6.25% RWA. As we showed before, diabetes significantly delayed wound healing in WT B6 mouse corneas (45.15 ± 6.75% RWA; P < 0.01); this delay, however, was markedly attenuated in Ephx2^−/− mice (13.5 ± 5.95% RWA).

Immunohistochemistry analysis revealed that there was little EPHX2 staining in normoglycemic corneas, but intense staining of EPHX2 in the entire epithelia of both unwounded and wounded diabetic corneas was observed (Fig. 3C).

Epithelial wounding is known to activate the JAK/STAT3 pathway, leading to wound closure in vitro and in vivo (30). EETs have been shown to mediate STAT3 signaling in cardiomyocytes (31). Using STAT3 phosphorylation as a marker of activation, we assessed the effects of hyperglycemia and EPHX2 expression on wound-induced STAT3 signaling (Fig. 4). Phosphorylation at both T705 and S727 can be detected at basal levels in unwounded (0 h) WT and Ephx2^−/− mouse corneas with or without diabetes. Wounding-induced phosphorylation (24 h postwounding) was observed at the T705 site in nondiabetic WT (6.26-fold over the control, 0-h WT-N, value set as 1 vs. 24-h WT-N) and Ephx2^−/− (13.1-fold, 24-h KO-N) mice, but not diabetic mouse corneas (1.05-fold, 24-h WT-D). Ephx2 depletion prevented hyperglycemia-suppressed STAT3 phosphorylation (13.4-fold, 24-h KO-D). Similar patterns were observed for site S727 as well.

Since HO-1 induction was found to be Src/STAT3 dependent in breast cancer cells (32), we therefore assessed HO-1 expression in response to wounding and/or hyperglycemia. Without wounding, HO-1 levels were low in WT and Ephx2^−/− mice with or without diabetes. Wounding induced HO-1 expression in normal WT and Ephx2^−/− (∼2.7-fold) but not diabetic mice, whereas Ephx2 depletion reversed the diabetes-induced suppression of HO-1 expression at the mRNA (a 2.53-fold increase) (Fig. 5A) and protein level (Fig. 5B and C).

There were contradicting reports regarding the effects of Ephx2 depletion on kidney function (20,27), presumably because of the phosphatase activity of the EPHX2 gene. To test the effects of sEH inhibition on corneal wound healing, we used two structurally different sEH-specific inhibitors, t-AUCB (33) and GSK2256294A (34) (Fig. 6). These two inhibitors had no noticeable effects on wound healing in normal corneas (Fig. 6A–D, panels 1, 2, 1′, and 2′). The delayed wound healing in diabetic corneas was significantly accelerated by t-AUCB (from 36.82 ± 13.03% to 15.57 ± 6.78% RWA) and GSK (from 33.27 ± 11.46% to 16.51 ± 4.22 RWA) (Fig. 6A–D). Similar to Ephx2 depletion, inhibition of sEH also promoted HO-1 expression in diabetic corneas, whereas it exhibited no effects on normal corneas in response to wounding (Fig. 6E and F).

Having shown that inhibition of sEH increased wound-induced HO-1 expression in diabetic corneas, we next investigated its role in accelerating wound healing in diabetic corneas using cobalt protoporphyrin (CoPP), a well-known HO-1 inducer (35). As expected, CoPP applied through subconjunctival injection stimulated HO-1 expression at the protein levels in healing diabetic corneas, whereas this expression was undetectable in the control, vehicle-treated eyes (Fig. 7A). These HO-1–expressing epithelia had a significantly higher healing rate compared with vehicle-treated diabetic corneas (28.78 ± 4.18% vs. 41.4 ± 6.36% RWA) (Fig. 7B and C).

We previously showed in rodent models of diabetes that there were decreases in the density of sensory nerve fibers/endings in the corneas (7,24,25). Using whole-mount confocal microscopy, we first examined sensory nerves in unwounded corneas. There was a significant decrease in the density of nerves at the center of the cornea of diabetic WT mice (Fig. 8A, 1 and 3), 71.9 ± 2.9% of the control, nondiabetic WT mice (set as value 1 or 100%) (Fig. 8B). Whereas Ephx2 depletion had no significant effects on NL corneas (Fig. 8A, 2 and 4), it increased the density of nerve fibers/endings of diabetic mouse corneas (105% of the control with no statistical significance) (Fig. 8B).

The regeneration of the basal nerve plexus was also examined at 1 month postepithelium wounding, at which time the wound center remained uncovered by the nerve plexus (Fig. 8A, 5–8). Whereas in the WT versus diabetic cornea, the region not covered by nerves was larger with a reduced nerve density (52.9 ± 2.2% vs. 38.9 ± 6.9%; P < 0.05) (Fig. 8A, compare 5 and 7), in Ephx2^−/− corneas, the densities of sensory nerves in normal and diabetic corneas (Fig. 8A, 6 and 8) were similar (51.1 ± 2.2% vs. 59.5 ± 10.%) (Fig. 8B).

In this study, we evaluated EPHX2 expression and role in the pathogenesis of DK in a B6 mouse model of human diabetes. We showed for the first time that EPHX2 expression was elevated in diabetic corneas as well as retinas with or without wounding. Unlike previous reports showing the failure of STZ to induce diabetes in Ephx2^−/− mice, we observed no significant differences in the course of STZ-induced diabetes between WT B6 and Ephx2^−/− mice. However, unlike WT mice, Ephx2^−/− mice did not develop diabetes-associated dry eye symptoms. Moreover, Ephx2 depletion increased the rate of epithelial wound closure, ameliorated sensory nerve degeneration in unwounded corneas, and enhanced their regeneration in wounded diabetic corneas while exhibiting no detectable effects on the NL corneas. Importantly, two structurally different sEH inhibitors accelerated delayed epithelial wound healing in the diabetic corneas to a level similar to that of NL mice. Inhibition of sEH also restored hyperglycemia-suppressed expression of HO-1, a factor necessary for proper wound healing in the cornea. Our study revealed that elevated EPHX2 in diabetic corneas is a pathogenic factor for DK and suggested that supplementing EETs and/or inhibiting sEH may prevent or treat DK.

EPHX2 has been found to be expressed ubiquitously in many tissues. It has, however, also been detected as a stress response gene associated with many human diseases (14–17). In the literature, results of EPHX2 expression in the diabetic mouse kidney were mixed, from a decrease in the cytosol (36), no significant changes (20), or an increase (37,38) in whole kidney extracts. We showed that the enzymatic activity of sEH was also elevated during the course of STZ-induced diabetes in CECs. This is consistent with the report showing a great decrease in the EETs/DHETs in diabetic, compared with normal, mice (20). The elevated EPHX2 expression in diabetic tissues is consistent with its role as a stress response gene (39,40).

EETs have been shown to promote organ and tissue regeneration, including liver regeneration, kidney and lung compensatory growth, corneal neovascularization, and retinal vascularization (41). Our study showed a trend of decreases in the levels of 8,9-EET as well as 11,12-EET in diabetic, compared with normal, CECs. Local administration of EETs accelerates wound epithelialization and neovascularization in a mouse ear wound model (42). Our study, however, suggested that in the normoglycemic mice, lowering sEH had no effect on epithelial wound closure, suggesting that supplementing extra EETs may not have beneficial effects on wound healing in normoglycemic corneas. Given the fact that EETs exhibit potent protective effects, including anti-inflammation, we speculate that overexpression of EPHX2 may contribute to low-grade inflammation in some tissues, such as the kidney and the cornea, but not in others, such as the liver; although it has been linked to the hepatic inflammatory response in fatty liver disease (37).

In addition to being potentially involved in tissue inflammation, our study also suggests that the elevated EPHX2 expression may contribute to the delayed epithelium wound closure and sensory nerve regeneration in diabetic corneas. Most studies of EPHX2 focused on its effects on injuries associated with endothelia, including ischemic cardiomyopathy (43), vascular remodeling (44), and renal injury during diabetes (20), and associated with hypertension (28). The cornea is an avascular tissue, and as such, the effects of diabetes-induced upregulation on impaired wound healing are likely due to the direct effects on CECs. However, it is not clear whether the defects were due to the general health affected by a decrease in EETs, the increase in the ratio of EETs to DHETs in epithelial cells, or the defects in the wound response. The fact that sEH inhibitors accelerated epithelial wound closure in diabetic corneas suggests that the cellular levels of EETs play a beneficial role for wound closure.

Although most studies show similar beneficial effects of Ephx2 deficiency and pharmacological inhibition of sEH, Ephx2 depletion has been shown to worsen angiotensin II–induced cardiac dysfunction since it aggravated myocardial fibrosis and increased cardiac inflammation (45). The adverse effects of Ephx2 depletion observed on cardiac fibrosis may be related to the depleted lipid phosphatase and sEH activities (45). In our model, both Ephx2 deletion and sEH inhibition exhibited similar effects on epithelial wound closure, indicating the importance of sEH activity or the cellular concentrations of EETs in mediating the epithelial wound response, which was impaired in diabetic tissues.

Our study also revealed a correlation between sEH activity and the expression of HO-1, a stress-inducible protein with a potential anti-inflammatory effect. HO-1 has been shown to play an important role in skin injury and wound healing (46). The induction of HO-1, the rate-limiting enzyme in heme degradation, represents a key event in cellular responses to pro-oxidative and proinflammatory insults (47). In the cornea, the increased expression of HO-1 modulates inflammation and promotes wound closure (48). HO-1 and EETs were found to influence each other’s expression and to attenuate diabetes and metabolic syndrome complications (49). Our study for the first time showed that hyperglycemia suppressed the wound-induced expression of HO-1, suggesting a potential contribution of this defect to delayed wound healing in diabetic corneas. Importantly, application of CoPP induced robust expression of HO-1, resulting in significant improvement of corneal epithelium wound closure. Whether EETs and HO-1 work synergistically, independently, or mutually exclusively in promoting corneal wound healing in diabetic corneas remains elusive and warrants further investigation.

Our previous studies showed that diabetes caused defects in sensory nerve structure and function in the corneas. Using an Ephx2-deficient mouse model, we showed that although no differences in induction of diabetes by STZ were found between WT and Ephx2^−/− mice, the hyperglycemia-induced sensory nerve degeneration was prevented in Ephx2^−/− mice. In humans, almost half the patients who undergo laser in situ keratomileusis (LASIK) experience dry eye after the procedure (50). It is believed that the alteration of corneal nerves after LASIK is the most likely cause of the subjective symptoms of LASIK-induced dry eye. The corneal sensitivity and the clinical indicators of dry eye usually start to improve over the first postoperative month but require a year to normalize due to the partial recovery of the corneal nerve plexus (50). Hence, regeneration of the subbasal nerve plexus is a slow process. To that end, we examined the subbasal nerve plexus at the center of the cornea 1 month post–epithelium debridement and found slower recovery of sensory nerve endings in the diabetic mouse corneas compared with normal controls in WT but not Ephx2^−/− mice, suggesting sustained positive effects of EETs on nerve regeneration under pathogenic conditions. Thus, exogenous EETs or inhibition of sEH activity may be used as long-term therapy to help functional recovery of sensory nerves in patients with diabetes.

Funding. The authors acknowledge support from the NIH/National Eye Institute (NEI) (R01EY10869 and EY17960 to F.-S.Y.), NEI Core (to WSU; p30 EY04078), and Research to Prevent Blindness (to Kresge Eye Institute). The WSU Lipidomics Core Facility is supported by a grant from the NIH National Center for Research Resources (S10RR027926).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. H.S., P.L., and C.Y. performed laboratory testing and edited and checked accuracy of the manuscript. N.G. performed laboratory testing. J.W. and X.F. contributed to the discussion and reviewed and revised the manuscript. F.-S.Y. was responsible for study design and recruitment, contributed to sample collection and data analysis, and reviewed and edited the manuscript. F.-S.Y. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. This study was presented at the Annual Meeting of ARVO, Seattle, WA, 1–5 May 2016.