Growth State-Dependent Expression of Arachidonate Lipoxygenases in the Human Endothelial Cell Line EA.hy926

Endothelial cells regulate vascular homeostasis through the secretion of various paracrine molecules, including bioactive lipids, but little is known regarding the enzymes responsible for generating these lipids under either physiological or pathophysiological conditions. Arachidonate lipoxygenase (ALOX) expression was therefore investigated in confluent and nonconfluent EA.h926 endothelial cells, which represent the normal quiescent and proliferative states, respectively. mRNAs for ALOX15, ALOX15B, and ALOXE3 were detected in EA.hy926 cells, with the highest levels present in confluent cells compared to nonconfluent cells. In contrast, ALOX5, ALOX12, and ALOX12B mRNAs were not detected. At the protein level, only ALOX15B and ALOXE3 were detected but only in confluent cells. ALOXE3 was also observed in confluent human umbilical artery endothelial cells (HUAEC), indicating that its expression, although previously unreported, may be a general feature of endothelial cells. Exposure to laminar flow further increased ALOXE3 levels in EA.hy926 cells and HUAECs. The evidence obtained in this study indicates that proliferative status and shear stress are both important factors that mediate endothelial ALOX gene expression. The presence of ALOX15B and ALOXE3 exclusively in quiescent human endothelial cells suggests their activity likely contributes to the maintenance of a healthy endothelium.


Introduction
The endothelial cells lining the inner surface of blood vessels are responsible for regulating vascular homeostasis [1], which includes modulation of vascular tone and blood flow, movement of nutrients and waste products between the blood and the underlying tissues, thrombogenicity, hemostasis, and inflammation [2]. To maintain this homeostatic balance, endothelial cells must be able to rapidly respond to various physiological signals. Endothelial cells are also major players in the processes that are activated when vascular tissue is damaged by pathophysiological stressors, such as oxidant damage, vascular surgical procedures, hyperlipidemia, hypertension, and disturbed blood flow [3,4]. Under these conditions, endothelial cells respond by modulating their underlying properties to enable migration, proliferation, and extracellular matrix remodeling [5]. This alteration of growth state, from quiescent to proliferating, which has been termed activation or endothelial to mesenchymal transition (EndMT), is essential for efficient repair of injured vascular tissue [6,7]. Growth state also plays a role in the response to changes in blood flow and remodeling of the blood vessel [8,9]. Endothelial cell activation results in various physical changes to the cell that affect its adhesion and barrier properties, as well as alterations in the 1:4 subculture of a confluent plate of cells. THP-1 cells were grown in RPMI-1640 medium supplemented with 0.05 mM 2-mercaptoethanol and 10% FBS [33]. Pooled human umbilical artery endothelial cells (HUAECs) were purchased commercially (Catalog number: C-12202; Cedarlane, Burlington, ON, Canada) and grown in Vascular Cell Basal Media (Catalog number: PCS-100-030; ATCC) supplemented with Endothelial Cell Growth Kit-VEGF (Catalog number: PCS-100-041; ATCC). The HUAECs were subcultured at 80% confluency and used within 3-5 passages.

RNA Extraction, cDNA Synthesis, and Reverse-Transcription PCR (RT-PCR)
Total RNA isolation, first-strand cDNA synthesis, and RT-PCR were performed in a 20 µL reaction mix as previously described [34]. The RT-PCR products were subsequently visualized after agarose gel electrophoresis. Primer sequences are provided in Table 1.

Dicer-Substrate siRNA (DsiRNA) Mediated Knockdown
Three 27 nucleotide siRNA duplexes specific for ALOXE3 (Table 1) and a TYE-563 tagged transfection control were purchased (DsiRNA TriFECTa Kit; Integrated DNA Technologies, Coralville, IA, USA) and reconstituted according to the manufacturer's protocol. Transfection of the DsiRNAs into EA.hy926 cells was achieved using Lipofectamine Mes-sengerMAX Reagent (ThermoFisher Scientific, Waltham, MA, USA). The transfected cells were incubated for 48 h, at which time brightfield and fluorescent images were captured with an Axiovision 4.1 Zeiss microscope. Cells were subsequently harvested, and ALOXE3 was measured.

Western Blotting
The proteins-of-interest were analyzed by immunoblotting after SDS-PAGE separation of total proteins and transfer to nitrocellulose membrane. Relative quantification was obtained using ImageJ Software (version 1.48) [35], as previously described [36]. Table 1 lists the primary antibodies used in this study.

Endothelial Cell Perfusion for Simulation of Laminar Flow
Exponentially growing EA.hy926 cells or HUAEC were seeded into the µ-Slide I Luer channel slides (ibidi USA Inc, Fitchburg, WI, USA) with a channel height of 0.4 mm, as per the manufacturer's instructions. The cell seeding parameters were adjusted so confluency in the µ-Slide I Luer channel slides was reached within 24 h. Subsequently, the µ-Slide I Luer channel slides were individually connected to the perfusion system and attached to the ibidi fluidic unit and pump. The flow rate (ϕ: mL/min) in the µ-Slide I Luer slides was held at a constant 32 mL/min to achieve a shear force (τ: dyn/cm 2 ) of 30 Dyn/cm 2 based on the formula τ = η × 131.6 × ϕ, where η = dynamical viscosity (Dyn-s/cm 2 ). The viscosity of the cell culture medium with 10% serum was considered to be 0.0072 Dyn-s/cm 2 . The cells were exposed to flow for 24 h while the control cells were maintained under static conditions over the same period. The cells were imaged via brightfield microscopy, then harvested for Western blotting.

Statistical Analysis
Statistical analysis was performed using Prism version 7.00 (GraphPad Software, San Diego, CA, USA). The means of more than two groups were compared using oneway ANOVA (randomized), followed by Dunnett's post hoc multiple comparison test to compare the means of multiple experimental groups against a control group mean. Comparisons between two groups were performed using Student's t-test (unpaired). Differences were considered significant at p < 0.05.

Results
3.1. RT-PCR Detects ALOXE3, ALOX15, and ALOX15B mRNAs in EA.hy926 Cells ALOX gene expression by EA.hy926 endothelial cells was profiled using reverse transcription polymerase chain reaction (RT-PCR) with primers (Table 1) designed to specifically amplify mRNA transcribed from the 5 ALOX genes (ALOXE3, ALOX12, ALOX12B, ALOX15, ALOX15B) clustered on human chromosome 17p13.1, as well as the ALOX5 gene located on chromosome 10q11.21. RNA was isolated from EA.hy926 cells that were 10-20% and 100% confluent, conditions that we have previously shown represent proliferative and quiescent endothelial cells, respectively [32]. Agarose gel electrophoresis of the RT-PCR products revealed the presence of ALOXE3, ALOX15, and ALOX15B mRNA in 100% confluent EA.hy.926 cells ( Figure 1A), while ALOX5, ALOX12, and ALOX12B were not detected. Interestingly, no ALOX amplification products were obtained with RNA isolated from the 10-20% confluent EA.hy926 cells ( Figure 1B). The absence of ALOXE3/ALOX15/15B-specific amplification products in the non-confluent EA.hy926 cells suggests growth state, possibly linked to cell contact, may modulate ALOX expression in this cell type. In comparison, THP-1 cells expressed ALOX5 and ALOX15 ( Figure 1C), and HepG2 cells expressed ALOXE3, ALOX5, and ALOX15 ( Figure 2D). THP-1 and HepG2 were chosen as references for comparison with EA.hy926 cells because the Human Protein Atlas (HPA) transcriptome data archive describes positive ALOX expression in these cell lines [27]. These data, along with the results presented in our previous study [34], validate the primer pairs designed for assessing expression of these genes, and thus confirm that ALOXE3, ALOX15, and ALOX15B mRNA were only detected in confluent EA.hy926 endothelial cells. The complete ALOX gene expression profile for the various cell types examined is summarized in Figure 1E. 1 and HepG2 were chosen as references for comparison with EA.hy926 cells because the Human Protein Atlas (HPA) transcriptome data archive describes positive ALOX expression in these cell lines [27]. These data, along with the results presented in our previous study [34], validate the primer pairs designed for assessing expression of these genes, and thus confirm that ALOXE3, ALOX15, and ALOX15B mRNA were only detected in confluent EA.hy926 endothelial cells. The complete ALOX gene expression profile for the various cell types examined is summarized in Figure 1E.   The HPA transcriptome database was screened for cancer cell lines expressing highlevels of ALOX15 to use as positive controls for ALOX15 protein expression. High ALOX15 mRNA levels were identified in metastatic breast tumor-derived epithelial-like MCF7 cells [40]. MDA-MB231 cells, which were established from pleural effusions from breast cancer patients [41], were also used for this purpose. Immunoblotting using anti-ALOX15 antibody revealed the presence of two immunoreactive bands in the 63-75 kDa range in lysates from both MCF7 and MDA-MB231 cells ( Figure 2B).

Differential Cadherin (CDH5) Expression Distinguishes the Activated (Non-Confluent) and Quiescent (Confluent) EA.hy926 Cell Growth States
To confirm that differences in growth state influence ALOX expression in EA.hy926 endothelial cells, immunoblotting for proteins capable of distinguishing the activated versus the quiescent state was performed. Previously, sub-confluent and confluent EA.hy926 cells were characterized in terms of DNA synthesis and the presence of cell division and adherent junction (CDH5)-associated protein [32], an endothelial-specific adhesion molecule located at adherens junctions [37]. Bright-field images ( Figure S1A) provided visual evidence of significant cell-cell contact in the confluent cell cultures, whereas non-confluent cells showed substantially fewer contacts. The level of CDH5 was subsequently examined to confirm the state of the cell-cell contact, since the abundance CDH5 was previously shown to be lower in sub-confluent cells compared to confluent cells. We reported that the relative abundance of CDH5 gradually increases following initial plating of the cells, reaching a peak after 7 days of culture [32]. In accordance with these observations, CDH5 levels were much higher in confluent versus non-confluent EA.hy926 cells ( Figure S1B). Likewise, CDH13 (T-Cadherin) was markedly higher in confluent cells ( Figure S1B). Furthermore, while the total protein profile showed distinct differences between confluent and non-confluent cells ( Figure S1B) indicative of significant differences in the protein content of these cell states, the total protein levels were similar, thus confirming that CDH5 and CDH13 were significantly reduced in non-confluent cells. Overall, these results verify that ALOX mRNA levels correlate with the degree of EA.hy926 cell confluency, thus suggesting cell-cell contact and cell growth state are important factors for modulating ALOX gene activity.

ALOX15B mRNA, but Not ALOX15 mRNA, Is Translated in EA.hy926 Cells
Immunoblotting with an anti-ALOX15 antibody was used to examine the protein levels of this enzyme in confluent EA.hy926 cells, the condition under which its mRNA was detected. Surprisingly, the anti-ALOX15 antibody (MA5-25653, Invitrogen) failed to amplify a band in EA.hy926 cells (Figure 2A). Since ALOX15 has been previously reported in endothelial cells [38,39], further investigation of this discrepancy was carried out. The HPA transcriptome database was screened for cancer cell lines expressing highlevels of ALOX15 to use as positive controls for ALOX15 protein expression. High ALOX15 mRNA levels were identified in metastatic breast tumor-derived epithelial-like MCF7 cells [40]. MDA-MB231 cells, which were established from pleural effusions from breast cancer patients [41], were also used for this purpose. Immunoblotting using anti-ALOX15 antibody revealed the presence of two immunoreactive bands in the 63-75 kDa range in lysates from both MCF7 and MDA-MB231 cells ( Figure 2B).
As a positive control, C-terminal Flag-tagged ALOX15 (ALOX15-Flag) cDNA (Transcript ID: ENST00000293761.8) was synthesized and cloned into pcDNA3.1(+). This construct was transiently transfected into HEK293 cells. Anti-Flag antibody immunoblotting revealed the appearance of two bands in overexpressed HEK293 cells, a predominant 74 kDa band and a less abundant 58 kDa band ( Figure 2C). These bands were not detected in HEK293 cells transfected with an empty vector. Immunoblotting using the anti-ALOX15 antibody also detected both p74 and p58 ALOX15-Flag proteins in overexpressed HEK293 cells only, indicating that HEK293 cells do not contain native ALOX15. The theoretical molecular weight of ALOX15 (protein ID: ENSP00000293761) was calculated to be 74.8 kDa with the ExPASy Compute pI/MW tool [42]. ALOX15 has an in-frame alternative second AUG start codon that can generate a shorter 57.5 kDa protein ( Figure S2). Thus, the presence of the 74 kDa and 58 kDa proteins in ALOX15-Flag overexpressed cells indicates that both start codons may be used during translation, however, the relative band intensities suggest that usage of the alternative second start codon is less than the first start codon. Overall, these results indicate that ALOX15 protein is not produced in confluent EA.hy926 cells.
Since the vascular localization of ALOX15B is predominantly associated with immune cells [43], the presence of ALOX15B mRNA in EA.hy926 cells was not expected. Immunoblotting was therefore used to determine if ALOX15B protein was also present in these cells. However, due to the similarities between ALOX15 and ALOX15B, the specificity of the anti-ALOX15B antibody was examined. C-terminal Flag-tagged ALOX15B (ALOX15B-Flag) cDNA (Transcript ID: ENSP00000369520) was synthesized and cloned into pcDNA3.1(+), then transiently transfected into HEK293 cells. No bands were detected in untransfected HEK293 cells using either anti-FLAG or anti-ALOX15B antibodies, indicating the native protein is not present in these cells ( Figure 3A), in agreement with the HPA database. However, the anti-ALOX15B antibody (sc-271290, Santa Cruz Biotechnology) detected 3 bands at 73 kDa, 67 kDa, and 51 kDa ( Figure 3A). These bands were distinct from the 58 kDa and 73 kDa bands detected with anti-ALOX15 antibody. Since the ALOX15 and ALOX15B bands were also unique according to the anti-FLAG antibody ( Figure 3A), it could be confidently concluded that both proteins can be distinguished by immunoblot analysis.
The 73 kDa ALOX15B band corresponds to the theoretical molecular weight of ALOX15B (72.5 kDa) calculated with the ExPASy Compute pI/MW tool [42]. However, ALOX15B has in-frame alternative second and third AUG start codons that can generate 56.5 and 50.8 kDa proteins, respectively ( Figure S3). The presence of p73, p67, and p51 bands in ALOX15-Flag overexpressed cells indicated that all three start codons are used during translation, although it would appear that the second alternative start codon was the most active based on their relative band intensities ( Figure S3A). Of note, the 57 kDa ALOX15B-Flag appeared at 67 kDa with respect to the molecular weight markers, likely due to the presence of post-translational modifications.
When applied to lysates of EA.hy926 cells, the ALOX15B antibody visualized a single band of 67 kDa ( Figure 3B) in samples from confluent cells, which was absent from the non-confluent cell samples. While these results might indicate only one start codon is active in endothelial cells, the weak band intensity suggests the cellular content of ALOX15B is low and the presence of the 73 and 51 kDa bands might be below the limit of detection. However, based on these results, it is reasonable to suggest that the ALOX15B mRNA in confluent EA.hy926 cells is translated into protein. However, based on these results, it is reasonable to suggest that the ALOX15B mRNA in confluent EA.hy926 cells is translated into protein.

Identification of ALOXE3 Protein in Endothelial Cells
Unlike other ALOXs, ALOXE3 has never been reported in endothelial cells. In order to corroborate the presence of ALOXE3 protein, immunoblotting was selected as the most efficient method. An approach to validating the ALOXE3 antibody similar to that of ALOX15 and ALOX15B was considered, but the ALOXE3 antibody (PA5-21833, Invitrogen) detected 3 bands (70, 100 and 150 kDa) in HEK293 cells ( Figure S4A), which is not surprising since the HPA database reports ALOXE3 is expressed in these cells. However, since the antibody also recognized a similar trio of bands in EA.hy926 cells (not shown), use of transient transfection for validation was considered unfeasible. Consequently, an mRNA knockdown strategy was employed to establish which band(s) represent ALOXE3,

Identification of ALOXE3 Protein in Endothelial Cells
Unlike other ALOXs, ALOXE3 has never been reported in endothelial cells. In order to corroborate the presence of ALOXE3 protein, immunoblotting was selected as the most efficient method. An approach to validating the ALOXE3 antibody similar to that of ALOX15 and ALOX15B was considered, but the ALOXE3 antibody (PA5-21833, Invitrogen) detected 3 bands (70, 100 and 150 kDa) in HEK293 cells ( Figure S4A), which is not surprising since the HPA database reports ALOXE3 is expressed in these cells. However, since the antibody also recognized a similar trio of bands in EA.hy926 cells (not shown), use of transient transfection for validation was considered unfeasible. Consequently, an mRNA knockdown strategy was employed to establish which band(s) represent ALOXE3, and HEK293 cells were used for this purpose since they are much easier to transfect than EA.hy926 cells.
Dicer-Substrate siRNAs (DsiRNAs), which are chemically synthesized 27mer RNA duplexes, exhibit RNA knockdown potency up to 100-fold more than conventional 21mer siRNAs [44]. Three sets of DsiRNA, each targeting a different exon of the ALOXE3 gene ( Figure 4A), were designed and transfected into HEK293 cells. Immunofluorescence images of the fluorescence-conjugated control DsiRNA (Dsi NC) revealed high transfection efficiency was achieved using lipofectamine ( Figure 4B). Immunoblotting for ALOXE3 revealed knock down by the three DsiRNAs only affected the 100 kDa band ( Figure 4C), while the levels of the 70 and 150 kDa bands remained constant. These results indicate that the 70 and 150 kDa immunoreactive bands are non-specific and that the 100 kDa immunoreactive band represents the ALOXE3 protein. Quantification of the 100 kDa band intensities after normalization to GAPDH revealed that DsiRNA3 was the most effective in knocking down ALOXE3 ( Figure 4D). Interestingly, the identification of the 100 kDa band as ALOXE3 (protein ID: ENSP00000314879) suggests it may be post-translationally modified since the apparent molecular mass is higher than the predicted size of 80 kDa, based on calculations with the ExPASy Compute pI/MW tool [42]. However, use of an in-frame alternative second AUG start codon may cause translation of a 76.2 kDa (theoretical) ALOXE3 protein ( Figure S4B), even though transcriptional isoforms are not listed in the Ensembl human genome assembly (GRCh38.p13) database [45].   Immunoblotting for ALOXE3 detected the three immunoreactive bands (70, 100, and 150 kDa) in confluent EA.hy926 cells observed in HEK293 cells but failed to detect the 100 kDa band representative of ALOXE3 in non-confluent EA.hy926 cells ( Figure 5A). These results correlated with the observed differences in the levels of ALOXE3 mRNA under these same growth conditions ( Figure 1A,B). Based on these observations, it was concluded that ALOXE3 is only expressed in quiescent EA.hy926 endothelial cells.
To verify that the results obtained with EA.hy926 cells were representative of endothelial cells, ALOXE3 protein levels were also measured in primary human endothelial cells derived from umbilical cord arteries (HUAEC). As was seen with EA.hy926 cells, immunoblotting detected ALOXE3 protein in confluent HUAEC cells ( Figure 5A). Furthermore, the relative abundance of ALOXE3 in HUAEC cells was similar to that of EA.hy926 cells when normalized relative to total protein ( Figure 5B). Thus, ALOXE3 is likely normally expressed in quiescent endothelial cells but not when they are in the activated state.

Shear Force Influences the Levels of ALOXE3 Protein in Endothelial Cells
In addition to cell contact, endothelial cell growth and activation can be influenced by blood flow. Specifically, shear stress can induce endothelial cell quiescence much like cell contact [46]. To see if ALOXE3 expression can be modulated by shear stress, confluent EA.hy926 were exposed to shear forces of 0, 8, 14, and 30 Dyn/cm 2 for 24 h, which approximate the physiological forces observed in veins, ascending aorta, and arteries, respectively [9]. Brightfield microscopy revealed that application of shear force for 24 h led to a loss of confluency and formation of isolated patches of 4-20 cells ( Figure 6A). This occurred under all conditions, however, only 30 Dyn/cm 2 is shown. In contrast, HUAECs exposed to 30 Dyn/cm 2 (lower shear forces were not tested) remained confluent ( Figure 6A). These results suggest that there is a cell-type-specific response with respect to cell-cell contact and adherence under shear stress. This difference may also be due to the transformed somatic hybrid nature (lung epithelial and endothelial) of EA.hy926 cells compared to the primary HUAEC. However, immunoblotting revealed that ALOXE3 levels increased in response to shear force in both EA.hy926 cells and HUAEC, regardless of the force applied ( Figure 6B). Quantification of the immunoblots revealed ALOXE3 was significantly increased by 3.2-fold under shear stress conditions ( Figure 6C). These novel findings indicate that the abundance of ALOXE3 protein in endothelial cells is responsive to both shear stress and cell growth state. To verify that the results obtained with EA.hy926 cells were representative of endothelial cells, ALOXE3 protein levels were also measured in primary human endothelial cells derived from umbilical cord arteries (HUAEC). As was seen with EA.hy926 cells, immunoblotting detected ALOXE3 protein in confluent HUAEC cells ( Figure 5A). Furthermore, the relative abundance of ALOXE3 in HUAEC cells was similar to that of EA.hy926 cells when normalized relative to total protein ( Figure 5B). Thus, ALOXE3 is likely normally expressed in quiescent endothelial cells but not when they are in the activated state. formed somatic hybrid nature (lung epithelial and endothelial) of EA.hy926 cells compared to the primary HUAEC. However, immunoblotting revealed that ALOXE3 levels increased in response to shear force in both EA.hy926 cells and HUAEC, regardless of the force applied ( Figure 6B). Quantification of the immunoblots revealed ALOXE3 was significantly increased by 3.2-fold under shear stress conditions ( Figure 6C). These novel findings indicate that the abundance of ALOXE3 protein in endothelial cells is responsive to both shear stress and cell growth state.

Discussion
This is the first study to comprehensively profile ALOX gene expression in endothelial cells. The evidence indicates that the ALOX15, ALOX15B, and ALOXE3 genes, and no others, are transcribed in EA.hy926 cells. However, at the protein level, only ALOX15B and ALOXE3 are detected. Furthermore, ALOX15B and ALOXE3 are found primarily in confluent endothelial cells relative to non-confluent cells. Since ALOXE3 has not been previously reported in vascular tissues, its presence in HUAEC was examined as confirmation that this protein is also found in primary endothelial cells. These data are the first to imply that ALOXE3 has a function in quiescent endothelial cells, which is the state these cells normally occupy in healthy blood vessels. However, the fact that ALOXE3 levels are increased in both EA.hy926 cells and HUAEC after exposure to flow suggests cell-cell contact and shear stress may independently regulate this gene. Finally, it is unlikely that the ALOX-derived bioactive lipids present in the circulation are a result of endothelial cell activity since the major products detected in plasma and serum originate via ALOX5 and ALOX15 [47,48], which are not detectable at the protein level in either growing or confluent endothelial cell populations.
The biological function of the ALOX enzymes is to synthesize a variety of bioactive lipids, primarily from PUFA, that modulate the response of many different cells and tissues.
For this reason, production of these lipid mediators is sensitive to both the species of PUFA available as a substrate and the amount of enzyme in a cell. With respect to endothelial cells, ALOX-derived bioactive lipids can influence many functions, including cell proliferation, endothelial-monocyte interactions, barrier function, cytokine production, and vasoreactivity [49,50]. While their effects on endothelial cells have been well documented, it still remains uncertain which cells are the source of these bioactive lipids. To date, localization of ALOX within the vasculature has been examined indirectly, with conclusions usually based on a reported association between a given lipid species and the ALOX presumed to be responsible for its synthesis [38,51,52]. However, while this approach is often employed, an individual ALOX produces multiple products including those that have been linked to other ALOX isoforms [43]. Likewise, the application of inhibitors for the same purpose [23] is limited due to uncertainties regarding inhibitor specificity [53]. At the same time, direct examination of ALOX expression by endothelial cells, whether by antibody staining for protein or PCR amplification for mRNA, has been sparse [39,54]. Furthermore, the cell culture conditions employed vary between the published studies, which makes meaningful comparisons difficult. Thus, more specific information is essential if an understanding is to be achieved regarding the endothelial origin of mediators that contribute to vascular function.
A major strength of the current study is its use of cells that occupy two well characterized growth states that have physiological relevance [32]: subconfluent endothelial cells that are actively growing, and confluent endothelial cells that are quiescent. In agreement with our previous report [32], confluent cells exhibit higher levels of endothelial-specific markers, resulting in a protein profile characteristic of the endothelial phenotype (see Supplemental Figure S1). In comparison, growing cells express mesenchymal markers indicating they have undergone partial EndMT and are thus in the activated state [55]. Given this ability to manipulate the cell characteristics by eliciting a transition between quiescent and activated endothelial cells, it has been possible to establish that expression of ALOX15B and ALOXE3 is sensitive to phenotype switching. This is the first indication that either ALOX might have a role in endothelial function, although RNA for both ALOX15B and ALOXE3 was identified in endothelial cells of a limited number of tissues by the HPA database. At the same time, the absence of the other ALOXs from endothelial cells raises questions regarding their contribution to the endothelial function.
While the ALOX15 protein was not detected in either growth state, in agreement with our finding that endothelial cells are not sensitive to the effects of the ALOX15 inhibitor PD146176 [56], the fact that its mRNA is present in quiescent endothelial cells suggests there may be conditions that will lead to its translation. Pertinent to this study, the earliest reports of endothelial expression of ALOX15 included IL-4 in the culture medium [54], and it was subsequently shown that IL-4 induces ALOX15 expression in confluent cells [39]. It may thus be speculated, based on the proliferation-inducing effects of ALOX15 and its metabolites [57], that the presence of ALOX15 mRNA in confluent cells provides a means of rapid endothelial cell activation in response to vascular damage without the need to wait for induction of its gene.
Unlike ALOX15, which is broadly expressed, ALOX15B mRNA and protein appear to be restricted to a small number of tissues [27]. Interestingly, even though ALOX15B and ALOX15 generate the same products, albeit in different proportions [58], the evidence indicates that ALOX15B suppresses cell growth [59], particularly as it relates to the prostate [60,61]. However, to date, the vascular presence of ALOX15B has largely been linked to macrophages [62], which would explain its localization in atherosclerotic lesions [63,64]. It is also possible that endothelial ALOX15B may have been misidentified as ALOX15 since the discovery of ALOX15B is more recent and antibody specificity would not have been validated. Regardless, it is now possible to state that ALOX15B appears to be regulated by growth state, however, further investigation to determine whether there is an inverse relationship between ALOX15B and ALOX15 as suggested by these results will require a more sensitive detection method.
The major outcome of this investigation is the discovery that endothelial cells express ALOXE3, an isoform first identified in epithelial cells of the skin [25,65,66]. However, unlike the other ALOX enzymes, PUFAs are not substrates for ALOXE3. Rather, ALOXE3, originally identified as hepoxilin synthase, catalyzes the conversion of ALOX12 and ALOX15 products into hepoxilins [66,67], which are then rapidly metabolized into trioxilins. Since hepoxilins have been shown to affect multiple cellular functions analogous to those active in the endothelium [68], ALOXE3 may contribute to a variety of endothelial processes, including modulation of tight junctions. For instance, products of ALOX15 (15-HETE) and ALOX12 (12-HETE) disrupt endothelial tight junctions [49,69] and increase vascular permeability [70]. In the skin, production of hepoxilin (HXA3) by ALOXE3 from the ALOX12 product 12-HpETE is essential for epidermal barrier formation [66]. ALOXE3 may have a similar role in endothelial cells, which would be to divert metabolism of 15-HpETE and 12-HpETE from barrier-disrupting 15-HETE and 12-HETE to hepoxilins that help preserve endothelium integrity. Based on our results, ALOXE3 could work in tandem with ALOX15B to control endothelial barrier function, as it does with ALOX12 in epithelial cells [65,71], since the available evidence indicates aortic lysates produce hepoxilins only when ALOX12 is included in the reaction system [72].
While the link between ALOXE3 and other endothelial functions has not been investigated directly, it has been established that hepoxilins and/or trioxilins induce vasodilation, prevent platelet aggregation, suppress endothelial-monocyte adhesion, inhibit fibrosis, and suppress inflammation [73][74][75]. In contrast, ALOX12 and ALOX15 products (12-HETE and 15-HETE, respectively) enhance monocyte adhesion [76,77], cause vasoconstriction, and induce inflammation [78]. These opposing actions of 12-/15-HETE and the hepoxilins enable us to extrapolate that ALOXE3 is essential for the lipid mediator class switch that occurs during initial inflammatory and terminal resolution stages of the vascular response to injury [11], wherein endothelial cell phenotype shifts from its quiescent state to the active state and back again. It is likely that this dichotomy also explains why the contribution of ALOX12 to vascular function can be different from the effects of 12-HETE treatment [79], as well as explain why ALOX12 has been positively linked to sex-dependent cardioprotection [47].
An examination of the intracellular pathways that regulate ALOX15 and ALOXE3 levels in endothelial cells has not been included in this study. In large part, this deficiency is due to the lack of information currently available regarding lipoxygenase expression, and in the absence of a credible target such an investigation would be impractical. However, based on the findings of this study, some speculation on this topic is possible. Shear stress as a result of laminar flow is a critical factor in maintaining the endothelial state. Specific mechanosensitive ion channels, such as PIEZO1, convey the mechanical effects of flow to a CaMKII-ERK5 signal transduction cascade that modulates expression of the endothelialspecific transcription factors KLF-2 and KLF-4 [80]. Furthermore, recent evidence suggests that ERK5 operates by phosphorylating MEF2, the transcription factor that controls KLF-2 and KLF-4 gene activity [81]. While a role for KLF-2 and KLF-4 in ALOX15B or ALOXE3 gene expression has not yet been identified, Segre et al. [82] have reported that KLF-4 governs the epidermal barrier function, a process which also involves ALOXE3 [63]. Conversely, disturbed flow leads to a reduction in KLF-2 and KLF-4 and conversion to the activated state [83], where ALOX15B and ALOXE3 are lacking (Figure 7). Consequently, as described above, 12-HpETE and 15-HpETE would preferentially be metabolized to 12-HETE and 15-HETE, respectively, which disrupt barrier function via PKCε-mediated phosphorylation of ZO-1 [69]. The loss of barrier function is considered to be the first step in EndMT, which is essential for vascular repair, but it can also result in endothelial dysfunction and atherosclerosis [55]. It is likely that modes of injury other than disturbed flow, including mechanical damage, also lead to EndMT, although they would engage via a distinct set of signal transduction pathways [84]. Once repair is complete, other lipid metabolizing enzymes contribute to the resolution phase of vascular wound healing [11], and these molecules likely promote restoration of the quiescent endothelial state and re-expression of KLF-2 and KLF-4.
to be the first step in EndMT, which is essential for vascular repair, but it can also result in endothelial dysfunction and atherosclerosis [55]. It is likely that modes of injury other than disturbed flow, including mechanical damage, also lead to EndMT, although they would engage via a distinct set of signal transduction pathways [84]. Once repair is complete, other lipid metabolizing enzymes contribute to the resolution phase of vascular wound healing [11], and these molecules likely promote restoration of the quiescent endothelial state and re-expression of KLF-2 and KLF-4. Figure 7. Schematic illustrating possible relationships between the pathways that regulate endothelial cell state in relation to ALOX expression and activity. "Blocks" is presented in green because this action is beneficial, while "promotes" is colored red because this action is associated with the development of vascular disease.
ALOXs produce many different bioactive lipid mediators [85] and while the individual products exhibit distinct biological effects on endothelial cells [50] their aggregate effect is difficult to determine since it would be a function of mediator potency, relative concentrations, and the physiological state of the cell (presence/absence of specific Figure 7. Schematic illustrating possible relationships between the pathways that regulate endothelial cell state in relation to ALOX expression and activity. "Blocks" is presented in green because this action is beneficial, while "promotes" is colored red because this action is associated with the development of vascular disease. ALOXs produce many different bioactive lipid mediators [85] and while the individual products exhibit distinct biological effects on endothelial cells [50] their aggregate effect is difficult to determine since it would be a function of mediator potency, relative concentrations, and the physiological state of the cell (presence/absence of specific receptors and intracellular signaling mediators). However, while there is a good understanding of how bioactive lipid mediators derived from PUFA contribute to the onset and resolution of inflammation in response to injury, [86] little is known of the mechanisms that regulate the ALOX enzymes responsible for generating these lipid mediators.

Conclusions
The ability to model distinct endothelial cell phenotypes has provided a unique perspective regarding the contributions of ALOXs to the vasculature in both the healthy (quiescent) and activated (growing) states. Furthermore, it is evident that the distinct properties exhibited under these cell culture conditions reflect to a reasonable degree the physiology of the normal and damaged endothelium and that it is also possible to emulate the effect of flow. Altogether, the information gained through this study has made it possible to propose a new paradigm for the physiological role of ALOXs in endothelial cells, with ALOXE3 contributing to phenotype switching and thus influencing multiple endothelial cell functions through its ability to alter the cellular response to other ALOX products. Finally, establishing that ALOX15B and ALOXE3 are only expressed in quiescent endothelial cells or endothelial cells exposed to flow suggests that these enzymes may be suitable targets for therapeutic interventions intended to manage vascular inflammation. This would provide a unique approach for averting endothelial dysfunction and thus interrupting the progression to ischemic heart disease.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/cells11162478/s1, Figure S1: Marker proteins of confluent and non-confluent EA.hy926 cells; Figure S2: Alternative AUG start codon in ALOX15 gene; Figure S3: Alternative AUG start codons in ALOX15B gene; Figure S4: Immunoblotting for ALOXE3 in HEK293 cells and alternative AUG start codon in ALOXE3 gene.