ADAM10 and ADAM17 proteases mediate proinflammatory cytokine-induced and constitutive cleavage of endomucin from the endothelial surface

Contact between inflammatory cells and endothelial cells (ECs) is a crucial step in vascular inflammation. Recently, we demonstrated that the cell-surface level of endomucin (EMCN), a heavily O-glycosylated single-transmembrane sialomucin, interferes with the interactions between inflammatory cells and ECs. We have also shown that, in response to an inflammatory stimulus, EMCN is cleared from the cell surface by an unknown mechanism. In this study, using adenovirus-mediated overexpression of a tagged EMCN in human umbilical vein ECs, we found that treatment with tumor necrosis factor α (TNF-α) or the strong oxidant pervanadate leads to loss of cell-surface EMCN and increases the levels of the C-terminal fragment of EMCN 3- to 4-fold. Furthermore, treatment with the broad-spectrum matrix metalloproteinase inhibitor batimastat (BB94) or inhibition of ADAM metallopeptidase domain 10 (ADAM10) and ADAM17 with two small-molecule inhibitors, GW280264X and GI254023X, or with siRNA significantly reduced basal and TNFα-induced cell-surface EMCN cleavage. Release of the C-terminal fragment of EMCN by TNF-α treatment was blocked by chemical inhibition of ADAM10 alone or in combination with ADAM17. These results indicate that cell-surface EMCN undergoes constitutive cleavage and that TNF-α treatment dramatically increases this cleavage, which is mediated predominantly by ADAM10 and ADAM17. As endothelial cell-surface EMCN attenuates leukocyte–EC interactions during inflammation, we propose that EMCN is a potential therapeutic target to manage vascular inflammation.

Interactions between inflammatory cells and endothelial cells (ECs), 4 a crucial step in vascular inflammation, can result in leukostasis and capillary occlusion (5) as well as migration of inflammatory cells across the endothelium into tissues, leading to tissue damage (6). It has been well documented that this multistep process is mediated by adhesion molecules and their binding partners on inflammatory cells and ECs that are induced by inflammatory stimuli (7).
It has been reported that, coincident with increased expression of adhesion molecules on the luminal endothelial cell surface during inflammation, there is a "thinning" of the glycocalyx, a layer of glycoproteins, proteoglycans, glycosaminoglycans, and associated plasma proteins (8 -10). The glycocalyx has been speculated to form a barrier that prevents adhesion of circulating blood cells to the quiescent endothelium layer, presumably because of the repulsive properties of the negatively charged dense sugar epitopes (11). Moreover, thinning of this glycocalyx layer is associated with release of inflammatory cytokines, chemokines, and their receptors and, more importantly, exposure of adhesion molecules on the surface of the endothelium to their binding partners on inflammatory cells (12).
Endomucin (EMCN), a component of the glycocalyx, is a single-transmembrane, heavily O-glycosylated sialomucin expressed by endothelial cells, specifically by capillaries and venous but not most arterial ECs (13,14). Recently, we demonstrated that the presence of EMCN on the endothelial surface prevents interactions between inflammatory cells and quiescent, noninflamed ECs. Moreover, we have shown that stimulation of ECs with inflammatory cytokines such as tumor necrosis factor ␣ (TNF␣), which is known to induce expression of proadhesion molecules on the endothelium, also leads to a significant decrease in apically localized EMCN as well as a decline in EMCN mRNA. Together, these changes facilitate contact between circulating leukocytes and the inflamed endothelium (15). The finding that EMCN overexpression in vitro or in vivo can suppress leukocyte-EC interactions under inflammatory conditions suggests that preservation of cell-surface EMCN may be a novel approach to regulating inflammation. However, the mechanism by which cell-surface EMCN levels are regulated is unknown and, thus, the topic of this investigation.
Matrix metalloproteinase (MMP) superfamily members, including MMPs and a disintegrin and metalloproteinase (ADAM), are secreted or membrane-bound Zn 2ϩ -dependent proteinases. One of the most well-studied actions of these enzymes is their ability to catalyze cleavage of the ectodomain of transmembrane proteins and reduce their cell surface level. Here we used overexpression of a tagged EMCN in conjunction with inhibitors against pan-MMPs, inhibitors, or siRNA against ADAM10 and/or ADAM17 to investigate whether EMCN is cleaved and whether these enzymes are involved in its cleavage from the endothelial surface. The results of these studies provide mechanistic insights into the posttranslational regulation of EMCN.

Treatment of EC with TNF␣ or pervanadate reduced EMCN cell surface levels
Treatment of HUVECs for 24 h with TNF␣, a potent inflammatory cytokine, at a dose as low as 0.1 ng/ml led to a dose-de-pendent reduction in total EMCN protein and mRNA levels ( Fig. 1, a and b). TNF␣ at 1 ng/ml reduced cell-surface EMCN by ϳ85% (Fig. 1c). The concentrations of TNF␣ here are within the range of plasma levels in patients with vascular inflammation (16,17), so the results highlight the relevance of cell-surface EMCN reduction in inflammation.
Pervanadate, a strong oxidant, has been employed to study the cleavage of glycocalyx components and cell-surface proteins, including syndecan-1, ICAM-1, and occludin, potentially through oxidation and inactivation of protein tyrosine phosphatase (18 -20). We therefore examined the effects of pervanadate treatment on EMCN. Pervanadate (50 M) led to an ϳ90% reduction of cell-surface EMCN protein (Fig. 1d) and an approximately 50% decrease of total EMCN protein in HUVECs after 30-min exposure (Fig. 1e). As expected, EMCN mRNA levels were not affected during the short time frame of this treatment (Fig. 1f).

Release of an EMCN C-terminal fragment (EMCN-CTF) by TNF␣ or pervanadate
It has been well documented that inflammatory stimuli such as TNF␣, interferon-␥, and lipopolysaccharide induce proteol- Figure 1. TNF␣ or pervanadate treatment leads to reduced cell-surface EMCN and down-regulation of EMCN mRNA in HUVECs. a-f, confluent HUVECs were treated with TNF␣ at the indicated concentrations for 24 h or with pervanadate (50 M) for 30 min after overnight serum starvation. Cell-surface protein was isolated by biotinylation, whereas total protein was harvested directly. Total (a and e) and cell surface (c and d) EMCN levels were determined by Western blotting. EMCN mRNA levels were evaluated by Real Time quantitative PCR (b and f). TNF␣ caused a dose-dependent reduction in total EMCN protein (a) and mRNA (b). TNF␣ treatment led to a reduction in cell-surface EMCN at 1 ng/ml (c). Pervanadate treatment resulted in decreased cell-surface EMCN (d) and decreased total EMCN protein (e) but did not affect EMCN mRNA levels (f). n ϭ 3, one-way ANOVA or Student's t test. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; ****, p Ͻ 0.0001; ns, not significant.

EMCN cleavage by ADAM10 and ADAM17 under inflammation
ysis and shedding of numerous cell-surface proteins (21)(22)(23). The ability of pervanadate to dramatically reduce cell-surface EMCN in 30 min and without affecting mRNA levels strongly suggests proteolytic cleavage of EMCN. To test this possibility, an adenoviral construct was designed to express a tagged EMCN with a myc peptide epitope at the N terminus and a DDK peptide epitope at the C terminus (Fig. 2a). The adenoviral constructs with or without tagged EMCN were transfected into HUVECs, cell lysates were harvested, and expression of the tagged EMCN was evaluated using antibodies against fulllength EMCN as well as against the specific tags. Increased total EMCN protein was observed in lysates of cells transfected with and overexpressing the tagged EMCN (Fig. 2b). The anti-myc and anti-DDK antibodies reacted with a protein from cell lysate of HUVECs transfected with the tagged EMCN construct migrating at a molecular mass similar to that of full-length EMCN because of the small size of a Myc tag and a DDK tag (ϳ1.2 kDa and ϳ1 kDa, respectively). This protein was absent from the lysates of HUVECs transfected with the empty construct (Fig. 2b), verifying expression of myc-and DDK-tagged EMCN.
To evaluate the potential cleavage of EMCN, an adenovirus was used to overexpress the tagged EMCN (AdEMCN) in cells that were then treated with TNF␣ or pervanadate. A fragment of ϳ12 kDa that corresponded to the size of the EMCN-CTF was detected with the DDK antibody in cell lysates from HUVECs transduced with AdEMCN; this fragment was not detectable in lysates from cells transduced with a control adenovirus expressing GFP (AdGFP) (Fig. 2c). The levels of the 12-kDa EMCN-CTF fragment were increased to 400% following treatment with TNF␣ (1 ng/ml, 24 h) (Fig. 2c) or to 280% with pervanadate (50 M, 30 min) (Fig. 2d) compared with untreated controls. The presence of EMCN-CTF in the overexpression system without stimuli suggests a low level of constitutive EMCN cleavage, and the increased EMCN-CTF with treatment by TNF␣ or pervanadate indicates induced cleavage of EMCN.

Pan-MMP inhibition partially rescued cell-surface EMCN reduction by TNF␣
The observation that TNF␣ treatment of HUVECs led to elevated levels of EMCN-CTF supported the hypothesis that inflammation induces cleavage of EMCN. The involvement of MMP family members in proteolytic cleavage of cell-surface proteins during inflammation has been well studied (12,24). To elucidate the mechanism of EMCN cleavage, we first

EMCN cleavage by ADAM10 and ADAM17 under inflammation
examined the effect of a broad-spectrum MMP inhibitor, BB94 (batimastat), a hydroxamate-type inhibitor of MMPs that binds competitively to the active sites of MMPs and ADAMs.
In the absence of stimulation, pretreatment of HUVECs with BB94 increased cell-surface EMCN to 150% (Fig. 3a) and total EMCN protein to 110% (Fig. 3b) of the DMSO vehicle control, respectively. Considering that BB94 alone reduced EMCN mRNA to 70% of the vehicle control ( Fig.  3d), BB94 most likely preserves cell-surface EMCN by attenuating constitutive EMCN protein cleavage. With TNF␣ stimulation, pretreatment of HUVECs with BB94 preserved cellsurface EMCN to 67% of the DMSO control ( Fig. 3a) and total EMCN protein to 60% of the DMSO control ( Fig. 3b) compared with 20% and 33%, respectively, without BB94 pretreatment. Inhibition of TNF␣-induced cell-surface EMCN reduction by BB94 pretreatment was also observed by immunostaining of cell-surface EMCN on nonpermeabilized samples (Fig. 3c). Treatment with GM6001, another broad-spectrum inhibitor of MMPs, provided similar preservation of total EMCN protein compared with the TNF␣-alone group (Fig. S1). Given that BB94 had no effect on the TNF␣-induced reduction in EMCN mRNA (Fig. 3d), the effects of BB94 were likely posttranslational. BB94 pretreatment also completely blocked the pervanadate-induced release of EMCN-CTF and reduction of total EMCN protein (Fig. S2). Together, these results support a role of members of the MMP family in not only constitutive cleavage of cell-surface EMCN but also under inflammatory conditions. Cell-surface protein (a) and total cell lysates (b) were harvested for protein analysis. Cells were fixed and stained for EMCN (c, red) or mRNA was harvested for gene expression analysis (d). BB94 blocked TNF␣-induced reduction of cell surface EMCN (a) as well as total EMCN protein (b). This was confirmed by immunostaining of EMCN (c, red). (d) BB94 pretreatment did not alter TNF␣-induced reduction in EMCN mRNA (d). Scale bar ϭ 50 m. n ϭ 3, one-way ANOVA. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; ****, p Ͻ 0.0001; ns, not significant.

EMCN cleavage by ADAM10 and ADAM17 under inflammation Inhibition of ADAM10 and ADAM17 using small-molecule inhibitors prevented loss of cell-surface EMCN
ADAMs, a family of transmembrane MMPs, have been heavily implicated in shedding/cleavage of adhesion molecules and chemokines during inflammation (24). In particular, ADAM10 and ADAM17 have been shown to mediate proteolytic cleavage of CX3CL1, VCAM-1, ICAM-1, and JAM-A in ECs (25)(26)(27)(28)(29). In our work, we noted a significant but transient increase in ADAM17 mRNA levels in TNF␣-treated HUVECs after 2 h (Fig. S3), suggesting potential involvement of ADAM17 in TNF␣-induced EMCN cleavage.
To determine whether these members of the ADAM family were involved in preserving cell-surface EMCN, GW280264X (GW) and GI254023X (GI) were used to discriminate the activities of ADAM10 and ADAM17 (30). GW inhibits ADAM10 and ADAM17, whereas GI preferentially inhibits ADAM10 (IC 50 , 5.3 nM) compared with ADAM17 (IC 50 , 541 nM). In these studies, the cells were pretreated for 30 min with GW, GI, or a DMSO control following treatment with or without TNF␣.
In the absence of TNF␣ stimulation, pretreatment with GW or GI significantly increased cell-surface EMCN protein to 231% and 208% of the vehicle control, respectively. Given that neither GW nor GI significantly altered total EMCN protein levels posttranslationally (Fig. 4b), this result supports a critical role of at least ADAM10 in preserving cell surface levels of EMCN (Fig. 4a). In the presence of TNF␣, although pretreatment with GW or GI did not influence the TNF␣-induced reduction in EMCN mRNA, it did lead to attenuated reduction in total EMCN protein levels from 28% in the TNF␣-alone group to 71% and 54% of the vehicle control, respectively (Fig.  4b). More importantly, under TNF␣ treatment, GW and GI normalized the levels of cell-surface EMCN from 61% in TNF␣alone group to 183% and 174% of the vehicle control, respectively (Fig. 4a), which was corroborated by immunostaining of cell-surface EMCN on nonpermeabilized samples (Fig. 4c). Those results suggest that inhibition of at least ADAM10 preserves cell-surface EMCN at basal levels and under inflammatory conditions. , or DMSO as a vehicle control for 30 min before TNF␣ treatment (1 ng/ml, 24 h). Cell-surface protein (a) and total cell lysates (b) were harvested for protein analysis. Cells were fixed and stained for EMCN protein expression (c, red), or mRNA was harvested for gene expression analysis (d). GW and GI blocked TNF␣-induced reduction of cell-surface EMCN protein (a) and total EMCN protein (b), which was also evident by immunostaining of EMCN (c, red). Neither GW nor GI treatment affected TNF␣-induced reduction of EMCN mRNA (d). Scale bar ϭ 50 m. n ϭ 3, one-way ANOVA. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; ns, not significant.

EMCN cleavage by ADAM10 and ADAM17 under inflammation
To determine whether the effects of GW and GI on preserving cell-surface EMCN were due to EMCN cleavage, levels of EMCN-CTF were determined in TNF␣-stimulated HUVECs overexpressing tagged EMCN with and without GW or GI pretreatment. In HUVECs overexpressing tagged EMCN, GI completely blocked TNF␣-induced release of EMCN-CTF, whereas GW significantly reduced TNF␣-induced release of EMCN-CTF from 338% to 182% of the vehicle control (Fig. S4). Although not statistically significant, without TNF␣ stimulation, there was a trend toward reduction of constitutive cleavage of EMCN by both inhibitors as well (Fig. S4). This result supports a critical role of at least ADAM10 in constitutive and TNF␣-induced cleavage of EMCN.

Inhibition of ADAM10 and ADAM17 using siRNA prevented loss of cell-surface EMCN
To define the requirement for ADAM10 and ADAM17 in TNF␣-induced EMCN ectodomain cleavage, HUVECs were transfected with siRNA (50 nM) against ADAM10 and ADAM17 alone or in combination. Successful knockdown of ADAM10 and ADAM17 was verified by mRNA and protein 24 h after transfection; combination siRNA transfection did not impact ADAM knockdown (Fig. S5, a-d). Representative Western blot images of ADAM10 and ADAM17 knockdown are shown in Fig. 5d.
Although TNF␣ stimulation reduced cell-surface EMCN levels in HUVECs transfected with siNT and siADAM17, only siADAM10 alone (28% increase) or in combination with siADAM10/siADAM17 (63% increase) resulted in significantly higher levels of surface EMCN following TNF␣ stimulation compared with siNT controls. These results support a role of ADAM10, alone or in combination with ADAM17, in regulating TNF␣-induced EMCN shedding. Furthermore, when looking at the role of ADAM10 and ADAM17 in constitutive EMCN shedding, knockdown of ADAM10 (46%) and ADAM17 (35%) alone and in combination (68%) resulted in elevated EMCN surface expression (Fig. 5a). This suggests that, although ADAM10 plays a prominent role in TNF␣-induced EMCN surface shedding, constitutive shedding of surface EMCN appears Cell-surface (a) and total (b) protein lysate was harvested for analysis using Western blotting. Total RNA was collected in RNA-Bee for quantitative PCR analysis (c). Representative Western blotting of ADAM10 and ADAM17 knockdown was performed using siRNA alone and in combination (d). Combination treatment with siADAM10 and siADAM17 blocked the TNF␣-induced reduction of total and cell-surface EMCN protein. n ϭ 7-12, one-way ANOVA. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; ****, p Ͻ 0.0001; ns, not significant.

EMCN cleavage by ADAM10 and ADAM17 under inflammation
to be regulated by ADAM10 and ADAM17 alone and in combination.
When looking at the role of ADAM10 and ADAM17 in regulating total EMCN protein following TNF␣ stimulation, only combined knockdown of ADAM10 and ADAM17 (148%) increased total EMCN protein expression compared with siNT controls. The finding that ADAM10 alone prevented only TNF␣-induced loss of cell-surface but not total EMCN protein suggests a role of ADAM10 in cleaving EMCN's cell-surface ectodomain under proinflammatory conditions. Accordingly, the results showed that TNF␣ stimulation reduced EMCN protein levels in HUVECs transfected with siNT (48% decrease), siADAM10 (34% decrease), and siADAM17 (46% decrease). Interestingly, combined knockdown of ADAM10/ADAM17 increased EMCN total protein expression in the absence of TNF␣ (126% increase). The finding that only combination knockdown of ADAM10 and ADAM17 increased EMCN total protein levels in the presence and absence of TNF␣ suggests a dual role of ADAM10 and ADAM17 in posttranscriptional regulation of EMCN expression (Fig. 5b).

Discussion
Proteolytic cleavage is an important posttranslational modification involved in regulating biological functions of cell-surface proteins. EMCN, an EC-specific cell-surface sialomucin and part of the glycocalyx, plays a critical role in regulating the interactions between ECs and inflammatory cells (15). This study provides evidence that EMCN undergoes both constitutive and induced cleavage. The detection of the C-terminal fragment of EMCN (i.e. the transmembrane and intracellular domains) at low levels in cell lysate of unstimulated cultured ECs and significant increased levels in the presence of TNF␣ or pervanadate indicate constitutive cleavage of EMCN under quiescent conditions and induced cleavage under inflammatory conditions, suggesting that EMCN could be regulated posttranslationally through cleavage. CD43 (leukosialin), the major membrane sialoprotein on leukocytes, is also proteolytically cleaved during inflammatory activation (31,32). Given the repulsive barrier function of sialomucins, shedding of the extracellular domain of EMCN and CD43 from the cell surface would facilitate EC-leukocyte interactions during inflammation.
Heparan sulfate proteoglycans, such as syndecans, another major component of the endothelial glycocalyx, are similarly cleaved during inflammation. Mice lacking syndecan-1 have increased leukocyte-EC interactions, suggesting a role of syndecan-1 in suppressing these interaction under normal condi-tions (33,34). At the same time, the abundant heparan sulfate in syndecan-1 also binds to proinflammatory cytokines and chemokines. Inhibition of syndecan-1 cleavage has been shown to impede removal of CXC chemokines bound to syndecan-1 in lipopolysaccharide-challenged mice and to exacerbate neutrophilic inflammation (35).
Given that the estimated molecular mass of EMCN-CTF released by TNF␣ or pervanadate treatment is about 12 kDa and the intracellular and transmembrane domains of EMCN combined are about 8 kDa, the cleavage site for EMCN-CTF is predicted to be in the extracellular domain. However, we have been unable to detect the N-terminal fragment of EMCN in cell lysate or conditioned medium, leaving the fate of the cleaved EMCN extracellular N-terminal fragment undetermined. It is possible that an additional cleavage event removes the myc tag from the N-terminal fragment during proteolytic processing of EMCN, that the extracellular fragment of EMCN undergoes degradation upon initial cleavage, and/or that the fragment is internalized and degraded; any of these events would render the N-terminal fragment nondetectable using our current methods. Many substrates of ADAMs, including Notch and amyloid precursor protein, undergo more than one cleavage event (24). Further studies using EMCN with the tag positioned at various locations of its extracellular domain in the presence of protease inhibitors may circumvent this limitation and provide more information about EMCN extracellular fragment.
MMPs and ADAMs have been shown to mediate shedding of the EC glycocalyx under inflammatory and ischemia conditions (8,36,37). Use of pan-MMP inhibitors pointed to involvement of the MMP superfamily in EMCN cleavage. Then, motivated by prior observations of the roles of specific ADAM family members in shedding of cell-surface proteins (24), we further employed small-molecule inhibitors as well as siRNA against ADAM10 and ADAM17 alone and in combination to dissect their relative contributions in EMCN shedding. The results indicate a critical role of ADAM10 and ADAM17 in constitutive shedding of cell-surface EMCN. For induced shedding of cell-surface EMCN, however, although both ADAM10 and 17 appear to be involved in pervanadate-induced cell-surface EMCN reduction (Fig. S6), only ADAM10 or ADAM10 in combination with ADAM17 is responsible for TNF␣-induced shedding (Figs. 4 and 5), suggesting that the mediators of EMCN shedding could be stimulus/context-dependent.
Interestingly, differential regulation by ADAMs has been observed for the cleavage of TNF␣ itself. Whereas constitutive cleavage of TNF␣ in macrophages, NIH3T3 cells and ECs was found to be mediated by both ADAM10 and ADAM17 (38), phorbol ester or pervanadate stimulation of TNF␣ shedding in mouse embryonic fibroblasts was primarily mediated by ADAM17 (39). In addition to EMCN as a substrate in ECs, it has been shown that ADAM10 or ADAM17 or both cleave EC junction proteins, VE-cadherin (40) and JAM-A (29), as well as, adhesion molecules between inflammatory cells and ECs, vascular cell adhesion molecule-1 also known as VCAM-1 (27) and intracellular adhesion molecule-1 or ICAM-1 (28). Furthermore, in a murine model of acute pulmonary inflammation induced by lipopolysaccharide leukocyte-specific knockout of ADAM10 (41) or endothelial-specific deletion of ADAM17 (22)

EMCN cleavage by ADAM10 and ADAM17 under inflammation
prevents leukocyte recruitment, supporting the critical role of ADAM10 and 17 in regulating inflammation. Further investigation of the effects of ADAM10 and ADAM17 deletion on their substrates in vivo could provide valuable information regarding their temporal and cell-type modulation to maximize the beneficial effects.
This study also defines a role of ADAM family members (ADAM10 and ADAM17) in regulating EMCN transcription in addition to transcriptional control of EMCN by GATA2, which has been reported previously (42). This is based on our data, in which cells depleted of both ADAM10 and ADAM17 showed an increase in EMCN mRNA in the presence and absence of 24 h of TNF␣ stimulation. These findings suggest that the ADAM family members ADAM10 and ADAM17 may regulate EMCN transcription under constitutive and TNF␣-stimulated conditions.
Because of its role as a critical contributor to the pathogenesis of numerous diseases, vascular inflammation has been the focus of much research. Onset of inflammation includes increased soluble proinflammatory mediators and increased adhesion molecules on circulating inflammatory cells as well as endothelial cells at sites of inflammation. So far, the mainstay of inflammation-targeting drugs has been nonsteroidal anti-inflammatory drugs and proinflammatory cytokine-neutralizing antibodies (for example, TNF␣), which focus on soluble mediators in inflammation. Although they alleviate the symptoms of some inflammatory diseases, they also systematically suppress critical components of the host immune response and increase the possibility of adverse effects (43). EMCN, an EC-specific cell-surface protein, is a more specific target, and our results suggest that regulation of cell-surface EMCN and its cleavage may be an attractive therapeutic approach.

Expression vector and adenovirus for tagged endomucin
A cDNA fragment encoding a double-tagged human EMCN protein (Myc tag at the N-terminal of the full-length human EMCN after the signal peptide sequence and DDK tag at its C terminus) was synthesized by Genscript (Piscataway, NJ). This cDNA was then cloned into the pcDNA 4 plasmid with a pCMV promoter via EcoRV and NotI cloning sites as the expression vector. The adenovirus expressing tagged EMCN was generated and titered by Vector Biolabs (Malvern, PA). Briefly, this cDNA was inserted into a shuttle vector with a pCMV promoter via EcoRV and NotI, and then the expression cassette was transferred into an adenoviral vector (Ad5 with E1/E3 deletion) by homologous recombination. The adenoviral vector was linearized and transfected into 239 cells to generate the adenovirus. An adenovirus expressing GFP under the same pCMV promoter was used as the control.

Cell culture
HUVECs (provided by the Center for Excellence in Vascular Biology, Brigham and Women's Hospital, Boston, MA) were cultured on 0.2% gelatin-coated T-75 flasks in EGM-2 medium (EBM-2 with the SingleQuot kit, Lonza, Walkersville, MD) supplemented with 20% FBS and L-glutamine and maintained at 37°C with 5% CO 2 . For all experiments, HUVECs between passages 4 and 6 were used. Prior to inflammatory stimulus treatments, confluent HUVECs that were confluent for less than 24 h were incubated overnight in low-serum medium (EBM-2 with 0.5% FBS and L-glutamine).

Plasmid transfection and adenoviral transduction
To confirm the expression of tagged EMCN, pcDNA plasmids with or without tagged EMCN cDNA were introduced into HUVECs using Amaxa HCAEC Nucleofactor Kit (Lonza). After trypsinization, 1.25 ϫ 10 6 cells were resuspended in 100 l of Nucleofactor transfection reagent with 2 g of the pcDNA vector, including the tagged EMCN or empty vector. The cells were then electroporated with the program S-005, recovered in EGM-2, and distributed into 12-well plates coated with 0.2% gelatin. Twenty-four hours after plating, the cells were incubated in low-serum medium for 48 h before cell lysates were harvested.
For the remainder of the experiments, tagged EMCN was delivered into HUVECs via the adenovirus. AdEMCN or

EMCN cleavage by ADAM10 and ADAM17 under inflammation
AdGFP was added to HUVECs at a multiplicity of infection of 30 24 h after the cells were plated. Another 24 h later, the cells were incubated overnight in low-serum medium before treatment.

Protein isolation and protein expression analysis
Cell-surface protein isolation by biotinylation-To determine the presence and relative protein levels of EMCN on HUVEC cell membranes, the Pierce Cell Surface Protein Isolation Kit (Thermo Fisher Scientific, Waltham, MA) was used according to the manufacturer's instruction. Briefly, cell-surface protein was biotinylated, purified from cell lysates using a NeutrAvidin-agarose affinity column, and subjected to Western blot analysis.
Total protein isolation-After the designated treatments, cells were collected, lysed in cell lysis buffer (Cell Signaling) containing protease and phosphatase inhibitors (Roche) with high-frequency ultrasonication, and clarified at 14,000 rpm for 20 min at 4°C. Total protein was then quantified by Pierce BCA assay (Thermo Fisher Scientific) and subjected to Western blot analysis.
Western blot analysis-Equal amounts of total protein or equal volumes of biotinylated protein were separated by SDS-PAGE under reducing conditions and transferred to PVDF membranes (Immobilon-P, Millipore, Billerica, MA). For proteins with a molecular mass below 20 kDa, equal amounts of protein (40 g) were separated by SDS-PAGE under reducing conditions and transferred to nitrocellulose membranes (N8017, Sigma). Membranes were blocked in TBS containing 0.1% Tween (TBS-T) and 5% nonfat milk for 1 h at room temperature, washed with TBS-T, and then probed with a primary antibody in TBS-T with 1% BSA overnight at 4°C. Secondary antibody incubations were performed with the corresponding horseradish peroxidase-conjugated antibody at room temperature for 1 h with thorough washes before and after. Proteins were detected by chemiluminescence with SuperSignal West Pico or Dura (Thermo Scientific). ␣-Tubulin was used as the loading control for total protein samples. For cell-surface assays, ␣-tubulin from the flow-through or CD31 from the biotinylated membrane fraction was used as a loading controls.
PCR cycling was performed at 95°C for 10 min, followed by 95°C for 15 s and 60°C for 1 min for a total of 40 cycles. The cycle threshold values corresponding to the PCR cycle number at which fluorescence emission in real-time reaches a threshold above the baseline emission were automatically calculated by a LightCycler 480 II (Roche) according to the manufacturer's instructions. PCR reactions where cDNA templates were replaced with RNase-free water were used as negative controls. Amplification of HPRT was performed on each sample as the housekeeping gene for normalization. Each biological sample was assayed in triplicate for each set of primers, and the averaged cycle threshold value of the triplicate was used to represent each biological sample for further analysis.

Immunostaining and fluorescence microscopy analysis
For localization of EMCN on cultured cells, HUVECs grown on fibronectin-treated glass coverslips were fixed in 1% formalin overnight at 4°C, washed three times for 5 min in PBS, and blocked with 3% secondary serum species and 1% BSA in PBS for 1 h at room temperature. Coverslips were then incubated in rat anti-human endomucin at 4°C overnight, followed by three 5-min washes in PBS. The cells were then incubated in DyLight 550 goat anti-rat secondary antibody for 1 h at room temperature and washed in PBS. DAPI was used to stain cell nuclei. Fluorescent images were taken with a Zeiss fluorescence microscope (Axioskop 2 MOT Plus, ϫ40 objective).

Statistical analysis
Statistical analysis was carried out with Prism 4.0b for Macintosh. All data are shown as mean Ϯ S.E. The mean values from each group were compared by Student's t test between two groups or one-way analysis of variance (ANOVA) with Tukey's post hoc test for more than two groups, except for dose-response (Fig. 1, a and b) and time point experiments (Fig. S3) with Dunnett's post hoc test. In all tests, p Ͻ 0.05 was considered statistically significant.

Data availability
All data are contained within the article.