Unbiased proteomics identifies plasminogen activator inhibitor-1 as a negative regulator of endothelial nitric oxide synthase

Significance

Endothelial nitric oxide synthase (eNOS) is a fundamental mediator of vascular function influencing cardiovascular homeostasis through the generation of nitric oxide (NO). eNOS activity can be regulated by protein–protein interactions and here using unbiased proteomics, we uncover plasminogen activator inhibitor-1 (PAI-1) as a potent negative regulator of eNOS function. Direct eNOS/PAI-1 binding inhibits NO production and bioavailability. Moreover, knockdown or antagonism of PAI-1 increases NO production. Mechanistically, we show that secreted PAI-1 is endocytosed and interacts with eNOS and pharmacological inhibition of PAI-1 improves endothelial function. Thus, this work identifies a nonproteolytic role for PAI-1, and antagonism of PAI-1 may serve as an approach to promote endothelial function and homeostasis.

Abstract

Nitric oxide (NO) produced by endothelial nitric oxide synthase (eNOS) is a critical mediator of vascular function. eNOS is tightly regulated at various levels, including transcription, co- and posttranslational modifications, and by various protein–protein interactions. Using stable isotope labeling with amino acids in cell culture (SILAC) and mass spectrometry (MS), we identified several eNOS interactors, including the protein plasminogen activator inhibitor-1 (PAI-1). In cultured human umbilical vein endothelial cells (HUVECs), PAI-1 and eNOS colocalize and proximity ligation assays demonstrate a protein–protein interaction between PAI-1 and eNOS. Knockdown of PAI-1 or eNOS eliminates the proximity ligation assay (PLA) signal in endothelial cells. Overexpression of eNOS and HA-tagged PAI-1 in COS7 cells confirmed the colocalization observations in HUVECs. Furthermore, the source of intracellular PAI-1 interacting with eNOS was shown to be endocytosis derived. The interaction between PAI-1 and eNOS is a direct interaction as supported in experiments with purified proteins. Moreover, PAI-1 directly inhibits eNOS activity, reducing NO synthesis, and the knockdown or antagonism of PAI-1 increases NO bioavailability. Taken together, these findings place PAI-1 as a negative regulator of eNOS and disruptions in eNOS–PAI-1 binding promote increases in NO production and enhance vasodilation in vivo.

The study of endothelial nitric oxide synthase (eNOS) and its product, nitric oxide (NO), has been at the cornerstone of understanding how the vascular endothelium regulates blood vessel function. The regulation of eNOS is multifaceted, tightly controlled at the level of transcription, co- and posttranslation, phosphorylation, and influenced by various protein–protein interactions (1–3). Moreover, the uncoupling and deregulation of eNOS has been characterized across a variety of conditions and pathologies, including hypertension, diabetes, ischemia–reperfusion, metabolic syndrome, myocardial infarction, and atherosclerosis (1). Thus, preserving effective eNOS function, and deeper insights into the interactions and mechanisms that promote or disrupt this important enzyme, is critical for cardiovascular homeostasis.

Diverse protein–protein interactions, both direct and indirect, have been shown to modulate eNOS activity. For example, the binding of caveolin-1 (Cav-1) to plasmalemmal anchored eNOS negatively regulates eNOS function (4) and upon activation of endothelium, the increases in intracellular Ca²⁺ promotes the activation of calmodulin (CaM) which rapidly displaces the eNOS–Cav-1 interaction through the direct binding of CaM at the N-terminal oxidase domain of eNOS (5, 6). This association sets the stage for the coordination of reduced nicotinamide-adenine-dinucleotide phosphate (NADPH)-dependent electron flux across the C-terminal reductase domain to the N-terminal heme moiety (7). The reduction of heme at this site binds molecular oxygen hydroxylating l-arginine to form the intermediate N^ω-hydroxy-l-arginine which is subsequently oxidized to form l-citrulline and NO (1, 7). Additional interactions of eNOS with HSP90 (8) promotes a conformational change that increases eNOS activity and this chaperone can stabilize heme in NOS enzymes (9, 10) and serve as a molecular scaffold recruiting kinases, such as AMPK and Akt (11, 12). These kinases can phosphorylate eNOS throughout various sites (Ser¹¹⁷⁷, Ser⁶¹⁵, Ser⁶³³) and promote a dramatic increase in NO synthesis. Hemoglobin (HB) α was recently shown to influence NO release through a direct protein–protein interaction with eNOS. HB α’s oxidation state (ferrous [Fe²⁺] or ferric [Fe³⁺]) mediates NO release with Fe²⁺ HB α scavenging NO rapidly while Fe³⁺ HB α promotes NO release and diffusion across the endothelium (13, 14). Dynamic changes to the eNOS interactome and subsequent changes to NO biosynthesis have been observed in response to various stimuli.

Recently, our laboratory has used eNOS tagging and copurification techniques followed by mass spectrometry (MS) to identify proteins, in an unbiased manner, that stably interact with eNOS (15). In order to appreciate the dynamic nature of protein–protein interactions regulating eNOS, we refined the initial experiments to examine rapid interactions in the “basal” versus “active” state of the enzyme using stable isotope labeling with amino acids in cell culture (SILAC) and MS to quantify new eNOS interacting proteins. Using this approach, we have uncovered several eNOS interactors and report here that the secreted protein, plasminogen activator inhibitor-1 (PAI-1), can be endocytosed by endothelial cells (ECs) and interact with eNOS. The binding of PAI-1 to eNOS is regulated by growth factors and angiotensin II, and PAI-1 directly suppresses eNOS activity and NO bioavailability. In addition, chemical inhibition of PAI-1 reduces its interaction with eNOS and improves endothelium-dependent relaxations of blood vessels. Thus, in addition to PAI-1 serving as a negative regulator of fibrinolysis, intracellular PAI-1 can interact with eNOS to reduce NO production by the endothelium and likely contribute to endothelial dysfunction in disease associated with elevated levels of PAI-1.

Results

Using SILAC to Uncover Potential eNOS Interactors under Basal and Stimulated Conditions.

To identify proteins that interact with eNOS under basal versus stimulated conditions, EA.hy926 endothelial cells, were stably labeled with light and heavy isotopes of amino acids (stable isotopic labeling with amino acids in cell culture [SILAC]) for multiple passages (more than six passages where more than 95% of the proteins incorporated the labels). Both light and heavy isotopically labeled cells were then infected with adenoviruses expressing GFP or eNOS-GFP. GFP was immunoprecipitated from lysates of light or heavy cells expressing GFP or eNOS-GFP after serum starvation or after stimulation with 10% fetal bovine serum (FBS) and endothelial cell growth supplement (ECGS) for 15 min (SI Appendix, Fig. S1). Proteins significantly enriched in the eNOS-GFP immunoprecipitates (i.e., eNOS interactors) before and after stimulation were identified by mass spectrometry (Fig. 1A, SI Appendix, Table S1, and Dataset S1). Only proteins significantly enriched in eNOS-GFP immunoprecipitates in comparison to GFP immunoprecipitates under either starved or stimulated conditions were regarded as potential eNOS interactors (SI Appendix, Fig. S2 A and B and Tables S2 and S3 and Datasets S1–S3). Prior to choosing a candidate for further investigation, we investigated the availability of biochemical tools for each of the enriched proteins, including validated antibodies, siRNAs, and recombinant proteins that would be vital tools necessary for the proposed experiments aimed at elucidating whether or not these candidates potentially interact with or influence eNOS. Due to the availability of these biochemical tools and the abundance of published reports demonstrating strong correlations between eNOS, NO, and PAI-1 intimately involved across various pathologies, including hypertension, endothelial dysfunction, senescence, and fibrosis (16–21), PAI-1 was chosen for further investigation into its biological significance as a component of the eNOS interactome complex.

Fig. 1.

Identification of potential eNOS interactors using SILAC and MS. (A) SILAC/MS plot of protein intensities against normalized heavy/light SILAC ratios from EA.hy926 SILAC cells expressing an eNOS-GFP construct under starvation compared to 10% FBS + ECGS conditions. Protein intensities are plotted against heavy/light SILAC ratios. Significantly enriched proteins are colored in dark blue (P < 0.05); other proteins are shown in light blue and PAI-1 is shown in red. (B) PAI-1 directly interacts with eNOS. Representative Western blot of eNOS immunoprecipitates showing the association of PAI-1 with eNOS in HUVECs and (C) purified protein preparations showing a direct protein–protein interaction. (D) Representative Far-Western immunoblot image confirming binding between eNOS and PAI-1 using purified anti–PAI-1 and anti-eNOS antibodies and corresponding secondary antibodies for fluorescent detection.

PAI-1 Directly Interacts with eNOS.

Several reports have documented a close correlation between eNOS, NO, and PAI-1. eNOS, hemoglobin, and PAI-1 are colocalized in myoendothelial junctions of microvessels and NO can regulate PAI-1 mRNA levels and its secretion from cells (14, 19, 21–24). Furthermore, PAI-1-deficient mice are protected against hypertension and vascular disease in the setting of long-term NOS inhibition (18, 25). Nevertheless, a direct link as to whether or not PAI-1 influences eNOS function through a protein–protein interaction mechanism has never been tested.

To confirm the presence of eNOS–PAI-1 complexes in primary cultures of HUVECs, coprecipitation experiments were done. Immunoprecipitation (IP) of eNOS resulted in the coassociation of eNOS with PAI-1 and a known eNOS interactor, HSP90 (1), whereas immunoprecipitation with an isotype-matched control Ab (IgG) did not (Fig. 1B). To further distinguish whether or not PAI-1 directly interacts with eNOS, purified recombinant eNOS and recombinant PAI-1 in its latent form were used. Incubation of purified eNOS and purified PAI-1 revealed the formation of a direct eNOS/PAI-1 complex as evident by immunoisolation and Western blotting (Fig. 1C) and similar results were seen with active versus latent PAI-1. Furthermore, Far-Western immunoblotting using recombinant PAI-1 electrophoresed in a SDS/PAGE gel followed by overlaying the blot with purified NOS protein demonstrated direct eNOS binding to PAI-1 (Fig. 1D). These independent approaches confirm the proteomic results and support a direct protein–protein interaction between eNOS and PAI-1 in cells and in solution.

The Interaction between PAI-1 and eNOS Is Regulated and PAI-1 Inhibits eNOS Activity.

To investigate whether the interaction between PAI-1 and eNOS is dynamically influenced by exogenous stimuli, HUVECs were treated with conditions known to alter eNOS activity followed by immunoprecipitation. First, the conditions used in the initial SILAC experiments to identify the eNOS interactome were used. Treatment of serum-starved HUVECs with 10% FBS/ECGS for 20 min resulted in a twofold increase in PAI-1, similar to results using SILAC (Fig. 2 A, Left). The increased interaction correlated with eNOS phosphorylation at the stimulatory site, Ser¹¹⁷⁷, and a reduction in phosphorylation at the inhibitory site, Thr⁴⁹⁵, as quantified (Fig. 2 A, Center and Right graph). To test whether agents that promote eNOS uncoupling influence the eNOS–PAI-1 interaction, HUVECs were treated with angiotensin II (Ang II). As seen in Fig. 2B, Ang II promoted the dissociation of PAI-1 from eNOS, a reduction in eNOS phosphorylation at Ser¹¹⁷⁷ and an enhancement of eNOS phosphorylation at Thr⁴⁹⁵. These observations suggest that the direct interaction between eNOS and PAI-1 can undergo a dynamic association and dissociation acutely, depending on the stimulus.

Fig. 2.

PAI-1 inhibits eNOS. (A) Representative immunoblot analyses of eNOS immunoprecipitates showing an increase in eNOS–PAI-1 association after 20 min of 10% FBS + ECGS stimulation (Left) and immunoblots showing the changes in eNOS phopshorylation in whole-cell lysates (WCLs) with respective averaged densitometric data (Right) during this treatment period. (B) eNOS immunoprecipitates showing a decrease in eNOS–PAI-1 association after 20 min of Angiotensin II (Ang II) (1 μM) (Left) and respective WCL immunoblots showing the eNOS phosphorylation with respective averaged densitometric data (Right) after Ang II treatment. (representative images, n = 4, mean ± SEM, n = 3 to 5, *P < 0.05 vs. starved or vehicle treatment). (C) eNOS activity (NO synthesis and NAPDH oxidation) of purified eNOS in response to increasing concentrations of PAI-1. Measurements were done at 25 °C and the values (min⁻¹) represent the average ± SD for three independent measurements (*P < 0.05 vs. 0 ng PAI-1). (D) Representative Western blot of Ctrl or PAI-1 siRNA-treated HUVECs for 36 h. (E) eNOS activity assays conducted in HUVECs treated with Ctrl or PAI-1 siRNA for 36 h followed by a 1-h assessment of l-[³H]arginine into l-[³H]citrulline conversion as an index of NOS activity in the presence and absence of the NOS inhibitor l-NAME (1 mM), ionomycin (1 μM), or l-NAME (1 mM) + ionomycin (1 μM) cotreatment (n = 4, mean ± SEM, *P < 0.05 vs. basal Ctrl siRNA, ^#P < 0.05 vs. basal PAI-1 siRNA).

In order to determine if PAI-1 directly impacts eNOS activity, enzymatic assays were performed to quantify NO synthesis and NADPH oxidation (as an index of eNOS reductase function) in the absence and presence of the allosteric activator of eNOS, CaM, using purified components. As seen in Fig. 2C, latent PAI-1 dose-dependently inhibited CaM-stimulated NO synthesis and NADPH oxidation with maximal inhibition observed at 100 ng of PAI-1 in the reaction mix (∼50% inhibition). PAI-1 did not directly inhibit NADPH oxidation in the absence of CaM. This suggests that PAI-1 inhibits eNOS in the activated state.

To examine if this in vitro mechanism is similar in intact cells, eNOS activity was measured in HUVECs treated with control or PAI-1 siRNA (Fig. 2D). PAI-1–depleted HUVECs displayed an increase in basal eNOS activity (Ctrl siRNA 1 ± 0.1-fold versus PAI-1 siRNA 1.92 ± 0.2-fold, respectively; P < 0.05) and an elevated response after stimulation with the calcium ionophore, ionomycin (Ctrl siRNA 2.38 ± 0.3-fold versus PAI-1 siRNA 3.56 ± 0.3-fold, respectively; P < 0.05) (Fig. 2E). All measurements of NOS activity were abrogated by the NOS inhibitor, ʟ-nitroarginine methyl ester (ʟ-NAME). Collectively, these data support the notion that endogenous PAI-1 negatively regulates eNOS function.

PAI-1 Colocalizes with eNOS in HUVECs.

PAI-1 is a secreted protein mediating its primary function as a serine-protease inhibitor targeting the tissue‐type and the urokinase‐type plasminogen activator (tPA and uPA, respectively) thus impeding fibrinolysis in the circulation (26, 27). In addition to this well-described function of PAI-1, there are recent reports demonstrating nonproteolytic, intracellular functions of PAI-1 showing that PAI-1 regulates cell signaling, survival, and migration (28). However, the mechanisms of this intracellular role of PAI-1 are not clear or well characterized. The compartmentalization of eNOS has been extensively characterized and eNOS is a dually acylated, peripheral membrane protein primarily found on the cytoplasmic face of the Golgi complex and on the plasma membrane in lipid rafts or in caveolae (1, 29, 30). While several reports have documented intracellular PAI-1 found in the Golgi as expected for a secreted protein as well as a diffusely distributed pool of cytosolic PAI-1 (31–33), we sought out to establish whether or not PAI-1 can interact with eNOS, given the obvious differences in compartmentalization of the two proteins.

Immunofluorescence confocal microscopy experiments confirmed the colocalization of PAI-1 in HUVECs with the Golgi marker GOLM1 (Fig. 3A, as quantified in SI Appendix, Fig. S3A) as well as a cytosolic pool. As expected, eNOS was also colocalized with GOLM1 (Fig. 3B, as quantified in SI Appendix, Fig. S3B) and PAI-1 colocalized with eNOS in HUVECs (Fig. 3C, as quantified in SI Appendix, Fig. S3C). To assess whether the absence of eNOS or PAI-1 altered the localization of either protein we employed the use of siRNA. Notably, the knockdown of eNOS (SI Appendix, Fig. S4A) in HUVECs did not affect the localization of PAI-1 in the perinuclear compartment or cytosol (SI Appendix, Fig. S4B). Similarly, the knockdown of PAI-1 (SI Appendix, Fig. S4C) did not alter the perinuclear pattern of eNOS labeling (SI Appendix, Fig. S4D).

Fig. 3.

eNOS and PAI-1 colocalize outside of the Golgi apparatus. (A) HUVECs fixed and stained for PAI-1 (green), GOLM1 (Golgi marker, red), and nuclear Hoechst staining (blue). (B) HUVECs fixed and stained for eNOS (green), GOLM1 (Golgi marker, red), and nuclear Hoechst staining (blue). (C) HUVECs fixed and stained for PAI-1 (green), eNOS (red), and nuclear Hoechst staining (blue). (Scale bars in merged images, 5 μm.) (D) Representative PLA image for protein interactions between eNOS and PAI-1 in Ctrl siRNA, (E) eNOS and (F) PAI-1 knockdown. (G) Quantified eNOS and PAI-1 PLA signal (arbitrary unit [A.U.]) of HUVECs treated with Ctrl, eNOS, and PAI-1 siRNA (n = 60 individual cells across four independent experiments). (Scale bars in merged images, 5 μm.)

To better assess if the perinuclear colocalization of eNOS and PAI-1 is indicative of a potential interaction in cells, proximity ligation assays (PLAs) were conducted. Analysis of PLA studies using well-characterized eNOS and PAI-1 antibodies and respective PLA probes in control siRNA-treated HUVECs revealed a robust perinuclear PLA signal (PLAS) of the eNOS–PAI-1 complex (Fig. 3D). Importantly, control experiments knocking down either eNOS or PAI-1 with a specific siRNA significantly reduced PLAS compared to control siRNA-treated cells (Ctrl siRNA 347 ± 14 versus eNOS siRNA 63 ± 5 and PAI-1 siRNA 32 ± 2 PLAS, respectivel; P < 0.05) (Fig. 3 D–F). Taken together, these data suggest that a portion of intracellular PAI-1 can interact with eNOS in cells.

eNOS Promotes PAI-1 Localization to the Cytosolic Face of the Golgi.

Next, to specifically evaluate whether the presence or absence of eNOS influences intracellular PAI-1 localization, COS7 cells that lack endogenous eNOS were used for reconstitution experiments. Overexpression of a HA-tagged PAI-1 (PAI-1-HA) in COS7 cells revealed both a Golgi (GOLM1 colocalized) and a diffuse, cytosolic HA signal (SI Appendix, Fig. S5A). Cells expressing wild-type (WT) eNOS had the expected perinuclear/Golgi eNOS signal that colocalized with GOLM1 (SI Appendix, Fig. S5B), and cotransfection of both PAI-1-HA and eNOS constructs confirmed the observation in HUVECs of clear colocalization (SI Appendix, Fig. S5C as quantified in SI Appendix, Fig. S6A). Furthermore, PLA experiments conducted under these cotransfected conditions showed a robust perinuclear and peripheral membrane-rich PLAS (SI Appendix, Fig. S5D). To determine whether the cellular compartmentalization of eNOS influenced the colocalization of eNOS with PAI-1, COS7 cells were transfected with PAI-1-HA and eNOS mutants which have been shown to localize to the cytosol (G2A eNOS, a nonacylated mutant) (29, 34) or the nucleus (nuclear localization signal [NLS] eNOS, the nonacylated mutant targeted to the nucleus) (35). As shown in SI Appendix, Fig. S5E, cells transfected with G2A eNOS showed a significant colocalization with cotransfected PAI-1-HA (as quantified in SI Appendix, Fig. S6B) and colocalization studies were confirmed by PLA (SI Appendix, Fig. S5F). In contrast, cotransfection of NLS eNOS (nuclear localization) with PAI-1-HA did not result in colocalization or a PLAS with the two proteins (SI Appendix, Fig. S5G as quantified in SI Appendix, Fig. S6C and for PLA, SI Appendix, Fig. S4H).

In order to address the source of PAI-1 that interacted with eNOS, COS7 cells lacking or expressing eNOS were incubated with conditioned media prepared from COS7 cells expressing PAI-1-HA. As shown in Fig. 4A, unmanipulated COS7 treated with media containing PAI-1-HA had a dominant cytosolic HA immunofluorescent signal. Western blotting of PAI-1-HA-treated COS7 cells confirmed the uptake/endocytosis of exogenous PAI-1-HA (see lanes 5 and 6 in Fig. 4B). In contrast, incubation of COS7 cells expressing eNOS with media containing PAI-1-HA revealed two distinct pools of PAI-1-HA signal, a cytosolic pool and a pool that colocalized eNOS in the perinuclear region of cells (Fig. 4C as quantified in SI Appendix, Fig. S6D). Moreover, PLA experiments under these conditions confirmed a robust perinuclear and peripheral membrane-rich PLAS indicative of a strong PAI-1/eNOS interaction (Fig. 4D). This observation clearly suggests that when PAI-1 is secreted from cells, it can be endocytosed in recipient cells and colocalize with eNOS. In HUVECs, this is likely an autocrine pathway since these cells express both eNOS and PAI-1.

Fig. 4.

Assessment of PAI-1 endocytosis and localization. (A) Representative image of COS7 cells treated with condition media from HA-tagged PAI-1 overexpressing cells for 6 h; HA (green), eNOS (red), GOLM1 (cyan), and nuclear Hoechst staining (blue). (B) Representative Western blot examining the endocytosis of HA-tagged PAI-1-enriched media after 6 h in COS7 cells. (C) Localization and (D) PLA imaging of PAI-1 in cells transfected with WT eNOS cDNA and fed condition media enriched in HA-tagged PAI-1 for 6 h; HA (green), eNOS (red), and nuclear Hoechst staining (blue). (Scale bars in merged images, 5 μm.)

Antagonism of PAI-1 Promotes NO Bioavailability and Enhances Endothelium-Mediated Vasodilation.

To examine the role of PAI-1 antagonism on vascular function and EC-mediated vasodilation we employed the use of a PAI-1 antagonist, TM5441. TM5441 is a small molecule inhibitor of PAI-1 that binds to a common binding region at the S4A of the beta sheet on PAI-1 and mimics the inhibited state of PAI-1 (similar to the confirmation induced with its reactive center loop [RCL] folded inwards) thereby preventing its covalent binding to tPA or uPA (36). As seen in Fig. 5A, PAI-1 antagonism prevents the eNOS–PAI-1 interaction using recombinant, purified proteins (1.0 ± 0.19 vs. 0.39 ± 0.04, P < 0.05) and prevents the PAI-1-mediated inhibition of NO synthesis (Fig. 5B), suggesting that TM5441 binding to latent PAI-1 alters its conformation to blunt eNOS–PAI-1 complex formation. To examine if this pathway is operational in intact blood vessels, vasomotor function was analyzed in cannulated third order mesenteric arterioles. Treatment with TM5441 increases acetylcholine (Ach) induced vasodilation as indicated by a leftward shift in the dose–response curve (Fig. 5C) and the contribution of eNOS to this dilatory response was rightward shifted when treated with l-NAME. How TM54441 blunts the interaction of eNOS with latent PAI-1 and the differences in concentrations of TM5441 required for in vitro versus in vivo action are not known and need to be explored by additional experiments. Collectively, these data show that endogenous PAI1 negatively regulates eNOS function in vitro and PAI-1 inhibition improves endothelial cell responsiveness in blood vessels.

Fig. 5.

PAI-1 antagonism promotes NO production and enhances EC-mediated vasodilation. (A) Representative Western blot of eNOS immunoprecipitates showing the association of recombinant PAI-1 with eNOS in the presence and absence of the PAI-1 antagonist TM5441 (n = 5, mean ± SEM *P < 0.05). (B) eNOS activity assays assessing the ability of the TM5441 to prevent the PAI-1–mediated inhibition of NO synthesis in purified protein preparations. l-[³H]arginine to l-[³H]citrulline conversion used as an index of eNOS activity in (n = 5, mean ± SEM, *P < 0.05 vs. eNOS, ^#P < 0.05 vs. eNOS + PAI-1). (C) Pressure myography cumulative concentration–response curve to acetylcholine in third order mesenteric arteries treated with control (dimethyl sulfoxide [DMSO]), TM5441 (10 μM), or l-NAME (300 μM) (mean ± SEM, n = 4–6 *P < 0.05 vs. control [DMSO]).

Discussion

The central goal of this study was to explore the eNOS interactome using an unbiased, quantitative approach. Here, we uncover several potential regulators of eNOS and deeply characterize PAI-1 as a potent negative regulator of eNOS via a direct protein–protein interaction. Prior work from our laboratory using a proteomics-based approach identified SDF2 as a novel interactor with HSP90 and, in turn, with eNOS (15). Here, using a more rapid affinity purification workflow in SILAC-labeled ECs, several additional candidates were identified that interact with eNOS in the basal and/or stimulated state of the endothelium (SI Appendix, Table S1). From this list, we chose PAI-1 as a protein of interest meriting further investigation due to both the availability of biochemical tools that are necessary for such an investigation and the consistent data from various reports implicating an inverse correlation between PAI-1 levels and NO (20, 21, 24). Indeed, our data confirm an intimate relationship between eNOS and PAI-1 and show using a variety of independent approaches that endocytosed, intracellular latent PAI-1 can directly interact with eNOS and reduce NO production. These data suggest that the autocrine production of PAI-1 by the endothelium can exert distinct actions aside from modulating fibrinolysis via inhibition of tPA and uPA, uncovering a function of PAI-1 as a mediator of endothelial function through direct effects on eNOS activity and NO production.

Canonically, PAI-1’s role in the vasculature has been intimately tied to the regulation of fibrinolysis. The degradation of fibrin relies on the activity of multiple plasminogen activators, including PAI-1 targets, u-PA and t-PA, preventing the cleavage of a specific Arg-Val (R561/V562) peptide bond located within the protease domain of plasminogen, thus preventing plasminogen to plasmin conversion (27, 37). PAI-1 is produced and secreted by a variety of cells, including endothelial cells in an active conformation. Active PAI-1 is thermodynamically unstable and is spontaneously converted to an inactive/latent thermodynamically favorable form. This process typically occurs spontaneously at a rapid rate with active PAI-1 having a latency transition half-life (t1/2) of around 2 h at 37 °C (38–41). Extracellular binding of PAI-1 to proteins such as vitronectin and low-density lipoprotein receptor-related protein (LRP1) stabilizes PAI-1 and its inhibitory activity, maintaining PAI-1 in an active state providing a severalfold increase in protein stabilization (41, 42). Interestingly, several PAI-1 mutations have been demonstrated to promote the stability of active PAI-1 with some mutations ranging from 9- to 300-fold increases in PAI-1 stability (38). Nevertheless, active PAI-1 which is endocytosed after interacting with its target, uPA, as a complex containing uPA:PAI-1:uPAR that is internalized (43). While studies have shown that after internalization, uPAR can be recycled to the cell surface, the fate of the covalently bound tPA:PAI-1 internalized complex remains unclear. The data presented here suggest an intracellular role for PAI-1, presumably in its latent conformation, since intracellular pH and temperature favor latent PAI-1 and the intracellular microenvironment lacks conditions that could convert latent PAI-1 to its active form (41). Recent investigations have uncovered several nonproteolytic functions of PAI-1, including the promotion of cell migration through a Jak/Stat1-mediated signaling mechanism brought upon after LRP1/PAI-1 binding that is independent of PAI-1/uPA/uPAR complex formation (28). Conversely, the vitronectin–PAI-1 interaction reduces cell motility by preventing LRP1/PAI-1 binding (44). Interestingly, intracellular PAI-1, which is conventionally thought to be degraded after endocytosis, has been shown to interact with and inhibit furin proprotein convertase reducing the furin-dependent maturation and activity of the insulin receptor and ADAM17 (32). Our data suggest an additional intracellular and nonproteolytic role for PAI-1, wherein PAI-1 protein dynamically binds to and promotes the inhibition of eNOS.

The interaction between PAI-1 and eNOS is striking and suggests the presence of a potential mechanistic feedback and feedforward relationship, wherein increased intracellular PAI-1 reduces NO production which in turn promotes an increase in PAI-1 secretion. Our data show that the direct interaction and pairing of PAI-1 with eNOS, resulting in a dose-dependent suppression of NOS activity and NO production in endothelial cells (SI Appendix, Fig. S7). These data are consistent with a recent study demonstrating that delivery of recombinant PAI-1 to carotid arteries induces myoendothelial junction formation, reduces NO signaling, and enhances endothelial-derived hyperpolarization signaling (45). Moreover, recent studies suggest that PAI-1 is not only a critical marker and mediator of senescence (46, 47), but that it is intimately linked to aging (48), and individuals with a null mutation in PAI-1 have a longer lifespan (49). We hypothesize that the observed reduction in NO examined with aging may (50), in part, be due to elevations in PAI-1 that are also correlated with aging (46, 51). Not only has PAI-1 been implicated in aging, it has also been suggested as a key factor correlating with arterial stiffness and factors that contribute to the development of atherosclerosis (17, 52). Interestingly, since we show that the PAI-1 antagonist TM5441 enhances endothelial-dependent relaxation of mesenteric arterioles, perhaps this effect contributes to the protective actions of PAI-1 antagonism in models of fibrosis, insulin resistance, and aging (53–57). However, the enhanced levels of NO mediated by PAI-1 inhibition cannot explain all of the beneficial effects, since TM5441 reduces blood pressure and fibrosis in mice treated with a NOS inhibitor (17). Future studies are necessary to determine the intricate nature of this interaction and other potential mediators influencing this intracellular protein–protein interplay. Identification of the protein–protein interface and characterization of specific critical residues involved in the eNOS–PAI-1 interaction may provide clues as to how to potentially disrupt their interaction and preserve/augment NO production and bioavailability. The implications of this unsuspected protein–protein interaction are interesting and add a layer of complexity to how eNOS and endothelial function is regulated.

Methods

Cell Culture.

As previously described, the EA.hy926 cell line was purchased from the American Type Culture Collection (CRL-2922) and grown in Dulbecco’s Modified Eagle Medium (DMEM) (11965092, Thermo Fisher Scientific) containing 10% FBS, penicillin (100 U/mL), streptomycin (0.1 mg/mL), 2 mM glutamine, and hypoxanthine–aminopterin–thymidine (HAT). HUVECs were obtained from the Yale University Vascular Biology and Therapeutics Core facility, plated on 0.1% gelatin-coated dishes in M199 medium supplemented with ECGS, 10% FBS, penicillin (100 U/mL), streptomycin (0.1 mg/mL), and 2 mM glutamine, and used between passages 2 and 4. COS7 cells were cultured in DMEM supplemented with 10% FBS, penicillin (100 U/mL), streptomycin (0.1 mg/mL), and 2 mM glutamine (15). In some experiments, HUVECs were transfected with 40 nM siRNA targeting PAI-1 (HSS107569, Thermo Fisher Scientific), eNOS (s9622, Thermo Fisher Scientific), or a control siRNA using Lipofectamine RNAiMAX (13778075, Thermo Fisher Scientific) in Opti-MEM medium (31985-070, Thermo Fisher Scientific). After 5 h, the media were supplemented with media containing 2× growth factors overnight.

Generation of SILAC Endothelial Cells.

SILAC in EA.hy926 cells was generated by passaging EA.hy926 cells for at least 10 times in SILAC media (Thermo Fisher Scientific) under the following conditions: DMEM for SILAC without l-arginine and l-lysine supplemented with unlabeled l-arginine l-lysine for the “light” version, or supplemented with labeled l-lysine-2HCl, 13C6, 15N2, and l-arginine-HCl, 13C6, 15N4 for the “heavy” version. Cells were grown in SILAC media containing 10% dialyzed FBS, penicillin (100 U/mL), streptomycin (0.1 mg/mL), 2 mM glutamine, and HAT. The successful generation of the radiolabeled amino acid cell line was confirmed by mass spectrometry.

Transduction of Cells with Adenoviruses.

SILAC EA.hy926 cells were infected with 80 multiplicity of infection (MOI) of replication-deficient adenoviruses to express GFP or eNOS-GFP (Iowa Viral Vector Core). After 16 h, all media was changed to fresh SILAC media and cells were allowed to grow to 100% confluency for 48 h. Cell were then starved overnight in SILAC DMEM containing 0.5% FBS and then lysed or treated with 10% FBS + ECGS for 15 min before lysis. After two washes in ice-cold phosphate-buffered saline (PBS), cells were lysed with tandem affinity purification lysis buffer (1% Nonidet P-40, 20 mM Tris [pH 8], 150 mM NaCl, 10% glycerol [vol/vol], 1 mM DTT, 1 mM CaCl₂, 10 mM NaF, 0.25 mM Na₃VO₄, 5 nM calyculin A, 50 mM β-glycerophosphate, and EDTA-free Complete Protease Inhibitors [Roche]) and protein lysates were collected with the aid of cell scrapers. Lysates were incubated on an end-over-end rocker for 1 h at 4 °C before centrifugation at 13,000 rpm for 15 min at 4 °C. Protein concentrations were measured and 10 mg of protein lysate/group was used for immunoprecipitation studies conducted using GFP-nAb Agarose (Allele Biotechnology) for 1 h at 4 °C followed by washing three times with TAP lysis buffer without Nonidet P-40 and protease inhibitors. After the final wash, equal volumes of light and heavy GFP immunoprecipitates were mixed as follows:

• 1)Heavy eNOS-GFP 10% FBS + ECGS + light eNOS-GFP starvation,
• 2)Heavy eNOS-GFP starvation + light GFP starvation,
• 3)Heavy eNOS-GFP 10% FBS + ECGS + light GFP 10% FBS + ECGS.

Biological duplicates for each group were generated, mixed as indicated above, and used for mass spectrometry.

Chromatography and Mass Spectrometry.

Reversed-phase chromatography was performed on a Thermo Easy nLC 1000 system connected to a Q Exactive HF mass spectrometer (Thermo) through a nano-electrospray ion source. A total of 7.5 µL of peptides per sample was separated on a 50-cm column (New Objective) with an inner diameter of 75 μm packed in house with 1.9 μm C18 resin (Dr. Maisch GmbH). Peptides were eluted from the column with a segmented gradient of acetonitrile from 5 to 60% in 0.1% formic acid for 100 min at a constant flow rate of 250 nL/min. The column temperature was kept at 45 °C by an oven (Sonation GmbH) with a Peltier element. Eluted peptides from the column were directly electrosprayed into the mass spectrometer. Mass spectra were acquired on the Q Exactive HF in data-dependent mode to automatically switch between full-scan MS and up to 10 data-dependent MS/MS scans. The maximum injection time for full scans was 20 ms, with a target value of 3,000,000 at a resolution of 60,000 at m/z = 200 (300 to 1,650 m/z). The 10 most intense multiple-charged ions (z ≥ 2) from the survey scan were selected with an isolation width of 1.4 Th and fragmented with higher energy collision dissociation with normalized collision energies of 27 (58). Target values for MS/MS were set at 50,000 with a maximum injection time of 120 ms at a resolution of 15,000 at m/z = 200. To avoid repetitive sequencing, the dynamic exclusion of sequenced peptides was set at 30 s.

Raw data were processed by the MaxQuant version 1.4.1.4 software package with its integrated ANDROMEDA search algorithms (59, 60). Peak lists were searched against the UniProt database for Homo sapiens with common contaminants added. The search included carbamidomethylation of cysteines as fixed modification, oxidized methionine (M), and acetylation (protein N-terminal) as variable modifications, “match between runs” (MBR) with a matching time window of 0.5 min and the requant option was enabled. Maximum allowed mass deviation was set to 6 ppm for MS peaks and 20 ppm for MS/MS peaks. The minimum peptide length was set to 6 and maximum missed cleavages were 2. The false discovery rate was determined by searching a reverse database. Maximum false discovery rates were 0.01 both on peptide and protein levels. Significances were calculated according to significance A (59). All calculations and plots were done with the R software package (www.r-project.org). The log10 of intensities represents log10 of the summed up eXtracted Ion Current (XIC) of all isotopic clusters associated with the identified amino acid sequence. All proteomics experiments were performed in duplicates which were analyzed with MaxQuant. Each experiment was analyzed separately and results were based on the combined analysis by MaxQuant.

Immunofluorescence.

For imaging studies, HUVECs or COS7 cells grown on coverslips were fixed with ice-cold methanol for 20 min, washed three times with PBS, and incubated with blocking buffer (PBS containing 3% bovine serum albumin [BSA]) for 1 h at room temperature. Cells were then washed three times with PBS and incubated with the PAI-1 (mouse monoclonal, C-9 sc-5297, Santa Cruz Biotechnology), anti-HA (rat monoclonal, 11867423001, Roche Diagnostics), eNOS (rabbit polyclonal, NB120-15280, Novus Biologicals or mouse monoclonal, A-9 sc-376751, Santa Cruz Biotechnology) or the Golgi-marker GOLM1 (rabbit polyclonal, NBP1-50627, Novus Biologicals) for 24 h. After washing, Alexa Fluor 488- (anti-mouse, anti-rabbit, anti-rat), 568- (anti-mouse, anti-rabbit), and 647- (anti-goat)-conjugated secondary antibodies (Life Technologies, Thermo Fisher Scientific) were used as required for 1 h followed by washing and Hoechst (H3570, Invitrogen Thermo Fisher Scientific) nuclear staining for 5 min. After Hoechst staining, coverslips were mounted on slides using Dako fluorescence mounting medium (S3023, Agilent). Images were acquired and analyzed using the Leica SP5 confocal microscope with the Leica Application Suite X (LAS X) software (Leica Microsystems). Images reflect a z-stack compression.

Plasmid Constructions and Cell Transfection.

To generate HA-tagged PAI-1 construct, PCR was performed using the PAI-1 cDNA from DNASU (FLH057235.01X) as a template using primers: forward (introducing BamHI and Kozak sequence) CAG​GAT​CCG​CCA​CCA​TGC​AGA​TGT​CTC​CAG​C and reverse (introducing EcoRI, HA tag and stop codon) TAG​AAT​TCC​TAA​GCG​TAG​TCT​GGG​ACG​TCG​TAT​GGG​TAG​GGT​TCC ATCACTTGGC. The PCR fragment was then digested with restriction enzymes BamHI and EcoR1 and inserted into the pcDNA3. The construct was confirmed by DNA sequencing and Western blot. Characterization of the WT, G2A, and NLS eNOS mutant cDNAs has been previously described (29, 35). CO7 cells underwent transfection using Lipofectamine 2000 as previously described (15). After incubating for 5 h, the transfection media (Opti-MEM, 31985–070, Thermo Fisher Scientific) was replaced with complete growth media and cells were grown for 36 h. The following plasmids, alone or in combination, were used for transfection: pcDNA3 empty plasmid (1 μg), human HA-tagged PAI-1 in pcDNA3 (1 μg), WT eNOS in pcDNA3 (1 μg), G2A eNOS in pcDNA3 (1 μg), and NLS eNOS (1 μg) cDNAs for 36 h depending on the experiment.

PLA.

HUVECs underwent protein-specific siRNA treatment as listed above for 36 h. For COS7 experiments, cells were cotransfected with HA-tagged PAI-1 and either WT eNOS, G2A eNOS, or NLS eNOS mutant cDNAs for 36 h. Coverslips containing cells were then fixed with ice-cold methanol for 20 min, washed three times with PBS, and incubated with blocking buffer (PBS containing 3% BSA) for 1 h at room temperature. Cells were then washed three times with PBS and incubated with the PAI-1 (mouse monoclonal, C-9 sc-5297, Santa Cruz, Biotechnology) and eNOS (rabbit polyclonal, NB120-15280, Novus Biologicals) for 24 h. Next, the Duolink in situ red mouse/rabbit PLA kit (DUO92101, Sigma) was used following manufacturer’s instructions. In brief, primary antibodies were washed two times for 5 min with 1× wash buffer A (provided) and incubated with PLA probes (probe anti-rabbit PLUS, affinity-purified donkey anti-rabbit IgG [H+L] and probe anti-mouse MINUS, affinity-purified donkey anti-mouse IgG [H+L]) for 1 h at 37 °C. Slides were then washed two times for 5 min with 1× wash buffer A at room temperature and incubated with ligase in a preheated humidity chamber for 30 min at 37 °C. Following ligation, slides were washed two times for 5 min with 1× wash buffer A at room temperature and incubated with polymerase in amplification solution (provided) in a preheated humidity chamber for 100 min at 37 °C. Finally, slides were washed two times for 10 min in 1× wash buffer B (provided) at room temperature and one time with 0.01× wash buffer B for 1 min. Hoechst staining was then conducted for 5 min and coverslips were mounted on slides using Dako fluorescence mounting medium (S3023, Agilent). Images were acquired and analyzed using the Leica SP5 confocal microscope with the Leica Application Suite X (LAS X) software (Leica Microsystems). Images reflect a z-stack compression. Quantification of PLA signal denoting eNOS–PAI-1 protein–protein interactions was done using the Duolink ImageTool (Sigma).

Immunoblotting.

For IP, HUVECs were collected after treatment under the following conditions: basal (unstimulated), starved (3 h), vehicle (20 min), 10% FBS + ECGS (20 min), and 1 μM Angiotensin II (Ang II) (20 min). As previously described, HUVECs were washed twice with ice-cold PBS and then collected for IP and Western blot studies in TAP lysis buffer (1% Nonidet P-40, 20 mM Tris [pH 8], 150 mM NaCl, 10% glycerol [vol/vol], 1 mM dithiothreitol (DTT), 1 mM CaCl₂, 10 mM NaF, 0.25 mM Na₃VO₄, 5 nM calyculin A, 50 mM β-glycerophosphate, and EDTA-free Complete Protease Inhibitors [Roche]) with the aid of cell scrapers (15). Protein concentrations were calculated and 300 μg of cell lysates was incubated with 2 μg of eNOS primary antibody (rabbit polyclonal, C-20, sc-654, Santa Cruz Biotechnology) overnight with end-to-end rotation. Samples were then incubated with Protein G-Sepharose beads (10-1241, Thermo Fisher Scientific), gently rotated with end-to-end rotation for 2 h, and washed five times to eliminate unbound proteins. Samples were separated by sodium dodecyl sulfate polacrylamide gel electrophoresis (SDS/PAGE) and then transferred to 0.45-μm nitrocellulose membranes (Bio-Rad) (15). For control IgG lanes, rabbit IgG (2729S, Cell Signaling Technology) was used. Preliminary data suggested that the conformation of PAI-1 (active vs. latent) had no effect on the observed eNOS/PAI-1 binding. Since the dominant postendocytosis pool of PAI-1 found inside the cell is latent (i.e., it reflects the most accurate intracellular configuration of the protein), it suggests that the interaction between PAI-1 and eNOS does not rely on the configuration of PAI-1’s reactive center loop (RCL) (41).

For Far-Western immunoblotting, 50 ng of purified latent PAI-1 (CC4075, EMD Millipore) and 5 μg of albumin (AB01088-00100, AmericanBio) were processed and run on SDS/PAGE and transferred to a 0.45-μm nitrocellulose membrane. After renaturation with blocking buffer, membranes were incubated with 100 ng/mL of WT eNOS in lysis buffer for 24 h. Unbound WT eNOS was washed off in TBS and membranes were incubated with primary antibodies for eNOS (rabbit polyclonal, NB120-15280, Novus Biologicals) and PAI-1 (mouse monoclonal, C-9 sc-5297, Santa Cruz Biotechnology) overnight and followed by the use of secondary antibodies (LI-COR) for eNOS–PAI-1 complex formation detection.

Western blotting was conducted as previously described using the following primary antibodies: eNOS (sc-376751, Santa Cruz Biotechnology), Ser¹¹⁷⁷ peNOS (612393, BD Biosciences), Thr⁴⁹⁵ peNOS (9574S, Cell Signaling Technology), Ser⁶¹⁷ peNOS (07-561, EMD Millipore), Ser⁶³⁵ peNOS (07-562, EMD Millipore), PAI-1 (sc-5297, Santa Cruz Biotechnology), HSP90 (sc-13119, Santa Cruz Biotechnology), rabbit IgG (2729S, Cell Signaling Technology), AKT (9272S, Cell Signaling Technology), Ser⁴⁷³ pAKT (4060S, Cell Signaling Technology), Thr³⁰⁸ pAKT (2965S, Cell Signaling Technology), and anti-HA (11867423001, Roche Diagnostics). The appropriate LI-COR secondary IRDye antibodies and LI-COR Odyssey Infrared Imaging System (LI-COR) were used to detect and quantify immunoblots.

Expression and Purification of eNOS.

Full-length eNOS was purified as described previously (61, 62). Purity of protein was assessed by SDS/PAGE and spectral analysis.

NO Synthesis and NADPH Oxidation Measurement.

Steady-state activities of eNOS was determined at 25 °C using the spectrophotometric oxyhemoglobin assay as described (61, 63, 64). Before measuring steady-state activities, eNOS proteins were incubated with latent PAI-1 for 30 min at room temperature (65). For studies using the PAI-1 antagonist (TM5441, Tocris/Bio-Techne), samples were preincubated with TM5441 for 5 to 10 min prior to the addition of PAI-1.

eNOS Activity Assay (In Vitro).

HUVECs underwent PAI-1-specific siRNA treatment (40 nM) for 36 h. Cells were serum starved for 2 h in M199 containing neither FBS nor ECGS and supplemented with 10 μM sepiapterin (Sigma), followed by treatment with basal media or ionomycin (1 µM) containing l-[³H]arginine for 1 h. For all experiments, cells were incubated with 1 mM l-NAME for 30 min prior to basal media or ionomycin treatment. All reactions were terminated using lysis buffer and stop buffer provided in the NOS activity assay kit (781001, Cayman Chemicals). Samples were then incubated with equilibrated resin and contents were transferred to spin columns and collection tubes. All samples were microcentrifuged at full speed for 30 s and the flowthrough was transferred to scintillation vials. Scintillation fluid was added to the vials and the conversion of l-[³H]arginine to l-[³H]citrulline was quantified using a liquid scintillation counter. Specific NOS activity was determined by subtracting total counts from l-NAME-inhibited sample and background counts. All sample values were normalized to total protein concentration and values are expressed as eNOS activity relative to Ctrl siRNA-treated HUVECs under basal conditions.

Pressure Myography of Cannulated Arteries.

As described previously (45, 66), third order mesenteric arteries were removed, cannulated, and pressurized to 60 mmHg. Arteries were washed with a Ca²⁺-free Krebs-Hepes solution to obtain maximal passive diameter of the vessels. Internal diameter was measured throughout the experiments using the DMT MyoVIEW software. Relaxation to Ach was calculated as a % relaxation: % relaxation = [(D-Ach − D-PE) × 100]/(D-max-D-PE), where DPE was the diameter of the artery 10 min after application of 2 µmol/L phenylephrine (PE); D-initial was the diameter prior to the addition of PE; D-Ach was the diameter of the artery after application of a given dose of Ach; and D-max was the maximal diameter of the artery measured at the end of experiment. Doses of TM5441 exceeding 100 μM reduced overall tone of this vascular preparation.

Statistical Analysis.

Results are presented as means ± SEM. All experiments in which the effects of two variables were tested were analyzed by two-way analysis of variance (ANOVA) followed by Bonferroni post hoc test. Differences between two groups were compared by unpaired Student’s t test. P ≤ 0.05 was considered significant.

Data Availability Statement.

All data are available in the paper and in SI Appendix. Unique reagents will be readily available to the scientific community.