Heparan sulfate proteoglycan glypican-1 and PECAM-1 cooperate in shear-induced endothelial nitric oxide production

This study aimed to clarify the role of glypican-1 and PECAM-1 in shear-induced nitric oxide production in endothelial cells. Atomic force microscopy pulling was used to apply force to glypican-1 and PECAM-1 on the surface of human umbilical vein endothelial cells and nitric oxide was measured using a fluorescent reporter dye. Glypican-1 pulling for 30 min stimulated nitric oxide production while PECAM-1 pulling did not. However, PECAM-1 downstream activation was necessary for the glypican-1 force-induced response. Glypican-1 knockout mice exhibited impaired flow-induced phosphorylation of eNOS without changes to PECAM-1 expression. A cooperation mechanism for the mechanotransduction of fluid shear stress to nitric oxide production was elucidated in which glypican-1 senses flow and phosphorylates PECAM-1 leading to endothelial nitric oxide synthase phosphorylation and nitric oxide production.


AFM pulling on glypican-1 but not PECAM-1 induces NO production in HUVEC.
To test that NO production can be induced using AFM pulling in HUVEC, the same methods as previously described 19 were repeated in human endothelial cells. First, adhesion forces were measured using silicon nitride AFM cantilevers with pyramidal tips to confirm probe-HUVEC interaction. Probes functionalized with anti-glypican-1 and anti-PECAM-1 had adhesion forces of 130 pN and 165 pN, respectively, significantly higher than the adhesion force of the rabbit IgG isotype control (Supplemental Figure S1).
Confluent HUVEC were stimulated with rabbit anti-glypican-1, rabbit anti-PECAM-1, and anti-rabbit IgG isotype control functionalized tipless AFM probes for 30 min (Fig. 1). The 30 min time point was chosen because a significant reduction in NO production after 30 min of shear stress with PECAM-1 depletion has been observed in HUVEC 12 . Pulling on the HUVEC surface for 30 min with glypican-1 probes induced 28% mean activation, significantly increased from static control regions located 2 mm from the probing site on the same slide. The same method using PECAM-1 probes did not yield significant activation; neither did pulling with the isotype control. Free glypican-1 antibody was also tested for its ability to stimulate NO production without the application of a pulling force. 30 min incubation with anti-glypican-1 had no significant effect on HUVEC NO production compared to untreated controls (Supplemental Figure S3).
To confirm the HUVEC results, NO production after pulling with PECAM-1 probes was measured in another cell line, rat fat pad endothelial cells (RFPEC). AFM pyramidal probes functionalized with PECAM-1 antibody exhibited strong binding to the RFPEC surface (Supplemental Figure S2). However, tipless triangular cantilevers functionalized with PECAM-1 antibody did not induce activation of NO production (Supplemental Figure S4). 'Pointed' probes (triangular cantilevers with pyramidal tips) were also functionalized with the same PECAM-1 antibody to test whether a smaller tip shape would yield activation, but pulling with pointed PECAM-1 probes had similarly insignificant activation. Anti-goat-IgG functionalized tipless triangular cantilevers included as isotype controls did not induce significant activation. We conclude that AFM pulling with anti-PECAM-1 probes has no significant effect on NO production (Supplemental Figure S4). COX-2 production was also measured in RFPEC to serve as a positive control for PECAM-1 pulling, validating the efficacy of PECAM-1 pulling as a sufficient mimic of shear-induced mechanical force. Shear-induced COX-2 production was initially assessed using immunofluorescent staining in RFPEC monolayers after 30 min of fluid shear stress application (Supplemental Figure S5A) and found to be increased significantly with shear (dashed line in Supplemental Figure S5B). Pulling on RFPEC with tipless AFM probes functionalized with PECAM-1 antibody for 30 min resulted in COX-2 activation (Supplemental Figure S5B). Isotype control and bare probe pulling had no significant effect on COX-2 production. Although prior studies have shown that www.nature.com/scientificreports/ the glycocalyx is not active in shear-induced COX-2 production 20 , AFM pulling was also conducted using a glypican-1 antibody functionalized probe and no significant effect on COX-2 production was observed (Supplemental Figure S5B). This is consistent with prior work showing that glycocalyx-degrading enzymes do not affect shear-induced PGI 2 or COX-2 production 20, 22 . These negative findings also show that force application  Production of NO is impaired in glypican-1 knockout mice. Glypican-1 knockout mice (Gpc1−/−) were bred to further characterize the relationship between glypican-1 and NO production in an animal model. Reduced glypican-1 expression was confirmed by immunostaining in Gpc1−/− mice (Supplemental Figure S6). Since phosphorylation of eNOS is required for NO production, we measured peNOS expression in Gpc1−/− mice. Immunostaining revealed that Gpc1−/− mice had significantly reduced peNOS expression in the descending aorta ( Fig. 2A). Mean peNOS expression in Gpc1−/− mice normalized to mean WT expression was 0.77. PECAM-1 expression was not significantly different from that of WT control mice (Fig. 2B). In a second experiment, mice were administered dobutamine, a β-adrenergic agonist that has been shown to increase cardiac flow in mice 30 . WT animals treated with dobutamine for 30 min had significantly increased peNOS expression compared to WT animals treated with saline sham (Fig. 2C). Normalized mean peNOS expression was 1.54 in dobutamine treated WT mice. However, Gpc1−/− mice given dobutamine had no significant change in peNOS (Fig. 2C).

PECAM1 siRNA transfection blocks shear-induced NO production in HUVEC.
To confirm the role of PECAM-1 in force-induced NO production, siRNA transfections were conducted to deplete PECAM1 mRNA and protein in HUVEC, and shear stress was applied using the 6-well rotating disc apparatus. Real time-PCR (qPCR) showed a significant decrease in PECAM1 gene expression (to 2%) for HUVEC transfected with PECAM1 siRNA compared to those transfected with a duplex containing a randomized sequence not present in humans, termed "Negative Control siRNA" and set to 100% (Fig. 3A). Glypican-1 expression was not significantly affected by PECAM1 siRNA; GPC1 mRNA did not significantly change in HUVEC transfected with PECAM1 siRNA compared to controls (Fig. 3B). Western blot confirmed a lack of protein production, showing a significant decrease (to 9%) in PECAM-1 protein expression for HUVEC transfected with PECAM1 siRNA (Fig. 3C). NO was again measured using DAF-2 DA and the shear-induced DAF-2 T levels were normalized to static control levels for each set. Fluid shear stress induced a 2.51-fold increase in NO production in negative control transfected cells compared to static control (Fig. 3D). In PECAM-1 knockdown cells, there was no significant change in static NO production, with a mean value of 0.94-fold compared to negative control transfected cells. After 30 min of shear stress, PECAM-1 knockdown cells did not significantly increase their NO production, with a mean increase of only 1.36-fold, which was not significantly different from that of static PECAM-1 knockdown cells (Fig. 3D).
PECAM1 siRNA transfection blocks NO production elicited from AFM pulling on glypican-1. To further understand the effect of PECAM-1 knockdown on the ability of endothelial cells to produce NO in response to mechanical force, AFM pulling on glypican-1 was conducted in HUVEC transfected with control and PECAM1 targeted siRNA (Fig. 4). Control cells stimulated with AFM pulling by anti-glypican-1 functionalized probes exhibited a significant increase in NO production, with 39% mean activation, consistent with that shown in Fig. 1 for untreated HUVEC. In contrast, PECAM1 siRNA transfected cells stimulated with AFM pulling by anti-glypican-1 functionalized probes did not have any significant response (Fig. 4).
Blocking HS-to-PECAM-1 extracellular interaction does not affect shear-induced NO production. The data in Figs. 1, 2, 3, 4 imply that there is an interaction between glypican-1 and PECAM-1 and that glypican-1 is upstream of PECAM-1. To understand the interaction between PECAM-1 and glypican-1 in shearinduced NO production, heparin and surfen blocking treatments were used to perturb extracellular interactions via HS chains. Figure 5 shows shear-induced NO production in HUVEC after 30 min of shear with heparin and surfen treatments (Fig. 5). In untreated HUVEC, 30 min of shear caused a 2.20-fold change in NO production. When heparin was used to treat HUVEC monolayers for 15 min prior to shear induction, static and shear NO levels were 0.94-fold and 2.05-fold, respectively. Similarly, when surfen was used to treat HUVEC monolayers for 10 min prior to shear induction, static and shear NO levels were 0.68-fold and 1.58-fold, respectively. Neither heparin nor surfen had any effect on the shear-induced increase in NO production (Fig. 5). To confirm the lack of change from heparin or surfen treatment, these experiments were repeated in RFPEC and similar trends were observed (Supplemental Figure S7).

Cleaving HS at the HUVEC surface blocks shear-induced PECAM-1 activation via tyrosine phosphorylation.
To investigate intracellular interactions between the glycocalyx and PECAM-1, heparinase III cleaving enzyme was used to remove HS (the only GAG on glypican-1) from the surface of HUVEC and total tyrosine phosphorylation of PECAM-1 was assessed (Fig. 6). Untreated HUVEC exhibited a 2.10-fold increase in shear-induced PECAM-1 activation compared to untreated static control cells (Fig. 6A). HUVEC treated with heparinase III had only a 1.50-fold increase after fluid shear stress was applied that was not significantly different from that of static heparinase treated controls. PECAM-1 tyrosine residue 686 (Tyr686) was investigated in the same manner to understand how a specific phosphorylation site on PECAM-1 was affected by heparinase III treatment (Fig. 6B). Untreated HUVEC exhibited 1.42-fold increase in shear-induced PECAM-1 Tyr686 activation compared to untreated static control cells. HUVEC treated with heparinase III had no significant increase in levels of PECAM-1 Tyr686 activation after fluid shear stress was applied.

Discussion
The goal of the present study was to elucidate the roles of PECAM-1 and glypican-1 in vascular flow-induced mechanotransduction of NO. Similar to prior work, fluid shear stress was applied to confluent monolayers and NO production activated globally across the monolayer surface was measured. Additionally, novel AFM pulling methods were applied to PECAM-1 and glypican-1 individually to measure NO production activated locally by individual surface protein molecules. Glypican-1's role in NO production was further confirmed in glypican-1 knockout animals. A significant effort was focused on understanding the interdependence between PECAM-1 and glypican-1 in the production of NO by repeating AFM pulling experiments on PECAM-1 knockdown cells. Interestingly, we discovered that PECAM-1 and glypican-1 participate cooperatively in the mechanotransduction of fluid shear stress. While the concept of cooperative mechanosensing is not new 31  www.nature.com/scientificreports/ glypican-1 and PECAM-1 has not been described previously. AFM pulling with glypican-1 probes induced a significant increase in NO production in human cells, consistent with prior work in rat cells 19 , that was blocked in PECAM-1 knockdown cells. Yet direct pulling on PECAM-1 in wild type cells did not elicit a NO response. These findings were consistent across experiments in two cell types, HUVEC and RFPEC. Additional pulling with pointed AFM probes was also conducted as a second approach to ensure access to junctional regions where PECAM-1 is localized and, similarly, did not elicit a NO response. Confirming these in vitro observations,   www.nature.com/scientificreports/ glypican-1 knockout mice had reduced levels of phosphorylated eNOS at baseline as well as after dobutamineinduced increases in blood flow compared to WT mice. Having established that glypican-1 and PECAM-1 act in series to produce NO, and glypican-1 is the upstream shear sensor, we further investigated the mechanism linking HS proteoglycan glypican-1 and PECAM-1. No effect was observed on NO production by blocking extracellular binding between HS and PECAM-1 with heparin and surfen. Shear-induced intracellular PECAM-1 activation was also assessed after HS removal from the EC luminal surface. We used heparinase III to selectively remove HS from the EC surface 32 to block shear stress activation of glypican-1. Since HS is the only GAG on glypican-1, removal of HS blocks shear-induced glypican-1 signaling. This is consistent with earlier studies that showed heparinase III treatment blocked shear-induced NO production in EC 21 . Of course, removal of HS also blocks shear-induced syndecan-1 signaling, but we showed previously that direct pulling on syndecan-1 with AFM probes does not elicit a NO response 19 . We observed that heparinase III-treated HUVEC have diminished shear-induced PECAM-1 tyrosine phosphorylation, with specific impairment occurring at Tyr686. Based on this HS-dependent PECAM-1 activation and the PECAM-1-dependent glypican-1 mechanosensing pathway, we conclude that fluid shear stress sensed through the heparan sulfate chains of glypican-1 is transduced to the interior of the cell, initiating a biochemical signaling cascade that phosphorylates PECAM-1. Activation of PECAM-1 leads to phosphorylation of eNOS 13,31 , which ultimately increases NO production. This process is outlined in the cartoon below, summarizing our overall findings (Fig. 7).
Further work is necessary to explore the intermediate signaling molecules controlling glypican-1 to PECAM-1 activation, but examining prior studies dealing with the signaling to and from these proteins allows us to form a clearer image of how this pathway may function. It is well established that PECAM-1 is activated by c-Src 12,28,29 . Furthermore, Bianco et al. found that glypican-1 and c-Src colocalize in lipid rafts on mouse mammary epithelial cells and that an epidermal growth factor binds to glypican-1 inducing activation of cytoplasmic c-Src 33 . There is also significant evidence for PECAM-1-to-eNOS signaling. PECAM-1 is associated with eNOS in the cell membrane of wild type mice 31 and is necessary for the shear-induced phosphorylation of eNOS at Ser1177 12,34 .
Shear stress likely plays a role not only in stimulating glypican-1 mechanically, but in amplifying the NO response by increasing the cooperation between glypican-1 and PECAM-1. Past work has shown that HS and www.nature.com/scientificreports/ glypican-1 (not syndecan-1) move toward cell junctions in response to 30 min of shear stress, a response that is mediated by glypican-1's GPI-linkage to mobile lipid rafts, which also reorganize toward the cell junction within 30 min of shear 25 . The colocalization of eNOS and PECAM-1 has also been shown to increase with shear stress and peak at the 30 min time point investigated in this study 12 . Thus the detailed mechanism for cooperation between glypican-1 and PECAM-1 in shear-induced NO production discovered here may include glypican-1's GPI-linkage to lipid rafts and associated intracellular eNOS movement to cellular junctions in response to fluid shear, forming a signaling complex with PECAM-1, and more rapidly engaging in intracellular signaling to activate PECAM-1's cytoplasmic tail and engage in eNOS phosphorylation.

Methods
Cell culture. HUVEC (Lonza C2519AS) were statically cultured at 37 °C with 5% CO 2 in EGM-2 (Lonza CC-3162), plated on fibronectin coated surfaces, and grown to confluency. At least 1 h prior to the onset of experiments, culture media was changed to experimental media consisting of phenol red-free, serum-free DMEM supplemented with 1% bovine serum albumin (BSA, Sigma A7284). For fluorescent monitoring of NO production, 5 µM DAF-2 DA (Sigma D225) and 1 µM NO cofactor tetrahydrobiopterin (BH4, Sigma T4425) were added to experimental media and incubated for 40 min prior to stimulation. To investigate whether extracellular binding between PECAM-1 and HS influence shear-induced NO release, heparin and surfen blocking treatments were employed. When added to cell culture media, heparin competitively binds to the extracellular Ig-like domains of PECAM-1 because of its structural similarity to HS and reduces HSto-PECAM-1 binding [35][36][37] . Surfen (bis-2-methyl-4-amino-quinolyl-6-carbamide) is a small molecule antagonist of HS that disrupts HS binding to other proteins 12,31,33,38 . Heparin and surfen treatments were used in the present study to block HS-to-PECAM-1 extracellular binding and shear-induced NO production was measured. When blocking treatments were used they were added to experimental media: 2 h 1215 mU/mL heparinase III (Ibex 50-012-001) treatment 12,[31][32][33]38 was conducted prior to DAF-2 DA incubation and stimulation while 10 min 500 µg/mL heparin (Sigma H3149-10KU) and 10 µM surfen (Sigma S6951) treatments 11,39,40 were added following DAF-2 DA incubation and prior to stimulation.

Mice. All animal protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of
The City College of New York, all experiments were performed in accordance with relevant guidelines and regulations, and reporting follows the recommendations in the ARRIVE guidelines. A pair of heterozygous mutants on the CD1 background (Gpc1−) was a gift from Arthur Lander's lab 41 . Gpc1-were bred to produce homozygous mutants (Gpc1−/−) and wild types (Gpc1+/+). To assess the effects of increases in blood flow 8-week-old mice were treated with dobutamine hydrochloride (Cat. No. 0515; Tocris), following previously published methods 20 . Briefly, mice were given a single intraperitoneal injection of Dobutamine at a dose of 1.5 μg/g body weight, vehicle control mice were given an equivalent volume of 0.9% saline at 5 μl/g body weight. Following 30 min of Dobutamine injection, mice were euthanized by CO 2 asphyxiation and the vasculature was fixed by pressure per- www.nature.com/scientificreports/ fusion as previously described 42 . Briefly, a mid-line surgical incision was made from the abdominal wall to the thoracic wall and the heart was exposed. The inferior vena cava and right atrium were severed and 30 mL of PBS containing 1% BSA were pressure perfused to clear the blood. The vessels were then pressure perfused for 5 min with PBS containing 2% paraformaldehyde. The aorta was dissected and stored in PBS until immunostained. Gene expression using RT-qPCR. RNA was isolated and purified from transfected cells using RNeasy kit (Qiagen) according to the manufacturer's protocol. Purified RNA was converted to cDNA by reverse transcription using Omniscript RT kit (Qiagen). RT-qPCR was performed using SYBR Green PCR Master Mix (Applied Biosystems) on the ABI PRISM 7000 sequence detection system (Applied Biosciences) to determine PECAM1 siRNA knockdown efficiency and check for off-target effects on glypican-1. The expression of housekeeping genes β-actin and GAPDH were used to normalize PECAM1 and GPC1 gene expression. Pellets were mixed with sample buffer, boiled 3 min, and loaded to 7% handcast SDS-PAGE gels. Protein content of lysates was determined using Bradford protein determination assay and 30 µg of total sample protein was mixed with sample buffer, heated to 100 °C for 5 min, and loaded to 7% hand-cast SDS-PAGE gel. Protein bands were separated by electrophoresis and transferred to PVDF membranes using standard methods. Membranes were blocked with 5% BSA in TBS-T solution at room temperature for 45 min. Primary antibodies were diluted in blocking solution and incubated at 4 °C on a rocker platform overnight. Anti-PECAM-1 (Proteintech 11265-1-AP), anti-phosphotyrosine clone 4G10 (Millipore 05-321), anti-PECAM-1 Tyr713 (Sigma SAB4503968-100UG), and β-actin (Sigma A5441) antibodies were used at 1:1,000. Tyrosine 713 is a synonym for tyrosine 686 in the mature PECAM-1 protein, once the 27 amino acid long signal peptide is cleaved. Secondary antibody was used at 1:10,000 and diluted in 3% BSA TBS-T solution.
Fluid shear stress application. Six-well rotating flat cylindrical disc and plate systems were used to apply 20 dynes/cm 2 fluid shear stress via steady laminar flow to cultured EC in a cell culture incubator as described previously 43 .
AFM force mapping and tensile force application. An Asylum MFP-3D atomic force microscope (AFM) was used to conduct experiments following the same procedure as recently published 19 . Briefly, silicon nitride cantilevers with pyramidal tips (k = 0.09 N/m, Asylum Research BL-TR400PB) were functionalized with antibodies and used to determine adhesion forces across several regions of the confluent cell monolayer. Flexible PEG linker molecules (NHS-PEG-Acetal linkers from Linz, Austria) were used to tether anti-glypican-1 (Proteintech 16700-1-AP), anti-PECAM-1 (Proteintech 11265-1-AP), and anti-rabbit IgG (Cell Signaling 2729S) antibodies to probes. Silicon nitride tipless cantilevers (k = 0.08 N/m, Nanoworld PNP-TR-TL) with the same functionalization scheme and antibodies were used with DAF-2 DA loaded cell monolayers to measure induced NO. Fibronectin coated glass gridded slides (EMS 63405-01) were used to track force mapping location for subsequent fluorescence imaging and comparison to unprobed regions 2 mm away from probing site on the same slide. The AFM force map settings were as previously described 19 . In adhesion force experiments with pyramidal tips 3 nN maximum indentation force was used, while in DAF-2 induction experiments 10 nN maximum indentation force was used with 10 s dwell time at the endothelial surface. Indentation (z-axis) speed was 1.6 µm/s and XY speed was 400 nm/s in all cases. Probing area in both studies was 20 µm × 20 µm and the number of indentations per scan was adjusted to fit 30-min total elapsed time (10 points × 9 lines).

Immunofluorescence and confocal imaging of cultured cells. Following mechanical stimulation
with either fluid shear stress application or AFM force application, cell monolayers loaded with DAF-2 DA and expressing DAF-2 T were fixed in 2% paraformaldehyde (PFA) for 10 min, counterstained with DAPI, and mounted for confocal imaging. Cyclooxygenase-2 (COX-2) staining was also conducted after mechanical stimulation. www.nature.com/scientificreports/ and the point of intersection between static and activated regions on pixel intensity histogram curves were used to threshold images before contrast stretching back to original size. Mean fluorescent intensities (MFIs) were calculated from the average grey value of each of these images and MFIs were normalized by the average MFI of static control samples. To calculate percent activation, normalized MFIs were converted to percentages by subtracting 1 and multiplying by 100.
En face immunostaining of the mouse endothelium. The expression of specific proteins was assessed at the endothelium of the thoracic descending aorta by en face immunostaining. Fixed arteries were processed for immunostaining using specific antibodies for phospho-eNOS (Cat. No. 074281, Millipore), PECAM-1 (553373, BD Pharmingen), glypican-1 (Cat. No. SC66909, Santa Cruz) and the nuclei were stained using DAPI (Cat. No. D1306, Thermo Fisher Scientific). The stained tissue was mounted on a microscope slide en face, followed by confocal laser imaging using a ZEISS LSM 800 microscope. Protein expression was quantified as fluorescent intensity using Image J.
Data analysis. Results are presented as mean + SEM for all data sets and differences were considered significant when p < 0.05. GraphPad Prism was used to perform statistical analyses. When comparing two groups, statistical analyses were performed using unpaired two-tailed student's t-test. When comparing one group to an artificially homogenous group (such as activation percentages for AFM probed regions vs 0% static regions), one sample t-test was used. When comparing more than two groups, one-way ANOVA was performed followed by the Sidak method for multiple comparison testing.

Data availability
All data generated or analysed during this study are included in this published article (and its Supplementary Information files).