## Introduction

The endothelial glycocalyx (GCX) is a glycosaminoglycan and proteoglycan-rich functional layer lining the endothelium which promotes homeostasis through its role as a mechanosensor and as a barrier to inflammatory cell attachment. Disruption of the GCX in pathological conditions such as type 2 diabetes (T2D) and atherosclerosis1,2 causes endothelial cell (EC) dysfunction and thus, leads to cardiovascular disease progression. In particular, in vitro and in vivo degradation of heparan sulfate (HS), the main component of EC GCX, has been shown to impair shear-regulated nitric oxide (NO) production and promote adhesion of inflammatory cells3,4,5. Given its central role in EC homeostasis, the GCX is considered a potential therapeutic target to prevent or alleviate EC dysfunction and therefore benefit clinical outcomes6,7.

Empagliflozin (EMPA; Jardiance ®) belongs to the sodium-glucose cotransporter 2 inhibitor (SGLT2i) class of drugs used in the treatment of T2D and more recently heart failure. This class has emerged as a promising new therapy for cardiovascular diseases as treatment with a SGLT2i has shown impressive beneficial outcomes in recent clinical trials (e.g. EMPA-REG OUTCOME (NCT01131676)8, DAPA-HF (NCT03036124) and EMPEROR-Reduced (NCT03057977)9,10). Among these, the EMPA-REG OUTCOME trial demonstrated significant reductions in the rate of hospitalization for heart failure and cardiovascular death in T2D patients treated with EMPA when compared with placebo8. Subsequent animal studies have suggested pleiotropy and extended protective effects by EMPA to atherosclerosis11, heart failure12,13 and ischemia–reperfusion injury14 in non-diabetic conditions, suggesting that the cardiovascular protection is not achieved solely through glycosuria.

As endothelial dysfunction is a central step in the development of cardiovascular complications, it has been hypothesized that SGLT2i can prevent or overcome EC dysfunction. Treatment with EMPA has been shown to improve EC-dependent vasorelaxation in diabetic rodents15,16,17. El-Daly et al. showed similar results with ex vivo mice aortic tissue in response to hyperglycemic conditions18. Moreover, EMPA reduced the inflammatory response and the adverse vascular remodeling in Ang II-induced AAA of ApoE(−/−) mice. Adhesion of inflammatory cells and adhesion molecule expressions in isolated mouse ECs were also reduced19. Recent in vitro studies suggested that EMPA restored NO bioavailability and reduced oxidative stress in TNF-α treated ECs20,21. Although cardiac and vascular protective mechanisms have been proposed22, the exact mechanism by which EMPA promotes EC homeostasis and in turn contributes to better vascular health remains to be clarified.

Considering the importance of GCX health in EC function and the impressive ability of EMPA to improve cardiovascular outcomes, we hypothesized that EMPA can overcome EC dysfunction caused by GCX disruption which could, in part, explain the improvement in vascular health. We previously showed that EMPA can normalize the inflammatory response of ECs possibly through shear mechanotransduction of a restored GCX after acute degradation23. As the GCX is a dynamic surface layer with potential of regrowth, this work aims at exploring more closely the signaling pathways involved under sustained GCX disruption to identify other anti-inflammatory mechanisms of EMPA.

## Results

### EMPA induces an anti-inflammatory response independent of HS

Sustained HS degradation was used to evaluate the inflammatory response of ECs to EMPA independent of HS. Inflammatory cell adhesion assays were performed on TNF-α treated cells similar to previous studies3,23,24. EMPA significantly reduced NB4 cell adhesions compared to control (Fig. 1, EMPA vs CTL *p < 0.001). While sustained degradation caused significant increases in NB4 cell adhesion (sDEG vs CTL *p < 0.001) showing an enhanced inflammatory response, EMPA reduced the adhesions to the same level as non-degraded samples (sDEG-EMPA vs sDEG #p < 0.001) and below control (sDEG-EMPA vs CTL *p < 0.001). This suggests that EMPA prevents the inflammatory response of ECs to TNF-α stimulation and HS degradation. As it persists under sustained degradation, it can be assumed that EMPA can reduce inflammation independently of HS.

### ICAM-1 and eNOS are not implicated in the HS independent EMPA anti-inflammatory response

The expression of eNOS and ICAM-1 were assessed to characterize the inflammatory response induced by sustained HS degradation and their potential role in the recovery effect provided by EMPA. After 24 h of culture, eNOS expression in all static and shear TNF-α + conditions were reduced compared to TNF-α – controls (Fig. 2a, *p < 0.001). There was no observed effect of EMPA nor degradation on eNOS expression. Shear significantly increased the phospho-eNOS Ser1177 levels compared to the static control in absence of TNF-α (Fig. 2b, CTL TNF-α – shear vs CTL TNF-α – static #p < 0.001). TNF-α stimulation impaired the shear-induced increase of phospho-eNOS Ser1177 to similar levels in all shear conditions suggesting reduced eNOS activation (*p < 0.001). Again, no significant impact on phospho-eNOS Ser1177 caused by EMPA or degradation were found. Similarly, TNF-α induced the expression of ICAM-1 to similar levels in all conditions (Fig. 2c, *p < 0.001) with no significant effect of shear, EMPA or degradation. Interestingly, the phospho-eNOS Ser1177/total eNOS ratio remained unchanged with or without TNF-α in shear condition while significantly higher ratios were found in static culture (Fig. 2d, *p < 0.01). Taken together, these results suggest that the activation of eNOS and the expression of ICAM-1 are not directly implicated in the HS independent recovery of EC dysfunction by EMPA.

### The induction of the unfolded protein response caused by sustained HS degradation is partially restored by EMPA

To explore the possible signaling pathways impacted by the sustained degradation of HS and EMPA beyond the traditional mechanobiology and anti-inflammatory role of the GCX, a genome-wide transcriptome analysis was performed. We sought to better understand the mechanisms by which EMPA prevents the increased inflammatory response triggered by the sustained HS degradation under shear culture. EC dysfunction associated genes were screened and grouped to identify the signaling pathways involved based on their fold changes.

Shear exposure for 24 h showed regulation of hallmark genes such as eNOS (NOS3), KLF2 and COX-2 (PTGS2) while downregulating inflammatory signaling genes (IL1B, IL6) compared to the static control (CTL vs CTL (Static)). mRNA expression of ICAM-1 was also slightly downregulated by shear (Supplementary Table S1).

### N-Acetyl-l-cysteine treatment reduces NB4 cell adhesion similar to EMPA

To further study the reduction in NB4 cell adhesion caused by EMPA, we sought to determine if the effect was mediated through the regulation of oxidative stress under shear and sustained HS degradation. Prior to the adhesion assay, ECs were treated with N-Acetyl-l-cysteine (NAC) during shear exposure for 24 h. Unlike EMPA, NAC did not significantly reduce the number of adhered NB4 cells compared to the control (Fig. 5, NAC vs CTL). However, under sustained degradation, treatment with NAC resulted in a significant reduction in adhesions (sDEG-NAC vs sDEG $p < 0.001), comparable to the effect provided by EMPA (sDEG-EMPA vs sDEG$p < 0.001). This could suggest that the mitigation of oxidative stress during shear exposure is a potential mechanism by which EMPA-treated ECs reduce their inflammatory response to sustained HS degradation.

## Discussion

The disruption of the endothelial GCX has been associated with pathological conditions of the vasculature and impaired endothelial function making the role of a healthy GCX of interest in multiple diseases and treatment options. Normal GCX mechanotransduction helps maintain EC homeostasis through shear-regulated functions such as NO-driven vasodilation, cell barrier function and inflammatory cell transmigration. In this regard, treatment with SGLT2i has been demonstrated to improve vascular functions in vivo and in vitro in non-hyperglycemic conditions through improvement of EC homeostasis13,17,19,20,21,26,27. Recent studies have demonstrated that EMPA can improve EC function but the mechanisms involved remain elusive.

Given the potential of EMPA to remediate EC dysfunction, we sought to study the effects of EMPA in a pro-inflammatory chronic GCX disruption model by using sustained HS degradation. In this study, we were able to show the impact of sustained GCX disruption on the induction of ER stress and a possible mechanistic effect of EMPA to attenuate the associated inflammatory response which could contribute to the improved vascular health observed clinically8,9,10.

Under normal conditions, the GCX at the apical surface of ECs allows the transduction of shear stress through the cytoplasmic membrane, transforming sensing forces into biochemical signals28. Thus, laminar shear stress induces the adaptive remodeling of the cytoskeleton i.e. cell elongation and the production of NO through the activation (phosphorylation Ser1177) of eNOS29,30. Upon disruption of the surface GCX, these transduced effects are impaired, leading ECs to a pro-inflammatory state, promoting inflammatory cell adhesion3. Hyperglycemic conditions have been shown to cause vascular GCX disruption in vivo1,31 and in vitro32 thus treatment with EMPA was studied for its potential benefits on HS integrity, which could prevent EC dysfunction.

The sustained HS degradation in our study reduced the HS mean intensity at EC surface by approximatively 85% and prevented regrowth during the 24 h culture period (Supplementary Fig. S1) which is in line with previous in vitro studies33,34. EMPA did not improve HS intensity in control condition or under sustained degradation. Similarly, treatment with EMPA did not overcome the shear-induced elongation loss under sustained HS degradation (Supplementary Fig. S2) but resulted in significantly less NB4 cells adhered to the EC surface under degraded condition (Fig. 1). Taken together, these results define the mechanosensory and anti-inflammatory role of the GCX as HS-dependent and indicate that EMPA can reduce inflammation independently of HS-mediated EC functions.

The impact on inflammation downstream of known GCX mechanotransduction was assessed through eNOS homeostasis and ICAM-1 expression to characterize the inflammatory response of treated ECs. Degradation of HS has been linked to reduced shear-induced NO production in ECs5,35. Ebong et al. and others further demonstrated that the impairment of NO production is mediated through reduced phospho-eNOS Ser1177 levels and is concomitant with loss of EC remodeling30,36. In our case, stimulation with TNF-α resulted in significant reductions in eNOS and phospho-eNOS Ser1177 levels with no further decrease due to HS degradation, indicating that TNF-α may have eclipsed the impact of degradation on NO production (Fig. 2). Nonetheless, no improvement of phospho-eNOS/eNOS levels were found with EMPA, indicating that eNOS activation might not contribute directly to the anti-inflammatory effect of EMPA. Although we did not find significant changes in the expression level of phospho-eNOS/eNOS induced by EMPA, recent studies have shown improved NO production and EC-dependent vasodilation in response to EMPA. Treatment with 1 μM EMPA restored NO bioavailability in TNF-α stimulated cardiac microvascular ECs (CMECs)21 and human coronary arterial ECs (HCAECs) while eNOS phosphorylation levels remained unchanged under static conditions20. Conflicting results exist regarding the effects of EMPA on inflammation and adhesion molecule expressions. Ortega et al. showed the reduction in ICAM-1 and VCAM-1 expression in aortic sections of Ang-II stimulated mice19. Moreover, reduction of inflammatory cell adhesion in a dose-dependent manner was demonstrated with Ang-II stimulated HUVECs. Expression of ICAM-1 and VCAM-1 was also found to be reduced in CMECs isolated from EMPA-treated mice17. Conversely, expression of ICAM-1 and VCAM-1 remained unchanged in TNF-α stimulated CMECs21 and HCAECs20 treated with 1 μM of EMPA. We previously showed that ICAM-1 expression correlates with NB4 cell adhesion3, however, the protein expression of ICAM-1 (Fig. 2) were unaffected by sustained HS degradation or EMPA suggesting that regulation of the adhesion molecule ICAM-1 may not contribute to the variations in the number of NB4 cell adhesions.

Treatment with EMPA caused small but significant reductions in the transcriptional expression of ATF3, DDIT3 (CHOP), HSPA5 (Grp78), PPP1R15A, and TRIB3 under HS degradation suggesting a dampened apoptotic response (Fig. 3). Other studies have also found similar dampening effects of EMPA on ER stress. Treatment of ApoE(−/−) and C57Bl/6 mice with EMPA for 5 weeks resulted in the reduction of hepatic ER stress as shown by the downregulation of ER stress signaling genes including HSPA5 (Grp78), ERN1 (IRE1), XBP1, DDIT3 (CHOP), ATF4, ATF6 and GADD45A and the reduced ratio of phospho-eIF2α/eIF2α46,47. A study by Zhou and Wu demonstrated the potency of EMPA on ER stress related cardiomyopathy of diabetic rats48. Treatment with EMPA for 8 weeks resulted in improved cardiac functions along with reduced cardiomyocyte apoptosis and protein expression of ER stress markers (Grp78, CHOP and cleaved caspase-12) suggesting a lesser impact of ER stress on the cardiomyocytes.

ER stress is inherently tied to oxidative stress as the adaptive protein folding is coupled to ROS production49. Additionally, oxidative stress can be exacerbated through the action of TXNIP which is related to ER stress by its concomitant activation. TXNIP is a critical regulator of ROS signaling implicated in diverse vascular pathologies due to its role in downstream inflammation and apoptosis. Through its inhibitory effect, TXNIP can impair the thiol-redox system thioredoxin (TRX), promoting oxidative stress. Lerner et al. demonstrated how activation of the UPR sensors, IRE1α and PERK, resulted in increased TXNIP mRNA levels and increased ROS levels in mouse embryonic fibroblasts50. Deletion of IRE1α, PERK or TXNIP abolished the detrimental effects associated with unresolved ER stress. Moreover, disturbed flow has been shown to upregulate TXNIP protein expression which caused increased leukocyte–EC adhesion through increased ICAM-1 and VCAM-1 expression51. TXNIP is also known to interact with and activate the NLRP3 inflammasome resulting in enhanced inflammation as well as promoting apoptosis through the dissociation of TRX from the apoptosis signal-regulating kinase 1 (ASK1)52,53.

Given the detrimental effects associated to the expression of TXNIP and its implication in ER stress, we hypothesize that it plays a regulatory role in the effects provided by EMPA. Indeed, TXNIP mRNA levels were downregulated by EMPA after 6 h in static condition (Fig. 4). Sustained HS degradation caused the loss of the shear-induced downregulation of TXNIP mRNA and an increase in TXNIP protein. EMPA respectively reduced and normalized the mRNA and protein expression, which correlates with the trend observed in NB4 cell adhesion. These results are in accordance with previous studies showing the role of shear-regulated TXNIP in the inflammatory response of HUVECs52 and the correlation between TXNIP expression and leukocyte–endothelium adhesiveness51.

The transcriptomic profile of degraded samples (sDEG and sDEG-EMPA vs CTL) was also characterized by the differential regulation of redox genes (Supplementary Table S1) suggesting an adaptive response to oxidative stress. While we were unable to show the direct reduction of ROS by EMPA in static culture (Supplementary Fig. S3) or under shear (data not shown), a few studies have shown that SGLT2i can effectively reduce oxidative stress in different conditions. Pre-clinical studies demonstrated the potential of EMPA to reduce whole blood oxidative burst, cardiac NAPDH oxidase activity as well as aortic and endothelial ROS levels in diabetic rats15,16. Zhou et al. also showed reduction in intracellular and mitochondrial ROS by EMPA in CMECs isolated from diabetic rats17. In vitro treatment of CMECs/HCAECs with 1 μM EMPA resulted in the reduction of TNF-α induced oxidative stress as measured by intracellular and mitochondrial ROS suggesting that EMPA has a protective effect against ROS production in normoglycemic conditions while a direct anti-oxidant effect was excluded by the authors.

By virtue of its role in EC homeostasis, the disruption of the GCX also coincide with increased ROS production, promoting a vicious circle of degradation as ROS will further promote GCX disruption54,55. In this context, unresolved ER stress with increased TXNIP expression can cause ROS production and may result in increased NB4 cell adhesion. To test this, NAC was used to promote ROS scavenger and mitigate oxidative stress during shear exposure. Perfusion of NAC has been previously shown to prevent hyperglycemic damage of the glycocalyx1 and to reduce the level of hyaluronan shedding during renal fluid resuscitation56. In our case, incubation with NAC resulted in similar reductions of adhesion compared to EMPA suggesting an oxidative stress mediated mechanism by EMPA. However, the exact impact of EMPA on the interplay between ER stress and TXNIP, resulting in reduced oxidative stress and in turn reduced inflammation, remains to be clarified. TXNIP knockout ECs could represent an interesting model to test the EMPA effect dependency on TXNIP expression.

The study of the glycocalyx in vitro as well as conclusions drawn from transcriptional data comes with inherent limitations. First, the glycocalyx of ECs in culture is known to lack the thickness of its in vivo counterpart which may lessen the impact of degradation on inflammatory cell adhesion57. Although different from chronic GCX disruption, sustained enzymatic degradation of HS was used as an inflammatory stimulus to mimic EC GCX shedding as found in diabetic or curved/branching vasculature. Furthermore, while the effect of HS degradation was the focus of this study, degradation of other components of the GCX can also impact EC homeostasis which could induce different signaling cascades and therefore change the role of EMPA in the restoration of GCX functions. Although the use of TNF-α under sustained HS degradation does not allow to fully dissociate their respective impact on EC inflammation in all experiments, TNF-α addition was required to perform adhesion assays and was therefore use in subsequent experiments to properly interpret the results. The single time-point studied does not reflect transient response or feedback regulation between ER stress and inflammation. Moreover, a suprapharmacological concentration of EMPA was used compared to the reported serological concentrations of treated patients58 in order to study the reduction in inflammation previously reported and to highlight the potential cellular mechanisms that may occur under lower serological concentrations over longer periods of time.

In conclusion, the present study suggests a possible mechanism by which EMPA may contribute to better endothelial health. Our results point toward the potential of EMPA to attenuate EC inflammation independently of HS-mediated functions, as shown by the reduction of inflammatory cell adhesion under sustained HS degradation. Downregulation of pro-apoptotic signaling and TXNIP expression by EMPA may contribute to oxidative stress mitigation therefore limiting oxidative damage under unresolved ER stress and resulting in reduced inflammatory cell adhesion. Future work should aim at studying the regulatory mechanisms associated with the expression of TXNIP on the inflammatory response through the NLRP3 inflammasome and apoptotic signaling. A better understanding of the effects of SGLT2i on GCX-mediated EC functions could benefit new strategies to limit EC dysfunction in pathological conditions where chronic GCX damage is found.

## Materials and methods

### Cell culture and 3D culture model

Human abdominal aortic endothelial cells (HAAECs, ATCC, Coriell, CRL-2472) were cultured in endothelial cell growth medium (Promocell, C-22020), supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin (Invitrogen) in culture flasks coated with 0.1% pig gelatin at 37 °C and 5% CO2. Cells were grown to confluence and harvested from the flask with TrypLE (ThermoFisher, 12605028). Experiments were performed with cells at passage 5 and 6.

Tubular cell culture models were produced as previously described59,60. Briefly, the silicon elastomer Sylgard™ 184 (Paisley, AVDC0184-39) was prepared by mixing the base and curing agent in a 10:1 ratio. Sylgard™ 184 was cured around a polished stainless steel rod in plexiglass molds. Following complete curing, the rod was removed, creating a straight hollow tube where cell can be seeded and grown. In preparation to cell culture, models were sterilized by boiling and coated with a 40 μg/mL human fibronectin (Akron Biotech, AK9715-0005) solution overnight. HAAECs were seeded at a concentration of 1 × 106 cells/mL and grown for 48 h. Models were mounted on a rotor to ensure complete and even coverage of the inner surface. Cells were then cultured for 24 h in static or shear condition (laminar shear stress of 10 dyne/cm2) with or without 50 μM EMPA.

GCX disruption was performed by enzymatic degradation of HS. HAAECs were treated with 0.5 U/mL of heparinase III (Sigma unit, enzyme 4.2.2.8, IBEX Pharmaceuticals, 50-012) in FBS-free media for 2 h followed by the addition of 0.1 U/mL in the medium during the 24 h of static or shear culture to achieve sustained degradation (sDEG) and to prevent HS regrowth. The initial condition of degradation (DEG T0) was assessed immediately after the initial 2 h exposure to heparinase III.

### Heparan sulfate immunostaining and confocal microscopy

Cells were fixed in situ with a 2%/0.1% paraformaldehyde/glutaraldehyde solution in PBS for 30 min without prior washes to prevent HS disruption. Sections of the models were cut and samples were blocked with a 2% normal goat serum (NGS) in PBS. Primary antibody for HS (1:100, mouse, 10E4, Amsbio, 370255) was diluted in 1% NGS/PBS and incubated overnight at 4 °C. Cells were rinsed with PBS and incubated for 1 h with the secondary antibody (ThermoFisher, goat anti-Mouse IgG, IgM (H + L) Alexa Fluor 488, A-10680) and DAPI (1:1000) diluted in 1% NGS/PBS. Cells were rinsed with PBS and the model section were cut, mounted using 0.2% Dabco/Glycerol (1:5, Sigma) and imaged via laser scanning confocal microscopy (Zeiss Exciter 800). Z-stack images were acquired at 10 × magnification and maximum intensity projections were created to quantify the mean gray value of each image using ImageJ.

### Cell morphology and shape index

Following the static or shear culture with or without EMPA, cells were fixed in situ with 1% paraformaldehyde in PBS for 20 min, rinsed with PBS and stained with a 4% crystal violet (BD Biosciences) solution for 10 min. Cell morphology was assessed through light microscopy (Leica DMIL microscope and Leica DC300 camera) images and quantified by computing the shape index of individual cells using Matlab scripts59. The shape index (SI) is used as a metric of cell elongation where a perfectly round cell has a SI close to 1 and a straight elongated cell, a SI closer to 0.

Adhesion assays were performed as previously described24. Acute promyelocytic leukemia (NB4) cells were cultured in suspension and maintained at a concentration of 2 × 105 to 1 × 106 cells/mL in RPMI 1640 medium with 2 mM l-glutamine supplemented with 10% FBS and 1% penicillin–streptomycin. 48 h prior to the adhesion assay, NB4 cells were stimulated with 10–6 M all-trans-retinoic acid (ATRA, Sigma, R2625) to induce differentiation into granulocytes. Adhesion assays were conducted following 24 h of shear exposure and tumor necrosis factor-alpha (TNF-α; PeproTech, 300-01A, 10 ng/mL) stimulation of HAAECs to induce NB4 cell adhesion. As previously shown, TNF-α stimulation is necessary to reach significant numbers of NB4 cell adhesions24. Also, to assess the effect of oxidative stress on adhesion, HAAECs were treated with 1 mM of N-Acetyl-l-cysteine (NAC, Sigma, A9165) for 24 h prior to the adhesion assay. A NB4 cell suspension at 5 × 105 cells/mL was perfused through the models at a shear stress of 1.25 dyne/cm2 for 1 h. PFA 1% was then perfused to fix the adhered NB4 cells to the endothelial monolayer while removing non-adherent NB4 cells. The number of adhered NB4 cells was determined by manually counting cells from light microscope images (3–5/sample) at a magnification of 4 ×.

### Protein collection and western blot

HAAECs were washed with cold PBS and lysed in situ in cold RIPA lysis buffer (ThermoFisher, 89900) with 2% Halt™ protease and phosphatase inhibitor cocktail (ThermoFisher, 78440). Samples were vortexed and kept on ice before centrifuging at 13,000 RPM for 10 min. Protein quantification was performed through colorimetric assays and protein quantities were normalized between samples. Loading samples were prepared with Bolt LDS sample buffer and reducing agent (ThermoFisher, B0007, B0009). Gel electrophoresis were performed using the Mini Blot Gel tank and pre-cast Bolt 4–12% Bis–Tris Plus gels loaded with 20 μg of protein. Protein transfer was carried out using the Mini Blot module and 0.45 μm nitrocellulose membranes following manufacturer instructions. Following the transfer, membranes were blocked with 5% bovine serum albumin (BSA, BioShop, ALB001) in 0.1% Tween-20 TBS (TBST) for 1 h. Primary antibodies were diluted in blocking buffer and incubated at 4 °C overnight on a shaking plate. Antibodies consisted of eNOS (1:500, BD Biosciences, 610297), phospho-eNOS Ser1177 (1:500, Abcam, ab184154), ICAM-1 (1:5000, Abcam, ab109361), GAPDH (1:10,000, ThermoFisher, AM4300) and TXNIP (1:1000, Abcam, ab188865). After 3 washes in TBST, the membranes were incubated with horseradish peroxidase secondary antibodies (1:40,000, ThermoFisher) diluted in blocking buffer for 1 h at room temperature. Membranes were washed 3 times in TBST and signal detection was achieved with SuperSignal™ West Pico PLUS Chemiluminescent Substrate and UVP Biospectrum 810 MultiSpectral Imaging System. Proteins were quantified by densitometry using ImageJ and normalized to the loading control (GAPDH).

### Transcriptome analysis

The transcriptome profiles of HAAECs under two sets of conditions were studied. Cells were statically cultured with or without EMPA for 6 h or cultured under similar conditions to the adhesion assay (24 h shear culture, TNF-α stimulation, CTL (Static), CTL, sDEG and sDEG-EMPA). Following treatment, cells were lysed in situ and RNA isolation was performed using the RNeasy Plus Micro kit (QIAGEN) according to manufacturer instructions. Sample RNA integrity was confirmed with the RIN using a BioAnalyzer (Agilent). The genomic transcriptome profile of each sample was determined with a Human Clariom S Array and analyzed using the Transcriptome Analysis Console software (SST-RMA method, Applied Biosystems, v4.0.2). Gene lists were generated by filtering minimal threshold expression and fold changes (≤ − 1.5 or ≥ 1.5). The open-source pathway database REACTOME was used to analyze the gene lists and to identify potential pathways based on the number of matched entities61.

### Reactive oxygen species measurement

Reactive oxygen species (ROS) production was measured using the 2′,7′-dichlorofluorescin diacetate (DCFDA) assay kit (Abcam, ab113851) according to the manufacturer instructions. Briefly, ECs were cultured in 96-well plates at a density of 30,000 cells/cm2 and stained with 10 μM H2DCFDA for 30 min prior to treatments. Tert-Butyl hydroperoxide (TBHP) at 50 μM was used as positive control, with or without 50 μM EMPA for 6 h. Pretreatment with EMPA (preEMPA + TBHP) was also tested by incubating with EMPA for 18 h prior to the assay. TNF-α (10 ng/mL) and sDEG over 24 h were also tested. After incubation, ROS production was measured by fluorescence unit on a Berthold fluorescence microplate reader with the FITC filter set. Measurements were normalized to the control sample i.e. H2DCFDA stained cells with no TBHP or EMPA treatment.

### Statistical analysis

Statistical analysis was performed using GraphPad Prism™ software. Results are presented as the mean values ± standard deviation of 3 independent experiments. Each experiment included 2–6 replicates for each condition and the mean was used in the analysis. When applicable, the normality of the data was confirmed by D’Agostino and Pearson test. Otherwise, the data was assumed to be normally distributed. Mean values were compared using t tests, one-way or two-way analysis of variance (ANOVA) followed by Bonferroni post-test with a 95% confidence interval. Statistical analysis of the transcriptome data was carried out through the Transcriptome Analysis Console software using default settings with eBayes analysis. p-values less than 0.05 were considered significant.