Endothelial cells form a monolayer called the endothelium covering the inner surface of blood vessels. When discovered in 1865, the endothelium was described as an inert barrier between blood and tissues. However, it is now well recognized that endothelial cells are capable of plasticity and phenotypical changes according to the microenvironment. At the whole-organism level, endothelial cells share essential functions in a multitude of physiological processes. They regulate vasomotor tone, vascular permeability, angiogenesis, immune cell trafficking, inflammation, coagulation and transport of nutrients, hormones, growth factors and particles. Importantly, the endothelium also exerts a number of organ-specific functions which conditioned the heterogeneity of endothelial cells (13). However, when dysfunctional, they strongly contribute to the pathogenesis of cardiovascular disease, including VTE (14).

In physiological conditions, endothelial cells guarantee an anti-coagulant and anti-inflammatory state through the integration of complex mechanisms to avoid contact between proteins from the coagulation cascade and blood cells. An intact endothelium actively regulates the coagulation response through potent inhibitory processes. Endothelial cells express various anticoagulant molecules, such as tissue factor pathway inhibitor (TFPI), thrombomodulin (TM), endothelial protein C receptor (EPCR), and heparin-like proteoglycans constituting the glycocalyx. The endothelium also expresses ectonucleoside triphosphate diphosphohydrolase-1 (ENTPDase1/CD39) which metabolizes ATP into adenosine preventing platelet activation and aggregation (15). Other mechanisms include endothelial synthesis of nitric oxide (NO) and prostacyclin I2 (PGI₂). These molecules inhibit platelet adhesion and aggregation and vasodilation of the vessel (16).

TFPI limits the action of tissue factor (TF), responsible for the activation of the extrinsic pathway. TFPI acts via factor Xa (FXa) to inhibit the formation of the TF-FVIIa complex (17). In addition, the TM/thrombin complex binds and activates protein C, which in turns binds to protein S and inactivates FVa and FVIIIa (18). Endothelial cells also regulate coagulation by producing the serine proteases, tPA (tissue-type plasminogen activator) and uPA (urokinase-type plasminogen activator) that cleave plasminogen into plasmin (19, 20). Finally, the glycocalyx, which is negatively charged, repels negative molecules thus preventing activation of the coagulation cascade (21). The glycocalyx also fosters the formation of complexes between antithrombin III and thrombin (factor II) or with other serine proteases (factor IXa, Xa, XIa and XIIa) (22).

Additionally, the glycocalyx provides the endothelium with a negatively charged surface contributing to the anti-adhesive properties of the endothelial cell surface. Initial platelet adhesion to VWF is also prevented by a disintegrin-like and metalloprotease with thrombospondin type I repeats−13 (ADAMTS13) cleaving VWF multimers released from the Weibel-Palade bodies (WPBs) (23, 24). In vivo studies using experimental models of inflammation showed that ADAMTS13 deficiency is associated with enhanced WPB secretion and increased leukocyte adhesion and extravasation (25). This was associated with increased endothelium-platelet interactions and spontaneous thrombus formation in veins (26). Importantly, platelet depletion in ADAMTS13 deficient mice reduced leukocyte rolling in unstimulated veins suggesting that platelet and endothelial VWF promote leukocyte arrest thus facilitating thrombus formation (25).

Sustained activation of endothelial cells by inflammatory stimuli, including circulating pathogen-associated molecular pattern (PAMPs), DAMPs, cytokines, chemokines, complement proteins and ROS cause alteration of the endothelial function (27, 28). The end result is an impairment of endothelial anti-coagulant, anti-inflammatory and immune-dependent properties, which are hallmarks of endothelial dysfunction (Figure 1). Experimental models of VTE showed that interferon-γ (IFN γ), transforming growth factor-β (TGF β), tumor necrosis factor-α (TNF α), interleukin-6 (IL-6), IL-17, IL-9, IL-1β contribute to thrombus formation and resolution through endothelial activation. Chemokines such as CXCL8/CXCL1 and CCL2 are also involved (29). It is important to note that endothelial cells interact with the complement system via the expression of specific receptors (C1q, C3a or C5a receptors) and complement proteins (C1, C3, C5, factor B) further contributing to inflammation (30). Lastly, oxidative stress mediated by ROS induce several signaling pathways leading to endothelial dysfunction (31).

Due to its localization, the endothelium plays a crucial role in supporting thrombus formation and resolution in collaboration with platelets and leukocytes. During the initiation phase of thrombus formation, endothelial cells, platelets and leukocytes (especially neutrophils and monocytes) activate each other leading to a highly pro-coagulant environment and the recruitment of additional blood cells. Interactions and mutual activation during thrombo-inflammation between endothelial cells, neutrophils, monocytes and platelets has been extensively reviewed elsewhere (32–35). Briefly, leukocyte recruitment to regions of thrombo-inflammation is primarily initiated by direct interaction between leukocytes and the inflamed endothelium. The activated endothelium expresses P-selectin allowing the rolling and the adhesion of neutrophil via the P-selectin glycoprotein ligand-1 (PSGL1). The upregulated expression of VCAM-1 (vascular cell adhesion protein) and ICAM-1 (intercellular adhesion molecule) allows a stronger adhesion of neutrophils on the endothelial surface. A key process linking inflammation to thrombosis is the formation of neutrophil extracellular traps (NETs) by diverse stimuli including cytokines, activated platelets via P-selectin, DAMPs and ROS. NET formation is conditioned to chromatin decondensation via H3 and H4 histone citrunillation by peptidylarginine deiminase 4 (PAD4). NETs trigger coagulation via the release of enzyme including neutrophil elastase (NE), cathepsin-G or myeloperoxidase (MPO). This induces FXII activation and inactivation and degradation of TFPI and TM. NETs also contain TF and favor platelet recruitment and activation via VWF, histones, fibrinogen and fibronectin (1, 36–38).

The inflamed endothelium releases WPBs containing VWF, P-selectin and other pro-coagulant and pro-inflammatory components (cytokines and chemokines) in the extra-cellular environment (23). Platelet adhesion and activation are related to integrin binding on endothelial cell surface. Interaction of GPIbα with VWF is crucial for platelet accumulation along the inflamed endothelium mostly in arterial thrombosis. Hence, high shear stress exposes A1 domain of VWF multimers, important for GPIbα binding. This mechanism was also found to drive the onset of thrombus initiation in venous thrombosis in the stenosis model of VTE. To explain these interactions, it was suggested that GPIbα interacts with VWF multimers with pre-exposed A1 domain (39). Platelet GPIIb/IIIa binds to fibrinogen, thereby crosslinking platelets with platelets and platelets with endothelial cells. Once activated, platelets release their granule content further fueling platelet and endothelial activation. Among these factors, selectins are important for platelet-endothelial and platelet-leukocyte interactions. In addition, platelet secrete several soluble factors affecting leukocyte function. These molecules including IL-8, platelet factor 4 (PF4), CCL5, CCL7, CCL3, CXCL4 and CXCL5 are important for the recruitment and tethering of neutrophils to the nascent thrombus in association with selectins. Importantly, enzymes secreted by the activated neutrophils mediate fibrin formation by platelets. The integrin CD11b/CD18, activated following neutrophil adhesion to platelet, further develops the pro-coagulant activity of activated platelets (9, 34, 40). Platelets also participate to neutrophil activation and NET formation. The reciprocal interactions between platelets and neutrophils have been demonstrated both in in vitro and in vivo experimental models including in VTE (2).

In parallel, monocytes are recruited to the inflamed endothelium through a CCL2/CCR2 mechanism and adhere to the endothelium through P- and E-selectin/PSGL-1 interactions. They initiate coagulation through TF expression, release of TF-positive microvesicles and interactions with platelets (41, 42). Monocytes/macrophages are also important for thrombus resolution by clearance of apoptotic and necrotic cells and matrix debris, profibrinolytic activity and thrombus neovascularization. Monocytes/macrophages are an heterogenous population of leukocytes. Importantly, it was recently suggested that monocyte conversion from pro-inflammatory Ly6Chi to patrolling Ly6Clo monocytes is important for thrombus resolution (43, 44). Finally, endothelial cell, platelet and leukocyte activation culminate to fibrin formation stabilizing the thrombus (Figure 1) (2, 45–48).

During tissue injury, purine nucleotides (ATP, ADP) are released in the extravascular environment. These molecules are recognized as danger signals by a variety of cells including endothelial cells, platelets, neutrophils and monocytes and activate inflammatory and thrombotic pathways (55, 56). The purine nucleotide metabolism is regulated by two systems that have opposite actions: CD73/CD39 that hydrolyzes ADP into adenosine has an anti-inflammatory and anti-thrombotic role; and P2X7 receptor, that binds ATP, activates the leucine-rich-containing family, pyrin domain containing 3 (NLRP3) inflammasome pathway and promotes pro-inflammatory and pro-thrombotic effects (15).

The inflammasome is extensively involved in non-sterile inflammation-dependent thrombosis. Recently, Wu et al. described a mechanism by which bacterial infection induces blood coagulation through activation of inflammatory responses by the host. They identified that inflammasome activation enhances the release of TF-positive microvesicles from macrophages in a mechanism dependent on pyroptosis. This ultimately leads to systemic activation of the coagulation (57). The importance of the inflammasome in different thrombotic-associated diseases along with infection or not has been reviewed elsewhere (53).

Activation of the inflammasome has also been demonstrated in sterile inflammation. Hence, microparticles from mice deficient for CD39 induce a pro-inflammatory and pro-thrombotic response from endothelial cells illustrated by increased expression of the adhesion molecules, ICAM-1 and VCAM-1, and the release of VWF and TNFα (54). More importantly, CD39 has an important protective role in venous thrombogenesis. Hence, haplo-insufficient mice develop much larger thrombus characterized by increased leukocyte recruitment, NET formation and increased fibrin and TF expression. Mechanistically, CD39 deficiency was associated with accumulation of extracellular ATP, activation of the NLRP3 inflammasome and subsequent IL-1β secretion (15, 55). Hence, when purine nucleotides are not metabolized by CD39, they might accumulate in the circulation and activate the P2X7 receptor expressed by endothelial or circulating cells leading to activation of NLRP3. Thus, CD39 has an important role in inhibiting NLRP3 and IL-1β signaling during VTE.

Additionally, activation of endothelial P2X7 induces p38 phosphorylation leading to IL-8, P-selectin and E-selectin expression in atherosclerotic plaques (56). The lack of P2X7 in mice abolished inflammasome activation and monocyte recruitment to atherosclerotic lesions (49). P2X7 receptors were also found to aggravate cardiac dysfunction in myocardial infarction, hypertension and diabetic retinopathy. Interestingly, P2X7 exerts pro-coagulant functions by inducing platelet activation and TF expression in macrophages, by enhancing the formation of the prothrombinase complex and by inducing the generation of phosphatidyl serine rich microvesicles. In an experimental model of arterial thrombosis, time to occlusion was delayed in P2X7 deficient mice compared to control animals (50). Since P2X7 is crucial for IL-1β production, this study suggests an important link between TF expression and activation of the inflammasome. Interestingly, it has recently been shown that in hypoxic conditions, such as exposure to high altitude, occurrence of thromboembolic events was increased. Thrombosis was the result of inflammation and enhanced expression of NLRP3, IL-1β, IL-18 and caspase-1 through a mechanism dependent of HIF1α (51). IL-1β release plays an important role in the production of additional pro-inflammatory mediators and upregulation of adhesion molecules on endothelial cells through autocrine and paracrine effects that would contribute to thrombus formation (53).

HMGB1 is a non-histone DNA-binding nuclear protein implicated in the regulation of DNA structure to facilitate transcription, replication, and repair (52). HMGB1, which is secreted by stimulated macrophages and activated platelets, is also one of the prototypical DAMPs. HMGB1 binds to RAGE, expressed on the endothelium, smooth muscle cells and monocytes/macrophages (52). Once release in the extracellular compartment, HMGB1 regulates leukocyte recruitment through RAGE-dependent production of cytokines, including TNFα, IL-6, CCL3, CCL4 and CXCL12, and expression of ICAM-1 and VCAM-1 (58). In a mouse model of VTE, HMGB1 accumulates at the endothelial surface as early as 1 h after induction of thrombosis. Increased HGMB1 deposition overtime was correlated with platelet-leukocyte aggregates. Pharmacological inhibition of HMGB1 resulted in significant decreased in thrombus size. Bone marrow or platelet transplantation experiments showed that platelets are the main source of HMBG1 thereby promoting monocyte recruitment and NET formation (59). In addition, Dyer et al. (60) confirmed that platelet specific deletion of HMGB1 reduced NET formation and corresponding thrombus size. Two studies further characterized the mechanism of action of HMGB1 by demonstrating that it is involved in increased cytokine secretion, especially IL-8. This was correlated with higher ICAM-1 and VCAM-1 expression, allowing neutrophil recruitment and adhesion to endothelial cells (52, 61).

Oxidative stress is a major contributor to endothelial dysfunction in cardiovascular diseases. ROS have deleterious effects when produced in excess and participate in the production and secretion of cytokines linking them to inflammation and endothelial dysfunction. Major sources of ROS are NADPH oxidase, the mitochondria and eNOS (endothelial nitric oxide synthase) uncoupling. Hence, mitochondria are important contributors to intracellular ROS production and play a central role in endothelial function (62). As mentioned below, mitochondrial dysfunction and associated liberation of mitochondrial DNA (mtDNA) was shown to promote inflammation (63). In endothelial cells, mitochondrial ROS production is also amplified by activation of NLRP3 and in turns NLRP3 promotes ROS generation. NLRP3 activation increases the secretion of pro-inflammatory molecules, such as IL-1β, IL-18 and HMGB1, which might eventually lead to endothelial cell death through pyroptosis. In addition, eNOS uncoupling is an important characteristic of endothelial dysfunction. Hence, eNOS normally prevents platelet aggregation and adhesion. NO has also anti-inflammatory actions by inhibiting leukocyte interactions with the vessel wall, thereby reducing pathological inflammation and thrombosis. Importantly, in an experimental model of VTE, thrombus burden was reduced by inhibition of TF expression consequent to eNOS activation (64). Ischemic conditions created by the formation of thrombi in veins is also associated with the production of ROS from endothelial cells and recruited leukocytes. In turns, ROS promote the formation of NETs by neutrophils, induce expression of cytokines and adhesion molecules by the endothelium and TF expression by monocytes (9). Thus, ROS are involved in almost all deleterious effects promoting thrombosis and endothelial dysfunction including EndMT. ROS mediate their effect either by directly influencing TGFβ signaling or indirectly by activating NFκB and cytokine production. TGFβ is also inducing oxidative stress in endothelial cells through mitochondrial dysfunction (65).

The role of cytokines and chemokines in VTE has been reviewed elsewhere (29). We will here only discuss IL-1β and TNFα that are important regulators of thrombo-inflammation and endothelial phenotype. IL-1β is involved in many pathologies related to inflammation and results from NLRP3 activation. IL-1β is produced by endothelial cells and platelets and directly affects endothelial functions. For example, IL-1β increases production of ROS, pro-coagulants mediators and impairs vasodilatation. IL-1β expression is initiated following activation of the endothelium by a priming signal such as TNFα leading to NFκB activation; and a second activation signal such as extracellular ATP or DAMPs. The first signal is important for expression of NLRP3, pro-caspase-1 and IL-1β and IL-18 in their inactive forms. The second signal allows the assembly of the inflammasome which leads to the cleavage of pro-caspase-1 into caspase-1. When pro-IL-1β is cleaved by caspase-1 to its active form, it is secreted in the extravascular environment and binds to its receptor. It has been shown in atherosclerosis that IL-1β directly activates endothelial cells and increases the production of thrombogenic mediators such as IL-6 (66).

TNFα is a central cytokine in the regulation of inflammation. TNFα has two receptors: TNF receptor type 1 (TNFR1) or TNF receptor type 2 (TNFR2). TNFR1 is ubiquitously expressed whereas TNFR2 is mostly expressed by immune and endothelial cells. TNFα receptor activation is associated with NFκB translocation to the nucleus and expression of pro-inflammatory genes (29). In presence of TNFα, the expression of P-selectin is upregulated on HUVEC surface allowing platelet rolling suggesting a pro-thrombotic role. Moreover, TNFα downregulates TM expression and increased TF and plasminogen activator inhibitor-1 (PAI-1) expression. All these endothelial changes might result in an acceleration of blood clotting induce by TNFα (67).

Importantly, IL-1β and TNFα are also potent inducers of EndMT. This will be discussed in the paragraph below together with TGFβ.

As for purine nucleotides, intracellular DNA, RNA, rRNA and miRNA become disseminated into the circulation or retained at the site of cell activation and/or injury in the extracellular space. These extracellular nucleic acids elicit profound pro-inflammatory and pro-thrombotic effects. Analysis of extracellular DNA provided insight into their alarmin properties including highly pro-coagulant NETs and extracellular mitochondrial DNA, derived from cell injury and recognized as a DAMP by TLR9 to foster sterile inflammatory responses. Endogenous nuclear and mitochondrial DNA are also recognized by intracellular receptors including the stimulator of interferon genes (STING). STING is ubiquitously expressed in a variety of cells including endothelial cells. Recent studies demonstrated that STING is activated by mtDNA, free fatty acids and TNFα in endothelial cells resulting in control of immune cell transmigration (68–70). STING-induced endothelial dysfunction has been associated with elevated expression of endothelial inflammatory markers including TF and adhesion molecules. Importantly, thrombotic coagulopathy in COVID-19 patients have been attributed to STING over-activation and subsequent endothelial dysfunction in these patients (69). Despite a potential involvement in coagulopathy related to endothelial cell dysfunction, this mechanism has yet not been involved in VTE and would require future research.

As extensively studied by Klaus Preissner's group, different forms of extracellular RNA (eRNA), derived from activated or injured cells, are detectable in the extracellular compartment and contribute to the pathogenesis of cardiovascular diseases [reviewed in (71)]. Although, the exact nature of these eRNA remain to be defined, they contribute to a variety of mechanisms also involved in VTE. eRNA were first found to have pro-coagulant functions in arterial thrombosis by activating the contact phase pathway. It was later describe that they contribute to endothelial activation, leukocyte recruitment, vascular permeability, angiogenesis, macrophage polarization and cell death (72). A recent study showed that acute hypoxia induces TF expression in the vasculature in a TLR3-dependent manner. However, no functional characterization of the direct involvement of this pathway in VTE was investigated in this study (73). Table 1 summarizes endothelial responses to molecules from the microenvironment and which thrombo-inflammatory events are induced following endothelial activation.

TGFβ is the most described inducer of EndMT but IL-1β, TNFα, IFNγ, hypoxia, high glucose and abnormal shear stress are also potent activators of this extreme phenomenon in the spectrum of endothelial activation (88). TGFβ binds with high affinity to type II serine/threonine kinase receptor (TGFβRII) and to activin-like kinases (ALK) 1 and ALK5 that are the predominant type I receptors (TGFβRI) in endothelial cells (76, 89). TGFβ signals through receptor complexes combining 2 type I and 2 type II components. Type I activation results in the activation of the canonical SMAD signaling pathway. Activated SMAD proteins form heterodimeric complexes with SMAD4 and translocate into the nucleus (82). There, they interact with various transcription factors such as Snail, Twist, Zeb and Slug and regulate the transcription of genes involved in EndMT (75). TGFβ might also induce non-canonical SMAD-independent signaling pathways including mitogen-activated protein kinase (MAPK; p38, JNK, ERK), phosphatidylinositol-3-kinase/Akt (PI3K/Akt), mammalian target of rapamycin (mTOR), Hippo/YAP, β-catenin/Wnt, protein kinase C and Rho-like GTPase (76, 78, 84). There are multiple signaling pathways induced by TGFβ alone or in combination with other inflammatory cytokines, including IL-1β and TNFα, hypoxia, abnormal shear stress or high glucose leading to EndMT (88, 90, 91). For example, during hypoxia the degradation of the hypoxia-inducible factor (HIF)1α is inhibited which then accumulates into the nucleus. Interestingly, HIF1α regulates the expression of Snail, Twist and ALK5 which might explain why hypoxic endothelial cells are sensitive to TGFβ signaling (76). Other cytokines can act as stimuli for EndMT, such as IFNγ, which acts via the Janus kinase (JAK) pathway and activation of STAT. Elevated HDL levels show anti-fibrotic effects by blocking the TGFβ/SMAD/Slug/ZEB signaling pathway. On the contrary, high glucose conditions have been shown to induce EndMT through the Akt/PI3K/NFκB pathway (82). Several studies showed that cytokine combination is more powerful in inducing EndMT. It has been particularly demonstrated for TGFβ, IL-1β and TNFα that converge to Sp1-dependent expression of EndMT-related genes (4). Since TGFβ, IL-1β, TNFα are released upon platelet activation along with other pro-inflammatory cytokines, it reinforces the important role of platelets in both thrombosis and inflammation and suggests that platelets participate in EndMT.

EndMT is a tightly regulated process in normal physiology. However, endogenous factors that inhibit EndMT have been less investigated. Several studies have highlighted the important role played by the fibroblast growth factor (FGF) in the regulation of TGFβ signaling (92). Hence, decreased FGF signaling leads to increased expression of TGFβ ligands and receptors and enhances TGFβ signaling resulting in EndMT (93). FGF controls EndMT and TGFβ signaling by maintaining high expression of Let-7 family of miRNAs, which downregulates ALK5 expression. However, in inflammatory conditions where FGF signaling is often reduced, the dramatic decrease of Let-7 miRNAs leads to rapid increase of ALK5 and amplification of TGFβ signaling. In addition, FGF directly inhibits ALK5-dependent expression of EndMT genes through activation of the Ras/MEK1 pathway (94). Accordingly, disruption of FGF signaling has been associated with aggravation of EndMT and atherosclerosis progression (95). BMP7 signaling is a second pathway that negatively regulates EndMT. It can antagonize TGFβ signaling through activation of SMAD1 leading to SMAD2/3 inhibition or by induction of ID proteins which can heterodimerize with SMAD2/3, thus inactivating their transcriptional activities (96).

Different types of non-coding RNA such as circular RNAs (circRNAs), long-non coding RNAs (lncRNAs) and miRNAs have been implicated in the regulation of EndMT. CircRNA are widely described as playing a role in epithelial-to-mesenchymal transition (EMT), a cellular transition process similar to EndMT and were found upregulated by 20-fold after TGFβ1 treatment (97). The circRNA DLGAP4 (DLG-associated protein 4) expressed in mice brain endothelial cells inhibits EndMT by interacting with miR-143 and thereby regulating expression of mesenchymal markers (98). Moreover, the CircHECTD1 was found to regulate the migratory capacity of endothelial cells by decreasing the expression of HECTD1 protein (99). The lncRNAs MALAT1 and GATA6-AS are also regulating EndMT. MALAT1 expression is increased in endothelial progenitor cells upon TGFβ1 treatment leading to downregulation of miR-145 and expression of SMAD3 and TGFβR2, thereby facilitating EndMT. GATA6-AS decreases TGFβ2-induced EndMT in human umbilical vein endothelial cells (HUVECs) by reducing SMA and calponin expression thus preventing VE-cadherin loss. GATA6-AS also interacts with LOXL2, which regulates endothelial gene expression via changes in histone methylation (H3K4me3) (97). However, little is currently known about how circRNAs and lncRNAs are regulated and functionally relevant in EndMT-related cardiovascular diseases.

Similarly, TGFβ signaling is directly controlled by miRNA that either downregulate growth factor receptor-bound 2 (Grb2), involved in fibrosis, (i.e. miR-200a) or the transcription factors Snail1 (i.e. miR-200b and miR-532) and Snail2 (i.e. miRNA-630) (84, 97, 100). TGFβ can also directly modulate expression of miRNAs that suppress inhibitors of mesenchymal gene expression or that affect endothelial gene expression. Normally, mesenchymal gene expression is kept inactive by transcriptional repressors SKI [SKI proto-oncogene (c-Ski)] and the ternary complex factor ELK1. SKI that inhibits the formation of the SMAD complex is the target of miR-155 and ELK1 is repressed by miRNA-27b thus inducing mesenchymal gene transcription. Finally, TGFβ increases the expression of miRNAs that suppress endothelial protein expression, including miR-21. PTEN is a target of miR-21 and an inhibitor of Akt which facilitates EndMT. Several other miRNAs were found to be modulated in EndMT but their precise role has not been defined yet (82, 100).

Interestingly, as for other cardiovascular diseases, miRNAs have been implicated in the pathogenesis of VTE. To date, only 4 studies have quantified miRNA in patients with venous thrombosis in an attempt to correlate miRNA expression to thrombotic events (101–104). Other studies have investigated miRNA expression profile in pulmonary embolism, recurrent venous thrombosis or experimental model of VTE. The most recent study in VTE patients showed that 9 miRNAs were significantly associated with venous thrombosis. Four of these miRNAs were found to regulate proteins of the coagulation cascade and might represent predictors of thrombotic events (104). Although this required further investigation, it has important clinical perspectives since miRNAs can be easily modulated. In addition, some miRNAs associated with VTE in these different studies were also found to be associated with EndMT in separate works (85, 105). For instance, miR-27b that promotes EndMT was found upregulated in pulmonary embolism. In chronic obstruction pulmonary disease, miR-145 was found to negatively regulate TGFβ signaling by decreasing SMAD3 phosphorylation, thus potentially interfering with EndMT (106). Interestingly, miR-145 was also associated with inhibition of TF expression thus preventing venous thrombosis in experimental models (107). These works suggest that miRNAs might link EndMT to VTE. Although, validation of causal implication of these miRNAs in EndMT-related VTE is required, it opens interesting research avenues.

Histone methylation, which is associated with both transcriptional activation and repression, is an important regulator of EndMT. EZH2 (enhancer of homologous zest 2) is the major histone methyltransferase responsible for the deposition of trimethylation marks on lysine 27 of H3 (histone 3; H3K27me3) leading to transcriptional repression. Co-stimulation of endothelial cells with TGFβ2 and IL-1β results in decreased expression of EZH2 and number of H3K27me3 marks at the SMA promoter. This induces SMA expression enabling EndMT. In addition, during cardiac development, histone deacetylase 3 (HDAC3) recruits EZH2 to prevent transcription of TGFβ1, and block physiological EndMT, which is an essential step to complete cardiac development. The association between HDAC3 and EZH2 might therefore be a mechanism involved in EndMT in cardiovascular diseases (97). Another histone demethylase that has been involved in the epigenetic control of EndMT is Jumonji domain-containing protein 2B (JMJD2B). This protein transcriptionally activates gene expression by demethylation of the repressive histone mark H3K9me3 and contributes to the methylation of the activating histone mark H3K4me3. JMJD2B is activated by EndMT-promoting stimuli including TGFβ1, TGFβ2, IL-1β and hypoxia. Using ChIP-sequencing, changes in H3K9me3 marks were found at various promoters upon induction of EndMT. Normally, expression of mesenchymal gene such as calponin is repressed by H3K9me3 marks. However, this repression can be reversed by JMJD2B-mediated demethylation following induction of EndMT (108).

Because aberrant HDAC expression and activity can promote EMT in cancer and that HDAC inhibitors prevent EMT, these proteins have been studied in the context of EndMT. Acetylation of H4 (histone 4) positively regulates SMAD3 upon combined TGFβ and Notch signaling stimulation. However, it remains to be defined which histone acetyltransferase (HAT) may be involved in this mechanism. Interestingly, treatment with TGFβ2 of cardiac endothelial cells induces histone acetyltransferase p300 expression, which is well known to be upregulated in fibrotic tissues. Therefore, p300 might be responsible for H4 acetylation and upregulation of specific SMAD3 target genes upon TGFβ1 stimulation (97). HDACs have also been studied in the regulation of gene expression during EndMT. As mentioned above, HDAC3 appears involved in repression of mesenchymal gene expression. In coronary endothelial cell lines, increased HDAC9 expression correlated with a generalized reduction of histone acetylation during EndMT. Importantly, class IIa HDAC inhibition prevented EndMT, whereas HDAC9 over-expression promoted EndMT. Endothelial-specific Hdac9 deficiency in mice was associated with reduced EndMT and a more stable atherosclerotic plaque phenotype (109). Others HDACs have been involved in EMT but their role in EndMT has not been studied yet.

DNA methylation refers to the presence of methyl 1 groups on cytosine bases in the CpG islands of DNA. These CpG islands are mainly located in gene promoter regions, which are responsible for transcriptional activation. Addition of methyl groups via DNA methyltransferases (DNMTs) to these CpG islands is the most potent epigenetic regulatory mechanism to stably silence gene expression. Treatment with TGFβ1 of human coronary endothelial cells resulted in aberrant methylation of the promoter RASAL1 (RAS protein activator like 1) (a Ras signaling inhibitor), which contributes to EndMT (97).

Regulation of EndMT involving epigenetic mechanisms to push toward expression of mesenchymal genes still requires investigation (Figure 3). A better understanding of the different regulatory networks involved might be useful to determine if maladaptive EndMT can be prevented or reversed to reduce or abrogate vascular remodeling. Aberrant regulation of epigenetic mechanisms is also strongly associated with inflammation and thrombosis. Some studies have investigated how DNA methylation and histone modifications were involved in VTE. Expression of several proteins from the coagulation cascade including FVII, FVIII and tPA was regulated by DNA methylation mechanisms leading to alteration of their plasmatic levels (110). Whole-blood DNA methylation analysis identified a potential association between methylation marks and quantitative biomarkers of thrombotic disorders (111). Regarding histone modifications, one particular histone modification, the citrullination, occurring in the formation of NETs have been extensively implicated in VTE. More importantly, studies have shown that the tPA gene expression was sensitive to histone modifications in endothelial cells and modified by HDAC inhibition (112, 113). In vivo, HDAC inhibition with valproic acid was associated with reduced thrombus burden supporting the idea that it can act as an anti-thrombotic agent (114). Elucidation of the epigenetic pathways involved in EndMT-related VTE will help in the discovery of new therapeutic targets.

Atherosclerosis is characterized by the accumulation of extracellular matrix proteins, mainly collagen and fibronectin, contributing to structural vascular remodeling, thickening of the arterial wall, and plaque formation. EndMT is thought to be one key process that leads to the differentiation of endothelial cells into pro-atherogenic cells leading to plaque formation (88). Recently, single-cell transcriptomic analyses supported evidence that EndMT features were observed in atherosclerotic plaque (121). Since, EndMT has been mostly associated and characterized in area of the vasculature exposed to disturbed flow where atherosclerotic plaque tends to form. Many sensors and pathways involved in transducing mechanical forces in endothelial cells contribute to endothelial dysfunction. Demos et al. recently identified Piezo, a mechanosensor that differentially modulates responses to laminar and disturbed flow. Piezo is a nonselective cation channel that activates PI3K-eNOS pathway which preserved the endothelial phenotype. On the contrary, disturbed flow activates the NFκB pathway contributing to the endothelial pro-atherogenic phenotype (122). The TGFβ-Alk5 is also important for shear-dependent activation of EndMT. Hence, endothelial-specific deletion of both TGFβR1 (Alk5) and TGFβR2 reduced plaque growth and induced plaque regression revealing an important link between EndMT and atherosclerosis (123). A direct connection between mechanical forces, mechanosensing and EndMT in the regulation of endothelial plasticity has recently been investigated. Results from this study showed that inhibition of Alk5 using siRNA in endothelial cells was associated with reduced flow-induced phosphorylation of SMAD2 and downstream expression of mesenchymal genes. Mechanistically, Alk5 association with Shc (Src homology and collagen) modulated EndMT and plaque formation in atheroprone areas (124). Other studies have confirmed that EndMT features were associated with plaque formation induced by disturbed flow, including expression and activation of Twist and Snail (125). It remains that other mechanosensors are probably involved in EndMT in atherosclerosis and would require further work.

PAH is a complex pathophysiological adaptation characterized by vascular remodeling and neointimal thickening of pulmonary vessels with increased migratory, inflammatory and metabolically impaired states of smooth muscle cells following endothelial dysfunction and infiltration of immune cells. Direct link between PAH and EndMT has been evidenced in human and experimental models of pulmonary hypertension (126). Mechanistically, studies have demonstrated that EndMT is induced by several factors in the context of PAH including hypoxia, inflammation, TGFβ signaling and ROS production. As mentioned, hypoxia induces HIF1α that can directly bind to Twist and promotes the expression of mesenchymal genes. However, it was found that silencing HIF2α, which is highly expressed in endothelial cells from PAH patients, reduced Snail and Slug expression in association with increased expression of mesenchymal makers and decreased endothelial markers. Interestingly, endothelial-specific deletion of HIF2α prevented the development of hypoxia-dependent pulmonary hypertension in mice (127). However, it remains to determine if HIF2α deletion in vivo would rescue or prevent the development of EndMT. Pulmonary arterial endothelial cells appear to respond differently to IL-1β, TGFβ1, TGFβ2 than other vascular beds. Inflammation contributes to dysregulation of TGFβ-BMP signaling, which has been associated with predisposition to PAH. BMPR2 signaling in particular is responsible for the activation of SMAD1/5/8 that has antagonistic effects on TGFβ signaling. Loss of function mutations in BMPR2 gene are associated with increased EndMT-related expression of Twist1 and decreased expression of VE-cadherin. Reduced expression and function of BMPR2 is also associated with BMP9-induced EndMT through higher IL-6 expression. Interestingly, neutralizing IL-6 blocked BMP9 and consequent EndMT (96, 128).