Endothelial Epigenetics in Biomechanical Stress: Disturbed Flow–Mediated Epigenomic Plasticity In Vivo and In Vitro

The interface between hemodynamics and the endothelium is profoundly important for the localization of atherosclerosis. Regional transient flow separations, flow reversals, and multidirectional oscillations, collectively referred to as disturbed flow (DF), occur at and near arterial branches, bifurcations and vessel curvatures prone to the development of atherosclerotic lesions.^1,2 Disturbed arterial flow is associated with metabolic stress in the endothelium that sensitizes the cells to site-specific inflammatory and other propathological changes.^3,4 In contrast, nearby regions of pulsatile unidirectional flow (undisturbed flow [UF]) are atheroprotected (Figure 1). Two broad flow-related mechanisms promote differential endothelial phenotypes at susceptible versus protected sites. First, different patterns of biomechanical cell deformation by hemodynamic forces, primarily shear stress,⁵ regulate endothelial mechanotransduction pathways, and second, flow-related differences in the transport characteristics of labile peptides, metabolic intermediates, and free radicals induce potent signaling interactions with the endothelium.^6,7 The mechanisms are not mutually exclusive. This ATVB in Focus series (http://atvb.ahajournals.org/cgi/collection/biomechanical_stress_in_vascular_remodeling) is directed to recent advances in vascular cell and molecular responses to biomechanical stress. Several other reviews in the series^8–10 discuss mechanisms of atherosusceptibility in relation to spatiotemporal hemodynamics. In our contribution, we introduce and discuss emerging epigenomic and epigenetic mechanisms, particularly transcriptional regulation by DNA methylation, in the determination of arterial endothelial phenotype in regions of flow-mediated biomechanical stress and during exposure of endothelial cells to DF and UF in vitro (Figure 1).

Epigenetics encompasses heritable changes in nuclear chromatin leading to gene expression changes that cannot be attributed to changes in the primary DNA sequence.¹¹ Inherited differences in phenotype that occur when the DNA sequence is identical or unchanging include, for example, differential characteristics of monozygous twins, the editing of gene expression essential for precursor cell differentiation through epigenetic gene silencing, progressive changes of chromatin function during normal aging, the normal development and pathological disruptions of neural circuits, and the inactivation of the X chromosome in females.^12,13 Unlike the stable DNA code, the epigenomic code is dynamic¹² with the potential to figure prominently in disease susceptibility and pathogenesis arising from environmental influences independent of, or in concert with, mutations and single nucleotide polymorphisms linked to many complex diseases including cardiovascular dysfunction.¹⁴ Furthermore, a significant fraction of interspecies differences in gene expression is likely attributable to epigenetic variations arising from noncoding regions of the genome than to mutations in the coding genes.^14,15

The efforts of the Encyclopedia of DNA Elements and International Human Epigenome Consortia have uncovered thousands of new putative epigenetic regulatory mechanisms in the noncoding regions of the genome. The epigenome is essentially the interface between the genome and the environment. Hemodynamics is a constantly changing environment to which the arterial circulation is exquisitely sensitive through multiple mechanisms of gene regulation. Given the associative relationships between risk factors and cardiovascular disease, the interplay of mechanotransduction and epigenomic regulation of vascular phenotypes that may lead to novel therapeutic interventions warrants further investigation.

Three major groups of integrated epigenetic mechanisms regulate gene expression: DNA methylation, histone modifications, and RNA-associated gene regulation (Figure 2). Several excellent integrated reviews of cardiovascular epigenetics^14,16 in the literature include an epigenetics primer for vascular biologists.¹⁷

Unsurprisingly, DNA methylation changes and chromatin structural modifications are readily detectable in atherosclerotic tissue. Administration of the HDAC inhibitor trichostatin A to LDLR^−/− mice exacerbated atherosclerosis.³⁸ Unfortunately, interpretation of cell-specific mechanisms is confounded by the comparison between complex lesions that develop in the arterial intima versus undiseased vessel wall composed almost entirely of medial smooth muscle cells. For example, a significant hypomethylation of CpG dinucleotides was measured in the coding region of extracellular superoxide dismutase from rabbit aortic atherosclerotic lesion when compared with normal intima-media.³⁹ However, this likely reflects a comparison between monocyte-derived macrophages in the lesion versus the smooth muscle cell–rich media in normal aorta. More instructive epigenetic data have been obtained when the tissue cell types can be identified or when in vitro manipulations of cell cultures can be used. Hiltunen and Yla-Herttuala⁴⁰ likened atherosclerotic lesions to benign vascular tumors because of the proliferation and monoclonality of intimal smooth muscle cells; they demonstrated the development of hypomethylation during transition of smooth muscle cells from a contractile to a synthetic, proliferative state in vitro. Lund et al⁴¹ reported aberrant methylation patterns in apoE^−/− mice aortas before any sign of lesion development and separately demonstrated that atherogenic lipoproteins induced global DNA hypermethylation in a human monocytic cell line. An important recent study by Dunn et al⁴² induced DF and subsequently atherosclerosis in mouse carotid artery, a vessel normally protected from lesion development. However, infusion of the pan-DNMT inhibitor 5-Aza significantly reduced atherosclerosis suggesting an epigenetic mechanism linking DF and DNMT-mediated DNA methylation to overall lesion progression (expanded discussion below).

Whole or partial genome sequencing of the endothelial DNA methylome in vivo provides a high-level view of site-specific differential methylation regions (DMRs) in endothelium in vivo without discriminating between developmentally stable epigenetics and that which may be under dynamic regulation. We performed comparative genome sequencing of endothelium from well-characterized arterial flow sites of disturbed (susceptible) and undisturbed (protected) blood flow in swine to create a DMR atlas. Using methylated DNA immunoprecipitation sequencing, the genome-wide DNA methylation patterns in endothelial cells from pig aortic arch (atherosusceptible) and descending thoracic aorta (atheroresistant) were examined in detail. A total of 5517 DMRs with an average length of 804±45 bp were identified in somatic chromosomes representing ≈0.2% of the (swine) genome. A total of 4019 regions were hypomethylated, whereas 1498 regions were hypermethylated in aortic arch relative to thoracic aortia endothelium. Methylated regions and DMRs predominantly overlapped with gene- and CGI-rich regions. Enrichment in the 5′-untranslated region suggests a functional role for DMRs in the transcriptional activity of genes. DMR-associated genes are involved in endoplasmic reticulum stress and superoxide radical degradation, pathways that have been identified as common features in DF atherosusceptible regions of swine arteries.^43,44 The association of DMRs with specific cytosines/CpG islands in and near the promoters of annotated genes provided insightful guidance for mechanistic experiments. Methylated DNA immunoprecipitation sequencing experiment annotation and data have been uploaded into ArrayExpress (https://www.ebi.ac.uk/arrayexpress/) with accession number E-MTAB-1930 and into the European Nucleotide Archive (http://www.ebi.ac.uk/ena/) with accession ERP004025.

Flow characteristics have been successfully manipulated in mouse carotid arteries by selective ligation of branch arteries. Dunn et al⁴² introduced DF in a normally UF region to demonstrate upregulation of DNMT1 in endothelium. Using partial genome sequencing they identified hypermethylation of a subset of potentially important endothelial molecules, including transcription factors, by cross-referencing transcriptomic and DNA methylomic datasets. Several DMRs in the promoters of the gene subset contained cAMP Response Elements (CRE) sites that were hypermethylated suggesting a possible cooperative regulatory role for CRE binding protein. Using in situ immunostaining protocols in rats, Zhou et al (2014)⁴⁵ also reported increased endothelial DNMT1 expression and enhanced DNA methylation in partially ligated mouse carotid arteries.

Prescient in vitro experiments by Illi et al⁵⁶ first established the molecular basis for epigenetic histone modification by shear stress in endothelial cells. The application of laminar shear stress to HUVECs activated histone H3 phosphorylation and acetylation and H4 acetylation that were further increased by the HDAC inhibitor trichostatin A. Flow stimulated CREB (CRE binding protein)/coactivator transcriptional complex formation that resulted in enhanced HAT activity; the HAT/HDAC balance presumably determines net acetylation. Selective inhibition studies suggested that the flow-dependent phosphorylation and acetylation of histones were independently regulated by different signaling pathways. Parallel experiments on chromatin remodeling by growth factors in the same study suggested that although shear stress and growth factors may share signaling pathways, histone modifications remain distinct to each stimulus.

Zeng et al⁵⁷ investigated the mechanisms by which flow regulates p53 and p21^Waf1 in endothelial cells. Laminar flow suppression of endothelial cell cycle at the transition from the G1 to S phase is associated with an increase in the expression of p21^Waf1. Undisturbed laminar flow led to an increase in p53 acetylation at Lys-382 but a deacetylation at Lys-320 and Lys-373 contributing to the activation of p21^Waf1. HDAC1 coimmunoprecipitated with p53 and HDAC activity of the immunoprecipitate was increased significantly by laminar flow.

Chen et al⁵⁸ demonstrated that the binding of nuclear factor-κB subunits p50 and p65 to NOS3 promoter by steady UF was dependent on activation of p300/HAT that acetylates p65 and histones close to the NOS3 shear stress response element. P300/HAT activation was essential for increased NOS3 transcription, an important protective response associated with undisturbed laminar flow. Inhibition or knockdown of p300/HAT completely prevented shear-induced NOS3 increase.

The above studies compared static conditions to simple laminar flow rather than measurements of DF versus UF analogous to arterial flow in vivo (which is never static). The first comparative study of DF effects on HDAC activity arose from immunohistochemical staining of endothelial HDAC expression in situ. Having noted by immunostaining that HDAC3 expression of mouse arterial endothelial cells was enhanced at DF near branches, Xu and colleagues⁵⁹ demonstrated that DF, but not undisturbed shear stress, increased HDAC3 expression in HUVEC. Neither flow profile increased deacetylase activity. HDAC3 knockdown, however, promoted apoptosis. Stabilization of HDAC3 in DF was potentiated by a strengthening of a complex formed between HDAC3 and the serine/threonine protein kinase Akt via increased phosphorylation of HDAC3. Overexpression of HDAC3 upregulated Akt phosphorylation and Akt kinase activity, and overexpression of Akt partially rescued the HDAC knockdown-induced apoptosis. Therefore, DF-induced HDAC3 seems to protect against apoptosis, promoting cell survival in a complex biomechanical stress environment.

Lee et al⁶⁰ expanded the involvement of HDACs in endothelial responses to flow, demonstrating that DF characteristics increase expression and activities of multiple HDACs. In direct comparisons of oscillatory and pulsatile undisturbed (unidirectional) shear stress, the expression and nuclear accumulation of HDAC1/2/3 and HDAC5/7 were upregulated in DF, whereas UF promoted nuclear export of HDACs 5 and 7. Immunostaining in rat aortic arch and in a stenosed DF region of abdominal aorta identified local HDAC2/3/5 expression. HDAC1/2/3 is also involved with endothelial atheroprotection through regulation of oxidative status. KLF2 and the antioxidant transcription factor NF-E2–related factor 2 (Nrf2) are upregulated by UF but suppressed by DF; these transcription factors are estimated to regulate a significant fraction of atheroprotective shear stress–induced gene sets in endothelium.⁶¹ Nrf2 binds to the antioxidant response element in the promoter region of antioxidant genes, including NAD(P)H quinone oxidoreductase-1, to induce their expression.⁶² Coimmunoprecipitation demonstrated that DF increased the association of Nrf2 with HDAC-1/2/3 resulting in deacetylation of Nrf2, reduced binding to NAD(P)H quinone oxidoreductase-1 antioxidant response element, lowered expression of NAD(P)H quinone oxidoreductase-1 and consequently greater susceptibility to oxidative stress.⁶¹ A similar approach demonstrated a DF-induced HDAC 3/5/7 association with MEF2, a transcription factor that binds to the KLF2 promoter; deacetylation of MEF2 led to inhibition of KLF2 expression.

Sirtuin 1 (SIRT1), a NAD-dependent HDAC that protects against oxidative stress by deacetylating forkhead transcription factors^63,64 and p53,⁶⁵ thereby inhibiting their ability to induce apoptosis, is also implicated in flow-mediated regulation of endothelial NOS.⁶⁶ Pulsatile unidirectional UF induced SIRT1 deacetylase activity in HUVECs whereas expression was unchanged by DF (oscillating shear stress). Consistent with the in vitro findings, endothelial SIRT1 expression in the undisturbed region of mouse thoracic aorta was elevated in comparison to the aortic arch (DF). Sirt1 elevation was linked to increased mitochondrial biogenesis and its attendant key regulators. SIRT1 was shown to associate with endothelial NOS resulting in deacetylation and increased availability of nitric oxide. Because DF is associated with chronically increased ROS and oxidative stress induces SIRT1, why is SIRT1 not elevated in DF? Chen et al⁶⁶ suggest that in contrast to oscillatory DF, pulsatile UF has a transitory effect on ROS elevation that is more effective in inducing SIRT1. Consistent with this idea is that short exposure to hydrogen peroxide induces SIRT1, whereas longer exposure does not. They suggest that high ROS levels in DF activate poly(ADP-ribose) polymerase-1 resulting in NAD+ depletion that inhibits SIRT1. Although further clarification of these competing mechanisms in a complex cell survival scenario is warranted, it is nevertheless clear that atheroprotected regions of mouse aorta express much higher levels of SIRT1 than the atherosusceptible aortic arch despite much higher ROS in the arch.

The endothelium is a dynamic complex metabolic interface, the biology of which is exquisitely sensitive to local biophysical forces generated by constantly flowing blood. Flow separation–induced disturbances result in adaptive cellular responses that predispose the artery to atherogenesis at specific sites but do not induce pathology unless accompanied by additional risk factors. Instead, susceptibility is related to the chronic adaptation of cellular phenotype that allows continuation of the major functions of endothelium. Whether this is dysfunction or an adaptive resetting of the rate constants for many different cellular pathways (allowing an extended but still normal range of functional activity) is a semantic point. Susceptible phenotype matters because it is the template that determines transition to pathology, whereas elsewhere the artery wall is protected against lesions either by the absence of susceptibility or the presence of additional specific protective mechanisms.

There have been extensive cell and molecular studies of biomechanical responses in endothelial cells. Many have investigated the effect of simple (often steady) flow versus no-flow for which there is no arterial flow equivalent; yet, remarkably, pathways and major regulatory molecules have been identified of great relevance to regional arterial differences in vivo, for example, KLF2.⁶⁷ More recently, flow characteristics in vitro and in vivo are being closely matched to allow direct comparison of DF versus UF. Furthermore, with advances in 4-dimensional flow visualization and computation,⁶⁸ the dynamic environment of endothelium at specific arterial sites is established. Cell phenotype profiles obtained from in vivo sites can then be used to establish candidate targets of interest for in vitro investigations (eg, miRNAs).^28,31 Access to efficient and economical sequencing methodologies has expanded the opportunities to analyze flow-related regulation of endothelial biology in great detail. Although it was realized a decade ago that epigenetic/epigenomic pathways can regulate endothelial responses to shear stress, only relatively recently has the emergence of Encyclopedia of DNA Elements and the Human Epigenome Consortium revealed the vast potential of dynamic epigenomic regulation of gene expression, including that by physical stimuli such as biomechanical stress. The recent demonstrations of endothelial KLF4 and HOXA5 proximal promoter DNA hypermethylation by DF^42,49,69,70 are an emerging epigenomic mechanism of flow-associated transcriptional regulation to add to post-transcriptional regulation by histone remodeling and noncoding RNAs. Further studies will determine whether transcriptional regulation by DNA methylation plasticity to flow is common at the many DMRs identified in the endothelial methylome. Collectively, epigenetics has greatly expanded our understanding of the complex hemodynamic-arterial endothelial biology interface and its role in susceptibility to cardiovascular disease.