There is mounting evidence that TET2 plays an important functional role in the vascular system, with previous studies identifying roles for TET2 in regulating the phenotypic plasticity of vascular smooth muscle cells [19] and regulating autophagy and the production of vasoactive substances by ECs [20, 22]. Studies have demonstrated the importance of TET2 in vascular disease, as downregulation of TET2, and correspondingly, decreased abundance of 5hmC (the first product of TET catalytic activity) are associated with the severity of atherosclerosis and vascular injury in humans and murine models [19–21]. However, the specific targets of TET2-mediated gene regulation in ECs and the role it plays in regulating vascular function remain largely unexplored. Here, using unbiased transcriptomic analyses, we identified that TET2 regulates the expression of interferon-responsive genes in murine and human ECs. We provide evidence that TET2 may be an important regulator of the IFNγ response in ECs, and that IFNγ-induced upregulation of CXCL9 and CXCL10 is likely dependent, in part, upon TET2 catalytic activity. We also provide evidence that decreased TET2 activity in hyperglycaemic conditions may have a suppressive effect on the expression of these cytokines.
In this study, we also assessed the role of endothelial-expressed TET2 in the control of vascular tone using ex vivo tension measurements of aortic rings from WT and EC-specific TET2 KO mice in response to PE and ACh. Although previous studies have identified TET2-dependent regulation of eNOS, CSE and endothelin-1 [20, 22], we observed no difference in the response of aortic rings to these stimuli between WT and TET2 KO mice under baseline conditions. It is possible that TET2 might contribute to the regulation of vascular tone in other vascular beds. However, in our other studies, we have not found evidence of altered blood pressure resulting from endothelial-specific depletion of TET2 (under baseline conditions; unpublished observations), as might be expected to result from the altered expression of vasoactive substances, particularly within endothelial cells of resistance vessels[41]. It is, however, noteworthy that the TET2-mediated regulation of both eNOS and endothelin were observed under conditions of low shear stress, while regulation of the CSE/H2S system was observed in the setting of elevated oxidised LDL [20, 22]. Further, our study identified the functional significance of TET2 in the context of an inflammatory stress (IFNγ treatment). Thus, TET2 is emerging as a dynamic epigenetic modifier that is critical in homeostatic and pathological adaptive cellular responses [42–46]. Future studies should therefore assess the involvement of endothelial TET2 in regulating vascular tone under stressed conditions, such as in atherosclerosis-prone mice.
Our data shows that CXCL9, CXCL10 and CXCL11 are amongst the most abundant cytokines released by HUVEC upon IFNγ treatment and that TET2 may act as an important transcriptional regulator of these cytokines, as their upregulation during IFNγ treatment was blunted by TET2 silencing. CXCL9, CXCL10 and CXCL11 are involved in the chemoattraction of leukocytes, particularly CD8 + and Th1 T cell populations as they abundantly express the CXCR3 receptor for which these cytokines are ligands [47]. In the context of atherosclerosis, high levels of CXCR3 ligands have been observed in atherosclerotic plaques and in particular, a high abundance of CXCL9 and CXCL10 is associated with plaque instability [48–50]. Plaques with high CXCL10 expression have an increased presence of larger lipid cores and altered expression of genes associated with lymphocyte and macrophage recruitment and activation [51]. Furthermore, IFNγ-driven CXCL10 release from smooth muscle cells has been shown to inhibit endothelial recovery following arterial injury, so may contribute to the persistent endothelial dysfunction seen in atherosclerosis [31]. Consistent with a detrimental role for CXCL10 in the context of atherosclerosis, genetic deletion of CXCL10 in ApoE-/- mice or pharmacological inhibition of CXCR3 in low density lipoprotein receptor (LDLR)-/- mice have been shown to reduce the development of atherosclerotic lesions [52, 53]. Our data, suggesting that TET2 is involved in the upregulation of these CXCR3 ligands in response to IFNγ in ECs, might therefore seem surprising, given the growing evidence for athero-protective roles for TET2 [15, 18, 19, 22, 54, 55]. However, these cytokines have also been demonstrated to have roles in wound healing [56], and to both promote or inhibit fibrosis, dependent upon tissue type and experimental model [57]. They are also known to have critical protective roles in the immune system, in the recruitment of activated CD4 + and CD8 + T cells to sites of infection and consequent enhancement of anti-tumour immunity, together with restraint of tumour growth (reviewed in [58]). Further, CXCR3 ligands have antiviral properties [59] and in this regard it may be relevant that TET2 has been shown in dendritic cells to be recruited to CpG-containing islands within interferon response genes in response to viral infection [60]. Further, mice deficient in CXXC5, the epigenetic regulator responsible for recruiting TET2, had compromised IFN responses and became highly vulnerable to viral infection [60]. Whether the CXCR3 ligands were involved in these responses, however, was not determined in this study. Therefore, the precise roles of the endothelial-expressed CXCR3 ligands, in response to TET2-dependent IFNγ-induction, need to be determined in the context of vascular disease prevention and/or development.
In our study, 5hmC deposition was, on average, markedly increased in HUVEC upon IFNγ treatment across genes and the flanking 3Kb up/downstream. This is consistent with IFNγ acting to regulate transcription and indicates that TET enzymes may be involved in DNA demethylation downstream of IFNγ. However, by contrast to a previous study in vascular smooth muscle cells, here, the change in 5hmC level was not associated with a change in TET2 expression. In vascular smooth muscle cells, IFNγ was shown to repress TET2 expression and decrease 5hmC [61]. In pancreatic b-cells, interferon treatment (in this case, IFNα) has previously been associated with hypomethylation and increased inflammatory gene expression by a TET2-dependent mechanism [62]. This discrepancy could be explained by a cell-type dependent effect of IFN on the (hydroxy)methylome or may depend on the type, concentration and duration of IFN treatment. Given that the removal of DNA methylation is a dynamic process involving TET-mediated conversion of 5mC to 5hmC, 5fC and 5caC in a sequential manner, it is not necessarily unexpected that the levels of these marks differ between studies as their further processing makes their relative concentrations highly dynamic.
Consistent with a role for DNA demethylation in the regulation of CXCL9 and CXCL10 expression, we demonstrated that their expression could be upregulated by treatment with 5azaC. Furthermore, by hMeDIP-sequencing we identified a region between the CXCL9 and CXCL10 gene loci, known to regulate these genes, in which 5hmC abundance was increased upon IFNγ treatment. This observation provides further evidence that the region may act as an enhancer, and that altered (hydroxy)methylation of the region may be involved in gene regulation, in addition to the histone modifications previously reported [35–37]. CXCL11 expression was not significantly changed by 5azaC treatment in HUVEC, but nonetheless its expression significantly decreased upon TET2 silencing at the mRNA and protein level. However, the change in expression of CXCL11 was smaller in magnitude than that of CXCL9 or CXCL10. Given that the regulatory region identified lies in closer proximity to the CXCL9 and CXCL10 genes, the expression of CXCL11, which is more distant from the regulatory region may be less affected by changes in the methylation status of the identified locus and account for these differences. Intriguingly, studies of CXCR3 ligand signalling pathways indicate that CXCL11 invokes an alternative signalling cascade to CXCL9 and CXCL10 [63, 64]. Whereas CXCL9 and CXCL10 have predominantly pro-inflammatory effects downstream of ligand binding (promoting Th1/Th17 cell development), CXCL11 instead signals to promote the Tr1 subtype and promotes CXCR3 receptor internalisation, thereby promoting the resolution of inflammation [63, 64]. Differential regulation of these cytokines by DNA methylation could plausibly be of biological importance in coordinating the timing of pro- and anti-inflammatory signals.
Although TET1 and TET3 are also expressed by ECs and could plausibly contribute to the observed 5hmC change, our data from qPCR experiments and data mined from 128 publicly accessible RNA sequencing databases (accessed via BulkECexplorer [65]) indicate that, on average, TET2 transcripts are more abundant than either TET1 or TET3 in HUVEC (Supplementary Fig. 5D-G). However, it cannot be ruled out that TET1 or TET3 may also contribute to the altered hydroxymethylation patterns observed here. Unique to the three TET family members, TET2 does not possess a CXXC DNA-binding domain [66]. TET2 is therefore understood to partner with other proteins to facilitate its interaction with DNA. A variety of binding partners including transcription factors have been identified to recruit TET2 and activate expression of target genes [25, 54, 66–77]. Notably, in THP-1 and B16-OVA cells, an interaction between STAT1 and TET2 was identified, which could be increased by IFNγ [75]. The authors suggest that STAT1 recruits TET2 to target genes including chemokine and PD-L1 genes where they become hydroxymethylated, associated with transcriptional activation [75]. The transcription factor STAT1 is known to mediate interferon signalling via the IFN/JAK/STAT pathway [78]. Considering that STAT1 and TET2 interact, and that multiple TET2-regulated interferon responsive genes identified in our study possess a STAT1 binding element, including CXCL9 and CXCL10 [79–82], it is plausible that STAT1 recruits TET2 to mediate hydroxymethylation and regulate gene expression in ECs. However, this remains to be tested experimentally in this cell type.
Reports from studies in multiple cell types provide evidence that high glucose exposure perturbs the expression of genes involved in interferon signalling pathways [83–88]. Furthermore, there are reports that differential methylation of interferon response genes under high glucose conditions correlates with their altered expression in particular cell types and in blood samples of diabetic individuals [83–86]. Thus, they may be subject to epigenetic regulation. Our data support the notion that TET2 may, at least in part, transcriptionally regulate endothelial IFNγ responses in hyperglycaemia via its demethylase activity, providing mechanistic insight into previous findings. As described above, impaired IFNγ responses in ECs may contribute to the balance of pro- and anti-inflammatory immune cell recruitment at sites of vascular injury. Furthermore, IFNγ and high glucose are reported to synergise to exacerbate endothelial dysfunction, assessed by impaired NO production and calcium dynamics, increased superoxide anion production and cell death [88]. IFNγ has also been shown to perturb metabolic pathways in human coronary artery ECs, including basal glycolysis, upon which ECs are highly dependent [89]. These metabolic alterations correlated with impaired proliferation and migratory capacity following IFNγ treatment. Therefore impaired IFNγ responses in hyperglycaemia may also contribute to metabolic dysregulation, further contributing to endothelial dysfunction. Our data are consistent with a (post-translational) loss of TET2 activity, impacted by high glucose exposure, that contributes to hyperglycaemia-induced endothelial dysfunction via the dysregulated transcriptional regulation of endothelial interferon responses. Therefore, further research into the potential of TET2 as a therapeutic target for limiting endothelial dysfunction in hyperglycaemia-induced vascular disease is warranted.
A noteworthy example of the importance of TET2, specifically expressed in myeloid cells in the aetiology of vascular disease is that individuals with CHIP who possess loss-of-function TET2 mutations in haematopoietic stem cells are at increased risk of coronary heart disease [15]. When atherosclerosis-prone mice were engrafted with bone marrow from TET2-deficient mice, accelerated atherosclerosis was observed [15]. The origin and characteristics of circulating endothelial progenitor cells (EPCs) remain subjects of controversy [90]. However, somatic TET2 mutations observed in expanded haematopoietic stem cell populations have also been identified in a population of (CD105/146+ ve/CD45− ve) EPC-like circulating cells in some patients, indicating the existence of a common precursor cell [17, 91, 92]. Whether (clonally-expanded) EPCs may also harbour TET2 mutations and contribute to vascular disease in other clinical settings remains an unexplored possibility.