Shock drives a STAT3 and JunB-mediated coordinated transcriptional and DNA methylation response in the endothelium Free

Shock-induced endothelial dysfunction is a critical factor leading to organ failure (Ince et al., 2016) and acts by promoting vascular leak, leukocyte adhesion, thrombosis and loss of vasoreactivity (Peng et al., 2019). Although advances in critical care medicine have resulted in an increased survival rate for subjects with critical illness (Griffiths et al., 2013; Hofhuis et al., 2021; Rudd et al., 2020), survivors continue to succumb to sequelae months and even years after the initial acute episode, a condition known as post-intensive care syndrome (PICS). This syndrome involves long-term physical, cognitive and mental deficits and carries increased 5-year mortality rates (Griffiths et al., 2013; Herridge et al., 2011; Hill et al., 2016; Morgan, 2021). The mechanisms driving PICS are not understood. Extensive efforts have been made to define the role of epigenetic regulation, including DNA methylation, in sepsis.

Shock is associated with an acute release of cytokines, a process called cytokine storm (Hill et al., 2016). Among all the cytokines changed during a cytokine storm, the circulating levels of IL-6 are particularly informative, as they correlate with systemic disease severity (Hirano, 2010; Hunter and Jones, 2015) and are highly predictive of mortality (Takahashi et al., 2016). Moreover, IL-6 signaling induces its own expression through a positive feedback loop that is associated with worse outcomes (Hirano, 2010; Hunter and Jones, 2015; Martino et al., 2021). Many cells, including the endothelium, lack sufficient expression of the gp80 (α subunit; also known as IL6R) subunit of the IL-6 receptor (Hirano, 2010). The effects of this cytokine in the endothelium are mediated by IL-6 trans-signaling (Hirano, 2010; Kang and Kishimoto, 2021), which involves a ternary binding between the cytokine, a circulating soluble form of the IL-6-binding receptor subunit (sIL6Rα), and the transmembrane receptor subunit gp130 (also known as IL6ST) to induce JAK kinase activity and STAT3 phosphorylation. We have previously demonstrated a requirement for sIL6Rα in the human umbilical vein endothelial cell (HUVEC) response to IL-6 (Alsaffar et al., 2018; Martino et al., 2022, 2021). This pathway is inhibited by a negative feedback loop involving de novo synthesis of SOCS3 (Hirano, 2010; Hunter and Jones, 2015). Previous work from our laboratory has demonstrated that IL-6 signaling promotes STAT3-dependent sustained barrier function loss in human endothelial cells (EC) primary cultures (Alsaffar et al., 2018), and we recently showed in primary ECs that this signaling axis functions as a crucial autocrine loop induced by endotoxin [lipopolysaccharide (LPS) from Escherichia coli] or TNF (Martino et al., 2022). Furthermore, we demonstrated that endothelial-specific, tamoxifen-inducible SOCS3 knockout (SOCS3iEKO) mice display mortality as soon as 18 h after an endotoxin challenge, which is preceded by increased IL-6 endothelial levels and severe vasculopathy leading to kidney failure (Martino et al., 2021). The organ damage was associated with a kidney proinflammatory transcriptional profile consistent with increased leukocyte adhesivity and thrombogenicity of the endothelial surface, including high levels of COX2, VWF and P-selectin (Martino et al., 2021). The specific mechanisms driving SOCS3-dependent endothelial dysfunction in this context, however, remains unknown.

Circulating cells in septic patients show substantial DNA methylation changes (Binnie et al., 2020; Cao et al., 2020; Hopp et al., 2018), which have been suggested to contribute to immune activation and tolerance (Cross et al., 2019; Falcao-Holanda et al., 2021). It is conceivable that similar epigenetic changes occur within the affected organs, but limitations in the availability of septic biopsies greatly restricts our understanding beyond circulating cells. Despite its clinical implications, the endothelial epigenetic changes and the transcription factor (TF) network that mediate endothelial gene expression changes in response to shock are not fully understood. A better understanding of the epigenetic, TF and transcriptional changes driving endothelial dysfunction might identify novel biomarkers to predict or provide potent therapeutic targets to treat long-term endothelial disfunction.

In this study, we sought to determine the transcriptional and epigenetic changes that occur in endothelial cells in response to proinflammatory stimuli initiated by endotoxin. By combining EC isolation protocols for DNA methylomics with translating ribosome affinity purification for endothelial translatome studies, we provide new insights into endothelial-specific changes in a failing organ to enable a new understanding of the vascular response to inflammatory stimuli. We performed cross-omics analyses of transcriptome and DNA methylome data obtained using DNA collected from isolated kidney endothelium of wild-type (WT) and SOCS3iEKO mice challenged with endotoxin, and from HUVECs treated with IL-6. This analysis allowed us to obtain a comprehensive assessment of gene regulatory relationships. In addition, we performed motif enrichment and discriminant regulon expression analysis to identify the TFs associated with these responses. We found that a single endotoxin challenge leads to substantial DNA methylation changes in the kidney endothelium and SOCS3iEKO promotes a divergent response in DNA methylation in following exposure to LPS. Furthermore, we found that increased STAT and AP1 (JunB, BATF and Fos) transcriptional activity was associated with DNA methylation changes, suggesting a potential mechanism driving transcription-induced gene-specific methylation changes. In vitro, we demonstrated that IL-6 induces similar changes in DNA methylation and that many changes in the endothelial methylome remain in place even without continued IL-6 signaling. Motif enrichment analysis suggested that not only STAT3, but also JunB, BATF and Fos were associated with both epigenetic and transcriptional changes. Depletion of STAT3 and JunB using siRNA in HUVECs demonstrated a role for these two TFs in many of these changes. Together, these findings demonstrate that the endothelium responds to shock with transcriptional and DNA methylation changes, at least in part via the coordination of multiple TFs, including STAT1, STAT3 and the AP1 family members Fos, JunB and BATF. These findings point to an unexpected link between the activity of these TFs and the regulation of DNA methylation. Although the converse is well understood (e.g. how DNA methylation affects gene expression), little is known about how a cell determines which loci to regulate epigenetically. Here, we provide evidence in support of the hypothesis that altered transcriptional activity enables long-term changes via epigenetic modifications.

The epigenetic program in the endothelium of failing organs is not known. To begin addressing this question, we first challenged young adult mice with a single dose of endotoxin (250 μg/mouse of lipopolysaccharides from E. coli 0111:B4; LPS) or saline solution and obtained kidney endothelial DNA for methylation assays 14 h post endotoxin injection. We previously demonstrated that this sub-lethal treatment leads to severe but transient kidney injury in WT mice (Martino et al., 2021). As expected, we observed that LPS induced a systemic response 14 h post challenge, as measured by an increase in the severity score and hypothermia (Fig. 1A,B). We enriched kidney EC from single-cell suspensions using a two-step approach, first using negative selection of non-EC using EpCAM- and CD45-labeled magnetic beads, followed by a positive selection with PECAM (also known as PECAM1)-labeled beads (Fig. 1C). EC enrichment was confirmed by measuring the expression levels of CDH1 (an epithelial marker) and VWF (an endothelial marker) in the RNA obtained from the same kidney suspension before and after magnetic bead enrichment (Fig. 1D). We then performed DNA bisulfite conversion and used bead arrays to investigate the DNA methylation status of 285,000 CpG sites across the entire mouse genome. The endothelium from endotoxemic kidneys displayed 1792 CpG sites with significantly higher methylation levels (hypermethylated) and 804 CpG sites with significantly lower methylation levels (hypomethylated) than control endothelium (Fig. 1E; Table S1). Gene ontology (GO) analysis of the differentially methylated genes showed an enrichment in expected categories, such as lymphocyte activation, cytokine signaling and cell adhesion, as well as in epigenetic regulation, such as DNA methylation itself and heterochromatin assembly (Fig. 1F).

We have previously shown that SOCS3iEKO mice succumb to the same LPS challenge as above, at least in part due to kidney failure (Martino et al., 2021). This response is associated with increased IL-6 signaling. To determine whether loss of SOCS3 leads to altered DNA methylation in response to inflammatory signaling, we measured DNA methylation in the kidney endothelium in control or SOCS3iEKO mice challenged or not with LPS. Proinflammatory signaling in kidney endotoxemic endothelium of SOCS3iEKO mice was confirmed by increased expression of IL-6 and COX2 (Fig. 2A). IL6 was induced 345±112-fold in WT mice and 2931±799-fold in SOCS3iEKO mice (mean±s.e.m.; SOCS3iEKO/WT=8.5-fold, P=0.03), whereas COX2 was induced 3.47±0.80-fold and 9.9±0.97-fold, respectively (SOCS3iEKO/WT=2.9-fold, P=0.007). Notably, lack of endothelial SOCS3 led to a highly divergent response, with 3564 differentially methylated positions (DMPs) in response to LPS in SOCS3iEKO, 2189 CpG sites that were significantly hypomethylated and 1374 sites that were significantly hypermethylated compared to the changes with control mice (Fig. 2B; Table S2). Gene ontology analyses revealed further changes in inflammation, leukocyte differentiation and cytokine signaling, as well as changes in the mitotic cell cycle, MAPK signaling, regulation of cell death and vascular remodeling (Fig. 2C). Notably, together with epigenetic regulation genes, we found multiple hypermethylated genes involved with regulation of MECP2, a 5-methylcytosine-binding protein, and many hypomethylated genes associated with the transcriptional response mediated by MECP2 (Fig. 2C). To gain insight into potential roles for acute transcription and identify the TFs binding to regulatory elements driving these changes in endothelial DNA methylation, we performed motif enrichment analysis of these differentially methylated genes (Fig. 2D). We have previously shown that loss of SOCS3 leads to the expression of a type I interferon-like gene signature in failing organs (Martino et al., 2021). Consistent with this response, hypermethylated CpG sites of LPS-treated endothelial cells in SOCS3iEKO mice were enriched in motifs binding several members of the interferon regulatory family (IRF) family. Notably, we also identified a strong enrichment in motifs binding AP1 family members JunB, ATF3 and BATF. All three genes are highly induced by IL-6 signaling in endothelial cells (Martino et al., 2021). In contrast, hypomethylated CpGs contained motifs for several homeobox factors. Other binding sites, such as that for CTCF, were present in both groups (Fig. 2D). Collectively, these results reveal specific TFs associated with both groups, suggesting a potential mechanism driving gene-specific methylation changes through TF binding.

The changes in the response of SOCS3iEKO mice suggested a role for IL-6-STAT3 signaling in this context. To directly test this, we challenged cultured HUVECs with a combination of recombinant IL-6 and sIL-6Rα (Alsaffar et al., 2018; Garcia-Alonso et al., 2019; Martino et al., 2021; O'Brien et al., 2021) [hereafter denoted IL-6+R, mimicking IL-6 trans-signaling (Kang and Kishimoto, 2021)] and obtained genomic DNA and total RNA 6–72 h post-challenge. Previous work demonstrated strong IL-6+R-induced transcriptional response and barrier function loss within this timeframe (Alsaffar et al., 2018; Martino et al., 2021). We then performed methylomics assays after genomic DNA bisulfite conversion and RNA-seq from the same samples. No significant changes in DNA methylation were detected 6 or 24 h post challenge (data not shown). Sustained gene expression changes 72 h after IL-6+R was demonstrated by RT-qPCR (Fig. 3A). Methylomics analysis 72 h post-IL-6+R identified 431 differentially methylated CpG positions (adjusted P<0.05) (Fig. 3B; Table S3). Of these, 204 CpG sites were hypomethylated and 227 sites were hypermethylated. GO analysis of the subset of differentially methylated sites located at genes revealed enriched biological processes that closely mimicked those identified in endotoxemic kidney endothelium, including innate immunity, cytokine signaling and cell adhesion, and endothelial cell differentiation (Fig. 3C). RNA-seq of the same samples resulted in the IL-6+R-induced upregulation of 316 genes [log fold change (logFC)≥1.5; false discovery rate (FDR)<0.05], and downregulation of 195 genes (logFC≤−1.5, FDR<0.05) (Fig. 3D; Table S4). The over-represented GO categories among the differentially expressed genes were similar to those found in the methylomics analysis, including categories related to cell migration, cell proliferation and response to cytokines (Fig. 3E). As shown in Fig. 3F, a comparison between the gene subsets with differential methylation and altered expression in HUVECs showed significant overlap. We identified 145 differentially expressed genes with 269 DMPs, comprising 38 genes hypermethylated and downregulated and 107 genes hypomethylated and upregulated (Table S5).

RT-qPCR on selected targets was performed on new biological replicates to confirm this finding (Fig. 4A). SERPINA3, NOSTRIN and PLCE1 genes each showed three hypomethylated CpGs in response to an IL-6+R challenge by 72 h, and consistently, their expression levels were upregulated by 72 h of IL-6+R treatment. Conversely, TNFSF4 and NAV2 showed three and two hypermethylated CpGs, respectively, and their expression was downregulated by 72 h of IL-6+R treatment. In all cases, the expression of these genes was not substantially altered by a short IL-6+R treatment, suggesting that the changes observed upon sustained treatment cannot be simply a direct effect of IL-6 transcriptional control (Fig. 4A). Analysis of publicly available chromatin immunoprecipitation (ChIP) data (ChIP-Atlas Peak Browser data; Zou et al., 2022) shows that STAT3 directly binds to the promoter regions on these genes (Fig. S1A). Notably, the expression changes increased over time and did not reach a maximum until 72 h of treatment. This behavior is in stark contrast to the fast response of well-known direct STAT3 targets such as SOCS3, COX2 and IL6 (Alsaffar et al., 2018; Martino et al., 2022, 2021). To test whether DNA methylation changes can directly alter the expression of these genes, we treated HUVECs for 72 h with the methyl transferase inhibitor 5-aza-2′-deoxycytidine (5-AZA). This inhibitor led to a significant increase in the expression of SERPINA3, NOSTRIN, PLCE1 and TNSF4, whereas it did not affect the levels of NAV2 (Fig. 4B). STAT3 is the main TF downstream of IL-6. To test whether STAT3 was required for the sustained IL-6-induced change in differentially methylated genes, we performed STAT3 gene knockdown and measured gene expression in cells treated or not with IL-6+R for 72 h (Fig. 4C). STAT3 knockdown abrogated the IL-6+R-induced increase in SERPINA3, NOSTRIN and PLCE1, and the decrease in expression of TNFSF4 and NAV2, demonstrating a causal, albeit probably indirect, role of STAT3 activation in this transcriptional response.

We then investigated whether the DNA methylation patterns were maintained after IL-6 signaling ended. HUVECs were incubated for 72 h with IL-6+R or saline as above. Then, cells were either lysed immediately or washed and incubated for an additional 96 h in regular growth medium in the absence of cytokine (Fig. 5A). DNA methylation status was assessed by bisulfite conversion and bead arrays as above. Notably, we found that nearly 40% (167 CpGs) of the DMP observed after 72 h of IL-6 remained altered post-washout (with a mean difference of less than 2%) (Fig. 5B; Table S6), suggesting that many changes in the endothelial methylome remain in place for prolonged periods, potentially inducing long-lasting transcriptional changes. STAT3 phosphorylation (Fig. 5C) and IL-6 mRNA expression (Fig. 5D) returned quickly back to control levels after the treatment wash, demonstrating complete removal of the cytokine.

To gain insights into potential roles for cytokine-induced TF binding, beyond STAT3, to regulatory elements driving the changes in the endothelial DNA methylation, we performed analysis of TF motif enrichment of the IL-6+R differentially methylated gene set (Fig. 6A). Notably, the motif enrichment closely mimics the findings obtained previously from endotoxemic kidney endothelium. Motifs enriched in hypomethylated genes included those binding STAT1 and STAT3, as well as ETS and homeobox families; those enriched in hypermethylated genes included those binding AP1 family members (including JunB, Fos and BATF) and IRFs. Our results also show enrichment in AP1 binding sites near hypermethylated positions that remained differentially methylated 96 h after IL-6 wash (Fig. 6B), strongly arguing for a sustained epigenetic response to the activation of these TFs.

We then performed discriminant regulon expression analysis of the HUVEC RNA-Seq data to test whether these TFs were associated with gene expression changes (Fig. 6C). Increased STAT activities further confirmed the sustained IL-6 signaling 72 h post challenge. Other TFs identified closely overlapped with the motifs associated with differentially methylated genes, including the AP1 member JunB and several IRFs.

We had previously identified JunB as a gene highly induced by a 3 h IL-6 treatment in a Jak-dependent manner (GEO series GSE163649; Martino et al., 2021). Consistent with a potential role for JunB in these responses, we found that JunB mRNA (Fig. 7A) and protein (Fig. 7B) levels quickly increase upon an IL-6+R treatment. Given its constitutively nuclear localization in HUVECs, it is likely that protein levels are the main mechanism of regulation for this TF (Fig. 7C). JunB expression was downstream of STAT3, because STAT3 knockdown abrogated JunB increases in gene expression (Fig. 7D). We then performed knockdowns to test a causal role for JunB in IL-6+R responses. As shown in Fig. 7E, siRNAs against JunB led to an ∼90% decrease in protein expression. COX2 and PCDH17 were identified as potential JunB targets, according to the ChIP-X enrichment analysis (CHEA) database (Rouillard et al., 2016; https://maayanlab.cloud/Harmonizome/) (see also Fig. S1B). As shown in Fig. 7F, JunB knockdown in HUVECs abrogated IL-6-induced increase in COX2 and PCDH17 expression. In contrast, this knockdown did not prevent the increase in expression of SOCS3, a direct STAT3 target gene (Fig. 7F).

We then performed translating ribosome affinity purification (TRAP) RNA sequencing (TRAP-Seq) from kidneys derived from LPS-treated control or SOCS3iEKO mice. Data from WT and SOCS3iEKO kidney TRAP-Seq shows that although loss of SOCS3 did not lead to significant changes in the endothelial translatome of otherwise healthy mice, it greatly affects the gene expression signature upon an LPS challenge (Fig. 8A). Similar to our findings using magnetic bead sorting of endothelial cells (Fig. 2A), we found that SOCS3 depletion exacerbated the LPS-induced increase in IL-6 and COX2 expression (Table S7). We observed a strong transcriptional response to LPS (565 genes were significantly upregulated, whereas 2688 genes were downregulated), demonstrating a broad endothelial response to systemic inflammation (Fig. 8B). Notably, many of the genes highly induced by LPS >2-fold in SOCS3iEKO but not in WT mice include those we previously found to be associated with an acute response to IL-6 in HUVECs (Martino et al., 2021), including LIPG, RHOU, DDX58, F3, BATF3 and EPSTI1 (Table S7). Genes induced by LPS in WT mice that also showed at least a further 2-fold increase in SOCS3iEKO versus WT mice include PTGS2 (coding for COX2), CXCL10, CEBPD, MX1 and IL6 itself. These genes were also found previously to be quickly induced by IL-6+R in HUVECs (GSE163649; Martino et al., 2021). Metascape analysis of this data again showed enrichment in processes such as the inflammatory response and cell cycle regulation (Fig. 8C). Notably, motif enrichment analysis demonstrated increased activity of not only STAT1 and STAT3, but also AP1 members BATF, Fos, and JunB and IRF (Fig. 8D).

Similar to our findings in HUVECs, the gene subsets with differential methylation and altered expression in endotoxemic kidneys showed significant overlap. As shown in Fig. 8E, 89 genes were identified with both DNA methylation and gene expression variations (Table S8), comprising 60 hypermethylated positions and downregulated expression and 37 hypomethylated positions and upregulated expression (Fig. 8F). Consistent with a role for STAT3, IRF and AP-1 promoting gene expression, many genes associated with STAT3 binding showed decreased methylation and increased expression (Table S8).

Here, we provide evidence for an IL-6-mediated coordinated response in endothelial cells that comprises both epigenetic and transcriptional changes. We show that acute transcription through the AP1, STAT and IRF families is associated with DNA methylation changes in multiple pro-inflammatory genes, which suggests a coordinated response to acute events with the potential of modulating long-term vascular consequences. This bi-directional response, in which not only DNA methylation affects gene expression, but also acute transcriptional events lead to epigenetic marks, could explain at least in part the chronic sequelae after shock. Although the effects of epigenetic changes in TF binding and activity are well understood, little is known regarding the effects of transcription on epigenetic changes and how a cell determines the loci that will be subject to epigenetic modifications. Determining these mechanisms is crucial for understanding the epigenetic responses to acute stimuli. Here, we provide evidence to support the notion that acute inflammatory challenges lead to epigenetic changes in inflammatory genes via the binding of specific TFs. Furthermore, our in vitro experiments after IL-6 removal suggest that this cytokine might elicit transcriptional changes long after the resolution of inflammation.

Endothelial dysfunction is a common feature of many inflammatory diseases (Peng et al., 2019). Septic shock survivors have increased risk of developing long-term complications, including endothelial dysfunction, which can have significant health consequences (Mikkelsen et al., 2020; Mostel et al., 2019; Shankar-Hari and Rubenfeld, 2016). However, little is known about the regulation within the endothelium of injured organs of specific TF networks and epigenetic responses during and after resolution of inflammation. A better understanding of these mechanisms is key for designing therapies for reducing the incidence of endothelial dysfunction and its complications in septic shock survivors. DNA methylation modifications in septic patients are predominantly examined in mixed populations of circulating cells (Binnie et al., 2020; Falcao-Holanda et al., 2021; Hopp et al., 2018). One study identified 668 methylation sites that were altered between patients with sepsis from those with non-septic critical illness (Binnie et al., 2020). Indeed, across functional enrichment analysis, DNA methylation alterations have been identified in methyltransferase activity, cell adhesion and cell junctions (Binnie et al., 2020). Furthermore, several studies have shown that the expression of enzymes involved in epigenetic modifications and chromatin remodeling varies with severity and cell type (El Gazzar et al., 2008; Hopp et al., 2018; Novakovic et al., 2016). In a recent study, it was found that alterations in DNA methylation in human monocytes following a septic episode correlate with circulating IL-10 and IL-6 levels, suggesting a potential mechanism downstream involving the generation of defective DNA methylation alterations (Lorente-Sorolla et al., 2019). Notably, this study also showed significant changes in the methylation status of IRF and STAT response genes (Lorente-Sorolla et al., 2019). A second study showed that LPS leads to changes in monocyte DNA methylation that are concomitant with the upregulation of inflammatory-related genes, and these involve the JAK2-STAT pathway (Morante-Palacios et al., 2021). In addition, a mouse study identified 200 genes with promoters that were differentially methylated in whole kidneys 24 h after ischemia and reperfusion injury, of which 79 maintained the difference 7 days after reperfusion (Zhao et al., 2017). Another report showed 1721 genes are differentially methylated in rat lung tissue in response to LPS-induced acute lung injury (Zhang et al., 2013). Together, these studies argue for an important role for DNA methylation in response to injury, but the cell-type-specific changes and their mechanisms within the failing organs remain unknown. Our innovative approaches of enriching endothelial cells from failing kidneys for methylome analyses and obtaining the endothelial translatome through a translating ribosome affinity purification strategy allowed us to directly assess the response to an acute inflammatory challenge within the kidney endothelium and provide new insights into the mechanisms driving acute kidney failure.

We identified multiple modifications in the DNA methylation on the kidney endothelium in the endotoxemic mice that were associated with an increased inflammatory response. We focused on the kidney because of its well-characterized injury in mouse models of systemic inflammation (Craciun et al., 2014; Martino et al., 2021) and because we have previously shown a dramatic loss of kidney function in endotoxemic SOCS3iEKO mice (Martino et al., 2021). Consistent with this, loss of endothelial SOCS3 led to a differential epigenetic signature in response to LPS compared to WT mice. A GO analysis showed an enrichment of pathways downstream of inflammatory cytokines and cell adhesion, consistent with an acute severe inflammatory reaction. Also, the GO analysis showed an enrichment of pathways associated with chromatin organization and regulation of epigenetics. Notably, IL-6 treatment on cultured endothelial cells led to differential methylation of a similar proinflammatory and epigenetic regulation gene set. This specific epigenetic response, which was consistent with the transcriptional changes induced by LPS and IL-6, led us to investigate a potential mechanism that coordinates these two responses. Bioinformatics analysis of the SOCS3iEKO kidney epigenetic data suggested an enrichment of binding sites for several families of TFs, most notably AP1 and IRF. A strikingly similar enrichment was observed in IL-6-treated HUVECs, consistent with a key role for IL-6 in the endothelial response to IL-6 (Martino et al., 2022, 2021). Bioinformatic analysis of the endothelial transcriptional response both in vivo and in vitro again suggested an increased activity of AP1 and IRF TF families, together with the expected STAT3 activity (Alsaffar et al., 2018; Martino et al., 2021; Zouein et al., 2019). Here, we confirmed that STAT3 and DNA methylation both regulate the levels of expression of several differentially methylated genes, suggesting a novel role for STAT3 not only in regulating gene expression, but also in enabling epigenetic changes at these loci.

The expression of many AP1 members, including JunB, Fos and BATF, is quickly induced by an IL-6 challenge in HUVECs and is strongly induced in the endotoxemic kidney endothelium. AP1 TFs are involved in activating genes involved in the inflammatory response, including those for cytokines, chemokines and adhesion molecules (Atsaves et al., 2019; Kyriakis, 1999; Lee et al., 2008; Papavassiliou and Musti, 2020; Szremska et al., 2003), and AP1 inhibitors might reduce inflammation in various disease models (Atsaves et al., 2019; Vasanwala et al., 2002). Our study observed enrichment in AP-1-binding sites near to hypermethylated positions at 72 h after an IL-6+R treatment that remained differentially methylated 96 h after washing. To begin assessing a causal role for these TFs, we knocked down JunB in HUVECs. Notably, the increased expression of COX2 was abolished in cells depleted of this TF. Many other genes associated with AP1 binding, however, did not show a change upon the depletion. This suggests a strong specificity for AP1 members in the regulation of different AP1-responsive genes. Alternatively, gene compensation from other AP1 members might have masked any JunB role. The specific role of JunB in regulating the immune response can vary depending on the specific cell type, stage of development and the presence of other regulatory elements (Carr et al., 2017; Perez-Benavente et al., 2022; Wu et al., 2019). The exact mechanism by which JunB regulates COX-2 expression is not well understood (Chen et al., 2005; Looby et al., 2009). Further work is required to fully understand the roles for AP1 members in this context.

In conclusion, we demonstrate that the epigenetic changes in the mouse kidney endothelium in response to endotoxemia are dependent on the expression of the IL-6-negative regulator SOCS3. In vitro, we show that IL-6 signaling directly promotes a sustained alteration in DNA methylation that might mediate many of the changes in endothelial gene expression. Our cross-omics analyses of transcriptome and DNA methylome analysis suggest the potential involvement of specific TFs in the regulation of these DNA methylation changes, providing potential candidate targets to limit the long-term damage induced by shock. This study highlights the importance of epigenetic modifications in endothelial function and provides a foundation for developing novel therapeutic interventions for inflammatory diseases.

The commercial sources for critical reagents and their catalog numbers are listed in Table S9. Table S10 provides a list of sequences for RT-qPCR primers. Table S11 lists all antibodies used.

All animal experiments were approved and conducted in accordance with the Albany Medical College IACUC guidelines for animal care and were performed in the Animal Research Facility at Albany Medical College. Mice were housed in specific pathogen-free rooms with 12-h light–12-h dark cycle and controlled temperature and humidity. Mice were kept in groups of five or fewer in Allentown cages with access to food and water ad libitum. Endothelial-specific, tamoxifen-inducible SOCS3 knockout (SOCS3iEKO) mice on a C57Bl6/J background were described previously (Martino et al., 2021). Briefly, all mice carried two copies of a Cdh5-CreERT2 endothelial driver (Wang et al., 2010) and ROSA26-tdTomato reporter (Madisen et al., 2010) by crossing B6.Tg(Cdh5-cre/ERT2)1Rha with B6.Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze}. A floxed SOCS3 transgene (Yasukawa et al., 2003) was introduced by breeding to B6;129S4-Socs3^tm1Ayos/J to generate the SOCS3iEKO. Control and SOCS3iEKO littermates were obtained by breeding two SOCS3^fl/+ heterozygous mice. Genotyping primers have been described previously (Martino et al., 2021). All mice received tamoxifen [2 mg tamoxifen in 100 μl via intraperitoneal (i.p.) injection] when they were 6–9 weeks old for 5 consecutive days. All the experiments were conducted between 2 and 3 weeks after the end of tamoxifen treatment and the SOCS3 deletion gene was confirmed by post-tamoxifen tail digestion and PCR.

We used mice carrying a TRAP transgene for translating ribosome affinity purification assays and performed RNA isolation from the endothelium of endotoxin-treated mice. These transgenic mice have a ribosomal subunit that can be replaced by a floxed GFP-tagged version (EGFP/Rpl10a), rendering the ribosome amenable to immunoprecipitation via two specific monoclonal antibodies (Zhou et al., 2013). Mice carrying this allele were previously backcrossed to a C57Bl/6 background (a kind gift from Dr Patrick Murphy, Center for Vascular Biology, University of Connecticut, Farmington, USA). This transgene was used to replace the tdTomato marker on cdh5-CreERT2;SOCS3^fl/+ mice. Using this system under the endothelial-specific, tamoxifen-inducible cdh5-CreERT2 driver (Wang et al., 2010) we isolated mRNAs that were actively translated in the endothelium of the kidneys.

Severe, acute inflammation was induced by a single i.p. injection of a bolus of 250 μg (250 μl) LPS. Control mice were given 250 μl of sterile saline via i.p. injection. Disease severity was scored 14 h after the LPS challenge as described previously (Martino et al., 2021). Immediately after scoring, body weight and temperature were measured. Mice were then euthanized with an overdose of pentobarbital. Both males and females were used. Assignment to the saline or LPS groups was performed through randomization of mice within each genotype and sex for each litter. No animals were excluded from the analysis. All handling, measurements, and scoring were performed by a researcher masked to treatment and genotype groups and based on mouse ID (ear tags). Experimental groups were unmasked at the end of each experiment.

After euthanasia, the animals were cannulated on the left ventricle, and the right atrium was nicked to allow perfusion with 5 ml/min PBS for 3 min. Kidneys were collected and minced into small pieces (∼1 mm³). The tissue was transferred to 50 ml conical tubes with 5 ml of PBS containing Ca²⁺ and Mg²⁺ plus 100 mg of type 1 collagenase and 100 mg dispase II for 10 min at 37°C. The slurry of tissue and buffer was triturated by passage through a 14-gauge cannula attached to a 10 ml syringe approximately ten times, followed by a 20-gauge cannula attached to a 10 ml syringe. The cell suspensions were filtered through a 70 µm cell strainer, resulting in single-cell suspensions prior to centrifugation at 400 g for 8 min at 4°C. Pellets were resuspended with 1.5 ml of ice-cold PBS+0.1% BSA. One aliquot (75 μl) per sample was lysed in TRIzol to obtain whole-kidney RNA.

The remainder of the cell suspension was transferred to RNase-free polystyrene tubes containing 60 µl of streptavidin-conjugated dynabeads bound to biotinylated EpCAM and biotinylated CD45 antibodies. The cell suspensions were incubated with these beads for 10 min at room temperature with rotation. Following this, tubes were placed on a Dynal MPC-S magnet for 5 min. The supernatants were transferred to RNase free polystyrene tubes containing 45 µl of streptavidin-conjugated dynabeads bound to biotinylated CD31 antibodies and incubated for 10 min at room temperature with rotation. Following this, tubes were placed on a Dynal MPC-S magnet for 5 min and the supernatant was discarded. The remaining cells bound to the beads were washed six times with PBS containing 0.1% BSA to remove any contaminating cells. After removal of the final wash, the cells were resuspended in 400 µl PBS. From this, 200 µl was transferred to a RNase-free tube and DNA was isolated using the DNeasy Blood & Tissue Kit (Qiagen, USA). The other 200 µl were used for RNA isolation following the RNeasy Plus Mini Kit (Qiagen) protocol. VWF and CDH1 expression were measured by RT-qPCR. The ratios of VWF and CDH1 expression in the enriched versus the total RNA from the same organ were calculated to confirm the enrichment of the endothelial cells following this protocol.

HUVECs were isolated in-house according to established protocols (Gimbrone et al., 1974; Jaffe et al., 1973; Valle et al., 2019) as described previously (Alsaffar et al., 2018; Martino et al., 2022, 2021). The identity and purity of the HUVEC isolations were confirmed for each isolation by checking for more than 99% positive immunostaining with endothelial cell markers (FITC-Ulex europaeus lectin, VE-cadherin) and more than 99.9% negative for α-smooth muscle actin. Cells were assayed between passages 3 and 8. To induce IL-6 signaling, cells were plated at full confluence at a density of 8×10⁴ cells/cm² on plates precoated for 30 min with 0.1% gelatin and incubated at least 48 h prior to the start of experiments. Cells were then treated with a combination of 200 ng/ml recombinant human IL-6 and 100 ng/ml sIL-6Rα (IL-6+R) or PBS (control) for 72 h. After 72 h, a subset of control and IL-6+R-treated cells were washed and incubated for 96 h in EGM-2 growth medium. Treated HUVECs and controls were immediately lysed in TRIzol or centrifuged and then stored until DNA extraction.

HUVECs were seeded in a six-well plate and exposed to 5-aza-2′-deoxycytidine (5-AZA; Sigma-Aldrich) at a concentration of 5 μM for 72 h; new 5-AZA was added each 24 h. The control group was treated in parallel with PBS. RNA was harvested for downstream analysis.

Immunofluorescence studies were performed by seeding 8×10⁴ cells/well on 8-well μ-slide chambers (Ibidi) precoated with 0.1% gelatin. At 48 h after seeding, cells were treated as indicated. Cells were then fixed with 4% paraformaldehyde (Affymetrix) in PBS for 30 min at 4°C, washed twice with PBS, and processed for immunofluorescence at room temperature. Briefly, cells were permeabilized with 0.1% Triton X-100 (Sigma) in PBS (PBS-TX) for 15 min, and blocked with 5% bovine serum in PBS-TX for 1 h. Antibodies were incubated for 2 h at room temperature. Slides were then washed in PBS-TX and stained with Alexa Fluor-conjugated secondary antibodies and 0.5 μg/ml DAPI for 1 h at room temperature. Slides were then washed in PBS. Images were taken with 20× and 63× magnification objectives using a Zeiss AXIO Observer Z1 microscope.

HUVECs grown on multi-well plates or single-cell suspensions of mouse endothelial cells were lysed with TRIzol reagent (Thermo Fisher Scientific). RNA was extracted with chloroform and precipitated with isopropanol as per the manufacturer's instructions. A total of 400 ng of RNA was used to prepare cDNA using Primescript RT Master Mix (Clontech) at 42°C following manufacturer's instructions. The cDNA was diluted 10-fold in nuclease-free water. Then, 2 μl of cDNA was used per PCR reaction. qPCR was performed in a StepOnePlus (Applied Biosystems) instrument using 10 μl SYBR green-based iTaq supermix (Bio-Rad), 7.8 μl of water, and 2 pmol primers (0.2 μl) (Thermo Fisher Scientific). Fold induction was calculated via the ΔΔCt method using GAPDH (human) or HPRT (mice) as the housekeeping gene.

DNA was isolated using the DNeasy Blood & Tissue Kit (Qiagen, USA) according to the manufacturer's instructions, including optional treatment with 100 mg/ml RNase A. The concentration of DNA was measured by means of the Qubit dsDNA BR Assay Kit (Molecular Probes). The DNA samples were stored at −80°C. Bisulfite conversion of 500 ng of genomic DNA was performed using an EZ DNA Methylation-Gold™ Kit (Zymo Research) following manufacturer's instructions. Analysis of DNA methylation was carried out using Illumina Infinium MethylationEPIC BeadChip arrays (Bibikova et al., 2011) for human genomes (>850,000 sites) and Infinium Mouse Methylation BeadChip arrays for mouse genomes (>285,000 sites) by the Epigenomic Services from Diagenode.

The methylation data obtained in this publication have been deposited in NCBIs Gene Expression Omnibus and are accessible through GEO Series accession number GSE223381 for HUVECs and GSE223336 for mice data. Methylation array data were processed with the statistical language R (version 4.2.2) and Bioconductor packages following workflows described in the linked code below. Briefly, processing of the raw HUVEC methylation data was performed with the Chip Analysis Methylation Pipeline (ChAMP) package (version 2.21.1) (Tian et al., 2017). Raw methylation data were imported by the minfi method (Aryee et al., 2014; Fortin et al., 2017). Probes with a detection P>0.01 in one or more samples, probes with a bead count less than three in at least 5% of samples, non-CpG probes, probes that align to multiple locations and sex chromosome-specific probes were removed from subsequent analyses. As a result, the remaining 727,127 probes were utilized for data analysis. Data normalization was performed by the BMIQ method (Teschendorff et al., 2013). Batch effect prediction was made by singular value decomposition (SVD) (Morris et al., 2014). Mouse probe, and intensity data for each sample were imported into R using the package RnBeads (version 2.10.0) (Assenov et al., 2014), followed by preprocessing to filter probes outside the CpG context, SNP probes, sex chromosome probes, probes without intensity value, and probes with a low standard deviation (Aryee et al., 2014; Chen et al., 2013; Triche et al., 2013). The RnBeads.mm10 package was used to annotate the location of each probe. Downstream data analysis was performed using β values. The β value is the ratio of the methylated probe intensity to the overall intensity (the sum of the methylated and unmethylated probe intensities). β values range from 0 to 1, in which 0 is no methylation and 1 is complete methylation and were used to derive heatmaps and further analysis (Du et al., 2010). Differential methylation between groups was defined as Δβ value difference and adjusted P-value (Benjamini–Hochberg method, FDR <0.05).

Gene ontology (GO) analysis of the differentially hypomethylated or hypermethylated CpG sites associated with gene locations was performed using Metascape (version 3.5) (Zhou et al., 2019) and selecting GO terms with a P-value less than 0.05. Motif enrichment of the same CpG sets was done using the findMotifsGenome.pl program of HOMER (version 4.11) (Heinz et al., 2010) and selecting a 500-bp window upstream and downstream of the differentially methylated CpG sites. Annotated CpGs in the EPIC array were used as background.

The bulk HUVEC RNA-seq data obtained in this publication have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE225236. RNA-seq libraries were prepared using Illumina's TruSeq protocol and were sequenced on an Illumina NextSeq 500. Reads were aligned to the hg38 genome using Rsubread v1.5.3 (Liao et al., 2019). Gene counts were quantified by Entrez Gene IDs using featureCounts and Rsubread's built-in annotation (Liao et al., 2014). Gene symbols were provided by NCBI gene annotation. Genes with count-per-million above 0.5 in at least three samples were kept in the analysis. Differential expression analysis was performed using limma-voom (Law et al., 2014).

Gene Set Enrichment Analysis (GSEA) was performed comparing HUVECs treated with IL-6 or PBS for 72 h. As gene sets collection, hallmarks (H) from the Molecular Signatures Database (MSigDB) were selected, adding the specified custom genesets. GSEA analysis and graphs were created with the ClusterProfiler (Yu et al., 2012) and enrichplot (https://bioconductor.org/packages/enrichplot) Bioconductor packages.

The total and TRAP RNA isolation was performed following established protocols (Heiman et al., 2014; Zhou et al., 2013) with small modifications. At 14 h fter LPS treatment, mice were euthanized with an overdose of pentobarbital. Immediately upon loss of paw reflex, mice were perfused intracardially with 10 ml of ice-cold Hanks’ balanced salt solution (HBSS) containing 100 μg/ml CHX. We immediately removed the kidneys and placed them in RNase Away-sprayed plates filled with ice-cold dissection buffer (2.5 mM HEPES-KOH pH 7.4, 35 mM glucose, 4 mM NaHCO₃ in HBSS) and minced with razor blades pre-rinsed with RNase Away solution. Approximately 100 mg of tissue of minced tissue was transferred to RNAse-free tubes containing lysis buffer (20 mM HEPES KOH pH 7.4, 5 mM MgCl₂, 150 mM KCl, 0.5 mM DTT, 100 µg/ml CHX, 40 U/ml RNasin, 1× Roche Mini Complete, EDTA-Free protease inhibitors in mQ H₂O) and homogenized using disposable RNase-free pestles. After centrifugation at 4°C, 2000 g for 10 min, the supernatant was mixed with a 1/9 volume of 10% Igepal and a 1/9 volume of 300 mM 1,2-diheptanoyl-sn-glycero-3-phosphocholine (DHPC), followed by centrifugation at 4°C, 20,000 g for 10 min. A portion of the supernatant was set aside as total RNA by adding 500 μl of TRIzol. The remaining supernatant was transferred to tubes containing GFP-conjugated Protein G beads and incubated. After incubation, we washed the beads, added BME in RLT buffer, incubated and removed the supernatant. The eluted RNA was further purified using the RNeasy Micro Kit. RNA quantity was determined using a Qubit Flex Fluorometer (Thermo Fisher Scientific). Library preparation was performed using an Ion Chef System (Thermo Fisher Scientific) followed by sequencing using an Ion GeneStudio S5 Plus System (Thermo Fisher Scientific) both following manufacturer's suggested protocols for the Ion AmpliSeq Transcriptome Mouse Gene Expression Kit (Thermo Fisher Scientific). Differentially expressed genes were identified using the Transcriptome Analysis Console (TAC) software (version 4.0.2) with the ampliSeqRNA plugin (Thermofisher Scientific). The TRAP/RNA-Seq data has been deposited in NCBI's Gene Expression Omnibus and is accessible through GEO Series accession number GSE229292. Metascape was performed as described above.

siRNAs against STAT3 and JunB oligonucleotides were obtained as a set of individual siRNAs (JunB) or a pool of four sequences (STAT3) from Horizon Discovery (On target plus SiRNA). Cells were transfected with individual siRNA complexed with Lipofectamine siRNA iMAX (Invitrogen) in suspension and seeded at 10⁵ cells/cm². Controls were transfected with On target plus nontargeting control pooled duplexes (Horizon Discovery). Knockdown efficiency was determined by western blotting and qPCR analysis.

HUVECs were lysed in 1× Laemmli buffer containing protease and phosphatase inhibitors [cOmplete protease inhibitor mixture (Roche Applied Science), PhosSTOP phosphatase inhibitor mixture (Roche Applied Science), 0.1 M NaF, 0.1 mM phenylarsine oxide, 10 mM pyrophosphate and 0.1 mM pervanadate (Sigma-Aldrich)]. After boiling, a total of 15 µl of cell lysate per lane was loaded on standard SDS-PAGE gels and transferred to nitrocellulose membranes (Bio-Rad). Membranes were blocked with 5% nonfat dry milk or 5% BSA in PBS containing 0.1% Tween and incubated overnight at 4°C with respective primary antibodies. Secondary HRP-conjugated anti-mouse IgG or anti-rabbit IgG antibodies were incubated for 1 h at room temperature. Membranes were developed using Clarity or Clarity Max (Bio-Rad) chemiluminescent substrates and a Chemidoc MP Imaging System (Bio-Rad). Full uncropped images of blots shown in this paper are presented in Fig. S2.

All in vitro data represent at least three independent experiments, usually run in duplicate or triplicate each. Each ‘n’ represents an independent well. Statistical significance was performed by two-tailed one- or two-sample unpaired Student's t-test or Mann–Whitney's U test, one-way ANOVA with Dunnett's post-hoc, or two-way ANOVA with Sidak's post-hoc tests. All sequencing P-values are adjusted for multiple comparisons (Benjamini–Hochberg method, FDR).

All analysis for the assessment of differential DNA methylation (bisulfite data) and differential expression (RNA-seq data) were processed in R. Code files are available from the GitHub repository https://github.com/ramonbossardi/HUVEC_methylation_geneexpression.

The authors thank the Animal Resource Facility and the Imaging Core Facility at Albany Med for support.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.261323.reviewer-comments.pdf.