HUVEC (#CL002-2XT25, Himedia) were cultured in HiEndoXL^™ Endothelial Cell Expansion medium (#AL517; Himedia) supplemented with 4% endothelial growth supplement, 3% fetal bovine serum (#RM1112, FBS; Himedia), and 1X penicillin streptomycin glutamine (#10378, PS; Gibco). Cells were incubated at 37°C with a constant 5% CO₂ condition. HUVEC from passages 4–9 were used to perform all the experiments. HUVEC were given at least 48 h to synchronize and reach certain confluence prior to applying any treatment conditions. Human primary kidney glomerular endothelial cells (HGMEC) (#H-6014G; Cell Biologics) were cultured in complete human endothelial cell medium (#H1168; Cell Biologics). HGMEC from passages 4–6 were used for all the reported data. Upon receiving both HUVEC and HGMEC from the respective supplier, we tested the purity of these cells through immunofluorescence analysis by immuno-staining the cells with CD31 and/or CD144 antibody followed by imaging to identify positive cells. During IHG treatment, both HUVEC and HGMEC were kept in low-growth-factor– and low-serum–containing media. For treatment, cells were seeded in a six-well plate with count of 1 × 10⁵ cells per well. The IHG, THG, and CHG treatment groups were exposed to a 3-day, 25 mM glucose treatment medium for 12-h pulsatile intervals, early 24 h, and constant 72 h duration, respectively.

All the experimental protocols were conducted in accordance with the guidelines of the Brazilian College for Animal Experimentation and were approved by the Institutional Animal Care and Research Advisory Committee (CEUA #5348280918). Male diabetic and obese mice (Ob/Ob) (BTBR.Cg-Lepob/WiscJ; The Jackson laboratory) on a BTBR background and wild-type (WT) mice were employed in this study. Mice were weaned at 21 days of age and genotyped following the protocol provided by The Jackson Laboratory. Animals were maintained in the Pharmacology Department Animal Care Facility, University of Sao Paulo, Sao Paulo, SP, Brazil. Animal were housed in cages (three to four animals/cage) and maintained at a 12:12 h light–dark cycle and 23°C ± 2°C with free access to standard diet (Nuvilab-Nuvital Nutrients Ltd., Parana, Brazil). Mouse kidney sections from WT and Ob/Ob male mice, 14 weeks old, were formalin-fixed and paraffin-embedded, followed by heat-induced epitope retrieval with citrate buffer. The sections were blocked (#X909, Protein Block; Dako), and incubated overnight at 4°C with the first primary antibody. On the subsequent day, after repeated washes with TBS-T, sections were incubated with secondary antibody for 2 h. Next, the second first antibody was added, and sections were incubated for 2 h at room temperature, followed by incubation with the second secondary antibody for another 2 h. DAPI was used at a concentration of 1:20,000. Images were either taken at 60X in the Microscopy Nikon Eclipse E600 with an Olympus DP72 camera or taken at 40X in the Zeiss LSM 780 Microscope with ZEN 2011 SP7 FP3 (black) software version 14.0.21.201. The antibodies used are specified in Supplementary Table S1. For colocalization analysis of CD31 and α-SMA–stained kidney tissues, we utilized the plugin of ImageJ named Coloc2, which analyzes different parameters of colocalization, including the Pearson correlation coefficient, indicating the degree of overlap between two fluorescent labeled proteins. Colocalization analyses were performed by selecting only the glomerular area of the tissue, and therefore, Pearson correlation values indicate the level of colocalization in the glomerular part of the kidney sections.

All experimental procedures for the rodent studies were approved by the Institutional Animal Ethics Committee of BITS Pilani, Pilani Campus. Male Wistar rats aged 12–16 weeks were selected for the ex vivo experiment. The rats were fed on a normal chow diet. Rats were anesthetized and dissected by an incision from the ventral end. The hearts and aortas were perfused with PBS to remove any blood cells from the vessels. The primary aortas were then collected, and the fatty tissue layers were carefully removed, followed by cutting into cylindrical pieces measuring 2 mm lengthwise to obtain aortic rings. These rings were washed with PBS and further cultured in HiEndoXL^™ endothelial cell expansion medium in 24-well plates for the experiment. Prior to initiating any treatment condition, these aortic rings were allowed a 12-h incubation period in the complete growth medium. The aortic rings were subjected to an intermittent glucose treatment along with a combination of OICR-9429 (10 μM). After 3 days, the tissue fractions were homogenized, suspended in RIPA lysis buffer, and sonicated. Protein estimation was done by Bradford assay, followed by SDS-PAGE and immunoblotting studies.

HUVEC were cultured on gelatin (#TC041; HiMedia) coated coverslips up to 70% confluency. After completion of treatment, cells were washed with PBS and subjected to 10 min fixation with 4% ice-cold paraformaldehyde. Cells were incubated with 0.1% Triton X for permeabilization. Cells were blocked with blocking buffer (1% BSA) for 1 h, followed by co-staining with α-SMA and VE-Cadherin. Cells were washed with PBS to remove unbound primary antibody and were subsequently incubated with Alexa fluor 555/488-conjugated antirabbit/mouse secondary antibody (1:4000) for 2 h, followed by incubation with DAPI (#D9542; Sigma-Aldrich) for nuclear staining. Fluorescence images were captured using a Zeiss ApoTome.2 microscope (Carl Zeiss, Jena, Germany), and intensities were measured using ImageJ software. The antibodies used are specified in Supplementary Table S1.

For predesigned human-specific KMT2D (MLL2) small interfering RNA (Silencer^® MLL2 siRNA #AM16708) and control siRNA (SignalSilence^® Control siRNA #6568; Cell signaling Technology), MLL2 siRNA was used at a concentration of 40 nM in combination of lipofectamine 2000 (#11668; Invitrogen, Thermo Fisher Scientific). The Scramble siRNA and MLL2-specific siRNA were incubated in Opti-MEM^™ reduced serum medium (#31985; Gibco) for 4 h and subjected to the abovementioned treatment conditions. Scrambled siRNA was used as a negative control. In combination with MLL2 siRNA inhibition, OICR-9429 (#SML1209; Sigma Aldrich) was used at a concentration of 10 µM. DMSO was used as a vehicle control. Cells were harvested for protein studies.

Media was removed, and cells were briefly washed with sterile 1X PBS. Cells were incubated in RIPA buffer (#R0278; Sigma Aldrich) containing protease inhibitor (#P8340; Sigma Aldrich) for lysis. After 1 h in ice, cells were scraped using a plastic scraper, followed by repeated cycles of sonication (15 kW, two cycles of 15 s each). Cell debris and whole cell lysate were separated by centrifugation (10,000 g for 10 min). Protein concentration was estimated with Bradford assay (Absorbance at 595 nm). Protein samples were processed for SDS polyacrylamide gel electrophoresis with 10–250 KDa prestained protein ladder (#PG-PMT2922; Genetix) as a molecular weight reference. Proteins were transferred onto a nitrocellulose membrane at 15 V, 2.5 A for 30 min. Membrane was blocked with 5% bovine serum albumin (BSA) for 1 h. Membrane was probed overnight at 4°C by incubation with different monoclonal human/rat specific primary antibodies as specified in Supplementary Table S1.

RNA isolation was performed following the manufacturer’s protocol (#15596, TRIzol^™ Reagent; Life Technologies, Thermo Fisher Scientific). HUVEC were cultured in six-well plates up to 70% confluency and subject to the earlier mentioned treatment regimes. After 72 h, treated cells were incubated in TRIzol^™ Reagent. Upon harvesting cells, an organic layer was separated using chloroform phase separation, and RNA in the aqueous layer was precipitated using isopropanol. Next, precipitated RNA was washed with 75% ethanol, and RNA pellets were air dried. Finally, RNA pellets were dissolved in sterile nuclease-free water, and quantity and quality were analyzed through a Nano-Drop spectrophotometer (SimpliNano; GE Lifesciences).

To measure the transcript level of different genes, we carried out reverse transcriptase-quantitative polymerase chain reaction (RT-qPCR). Total RNA (1 μg) was taken from the cDNA preparation using iScript^™ cDNA Synthesis Kit (#1708891; Bio-Rad Laboratories, Hercules, CA, United States). Prior to cDNA synthesis, isolated RNAs were preincubated with DNAse to remove any DNA contamination. Real-time PCR was performed using iTaq^™ Universal SYBR^® Green Supermix (#1725124; Bio-Rad Laboratories) with a total master mix volume of 10 μl. Analysis was carried out by calculating delta-delta Ct. GAPDH was used as the housekeeping gene.

A ChIP assay was performed using an Imprint^® Chromatin Immunoprecipitation Kit (#CHP1; Sigma-Aldrich). Cultured HUVEC (80% cell density) were subject to intermittent high-glucose treatment. Treated HUVEC were harvested (1 × 10⁶ cells), washed, and cross-linked with 1% formaldehyde in HiEndoXLTM endothelial cell expansion medium (10 min at room temperature). After washing in PBS, the cell pellet was resuspended in nuclei preparation buffer (200 µl per 10⁶ cells) and kept on ice for 10 min. The nuclear pellet thus obtained was resuspended in shearing buffer (100 µl per 10⁶ cells) supplemented with protease inhibitor cocktail (1 µl per ml of shearing buffer) and further sheared by sonication for 30 s (×40). The sheared chromatin (containing 100–500 bp long sheared genomic DNA) was immunoprecipitated with antibodies directed against H3K4me3 at a concentration of 1:50 (#9751; Cell Signaling Technology). The samples were then washed, reverse cross-linked, and treated with proteinase K to obtain purified DNA fragments. qPCR was performed using primers targeted to amplify regions of human Jagged1 and Jagged2 gene promoters.

Primer sequences to analyze the transcript level of different genes and ChIP-qPCR primer determine promoter level enrichment of H3K4me3 are as mentioned in Supplementary Table S2.

All the values are expressed as the mean ± SD. Statistical significance between groups were determined by one-way ANOVA with a false discovery rate (FDR) method called the two-stage linear step-up procedure of Benjamin, Krieger, and Yekutieli for comparisons of multiple groups and a two-tailed Student’s t-test for comparisons between two groups (or a Mann–Whitney U test for nonparametric data). In the multiple comparison analysis, a q-value representing the FDR of <0.05 (between groups) was considered as a criterion to consider the analysis for inclusion as statistically significant. Statistical analyses were performed using GraphPad Prism software. A p-value of less than .05 was considered statistically significant.

In a patient with type 2 diabetes, oscillating blood glucose level is apparent and shown to have more deleterious effects than constant high glucose on endothelial function and oxidative stress, two key factors associated with hyperglycemia-dependent cardiovascular complications (Quagliaro et al., 2003; Ceriello et al., 2008). Furthermore, our recent work indicates that pulses of high-glucose shock (we termed this intermittent high glucose) promotes an inflammatory switch of EC through epigenetically regulating KLF2 and KLF4 genes (Thakar et al., 2021). Because oscillating glucose levels causes potent endothelial dysfunction, we, therefore, undertook experiments to evaluate the effect of different high-glucose treatment conditions, such as intermittent high glucose (12 h of 25 mM glucose followed by 12 h of 5.5 mM glucose in consecutive three cycles totaling 72 h time), transient high glucose (24 h of 25 mM glucose followed by 48 h of 5.5 mM glucose) and constant high glucose (72 h of 25 mM high glucose) on acquistion of mesenchymal character. Through detection of mesenchymal markers, we established that intermittent high glucose maximally induced some of the mesenchymal-like biochemical character in EC as confirmed through the expression of α-SMA and Slug (Figure 1A). Interestingly, for the given time point, we were unable to detect any reduction in the expression of CD144 (Figure 1A) and CD31 (Figure 1B) and their expression remained unaltered. We also measured the level of other EndMT-associated transcripts and/or proteins, including N-Cadherine, Vimentin, Snail, Calponin, Versican, and FSP1. In so doing, we did not detect significant alteration in transcript and/or protein level expression of N-Cadherine, Vimentin, Calponin, Versican, and FSP1 (Supplementary Figures S1A,D–F) while elevating Slug transcript level (Supplementary Figure S1C). Surprisingly, unlike Slug, we observed a significant reduction in the level of Snail protein when HUVEC were challenged with intermittent high glucose (Supplementary Figure S1B). Because, unlike the EndMT phenotype of EC, we did not detect a complete change in the biochemical character of the intermittent high-glucose-challenged EC, we, therefore, named such phenomenon as “EC exhibiting partial mesenchymal character.”

To exclude the possibility of intermittent hyperglycemia-dependent induction of partial mesenchymal character being a simple bystander due to the osmolality effect, we next measured the level of CD31, CD144, α-SMA, and Slug in EC exposed to intermittent D-Mannitol. We fixed D-Mannitol concentration such that it achieves comparable osmolality in a cell system as similar to high-glucose treatment. Through such experiments, we detected no change in protein level expression of CD31, CD144, α-SMA, and Slug in HUVEC exposed to D-Mannitol (Supplementary Figure S2A).

Because we did not detect any alteration in CD144 and CD31 while the expression of α-SMA and Slug were found to be elevated, we wondered whether such phenomenon is due to the acquisition of the mesenchymal character of EC prior to losing their endothelial character. To confirm whether EC is undergoing mesenchymal transition prior to losing endothelial biomarkers, we then performed dual immunofluorescence staining to detect CD144 and α-SMA. In so doing, we detected that intermittent high glucose caused expression of α-SMA in HUVEC, which also expressed CD144 in comparison to normal glucose-incubated cells (Figure 1C). To ascertain these in vitro findings in an in vivo model of hyperglycemia, we further used WT and Ob/Ob mice. These mice exhibited heightened blood glucose levels (Supplementary Figure S3A) and body weight (Supplementary Figure S3B). Interestingly, Ob/Ob mice displayed significant loss of albumin through urine in comparison to their age- and sex-matched WT mice (Supplementary Figures S3C,D). In addition, we also found a significant accumulation of extracellular matrix in the glomeruli of Ob/Ob mice (Supplementary Figure S3E). We further performed CD31 and α-SMA co-staining in kidney tissues of WT and Ob/Ob mice and imaged the glomeruli of kidney tissue sections. Through such imaging, we further confirmed a considerable number of glomerular EC concomitantly expressing both CD31 and α-SMA in Ob/Ob mice (Figure 1D and Supplementary Figure S4A). Furthermore, colocalization analysis revealed significant colocalization of CD31 and α-SMA in endothelial cells in the glomerular region of kidney tissues from Ob/Ob mice (Figure 1D, bar graph and Supplementary Figures S4B,C).

Once we established that intermittent hyperglycemia most effectively caused expression of some the mesenchymal genes, we next set out to explore the possible underlying mechanisms. Alteration in epigenetic marks is prevalent in EC undergoing mesenchymal transition while hyperglycemia independently regulates multiple epigenetic mechanisms. A recent study by Glaser and coworkers describes site-specific reduction of repressive H3K9me3 marks at promoters of mesenchymal genes through promoter-level enrichment of the demethylase JMJD2B (Glaser et al., 2020). However, whether hyperglycemia alters epigenetic pathways to cause a mesenchymal phenotype in EC is yet to be established. To address this, we explored the methylation of histone 3 lysine 4 in cells and tissues upon hyperglycemia. Intermittent hyperglycemia treatment significantly elevated the level of activating the H3K4me3 mark in HUVEC (Figure 2A). To eliminate the possibility of an osmolality effect, we next measured the H3K4me3 level in EC incubated with intermittent D-Mannitol. By doing so, we detected no change in the H3K4me3 level in HUVEC exposed to D-Mannitol (Supplementary Figure S2B).

To identify the cause of such an increase in H3K4me3, we measured the level of the SET1/COMPASS family of proteins that are responsible for catalysis of H3K4me3. We detected a significant increase in MLL2 (Figure 2B) and WDR82 (Figure 2C) in HUVEC exposed to intermittent hyperglycemia, and such treatment did not alter the protein level expression of SET1A (Figure 2D), WDR5 (Figure 2D), SET1B (Figure 2E), and MLL1(NT) (Figure 2F). Surprisingly, despite the increase in cellular H3K4me3 level, MLL1(CT) and Menin protein significantly reduced upon intermittent hyperglycemia challenge (Figure 2D). To validate such findings in vivo and because we know hyperglycemia causes renal microvascular dysfunction, we performed dual staining of H3K4me3 with CD31 in Ob/Ob mice kidney tissues. Through such analysis, we detected an endothelial-specific heightened H3K4me3 level in kidney glomeruli (Figure 2G). Furthermore, in vitro analysis of differential intermittent high-glucose-challenged human glomerular EC (HGMEC) showed a heightened H3K4me3 level (Figure 2H) associated with an increase in MLL2 (Figure 2I) and WDR28 levels (Figure 2H). We also confirmed such an elevation in H3K4me3 (Figure 2J), MLL2 (Figure 2K), and WDR82 (Figure 2L) in the rat aorta exposed to the intermittent high-glucose treatment condition ex vivo.

We further examined the level of H3K4-specific lysine-specific demethylase 1A (LSD1), which preferentially caused demethylation of H3K4me1 and H3K4me2. Interestingly, we observed no alteration in the LSD1 protein level in HUVEC exposed to the intermittent high-glucose treatment condition (Supplementary Figure S5A). We also detected H3K4me1 and H3K4me2 in HUVEC challenged with intermittent high glucose and observed no detectable changes in H3K4me1 (Supplementary Figure S5B) and H3K4me2 (Supplementary Figure S5C).

Because we detected an increase in the expression of a few of the mesenchyme-associated genes, including α-SMA and Slug, we, therefore, decided to evaluate the promoter-level enrichment of H3K4me3 on α-SMA and Slug genes in EC challenged with intermittent hyperglycemia. ChIP-qPCR analysis of α-SMA and Slug gene promoters indicated no alteration of H3K4me3 enrichment on these promoters (Supplementary Figures S6A,B). Activation of Notch signaling is well-described as a driver of mesenchymal transition of EC (Chang et al., 2011). Therefore, we measured the level of active Notch signaling in EC exposed to intermittent hyperglycemia. Intermittent hyperglycemia-challenged EC exhibited Notch activation as detected through an increase in Notch1-intracellular domain (N1-ICD) level (Figure 3A) and heightened expression level of Notch downstream target gene Hes1 (Figure 3B). Next, we questioned whether such activation of Notch signaling upon intermittent hyperglycemia exposure could be associated with control of Notch ligands through promoter-level regulation by H3K4me3. Through ChIP-qPCR analysis, we indeed confirm that H3K4me3 was significantly enriched in the proximal promoter region of both Jagged1 (Figure 3C) and Jagged2 (Figure 3D) gene in EC challenged with intermittent hyperglycemia. Furthermore, concurrent increase in Jagged1 and Jagged2 transcripts (Figures 3E,F) and protein (Figures 3G,H) levels were detected, respectively, in intermittent high-glucose-treated EC. In addition, we measured the protein level expression of membrane-associated metalloproteinases, ADAM9, ADAM10, and ADAM17, which cause ligand-independent activation of Notch signaling. Through such analysis, we did not observe any alteration in ADAM9 (Supplementary Figure S6C), ADAM10 (Supplementary Figure S6D), and ADAM17 (Supplementary Figure S6E) metalloproteinases. We also detected the expression of other Notch ligands, such as Dll1 and Dll4. Although we are unable to detect Dll1 with the specified antibody (Supplementary Figure S6F), interestingly, intermittent high-glucose-treatment caused a significant reduction in another Notch ligand Dll4 (Supplementary Figure S6H), which is previously shown to have a more protective role in maintaining normal physiological functioning of EC (Benedito et al., 2009). Furthermore, we did not detect changes in many other Notch downstream genes, such as Cyclin D3 (Supplementary Figure S6F) and cMyc (Supplementary Figure S6G), indicating induction of Hes1 expression possibly as the primary target of Notch1 activation in current settings.

We then detected the changes in Notch signaling–associated genes in HUVEC that are exposed to intermittent D-Mannitol treatment condition as specified earlier. We could detect either unchanged or statistically insignificant changes in Jagged1 (Supplementary Figure S2C), Jagged2 (Supplementary Figure S2D), N1-ICD, and Hes1 (Supplementary Figure S2C) protein levels. We further confirm the increase in Jagged1 (Figure 3J) and Jagged2 (Figure 3J) protein levels in HGMEC exposed to intermittent high-glucose conditions. To provide evidence that such changes also occur at the tissue level, we detected active Notch1 product N1-ICD and its associated ligands Jagged1 and Jagged2 in rat aorta tissues that were challenged with intermittent hyperglycemia ex vivo. Parallel to in vitro observation, we detected significant increase in N1-ICD (Figure 3K) and Jagged1 (Figure 3L) level in groups in which rat aortic tissues were incubated in intermittent high-glucose treatment conditions. Interestingly, we could not detect any alteration in Jagged2 protein level (Figure 3M) in intermittent high-glucose treated rat aortic tissues.

As we established that H3K4me3-dependent transcriptional activation of Notch ligands Jagged1 and Jagged2 caused their de-repression followed by Notch signaling activation, we wondered whether inhibiting H3K4me3 catalysis through MLL proteins could abrogate intermittent high-glucose-driven Notch ligand expression and activation of Notch signaling. We took advantage of a well-established MLL inhibitor named OICR-9429 (Grebien et al., 2015) and studied its effect on intermittent high glucose-dependent Notch signaling in EC. Administration of OICR-9429 blocked both basal and intermittent high glucose-dependent increase in H3K4me3 level (Figure 4A). Concomitantly, EC treated with OICR-9429 and exposed to intermittent high glucose displayed lower levels of Jagged1 (Figure 4C), Jagged2 (Figure 4B), N1-ICD (Figure 4C), and Hes1 (Figure 4C) proteins compared with only intermittent high-glucose-challenged EC, indicating inhibition of Notch signaling activation in intermittent hyperglycemia-challenged HUVEC. Because, in our previous work, we established that intermittent high-glucose exposure to EC caused significant reduction in KLF2 and KLF4 level (Thakar et al., 2021) and such reduction in KLF2 and KLF4 were previously shown to be associated with EndMT (Moonen et al., 2015), we, therefore, measured the levels of KLF2 and KLF4 in EC exposed to intermittent hyperglycemia and OICR. Such experiment revealed that inhibition of H3K4me3 catalysis in intermittent hyperglycemia-challenged EC normalized KLF2 level (Supplementary Figure S7A) without reversing the level of KLF4 (Supplementary Figure S7B).

We next used a prevalidated human MLL2 specific siRNA and determined whether such targeted knockdown of MLL2 can attenuate intermittent high-glucose dependent de-repression of Jagged1 and Jagged2 to further diminish intermittent high-glucose-driven Notch activation. We detected a significant reduction in MLL2 protein level (Figure 4D) accompanied with reduction in H3K4me3 levels (Figure 4E) upon MLL2 knockdown in HUVEC. Furthermore, MLL2 knockdown diminished intermittent hyperglycemia-mediated increase in Jagged1 (Figure 4F), Jagged2 (Figure 4G), N1-ICD (Figure 4H), and Hes1 (Figure 4F) protein levels in HUVEC. In parallel to the OICR treated groups, MLL2 knockdown normalized the level of KLF2 in intermittent high-glucose-challenged EC (Supplementary Figure S7C).

Once we confirmed the reversal of expression for specified Notch-associated genes upon MLL pharmacological and molecular inhibition in intermittent hyperglycemia-challenged EC in vitro, we then studied the effect of MLL inhibition on H3K4me3 catalysis and Notch activation in an ex vivo model of rat aortic rings. Parallel exposure of intermittent hyperglycemia and OICR-9429 blocked the high-glucose-dependent increase in H3K4me3 levels in tissue lysates of rat aortic rings (Figure 5A). Upon OICR-9429 administration, intermittent high-glucose treatment failed to elevate the level of Notch ligands Jagged1 (Figure 5B) and further could not potentiate Notch receptor activation and downstream signaling as detected through N1-ICD and Hes1, respectively (Figure 5B).

Upon confirming Notch inactivation through MLL inhibition, we next set out to determine whether MLL inactivation could override the intermittent high-glucose effect on acquisition of the partial mesenchymal character of EC. Because we observed significant expression of α-SMA and Slug in intermittent high-glucose-subjected HUVEC, we, therefore, measured these two key mesenchymal markers in EC that are exposed to intermittent high glucose and either preconditioned with OICR-9429 or MLL2-specific siRNA. Catalytic inhibition of MLL through OICR-9429 administration to EC attenuated hyperglycemia dependent induction of mesenchymal genes α-SMA and Slug (Figure 6A) in EC without altering the level of endothelial surface markers CD31 and CD144 (data not shown).

We next analyzed the expression of these mesenchymal genes in EC in which MLL2 was knocked down using MLL2-specific siRNA and were challenged to an intermittent high-glucose condition. In parallel to OICR-9429 treatment, α-SMA and Slug expression levels diminished in intermittent hyperglycemia-exposed EC where MLL2 was knocked down in comparison to only intermittent high-glucose-treated EC (Figure 6B). However, no changes in CD31 and CD144 were detected between any of these treatment groups (data not shown). In addition, no changes in N-cadherin and vimentin were detected in EC that are either exposed to intermittent high glucose or challenged with intermittent high glucose in the presence of OICR or MLL2 siRNA (Supplementary Figures 7D,E)