MEF2 is Essential for Endothelial Homeostasis and the Atheroprotective Gene Expression Program

Abstract

Objective:

Atherosclerosis predominantly forms in regions of oscillatory shear stress while regions of laminar shear stress are protected. This protection is partly through the endothelium in laminar flow regions expressing an anti-inflammatory and anti-thrombotic gene expression program. Several molecular pathways transmitting these distinct flow patterns to the endothelium have been defined. Our objective is to define the role of the MEF2 family of transcription factors in promoting an atheroprotective endothelium.

Approach and Results:

Here we show through endothelial-specific deletion of the three MEF2 factors in the endothelium, Mef2a, -c, and -d, that MEF2 is a critical regulator of vascular homeostasis. MEF2 deficiency results in systemic inflammation, hemorrhage, thrombocytopenia, leukocytosis, and rapid lethality. Transcriptome analysis reveals that MEF2 is required for normal regulation of three pathways implicated in determining the flow responsiveness of the endothelium. Specifically, MEF2 is required for expression of Klf2 and Klf4, two partially redundant factors essential for promoting an anti-inflammatory and anti-thrombotic endothelium. This critical requirement results in phenotypic similarities between endothelial-specific deletions of Mef2a/c/d and Klf2/4. In addition, MEF2 regulates the expression of Notch family genes, Notch1, Dll1, and Jag1, that also promote an atheroprotective endothelium. In contrast to these atheroprotective pathways, MEF2 deficiency upregulates an atherosclerosis promoting pathway through increasing the amount of TAZ.

Conclusions:

Our results implicate MEF2 as a critical upstream regulator of several transcription factors responsible for gene expression programs that affect development of atherosclerosis and promote an anti-inflammatory and anti-thrombotic endothelium.

Graphical Abstract

Introduction

Distinct blood flow patterns within the vasculature are critical determinants of atherosclerosis ^1,2. Atherosclerosis and intimal smooth muscle (SM) preferentially develop in regions that experience disturbed flow (DF) ^3–8. By contrast, regions of the vasculature that experience laminar flow (LF) are protected against atherosclerosis ⁹. Endothelial cells (ECs) experiencing DF are more proliferative, have more actin stress fibers, have a less uniform structure, and express proteins promoting coagulation and inflammation ¹. Although the respective roles of LF and DF in atherosclerosis are well documented, the mechanisms for these distinct roles are less well understood ^2,10–14.

Several signaling pathways and transcription factors are involved in regulation of the LF and DF transcriptional programs ^2,15–17. One pathway for flow-regulated transcription is through the Myocyte Enhancer Factor 2 (MEF2) family of transcription factors. Critical roles for the four members of the MEF2 family (Mef2a - d) have been established in several cell types ^18–20. (Throughout, we will refer to MEF2 when discussing these factors as a group or their activity.) In ECs specifically, endothelial MEF2 inhibits migration of SM into the intima ²¹, regulates gene expression in tip cells during angiogenesis ²², suppresses inflammation and pathological neovascularization ^23,24, and ameliorates pulmonary hypertension ²⁵. MEF2 activation by LF is through two major pathways: In the first, LF activates Extracellular signal-regulated kinase 5 (ERK5), which phosphorylates Mef2a and Mef2c to induce transcription of Klf2 and Klf4 ^26,27. These transcription factors are important regulators of the anti-thrombotic and anti-inflammatory transcriptional program in response to LF ^34,35. In the second, LF causes phosphorylation of the class IIa histone deacetylases (HDACs) that form a complex with MEF2 factors leading to HDAC nuclear export and derepression of MEF2-dependent genes ^36,37. By contrast, DF actively represses MEF2 by three reported mechanisms: posttranscriptional modification of ERK5 to decrease activity ^10,38–40; increased nuclear localization of class IIa HDACs to repress transcription ^17,37,41,42; and inducing CpG methylation to reduce MEF2 binding ⁴³.

Statins have atheroprotective effects in addition to a reduction of LDL levels, in part through induction of Krüppel-like Factor (KLF) Klf2 and Klf4 ^28–33. Statins activate MEF2 through Ras homolog family member A (RhoA) inhibition and direct ERK5 activation ^29,33,44. Thus, when activated, MEF2 induces Klf2 and Klf4 through MEF2 sites in the promoter and distal enhancer, respectively ^29,32. Reflecting the dual nature of MEF2 activity, Tumor Necrosis Factor α represses KLF2 transcription by formation of a MEF2 repressor complex on the KLF2 promoter ¹⁷. Klf2 and Klf4 are critical regulators of vascular development ⁴⁵, cerebral cavernous malformations ^46–49, barrier function, thrombosis, and atherosclerosis ^26,50–54. Klf2 and Klf4 are partially redundant in that endothelial-specific postnatal deletion of individual factors causes relatively small changes in gene expression and function; however, combined, endothelial-specific deletion (referred to hereafter as KLF2/4iEKO) causes dysregulation of thousands of genes and rapid lethality ⁵⁵. Despite considerable in vitro evidence for MEF2 transcriptional regulation of Klf2 and Klf4 (hereafter referred to jointly as Klf2/4), there has been no in vivo evidence in support of this regulation. Indeed, we previously reported that mice with inducible, endothelial-specific deletion of Mef2c (CiEKO) produced vascular features similar to DF with enhancement of the actin cytoskeleton and SM migration, without significant changes on Klf2/4 expression ²¹. The normal expression of Klf2/4 in the absence of MEF2C could indicate that MEF2 proteins are partially redundant in the endothelium. Partial redundancy has been observed in other cell types in which combined deletion of two or three MEF2 factors was required to produce a phenotype ^22,56–58.

LF also promotes an atheroprotective endothelium through transcriptional and post-transcriptional activation of the Notch pathway. LF promotes Notch1 and Notch ligands expression, as well as Notch1 nuclear localization ^59–61. Endothelial-specific deletion of Notch1 increases endothelial proliferation, inflammation, and atherosclerotic plaque formation but does not affect Klf2/4 expression, showing that Notch can regulate these processes independent of Klf2/4 ^59,60. Mechanistically, proliferation is inhibited by Notch through upregulation of Gja4 that is in turn required for induction of the cell cycle inhibitor, Cdkn1b (p27)⁶⁰.

In distinction to these atheroprotective pathways, the effectors of the Hippo pathway, Yes-associated protein (YAP) and Transcriptional coactivator with PDZ-binding motif (TAZ), promote an atheroprone endothelium. In contrast to MEF2, YAP and TAZ are activated by DF and inhibited by LF. Moreover, statins are one of the most potent inhibitors of YAP/TAZ ^62–64. LF inhibits this pathway by increasing YAP phosphorylation and nuclear exclusion through an integrin-Gα13-RhoA pathway ^62,63. Although the pathway by which DF activates is less understood, it reduces YAP phosphorylation, increases nuclear localization and expression of target genes ^62,63. TAZ has increased expression in DF regions of the vasculature and in atherosclerotic plaques, but the mechanism is not well characterized ^62,63. Consequences of YAP/TAZ activation by DF are increased endothelial proliferation and activation of inflammatory pathways ^62,63. Gain and loss of function experiments demonstrate that this YAP/TAZ signaling promotes atherosclerotic lesion development ^62,63.

Here we report that Mef2a, -c, and -d are redundant regulators of the atheroprotective transcriptional program in the endothelium. Combined endothelial-specific deletion of Klf2 and Klf4 has similar, but not identical, features such as hemorrhage, rapid lethality, and results in overlapping changes in expression of atheroprotective genes. These similarities result from the essential requirement of MEF2 for expression of both Klf2 and Klf4 in the endothelium. In addition, MEF2 deficiency results in systemic inflammation, migration of SM into the intima, endothelial cytoskeletal changes, and increased endothelial proliferation, indicating broader actions of MEF2 in promoting an atheroprotective endothelium. These could partially be through induction of Notch pathway genes and reduction of TAZ. Thus, MEF2 has roles in three pathways implicated in either promoting (Klf2/4 and Notch) or antagonizing (YAP/TAZ) the atheroprotective gene expression program.

MATERIALS AND METHODS

The authors declare that all supporting data are available within the article and its online supplementary files.

Mice

For combined Mef2a, Mef2c and Mef2d targeting in the endothelium, mice carrying loxP-flanked (floxed) alleles of Mef2a^{flox 56}, Mef2c^{flox 65} and Mef2d^{flox 66} were bred with Cdh5(PAC)-CreER^{T2 67} to generate Cdh5(PAC)-CreER^T2+; Mef2a^flox/flox; Mef2c^flox/flox; Mef2d^flox/flox, Cdh5(PAC)-CreER^T2+; Mef2a^flox/flox; Mef2c^flox/flox; and Cre negative littermates.

To induce deletion in adult mice with Cdh5(PAC)-CreER^T2+, we performed serial intraperitoneal injections of 2 mg tamoxifen solution (Sigma-Aldrich T5648, 10 mg/ml prepared in sunflower-seed oil containing 10% Ethanol) every 24 hours for 5 days. The tamoxifen injections were initiated when the mice were at least 7 weeks old as shown in the schematic (Figure 1A). For the ease of presentation, we will refer to mice after tamoxifen treatment of Cdh5(PAC)-CreER^T2+; Mef2a^flox/flox; Mef2c^flox/flox; Mef2d^flox/flox as ACDiEKO, and of Cdh5(PAC)-CreER^T2+; Mef2a^flox/flox; Mef2c^flox/flox as ACiEKO. Control animals were either tamoxifen injected Cre-negative littermates in respective floxed allele combinations, or vehicle injected Cre-positive littermates. Recombination specificity was monitored with the Gt(ROSA)26Sor^{tm9(CAG-tdTomato)Hze} (The Jackson Laboratory stock # 007909)⁶⁸ allele. The efficiency of recombination was monitored by reverse transcriptase – quantitative PCR (RT-QPCR). Female and male mice were separately analyzed for the survival curve and no difference was observed in survival, indicating similar timing and pathological severity. Therefore, female and male mice were pooled for the rest of the experiments to reduce the number of mice required. All murine experiments were performed in accordance with approved protocols of the Albany Medical College Institutional Animal Care and Use Committee.

Figure 1. Combined endothelial-specific deletion of Mef2a/c/d causes bradycardia, hemorrhage, thrombocytopenia, leukocytosis, inflammation, and lethality.

A, Diagram showing the timeline for the deletion induction by tamoxifen injections. B, Survival curve for ACDiEKO mice showing a mean survival of 11 days after the initiation of tamoxifen treatments estimated by the Kaplan-Meier method (Controls: n=12 females, n=13 males; ACDiEKO: n=15 females, n=10 males; P < 0.0001 for both sexes compared to their respective controls). C, Representative dissection images and sections show hemorrhage in the lung of ACDiEKO mice harvested 10 days post tamoxifen. D, Heart rates measured with MouseOx Plus pulse oximeter revealed bradycardia (n=3 for Control, n=5 for ACDiEKO). E, Platelet numbers measured on day 9 post tamoxifen injection revealed thrombocytopenia (n=5 for Control, n=6 for ACDiEKO). F, Total white blood cell counts (WBC) were increased in ACDiEKO mice starting on day 6 (n=5 for Control, n=6 for ACDiEKO). G, Temperature measured on days 7, 8, and 9 after the start of tamoxifen showed reduced temperatures. Measurements below the lower limit for the thermometer (34 °C) are plotted at that limit (n=6 for Control, n=8 for ACDiEKO). H, IL6 RT-QPCR on RNA from kidneys, heart, liver and lung harvested 10 days after the start of tamoxifen showed increased IL6 expression in ACDiEKO mice (n=3). I & J, En face IF 10 days after the start of tamoxifen showing in I, higher VCAM1 expression in the endothelium of the ACDiEKO vena cava, and in J, greater CD45+ hematopoietic cells in ACDiEKO vena cava. I & J, VE-cadherin staining and cytoplasmic localization is increased and there is a greater density of nuclei (DAPI) in ACDiEKO. Values are shown as mean ± SEM, individual graphs represent separate groups. Statistical testing by Student’s t-test (** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001).

En face preparation and immunofluorescence labeling

En face immunofluorescence was performed as described ^21,69. Mice were euthanized with intraperitoneal injection of sodium pentobarbital (0.1mg/g of animal). A whole-body cardiac perfusion was performed by perfusing 0.9% sodium chloride saline followed by fixative (2% paraformaldehyde in phosphate buffered saline (PBS), pH 7.4) through the left ventricle of the heart after severing the right atrium. Thoracic aorta and/or vena cava were isolated and further fixed in 2% paraformaldehyde solution at 4 °C for 1.5 hours. Fixed tissues were permeabilized in a permeabilization buffer (PBS containing 0.2% Triton X-100) for 1 hour at room temperature and subsequently incubated with a blocking buffer (5.5% FBS in permeabilization buffer) for 1 hour at room temperature. Primary antibodies were diluted in a staining buffer (2.75% FBS in permeabilization buffer) and incubated with tissue for 16 hours at 4 °C with gentle agitation. phospho-MLC antibody was generated in-house previously ⁷⁰, and all antibody and labeling reagent information is in the supplemental materials and methods. Tissues were subsequently washed three times in the permeabilization buffer in 30-minute intervals. Secondary antibodies (conjugated with Alexa Fluor 647, Alexa Fluor 594, or Alexa Fluor 488 (ThermoFisher Scientific)) diluted as 1:200 working solution (10ug/ml) in the staining buffer were added after the third wash and incubated at room temperature for 3 hours. 4, 6-diamidino-2-phenylindole (DAPI) and Alexa Fluor phalloidin (ThermoFisher Scientific) were used in combination with the secondary antibodies when specified. Tissues were subsequently washed three times in the permeabilization buffer and one time in PBS with 30 minutes for each step. After the PBS wash, vascular tissues were bisected along the direction of flow, and mounted with a Prolong Diamond antifade mounting medium (ThermoFisher Scientific). Images were obtained using Zeiss LSM 880 with AiryScan on either LSM mode or AiryScan mode, with the manufacturer’s software (Zen Black for microscope operation and Zen Blue for image export and analysis).

Image analysis and quantification

The MFI quantification was performed with ImageJ (v.1.52n) by quantifying the Mean Grey Value for each field of view, and normalized it to the cell number in each field of view. For quantification of alignment angle, manual cell segmentation (Imaris, Bitplane) was applied to each processed en face confocal image acquired, based on endothelial-specific VE-Cadherin expression, which marked the endothelial cell-cell junction. The vectors for the cellular long-axis (ellipsoid axis c) were measured and the angle normalized to the long axis of the aorta was quantified as alignment angle for each segmented cell, and the histogram distribution for the cellular alignment angle was graphed. 3D rendering was performed with Imaris software (Bitplane) on acquired confocal Z-stack images.

For MFI quantification of TAZ within the nucleus of endothelium, TAZ immunofluorescence staining signals or endothelial nuclei were first 3D-rendered as a surface using Imaris (9.2.0) based on its fluorescence intensity. A MATLAB plugin “surface-to-surface” localization tool within Imaris was utilized to identify the TAZ colocalized to the nucleus of the endothelium. MFI for the colocalized signal was quantified using the statistics tools in Imaris and exported for statistical analysis with GraphPad Prism.

Endothelial RNA isolation from thoracic aorta and vena cava

RNA from mouse aortic endothelium was isolated using a previously described protocol with minor modifications ^21,71,72. Briefly, mice were euthanized by intraperitoneal injection of sodium pentobarbital and then the vasculature was perfused with saline solution for 2 minutes via the left ventricle after severing the right atrium. The straight portion of the thoracic aorta was isolated and cleared of periadventitial adipose tissue; the vena cava was isolated as well. The thoracic aorta was quickly flushed with 500 ul of TRIzol reagent (ThermoFisher Scientific) using a 25G needle, while the vena cava was flushed with 300 ul of TRIzol reagent. Eluate of respective vessels from a single animal was collected separately into their respective microcentrifuge tube. Chloroform (0.2X volume) was added to the eluate to separate the aqueous phase. GenElute-LPA linear polyacrylamide (Sigma-Aldrich) (1 ml) was added to the aqueous phase as a neutral RNA-carrier and isopropanol (1.25X volume) was subsequently added to the collected aqueous phase to precipitate the total RNA. RNA pellets were washed by 75% ethanol twice and reconstituted with DEPC-treated water. RNA quality was validated by spectroscopy using a NanoDrop 2000 (ThermoFisher Scientific).

Microarray analysis

For transcriptome array profiling, RNA was isolated 7 days after the start of tamoxifen treatment from the mouse thoracic aortic endothelium as described and was processed for the Affymetrix Clariom D Mouse Transcriptome Array (previously known as MTA 1.0). Briefly, 5ng total RNA was converted to fragmented biotinylated cDNA using the standard WT Pico protocol (Affymetrix). The labeled samples were hybridized to the Clarion D Mouse Transcriptome Array, and were subsequently scanned on a GCS3000 7G scanner using standard Affymetrix protocols. Raw CEL files for the ACDiEKO (n=5) or control (n=4) were normalized using Affymetrix Expression Console (build 1.4.1.46) with Signal-Space-Transformation (SST) algorithm and Robust Multichip Average (RMA) with median polish method summarization. Differential gene expression was statistically determined by one-way ANOVA between ACDiEKO and control samples, and the resulting p-values were corrected for multiple comparisons using Benjamini-Hochberg false discovery rate (FDR) with Affymetrix Transcriptome Analysis Console (v3.0). The microarray data has been deposited to Gene Expression Omnibus (GEO) with accession number GSE152884.

Reverse transcription and quantitative PCR

RNA was reverse transcribed using SuperScript III First-Strand Synthesis System or SuperScript IV First-Strand Synthesis System (ThermoFisher Scientific) with random hexamers following the manufacturer’s protocol. RT-QPCR was performed on a Mx3000P qPCR system (Agilent Technologies) with iQ SYBR Green Supermix (Bio-Rad). mRNA levels were normalized to the level of Hprt, and the relative expression fold-change between experimental and control was calculated with the comparative Ct method described previously⁷³. Primers information is available in the supplemental materials and methods.

RNA-sequencing data analysis

RNA-sequencing data from previously published datasets were obtained from GEO with accession numbers (GSE101826, GSE92965)^55,74. FASTQ files were extracted from SRA files with fastq-dump function in sra-tools (NCBI), The TruSeq sequencing adapters and low-quality reads were removed from FASTQ files with Trim Galore! (Babraham Bioinformatics) wrapper script using Cutadapt⁷⁵. The cleaned FASTQ files were quality checked using FastQC (Babraham Bioinformatics), then aligned to the mouse genome (Ensembl GRCm38 genome obtained from GENCODE) using HISAT2 (v.2.1.0)^76,77. Subsequently, transcript assembly was performed using StringTie (v.1.3.4)⁷⁸ with annotated transcriptome as reference. The assembled transcriptomes were quantified using the prepDE.py script provided by the StringTie developer to generate gene matrix files. EdgeR (v.3.26.1)⁷⁹ was used to compute Counts Per Million (CPM) as a normalized measurement for gene expression. Differentially expressed genes were tested using Fisher’s exact test, and multiplicity correction is performed with the Benjamini-Hochberg method on the p-values, to control the false discovery rate (FDR). Differentially regulated genes with FDR < 0.05 were considered significant. Gene Set Enrichment Analysis (GSEA)⁸⁰ was performed on expressed genes with a modified Kolmogorov-Smirnov statistics, and computes false discovery rate from an empirical distribution according to the software manual, with gene-set permutation value set to 1000 as recommended to calculate the enrichment score.^80,81 The Notch gene set is available in MSigDB ⁸⁰. The YAP/TAZ gene-sets used for the analysis were generated previously ^82–84, and available in the supplemental table III.

ChIP-sequencing analysis

The mouse Mef2a or Mef2c biotinylated factor ChIP-seq (bioChIP-seq) data was retrieved from GEO with accession number GSE124008. FASTQ files were processed as described for RNA-Seq and then aligned to the mouse genome (mm10 genome obtained from UCSC genome browser) using Bowtie2 (v2.3.4.3)⁸⁵ with “--local” option. SAM files output from Bowtie2 alignment were subsequently input to HOMER (v4.10)⁸⁶ to perform peak calling with biological replicates using getDifferentialPeaksReplicates.pl. The return peaks that have at least 2-fold enrichment with FDR <0.05 were converted to BED file output with pos2bed.pl script, and subsequently ported to UCSC genome browser ^87,88 for visualization. ENCODE ATAC-seq tracks from heart and lung tissue (source available in Online Supplemental Materials) are used to visualize regions with accessible chromatin.

Statistics analysis

All experiments were performed a minimum of three times (n≥3), and values are reported as means ± SEM. Two-way analysis of variance (ANOVA) followed by Tukey’s test was used to evaluate the statistical significance for comparisons with two independent variables. Two-sample Poisson rate test for frequency distribution comparison was performed with Minitab software. In addition, the two-tailed Student’s T-test was used for comparisons between two groups. Values of p<0.05 were considered statistically significant (* p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001). Although we used parametric tests for statistical analysis, we did not run tests for normality or variance as small group size would make results of such tests invalid.

RESULTS

Combined endothelial-specific deletion of Mef2a, -c, and -d is lethal.

To test the hypothesis that different MEF2 genes have redundant roles in the endothelium, we created a line for the combined, inducible, endothelial-specific deletion of Mef2a and Mef2c using Cdh5-Cre-ER^T2 (ACiEKO)^22,56,65,67. To assay deletion efficiency, we isolated RNA directly from aortic ECs and performed RT-QPCR to the targeted exons as previously described ²¹. This showed deletions of Mef2a/c but did not affect the levels of Mef2d, Klf2 or Klf4 (Figure I A in the Data Supplement). (Mef2b is below detection in our RT-QPCR assay). Because the Mef2a/c endothelial-specific deletion did not affect Klf2/4 expression and the global deletion of Mef2d is viable, without reported vascular dysfunction ⁶⁶, we created a line for deletion of all three Mef2s (Mef2a, -c, and -d) in the endothelium (ACDiEKO). RT-QPCR established that deletion of all three MEF2 is more than 90% effective in both the aorta and vena cava endothelium (Figure I B in the Data Supplement).

In contrast to viability of CiEKO and ACiEKO mice, ACDiEKO mice die rapidly, with a mean survival of 11 days post tamoxifen (Figure 1B). A consistent finding is pulmonary hemorrhage visible during dissection and histology analysis (Figure 1C). There was also hemorrhage in other tissues, most prominently the testis (Figure I C in the Data Supplement). We did not observe an increase in the ratio of wet to dry weight in ACDiEKO lungs compared to control, indicating lack of substantial edema (Figure I D&E in the Data Supplement). We further analyzed these mice for physiological parameters and observed that the heart rate decreases from day 7 onward (Figure 1D). As would be expected with hemorrhage, thrombocytopenia was observed on day 9 (Figure 1E) which corresponded with a significant decrease in erythrocytes (not shown). Consistent with thrombosis, we observed fibrin deposition in the vena cava by en face immunofluorescence (IF) (Figure I F&G in the Data Supplement).

The pathology observed for ACDiEKO shares many attributes to that described for combined, endothelial-specific deletion of Klf2 and Klf4 (KLF2/4iEKO) including pulmonary hemorrhage, bradycardia, and thrombocytopenia ⁵⁵. However, the amount of hemorrhage in other tissues such as the brain and skin appears less than in the KLF2/4iEKO and we did not observe the extensive edema exhibited by KLF2/4 deficiency. Still, there are distinct features of the ACDiEKO. We found a severe elevation of white blood cell counts in ACDiEKO mice (Figure 1F), whereas leukocyte counts were not changed in KLF2/4iEKO mice ⁵⁵. Since the elevation in white blood cell counts in ACDiEKO was an indication of systemic inflammation, we measured the temperature of ACDiEKO mice and found that it decreased from day 8 onward (Figure 1G). A similar reduction in temperature is observed in murine sepsis models but with a more rapid onset ⁸⁹. In further support of an inflammatory response without endothelial MEF2, we found increased interleukin (IL) 6 tissue expression (Figure 1H), increased endothelial VCAM1 protein expression (Figure 1I), and increased attachment of CD45 cells to the endothelium (Figure 1J).

MEF2 activity is required for endothelial expression of Klf2 and Klf4.

We then examined Klf2 and Klf4 expressions in the endothelium of the aorta and vena cava. Deletion of Mef2a/c/d reduced expression of both Klf2 and Klf4 to essentially the same extent as the MEF2 genes themselves were reduced (91–94% reduction of each in both thoracic aorta and vena cava) (Figure 2B and Figure IIA in the Data Supplement). These changes were observed in both the aorta with high LF and vena cava with low LF, indicating that MEF2 regulates Klf2/4 expression in both environments. Moreover, KLF4 protein levels in the ACDiEKO endothelium showed reduction in the thoracic aorta experiencing LF but not in the aortic inner curvature experiencing DF, where expression was already low (Figure 2C). The decrease in Klf4 protein was consistent but not as great as the mRNA decrease. This may be from a delay in protein decay following the mRNA decrease, or from the smaller dynamic range for protein quantification by en face immunofluorescence. Klf2/4 activate transcription of Nos3, Thbd and Dhh, but inhibit transcription of Edn1 and Sele (Figure 2D) ^26,90,91. We, therefore, assayed these genes for changes indicative of loss of Klf2/4 activity. As shown (Figure 2D and Figure IIC in the Data Supplement), there was decreased expression of Nos3, Thbd, Dhh, and increased Edn1 and Sele in the aorta and vena cava, findings that were consistent with a loss of Klf2/4 activity. The overall effect of these changes would be to create a more thrombogenic and activated endothelium, consistent with our observations of increased fibrin deposition, VCAM1 expression and leukocyte adherence in the vena cava (Figure IF in the data supplement and Figure 1I&J). Thus, triple-deletion of Mef2a/c/d reveals the essential requirement for MEF2 activity for transcription of Klf2/4 in the endothelium, explains the phenotypic similarities between ACDiEKO and KLF2/4iEKO, and supports a role for MEF2 as a critical nexus in transducing the LF signal into an atheroprotective gene expression program.

Figure 2. MEF2 activity is essential for endothelial expression of Klf2 and Klf4.

A, Diagram of in vitro derived model of LF regulation of Klf2/4 transcription. B, RT-QPCR on mRNA from the endothelium of the thoracic aorta 7 days after the start of tamoxifen showing decreased levels of Klf2 and Klf4 in ACDiEKO. C, En Face IF of Klf4 10 days after the start of tamoxifen showing reduced Klf4 protein in the endothelium of the thoracic aorta (LF) but no change in the inner curvature (DF). Two-way analysis of variance analysis (ANOVA) followed by Tukey’s test was performed on the measured values (* p ≤ 0.05, *** p ≤ 0.001. For thoracic aorta, n=3 for control and n=4 for ACDiEKO; for inner-curvature, n=3 for both control and ACDiEKO). D, Diagram of known positive and negative regulation by Klf2/4. These genes were assayed by RT-QPCR on ACDiEKO and Control 7 days after the start of tamoxifen and showed downregulation of the positively regulated genes (Dhh, Nos3 and Thbd) and upregulation of the negatively regulated genes (Edn1 and Sele) in ACDiEKO. Values are shown as mean ± SEM. Statistical testing by Student’s t-test (* p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, n=3 for Control, n=5 for ACDiEKO).

MEF2 deficiency alters endothelial morphology.

Enhanced stress fiber formation is a hallmark of endothelial-specific Mef2c deletion and siRNA-directed depletion in endothelial culture ²¹. En face imaging with phalloidin staining of F-actin showed that combined deletion of Mef2a/c/d caused a similar but more pronounced enhancement of the formation of actin stress fibers and increased myosin light chain phosphorylation (Figure 3A). Interestingly, we observed endothelial Acta2 (SM α-actin) staining and apparent integration into the endothelial cytoskeleton (Figure 3B). Substantial endothelial Acta2 mRNA upregulation was also observed by RT-QPCR along with a decrease in the endothelial marker, Pecam1 (Figure 3C). However, the mRNA level of mature smooth muscle marker, Myh11 and Cnn1 is unchanged (Figure III in Data Supplement). This could indicate an early step in endothelial to mesenchymal transition (EndoMT)^92,93. Nevertheless, we did not observe changes in other EndoMT-related genes such as S100a4, Vim, Snai1 or Snai2 (Figure III in Data Supplement). Remarkably, despite the drastic changes in the actin cytoskeleton, the alignment of ECs in relation to blood flow was unchanged (Figure 3D).

Figure 3. MEF2 deficiency alters endothelial morphology.

ACDiEKO and control vessels were collected 10 days post tamoxifen for imaging analysis, and 7 days post tamoxifen for RT-QPCR analysis. En face immunostaining of the thoracic aorta was performed with antibodies to VE-Cadherin (endothelial junctions), DAPI (nuclei) and either (A) phalloidin (F-actin) and pMLC, or (C) Acta2. Representative images of a single z-plane are shown (n=3). A, F-actin and phosphorylation of Myosin Light Chain (pMLC) is enhanced in the ACDiEKO ECs. B, En face staining to Acta 2 shows increased endothelial expression of smooth muscle actin (Acta2) by MEF2 deficiency. C, RT-QPCR quantification of mRNA from the thoracic aorta showing induction of Acta2 and reduction of Pecam1 (n=5 for ACDiEKO, n=3 for controls). D, The alignment of the long axis of ECs to the long axis of the aorta is the same for control and ACDiEKO (n=6 for control and n=5 for ACDiEKO, at least 200 cells were analyzed from at least three independent fields of view acquired from each mice). E, En face fluorescence staining of the vena cava with phalloidin (F-actin), DAPI (nuclei), and the tdTomato Cre reporter showing reduced cortical F-actin and increased nuclear density (n=3). F, RT-QPCR quantification of mRNA from the vena cava showing reduction of Pecam1 but no change in Acta2 (n=3 for controls, n=5 for ACDiEKO). G. En face IF with an isoform-specific antibody (clone 3E2, Sigma-Aldrich, F6140) to the alternatively spliced form of fibronectin containing EIIIA&B (FN-EDA) shows increased staining in both ACDiEKO aorta and vena cava. (Representative images, n=3) H. Quantification of the Mean Fluorescent Intensity (MFI) of the images in G showing upregulation of FN-EDA (n=3). Values are shown as mean ± SEM, individual graphs represent separate groups. Two-way ANOVA followed by Tukey’s test was performed on the measured values (* p ≤ 0.05,** p ≤ 0.01, *** p ≤ 0.001).

In contrast to aortic endothelium, vena cava endothelial cells lacked stress fibers but displayed altered cortical actin leading to an extremely disorganized endothelial layer. Compared to the normal cobblestone pattern, ACDiEKO cells formed aggregates with indistinct boundaries (Figure 3E) and the junctional marker, VE-Cadherin, displayed diffuse cytoplasmic staining rather than strong localization at cell junctions (Figure 1I&J, and Figure 3G). There was also a reduction in Pecam1 mRNA similar to the aorta, but no upregulation of Acta2 (Figure 3F). These distinct phenotypes in arterial and venous endothelium are intriguing and suggest functional differences in MEF2s roles in these vascular beds.

The alternatively spliced exons of fibronectin, EIIIA and EIIIB, are usually not expressed in the adult endothelium. However, induction of low, bidirectional flow by partial ligation of the carotid arteries induces alternative splicing of endothelial fibronectin for inclusion of EIIIA and EIIIB that is protective to hemorrhage and vessel remodeling ⁹⁴. We observed a similar induction of this isoform in ACDiEKO endothelium in both protein (Figure 3G&H) and mRNA (Figure IV in the Data Supplement), suggesting that MEF2 deficiency produces an endothelium similar to that from loss of LF.

Endothelial MEF2 activity inhibits intimal SM formation.

Endothelial-specific deletion of Mef2c causes migration of medial SM into the intima ²¹. This process in CiEKO mice is progressive with an average of 4.6 intimal SM clusters per aorta at 28 days and 28.0 at 84 days post deletion²¹. We examined intimal SM cluster formation in ACiEKO and ACDiEKO aortas and found that the formation in ACiEKO was similar to what we observed in CiEKO; however, formation in ACDiEKO was greatly accelerated with a mean of 28.8 clusters arising 10 days post tamoxifen, virtually the same as at 84 days for CiEKO ²¹ and greater than for ACiEKO (Figure VB in the Data Supplement). Moreover, the frequency distribution showed a greater occurrence of larger clusters in ACDiEKO at 10 days compared with ACiEKO at 84 days (Figure VC in the Data Supplement). These data further support a critical, partially redundant role for MEF2 factors in promoting an anti-atherosclerotic endothelium.

Endothelial proliferation is increased by MEF2 deficiency.

The number of ECs in the aorta and vena cava appeared to be increased in these ACDiEKO en face experiments (Figure 1&2) and anaphase chromosomes were frequently seen in EC of ACDiEKO (Figure 3A). These observations motivated us to quantify the EC density and to measure proliferation by immunofluorescence of phosphorylated histone H3. We found increased density of ECs and increased percent mitotic cells in both the aorta (Figure 4A–C) and vena cava (Figure 4E–G). RT-QPCR showed that there was a corresponding increase in Pcna expression (Figure 4D&H). These data indicate the MEF2 deficient endothelium is highly proliferative.

Figure 4. MEF2 deficiency increased endothelial cell proliferation.

Thoracic aortas (A) and vena cavas (E) from ACDiEKO and controls were collected 10 days post tamoxifen treatment for en face staining with DAPI (nuclei), antibody to phosphorylated Ser10 of histone H3 (p-H3) (mitotic cells), and tdTomato (n=7 for aortas, n=3 for vena cavas). Endothelial cell density was calculated by dividing the mean of cell count per field for each tile per vessel by the tiled area for (B) aortas and (F) vena cavas. The percent p-H3 positive for (C) aortas and (G) vena cavas. Pcna expression was determined by RT-QPCR on control and ACDiEKO samples 7 days after the start of tamoxifen for (D) aortas and (H) vena cavas. (n=3 for controls, n=5 for ACDiEKO) (* p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001).

Transcriptome analysis indicates that the majority of MEF2-dependent genes are part of a LF-MEF2-KLF2/4 axis.

To gain a better understanding of transcriptional dysregulation in the ACDiEKO endothelium, we performed transcriptional profiling with microarray on RNA isolated from the endothelium of the thoracic aorta. Using differential expression criteria of a fold change of ≥ 1.8 fold and a false discovery rate (FDR) < 0.05, there were 1629 differentially expressed genes (DEGs). Notably, there were no differences in markers of differentiated smooth muscle, immune, or mesenchymal cells, indicating that there is no significant contamination with other cell types despite the pathology evident in the ACDiEKO endothelium (Figure III in the data supplement). As discussed above, in vitro studies revealed a LF-MEF2-KLF2/4 regulatory axis in which LF increases MEF2 activity to induce transcription of Klf2 and Klf4 ²⁶. As shown in figure 2, our data strongly supports this model with MEF2 deficiency resulting in substantial decreases in both Klf2 and Klf4, along with corresponding changes in several Klf2/4-dependent genes. Based on these results, we decided to classify MEF2-dependent genes in relation to their dependency on LF and Klf2/4 as an initial framework for further analysis. Therefore, we provisionally categorized DEGs into 1) those that are members of a LF-MEF2 axis, 2) those that are further members of a LF-MEF2-KLF2/4 axis, and 3) those that are LF independent. For exploring the relationship between MEF2 and LF, we compared our ACDiEKO DEGs with those resulting from carotid partial ligation (CPL) (GSE101826) ⁷⁴. In the CPL procedure, the LF gene expression program, of which MEF2 is a part, in the carotid is lost and a DF expression pattern initiated. For exploring the relationship with Klf2/4, we compared DEGs in ACDiEKO aortic ECs and KLF2/4iEKO cardiac ECs (GSE92965) ⁵⁵. In this analysis we focused on the 800 ACDiEKO DEGs that were also quantified in the RNA-Seq data for both KLF2/4iEKO and CPL (Table I in the Data Supplement). Although the ACDiEKO and CPL RNA samples were extracted in situ from the endothelium of large arteries and the KLF2/4iEKO samples extracted from isolated cardiac endothelial cells, there is a remarkable concordance with 86% of the ACDiEKO DEGs changed in the same direction for ACDiEKO and CPL (Figure 5A). And, as expected, two of the most downregulated genes in this set are Klf2 and Klf4. This suggests that the overwhelming majority of ACDiEKO DEGs are part of a LF-MEF2 axis. The remaining 14% of ACDiEKO DEGs were not coordinately differentially expressed with CPL and could, therefore, be genes whose regulation by MEF2 does not require LF activation. Notably, 66% of DEGs were concordantly differentially expressed in ACDiEKO, CPL and KLF2/4iEKO, suggesting that the majority of genes are part of the LF-MEF2-KLF2/4 axis. The remaining 20% of the ACDiEKO DEGs were coordinately differentially expressed with CPL but not with KLF2/4iEKO DEGs, suggesting that these genes are regulated by LF through MEF2 but not part of the LF-MEF2-KLF2/4 axis. Although we refer to these as Klf2/4-independent for simplicity, it should be noted that MEF2 and Klf2/4 could jointly regulate these genes, either directly or indirectly, but that Klf2/4 deletion alone does not affect their expression.

Figure 5. Transcriptome analysis indicates that the LF-MEF2-KLF2/4 axis inhibits proliferation whereas MEF2-dependent genes outside this axis are involved in cytoskeletal processes.

A, Comparison of differentially expressed genes (|FC| ≥ 1.8, FDR<0.05) in ACDiEKO aortic endothelium, carotid partial ligation (CPL), and the endothelial specific deletion of Klf2 and Klf4 (KLF2/4iEKO) shows that the majority of genes differentially expressed by Mef2a/c/d-deletion are differentially expressed in the same direction by changing from LF to DF by carotid partial ligation and deletion of Klf2 and Klf4. B, Pathway analysis with Metascape on genes coordinately differentially expressed in ACDiEKO, CPL, and KLF2/4iEKO reveals an association with processes involved in cell proliferation. C, A heat map of proliferation associated genes identified by Metascape that were increased. Only increased genes are shown for legibility. D, Genes differentially expressed by Mef2a/c/d deletion that are not coordinately differentially expressed by Klf2/4 deletion are associated with cytoskeletal and developmental processes. E, A heat map of a subset (|FC| > 2.0) of differentially expressed genes analyzed in D.

MEF2-dependent genes that are part of the LF-MEF2-KLF2/4 axis inhibit endothelial proliferation and those independent of this axis regulate actin filament-based processes and morphogenesis.

We next analyzed biological process enrichment using Metascape ⁹⁵. Because 66% of DEGs appear to be part of a LF-MEF2-KLF2/4 axis, we separately analyzed these and the 34% that appear independent of this axis in order to separate processes that are Klf2/4 regulated from independent processes. Significantly, the top 10 enriched processes for the LF-MEF2-KLF2/4 axis all relate to cell proliferation (Figure 5B) and most of these were induced (Figure 5C). Thus, most of the genes in this putative LF-MEF2-KLF2/4 axis unexpectedly involve regulation of proliferation. This suggests that the LF-MEF2-KLF2/4 axis represses endothelial proliferation, which is consistent with the increased proliferation observed in DF compared to LF regions ¹. Importantly, this implies that there would be more proliferation in KLF2/4iEKO endothelium. Although measurement of proliferation was not reported for KLF2/4iEKO, the reported GSEA analysis revealed that the top dysregulated gene-sets are related to proliferation ⁵⁵.

In striking distinction with the LF-MEF2-KLF2/4 axis, the processes enriched for Klf2/4-independent DEGs were actin filament-based processes and morphogenesis (Figure 5D & E). The enrichment in actin filament-based processes is consistent with the massive increase in F-actin observed in the endothelium (Figure 3A) that is also observed to a lesser extent with Mef2c deletion alone, where there is no change in Klf2/4 expression ²¹. Similarly, migration of SM into the intima occurs, albeit at a lower rate, in the CiEKO and would be expected to be Klf2/4 independent ²¹. Of note, highly induced genes in the Klf2/4-independent set include Pdgfb, Tgfb2, and Nrg1, which may have a role in SM migration into the intima.

Notch pathway genes are reduced in the ACDiEKO endothelium.

Notch pathway genes are upregulated and Notch signaling is activated by LF; and this activation is at least partly responsible for repressing proliferation and inflammation in the endothelium experiencing LF ^59–61. Transcriptome analysis with GSEA revealed downregulation of Notch pathway genes in the ACDiEKO endothelium (Figure 6A). We validated a subset of these genes by RT-QPCR and found lower amounts of the receptor, Notch1, the ligands Dll1 and Jag1, and the downstream transcription factor, Hey2 (Figure 6B). There was also reduction of the gap junction gene, Gja5, which was shown to be Notch-dependent in ECs under LF ⁶⁰. Because MEF2 directly regulates transcription of the Notch ligand, Dll4, in the angiogenic tip cell through an intronic enhancer ²², we also examined its expression but found no change (Figure 6B). The decreased expression of Notch pathway genes in ACDiEKO endothelium suggests that LF controls at least some parts of this pathway through MEF2. To explore the possibility of direct regulation, we examined available ChIP-Seq data for binding of MEF2 family members to putative promoters or enhancers of these genes. For ChIP-Seq studies on human cell lines ^32,96,97, we used the ReMap2020 database and for mice we used the datasets for binding of Mef2a or Mef2c in fetal and adult hearts, which includes endothelial cells in a mix of cell types ^98,99. These data show binding peaks in several human cell lines to the established MEF2 sites in the KLF2 promoter and KLF4 enhancer at −148 kb (Figure VI A–B in the Data Supplement). Similarly, there is binding in murine hearts to the MEF2 sites in the Klf2 promoter and a region at −119 kb conserved with the human Klf4 enhancer (Figure VIIA–B in the Data Supplement). These results indicate that it is possible to identify regulatory elements critical for endothelial expression using other human cell types or in the murine heart. It is important to note that much MEF2 binding is dependent on interaction with cell-type specific transcription factors that may not be identified by this approach ^99,100. Nevertheless, we identified binding of MEF2 factors nearby or within the human NOTCH1, JAG1, and DLL1 genes that coincides with regions of high transcription factor binding density indicative of regulatory regions (Figure VIC–E in the Data Supplement). Similarly, in murine hearts we identified binding peaks to Notch1 and Jag1 but not to Dll1 that coincides with open chromosomal regions (Figure VIIC–E in the Data Supplement). Although the relevance of these binding sites remains to be determined, these data show that MEF2 factors bind in putative regulatory regions of these genes.

Figure 6. Notch pathway genes are reduced in ACDiEKO endothelium.

A, The enrichment plot of genes differentially expressed in ACDiEKO shows negative enrichment in the GO_NOTCH_BINDING signature. The core enrichment genes in this signature are shown in the heatmap. B, RT-QPCR assay on control and ACDiEKO aortic endothelium 7 days after the start of tamoxifen showing reduction in Notch pathway genes (Notch1, Dll1, and Jag1) and Notch target genes (Hey2 and Gja5). Values are shown as mean ± SEM. Statistical testing by Student’s t-test (** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001; n=3 for Control, n=5 for ACDiEKO).

MEF2 binding near DEGs.

We next used this approach to examine MEF2 factor binding to other DEGs to investigate possible direct regulation. This revealed that 66% of the Klf2/4-independent DEGs are the nearest gene to binding peaks for Mef2a/c in murine hearts, which is an enrichment over the 49% for the LF-MEF2-KLF2/4 axis DEGs; and both percentages are higher compare to the 24% for genes that were not differentially expressed. Because some of these binding peaks are at considerable distances from the DEGs, we next examined promoter regions where it is easier to attribute regulation to a nearby gene. This analysis showed that 21% of the Klf2/4-independent promoters and 15% of the LF-MEF2-KLF2 axis promoters display Mef2a/c binding compared to only 6% for non-DEGs (Table II in the Data Supplement). Although further studies are necessary to determine their functional importance, these results support a role for MEF2s in direct regulation of many DEGs. Interestingly, a significant number of DEGs that are members of the putative LF-MEF2-KLF2/4 axis bind to Mef2a/c and the percent binding is higher compare to that of non-DEGs, suggesting possible joint regulation of these genes by MEF2 and KLF2/4.

MEF2 deficiency increases YAP/TAZ activity through increasing the amount of TAZ.

Unexpectedly, we observed that several DEGs are known YAP/TAZ targets. We investigated this further with GSEA by comparing ACDiEKO differential gene expression to the multiple previously published YAP/TAZ signatures (Table III in the Data Supplement)^82–84. This revealed a remarkable enrichment indicative of increased YAP/TAZ activity by deletion of MEF2 (Figure 7A&B and Figure VIII in the Data Supplement). To further evaluate this increase, we examined by RT-QPCR the expression of known endothelial YAP/TAZ target genes Ctgf, Cyr61, Sele, Serpine1 and Tgfb2 ^62,101,102. This confirmed large induction of Ctgf, Cyr61, Tgfb2, Serpine1 (Figure 7C–F), and Sele (Figure 4D and Figure II in the Data Supplement) in the ACDiEKO endothelium. Interestingly, Sele and Serpine1 are induced by YAP/TAZ and repressed by Klf2/4 ^26,90. Thus, their induction could be a consequence of both YAP/TAZ activation and Klf2/4 loss. YAP/TAZ are regulated by nuclear/cytoplasmic shuttling and stability ^103,104. En face IF for YAP did not exhibit any increase in expression or nuclear localization (Figure IX in the Data Supplement). However, the total amount and nuclear localization of TAZ protein was substantially increased (Figure 7G–I) without altering its mRNA level (Figure 7J). This suggests that MEF2 deletion upregulates TAZ post-transcriptionally, possibly through protein stability or translational efficiency. Significantly, TAZ protein is higher in the aortic inner-curvature which experience DF and have lower MEF2 activity compared with the thoracic aorta ⁶³. This suggests that MEF2 may be a physiological repressor of TAZ by flow. Moreover, increased TAZ could be one mechanism for increased proliferation in ACDiEKO endothelium, similar to how YAP/TAZ activation by endothelial-specific deletion of the upstream inhibitory kinases, Lats1 and Lats2, increases proliferation ^101,102.

Figure 7. MEF2 deficiency increases YAP/TAZ activity and TAZ protein.

A, Genes differentially expressed in ACDiEKO are enriched in multiple YAP/TAZ signatures (all fulfill nominal p-value <0.01 and FDR q-value <0.1). B, Enrichment plot of genes differentially expressed in ACDiEKO show positive enrichment in the Cordenonsi YAP signature. The top 20 core enrichment genes in this signature are shown in the heatmap. (C-F) RT-QPCR quantification on aortic endothelium samples 7 days after the start of tamoxifen showing upregulation of YAP/TAZ regulated genes. G, En face IF showing increased TAZ protein in the ACDiEKO thoracic aorta 10 days after the start of tamoxifen. Representative images of a single z-plane are shown. H & I, Total and nuclear Mean Fluorescent Intensity (MFI) for TAZ as determined with Imaris software showing increased total and nuclear TAZ staining (N=3 for control and N=5 for ACDiEKO). J, RT-PCR of Wwtr1 (TAZ) showing no change in mRNA. Values are shown as mean ± SEM and statistical testing by Student’s t-test (* p ≤ 0.05, *** p ≤ 0.001, **** p ≤ 0.0001; n=3 for Control, n=5 for ACDiEKO).

DISCUSSION

Our results support the notion that MEF2 has roles in many of the functions that promote an atheroprotective endothelium. Specifically, endothelial MEF2 activity inhibits SM migration from the media to the intima (Figure V in the Data Supplement and our previous publication ²¹), enhancement of the actin cytoskeleton (Figure 3), thrombosis, inflammation (Figure 1 and Figure I in the Data Supplement), and proliferation (Figure 4). Moreover, we propose from comparison with CPL that 86% of the MEF2-dependent genes are putatively flow-regulated, indicating that LF is a major MEF2 activator in the endothelium. This is in line with in vitro experiments by which LF activates MEF2 ^26,36. MEF2’s endothelial functions are likely both direct and indirect. Direct regulation of many DEGs could be through MEF2 binding to their promoter and distal elements as shown by ChIP-Seq (Table II in the Data Supplement); however, evaluating the functions of these requires further studies. Defining indirect pathways is complicated by the complexity of the pathologies resulting from MEF2 deficiency as well as those resulting from deficiencies in other transcription factors or CPL ^55,74,94. In the CPL model specifically, macrophages and platelets are required for much differential expression and for alternative splicing of fibronectin and other genes ^74,94. A similar alternative splicing of fibronectin occurs in ACDiEKO (Figure 3G&H and Figure IV in the Data Supplement) but the role of macrophages and platelets remains to be tested, as does any role in differential expression. Acknowledging these caveats, we propose that there are at least three indirect transcriptional pathways through which MEF2 functions to produce an anti-thrombotic, anti-inflammatory, and anti-proliferative endothelium: Klf2/4, Notch, YAP/TAZ.

Application of LF in vitro induces transcription of Klf2 and Klf4 through MEF2 elements ^26,105,106. However, establishing this regulation in vivo has been challenging. We previously reported that endothelial-specific deletion of Mef2c does not alter Klf2/4 expression ²¹ and now report that combined deletion of both Mef2a and Mef2c also had no effect (Figure I in the Data Supplement). We have now determined that deletion of all three MEF2 factors expressed in the endothelium is required to alter Klf2/4 expression. The extent of this regulation is surprisingly comprehensive with MEF2 being responsible for expression of about 90% of both Klf2 and Klf4 mRNAs (Figure 2). Klf2 and Klf4 themselves redundantly regulate a large number of genes in the endothelium ⁵⁵. Thus, the severe reductions in both Klf2 and Klf4 by MEF2 deficiency translates into the large percentage of ACDiEKO DEGs (66%) that are similarly differentially expressed in KLF2/4iEKO and are, therefore, putatively MEF2-dependent through Klf2/4 in a LF-MEF2-KLF2/4 axis (Figure 5). There is notable transcription factor redundancy in this axis with deletion of all three MEF2s expressed in the endothelium being required to alter Klf2 and Klf4 expression, and for combined deletion of Klf2 and Klf4 being required to produce large scale changes in gene expression. Considering the number of genes regulated in this axis and the severe disruption of vascular homeostasis resulting from its dysfunction, this redundancy ensures the execution of critical functions are robust to perturbation of individual factors. While many of the genes in this axis are established Klf2/4 targets (Figure 2), it was an unexpected and novel finding that 46% of the genes jointly altered in ACDiEKO and KLF2/4iEKO are involved in proliferation. In fact, the top 10 enriched biological process categories in ACDiEKO are related to DNA replication and mitosis (Figure 5), which is consistent with the enhanced proliferation in the ACDiEKO endothelium (Figure 4). Although endothelial proliferation was not reported for the KLF2/4iEKO, the most enriched signatures by GSEA analysis are G2M checkpoint, E2F targets, and Myc targets ⁵⁵, suggesting that increased proliferation is also a feature of Klf2/4 deficiency. There are many similarities in the phenotypes of ACDiEKO and KLF2/4iEKO such as pulmonary hemorrhage, bradycardia, and thrombocytopenia. However, there are also important differences, perhaps the most striking are that we did not observe the increased vascular permeability and edema reported for KLF2/4iEKO, the hemorrhage may be less extensive, and the onset of lethality is later ⁵⁵. The causes of the differences in permeability and hemorrhage are unclear, and require further investigation. It is possible that the extent of hemorrhage may cover any measurable differences in vascular permeability between these mice. We speculate that lethality is a consequence of Klf2/4 deficiency and the relative delay in timing for ACDiEKO lethality is from the decrease in Klf2/4 expression following the loss of MEF2 activity. The true extent of the role of Klf2/4 deficiency in the ACDiEKO phenotype will require rescue experiments to express KLF2 or KLF4 in ACDiEKO to determine. Nevertheless, there are important differences between MEF2 deficiency and KLF2/4 deficiency that indicate additional MEF2 functions. For instance, migration of SM into the intima and enhanced actin cytoskeleton occur in mice with endothelial-specific deletion of Mef2c or both Mef2a and Mef2c without any changes in Klf2 or Klf4 expression. Because expression of Klf2/4 appears to be a sensitive indicator of LF activation of MEF2, the observation that these processes occur without changes in Klf2/4 levels suggests that they are MEF2-dependent but independent of LF activation or Klf2/4. Additionally, ACDiEKO increases WBC numbers, IL6 expression, and causes hypothermia, indicative of severe systemic inflammation, which was not reported in the KLF2/4iEKO, suggesting that this is through MEF2 functions independent of Klf2/4 or through combined loss of MEF2 and KLF2/4 activity.

Dysregulation of two other flow dependent pathways also results from MEF2 deficiency: Notch and YAP/TAZ. These changes could explain, at least in part, the phenotypic differences between ACDiEKO and KLF2/4iEKO mice. It should be noted, however, that these pathways were altered in the context of both MEF2 and Klf2/4 deficiency and could be the result of this combined loss rather than MEF2 deficiency alone. Notch1 acts as a mechanosensor and promotes an atheroprotective endothelium independent of Klf2/4 regulation ^59–61. Because Notch1 and some of its ligands are induced by LF, this pathway is both LF-regulated and a transducer of LF signals. Our analysis shows that MEF2 deficiency reduces the levels of Notch1 and the Notch ligands, Jag1 and Dll1, as well as Notch activity as determined by expression of its target genes Hey2 and Gja5 (Figure 6). This suggests that LF induction of these Notch pathway genes is downstream of MEF2. Although Dll4 is a direct MEF2 target in angiogenic tip cells with an essential MEF2 site in a Dll4 enhancer ²², MEF2 has not previously been implicated in the regulation of other Notch ligands. The mechanism for this regulation requires further study, but we speculate from our transcriptomic analysis that Dll1 may be MEF2-regulated through Klf2/4 because it is decreased by Klf2/4 deletion based on RNA-Seq. The Notch target, Hey2, is also downregulated by RNA-Seq, suggesting that Notch signaling is decreased by Klf2/4 deficiency. However, decreased expression of Notch1 and Jag1 were not observed in RNA-Seq data from KLF2/4iEKO, suggesting that they are possibly regulated by MEF2 either directly or through other intermediary factors. In support of direct regulation, members of the MEF2 family bind near and within NOTCH1 and JAG1 in several human cell lines and in murine hearts (Figures VI and VII in the Data Supplement). While intriguing, additional studies are required to clarify the role of these binding sites for flow regulation.

In contrast to the atheroprotective functions of Klf2/4 and Notch, YAP/TAZ promotes an atheroprone endothelium through increasing proliferation and inflammation ^62,63. LF reduces YAP/TAZ activity through an integrin / G_α13 pathway repressing Rho, resulting in LATS1/2 activation ^62,107. Rho repression of LATS1/2 requires F-actin but the mechanism remains to be defined ¹⁰⁴. Our data indicate that MEF2 deficiency upregulates YAP/TAZ regulated genes. A plausible mechanism for this could be through the large increase in F-actin caused by MEF2 deficiency (Figure 3A), which would presumably increase nuclear localization of both YAP and TAZ. However, our data indicates that MEF2 deficiency increases the total amount and nuclear localization of TAZ protein without altering the level of its mRNA or the nuclear localization of YAP. This suggests that MEF2 might regulate TAZ stability and hence activity while not affecting YAP. There is precedent for this in the regulation of TAZ stability by the PI3K pathway through a TAZ-specific N-terminal phosphodegron targeted for phosphorylation by GSK3 ¹⁰⁸ and in TAZ-specific degradation by the β-catenin destruction complex ^109,110. Because YAP/TAZ are mechanosensitive transcription factors with a defined, direct pathway for repression by LF, a role for MEF2 in their regulation would appear to be indirect. We speculate that MEF2 regulation of TAZ might be similar to the situation with Notch, which is also a mechanosensitive transcriptional coactivator, in controlling the expression or activity of components of the YAP/TAZ pathway to modulate their activity in response to LF. What these components are and how they are regulated by MEF2 remains to be determined.

In summary, from our results and previous studies, we propose that MEF2 is essential for promoting an anti-thrombotic, anti-inflammatory, and anti-proliferative endothelium. In part, this is through regulation of three pathways either directly or indirectly: Klf2/4, Notch, and YAP/TAZ. The chronic effect would be to promote an atheroprotective endothelium. However, as evidenced by the ACDiEKO phenotype where the acute effect of loss of these factors is increased inflammation and hemorrhage, MEF2 activity in the endothelium has an unappreciated role in modulating the inflammatory response that warrants further elucidation. As such, MEF2 is a compelling therapeutic target for atherosclerosis, thrombosis, and inflammation.

HIGHLIGHTS.

• Endothelial MEF2 activity is essential for vascular homeostasis and expression of Klf2 and Klf4 in the endothelium.

• MEF2 deficiency increases expression of YAP/TAZ target genes through increasing the amount of TAZ.

• MEF2 regulates endothelial expression of the flow-responsive Notch pathway genes, Notch1, Dll1, and Jag1.

Nonstandard Abbreviations and Acronyms

ACDiEKO

Inducible endothelial-specific deletion of Mef2a, -c, -d

ACiEKO

Inducible endothelial-specific deletion of Mef2a, -c

CiEKO

Inducible endothelial-specific deletion of Mef2c

CPL

Carotid partial ligation

DEG

Differentially expressed gene

DF

Disturbed flow

EC

Endothelial cell

ERK5

Extracellular signal-regulated kinase 5

HDAC

Histone deacetylase

KLF2

Krüppel-like factor 2

KLF4

Krüppel-like factor 4

KLF2/4iEKO

Inducible endothelial-specific deletion of Klf2, −4

LF

Laminar flow

MEF2

Myocyte enhancer factor 2

RhoA

Ras homolog family member A

RT-QPCR

Reverse transcriptase – quantitative PCR

SM

Smooth muscle

TAZ

Transcriptional coactivator with PDZ-binding motif

YAP

Yes-associated protein