Abstract

   The pleiotropic benefits of statins in cardiovascular diseases that are
   independent of their lipid-lowering effects have been well documented,
   but the underlying mechanisms remain elusive. Here we show that
   simvastatin significantly improves human induced pluripotent stem
   cell-derived endothelial cell functions in both baseline and diabetic
   conditions by reducing chromatin accessibility at transcriptional
   enhanced associate domain elements and ultimately at
   endothelial-to-mesenchymal transition (EndMT)-regulating genes in a
   yes-associated protein (YAP)-dependent manner. Inhibition of
   geranylgeranyltransferase (GGTase) I, a mevalonate pathway
   intermediate, repressed YAP nuclear translocation and YAP activity via
   RhoA signaling antagonism. We further identified a previously
   undescribed SOX9 enhancer downstream of statin–YAP signaling that
   promotes the EndMT process. Thus, inhibition of any component of the
   GGTase–RhoA–YAP–SRY box transcription factor 9 (SOX9) signaling axis
   was shown to rescue EndMT-associated endothelial dysfunction both in
   vitro and in vivo, especially under diabetic conditions. Overall, our
   study reveals an epigenetic modulatory role for simvastatin in
   repressing EndMT to confer protection against endothelial dysfunction.
     __________________________________________________________________

   Cardiovascular diseases (CVDs) are a leading cause of global
   mortality^[54]1. Statins, which are widely prescribed to reduce
   low-density lipoprotein cholesterol^[55]2, have pleiotropic activities
   that confer CVD protection beyond their lipid-lowering
   effect^[56]3–[57]5. While the precise mechanisms underlying these
   pleiotropic effects remain poorly understood, accumulating evidence
   suggests that epigenetic modifications play a key role in CVD
   initiation and progression^[58]6. For example, statins have been shown
   to protect against cancer and vascular diseases by remodeling chromatin
   accessibility through DNA methylation and histone
   acetylation^[59]7,[60]8. Thus, understanding of the role of statins in
   endothelial epigenetics could advance alternative therapeutic
   approaches to treat CVDs.

   Various human induced pluripotent stem cell (iPSC)-derived
   cardiovascular cells have been shown to accurately recapitulate drug
   responses in patients and to uncover pathogenic mechanisms of
   CVDs^[61]9. In this study, we integrated iPSC-derived endothelial cells
   (iPSC-ECs) and assay for transposase-accessible chromatin using
   sequencing (ATAC-seq) to show that simvastatin improved endothelial
   functions under baseline and diabetic or hyperglycemic (HG) conditions.
   It did so by decreasing chromatin accessibility at TEAD elements, such
   as the SOX9 enhancer region, and repressing the transcriptional
   activity of EndMT-regulating genes. We identified YAP as a direct
   regulator that enhanced chromatin accessibility at the SOX9 enhancer to
   upregulate EndMT genes. Simvastatin-regulated YAP activity was not
   affected by endothelial cell (EC) densities or mechanical
   perturbations. The inhibition of GGTase I, a key intermediate in the
   mevalonate pathway, was equally effective as simvastatin at improving
   endothelial functions by blocking YAP nuclear translocation and
   repressing EndMT in a RhoA-dependent manner. Genetic or pharmacological
   inhibition of any component of the GGTase–RhoA–YAP–SOX9 signaling axis
   effectively improved endothelial functions under both in vitro and in
   vivo diabetic conditions. In conclusion, our study identifies the
   mechanism underlying statin pleiotropy in endothelial protection via
   suppressing YAP-dependent chromatin remodeling at EndMT-regulating
   enhancer regions. The inhibition of YAP–SOX9 or their upstream
   regulators (that is, GGTase or RhoA) could represent alternative
   therapeutic strategies for treating endothelial dysfunction-associated
   CVDs, such as type 2 diabetes (T2D).

Results

Simvastatin improves endothelial functions and alters the transcriptome

   To investigate the effects of statin pleiotropy on endothelial
   function, we used an established protocol^[62]10 to generate iPSC-ECs
   from three healthy donors without familial hypercholesterolemia or CVDs
   ([63]Fig. 1a). These iPSC-ECs exhibited a typical ‘cobblestone’
   morphology and expressed high levels of EC-specific markers, such as
   CD31 and vascular endothelial (VE)-cadherin ([64]Fig. 1b). We first
   tested the effects of seven statins, namely, pravastatin, atorvastatin,
   mevastatin, fluvastatin, lovastatin, rosuvastatin and simvastatin, at
   three concentrations (0.1 μM, 1 μM and 10 μM) on iPSC-EC functions.
   Because expressing high levels of NOS3 is a hallmark of functional ECs,
   we quantified its mRNA levels in vehicle- and simvastatin-treated
   iPSC-ECs and observed that simvastatin was the most potent statin that
   can upregulate NOS3 expression in a dose-dependent manner ([65]Fig. 1c
   and [66]Extended Data Fig. 1a). As doses of simvastatin beyond 10 μM
   compromised cell viability and increased reactive oxygen species (ROS)
   production in iPSC-ECs ([67]Extended Data Fig. 1b,[68]c), we chose 1 μM
   simvastatin for the following functional assays and transcriptomic and
   epigenetic profiling in iPSC-ECs. Consistent with the upregulated
   expression of NOS3, simvastatin improved the capacity of in vitro tube
   formation ([69]Fig. 1d and [70]Extended Data Fig. 1d) and nitric oxide
   (NO) production ([71]Fig. 1e) in iPSC-ECs compared to the vehicle
   control.

Fig. 1 |. Simvastatin improves endothelial function by altering
chromatin-associated transcriptome profiles of iPSC-ECs.

   Fig. 1 |
   [72]Open in a new tab

   a, Schematic diagram of the experimental design. iPSC-ECs
   differentiated from three healthy donors (D1, D2 and D3) were treated
   with statins for functional assays and multiomic analysis. CHIR,
   CHIR99021. b, Representative immunofluorescent images showing iPSC-ECs
   (iECs) from three healthy donors (D1, D2 and D3) expressing the typical
   endothelial surface markers CD31 (green) and VE-cadherin (red). Scale
   bar, 50 μm. c, Simvastatin exhibits a dose-dependent upregulation of
   NOS3 in iPSC-ECs (n = 4 biological samples). d, Simvastatin-treated
   iPSC-ECs showing a higher density of capillary-like tubes than the
   vehicle (DMSO)-treated counterparts (n = 3 biological samples). e,
   Simvastatin-treated iPSC-ECs produce a higher level of NO than the
   vehicle-treated counterparts (n = 3 biological samples). f,
   Transcriptomic analysis showing differential expression of genes in
   iPSC-ECs after treatment with simvastatin versus vehicle control. g,h,
   Enrichment analysis showing upregulated (g) and downregulated (h) DEGs
   in simvastatin-treated iPSC-ECs compared to the vehicle control (red,
   terms associated with vascular function; blue, terms associated with
   epigenetic regulation). rRNA, ribosomal RNA. i,j, Circle plots showing
   upregulated (i) and downregulated (j) DEGs under the indicated
   ontologies in simvastatin- versus vehicle-treated iPSC-ECs. All data
   are presented as mean ± s.e.m. Unpaired two-sided Student’s t-test
   (c–e).

   Next, we conducted next-generation sequencing analysis on vehicle- and
   simvastatin-treated iPSC-ECs. Compared to the vehicle, simvastatin
   upregulated 1,019 genes and downregulated 1,600 genes in iPSC-ECs
   ([73]Fig. 1f and [74]Extended Data Fig. 1e). The upregulated genes were
   enriched in processes such as angiogenesis, anti-inflammatory response
   and extracellular matrix (ECM) organization ([75]Fig. 1g). Notably,
   these gene ontology (GO) terms for biological processes are associated
   genes that enhance endothelial functions ([76]Fig. 1i): ETS1 and PPARA
   encode negative regulators of the inflammatory response^[77]11; NOS3,
   TIE1 and EPHB4 encode mediators of angiogenesis^[78]12; VWF and KDR are
   mature endothelial marker genes^[79]13; and KLF2 and KLF4 encode
   critical endothelial transcription factors (TFs) involved in
   endothelial biology and physiology^[80]14. By contrast, simvastatin
   downregulated genes such as HMGB2, CHAF1B, SET and EED, which are
   associated with epigenetic regulation of transcription, nucleosome
   assembly and chromatin organization ([81]Fig. 1h,[82]j). No overt
   differences in cell proliferation between vehicle- versus
   simvastatin-treated iPSC-ECs were observed ([83]Extended Data Fig. 1f).
   GO terms for cellular component revealed that most downregulated genes
   were associated with the nucleus and chromosomes ([84]Extended Data
   Fig. 1g,[85]h). Collectively, these results suggest that the beneficial
   effects of simvastatin on ECs might be achieved, in part, by modulating
   the activity of associated genes at the chromatin level.

Simvastatin alters enhancer accessibility in iPSC-ECs

   While statins can regulate gene expression in cancer and
   atherosclerosis via epigenetic inhibition of histone
   deacetylase^[86]8,[87]15, how they alter the chromatin accessibility of
   target genes in vascular cells remains unknown. As we observed a
   significant downregulation of genes associated with epigenetic
   regulation in simvastatin-treated iPSC-ECs, we next conducted a
   genome-wide chromatin accessibility assessment in another set of
   vehicle- and simvastatin-treated iPSC-ECs using ATAC-seq. In line with
   our transcriptomic data ([88]Fig. 1h), we detected a reduced intensity
   of chromatin accessibility peaks (±5 kb of ATAC signal summit) in
   simvastatin- versus vehicle-treated iPSC-ECs ([89]Fig. 2a).
   Interestingly, our ATAC peak-annotation results showed that most of the
   accessible regions in iPSC-ECs were intergenic (36%) and intronic
   (45.7%), and 11% of ATAC peaks were promoters ([90]Fig. 2b). Therefore,
   most ATAC regions in iPSC-ECs were primarily associated with
   non-promoter regulation. To better understand the regulatory role of
   ATAC peak regions in iPSC-ECs, we next annotated our ATAC-seq data with
   the chromatin immunoprecipitation followed by sequencing (ChIP–seq)
   datasets of diverse histone marks in iPSC-ECs from the Encyclopedia of
   DNA Elements (ENCODE) ([91]Fig. 2c). Active enhancers are characterized
   by the concomitant presence of histone 3 lysine 4 (H3K4)
   monomethylation (H3K4me1) and histone 3 lysine 27 (H3K27) acetylation
   (H3K27ac), and poised enhancers harbor H3K27 trimethylation (H3K27me3)
   and H3K4me1 (ref. [92]16). Because our ATAC-seq peak regions in
   iPSC-ECs were substantially related to H3K27ac and H3K4me1 and devoid
   of H3K27me3 ([93]Fig. 2c), we infer that they are primarily located at
   active enhancer regions.

Fig. 2 |. Simvastatin reduces chromatin accessibility at TEAD elements and
represses YAP activity.

   Fig. 2 |
   [94]Open in a new tab

   a, Aggregated ATAC-seq signals across the peak summit (±5 kb) showing
   reduced accessibility of iPSC-ECs in the simvastatin group as compared
   to the vehicle control group (blue, vehicle control; red, simvastatin).
   CPM, counts per million. b, Annotation of all chromatin accessibility
   regions from the ATAC-seq data of simvastatin- and vehicle-treated
   iPSC-ECs. UTR, untranslated region. c, A heatmap showing the enrichment
   and depletion of various histone marks (from ENCODE; see [95]Methods
   for ENCODE accessions) at the peak regions of our ATAC-seq dataset.
   H3K4me3, H3K4 trimethylation; H3K36me3, histone 3 lysine 36
   trimethylation. d, ATAC-seq footprinting heatmap showing global
   differential accessibility scores between control- and
   simvastatintreated iPSC-ECs (FDR < 0.1). e, Motif-enrichment analysis
   at the ATAC-seq sites identified KLFs and TEADs as the most
   predominantly affected motifs with differential chromatin
   accessibility. TEAD, TEAD1 (Homer unpublished data); TEAD1 (Homer
   encode data). f, ATAC-seq footprinting heatmap showing differential
   accessibility scores between control- and simvastatin-treated samples
   at loci marked with TEAD or KLF motifs (fold change (FC) > 0.5 and FDR
   < 0.01). g, KEGG pathway analysis of the ATAC-seq peaks with TEAD
   motifs. cAMP, cyclic AMP; RAP1, repressor activator protein; TRP,
   transient receptor potential. h, Representative immunofluorescent
   images showing reduced YAP nuclear localization in iPSC-ECs treated
   with simvastatin versus vehicle control. Blue, DAPI; green, YAP. Scale
   bar, 100 μm. i, Quantitative data showing nuclear/cytoplasmic YAP
   ratios in iPSC-ECs treated with simvastatin and vehicle control (n = 11
   cells). j, Western blot analysis showing that simvastatin downregulated
   total YAP, SOX9 and CTGF protein levels in iPSC-ECs (n = 2 biological
   samples). k, Quantification of YAP, SOX9 and CTGF protein-level changes
   in simvastatin- and vehicle-treated iPSC-ECs. All data are presented as
   mean ± s.e.m. Unpaired two-sided Student’s t-test (i,k).

Simvastatin reduces chromatin accessibility by blocking YAP activity

   Next, we investigated the chromatin accessibility of ATAC peaks
   globally and sought to identify the increased and decreased chromatin
   accessible regions caused by simvastatin treatment ([96]Fig. 2d).
   Unsupervised motif-enrichment analysis showed that the Krüppel-like
   factor and specificity protein (KLF and SP) family of TFs were the most
   significantly enriched motifs in regions with increased accessibility
   ([97]Fig. 2e,[98]f). By contrast, TFs in the TEAD family (that is,
   TEAD, TEAD1, TEAD3 and TEAD4) were the most significantly enriched
   motifs in regions with decreased accessibility ([99]Fig. 2e,[100]f). In
   line with our initial ATAC peak annotations ([101]Fig. 2b), when we
   annotated these ATAC peaks based on their increased or decreased
   accessibility, we observed that simvastatin-altered chromatin regions
   remained primarily enhancer regions rather than promoter regions
   ([102]Extended Data Fig. 2a).

   TEADs are broadly expressed TFs that regulate differentiation,
   development and tumorigenesis. However, little is known about their
   role in the vascular system^[103]17. As KLF and SP regulate vascular
   functions by binding to the promoters and enhancers of genes with
   similar DNA sequences^[104]14, we hypothesized that TEADs regulate
   endothelial functions through a similar mechanism. Indeed, after we
   binned ATAC signals into enhancer regions and performed motif analysis,
   we observed that the motifs that were associated with vascular
   endothelial growth factor (VEGF) signaling and endothelial function,
   such as ETV2, JUNB and ATF3, showed increased accessibility at enhancer
   regions ([105]Extended Data Fig. 2b)^[106]18. Importantly, consistent
   with our global ATAC-seq analysis ([107]Fig. 2e,[108]f), the TEAD
   motifs remained the most significantly enriched motifs with decreased
   accessibility at enhancer regions ([109]Extended Data Fig. 2c).

   Next, we performed Kyoto Encyclopedia of Genes and Genomes (KEGG)
   pathway enrichment analysis of adjacent genes in the regions associated
   with TEAD and KLF motifs. Our data further confirmed that these motifs
   were enriched in pathways associated with mesenchymal cells such as
   Hippo and transforming growth factor (TGF)-β signaling ([110]Fig. 2g)
   and EC functions such as VEGF signaling ([111]Extended Data Fig. 2d),
   respectively. Similarly, GO pathway analysis revealed that
   simvastatin-altered accessibility regions were mainly associated with
   vascular functions, including angiogenesis and VEGF receptor signaling
   ([112]Extended Data Fig. 2d–[113]f). Collectively, our in-depth
   analysis of ATAC-seq data suggested that simvastatin exerts its
   pleiotropic effects on iPSC-ECs by reducing chromatin accessibility
   primarily at the enhancer regions associated with TEAD motifs.

   Although TEADs (TEAD1–TEAD4) can directly bind to DNA, they are not
   able to regulate gene expression independently due to the lack of
   transcription activation domains. When forming a complex in the
   nucleus, YAP and TEAD can activate target genes^[114]19. To determine
   the effect of simvastatin treatment on YAP activity in iPSC-ECs, we
   first assessed the subcellular distribution of YAP by performing
   immunofluorescence staining ([115]Fig. 2h,[116]i). At the baseline
   condition, both cytoplasmic and nuclear YAP were detectable in
   iPSC-ECs. However, simvastatin-treated iPSC-ECs showed a significant
   reduction in nuclear YAP, suggesting an inhibitory effect of
   simvastatin on YAP nuclear translocation. This finding was further
   supported by western blot results, which showed a significant decrease
   in total YAP protein levels ([117]Fig. 2j,[118]k). The major downstream
   proteins of YAP signaling, SOX9 and connective tissue growth factor
   (CTGF), were also downregulated by simvastatin in iPSC-ECs ([119]Fig.
   2j,[120]k). In sum, our results suggest that simvastatin can decrease
   chromatin accessibility at TEAD elements through inhibiting YAP
   signaling in iPSC-ECs.

   YAP activity was previously shown to be negatively regulated by cell
   density^[121]20. Therefore, we performed comparative experiments by
   treating confluently and sparsely seeded iPSC-ECs with vehicle or
   simvastatin to assess their contributions to YAP activity. Simvastatin
   significantly downregulated YAP signaling in both confluent and sparse
   iPSC-ECs, as well as reduced YAP and SOX9 protein levels, independent
   of cell density ([122]Supplementary Fig. 1a–[123]d). Simvastatin also
   downregulated YAP in primary ECs ([124]Supplementary Fig. 2a–[125]d),
   supporting simvastatin’s inhibitory role in endothelial YAP signaling.

Simvastatin inhibits TEAD–YAP activity

   To explore the kinetics of statin-induced transcriptome and epigenetic
   changes, iPSC-ECs treated with simvastatin for 0–72 h were subjected to
   both RNA-sequencing (RNA-seq) and ATAC-seq analyses. Principal
   component analysis (PCA) of RNA-seq data showed that simvastatin
   induced significant changes in global transcriptomic profiles
   ([126]Fig. 3a). Simvastatin-induced changes in global transcriptomic
   profiles were observed as early as 12 h and were sustained until the
   final examined time point at 72 h ([127]Extended Data Fig. 3a). KEGG
   pathway analysis showed that simvastatin upregulated the steroid
   biosynthesis pathway ([128]Fig. 3b). By contrast, simvastatin
   significantly downregulated the Hippo signaling pathway ([129]Fig. 3b)
   and concomitantly upregulated KLF2 and KLF4 and typical endothelial
   marker genes (NOS3, CDH5 and KDR) ([130]Fig. 3c and [131]Extended Data
   Fig. 3b). Genes downstream of the Hippo–YAP pathway^[132]21 were
   downregulated after simvastatin treatment ([133]Fig. 3d and
   [134]Extended Data Fig. 3c). Moreover, mesenchymal genes, such as
   TGFBR1, TWIST1, SNAI2 and SOX9, were also significantly downregulated
   by simvastatin ([135]Extended Data Fig. 3d).

Fig. 3 |. RNA-seq and ATAC-seq analyses of simvastatin-treated iPSC-ECs at
different time points.

   Fig. 3 |
   [136]Open in a new tab

   a, PCA plot of RNA-seq results of simvastatin-treated iPSC-ECs at 0 h,
   12 h, 24 h and 72 h. PC, principal component. b, KEGG pathway analysis
   of RNA-seq data showing multiple dysregulated pathways, including the
   Hippo pathway, in iPSC-ECs after 72 h of treatment with simvastatin. c,
   Normalized expression levels of endothelial marker genes NOS3 and KLF2
   from RNA-seq data of simvastatin-treated iPSC-ECs at different time
   points (n = 2 RNA-seq biological samples). d, Normalized expression
   levels of YAP downstream genes TGFB1 and FGF1 from RNA-seq data of
   simvastatin-treated iPSC-ECs at different time points. e, Heatmap of
   ATAC-seq data of simvastatin-treated iPSC-ECs at 0 h, 12 h, 24 h and 72
   h. f, TEAD motifs were significantly downregulated in iPSC-ECs after
   simvastatin treatment for 12 h, 24 h and 72 h versus 0 h. Values
   indicate −log[10] (P values). NS, not significant. g, Schematic design
   of a YAP–TEAD activity fluorescence reporter system. CMV,
   cytomegalovirus; IRES, internal ribosome entry site. h, Representative
   fluorescent images of iPSC-ECs transfected with a YAP^GFP–TEAD^mCherry
   reporter before treatment with simvastatin and vehicle control. i,
   Schematic design of a YAP–TEAD activity luciferase reporter system. j,
   Simvastatin reduced TEAD activity in iPSC-ECs compared to vehicle
   control over the course of 72 h using a TEAD response element-driven
   luciferase reporter (n = 5 biological samples). All data are presented
   as mean ± s.e.m. One-way ANOVA corrected with the Bonferroni method
   (j).

   To assess the correlations between simvastatin-induced changes in
   chromatin accessibility and gene expression over the course of 72 h, we
   further analyzed ATAC-seq data generated from the same set of iPSC-ECs.
   Our data showed that the changes in chromatin accessibility were
   initiated after 12 h of simvastatin treatment, which became more
   striking at 24 h and 72 h ([137]Fig. 3e). Our time-resolved ATAC-seq
   data showed that 24 h of simvastatin treatment resulted in the most
   significant chromatin changes, such as downregulation of TEAD motifs
   ([138]Fig. 3e,[139]f). Moreover, we observed a strong concordance
   between chromatin accessibility and gene expression changes in iPSC-ECs
   after 24 h of simvastatin treatment ([140]Extended Data Fig. 4a).
   Collectively, these data strongly support a protective role of
   simvastatin in iPSC-ECs by modulating chromatin accessibility and gene
   expression patterns that confer improved EC functions.

   After observing a significant decrease in nuclear YAP
   immunofluorescence intensity in iPSC-ECs following 24 h of simvastatin
   treatment ([141]Extended Data Fig. 4b,[142]c), we further leveraged a
   reporter system comprising a YAP-driven green fluorescent protein (GFP)
   tag and a TEAD-responsive element-mediated nuclear histone 2B–mCherry
   tag to examine whether simvastatin can directly inhibit YAP–TEAD
   activity ([143]Fig. 3g)^[144]22. In the YAP–TEAD reporter iPSC-ECs, a
   robust nuclear TEAD signal at baseline ([145]Fig. 3h) was significantly
   diminished after 24 h of simvastatin treatment ([146]Extended Data Fig.
   4d). We also found that simvastatin reduced TEAD activity after only 12
   h of treatment using a TEAD response element-driven luciferase reporter
   ([147]Fig. 3i,[148]j). In sum, these results support a direct role for
   simvastatin in repressing YAP–TEAD activity by blocking nuclear YAP
   translocation in ECs.

Simvastatin inhibits YAP by blocking geranylgeranyl pyrophosphate (GGPP)–Rho
signaling

   Growing evidence suggests that the mevalonate pathway diverges at
   farnesyl pyrophosphate, which either generates squalene for cholesterol
   biosynthesis or gives rise to GGPP, which modulates Rho
   signaling^[149]23. Interestingly, statins were shown to reduce YAP
   activity in breast cancer cells by inhibiting the geranylgeranylation
   of RhoA^[150]24. Therefore, we hypothesized that simvastatin suppresses
   endothelial YAP activity by inhibiting GGPP–Rho signaling. Compared to
   the vehicle control, simvastatin upregulated KLF2 and downregulated
   VCAM1, a YAP target gene associated with inflammation, in iPSC-ECs
   ([151]Fig. 4a). However, mevalonic acid (MA), a product of
   3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase, reversed the
   expression patterns of these genes in the presence of simvastatin
   ([152]Fig. 4a). The addition of GGPP but not squalene in the presence
   of simvastatin resulted in a similar gene expression pattern for KLF2
   and VCAM1 in iPSC-ECs, as did MA, suggesting that simvastatin improved
   endothelial functions by inhibiting GGPP. Simvastatin-induced reduction
   in the nuclear localization of YAP was reversed by MA and GGPP but not
   by squalene ([153]Fig. 4b,[154]c). While simvastatin enhanced
   tube-formation capacity, this effect was blunted by GGPP and MA but not
   by squalene ([155]Fig. 4d and [156]Extended Data Fig. 4e). MA and GGPP,
   but not squalene, abolished simvastatin-mediated improvements on
   endothelial NO release ([157]Fig. 4e).

Fig. 4 |. Simvastatin improves endothelial functions through the
GGTase–RhoA–YAP signaling axis.

   Fig. 4 |
   [158]Open in a new tab

   a, Quantitative PCR analysis showing expression levels of KLF2 and
   VCAM1 in iPSC-ECs treated with simvastatin, simvastatin + MA,
   simvastatin + GGPP, simvastatin + squalene, GGTi298, RhoAi and vehicle
   control (n = 3 biological samples). b, Representative immunofluorescent
   images showing YAP nuclear localization patterns (Cell Signaling
   Technology, 93622) in iPSC-ECs treated with simvastatin, simvastatin +
   MA, simvastatin + GGPP, simvastatin + squalene, GGTi298, RhoAi and
   vehicle control. Blue, DAPI; green, YAP. Scale bar, 100 μm. c,
   Nuclear/cytoplasmic YAP ratios in iPSC-ECs treated with simvastatin,
   simvastatin + MA, simvastatin + GGPP, simvastatin + squalene, GGTi298,
   RhoAi and vehicle control (n = 48 cells). d, Quantification of
   capillary-like tubular networks formed by iPSC-ECs treated with
   simvastatin, simvastatin + MA, simvastatin + GGPP, simvastatin +
   squalene, GGTi298, RhoAi and vehicle control (n = 3 biological
   samples). e, NO production by iPSC-ECs treated with simvastatin,
   simvastatin + MA, simvastatin + GGPP, simvastatin + squalene, GGTi298,
   RhoAi and vehicle control (n = 6 biological samples). f, Schematic
   overview of the regulation of YAP by the mevalonate pathway.
   Metabolites are shown in green boxes, and specific enzymes are also
   labeled. PP, pyrophosphate. All data are presented as mean ± s.e.m.
   One-way ANOVA corrected with the Bonferroni method (a,c–e).

   To further validate the causal role of simvastatin-mediated blockage of
   GGPP–RhoA signaling in YAP inhibition and improved endothelial
   functions, we used GGTi298, a specific GGTase inhibitor targeting the
   geranylgeranylation of RhoA, and a RhoA inhibitor (RhoAi) to treat
   iPSC-ECs. Indeed, both GGTi298 and RhoAi suppressed nuclear
   translocation of YAP ([159]Fig. 4b,[160]c), upregulated KLF2 ([161]Fig.
   4a) and downregulated VCAM1 in iPSC-ECs ([162]Fig. 4a). Moreover,
   GGTi298 and RhoAi improved tube formation and NO release in iPSC-ECs
   ([163]Fig. 4d,[164]e and [165]Extended Data Fig. 4e). In sum, these
   results suggest that simvastatin improved iPSC-EC functions by
   inhibiting the GGTase–RhoA–YAP signaling axis ([166]Fig. 4f).

Simvastatin represses YAP-dependent EndMT to improve EC function

   Endothelial dysfunction is an early manifestation of vascular
   dysfunction in T2D^[167]25. While T2D significantly increases the risk
   of developing coronary artery disease^[168]26, statins have been shown
   to effectively reduce major cardiovascular events in patients with
   T2D^[169]27. To understand the protective role of statins in
   endothelial dysfunction beyond their cholesterol-lowering effects, we
   cultured iPSC-ECs under diabetic conditions in vitro and sought to
   determine whether targeted inhibition of the GGTase–RhoA–YAP axis could
   rescue endothelial dysfunction. Specifically, iPSC-ECs from one male
   and one female donor were subjected to a diabetic milieu composed of
   high glucose, insulin and endothelin 1 (ET-1; ref. [170]28). While such
   an HG condition significantly impaired tube formation and NO production
   capacity in iPSC-ECs, these pathological phenotypes were improved by
   simvastatin, GGTi298 and RhoAi ([171]Fig. 5a–[172]c). Moreover,
   simvastatin, GGTi298 and RhoAi blunted the increase in ROS in iPSC-ECs
   maintained under the HG condition ([173]Extended Data Fig. 5a).

Fig. 5 |. Simvastatin rescues diabetes-induced endothelial dysfunction.

   Fig. 5 |
   [174]Open in a new tab

   a, Representative images of capillary-like tube densities in iPSC-ECs
   under conditions of HG, HG + simvastatin, HG + GGTi298, HG + RhoAi and
   vehicle control. Scale bars, 250 μm. b, Quantification of
   capillary-like tube numbers in iPSC-ECs under conditions of HG, HG +
   simvastatin, HG + GGTi298, HG + RhoAi and vehicle control (n = 3
   biological samples). D1, donor 1 iPSC-ECs; D2, donor 2 iPSC-ECs. c, NO
   production levels in iPSC-ECs under the control condition, the HG
   condition and that of HG + simvastatin reveal that simvastatin restores
   HG-impaired NO production (n = 3 biological samples). d, Transcriptome
   profiles of iPSC-ECs treated with HG, HG + simvastatin and vehicle
   control. e, Venn diagrams of transcriptomic analysis showing
   overlapping genes in HG versus vehicle control and HG versus HG +
   simvastatin. Top: overlapping upregulated genes. Bottom: overlapping
   downregulated genes. f, Gene set enrichment analysis plots of the EMT
   gene set in HG versus vehicle control (top) and HG versus HG +
   simvastatin (bottom). NES, normalized enrichment score, relative to
   vehicle control or HG + simvastatin. g, Heatmap of ATAC-seq data
   showing differential peaks in diabetic stress HG versus vehicle-treated
   iPSC-ECs. h, Motif analysis identifies both increased and decreased
   chromatin accessibility to the ATAC regions in HG-treated iPSC-ECs. i,
   KEGG pathway analysis shows dysregulated pathways in HG conditions
   compared to vehicle control-treated iPSC-ECs. HTLV-1, human T
   lymphotropic virus type 1; ARVC, arrhythmogenic right ventricular
   cardiomyopathy; AMPK, AMP-activated protein kinase. All data are
   presented as mean ± s.e.m. Unpaired two-sided Student’s t-test (b),
   one-way ANOVA corrected with the Bonferroni method (c).

   Next, we performed transcriptome analysis on iPSC-ECs exposed to the
   diabetic condition with or without simvastatin ([175]Fig. 5d).
   Remarkably, iPSC-ECs in both the vehicle and simvastatin–HG (HG +
   simvastatin) groups showed an almost 90% overlap of differentially
   expressed genes (DEGs) compared to those maintained under the HG
   condition, suggesting that simvastatin restored hyperglycemia-induced
   transcriptomic profiles toward the baseline condition ([176]Fig. 5e).
   We also observed a high degree of overlap in the top ten enriched gene
   sets between the vehicle and HG–simvastatin groups ([177]Extended Data
   Fig. 5b). Most of these gene sets were associated with metabolic
   pathways, such as glycolysis and mammalian target of rapamycin complex
   1 signaling. Notably, an interesting gene set involved in the
   epithelial-to-mesenchymal transition (EMT) process was significantly
   activated in the HG group as compared to the vehicle and HG–simvastatin
   groups ([178]Fig. 5f).

   EMT, which refers to the EndMT process in ECs, contributes
   substantially to CVDs^[179]29. Intriguingly, we observed that
   simvastatin effectively repressed several clusters of genes with
   critical EndMT functions that were activated in iPSC-ECs under the
   diabetic condition ([180]Extended Data Fig. 5c). These clusters
   included ECM-producing genes such as COL1A1 and COL3A1; ECM modulator
   genes such as PCOLCE2, POSTN and FBN1; cell adhesion genes such as
   ITGB5, CDH2, VIM and THY1; EMT-associated signaling genes such as
   TGFB1, DKK1, WNT5A, PDGFRB and FGF2; and classic EMT regulator genes
   such as MEST, NID2, SNAI2 and PRRX1. In accordance with
   simvastatin-mediated downregulation of EndMT genes in iPSC-ECs cultured
   under the diabetic condition, our ATAC-seq data showed that chromatin
   accessibility of loci associated with critical endothelial TFs and
   TEADs changed significantly after 7 d of exposure to the diabetic
   condition ([181]Fig. 5g,[182]h). While simvastatin reduced YAP–TEAD
   activity, increased accessibility at chromatin associated with TEAD
   motifs suggested that YAP–TEAD was activated under the diabetic
   condition. Moreover, KEGG pathway analysis of the neighboring genes of
   the ATAC peaks revealed that the Hippo signaling pathway was
   significantly upregulated in HG-treated ECs ([183]Fig. 5i).

   We hypothesized that the rescue effects of simvastatin on T2D-induced
   endothelial dysfunction occur through the inhibition of YAP signaling.
   Indeed, we observed that simvastatin significantly reduced the protein
   levels of YAP and its downstream target SOX9, which were elevated under
   diabetic conditions ([184]Fig. 6a,[185]b and [186]Extended Data Fig.
   5d,[187]e). Simvastatin, GGTi298 and RhoAi attenuated the accumulation
   of nuclear YAP induced by HG medium in iPSC-ECs ([188]Fig. 6c,[189]d).
   To further confirm the causal role of YAP inhibition in this process,
   we used CRISPR interference (CRISPRi) technology to knock down YAP
   expression ([190]Fig. 6e,[191]f). YAP repression via CRISPRi rescued
   diabetes-induced endothelial dysfunction in iPSC-ECs ([192]Fig. 6g).
   Moreover, the expression levels of VCAM1 and SOX9, two major YAP
   downstream genes, and SNAI2 and TWIST1, two classic EMT regulator
   genes, which were increased by HG medium, were restored to the baseline
   condition by simvastatin, GGTTi298, RhoAi and YAP repression ([193]Fig.
   6h,[194]i). These data collectively support the idea that simvastatin
   protected endothelial function not only at the baseline level but also
   under diabetic conditions by suppressing the GGTase–RhoA–YAP axis.

Fig. 6 |. Simvastatin rescues diabetes-induced endothelial dysfunction via
inhibition of the YAP-dependent mevalonate pathway bifurcation.

   Fig. 6 |
   [195]Open in a new tab

   a, Western blot analysis showing that simvastatin downregulated total
   YAP and SOX9 levels in iPSC-ECs. b, Densitometric quantification of YAP
   and SOX9 protein expression changes in iPSC-ECs treated with vehicle
   control, HG medium, HG + simvastatin, HG + GGTi298 and HG + RhoAi (n =
   2 biological samples). c, Representative immunofluorescent images
   showing YAP nuclear localization in iPSC-ECs treated with vehicle
   control, HG medium, HG + simvastatin, HG + GGTi298 and HG + RhoAi.
   Blue, DAPI; green, YAP. Scale bar, 100 μm. d, Nuclear/cytoplasmic YAP
   ratios in iPSC-ECs treated with vehicle control (n = 61 cells), HG
   medium (n = 58 cells), HG + simvastatin (n = 55 cells), HG + GGTi298 (n
   = 77 cells) and HG + RhoAi (n = 55 cells). e, Schematic design of
   CRISPRi to repress YAP1 transcription. Guided by a designed gRNA,
   CRISPRi uses a Tet-on controlled dCas9 (fused with a KRAB effector
   domain) to target the YAP1 promoter for transcriptional silencing upon
   addition of doxycycline. f, Quantitative PCR analysis showing
   significantly repressed YAP expression after targeted CRISPRi (n = 3
   biological samples). g, Capillary-like tube numbers in iPSC-ECs
   transduced with the YAP CRISPRi (Tet-on dCas9) system in the vehicle
   control condition with dCas9 off, HG + dCas9 off and HG + dCas9 on (n =
   3 biological samples). h, Quantitative PCR analysis showing the
   expression levels of VCAM1 and SOX9 in iPSC-ECs treated with diabetic
   medium (HG), HG + simvastatin, HG + GGTi298, HG + RhoAi and HG + YAP
   CRISPRi (n = 3 biological samples). i, Quantitative PCR analysis of
   SNAI2 and TWIST1 expression in iPSC-ECs treated with vehicle, HG and HG
   + YAP CRISPRi (n = 3 biological samples). All data are presented as
   mean ± s.e.m. Unpaired two-sided Student’s t-test (f), one-way ANOVA
   corrected with the Bonferroni method (b–d,g–i).

Simvastatin rescues endothelial dysfunction in diabetic mice

   To investigate the protective effects of simvastatin and GGTi298 on
   endothelial function in vivo, we administered each compound separately
   to db/db mice, a mouse model of leptin-deficient T2D, via daily oral
   gavage for 8 weeks. We then assessed changes in vascular function using
   wire myography, as previously described^[196]10. After 8 weeks of
   treatment ([197]Fig. 7a), we dissected descending thoracic aortas from
   mice of all groups to evaluate changes in vascular tone ([198]Extended
   Data Fig. 6a). At the baseline level, the aortic rings isolated from
   db/db mice showed a lower basal endothelial-dependent relaxation
   response to acetylcholine than those from db/+ mice, indicating
   impaired endothelial function in the diabetic mice ([199]Fig. 7b).
   Simvastatin significantly improved vascular function by a complete
   restoration of acetylcholine-induced endothelial-dependent relaxation
   ([200]Fig. 7b, top), whereas GGTi298 exhibited a milder rescue effect
   ([201]Fig. 7b, bottom). Interestingly, we also observed a decreased
   basal contractile response to ET-1 in db/db mouse aortas, suggesting
   that endothelial-modified vascular smooth muscle cell function was also
   compromised ([202]Extended Data Fig. 6b). However, both simvastatin and
   GGTi298 effectively restored ET-1-induced vasoconstriction to the level
   observed in control mice ([203]Extended Data Fig. 6b). The protective
   role of simvastatin on vascular smooth muscle cell function might be
   attributed to its anti-inflammatory effects^[204]30. As neither
   simvastatin nor GGTi298 improved aortic vasoconstriction or
   vasodilation in wild-type (WT) control mice ([205]Extended Data Fig.
   6c), our data collectively suggest that suppressing the GGTase–RhoA–YAP
   axis represents a promising strategy to potently rescue
   diabetes-induced endothelial dysfunction in vivo.

Fig. 7 |. In vivo validation of the vasculoprotective role of simvastatin in
a diabetic mouse model.

   Fig. 7 |
   [206]Open in a new tab

   a, Experimental design schematic showing four groups of mice with
   different treatments. b, Concentration–response relationship for
   acetylcholine (Ach)-induced endothelial-dependent aortic relaxation in
   mice treated with simvastatin (top) and GGTi298 (bottom). Each point
   represents the mean relaxation response ±s.e.m. (n = 3 biological
   samples). c, PCA showing that simvastatin- and GGTi298-treated db/db
   mice were close to non-diabetic control (db/+) mice. d, KEGG pathway
   analysis showing GGTi298- and simvastatin-repressed genes that were
   upregulated in db/db aortas. AKT, serine–threonine kinase; MAPK,
   mitogen-activated protein kinase; NF-κB, nuclear factor κB; PI3K,
   phosphoinositide 3-kinase; TNF, tumor necrosis factor. e, Schematic
   diagram of different regions of the aorta receiving different blood
   flow shear stress patterns. f, Representative brightfield images of
   iPSC-ECs under static, disturbed flow, disturbed flow + simvastatin,
   laminar flow and laminar flow + simvastatin conditions after 72 h. g,
   Nuclear/cytoplasmic YAP ratios in iPSC-ECs under conditions of
   disturbed flow (n = 117 cells), disturbed flow + simvastatin (n = 117
   cells), laminar flow (n = 127 cells) and laminar flow + simvastatin (n
   = 127 cells). All data are presented as mean ± s.e.m. Two-way ANOVA
   corrected with the Bonferroni method (b). Unpaired two-sided Student’s
   t-test (g). *P < 0.05.

Simvastatin represses YAP activity and EndMT in diabetic mice

   To better understand how simvastatin and GGTi298 alleviated
   diabetes-induced endothelial dysfunction in diabetic mice, we performed
   RNA-seq on aortic tissues collected from the same region for wire
   myography but from another set of mice ([207]Fig. 7a). PCA analysis
   showed that simvastatin- and GGTi298-treated db/db aortas were
   clustered together and closer to db/+ control aortas but separated from
   vehicle-treated db/db aortas ([208]Fig. 7c). Next, we analyzed the DEGs
   in db/db aortas treated with vehicle, simvastatin and GGTi298. Compared
   to those in vehicle-treated db/db aortas, 1,174 DEGs were found to be
   significantly restored by both simvastatin and GGTi298 ([209]Extended
   Data Fig. 7a). We divided the 1,174 DEGs into diabetes-upregulated and
   -downregulated categories to perform further KEGG pathway analyses. We
   found that diabetes-upregulated genes were significantly enriched in
   the Hippo signaling- and inflammation-associated pathways ([210]Fig.
   7d), which were repressed by simvastatin and GGTi298 ([211]Extended
   Data Fig. 7b). By contrast, genes downregulated by diabetes linked to
   vascular functions, such as blood pressure, vasoconstriction and the
   oxidation–reduction process, were upregulated by simvastatin and
   GGTi298 ([212]Extended Data Fig. 7c). In line with our iPSC-EC data,
   simvastatin and GGTi298 rescued endothelial marker genes such as Klf2,
   Nr2f2 and Klf5 that were repressed in db/db aortas ([213]Extended Data
   Fig. 7d). Collectively, these data demonstrated that simvastatin and
   GGTi298 alleviated diabetes-induced endothelial dysfunction by
   repressing YAP-dependent pathways, such as EndMT.

   To examine whether differential shear flow patterns could affect YAP
   signaling along the aorta in the presence of simvastatin ([214]Fig.
   7e), we applied disturbed (aortic arch, atheroprone) versus laminar
   flow (descending aorta, atheroprotected) to iPSC-ECs to mimic this
   physiological feature ([215]Fig. 7e). Specifically, iPSC-ECs treated
   with vehicle or simvastatin were concurrently exposed to laminar (12
   dyne cm^−2) or oscillatory shear stress (±4 dyne cm^−2) for 72 h as
   previously reported^[216]31. Although iPSC-ECs aligned with the
   direction of laminar flow and showed chaotic alignment under disturbed
   flow ([217]Fig. 7f), simvastatin significantly reduced nuclear YAP
   signals under both flow patterns ([218]Extended Data Fig. 7e and
   [219]Fig. 7g). Moreover, simvastatin reduced a critical YAP downstream
   target, SOX9, in iPSC-ECs under both flow conditions ([220]Extended
   Data Fig. 8a,[221]b). Collectively, these data reaffirmed a causal role
   of simvastatin-repressed YAP signaling and EndMT in improved
   endothelial function.

Simvastatin suppresses YAP–TEAD activity via a SOX9 enhancer

   We further investigated whether YAP can directly interact with EndMT
   genes in iPSC-ECs. We first conducted ChIP–seq analysis on simvastatin-
   and GGTi298-treated iPSC-ECs and examined YAP-binding signals at
   regions ranging from −5 kb to +5 kb from the peak summit. Both
   simvastatin and GGTi298 reduced YAP-binding signals ([222]Fig. 8a).
   Consistent with previous studies^[223]32, most of the YAP-binding
   regions (73.5%) in our ChIP–seq data were located at enhancer regions
   ([224]Fig. 8b). Simvastatin- and GGTi298-treated iPSC-ECs showed a more
   similar YAP-binding pattern than the control group ([225]Fig. 8c). TEAD
   motifs were the most enriched motifs in simvastatin- and
   GGTi298-treated iPSC-ECs ([226]Fig. 8d,[227]e). Interestingly, SOX9, a
   major cardiac fibrosis and EndMT TF^[228]33,[229]34, was identified in
   our motif analysis ([230]Fig. 8e). Next, we identified two major
   YAP-binding chromatin regions, chromosome (chr)17:71,249,620–71,250,020
   and chr9:136,589,603–136,590,003, that were significantly downregulated
   upon simvastatin and GGTi298 treatment ([231]Fig. 8f,[232]g and
   [233]Extended Data Fig. 8c). Among the two chromatin regions,
   chr9:136,589,603–136,590,003 is located at the NOTCH1 enhancer region,
   whereas chr17:71,249,620–71,250,020 is part of the SOX9 enhancer region
   ([234]Fig. 8g). Notably, gene expression of both Sox9 and Notch1 was
   activated in db/db aortas compared to db/+ control aortas
   ([235]Extended Data Fig. 8d), suggesting that the two EndMT regulators
   may be the targets of statins in diabetic arteries.

Fig. 8 |. Inhibition of the mevalonate pathway decreases YAP–TEAD binding at
the SOX9 enhancer.

   Fig. 8 |
   [236]Open in a new tab

   a, Average YAP ChIP–seq enrichment scores among vehicle-, GGTi298- and
   simvastatin-treated iPSC-ECs (blue, vehicle control; orange, GGTi298;
   red, simvastatin). b, Annotation of ChIP–seq YAP-binding regions. TTS,
   transcription termination site. c, Unsupervised hierarchical clustering
   of vehicle-, GGTi298- and simvastatin-treated iPSC-ECs using the
   differentially enriched binding regions. d, ChIP–seq footprinting
   heatmap showing differential binding regions between iPSC-ECs treated
   with vehicle versus GGTi298 or simvastatin (FDR < 0.1). e, Overlapping
   YAP-binding motifs in iPSC-ECs treated with GGTi298 and simvastatin
   versus vehicle. f, Top decreased overlapping YAP-binding peaks in
   iPSC-ECs treated with GGTi298 and simvastatin versus vehicle. g, Genome
   Browser snapshots of the decreased binding loci (SOX9 enhancer) after
   simvastatin and GGTi298 treatment (highlighted in a yellow frame). KO,
   knockout. h, Immunostaining of endothelial markers, CD31 and
   VE-cadherin in SOX9 enhancer-knockout iPSC-ECs. i, Quantitative PCR
   analysis showing decreased levels of SOX9 in SOX9 enhancer-knockout
   cells (n = 3 biological samples). j, Quantitative PCR analysis of SOX9
   gene expression in SOX9 enhancer-knockout and WT iPSC-ECs under
   diabetic conditions (HG) (n = 3 biological samples). k, Tube formation
   changes in HG-treated SOX9 enhancer-knockout and WT iPSC-ECs (n = 3
   biological samples). All data are presented as mean ± s.e.m. Unpaired
   two-sided Student’s t-test (i–k).

   As Notch receptor 1 (NOTCH1) has binary roles in regulating endothelial
   functions by promoting angiogenesis and positively regulating the EndMT
   process during both physiological and pathological conditions^[237]35,
   we decided to focus on understanding the role of the previously
   undescribed SOX9 enhancer in simvastatin-mediated endothelial
   protective effects. We knocked out the SOX9 enhancer region using
   CRISPR–Cas9 in one iPSC line. The single-guide RNA (sgRNA) species were
   designed to target the upstream and downstream regions of
   chr17:71,249,620–71,250,020 ([238]Supplementary Fig. 3a), and a 540-bp
   fragment of the SOX9 enhancer was successfully excised
   ([239]Supplementary Fig. 3b,[240]c). We confirmed that knocking out the
   SOX9 enhancer in iPSCs did not compromise endothelial differentiation,
   as shown by high levels of EC marker genes ([241]Fig. 8h). Compared to
   parental control (WT) iPSC-ECs, SOX9 enhancer-knockout iPSC-ECs showed
   significantly downregulated SOX9 expression ([242]Fig. 8i) and did not
   respond to HG-induced upregulation of SOX9 nor to simvastatin-induced
   downregulation of SOX9 ([243]Fig. 8j). SOX9 enhancer-knockout iPSC-ECs
   also failed to respond to HG-induced upregulation and
   simvastatin-induced downregulation of an EndMT master gene, SNAI2
   ([244]Supplementary Fig. 3d). By contrast, KLF2 expression was not
   compromised in SOX9 enhancer-knockout iPSC-ECs ([245]Supplementary Fig.
   3e). Finally, simvastatin failed to restore diabetes-induced
   angiogenesis defects in SOX9 enhancer-knockout iPSC-ECs ([246]Fig. 8k).
   Overall, our data suggest that simvastatin improved endothelial
   function by epigenetically blocking YAP–TEAD-mediated upregulation of
   chromatin accessibility at EndMT-regulating genes by repressing the
   GGTase–RhoA–YAP–SOX9 signaling axis ([247]Extended Data Fig. 9).

Discussion

   Endothelial dysfunction contributes to CVD-associated morbidity and
   mortality. In this study, we show in iPSC-ECs that YAP can translocate
   to the nucleus, where it binds TEAD to form a functional transcription
   complex that epigenetically upregulates genes associated with EndMT
   ([248]Extended Data Fig. 9). While sustained hyperglycemia can further
   activate YAP and EndMT to aggravate endothelial dysfunction, we showed
   that simvastatin effectively rescued this pathology by blocking the
   GGTase–RhoA–YAP axis.

   Cardiac-specific deletion of YAP decreased cardiomyocyte proliferation
   and led to myocardial hypoplasia, whereas endothelial-specific deletion
   of YAP compromised vascularization^[249]36,[250]37. However, YAP
   activation in adult tissues is generally believed to be pathogenic, as
   it was detected in mouse atherosclerotic plaques^[251]31, diabetic
   kidneys^[252]38 and diabetic cardiomyopathic hearts^[253]39. The
   YAP–TEAD complex has been reported to influence downstream gene
   transcription activity through regulating chromatin alterations at
   target loci^[254]40. Because ECs express higher levels of YAP than
   other cell types in the aorta^[255]41, endothelial YAP may maintain
   higher chromatin accessibility at target genes even under physiological
   conditions. Indeed, we observed that simvastatin improved endothelial
   functions not only under diabetic conditions but also at the baseline
   level. Because activated YAP can upregulate EndMT-regulating genes, it
   explains why ECs are prone to transitioning to a mesenchymal phenotype
   both in vitro and in vivo.

   Several studies showed that YAP activity was upregulated in HG
   conditions and T2D^[256]39,[257]42. We cross-validated the
   effectiveness of modulating simvastatin-mediated epigenetic repression
   of YAP signaling and EndMT in rescuing diabetes-induced endothelial
   dysfunction using simvastatin, GGTi298 and YAP CRISPRi. Additionally,
   we identified two previously undescribed enhancer regions (NOTCH1 and
   SOX9) as the direct binding sites of YAP. Using CRISPR technology, we
   further confirmed that the SOX9 enhancer is essential for the
   simvastatin-repressed EndMT process. Because our ChIP–seq data suggest
   that SOX9 motifs are also YAP targets, activated SOX9 may
   synergistically regulate downstream EndMT genes such as TWIST1, TWIST2,
   SNAI1 and SNAI2 with YAP^[258]34. EndMT has been recognized as
   therapeutically promising target for treating vascular dysfunction, and
   major EndMT drivers (for example, those encoded by SNAI1, SNAI2,
   TWIST1/2, ZEB1 and ZEB2) are controlled by epigenetic
   regulators^[259]43. Thus, our study provides an alternative strategy to
   modulate EndMT by epigenetically repressing the GGTase–RhoA–YAP–SOX9
   signaling axis.

   Human iPSC-ECs offer tremendous potential for revealing pathogenic
   mechanisms and assessing drug responses in a patient-specific manner.
   However, several technical limitations, as outlined below, need to be
   addressed in future research to make this in vitro model a more
   effective platform for disease modeling and drug testing. First,
   iPSC-derived cells are less mature than their primary counterparts,
   which may lead to incomplete recapitulation of disease phenotypes
   observed in patients^[260]44. Second, the reprogramming process of
   iPSCs inevitably eliminates disease phenotype-associated epigenetic
   alterations, making it challenging to investigate patient-specific
   disease phenotypes and drug responses caused by complex interactions of
   multiple genetic mutations and environmental risk factors^[261]45.
   Nevertheless, with the advance of bioengineering tools in promoting
   cell maturation and the emergence of multiomic studies focusing on
   identification of diabetes-associated variants in large-scale
   populations^[262]46, iPSC-ECs derived from diabetic patients will be a
   useful platform to decode the effects of patient-specific epigenetic
   changes on endothelial dysfunction.

   In summary, our findings uncover a previously unidentified interaction
   between statin-mediated chromatin remodeling and improved endothelial
   function via repressing the GGTase–RhoA–YAP–SOX9 signaling axis. The
   causal role of YAP activation in EndMT and vascular dysfunction could
   lead to the development of alternative biomarkers for CVD diagnosis and
   therapeutic strategies to prevent or treat endothelial
   dysfunction-associated CVDs.

Methods

Human iPSC reprogramming and culture

   Human iPSCs were generated from peripheral blood mononuclear cells
   collected from healthy donors using the CytoTune-iPS 2.0 Sendai
   reprogramming kit (Thermo Fisher Scientific). In brief, peripheral
   blood mononuclear cells were isolated from 20 ml of whole blood using
   Ficoll-Paque Premium (GE HealthCare) and cultured for 9 d in blood
   medium. A total of 1 × 10^6 cells were transfected with the Sendai
   reprogramming virus overnight. Cells were then transferred to a mouse
   embryonic fibroblast-coated plate and cultured in mTeSR medium
   (Stemcell Technologies). Human embryonic stem cell-like colonies were
   picked after day 20, and iPSC clones were transferred to
   Matrigel-coated plates (BD Biosciences) for expansion. Human iPSCs were
   passaged using Accutase (Global Cell Solutions) and 10 μM Y-27632
   (Selleckchem) and then refreshed with daily changes of mTeSR medium.
   Two clones of iPSCs from each donor were maintained and passaged at the
   same time. The maintenance medium for iPSCs was switched to chemically
   defined E8 medium (Gibco) after passage 3. The E8 medium was changed
   daily, and iPSCs were passaged when they reached >80% confluency.
   Aliquoted iPSCs were cryopreserved between passages 15 and 30 for
   downstream experiments. For human iPSC reprogramming, written consents
   were obtained from these individuals, and the studies were approved by
   the Stanford Administrative Panels on Human Subjects Research.

EC differentiation

   Endothelial differentiation was carried out in RPMI medium with B27
   supplement and no insulin (Gibco) when iPSCs (at passages 15–30)
   reached 80–90% confluency. In brief, iPSCs were treated with 6 μM
   CHIR99021 (Selleckchem) for 2 d and then 2 μM CHIR99021 for another 2
   d. Starting from day 4, the differentiation medium was changed to
   Endothelial Cell Growth Medium-2 (EGM-2) medium (Lonza) supplemented
   with VEGF (50 ng ml^−1, PeproTech), basic fibroblast growth factor (20
   ng ml^−1, PeproTech) and the TGF-β inhibitor SB431542 (10 μM,
   Selleckchem). On day 12, differentiated iPSC-ECs were enriched using
   CD144 microbeads (Miltenyi Biotec). Sorted iPSC-ECs were cultured in
   gelatin-coated plates using EGM-2 medium. Confluent iPSC-ECs were
   passaged using TrypLE (Gibco) and cryopreserved at passage 2 after
   derivation using Bambanker (Wako Chemicals). To account for technical
   variations (for example, culture medium, coating matrix and growth
   factors) on transcriptome profile changes, two clones of each iPSC line
   were differentiated into ECs in two independent batches. Transcriptomic
   analysis and functional experiments were conducted using recovered
   iPSC-ECs.

Drug and compound treatment

   iPSC-ECs were seeded at 30,000 cells per well in six-well plates or
   1,500 cells per well in 96-well plates depending on which experiment
   was carried out. The following drugs and compounds were used in this
   study: simvastatin (0.1–10 μM; Sigma-Aldrich, S6196), lovastatin
   (0.1–10 μM; Sigma-Aldrich, PHR1285), atorvastatin (0.1–10 μM;
   Sigma-Aldrich, PZ0001), rosuvastatin (0.1–10 μM; Sigma-Aldrich,
   SML1264), mevastatin (0.1–10 μM; Sigma-Aldrich, M2537), fluvastatin
   (0.1–10 μM; Sigma-Aldrich, SML0038), pravastatin (0.1–10 μM;
   Sigma-Aldrich, P4498), GGTi298 (1 μM; Sigma-Aldrich, G5169), (±)-MA
   5-phosphate (0.5 mM, 1 μM; Sigma-Aldrich, 79849), GGPP (20 μM;
   Sigma-Aldrich, G6025), squalene (20 μM; Sigma-Aldrich, S3626) and RhoAi
   C3 (2 μg ml^−1; Cytoskeleton, CT-04). Unless otherwise specified in the
   figure legends, iPSC-ECs were treated with drugs for 72 h.

Quantitative real-time PCR (rtPCR)

   Total RNA from iPSC-ECs was isolated using the RNeasy Plus Mini Kit
   (Qiagen), and reverse transcription was performed using the
   High-Capacity cDNA Reverse Transcription Kit (Life Technologies).
   Quantitative rtPCR reactions were carried out using TaqMan Universal
   PCR Master Mix (Applied Biosystems) according to the manufacturer’s
   protocol in a CFX96 real-time PCR detection system (Bio-Rad). Relative
   mRNA levels were normalized against glyceraldehyde 3-phosphate
   dehydrogenase (GAPDH) in each reaction. TaqMan primers and probes were
   ordered from Thermo Fisher Scientific (NOS3, Hs01574659_m1; KLF2,
   Hs00360439_g1; VCAM1, Hs01003372_m1; SNAI2, Hs00161904_m1; TWIST1,
   Hs01675818_s1; and SOX9, Hs00165814_m1). More than three biological
   replicates per group were performed.

In vitro model of hyperglycemia

   iPSC-ECs were seeded at 30,000 cells per well in six-well plates or
   1,500 cells per well in 96-well plates according to different
   experimental settings. To mimic the diabetic condition in vitro, cells
   were maintained in EGM-2 medium supplemented with 25 mM d-glucose
   (equivalent to 4.5 mg ml^−1 blood glucose in patients), 20 nM insulin
   and 10 nM ET-1 for 7 d with daily medium changes. Simvastatin, GGTi298
   and RhoAi were introduced to the diabetic condition for 72 h to rescue
   endothelial dysfunction. Mannitol (25 mM) was used in the vehicle
   control group to normalize osmolality.

Immunofluorescence staining and quantification

   iPSC-ECs were seeded onto gelatin-coated eight-chamber glass slides
   (Nunc) and maintained in EGM-2 medium. Cells were fixed using 4% PFA
   after 2 d of culture and then permeabilized with 0.1% Triton X-100
   (Sigma-Aldrich) and blocked with 2.5% donkey serum (Sigma-Aldrich) in
   PBS. Cells were then incubated with primary antibodies (anti-CD31
   (PECAM), R&D Systems; anti-VE-cadherin, R&D Systems; anti-YAP, Novus
   Biologicals; YAP, Cell Signaling Technology) in 2.5% donkey serum–PBS
   overnight at 4 °C. After three washes with PBS, cells were further
   incubated with secondary antibodies (donkey anti-mouse Alexa Fluor 488;
   donkey anti-rabbit Alexa Fluor 594, Life Technologies) at room
   temperature for 1 h. Next, slides were washed with PBS and mounted with
   SlowFade Gold Antifade Reagent with 4,6-diamidino-2-phenylindole (DAPI;
   Life Technologies) before imaging. All immunofluorescent images were
   captured using a Nikon Eclipse 80i fluorescence microscope and analyzed
   using ImageJ. The intensity of YAP-positive signals in the nuclear
   versus cytoplasmic fractions was analyzed using CellProfiler software.

Confluent and sparse iPSC-EC culture

   Five times more iPSC-ECs were seeded in the confluent condition (5,000
   cells per well) compared to the sparse condition (1,000 cells per well)
   in 96-well plates, and simvastatin was administered to both conditions
   at the same time. Cells were subjected to immunofluorescence staining
   and western blot analysis for YAP signaling.

Cell viability assay

   Differentiated iPSC-ECs were seeded on gelatin-coated 96-well plates at
   10,000 cells per well. Cells were treated with simvastatin at 0 μM, 0.1
   μM, 0.3 μM, 1 μM, 3 μM and 10 μM for 48 h. Cell viability after
   simvastatin treatment was evaluated using a CellTiter-Glo viability
   assay kit (Promega) according to the manufacturer’s procedures.
   Briefly, a volume of CellTiter-Glo reagent equal to the volume of
   culture medium was added to each well. The cells and reagents were
   mixed for 2 min on an orbital shaker. Next, the cells were lysed for
   another 10 min to stabilize the luminescent signal. Raw readings were
   recorded using a Cytation 5 plate reader or imager (BioTek
   Instruments).

Cell proliferation measurement

   The cell-proliferation assay was performed by tracking cell number
   changes over the course of simvastatin treatment for 3 d. In brief,
   cells were seeded in the same condition, and replicates were fixed
   every day during the 3-d treatment with simvastatin and dimethyl
   sulfoxide (DMSO). All fixed cells were stained with DAPI and counted to
   assess simvastatin’s effect on cell proliferation.

ROS measurement

   iPSC-ECs in different treatment groups were maintained in 96-well
   white-walled plates with clear bottoms (Falcon). ROS-Glo H[2]O[2]
   substrate solution (Promega) was prepared according to the
   manufacturer’s instructions. H[2]O[2] substrate solution (20 μl) was
   added to each well 2 h before the end of each treatment. The final
   volume in each well (H[2]O[2] substrate solution plus culture medium)
   was 100 μl, and plates were incubated at 37 °C in an incubator with 5%
   CO[2]. Next, 50 μl of supernatant was combined with 50 μl of ROS-Glo
   H[2]O[2] Detection solution in a separate opaque white plate. After 20
   min of incubation at room temperature, the relative luminescence units
   were measured using the GloMax-Multi detection plate reader (Promega).

In vitro tube formation assay

   iPSC-ECs were seeded at 40,000 cells per well in a Matrigel (growth
   factor reduced; Corning)-coated 24-well plate. Cells in each treatment
   group were further cultured in EGM-2 medium with 50 ng ml^−1 VEGF
   overnight at 37 °C. Images were taken, and tubular knots were
   quantified to compare the in vitro angiogenesis capacity of iPSC-ECs
   with different treatments.

NO measurement

   NO production in iPSC-ECs was measured with the Nitrate/Nitrite
   Fluorometric Assay Kit (Cayman Chemical). In brief, iPSC-ECs from
   different donors were seeded at the same density in gelatin-coated
   six-well plates and cultured for 3 d until the cells reached over 80%
   confluency. Next, cells were treated with 10 μM acetylcholine chloride
   (Sigma) overnight, and the supernatant was collected and stored at −80
   °C. The thawed supernatant was centrifuged to remove cell debris before
   adding nitrate reductase to convert nitrate to nitrite. Total nitrite
   was measured using a Cytation 5 Cell Imaging Multimode Reader (BioTek).
   The total NO concentration for each sample was calculated using a
   standard curve measured at the same time with samples.

YAP CRISPRi construct, virus production and transfection

   The sgRNA sequence targeting the YAP promoter was ligated into the
   pSLQ1373 vector with BstXI and Xhol (YAP CRISPRi sgRNAs:
   GCCGCCGCCAGGGAAAAGAA, GGCCTTCGCAGCCCCCGCAG). Lentivirus was produced
   using pSLQ1373 containing YAP sgRNA and packaging protein vectors
   (pMD2.G and pCMV-dR8.91). In brief, HEK293T cells were transfected with
   these three plasmids using TransIT-LT1 transfection reagents (Mirus).
   Lentiviral supernatants were collected after 48 h of transfection and
   used for transduction of an iPSC line containing a Tet-On 3G system
   driving dead Cas9 (dCas9)–Krüppel-associated box (KRAB)
   expression^[263]47. For YAP CRISPRi experiments, the puromycin-purified
   YAP CRISPRi iPSC line was differentiated into iPSC-ECs, and doxycycline
   was added into the iPSC-EC culture medium to turn on dCas9–KRAB
   expression. Transcription-repression efficiency was measured by
   quantifying the mRNA level of YAP using quantitative PCR.

Generation of a YAP–TEAD reporter line

   The YAP–TEAD reporter plasmid (pLL3.7 FLAG-YAP1-TEAD-P-H2B-mCherry) was
   purchased from Addgene (plasmid 128327). The lentiviral YAP–TEAD
   reporter was packaged into HEK293T cells using packaging vectors
   (pLL3.7 FLAG-YAP1-TEAD-P-H2B-mCherry, pMD2.G and pCMV-dR8.91), which
   were transduced into iPSCs (D1 iPSC line) using 5 μg ml^−1 polybrene
   (Sigma-Aldrich) overnight. The iPSCs carrying the YAP–TEAD reporter
   were purified using GFP selection and expanded for iPSC-EC
   differentiation for experiments. Similarly, a TEAD response
   element-driven luciferase reporter (BPS Bioscience) was transduced into
   iPSC-ECs as a lentivirus using 5 μg ml^−1 polybrene overnight. Next, an
   equal number of transfected iPSC-ECs was seeded into 96-well plates
   (5,000 cells per well), followed by treatment with vehicle control
   (DMSO) or simvastatin. Cells were collected for TEAD
   luciferase-activity analysis at 12 h, 24 h and 72 h after treatment
   using a ONE-Step Luciferase Assay (BPS Bioscience).

Laminar flow and disturbed flow experiments

   To mimic the effects of disturbed (aortic arch, atheroprone) versus
   laminar (descending aorta, atheroprotected) flow at different aortic
   regions on YAP signaling in the presence of simvastatin, iPSC-ECs were
   cultured under ibidi fluidic units (ibidi) according to the
   manufacturer’s instructions. In brief, 1 × 10^5 iPSC-ECs were plated on
   each μ-Slide I 0.4 Luer slide, which was attached in series to each
   fluidic unit with pumping EGM-2 medium. iPSC-ECs treated with vehicle
   or simvastatin were concurrently exposed to laminar shear stress (12
   dyne cm^−2) and oscillatory shear stress (disturbed, ±4 dyne cm^−2) for
   72 h. Slides were then collected for immunostaining and western blot
   analysis at the end of the experiment.

CRISPR knockout experiments

   SOX9 enhancer-knockout sgRNAs were designed using a Synthego design
   tool and synthesized by Synthego. For electroporation, 3.5 × 10^5 iPSCs
   were transfected with 1 μl Cas9 protein (Integrated DNA Technologies,
   20 μM) and 1 μl sgRNA (100 μM) in Resuspension Buffer R (Thermo Fisher)
   in a final volume of 10 μl. Cells were transfected with the Neon
   Transfection System (Thermo Fisher) following the electroporation
   procedure of 200 V, 20 ms and two pulses and then resuspended in E8
   medium supplemented with 10 μM Y-27632 for 24 h. The SOX9 enhancer
   primers and OneTaq DNA Polymerase (NEB) were used to amplify the
   targeted loci following the manufacturer’s instructions. After PCR
   amplification, electrophoresis on a 2% agarose gel was used to
   characterize the PCR products. The QIAquick PCR Purification Kit
   (Qiagen) was used to purify the PCR products, which were then examined
   using Sanger sequencing (sgRNA: top, AUUUUCCCACAGAGGUAAUG; bottom,
   AUUGCCGCAAAGGAGCUUUG; Genomic PCR primers: ACGTCTCTTGCTTCCCAAACA,
   TGGTTTGAGTTTGTTTTCTTCTTCA).

Subcellular fractionation and western blot analysis

   Cells were lysed on ice using RIPA buffer (Thermo Fisher) containing a
   protease inhibitor and phosphatase inhibitor cocktail (Sigma).
   Supernatants were collected by centrifugation at 14,000g for 20 min at
   4 °C. The Subcellular Protein Fractionation Kit for Cultured Cells
   (Thermo Scientific) was used to separate cytoplasmic and nuclear
   proteins. Protein concentrations were quantified using a Pierce^™ BCA
   protein assay kit (Thermo Scientific). Equal quantities of protein
   lysates (25 μg) were loaded on NuPAGE Bis-Tris Gels (Life
   Technologies), and resolved bands were transferred to PVDF membranes
   using the Trans-Blot Turbo Transfer System (Bio-Rad). PVDF membranes
   were incubated with primary antibodies (anti-YAP, 1:1,000, Novus
   Biologicals; anti-phospho-YAP, 1:1,000, Cell Signaling Technology;
   anti-lamin A/C, 1:500, Santa Cruz; anti-SOX9, 1:1,000, Cell Signaling;
   anti-CTGF, 1:1,000, Thermo Scientific; anti-histone H3, 1:1,000, Cell
   Signaling Technology) overnight at 4 °C, followed by incubation with
   HRP-conjugated secondary antibodies (1:5,000) for 1 h at room
   temperature and then detected using the ChemiDoc XRS+ System (Bio-Rad).
   An HRP-conjugated GAPDH antibody (Life Technologies) was used as a
   loading control for normalization.

Animals

   Six-week-old male C57BL/KsJ-db/db (db/db), non-diabetic (db/+) and WT
   mice (Jackson Laboratory) were used for experiments. No statistical
   methods were used to pre-determine sample size. Because men have a
   higher prevalence of T2D and benefit more from statin therapy in the
   management of cardiovascular risks than do women, to reduce
   variability, we used male mice and similar sample sizes based on our
   previous vascular animal studies^[264]48. This study was approved by
   the Stanford Administrative Panel on Laboratory Animal Care under
   protocol APLAC 33803, and the mice were housed in
   temperature-controlled environments. The drugs were dissolved in saline
   and administered through oral gavage once a day for 8 weeks.
   Specifically, five non-diabetic control (db/+) mice received saline as
   a control group, five db/db mice received saline as a vehicle control,
   five db/db mice received simvastatin treatment (25 mg per kg), and five
   db/db mice received GGTi298 treatment (5 mg per kg). Descending
   thoracic aortas from all groups were collected for a wire myography
   study and RNA-seq analysis at the end of treatment. We further included
   WT (C57BL/6) mice to test the effect of vehicle, simvastatin and
   GGTi298 on vascular function using wire myography.

Isometric tension studies on mouse descending thoracic aortas

   After 8 weeks of statin and GGTi298 treatment, the aortas from mice of
   all groups were collected, cleaned of excess connective tissue and fat,
   cut into ~2-mm rings and mounted on two parallel tissue supports in the
   vessel incubation chamber filled with KHB buffer oxygenated
   continuously with 95% O[2]–5% CO[2] at 37 °C (KHB buffer composition in
   mM: 131.5 NaCl, 5 KCl, 1.2 NaH[2]PO[4], 2.5 MgCl[2], 2.5 CaCl[2]; 2 g
   glucose and 2 g NaHCO[3] in 1 L with pH 7.4). Aortas were subjected to
   isometric tension studies, and the developed force traces were recorded
   by LabChart software (ADInstruments). After two pre-constriction
   treatments with the vasoconstrictor PGF2-α (1 μM), the aortic rings
   were equilibrated for 30 min under a resting optimal tension of 10 mN
   with the chamber buffer changed every 10 min. The optimal resting
   tension was verified in each set of experiments before generating
   concentration–response curves. ET-1 (0.1 nM to 1 μM) was added to
   induce a contractile response. At the plateau of vasocontraction,
   acetylcholine (1 nM to 1 μM) was added accumulatively to induce
   vasorelaxation. Relaxation responses were calculated as the percent
   reduction in tension from the pre-contracted state. Contractile
   responses were calculated as the force generated by aortic rings
   calibrated by the dry weight with units of mN mg^−1. Control responses
   were measured by adding the same amount of vehicle (water).

Transcriptomic analysis of AmpliSeq and RNA-seq data

   RNA was extracted from frozen iPSCs, iPSC-ECs and mouse aortas using
   the Qiagen RNeasy kit according to the manufacturer’s instructions. For
   AmpliSeq, 10 ng of RNA was used for cDNA synthesis with the SuperScript
   VILO cDNA Synthesis Kit (Thermo Fisher Scientific). The barcoded
   libraries were prepared with the Ion AmpliSeq Transcriptome Human Gene
   Expression Kit (Thermo Fisher Scientific) according to the
   manufacturer’s instructions. Combined libraries (50 pM) were loaded on
   Ion PI v3 chips using the Ion Chef automated platform (Thermo Fisher
   Scientific). AmpliSeq sequencing data were analyzed using the Ion
   Torrent Mapping Alignment Program (TMAP) as described in a previous
   study. To achieve both specificity and sensitivity, TMAP implemented a
   two-stage mapping approach with four alignment algorithms, BWA-short
   and long, SSAHA and Super-maximal Exact Matching, which were followed
   by the Smith–Waterman algorithm to find the final best mapping. RNA-seq
   analysis was conducted through Novogene. DEG analysis was performed
   with the R package DESeq2, which is available from Bioconductor. Genes
   with false discovery rate (FDR) or adjusted P value <0.05 are
   identified as DEGs. To identify potential function-enriched pathways,
   DEGs were analyzed for biological functions and pathway enrichment
   using the KEGG Orthology-Based Annotation System and gene set
   enrichment analysis.

ATAC-seq and ChIP–seq data collection and analysis

   For ATAC-seq samples, 50,000 iPSC-ECs treated with simvastatin or
   vehicle control were collected. The transposition reaction was
   conducted using Nextera Tn5 Transposase. Transposed DNA was purified
   and amplified according to Buenrostro’s protocol^[265]49. For ChIP–seq
   samples, vehicle-, simvastatin- and GGIT298-treated iPSC-ECs were
   collected to perform chromatin immunoprecipitation using an anti-YAP
   antibody (Novus Biologicals) with the Zymo-Spin ChIP Kit (Zymo
   Research). Sheared chromatin was cross-linked (a small portion was
   saved as the ChIP–seq input sample) and incubated with anti-YAP
   antibody-conjugated Dynabeads overnight in a cold room for
   immunoprecipitation. Immunoprecipitated DNA was eluted, de-cross-linked
   and purified following the instructions of the Zymo-Spin ChIP Kit to
   prepare ChIP–seq libraries. ATAC-seq sequencing reads in FASTQ format
   were controlled for quality and trimmed with trim_galore in paired-end
   mode. The trimmed reads were aligned to the human genome (hg38) using
   Burrows–Wheeler Aligner version 0.7.17 (BWA-MEM). Duplicate reads were
   subsequently removed with Picard MarkDuplicates. Peak calling for each
   ATAC-seq replicate was performed using macs2 (version 2.2.7.1) callpeak
   with the options ‘–nomodel –shift −37 –extsize 73 –SPMR’. The quality
   of ATAC-seq peaks was assessed with the R package ATACseqQC. For
   ChIP–seq, peak calling was performed using DFilter with the parameters
   ‘-ks=10 -bs=50 -lpval=6 -wig -nonzero -refine’. The signal was
   converted into bigWig format, and visual inspection of peak quality was
   performed using the WashU Epigenome Browser. To identify differential
   accessible peaks, the R package DiffBind was employed using the
   de-duplicated alignment file in BAM format and peaks called by macs2 as
   inputs. The consensus peak was determined by including peaks that were
   called in more than two of the replicates. Affinity binding matrices
   were computed with DiffBind. Differential accessibility analyses were
   carried out using DESeq2. Genomic annotation of peaks and differential
   motif analyses were performed with Homer2. Genomic cis regulatory
   functional prediction analyses of differential peaks were performed
   using GREAT with basal plus extension association rules.

Statistical analysis

   Statistical analyses were performed using Student’s t-test or one-way
   analysis of variance (ANOVA) corrected with the Bonferroni method.
   Multiple-group comparisons were conducted using two-way ANOVA corrected
   with the Bonferroni method. Sequencing data analysis was conducted with
   the Wald test and corrected for multiple testing using the Benjamini
   and Hochberg method. All error bars are defined as s.e.m. unless
   otherwise indicated. Statistical analysis was conducted on data from
   three or more biologically independent experimental replicates; western
   blot experiments were performed with two biological replicates.

Extended Data

Extended Data Fig. 1 |. Statins improve endothelial function and alter
epigenetic associated genes in iPSC-ECs.

   Extended Data Fig. 1 |
   [266]Open in a new tab

   a, NOS3 expression levels in iPSC-ECs after being treated with seven
   statins at different concentrations (0.1 μM, 1 μM, and 10 μM). Data are
   normalized to that of the vehicle control group; prava, pravastatin;
   atorva, atorvastatin; rosuva, rosuvastatin; meva, mevastatin; fluva,
   fluvastatin; lova, lovastatin; and simva, simvastatin (n = 6 biological
   samples). b, Simvastatin decreases the viability of iPSC-ECs at 10 μM
   (n = 3 biological samples). c, Simvastatin increases ROS production in
   iPSC-ECs at 10 μM (n = 3 biological samples). d, Representative images
   showing higher densities of capillary-like networks formed by iPSC-ECs
   from three healthy donors treated with simvastatin versus vehicle.
   Scale bars, 250 μm. e, Differentially expressed genes (DEGs) in
   simvastatin- versus vehicle-treated iPSC-ECs (FDR < 0.05). f,
   Simvastatin had no effect on iPSC-EC proliferation compared to vehicle
   (DMSO) (n = 4 biological samples). g, Enrichment (cellular components)
   analysis of upregulated DEGs shown in e. h, Enrichment (cellular
   components) analysis of downregulated DEGs shown in (e). All data are
   presented as mean ± SEM. Unpaired two-sided Student’s t-test (a,b,c).

Extended Data Fig. 2 |. Simvastatin alters chromatin accessibility in
iPSC-ECs.

   Extended Data Fig. 2 |
   [267]Open in a new tab

   a, Annotation of ATAC peaks with differential chromatin accessibility.
   b and c, Motif enrichment analysis at the ATAC-seq sites with
   differential chromatin accessibility at enhancer regions. d, KEGG
   pathway analysis of the ATAC-seq peaks with KLF motifs. e and f, GO
   pathway analysis of ATAC-seq peaks with KLF motifs and TEAD motifs.

Extended Data Fig. 3 |. RNA-seq and ATAC-seq analyses of iPSC-ECs treated
with simvastatin at different timepoints.

   Extended Data Fig. 3 |
   [268]Open in a new tab

   a, Heatmap of RNA-seq showing changes in transcriptomic patterns of
   iPSC-ECs after being treated with simvastatin at 0 h, 12 h, 24 h, and
   72 h. b, Normalized expression levels of endothelial marker genes
   (KLF4, CDH5, and KDR) from RNA-seq of simvastatin-treated iPSC-ECs at
   different timepoints. c, Normalized expression levels of YAP downstream
   genes (SMAD3, CTGF, GLI2, MCL1, RUNX1, BIRC2, TGFB2, and BIRC5) from
   RNA-seq of simvastatin-treated iPSC-ECs at different timepoints. d,
   Normalized expression levels of mesenchymal genes (TGFBR1, TWIST1,
   SNAI2, and SOX9) from RNA-seq of simvastatin-treated iPSC-ECs at
   different timepoints. All data are presented as mean ± SEM. n = 2
   RNA-seq biological samples.

Extended Data Fig. 4 |. Simvastatin inhibits YAP and TEAD activity in
iPSC-ECs.

   Extended Data Fig. 4 |
   [269]Open in a new tab

   a, Comparative analysis showing that simvastatin exposure to iPSC-ECs
   for 12 h, 24 h, and 72 h induces highly correlated changes in DEGs and
   ATAC signal alterations (values indicate log[2](fold change)). b,
   Immunofluorescence of YAP subcellular localization in iPSC-ECs treated
   with simvastatin versus vehicle control for 24 h. c, Quantification of
   nuclear/cytoplasmic YAP ratios in iPSC-ECs treated with simvastatin
   versus vehicle control for 24 h (n = 14 cells). d, iPSC-ECs treated
   with simvastatin (n = 101 cells) but not vehicle (n = 138 cells) for 24
   h showing significantly decreased nuclear TEAD activity. e,
   Representative images of capillary-like tubular networks formed by
   iPSC-ECs treated with simva, simva + MA, simva + GGPP, simva +
   squalene, GGTi298, RhoAi, and vehicle control. All data are presented
   as mean ± SEM. Unpaired two-sided Student’s t-test (c,d).

Extended Data Fig. 5 |. Simvastatin reverses hyperglycemia (HG)-induced
endothelial dysfunction by repressing YAP-mediated EndMT process.

   Extended Data Fig. 5 |
   [270]Open in a new tab

   a, ROS levels in iPSC-ECs treated with HG, HG + simvastatin, HG +
   GGTi298, HG + RhoAi, and vehicle control (n = 3 biological samples). b,
   Gene set enrichment analysis (GSEA) of the differentially regulated
   genes in HG versus vehicle control (top panel) and HG versus HG +
   simvastatin (bottom panel) reveals that HG-upregulated EndMT process is
   blunted by simvastatin. c, Heatmaps of
   epithelial-mesenchymal-transition gene sets in HG versus vehicle
   control (left) and HG versus HG + simvastatin (right) based on GSEA
   analysis. Representative western blot data (d) and densitometric
   quantification (e) showing the hyperglycemic condition (HG) decreases
   phosphorylated/total YAP in iPSC-ECs compared to the control condition.
   GAPDH serves as a loading control (n = 3 biological samples). All data
   are presented as mean ± SEM. Unpaired two-sided Student’s t-test (a).

Extended Data Fig. 6 |. In vivo validation of the endothelial protective
effects of simvastatin in a diabetic mouse model.

   Extended Data Fig. 6 |
   [271]Open in a new tab

   a, A representative trace of isometric tension in the mouse aorta. The
   aortic rings were equilibrated for 30 min under a resting tension of 10
   mN after two sessions of pre-constriction with the vasoconstrictor
   PGF2α (1 μM). For a vasoconstriction response, endothelin-1 (ET-1, 0.1
   nM to 1 μM) was used to induce a contractile response. At the plateau
   of maximal contraction, acetylcholine (Ach, 1 nM to 1 mM) was added
   accumulatively to initiate relaxation. b, Concentration-response
   relationship for ET-1-induced aortic constriction in mice treated with
   simvastatin (left panel) and GGTi298 (right panel). Developed tension
   was the force generated by aortic rings normalized to aortic tissue dry
   weight (mN/mg). Each point represents the mean developed contractile
   force ± SEM (n = 4 biological samples). c, Concentration-response
   relationship for ET-1-induced aortic constriction in wildtype mice
   treated with vehicle (saline), simvastatin (25 mg/kg), or GGTi298 (5
   mg/kg) for 8 weeks (left panel). Concentration-response relationship
   for Ach-induced endothelial-dependent aortic relaxation in wildtype
   mice treated with vehicle (saline), simvastatin (25 mg/kg), or GGTi298
   (5 mg/kg) for 8 weeks (right panel). Each point represents the mean
   constriction/relaxation response ± SEM (n = 4 biological samples). All
   data are presented as mean ± SEM.

Extended Data Fig. 7 |. RNA-seq analysis shows potent rescue effects of both
simvastatin and GGTi298 on diabetes-induced vascular dysfunction in mice.

   Extended Data Fig. 7 |
   [272]Open in a new tab

   a, Venn diagram showing overlapped DEGs of db/+ (vehicle), db/db +
   simvastatin, and db/db + GGTi298 compared to the db/db + vehicle group,
   respectively. b, EndMT-associated genes, such as Ctgf, Vcam1, Tgfb2,
   and Smad3, were rescued by simvastatin and GGTi298 in diabetic mouse
   aortas (n = 2 RNA-seq biological samples). c, KEGG pathway analysis
   showing both simvastatin and GGTi298 improved genes associated with
   endothelial functions that were downregulated in db/db mouse aortas. d,
   Endothelial marker genes, such as Klf2, Nr2f2, and Klf5, were restored
   by simvastatin (simva) and GGTi298 in db/db mouse aortas (n = 2 RNA-seq
   biological samples). e, Representative immunofluorescent images showing
   YAP nuclear localization patterns in iPSC-ECs treated with disturbed
   flow, disturbed flow + simvastatin, laminar flow, and laminar flow +
   simvastatin. Blue: DAPI. Green: YAP. Scale bars: 50 μm. All data are
   presented as mean ± SEM.

Extended Data Fig. 8 |. Simvastatin-regulated enhancers identified by YAP
ChIP-seq.

   Extended Data Fig. 8 |
   [273]Open in a new tab

   a, Western blot analysis showing simvastatin downregulated SOX9
   expression in iPSC-ECs under disturbed and laminar flow patterns. b,
   Densitometric quantification of SOX9 protein expression changes in
   simvastatin- and vehicle-treated iPSC-ECs under disturbed and laminar
   flow patterns (n = 2 biological samples). c, Genome Browser snapshots
   of decreased binding loci at the NOTCH1 enhancer region showing
   simvastatin and GGTi298 treatment inhibited YAP binding at this region
   (highlighted in a yellow frame). d, Normalized gene expression levels
   (from RNA-seq) of Notch1 and Sox9 in aortas (n = 2 RNA-seq biological
   samples) from diabetic (db/db) and control mice (db/+). All data are
   presented as mean ± SEM. Unpaired two-sided Student’s t-test (b,d).

Extended Data Fig. 9 |. Schematic summary demonstrates that simvastatin
rescues endothelial dysfunction by repressing YAP-mediated chromatin
remodeling of the EndMT process.

   Extended Data Fig. 9 |
   [274]Open in a new tab

   A schematic overview showing our proposed model. In normal endothelial
   cells (ECs), geranylgeranyltransferase (GGTase) mediates YAP activity
   through the mevalonate pathway. Active YAP creates an ‘open chromatin
   status’ at the enhancer regions of genes regulating EndMT, such as
   SOX9. The diabetic condition further enhances YAP activity, thereby
   exacerbating endothelial functions by further upregulating EndMT genes.
   In contrast, statins decrease mevalonate levels via the inhibition of
   HMG-CoA reductase, followed by suppression of GGTase-mediated RhoA
   geranylgeranylation, consequently attenuating YAP activity (dashed
   line). Reduced YAP activity makes chromatin less-open (‘closed status’)
   at the enhancer regions of genes associated with EndMT.
   Diabetes-induced endothelial dysfunction can be alleviated by
   suppressing YAP activity with statins, GGTi298 or RhoA inhibitor
   (RhoAi), leading to the downregulation of EndMT genes. P,
   phosphorylation.

Supplementary Material

   Supplementary Information
   [275]NIHMS1923558-supplement-Supplementary_Information.pdf^ (8.1MB,
   pdf)

Acknowledgements