Endothelial cell (EC)-specific CTGF/CCN2 Expression Increases EC Reprogramming and Atherosclerosis

Abstract

Arterial endothelial cells (ECs) reside in a complex biomechanical environment. ECs sense and respond to wall shear stress. Low and oscillatory wall shear stress is characteristic of disturbed flow and commonly found at arterial bifurcations and around atherosclerotic plaques. Disturbed flow is pro-inflammatory to ECs. Arteries also stiffen with aging and/or the onset of vascular disease. ECs sense and respond to stiffening in a pro-fibrotic manner. Thus, flow and stiffening disturbances elicit EC responses that promote pathologic arterial remodeling. However, the pathways elicited by ECs under pathologic stiffening and disturbed flow are not well understood.

The objective of this work was to discover and test the modifiability of key pathways in ECs. To do this we used the partial carotid ligation model to impose disturbed flow onto the precociously stiffened fibulin-5 knockout (Fbln5^−/−) mouse carotid arteries. Biomechanical testing demonstrated that Fbln5^−/− arteries under disturbed flow approximate the stiffness ratio of diseased human arteries, and the ECs in these Fbln5^−/− arteries underwent rapid reprogramming via endothelial to mesenchymal transition (EndMT). Under atherogenic conditions, disturbed flow Fbln5^−/− arteries developed more vulnerable plaques than the wild type (WT) mouse arteries. Connective tissue growth factor/cellular communication network factor 2 (Ctgf/Ccn2) was upregulated in vivo in ECs with aging, with stiffening in the Fbln5^−/− arteries, and increased again by disturbed flow under stiffened conditions, supporting CTGF as a key biomarker for flow and stiffening. This was validated by immunohistochemistry, which demonstrated increased CTGF deposition in areas of disturbed flow in patient carotid endarterectomy and peripheral artery disease (PAD) specimens. Finally, to test the role of CTGF in regulating and combining these processes, we created an EC-specific Ctgf knockout (Ctgf^ecko). We identified that carotid arteries under disturbed flow and atherogenic conditions in male Ctgf^ecko, but not female, mice had decreased plaque area compared to WT control mice. We then tested the Ctgf expression in the carotid endothelium exposed to disturbed or stable flow in WT and Fbln5^−/− mice. Here we found that under disturbed flow male mice had greater Ctgf expression than female mice.

This work demonstrates that stiffened + disturbed flow conditions drive EC reprogramming, that CTGF is increased by these conditions, and that this increase is more prominent in male carotid arteries. Future exploration of sex-based differences in these fibrotic pathways are warranted to develop targeted therapeutics to limit pathologic arterial remodeling under pathologically stiffened + disturbed flow environments.

Introduction

Arterial endothelial cells (ECs) reside in a complex biomechanical environment. ECs sense and respond to wall shear stress. Disturbed flow is characterized as low and oscillatory wall shear stress. This is pro-inflammatory to ECs and commonly occurs at arterial bifurcations and around atherosclerotic plaques. Atherosclerotic plaques preferentially develop at sites of disturbed flow, and these atherosclerotic plaques, progressively obstruct flow further leading to extension of disturbed flow to surrounding arterial segments.[1, 2] Diseased arteries are characterized by increased arterial stiffness or the loss of arterial elasticity/compliance. Arterial stiffening is a risk factor for cardiovascular disease and mortality.[3] Stiffening also influences EC signaling pathways towards atherosclerotic plaque formation and arterial thrombosis.[4–6] Thus, disturbed flow and arterial stiffening partner to pathologically remodel diseased arteries. However, the mechanisms by which ECs coordinate their pathologic response to arterial stiffening and disturbed flow are not well understood. We previously reported that disturbed flow stiffens otherwise healthy carotid arteries through flow-sensitive matricellular proteins that modulate fibrosis and fibro-inflammation, including thrombospondin-1 (TSP-1).[7] While both disturbed flow[7, 8] and arterial stiffening[9] independently promote fibrotic and inflammatory pathways in ECs, the regulation of these fibro-inflammatory molecular pathways are poorly understood. By creating translational and mechanically motivated models of stiffening + disturbed flow, we aim to discover not only biomarkers of these pathways but the modifiability of key pathway in ECs.

The objective of this work was to discover and test the modifiability of stiffening and disturbed flow pathways in ECs. To deliver the disturbed flow + stiffening environment of diseased arteries, we imposed disturbed flow via partial carotid ligation [7] on fibulin-5 null (Fbln5^−/−); these mice have deficits in elastin function due to the absence of the elastin-associated glycoprotein fibulin-5, causing precocious arterial stiffening that is similar to the pathologic stiffening of aging.[10, 11] Similarly aged wildtype (WT) mice under disturbed flow and unmanipulated 18 month old WT mice served as control groups for basal arterial stiffness and stiffening from aging, respectively. The elastic moduli of these carotid arteries were then compared to that of diseased and healthy human arteries using a stiffness ratio. We hypothesized that profibrotic pathways in ECs would be augmented by stiffening and again by disturbed flow, would direct EC reprogramming, and promote a more progressive flow-mediated atherogenic remodeling compared to control groups.

To discover the relevant pathways for ECs exposed to disturbed flow and pathologic stiffening, we identified EC genes upregulated by disturbed flow, by stiffening under stable flow, and by both stiffening and disturbed flow. We discovered that connective tissue growth factor/cellular communication network factor 2 (Ctgf/Ccn2) was increased in all conditions and augmented by disturbed flow in both aged WT and Fbln5^−/− mice. This was validated in human arterial tissue under disturbed flow conditions. ECs in Fbln5^−/− rapidly developed endothelial to mesenchymal transition reprogramming (EndMT), and under atherogenic conditions, developed vulnerable plaque phenotypes.

An EC-specific knockout of CTGF (Ctgf^ecko) was created to test the modifiability of CTGF on flow-mediated atherogenic remodeling. Unexpectedly, we found that male, but not female Ctgf^ecko mouse arteries, had less atherosclerotic plaque. After interrogating our models for sex-based differences, we found that in response to disturbed flow, ECs in male arteries had significantly greater Ctgf expression compared to ECs in female mouse arteries, supporting a sex-difference in Ctgf expression under stiffened + disturbed flow conditions.

Results

Fbln5^−/− carotid arteries mechanics

To combine disturbed flow and stiffening in one model, disturbed flow was imposed via partial carotid ligation on fibulin-5 null (Fbln5^−/−), which have precocious stiffening. Partial carotid ligation induced disturbed flow in the left common carotid artery (LCCA); stable flow was preserved in the unmanipulated right common carotid artery (RCCA). Flow profiles were verified using duplex ultrasound (Figure 1A–C). When comparing the stiffness of unmanipulated (stable flow) Fbln5^−/− RCCA to age matched WT RCCA from prior publication [7], Fbln5^−/− RCCA were significantly stiffer than WT RCCA (Figure 1D). After 4 weeks of disturbed flow, Fbln5^−/− LCCA were stiffer than WT LCCA over a broad range of testing pressures (10–100mmHg) (Figure 1E). To compare the stiffness parameters of Fbln5^−/− and WT arteries to diseased peripheral artery disease (PAD) arteries, we used the ϕ_E (disturbed flow/stable flow stiffness ratio) of respective murine carotid artery elastic moduli and compared this to the stiffness ratio derived from comparing the elastic moduli of PAD arteries/aged healthy arteries. Fbln5^−/− arteries under disturbed flow had a stiffness ratio of ~3; this approximates the stiffness ratio in PAD arteries (Figure 1F). To compare the stiffness profile of Fbln5^−/− carotid arteries with that of aged WT carotid arteries, we compared stable flow RCCA and disturbed flow LCCA to that of aged (80-week/18-month-old) WT mice from prior publication [7]. Fbln5^−/− RCCA were stiffer than WT carotid arteries at 60–90 mmHg testing conditions (Figure 1G). Fbln5^−/− LCCA were stiffer than WT carotid arteries at 20–90mmHg and 110mmHg (Figure 1H). The axial or physiologic stretch (axial stiffness) had the same profile with Fbln5^−/− RCCA stiffer than 80-week-old WT and Fbln5^−/− LCCA stiffer than Fbln5^−/− RCCA (Supplemental Figure 2). There were no differences in disturbed flow mediated stiffening between Fbln5^−/− female and male carotid arteries (Supplemental Figure 3). Similar to diseased arteries, Fbln5^−/− arteries also demonstrated EC dysfunction at baseline. This is demonstrated by impaired endothelial dependent vasodilation in arterial ring testing (Supplemental Figure 4A) that was ameliorated by exogenous sodium nitroprusside (Supplemental Figure 4B).

Figure 1: Fbln5^−/− Mimics the Hemodynamics of PAD.

A/B. Representative duplex ultrasonography images of Fbln5^−/− LCCA and RCCA two weeks after partial carotid ligation with representative disturbed flow (LCCA) and stable flow (RCCA) waveforms. C. LCCAs demonstrated persistent disturbed flow manifest by low and oscillatory wall shear stress throughout the cardiac cycle (red curve), which contrast sharply with stable flow RCCA (blue curve). D. Compliance-pressure curves was generated from biaxial biomechanical test to compare the basal stiffness of unmanipulated Fbln5^−/− RCCA (stable flow) to stable flow WT RCCA. The Fbln5^−/− RCCA provides a stiffer profile over 20–80 mmHg. E. To quantify disturbed flow mediated stiffening in Fbln5^−/− arteries, compliance-pressure curves of Fbln5^−/− LCCA were compared to age matched WT LCCA, which demonstrated that the disturbed flow Fbln5^−/− LCCA were circumferentially stiffer than the disturbed flow WT LCCA. F. To compare stiffness changes between WT (black line) and Fbln5^−/− arteries (dotted line), a stiffness ratio was calculated comparing disturbed flow stiffness (LCCA) to stable flow RCCA stiffness. This was then compared to the stiffness ratio of PAD arteries (compared to healthy arteries). Here the mean average stiffness ratio and standard deviation is superimposed as triangle/blue line that is more similar to that of Fbln5^−/− compared to WT arteries. G. Basal stiffness (compliance-pressure curve) of the stable flow, Fbln5^−/− RCCA were similar to old (80 weeks) WT arteries and stiffer over 60–90 mmHg. H. When comparing compliance curves of 80-week-old WT arteries to Fbln5^−/− LCCAs, Fbln5^−/− LCCAs were significantly stiffer. Young WT n=6, 3 males/3 females; Aged WT n=6, all males; Fbln5^−/− n= 6–7, 4 males/3 females. Data are mean ± SEM; Two-way ANOVA with Tukey’s post hoc test (* P<0.05)

Testing the effect of disturbed flow and stiffening on EC reprogramming/plasticity

To investigate the effect of disturbed flow on EC plasticity, en face immunofluorescence staining of WT and Fbln5^−/− carotid arteries was performed at 72 hours after partial carotid ligation. We found increased alpha-smooth muscle actin (α-SMA) in the Fbln5^−/− LCCA endothelium compared to WT LCCA and both Fbln5^−/− and WT RCCA, supporting a synergistic relationship of stiffening + disturbed flow (Figure 2A–E). Microarray analysis of EC-enriched RNA from WT and Fbln5^−/− carotid arteries after 24h of partial carotid ligation was conducted for further mechanistic discovery. The differentially expressed genes (DEGs) are delineated in the supplement. Pathway enrichment analysis of the microarray DEGs comparing Fbln5^−/− LCCA and RCCA identified the following pathways: cytoskeleton remodeling signaling pathways (TGF and Wnt), cell adhesion, and the EndMT pathways (Figure 2G), which were not found in comparison of WT LCCA and RCCA (Figure 2F). Venn diagram of microarray DEGs comparing flow mediated (comparing disturbed flow to stable flow) differences within the Fbln5^−/− and WT arteries identified 24 overlapping genes, and 138 gene differences within Fbln5^−/− group, while 398 within the WT group (Supplemental Figure 5A). Venn diagram also showed 240 genes differently expressed comparing EC genes from RCCA of WT and Fbln5^−/− under stable flow and 441 EC genes from the LCCA under disturbed flow, of which only 62 were overlapped between the two comparisons. (Supplemental Figure 5B).

Figure 2: EC Reprogramming is Increased in Fbln5^−/− Carotid Arteries Under Disturbed Flow.

A-D. En face immunofluorescence staining of the intima demonstrates robust α-SMA expression in the endothelium of Fbln5^−/− LCCA (D) after 72 hours under disturbed flow; this is not seen in the WT or Fbln5^−/− under stable flow (A/B), or either in the WT under disturbed flow (C). Scale: 50μm. E. Quantification of α-SMA immunofluorescence in en face staining confirmed that disturbed flow significantly increased α-SMA expression both in WT and Fbln5^−/− LCCA. Moreover, Fbln5^−/− RCCA and LCCA manifested increased α-SMA in endothelium than WT RCCA and LCCA respectively. N=3 per group; all males. F/G. EC-enriched RNA from the LCCA and RCCA of WT and Fbln5^−/− mice at 24h post partial carotid ligation were collected. This was used to generate pathway enrichment analysis of differentially expressed genes (DEGs) between WT LCCA/RCCA (F) and Fbln5^−/− LCCA/RCCA (G). Cell cycle modification and changes in the actin cytoskeleton predominated in WT LCCA/RCCA, while EndMT and TGF-beta/Wnt cytoskeletal remodeling were prominent in Fbln5^−/− LCCA/RCCA. N=3 per group. Each sample was a combination of 3 arteries.

To discover the relevant pathways, we queried EC genes upregulated by disturbed flow, by stiffening under stable flow, and by both stiffening and disturbed flow. We discovered 16 genes that were significantly changed by all three (disturbed flow, stiffening, and stiffening + disturbed flow) conditions (Figure 3A). Ctgf (connective tissue growth factor), Lmo4 (LIM domain only 4), Timp3 (tissue inhibitor of metalloproteinase 3), Clic4 (chloride intracellular channel 4), Slmap (sarcolemma associated protein), and Wdr7 (WD repat domain 7) were significantly elevated under all 3 conditions (Figure 3B). The top 10 significantly expressed GO terms are presented in Figure 3C/D. The stiffening differences under stable flow were identified by comparing Fbln5^−/− RCCA and WT RCCA and include: cell migration, cell polarity, cytoskeleton, stress fiber and actin filament binding. (Figure 3C). The differences in differentially stiffened arteries under disturbed flow were identified by comparing Fbln5^−/− LCCA and WT LCCA. Here, ECs also exhibited significant changes in stress fiber, actin cytoskeleton organization, focal adhesion and integrin binding over WT controls. (Figure 3D) Moreover, using data from a previous Jo lab publication, Ctgf was the only one of these 6 genes that was highly upregulated in aged EC.[12]

Figure 3: Microarray Analysis of Significantly Different EC-enriched RNA under Disturbed flow, Increased Stiffening, and Disturbed Flow + Increased Stiffening Conditions.

A. Venn diagram of different expressed genes (DEGs) of microarray analysis identified the 16 overlapped genes between 3 comparisons: (1) WT LCCA: RCCA (disturbed flow); (2) RCCA Fbln5^−/−: WT (stiffening); (3) LCCA Fbln5^−/−: WT (disturbed flow + stiffening). Ctgf (connective tissue growth factor), Lmo4 (LIM domain only 4), Timp3 (tissue inhibitor of metalloproteinase 3), Clic4 (chloride intracellular channel 4), Slmap (sarcolemma associated protein), and Wdr7 (WD repat domain 7) were significantly elevated under all 3 conditions. B. Bar graph demonstrating the direction and fold change of these 16 genes. C/D. Gene ontology (GO) enrichment analysis of DEGs found in RCCA and LCCA between Fbln5^−/− and WT. GO biological process, cellular component, and molecular function enrichment analyses were performed in RCCA (C) and LCCA (D) between Fbln5^−/− and WT using Database for Annotation, Visualization, and Integrated Discovery (DAVID). Top 10 Significantly enriched GO terms are presented with Benjamini-Hochberg FDR-corrected P-values.

Aging, stiff arteries, and disturbed flow increase CTGF expression

To validate which these 16 genes identified to be changed by flow (comparing WT LCCA to RCCA), stiffening (Fbln5^−/− RCCA to WT RCCA), and disturbed flow + stiffening (Fbln5^−/−: WT LCCA) in aged mice, we queried data previously published by the Jo laboratory, [12] comparing young (12 weeks) versus aged (18 months) WT carotid arteries. Here we identified that Ctgf was increased in old versus young arteries and stable flow/disturbed flow arteries in a stair-step fashion with the greatest expression in old + disturbed flow (Figure 4A). We further confirmed the Ctgf expression in endothelial cells from male and female WT and Fbln5^−/− LCCA and RCCAs via qPCR of the intima as published.[7] Overall, the Ctgf expression in LCCA under disturbed flow was higher than RCCA stable flow arteries, except in female WT mice, who had the lowest Ctgf expression. ECs from Male Fbln5^−/− LCCAs had the greatest increase in Ctgf expression. This was greater than both male WT and female Fbln5^−/− LCCAs under disturbed flow (Figure 4B). Thus, Ctgf was determined to be the gene of choice for validating our finding s in human arterial tissue. Using disturbed flow regions of arterial specimens from peripheral arterial disease and carotid stenosis patients, we used immunohistochemistry (IHC) staining to compare the abundance of CTGF to that of human arteries under stable flow conditions. (Figure 4E–G) CTGF abundance was increased in disturbed flow regions of peripheral arteries and carotid endarterectomy (CEA) samples, both in the endothelium, as well as the rest of the arterial wall. (Figure 4C–D)

Figure 4: CTGF is Flow-dependent, Age-dependent, and Stiffness-Dependent and Increased in Human Arteries under Disturbed flow.

A. Normalized copy number of Ctgf revealed that Ctgf was elevated under disturbed flow in the LCCA of young mice compared to stable flow RCCA in young mice. This increase is comparable to the basal expression in stable flow RCCA in old mice (80-week-old). LCCA in older mice under disturbed flow show significantly higher expression compared to both young disturbed flow and aged stable flow. Data are mean ± SEM; n=3 for each group, all males; paired t-test for young stable flow vs disturbed flow; unpaired t-test for young disturbed flow vs old disturbed flow. B. Real-time qPCR was conducted on EC-enriched RNA from WT and Fbln5^−/− carotid arteries after 24h partial carotid ligation. Pre-stiffened, Fbln5^−/− RCCA exhibit higher Ctgf expression levels than WT RCCA. Ctgf increases in male (but not female) WT arteries under disturbed flow (LCCA). Pre-stiffened Fbln5^−/− LCCA (disturbed flow) have increased Ctgf expression compared to RCCA (stable flow), but the magnitude of Ctgf expression is significantly increased in males. Data are mean ± SEM; n=3–5 for each group; One-way ANOVA with Tukey’s post hoc test between groups and paired t-test for LCCA vs RCCA within each animal; C/D. CTGF abundance is increased in both the intima and the rest of the artery in disturbed flow regions of PAD arteries and CEA plaques compared to aged stable flow arteries. n=4 for healthy control; n=7 for PAD; n=16 for CEA; data are mean ± SEM; one-way ANOVA with Tukey’s post hoc test. E-G. Representative images of IHC staining of CTGF in stable flow peripheral arteries, disturbed flow peripheral arteries and disturbed flow CEA plaques. Scale: 100μm.

Fbln5^−/− carotid artery plaque under disturbed flow + atherogenic conditions

To test the role of pathologic stiffening and disturbed flow on atherosclerotic plaque, atherogenic conditions were introduced into Fbln5^−/− and WT mice by PCSK9 infection + high fat diet. This was followed by partial carotid ligation of the LCCA 1 week later. Disturbed flow LCCAs in Fbln5^−/− and WT mice both developed plaques after 4 weeks (representative histology presented in Figure 5A–E), while Fbln5^−/− LCCA manifested increased atherosclerotic plaque area, lipid deposition and necrotic core, and CD68 infiltration (Figure 5F–I). We also observed differences in atherosclerotic plaque features were driven by sex. Male Fbln5^−/− mice exhibited significantly larger plaques compared to male WT mice (p=0.0008), whereas no significant difference was found between female Fbln5^−/− and WT mice (p=0.3074). In contrast, female Fbln5^−/− mice showed significantly higher lipid content (p=0.0145) and larger necrotic cores (p=0.0062) in their plaques compared to female WT mice; this was not observed in the comparisons of male mice. Fbln5^−/− arteries also had increased inner and outer diameter (Figure 5L/M) indicating outward remodeling with the growing plaque area that is seen in human atherosclerosis [13] and an increase in elastin breaks that likely contributes to further stiffening in Fbln5^−/− arteries (Figure 5N). In contrast, WT LCCA demonstrated a more stable fibrotic plaque phenotype with medial thickening and greater collagen content (Figure 5J and O/P). Atherogenic conditions increased mean arterial pressure (MAP) in WT (compared to Fbln5^−/−) male animals that did not exist under standard, non-atherogenic conditions (Supplemental Figure 6A/B). There were no differences in the serum lipid levels between Fbln5^−/− and WT animals (Supplemental Figure 6E/F).

Figure 5: Disturbed flow Promotes Vulnerable Plaques in Fbln5^−/− Compared to WT LCCA.

A-E. Representative images of ORO, H&E, Masson’s trichrome, elastin autofluorescence and IHC staining of macrophages (CD68) of the LCCA from WT and Fbln5^−/− male and female mice. F. Fbln5^−/− mice exhibit significantly larger atherosclerotic plaque area; this is driven by differences in the male animals. G. Lipid deposition within the plaques is greater in Fbln5^−/− mice; this is driven by differences in the female animals. H. Plaques in Fbln5^−/− mice display increased necrotic area.; this is driven by differences in female animals I. The percentage of macrophage infiltration within the plaque is higher in Fbln5^−/− mice. J/K. Fbln5^−/− arteries had thinner medial layers than WT with no differences in intima/media ratio. L/M. Fbln5^−/− LCCA demonstrated increased inner and outer diameters of their lumen and arterial walls than WT. N. Fbln5^−/− LCCA also exhibited increased elastin breaks compared to WT LCCA. O/P. In contrast, WT LCCA developed more collagen content in the plaque, with collagen content in the rest of the arterial wall being similar. WT n=12 (7 males/5 females) and Fbln5^−/− n=10 (5 males/5 females) in ORO quantification; WT n=8 (4 males/4 females) and Fbln5^−/− n=7 (4 males/3 females) in the remaining comparisons; male and female mice are represented with blue and pink symbols, respectively; data expressed as mean ± SEM; unpaired t-test; *P<0.05; †<.01; ‡<.001. Scale: 50μm.

Endothelial CTGF-targeted therapy favorably improves plaque area in male mice

To test the modifiability of CTGF signaling in ECs on atherosclerotic plaque composition, we generated EC-specific Ctgf knockout mice (Ctgf^ecko) and exposed these mice to atherogenic conditions. There were no differences between age matched WT and Ctgf^ecko in body weight before and after experiments (Supplemental Table 3). Male Ctgf^ecko mice developed less plaque area and had less macrophage in plaque compared to control animals (Figure 6F/I). Female Ctgf^ecko developed a trend (P=.06) towards decreased necrotic area compared to littermate controls but were otherwise similar to WT female mice.

Figure 6: Male Ctgf^ecko LCCA under Disturbed flow Have Decreased Atherosclerotic Plaque Area and Macrophage Infiltration.

A-E. Representative images of ORO, H&E, Masson’s trichrome, elastin autofluorescence and IHC staining of macrophages (CD68) of the LCCA from EC specific knockout of Ctgf (Ctgf^ecko) and WT (Ctgf^ecwt) animals. F. Male Ctgf^ecko mice displayed reduced atherosclerotic plaque area than WT. G/H. Lipid deposition and necrotic area were comparable between groups. I. Male Ctgf^ecko mice showed less macrophage infiltration than WT. J/K. WT and Ctgf^ecko mice exhibited similar medial thickness and intima/media ratio. L/M. The inner luminal and outer arterial wall diameters of LCCA were similar between groups. N-P. No differences were observed between groups in elastin breaks or the collagen content of the arterial intima or rest of arterial wall. Ctgf^ecwt (male n=5; female n=4) Ctgf^ecko (male n=5; female n=5). Data expressed as mean ± SEM; unpaired t-test; *P<0.05; †<.01; ‡<.001. Scale: 50μm.

Discussion

This is the first publication to discover that Ctgf is increased in the EC response to stiffening due to aging, in a murine model of precocious stiffening, and that Ctgf expression was further increased by disturbed flow under all conditions. This was validated in human diseased arteries under disturbed flow, and we identified that the magnitude of Ctgf expression is increased in males compared to female arteries. Thus, we have identified CTGF as a biomarker of disturbed flow that increases with arterial stiffening. Since ECs sense both shear stress and arterial stiffening, it is not surprising that ECs express CTGF. However, it is notable to identify similar stair-step increases in EC Ctgf expression in Fbln5^−/− compared to WT and disturbed flow compared to stable flow, as well as in aged arteries compared to young arteries.

We have also more broadly characterized mechanical changes by disturbed flow and identified key signaling pathways in flow-mediated remodeling of Fbln5^−/− mouse arteries. We found that adding disturbed flow to the pre-stiffened Fbln5^−/− carotid arteries mimics the diseased-state biomechanics. Since the Fbln5^−/− mimics vascular aging due to a defect in elastic fiber assembling,[10, 11] we were not surprised that unmanipulated Fbln5^−/− RCCA were similar to and somewhat stiffer than 80-week-old aged arteries. We were impressed that the addition of disturbed flow further stiffened these already stiff arteries. Moreover, Fbln5^−/− mice displayed EC-dependent relaxation dysfunction which approximates to endothelial cells in human PAD arteries, which is another feature of this model that is shared by patients with vascular disease.

EndMT EC reprogramming may not only contribute to stiffened arterial remodeling but also to the vulnerable plaque phenotype seen in Fbln5^−/− LCCAs.[15] The introduction of EndMT by stiffening and disturbed flow in precociously stiffened Fbln5^−/− carotid arteries may have relevance to infra-inguinal arterial disease seen in peripheral arterial disease (PAD). Here, the infra-inguinal muscular arteries of the leg are naturally stiffer than elastic arteries like the carotid.[14] Thus, these muscular arteries may be particularly sensitive to this pathway. In this work, the impact of disturbed flow under stiffened conditions increased the onset of EC reprogramming in Fbln5^−/− under disturbed flow (3 days in this work compared to two weeks in prior publication by the Jo lab).[8] Thus, the effect of PAD conditions (stiffening + disturbed flow) on EC plasticity will be an important future direction. How EC reprogramming and matricellular proteins interact with EC signaling cascades in diseased arteries, and how can the pathologic mechanisms be targeted for the development of next generation therapies for arterial health is a long-term goal of our collaboration.

This work also reveals that CTGF is a critical mediator of flow-mediated atherosclerotic plaque formation, [16–18] and that this appears to be most prominent in males. Interestingly CTGF may be an important regulator of endothelial reprogramming, [8] particularly EndMT. [19] This fits with a putative role of CTGF in unhinging fibrotic check points and spurring cell transitions to pathologic phenotypes.[20] Thus, this work supports the paradigm that ECs in already stiff arteries that are exposed to disturbed flow, are “tuned” to direct a unique molecular signature to disturbed flow, which is more pernicious than that seen in less stiff arteries. This appears to be further aggravated by matricellular proteins, like CTGF, that tune molecular pathways toward vicious cycles, acting as a permanent on-accelerator switch for pathologic remodeling. We confirmed the correlation of CTGF abundance under disturbed flow in PAD arteries and carotid endarterectomy samples. CTGF abundance has been found high in complicated plaques compared with fibrous plaques,[21] and our work supports benefit of interfering with CTGF expression, at least in males. However, the exact mechanisms by which CTGF or Ctgf expression limit the activation of fibro-inflammatory pathways and thereby direct formation/progression of plaque and with inflammation remain unknown.[22]

While the burden vascular disease with increasing age is shared by both sexes,[23] sex-based differences in arterial stiffening is an important area of research. Clinically, arterial stiffening in women is delayed with age proportionately to the delayed onset of arterial stiffening and EC dysfunction, ~10 years after men;[23] delays in stiffening are also seen in murine models.[24] Arterial stiffening is of particular interest to sex-based differences because not only is it a risk factor for cardiovascular disease and mortality,[3] but the association between arterial stiffening and mortality is ~2 fold greater in women.[25] It is possible that matricellular proteins, like CTGF, could be differentially regulated by age-related stiffening. In this work and under atherogenic conditions, Fbln5^−/− resulted in increased plaque area in male mice and increased lipid content/necrotic core in female mice. However, Ctgf^ecko resulted in favorable decreased atherosclerotic plaque size only in male mice. This indicates that sex differences may impact potential benefits of CTGF inhibition on atherosclerotic plaque biology. This is consistent with decreased augmentation of Ctgf expression seen in female compared to male WT and Fbln5^−/− mice demonstrated in this work. It is also consistent with decreased EC expression in EndMT genes in female donor compared to male donor human aortic ECs that we recently published.[26] One potential regulator of sex-dependent CTGF is YAP. It has been reported that human umbilical vein ECs under stable flow led to an increase in YAP nuclear localization in male but not female human umbilical vein ECs.[17] While we did not find YAP expression increased in our data sets, YAP is mechanosensitive and related to CTGF signaling,[16, 18] so the interaction of YAP under stiffening and disturbed flow may be one possible mechanism at play, particularly as it relates to male/female differences seen here. While this manuscript focused on the EC response to stiffened + disturbed flow conditions, further sex-based investigation of matricellular pathways under PAD conditions is warranted.

It is important to note limitations of this work that warrant future work. While the Fbln5^−/− animal model of stiffening + disturbed flow well approximates flow and solid mechanics of human arterial disease, no animal model perfectly replicates the human condition. Arterial stiffness and other mechanical parameters differ between mouse and human arteries.

However, relative changes between mice and human arteries do appear to be conserved.[27] Still, there are additional downstream pathways in our data set that are likely important to the EC contribution to arterial stiffening and flow-mediated atherosclerotic plaque under these conditions. This work presents new opportunities to integrate stiffening + disturbed flow into their discovery science to build a better understanding of therapeutic opportunities that exist under disease-centric mechanics. Also, relatively little is known about the sex-based differences in EC reprogramming, particularly if there is a sex-based difference that is cued by stiffening and disturbed flow. It is exciting to consider that matricellular differences in CTGF and perhaps other genes may play a role in the onset and progression of arterial stiffening through differences in EC phenotypes. Of note, the sex differences were identified in a subgroup of the original animal group sample size. Thus, the sample sizes are small. Since this is a current focus for our team, we will be building more robust samples and integrating scRNAseq to better understand this finding. In addition, this work mainly observed the effect of endothelial knockout of Ctgf on atherosclerotic plaque at late stage. The role of CTGF over time on the immune-related genes (ICAM-1, VCAM-1, CCL2) and inflammatory state of flow-mediated atherosclerotic plaque would be very interesting to pursue as a future goal. To date the Jo lab has identified a prominent role for EC transition to immune-like cells.[8] Future work investigating such EC subpopulations will best define this relationship.

In summary, this work characterized a novel model of pathologic stiffening + disturbed flow mechanics, which contributes a modifiable mechanism and useful model to the emerging collective work linking arterial stiffening and vascular disease.[28, 29] The combination of stiffening + disturbed flow leads to the particularly aggressive atherosclerotic plaque remodeling as evidenced by plaque area, plaque content, and elastin breaks. We identified unique molecular signatures related to these mechanics that involve precocious EndMT and an important role for CTGF in disturbed flow mediated arterial remodeling and in PAD vasculature. This study highlighted the idea that limiting stiffening (targeted CTGF therapies) may play a favorable role in ameliorating the impact of PAD on vascular health and may reverse the velocity of atherosclerotic remodeling and outcome of this process, and that CTGF’s role in pathologic arterial remodeling may have sex-dependent features.

Experimental procedures

Animals

All mouse studies performed here were approved by the Institutional Animal Care and Use Committee (IACUC) at Emory University and under the established guidelines and regulations consistent with federal assurance. Fibulin-5 knockout (Fbln5^−/−) [C57Bl/6 × 129S2/SvPas] mice were generated from a heterozygous breeding pair originally obtained from Dr. Hiromi Yanagisawa.[30] Wild-type littermates were used as control. Constitutive endothelial cell (EC)-specific Ctgf knockout (Ctgf^fl/fl; Cdh5 Cre^+/−) mice were generated by breeding Ctgf^fl/fl mice (generously provided by Dr. Andrew Leask) with transgenic Cdh5-cre^7Mlia mice (Jackson Lab). The Ctgf^fl allele was converted to null allele by EC-specific Cdh5 Cre-mediated excision to create Ctgf^ecko. WT (Ctgf^fl/fl; Cdh5 Cre^−/−) littermates were used as control group. The expression of Cre recombinase in endothelial cells was validated by breeding the Cre mice with Ai14 reporter mice (B6.Cg-Gt (ROSA)26Sortm14(CAG-tdTomato)Hze/J, Jackson Lab). Supplemental Figure 1 provides an overview of the animal experimental methodology and time points for arterial remodeling studies.

Partial carotid ligation model of disturbed flow

In the partial carotid ligation model, three of the four caudal branches of the left common carotid arteries (LCCA) were ligated as published,[31] resulting in a disturbed flow in LCCA. Right common carotid arteries (RCCA) were not manipulated and had stable flow. Partial carotid ligation was performed on mice as listed in Supplemental Figure 1. The in vivo hemodynamic environment in the carotid arteries was characterized after partial carotid ligation using ultrasonography. Blood pressure (BP) was assessed in all murine groups using the CODA noninvasive BP system (a tail-cuff method, Kent Scientific Corporation) as published previously. [7]

Biomechanical testing of arterial stiffness

To compare the biomechanical properties, the carotid arteries were harvested from mice quickly after CO₂ asphyxiation, perfused with saline, isolated, excised, and mounted on an ex vivo bioreactor/biomechanical testing device for cylindrical biaxial biomechanical testing as published.[11]

Microarray procedure, data analysis, and bioinformatics

To compare EC gene expression between groups, EC enriched RNA from 9 LCCA and 9 RCCA (n=3 animals for each sample; n=3 samples per group) from male Fbln5^−/− and WT mice were pooled for microarray analysis after 24h partial carotid ligation. RNA sample quality was confirmed, and then samples were amplified and analyzed as a microarray using the Mouse WG-6 v2 expression BeadChip array by the Emory Winship Cancer Genomics Core. The meta-data was uploaded to the GEO Repository (GSE222583). The normalized microarray data was analyzed statistically, and the differentially expressed genes (DEGs) with statistically significance were screened for greater than 1.5 folds change and P-value <0.05. DEGs were submitted to functional analysis using the Database for Annotation, Visualization and Integrated Discovery (DAVID). The pathway enrichment analysis was performed by mapping these differentially expressed genes to GeneGO’s MetaCore^™.

Quantitative PCR for EC-enriched RNA isolation from carotid arteries

Total RNA from intima (EC-enriched) was collected from Fbln5^−/− and WT carotid arteries 24 hours post-partial carotid ligation. The carotid lumen was flushed with QIAzol lysis reagent, and the eluate was used for RNA isolation using RNeasy mini kit as published.[7] Quantitative PCR was performed on target genes using One-Step Multiplex Supermix (Bio-Rad) on a Real-Time PCR system (ABI StepOne Plus). The primer sequences for the involved genes were as follows:

CTGF: forward 5’-TGCGAAGCTGACCTGGAGGAAA-3’

reverse 5’-CCGCAGAACTTAGCCCTGTATG-3’

18s: forward 5’-AGGAATTGACGGAAGGGCACCA-3’

reverse 5’- GTGCAGCCCCGGACATCTAAG-3’

En face staining for EC plasticity

Mice were euthanized by CO₂ inhalation at 72 hours post partial carotid ligation, arteries perfused with saline containing heparin, followed by a second perfusion with 10% neutral buffered formalin. Carotid arteries were carefully cleaned and dissected free of surrounding tissues and fats. The intima of LCCA and RCCA were stained en face with VE-cadherin antibody (BD Biosciences) and α-SMA antibody (ab5694, Abcam), then followed staining with DyLight Fluor secondary antibodies. The nuclei were counterstained with DAPI (Sigma) and mounted on glass slides using fluorescence mounting medium (Dako). En face images were collected with a Zeiss LSM 510 META confocal microscope. Fluorescence intensity of 10–12 images from each group was quantified by ImageJ software with the same threshold.

Murine model of atherosclerosis

Atherosclerosis in LCCA was induced in all murine groups using age and gender matched 12–20 weeks old mice as published.[32] Briefly, mice were injected once with AAV-PCSK9 (1×10¹¹ VG) and fed a high-fat diet (16% fat and 1.25% cholesterol, Research Diets Cat # D12336, USA) to create an atherogenic environment. After one week, mice were subjected to partial carotid ligation as above. Animals were euthanized 4 weeks post-surgery with CO₂ asphyxiation. After perfused with saline, carotid arteries were dissected out and cleaned. The blood serum was collected post-mortem through cardiac puncture for validation of hyperlipidemia.

Histological and immunohistochemical analysis of animal carotid arteries

Oil red O (ORO), H&E, Masson’s trichrome staining, and elastin autofluorescence were conducted on cross sections of common carotid arteries. Lipids were detected with ORO (Sigma) staining following the standard protocol as described.[31] Masson’s trichrome staining was used for the assessment of collagen composition of the atherosclerotic plaque. Elastin architecture was visualized by autofluorescence. Immunohistochemical (IHC) staining of CD68 (MCA1957, Bio-Rad) was applied to quantify macrophage positive area. The images were captured by an Olympus microscope (IX51). The assessments of histological characteristics were blinded and quantified using Image J program.

Human tissue validation testing

After informed consent was obtained, human artery samples from PAD and carotid endarterectomy (CEA) patients (Supplementary Tables 1 and 2) were collected from operative specimens under an IRB approved protocol (Emory University approved IRB protocol numbers: 51432 and 70813). Arterial specimens were segregated prior to testing into stable flow or disturbed flow environment based on whether there was unimpeded flow to the artery as previously published.[7] All disturbed flow arteries were in patients with proximal inflow occlusion and distal reconstitution through collateral pathways. All stable flow arteries had inline flow without obstructive PAD. IHC staining of CTGF (Catalog No. ab6992, Abcam) was performed on these tissues. Semi-quantitative mean gray value was quantified with ImageJ.

Statistics

Statistical comparisons were performed with GraphPad Prism (GraphPad Software), and statistical significance was set at P<0.05 (*P<0.05; †<.01; ‡<.001.) Parameters such as sample size, the number of replicates, the number of independent experiments, measures of center, dispersion and precision, and statistical significance are reported in figures and/or figure legends. Briefly, two-way ANOVA was used in pressure-compliance curves. Unpaired t-tests were performed to compare histological parameters of LCCA and RCCA. One-way ANOVA with Tukey’s post hoc tests were performed in statistical comparisons of multiple groups.

Highlights:

• This study discovers the endothelial cell molecular signature in response to an environment combined stiffening + disturbed flow. These pathways directed endothelial cells towards an endothelial to mesenchymal transition phenotype, and under atherogenic conditions, deliver a vulnerable plaque profile.

• This study identifies a pattern of Ctgf upregulation in endothelial cells under disturbed flow, endothelial cells in arteries stiffened by age or in a genetic model of precocious stiffening (Fbln5^−/−) and further augmented when disturbed flow is combined with stiffening. This was validated in arterial tissue under disturbed flow from peripheral arterial and carotid artery disease.

• Ctgf upregulation in the endothelium of carotid arteries under disturbed flow was significantly increased in male mice, and disturbed flow-mediated atherosclerotic plaque was decreased in male, but not female, Ctgf^ecko mouse carotid arteries compared to WT mice, supporting a sex-based regulation of Ctgf in flow-mediated arterial remodeling.

Data availability

The meta-data of microarray was uploaded to the NCBI Gene Expression Omnibus (GEO) Repository (GSE222583).