Aortic Fibrosis in Insulin-Sensitive Mice with Endothelial Cell-Specific Deletion of Ceacam1 Gene

(1) Background: Mice with global Ceacam1 deletion developed plaque-like aortic lesions even on C57BL/6J background in the presence of increased endothelial cell permeability and insulin resistance. Loss of endothelial Ceacam1 gene caused endothelial dysfunction and reduced vascular integrity without affecting systemic insulin sensitivity. Because endothelial cell injury precedes atherosclerosis, we herein investigated whether the loss of endothelial Ceacam1 initiates atheroma formation in the absence of insulin resistance. (2) Methods: Endothelial cell-specific Ceacam1 null mice on C57BL/6J.Ldlr−/− background (Ldlr−/−VECadCre+Cc1fl/fl) were fed an atherogenic diet for 3–5 months before metabolic, histopathological, and en-face analysis of aortae were compared to their control littermates. Sirius Red stain was also performed on liver sections to analyze hepatic fibrosis. (3) Results: These mice displayed insulin sensitivity without significant fat deposition on aortic walls despite hypercholesterolemia. They also displayed increased inflammation and fibrosis. Deleting Ceacam1 in endothelial cells caused hyperactivation of VEGFR2/Shc/NF-κB pathway with resultant transcriptional induction of NF-κB targets. These include IL-6 that activates STAT3 inflammatory pathways, in addition to endothelin-1 and PDGF-B profibrogenic effectors. It also induced the association between SHP2 phosphatase and VEGFR2, downregulating the Akt/eNOS pathway and reducing nitric oxide production, a characteristic feature of endothelial dysfunction. Similarly, hepatic inflammation and fibrosis developed in Ldlr−/−VECadCre+Cc1fl/fl mice without an increase in hepatic steatosis. (4) Conclusions: Deleting endothelial cell Ceacam1 caused hepatic and aortic inflammation and fibrosis with increased endothelial dysfunction and oxidative stress in the presence of hypercholesterolemia. However, this did not progress into frank atheroma formation. Because these mice remained insulin sensitive, the study provides an in vivo demonstration that insulin resistance plays a critical role in the pathogenesis of frank atherosclerosis.


Introduction
Atherosclerotic cardiovascular disease remains a major cause of death worldwide [1]. Several risk factors predispose to atherosclerosis, most importantly the metabolic syndrome: a constellation of cardiometabolic abnormalities that include visceral obesity, hypertension, and non-alcoholic fatty liver disease (NAFLD) [2]. Common molecular mechanisms include insulin resistance and dyslipidemia characterized by elevated serum triglycerides, decreased HDL cholesterol, and a remarkable increase in small dense LDL cholesterol [3,4]. Whether insulin resistance leads to atherosclerosis independently of accompanying dyslipidemia remains unclear, owing in part to the lack of an experimental model. Most commonly used atherosclerosis-prone mouse models are the apolipoprotein E-deficient (ApoE −/− ) and LDL receptor-deficient (Ldlr −/− ) mice, which exhibit a phenotype driven by extreme hypercholesterolemia that exceeds the level found in most patients with atherosclerosis and can act as the sole atherogenic factor that masks the potential effect of insulin resistance in the pathogenesis of atherosclerosis [5].
Insulin resistance is an independent risk factor for atherosclerosis [3,6,7]. It exerts an atherogenic effect on the vascular endothelium in association with endothelial dysfunction [4,8,9] that is manifested by compromised vasodilation and increased leukocyteendothelium adhesiveness [8,10], both of which are caused by reduced nitric oxide (NO) bioavailability [4]. Insulin resistance also increases the vasoconstrictor tone by upregulating NF-kB transcriptional activity to induce the expression of endothelin-1 [11] and of pro-inflammatory cell adhesion molecules, such as vascular cell adhesion molecule-1 (VCAM-1) [12].
Carcinoembryonic antigen-related cell adhesion molecule 1 (CEACAM1), a transmembrane glycoprotein and a common substrate of insulin receptor and vascular epidermal growth factor receptor 2 (VEGFR2) [13,14], is ubiquitously expressed. In hepatocytes, it regulates insulin action by promoting insulin clearance [15]. Accordingly, its global and liver-specific biallelic deletion cause hepatic insulin resistance and increased hepatic de novo lipogenesis that leads to VLDL-triglyceride redistribution to the white adipose tissue to contribute to visceral obesity, followed by lipolysis and inflammation with resultant systemic insulin resistance [16]. In addition, global Cc1 −/− null mice fed a regular chow diet manifested cardiac and endothelial dysfunctions, and aortic plaque-like lesions with macrophage and collagen deposition and restricted fat accumulation, likely owing to its limited pro-atherogenic hypercholesterolemia [17]. Conversely, liver-specific reconstitution of CEACAM1 restored insulin sensitivity and reversed hepatic steatosis in Cc1 −/− mice [18]. It also reversed their endothelial and cardiovascular dysfunctions in parallel to restoring nitric oxide (NO) bioavailability and reduced endothelial-leukocyte adhesion [19]. This prevented diet-induced non-alcoholic steatohepatitis (NASH) that is often linked to atherosclerosis [20]. In support of the role of hepatic CEACAM1 in atherosclerosis, mice with liver-specific Ceacam1 deletion developed NASH and atherogenic plaques when propagated on Ldlr −/− background and fed a high-cholesterol diet [21]. While this mouse model provided an in vivo evidence of how hyperinsulinemia-driven insulin resistance links NASH to atherosclerosis in mice with pro-atherogenic hypercholesterolemia [21], it did not dissect out the role of insulin resistance independently of dyslipidemia in the development of aortic plaques. On the other hand, ApoE −/− mice with whole-body insulin receptor haploinsufficiency developed hyperinsulinemia in response to impaired insulin clearance [22]. However, in the absence of insulin resistance and significant changes in dyslipidemia, they developed comparable levels of aortic plaques to their ApoE −/− controls [22]. While this mouse demonstrated that hyperinsulinemia without insulin resistance did not exert a pro-atherogenic effect, it failed to dissociate insulin resistance from dyslipidemia in the pathogenesis of atherosclerosis.
Endothelial dysfunction can also be caused by endothelial cell injury, which in turn can play an essential role in the pathogenesis of atherosclerosis [23]. Loss of endothelial Ceacam1 disturbed the endothelial barrier integrity and caused permeability owing to the dissociation of VE-cadherin/β-catenin complexes at cell junctions [24]. Endothelial cell injury in BL6.VECadCre+Cc1 fl/fl mice with endothelial-specific deletion of Ceacam1 was accompanied by persistent insulin sensitivity. In response to increased paracellular insulin transport, these mice exhibited intact glucose turnover [25]. They also manifested enhanced fat storage in adipocytes despite the development of oxidative stress and increased NF-kB-dependent production of pro-inflammatory cytokines (TNFα) and pro-fibrogenic effectors (endothelin-1) [25]. This appeared to be due to TNFα-mediated induction of the anti-inflammatory CEACAM1-4L variant in M2 macrophages associated with white adipose tissue to enhance innate immunity [26] and counter inflammation-driven insulin resistance [27].
To address whether endothelial CEACAM1 prevents atherosclerosis, the current study examined whether the insulin sensitive endothelial-Ceacam1 null mouse on the Ldlr −/− background (Ldlr −/− VECadCre+Cc1 fl/fl ) develop atherosclerosis. Because the combined deletion of Ldlr and Ceacam1 is not expected to alter insulin sensitivity [5], this model will predictably test whether hypercholesterolemia causes atherosclerosis in the absence of insulin resistance.  [25]. They also showed intact lipid homeostasis at 9 months of age, as assessed by normal hepatic (not shown) and plasma triacylglycerol ( Figure S2A(i)), and cholesterol (total, free, VLDL, LDL and HDL) levels compared to their littermate controls ( Figure S2A(ii-vi)). Mice with endothelial loss of Ceacam1 did not develop steatosis in aortae, as revealed by Oil red-O staining of aortic root sections ( Figure S2B(i)). Moreover, IHC analysis of aortic roots with CD68 ( Figure S2B(ii)) and Trichrome staining ( Figure S2B(iii)) revealed no induction in macrophage activation or fibrosis, respectively, in BL6.VECadCre+Cc1 fl/fl nulls relative to their controls. This is in marked contrast to the insulin resistant Cc1 −/− mouse that developed steatosis, inflammation, and fibrosis in aortic roots in the face of limited pro-atherogenic lipidemia when propagated on C57BL6/J background and fed a regular chow diet [17].
Failure of BL6.VECadCre+Cc1 fl/fl nulls to develop atherogenic-like histological abnormalities could stem from the absence of insulin resistance and pro-atherogenic hypercholesterolemia. To test this hypothesis, we propagated mice on the Ldlr −/− background and fed them a high-cholesterol (HC) atherogenic diet for 3 months starting at 6 months of age. Ldlr −/− VECadCre+Cc1 fl/fl nulls displayed a comparable age-related (Table 1) and dietinduced body weight gain to their littermate controls when fed an HC-rich diet ( Figure 1A). This was supported by normal visceral adipose mass in the null mice relative to controls at 9 months of age (Table 1). Steady-state plasma insulin levels remained normal in association with intact insulin secretion (inferred to from intact plasma C-peptide and fasting normoglycemia) and clearance (steady-state C-peptide/insulin molar ratio) (Table 1). Together with normal tolerance to exogenous insulin ( Figure 1B) and glucose (not shown), normal fasting insulinemia and random glycemia (Table 1) demonstrated persistent insulin sensitivity in Ldlr −/− VECadCre+Cc1 fl/fl mice, as expected of the Ldlr −/− background, which does not significantly cause/accelerate insulin resistance in the absence of other genetic factors [5].
Consistent with normal visceral adiposity, null mice exhibited intact plasma nonesterified fatty acids (NEFA) ( Table 1). They also exhibited intact hepatic and plasma triacylglycerol levels (Table 1). In contrast, plasma LDL cholesterol (LDL-C) and total cholesterol levels were significantly elevated in Ldlr −/− VECadCre+Cc1 fl/fl compared to controls (Table 1). This did not derive from changes in cholesterol synthesis, as supported by normal hepatic HMG-CoA reductase (HMGCR) activity in nulls compared to controls (Table 1). It likely derived from reduced LDL clearance as suggested by the higher plasma ApoB100, and proprotein convertase subtilisin/kexin type 9 (PCSK9) levels (Table 1), leading to reduction in LDL-C transport in null mice [28]. Consistently, the mRNA levels of hepatic Pcsk9 were~2-to-4-fold higher and those of hepatic LDL receptor-related protein 1 (Lrp1) that is involved in hepatocytic clearance of chylomicron remnants were~2-fold lower in nulls versus controls. Together, this demonstrated that Ldlr −/− VECadCre+Cc1 fl/fl developed pro-atherogenic dyslipidemia without insulin resistance. Except for andom blood glucose level that was assessed in blood drawn at 10:00 pm, male mice (n > 7/genotype) were fasted from 5:00 pm until 11:00 am in the morning of the following day. Retro-orbital blood was drawn, and tissues were collected. Visceral adiposity: % of gonadal plus inguinal white adipose tissue per body weight; the steady-state plasma C-peptide/insulin (C/I) molar ratio was calculated as a measure of insulin clearance; NEFA: non-esterified fatty acids. Plasma VLDL-C was calculated as triacylglycerolx0.2. Data were analyzed by one-way ANOVA with Tukey's for multiple comparisons and values were expressed as mean ± SEM. * p < 0.05 vs. Ldlr −/− VECad-Cc1 +/+ , † p < 0.05 vs. Ldlr −/− VECad+Cc1 +/+ , § p < 0.05 vs. Ldlr −/− VECad-Cc1 fl/fl .

Ldlr −/− VECadCre+Cc1 fl/fl Mice Did Not Develop Atheromatic Plaques
Oil red-O staining revealed no significant change in fat deposition in the aortic root cross-sections from HC-fed Ldlr −/− VECadCre+Cc1 fl/fl mice relative to their littermate controls ( Figure 1C). En-face analysis revealed no difference in the fat deposition along the whole aortae of Ldlr −/− VECadCre+Cc1 fl/fl and their aged-matched controls even after 5 months of HC intake ( Figure 1D and accompanying graph). Moreover, there was no gross aorta dilation or aneurysm formation in null mice. This histology developed in the presence of higher (~2-fold) plasma total-C and LDL-C/VLDL-C levels ( Figure 1E) and sustained insulin sensitivity in null mice by comparison to their control littermates (not shown).

Elevated Pro-Inflammatory State in Ldlr −/− VECadCre+Cc1 fl/fl Aortae
Inflammation is a key factor in the development of pro-fibrotic diseases such as atherosclerosis [29]. Consistently, deleting Ceacam1 from endothelial cells induced the macrophage pool (mRNA of F4/80) (Table S2) and its activation (IHC analysis of CD68- Figure 2A) in Ldlr −/− VECadCre+Cc1 fl/fl aortae compared to all three controls. qRT-PCR analysis revealed an increase in the mRNA levels of pro-inflammatory CD4+T and CD8+T cells, but not in the anti-inflammatory Treg pools (Foxp3) in Ldlr −/− VECadCre+Cc1 fl/fl aortae (Table S2).  (C) Western blot analysis was performed on aorta lysates by immunoblotting (Ib) with antibodies (α-) against phosphorylated (i) α-pNF-κB, (ii) α-pSTAT3, and (iii) α-pPKCζ and was normalized by immunoprobing parallel gels with antibodies against total NF-κB, STAT3, and PKCζ, respectively. Numbers (in kDa) to the right of the gels indicate the apparent molecular mass of the signaling proteins. Gels represent analysis on 2 mice/genotype performed on different sets of mice/protein.
Activated macrophages and increased inflammatory response was supported by the higher production of pro-inflammatory cytokines (IL-1β; IL-6; TNFα) (Table S2) and their release into the plasma ( Figure 2B). As we have reported in BL6.VECadCre+Cc1 fl/fl mice [25], this could stem from the basal activation (phosphorylation) of their transcriptional master regulator (NF-κB) ( Figure 2C(i)). As in response to insulin, VEGF-A, (40 ng/mL, 5 min) induced the pull-down of phosphorylated VEGFR2 in the α-Shc immunopellet in endothelial cells isolated from the hearts of 2-month-old Ldlr −/− VECadCre+Cc1 fl/fl mice relative to their control counterparts ( Figure 3A(i)). This activated MAP Kinase ( Figure 3A(ii)) and NF-κB ( Figure 3A(iii)) in null endothelial cells to a higher extent than controls.
Consistent with increased inflammation and vascular permeability, deleting Ceacam1 from endothelial cells induced leukocyte adhesion to the vessel wall of the nulls, as indicated by intravital microscopy of the carotid artery ( Figure 4B). This was supported by increased mRNA (Table S2) and protein levels ( Figure 4A(ii)) of the vascular cell adhesion As expected from IL-6 activation of STAT3 (signal transducer and activator of transcription 3), Ldlr −/− VECadCre+Cc1 fl/fl aortae exhibited higher basal STAT3 phosphorylation (activation) compared to controls ( Figure 2C(ii)). Together with NF-κB activation, this induced the mRNA levels of monocyte chemoattractant protein-1 (Mcp-1), Toll-like receptors 2/4 (Tlr-2/4), and the Cd11b+ macrophage pool (Table S2).
The data demonstrate that null deletion of Ceacam1 in endothelial cells caused a cellautonomous positive effect on NF-κB and Stat3 activation and their downstream transcriptional targets to induce an inflammatory microenvironment in Ldlr −/− VECadCre+Cc1 fl/fl aortae.

Increased Leukocyte Adhesion to the Vessel Walls of Ldlr −/− VECadCre+Cc1 fl/fl Mice
Endothelial CEACAM1 prevents vascular permeability by contributing to the regulation of endothelial barrier formation [24]. Accordingly, Ldlr −/− VECadCre+Cc1 fl/fl aortae displayed a 2-to-3-fold decrease in the mRNA levels of genes involved in the formation of tight (Zo-1, Claudin1, Claudin5, Occludin) and adherens junctions (VE-Cadherin, βcatenin), which contribute to the maintenance of vascular integrity together with VEGF-A, VEGFR2, VEGFR1 and angiopoietins (Ang-1/2) (Table S3). Western blot analysis confirmed reduction of VEGFR2 in Ldlr −/− VECadCre+Cc1 fl/fl aortae relative to controls ( Figure 4A(i)). Activation of PKCζ/NF-κB pathways ( Figure 2C Figure 2, mice were fed a HC for 3 months and their aorta lysates were subjected to immunoblotting with (i) α-VEGFR2 and (ii) α-VCAM1 followed by probing with α-Tubulin for normalization. Gels represent analysis on 2 mice/genotype performed on different sets of mice/protein. (B) Mice (n ≥ 5/genotype) were fed a HC for 5 months before their carotid arteries were isolated from the surrounding tissues and intravital microscopy of leukocyte adhesion was assessed, as described in the methods section. Cells that adhered to the vessel wall without rolling or moving for at least 3 sec (red arrows) were counted over the vessel observed by using an intravital microscope. Video images were analyzed offline for leukocyte adhesion. Total numbers were used for statistical analysis. Values were expressed as mean ± SEM and represented in the accompanying graph. * p < 0.05 vs. Ldlr −/− VECad+Cc1 +/+ (a, white bar) and § p < 0.05 vs. Ldlr −/− VECad-Cc1 fl/fl (b, grey bar).

Increased Oxidative Stress in Ldlr −/− VECadCre+Cc1 fl/fl Mice
Oxidative stress plays a central role in cardiovascular disorders [29]. 3 months of HC induced the mRNA levels of NOX4 and Gp91 in the aortae of Ldlr −/− VECadCre+Cc1 fl/fl mice (Table S3). It also increased their plasma NAD/NADH and 8-isoprostane levels with a reciprocal decrease in plasma nitric oxide (NO) levels (Table 1). Moreover, the mRNA level of aortic Niemann-Pick type C1 protein (NPC-1) was significantly reduced in Ldlr −/− VECadCre+Cc1 fl/fl mice (Table S3). This was associated with a marked reduction in their plasma GSH levels (Table 1), which could contribute to the robust response to the cytotoxic effects of TNFα.
Inflammation and oxidative stress mediate endothelial injury resulting from low NO bioavailability [4]. Consistently, 3 months of HC lowered aortic NO levels in Ldlr −/− VECadCre+Cc1 fl/fl mice (Table 1). As in insulin treatment of endothelial cells from BL6.VECadCre+Cc1 fl/fl mice [25], VEGF-A treatment induced binding of SHP2 to VEGFR2 ( Figure 5B) in the absence of its sequestration by CEACAM1 in Ldlr −/− VECadCre+Cc1 fl/fl endothelial cells ( Figure 5A). Consequently, phosphorylation of VEGFR2 ( Figure 5B) and of its Akt/eNOS downstream signaling pathway ( Figure 5C,D) was reduced, stemming NO production and its release from null endothelial cells ( Figure 5E).  Figure 2, mice were fed a HC for 3 months and their aorta lysates were subjected to immunoblotting with (i) α-VEGFR2 and (ii) α-VCAM1 followed by probing with α-Tubulin for normalization. Gels represent analysis on 2 mice/genotype performed on different sets of mice/protein. (B) Mice (n ≥ 5/genotype) were fed a HC for 5 months before their carotid arteries were isolated from the surrounding tissues and intravital microscopy of leukocyte adhesion was assessed, as described in the methods section. Cells that adhered to the vessel wall without rolling or moving for at least 3 sec (red arrows) were counted over the vessel observed by using an intravital microscope. Video images were analyzed offline for leukocyte adhesion. Total numbers were used for statistical analysis. Values were expressed as mean ± SEM and represented in the accompanying graph. * p < 0.05 vs. Ldlr −/− VECad+Cc1 +/+ (a, white bar) and § p < 0.05 vs. Ldlr −/− VECad-Cc1 fl/fl (b, grey bar). Consistent with increased inflammation and vascular permeability, deleting Ceacam1 from endothelial cells induced leukocyte adhesion to the vessel wall of the nulls, as indicated by intravital microscopy of the carotid artery ( Figure 4B). This was supported by increased mRNA (Table S2) and protein levels ( Figure 4A(ii)) of the vascular cell adhesion molecule (VCAM-1), a transcriptional target of NF-κB and a mediator of leukocyteendothelial adhesion, as previously shown in bovine aortic endothelial cells with SiRNAmediated downregulation of Ceacam1 [17].  (Table S3). It also increased their plasma NAD/NADH and 8-isoprostane levels with a reciprocal decrease in plasma nitric oxide (NO) levels (Table 1). Moreover, the mRNA level of aortic Niemann-Pick type C1 protein (NPC-1) was significantly reduced in Ldlr −/− VECadCre+Cc1 fl/fl mice (Table S3). This was associated with a marked reduction in their plasma GSH levels (Table 1), which could contribute to the robust response to the cytotoxic effects of TNFα.

Increased Oxidative Stress in Ldlr
Inflammation and oxidative stress mediate endothelial injury resulting from low NO bioavailability [4]. Consistently, 3 months of HC lowered aortic NO levels in Ldlr −/− VECadCre+Cc1 fl/fl mice (Table 1). As in insulin treatment of endothelial cells from BL6.VECadCre+Cc1 fl/fl mice [25], VEGF-A treatment induced binding of SHP2 to VEGFR2 ( Figure 5B) in the absence of its sequestration by CEACAM1 in Ldlr −/− VECadCre+Cc1 fl/fl endothelial cells ( Figure 5A). Consequently, phosphorylation of VEGFR2 ( Figure 5B) and of its Akt/eNOS downstream signaling pathway ( Figure 5C,D) was reduced, stemming NO production and its release from null endothelial cells ( Figure 5E).

Increased Inflammation and Fibrosis in the Livers of Ldlr −/− VECadCre+Cc1 fl/fl Mice
Disruption of the integrity of endothelial cells, the second largest cell population in the liver, facilitates the recruitment of neutrophils to contribute to hepatic fibrosis [33,34]. Thus, we next investigated whether endothelial Ceacam1 deletion altered liver histology. H&E analysis of liver sections from mice fed HC for 3 months revealed extensive parenchymal inflammatory foci formation (arrowheads) without an increase in steatosis in Ldlr −/− VECadCre+Cc1 fl/fl mice compared to controls ( Figure 7A panels d vs. a-c). Like global Cc1 −/− mice [20], this occurred without ballooning or changes in hepatocellular architecture. Consistently, hepatic mRNA of lipogenic genes such as Srebp-1c and fatty acid synthase (Fasn) ( Figure 7C) and hepatic triacylglycerol levels and HMGCR activity were normal (Table 1). Cc1 −/− mice [20], this occurred without ballooning or changes in hepatocellular architecture. Consistently, hepatic mRNA of lipogenic genes such as Srebp-1c and fatty acid synthase (Fasn) ( Figure 7C) and hepatic triacylglycerol levels and HMGCR activity were normal (Table 1). As in aortae, the macrophage pool in Ldlr −/− VECadCre+Cc1 fl/fl livers was induced and activated, as assessed by higher F4/80 and CD68 hepatic levels, respectively ( Figure 7C). In addition to activated endothelial cells, this could contribute to increased hepatic mRNA levels of pro-inflammatory genes (IL-1β; IL-6; TNFα). In addition, hepatic mRNA levels of IFNγ points to increased activation of lymphocytes in null versus control mice ( Figure  7C). Consistent with higher TNFα, hepatic Smad7 mRNA levels were lower in Ldlr −/− VECadCre+Cc1 fl/fl livers ( Figure 7C). Together with their higher hepatic Tgfβ mRNA levels, this suggests increased activation of the TGFβ signaling pathway to mediate hepatic fibrosis. Hepatic fibrosis was demonstrated by increased mRNA levels of α-SMA and Col6α3 pro-fibrogenic genes ( Figure 7C) and manifested by increased interstitial chicken-wire bridging pattern of collagen deposition in Sirius Red-stained Ldlr −/− VECadCre+Cc1 flfl liver sections ( Figure 7B panel d vs. a-c). Consistently, the hepatic mRNA levels of ET-1 and in its pro-fibrogenic ETAR receptors were induced with a reciprocal decrease in the anti-fibrogenic ETBR mRNA levels in Ldlr −/− VECadCre+Cc1 flfl livers compared to control mice ( Figure 7D). This could be partly caused by hyperactivation of NF-κB in endothelial cells in response to their loss of Ceacam1 gene [25,35]. As in aortae, the macrophage pool in Ldlr −/− VECadCre+Cc1 fl/fl livers was induced and activated, as assessed by higher F4/80 and CD68 hepatic levels, respectively ( Figure 7C). In addition to activated endothelial cells, this could contribute to increased hepatic mRNA levels of pro-inflammatory genes (IL-1β; IL-6; TNFα). In addition, hepatic mRNA levels of IFNγ points to increased activation of lymphocytes in null versus control mice ( Figure 7C). Consistent with higher TNFα, hepatic Smad7 mRNA levels were lower in Ldlr −/− VECadCre+Cc1 fl/fl livers ( Figure 7C). Together with their higher hepatic Tgfβ mRNA levels, this suggests increased activation of the TGFβ signaling pathway to mediate hepatic fibrosis. Hepatic fibrosis was demonstrated by increased mRNA levels of α-SMA and Col6α3 pro-fibrogenic genes ( Figure 7C) and manifested by increased interstitial chickenwire bridging pattern of collagen deposition in Sirius Red-stained Ldlr −/− VECadCre+Cc1 flfl liver sections ( Figure 7B panel d vs. a-c). Consistently, the hepatic mRNA levels of ET-1 and in its pro-fibrogenic ET A R receptors were induced with a reciprocal decrease in the anti-fibrogenic ET B R mRNA levels in Ldlr −/− VECadCre+Cc1 flfl livers compared to control mice ( Figure 7D). This could be partly caused by hyperactivation of NF-κB in endothelial cells in response to their loss of Ceacam1 gene [25,35].

Discussion
Dissecting out the role of altered metabolism [dyslipidemia and insulin resistance] from endothelial cell injury in the pathogenesis of atherosclerosis is challenging [36], in part owing to the causal role that dyslipidemia can play in endothelial cell injury [4]. The current studies revealed that feeding the insulin sensitive Ldlr −/− VECadCre+Cc1 fl/fl mice with a high cholesterol diet for 3-5 months caused hypercholesterolemia (mostly LDL-C) without insulin resistance, visceral obesity or increase in plasma NEFA and other dyslipidemia-related changes. Interestingly, no atherosclerotic plaques developed in Ldlr −/− VECadCre+Cc1 fl/fl mice, and their histological abnormalities were limited to inflammation and fibrosis in the aortae without fatty streaks. A similar phenotype developed in the livers of these mice in the absence of higher hepatic steatosis than that found in control mice. In contrast, hepatocyte-specific Ceacam1 deletion caused overt atheroma plaque formation and NASH in the presence of insulin resistance [21]. Together, this points to a key role for systemic insulin resistance and its associated metabolic derangements, such as visceral obesity, lipolysis and hepatic steatosis, in plaque formation in aortae.
Hypercholesterolemia in Ldlr −/− VECadCre+Cc1 fl/fl mice was marked by higher LDL-C levels likely resulting from reduced clearance instead of synthesis. Despite the adverse effect of CEACAM1 deletion on endothelial cell integrity and increased cell permeability in Ldlr −/− VECadCre+Cc1 fl/fl mice, consistent with previous reports [24], possible LDL-C trans-endothelial transport and redistribution to the subendothelial intima did not appear to occur to sufficient levels to result in fatty streaks. This provides in vivo evidence that fatty streaks do not advance to plaque formation in the absence of insulin resistance.
In addition to dyslipidemia, the pro-inflammatory state brought about by increased visceral obesity links endothelial cell injury and dysfunction to early atherosclerosis [4]. As with individual deletion of endothelial Ceacam1 [25], combined deletion with Ldlr caused NF-kB driven increases of inflammatory cytokines (such as TNFα and IL-6) in endothelial cells and release into the plasma without causing visceral obesity and insulin resistance (also marked by normal NEFA levels). The increase in endothelial TNFα may contribute to increased endothelial cell permeability [37] and disruption of the endothelial cell barrier [38] and subsequently, to the development of early atherosclerosis in Ldlr −/− VECadCre+Cc1 fl/fl mice.
Insulin resistance in endothelial cells plays an important role in atherosclerosis by reducing insulin signaling through the Akt/eNOS pathway [39]. This does not only reduce NO production and cause oxidative stress, but also reduces endothelin-1B receptor expression to mediate atherosclerosis [40]. Consistently, endothelial cell-specific Ceacam1 deletion reduced Akt/eNOS activation by VEGF-A, leading to reduced VEGFA-stimulated NO production, endothelin-1B receptor expression, redox imbalance and oxidative stress in double Ceacam1/Ldlr knockouts. As in response to insulin [25], a compromised Akt/eNOS pathway in endothelial cells was accompanied by an increased Shc/NF-kB-dependent pathway and VEGFA-stimulated endothelin-1 production. This induced aortic inflammation and fibrosis [in part by activating TGFβ pathways [41], consistent with endothelial gain-of-function of endothelin 1 [42] and reduced NO production [43].
In contrast to endothelial cells, deleting Ceacam1 in hepatocytes caused hyperinsulinemiadriven systemic insulin resistance and visceral obesity associated with features of nonalcoholic fatty liver disease and atherosclerosis when mice were backcrossed on the Ldlr null background and fed an atherogenic diet for 3 months [21]. These mice developed hypercholesterolemia due to increased cholesterol synthesis. Histopathologically, mice developed aortic plaques with lipid accumulation, inflammation and fibrosis [21]. The phenotype of mice with hepatocyte-specific deletion of Ceacam1 faithfully replicated the human disease in contrast to that of Ldlr −/− VECadCre+Cc1 fl/fl mice in which the metabolic derangement was restricted to elevation in total and LDL cholesterol without insulin resistance or visceral obesity. This points to targeting insulin resistance as an effective preventive/treatment measure against atherosclerosis. With insulin resistance constituting an independent predictor of endothelial dysfunction in individuals without classical risk factors for atherosclerosis [6], insulin sensitizers are increasingly being considered as a promising preventive approach [8].
Thus, endothelial cell-specific deletion of Ceacam1 on Ldlr null background caused hypercholesterolemia without insulin resistance. It also caused vascular inflammation and fibrosis without increased atheroma formation relative to controls. Mechanistically, this was underlined by reduced NO bioavailability and increased oxidative stress driven by blunted Akt/eNOS pathway in response to VEGF, as we previously reported for insulin [25]. This endothelial loss of Ceacam1 also caused hyperactivation of Shc/NF-kB signaling and its downstream pro-inflammatory and endothelin 1-dependent pro-fibrogenic pathways. Because CEACAM1 is a shared element between insulin and VEGF-A signaling, the data showed that altered signaling through CEACAM1-dependent pathways disrupts endothelial cells' response to insulin and VEGF-A, and in this cell-autonomous fashion, drives profound endothelial dysfunction, which in the absence of insulin resistance did not advance to overt atheroma formation.
In summary, the current studies showed that, in the absence of insulin resistance, hypercholesterolemia did not lead to overt atheroma formation or hepatic steatosis despite increased development of inflammation and fibrosis along the aortic wall and in the liver. This supports a critical role for hyperinsulinemia and insulin resistance in the pathogenesis of atherosclerosis and hepatic steatosis.
Research limitations and future perspectives: The current studies showed that despite increased LDL-C levels and its trans-endothelial transport through the endothelial cell that was rendered permeable by the loss of Ceacam1, fatty streaks did not develop on the aortic walls of Ldlr −/− VECadCre+Cc1 fl/fl mice. While it is possible that a much higher level of LDL-C is needed to cause fatty streaks in mice as compared to humans, these observations assign a stronger role for insulin sensitizers in the prevention of atherosclerosis. This would be consistent with outcome studies showing that lowering plasma cholesterol levels does not suffice to stop the progression of atherosclerosis in patients with metabolic disease [44]. It also supports the observation that pioglitazone, an insulin sensitizer and a PPARγ agonist, ameliorates atherosclerosis in at risk patients [8]. Whether this drug can prevent atherosclerosis by inducing Ceacam1 transcription [45] in endothelial cells and hepatocytes of patients with cardiovascular risk remains to be tested.
Animals were housed in a 12-h dark-light cycle and fed a standard regular chow diet until 6 months of age when male mice were fed ad libitum a high cholesterol atherogenic diet (HC) of 42% kcal from fat, 0.2% total cholesterol, and 2.7% kcal from carbohydrate (high sucrose 34% by weight) (Harlan Teklad, TD.88137, Haslett, MI, USA) for 3-5 months.
Body weight was assessed weekly. Of note, the age and the duration of feeding were optimized from several cohorts of mice that were fed HC diet for different periods of time and starting at different ages. Feeding HC for 3 months was found to be the shortest period to cause altered histology in the aortae of null mice. All procedures and animal experiments were approved by the Institutional Animal Care and Utilization Committee in all participating institutions.

En-Face Oil-Red-O Staining
As in [21], aortae were isolated starting from the aortic arch down to the femoral arteries and fixed in 10% formalin for 24 h. Longitudinally-opened aortae were cleaned from fat and surrounding connective tissues under a dissection microscope, rinsed with water and 60% isopropanol before undergoing staining with Oil Red-O (ORO) solution (Sigma-Aldrich) for 15 min, rinsed again, mounted to slide glass with lumen down, fixed with a cover slip and scanned with Olympus CKX41. Atherosclerotic lesions were defined as the percentage of total ORO-positive area/total surface area, determined by CellSens Standard software in area pixel 2 .

Intravital Microscopy of Leukocyte Adhesion in Carotid Artery
Following sedation with ketamine and xylazine (100/10 mg/kg), mice were fixed on a 15-cm cell-culture lid in a supine position. As previously described [47], the right jugular vein and left carotid artery were exposed and injected with 100 µL of 0.5 mg/mL rhodamine 6G (R4127; Sigma-Aldrich, St. Louis, MO, USA) to label cells, including leukocytes, with mitochondria. A U-shaped black plastic was placed under the carotid artery (~4-5 mm) to block background fluorescence during real life observation using an intravital microscope (Leica DM6 FS, Buffalo Grove, IL, USA). Video images were captured with a color digital camera (Teledyne QImaging, Surrey, BC, Canada) and analyzed offline for leukocyte adhesion. Cells that adhered to the vessel wall without rolling or moving for at least 3 s were counted over the vessel observed. Total numbers were used for statistical analysis.

Real-Time Quantitative RT-PCR
Total RNA was isolated from aortae with RNeasy fibrous tissue kit (74704, QIAGEN, Germantown, MD, USA). cDNA was synthesized by iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA), using 1 µg of total RNA and oligodT primers. qRT-PCR was performed using Fast SYBR Green Master Mix by the ABI StepOnePlus Real-Time PCR System (Applied Biosystems, Beverly, MA, USA), as described [46]. All primers (Table S1) were used at a final concentration of 10 µM. mRNA was normalized relative to ribosomal 18S.

Statistical Analysis
Data were analyzed by one-way analysis of variance (ANOVA) with Tukey's for multiple comparisons using GraphPad Prism 7 software. p < 0.05 was considered statistically significant.