Insulin Sensitivity Is Retained in Mice with Endothelial Loss of Carcinoembryonic Antigen Cell Adhesion Molecule 1

CEACAM1 regulates endothelial barrier integrity. Because insulin signaling in extrahepatic target tissues is regulated by insulin transport through the endothelium, we aimed at investigating the metabolic role of endothelial CEACAM1. To this end, we generated endothelial cell-specific Ceacam1 null mice (VECadCre+Cc1fl/fl) and carried out their metabolic phenotyping and mechanistic analysis by comparison to littermate controls. Hyperinsulinemic-euglycemic clamp analysis showed intact insulin sensitivity in VECadCre+Cc1fl/fl mice. This was associated with the absence of visceral obesity and lipolysis and normal levels of circulating non-esterified fatty acids, leptin, and adiponectin. Whereas the loss of endothelial Ceacam1 did not affect insulin-stimulated receptor phosphorylation, it reduced IRS-1/Akt/eNOS activation to lower nitric oxide production resulting from limited SHP2 sequestration. It also reduced Shc sequestration to activate NF-κB and increase the transcription of matrix metalloproteases, ultimately inducing plasma IL-6 and TNFα levels. Loss of endothelial Ceacam1 also induced the expression of the anti-inflammatory CEACAM1-4L variant in M2 macrophages in white adipose tissue. Together, this could cause endothelial barrier dysfunction and facilitate insulin transport, sustaining normal glucose homeostasis and retaining fat accumulation in adipocytes. The data assign a significant role for endothelial cell CEACAM1 in maintaining insulin sensitivity in peripheral extrahepatic target tissues.


Introduction
Carcinoembryonic antigen-related cell adhesion molecule-1 (CEACAM1) is a transmembrane glycoprotein that undergoes phosphorylation by insulin and vascular endothelial growth factor receptor-2 (VEGFR2) [1,2]. It is highly expressed in hepatocytes, where it promotes insulin clearance to sustain insulin sensitivity [3]. This is mediated by partaking in the insulin receptor internalization complex and increasing the rate of its cellular uptake.
Mice were housed in pathogen-free conditions and fed ad libitum a standard chow (Harlan Teklad 2016; Harlan, Haslett, MI, USA). The Animal Care and Utilization Committee of each participating institution approved all experiments.

Glucose and Insulin Tolerance Tests
Awake 6 h-fasted mice were injected intraperitoneally with 1.5 g/kg BW dextrose solution or 0.75 units/kg BW human regular insulin (Novo Nordisk, Princeton, NJ, USA) before their tail blood glucose was measured at 0-180 min post-injection [18].

Insulin Treatment and Media Analysis
Primary liver endothelial cells (LEC) were seeded into 6-well plates at a density of 3 × 10 5 cells/well in complete DMEM-F12 medium (Gibco) for 24 h. They were then serumstarved in phenol red-free medium (Invitrogen)-25 mM HEPES and 0.1% BSA for 2 h before treating with BSA-free insulin (100 nM in 25 mM HEPES) or vehicle at 37 • C for 5 min (for Western blot analysis of cell lysates) or for 20 min to analyze NO levels in 20 µL of media. Fluorescence was read using the Synergy H1 hybrid microplate reader (BioTek Instruments, Winooski, VT, USA) at 360 nm excitation and 430 nm emission wavelengths [21].

Semi-Quantitative Real-Time PCR (qRT-PCR) Analysis of mRNA
Total RNA was isolated from primary cells, livers, and skeletal muscle with a Nu-cleoSpin RNA Kit (740955.50, Macherey-Nagel, Bethlehem, PA, USA) and from visceral white adipose tissue (WAT) using TRIzol reagent (15596026, Ambion, Life Technologies, CA, USA). cDNA was synthesized using Superscript III (Bio-Rad, Hercules, CA, USA), and 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 [18]. All primers (Table S1) were used at a final concentration of 10 µM. mRNA was normalized relative to ribosomal 18S or 36B4 (for white adipose tissue). Values were expressed as mean ± SEM.

Statistical Analysis
Data were analyzed using one-way ANOVA analysis with post-hoc Tukey's test for multiple comparisons, or two-tailed Student's t-test, using GraphPad Prism6 software. Data were presented as mean ± SEM. p < 0.05 was considered statistically significant.

Results
3.1. Specific Deletion of Ceacam1 in Endothelial Cells of VECadCre+Cc1 fl/fl Mice mRNA levels were determined by qRT-PCR analysis that demonstrated almost complete deletion of Ceacam1 from endothelial cells derived from the hearts and livers of VECadCre+Cc1 fl/fl mice ( Figure 1A). In contrast, Ceacam1 mRNA was intact in hepatocytes, hepatic stellate cells (HSC), and bone marrow macrophages ( Figure 1A). Immunoblotting (Ib) cell lysates of liver endothelial cells (LEC) with CEACAM1 antibody (α-CC1) and normalizing against GAPDH to assess total protein loaded on the SDS-gel ( Figure 1B) confirmed the loss of CEACAM1 in mutant LEC as compared to wild-type controls ( Figure 1B(i)). 1:100 followed by Texas Red-conjugated goat anti-rat IgG antibody (1:1000; Invitrogen). For M1 macrophages, anti-CD11c Armenian hamster anti-mouse monoclonal antibody (1:100; Abcam) and the secondary was Dylight 649 goat anti-Armenian hamster antibody (Invitrogen) at 1:1000 dilution. For imaging, samples were placed on a coverslip and visualized using a Leica TCS SP5 laser-scanning microscope (Leica Microsystems, Buffalo Grove, IL, USA) [24].

Statistical Analysis
Data were analyzed using one-way ANOVA analysis with post-hoc Tukey's test for multiple comparisons, or two-tailed Student's t-test, using GraphPad Prism6 software. Data were presented as mean ± SEM. p < 0.05 was considered statistically significant.

Specific Deletion of Ceacam1 in Endothelial Cells of VECadCre+Cc1 fl/fl
Mice mRNA levels were determined by qRT-PCR analysis that demonstrated almost complete deletion of Ceacam1 from endothelial cells derived from the hearts and livers of VECadCre+Cc1 fl/fl mice ( Figure 1A). In contrast, Ceacam1 mRNA was intact in hepatocytes, hepatic stellate cells (HSC), and bone marrow macrophages ( Figure 1A). Immunoblotting (Ib) cell lysates of liver endothelial cells (LEC) with CEACAM1 antibody (α-CC1) and normalizing against GAPDH to assess total protein loaded on the SDS-gel ( Figure 1B) confirmed the loss of CEACAM1 in mutant LEC as compared to wild-type controls ( Figure  1B male mice at 2 months of age (n = 5/genotype), except for hepatic stellate cells that were derived from male mice at 8 months of age. Ceacam1 mRNA levels were analyzed by qRT-PCR in triplicate and normalized to 18S. Values are expressed as mean ± SEM. * p < 0.05 vs. all control groups; Negl, negligible. (B). Liver endothelial cells (LEC) combined from several wild-type (VECad-Cc1 +/+ ) and null (VECad+Cc1 fl/fl ) mice were analyzed by immunoblotting (Ib) with specific antibodies (α-) to detect specific proteins and normalized against levels of loaded proteins by immunoblotting parallel gels or lower half-gels with α-GAPDH or α-Tubulin. The apparent molecular mass (kDa) is indicated at the right-hand side of each gel. Gels represent two separate/repeated experiments.

VECadCre+Cc1 fl/fl Mice Display Insulin Sensitivity
Compared to control littermates, VECadCre+Cc1 fl/fl nulls exhibited normal body weight for up to 12 months of age ( Figure 2A, data of 8-month-old mice are shown). This was associated with normal visceral adipose mass ( Figure 2B). Consistently, plasma levels of NEFA, adiponectin, and leptin were also normal in nulls relative to control littermates ( Figure 2C-E, respectively). In addition to normal plasma NEFA and adiponectin, null mice exhibited fed (and fasting) normoglycemia ( Figure 2I-J), suggesting insulin sensitivity. Consistently, null mice remained tolerant to exogenous insulin and glucose until 12 months of age ( Figure  3A and 3B, respectively). A 2h-hyperinsulinemic/euglycemic ( Figure 3C(i-ii)) clamp analysis was performed on 7-month-old mice. The glucose infusion rate required to maintain euglycemia (iii), whole-body glycogen synthesis (iv), and glycolysis (v) were normal in Null mice displayed normal steady-state plasma insulin ( Figure 2F) and C-peptide ( Figure 2G) levels. This was associated with normal insulin clearance (calculated as steadystate C-peptide/insulin molar ratio) ( Figure 2H), consistent with the intact expression of CEACAM1 in hepatocytes, the major site of insulin clearance and home for the highest level of CEACAM1 expression [3].
In addition to normal plasma NEFA and adiponectin, null mice exhibited fed (and fasting) normoglycemia ( Figure 2I,J), suggesting insulin sensitivity. Consistently, null mice remained tolerant to exogenous insulin and glucose until 12 months of age ( Figure 3A,B, respectively). A 2 h-hyperinsulinemic/euglycemic ( Figure 3C(i,ii)) clamp analysis was performed on 7-month-old mice. The glucose infusion rate required to maintain euglycemia (iii), whole-body glycogen synthesis (iv), and glycolysis (v) were normal in null mice relative to wild-type controls. The ability of insulin to suppress hepatic glucose production (vi,vii) and induce whole-body glucose turnover (Rd) (viii) and glucose uptake in the gastrocnemius muscle (ix) and in white and brown adipose tissue (x,xi) were all intact in VECadCre+Cc1 fl/fl relative to VECadCre-Cc1 +/+ controls.  At the cellular level, Western blot analysis revealed that insulin induced IR beta (IRβ) phosphorylation to the same extent in mutant and wild-type LEC, as indicated by immunoblotting cell lysates with antibodies against phosphorylated IRβ (α-pIRβ) antibody normalized to IRβ ( Figure 4A). Like IRβ, insulin receptor alpha (IRα) protein level was intact in LEC from null mice ( Figure 1B(ii)).

Increased NF-κB-Mediated Systemic Inflammation in VECadCre+Cc1 fl/fl Mice
Null LEC (Figure 4D), as well as endothelial cells isolated from heart (not shown), exhibited a higher basal and insulin-stimulated NF-κB activation (phosphorylation) than wild-type LEC. This could result from the higher activation of MAP kinase both basally and in response to insulin ( Figure 4C), which stemmed in turn, from lower Shc sequestration in the absence of CEACAM1 [25] and the reciprocal increase in its binding to IRβ as . NF-κB insulin signaling pathway in liver endothelial cells. Primary liver endothelial cells (LEC) from 2-month-old male WT (VECad−Cc1 +/+ ) and null (VECad+Cc1 fl/fl ) mice were treated with (Ins, 100 nM) or without insulin for 5 min before their lysates were subjected to immunblotting (Ib) with antibody against (A) phospho-IRβ (α-pIRβ), (C) phopsho-MAP kinase (α-pMAPK) and (D) phopho-NF-κB (α-pNF-κB) and in parallel gels, with their specific antibodies for normalization. (B) lysates were subjected to immunoprecipitation (Ip) with Shc antibody followed by immunoblotting (Ib) with antibodies against phospho-CEACAM1 (pCC1) and IRβ or Shc to assess the amount of pCC1 and pIRβ in the Shc immunopellet. Gels represent two separate experiments.

Increased NF-κB-Mediated Systemic Inflammation in VECadCre+Cc1 fl/fl Mice
Null LEC (Figure 4D), as well as endothelial cells isolated from heart (not shown), exhibited a higher basal and insulin-stimulated NF-κB activation (phosphorylation) than wild-type LEC. This could result from the higher activation of MAP kinase both basally and in response to insulin ( Figure 4C), which stemmed in turn, from lower Shc sequestration in the absence of CEACAM1 [25] and the reciprocal increase in its binding to IRβ as co-immunoprecipitation experiments revealed ( Figure 4B).
NF-κB could drive the transcription and release of pro-inflammatory cytokines from LEC to contribute to elevated plasma IL-6 and TNFα levels in null mice starting at 8 months of age ( Figure 2K,L, respectively). NF-κB could also induce the transcription of VCAM1 [26,27] to elevate its mRNA (by~4-fold) ( Table 1) and protein levels ( Figure 1B(iii)), as well as the mRNA of ICAM1 and P-Selectin [28] (by~3-5-fold) ( Table 1); all being vascular mediators of leukocyte-endothelial adhesion. The increase in VCAM1 protein level in VECadherin+Cc1 fl/fl LEC is consistent with our previous findings in bovine aortic endothelial cells (BAEC) with siRNA-mediated downregulation of Ceacam1 [16]. Consistent with the inducing effect of NF-κB on the transcription of matrix metalloprotease (Mmp) 2 and 9 [29], mRNA levels were increased by~2-fold in the heart and skeletal muscle of VECadherin+Cc1 fl/fl relative to their three controls ( Table 2). This could lower the mRNA of tight junctions and adherent molecules such as ZO-1/2, claudin 1/3, and occludin in null mice ( Table 2) to contribute to endothelial barrier dysfunction in these tissues [30].

Compromised Endothelial Barrier and Lower Nitric Oxide Production in VECadCre+Cc1 fl/fl Mice
Consistent with CEACAM1 regulating endothelial barrier [15], the mRNA levels of β-catenin and other genes involved in adhesion and tight junctions (Vecadherin, occludin, ZO-1, claudin-1, and -5) critical to cell-to-cell contact, were all reduced by~2-to-3-fold in null LEC (Table 1). VECadherin protein levels were lower in LEC ( Figure 1B(iv)) from VECadCre+Cc1 fl/fl mice. This could be partly caused by the rise in ADAM10 ( Figure 1B(v)), a metalloprotease that cleaves VECadherin to promote vascular permeability and leukocyte trans-endothelial migration [31]. In addition, VEGFR2 protein levels were markedly reduced in null LEC ( Figure 1B(vi)). qRT-PCR analysis showed~three-to-five-fold lower levels of mRNA of Vegf-A and its receptor (Vegfr2) in addition to Vegfr1 in null LEC (Table 1). In contrast, mRNA levels of Vegf-C and Vegf-D that are mainly involved in lymphatic vessel formation were unaltered ( Table 1). The differential downregulatory effect on Vegf-A transcription could be due, at least partly, to compromised Akt/eNOS signaling ( Figure 5A(iii,iv)), leading to lower NO production and release in the media of LEC both basally and in response to insulin ( Figure 5B), and to reduced β-catenin mRNA levels in LEC cells (Table 1) [32]. This would contribute to the significant drop in plasma NO levels in VECadCre+Cc1 fl/fl mice beginning at 6 months of age ( Figure 5C). Reduced Akt/eNOS phosphorylation could result from limited SHP2 phosphatase sequestration in the absence of CEACAM1 [16] and its reciprocal targeting to IRS-1, as co-immunoprecipitation analysis showed ( Figure 5A(i,ii)). were treated with (Ins, 100 nM) or without insulin for 5 min before proteins were lyzed and (i) subjected to immunoprecipitation (Ip) with SHP2 antibody followed by immunoblotting (Ib) with antibodies against phospho-CEACAM1 (pCC1) and CEACAM1 (CC1) or SHP2 to assess the amount of CC1 and pCC1 in the SHP2 immunopellet. (ii) Similar co-immunoprecipitation experiments were performed to evaluate the association between IRS1 with SHP2. (iii,iv) Lysates were also subjected to immunoblotting with α-phospho-antibodies to account for activation of Akt and eNOS, respectively, as in the legend in Figure 4.  Figure 5. Insulin signaling leading to nitric oxide production in liver endothelial cells. (A) Primary liver endothelial cells (LEC) from 2-month-old WT (VECad−Cc1 +/+ ) and null (VECad+Cc1 fl/fl ) mice were treated with (Ins, 100 nM) or without insulin for 5 min before proteins were lyzed and (i) subjected to immunoprecipitation (Ip) with SHP2 antibody followed by immunoblotting (Ib) with antibodies against phospho-CEACAM1 (pCC1) and CEACAM1 (CC1) or SHP2 to assess the amount of CC1 and pCC1 in the SHP2 immunopellet. (ii) Similar co-immunoprecipitation experiments were performed to evaluate the association between IRS1 with SHP2. (iii,iv) Lysates were also subjected to immunoblotting with α-phospho-antibodies to account for activation of Akt and eNOS, respectively, as in the legend in Figure 4. Gels represent two separate experiments. (B) levels of nitric oxide (NO) were determined in triplicate in the media of cells treated with or without insulin for 20 min. Values are expressed as mean ± SEM. * p < 0.05 vs. no insulin/same genotype; † p < 0.05 vs. VECadCre−Cc1 +/+ no insulin; ‡ p < 0.05 vs. VECadCre−Cc1 +/+ plus insulin.(C) Male mice (3-8 months of age, n ≥ 5/genotype) were fasted overnight before blood was drawn, and plasma was processed to assess NO levels. Values are expressed as mean ± SEM. * p < 0.05 vs. VECadCre−Cc1 +/+ , † p < 0.05 vs. VECadCre+Cc1 +/+ , ‡ p < 0.05 vs. VECadCre−Cc1 fl/fl .

Increased Fat Accumulation and Insulin Sensitivity in Adipocytes of VECadCre+Cc1 fl/fl Mice
Histological analysis of H&E-stained sections from WAT showed a significant expansion of adipocytes in 12-month-old VECadCre+Cc1 fl/fl mice relative to their age-matched wild-type counterparts ( Figure 6A). Immunofluorescence (IF) analysis with anti-F4/80, a macrophage marker, showed multiple areas with crown-like structures (CLS)-containing macrophages (green) surrounding adipocytes (gray) with small pieces of lipid inside the macrophages in wild-type mice ( Figure 6B, yellow arrows). In contrast, VECadCre+Cc1 fl/fl mice exhibited minimal CLS formation with no evidence of lipids inside the macrophages in these structures. Histological analysis of H&E-stained sections from WAT showed a significant expansion of adipocytes in 12-month-old VECadCre+Cc1 fl/fl mice relative to their age-matched wild-type counterparts ( Figure 6A). Immunofluorescence (IF) analysis with anti-F4/80, a macrophage marker, showed multiple areas with crown-like structures (CLS)-containing macrophages (green) surrounding adipocytes (gray) with small pieces of lipid inside the macrophages in wild-type mice ( Figure 6B, yellow arrows). In contrast, VECadCre+Cc1 fl/fl mice exhibited minimal CLS formation with no evidence of lipids inside the macrophages in these structures.  Given that TNFα can induce CEACAM1 transcription via activating NF-κB [33], we then examined CEACAM1 mRNA and protein levels in WAT. As expected from the low expression of CEACAM1 in WAT under physiological conditions [34], Western blot analysis did not detect CEACAM1 protein in wild-type WAT lysates ( Figure 6D). In contrast, WAT from VECadCre+Cc1 fl/fl mice exhibited a substantial increase in CEACAM1 protein content ( Figure 6D). This appears to be due to a preferential~2-fold increase in the mRNA of Ceacam1-4L, but not its alternative spliced variant, Ceacam1-4S [35], as expected from elevated levels of interferon response factor (Irf)-1 [36] and Irf-3 [37] in the WAT of null mice relative to their controls (Table 2). Consistently, IF analysis of CEACAM1 revealed a 2.5-fold increase in CEACAM1 expression (green) in VECadCre+Cc1 fl/fl than wild-type WAT, and its parallel increase in its co-localization with the anti-inflammatory M2 macrophages (red), as the overlay revealed ( Figure 6E). In contrast, wild-type WAT showed more of the proinflammatory M1 macrophages (magenta) than the null ( Figure 6E). This is consistent with the anti-inflammatory function of CEACAM1-4L that is mediated by two immunoreceptor tyrosine-based inhibitory motifs in its cytoplasmic tail [38]. Accordingly, the mRNA levels of the anti-inflammatory IL-10 cytokine were elevated in null WAT (Table 2) without any change in the expression of pro-inflammatory cytokines such as IL-6, TNFα, IL-1β, and IFNγ, and in the T regulatory FoxP3 in null by comparison to wild-type WAT ( Table 2). Consistent with normal TNFα levels, Smad7 expression was intact in null WAT, likely suggesting an unaltered TGFβ signaling pathway, which could, in turn, maintain normal mRNA levels of pro-fibrogenic genes, such as α-Sma, Col1α1, and Col6α3 in WAT of null mice (Table 2). Furthermore, Trichome staining revealed a modest decline in collagen deposition in WAT of VECadCre+Cc1 fl/fl relative to wild-type mice ( Figure 6C).

Discussion
In hepatocytes, CEACAM1 promotes insulin sensitivity by increasing the rate of receptor-mediated insulin uptake followed by degradation [4]. Accordingly, liver-specific and global Cc1 −/− null mice developed chronic hyperinsulinemia and insulin resistance emanating from impaired insulin clearance. Cc1 −/− also manifested endothelial and cardiovascular dysfunction and leukocyte-endothelial adherence [7]. Exclusive reconstitution of CEACAM1 in hepatocytes of Cc1 −/− mice reversed these metabolic [8] and cardiovascular abnormalities, including restoration of NO bioavailability and stemming endothelialleukocyte adhesion [7]. This emphasized the significant contribution of hyperinsulinemiadriven insulin resistance in endothelial and cardiovascular dysfunction.
In contrast to liver-specific Ceacam1 deletion or inactivation, the current studies showed that endothelial loss of Ceacam1 did not alter insulin metabolism or sensitivity, as buttressed by normoinsulinemia and normoglycemia (both fed and fasting) in VECadherin+Cc1 fl/fl mice. Consistently, insulin receptor expression and phosphorylation in isolated liver endothelial cells were not impacted by the loss of endothelial CEACAM1. This is in contrast to the remarkable hyperinsulinemia-driven downregulation of insulin receptors in the hepatocytes of liver-specific inactive CEACAM1 mutants, as analyzed by Scatchard plot [6]. Normal insulin metabolism in VECadherin+Cc1 fl/fl mice is consistent with: (1) the rapid and passive transport of secreted insulin from the portal vein to hepatocytes through fenestrae in the liver endothelium to be targeted to degradation, (2) intact CEACAM1 in hepatocytes, main sites of insulin extraction and (3) insulin is not significantly degraded in endothelial cells [11].
Ablating insulin receptors in endothelial cells adversely affected PI3-kinase-dependent insulin signaling pathways without altering glucose homeostasis under normal feeding conditions [39]. Whereas deleting endothelial Ceacam1 did not affect insulin-stimulated insulin receptor phosphorylation in isolated liver endothelial cells, it deactivated the IRS-1/Akt/eNOS pathway to reduce eNOS-mediated NO production, a hallmark of endothelial dysfunction, as was manifested in global Cc1 −/− nulls by reduced endothelium-dependent relaxation in aortae [16] and microvasculature [7]. As previously shown in BAEC with siRNA-mediated downregulation of Ceacam1 [16] and in the heart and aortae of Cc1 −/− mice [7], lowered eNOS activity was mediated by reduced sequestration of SHP2 phosphatase and its reciprocal increased binding to IRS-1. Recent studies in isolated myocardial cells from Cc1 −/− mice showed that compromised eNOS activity was also mediated by its depalmitoylation and increased caveolae-mediated redistribution to the perinuclear region [13]. Lower NO production would be expected to cause vasoconstriction, as in global Cc1 −/− mice [7], which would, in turn, limit blood flow and insulin delivery to skeletal muscle where insulin transport into the interstitial space is tightly regulated by the endothelial lining of its vasculature [40]. However, hyperinsulinemic-euglycemic clamp analysis showed normal insulin-stimulated glucose uptake in skeletal muscle (and adipose tissue) of VECadherin+Cc1 fl/fl mice. Thus, it is likely that counterregulatory mechanisms developed to limit the potential adverse effect of vasoconstriction on insulin delivery to their extrahepatic target cells. Insulin is transported to myocytes either by a transcytosis mechanism inside the endothelial cell, or via a paracellular route through pores between adjacent endothelial cells [9]. Whether the loss of endothelial CEACAM1 impacted transcellular insulin access to the interstitial space, as expected from the loss of NO production [41] in endothelial cells isolated from VECadherin+Cc1 fl/fl mice, remains to be determined. However, like myocardial endothelial cells of Cc1 −/− mice that exhibited destabilization of VECadherin/β-catenin complexes at the adherent junctions [13], VECadherin+Cc1 fl/fl endothelial cells displayed a loss of VECadherin and β-catenin relative to wild-type mice. This is consistent with increased endothelial barrier dysfunction and vascular leakiness, which is expected to be intensified by the simultaneous rise of the inflammatory microenvironment caused by NF-κB-dependent production of TNFα in VECadherin+Cc1 fl/fl mice [42]. Moreover, the loss of endothelial CEACAM1 yielded an increase in Shc binding to IRβ and activation of its downstream NF-κB pathway that would in turn, drive the transcription and release of MMP2 and 9 [29] in the heart and skeletal muscle of VECadherin+Cc1 fl/fl relative to their three controls. This could lower the expression of tight junctions and adherent molecules (i.e., ZO-1/2, claudin 1/3 and occludin) in null mice to contribute to endothelial barrier dysfunction in extrahepatic tissues [30]. Together, the data suggest that like endothelial cells from global Cc1 −/− mice [13], VECadherin+Cc1 fl/fl developed vascular permeability that would facilitate paracellular transport of insulin (~6 kDa) in extrahepatic peripheral tissues to support their normal glucose homeostasis.
VECadherin+Cc1 fl/fl mice displayed an expansion of adipocytes to accommodate the storage of extra fat that is likely transported through the more permeable endothelial cell barrier [43,44]. However, these mice did not display a significant increase in visceral obesity, in association with normal plasma NEFA, adiponectin, and leptin levels. Like skeletal muscle, insulin-stimulated glucose uptake in white and brown adipose tissue was intact in VECadherin+Cc1 fl/fl mice, as opposed to global Cc1 −/− mice. This is consistent with the rise in plasma TNFα and IL-6 emanating mostly from NF-κB activation in endothelial cells without involving adipokines released from adipose tissue macrophages that are typically associated with systemic insulin resistance and visceral obesity [45]. Normal plasma NEFA may have played an important part in insulin sensitivity, as supported by studies showing that treatment with acipimox, a lipolysis inhibitor, restored insulin's suppression of endogenous glucose production without altering adipokines (IL-6 and TNFα) or adiponectin levels in subjects with a strong family history of type 2 diabetes [46]. The rise in plasma TNFα may have served to preferentially induce the expression of the Ceacam1-4L variant through activating NF-κB in the anti-inflammatory M2 macrophages of VECadherin+Cc1 fl/fl to enhance innate immunity and limit systemic inflammation that would otherwise drive insulin resistance. The resultant low-level inflammation in the adipocytic microenvironment of VECadherin+Cc1 fl/fl would limit fibrosis to bestow plasticity and enhance the expansion of adipocytes to store excess fat [47]. While we have observed that overexpression of CEACAM1 in hepatocytes limits fibrosis in adipose tissue [24], the current studies provide a first in vivo demonstration of the anti-fibrogenic role of CEACAM1-4L in M2 macrophages.

Conclusions
In summary, deleting Ceacam1 in endothelial cells did not adversely affect systemic insulin sensitivity despite modulating post-receptor signaling to reduce NO production in addition to stimulating an NF-κB-dependent pro-inflammatory microenvironment in the endothelium. Sustained insulin sensitivity is in part due to intact insulin metabolism and increased fat accumulation in adipocytes. The studies provide an in vivo demonstration of a distinct endothelial cell CEACAM1-dependent pathway playing a role in regulating insulin sensitivity in extrahepatic cells independent of its role in insulin clearance in hepatocytes.

Supplementary Materials:
The following supporting information is available online at https:// www.mdpi.com/article/10.3390/cells10082093/s1, Figure S1: Mice genotyping, Table S1: Real-time PCR primer sequences from mouse genes. contributed to data analysis, assembled and organized the figures, and contributed to drafting the manuscript. J.K.K. contributed to data analysis and to the editing of the manuscript. S.M.N. oversaw the work, including its conception and study design, analyzed data, led scientific discussions, and reviewed/edited the manuscript. All authors have read and agreed to the published version of the manuscript.