Pro-Inflammatory Endothelial Activation Detected by Molecular Imaging in Obese Non-Human Primates Coincides with the Onset of Insulin Resistance and Progressively Increases with Duration of Insulin Resistance

Abstract

Background

Inflammation and insulin resistance (IR) are associated processes that potentiate risk for cardiovascular disease in obesity. The temporal relation between IR and inflammation is not completely characterized. We hypothesized that endothelial cell adhesion molecule (ECAM) expression in large arteries is an early event that coincides with diet-induced obesity and IR in primates.

Methods and Results

Ten adult male rhesus macaques were studied at baseline and every 4-6 months on high-fat diet (HFD) for 2 years. Truncal fat, carotid intima-media thickness (IMT), plasma inflammatory biomarkers, and carotid P-selectin and VCAM-1 expression by contrast-enhanced ultrasound molecular imaging were assessed. Intravenous glucose tolerance test (IVGTT) was performed at baseline, 4 and 18 months. HFD produced a rapid increase (p<0.01) in weight, truncal fat, and degree of IR indicated by the insulin area-under-the-curve and glucose disappearance rate on IVGTT; all of which worsened minimally thereafter. Molecular imaging detected a progressive increase in ECAM expression over time (5-7-fold greater than control agent signal at 2 yrs, p<0.01). Changes in IMT were not detected until 2 years and, while there was a trend toward an increase in plasma markers of inflammation (MCP-1, CRP), the pattern of increase varied considerably over time.

Conclusions

In primates with diet-induced obesity, endothelial inflammatory activation is an early event that occurs coincident with the development of IR and long before any measurable change carotid IMT. Endothelial activation is more related to the duration rather than severity of IR and is not mirrored by changes in plasma biomarkers.

Systemic inflammation and insulin resistance (IR) are mutually amplifying processes that together increase risk for cardiovascular disease and the development of type 2 diabetes mellitus in obese individuals.^{1, 2} On one hand, pro-inflammatory cytokines produced by adipocytes and resident macrophages within inflamed adipose tissue contribute to IR in part by interfering with normal insulin receptor intracellular signaling.^{3, 4} Conversely, obesity-related IR promotes inflammation through multiple pathways including an increase in oxidative stress and the production of pro-inflammatory adipokines, advanced glycation end products (AGEs), and free fatty acids.^5-8

Activation of nuclear factor κB (NFκB) is one of the master pathways that link obesity, IR, and inflammation.⁴ Activation of NFκB leads to transcription and expression of endothelial cell adhesion molecules (ECAMs), which represents an important mechanism by which obesity increases risk for atherosclerotic disease.⁹ In mice fed a high fat diet (HFD), NFκB activation, elevation of mRNA for ECAMs, and dysregulation of nitric oxide (NO) occurs approximately at the same time that obesity develops but prior to any detectable increase in plasma biomarkers for inflammation.¹⁰ There is even evidence that vascular IR, measured by phospho-IκBα, precedes the development of IR in the liver and adipose tissue.¹⁰ However, the temporal onset of vascular inflammatory activation, endothelial dysfunction, and circulating biomarkers in humans or larger mammalian models of obesity and IR has not been characterized.

The aim of this study was to temporally evaluate pathologic changes in endothelial phenotype (ECAM expression, NO bioactivity) that contribute to atherosclerosis using non-invasive molecular imaging and metabolic profiling in a primate model of diet-induced obesity (DIO). We hypothesized that endothelial inflammatory changes would occur coincident with the development of obesity and IR, and prior to changes in circulating markers of inflammation and carotid intimal thickening. We also sought to determine whether the degree of ECAM expression correlated with either the degree or severity of IR.

METHODS

Study Design

The study was approved by the Animal Care and Use Committee of the Oregon National Primate Research Center and conformed to USDA and AAALAC guidelines for non-human primate care. Ten adult male rhesus macaques (Macaca mulatta) 9-11 yrs of age at baseline were studied. At baseline, animals were fed a standard primate chow diet consisting of 58% carbohydrates, 27% protein, and 15% fat by caloric content. Animals were studied at baseline and at 4, 8, 12, 18, and 24 months after starting a HFD consisting of 46% carbohydrates, 18% protein, and 36% fat by caloric content supplemented with a 100 g fructose drink four times weekly. At each study interval, animals underwent a 3-day evaluation that included (1) dual X-ray absorptiometry (DEXA) for truncal fat quantification, (2) contrast-enhanced ultrasound molecular imaging of carotid artery ECAM expression, (3) high-frequency ultrasound for carotid intima-media thickness (IMT) and arterial compliance, (4) ultrasound of the brachial artery with flow-mediated vasodilation (FMD) for NO bioactivity, and (5) venous blood sampling for inflammatory biomarkers and plasma lipids. Intravenous glucose tolerance test (IVGTT) was performed at baseline and at 4 and 18 months after HFD. Three non-obese control primates that were age-matched (12 yrs) to the final imaging period were studied by molecular imaging. These animals were fed chow diet and had identical living environments to obese macaques. Anesthesia was induced with ketamine (10mg/kg I.M.) and maintained with isoflurane (1.0-1.5%) for all studies except for IVGTT and DEXA during which telazol (5mg/kg I.M.) was used.

Intravenous Glucose Tolerance Test

After an overnight fast, animals received dextrose (600 mg/kg, I.V.) over 1 min. Venous blood samples were obtained at baseline and at 1, 3, 5, 10, 20, 40, and 60 min after dextrose injection for measurement of blood glucose and plasma insulin concentration by radioimmunoassay. Time-concentration curves were plotted to derive the insulin area-under-the-curve (AUC). Glucose disappearance rate (K_G) was calculated by: K_G = Ln (20min glucose) – Ln (5min glucose)/15 min × 100%. ^{11, 12}

Body Composition

Body composition was assessed by dual X-ray absorptiometry (DEXA) (Discovery A, Hologic Inc.) to obtain muscle mass and fat mass. Percent truncal fat was calculated by dividing the truncal fat mass by the total truncal mass.

Carotid Intima-medial Thickness and Compliance

Two-dimensional long-axis ultrasound imaging of the distal common carotid artery was performed with a linear array transducer (15L8, Sequoia, Siemens Medical Systems, Mountain View, CA) at 10MHz using a mechanical index 1.3 and a dynamic range of 65 dB. The average far-wall intima-medial thickness (IMT) at end-diastole was measured at 9 separate sites within 1 cm proximal to the carotid bifurcation. M-mode ultrasound was used to measure the diastolic (D_D) and systolic diameter (S_D) and was averaged over 3 cardiac cycles. The following vascular mechanical properties were determined; vascular strain, calculated by: (S_D–D_D)/D_D and the elastic modulus (EM), calculated by: ([P_S – P_D]/ [S_D – D_D]/D_D) × 0.5D_D/IMT where P_S and P_D are systolic and diastolic blood pressure.

Carotid Molecular Imaging

For contrast enhanced ultrasound (CEU) molecular imaging, microbubbles targeted to P-selectin (MB_P), VCAM-1 (MB_V), or control microbubbles (MB_C) were prepared by conjugating biotinylated anti-human P-selectin mAb (AK4, Biolegend), anti-human VCAM-1 (1.G11B1, ABD Serotec), or IgG1 isotype control (BD biosciences) to the surface of lipid-shelled decafluorobutane microbubbles containing a bi-functional molecular spacer with a membrane anchor and polyethyleneglycol-streptavidin moiety (Micromarker 2, VisualSonics Inc., Toronto, Canada). Longitudinal-axis imaging of the distal common carotid artery and bifurcation was performed using multi-pulse amplitude-modulation imaging at 7MHz, a mechanical index of 0.25, and a dynamic range of 55dB. Imaging was performed 5 min after the intravenous injection of 1×10⁸ MB_C, MB_P, or MB_V, performed in a random order. Several frames were digitally averaged. Signal from any remaining freely circulating microbubbles was eliminated by digital subtraction of several averaged frames obtained after a high-power (mechanical index 1.9) pulse sequence.¹³ Regions-of-interest were placed on the near and far wall of the common carotid artery with the former extending into the proximal portion of the internal carotid. Data were averaged for the right and left carotid artery.

Flow Mediated Vasodilation

Ultrasound imaging (10 MHz) of the brachial artery in its long-axis was performed to measure diameter and centerline average peak velocity using angle-corrected pulsed-wave Doppler. Data were acquired at baseline and then every 30 s for 3 min after release of a forearm cuff inflated to 60 mm Hg above systolic pressure for 5 min. Brachial artery diameter was measured by averaging nine separate end-diastolic measurements over three cardiac cycles. Data were expressed as absolute change in diameter and percent change from baseline. Shear stress was calculated immediately after cuff deflation as 8 ×μ × V/D where μ is the viscosity of blood and assumed to be 0.035dyne × s/cm², V is Doppler peak average velocity, and D is vessel diameter at baseline. ¹⁴

Serum Lipids and Biomarkers

Venous blood samples were used to measure plasma lipids, hemoglobin-A1, and inflammatory biomarker profiles using a multi-analyte immnunoassay (Human MAP V.1.6 Rules Based Medicine, Inc.).

Targeting Ligand Cross-reactivity With Macaque Epitopes

The cross reactivity of the anti-human P-selectin mAb used for microbubble targeting was tested by assessing immunofluorescence on activated platelets. Heparanized blood samples from normal rhesus macaques were obtained. Immunostaining of platelets, which were aggregated through vortex shear, was performed using the AK4 primary mAb and FITC-labeled polyclonal rat anti-mouse secondary Ab (BD Biosciences) (5 μg/mL). Control experiments were performed with secondary Ab alone. Cross reactivity of the VCAM-1 mAb was performed by immunohistochemistry of tissue bank specimens of spleen and carotid artery from rhesus macaques fed a HFD for 2 yrs. Sections (5 μm) were fixed in formalin and paraffin embedded. After antigen retrieval, endogenous peroxidases were blocked for 15 minutes in 2% H₂0₂ in .05M potassium phosphate buffered saline (KPBS). Sections were then treated with Avidin/Biotin blocking reagents according to kit directions (Vector Labs Cat# SP2001), prior to blocking for 30 minutes in KPBS containing 0.4% Triton-X and 2% normal donkey serum. Primary staining was performed with the biotinylated mouse-anti-VCAM-1 mAb (1:20 dilution), rinsed, and stained using Vectastain Elite ABC kit (Vector Labs Cat# PK-6100) and a DAB Peroxidase Substrate kit (Vector Labs Cat# SK-4100). Sections were counterstained with hematoxylin.

Statistical Analysis

Data were analyzed on Prism, v5.0 and are expressed as ± SD unless stated otherwise. D'Agostino and Pearson omnibus test were used to assess data normality. Changes from baseline through18 months were analyzed with one-way repeated measures ANOVA with a Bonferroni's post-test (parametric), a Friedman test with a Dunn's post-test (non-parametric) for multiple comparisons, or a test for linear trend. Due to a loss in follow-up of 2 animals after the 18 month time point, 18 and 24 month differences were assessed with either a student's paired t-test or a Wilcoxon matched-pairs signed rank test. IVGTT insulin AUC data was natural log transformed and fit to a quadratic model and K_G was fit to a cubic model to assess changes across follow-up. Correlation of parametric data was assessed by a Pearson's product and non-parametric data with a Spearman's test. Best-fit linear regression analysis was performed on data when significant correlations were observed. Only correlations with a significant linear association are reported.

RESULTS

Diet-induced Changes in Morphometry, Lipids, and Insulin Sensitivity

All animals were studied at every pre-defined interval except for two animals that were not available for study at the final 24 month time point. The mean caloric intake increased by 68% from 711±167 kcal/day on baseline chow diet to 1195±67 kcal/day when measured 4 months after starting HFD.

Total body mass and truncal fat on DEXA both increased significantly over the first 4 months after animals were started on HFD (Table 1). Over the remainder of the study period there were only small further increases in mass and truncal fat. After 4 months of HFD there were significant increases in serum cholesterol that remained relatively stable thereafter while a peak and fall was noted in serum triglycerides (Table 1). On IVGTT there was evidence for the development of IR after 4 months of HFD. This early onset of IR was reflected by an increase in the insulin AUC and a decrease in the glucose disappearance rate (K_G) with little change in these parameters at 18 months (Table 1). It was not until 18 months that the gradual rise in glycosylated hemoglobin was significantly elevated beyond baseline values (Table 1).

Table 1.

Body Morphometries, Lipid Profiles, and Metabolic Changes to HFD

	Baseline N= 10	4 months N= 10	8 months N= 10	12 months N= 10	18 months N= 10	24 months N= 8	ANOVA, Freidman, or t-test
Weight, Kg	11.0 ± 1.9	13.2 ± 2.6*	13.5 ± 2.7*	13.7 ± 2.9*	14.5 ± 2.6*†	14.1 ± 2.3	p < 0.0001
Truncal Fat, %	19.2 ± 9.4	34.0 ± 10.7*	34.5 ± 10.8*	36.6 ± 11.0*	40.4 ± 9.8*†	38.3 ± 8.3	p < 0.0001
Total Cholesterol, mg/dL	134 ± 21	192 ± 31*	195 ± 39*	184 ± 51*	185 ± 46*	188 ± 42	p < 0.0001
LDL Cholesterol, mg/dL	58 ± 11	82 ± 16‡	87 ± 20‡	67 ± 25	71 ± 24	71 ± 23	p = 0.005
HDL Cholesterol, mg/dL	65 ± 15	93 ± 25*	91 ± 31*	88 ± 37*	91 ± 24*	97 ± 30	p = 0.0005
Triglycerides, mg/dL	42 ± 22	70 ± 56	53 ± 41	121 ± 97*§	76 ± 66	81 ± 99	p = 0.004
Ln Insulin AUC, μg/mL χ min	8.20 ± 0.59	8.53 ± 0.68#	-------	-------	8.67 ± 0.67	-------	p = 0.04
KG (%/min)	3.16 ± 1.24	2.40 ± 0.81#	-------	-------	2.86 ± 0.72	-------	p = 0.02
Hbalc, %	6.2 ± 1.0	6.3 ± 0.4	6.5 ± 0.4	6.6 ± 0.8	6.7 ± 0.2‡	6.6 ± 0.2	p = 0.009

Data are mean (±SD).

^*Repeated measures one-way ANOVA with Bonferroni's post-test compared to baseline

^†compared to 4 months

^§compared to 8 months.

^‡Freidman test with Dunn's post-test compared to baseline.

^#Student's paired t-test compared to baseline.

Hemodynamics, Vascular Function and Arterial Remodeling

Systolic and diastolic arterial blood pressure increased significantly in the first four months and remained elevated over the remaining study period while pulse pressure did not change (Table 2). There were also mild progressive increases in brachial and carotid artery dimensions over time (Table 2). Although there was no significant change in carotid arterial strain, there was a significant increase in EM early after initiating HFD indicating an increase in vascular stiffness (Figure 1A). This increase in vascular stiffness demonstrated a linear relationship with the degree of hyperglycemia as measured by the glucose AUC during IVGTT (R² = 0.39, p < 0.001). There was a linear trend for a slight increase in carotid IMT over the two year study period (Figure 1B and 1C). However, IMT was not significantly different from baseline until two years after starting HFD at which time it had increased by an average of 0.05±0.06 mm.

Table 2.

Blood Pressure, Carotid Artery, and Brachial Artery Changes to HFD

	Baseline N= 10	4 months N= 10	8 months N= 10	12 months N= 10	18 months N= 10	24 months N= 8	ANOVA or Freidman
Systolic BP, mmHg	78 ± 9	95 ± 13*	94 ± 13	90 ± 12	95 ± 15	90 ± 8	p = 0.02
Diastolic BP, mmHg	32 ± 6	46 ± 8*	42 ± 14	42 ± 9	45 ± 12*	48 ± 8	p = 0.003
Pulse Pressure, mmHg	46 ± 8	49 ± 9	51 ± 8	48 ± 7	49 ± 6	42 ± 9	p = 0.56
Brachial Artery Diastolic Diameter, mm	2.1 ± 0.2	2.1 ± 0.3	2.1 ± 0.3	2.2 ± 0.3	2.3 ± 0.3†	2.3 ± 0.2	p = 0.01
Carotid Artery Diastolic Diameter, mm	3.5 ± 1.1	4.9 ± 0.7†	5.5 ± 1.1†	5.3 ± 0.7†	5.5 ± 0.6†	6.3 ± 1.0‡	p < 0.0001
Carotid Artery Systolic Diameter, mm	4.3 ± 1.3	5.7 ± 0.7†	6.4 ± 1.3†	6.4 ± 0.8†	6.5 ± 0.7†	7.7± 1.3	p < 0.0001
Carotid Artery Strain, %	22 ± 6	19 ± 5	18 ± 5	20 ± 4	20 ± 6	21 ± 6	p = 0.22
Carotid Artery Compliance, mm2/mmHg	0.10 ± 0.28	0.17 ± 0.47	0.29 ± 0.85	0.17 ± 0.49	0.20 ± 0.58	0.40 ± 1.0	p = 0.17

Data are mean (±SD). Abbreviations: BP; Blood pressure. Data are mean (±SD).

^*Freidman test with Dunn's post test compared to baseline and

^†repeated measures one-way ANOVA with Bonferroni's post test compared to baseline.

^‡Difference between 18 and 24 month values, p = 0.03, student's paired t-test.

Figure 1.

Ultrasound-based measurements of vascular morphology and function. Data represent mean (± SEM) values for (A) elastic modulus (EM); * repeated measures Freidman test with Dunn's post test baseline compared to 8 months, p = 0.003; (B) carotid intima-media thickness (IMT); † p = 0.01 for repeated measures test for linear trend to 18 months; ‡ p=0.05 for Wilcoxon matched-pairs signed rank test baseline to 24 months; (D and E) flow mediated vasodilation quantified as percent and absolute change in brachial artery diameter change; † p=0.04 (%) and 0.07 (absolute) for repeated measures test for linear trend baseline to 18 months, respectively; ‡ p=0.09 for Wilcoxon matched-pairs signed rank test; § p=0.08 by paired t-test; and (F) brachial artery shear stress during post-ischemic hyperemia. Panel (C) provides representative image of IMT measurement.

Bioactivity of NO from endothelial shear response was tested at each study interval by FMD. There was a significant trend for a progressive decrease in FMD beginning after the 4 month time interval (Figure 1D and 1E). However, FMD was not significantly different from baseline until 18 months to 2 years indicating that the temporal decrease was gradual. Brachial artery hyperemic shear stress measured immediately upon cuff deflation did not significantly change over time (Figure 1F), confirming that changes in shear were not responsible for changes in FMD.

Endothelial Activation and Serum Biomarkers Inflammation

Cross reactivity of the targeting ligands used for molecular imaging in rhesus macaques was confirmed using immunohistochemistry (see online Supplemental Figures). CEU molecular imaging of the carotid arteries was well tolerated in all animals without hemodynamic change. Representative two-dimensional and CEU targeted images are shown in figure 2A. At baseline, signal enhancement at the vascular wall for P-selectin and VCAM-1-targeted microbubbles was low and similar to that for control microbubbles (Figure 2B). Molecular imaging signal for both of these ECAMs increased progressively after initiating HFD whereas signal for control microbubbles did not change. A significant increase in signal was seen at 4 months for P-selectin and 8 months for VCAM-1. Molecular imaging signal for ECAMS was not elevated in control animals that were age-matched to the final study period (Figure 2C). There was a modest but significant correlation between P-selectin signal enhancement and both plasma insulin (r= 0.37, p<0.01) and truncal fat (r= 0.39, p<0.01). There were no significant relations between ECAM expression on molecular imaging and the degree of IR indicated by fasting insulin, or the IVGTT insulin AUC or glucose disappearance rate.

Figure 2.

Ultrasound-based molecular imaging. (A) Example of a carotid artery by 2-D B-mode ultrasound imaging (left) and color-coded contrast-enhanced molecular imaging of P-selectin (right) demonstrating P-selectin signal at the endothelial surface in a subject after one year of high-fat diet. (B) Mean (± SEM) background subtracted video intensity for; P-selectin-targeted (MB_P), VCAM-1-targeted (MB_V) and control (MB_C) microbubbles. * p < 0.01 for MB_P versus MB_C at 4 to 24 months. † p < 0.01 for MB_V vs MB_C at 8 to 24 months. (C) Mean (± SEM) background subtracted video intensity for; P-selectin-targeted (MB_P), VCAM-1-targeted (MB_V) and control (MB_C) microbubbles in primates exposed to a HFD for 24 months (average age 12 yrs) compared to age-matched controls on chow diet.

On biomarker profile, there were significant and sustained reductions in adiponectin and RANTES (Regulated upon Activation, Normal T-cell Expressed, and Secreted) after 4 months of HFD (Table 3). There was a significant increase in MCP-1 at 4 months, followed later by increases in C-reactive protein (CRP) and IL-18. However there was substantial variation in these markers over time. Soluble VCAM did not follow the same temporal as that observed with endothelial VCAM-1 signal on molecular imaging with levels declining over time. There were no significant associations observed between circulating plasma biomarkers and molecular imaging ECAM expression.

Table 3.

Inflammatory and Thrombotic Biomarker Changes to HFD

	Baseline N= 10	4 months N= 10	8 months N= 10	12 months N= 10	18 months N= 10	24 months N= 8	ANOVA or Freidman
Adiponectin, μg/mL	11.7 ± 2.4	8.4 ± 1.7*	8.4 ± 1.9*	8.5 ± 1.8*	6.6 ± 1.3*	6.7 ± 1.7	p = 0.003
RANTES, ng/mL	10.7 ± 4.3	3.1 ± 1.3	2.5 ± 0.4	2.3 ± 0.6	2.3 ± 0.5‡	1.9 ± 0.7	p = 0.01
MCP-1, pg/mL	246 ± 41	430 ± 74*	386 ± 67	476 ± 72*	344 ± 48	367 ± 27	p = 0.006
CRP, μg/mL	0.17 ± 0.04	0.40 ± 0.17	0.49 ± 0.16	1.39 ± 0.83†	0.10 ± 0.03	0.09 ± 0.03	p = 0.002
Interleukin-18, pg/mL	116 ± 12	144 ± 17	303 ± 144	307 ± 110†	173 ± 26	159 ± 14	p = 0.01
sVCAM, ng/mL	212 ± 18	206 ± 12	212 ± 17	213 ± 14	155 ± 10*	167 ± 9	p < 0.0001
Interleukin-8, pg/mL	1704 ± 509	1189 ± 324	941 ± 128	986 ± 206	788 ± 93	761 ± 204‡	p = 0.02
vWF, μg/mL	54 ± 4	59 ± 5	56 ± 6	64 ± 7	79 ± 10*	83 ± 9	p < 0.0001

Data are mean ± SE. Abbreviations: CRP; C-reactive protein, MCP; monocyte chemotactic protein

^*Repeated measures one-way ANOVA with Bonferroni's post test and

^†Freidman test with Dunn's post test from baseline to 18 months. Each time point compared to baseline.

^‡ANOVA post-test for linear trend

DISCUSSION

Technologies for molecular imaging of inflammation in atherosclerosis are being developed to improve clinical care by early identification of high-risk patients. These same techniques are also able to provide insight into disease pathophysiology and treatment in pre-clinical models of disease. The basis of this study was to provide further information on the complex arrangement between IR and inflammation and to better understand the early events that may lead to atherosclerosis in an animal model designed to replicate patients who develop obesity and IR due to diet and inactivity. Our data indicate that endothelial activation, manifest by ECAM expression: (a) occurs in concert with the development of obesity and IR, (b) progresses rapidly with time, and (c) precedes changes in IMT by over a year. The progressive pattern of endothelial activation on molecular imaging was not mirrored by the temporal changes seen in circulating markers of inflammation. Moreover, the progressive and steep increase in ECAM expression occurred without much further progression in the degree of IR.

In this study we have defined the endothelial expression of P-selectin and VCAM-1 in a high risk, pro-atherosclerotic phenotype that mimics the human conditions of obesity and IR. Both P-selectin and VCAM-1 were chosen as ECAM targets as they have been shown in some studies to be present in early atherosclerosis in animal models and humans with atherosclerotic disease.^{13, 15-18} Although our histology results demonstrated VCAM-1 on the plaque endothelial surface, the bulk of data argues against significant macrovascular VCAM-1 expression in humans. However, our main objective was not to identify the specific ECAMs involved in atherogenesis; rather it was to characterize the temporal relationship between diet-induced obesity and IR, and inflammatory endothelial activation.

The model of diet-induced obesity in macaques fed a HFD has been shown to produce a gradual worsening in neointimal formation that would be classified as mild to moderate on histology by 14-28 months.¹⁹ In our experience, there is substantial inter-subject variability in the metabolic response to HFD in rhesus macaques with some animals showing relative resistance to the effects of diet while other animals develop rather early and severe IR. In the current study, we did find heterogeneity in the degree of IR on IVGTT at 4 and 18 months. Yet in almost all study subjects the relative worsening of the insulin AUCs and glucose disappearance rate on IVGTT was greatest in the first 4 months and progressed more gradually after that.

Inflammation and the production of pro-inflammatory cytokines play a role in the development and/or worsening of diet-induced IR. For example, cytokines such as TNFα, IL-1, and IL-6 have been shown to increase the activity of serine kinases and/or suppressor of cytokine signaling (SOCS) pathways which together inactivate and degrade insulin receptor substrates which leads to IR.²⁰ On the other hand, diet-induced obesity and IR can produce vascular inflammation by many different potential pathways. Direct glucotoxicity, toxic effects of fatty acids mediated by toll-like receptor-4, and increased production of oxygen free radicals in IR are all potential mediators of the vascular changes that predispose to atherosclerosis.²¹ Vascular tissue IR is associated with upregulation of inflammatory genes independent of IR in other tissues such as adipose and skeletal muscle.¹⁰ Not all of these processes would necessarily be reflected by standard serologic markers of total body inflammation or cytokines. For this reason we directly assessed the endothelial expression of P-selectin and VCAM-1 expression, which are critical participants in the vascular inflammatory cascade that initiate atherosclerotic disease.^{13, 22-24}

Large prospective case-control studies in humans have demonstrated that baseline inflammatory markers serve as a predictor for future development of diabetes mellitus.²⁵ Our study in non-human primates was designed to examine some of the very early events defining the relationship between vascular inflammation and altered total body glucose homeostasis. Our findings indicate that carotid endothelial ECAM expression develops in concert with the onset of obesity and systemic IR and the timing and extent of ECAM expression is not necessarily reflected by often-used blood markers of systemic inflammation. We investigated whether there were correlations between ECAM expression and the degree of IR. Associations were found only between P-selectin and the degree of abdominal obesity and basal insulin levels, yet even these associations were weak. The lack of strong associations were largely because measures of adiposity and IR tended to worsen dramatically in the first 4 months and then level off whereas ECAM expression progressively increased over time. These findings together suggest that endothelial activation is as much dependent on the duration of IR as the severity of IR. The FMD experiments in our study were designed to test the temporal relation between ECAM expression and bioavailability of endothelial-derived NO. Again, many events lead to reduced NO bioactivity in IR such as: (1) reduced NO production through impaired insulin receptor signaling of the phosphoinositide 3-kinase pathway, (2) increased production of asymmetric dimethylarginine, the endogenous inhibitor of NO synthase, and (3) increased oxidative stress which promotes conversion of NO to peroxynitrite.²⁶ In mice with diet-induced obesity, reduced production of NO has been shown to temporally correlate with ECAM expression.¹⁰ We found it interesting that abnormal NO production occurred later and did not necessarily correlate directly with the degree of IR. However, it should be cautioned that animals used in this study had been activity-restricted for >1 yr prior to starting HFD which we have previously shown leads a substantial decrease in FMD response compared to animals with normal activity patterns.²⁷ Our study tested the temporal relation between carotid artery molecular changes, intima-medial thickness on ultrasound, and endothelial inflammatory activation in a primate model. Since endothelial activation is one of the very early events that lead to neo-intimal proliferation, it was not surprising that ECAM expression preceded any significant increases in IMT. Interestingly though, we did find an early increase in arterial stiffness that coincided with ECAM expression. Our histologic data, suggests that this dysfunction may be due to ultrastructural changes in the vessel wall content that can occur due to inflammation, hyperglycemia, or abnormal neurohormonal environment, though purely functional changes cannot be excluded either. The changes observed in arterial stiffness across follow-up displayed a strong linear relationship to the degree of hyperglycemia and are consistent with observational findings of a stepwise increase in arterial stiffness in humans with glucose intolerance and non-insulin-dependent DM.²⁸

There are several aspects of this study that may impact clinical practice. We found that systemic IR and the timing and extent of ECAM expression is not necessarily reflected by often-used blood markers of systemic inflammation. C-reactive protein (CRP) is the most often-used marker of vascular inflammation in atherosclerosis and has been validated in rhesus macaque cohorts as a reasonable pentraxin to follow the progression inflammation. ^29-31 Interestingly, in our cohort of primates, CRP, IL-18, and MCP-1 all peaked by 1 year and did not follow the same progressive pattern found for endothelial ECAM expression. Our results in primates suggest that circulating biomarkers of inflammation may not necessarily correlate with the degree of endothelial activation. It should be noted, however, that large population studies have demonstrated that elevations of CRP have been associated with an increased relative risk for developing complications of atherosclerotic disease.³² Our results also suggest that once IR develops, ECAM expression progressively increases even if the degree of IR does not rapidly progress. Studies examining the effect of dietary intervention and/or exercise training will be needed to determine whether ECAM expression can be reduced or reversed by eliminating the causal factors for IR. Studies have demonstrated that absolute carotid IMT as well as the rate of carotid IMT progression over time is associated with an increased risk of coronary revascularization, non-fatal myocardial infarction, and cardiovascular death. ^{33, 34} Identification of the endothelial molecular signal that precedes these morphometric changes has the potential to identify those at highest risk for the development of atherosclerotic disease, or can be utilized in a temporal manner to evaluate therapeutic response to lifestyle, diet, and pharmacologic interventions.

There are several important limitations of this study, the most notable of which is the number of primates studied across the follow-up. Our initial cohort of 10 primates was reduced to 8 animals after 18 months. The two animals lost to follow-up had the largest gains in both weight and truncal fat by 18 months and each demonstrated severe IR. This loss limited the use of paired statistical testing over time and likely contributed to the decline in markers of IR and adiposity between 18 and 24 months. While the biomarkers reported in our study have been shown to have cross-reactivity to human markers, the mean ratios of human to primate levels vary from 0.4 to 3.3 and thus the absolute values reported should not be directly compared to human values. In addition, the cross reactivity studies were performed in Macaca fascicularis (Cynomolgus monkeys) while our study was performed in a cohort of rhesus macaques (Macaca mullata). Although these are different species, they are both of the same genus and are the same old-world primate lineage. It is estimated that there is no more than 1% difference in their genetic sequence. ³⁵ Additionally, while dietary caloric intake was markedly increased after 4 months of a HFD and remained elevated and unchanged over 12 months, we do not have data on caloric intake after one year. Finally, we do not have confirmatory histology data in this study since these animals are part of a shared animal resource and because of an effort to avoid animal sacrifice for ethical reasons. We are, however, reassured by the fact that CEU molecular imaging signal for both P-selectin and VCAM-1 has been shown to quantify the degree of vascular inflammation in murine models of progressive atherosclerosis and to detect the earliest atherosclerotic process even before the development of neointima or fatty streaks.¹³

In summary, endothelial inflammatory activation is an early event in the development of diet induced obesity and IR without DM in primates and occurs prior to substantial increases in circulating inflammatory biomarkers or carotid IMT. Endothelial cell adhesion molecule expression progressively increases over time despite a relatively constant degree of IR. From a methodologic standpoint, we have also shown that CEU molecular imaging of large vessel endothelial phenotype in primates can provide an early indication of pro-atherosclerotic phenotype long before changes in morphology can be detected.