Impact of Free Fatty Acids on Vascular Insulin Responses Across the Arterial Tree: A Randomized Crossover Study

Abstract

Context

Vascular insulin resistance is commonly observed in obesity and diabetes; yet, insulin action across the vascular tree and the relationship between insulin responses at different vascular locations remains incompletely defined.

Objective

To elucidate the impact of elevated free fatty acids (FFAs) on insulin action across the arterial tree and define the relationship among insulin actions in the different arterial segments.

Methods

This randomized crossover study assigned healthy lean adults to 2 separate admissions with euglycemic insulin clamp superimposed for the final 120 minutes of 5-hour lipid or matched-volume saline infusion. Vascular measures including peripheral and central arterial blood pressure, brachial artery flow-mediated dilation (FMD), carotid femoral pulse wave velocity (cfPWV), augmentation index (AIx), pulse wave separation analysis, subendocardial viability ratio (SEVR), and skeletal and cardiac muscle microvascular perfusion were determined before and after insulin clamp. Insulin-mediated whole body glucose disposal was calculated.

Results

Insulin enhanced FMD, AIx, reflection magnitude, and cardiac and skeletal muscle microvascular perfusion. Elevation of plasma FFA concentrations to the levels seen in the postabsorptive state in people with insulin resistance suppressed SEVR, blunted insulin-induced increases in FMD and cardiac and skeletal muscle microvascular blood volume, and lowered insulin's ability to reduce AIx and reflection magnitude. In multivariate regression, insulin-mediated muscle microvascular perfusion was independently associated with insulin-mediated FMD and cfPWV.

Conclusion

Clinically relevant elevation of plasma FFA concentrations induces pan-arterial insulin resistance, the vascular insulin resistance outcomes are interconnected, and insulin-mediated muscle microvascular perfusion associates with cardiovascular disease predictors. Our data provide biologic plausibility whereby a causative relationship between FFAs and cardiovascular disease could exist, and suggest that further attention to interventions that block FFA-mediated vascular insulin resistance may be warranted.

insulin, free fatty acids, endothelium, flow-mediated dilation, arterial stiffness, microvasculature

Insulin is a highly vasoactive hormone, enhancing tissue perfusion and aiding its own delivery and orchestrating vital nutrient and oxygen transport to target tissues like muscle via a receptor-mediated vasodilatory action on the resistance arterioles and the microvasculature (1, 2). Additionally, insulin acts on the conduit arteries to regulate compliance. In healthy condition, insulin signals through the phosphatidylinositol 3-kinase/endothelial nitric oxide (NO) synthase pathway, to produce NO and trigger vasodilation. In the insulin-resistant states, insulin's signaling via this pathway is blunted, and insulin may even induce vasoconstriction. Endothelial cells are extensively differentiated across vascular beds and their gene expression regulated by local environmental cues based on vascular tree location and surrounding tissues. Thus, insulin could stimulate varied effects based on vascular region and the presence or absence of insulin resistance.

Based on epidemiologic studies in populations with diabetes or insulin resistance, numerous vascular measures hold predictive value for cardiovascular (CV) events or mortality. Cardiac perfusion supply and demand ratio predicts CV outcomes/mortality (3), and central and/or peripheral large artery stiffness heightens CV event risk (4) in cohorts with type 2 diabetes. Furthermore, impaired endothelium-dependent vasodilation at conduit arteries and resistant arterioles has a strong direct relationship with future CV events (4, 5). Interestingly, insulin has been shown to relax the vasculature at several of these levels including large peripheral arteries (6) and conduit arteries/resistance arterioles (7) in health but not in insulin-resistant conditions (7, 8). Mounting evidence confirms that insulin also enhances microvascular perfusion in cardiac and skeletal muscle in health (9) but that this effect is lost in insulin-resistant conditions like metabolic syndrome (7, 10, 11). Insulin resistance at these levels likely transpires early in the pathogenesis of vascular dysfunction and is potentially reversible (12). Nevertheless, the relationships between these vascular measures and pathogenesis of these changes remain incompletely defined.

Free fatty acids (FFAs) are implicated in insulin resistance pathogenesis and independently associated with all-cause and CV morality in individuals with coronary artery disease. Chronic insulin-resistant states like type 2 diabetes are associated with elevated plasma FFA concentrations (13). Data from the large epidemiologic Cardiovascular Health Study indicates each SD (0.2 mEq/L) higher FFA associates with a 12% increased risk of future heart failure (14). Elevated circulating FFA levels impair endothelium-dependent vasodilation, and reduce insulin-mediated vasodilation, NO production, and glucose uptake (15). We have previously demonstrated in healthy humans that raising plasma FFA concentrations to the levels seen in poorly controlled diabetes induces microvascular insulin resistance in both cardiac and skeletal muscle (16).

In the current study we aimed to examine the impact of clinically relevant elevation of plasma FFA concentrations on insulin responses across the arterial tree as compelling data have indicated that FFAs potently induce both endothelial dysfunction and insulin resistance, with focus on measures known to be frequently impaired in insulin resistant states and/or predictive of CV outcomes: (1) whole heart perfusion, (2) large artery stiffness, (3) conduit artery or resistance arteriole endothelium-dependent dilatation, and (4) cardiac and skeletal muscle microvascular perfusion. This approach would add to the literature a concurrent examination of FFAs and insulin interaction at multiple arterial levels. We hypothesized that FFA-induced vascular insulin resistance varies depending on the vessel size and function, and the vascular insulin resistance outcomes are interconnected. Furthermore, we speculated that, if insulin responses were differentially impacted by FFAs across the vascular tree, peripheral vascular regions that directly supply the target tissue (ie, the muscle microvasculature) would be more affected than central vascular responses.

Materials and Methods

Study Subjects

This crossover study allocated healthy, lean (body mass index 18-25 kg/m²) participants (ages 18-35 years) 1:1 to 2 separate admissions, scheduled 2 to 4 weeks apart, in randomized order so that participants served as their own control. People with chronic medical illness such as type 1 diabetes, type 2 diabetes, hypertension, hypotension (blood pressure <100/60 mmHg), hyperlipidemia, chronic or acute cardiac, pulmonary, liver or kidney diseases, or intracardiac or intrapulmonary shunts, current smokers (within the past 6 months), use of medications that may affect vascular function (such as angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, glucagon-like peptide receptor agonists, statins, fibrates, fish oil, aspirin, vitamin C, and vitamin E), pregnancy, breastfeeding, or known hypersensitivity to soy beans or eggs (contained in lipid emulsion) or perflutren (contained in Definity microbubbles) were excluded from the study.

Study Protocol

The study is reported in accordance with Consolidated Standards of Reporting Trials (CONSORT) recommendations (Fig. 1). All participants were screened and studies conducted at the University of Virginia Clinical Research Unit. Initial screening visit consisted of obtaining written informed consent, history and physical examination, and fasting blood work, including comprehensive metabolic profile, complete blood count, lipid profile, prothrombine time, partial thromboplastin time, and, if female, a serum beta-human chorionic gonadotropin pregnancy test. If eligibility criteria were satisfied, participants were invited to participate in the study. After a second visit to measure cardiorespiratory fitness (VO₂max) using the treadmill Bruce protocol and body composition by air displacement plethysmography (Bod-Pod, Life Management, Concorde, CA), each participant underwent 2 separate admission studies in randomized order (Admission Protocol, Fig. 2A).

Figure 1.

CONSORT flow diagram.

Figure 2.

(A) Study protocol for admissions. BD, blood draw; GLP-1, glucagon like peptide-1; NO, nitric oxide; FFA, free fatty acids. (B) Mean glucose infusion rate during euglycemic insulin clamp with standard error of mean (bars) during euglycemic insulin clamp. (C) Mean FFA concentrations with standard error of mean during euglycemic insulin clamp. Open circles = saline admission, black squares = lipid admission. *P < .0001, 2-way ANOVA analysis comparing final 40 minutes of euglycemic insulin clamp between admissions. #P < .0001, paired t-test comparing time 0 and time 120 minutes.

For each admission study, participants were admitted to the University of Virginia Clinical Research Unit at 07:00 hours on admission days after a fast beginning at 22:00 hours the previous night and abstaining from caffeine and exercise for 24 hours prior to admissions. Participants reclined supine, in a temperature-controlled room, and received an intravenous infusion of either normal saline or lipid (20% intralipids + heparin 12 units/kg/hour) for 5 hours (time −180 minutes to time 120 minutes) at a rate of 45 mL/hour × 1 hour and then 30 mL/hour × 4 hours, with a primed (2 mU/kg/minute for 10 minutes) then continuous (1 mU/kg/minute for 110 minutes) insulin infusion superimposed on the final 120 minutes (time 0-120 minutes). During the insulin clamp, plasma venous glucose was measured every 5 minutes and dextrose (20%) was infused at a variable rate to maintain plasma glucose within 10 mg/dL of starting estimated arterial glucose concentrations, with venous glucose concentration corrected for glucose infusion rate as previously described (9, 11). We did not utilize hand heating to promote venous blood arterialization because of the potential confounding influence of temperature on contralateral limb blood flow and vasodilation (17, 18). The steady-state glucose infusion rate was determined by the glucose infusion rates over the final 40 minutes, expressed as mg/kg/minute. At the rate used, lipid infusion raised plasma FFAs concentration to ∼2 mM in 3 hours and then fell back to ∼1 mM (19), levels seen postabsorptively in people with type 2 diabetes (13) or metabolic syndrome (7). Vascular measurements including peripheral and central arterial blood pressures and the vascular measures described below were performed before and at the end of the euglycemic insulin clamp (Fig. 2A). Venous blood samples were collected to determine plasma concentrations of insulin, FFAs, NO, and glucagon-like peptide 1 (GLP-1) levels at times −180, 0, and 120 minutes.

The study protocol was carried out in accordance with the Declaration of Helsinki (2013) of the World Medical Association and approved by the University of Virginia Institutional Review Board. Written informed consent was obtained from each subject before study enrollment during the screening visit.

Vascular Outcome Measures

Subendocardial viability ratio

The subendocardial viability ratio (SEVR), or Buckberg index, is calculated using the aortic pressure waveform by ratio of area under the curve during diastole to the area under the curve during systole, expressed as a percentage. It estimates myocardial perfusion to cardiac contraction pressure because coronary artery filling occurs in diastole (20).

Carotid femoral pulse wave velocity and augmentation index

Both the carotid femoral pulse wave velocity (cfPWV) and the radial artery augmentation index (AIx) were assessed using SphygmoCor tonometer (AtCor Medical; Naperville, IL) as measurements of arterial stiffness. Both were performed after participants rested supine for 15 minutes minimum. For cfPWV, which estimates central (aortic) arterial stiffness, distance from the suprasternal notch to both the right carotid pulse and right femoral pulse were measured. Arterial waveforms were captured for 10 seconds at each location in duplicate and the mean value reported. The technique was performed by an investigator who previously demonstrated good intraobserver reliability with a coefficient of variation of 3.63%. AIx, a measure of arterial stiffness at a muscular and more peripheral artery, was determined at the radial artery over a 30-second interval and calculated by the difference between the systolic pressure and the inflection pressure divided by the pulse pressure, expressed as a percentage. Mean values were calculated and corrected for a heart rate of 75 beats/minute.

Wave separation analysis

Radial artery pressure waves were recorded through aplanation tonometry using a Sphygmocor tonometer and used to synthesize the central aortic pressure waveform by averaging 20 aggregated sequential waveforms and performing a general transfer function. Waves were decomposed to their forward (Pf) and backward (Pb) components, and reflection magnitude (RM) was calculated as the ratio of Pb to Pf wave amplitudes.

Flow-mediated dilation and postischemic flow velocity

Left brachial artery flow-mediated dilation (FMD) and postischemic flow velocity (PIFV) were measured at approximately 5 cm proximal to the antecubital fossa using the EPIQ 7 CV ultrasound (Philips Medical Systems; Andover, MA) and linear array probe (L12-3). Forearm blood pressure cuff was insufflated to 240 mmHg, held for 5 minutes, then rapidly deflated. Blood flow velocity and vessel diameter were captured prior to insufflation and for 120 seconds following cuff release. The technique was performed by a single experienced investigator who has demonstrated good intraobserver reliability with a coefficient of variation of 7.41%. Images were analyzed using Brachial Analyzer edge detection software (Medical Imaging Applications; Coralville, IA).

Microvascular perfusion in cardiac and skeletal muscle

Cardiac (interventricular septum, 4-chamber view) and forearm flexor muscle (∼5 cm distal to the antecubital fossa) microvascular perfusions were determined by contrast-enhanced ultrasound using an EPIQ 7 ultrasound system. Definity microbubbles (Lantheus Medical Imaging, North Billerica, MA) were infused intravenously for 4 minutes prior to image acquisition to allow the contrast agent to reach steady-state concentration in the circulation. We used a low mechanical index (MI = 0.10) continuous imaging for 20 seconds (myocardium) or 30 seconds (forearm) at a frame rate of 15 per second with a flash at 0.88 MI to initiate a replenishment curve (21). Four forearm and 4 myocardial replenishment curves were acquired and analyzed using Q-Lab software (Philips Research; Cambridge, MA) yielding measures of microvascular blood volume (MBV, in video intensity units) and microvascular flow velocity (MFV, per second).

Biochemical Analysis

Comprehensive metabolic panels, complete blood counts, lipid profiles, and pregnancy tests were performed at the University of Virginia Clinical Chemistry Laboratory. Plasma glucose was determined using a YSI glucose analyzer (Yellow Spring Instruments). Plasma insulin concentrations were assayed using an enzyme-linked immunosorbent assay kit (ALPCO Diagnostics, Windham, NH). Plasma FFA concentrations were assayed with an in vitro enzymatic colorimetric assay using a Wako HR Series NEFA-HR kit (Wako Diagnostics, Richmond, VA). Plasma total GLP-1 concentrations were measured using an enzyme-linked immunosorbent assay kit (EZGLPIT-36K, RRID:AB_281378, Burlington, MA). Serum NO levels were quantified using the 280i Nitric Oxide Analyzer (GE Analytical Instruments), as described previously (22, 23). Homeostatic model assessment for insulin resistance (HOMA-IR) was calculated (HOMA-IR = fasting glucose in mg/dL × fasting insulin in mU/L/405).

Statistical Analysis

Sample size was determined based on results from our prior studies (16, 24) which demonstrated a 25% to 30% differential treatment effect in insulin-mediated changes in the geometric mean of skeletal and cardiac muscle MBV. With a 2-tailed α of .05, a sample size of 17 would be required to have >85% power (1 – β) to detect a difference in insulin-mediated change in both skeletal and cardiac muscle MBV. Descriptive data are presented as mean ± standard error of mean (SEM) for continuous data and as frequencies and percentages for categorical data. Statistical analyses were completed with GraphPad Prism 9.3 Software (San Diego, CA).

Pairwise/multiple comparisons

Pairwise comparisons of mean vascular measures before and after insulin and preinsulin values between admissions were performed with Student's paired t-test. Comparisons between glucose infusion rates, blood pressures, and biochemical values across 3 time points were performed with the 1-way analysis of variance (ANOVA) test. Comparisons across time (time 0 vs 120 minutes) and admission (saline vs lipid) factors were tested with 2-way ANOVA testing or mixed-effects analysis. Statistical significance was set at the P < .05 level.

Univariate and multivariate analyses

To determine the relationships between microvascular insulin responses/endothelial function and vascular outcome measures of conduit and resistance vessels, both univariate and multivariate correlation analyses were performed. The univariate correlations tested the bivariate relationships between insulin-stimulated changes in cardiac microvascular perfusion and baseline (saline admission, preinsulin) FMD, SEVR, cfPWV, AIx, glucose infusion rate, and plasma concentrations of FFAs, insulin, and triglycerides. The bivariate association was quantified by Pearson correlation coefficient (r). The multivariate regression examined the unique relationships between skeletal muscle insulin-mediated changes in MBV (dependent variable) and insulin-induced changes in FMD, AIx, and cfPWV (independent variables) in both saline and lipid admissions. The null hypothesis for each variable was there was no relationship between insulin-mediated changed in microvascular parameters and other variables. If P < .05, the null hypothesis was rejected.

Results

A total of 18 participants (50% female) completed the study (CONSORT flow diagram, Fig. 1). Participants were young adults (mean age 23 years) with normal body mass index, hemoglobin A1c, blood pressure, cholesterol level, and good fitness level (mean VO₂max 44 mL/kg/min) (Table 1). Baseline biochemical analyses were not significantly different between admissions. Lipid infusion raised plasma FFA concentrations to ∼1.8 mM and triglyceride levels 2-fold as expected, without altering basal plasma levels of NO or GLP-1 (Table 2). Lipid infusion alone modestly raised insulin levels from baseline 3.3 mU/L to 5.4 mU/L (P = .03). Insulin clamp lowered FFA concentrations in both admissions (Fig. 1C). Lipid infusion did not alter either peripheral or central blood pressures. Similarly, insulin infusion, either alone or on top of lipid infusion, had no effect on peripheral and central blood pressures. However, steady-state glucose infusion rates were significantly lower during lipid admission than the saline admission (mean 5.5 vs 6.2 mg/kg/minute, P < .0001, Fig. 2B), indicating FFA-induced metabolic insulin resistance.

Table 1.

Baseline characteristics

Characteristic	Combined (mean ± SD)	Females(mean ± SD)	Males(mean ± SD)
Sex (number/% female)	18	9/50%	9/50%
Age (years)	23 ± 3.4	22.9 ± 2.5	23.1 ± 4.3
Body mass index (kg/m2)	22.3 ± 1.4	22 ± 1.6	22.6 ± 1.3
Percent body fat (%)	20.8 ± 8.7	27.9 ± 4.1	13.8 ± 5.6
Fasting blood glucose (mg/dL)	88.9 ± 6	85.4 ± 6.1	92.3 ± 3.4
Systolic blood pressure (mmHg)	114 ± 9.6	107 ± 7.4	121 ± 6.1
Diastolic blood pressure (mmHg)	68.9 ± 8.1	65.9 ± 7.5	71.9 ± 7.8
Total cholesterol (mg/dL)	158.5 ± 31.8	160.6 ± 32.4	156.4 ± 33.1
LDL-cholesterol (mg/dL)	88.3 ± 24.7	86.3 ± 20.9	90.3 ± 29.1
HDL-cholesterol (mg/dL)	58.1 ± 14.1	62.4 ± 14.4	53.7 ± 13
Triglycerides (mg/dL)	74.3 ± 44	72.7 ± 40	76 ± 50.1
VO2max (mL/kg/min)	44 ± 5.9	40.8 ± 5.7	47.2 ± 4.3
HOMA-IR	0.75 ± 0.08	0.66 ± 0.02	0.84 ± 0.15

Abbreviations: LDL, low-density lipoprotein; HOMA-IR, homeostatic model assessment for insulin resistance.

Table 2.

Central and peripheral blood pressures and biochemical analyses

	Saline admission		Lipid admission		P values
	Time 0 minutes	Time 120 minutes	Time 0 minutes	Time 120 minutes	Time, saline admission	Time, lipid admission	Time × admission factors
Biochemical analyses
FFA (mEq/L)	0.45 ± 0.3	0.04 ± 0.004	1.8 ± 0.1	1.0 ± 0.06	<.0001	<.0001	<.0001
GLP-1 (pM)	12.7 ± 1.4	8.1 ± 0.9	12.1 ± 1.3	9.1 ± 1.3	<.0001	<.0001	.53
Insulin (pM)	3.0 ± 0.3	66.7 ± 4.0	5.3 ± 0.9	68.1 ± 4.7	<.0001	<.0001	.89
Triglyceride (mg/dL)	59.6 ± 4.9	42.5 ± 3.2	122.1 ± 13.6	115.6 ± 10.1	<.0001	<.0001	<.0001
NO (μM)	25.5 ± 3.5	22.3 ± 3.0	28.6 ± 3.9	25.0 ± 3.3	<.0001	<.0001	.88
Peripheral blood pressures (mmHg)
SBP	108.4 ± 2.1	108.6 ± 2.3	109.5 ± 2.6	110.5 ± 2.4	.93	.57	.77
DBP	62.8 ± 1.8	61.1 ± 1.2	63.3 ± 1.6	64.3 ± 1.3	.25	.44	.12
MAP	76.6 ± 1.7	74.4 ± 1.2	77.1 ± 1.8	77.5 ± 1.5	.12	.76	.16
PP	46.7 ± 2.3	47.4 ± 2.1	46.2 ± 1.8	46.2 ± 2.0	.76	.99	.78
Central blood pressures (mmHg)
SBP	93.1 ± 1.8	91.1 ± 1.8	94.1 ± 2.2	93.2 ± 1.9	.18	.52	.63
DBP	63.4 ± 1.9	62.1 ± 1.3	63.7 ± 1.6	65.1 ± 1.3	.38	.27	.13
MAP	76.6 ± 1.7	74.4 ± 1.2	77.1 ± 1.8	77.5 ± 1.5	.12	.76	.16
PP	29.7 ± 1.4	29.4 ± 2.0	29.9 ± 1.3	28.2 ± 1.3	.88	.17	.58

Time 0 minutes indicates after 180 minutes of saline/lipid infusion and pre-euglycemic insulin clamp. Time 120 indicates posteuglycemic insulin clamp. P values represent results from paired t-tests, 1-way ANOVA, or 2-way ANOVA as applicable. Baseline (prelipid infusion) biochemical results are presented in the text.

Abbreviations: DBP, diastolic blood pressure; FFA, free fatty acid; GLP-1, glucagon-like peptide-1; MAP, mean arterial pressure; NO, nitric oxide; PP, pulse pressure; SBP, systolic blood pressure.

Arterial Stiffness and Wave Separation Analysis

Insulin infusion, either alone or on top of lipid infusion, did not alter cfPWV (Table 3). However, insulin significantly reduced AIx, and insulin's ability to reduce AIx was attenuated by lipid infusion (Table 3). Dissecting the arterial wave forms through the wave separation analysis showed that RM showed a similar pattern of decreasing in response to insulin, which was attenuated by lipid infusion. Neither Pf nor Pb waveforms were significantly changed by insulin in either lipid or saline admission.

Table 3.

Conduit artery endothelial function, arterial stiffness, and SEVR

	Saline admission			Lipid admission			P value
	Time 0 minutes	Time 120 minutes	Change	Time 0 minutes	Time 120 minutes	Change	Time, saline	Time, lipid	Time × admit factors
PIFV (cm/sec)	262.2 ± 15.8	267.7 ± 14	5.4 ± 9.2 (−13.9-24.8)	297.6 ± 18	286.5 ± 9.9	−11.1 ± 16.3 (−45.5-23.4)	.56	.51	.38
cfPWV (m/sec)	5 ± 0.1	5 ± 0.1	−0.01 ± 0.1 (−0.2-0.2)	5.2 ± 0.2	4.9 ± 0.1	−0.3 ± 0.2 (−0.6-0.04)	.92	.08	.18
AIx (%)	−2.4 ± 2.2	−6.4 ± 2.5	−3.9 ± 1.3 (−6.6-−1.2)	−3.6 ± 2.5	−6.8 ± 2	−3.2 ± 1.5 (−6.3-0.3)	.006	.07	.73
Pf	26.7 ± 1	27.5 ± 1.2	0.8 ± 1.4 (−2.2-3.7)	25.9 ± 1	26.7 ± 1.3	0.1 ± 1.1 (−2.2-2.5)	.58	.9	.87
Pb	13.1 ± 0.8	11.9 ± 0.8	−1.2 ± 0.9 (−3.0-0.7)	11.9 ± 0.6	11.3 ± 0.6	−0.7 ± 0.5 (−1.9-0.5)	.21	.22	.71
RM	47.4 ± 2	41.1 ± 1.7	−6.3 ± 1.8 (−10-−2.5)	45.2 ± 1.6	42 ± 2	−2.7 ± 2.4 (−7.8-2.2)	.002	.26	.14
SEVR (%)	194.4 ± 6.8	196.7 ± 7.2	2.3 ± 3.6 (−5.3-9.9)	183.9 ± 7.3a	180 ± 7.9	−3.9 ± 4.3 (−12.9-5.2)	.54	.38	.27

P values indicate paired t-test comparing time (pre- vs post-120 minute euglycemic insulin clamp) for each admission (saline vs lipid) and 2-way ANOVA or mixed effects analysis of time (0 vs 120 minutes) by admission (saline vs lipid) factors. Bolded values indicate significant P values.

Abbreviations: AIx, augmentation index; cfPWV, carotid femoral pulse wave velocity; Pb, backward wave; Pf, forward wave; PIFV, peak postischemic flow velocity; RM, reflection magnitude; SEVR, subendocardial viability ratio.

^aP < .05 comparing time 0 minutes SEVR between admissions.

Conduit Artery and Resistance Arteriole Endothelial Function

Insulin significantly enhanced FMD and lipid infusion extinguished this insulin-mediated effect, reaching significance in time by admission factor analysis (Fig. 3A). Postischemic peak flow velocity was not significantly changed by insulin infusion in either admission; however, time 0 peak flow velocity trended toward higher in the lipid admission (P = .05, time 0 PIFV, lipid vs saline) (Table 3).

Figure 3.

(A) FMD. (B) Cardiac MBV. (C) Cardiac MFV. (D) Skeletal muscle MBV. (E) Skeletal muscle MFV. Open circles = 0 minutes (pre-insulin clamp). Open triangles = 120 minutes (postinsulin clamp). (F) Scatter plot showing the relationship between insulin-mediated FMD and insulin-mediated skeletal muscle MBV. Open circles = saline admission, black squares = lipid admission. Solid line = best fit line, dotted line = 95% CI of best fit line. FMD, flow-mediated dilation; MBV, microvascular blood volume, MFV, microvascular flow velocity.

Insulin-Mediated Cardiac Perfusion

We assessed both insulin-mediated changes in cardiac microvascular perfusion and SEVR, which estimates myocardial perfusion to cardiac contraction pressure. Similar to prior reports, insulin significantly enhanced MBV during the saline admission and lipid infusion extinguished this effect (Fig. 3B), reaching significance for difference in time (preinsulin vs postinsulin) by admission factors (saline vs lipid infusion). However, insulin, either alone or on top of lipid infusion, did not alter cardiac MFV (Fig. 3C). These results indicate that in our study population insulin enhances cardiac microvascular perfusion and this action is blocked by lipid infusion. Insulin infusion did not alter SEVR, but raising plasma concentrations to ∼1.8 mM via lipid infusion significantly decreased SEVR compared with saline (Table 3). Superimposing insulin infusion on top of lipid infusion did not further depress or correct the decrease in SEVR.

Insulin-Mediated Skeletal Muscle Microvascular Perfusion

As shown in Fig. 3D, insulin infusion resulted in an overall significant increase in muscle MBV and lipid infusion attenuated this effect. However, there was no significant difference between time by admission factors, likely due to a more heterogeneous responses to insulin in the skeletal muscle microvasculature. Indeed, during saline admission, insulin actually resulted a net decrease in MBV in 3 subjects. In the presence of lipid infusion, insulin depressed MBV in 6 participants. Skeletal muscle MFV was significantly increased by insulin and this augmented flow velocity was extinguished during lipid admission, reaching significance for difference in time by admission factors (Fig. 3E).

Correlation and Linear Regression Analysis

In univariate analyses examining relationships between change in heart MBV and baseline vascular and biochemical measures, insulin-stimulated change in heart MBV correlated only with baseline FMD (r = 0.51, P = .04) but not SEVR. In a multivariate regression analysis, we examined the unique relationship between insulin-stimulated difference in skeletal muscle MBV and insulin-stimulated differences in other vascular measures (FMD, AIx, and cfPWV) in the presence and absence of elevated FFAs. The overall regression was statistically significant (R² = 0.38, F(3, 29) = 5.97, P = .003). Insulin-mediated difference in FMD (β = .06, P = .001) and insulin-mediated difference in cfPWV (β = −.31, P = .03) significantly predicted skeletal muscle MBV. However, insulin-mediated change in AIx did not significantly predict insulin-mediated change in skeletal muscle MBV (β = .007, P = .6). Scatter plot examining the relationship between insulin-mediated change in FMD and skeletal muscle MBV (Fig. 3F) indicates a close relationship between the 2 variables in both admissions.

Discussion

FFAs play a pivotal role in insulin resistance pathogenesis and the CV complications of diabetes. We have previously shown in healthy humans that elevation of plasma FFA concentrations to levels seen in extremely poorly controlled diabetes via intralipid infusion induces metabolic insulin resistance as well as microvascular insulin resistance in both cardiac and skeletal muscle. The current results demonstrate that insulin exerts a vasodilatory action on all segments of the arterial tree and FFAs at concentrations similar to those in the postabsorptive state in type 2 diabetes (13) or metabolic syndrome (7) effectively impair insulin action in large conduit arteries in addition to cardiac and skeletal muscle microvasculature, along with a reducing noninsulin stimulated whole heart perfusion. Additionally, multivariate regression analysis showed that insulin-mediated skeletal muscle microvascular perfusion was independently associated with insulin-mediated FMD and cfPWV. Our findings were clearly in line with our hypothesis that FFAs would induce endothelial dysfunction and that insulin-mediated vascular responses are interconnected across the vascular tree and our speculation that insulin action at the peripheral/tissue vessels might be more selectively impacted than central arteries given the overwhelming evidence that FFAs induce whole body insulin resistance and microvascular insulin action predicts insulin's metabolic action in humans (1, 25).

FMD, a surrogate of endothelial function, has been associated with traditional CV risk factors and impaired brachial FMD predicts future CV events in older adults (5). The impact of insulin on FMD has been controversial. Some studies reported that in healthy humans modest hyperinsulinemia, mimicking fasting hyperinsulinemia of insulin-resistant states, abolishes endothelium-dependent vasodilation in large conduit arteries (26), independent of insulin sensitivity or lipid profile (27). Limitations of these studies are the small sample size and use of the hot-hand method during euglycemic clamp, which has a potential influence on blood flow and vasodilation (17, 18), and both studies were published prior to consensus FMD methodology guidelines (28), which has led to improved reproducibility of the technique and standardized methodologies. In contrast to these findings but in line with results from our current study, others have shown that insulin at similar concentrations potently enhances endothelial function in the brachial artery in healthy humans (7) and insulin's ability to improve FMD is lost in insulin-resistant humans (7, 29). Importantly, raising plasma FFA concentrations via orally ingested saturated fatty acids also led to a reduction in FMD posteuglycemic insulin clamp (30).

In the current study insulin infusion significantly increased FMD, which was reversed in the presence of elevated FFA concentrations. The finding that insulin-mediated FMD was independently associated with insulin-mediated microvascular perfusion in muscle is important as FMD reflects endothelial function mainly in the brachial artery and with some component of resistance arterioles, and tissue perfusion, including microvascular perfusion, is mostly influenced by resistance and microvascular arterioles. Furthermore, both FMD and insulin-stimulated microvascular perfusion are predominately NO dependent. Indeed, we have previously observed in healthy humans and participants with metabolic syndrome a significant positive correlation between FMD and glucose infusion rate during the insulin clamp (7), and glucose infusion rate and insulin-mediated changes in MBV are mutually predictive in a cohort including healthy, obese, and type 1 diabetes populations (25). While we did not in the current study observe a change in PIFV, which is affected by endothelial function in resistance as well as microvascular arterioles, insulin significantly increased PIFV after a high-fat meal (7).

Both AIx and cfPWV measure arterial stiffness and predict CV events and mortality. The cfPWV in particular is considered a surrogate for subclinical atherosclerosis (31). While cfPWV has not been shown to be affected by acute insulin infusion (7, 9), insulin infusion acutely lowers AIx in healthy but not obese individuals (6, 8). In concert with these earlier findings, we in the current study observed a significant reduction in AIx after insulin infusion and this effect was clearly blunted in the presence of elevated FFAs. Interestingly, waveform analysis showed that insulin infusion did not affect either the Pf or Pb components but significantly lowered the RM. This is of potential importance as in humans RM is strongly predictive of incident heart failure and CV events (32, 33) and independently associated with all-cause mortality (34). Clearly these salutary actions of insulin in the large peripheral artery are abrogated in the presence of lipid infusion, consistent with prior findings that plasma FFA levels positively associate with arterial stiffness (35).

SEVR, defined as diastolic to systolic pressure–time integral ratio, is a useful tool to estimate myocardial oxygen supply and demand. Reduced SEVR is associated with unfavorable CV risk profile in women with type 2 diabetes (36) and predicts fatal and nonfatal CV events in patients with nondialysis-dependent chronic kidney disease (37). The observation of reduced SEVR in the presence of elevated FFAs is important as plasma concentrations of FFAs increase in both the insulin-resistant and insulin-deficient state and contribute to insulin resistance and myocardial dysfunction. This potentially creates a detrimental ischemic cycle in situations of impaired myocardial perfusion as FFA oxidation requires more oxygen and energy than glycolysis. Humans with metabolic syndrome tend to have lower SEVR, and this decrease worsens as the number of metabolic syndrome components increases (38).

It is notable that in the current study insulin increased myocardial perfusion by increasing both cardiac MBV and MFV but failed to alter SEVR. Nor did microvascular myocardial perfusion correlate with SEVR in this study. Whether this lack of correlation is a consequence of SEVR being an assessment of entire heart perfusion including larger coronary vessels, in addition to microvessels, or the small sample size requires further study. Nevertheless, in other much larger populations with primary hypertension, SEVR also did not relate to markers of insulin resistance (39). On the other hand, lipid infusion not only blunted insulin-mediated cardiac microvascular perfusion but also significantly depressed SEVR. This is of potential significance physiologically and clinically given people with type 2 diabetes tend to have reduced microvascular perfusion from either coronary atherosclerotic changes and/or microvascular rarefaction. The lack of insulin-mediated coronary microvascular perfusion due to insulin resistance and reduction in SEVR from elevated plasma FFAs would further suppress coronary perfusion and oxygen supply. In ischemic dog hearts, FFAs play a significant role in increasing myocardial oxygen requirement and depressing contractility (40).

This study has several limitations. Firstly, it focuses on young healthy individuals in order to avoid many confounding factors that could affect interpretation of results. This certainly limits the generalizability to metabolically diverse, chronic insulin-resistant conditions. Secondly, acute lipid infusion was used to raise plasma FFA concentrations and the study is not a longitudinal one. As such, the study condition may not authentically imitate conditions of chronically elevated FFA. We opted this approach as FFAs are the most studied insulin resistance inducing factors and linked to CVD, and the current study design would help to specifically isolate the impact of FFAs on insulin's vascular actions and avoid other insulin resistance–inducing factors associated with chronic insulin resistant conditions. Nonetheless, our results did show that insulin-stimulated changes in NO-dependent endothelial function, AIx, and RM are more sensitive and malleable vascular outcome measures than the other vascular outcomes that were unchanged by insulin or lipid. The positive association between insulin-mediated microvascular perfusion and FMD adds to our understanding of endothelial function in health and with acute insulin resistance across vascular beds. It also raises a potential mechanism whereby whole body insulin resistance, accompanied by impaired microvascular insulin responses at the skeletal muscle, is a marker of future CV disease risk through a shared pathogenesis which triggers endothelial dysfunction at multiple vascular regions in parallel.

While elevated FFAs provide a useful vascular insulin resistance model and have mechanistic implications for early insulin resistance and CV outcomes, drugs associated with FFA lowering, such as acipimox (41, 42), thiazolidinediones (43), and fibrates (44, 45), demonstrate lackluster long-term CV benefits despite evidence of improvement in whole body insulin sensitivity. This is not surprising as CV outcomes are determined by numerous contributing factors and a simple suppression of plasma FFAs concentrations may not be sufficient to alter the CV outcomes associated with insulin resistant conditions. Additionally, each of the drugs tested has many other effects in addition to suppressing plasma FFAs concentrations. Further study is warranted to leverage multiple factors or signaling pathways for future drug development.

In conclusion, raising plasma concentrations of FFAs to the levels seen in the postabsorptive state in type 2 diabetes or metabolic syndrome reduced insulin's ability to increase FMD or reduce AIx and RM, blunted insulin-mediated changes in cardiac and skeletal muscle MBV, and lowered SEVR (Fig. 4). Additionally, our data suggest that impaired insulin-mediated FMD and insulin-mediated microvascular perfusion are interconnected and precedes impaired noninsulin-stimulated FMD in the setting of FFA-induced vascular insulin resistance and dysfunction. While further study is needed to discern the contributory mechanisms, these findings suggest that FFAs are critically important in modulating insulin actions in all segments of the arterial tree and blocking the deleterious action of FFAs on vascular insulin action could be part of the therapeutic strategy in preventing vascular insulin resistance and associated CV complications.

Figure 4.

Clinically relevant elevation of plasma FFA concentrations induces pan-arterial insulin resistance and measures of nitric oxide–dependent endothelial function are closely related. AIx, augmentation index; cfPWV, carotid femoral pulse wave velocity; FMD, flow-mediated dilation; MBV, microvascular blood volume; SEVR, subendocardial viability ratio.

Funding

This work was supported by National Institutes of Health grants R01DK102359 and R01DK125330 (to Z.L.) and K23DK131327 (to KML).

Disclosures

All authors have no conflicts relevant to this article.

Data Availability

The dataset generated during the current study are available from the corresponding author upon reasonable request.

Abbreviations

 
• AIx

  augmentation index

 
• ANOVA

  analysis of variance

 
• cfPWV

  carotid femoral pulse wave velocity

 
• CV

  cardiovascular

 
• FFA

  free fatty acid

 
• FMD

  flow-mediated dilation

 
• GLP

  glucagon-like peptide

 
• HOMA-IR

  homeostatic model assessment for insulin resistance

 
• MBV

  microvascular blood volume

 
• MFV

  microvascular flow velocity

 
• MI

  mechanical index

 
• NO

  nitric oxide

 
• Pb

  backward

 
• Pf

  forward wave

 
• PIFV

  postischemic flow velocity

 
• RM

  reflection magnitude

 
• SEVR

  subendocardial viability ratio