Relationship of Intramyocellular Lipid to Insulin Sensitivity May Differ With Ethnicity in Healthy Girls and Women

Abstract

The prevalence of type 2 diabetes is greater among African Americans (AA) vs. European Americans (EA), independent of obesity and lifestyle. We tested the hypothesis that intramyocellular lipid (IMCL) or extramycellular lipid (EMCL) would be associated with insulin sensitivity among healthy young women, and that the associations would differ with ethnic background. We also explored the hypothesis that adipokines and estradiol would be associated with muscle lipid content. Participants were 57 healthy, normoglycemic, women and girls mean age 26 (±10) years; mean BMI 27.3 (±4.8) kg/m²; 32 AA, 25 EA. Soleus IMCL and EMCL were assessed with ¹H magnetic resonance spectroscopy (MRS); insulin sensitivity with an insulin-modified frequently sampled intravenous glucose tolerance test and minimal modeling; body composition with dual-energy X-ray absorptiometry; and intra-abdominal adipose tissue (IAAT) with computed tomography. Adiponectin, leptin, and estradiol were assessed in fasting sera. Analyses indicated that EMCL, but not IMCL, was greater in AA vs. EA (2.55 ± 0.16 vs. 1.98 ± 0.18 arbitrary units, respectively, P < 0.05; adjusted for total body fat). IMCL was associated with insulin sensitivity in EA (r = −0.54, P < 0.05, adjusted for total fat, IAAT, and age), but not AA (r = 0.16, P = 0.424). IMCL was inversely associated with adiponectin (r = −0.31, P < 0.05, adjusted for ethnicity, age, total fat, and IAAT). In conclusion, IMCL was a significant determinant of insulin sensitivity among healthy, young, EA but not AA women. Further research is needed to determine whether the component lipids of IMCL (e.g., diacylglycerol (DAG) or ceramide) are associated with insulin sensitivity in an ethnicity-specific manner.

INTRODUCTION

The prevalence of type 2 diabetes is greater among African Americans (AA) vs. European Americans (EA) (1). Although this ethnic difference cannot be completely explained by obesity or lifestyle (2), lower insulin sensitivity among AA certainly may play a role. It has been extensively documented that insulin sensitivity is lower among AA vs. EA (3–7). However, the physiological basis for lower insulin sensitivity among AA is not known, and cannot be attributed to greater total or central adiposity.

Intramyocellular lipid (IMCL) is inversely associated with insulin sensitivity (8–10). IMCL is a mixture of lipid species that includes triacylglycerol, diacylglycerol (DAG), ceramide, and long-chain acyl-coenzyme A. Although triacylglycerol is present in the largest absolute amount, evolving evidence indicates that the other fatty acid metabolites, in particular DAG and ceramide, are actually the active agents conferring insulin resistance (11,12). Associations between IMCL and insulin sensitivity have been observed primarily in studies involving white populations (8–10). Whether this relationship occurs among AA has not been investigated. However, the accumulation of extramyocellular, “intermuscular,” lipid (EMCL), as assessed with a variety of techniques, is higher among subjects of African ancestry than other ethnic groups (13–15), and has been associated with insulin sensitivity (14) and type 2 diabetes (15). Research is needed to determine the physiological relevance of EMCL, and to determine whether either IMCL or EMCL explains lower insulin sensitivity among AA.

The physiological basis for individual differences in accumulation of lipid within skeletal muscle is not well understood, but may relate to the endocrine environment. The adipocyte-derived hormones adiponectin and leptin play a role in intramuscular lipid oxidation via stimulation of 5′-AMP-activated protein kinase (16), and thereby may contribute to regulation of intramuscular lipid stores. Women display greater IMCL than men (17), a difference that has been attributed to estradiol (18). AA have lower adiponectin (19) and higher estradiol (20) than EA, differences that may affect muscle lipid accumulation. The potential role of the endocrine environment in muscle lipid accumulation has not been extensively explored.

The objective of this study was to examine associations of skeletal muscle lipid and insulin sensitivity within healthy, normoglycemic premenopausal AA and EA women and girls. We examined both IMCL and EMCL using magnetic resonance spectroscopy (MRS). We tested the specific hypothesis that IMCL and/or EMCL would be higher among AA, and would be associated with insulin sensitivity among AA. We also explored the hypothesis that the endocrine environment (adiponectin, leptin, estradiol) would be associated with muscle lipid content.

Methods and Procedures

Participants

Participants were recruited from two ongoing studies (20,21). Inclusion criteria were female sex, age ≤41 years, nonsmoker, normoglycemic, and having achieved pubertal stage 5 (22). A portion of the subjects were recruited to be sedentary (<2 h per week of regular exercise), and to have a family history of obesity in at least one first-degree relative. In this portion of subjects, the range of maximal oxygen consumption during a treadmill walking test was 23.4–46.9 ml/kg/min (mean ± s.d.: 31.28 ± 4.84 ml/kg/min; only one woman had a maximal VO₂ >36 ml/kg/min). The remaining subjects were monitored with accelerometers. Within this group, total physical activity ranged from 7.6 to 82 min/day, >95% of which was in the “moderate” range. All but two individuals had <45 min per day. Two had activity counts equaling ~80 min/day. Exclusion criteria were presence or history of chronic illness or syndromes, major illnesses since birth, or use of medications known to affect body composition or physical activity. Testing was conducted at the baseline evaluation (no intervention had occurred). Both studies received approval by the Institutional Review Board for Human Use at the University of Alabama at Birmingham (UAB). For subjects who were minors, informed consent was obtained from both the minor and their parents.

Protocol

Data were collected during two testing visits. The first visit involved assessment of muscle lipid with MRS on an outpatient basis. The second visit involved assessment of insulin sensitivity, fasting hormone concentrations, body composition, and fat distribution, and involved an in-patient overnight stay in UAB’s General Clinical Research Center (GCRC). Diet was provided for the duration of the visit, and participants engaged in no physical activity. Subjects were given a standardized dinner meal at ~1800 h. A small snack was provided at ~1930 h and consumed by 2000 h. Only water or noncaffeinated, noncaloric beverages were allowed after 2000 h. Insulin sensitivity testing was conducted the following morning at ~0700 h.

Body composition

Total and regional body composition were measured through the Clinical Nutrition Research Center (CNRC) Metabolism Core Facility at UAB by dual-energy X-ray absorptiometry (DXA) using a GE Lunar Prodigy densitometer (GE Lunar, Madison, WI). Subjects were scanned wearing light clothing, lying supine with their arms resting at their sides. All scans were performed and analyzed according to manufacturer’s instructions with enCORE 2002 Version 6.10.029 software (GE Medical Systems LUNAR, Madison, WI). Whole-body composition analysis was performed using the “Standard” analysis sequence per manufacturer’s instructions. Leg fat mass was obtained from the standard analysis output, and reflected adipose tissue from both lower limbs. Specialized analysis was performed for calf fat mass using a “Region of Interest” (ROI) “Custom” analysis sequence. Right calf fat mass was quantified for comparison with the MRS data, which were acquired on the right calf. To assess calf tissue, a box was manually drawn with the upper boundary of the ROI placed immediately distal to the patella and the lower boundary placed below the sole of the foot.

Body fat distribution

Computed tomography was used to assess intra-abdominal and subcutaneous abdominal adipose tissue (IAAT and SAAT, respectively). Scans were performed using a GE HiLight/Advantage scanner (GE Medical Systems, Milwaukee, WI). Radiographic factors were 120 kV(p) and 40 mA. A single, 5 mm scan for 2 s approximately at the L4-L5 region was obtained while the subject was lying in a supine position with their arms stretched above their heads. A range of −190 to −30 Hounsfield units (HU) was used to measure cross-sectional area of adipose tissue. IAAT/SAAT within the HU range selected was characterized using an automated computer program. Analysis was performed on a Macintosh computer using the public domain NIH Image version 1.6 program (US National Institutes of Health, Bethesda, MD and available at http://rsb.info.nih.gov/nih-image/). Repeat analyses of 40 scans produced a coefficient of variation of <2%. Computed tomographic scans were not available on three participants. For statistical analysis, computed tomographic data on 30 AA and 24 EA were used.

Muscle lipid assessment

All measurements were performed at the UAB Center for Nuclear Imaging Research (CNIR) on a 4.1 T whole-body imaging and spectroscopy system interfaced to a Bruker console (Bruker Medical, Billerica, MA). IMCL and EMCL were measured using a slice selective 2D MRS sequence. Measurements were obtained from the right calf with the subjects lying in the supine position. Leg orientation, in reference to the magnetic field (B₀) and coil position, was determined from our earlier work and provided optimal splitting of IMCL and EMCL (23).

All MRS data sets were converted and reconstructed using a custom-built software package written in the MatLab version 6.1 (The MathWorks, Natick, MA). Individual spectra and spectra from ROIs were manually chosen and extracted for spectral analysis as previously described (23). For IMCL determination, a small summed ROI from a 6 × 6 set of voxels that was free from marbling and produced adequate peak splitting was identified in the soleus muscle. For EMCL determination, a large summed ROI of the soleus was manually drawn and included all voxels inside the facial layer of the muscle. The ROI was drawn one voxel within the fascia to avoid any potential contamination from fat associated with the fascia. Clear splitting of spectral peaks, as shown in Newcomer et al. (23), allowed this whole-muscle measure to reflect only EMCL (not IMCL). This measure of whole-muscle EMCL included visible “marbling” within the muscle. Spectra were analyzed by fitting the peak positions and areas through time domain fitting using Javabased magnetic resonance user interface (jMRUI) version 2.1 (European Consortium of Scientists). IMCL and EMCL in the soleus spectra were fit using previously published fitting models and sets of prior knowledge information (23). Lipid peak amplitudes were normalized to an external peanut oil phantom. Peak areas are expressed in arbitrary units relative to the phantom signal amplitude.

Insulin sensitivity

A frequently sampled, intravenous glucose tolerance test was conducted at ~0700 h after an overnight fast. Tests were performed on an in-patient basis, with the subjects admitted to the GCRC the day before the test, thus eliminating the potential for an acute bout of physical activity to affect the results. Glucose (50% dextrose at 300 mg/kg) was administered intravenously at time “zero.” Insulin (0.02 U/kg) was administered at 20 min following glucose infusion. A 32-sample protocol was used for the study involving adults (21), whereas a 21-sample protocol was employed with the young adult subjects to remain consistent with the protocol used in previous study visits (20). Sera were analyzed for glucose and insulin, and these values were entered into the MinMod computer program (version 3; Richard N. Bergman, University of Southern California, Los Angeles, CA) for determination of the insulin sensitivity index (S_I) (24). S_I data were not available on three participants. One S_I value was identified as an outlier (>3 s.d. from mean). For statistical analysis, data from 30 AA and 23 EA were used.

Serum analyses

All analyses were conducted in the Metabolism Core Laboratory of the GCRC and CNRC. Glucose concentration was assessed using the glucose oxidase method (Analox, Lunenburg, MA). Insulin was assayed in duplicate 100 μl aliquots using double-antibody RIA with reagents obtained from Linco Research (St Charles, MO). The assay sensitivity was 2.9 μIU/ml, mean intra-assay and interassay coefficients of variation were 4.0 and 3.5%, respectively. Adiponectin, leptin, and estradiol were measured by RIA as described (19,25).

Statistical analysis

Descriptive characteristics (mean, s.d.) were determined for the study population in all subjects combined and by ethnic group. ANOVA and analysis of covariance were used to compare characteristics between ethnic groups. IMCL and EMCL were compared both unadjusted and adjusted for total body fat in order to account for the potential confounding effect of total body adiposity when examining ethnic differences in muscle lipid content.

Pearson partial correlation analysis, adjusting for ethnicity and age, was used to examine relationships among body composition, fat distribution, and S_I. Pearson partial correlation analysis, adjusting for ethnicity, age, total body fat mass, and IAAT, was used to examine relationships among IMCL/EMCL, hormone concentrations, and S_I. Multiple linear regression analysis was used to identify variables independently related to S_I. Following inspection of multiple linear regression results, further partial correlation analysis was conducted within each ethnic group to examine the association between S_I and IMCL after adjusting for total body fat, IAAT, and age. Variables were log-transformed to ensure normality of distribution. All statistical tests were two-sided and were performed using a significance level of 0.05. Analyses were performed using SAS 9.2 (SAS Institute, Cary, NC).

RESULTS

Descriptive characteristics of the study population are presented in Table 1. EA women had more IAAT than AA, whereas AA had greater EMCL (after adjusting for total body adiposity). S_I was higher among EA vs. AA. No other differences between the two groups were detected. BMI ranged from 17.5 to 34.4 kg/m² within EA, and from 18.5 to 43.7 kg/m² within AA. %Fat ranged from 25.7 to 54.8 within EA, and from 23.4 to 51.3 within AA.

Table 1.

characteristics of the study population (mean ± s.d.)

	All (n = 57)	AA (n = 32)	EA (n = 25)
Age (year)	26.2 ± 10.5	23.78 ± 9.5	29.19 ± 11.0
Weight (kg)	74.4 ± 14.8	73.0 ± 15.1	76.2 ± 14.5
Height (cm)	164.8 ± 7.3	163.2 ± 6.8	166.8 ± 7.5
BMI (kg/m2)	27.3 ± 4.8	27.4 ± 5.1	27.3 ± 4.5
Total body fat mass (kg)	30.0 ± 10.3	28.5 ± 10.1	32.0 ± 10.5
Lean body mass (kg)	41.0 ± 5.2	41.0 ± 5.4	41.0 ± 5.2
% Fat	40.9 ± 7.7	39.7 ± 7.6	42.4 ± 7.8
Leg fat mass (kg)	12.6 ± 4.5	12.5 ± 4.4	12.7 ± 4.8
Calf fat mass (kg)a	1.4 ± 0.4	1.4 ± 0.4	1.4 ± 0.5
SAAT (cm2)b	314.1 ± 166.0	284.5 ± 149.6	351.2 ± 180.8
IAAT (cm2)b	60.9 ± 35.0	47.2 ± 25.4	78.1 ± 38.1*
IMCL (AU)	0.92 ± 0.40	0.90 ± 0.42	0.95 ± 0.37
EMCL (AU)	2.30 ± 0.97	2.49 ± 1.08	2.05 ± 0.76
Adjusted IMCL (AU)c	0.92 ± 0.07	0.92 ± 0.07	0.92 ± 0.07
Adjusted EMCL (AU)c	2.30 ± 0.17	2.50 ± 0.15	1.98 ± 0.17**
SI (×10−4 min−1/(μIU/ml))d	3.20 ± 1.64	2.69 ± 1.22	3.87 ± 1.89*
Adiponectin (μg/ml)e	10.2 ± 4.0	10.0 ± 3.9	10.4 ± 4.1
Leptin (ng/ml)e	23.5 ± 12.4	22.4 ± 9.5	25.0 ± 15.4
Estradiol (pg/ml)f	41.3 ± 34.6	39.3 ± 35.0	43.8 ± 34.8

AA, African Americans; EA, European Americans; EMCL, extramyocellular lipid; IAAT, intra-abdominal adipose tissue; IMCL, intramyocellular lipid; SAAT, subcutaneous abdominal adipose tissue; S_I, insulin sensitivity.

^an = 52, right calf only;

^bn = 54;

^cAdjusted for total body fat mass (±s.e.m.);

^dn = 53;

^en = 56;

^fn = 51.

^*P < 0.01 vs. AA;

^**P < 0.05 vs. AA.

Among all women combined, S_I was associated with IAAT, but not with other adipose tissue depots (Table 2). IMCL and EMCL were associated with all adipose tissue depots examined, except for IAAT. IMCL and EMCL were associated with each other. Serum adiponectin concentration was inversely associated with IMCL (Table 3), after adjusting for ethnicity, age, total fat mass, and IAAT. Serum estradiol was positively associated with EMCL. No other associations of S_I or the MRS measures with hormone outcomes were significant.

Table 2.

Pearson partial correlation analysis of insulin sensitivity (S_I) and muscle lipid measures with total and regional adipose tissue depots

	SI	IMCL	EMCL
Total body fat mass	−0.19	0.42a	0.30b
Leg fat mass	−0.07	0.42a	0.39a
Calf fat mass	−0.10	0.39a	0.39a
SAAT	−0.02	0.30b	0.33b
IAAT	−0.49c	0.19	0.15
IMCL	−0.14	—	0.29b
EMCL	−0.09	0.36b	—

Data are adjusted for ethnicity and age. All variables are log transformed.

EMCL, extramyocellular lipid; IAAT, intra-abdominal adipose tissue; IMCL, intra-myocellular lipid; SAAT, subcutaneous abdominal adipose tissue.

^aP < 0.01;

^bP < 0.05;

^cP < 0.001.

Table 3.

Pearson partial correlation analysis of insulin sensitivity (S_I) and muscle lipid measures with serum analytes

	SI	IMCL	EMCL
Adiponectin	0.16	−0.31a	0.15
Leptin	−0.13	−0.26b	0.22
Estradiol	−0.01	0.15	0.35a

Data are adjusted for ethnicity, age, total body fat, and intra-abdominal adipose tissue. All variables are log transformed.

EMCL, extramyocellular lipid; IMCL, intramyocellular lipid.

^aP < 0.05;

^bP = 0.08.

In multiple linear regression analysis, S_I was best explained by a model containing IMCL, IAAT, ethnicity, and the interaction term “ethnicity × IMCL” (Table 4). S_I was significantly lower among AA vs. EA, and significantly associated with IAAT. The significant interaction term indicated that S_I was associated with IMCL in EA, but not AA women. Subsequent partial correlation analysis within each ethnic group verified that S_I was associated with IMCL in EA (r = −0.54, P < 0.05), but not AA (r = 0.16, P = 0.424) women, after adjusting for total body fat, IAAT, and age. Simple correlation analysis likewise indicated that IMCL was associated with S_I in EA but not AA (Figure 1a,b). EMCL was not associated with S_I in either group after adjusting for total body fat, IAAT, and age (EA: partial r = 0.12, P = 0.614; AA: partial r = −0.06, P = 0.760).

Table 4.

Results of multiple linear regression analysis for the dependent variable log S_I; R² = 0.35

	Parameter estimate	s.e.m.	P
Intercept	0.50	0.57	0.382
IMCL	0.70	0.56	0.213
IAAT	−0.08	0.03	0.005
Ethnicitya	2.11	0.88	0.021
Ethnicity × IMCL	−1.98	0.85	0.024

IAAT, intra-abdominal adipose tissue; IMCL, intramyocellular lipid.

^aAfrican Americans coded as 0, European Americans coded as 1.

Figure 1.

IMCL is associated with S_I in (a) European American (r = −0.43, P < 0.05) but not (b) African American (r = 0.11, P = 0.561) women. Data shown are unadjusted. AU, arbitrary units; IMCL, intramyocellular lipid; S_I, insulin sensitivity.

DISCUSSION

The main objective of this study was to determine whether insulin sensitivity was related to IMCL or EMCL among healthy premenopausal women. In particular, we were interested in determining whether intramuscular lipid accumulation might contribute to lower insulin sensitivity among AA. We found that EMCL was higher among AA, but was not associated with S_I. IMCL was significantly associated with S_I only among EA. These observations suggest that IMCL may be a significant determinant of insulin sensitivity only among individuals of European ancestry.

Many, but not all (26–28), studies have reported an inverse association between IMCL and insulin sensitivity among sedentary individuals. This association has been observed in subjects who were lean and healthy (8,28,29), or heterogeneous regarding metabolic health (9,30–32). In contrast, within obese subjects, the association may be absent (27,28), and endurance athletes demonstrate a positive association between IMCL and insulin sensitivity (33). Our findings here contribute to this body of research in that we observed a strong association between IMCL and S_I among young, healthy, EA women.

However, our results indicated a clear ethnic difference in the association of IMCL with S_I, with no association being apparent among AA women. IMCL levels did not differ with ethnicity, indicating that the potential adverse effect of IMCL on insulin sensitivity differs with ethnic background. IMCL is a heterogeneous mixture of lipids, with triacylglycerol being the predominant species. Recent evidence has indicated that intramyocellular triacylglycerol is not the metabolically active molecule that confers insulin resistance. Lipid species that are more likely to play a mechanistic role in interfering with insulin signaling include DAG, a fatty acid derivative that can be converted to triglyceride via the enzyme DAG acyltransferase (11), and ceramide, which has been inversely associated with skeletal muscle insulin resistance, and is higher among insulin resistant vs. insulin-sensitive individuals (12). Thus, it is possible that ethnic differences exist in DAG, ceramide, or other fatty acid metabolites.

Alternatively, it is possible that the nature of whole-body insulin sensitivity, and thereby its determinants, differs with ethnicity. Whole-body insulin sensitivity, as assessed in this study with the S_I index, reflects the combined effects of both skeletal muscle and hepatic processes. Our previous work has suggested that the liver makes a larger contribution to whole-body S_I among AA than EA (34). It is possible that whole-body insulin sensitivity among EA primarily reflects insulin-stimulated skeletal muscle glucose uptake, as opposed to insulin suppression of hepatic glucose production, perhaps explaining the larger impact of IMCL on S_I.

The regulation of IMCL is incompletely understood. IMCL has been positively associated with IAAT in some (9,28) but not all studies (28), and it has been proposed that both “ectopic” depots accumulate when the storage capacity of subcutaneous adipose tissue is exceeded (35). We observed positive associations between IMCL and subcutaneous adipose depots, suggesting that, among largely nonobese subjects, IMCL may be deposited in parallel with deposition of lipid into subcutaneous adipose stores, not necessarily secondary to an “overflow.” Further, our finding of no association between IMCL and IAAT fails to support a coordinated deposition of these depots. It is possible that metabolic phenotype and/or obesity status determines whether lipid is deposited in subcutaneous vs. ectopic stores, as has been suggested (28).

Deposition of IMCL may occur secondary either to a shift in tissue-specific uptake of fatty acids from the circulation (36), or to a deficit in fatty acid oxidation within myocytes. Both processes may be affected by adiponectin, which promotes hyperplastic expansion of subcutaneous adipose tissue by increasing PPAR-γ activity (37). Both adiponectin and leptin activate 5′-AMP-activated protein kinase, an enzyme that plays a pivotal role in skeletal muscle fatty acid oxidation. We previously reported an inverse association between circulating leptin and muscle lipid accumulation among postmenopausal women (38). In the present study, we observed inverse associations of adiponectin (P < 0.5) and leptin (P = 0.08) with IMCL. These observations support a role for both hormones in minimizing accrual of IMCL.

Women consistently have higher IMCL than men (reviewed in ref. 17), leading to speculation that estrogen may promote deposition of IMCL to provide substrate for fat oxidation during exercise, which is higher among women than men. This concept is supported by the observation that estrogen administration to male rats increased IMCL (18). In this study, we observed a significant association between estradiol and EMCL (r = 0.35; P < 0.05). This observation supports a role for estrogens in promoting uptake of lipid by skeletal muscle, and adds to the current body of knowledge by documenting the association of estradiol with extramyocellular lipid depots.

In contrast to IMCL, EMCL has been little studied. Only one study that we are aware of, involving lean and obese premenopausal women, has reported an independent inverse association between EMCL and insulin sensitivity as assessed with intravenous glucose tolerance test (14). An inverse univariate association of EMCL with insulin sensitivity as assessed with clamp was reported in a heterogeneous group of men and women, but it was not stated whether the relationship was independent of total or abdominal adiposity, or whether metabolic status affected the association (39). Within obese premenopausal women, EMCL was not associated with an OGTT-based measure of insulin sensitivity (40). Within men of African ancestry, EMCL was greater among those with type 2 diabetes (15). We observed that EMCL was higher among AA vs. EA, as has been observed elsewhere (13–15). In our study, we observed close correlations of EMCL with both IMCL and subcutaneous adipose depots. However, EMCL was not associated with S_I among EA or AA. These observations suggest that, in young, healthy, girls and women, EMCL does not share the adverse metabolic properties of IMCL. Likewise, our data suggest that EMCL is not regulated by adipokines, as we observed significant inverse associations of adipokines with IMCL but not EMCL. The possible population-specific association of EMCL with metabolic disease deserves further study.

Strengths of the study included robust measures of insulin sensitivity, IMCL and EMCL, body composition, and body fat distribution. Limitations included the relatively small sample size, the cross-sectional nature of the study, and the inclusion of only healthy young women. Results may not be generalizable to other populations. Further, lipid within only the soleus was investigated; associations involving muscle groups with lower oxidative capacity may differ.

In conclusion, among healthy young women, IMCL was associated with insulin sensitivity among EA but not AA. The physiological basis for this ethnic difference, and its implications for disease risk, deserves further study. Our results suggest that greater risk for type 2 diabetes among AA is not related to IMCL; however, it is important to investigate other intramyocellular fatty acid metabolites such as DAG and ceramide. The metabolic implications of greater EMCL observed among AA should be explored.