SEARCH

SEARCH BY CITATION

Keywords:

  • CT scan;
  • high-throughput risk screening;
  • non-apparent visceral obesity;
  • plasma amino acid.

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conflicts of Interest Statement
  8. Acknowledgements
  9. References

What is already known about this subject

  • • 
    Asians with metabolic complications associated with obesity, a low body mass index and a low waist circumference have a greater proportion of visceral adipose tissue for a given amount of total body fat compared with Europeans.
  • • 
    Apparent obese humans and obese animal models show an elevation of branched-chain amino acid levels in plasma.
  • • 
    A multivariate logistic regression model of plasma free amino acids has been used to screen for several types of cancers in clinical settings.

What this study adds

  • • 
    A specific formula incorporating six amino acid values (Ala, Gly, Glu, Trp, Tyr and branched-chain amino acid) was developed for discrimination of subjects with high visceral fat area by multivariate logistic regression analyses.
  • • 
    The generated amino acid formula was strongly correlated with visceral fat area in both apparent and non-apparent obese subjects.
  • • 
    Measuring plasma free amino acids can be used to distinguish the non-apparent visceral obesity in clinical settings in Asian populations.

Summary

Metabolic complications associated with obesity are becoming more common among Japanese subjects. However, visceral fat accumulation is not always apparent by measuring body mass index (BMI) or waist circumference in Asian populations because of the physiological characteristics particular to those ethnicities. Excess visceral fat accumulation raises the odds ratio for developing cardiovascular disease. Thus, high-throughput determination of the amount of abdominal adipose tissue is necessary. We hypothesized that accumulating visceral fat alters the peripheral amino acid profile and that a multivariate logistic regression model of plasma free amino acids can distinguish visceral obesity. A total of 1449 Japanese subjects (985 males and 464 females) who had undergone a comprehensive health screening were enrolled in this study. The visceral fat area was determined using computed tomography imaging, and a plasma free amino acid index to identify high visceral fat areas (≥100 cm2) was developed. The sensitivity and specificity values of the generated amino acid index were 80% and 65%, respectively. In particular, the sensitivity of the generated index to identify subjects with non-apparent visceral obesity (BMI < 25 kg m−2; visceral fat area ≥ 100 cm2) was much greater than that of the waist circumference (73% vs. 46%, respectively). This index's high sensitivity and specificity may be the result of specific alterations in the patients' amino acid profiles, which were specifically correlated with the visceral fat areas and not with subcutaneous fat areas. This profile can be used as a predictor of elevated visceral obesity and a risk assessment tool for metabolic complications in Asian populations.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conflicts of Interest Statement
  8. Acknowledgements
  9. References

The prevalence of obesity is increasing worldwide, and Aboriginal, Chinese and South Asian populations are all affected by the obesity epidemic (1). Obesity has become a major health concern because it is related to a number of cardiovascular and metabolic disorders, including insulin resistance, dyslipidemia and non-alcoholic fatty liver disease (NAFLD) (2,3). The degree of excess weight is not the only determinant for the risk of overweight-related complications because the distribution pattern of fat throughout the body is an important determinant of future complications (2–4). Visceral fat tissue, rather than subcutaneous fat tissue, secretes tumour necrosis factor-α, interleukin-6 and other adipokines that strongly correlate with the development of metabolic complications (2,3) and with the occurrence of cardiovascular disease (2,3). In fact, visceral fat accumulation is associated with higher risks of myocardial infarction and cerebral infarction even if the subject does not have hypertension, diabetes or hyperlipidemia (5).

Race and ethnicity affect an individual's susceptibility to becoming overweight or obese and to developing co-morbid complications (2). Asians with metabolic complications associated with obesity (dyslipidemia, NAFLD, insulin resistance and type 2 diabetes), a low body mass index (BMI) and a low waist circumference (WC) have a greater proportion of visceral adipose tissue for a given amount of total body fat compared with Europeans (6,7). Thus, it is difficult to predict obesity-related disorders among Japanese subjects because the visceral fat accumulation is not always apparent when using BMI or WC measurements. Therefore, it is imperative to determine the amount of abdominal adipose tissue in Japanese subjects via high-throughput measurements to evaluate the risks of metabolic complications associated with obesity.

An elevation of plasma branched-chain amino acid (BCAA) levels has been reported in apparent obese humans and animal models of obesity (8,9). The decline in the BCAA-catabolizing capacity of adipose tissue determines the circulating BCAA levels in obese mice (10). The progress in metabolomics has enabled healthcare providers to obtain high-throughput measurements of several amino acids (11–13). Moreover, a recently developed, novel multivariate logistic regression model of plasma free amino acids has been used to discriminate various disease states in rat models of type 1 and type 2 diabetes (14), to determine the progression of liver fibrosis in chronic hepatitis C in humans (15) and to screen for several types of cancers (13,16). Thus, we hypothesized that an accumulation of visceral fat changes the peripheral amino acid profile and that the multivariate logistic regression model of plasma free amino acids can distinguish non-apparent visceral obesity. Our study aimed to examine the use of systemic amino acid changes to predict the visceral fat area (VFA) in clinical settings, which could advance the diabetes/metabolic syndrome risk assessment for Asian populations.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conflicts of Interest Statement
  8. Acknowledgements
  9. References

Subjects

A total of 1449 Japanese subjects (985 males, mean age of 58.0 ± 0.4 years; and 464 females, mean age of 58.5 ± 0.5 years) who had undergone the Ningen Dock (or human dry dock) comprehensive medical check-up system (17), which is unique to Japan, between January 2008 and June 2009 at the Center for Multiphasic Health Testing and Services, Mitsui Memorial Hospital and the Kameda Medical Center Makuhari were enrolled. Subjects were not provided with medical treatment before the examination and blood sampling. This study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethical Committees of Mitsui Memorial Hospital and Kameda Medical Center Makuhari. All subjects gave their informed consent for inclusion before they participated in the study. This study population did not show a predisposition to any particular disease and did not have serious health problems.

The subjects were divided into four groups according to their BMI and VFA values, which were calculated from computed tomography (CT) images (Fig. 1). Obesity was defined as a BMI of 30 kg m−2 or greater, and overweight was defined as a BMI of 25 kg m−2 or greater. Thus, a BMI of 25 kg m−2 was used as a threshold as follows: high BMI (≥25 kg m−2; n = 460) and low BMI (<25 kg m−2; n = 989). Previous studies (18) have reported that the mean number of metabolic risk factors in Japanese subjects with VFA values greater than or equal to 100 cm2 is significantly higher than those with VFA values less than 100 cm2 independent of BMI. Therefore, a VFA value of 100 cm2 was used as a threshold as follows: high VFA (≥100 cm2; n = 867) and low VFA (<100 cm2; n = 582). The following four categories, which were defined based on the BMI and VFA values, were used in this study: healthy subjects, HS (BMI < 25 kg m−2; VFA < 100 cm2; n = 522); apparent obese subjects, AO (BMI ≥ 25 kg m−2; VFA < 100 cm2; n = 60); non-apparent visceral obese subjects, NAVO (BMI < 25 kg m−2; VFA ≥ 100 cm2; n = 467); and obese subjects, Obesity (BMI ≥ 25 kg m−2; VFA ≥ 100 cm2; n = 400).

image

Figure 1. Visceral obesity is not always apparent in Japanese subjects. (a) Subjects were divided into four groups according to the body mass index (BMI) and visceral fat area (VFA) values as follows: HS, healthy subjects (BMI < 25; VFA < 100 cm2; n = 522); AO, apparent obese subjects (BMI ≥ 25; VFA < 100 cm2; n = 60); NAVO, non-apparent obese subjects (BMI < 25; VFA ≥ 100 cm2; n = 467); and Obesity, obese subjects (BMI ≥ 25; VFA ≥ 100 cm2; n = 400). The ages are presented as the mean ± standard deviation. (b) Representative images of visceral fat in the HS, AO, NAVO and Obesity subjects. The distribution of abdominal fat was measured by the FatScan software based on computed tomography scans at the level of the umbilicus. The VFA is represented in red, and the subcutaneous fat area is represented in pink.

Download figure to PowerPoint

Analyses of metabolic parameters

Blood samples were taken from the subjects after an overnight fast. Serum levels of total cholesterol, high-density lipoprotein (HDL) cholesterol and triglycerides were determined enzymatically. Plasma glucose was measured with the hexokinase method, and haemoglobin A1c (HbA1c) was determined using the latex agglutination immunoassay. The systolic and diastolic blood pressures were measured twice on the same day, and the mean value was used in the analyses. The BMI was calculated as weight in kilograms divided by height in meters squared. The Japanese Committee of the Criteria for Metabolic Syndrome (JCCMS) adopted the WC, which was measured at the level of the umbilicus with a non-stretchable tape in late expiration while standing, as an essential component risk factor for diagnosing metabolic syndrome, and WC cut-off points of 85 and 90 cm for males and females, respectively, were selected as thresholds in Japan (19).

Measurement of the abdominal fat area by CT scan

The subcutaneous fat area (SFA) and VFA visualized on a CT scan at the level of the umbilicus were measured using FatScan software (N2 System Co., Osaka, Japan). All CT scans were performed in the supine posture at the umbilical level using a CT scanner (SOMATOM Sensation Cardiac 64, Siemens, Munich, Germany) based on the Japanese guidelines for obesity treatment by the Japan Society for the Study of Obesity. The VFA was defined as the intraperitoneal fat bound by the parietal peritoneum or transverse fascia excluding the vertebral column and paraspinal muscles (Fig. 1b). The SFA was defined as the fat superficial to the abdominal and back muscles. A region of interest drawn around the external margin of the dermis was used to calculate the total abdominal fat area. The SFA was obtained by subtracting the VFA from the total abdominal fat.

Plasma amino acid profiling

Plasma samples were prepared as previously described (11–13), and amino acid analysis was performed according to the recently established method of high-performance liquid chromatography (HPLC)/electrospray ionization (ESI)/mass spectrometry (MS) following derivatization (11–13). Briefly, an MSQ Plus LC/MS system (Thermo Fischer Scientific, Waltham, MA, USA) equipped with an ESI source was used in the positive ionization mode for selected ion monitoring. The Xcalibur (ThermoFisher Scientific Inc., Waltham, MA, USA) software (version 1.4 SR1) was used for the data collection and processing. The HPLC separation system consisted of an L-2100 (pump), an L-2200 (autosampler) and an L-2300 (column oven) (Hitachi High-Technologies Corporation, Tokyo, Japan). A Wakosil-II 3C8-100HG column (100 mm × 2.1 mm; 3 µm) (Wako Pure Chemical Industries, Osaka, Japan) was used for the separation. The mobile phase included eluent A (25 mM ammonium formate in water) and eluent B (water : acetonitrile = 40:60). In this study, the following 21 compounds were measured: alanine (Ala), alpha-aminobutyric acid (ABA), arginine (Arg), asparagine (Asn), citrulline (Cit), glutamate (Glu), glutamine (Gln), glycine (Gly), histidine (His), isoleucine (Ile), leucine (Leu), lysine (Lys), methionine (Met), ornithine (Orn), phenylalanine (Phe), proline (Pro), serine (Ser), threonine (Thr), tryptophan (Trp), tyrosine (Tyr) and valine (Val). The BCAA level was calculated as the sum of the Val, Leu and Ile levels.

Statistical analyses

Dunnett's test was used to analyse differences in demographic variables, biochemical variables and plasma amino acid concentrations among the four groups categorized by the VFA and BMI values. Significance was set at P < 0.05. To evaluate the correlations among the variables, VFA and SFA, Pearson's correlation coefficients were calculated.

Multivariate logistic regression analyses for discriminating high VFA subjects

We previously constructed and tested a diagnostic index based on plasma amino acid concentrations, known as the ‘AminoIndex technology’(13), to compress multidimensional information from plasma amino acid profiles into a single dimension and to maximize the differences between the patients and the controls. The search for an optimal index to discriminate the high VFA subjects was performed using a previously described algorithm (13) that was based on a multivariate logistic regression model with the amino acid concentrations as variables (14–16). All possible combinations of variables were investigated with cross-validation, and an assessment of the discriminatory power was performed with the area under the receiver operating characteristic (ROC) curve (ROC_AUC) values (20). In this study, the maximum number of variables for each regression formulation was restricted to six to limit the degrees of freedom and to avoid over-fitting. In addition, the variance inflation factor was calculated to determine the degree of multi-collinearity when its cut-off value was set at 10. The best model was defined as the model with the minimum Akaike's information criterion (21). All of the statistical and multivariate analyses were performed with MATLAB (The MathWorks, Natick, MA, USA), GraphPad Prism (GraphPad Software, La Jolla, CA, USA), JMP 9.0.0 program (SAS Institute Inc., Cary, NC, USA) and SigmaPlot 12.1 (Systat Software, Inc., Chicago, IL, USA).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conflicts of Interest Statement
  8. Acknowledgements
  9. References

Multivariate logistic regression analysis to discriminate the high VFA group

The age, sex and body weight profiles are shown in Fig. 1a, and representative CT scan images of the VFAs are shown in Fig. 1b. Table 1 summarizes the metabolic parameters, and Table 2 shows the plasma amino acid concentrations. Intriguingly, the plasma amino acid profiles (Table 2) were highly altered in the NAVO and Obesity populations, but only the Pro level was changed in the AO subjects compared with the HS population. The plasma levels of Glu, Ser, Pro, Gly, His, Ala, ABA, Tyr, Val, Orn, Met, Lys, Ile, Leu, Phe and Trp were changed in the NAVO and Obesity populations compared with the HS population. For discrimination between the high and low VFA groups, a formula incorporating six amino acid values (Ala, Gly, Glu, Trp, Tyr and BCAA) was developed, and the following formula was modelled:

  • image
Table 1. Characteristics of the subjects enrolled in the study
GroupHSAONAVOObesity
  1. Dunnett's test was used to detect significant differences between the healthy subjects and the other groups.

  2. Significant differences are shown as *P < 0.05, **P < 0.01 and ***P < 0.001.

  3. AO, apparent obese subjects; BMI, body mass index; DBP, diastolic blood pressure; HbA1c, haemoglobin A1c; HDL, high-density lipoprotein; HS, healthy subjects; LDL, low-density lipoprotein; NAVO, non-apparent obese subjects; Obesity, obese subjects; SBP, systolic blood pressure; SFA, subcutaneous fat area; VFA, visceral fat area; WC, waist circumference.

n 52260467400
(Male, Female)(252, 270)(26, 34)(92, 375)(94, 306)
Age (years)56.4 ± 0.650.2 ± 1.4***62.3 ± 0.5***57.0 ± 0.6
Body weight (kg)56.5 ± 0.471.2 ± 1.2***63.6 ± 0.3***76.5 ± 0.5***
BMI (kg m−2)21.2 ± 0.126.6 ± 0.2***23.1 ± 0.1***27.6 ± 0.1***
WC (cm)77.9 ± 0.389.9 ± 0.8***85.1 ± 0.2***94.8 ± 0.3***
Glucose (mg dL−1)93.2 ± 0.5101.1 ± 3.7**102.7 ± 1.0***107.3 ± 1.1***
HbA1c (%)5.3 ± 0.05.6 ± 0.2*5.5 ± 0.0***5.6 ± 0.0***
SBP (mmHg)118.1 ± 0.8123.0 ± 1.8127.9 ± 0.9***132.4 ± 0.8***
DBP (mmHg)74.0 ± 0.577.0 ± 1.179.9 ± 0.5***83.1 ± 0.5***
HDL cholesterol (mg dL−1)66.1 ± 0.756.3 ± 1.8***57.2 ± 0.6***52.8 ± 0.6***
LDL cholesterol (mg dL−1)121.7 ± 1.2124.6 ± 3.5126.9 ± 1.4*127.6 ± 1.6**
Triglycerides (mg dL−1)92.6 ± 2.2107.5 ± 8.9134.4 ± 5.6***160.4 ± 5.6***
SFA (cm2)119.8 ± 2.3214.8 ± 9.3***148.7 ± 2.1***225.7 ± 4.0***
VFA (cm2)61.0 ± 1.080.7 ± 2.0***143.4 ± 1.6***168.9 ± 2.4***
Table 2. Plasma amino acid profile for each group
GroupHSAONAVOObesity
  1. Dunnett's test was used to detect significant differences between the healthy subjects and the other groups.

  2. Significant differences are shown as *P < 0.05, **P < 0.01 and ***P < 0.001.

  3. ABA, alpha-aminobutyric acid; AO, apparent obese subjects; HS, healthy subjects; NAVO, non-apparent obese subjects; Obesity, obese subjects.

n 52260467400
(µmol L−1)    
ABA22.0 ± 0.321.9 ± 0.823.3 ± 0.3**24.2 ± 0.4***
Alanine363.5 ± 3.5376.5 ± 10.6404.3 ± 3.9***427.9 ± 4.5***
Arginine97.2 ± 0.894.5 ± 2.197.9 ± 0.897.3 ± 0.9
Asparagine45.3 ± 0.344.7 ± 0.944.8 ± 0.344.0 ± 0.4*
Citrulline33.7 ± 0.432.0 ± 0.935.0 ± 0.4*32.5 ± 0.4
Glutamine591.3 ± 3.5594.7 ± 11.1582.4 ± 3.4575.6 ± 3.9**
Glutamate36.7 ± 0.742.1 ± 2.449.5 ± 0.8***60.6 ± 1.2***
Glycine259.4 ± 2.8249.1 ± 7.5231.0 ± 2.1***223.5 ± 2.5***
Histidine79.7 ± 0.580.9 ± 1.382.5 ± 0.5***82.6 ± 0.6***
Isoleucine65.7 ± 0.670.3 ± 2.075.2 ± 0.8***80.2 ± 0.9***
Leucine113.6 ± 1.0118.1 ± 3.1128.9 ± 1.2***135.5 ± 1.2***
Lysine198.5 ± 1.5195.4 ± 4.2209.8 ± 1.4***213.5 ± 1.6***
Methionine27.0 ± 0.327.5 ± 0.628.0 ± 0.2*28.8 ± 0.3***
Ornithine54.2 ± 0.652.9 ± 1.759.5 ± 0.7***59.1 ± 0.7***
Phenylalanine61.9 ± 0.563.2 ± 1.065.5 ± 0.5***67.8 ± 0.6***
Proline140.7 ± 1.7156.0 ± 6.2*154.3 ± 1.8***165.3 ± 2.1***
Serine115.3 ± 0.8113.0 ± 2.4109.4 ± 0.8***109.3 ± 0.9***
Threonine115.3 ± 1.1119.6 ± 4.1116.4 ± 1.1118.9 ± 1.3
Tryptophan57.8 ± 0.459.2 ± 1.160.9 ± 0.5***61.8 ± 0.5***
Tyrosine73.2 ± 0.676.8 ± 2.181.5 ± 0.7***86.4 ± 0.8***
Valine224.2 ± 1.7227.1 ± 4.3252.9 ± 2.0***260.7 ± 2.2***

Performance of WC and plasma amino acids for discriminating between the high and low VFA groups

The sensitivity and specificity values for the WC cut-off points are shown in Table 3. The cut-off values of 85 cm for males and 90 cm for females, which have been selected by the JCCMS, were almost optimal to detect visceral obesity in males and optimal in females in this study. Figure 2 shows the distribution plots and ROC curve for the WC and amino acid index. The ROC_AUC of the obtained amino acid index was 0.81 (95% confidence interval: 0.78–0.83), and the sensitivity and specificity values of the optimal cut-off point of the generated index were 80% and 65%, respectively. Although the ROC_AUC values were similar between the WC and the amino acid index, it is plausible that the sensitivity to detect a VFA accumulation in the NAVO population was higher in the diagnosis by the amino acid index. Thus, we next visualized the distribution plots using the axis of BMI and the amino acid index and evaluated the sensitivity in Fig. 3. Marked distinctions in the sensitivity between the amino acid index and the WC were found in the NAVO (73% vs. 46%, respectively) and AO (48% vs. 25%) populations. Comparison of the plots of the index obtained from the plasma amino acids (Fig. 3c) and the WC (Fig. 3d) against the BMI shows clearly that the AO and NAVO populations were plotted in the appropriate regions by the generated amino acid index compared with the WC.

Table 3. The sensitivity and specificity of the waist circumference cut-off value to discriminate the high VFA group (≥100 cm2)
  1. Sensitivity, TP/(TP + FN); specificity, TN/(TN + FP); PPV, TP/(TP + FP); NPV, TN/(TN + FN); and efficiency, (TP + TN)/(TP + FP + TN + FN).

  2. TP, true positive, TN, true negative; FP, false positive; FN, false negative; PPV, positive predictive values; NPV, negative predictive values, in each cut-off point.

Male     
Cut-off (cm)Sensitivity (%)Specificity (%)PPV (%)NPV (%)Efficiency (%)
 904593944360.0
 857475875674.2
 809542797878.4
 759917739174.0
Female     
 Cut-off (cm)Sensitivity (%)Specificity (%)PPV (%)NPV (%)Efficiency (%)
 904792817274.1
 856481697773.9
 808864628973.7
 759641529463.4
image

Figure 2. The ROC curve of the WC and the multivariate regression model of plasma amino acids for discrimination between the high and low VFA groups. (a) Distribution plots of the WC against the high (≥100 cm2) and low (<100 cm2) VFA groups in a male Japanese population. The dotted line represents the cut-off point of 85 cm, which follows the guideline by the Japanese Committee of the Criteria for Metabolic Syndrome (JCCMS). (b) The ROC curve of the WC to distinguish the high VFA group from the low VFA group in a male Japanese population. The circle represents the cut-off point of 85 cm. (c) Distribution plots of the WC against the high and low VFA groups in a female Japanese population. The dotted line represents the cut-off point of 90 cm, which follows the guidelines of the JCCMS. (d) The ROC curve of the WC to distinguish the high VFA group from the low VFA group in a female Japanese population. The circle represents the cut-off point of 90 cm. (e) Distribution plots of the multivariate regression model of plasma amino acids against the high and low VFA groups. The dotted line represents the cut-off point of 0.050. (f) The ROC curve of the multivariate regression model of plasma amino acids to distinguish the high VFA group from the low VFA group. The circle represents the cut-off point of 0.050. AO, apparent obese subjects; HS, healthy subjects; NAVO, non-apparent obese subjects; Obesity, obese subjects; ROC, receiver operating characteristic; VFA, visceral fat area; WC, waist circumference.

Download figure to PowerPoint

image

Figure 3. The multivariate regression model of plasma amino acids is more capable of discriminating non-apparent visceral obesity than waist circumference. (a) The sensitivities of the multivariate regression model of plasma amino acids in each group were as follows: HS (66%), AO (48%), NAVO (73%) and Obesity (88%). (b) The sensitivities of the WC in each group were as follows: HS (90%), AO (25%), NAVO (46%) and obesity (93%). (c) Scatter plots of the BMI against values of the multivariate regression model of plasma amino acids using a cut-off value of 0.050. (d) Scatter plots of the BMI against the WC, which had the WC cut-off points for males and females set at 85 and 90 cm, respectively, following the guidelines of the Japanese Committee of the Criteria for Metabolic Syndrome. AO, apparent obese subjects; BMI, body mass index; HS, healthy subjects; NAVO, non-apparent obese subjects; Obesity, obese subjects; WC, waist circumference.

Download figure to PowerPoint

The WC and BMI were highly correlated with the SFA

Pearson's correlation coefficients of general metabolic variables versus the VFA and SFA were summarized, and the discriminative performance for each parameter was evaluated by ROC analysis, as shown in Table 4. As expected, there were close relationships between the Welch t-test and the ROC_AUC due to the binomial distribution in the two groups. Figure 4 shows scatter plots of the WC and BMI against the VFA and SFA. Specifically, strong correlations were found in the SFA versus the WC (r = 0.71) and in the SFA versus the BMI (r = 0.72), and both of these correlation coefficients were greater than the correlation coefficients of the VFA versus the WC (r = 0.66) and the VFA versus the BMI (r = 0.61).

Table 4. The ability of each metabolic variable to identify the high VFA group (≥100 cm2) and its Pearson's correlation coefficients with the VFA and SFA
 ROC_AUC (95%CI)Coefficient of correlation
VFASFA
  1. The ROC_AUC values were evaluated with 95% CI by the following P values: *P < 0.05, **P < 0.01 and ***P < 0.001. The variables were ordered by the ROC_AUC values. Pearson's correlation coefficients were used to evaluate the correlations between the metabolic variables and the VFA and SFA.

  2. ABA, alpha-aminobutyric acid; BMI, body mass index; CI, confidence interval; DBP, diastolic blood pressure; HbA1c, haemoglobin A1c; HDL, high-density lipoprotein; LDL, low-density lipoprotein; ROC_AUC, area under the receiver operating characteristic curve; SBP, systolic blood pressure; SFA, subcutaneous fat area; VFA, visceral fat area; WC, waist circumference.

WC (cm)0.84 (0.82–0.86)***0.660.71
BMI (kg m−2)0.82 (0.80–0.84)***0.610.72
Triglycerides (mg dL−1)0.71 (0.68–0.74)***0.300.08
Glucose (mg dL−1)0.69 (0.66–0.71)***0.310.09
HDL cholesterol (mg dL−1)0.68 (0.66–0.71)***−0.33−0.15
SBP (mmHg)0.68 (0.66–0.71)***0.330.17
DBP (mmHg)0.68 (0.65–0.71)***0.330.14
HbA1c (%)0.64 (0.61–0.67)***0.240.07
LDL cholesterol (mg dL−1)0.55 (0.52–0.58)**0.070.14
Glutamate0.75 (0.73–0.78)***0.490.21
Valine0.71 (0.68–0.74)***0.420.08
Leucine0.70 (0.68–0.73)***0.390.04
Isoleucine0.69 (0.66–0.72)***0.390.06
Tyrosine0.68 (0.65–0.71)***0.400.18
Alanine0.67 (0.64–0.70)***0.330.15
Glycine0.66 (0.63–0.69)***−0.31−0.12
Phenylalanine0.63 (0.61–0.66)***0.270.07
Proline0.63 (0.60–0.66)***0.240.05
Lysine0.62 (0.59–0.65)***0.230.02
Ornithine0.61 (0.59–0.64)***0.200.02
Tryptophan0.60 (0.57–0.63)***0.200.02
Serine0.59 (0.56–0.62)***−0.17−0.04
Methionine0.58 (0.55–0.61)***0.16−0.01
Histidine0.57 (0.54–0.60)***0.140.06
ABA0.57 (0.54–0.60)**0.110.06
Glutamine0.55 (0.52–0.58)*−0.09−0.05
Asparagine0.53 (0.50–0.56)−0.08−0.15
Threonine0.52 (0.49–0.55)0.050.00
Arginine0.51 (0.48–0.54)0.01−0.08
Citrulline0.51 (0.48–0.54)0.01−0.12
Amino acid index0.81 (0.78–0.83)***0.610.23
image

Figure 4. WC and BMI values are correlated with both the VFA and the SFA, but the multivariate regression model of plasma amino acids is only correlated with the VFA. (a) WC (cm), BMI (kg m−2), Glu (µmol L−1), Val (µmol L−1), Leu (µmol L−1), Ile (µmol L−1), Tyr (µmol L−1), Ala (µmol L−1) and Trp (µmol L−1) are plotted against the VFA (cm2) and SFA (cm2). The plots against the VFA have positive Pearson's correlation coefficients, and the plot of Gly (µmol L−1) against the VFA has a negative Pearson's correlation coefficient. (b) Scatter plots of the multivariate regression model of plasma amino acids against the VFA (r = 0.61) and (c) against the SFA (r = 0.23). The 95% probability ellipses are represented by circles. Ala, alanine; AO, apparent obese subjects; Glu, glutamate; Gly, glycine; HS, healthy subjects; Ile, isoleucine; Leu, leucine; NAVO, non-apparent obese subjects; Obesity, obese subjects; SFA, subcutaneous fat area; Trp, tryptophan; Tyr, tyrosine; Val, valine; VFA, visceral fat area; WC, waist circumference.

Download figure to PowerPoint

The plasma amino acid profile and regression model were correlated with the VFA but not with the SFA

The plasma amino acid profiles are shown in Table 4, and the scatter plots of the amino acid levels against the VFA and SFA are illustrated in Fig. 4a. Most of the amino acid levels were positively correlated with the VFA. In contrast, Gly, Ser, Gln and Asn were significantly lower in the high VFA group than in the low VFA group, and a negative correlation with VFA existed for these amino acids. Glu and Val produced higher Pearson's correlations (more than 0.4), especially for the VFA, and lower correlation coefficients were found for the SFA. Additionally, Leu, Ile, Tyr and Ala had similar correlations that were specific to the VFA and higher than the SFA. Figure 4b and c show the scatter plots of the values of the multivariate regression model of plasma amino acids against the VFA and SFA, respectively. Specifically, strong correlation coefficients were found between the obtained index and the VFA (r = 0.61), and a relatively weak correlation was found between the amino acid index and the SFA (r = 0.23). By comparing the plots of the WC (Fig. 4a) and the index obtained from the amino acid levels (Fig. 4b) against the VFA, it appears that the amino acid index includes many more NAVO subjects at the higher values of the y-axis and AO subjects at the lower values of the y-axis.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conflicts of Interest Statement
  8. Acknowledgements
  9. References

In Asian individuals, including Japanese individuals, who have metabolic complications associated with obesity that are not always apparent from BMI or WC measurements (7,22), a high-throughput method for determining the amount of abdominal adipose tissue is critically needed (2). Abdominal CT scans (Fig. 1) and WC (Fig. 2) are the most frequently used parameters for VFA predictions in clinical settings (23,24). However, CT scanners are expensive and are low-throughput in nature. Additionally, CT scanners frequently have mechanical problems that require money, time and labour to repair, and these dysfunctions potentially present radiation threats. The WC is a non-invasive method to measure the VFA, but it is subject to variations depending on the person taking the measurement and on the measurement site used. Moreover, WC measurements include both the subcutaneous fat and the visceral fat (Fig. 3), which may lead to an underestimation of the VFA (25). The data in Fig. 2a and c show that it is fundamentally difficult to discriminate between the NAVO and AO populations and the other populations by measuring the WC. Figure 3 shows the low sensitivity of the WC for detecting NAVO in Japanese subjects (46% by WC vs. 73% by the amino acid index), thus indicating that the WC is an inadequate measure of body fat distribution for specific regions of abdominal adiposity.

Conversely, the plasma amino acid levels were specifically altered in accordance with the degree of VFA in this study, and the levels were sufficiently sensitive to detect NAVO with a multivariate logistic regression model. Amino acid levels in biological fluids change in response to metabolic alterations during the course of various diseases, and Fischer's ratio is one of a few classic indicators used to monitor hepatic encephalopathy (26). In this study, a formula incorporating six amino acid measurements (Ala, Gly, Glu, Trp, Tyr and BCAA), with an ROC_AUC of 0.81 (Fig. 2f), was developed and was successful at accurately detecting NAVO subjects (Figs 2e and 3). The amino acid index was able to distinguish the NAVO and AO subjects from other populations, possibly because of the marked differences in the correlation coefficients for the VFA and SFA. Table 4 and Fig. 4 show that the amino acid index specifically correlated with the VFA (correlation coefficients of 0.23 and 0.61 with regard to the SFA and VFA, respectively). The correlation coefficient of the WC against the SFA was as high as 0.71, which contributed to the low sensitivity and specificity for discriminating the NAVO and AO populations from the other populations (Fig. 3).

It is important to understand why some plasma amino acid levels vary depending on the degree of VFA. Many studies have supported the hypothesis of a specific role for intra-abdominal fat accumulation in the link between visceral obesity and metabolic dysfunction (4). In particular, the removal of visceral adipose tissue by an omentectomy results in decreased glucose and insulin levels in humans (27); however, the removal of subcutaneous adipose tissue by liposuction does not always improve these levels (28). The causative mechanism is attributed to the release of adipocytokines and free fatty acids from the visceral adipose tissue, and these molecules exert adverse effects on hepatic metabolism by draining into the portal vein (2,3). This type of metabolic disturbance might affect the systemic amino acid concentrations. Although our present study (Fig. 4, Tables 2 and 4) provides strong evidence for the importance of plasma BCAAs as specific markers for visceral fat accumulation, it is important to note that our results do not indicate that the ingestion of dietary protein or BCAAs accelerates the accumulation of visceral adipose tissue. Many papers report that the dietary ingestion of BCAAs is responsible for some of the beneficial effects of high-protein diets (29), including improved body weight control by muscle volume tuning (30), and amelioration of hepatic de novo lipogenesis and prevention of hepatic steatosis (31). Additionally, human population-based studies on dietary BCAA intake and body weight have shown that a higher dietary BCAA intake is associated with a lower prevalence of an overweight status and obesity among multi-ethnic populations, including the Japanese (29). This result may seem to be paradoxical; therefore, further studies should be performed regarding the relationship between peripheral BCAA alterations and dietary BCAA ingestion. Gly and Ser, two glucogenic amino acids, had negative correlations with the VFA (Fig. 4, Tables 2 and 4). One possible reason for the reduction of Gly and Ser in the high VFA group is an increased consumption of these amino acids by enhanced gluconeogenesis (32) together with a decreased production by the enhancement of glyceroneogenesis. It has been reported that glucose production from both Gly and Ser in hepatocytes is increased in diabetic individuals, whereas this type of glucose production is low in healthy conditions (32). In this study, we tested another index with Gly in addition to the BCAAs as explanatory variables, and we confirmed that the discriminatory power to distinguish between the high and low VFA groups existed (data not shown). Although the levels of several other amino acids, including Tyr and Phe, changed depending on the presence of VFA, the reasons are unclear. Thus, the mechanisms underlying the alterations in plasma amino acid levels require further investigation.

The Framingham Offspring Study showed that the plasma levels of BCAAs, Tyr and Phe were significantly associated with the future diagnosis of diabetes and that a combination of these amino acids could predict a future diagnosis of diabetes with a fivefold higher risk for individuals in the top quartile (33). Furthermore, Shah et al. reports that the amino acid profiles may help identify individuals who would be most likely to benefit from a moderate weight loss program (34). In view of these reports, further research on the sequential monitoring of peripheral amino acid changes together with other metabolic variables in these Japanese subjects are important. Whether the examination of these peripheral amino acid alterations can predict future occurrences of metabolic complications better than the measurement of WC or even a CT scan is of interest for future examination.

In conclusion, our results suggest that the measurement of different amino acid levels in the plasma is a useful approach for understanding the metabolic implications of obesity. Furthermore, our results indicate that this amino acid index can be used as a predictor of elevated visceral obesity in Asian populations. This type of method is simple and versatile for diagnosing metabolic dysfunction. Thus, its applicability to races and ethnicities other than Japanese is worth studying. Further studies should be performed to provide more evidence regarding the relationship between metabolic dysfunction and systemic amino acid changes.

Conflicts of Interest Statement

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conflicts of Interest Statement
  8. Acknowledgements
  9. References

TT, KN, HM and HY are employees of Ajinomoto Co., Inc. MY received research grants and consultancy fees from Ajinomoto, Co., Inc. MY, TT and HY have applied for patents for plasma amino acid profiling using multivariate analysis as a diagnostic tool for visceral fat accumulation and obesity evaluation (WO2009/001862 and WO2010/095682). No other potential conflicts of interest relevant to this paper are declared.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conflicts of Interest Statement
  8. Acknowledgements
  9. References

We thank the following people: Dr Y. Noguchi, Dr T. Ando, Dr T. Kobayashi, Dr A. Imaizumi, Dr M. Takahashi, Mr N. Ono and Dr T. Muramatsu for useful discussions; Dr T. Yamamoto, Dr K. Shimbo, Mr H. Yoshida, Ms M. Nakamura, Mr K. Nakamura and Ms N. Kageyama for amino acid analyses; and Ms N. Takahashi, Ms M. Takasu and Ms M. Suzuki for data acquisition.

The authors' responsibilities were as follows: MY and HY designed the research; MY, YI, TM, MT, AT, ET and MO conducted the research; HM provided essential materials and conducted the research; TT analysed the data; MY, TT and KN wrote the paper; and KN had the primary responsibility for the final content. MY and TT contributed equally to this work. All authors read and approved the final manuscript.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Conflicts of Interest Statement
  8. Acknowledgements
  9. References