Plasma Amino Acid Profile, a Biomarker for Visceral Adipose Tissue that can substitute for Waist Circumference in Japanese Americans

Abstract

Background

Greater visceral fat area (VFA) is associated with cardiometabolic outcomes. We sought to identify cross-sectional and longitudinal associations between amino acid (AA) levels and VFA in Japanese-Americans.

Methods

From the cohort of 342 Japanese-American participants (51% men) in a study of diabetes risk factors who were free from diabetes, we measured levels of 20 AA by mass spectrometry, height, weight, waist circumference (WC), VFA, subcutaneous fat area by single-slice CT at the umbilicus. Using AA significantly associated with VFA in univariate analyses, we created a VFA prediction index, termed the 4A index. We compared area under receiver-operating characteristic curve (AUROC) of the 4A index to WC and an existing AA index (Yamakado et al. Clin Obes 2012) in classifying VFA at different cutoff values. We fit age-adjusted linear regression models to evaluate associations between AA levels and change in VFA over 5 years.

Results

All 20 AA levels significantly detected VFA excess, but WC was better. The 4A index performed better than Yamakado index at classifying VFA ≥100 cm² (0.798, 0.807 vs. 0.677, 0.671 for men and women, p <0.0033) and VFA ≥ sex-specific median values (0.797, 0.786 vs. 0.676, 0.629 for men and women, p <0.0017). AA significantly associated with change in VFA over 5 years were asparagine, glutamate, glutamine, glycine, methionine, proline, threonine in men; and histidine, isoleucine, tyrosine in women (p <0.05).

Conclusions

The 4A index can serve as a biomarker for VFA in Japanese-Americans and be considered for this purpose when WC is not available.

Introduction

Visceral adiposity is a key determinant of cardiometabolic diseases ¹. Accurate measurement of visceral fat area (VFA) is important in prevention and treatment of cardiometabolic outcomes ². There are limitations associated with the existing methodology in quantifying VFA. Development of imaging techniques such as computed tomography (CT) and magnetic resonance imaging (MRI) led to the ability to selectively and accurately measure VFA ³; however, cost has restricted its routine clinical use for this purpose. Despite their common use as adiposity surrogates, anthropometric measurements, such as waist-to-hip ratio and waist circumference, have limitations in populations with different body fat distribution ⁴. Notably, Asians have higher percentage of body fat, especially visceral adipose tissue, than Caucasians despite lower BMI and waist circumference, thus creating challenges for the detection of visceral fat accumulation in Asian subjects ⁵. Therefore, the identification of a new, effective and reliable adiposity biomarker is necessary. In recent years, metabolomics has emerged as a powerful tool in disease diagnosis and in furthering our understanding of underlying metabolic derangements in conditions such as obesity ⁶.

In 1969, Felig et al. first reported the observation of higher branched-chain amino acids (BCAA), tyrosine, phenylalanine, and decreased glycine levels in obese subjects ⁷. Since then, more studies have further investigated the link between amino acid (AA) profiles and obesity ^8-13. In particular, two studies have been conducted in individuals of Japanese descent ^9,10; however, they were cross-sectional, limiting assessment of potential causal relationships. Further, age, a key risk factor for higher VFA, was not included in the prediction models ¹⁴, and cut-points used for excess VFA (100 cm²) in these studies were high, especially in women ^15,16. To address these gaps, we sought to develop a new adiposity index incorporating plasma AA levels, age, and sex that could reliably predict VFA in Japanese Americans, and to identify the temporal relationship between AA levels and VFA to improve understanding of the temporal associations between AA levels and VFA.

Material and methods

1.1. Study population and design

Participants for this cohort, termed the Japanese American Community Diabetes Study (JACDS) Biomarker Discovery Project (BDP), were taken from the parent JACDS study, a prospective community-based cohort of second- and third- generation Japanese Americans of 100% Japanese ancestry in King County, Washington State. JACDS was designed to investigate risk factors for and prevalence of type 2 diabetes and related conditions in Japanese Americans, and participants were recruited between 1983 and 1991 (n = 761). Details about selection and recruitment have been described previously ¹⁷. Diabetes was defined according to World Health Organization diagnostic criteria (fasting glucose ≥126 mg/dL, glucose ≥200 mg/dL two hours after a 75-gram glucose load) ¹⁸. Follow-up visits were conducted at 5–6 years and 10–11 years after enrollment. As no stored plasma samples were available from the original (entry) visit, the current (BDP) cohort comprises all participants with available fasting plasma samples from the 5–6 year visit who were free from diabetes at that time (n = 342). Thus, for the purposes of the current analysis, the original JACDS 5–6 year visit is termed the BDP baseline visit, and the JACDS 10–11 year visit is termed the BDP follow-up visit. The study was approved by the University of Washington Human Subject Division, and written informed consent was obtained from the participants.

1.2. Data collection

Evaluations were performed at the General Clinical Research Center at the University of Washington, Seattle. Information on age and sex was obtained by interview. Trained staff measured height, weight, and waist circumference (WC). Weight was measured using a digital scale after shoes and outer clothing were removed. Body mass index (BMI) was defined as weight in kilograms divided by the square of height in meters. WC was measured at the narrowest point between the iliac crest and the lowest rib. Visceral fat area (VFA) and subcutaneous fat area (SFA) in squared centimeters were obtained by single-slice CT at the level of umbilicus. VFA was defined as fat within the confines of the transversalis fascia while SFA was defined as fat superficial to the abdominal and back muscles ¹⁹.

At the BDP baseline visit, study participants underwent a 75-gram oral glucose tolerance test (OGTT). Normal glucose tolerance (NGT) was defined as fasting glucose <100 mg/dL and glucose <140 mg/dL at two hours. Impaired fasting glucose/impaired glucose tolerance (IFG/IGT) was defined as fasting glucose ≥100 mg/dL and <126 mg/dl or glucose ≥140 mg/dL and <200 mg/dl at two hours.

At the BDP baseline visit, trained phlebotomists collected fasting blood samples. Plasma was isolated and stored at −80°C. Amino acids were measured using targeted metabolomics protocols with GC-TOF-HRMS, hydrophilic (HILIC)-LC-HRMS, and HILIC-LC-MS/MS instruments as previously reported ²⁰. Amino acids detected in JACDS fasting plasma samples included the 20 amino acids found in human proteins: alanine (Ala), arginine (Arg), asparagine (Asn), aspartate (Asp), cystine, glutamine (Gln), glutamate (Glu), glycine (Gly), histidine (His), isoleucine (Ile), leucine (Leu), lysine (Lys), methionine (Met), phenylalanine (Phe), proline (Pro), serine (Ser), threonine (Thr), tryphtophan (Trp), tyrosine (Tyr), and valine (Val).

1.3. Statistical analysis

Study population characteristics were stratified by sex and described as median and interquartile range unless specified otherwise. Change in VFA and SFA between the BDP baseline and follow-up visits was calculated as follow-up minus baseline measurements. Significant differences in means of these continuous variables between men and women were determined using Student’s t-test. Plasma levels of 20 AA were compared for low vs. high VFA dichotomized at sex-specific median levels in men (99.3 cm²) and women (65.6 cm²) using two-sample Wilcoxon – Mann-Whitney tests.

For each AA, we fit sex-stratified, age-adjusted linear regression models to predict VFA and used the predicted value to classify low vs. high VFA defined by sex-specific median values using the area under receiver-operating characteristic (AUROC) curves analyses. In sensitivity analyses, we also fit sex-stratified linear regression models for each AA to evaluate the presence of multiplicative first-order interactions between glycemic status (IFG/IGT vs. NGT) and AA levels on VFA.

Correlations between AA and adiposity-related variables (body weight, BMI, WC, VFA, and SFA) were assessed by univariate Spearman’s correlation coefficients. Using AA that were significantly correlated with VFA in men and women, we fit multivariate logistic regression models to create a new VFA prediction index, which we termed the Anh Amino Acid Adiposity (4A) index.

For comparison to our novel index, we selected the following existing measures that have been used in prediction of VFA: WC and a previously reported AA index by Yamakado et al. ⁹. Developed as a single-dimension index to discriminate subjects with high VFA (defined as ≥100 cm²), the Yamakado index was defined as:

[−3.5250] + [0.0379] Glu + [−0.0070] Gly + [0.0034] Ala + [0.0196] Tyr + [−0.0216] Trp + [0.0054] [Ile + Leu + Val] (μmol/L).

We compared performance of the 4A index to existing adiposity indices using AUROC. First, we compared the 4A index to the Yamakado index at two different VFA cut-points: 100 cm² and sex-specific median values. Since the Yamakado index does not include a term for age, we did not incorporate age in this initial logistic regression model.

At classifying VFA ≥100 cm², the 4A index was defined as:

• For men: [1.8278] + [−0.0556] Asn + [0.0215] Asp + [−0.0213] Glu + [−0.0551] Gln + [−0.0524] Gly + [0.0026] His + [0.0173] Ile + [−0.0133] Leu + [−0.0251] Met + [−0.0606] Ser + [0.0992] Tyr + [0.0933] Val (μmol/L).

• For women: [−0.8499] + [−0.0930] Asn + [−0.0574] Asp + [−0.0040] Glu + [−0.0440] Gln + [−0.0203] Gly + [0.0398] His + [0.0442] Ile + [0.0420] Leu + [−0.0306] Met + [−0.0221] Ser + [0.0829] Tyr + [0.0271] Val (μmol/L).

At classifying VFA ≥ sex-specific median values, the 4A index was defined as:

• For men: [1.6998] + [−0. 0516] Asn + [0. 0162] Asp + [−0. 0199] Glu + [−0. 0507] Gln + [−0. 0531] Gly + [0. 0023] His + [0. 0339] Ile + [−0. 0269] Leu + [−0. 0279] Met + [−0. 0624] Ser + [0. 1035] Tyr + [0. 0865] Val (μmol/L).

• For women: [−1.0098] + [−0. 1355] Asn + [−0. 0220] Asp + [−0.0011] Glu + [0. 0299] Gln + [−0. 0093] Gly + [0. 0504] His + [0. 1980] Ile + [−0. 0259] Leu + [−0. 0042] Met + [−0. 0399] Ser + [0. 0142] Tyr + [−0. 0614] Val (μmol/L).

Next, we compared performance of the 4A index and WC using sex-specific median cut-points in age-adjusted models. Then, we compared the performance of a measure that included either WC and 4A index versus either WC or 4A index alone.

Lastly, in a longitudinal analysis, we fit age- and sex-adjusted linear regression models for each AA to evaluate associations between BDP baseline AA levels and change in VFA in 5 years. Statistical significance was set at p <0.05. All statistical analyses were performed using Stata (version 16; StataCorp, College Station, TX). The figure was created using Stata’s coefplot command ²¹.

Results

Participants were 51% men (n = 176) and 49% women (n = 166) who did not differ by age, but women weighed less and had a lower BMI (Table 1). Median VFA was significantly higher in men than women, while the opposite was the case for SFA. Fasting glucose did not differ by sex, but the 120-minute glucose was significantly higher in women.

Table 1:

Clinical characteristics of participants in the Japanese American Community Diabetes Study Biomarker Discovery Project who were free from diabetes at baseline, stratified by sex

	MEN (n = 176)	WOMEN (n= 166)	p value
Median age, years (IQR)	55.5 (44.8 – 66.4)	52.6 (45.5 – 67.4)	0.6356
Median body weight, kg (IQR)	71.8 (64.5 – 77.5)	55.4 (50.0 – 60.0)	<0.0001
Median body mass index, kg/m2 (IQR)	25.2 (23.2 – 27.3)	23.2 (21.2 – 25.4)	<0.0001
Median waist circumference, cm (IQR)	89.5 (84.2 – 94.0)	84.6 (77.7 – 91.4)	<0.0001
Median subcutaneous fat area (SFA), cm2 (IQR)	140.6 (107.5 – 189.1)	194.9 (139.5 – 249)	<0.0001
Median change in SFA, cm2 (IQR)	8.5 (−9.4 – 27.5)	10.7 (−17.5 – 36.9)	0.8820
Median visceral fat area (VFA), cm2 (IQR)	99.3 (68.6 – 122.9)	65.6 (42.9 – 101.8)	<0.0001
Median change in VFA, cm2 (IQR)	11.8 (−10.1 – 29.3)	2.8 (−8.8 – 17.1)	0.0318
Median fasting glucose, mg/dL (IQR)	97 (91 – 103)	95 (90 – 102)	0.1896
Median glucose at 120 min_OGTT, mg/dL (IQR)	132 (116 – 151)	141 (122 – 161)	0.0150
Impaired Fasting Glucose/Impaired Glucose Tolerance, % (n)	59% (103)	63% (104)	0.5029

In univariate comparisons, high VFA was associated with higher levels of glutamate and lower levels of asparagine, glycine, and serine (p<0.05) in both sexes. High VFA was associated with higher plasma levels of individual BCAA—isoleucine, leucine, and valine—in women only (all p <0.01) (Table 2). In sex-stratified and age-adjusted linear regression models, every AA significantly discriminated between high and low VFA; however, the AUROC of WC (0.824 for men and 0.861 for women) was significantly greater than the AUROC of any given AA on its own (0.650–0.742 for men, 0.677–0.735 for women, all p <0.0222) (Table 3). In addition, the AUROC of WC was also greater to a statistically significant degree when compared to the sum of all AAs (Table 3).

Table 2:

Comparison of baseline plasma amino acid levels dichotomized at median values of VFA among men and women free from diabetes in Japanese American Community Diabetes Study Biomarker Discovery Project, using two-sample Wilcoxon – Mann-Whitney test (n = 342)

	Men (n = 176)				Women (n = 166)
	VFA <99.3 cm2		VFA ≥99.3 cm2		VFA <65.6 cm2		VFA ≥65.6 cm2
Alanine	46.8 ± 11.1		47.9 ± 11.9		45.3 ± 12.0	*	50.5 ± 11.8
Arginine	30.0 ± 7.4	**	27.3 ± 5.7		28.8 ± 6.4		29.1 ±6.6
Asparagine	34.1 ± 9.2	***	28.0 ± 7.7		33.6 ± 8.8	***	29.4 ± 7.1
Aspartate	21.4 ± 8.8	*	24.0 ± 9.5		21.0 ± 10.2		22.0 ± 10.5
Cystine	26.6 ± 7.5		25.3 ± 7.9		21.7 ± 5.7	***	25.6 ± 7.1
Glutamate	58.4 ± 39.5	***	72.2 ± 35.5		47.8 ± 38.5	*	57.1 ± 44.3
Glutamine	40.6 ± 14.0	***	32.6 ± 12.2		40.6 ± 12.5		38.1 ± 12.5
Glycine	35.0 ± 7.8	***	29.8 ± 5.8		36.5 ± 10.3	**	33.1 ± 10.4
Histidine	39.9 ± 5.9	**	37.7 ± 5.1		38.3 ± 5.5		38.8 ±5.0
Isoleucine	36.7 ± 7.5		37.9 ± 6.5		31.0 ± 5.9	***	35.0 ± 6.3
Leucine	33.4 ± 6.9		33.8 ± 5.8		30.2 ± 5.6	**	33.5 ± 6.5
Lysine	34.7 ± 6.8		34.2 ± 5.9		33.4 ± 6.7		34.7 ± 5.7
Methionine	39.7 ± 12.2	**	34.5 ± 11.8		35.7 ± 11.4		33.8 ± 10.9
Phenylalanine	31.7 ± 5.0		31.0 ± 4.9		30.1 ± 5.1		30.9 ± 4.8
Proline	35.6 ± 10.3		33.5 ± 9.1		32.8 ± 10.4	**	37.4 ± 11.9
Serine	29.7 ± 5.6	***	27.0 ± 4.6		29.7 ± 6.9	*	27.3 ± 5.9
Threonine	34.6 ± 8.7	**	31.5 ± 5.4		33.1 ± 8.4		31.9 ± 7.2
Tryptophan	29.1 ± 6.1		28.4 ± 5.1		26.5 ± 5.0		25.8 ± 4.4
Tyrosine	35.6 ± 7.2		37.5 ± 6.8		33.4 ± 6.8		35.0 ± 7.5
Valine	33.3 ± 6.4		34.8 ± 5.9		29.3 ± 5.7	**	31.8 ± 5.3

^*p <0.05 for dichotomized VFA within each sex category

^**p <0.01 for dichotomized VFA within each sex category

^***p <0.001 for dichotomized VFA within each sex category

Table 3:

AUROC and p-values comparing age-adjusted levels of plasma amino acids to waist circumference in prediction of low vs. high VFA dichotomized at median values among men (99.3 cm²) and women (65.6 cm²) who were free from diabetes in the Japanese American Community Diabetes Study Biomarker Discovery Project (n = 342)

	Men (n = 176)			Women (n = 166)
	AUROC	95% CI	p value	AUROC	95% CI	p value *
Alanine	0.667	0.585 – 0.749	0.0001	0.718	0.640 – 0.795	0.0006
Arginine	0.669	0.588 – 0.750	0.0002	0.679	0.598 – 0.761	<0.0001
Asparagine	0.695	0.615 – 0.774	0.0007	0.705	0.625 – 0.786	0.0001
Aspartate	0.654	0.570 – 0.737	<0.0001	0.688	0.608 – 0.769	<0.0001
Cystine	0.658	0.575 – 0.741	<0.0001	0.723	0.645 – 0.801	0.0008
Glutamate	0.654	0.571 – 0.738	<0.0001	0.685	0.604 – 0.765	<0.0001
Glutamine	0.668	0.586 – 0.749	0.0001	0.684	0.603 – 0.765	<0.0001
Glycine	0.742	0.669 – 0.815	0.0222	0.701	0.621 – 0.782	0.0001
Histidine	0.663	0.581 – 0.746	0.0001	0.691	0.610 – 0.772	0.0001
Isoleucine	0.698	0.620 – 0.777	0.0010	0.735	0.659 – 0.811	0.0016
Leucine	0.692	0.613 – 0.771	0.0008	0.707	0.627 – 0.786	0.0002
Lysine	0.658	0.576 – 0.741	<0.0001	0.680	0.599 – 0.761	<0.0001
Methionine	0.650	0.567 – 0.733	<0.0001	0.687	0.606 – 0.767	<0.0001
Phenylalanine	0.655	0.573 – 0.738	<0.0001	0.680	0.598 – 0.762	<0.0001
Proline	0.651	0.568 – 0.734	<0.0001	0.695	0.615 – 0.775	<0.0001
Serine	0.697	0.619 – 0.775	0.0016	0.691	0.609 – 0.772	0.0001
Threonine	0.681	0.601 – 0.761	0.0005	0.683	0.602 – 0.765	<0.0001
Tryptophan	0.661	0.579 – 0.743	<0.0001	0.677	0.596 – 0.758	<0.0001
Tyrosine	0.689	0.609 – 0.769	0.0005	0.686	0.604 – 0.767	<0.0001
Valine	0.720	0.644 – 0.795	0.0080	0.726	0.649 – 0.803	0.0008
Total AA	0.651	0.567 – 0.734	<0.0001	0.687	0.606 – 0.768	<0.0001

^*p value compared with predicted VFA using WC (AUROC 0.824 for men and 0.861 for women)

Asparagine and glycine showed the strongest negative correlations with VFA, r = −0.364 (p <0.001) and −0.336 (p <0.001) respectively. Isoleucine, leucine, valine, and glutamate had the strongest positive correlations with VFA (r = 0.292, 0.207, 0.279 and 0.330 respectively, all p <0.001). Similar associations were demonstrated between other adiposity measures (body weight, BMI, WC, SFA) and both BCAA (positive correlation) and glycine (negative correlation). Interestingly, glutamate had positive correlations with all indices of adiposity, except for SFA (Table 4). In a sensitivity analysis (data not shown), associations between AA levels and VFA did not differ in subjects with NGT compared to those with IFG/IGT in women. In men, associations between AA level and VFA differed significantly by glycemic status for arginine, asparagine, cystine and glycine, r = −0.321 (NGT) vs. 0.0001 (IFG/IGT), p = 0.009; −0.579 (NGT) vs. −0.251 (IFG/IGT), p = 0.012; −0.266 (NGT) vs. 0.120 (IFG/IGT), p = 0.032; and −0.559 (NGT) vs. −0.227 (IFG/IGT), p = 0.038 respectively.

Table 4:

Spearman correlation coefficients of baseline plasma amino acid levels with various indices of adiposity in participants free from diabetes in the Japanese American Community Diabetes Study Biomarker Discovery Project (n = 342)

	Body weight		BMI		WC		VFA		SFA
Alanine	0.089		0.123	*	0.101		0.100		0.197	***
Arginine	−0.080		−0.079		−0.112	*	−0.097		−0.031
Asparagine	−0.092		−0.141	*	−0.309	***	−0.364	***	−0.134	*
Aspartate	0.041		0.072		0.151	**	0.145	**	0.022
Cystine	0.191	***	0.185	***	0.161	**	0.095		0.074
Glutamate	0.181	***	0.189	***	0.316	***	0.330	***	0.055
Glutamine	−0.067		−0.080		−0.229	***	−0.301	***	−0.009
Glycine	−0.206	***	−0.253	***	−0.299	***	−0.336	***	−0.204	***
Histidine	0.069		0.057		−0.058		−0.108	*	0.079
Isoleucine	0.315	***	0.302	***	0.286	***	0.292	***	0.130	*
Leucine	0.181	***	0.236	***	0.233	***	0.207	***	0.142	**
Lysine	−0.014		0.011		0.049		0.055		0.047
Methionine	0.101		0.041		−0.093		−0.138	*	0.025
Phenylalanine	0.084		0.119	*	0.085		0.057		0.096
Proline	0.018		0.070		0.063		0.040		0.116	*
Serine	−0.042		−0.064		−0.134	*	−0.210	***	−0.058
Threonine	−0.005		−0.021		−0.079		−0.095		−0.010
Tryptophan	0.210	***	0.149	**	0.019		0.044		0.001
Tyrosine	0.221	***	0.242	***	0.224	***	0.193	***	0.148	**
Valine	0.350	***	0.349	***	0.268	***	0.279	***	0.164	**

^*p <0.05

^**p <0.01

^***p <0.001

The AA that were significantly correlated with visceral adiposity in univariate analyses (asparagine, aspartate, glutamate, glutamine, glycine, histidine, isoleucine, leucine, methionine, serine, tyrosine and valine) were included in our sex-stratified multivariate logistic regression models to predict VFA. Compared to the Yamakado index, AUROC for the fitted estimate from the novel 4A index was significantly greater at a cut point of 100 cm² (4A index 0.798 and 0.807 vs. Yamakado index 0.677 and 0.671 for men and women respectively, p <0.0033) as well as at sex-specific median VFA cut-points (4A index 0.797 and 0.786 vs. Yamakado index 0.676 and 0.629 for men and women respectively, p <0.0017). The AUROC for the novel index was not statistically different from WC alone (4A index 0.798 vs. WC 0.825, p = 0.448 for men; 4A index 0.814 vs. WC 0.865, p = 0.145 for women). The novel index in combination with WC performed significantly better at classifying high VFA based on median cut-points than WC alone in men (4A index_WC 0.871 vs. WC 0.825, p = 0.022 for men; 4A index_WC 0.889 vs. WC 0.865, p = 0.099 for women).

In sex-stratified linear regression models for each amino acid, AA that were associated with VFA over 5 year follow-up independent of age and baseline VFA were asparagine, glutamate, glutamine, glycine, methionine, proline and threonine in men; and histidine, isoleucine and tyrosine in women (all p <0.05), as illustrated in Figure 1.

Figure 1:

Regression coefficients from linear regression model testing associations of baseline plasma amino acids and change in VFA in 5 years in Japanese Americans, independent of baseline VFA (n = 342)

Discussion

In this prospective community-based study of 342 Japanese Americans, we have developed a novel index based on plasma AA levels and found that it performed better than a previously reported AA-based index in classifying low vs. high VFA. When age, which is a strong predictor of VFA ¹⁴, is included in the model, the index performs similarly to WC and may be useful in datasets which do not include anthropometric measurements. Further, levels of seven AA including asparagine, glutamate, glutamine, glycine, methionine, proline and threonine were prospectively associated with change in VFA over 5 years of follow-up in men. The three AA histidine, isoleucine and tyrosine were prospectively associated with change in VFA in women. To our knowledge, this is the first longitudinal study of plasma AA levels and their relationship to VFA in Asian Americans.

Newgard et al. reported that the BCAA, phenylalanine, tyrosine, glutamate/glutamine, aspartate/asparagine, and arginine were elevated in obese subjects ⁸. Similarly, other investigators showed positive associations between BCAA, glutamate, tyrosine and visceral adipose tissue area ¹¹. However, in these studies participants were primarily Caucasian or African Americans. While we believe ours is the first study in Asian Americans, two studies have evaluated associations of plasma AA with VFA in Japanese living in Japan ^9,10. In a crosssectional study of 83 Japanese subjects with normal glucose tolerance, Takashina et al. concluded that visceral obesity was positively associated with BCAA and glutamate; and negatively associated with asparagine, glutamine, glycine and serine ¹⁰. Our results are consistent with these findings; for example, BCAA and glutamate had the strongest positive correlation with VFA, while asparagine and glycine demonstrated the strongest negative relationship with VFA.

In a larger cross-sectional analysis, Yamakado et al. enrolled 1449 Japanese participants (68% male, mean age 58 years) to examine whether plasma AA levels could classify low vs. high VFA ⁹ As in the current study, Yamakado et al. fit a multivariate logistic regression model to develop a single dimension AA index, consisting of alanine, glycine, glutamate, tryptophan, tyrosine, and BCAA, with AUROC of 0.81 (p <0.001) to detect VFA ≥100 cm^{2 9} In direct comparison, our novel 4A index outperformed the Yamakado index at predicting visceral adiposity. One difference was that age—a strong risk factor for high VFA—was included as a covariate in our model compared to Yamakado’s model. Another difference between our study and those of Yamakado et al. and Takashina et al. was we only selected the 20 proteinogenic AA while they also included several non-proteinogenic AA (ornithine, citrulline, alpha-aminobutyric acid).

While we cannot imply cause and effect from our observation of relationships between certain AA and VFA change, there may be mechanistic underpinnings for the BCAA, glutamate and glycine associations. In line with our observations, the BCAA—isoleucine, leucine, and valine—and their catabolite glutamate have been correlated with visceral obesity defined by VFA in previous cross-sectional studies ^9-12 Other investigations have demonstrated a possible involvement of amino acids in visceral fat deposition over time. Glutamate and glutamine are thought to be involved in the regulation of food intake. Glutamate is a substrate in the synthesis of gamma aminobutyric acid, an appetite enhancer, while glutamine increases the secretion of glucagon-like peptide-1, an appetite suppressor ^22,23. Decrease in intra-abdominal fat and nonesterified fatty acids, and increase in hepatic mitochondrial respiration were observed in sucrose-fed rats treated with glycine, suggesting a possible link between glycine and its effect on fatty acid oxidation, circulating nonesterified fatty acids, and ultimately intra-abdominal fat accumulation ²⁴. In this study, no significant difference in energy intake or body weight was demonstrated among groups (sucrose-fed rats with or without glycine, control rats with or without glycine), but differences in energy expenditure was not discussed ²⁴. These results therefore suggest a role for the AA glycine in body fat distribution in rats. Whether glycine, glutamine and glutamate concentrations will affect visceral fat accumulation in humans will require further research.

Available evidence also suggests a role for fat tissue in amino acid metabolism. Obese and insulin-resistant individuals have increased de novo production of BCAA by the gut microbiome and reduced utilization of BCAA in liver and adipose tissue ¹³. Additionally, two main catabolizing enzymes, branched-chain aminotransferase (BCAT) and branched-chain keto-acid dehydrogenase complex (BCKDC), are present in brain, heart, liver, kidney, stomach, colon, skeletal muscle, adipose tissues and are involved in the first two steps in the pathway of BCAA catabolism ²⁵. Both animal and human studies suggested reduced gene expression of these two enzymes in adipocytes of obese subjects ^26,27. Lackey et al. and Boulet et al. determined that expression of BCKDC was reduced in visceral adipose tissue of obese individuals ^11,28. This might contribute to the higher circulating levels of BCAA and glutamate observed in individuals with high VFA.

Our study indicated sex differences in the plasma levels of certain AA and their associations with high VFA, such as higher levels of glutamate seen in men, or higher levels of BCAA associated with higher VFA only in women. Previous studies have reported sex differences in several AA ^29-33. Volpi et al. reported leucine oxidation is lower in women than in men independent of body composition expressed by either total body weight or fat-free body mass in the basal state ³¹. Another study suggested that components of the amino acid handling system are sex specific, for instance, estrogen induces gene expressions of sodium-coupled neutral AA transporters in breast cancer cells ³². This might indicate that sex hormones play a causal role in AA metabolism ^31-33. Prior research in this area has been mostly conducted in animal models, and the few studies in humans have been limited by small sample sizes and cross-sectional study designs ³³; therefore, the mechanisms underlying sex differences in AA metabolism are still subject to future research.

There are several strengths to this study. First, we used a relatively large cohort of Japanese American participants with well-characterized anthropometric, glycemic and plasma amino acid measurements, allowing more accuracy in interpreting the results. Second, this is to our knowledge the first longitudinal study of plasma AA and VFA. Third, recognizing the limitation of VFA cut-point of 100 cm² as a universal value for visceral obesity detection and the fact the optimal VFA cut-point for diagnosis of metabolic syndrome is lower in women than men ^15,16, we used a sex-specific median VFA cut-point to identify individuals with excess visceral fat. Lastly, the final models were adjusted for age to improve their discriminative ability as aging is well recognized to influence the magnitude of visceral adiposity ¹⁴.

There are also limitations to our study. First, the results were derived from a single cohort and we were not able to validate the models using an external cohort. Hence, external validation of our findings is warranted that ideally would include comparisons of the 4A index to both WC and the Yamakado index in the estimation and prediction of VFA. Second, VFA was determined from single-slice CT scan at the level of umbilicus (L4-5); however, single-slice CT correlates well with directly ascertained visceral fat volume ³⁴. Third, we were unable to control for dietary AA intake; however, multiple previous studies have shown no association of dietary AA intake with plasma AA levels ^11,35, suggesting that dietary AA intake may not be an important confounder of the association of plasma AA levels with VFA. Finally, we did not evaluate the change in AA levels over time and its impact on visceral adiposity, and this should be investigated in future studies.

In conclusion, we have found in Japanese Americans without diabetes that the levels of a number of amino acids are associated with the amount of visceral adipose tissue and that an index based on age-adjusted levels of these amino acid levels can serve as a marker for visceral adiposity. Furthermore, we observed multiple associations between AA levels and visceral fat accumulation over time, suggesting the possibility that AA metabolism is linked with this fat depot. Future research is necessary to validate these results in other populations and to investigate potential underlying mechanisms.

Table 5:

Comparison of AUROC to predict baseline VFA in Japanese Americans, using WC, 4A index and Yamakado AA index dichotomized at different VFA cutoff values (n = 342)

Without age adjustment	VFA 100 cm2
	MEN (n = 176)			WOMEN (n = 166)
	AUROC	95% CI	p value	AUROC	95% CI	p value
Yamakado index	0.677	0.598 – 0.756	-	0.671	0.581 – 0.761	-
4A index	0.798	0.733 – 0.863	0.0015	0.807	0.736 – 0.878	0.0033
Without age adjustment	VFA 99.3 cm2			VFA 65.6 cm2
	MEN (n = 176)			WOMEN (n = 166)
	AUROC	95% CI	p value	AUROC	95% CI	p value
Yamakado index	0.676	0.596 – 0.755	-	0.629	0.544 – 0.714	-
4A index	0.797	0.731 – 0.862	0.0017	0.786	0.716 – 0.856	0.0009
With age adjustment	VFA 99.3 cm2			VFA 65.6 cm2
	MEN (n = 176)			WOMEN (n = 166)
	AUROC	95% CI	p value	AUROC	95% CI	p value
WC	0.825	0.765 – 0.885	-	0.865	0.809 – 0.921	-
4A index	0.798	0.732 – 0.863	0.4481	0.814	0.748 – 0.879	0.1447
4A index + WC	0.871	0.820 – 0.923	0.0222	0.889	0.839 – 0.939	0.0995

Abbreviations:

4A index

Anh Amino Acid Adiposity index

AA

Amino acid

AUROC

Area under receiver-operating characteristics

BCAA

Branched-chain amino acid

BDP

Biomarker Discovery Project

IFG

Impaired fasting glucose

IGT

Impaired glucose tolerance

JACDS

Japanese American Community Diabetes Study

NGT

Normal glucose tolerance

SFA

Subcutaneous fat area

VFA

Visceral fat area

WC

Waist circumference

Data availability

Data are available on reasonable request from the corresponding author.