Plasma Amino Acids During 8 Weeks of Overfeeding: Relation to Diet Body Composition and Fat Cell Size in the PROOF Study

Abstract

Objective:

Different amounts of dietary protein during overfeeding produced similar fat gain but different amounts of gain in fat-free body mass. Protein and energy intake may have differential effects on amino acids during overfeeding.

Methods:

Twenty-three healthy adult men and women were overfed by 40% for 8 weeks with 5%, 15%, or 25% protein diets. Plasma amino acids were measured by gas chromatography and mass spectrometry at baseline and week 8. Body composition was measured by dual-energy x-ray absorptiometry, fat cell size (FCS) from subcutaneous fat biopsies, and insulin resistance by euglycemic-hyperinsulinemic clamp.

Results:

The following three amino acid patterns were seen: increasing concentration of five essential and three nonessential amino acids with increasing protein intake, higher levels of six nonessential amino acids with the low-protein diet, and a pattern that was flat or “V” shaped. Dietary fat and protein were both correlated with changes in valine, leucine/isoleucine/norleucine, phenylalanine, and tyrosine, but energy intake was not. The change in fat mass and weight was related to the change in several amino acids. Baseline FCS and the interaction between glucose disposal and FCS were associated with changes in several amino acids during overfeeding.

Conclusions:

Overfeeding dietary protein affects the levels of both essential and nonessential amino acids.

Introduction

Obesity is a major public health problem affecting more than one-third of the adult American population (1). It results from a positive energy balance influenced by a variety of environmental (2,3) and genetic factors (4,5). Understanding how obesity develops and what can be done to reduce its public health impact provides an important challenge.

Many approaches have been used to unravel the mystery of obesity. One of these is the use of conscious overfeeding (6–8). This strategy increases body weight, (9,10), body fat (10,11), fat cell size (FCS) (11,12)), fat-free body mass (10), and energy expenditure (8,11) and has a variety of effects on other metabolic phenotypes (6–14).

One factor in the response to overfeeding is the level of protein in the diet (10,12,13). A low-protein diet (LPD) increases the percentage of body fat (13), whereas a high-protein diet (HPD) reduces fat storage and increases fat-free mass in muscle in relation to protein intake (12). The increase of body fat during overfeeding is directly related to overfed energy but not overfed protein (10). The increase in body fat significantly increases FCS and fat accumulation in visceral adipose tissue and deep subcutaneous adipose tissue, and it is associated with increased insulin resistance (13). FCS at baseline is related to the increase in visceral fat and to insulin resistance (13). The nitrogen in the HPD is stored mainly in muscle (12). Changing the level of dietary protein also alters the amount of essential amino acids available to the body. This overfeeding study thus offers an opportunity to examine the effects of dietary protein on changes in amino acids.

Patterns of plasma amino acids have been associated with insulin resistance and the risk for developing diabetes as well as with the level of body fatness and dietary intake. By using metabolomics, Newgard et al. (15,16) identified branched chain amino acids as a signature for insulin resistance and for the risk of developing diabetes, a finding that has been repeatedly replicated (17–19). The metabolism of branched chain amino acids by adipose tissue may play a role in the development of insulin resistance through reduced expression of mitochondrial branched chain aminotransferase in adipose tissue (20) or other mechanisms within the fat (21). However, whether the insulin resistance or the change in branched chain amino acid levels comes first remains an open question (22). Decreased levels of glycine have also been reported in metabolomic analysis of blood from individuals with obesity or insulin resistance in comparison with lean controls (23,24). Other reports have shown changes in several other amino acids, including cysteine, glutamate, phenylalanine, threonine, tryptophan, tyrosine, methionine, and citrulline (25–27). In this secondary analysis of the Protein Overfeeding (PROOF) study, we have used metabolomics to test the hypothesis that during overfeeding, dietary protein affects the levels of essential amino acids to a greater extent than other amino acids.

Methods

Participants

Twenty-three healthy men and women between the ages of 18 and 35 years with a BMI of 19.7 to 29.6 kg/m², who did not participate in regular physical activity (i.e., less than 2 hours of moderate to vigorous exercise per week), were included. Activity was measured by accelerometry, which increased during overeating but did not differ by protein intake (27). Euglycemic hypersinsulinemic clamps were performed at baseline and again during the 8th week. Participants were informed that they would be randomly assigned to one of three different overfeeding diets, would have to live as inpatients for approximately 12 weeks, and would be required to eat all foods and only those foods provided by the metabolic kitchen. All participants signed a consent form approved by the Pennington Biomedical Research Center Institutional Review Board. This trial was registered at ClinicalTrials.gov (NCT00565149).

Protocol

Details of this study were previously described (10,12,13). Briefly, it was a randomized, parallel arm, inpatient study, which is depicted in Figure 1. Overfeeding was planned at ~40% above energy requirements for weight maintenance. Diets contained either 5% (LPD), 15% (normal-protein diet [NPD]), or 25% (HPD) protein. During the final 24-hour period, the baseline diet was restored. The timing of the dual-energy x-ray absorptiometry (DXA) and computed tomography (CT) scans and amino acid sampling is indicated in Figure 1.

Figure 1.

Diagram of the protocol to study the effect of three levels of dietary protein on the plasma amino acids of healthy individuals during 8 weeks of overfeeding: the PROOF study.

The first 13 to 25 days of the inpatient stay were used to establish energy requirements for weight maintenance and, once weight stability was achieved, to perform baseline measurements of body composition by DXA and abdominal CT scan and to collect plasma samples for analysis of amino acids. The details of how energy requirements were estimated are described elsewhere (10).

Diets

All food was prepared by the metabolic kitchen and provided to the participants in a 5-day menu rotation over the entire inpatient stay (10). All meals were prepared in duplicate, and food composites were prepared, frozen, and later analyzed for nutrient composition (10). The same foods in different quantities were used for the experimental diets. Baseline and overfed diet composition can be found in the box inserts in Figure 1. Meals were consumed while being supervised by kitchen or other inpatient personnel. There were three experimental diets providing 5% of energy from protein, 15% of energy from protein, or 25% of energy from protein. Baseline protein intake was 90 ± 16 g/d, and it fell to 47 ± 4.7 g/d in the 5% diet group and rose to 140 ± 29 g/d and 228 ± 48 g/d in the 15% (NPD) and 25% (HPD) protein groups, respectively. Absolute intake of carbohydrates remained constant throughout overfeeding. At baseline, carbohydrates provided 50% of calories but fell to 36% of calories during overfeeding. Fat was 59% in the LPD, 49% in the NPD, and 39% in the HPD.

Body composition

Body weight was measured each morning during the study. Total body fat and lean body mass were measured by DXA (Hologic QDR 4500; Hologic Inc., Waltham, Massachusetts) at baseline and every 2 weeks thereafter (12).

CT measurement of adipose tissue volume

Volunteers were supine on the CT scanner table with the arms extended over their heads. For abdominal adiposity, eight contiguous images were obtained every 5 cm, with five images obtained above and two below a slice centered on L4-L5 intervertebral disc. Adipose tissue cross-sectional areas were determined by using the Analyze PC software package (Biomedical Imaging Resource, Mayo Clinic, Rochester, Minnesota), and visceral, subcutaneous-deep, and subcutaneous-superficial fat areas were calculated by using a triangulation method (28).

Adipose tissue biopsy for FCS

Approximately 500 mg of adipose tissue was removed through a 1-cm skin incision made 5 cm from the umbilicus. Approximately 50 mg of tissue was placed in osmium tetroxide for the determination of FCS (28). The adipocyte cell number (cells per milligram wet weight of tissue) was determined from the amount of sample (milliliters), the quantity of cells (per milliliters), and the tissue weight (milligrams).

Laboratory measurements

Glucose was measured by using a glucose oxidase electrode (Beckman Coulter DxC 600 Pro; Beckman Coulter Inc., Brea, California), and insulin was measured by immunoassay (Siemens IMMULITE 2000; Siemens Medical Solutions USA Inc., Los Angeles, California). Concentrations of 20 amino acid species were measured in plasma at baseline (obtained prior to overfeeding) and during the last week of overfeeding. Plasma from two fasted blood samples collected 30 minutes apart during the basal period of the euglycemic hyperinsulinemic clamp were pooled and frozen at −80°C until used to measure amino acids by gas chromatography and tandem mass spectrometry (29). This system did not separate lysine from glutamine, ornithine from asparagine, or leucine from isoleucine or norleucine, and they have each been listed together. At baseline, there were no differences between amino acids by diet assignment.

Statistical analysis

Protein diets were randomly assigned. Baseline data are expressed as mean ± SD and change from baseline as mean ± SE. Effects of treatment on the change from baseline of body weight, body composition, and glucose metabolism were evaluated by using regression analysis and with baseline covariates. Analysis of variance was used to compare treatment effects, and where significant differences existed, post hoc comparisons were done with the Tukey honest significant difference test. Alpha was set at P ≤ 0.05 and not adjusted for multiple comparisons because this was a secondary analysis. Analyses were done with the JMP 7 statistical package (SAS Institute Inc., Cary, North Carolina).

Results

Participants

Twenty-three of the initial twenty-five participants in the PROOF study (10) had plasma amino acid concentrations measured at baseline and again during the 8th week and are the ones included in this secondary analysis. At baseline, the subjects were 24 ± 4 (mean ± SD) years old, stood 171 ± 12 cm tall, weighed 74.4 ± 14.0 kg, had 56.6 ± 12.4 kg of fat-free mass, and had 18.5 ± 6.4 kg of fat mass, and they had a BMI of 24.4 ± 3.5 in the LPD group, 25.5 ± 3.0 in the NPD group, and 25.8 ± 3.0 in the HPD group (P = 0.68). Fifteen were African American, six were White, and two were of other race/ethnicity. Race was not significantly different between groups (P = 0.15). There were three women in each group. Overfed energy intake per kilogram baseline weight was 45.8 ± 1.7 kcal/kg in the LPD group, 45.2 ± 1.6 kcal/kg in the NPD group, and 46.0 ± 2.0 kcal/kg in the HPD group (P = 0.95).

Relationship between amino acids at baseline or 8 weeks and dietary intake of energy, protein, and fat

At baseline, energy intake, protein intake, and fat intake were each positively associated with baseline proline (r = +0.44; P = 0.030), methyl histidine (r = +0.60; P = 0.0025), leucine/isoleucine/norleucine (r = +42; P = 0.049), and tyrosine (r = +0.44; P = 0.036). After 8 weeks of overfeeding, several amino acids were still related to overfed energy intake (methyl histidine r = +0.45; P = 0.030; tyrosine r = +0.042; P = 0.028; tryptophan r = +0.46; P = 0.048) and to overfed protein (methyl histidine r = +0.054; P = 0.0072; phenylalanine r = +0.043; P = 0.044; tyrosine r = +0.45; P = 0.029; and valine r = +0.45; P = 0.028), but none were related to overfed fat intake.

Relation of assigned diet and body composition to change in amino acids

Table 1 shows the concentration of 20 amino acids at baseline and the change from baseline for the LPD, NPD, and HPD diets. Three patterns are evident (Figure 2). The top panel depicts the eight amino acids (three nonessential) with a gradient rising as dietary protein intake increased. In this group, the concentration at 8 weeks compared with baseline was reduced in the LPD group and increased at 8 weeks in the HPD group, with the NPD group being intermediate. Phenylalanine, tyrosine, and valine were statistically significantly different (P < 0.05). Serine differed from the other seven amino acids in this group with lower values at 8 weeks than at baseline on all three diets. The middle panel shows six amino acids whose 8-week concentrations were higher in the LPD group than in the NPD or HPD groups, with alanine, glycine, and glutamine/lysine being statistically significant (P < 0.05). The lower panel shows six amino acids (two essential), of which four had a “V” shaped pattern and two (histidine and proline) showed very little change. Only methyl histidine was significantly different between diets (P = 0.015).

TABLE 1.

Baseline amino acid concentrations and change from baseline by diet group

		Change from baseline
Amino acid	Mean ± SD	LPD	NPD	HPD	Diet effect (P)
Glycine, pmol/μL	782 ± 500	71.7 ± 98.2	−154 ± 67.6	−145 ± 82.7	0.012 (LPD > NPD)
Alanine/sarcosine, pmol/μL	285 ± 177	51.6 ± 31.2	−71.3 ± 29.4	−11.9 ± 36.1	0.020 (LPD > NPD)
Serine, pmol/μL	237 ± 162	−35.4 ± 20.2	−23.2 ± 19.1	−20.7 ± 23.4
Proline, pmol/μL	1,532 ± 671	−87.7 ± 101.6	−83.7 ± 95.8	−100.1 ± 117.4
Valine, pmol/μL	552 ± 128	−92.2 ± 29.6	1.00 ± 27.9	18.2 ± 34.2	0.010 (LPD < NPD)
Threonine, pmol/μL	103 ± 61.7	−22.3 ± 11.2	−7.00 ± 10.5	1.41 ± 12.9
Pipecolic acid, pmol/μL	271 ± 123	−2.35 ± 18.3	−21.1 ± 17.3	−15.1 ± 21.1
Leu/isoleu/norleu, pmol/μL	261 ± 119	−35.4 ± 17.5	−15.6 ± 16.5	9.53 ± 20.2
Asparagine/ornithine, pmol/μL	203 ± 134	14.8 ± 24.4	−50.9 ± 23.0	−15.8 ± 28.2
Glutamine/lysine, pmol/μL	3,457 ± 2,028	662 ± 349	−673 ± 329	108 ± 403	0.040 (LPD > NPD)
Methionine, pmol/μL	54.2 ± 18.2	−1.12 ± 3.84	−6.58 ± 3.62	3.77 ± 4.44
Histidine, pmol/μL	533 ± 221	−52.8 ± 41.4	−51.7 ± 39.1	29.5 ± 47.9
Phenylalanine, pmol/μL	92.6 ± 43.6	−17.0 ± 8.23	−10.4 ± 7.76	13.8 ± 9.50	0.014 (HPD > NPD); 0.037 (LPD < NPD)
Methyl histidine, pmol/μL	7.04 ± 2.66	−0.68 ± 0.70	−1.30 ± 0.66	1.57 ± 0.80	0.013 (HPD > NPD)
Arginine, pmol/μL	178 ± 107	−6.13 ± 16.0	−21.2 ± 15.0	15.6 ± 18.4
Citrulline, pmol/μL	45.0 ± 25.1	4.80 ± 4.21	−4.50 ± 3.96	−1.86 ± 4.86
Tyrosine, pmol/μL	201 ± 98.1	−29.0 ± 15.8	−12.9 ± 14.9	23.0 ± 18.2	0.044 (LPD < NPD)
Aspartic acid, pmol/μL	63.7 ± 37.0	−6.9 ± 6.93	−1.58 ± 6.54	6.14 ± 8.00
Glutamic acid, pmol/μL	209 ± 127	18.7 ± 17.1	−24.5 ± 16.1	−14.3 ± 19.8
Tryptophan, pmol/μL	60.0 ± 20.4	−6.24 ± 4.24	−4.77 ± 4.00	1.17 ± 4.89

Fit Model Platform (SAS Institute Inc., Cary, North Carolina); change in amino acid adjusted for the baseline amino acid value. Leu, leucine; isoleu, isoleucine; norleu, norleucine.

Figure 2.

Patterns of change in plasma amino acids during overfeeding. (A) Pattern for eight amino acids (five essential) that had a stepwise increase with increasing dietary protein dosage. (B) Six amino acids for which the value of the low-protein diet was higher than both the normal- and high-protein diet. (C) Amino acids that did not fit into panels A or B and mostly had a “V” shape.

Relationship between changes in amino acids and dietary intake or components of body composition

The relationship between the change in amino acids during overfeeding and the intake of nutrients and the change in components of body composition are shown in Table 2, which is arranged by the pattern of response shown in Figure 2. Intake of protein and fat was related to changes in branched chain and aromatic amino acids, but total energy intake was not. Fat intake was inversely and significantly related to changes in leucine/isoleucine/norleucine (P = 0.051), phenylalanine (P = 0.0033), tyrosine (P = 0.0028), valine (P = 0.044), histidine (P = 0.017), and methyl histidine (P = 0.0029), whereas protein intake was positively and significantly related to leucine/isoleucine/norleucine (P = 0.0046), phenylalanine (P = 0.0086), tyrosine (P = 0.028), and valine (P = 0.026). Overfed energy intake was inversely related to changes in the following four amino acids: alanine (P = 0.044), glutamine/lysine (P = 0.025), methionine (P = 0.026), and methyl histidine (P = 0.016).

TABLE 2.

Change in amino acids by overfed energy, overfed fat, overfed protein, change in body weight, change in fat-free mass, and change in fat mass arranged by pattern of response to dietary protein

Change in amino acid	OF fat intake		OF protein intake		OF energy intake		Change in wt		Change in FFM		Change in FM
	r	P	r	P	r	P	R	P	r	P	r	P
Aspartic acid, pmol/μL
Leu/isoleu/norleu, pmol/μL	−0.45	0.051	+0.46	0.046							−0.53	0.021
Phenylalanine, pmol/μL	−0.61	0.0033	+0.55	0.0086							−0.55	0.018
Serine, pmol/μL											−0.62	0.0012
Threonine, pmol/μL											−0.46	0.033
Tryptophan, pmol/μL
Tyrosine, pmol/μL	−0.61	0.0028	+0.47	0.028							−0.54	0.014
Valine, pmol/μL	−0.43	0.044	+0.47	0.026
Alanine/sarcosine, pmol/μL					−0.48	0.044	−0.58	0.0058
Citrulline, pmol/μL
Glutamic acid, pmol/μL
Glycine, pmol/μL							−0.51	0.014
Lysine/glutamine, pmol/μL					−0.56	0.025	−0.51	0.021			−0.45	0.037
Ornithine/aspx, pmol/μL							−0.46	0.043			−0.50	0.017
Arginine, pmol/μL
Histidine, pmol/μL	−0.52	0.017									−0.63	0.0027
Methionine, pmol/μL					−0.49	0.026					−0.60	0.0049
Methyl histidine, pmol/μL	−0.61	0.0029			−0.51	0.016					−0.50	0.023
Pipecolic acid, pmol/μL							−0.61	0.0091			−0.66	0.0007
Proline, pmol/μL							−0.46	0.036			−0.61	0.0050

Platform was Fit Model (SAS Institute Inc., Cary North Carolina). Y was change in amino acid adjusted for the baseline value of each parameter. One heavy subject added more fat than anyone else, and the distribution could not be normalized with log transformation or square root transformation, and this subject was removed from the analysis of change in fat mass.

OF, overfed; FFM, fat-free mass; FM, fat mass; leu, leucine; isoleu, isoleucine; norleu, norleucine; wt, weight.

Some measurements of body composition were also correlated with changes in amino acids (Table 2). The change in body weight was negatively associated with changes in a number of amino acids, including alanine (P = 0.0058), glycine (P = 0.014), glutamine/lysine (P = 0.021), asparagine/ornithine (P = 0.043), pipecolic acid (P = 0.0091), and proline (P = 0.036), implying a relationship with either change in fat-free mass or fat mass. Surprisingly, fat-free mass was not related to changes in any amino acids. In contrast, the increase in fat mass was inversely and significantly related to a lower 8-week level of leucine/isoleucine/norleucine (P = 0.021), phenylalanine (P = 0.018), serine (P = 0.0012), threonine (P = 0.033), tyrosine (P = 0.014), lysine/glutamine (P = 0.037), ornithine/asparagine (P = 0.017), histidine (P = 0.0027), methionine (P = 0.0049), methyl histidine (P = 0.023), pipecolic acid (P = 0.0007), and proline (P = 0.013). Changes in amino acids were not related to changes in visceral adipose tissue, subcutaneous adipose tissue, or deep subcutaneous adipose tissue (data not shown).

Relationship of changes in amino acids to FCS and to glucose disposal during the insulin clamp

At baseline and after 8 weeks of overfeeding, glucose disposal during step 2 of the insulin infusion (glucose disposal rate [GDR]) was positively related to fat-free mass (baseline: r = 0.51; P = 0.017; 8 weeks: r = 0.53; P = 0.012) but not to fat mass or FCS. The change in GDR from baseline to 8 weeks adjusted for baseline GDR, however, was not related to the change in fat-free mass, fat mass, or FCS. Baseline FCS was not related to baseline fat mass or fat-free mass but was positively related to the change in fat mass (r = 0.54; P = 0.018). In addition, baseline FCS, adjusted for baseline fat mass, was positively associated with changes in several amino acids during overfeeding, including methionine (r = 0.60; P = 0.0051), histidine (r = 0.49; P = 0.030), phenylalanine (r = 0.45; P = 0.044), and tyrosine (r = 0.59; P = 0.0083). In a model examining the interaction between baseline GDR and baseline FCS adjusted for baseline fat mass, there were several interactions with changes in amino acids, including valine (P = 0.046), methionine (P = 0.027), histidine (P = 0.025), and tyrosine (P = 0.053). In this model, a higher interaction term was associated with higher 8-week values for the amino acid in question. In this analysis, two participants (10% of the total) who gained the most weight, most fat, and had the largest increase in FCS were prominent in the regression lines. If these two were omitted, significant differences would be eliminated. However, they have been left in the analysis because this was an overfeeding study in which larger differences may be the most meaningful.

Discussion

This secondary analysis of the PROOF study examined amino acid concentrations measured by gas chromatography and tandem mass spectrometry at baseline and during the 8th week of overfeeding in 23 healthy men and women who were overfed for 8 weeks at 40% above their baseline energy requirements with diets containing 5%, 15%, or 25% protein. Although the higher protein diets would provide increasing amounts of both essential and nonessential amino acids, three different patterns of response were observed. One was a positive gradient from lower to higher protein for five essential and three nonessential amino acids, including the branched chain amino acid valine, the aromatic amino acids phenylalanine and tyrosine, and threonine and tryptophan. The second pattern included six amino acids for which the value on the LPD was higher than the other diets, including significant effects on alanine, glycine, and glutamine/lysine. The third pattern was either “V” shaped or flat and included two essential amino acids (histidine and methionine) and four others. Surprisingly, two essential amino acids, histidine and methionine, did not show the expected change with increasing dietary protein intake, and three amino acids, alanine, glycine, and glutamine/lysine, showed much higher levels in the group eating the LPD.

Newgard and his colleagues (15,16), by using principal components analysis, described a pattern of amino acids involving branched chain amino acids that predicted insulin resistance and the risk of diabetes. They went on to argue that the adipose tissue phenotype was essential to this pattern (15,17,20,21), which is consistent with our data. The relationship of branched chain amino acids to insulin resistance and the risk of developing type 2 diabetes has been confirmed in the pediatric age group (30,31) and in adults (23,26,32–36), and it has been reviewed in several publications (17,19,37,38). In the review and meta-analysis by Guasch-Ferre et al. (39), 19 of 33 analyses examined the association between amino acids and pre-diabetes or diabetes and found a positive association with the branched chain amino acids and aromatic amino acids, but they found an inverse association with glycine, glutamate, and glutamine. Eight prospective studies found a positive relationship between valine, tyrosine, leucine/isoleucine/norleucine, and phenylalanine and the risk of type 2 diabetes (37). Our data add to this model by showing that changes in five amino acids, phenylalanine, tyrosine, ornithine/asparagine, tryptophan, and methionine, are inversely related to FCS, and that there is an interaction between baseline FCS and insulin-stimulated GDR with the high-dose insulin clamp. Thus, during overfeeding, the concentration of these five amino acids was reduced in individuals with larger FCS, and this effect was reversed if the fat cells were more sensitive to insulin as measured by increased glucose disposal. Several studies (17,20,40) have shown altered processing of amino acids by fat cells in obesity. The present findings may be interpreted as showing that larger fat cells remove selected amino acids more than smaller ones after overfeeding, and that this is partially related to the insulin sensitivity of these cells.

Essential amino acids, by definition, come from the diet, and during this study, they, along with aspartic acid, serine, and tyrosine, were increased with the higher protein diets and decreased in those eating the LPD. Newgard et al. (15) showed that adding branched chain amino acids to the diet of mice increased insulin resistance. Earlier, Krebs et al. (41) showed that increasing plasma amino acids could acutely increase insulin resistance, and a review of nutrient overload, similar to the overfeeding study performed here, outlined the mechanisms by which protein could activate pathways that affect insulin resistance (42). The changes in plasma amino acids observed during overfeeding in the PROOF study represent, in part, the consequence of higher dietary amino acids. However, two essential amino acids, methionine and histidine, did not follow this expected change. Higher dietary branched chain amino acids may play a role in the risk of developing diabetes as shown in a follow-up of the Nurses Health Study and Health Professionals Follow-up Study (43). Our data show that increasing concentrations of essential amino acids in the diet produce larger gradients and positive values with the higher intake for the branched chain amino acids leucine/isoleucine/norleucine and valine and the aromatic amino acids tyrosine and phenylalanine. However, the issue of whether the change in amino acids or insulin resistance comes first is still unsettled (22).

Several amino acids, including alanine, citrulline, glutamate, glutamine/lysine, and asparagine/ornithine, were higher in those eating the LPD. These amino acids are involved in the tricarboxylic acid cycle and urea cycle. The individuals eating the LPD rapidly decreased their protein catabolism, as would be expected, because they were in negative nitrogen balance (12). The higher concentration of glutamic acid, citrulline, glutamine/lysine, and asparagine/ornithine in individuals eating the LPD may reflect their attempt to spare protein (12).

Another explanation is needed for glycine because it is not involved in the tricarboxylic acid or urea cycle. Glycine is the precursor for a variety of important metabolites, such as glutathione, porphyrins, purines, heme, and creatine. It also acts as a neurotransmitter in the central nervous system (23). Increased glycine synthesis seems an unlikely explanation because serine, from which glycine is formed, was not higher after overfeeding with any diet and behaved quite differently from glycine. However, our measurements were only of serum concentrations, and turnover, if measured, might give a different story. Glycine is the principal amino acid in collagen (23), and enhanced mobilization of bone in the LPD group could be another explanation for this rise. In the presence of excess energy and adequate amounts of dietary carbohydrates and minerals, there is no reason to expect mobilization of bone collagen. Glycine is also a biosynthetic intermediate in the formation of porphyrins and purines (23,44). The observed increase in glycine at the end of overfeeding could be the result of anaplerosis from the increased fat intake in the group eating the LPD. Reduced availability of succinyl-CoA because of the increased activity of the tricarboxylic acid cycle might result from the increased fat metabolism in the LPD, which is pari passu with a high-fat diet and might, thus, account for increased glycine concentration in the LPD.

Changes in body composition were also associated with changes in amino acids during overfeeding. One hypothesis for the increased level of branched chain amino acids in insulin resistant states is an alteration in adipose tissue, with a reduction in branched chain amino acid transport protein in adipose tissue being one potential explanation (22). In an earlier paper, we showed that baseline FCS was associated with insulin resistance and change in visceral fat (13). It was surprising that the changes in amino acids had no relationship to changes in fat-free mass. One implication of this data and the findings in the present paper is that an HPD may be undesirable because it increases the conditions that may enhance insulin resistance and, thus, type 2 diabetes.

There are two essential amino acids, methionine and histidine, that differ in their response to overfeeding. Methionine had a “V” shaped relationship, implying that its circulating level was modulated by factors other than diet, one of which could be FCS. Methionine is a sulfur-containing amino acid that is involved in the transfer of methyl groups (S adenosyl-methionine). Why its levels did not respond like other essential amino acids is an unanswered question. Histidine increased in the HPD but fell during overfeeding in both the LPD and NPD. Like methionine, the response of histidine is unresolved.

To our knowledge, this is the first study in which the effects of dietary protein on amino acid profiles have been examined during overfeeding. Plasma amino acids were examined in one other 5-day overfeeding study that used a high-fat diet fed to men who were either of normal or low birth weight (45). In this study, protein was 15% in the control condition and fell to 7.5% during overfeeding because all of the extra energy was fat. They found that both groups had increased plasma alanine and decreased leucine/isoleucine/norleucine and valine levels in response to overfeeding. Alanine similarly increased in our subjects overfed with the 5% higher fat (LPD) diet. They argue that the higher alanine level in men with low birth weight could be accompanied by an increased anaplerotic formation of oxaloacetate and, thereby, enhanced tricarboxylic acid cycle activity as well increased gluconeogenesis. Our data are consistent with this interpretation.

There were clear limitations to our study. First, it only included 23 men and women. However, this was compensated for by the tight control of food intake and activity during the 8 weeks they spent in the inpatient setting. Additional strengths were the methods used to measure the components of body composition and metabolic assessments. There were two participants (10% of the total) who gained the most weight and most fat and had the largest change in FCS. If they were omitted, many significant differences were eliminated. However, this was an overfeeding study, and they have been left in because these differences may be the most meaningful.

Conclusion

We have made three important observations. The first is that overfeeding at different levels of protein produced three quite different patterns of response. Second, the change in several amino acids during overfeeding was related to baseline FCS. Third, high levels of three amino acids with the LPD may reflect anaplerosis and alterations in the activity of the tricarboxylic acid and urea cycles.