Physiol. Genomics Ad Instruments
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Physiol. Genomics 34: 315-326, 2008. First published June 17, 2008; doi:10.1152/physiolgenomics.00007.2008
1094-8341/08 $8.00
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental Figures and Tables
Right arrow All Versions of this Article:
34/3/315    most recent
00007.2008v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Noguchi, Y.
Right arrow Articles by Takahashi, M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Noguchi, Y.
Right arrow Articles by Takahashi, M.
Received 9 January 2008; accepted in final form 17 June 2008.
Physiological Genomics 34:315-326 (2008)
1094-8341/08 $8.00 © 2008 American Physiological Society

Characterization of dietary protein-dependent amino acid metabolism by linking free amino acids with transcriptional profiles through analysis of correlation

Yasushi Noguchi 1, Nahoko Shikata 1, Yasufumi Furuhata 1, Takeshi Kimura 2 and Michio Takahashi 1

1 Research Institute for Health Fundamentals, Ajinomoto Company, Incorporated, Kawasaki, Kanagawa
2 Quality Assurance & External Scientific Affairs Department, Ajinomoto Company, Incorporated, Tokyo, Japan


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
This study aims to characterize diet-dependent amino acid metabolism by linking profiles of amino acids concentrations ("aminograms") with transcript datasets through the analysis of correlation. We used a dietary model of protein restriction-to-excess, where rats were fed diets with different levels of casein (5, 10, 15, 20, 30, 50, and 70%) for 2 wk. Twenty-five different amino acids in the plasma, liver, kidney, small intestine, and muscle and 71 gene transcripts in these compartments were measured together with general physiological variables. Under low-protein diet (LPD) conditions, the plasma aminogram for EAA was similar to that of the liver and the small intestine, respectively. Under the high-protein diet (HPD), however, the plasma aminogram for EAA became like that of muscle, while that of NEAA was similar with that of both liver and muscle. To assess the impact of gene expressions in each tissue on the plasma aminograms, correlations were obtained between aminograms and transcripts in each tissue under a diet with different protein levels. Based on the correlations obtained, amino acids and transcripts were systematically connected and then a metabolite-to-gene network was constructed for either LPD or HPD condition. The networks obtained and some other metabolically meaningful relationships such as ureagenesis and serine metabolism clearly illustrated activation of either body protein breakdown with LPD or amino acid catabolism with HPD.

metabolomics; aminogram; amino acids metabolism; dietary protein


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
THE INFLUENCE OF DIETARY PROTEIN INTAKE on physiology has been an important clinical issue (17, 22, 55). Effects of protein restriction or excess have been studied in terms of either protective or exacerbating effects on various diseases such as diabetes (44), liver disorders (25), and renal complications (55) or physiological events such as aging (20) and pregnancy (27). Since dietary protein intake inevitably alters amino acid metabolism, blood free amino acid profile (aminogram) is expected to transform in accordance with some principle. Earlier studies analyzed metabolic responses to dietary protein intake in terms of blood aminogram changes (14, 43), but any uniform trends or distinct relationships have not yet been proposed. Thus, development of a methodology to achieve this is wanted, particularly for assessing how to contribute protein nutritional condition for specific pathophysiological events.

In general, data of free amino acids concentrations in clinical conditions are derived from blood (18, 28), urine (50), or amniotic fluid (4); however, amino acids can also be measured, though less frequently, in other biological fluids such as breast milk (1) and cerebrospinal fluid (10) and in cells (16) and tissues such as liver and muscle (10, 19). The balance between branched amino acids (Leu, Val, and Ile; BCAA) and aromatic amino acids (Phe and Tyr; AAA), known as the Fischer's ratio, is one of a few established diagnostic markers utilizing plasma amino acid levels and is used for monitoring progression of liver fibrosis (11). Unlike Fischer's ratio, neither metabolomic data nor aminograms have been well utilized to monitor the progression of such diseases as renal failure (33), cancer (26), atherosclerosis (15), and diabetes(51), where specific amino acid abnormalities have been reported.

Several useful analytical approaches, including principal component analysis and partial least-squares discriminant analysis, have been reported for metabolomic data analysis (5, 23, 34). In this context, we previously demonstrated the usefulness of correlation analysis between metabolites and phenotypes and proposed a correlativity-based novel regression model for systematically identifying phenotypic indexes from blood aminogram data (36, 42). The reported methods including ours unfortunately have been designed primarily as one of many biomarkers, and thus there are few trials from which to deduce more fundamental metabolic implications from the transformations of aminograms.

In this paper, we expand correlation-based analysis of the metabolomic and phenotypic dataset for describing diet-dependent alterations of whole body amino acid metabolism. First, we analyze the correlation of plasma amino acids with tissue amino acids to assess the aminogram as a potential indicator of the whole body metabolism network. Then, aminograms were correlated with gene expression profiles to obtain metabolite-to-gene connectivity. The network structure in rats fed either low (LPD) or high protein diets (HPD) was then visualized. Though the method in this study was developed under specific nutritional conditions, it has the potential to reveal alterations in the amino acids metabolic network in various pathophysiological conditions.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Animals and diets.
Male young growing Fischer-344 rats were purchased from Charles River Laboratories (Kanagawa, Japan). They were maintained in a 12-h light/12-h dark lighting conditions. Rats were fed the diets specified below for 14 days and had free access to water. All studies were reviewed and approved by the Animal Care Committee of Ajinomoto.

A total of 35 rats were used in the experiments and were randomly divided into groups fed with 5, 10, 15, 20, 30, 50, or 70% protein diets (n = 5 each group), with body weights at the starting point of 88 ± 5, 89 ± 6 90 ± 4, 89 ± 5, 89 ± 3, 90 ± 4, and 89 ± 3 g (mean ± SD), respectively. The diet contained casein, cornstarch, cellulose powder, mineral premix, and soybean oil. The standard protein diet in this study contained 20% casein. A more detailed diet composition is shown in Table 1. Multiple parameters described below were measured in the same individual animals.


View this table:
[in this window]
[in a new window]

  		Table 1. Diet composition

 
Blood components and hormones.
Blood glucose, cholesterol, triglyceride, ammonia (NH3), and albumin were determined using DRI-CHEM5500 (FUJI Films, Tokyo, Japan). Plasma nonesterified fatty acid (NEFA) was determined using commercial kits (Wako Pure Chemical Industries, Osaka, Japan). β-Hydroxybutyrate (BHB), a marker of ketosis, was measured enzymatically by Ketone Test (Sanwa Kagaku Kenkyusho, Nagoya, Japan). Serum leptin and insulin were determined using commercial rat/mouse ELISA kits (Seikagaku-kogyo, Tokyo, Japan). Plasma IGF-1 was determined using a commercial rat IGF-1 EIA kit (Diagnostic Systems Laboratories, Webster, TX). Plasma total thyroxine (T4) was determined using a commercial ELISA kit (Alpha Diagnostics Intl., San Antonio, TX).

Sample preparation.
Plasma samples were prepared with EDTA as anticoagulant, treated with two volumes of 5% (wt/wt) trichloroacetic acid (TCA), and then centrifuged to remove protein as precipitate. The samples obtained were filtered through an Ultrafree-MC centrifugal filter (Millipore, Billerica, MA). To prepare deproteinized tissue extracts, tissues were homogenized with 5% TCA and processed with the same protocol used for plasma. All samples were kept at 4°C during all steps to minimize chemical reactions of thiol-metabolites and stored at –80°C.

Amino acid profiling.
Plasma amino acid concentrations were measured using an L-8800 automatic amino acid analyzer (Hitachi, Tokyo, Japan) as described previously (37). Briefly, amino acids separated by cation-exchange chromatography were detected spectrometrically after postcolumn reaction with the ninhydrin reagent. Tissue amino acids were analyzed by reverse-phase HPLC (Agilent Technologies, Palo Alto, CA) and triple-quadrupole tandem mass spectrometry (Perkin-Elmer, Shelton, CT) as described previously (21).

Real time-PCR.
Total RNA was extracted from the homogenized liver, kidney, muscle, and small intestine using an RNAeasy kit (Qiagen, Germantown, MD) following the manufacturer's instructions. mRNA was then extracted from total RNA preparations using an Oligotex kit (Qiagen) following the manufacturer's instructions. Quality and integrity of the RNA were checked by A260/280 ratio and on a formaldehyde-agarose gel, respectively.

Equal amounts of RNA were reverse transcribed using Superscript II reverse transcriptase (Invitrogen, Carlsbad, CA) as per the manufacturer's instructions. Primers for RT-PCR were designed using the primer design software Primer3, and then the sequence homology for related proteins was checked. 18S ribosomal RNA was used as an endogenous control. RT-PCR was performed on an ABI Prism 7700 Sequence Detection System (PE Applied Biosystems, Foster CA), and the data obtained were analyzed using the provided software. The reaction mixture consisted of 4 µl cDNA template, 10 µl of Syber Green PCR master mix (Roche Biochemicals), 2 µl of 0.25–1 µM forward primer, and 2 µl of 0.25–1 µM reverse primer in a 20 µl reaction volume. The PCR protocol consisted of one 10 min denaturation cycle at 95°C followed by 40 cycles of denaturation at 95°C for 15 s and annealing/extension at 60°C for 1 min. Standard curves for each gene and endogenous control 18S rRNA were obtained. The efficiency of PCR amplification was 100% and the R2 value between 0.995 and 0.999. All RT-PCR data were first normalized by 18S ribosomal RNA and then expressed as relative mRNA levels to 20% casein group. Primer pair sequences are listed in Table 2. We also examined that normalizing PCR signals to β-actin, which is another housekeeping gene, provides similar relative results as when the PCR signals are normalized to 18S rRNA.


View this table:
[in this window]
[in a new window]

  		Table 2. Primer sequences for RT-PCR

 
Correlation based metabolite-to-gene network analysis.
Correlation-based two-mode networks were constructed in accord with the previously reported method (6). Generally in the metabolomics research, the two-mode network data is generated to assemble information that links metabolites with an array of collective entities such data as gene expression, protein expression, and enzymatic activity. A two-mode dataset is able to be generated without an assumption of direct relationships between metabolites or direct connections between the collective entities, while in most biological datasets direct relationships are assumed or derived from data such as one-mode datasets. For instance, we can choose "transcript" as a collective entity and then define a model as a metabolite-by-gene expression matrix. Two-mode datasets can be given as rectangular (n x k) matrices of n metabolites and k transcripts. To uncover possible interactions between two entities i (= metabolite) and j (= transcript), the Pearson or Spearman's rank correlation coefficient {rho}ij has been calculated for each pair within the obtained dataset. The significance of each pair of metabolite and transcript was assessed using the false discovery rate (FDR), according to reported studies (41). Then by eliminating the associations with lower {rho}ij pairs (n = 20), we selected the networks that contained only highly significant correlated entities.

To measure centrality in two-mode datasets or a projected bipartite network, several standard measure of centrality can be used such as degree, closeness, betweenness, and eigenvector centrality (6). In the present study, we used degree as a standard measure of centrality, where the degree centrality of a node is defined as the number of edges incident upon that node. In the case of correlation of metabolite and transcript, this means that the degree of a metabolite is the number of transcripts that relates with the metabolite, while the degree of a transcript is the number of metabolites that relates with the transcript. Thus, degree has a clear and simple interpretation in the two-mode case. In addition, we introduced k-core decomposition as an additional dimension (2) to visualize rather complex networks. This decomposition, based on a recursive pruning of the least connected vertices, allows disentangling the hierarchical structure of networks by progressively focusing on their central cores. The k-core decomposition consists of k-cores that identify particular subsets; each one was obtained by recursively removing all the vertices having the degree smaller than k, until the degree of all the remaining vertices became larger than or equal to k (0, 1, 2,... kmax). Larger values of "coreness" simply correspond to vertices with larger degrees and more central positions in the given network (52).

Statistics.
Statistical significance among the groups was determined either by Tukey's honestly significant difference test after an ANOVA for multiple comparisons. Power analysis was carried out using JMP version 5.0 (SAS Institute, Cary, NC) statistical software, where both power and least significant number of each parameters were determined. All correlation analyses were performed using JMP. The FDR for each metabolite-to-gene correlate pairs were estimated as a q-value using software created by Alan Dabney and John Storey (39). The q-value for a particular correlation estimates the proportion, on average, of incurred false positives for correlations with a similar level of significance (47). For generation of metabolite-to-gene network images, UCINET version 6.34 (Analytic Technologies, Harvard, MA) was used.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Physiological variables in the rat dietary model of protein excess-to-restriction.
In this study, a dietary model of protein excess-to-restriction was employed, where rats were fed with different levels of casein (5, 10 15, 20 30, 50, and 70%; n = 5 per group) for 2 wk. The diet of 20% protein was assigned as the standard protein diet, as described in MATERIALS AND METHODS. Table 3 shows cumulative intakes of carbohydrate, fat, protein, and calories as well as physiological parameters such as body weight, tissue weight, blood components, and hormones at the end of 2 wk experiment. ANOVA of most parameters show clear statistical significances (P values) among seven rat groups. The total protein intake increased depending on the percentage of casein in the diet, while carbohydrate intake decreased, resulting in only slight variation among groups in total calorie and FAT intakes. Liver and kidney weights significantly increased parallel with an increase in dietary casein intake, while FAT weight decreased. Body weight and muscle weight altered in a biphasic manner; lower weights were observes either with both LPD (casein <20%) or HPD (casein >20%). Plasma triglyceride decreased parallel with an increase in dietary casein intake. Both plasma BHB and NH3 changed concavely with the lowest values at 20–30% casein diet.


View this table:
[in this window]
[in a new window]

  		Table 3. Physiological variables

 
We next prepared LPD (casein <20%) and HFD (casein >20%) groups by integrating each three groups (n = 15 each) from original dataset. Significance test using t-test indicates physiological changes between two groups particularly in dietary intakes, tissue weights, plasma triglyceride, and glucagon (Table 1). To justify obtained experimental groups, post hoc power analysis was carried out. As a result, higher powers between two groups were confirmed in significant parameters (Table 1). Several nonsignificant parameters such as plasma insulin, T4, IGF-1, and cholesterol showed biphasic changes with the peak at ~15–30% dietary casein, indicating clear metabolic differences occurred between protein excess and restriction in present model.

Dietary protein level altered correlativity between plasma aminogram and physiological variables as well as tissue aminograms.
We determined plasma aminograms from rats fed with different levels of casein (Fig. 1A). In plasma, aminograms show clear changes depending on dietary protein levels. Most plasma essential amino acids (EAA) show an upward trend, while that of nonessential amino acids (NEAA) show complex changes, such as concave and convex trends. The aminogram from an individual rat was analyzed to find out possible relationships between physiological and dietary variables and plasma amino acids profile. The intensity of the correlation between each of 25 amino acids and each physiological and dietary valuable was assessed separately in LPD and HPD (Fig. 1B). In all correlation types of analysis, a control group (20% casein, n = 5) was added to both LPD and HPD groups, to assess changes in correlations from control to excess or deficiency of dietary protein. A nonuniform trend of correlation between these two attributes emerged. For instance, a cluster including urea and BCAA highly correlated with protein intake, kidney weight, liver weight, and glucagon particularly in HPD, while the other cluster, including NEAA, glutamine, and serine, highly correlated with carbohydrate intake and fat weight in LPD (Fig. 1B). Furthermore, EAA, threonine, methionine, aromatic amino acid, tryptophan, and tyrosine had positive correlations with body and muscle weights and IGF-1 (Fig. 1B).


Figure 1
View larger version (57K):
[in this window]
[in a new window]

  		Fig. 1. Correlation analyses between plasma aminogram and physiologic valuables or tissue aminograms. A: changes in relative levels of amino acids in the plasma in rats fed with different percentages of dietary protein (1.0 = a level obtained by feeding standard protein diet containing 20% casein). LP, low protein; HP, high protein. B: cluster analysis between plasma amino acids and dietary or physiological parameters obtained in Fig. 1. Low protein and high protein diet dataset were formed by combining 5–20% casein groups (n = 20) and 20–70% casein groups (n = 20), respectively. The value of the correlation coefficient is indicated by the color range from 1 (highly positive: deep red) to –1 (highly negative: deep blue). C: genealogical illustrations of the similarity of dietary protein dependent changes of free essential (EAAs) and nonessential amino acids (NEAAs) obtained from the plasma and tissues.

 
We also obtained aminograms from the liver, kidney, muscle, and small intestine and then compared the similarity of the aminograms between plasma and each tissue, where EAA and NEAA were treated separately (Fig. 1C). Changes of tissue aminograms are shown in Supplemental Fig. S1.1 With a load of either LPD or HPD, plasma EAA formed a cluster with liver or muscle EAA, respectively (Fig. 1C). The least similarity with plasma EAA was found in small intestine EAA, tallying with the small impact of the small intestine on whole body EAA metabolism. With a load of LPD, in contrast, plasma NEAA formed a cluster with small intestine NEAA, and with a load of HPD, with liver and muscle NEAA (Fig. 1C). The present results suggest that blood aminograms at a steady state are strongly affected by tissue metabolic activities but that the extent of the contribution varies depending on nutritional states.

Dietary protein level altered expression of gene relevant to amino acid metabolism.
The impact of dietary protein ratios on gene expression levels was assessed by employing 71 genes that were involved primarily in the liver amino acid metabolic pathway of catabolism and de novo synthesis. We also analyzed genes known to be involved in the major pathways in the kidney, muscle, and small intestine (Fig. 2, A–D), as well as neutral amino acid transporter genes in the liver, kidney (Fig. 2E), IGF-1, and IGFBP1, -2, -3, and -5 genes in the liver (Fig. 2F), and proteolytic genes in the muscle (Fig. 2G). They are listed in Table 2.


Figure 2
View larger version (37K):
[in this window]
[in a new window]

  		Fig. 2. Dietary protein-dependent changes in mRNA expressions of the genes relevant to the amino acid metabolism. Changes in the relative mRNA expressions of aminotransferases (A), deaminases (B), methyltransferase (C), enzymes responsible for the urea cycle (D), neutral amino acids transporters (E), liver IGF-1 and IGFBPs (F), and muscle genes (G) relevant to proteolysis. Full names of the genes are listed in Table 2. Letters in parentheses attached to each gene correspond to the tissue names: liver (l), kidney (k), muscle (m) and small intestine (si). Values are expressed as means ± SE (n = 5). RT-PCR data was normalized by 18S ribosomal RNA. Relative mRNA expressions are expressed relative to feeding control protein group (20% casein).

 
In the liver, gene expressions of most aminotransferase and deaminase increased in parallel with an increase in protein intake levels (Fig. 2, A and B). Significant increases were observed in those of alanine aminotransferase (gpt1), ornithine aminotransferase (oat), serine dehydratase (sds), and glutaminase (gls2), but not in glutamate dehydrogenase (glud1) (Fig. 2, A and B). Simultaneously, gene expression levels of kidney glutaminase (gls1) and glud1 increased significantly with positive correlations with an increase in dietary protein intake, but muscle and small intestine genes in general did not show clear trends (Fig. 2, A and B). Gene expression levels of most liver methyltransferases and urea cycle increased. On the other hand, those of methionine adenosyltransferase-1A (mat1a) and ornithine decarboxylase (odc) changed concavely (Fig. 2, C and D). Those of liver neutral amino acid transporters, ATA1 (slc38a1), ATA2 (slc38a2), and ATA3 (slc38a4), increased with positive correlations with dietary protein level, while kidney slc38a1 and slc38a2 decreased (Fig. 2E). Those of liver IGF-1 (igfbp1), IGFBP-3(igfbp3), and IGFBP-5 (igfbp5) changed convexly with a peak at ~15% casein, while those of IGFBP-1 (igfbp1) and -2 (igfbp2) changed concavely (Fig. 2F). Those of most muscle proteolytic enzymes involved in the ubiquitin or calpain system increased with a load of LPD (Fig. 2G). Among cathepsin genes, cathepsin-L (castl) gene expression increased with LPD, while expressions of cathepsin-B (castb) and cathepsin-H (casth) increased with HPD (Fig. 2G).

Dietary protein level altered correlativity between amino acid and tissue transcript.
We estimated the degree of correlation between amino acid and transcript separately in LPD and HFD. Both LPD and HPD data contain control groups (20% casein, n = 5) to assess changes of metabolite-to-gene correlate pairs from control to excess or deficiency of dietary protein. Thus, both low protein and high protein data contain total n = 20 animal data. We determined both Pearson and Spearman's rank correlation, but because of nonlinear relationships in most of the scatter plots (data not shown), we choose Spearman {rho} for a comparison of metabolite-to-gene correlate pairs in the present study. Figure 3A shows histograms of Spearman {rho} between amino acid and gene, where a total 8,640 metabolite-to-gene pairs were generated from 120 plasma and tissue amino acids and 71 tissue transcripts. To examine whether dietary protein could alter correlativity between amino acid and tissue mRNA expression, metabolite-to-gene pairs were arranged in order of the delta {rho} between two groups from negative to positive maximum, a finding that demonstrates that that part of the correlations was clearly sensitive to dietary protein (Fig. 3B).


Figure 3
View larger version (22K):
[in this window]
[in a new window]

  		Fig. 3. Correlativity between plasma amino acids and gene expressions. A: histograms of Spearman's rank correlation ({rho}) of metabolite-to-gene pair calculated in low protein (sum of 5–20% casein groups, n = 20) and high protein (sum of 20–70% groups, n = 20). A total 8,640 metabolite-to-gene pairs were generated from 120 plasma and tissue amino acids and 71 tissue transcripts. B: dietary protein dependent changes in the correlations ({rho}) of metabolite and gene. Metabolite-to-gene pairs are placed in order of the {Delta}{rho} (high protein-low protein) from negative maximum to positive maximum (bottom panel). C: estimation of the false discovery rate (FDR). Comparisons of q-values and P values for the metabolite-to-gene correlates are shown in both low protein and high protein dataset. q-Values of 0.05 (i.e., for every 100 significant correlates, 5 false correlates are expected) in low protein and high protein are associated with a P value of ~0.0027 and 0.0064, respectively. D: distribution of metabolite-to-gene correlation ({rho}) to P value.

 
To assess the significance of correlate pairs, we determined an FDR, as estimated by a q-value, for each correlate. The q-value measures the predicted FDR associated with a significant test when multiple hypotheses are tested, i.e., a q-value of 0.05 implies that for every 100 significant correlates, five false correlates are expected. A comparison of multiple false discovery estimations on several gene expression data sets suggests that the q-value method has a high apparent power and strong control of the FDR. Predicted q-value in LPD and HPD data were compared with P values for ~8,640 metabolite-to-gene correlate pairs (Fig. 3C). In LPD data, a P value of 0.0027 has a corresponding q-value of 0.05, which on average was associated with a correlation ({rho}) of 0.62 (Fig. 3D). This estimation of the FDR was different from the distribution of that of HPD data. In HPD, a P value of 0.0064 has a corresponding q-value of 0.05, which on average was associated with a correlation ({rho}) of 0.59 (Fig. 3D).

Metabolite-to-gene network illustrated a dietary protein-dependent whole body metabolism.
Metabolite-to-gene correlate pairs were visualized as bipartite network graphs using the Spearman {rho} matrices of two different entities of the aminograms from blood and tissues and of the gene expression data (Fig. 4). For this analysis, a correlation cutoff was placed at 0.65, greater than the minimum {rho} of 0.63 indicated by a q-value of 0.05 (Fig. 3C), where the q-value estimation of the FDR suggests that correlations with an absolute value >0.63 are very likely to be reproducible. To find importance in the bipartite networks, we used degree as a standard measure of centrality, where the degree centrality of a node is defined as the number of edges incident upon that node. Furthermore, we applied k-core analysis, which is an iterative process, where for each iteration of k, given the network from the previous iteration, nodes with fewer than k connections are removed from the graph. This will generate in a sequence of subnetworks that gradually reveal the globally central region of the original network. In Fig. 4, each node is characterized by shape (metabolite or transcript), size (degree of correlation), and color ("coreness"). In addition, a red or purple line corresponds to the relationship inside or outside coreness-kmax, respectively. Correlation networks in LPD and HPD groups were developed using 5–20% casein group data as LPD (n = 20) and 20–70% casein group data as HPD (n = 20), respectively (Fig. 4, A and B). In LPD, network analysis pointed out the highest degree of correlation in liver serine with igfbp2 and igfbp3, where serine was also linked with its metabolic enzymes, liver sds, and kidney pgdh. Both liver sds and oat seem to be dominant factors in amino acid enzymes correlating multiple amino acids under LPD. Muscle proteolytic enzymes' genes such as ubb, capan1, and ctsl are also factors linking multiple amino acids under LPD. These transcripts made coreness-kmax and were linked to multiple amino acids (no indications in the figure). As a whole, most free amino acids have smaller degrees and corenesses. This network under LPD conditions appears to illustrate the activation of body protein breakdown and subsequent free amino acid generation. In contrast, HPD conditions provided most aminotransferases and deaminases with larger degrees and highest coreness. As plasma or tissue urea cycle metabolites and NH3 also bore the highest coreness, this network appears to illustrate activated whole body amino acid catabolism and ureagenesis (Fig. 4B). Additionally, tyrosine is another factor observed in HPD, where tyrosine has multiple links to its metabolic enzymes of aromatic amino acid pathways (Fig. 4B).


Figure 4
View larger version (31K):
[in this window]
[in a new window]

  		Fig. 4. Metabolite-to-gene correlation networks in rats fed low protein (A) or high protein (B) diet. Networks are visualized based on the Spearman {rho} matrices of 2 different entities: the aminogram from the blood and tissues and the tissue gene expression data. The node shape, size, and color indicate mode (metabolite or transcript), degree, and coreness, respectively. Red or blue lines signify the relationship connected each other inside or outside of the coreness-kmax, respectively. For this analysis, a correlation cutoff was placed at 0.65, greater than the minimum {rho} of 0.63 as indicated by a q-value of 0.05 (Fig. 3C), where the q-value estimation of the FDR suggests that correlations with an absolute value >0.63 are very likely to be reproducible. A total of 432 and 560 metabolite-to-gene correlates were applied to estimate networks in low protein and high protein data, respectively.

 
Based on the network obtained above, we confirmed highlighted correlate pairs particularly in ureagenesis and serine metabolic pathways (Fig. 5). Figure 5, A–C, represents plasma serine concentrations and correlation between plasma serine and expressions of genes related to the serine metabolism, respectively. Plasma serine level increased depending on the decrease in the level of protein intake (Fig. 5A). In LPD, expression of kidney 3-phoshoglycerate dehydrogenase (phgdh), a crucial enzyme in de novo serine synthesis, showed a clear positive correlation with plasma serine, while liver sds, a major serine catabolic enzyme, showed a negative correction with it (Fig. 5, B and C). In HPD, these metabolite-to-gene relationships showed quite different trends; the phgdh expression was repressed and lost correlation with plasma serine, while sds came to have a slight positive correlation with it (Fig. 5, B and C). Figure 5, D–F, represents plasma correlation plots between plasma urea and liver and kidney glutamate dehydrogenase and glutaminase expressions. Liver gls2 and kidney gls1 had clear positive correlations with plasma urea under HPD, but not LPD (Fig. 5D). Under HPD, kidney glud1 showed clear positive correlations with plasma urea, while liver glud1 had negative correlations (Fig. 5E). Here, muscle expressions of both genes did not show any clear correlation (data now shown). Correlation networks of both LPD and HPD groups clearly indicate higher dominance of sds and oat, where sds in particular shows a high degree of connectivity with urea in the HPD group. We found that liver sds and oat had the highest significant linear correlation with plasma urea among all pathway genes examined in this study, P = 0. 83 (n = 35, P < 0.0001) and 0.78 (n = 35, P < 0.0001), respectively (Fig. 5F).


Figure 5
View larger version (30K):
[in this window]
[in a new window]

  		Fig. 5. Changes in dietary protein-dependent correlation of plasma serine concentrations (A) and correlation plots of plasma serine against kidney 3-PGDH (B) liver SDH (C). Open and closed symbols represent the dietary conditions of LPD and HPD, respectively. Correlation plots of plasma urea against glutaminase (D) and GDH (E) expression in the liver and kidney. F: correlation plots of plasma urea against liver SDH (top) and OAT (bottom). RT-PCR data was normalized by 18S ribosomal RNA. Relative mRNA expressions were expressed relative to feeding control protein group (20% casein).

 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
Dietary model of "protein restriction-to-excess."
This study aimed to describe diet-dependent changes of amino acid metabolism by integrating aminograms and transcript profiles through correlation. A limitation in the correlation type of analysis exists in model validity, because the variables we were interested in from a nutritional aspect are sensitive to the degree of variation (45), and an introduction of an appropriate model will be crucial to overcome this limitation. From this, we chose a rat model of protein restriction-to-excess where rats were fed with a vastly wide range of casein contents in their diet, from 5 to 70%, for 2 wk and confirmed that this model could introduce enough changes in objective variables to be analyzed with validity. Most physiological parameters changed significantly with different types of trends such as upward, downward, convex, and concave; liver and kidney weights positively correlated with dietary protein levels, blood metabolites (glucose, cholesterol and BHB), and metabolic hormones (insulin, T4, IGF-1, and glucagon) changed so significantly that the metabolic changes due to dietary protein could be discussed in terms of the metabolic pathways and the endocrine regulatory system.

Contribution of dietary protein and tissue metabolism to plasma amino acids.
Steady-state plasma amino acid levels result from the rates of appearance (Ra) into and disappearance (Rd) from the plasma (9). Basically, Ra and Rd are tightly regulated by hormones and the tissue metabolism, and at the postabsorptive state, Ra equals Rd (9). To determine the interactions between such regulators and amino acids, aminograms obtained from the plasma, liver, kidney, small intestine, and muscle were subjected to correlation analyses. Most amino acids in the plasma correlated either negatively or positively with various physiological valuables, and the relationship was opposite between EAAs and nonessential ones (NEAAs) as a whole (data now shown).

Changes in plasma amino acids were shown to associate with physiological variables such as dietary protein, glucose, lipids, and hormones. EAAs, particularly BCAAs, showed positive correlations with protein intake, kidney mass, liver mass, NH3, and glucagons, indicating that those amino acid levels responded well to dietary protein intake. This is likely because the lack of de novo synthetic pathways for EAAs would reduce the contribution of tissue-originated EAAs to Ra, which exaggerates the contribution of dietary protein. By contrast, most plasma NEAAs showed clear positive correlations with carbohydrate intake, FAT mass, and plasma triglyceride. Negative correlations of NEAAs with protein uptake, glucagon, and liver and kidney weights appeared to illustrate that tissue-originated release of NEAAs would mainly depend on dietary carbohydrate intake or that NEAAs level would be regulated depending on tissue de novo synthetic flux from glucose. As for the relationships between NEAAs and lipids, the influences of amino acids on lipid synthesis have been reported in terms of control of "anaplerosis" and "cataplerosis" (7, 38) or the activation of hepatic lipogenic flux (3, 53). Thus, plasma NEAAs or EAA/NEAA ratio may be an indicator representing a balance of "anaplerosis" and "cataplerosis."

We hypothesized that plasma EAA levels are primarily maintained by postintestinal metabolism. To test this, we compared the plasma and tissue aminograms. The comparison was made separately for EAA and NEAA. As for EAA, the aminogram obtained from the intestine was consistently less correlated with the plasma aminogram than that from other tissues. Thus, the small intestine would have only minor effects on EAA metabolism. As for NEAA, however, the aminogram of the intestine had robust correlations with the plasma aminogram, particularly in LPD conditions, suggesting that NEAA was metabolically modified primarily by gastrointestinal metabolism.

Correlation analysis of aminogram and transcript profile.
Trials for integrative profiling of metabolites and transcripts (12) or proteins (30) have been reported. None of studies, however, have tried to link aminograms with transcripts expected to be involved in the metabolic pathways. To address this, we introduced in our correlation analysis study a repertoire of transcripts that were supposed to be involved not only in the liver but also kidney, muscle, and small intestine metabolic pathways on the assumption that their metabolism would significantly contribute to the whole body amino acid metabolism.

The present data showed comprehensible trends of amino acid-gene correlation. For instance, liver gls2, kidney glud1 and gls2, and small intestine glud1 positively and obviously correlated with plasma urea but not muscle tissue isozymes, but liver glud1 showed negative correlations with plasma urea. Furthermore, the power of these correlativities was modified by dietary protein levels. Because the nutritional status can alter the share of each tissue's metabolic contribution, the plasma aminograms, in turn, could be utilized as a marker for the metabolic status of each of the tissues.

In the present study, cytosolic pyridoxal-5' phosphate (PLP)-dependent enzyme sds showed significant increases in response to protein intake and highly correlates with plasma urea. Interestingly, sds expression highly correlated with oat expression. This enzyme is known also to be dependent on PLP and to control urea generation (31). Further, sds is reported to be induced by increased protein intake and correlates negatively with the nitrogen balance (24, 32). Based on the knowledge above, the present results could be interpreted to mean that cytosolic sds works together with mitochondrial oat and meets an increasing demand of ureagenesis activity under high protein feeding. Plasma serine level was shown to increase under low protein feeding, together with a sharp increase in the expression of kidney 3-phosphoglycerate dehydrogenase (phgdh) that was solely responsible for serine de novo synthesis. Thus, it is likely that the plasma serine level would be controlled mostly by its metabolic genes, phgdh and liver sds. We also detected this kind of relationship between metabolites and their synthetic genes of amino acids such as aromatic amino acids, glutamine, and glycine; thus, further analyses of metabolite-to-gene correlativity in different experimental models may be helpful to understand the network of metabolic pathways carried by different tissues.

A metabolite-to-gene network illustrated by correlation analyses.
Metabolic networks have been assembled by correlations using metabolic (46) or transcript data (40). Except for metabolite-gene networks reported in plants (35, 49), however, few trials have been carried out to combine different "types" of datasets. In this study, we utilized physiological variables, metabolites, and genes datasets and assembled the network by correlation analyses. To develop network structures, we used an analysis of two-mode network with k-core. An advantage in this kind of study is that the method allows for increasing the dimensions and integrating additional information into the resulting network graphs. As a result, a two-mode network illustrated direct metabolite-to gene interactions well, and k-core analysis was shown to be helpful to extract key structures in a given complex network, although the method had a potential limitation of not detecting interactions between the same entities with different origins (2, 6). More sophisticated methodologies have been developed, but it is always practical to apply these to biological datasets because of assumptions not suitable for biological experimentation (8, 29).

Under LPD conditions, the network structure highlighted the coreness-kmax of serine, which was associated with IGFBP-2, IGFBP-3, and proteolytic genes, while under HPD conditions, the network structure highlighted the coreness-kmax of urea cycle metabolites, which was associated with aminotransferases and deaminases. Contributions of both IGFBPs and proteolytic genes have been reported to adapt to protein restriction conditions (13, 48). Moreover, our network model successfully illustrated that the liver or the muscle genes responded to the peripheral availability of amino acids with activation or inhibition of ureagenesis in protein-excess feeding or protein-restriction diet, respectively. Information about the interaction between metabolites and gene expression will provide clues, revealing underlying specific regulatory mechanisms.

Recent technology improvements in NMR or mass spectrometry enable investigators to analyze biological molecules comprehensively in biological fluids, where blood proteomics and metabolomics have been reported for disease marker discovery. Of course, these approaches could be much more powerful for exploring phenotypic markers than the sort of focused analysis used in the present study. In many cases, however, it is difficult to understand underlying mechanisms just from obtained comprehensive "omics" data, whereas the focused analysis has the advantage in integrating different types of analyses such as gene, protein, and metabolite and also in speculating mechanisms from projected relational networks. Therefore, the detailed analysis of particular metabolic pathways by targeted analysis of lipids or amino acids with transcript profiles in related pathway genes might be useful to understand disease-specific regulatory mechanisms, as well as to assess drugs and nutrients, presumably in recent problems of metabolic syndrome such as obesity, diabetes, and hyperlipidemia. Most recently, a new methodology of stable isotopic flux analysis for comprehensive pathway analysis has been reported (54). A combination of relational networks and such flux analysis may certainly lead to establish new maps of metabolism.


   ACKNOWLEDGMENTS
 
We thank Kazutaka Shimbo for help with amino acid analysis.


   FOOTNOTES
 
Address for reprint requests and other correspondence: Y. Noguchi, Research Inst. for Health Fundamentals, Ajinomoto, 1-1 Suzuki-cho, Kawasaki-ku, Kawasaki 210-8681, Japan (e-mail: yasushi_noguchi{at}ajinomoto.com).

The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1 The online version of this article contains supplemental material. Back


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 

  1. Agostoni C, Carratu B, Boniglia C, Riva E, Sanzini E. Free amino acid content in standard infant formulas: comparison with human milk. J Am Coll Nutr 19: 434–438, 2000.[Abstract/Free Full Text]
  2. Alvarez-Hamelin JI, Dall'Asta L, Barrat A, Vespignani A. k-core decomposition: a tool for the visualization of large scale networks. Advances in Neural Information Processing Systems 18 41: http://arxiv.org/abs/cs.ni/050 4107, 2006.
  3. Baquet A, Lavoinne A, Hue L. Comparison of the effects of various amino acids on glycogen synthesis, lipogenesis and ketogenesis in isolated rat hepatocytes. Biochem J 273: 57–62, 1991.[Web of Science][Medline]
  4. Bernstein IM, Silver R, Nair KS, Stirewalt WS. Amniotic fluid glycine-valine ratio and neonatal morbidity in fetal growth restriction. Obstet Gynecol 90: 933–937, 1997.[CrossRef][Web of Science][Medline]
  5. Bijlsma S, Bobeldijk I, Verheij ER, Ramaker R, Kochhar S, Macdonald IA, van Ommen B, Smilde AK. Large-scale human metabolomics studies: a strategy for data (pre-) processing and validation. Anal Chem 78: 567–574, 2006.[Medline]
  6. Borgatti SP, Everett MG. Network analysis of 2-mode data. Social Network 19: 243–269, 1997.[CrossRef]
  7. Brunengraber H, Roe CR. Anaplerotic molecules: current and future. J Inherit Metab Dis 29: 327–331, 2006.[CrossRef][Web of Science][Medline]
  8. Brynildsen MP, Wu TY, Jang SS, Liao JC. Biological network mapping and source signal deduction. Bioinformatics 23: 1783–1791, 2007.[Abstract/Free Full Text]
  9. Cynober LA. Plasma amino acid levels with a note on membrane transport: characteristics, regulation, and metabolic significance. Nutrition 18: 761–766, 2002.[CrossRef][Web of Science][Medline]
  10. Degrell I, Hellsing K, Nagy E, Niklasson F. Amino acid concentrations in cerebrospinal fluid in presenile and senile dementia of Alzheimer type and multi-infarct dementia. Arch Gerontol Geriatr 9: 123–135, 1989.[CrossRef][Web of Science][Medline]
  11. Fischer JE, Funovics JM, Aguirre A, James JH, Keane JM, Wesdorp RI, Yoshimura N, Westman T. The role of plasma amino acids in hepatic encephalopathy. Surgery 78: 276–290, 1975.[Web of Science][Medline]
  12. Ghassemian M, Lutes J, Tepperman JM, Chang HS, Zhu T, Wang X, Quail PH, Lange BM. Integrative analysis of transcript and metabolite profiling data sets to evaluate the regulation of biochemical pathways during photomorphogenesis. Arch Biochem Biophys 448: 45–59, 2006.[CrossRef][Web of Science][Medline]
  13. Glass DJ. Signalling pathways that mediate skeletal muscle hypertrophy and atrophy. Nat Cell Biol 5: 87–90, 2003.[CrossRef][Web of Science][Medline]
  14. Graham GG, Baertl JM, Placko RP. Nitrogen balance and plasma aminogram in measuring supplemental effect of amino acids for children. J Agric Food Chem 20: 506–508, 1972.[CrossRef][Web of Science][Medline]
  15. Guthikonda S, Haynes WG. Homocysteine as a novel risk factor for atherosclerosis. Curr Opin Cardiol 14: 283–291, 1999.[CrossRef][Web of Science][Medline]
  16. Halmi KA, Struss AL, Owen WP, Stegink LD. Plasma and erythrocyte amino acid concentrations in anorexia nervosa. JPEN J Parenter Enteral Nutr 11: 458–464, 1987.[Abstract]
  17. Holm EA, Solling K. Dietary protein restriction and the progression of chronic renal insufficiency: a review of the literature. J Intern Med 239: 99–104, 1996.[CrossRef][Web of Science][Medline]
  18. Holt LE Jr, Snyderman SE, Norton PM, Roitman E, Finch J. Plasma Aminogram In Kwashiorkor. Lancet 1: 1278–1279, 1964.[Web of Science][Medline]
  19. Hsu JM, Anthony WL, Rider AA. Free amino acids in plasma and tissue of rats born to underfed dams. Proc Soc Exp Biol Med 148: 1087–1089, 1975.[CrossRef][Medline]
  20. Itoh H, Kaneko M, Ohshima S, Shumiya S, Sakaguchi E. Effect of low protein and low energy diet on physiological status and digestibility of F344 rats. Exp Anim 51: 485–491, 2002.[CrossRef][Web of Science][Medline]
  21. Iwatani S, Van Dien S, Shimbo K, Kubota K, Kageyama N, Iwahata D, Miyano H, Hirayama K, Usuda Y, Shimizu K, Matsui K. Determination of metabolic flux changes during fed-batch cultivation from measurements of intracellular amino acids by LC-MS/MS. J Biotechnol 128: 93–111, 2007.[CrossRef][Web of Science][Medline]
  22. Jackson AA. Limits of adaptation to high dietary protein intakes. Eur J Clin Nutr 53, Suppl 1: S44–S52, 1999.
  23. Jansen JJ, Hoefsloot HC, Boelens HF, van der Greef J, Smilde AK. Analysis of longitudinal metabolomics data. Bioinformatics 20: 2438–2446, 2004.[Abstract/Free Full Text]
  24. Jean C, Rome S, Mathe V, Huneau JF, Aattouri N, Fromentin G, Achagiotis CL, Tome D. Metabolic evidence for adaptation to a high protein diet in rats. J Nutr 131: 91–98, 2001.[Abstract/Free Full Text]
  25. Johansson U, Hagenfeldt L, Persson A, Siw Eriksson L. Parenteral nutrition in patients with liver cirrhosis. Effects on circulating levels of glucose and hormones and on cerebral function. Clin Nutr 8: 321–327, 1989.[CrossRef][Web of Science][Medline]
  26. Lai HS, Lee JC, Lee PH, Wang ST, Chen WJ. Plasma free amino acid profile in cancer patients. Semin Cancer Biol 15: 267–276, 2005.[CrossRef][Web of Science][Medline]
  27. Langley-Evans SC, Gardner DS, Jackson AA. Maternal protein restriction influences the programming of the rat hypothalamic-pituitary-adrenal axis. J Nutr 126: 1578–1585, 1996.[Abstract/Free Full Text]
  28. Lyttle JD, Goettsch E, Greeley DM, Grim WM, Dunbar P. Amino Acid Studies. II. Plasma amino acid retention, as evidence of impaired liver function investigations in children with nephrosis and liver disease. J Clin Invest 22: 169–181, 1943.[CrossRef][Web of Science][Medline]
  29. Ma HW, Zhao XM, Yuan YJ, Zeng AP. Decomposition of metabolic network into functional modules based on the global connectivity structure of reaction graph. Bioinformatics 20: 1870–1876, 2004.[Abstract/Free Full Text]
  30. Morgenthal K, Wienkoop S, Wolschin F, Weckwerth W. Integrative profiling of metabolites and proteins: improving pattern recognition and biomarker selection for systems level approaches. Methods Mol Biol 358: 57–75, 2007.[Medline]
  31. Morris JE, Peraino C. Immunochemical studies of serine dehydratase and ornithine aminotransferase regulation in rat liver in vivo. J Biol Chem 251: 2571–2578, 1976.[Abstract/Free Full Text]
  32. Moundras C, Remesy C, Demigne C. Dietary protein paradox: decrease of amino acid availability induced by high-protein diets. Am J Physiol Gastrointest Liver Physiol 264: G1057–G1065, 1993.[Abstract/Free Full Text]
  33. Nakanishi T, Otaki Y, Hasuike Y, Nanami M, Itahana R, Miyagawa K, Nishikage H, Izumi M, Takamitsu Y. Association of hyperhomocysteinemia with plasma sulfate and urine sulfate excretion in patients with progressive renal disease. Am J Kidney Dis 40: 909–915, 2002.[CrossRef][Web of Science][Medline]
  34. Nicholson JK, Wilson ID. Opinion: understanding ‘global’ systems biology: metabonomics and the continuum of metabolism. Nat Rev Drug Discov 2: 668–676, 2003.[CrossRef][Web of Science][Medline]
  35. Nikiforova VJ, Daub CO, Hesse H, Willmitzer L, Hoefgen R. Integrative gene-metabolite network with implemented causality deciphers informational fluxes of sulphur stress response. J Exp Bot 56: 1887–1896, 2005.[Abstract/Free Full Text]
  36. Noguchi Y, Sakai R, Kimura T. Metabolomics and its potential for assessment of adequacy and safety of amino acid intake. J Nutr 133: 2097S-2100S, 2003.[Abstract/Free Full Text]
  37. Noguchi Y, Zhang QW, Sugimoto T, Furuhata Y, Sakai R, Mori M, Takahashi M, Kimura T. Network analysis of plasma and tissue amino acids and the generation of an amino index for potential diagnostic use. Am J Clin Nutr 83: 513S–519S, 2006.[Abstract/Free Full Text]
  38. Owen OE, Kalhan SC, Hanson RW. The key role of anaplerosis and cataplerosis for citric acid cycle function. J Biol Chem 277: 30409–30412, 2002.[Free Full Text]
  39. Q Value. Q value software. [http://faculty.washington.edu/jstorey/qvalue/].
  40. Reverter A, Hudson NJ, Wang Y, Tan SH, Barris W, Byrne KA, McWilliam SM, Bottema CD, Kister A, Greenwood PL, Harper GS, Lehnert SA, Dalrymple BP. A gene coexpression network for bovine skeletal muscle inferred from microarray data. Physiol Genomics 28: 76–83, 2006.[Abstract/Free Full Text]
  41. Ring BZ, Chang S, Ring LW, Seitz RS, Ross DT. Gene expression patterns within cell lines are predictive of chemosensitivity. BMC Genomics 9: 74, 2008.[CrossRef][Medline]
  42. Sakai R, Miura M, Amao M, Kodama R, Toue S, Noguchi Y, Kimura T. Potential approaches to the assessment of amino acid adequacy in rats: a progress report. J Nutr 134: 1651S–1655S, 2004.[Abstract/Free Full Text]
  43. Snyderman SE, Holt LE Jr, Nortn PM, Roitman E, Phansalkar SV. The plasma aminogram I. Influence of the level of protein intake and a comparison of whole protein and amino acid diets. Pediatr Res 2: 131–144, 1968.[CrossRef][Web of Science][Medline]
  44. Solerte SB, Gazzaruso C, Schifino N, Locatelli E, Destro T, Ceresini G, Ferrari E, Fioravanti M. Metabolic effects of orally administered amino acid mixture in elderly subjects with poorly controlled type 2 diabetes mellitus. Am J Cardiol 93: 23A–29A, 2004.[CrossRef][Web of Science][Medline]
  45. Steinfath M, Repsilber D, Scholz M, Walther D, Selbig J. Integrated data analysis for genome-wide research. EXS 97: 309–329, 2007.[Medline]
  46. Steuer R, Kurths J, Fiehn O, Weckwerth W. Observing and interpreting correlations in metabolomic networks. Bioinformatics 19: 1019–1026, 2003.[Abstract/Free Full Text]
  47. Storey JD, Tibshirani R. Statistical significance for genomewide studies. Proc Natl Acad Sci USA 100: 9440–9445, 2003.[Abstract/Free Full Text]
  48. Straus DS, Takemoto CD. Effect of dietary protein deprivation on insulin-like growth factor (IGF)-I and -II, IGF binding protein-2, and serum albumin gene expression in rat. Endocrinology 127: 1849–1860, 1990.[Abstract/Free Full Text]
  49. Szymanski J, Bielecka M, Carrari F, Fernie AR, Hoefgen R, Nikiforova VJ. On the processing of metabolic information through metabolite-gene communication networks: an approach for modelling causality. Phytochemistry 68: 2163–2175, 2007.[CrossRef][Web of Science][Medline]
  50. Torok B, Soltesz G, Szmolenszky T, Roth E. The aminogram of the urine produced by the preserved kidney. Int Urol Nephrol 11: 67–70, 1979.[CrossRef][Web of Science][Medline]
  51. Ventura P, Panini R, Emiliani S, Salvioli G. Plasma homocysteine after insulin infusion in type II diabetic patients with and without methionine intolerance. Exp Clin Endocrinol Diabetes 112: 44–51, 2004.[CrossRef][Web of Science][Medline]
  52. Wachi S, Yoneda K, Wu R. Interactome-transcriptome analysis reveals the high centrality of genes differentially expressed in lung cancer tissues. Bioinformatics 21: 4205–4208, 2005.[Abstract/Free Full Text]
  53. Wong MS, Raab RM, Rigoutsos I, Stephanopoulos GN, Kelleher JK. Metabolic and transcriptional patterns accompanying glutamine depletion and repletion in mouse hepatoma cells: a model for physiological regulatory networks. Physiol Genomics 16: 247–255, 2004.[Abstract/Free Full Text]
  54. Yoo H, Antoniewicz MR, Stephanopoulos G, Kelleher JK. Quantifying reductive carboxylation flux of glutamine to lipid in a brown adipocyte cell line. J Biol Chem, in press.
  55. Zeller KR. Low-protein diets in renal disease. Diabetes Care 14: 856–866, 1991.[Abstract]




This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental Figures and Tables
Right arrow All Versions of this Article:
34/3/315    most recent
00007.2008v1
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Noguchi, Y.
Right arrow Articles by Takahashi, M.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Noguchi, Y.
Right arrow Articles by Takahashi, M.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Visit Other APS Journals Online
Copyright © 2008 by the American Physiological Society.