AJP - Endo Fuel your research with LabChart
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Am J Physiol Endocrinol Metab 292: E1637-E1646, 2007. First published February 6, 2007; doi:10.1152/ajpendo.00670.2006
0193-1849/07 $8.00
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental Tables
Right arrow All Versions of this Article:
292/6/E1637    most recent
00670.2006v1
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 ISI Web of Science
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 ISI Web of Science (2)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Si, Y.
Right arrow Articles by Lee, K.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Si, Y.
Right arrow Articles by Lee, K.

Flux profile and modularity analysis of time-dependent metabolic changes of de novo adipocyte formation

Yaguang Si,1 Jeongah Yoon,2 and Kyongbum Lee2

Departments of 1Biology and 2Chemical and Biological Engineering, Tufts University, Medford, Massachusetts

Submitted 8 December 2006 ; accepted in final form 31 January 2007


   ABSTRACT
TOP
 ABSTRACT
METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
White adipose tissue (WAT) mass is the main determinant of obesity and associated health risks. WAT expansion results from increases in white adipocyte cell number and size, which in turn reflect a series of shifts in the cellular metabolic state. To quantitatively profile the metabolic alterations occurring during de novo adipocyte formation, metabolic flux analysis (MFA) was used in conjunction with a novel modularity analysis algorithm on differentiating 3T3-L1 preadipocytes. Use of a type I collagen gel as an effective long-term culture substrate was also assessed. The calculated flux distributions predicted the sequential activation of several intracellular cross-compartmental pathways, including lipogenesis, the pentose phosphate pathway, and the malate cycle, in good agreement with earlier isotopic tracer experiments and gene profiling studies. Partition of the adipocyte metabolic network into highly interacting reaction subgroups suggested a functional reorganization of the major pathways consistent with the lipid-loading phenotype of the adipocyte. Flux and modularity analysis results together point to the flux distribution around pyruvate as a key indicator of adipocyte lipid accumulation.

adipocyte metabolism; metabolic profile; network modularity

AS A CHRONIC CONDITION, OBESITY increases the risk for many diseases, including diabetes and some forms of cancer (48). The effects of dietary therapies are mostly reversible, and <10% of those who lose weight are able to maintain the weight loss (4). With few drugs (two as of 2005, Ref. 32) approved for long-term treatment, there is a pressing need to develop additional alternative therapeutics or therapies. Solid epidemiological data support the pivotal role of body fat (white adipose tissue; WAT) mass in the development of the obesity-related risk factors. Adiposity involves increases in both fat cell size (hypertrophy) and number (hyperplasia) (1), where hypertrophy usually precedes hyperplasia in a cyclical manner (11, 19, 20). Hypertrophy occurs through lipid accumulation and directly relates to the primary metabolic function of the WAT. Because mature adipocytes generally do not divide, hyperplasia involves the formation of new adipocytes, or adipogenesis, via the recruitment and differentiation of locally present precursor cells.

In recent years, it has become clear that the WAT not only acts as a storage depot but actively regulates its own as well as whole body metabolism through paracrine and endocrine mediators (36, 41). In this light, one approach to controlling adiposity could be to target metabolic processes that are directly tied to adipogenesis and lipid accumulation. This approach requires a systematic investigation of WAT-specific metabolism, preferably in isolation from confounding systemic influences. In this work, we characterized the metabolism of adipocytes derived from the 3T3-L1 cell line, which is widely used to study in vitro adipogenesis. Recently, gene chip analyses have been performed to globally characterize transcription-level changes of adipocyte differentiation (40). Less work has been done to comprehensively profile metabolic activity changes, which take place not only during early differentiation but also throughout the ensuing stages of hypertrophic growth (25). The goal of this study was to address this knowledge gap. Because this study required a stable experimental setting for long-term culture, we have also examined the use of type I collagen gel as a culture substrate. Although standard treated tissue culture plastic (TCP) yields satisfactory differentiation outcomes, its ability to anchor lipid-filled adipocytes and thus prevent cell loss is limited. Experiences with other anchorage-dependent cells suggest that an extracellular matrix (ECM) may improve adhesion as well other differentiated cell functions, but few data are available on the metabolic effects. Our results showed that the flux distribution of the adipocyte metabolic network continues to adjust well after terminal differentiation has occurred. The major effect of the collagen gel was to enhance differentiation-related metabolic activities while inhibiting preadipocyte proliferation.


   METHODS
TOP
 ABSTRACT
 METHODS
RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 
Materials. 3T3-L1 preadipocytes were obtained from American Type Culture Collection (Manassas, VA). Tissue culture reagents including DMEM, FCS, FBS, human insulin, and penicillin/streptomycin were purchased from Invitrogen (Carlsbad, CA). Type I rat tail collagen was purchased from Roche Applied Science (Indianapolis, IN). Unless otherwise noted, all other chemicals were purchased from Sigma (St. Louis, MO).

Cell culture and differentiation. 3T3-L1 preadipocytes were first expanded in T-75 flasks with a standard growth medium (DMEM with 10% FCS), passed once, and randomly divided into two groups. One group was seeded in 12-well TCP plates. The other group was seeded in 12-well plates whose wells had been coated with a collagen gel layer. The collagen gel mixture consisted of type I rat tail collagen (1.2 mg/ml in 1 mM HCl), 10x DMEM, and 75 mg/ml NaHCO3 at a volume ratio of 8.5:1:0.5, respectively. Each well was coated with 0.5 ml of this mixture, which was allowed to gel at 37°C for 90 min. The seeding density was 2 x 104 cells/well for both the TCP and collagen gel cultures. Differentiation was induced by exposing cultures to DMEM supplemented with 10% FBS and an adipogenic cocktail consisting of 1 µg/ml insulin, 0.5 mM IBMX, and 1 µM dexamethasone (DEX). On day 2, the induction medium was replaced with a second medium without IBMX and DEX. On day 4, the second induction medium was replaced with a basal maintenance medium consisting of DMEM and FBS. This basal medium was replenished every other day during the remainder of the culture experiments.

Microscopy. At the indicated time points, cellular morphology was recorded using phase-contrast microscopy (Nikon-US, Melville, NY). Intracellular lipid droplets were visualized by staining with Oil Red O (16).

Glycerol-3-phosphate dehydrogenase enzyme activity and leptin. Glycerol-3-phosphate dehydrogenase (GPDH; EC 1.1.1.8 [EC] ) activities were measured in situ (39) by monitoring the rate of change of NADH absorbance at 340 nm. Leptin was measured on spent medium samples by ELISA (DuoSet System; R&D Systems, Minneapolis, MN).

Metabolite assays. Metabolite measurements were performed on cell lysates and spent medium samples. Cells cultured on the collagen gel were first released by collagenase (aqueous solution, 0.1% wt/vol) treatment. Cell were then lysed in situ with an 0.1% SDS buffer and briefly sonicated. Immediately after collection, the spent medium samples were cleaned of cell debris by a brief centrifugation step. Triglyceride (TG) levels were measured, using an assay kit (Sigma) that is based on the release of glycerol from TG by lipoprotein lipase (LPL). Free glycerol (unbound to TG) was measured by substituting water for the LPL in the assay reagent mixture. Free fatty acids (FFA) were measured using an assay kit (Roche Applied Science) that is based on the enzymatic conversion of FFA into acyl-CoA. Glucose and lactate concentrations were measured using enzymatic assays based on the methods of Trinder (43) and Loomis (28), respectively. Concentrations of beta-hydroxybutyrate and acetoacetate were determined, using the method of Olsen (35), by monitoring, respectively, the production or consumption of NADH in a reaction with beta-hydroxybutyrate dehydrogenase. Amino acids were quantified by HPLC using fluorescence-based detection following precolumn derivatization of primary or secondary amines with 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate (8). All metabolite data were normalized by the corresponding cell sample DNA content, which was determined with a fluorescence-based assay using the Hoechst dye.

Stoichiometric model. A stoichiometric model of adipocyte central carbon metabolism was constructed as follows. First, mouse-specific lists of enzyme-mediated reactions were collected from an annotated genome database (29). Second, stoichiometric information was added for each of the collected enzymes by cross-referencing their common names and enzyme commission names (21). Third, biochemistry textbooks (27, 42) and the published literature (12) were consulted to eliminate enzymes thought to be inactive in the fed state. Finally, the following assumptions were applied: 1) glycogen synthesis is negligible (22); 2) fatty acid oxidation is small compared with both lipogenesis and lipolysis (47); 3) the pentose phosphate pathway (PPP) is operating in the oxidative mode (6); and 4) net protein synthesis is small compared with metabolic fluxes (17). The stoichiometric model was rendered into a compound, directed graph, visualized using the Bioinformatics toolbox of MATLAB (MathWorks, Natick, MA), and corrected for missing steps and nonsensical dead ends. Reversible reactions flanked by irreversible reactions were assigned directionality so as to ensure unidirectional metabolic flux between the flanking reactions. The final adipocyte model consisted of the following pathways: anaplerosis, glycolysis, glycerogenesis, ketone body synthesis, lipogenesis, lipolysis, the malate cycle, the PPP, and the tricarboxylic acid (TCA) cycle (Supplemental Table S1; supplemental data are available at the online version of this article).

Flux analysis. Intracellular fluxes were estimated by solving a constrained nonlinear optimization problem as described previously (34)

Formula

Formula 1(1)

Formula 2(2)
where the objective is to minimize the sum squared error between experimentally observed and calculated exchange fluxes. Equation 1 expresses the balances around intracellular metabolites using an M x N stoichiometric matrix S and an N x 1 steady-state flux distribution vector v. Inequality (Eq. 2) expresses constraints derived from the Gibbs free energy change ({Delta}G) form of the Second Law.

Network partition. Reaction clusters reflecting significant metabolic modules were characterized using an algorithm for top-down partition of directed and weighted graphs (50). The algorithm iteratively applies the following two steps until all edges in the graph have been removed. 1) Shortest, weighted paths through the network are calculated using Dijkstra's algorithm (9). Edge-weights were assigned such that the length of an edge was inversely proportional to the corresponding reaction flux. 2) The edge-betweenness centrality index is calculated for all edges, and the edge with the highest index value is removed. This index measures how frequently an edge lies on the shortest paths between all pairs of vertices. It has been shown that the edges with highest betweenness values are most likely to lie between subgraphs rather than inside a subgraph (33). Successively removing edges with the highest edge-betweenness values should eventually isolate subgraphs consisting of vertices that share connections only with other vertices in the same subgraph.

Module scoring. We applied a pair of projection-based scoring metrics to guide in the selection of a partition that best reflects the underlying modularity of the network for a given set of edge-weights (i.e., flux distribution). Details of the scoring algorithm have been presented elsewhere (51). Each subgraph of a partition was represented by a 1 x R binary reaction composition vector (RCV), where R is the total number of reactions included in the stoichiometric model. An element was set to one if both the reactants and products of the corresponding reaction were present as nodes in the module; otherwise, it was set to zero. Each RCV was projected onto a pathway space defined by the elementary flux modes (EFMs) (37) of the parent network. Our goal was to assess the reaction composition of a subgraph, or "module." Thus each EFM was also formatted as a 1 x R binary pathway inventory vector (PIV) by replacing all nonzero entries with one. A projection score was computed for every pairwise combination of RCV and PIV

Formula 3(3)
In Eq. 3, PSFormula 3, RCVFormula 3, and NFormula 3 are, respectively, the projection score, reaction composition vector, and number of nodes in module i at iteration k; Lk is the number of modules at k; PIVj is the jth PIV, and m is the total number of EFMs. The overall projection score of an iteration k was calculated by averaging the "best-match" projection scores of this iteration

Formula 4(4)
In cases where the projection score identified more than one best-match PIV for a given subgraph, a 1 x R consensus pathway fragment (CPF) vector was calculated by identifying the reactions conserved across every best-match PIV. The similarity between a subgraph and its CPF was assessed by the match score

Formula 5(5)
where MSFormula 5 was the match score of module i at iteration number k, and WFormula 5 was the number of mismatches between the module RCVFormula 5 and the corresponding CPF. A mismatch occurs if a reaction is found in the module but not the fragment vector, or if a reaction is found in the fragment but not the module vector. The overall match score of an iteration was calculated by averaging the individual module match scores

Formula 6(6)

Statistics. Comparisons between two experimental groups were performed using one-way ANOVA. Group means were deemed to be statistically significantly different when P < 0.01.


   RESULTS
TOP
 ABSTRACT
 METHODS
 RESULTS
DISCUSSION
 GRANTS
 REFERENCES
 
Cell proliferation and differentiation. Intracellular lipid droplets first became visible on day 4 postinduction. Round and lipid-filled adipocytes appeared by day 8 (Fig. 1). The number of cells, as measured by its DNA content, continued to vary throughout the course of differentiation for both the TCP and collagen cultures (see GoFig. 3A), suggesting that additional rounds of clonal expansion took place after the addition of the induction cocktail. Cell proliferation was most active during the first 4 days. During this period, a threefold increase was observed for both TCP and collagen cultures. The DNA contents were consistently less on the collagen gel across all time points, presumably because the proliferation rate was slower during the preadipocyte expansion. Although the seeding density was the same, the DNA content at the time of induction (day 0) was 56% less on collagen. After reaching a peak (day 4 for TCP and day 8 for collagen culture), culture sizes slightly decreased (ca. 14~15%) by the next time point and remained stable thereafter. Prior reports have noted a similar decrease and attributed this to cell detachment during medium change (38). Motivated by these findings, all metabolite measurements in this study were normalized by the DNA content of the corresponding cell sample.


Figure 1
View larger version (106K):
[in this window]
[in a new window]

  		Fig. 1. Representative Oil Red O-stained images of newly differentiating 3T3-L1 adipocytes on days 0, 4, 8, 12, and 16 postinduction (top to bottom). Left- and right-hand columns refer to tissue culture plastic (TCP) and type I rat tail collagen gel cultures, respectively.

 

Figure 2
View larger version (13K):
[in this window]
[in a new window]

  		Fig. 2. Effect of culture substrate on glycerol-3-phosphate dehydrogenase (GPDH) activity (A; n = 6) and leptin secretion (B; n = 4). Data shown are means ± SD.

 

Figure 3
View larger version (24K):
[in this window]
[in a new window]

  		Fig. 3. Time profiles of metabolite uptake, output, or accumulation. Data shown are means ± SD (n = 6) of total DNA per well (A), intracellular triglyceride (TG; B), cellular free fatty acid (FFA; C), glucose uptake (D), lactate output (E), glycerol output (F), lactate-to-glucose ratio (G), and amino acid exchange (H). GLN, glutamine; ALA, alanine.

 
Adipogenic differentiation was assessed by GPDH activity (Fig. 2A) and leptin secretion (Fig. 2B). Enzyme activities for the TCP cultures were first normalized by total protein for quantitative comparisons against published data. We found that GPDH activity increased significantly from 0.01 to 0.29 mU/µg protein between days 4 and 12, showing good agreement with the literature (38). The effect of the culture substrate was assessed based on DNA-normalized activity, because the residual ECM proteins remaining after collagenase digest confounded the measurements on total cellular protein. No significant activity was detected on either substrate at day 0. GPDH activity (in mU/µg DNA) significantly increased to similar extents on both collagen and TCP between days 4 and 8. By day 12, GPDH activity was almost threefold greater on collagen. Significant amounts of leptin were not detected in the medium samples of either collagen or TCP culture until day 8. Leptin secretion (in pg·µg DNA–1·2 days–1) peaked before day 12 for cells on TCP but continued to increase on collagen. At day 12, leptin secretion by cells on collagen was about threefold greater compared with the cells on TCP.

Metabolite uptake and output. We examined the rates of uptake, output, and accumulation of 26 primary metabolites to broadly sample the activities of major metabolic pathways in adipocytes. Figure 3 shows the time profiles of selected lipid and carbohydrate metabolites that accounted for the bulk of the total carbon flux. Consistent with the morphological changes (Fig. 1), intracellular TG was first detected at day 4 and rapidly increased thereafter (Fig. 3B). The increase was more rapid in cells on collagen; at day 8, there was a 34% difference in TG content (in mmol/g DNA), which grew to 43% at day 12. A similar trend was observed for intracellular FFA, with concentrations in the collagen cultures 60 and 62% larger compared with TCP at days 8 and 12, respectively (Fig. 3C).

The main carbon source for de novo lipid synthesis was medium glucose, because FFA was not added to the culture medium. For both collagen and TCP cultures, glucose consumption (in mmol·g DNA–1·2 days–1) increased through the first 8 days but declined thereafter (Fig. 3D). The collagen cultures consumed more glucose at all times, with differences ranging from 27% at day 4 to 109% at day 12. The lactate output profile mirrored the glucose consumption profile. Peak output was observed at day 8, with steep declines thereafter for both TCP and collagen cultures (Fig. 3E). Lactate output (in mmol·g DNA–1·2 days–1) was greater for the collagen cultures, with differences ranging from 14% at day 4 to 150% at day 12. The molar ratio of lactate output to glucose uptake declined throughout the study period for both cultures, from 2.4 (TCP) and 1.5 (collagen) on day 0 to 0.5 (TCP) and 0.6 (collagen) on day 16 (Fig. 3F). A significant increase in glycerol output (a measure of lipolysis) was observed for both cultures by day 4 (Fig. 3G). Similar to the other metabolites, the collagen cultures exhibited a higher rate of output throughout the study period, albeit with a delayed peak time (day 12) compared with TCP (day 8).

Amino acids were generally utilized to smaller extents compared with the above lipid and carbohydrates. The rates of output or uptake of most amino acids were on the order of a few millimoles per grams DNA per 2 days (mmol·g DNA–1·2 days–1) and remained relatively stable over the duration of the study period. One exception was glutamine (GLN), whose output rate (in mmol·g DNA–1·2 days–1) increased significantly from –4.3 and 11.3 at day 4 to 29 and 48 at day 8, respectively, for the TCP and collagen cultures. The output of alanine (ALA) decreased significantly from 15 and 25 at day 4 to 6.6 and 12 at day 8 (Fig. 3H). The largest uptake rates were observed for the branched chain amino acids (BCAAs) and serine, a glucogenic amino acid (Supplemental Table S2).

Flux profiles. The significant increase in per well DNA content from day 0 to 4 indicated that the cultures initially consisted mainly of preadipocytes, because terminally differentiated adipocytes are thought to be incapable of cell division. Our stoichiometric model was developed for adipocytes; thus metabolic flux analysis (MFA) was applied to the metabolite data of days 4, 8, 12, and 16 postinduction. Complete results of the flux calculation can be found in Supplemental Table S3. To obtain an overview, we first performed a series of statistical comparisons between the flux profiles of the collagen and TCP cultures at various times using ANOVA. These comparisons were performed with and without normalization by the glucose flux, which was the main carbon source, to determine whether differences between the two culture substrates represented a general upregulation of metabolic activity or redistribution of the carbon fluxes or both. For all time points, the number of significantly different reaction fluxes was decreased by the normalization (Table 1), suggesting that the flux differences between the two substrate groups likely reflected an overall upregulation of metabolic activity on the collagen gel compared with TCP. Motivated by these observations, the remainder of this section focuses on the flux distribution changes in adipocytes growing on TCP (Fig. 4). As suggested by the glucose uptake profile, flux through glycolysis first increased significantly, 1.5-fold, from day 4 to 8 and then decreased 1.6-fold from day 8 to 12. Flux through the PPP increased from zero to 88 mmol·g DNA–1·2 days–1 between days 4 and 8 and leveled off thereafter. Consistent with these changes, our model calculated a steady decrease in flux through the reactions of glycolysis from day 8 to day 12. Taken together with a large decrease in lactate dehydrogenase flux, our calculations pointed to a shift in glucose utilization away from glycolysis to lipogenesis and the TCA cycle.


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

  		Table 1. Comparison of glucose-normalized fluxes in TCP and collagen gel cultures

 

Figure 4
View larger version (32K):
[in this window]
[in a new window]

  		Fig. 4. Time profile of flux distribution through major metabolic pathways in adipocytes in TCP cultures. Data shown correspond to days 4, 8, and 12 postinduction (top to bottom). Extracellular metabolites are shown in bold; measured metabolites are indicated by italics. Values shown are means ± SD (n = 6). All units are in mmol·g DNA–1·2 days–1. PYR, pyruvate; G3P, glycerol-3-phosphate; R5P, ribulose 5-P; subscript m denotes mitochondrial metabolites.

 
The pyruvate (PYR) flux entering the mitochondria more than doubled from day 4 to 8, after which it dipped by 16%. Within the mitochondria, PYR can enter the TCA cycle either as acetyl-CoA (ACC) via condensation with oxaloacetate (OAA) or as OAA via a carboxylation reaction. From day 4 to 8, the condensation and carboxylation reactions increased two- and sevenfold, respectively. Increased PYR flux entering the TCA cycle from day 4 to day 8 was correlated with a ninefold increase of flux transporting citrate from the mitochondria into the cytosol. As a result, the net carbon flux leading to the formation of the mitochondrial reducing cofactors NADH and FADH2 stayed relatively unchanged. Flux through oxidative phosphorylation also remained stable, as did the predicted O2 uptake rate. Significant changes were calculated for CO2 output, which first increased 1.6-fold from day 4 to 8 and then decreased 1.3-fold from day 8 to 12.

In the cytosol, citrate is cleaved to regenerate ACC, an important substrate for fatty acid synthesis. The flux leading to ACC formation increased 12-fold from day 4 to 8 and leveled off thereafter. Malic enzyme and cytosolic isocitrate dehydrogenase (IDHc) generate NADPH, an important cofactor for fatty acid synthesis. From day 4 to 8, fluxes through malic enzyme and IDHc significantly increased three- and fivefold, respectively. Phosphoenolpyruvate carboxykinase (PEPCK) catalyzes the conversion of cytosolic OAA into phosphoenolpyruvate, a key step in glycerogenesis. Flux through PEPCK increased 12-fold from day 4 to 8 and leveled off thereafter.

Significant changes for the reaction fluxes around TG were estimated mainly for the period between days 4 and 8. During this time, fluxes through fatty acid synthesis, TG synthesis, and lipolysis were all upregulated. Fatty acid synthesis increased nearly 12-fold, whereas TG synthesis and lipolysis both increased 3- to 4-fold. The rate of TG accumulation, calculated as the net difference between lipogenesis and lipolysis, increased 13-fold. In adipocytes, where NADPH is mainly used for de novo fatty acid synthesis, the cofactor can be supplied by way of the PPP, the malate cycle, or IDHc. The PPP was inactive early on (day 4) but supplied the majority (60%) of NADPH for de novo fatty acid synthesis by day 8.

Amino acid fluxes were considered by grouping them according to their points of contact with the TCA cycle and glycerogenesis. Overall, amino acids were not significantly metabolized, and we mention here only those whose flux profiles exhibited significant changes over time. The conversion of PYR into ALA decreased significantly from 15 (mmol·g DNA–1·2 days–1) at day 4 to 6.6 at day 8, contributing to the relative increase in PYR flux into the TCA cycle. Glutamate (GLU) uptake and histidine conversion into GLU both increased significantly, five- to sixfold, from day 4 to 8, while the conversions of GLU into 2-oxoglutarate and proline both decreased. The net effect was to provide additional substrates for GLN synthesis. The BCAAs isoleucine and valine enter the TCA cycle by donating their carbon moieties to succinyl-CoA. These fluxes increased significantly (1.6- and 1.7-fold, respectively) from day 4 to 8 but decreased thereafter. A similar profile was observed for leucine, another BCAA, whose deamination products become ACC and acetoacetyl-CoA.

Modularity. To explore how the flux distribution changes correlated with coordinated activity adjustments of the major pathways, we examined the functional organization of the adipocyte metabolic network by applying our partition algorithm. This approach to grouping metabolic reactions is fundamentally different from both purely topological as well as correlation analyses (e.g., clustering), because significant groupings are considered only if the components are directly connected to each other via enzymatic reactions and if these enzymes are highly active. Our algorithm found three types of modularity patterns corresponding to the different stages of adipogenesis: newly differentiating (day 4), maturing (day 8), and fully differentiated adipocyte (days 12 and 16; the modules for days 12 and 16 were identical). The day 4 partition consisted of three subgraphs, or modules (Fig. 5A). The largest subgraph encompassed the metabolites of the TCA cycle, malate cycle, glycolysis, and PPP. The two smaller subgraphs each included amino acids and their catabolic products, where the smallest involved the ketone body metabolism intermediates and the other involved entry points into the TCA cycle. For day 8, only one module was identified, essentially consisting of all major pathways needed for lipogenesis (Fig. 5B). This module was similar to that found for days 12 and 16 (Fig. 5C), except that segments of the malate cycle were disconnected. The addition of a complete malate cycle at days 12 and 16 likely indicates a greater demand for NADPH and is consistent with the continued increase in lipid loading through day 16 observed in this study (Figs. 1 and 3B) as well as earlier observations that the malate cycle quantitatively contributes to the pool of reducing equivalents required for lipid biosynthesis in mature adipocytes (13, 31, 52). The overall trend from day 4 to 16 was for the various pathways to coalesce into a single module whose main biochemical function was to convert input (glucose) into TG.


Figure 5
View larger version (19K):
[in this window]
[in a new window]

  		Fig. 5. Modularity of core adipocyte metabolic network on day 4 (A), day 8 (B), and day 12 (C) postinduction. Results for days 12 and 16 were identical. Arrow thickness is proportional to corresponding reaction flux.

 

   DISCUSSION
TOP
 ABSTRACT
 METHODS
 RESULTS
 DISCUSSION
GRANTS
 REFERENCES
 
Previous work on gene expression (40), differentiation kinetics (38), and cell aging (25) indicated that biochemical changes occur throughout the life cycle of the adipocyte. Here, we extend on these studies to show that adipogenesis involves significant, sequential alterations to the distribution of reaction fluxes around several of the major intermediates of adipocyte primary metabolism. Results of our modularity analysis suggest that these flux adjustments reflect a restructuring of the metabolic network's functional layout. These observations underscore the dynamic character of adipocyte growth, which is not explained by a uniform activation of lipogenic reactions. In this discussion, we present evidences collected from the literature that support the flux estimates of this study, point out the possible limitations of our current model, and highlight the potential impact of PYR reaction fluxes on adipocyte lipid accumulation.

Incorporating free energy directionality constraints into an optimization-based framework for MFA, we obtained robust estimates of all intracellular reaction fluxes in our model, including the intercompartmental exchange fluxes between mitochondria and cytosol. As a summary assessment, we compared carbon distributions derived from the flux estimates against published data and found generally good agreement (Table 2). Quantitative agreement with literature data was also found for the calculated fluxes of several pathways whose intermediates were not directly measured in this study. For example, our model indicated that the PPP supplies 60% of the NADPH needed for fatty acid synthesis in mature adipocytes, which is in excellent agreement with the results of isotopic tracer studies (14, 22, 24). A current estimate (46) for the oxygen uptake rate of day 14 postinduction 3T3-L1 cells is 2.9 nmol/min per 4x105 cells, or 1,147 mmol·g DNA–1·2 days–1 (assuming a conversion factor of 18.2 pg DNA/cell; unpublished observation), which is quantitatively similar to the rates calculated in this study (626 and 801 mmol·g DNA–1·2 days–1, respectively, for TCP and collagen cultures at day 16).


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

  		Table 2. Comparison of calculated and literature-derived carbon distributions

 
We have also examined the predicted timing of flux changes in the context of previously described gene activation/repression profiles of in vitro adipogenesis. According to the model calculations, ACC flux and de novo formation of fatty acids both increase significantly from day 4 to 8, consistent with a previous report on the activation time profile of ATP-citrate lyase, ACC carboxylase, and fatty acid synthetase (30). Fatty acid synthesis depends on the availability of cytosolic NADPH, which is supplied primarily through glucose-6-phosphate dehydrogenase (G6PDH), malic enzyme, and cytosolic IDHc (26, 44). A transcriptional profiling experiment (40) has shown that G6PDH expression is nearly undetectable until day 4, after which time it sharply increases and then levels off, which was also described by the MFA results obtained here. The flux time profiles also correctly predicted increases in the activities of malic enzyme (15) and IDHc (26) between days 4 and 8 following adipogenic induction. Run-off assays on nuclei isolated from 3T3-F442A cells (10) showed that the transcription of the glycerogenesis enzyme PEPCK is a differentiation-dependent event. Consistent with this observation, our model results indicated a sharp (12-fold) increase in PEPCK flux between days 4 and 8 (Supplemental Table S3a, reaction no. 21), which also coincided with significant induction of other differentiation markers (Fig. 2). A similar activation was also predicted for a regulated TCA cycle enzyme, pyruvate dehydrogenase (PDH; Supplemental Table S3a, reaction no. 10), which has been shown to be significantly induced in 3T3-L1 cells on treatment with a differentiation cocktail (18).

The MFA results pointed to a strong correlation between the redistribution of reaction fluxes around the PYR node and induction of lipogenic activity, even though the PYR reactions did not share direct stoichiometric links with lipogenesis. Between days 8 and 12, a stable rate of TG synthesis was maintained essentially unchanged despite a 40% drop in glucose uptake. During this time, fluxes through PPP and PDH remained statistically unchanged, but lactate production decreased by 60%. Overall, the flux ratio of lactate output to glucose uptake decreased continuously throughout the 16-day study period (Fig. 3F), indicating a shift in carbon utilization from lactate fermentation to aerobic respiration. Interestingly, clinical studies have shown that the lactate output of white adipocytes isolated from human subjects correlates negatively with the subject's body mass index (BMI; r = –0.52) (2). Because BMI correlates with obesity, which in turn is associated with increased numbers of hypertrophic adipocytes, these findings suggest a link between decreased anaerobic metabolism and de novo fat cell formation.

A second aim of this study was to characterize the effects of the culture substrate on adipocyte differentiation and metabolism. Our results pointed to a trade-off between differentiation and proliferation. The collagen gel cultures exhibited higher GPDH activity, leptin synthesis, and TG accumulation on a per cell basis but achieved lower cell densities. A similar result was reported by Yonemitsu et al. (49), who use a "ceiling culture" system to show that stromal cells isolated from subcutaneous fat of newborn rats differentiated to a significantly greater extent on type I collagen gel compared with TCP. One possible explanation for this phenomenon involves biophysical stimulation of mitogenic signaling pathways. Preadipocytes attached to the collagen gel deform from a flat to a spindle-like shape, which could in turn modify the MAPK pathway such that differentiation is promoted and proliferation suppressed (3). A modification of the MAPK pathway could also affect TG hydrolysis, and therefore net lipid accumulation, as the activity of hormone-sensitive lipase (HSL) is influenced by the phosphorylation states of extracellular signal-regulated kinases 1 and 2 (7). A different observation was described by Viravaidya and Shuler (45), who found that Matrigel (a mixture including laminin, type IV collagen, and heparan sulfate) enhanced both proliferation and differentiation of 3T3-F442A preadipocytes compared with type I collagen and poly-D-lysine. One possible reason for this discrepancy is that the Matrigel results were based on an evaluation of adipogenesis using the Oil Red O stain and did not involve a normalization of the accumulated lipid with respect to cell number.

One limitation of the current flux model is that it does not carry a full balance on ATP. As the "energy currency" of the cell, ATP provides a direct stoichiometric link between the reactions of energy metabolism and various cellular functions (5). Therefore, future work will incorporate additional ATP consumption reactions, e.g., biosynthesis of macromolecules, which should further improve the accuracy and predictive capacity of our model, especially with respect to the estimation of fluxes in and around oxidative phosphorylation. Future work can also include tracer experiments whereby the literature-based model assumptions, e.g., low beta-oxidation activity (47), can be directly tested.

In conclusion, we found that the combination of metabolite measurements and MFA provides an efficient approach for characterizing the metabolic profile of de novo adipocyte formation. Our results predicted a global distribution of carbon flux that was consistent with reported values obtained using isotopic tracers, which are considered to be accurate but experimentally much more costly. The flux distribution profiles obtained in this study clearly demonstrate that adipocytes exhibit distinct metabolic phenotypes depending on their stage of development, which could have significant implications for both basic and applied studies involving adipogenesis. For example, pharmacological agents designed to inhibit lipid loading will likely achieve differential results in fully mature and newly differentiated adipocytes, as these cells differ significantly in the relative engagements of lipogenic reactions. In this light, a better understanding of these metabolic phenotype differences could aid in the development of obesity intervention strategies (possibly involving multiple agents) that target a selected subpopulation of adipocytes in the WAT. Our findings, which reflect a comprehensive analysis of the changing flux distributions in differentiating and lipid loading adipocytes, pointed to the reactions of pyruvate transport and/or oxidation as key steps that contribute to the lipogenic phenotype of mature adipocytes. Results from the modularity analysis suggest that the malate cycle is another key pathway. Whether these reaction fluxes can be modulated to specifically control adipocyte differentiation or lipid loading remains to be explored. Prospectively, genes or enzymes for these reactions could provide new molecular targets to develop strategies for pharmacological or nutritional management of obesity.


   GRANTS
TOP
 ABSTRACT
 METHODS
 RESULTS
 DISCUSSION
 GRANTS
REFERENCES
 
This work was supported in part by National Institute of Diabetes and Digestive and Kidney Diseases Grant 1-R21-DK-67228-01 and a Tufts University Faculty Research Award to K. Lee.


   ACKNOWLEDGMENTS
 
We gratefully acknowledge Andrew Wood for help with the Oil Red O stains.


   FOOTNOTES
 

Address for reprint requests and other correspondence: K. Lee, Dept. of Chemical and Biological Engineering, Tufts Univ., 4 Colby St., Rm. 142, Medford, MA 02155 (e-mail: kyongbum.lee{at}tufts.edu)

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.


   REFERENCES
TOP
 ABSTRACT
 METHODS
 RESULTS
 DISCUSSION
 GRANTS
 REFERENCES
 

  1. Boone C, Mourot J, Gregoire F, Remacle C. The adipose conversion process: regulation by extracellular and intracellular factors. Reprod Nutr Dev 40: 325–358, 2000.[CrossRef][ISI][Medline]
  2. Bottcher H, Furst P. Decreased white fat cell thermogenesis in obese individuals. Int J Obes Relat Metab Disord 21: 439–444, 1997.[CrossRef][ISI][Medline]
  3. Boudreau NJ, Jones PL. Extracellular matrix and integrin signalling: the shape of things to come. Biochem J 339: 481–488, 1999.[CrossRef][ISI][Medline]
  4. Bray GA, Greenway FL. Current and potential drugs for treatment of obesity. Endocr Rev 20: 805–875, 1999.[Abstract/Free Full Text]
  5. Buttgereit F, Brand MD. A hierarchy of ATP-consuming processes in mammalian cells. Biochem J 312: 163–167, 1995.[ISI][Medline]
  6. Cabezas H, Raposo RR, Melendez-Hevia E. Activity and metabolic roles of the pentose phosphate cycle in several rat tissues. Mol Cell Biochem 201: 57–63, 1999.[CrossRef][ISI][Medline]
  7. Carmen GY, Victor SM. Signalling mechanisms regulating lipolysis. Cell Signal 18: 401–408, 2006.[CrossRef][ISI][Medline]
  8. Cohen SA, De Antonis KM. Applications of amino acid derivatization with 6-aminoquinolyl-N-hydroxysuccinimidyl carbamate. Analysis of feed grains, intravenous solutions and glycoproteins. J Chromatogr A 661: 25–34, 1994.[CrossRef][Medline]
  9. Cormen TH, Leiserson CE, Rivest RL, Stein C. Introduction to Algorithms. Cambridge, MA: The MIT Press, 2001.
  10. Dani C, Doglio A, Grimaldi P, Ailhaud G. Expression of the phosphoenolpyruvate carboxykinase gene and its insulin regulation during differentiation of preadipose cell lines. Biochem Biophys Res Commun 138: 468–475, 1986.[CrossRef][ISI][Medline]
  11. Faust IM, Johnson PR, Stern JS, Hirsch J. Diet-induced adipocyte number increase in adult rats: a new model of obesity. Am J Physiol Endocrinol Metab Gastrointest Physiol 235: E279–E286, 1978.[Abstract/Free Full Text]
  12. Fell DA, Small JR. Fat synthesis in adipose tissue. An examination of stoichiometric constraints. Biochem J 238: 781–786, 1986.[ISI][Medline]
  13. Flatt JP. Conversion of carbohydrate to fat in adipose tissue: an energy-yielding and, therefore, self-limiting process. J Lipid Res 11: 131–143, 1970.[Abstract]
  14. Flatt JP, Ball EG. Studies on the metabolism of adipose tissue. Xv. An evaluation of the major pathways of glucose catabolism as influenced by insulin and epinephrine. J Biol Chem 239: 675–685, 1964.[Free Full Text]
  15. Goodridge AG, Fisch JE, Glynias MJ. Regulation of the activity and synthesis of malic enzyme in 3T3-L1 cells. Arch Biochem Biophys 228: 54–63, 1984.[CrossRef][ISI][Medline]
  16. Hauner H, Skurk T, Wabitsch M. Cultures of human adipose precursor cells. In: Adipose Tissue Protocols, edited by Ailhaud G. Totowa, NJ: Humana, 2001, p. xiii, 334.
  17. Herrera MG, Philipps GR, Renold AE. Stimulation of metabolic activity of adipose tissue from fasted rats by prolonged incubation in vitro. I. Requirement for glucose and insulin. Biochim Biophys Acta 106: 221–233, 1965.[Medline]
  18. Hu CW, Utter MF, Patel MS. Induction of pyruvate dehydrogenase in 3T3-L1 cells during differentiation. J Biol Chem 258: 2315–2320, 1983.[Abstract/Free Full Text]
  19. Johnson PR, Stern JS, Greenwood MR, Hirsch J. Adipose tissue hyperplasia and hyperinsulinemia on Zucker obese female rats: a developmental study. Metabolism 27: 1941–1954, 1978.[ISI][Medline]
  20. Johnson PR, Zucker LM, Cruce JA, Hirsch J. Cellularity of adipose depots in the genetically obese Zucker rat. J Lipid Res 12: 706–714, 1971.[Abstract]
  21. Kanehisa M, Goto S, Hattori M, Aoki-Kinoshita KF, Itoh M, Kawashima S, Katayama T, Araki M, Hirakawa M. From genomics to chemical genomics: new developments in KEGG. Nucleic Acids Res 34: D354–D357, 2006.[Abstract/Free Full Text]
  22. Kather H, Rivera M, Brand K. Interrelationship and control of glucose metabolism and lipogenesis in isolated fat-cells. Control of pentose phosphate-cycle activity by cellular requirement for reduced nicotinamide adenine dinucleotide phosphate. Biochem J 128: 1097–1102, 1972.[ISI][Medline]
  23. Katz J, Landau BR, Bartsch GE. The pentose cycle, triose phosphate isomerization, and lipogenesis in rat adipose tissue. J Biol Chem 241: 727–740, 1966.[Abstract/Free Full Text]
  24. Katz J, Rognstad R. The metabolism of tritiated glucose by rat adipose tissue. J Biol Chem 241: 3600–3610, 1966.[Abstract/Free Full Text]
  25. Kim MS, Yoon CY, Jang PG, Park YJ, Shin CS, Park HS, Ryu JW, Pak YK, Park JY, Lee KU, Kim SY, Lee HK, Kim YB, Park KS. The mitogenic and antiapoptotic actions of ghrelin in 3T3-L1 adipocytes. Mol Endocrinol 18: 2291–2301, 2004.[Abstract/Free Full Text]
  26. Koh HJ, Lee SM, Son BG, Lee SH, Ryoo ZY, Chang KT, Park JW, Park DC, Song BJ, Veech RL, Song H, Huh TL. Cytosolic NADP+-dependent isocitrate dehydrogenase plays a key role in lipid metabolism. J Biol Chem 279: 39968–39974, 2004.[Abstract/Free Full Text]
  27. Lehninger AL, Nelson DL, Cox MM. Principles of Biochemistry. New York: WH Freeman, 2005.
  28. Loomis ME. An enzymatic fluorometric method for the determination of lactic acid in serum. J Lab Clin Med 57: 966–969, 1961.[ISI][Medline]
  29. Ma H, Zeng AP. Reconstruction of metabolic networks from genome data and analysis of their global structure for various organisms. Bioinformatics 19: 270–277, 2003.[Abstract/Free Full Text]
  30. Mackall JC, Student AK, Polakis SE, Lane MD. Induction of lipogenesis during differentiation in a "preadipocyte" cell line. J Biol Chem 251: 6462–6464, 1976.[Abstract/Free Full Text]
  31. Martin BR, Denton RM. The intracellular localization of enzymes in white-adipose-tissue fat-cells and permeability properties of fat-cell mitochondria. Transfer of acetyl units and reducing power between mitochondria and cytoplasm. Biochem J 117: 861–877, 1970.[ISI][Medline]
  32. Melnikova I, Wages D. Anti-obesity therapies. Nat Rev Drug Discov 5: 369–370, 2006.[CrossRef][ISI][Medline]
  33. Newman ME, Girvan M. Finding and evaluating community structure in networks. Phys Rev E Stat Nonlin Soft Matter Phys 69: 026113, 2004.[Medline]
  34. Nolan RP, Fenley AP, Lee K. Identification of distributed metabolic objectives in the hypermetabolic liver by flux and energy balance analysis. Metab Eng 8: 30–45, 2006.[CrossRef][ISI][Medline]
  35. Olsen C. An enzymatic fluorimetric micromethod for the determination of acetoacetate, -hydroxybutyrate, pyruvate and lactate. Clin Chim Acta 33: 293–300, 1971.[CrossRef][ISI][Medline]
  36. Rajala MW, Scherer PE. Minireview: The adipocyte–at the crossroads of energy homeostasis, inflammation, and atherosclerosis. Endocrinology 144: 3765–3773, 2003.[Abstract/Free Full Text]
  37. Schuster S, Hilgetag C, Woods JH, Fell DA. Reaction routes in biochemical reaction systems: algebraic properties, validated calculation procedure and example from nucleotide metabolism. J Math Biol 45: 153–181, 2002.[CrossRef][ISI][Medline]
  38. Song H, O'Connor KC, Papadopoulos KD, Jansen DA. Differentiation kinetics of in vitro 3T3-L1 preadipocyte cultures. Tissue Eng 8: 1071–1081, 2002.[CrossRef][ISI][Medline]
  39. Sottile V, Seuwen K. A high-capacity screen for adipogenic differentiation. Anal Biochem 293: 124–128, 2001.[CrossRef][ISI][Medline]
  40. Soukas A, Socci ND, Saatkamp BD, Novelli S, Friedman JM. Distinct transcriptional profiles of adipogenesis in vivo and in vitro. J Biol Chem 276: 34167–34174, 2001.[Abstract/Free Full Text]
  41. Spiegelman BM, Flier JS. Obesity and the regulation of energy balance. Cell 104: 531–543, 2001.[CrossRef][ISI][Medline]
  42. Stryer L. Biochemistry. New York: WH Freeman, 1995.
  43. Trinder P. Determination of blood glucose using an oxidase-peroxidase system with a non-carcinogenic chromogen. J Clin Pathol 22: 158–161, 1969.[Abstract/Free Full Text]
  44. Veech RL, Eggleston LV, Krebs HA. The redox state of free nicotinamide-adenine dinucleotide phosphate in the cytoplasm of rat liver. Biochem J 115: 609–619, 1969.[ISI][Medline]
  45. Viravaidya K, Shuler ML. The effect of various substrates on cell attachment and differentiation of 3T3-F442A preadipocytes. Biotechnol Bioeng 78: 454–458, 2002.[CrossRef][ISI][Medline]
  46. von Heimburg D, Hemmrich K, Zachariah S, Staiger H, Pallua N. Oxygen consumption in undifferentiated versus differentiated adipogenic mesenchymal precursor cells. Respir Physiol Neurobiol 146: 107–116, 2005.[CrossRef][ISI][Medline]
  47. Wang T, Zang Y, Ling W, Corkey BE, Guo W. Metabolic partitioning of endogenous fatty acid in adipocytes. Obesity Res 11: 880–887, 2003.[ISI][Medline]
  48. Wickelgren I. Obesity: how big a problem? Science 280: 1364–1367, 1998.[Free Full Text]
  49. Yonemitsu N, Bittinger F, Gerheld T, Sugihara H, Mittermayer C, Kirkpatrick CJ. Influence of extracellular matrix on the lipogenesis of cultured white fat cells. Acta Histochem Cytochem 31: 9–15, 1998.
  50. Yoon J, Blumer A, Lee K. An algorithm for modularity analysis of directed and weighted biological networks based on edge-betweenness centrality. Bioinformatics 22: 3106–3108, 2006.[Abstract/Free Full Text]
  51. Yoon J, Lee K. Metabolic flux profiling of reaction modules in liver drug transformation. Pac Symp Biocomput 12: 193–204, 2007.
  52. Young JW, Shrago E, Lardy HA. Metabolic control of enzymes involved in lipogenesis and gluconeogenesis. Biochemistry 3: 1687–1692, 1964.[CrossRef][Medline]




This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Supplemental Tables
Right arrow All Versions of this Article:
292/6/E1637    most recent
00670.2006v1
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 ISI Web of Science
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 ISI Web of Science (2)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Si, Y.
Right arrow Articles by Lee, K.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Si, Y.
Right arrow Articles by Lee, K.

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
Visit Other APS Journals Online
Copyright © 2007 by the American Physiological Society.