Metabolic and transcriptional patterns accompanying glutamine depletion and repletion in mouse hepatoma cells: a model for physiological regulatory networks
Matthew S. Wong1,
R. Michael Raab1,
Isidore Rigoutsos2,
Gregory N. Stephanopoulos1 and
Joanne K. Kelleher1,3
1 Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139
2 Bioinformatics and Pattern Discovery Group, IBM T. J. Watson Research Center, Yorktown Heights, New York 10598
3 Department of Physiology, George Washington University School of Medical and Health Sciences, Washington, District of Columbia 20037
 |
ABSTRACT
|
---|
An important objective in postgenomic biology is to link gene expression to function by developing physiological networks that include data from the genomic and functional levels. Here, we develop a model for the analysis of time-dependent changes in metabolites, fluxes, and gene expression in a hepatic model system. The experimental framework chosen was modulation of extracellular glutamine in confluent cultures of mouse Hepa1-6 cells. The importance of glutamine has been demonstrated previously in mammalian cell culture by precipitating metabolic shifts with glutamine depletion and repletion. Our protocol removed glutamine from the medium for 24 h and returned it for a second 24 h. Flux assays of glycolysis, the tricarboxylic acid (TCA) cycle, and lipogenesis were used at specified intervals. All of these fluxes declined in the absence of glutamine and were restored when glutamine was repleted. Isotopomer spectral analysis identified glucose and glutamine as equal sources of lipogenic carbon. Metabolite measurements of organic acids and amino acids indicated that most metabolites changed in parallel with the fluxes. Experiments with actinomycin D indicated that de novo mRNA synthesis was required for observed flux changes during the depletion/repletion of glutamine. Analysis of gene expression data from DNA microarrays revealed that many more genes were anticorrelated with the glycolytic flux and glutamine level than were correlated with these indicators. In conclusion, this model may be useful as a prototype physiological regulatory network where gene expression profiles are analyzed in concert with changes in cell function.
correlation; flux; microarray; carbon-13; carbon-14
 |
INTRODUCTION
|
---|
ADVANCES IN PHYSIOLOGICAL GENOMICS require connecting changes in gene sequence or gene transcription to function. A first approach to the global investigation of these relationships is to perturb the physiology of an organism and observe time-dependent changes in gene expression. Transcriptional profiling studies of this type have begun to identify groups of coexpressed and interrelated genes constituting transcriptional regulatory networks (3, 13, 24). To complement this transcriptional or horizontal level of organization, a second vertical category of networks may be considered that investigates responses to perturbations in physiology over time, but includes data from the transcriptional level and the level of cell function. This second category, which we term "physiological regulatory networks," provides a structure for directly incorporating changes in physiology in the development of regulatory networks. As a prototype for mammalian cell physiological regulatory networks, we have examined the time-dependent response of confluent cultured mouse hepatoma cells to changes in the glutamine concentration of the medium. Using this model, we report quantitative changes in gene expression, metabolic fluxes, and metabolite levels for pathways directly involving glutamine and its metabolites.
Most mammalian cells in culture require super-physiological levels of glutamine for optimal growth. High rates of glutamine consumption are commonly observed in tumor cell lines, hybridomas, and other rapidly proliferating cells (19). The importance of glutamine is further demonstrated by the observation that changes in the extracellular glutamine concentration cause metabolic shifts in mammalian cell culture (17, 18). For example, in continuous hybridoma cultures, a step change from approximately 0 to 0.9 mM extracellular glutamine (with excess glucose present) produced a marked increase in the consumption of glucose, glutamine, and oxygen and in the production of ammonium, alanine, and lactate (18). This observation that glutamine is required for high rates of glycolysis has been observed in a variety of mammalian cells. Rat lymphocytes increased the consumption of glucose when glutamine was in the medium (1). C6 rat glioma cells increased glucose consumption by 60% when transferred from glutamine-free medium to 4 mM glutamine (20). When both glucose and glutamine are available, glutamine provides a significant fraction of the cellular energy requirements, calculated as 40% for normal diploid fibroblast cells (27) and Chinese hamster cells (8) and as high as 70% for HeLa cells (22). This capacity for elevated glutamine consumption may be a consequence of a distinct high-capacity glutamine transport system, which has been documented for human hepatoma cells (2). Once inside the cell, glutamine carbon has several fates. It may enter the tricarboxylic acid (TCA) cycle at oxaloacetate, but, for each molecule entering the cycle, one four- or five-carbon moiety must exit to maintain steady state. The flux of glutamine carbon to purine biosynthesis is an example of one such biosynthetic exit path. A second pathway is exit from the TCA cycle via malic enzyme to pyruvate, which can either be converted to lactate or acetyl-CoA. Both of these paths allow the cell to maintain steady-state TCA cycle metabolite levels, and the latter pathway is required for complete oxidation of glutamine. Finally, glutamine can be converted to citrate via reductive carboxylation where isocitrate dehydrogenase carries flux in reverse of the TCA cycle direction. This last pathway contributes significantly to de novo lipogenesis in rat hepatoma (10) and to gluconeogenesis in normal perfused rat liver (7).
To develop a prototype for physiological regulatory networks, we chose Hepa1-6 cells, a mouse hepatoma cell line. This cell line arose from a C57/L mouse and retains many liver-specific phenotypes, including the secretion of several serum proteins (6). However, it does not store glycogen and has a low activity of glucose-6-phosphatase (25), suggesting an absence of glucose production. We designed studies to simultaneously probe the changes in metabolic function and gene expression following removal of glutamine from the medium for 24 h followed by 24 h of 4 mM glutamine. This glutamine depletion/repletion protocol provided the basis for investigating both metabolic physiology and gene expression under well-controlled conditions.
 |
MATERIALS AND METHODS
|
---|
Cell line and culture conditions.
The mouse hepatoma line Hepa1-6 was obtained from the ATCC (Manassas, VA) and maintained in DMEM containing 25 mM glucose and 4 mM glutamine and supplemented by 10% (vol/vol) fetal bovine serum (FBS) (Sigma, St. Louis, MO) and 1% (vol/vol) penicillin-streptomycin (GIBCO). Confluent cells grown in this medium were used to investigate the effects of removing glutamine from the medium for 24 h followed by glutamine repletion to 4 mM and incubation for 24 h. The transcriptional and metabolic activities of cells undergoing this 48-h glutamine depletion/repletion protocol were compared with control cells maintained in 4 mM glutamine for the entire 48 h. To reduce the unknown amount of hormones and endogenous lipids in the medium during the 48-h glutamine depletion/repletion, FBS was replaced with 10% controlled processed serum replacement 3 (CPSR3) (Sigma) in both the control and experimental samples. Metabolites and actinomycin D were obtained from Sigma. Stable isotopes were purchased from Isotec (Miamisburg, OH), and radioisotopes were from American Radiolabeled Chemicals (St. Louis, MO).
Isotopic metabolic flux assays.
Isotopic metabolic flux assays were conducted at specified intervals across the 48-h glutamine depletion/repletion protocol. Fluxes were monitored as the release of labeled atoms from labeled glucose as it was metabolized, either 3H2O from 3H-labeled glucose or 14CO2 from 14C-labeled glucose. The forward flux of hexose-phosphate isomerase was estimated from release of 3H2O from [2-3H]glucose while the flux through the normally irreversible glycolytic step, phosphofructokinase, followed by triose-phosphate isomerase, was estimated from release of 3H2O from [3-3H]glucose. The flux through pyruvate dehydrogenase (PDH) was monitored by 14CO2 production from [3,4-14C]glucose. The oxidation of glucose in the TCA cycle was monitored by 14CO2 production from [6-14C]glucose. It should be noted that flux estimates using these labeled precursors supplied exogenously do not include flux of preexisting intracellular metabolites through the same reactions. Modifications of the traditional versions of these assays (5) were developed to allow the assays to be performed in a higher throughput 96-well format and to limit the incubation time for the assays to 1 h. The absolute flux values of these assays varied between experiments but were self-consistent within each experiment. The absolute flux values may not compared between experiments or between assays.
3H2O and 14CO2 flux indicator assays.
At specified intervals, the medium prescribed by the glutamine depletion/repletion protocol was removed from designated wells and replaced with isotope flux assay medium consisting of DMEM modified to contain 0.6 mM glucose and 1 mM glutamine. In addition, each well contained 0.10.2 µCi of one 3H- or 14C-labeled glucose isotope at a specific activity of 0.8 to 1.6 Ci/mol. To quantify 3H2O production from 3H-labeled glucose, the traditional assay using a borate resin to trap the labeled glycolytic compounds (8) was modified for use in a 96-well format. After a 1-h incubation with isotopes in the 96-well plate, the medium was removed and dispensed onto 0.4 ml of Dowex 1 x 400 borate resin (Sigma) in Spin-X centrifuge filter tubes (Costar). The tubes were agitated, incubated for 30 min, and centrifuged for 5 min at 5,000 rpm. The filtrate, which contained the 3H2O, was quantified by liquid scintillation counting.
To allow for 14CO2 collection, breakaway 96-well clusters (Costar) were used for the [14C]glucose studies. To measure the 14CO2 produced, individual wells were suspended in a 7-ml scintillation vial and gassed with 95% O2-5% CO2 immediately following addition of the [14C]glucose the wells. The vial was then closed with a rubber septum cap and incubated for 1 h at 37°C. After the 1-h incubation, a syringe deposited
20 µl of 30% perchloric acid through the septum into the well to terminate metabolic activity. A second syringe delivered 0.2 ml of hyamine hydroxide (ICN Pharmaceuticals, Costa Mesa, CA) to the bottom of the scintillation vial to absorb the CO2. The trapped 14CO2 was quantified by liquid scintillation counting.
13C metabolite pool measurements.
Metabolite pools were measured using a gas chromatography-mass spectroscopy (GC-MS) method that included the addition of heavy 13C-labeled internal standards at the time of cell lysis. The internal standards were heavy 13C-labeled versions of each compound to be quantified. The area of each internal standard was compared with its naturally labeled counterpart. At the time of the assay, metabolism was terminated with addition of 3 ml of 2% perchloric acid. The internal standards were added. Intracellular anions and cations were isolated by ion-exchange chromatography (26). These fractions were dried and derivatized with N-methyl-N-(tert-butyldimethylsilyl)trifluoroacetamide + 1% tert-butyldimethylchlorosilane (MTBSTFA + 1% TBDMCS) (Pierce, Rockford, IL). Seventy microliters of MTBSTFA + 1% TBDMCS and 50 µl of dimethylformamide (DMF) (Pierce) were added to the dried sample. The sample was then capped, vortexed, and heated at 70°C for 30 min. The samples were analyzed with a Varian model Saturn 2000 GC-MS in electron ionization mode. One microliter of each sample was injected onto a 30-m CP-SIL 8 CB low-bleed column (Varian, Walnut Creek, CA). The GC oven temperature was held at 140°C for 2 min after sample injection, before increasing it at a rate of 3°C/min to a final temperature of 250°C. This final temperature was held for 6.33 min for a total run time of 45 min.
3H2O lipogenesis indicator assay.
Total lipid synthesis was estimated by the 3H2O incorporation method (16) using a protocol designed for cultured cells (4) and modified for a 1-h incubation. Cells in six-well plates were incubated as prescribed by the glutamine depletion/repletion protocol. At 4-h intervals, the medium in individual wells was replaced with isotope flux assay medium containing 23 mCi 3H2O in 1.5 ml. The assay was terminated with 2% perchloric acid. Saponifiable lipids were extracted and quantified by liquid scintillation counting.
Isotopomer spectral analysis.
Isotopomer spectral analysis (ISA) provides a method for quantifying the sources of carbon for de novo lipogenesis using stable isotope incorporation into products and analysis by nonlinear regression (14). Unlike the 1-h radioisotope assays, ISA was performed over a 24-h period to estimate the overall effect of glutamine depletion on lipogenesis. Cells were preincubated for 24 h in media with 4 mM or 0 mM glutamine and then transferred to media containing one 13C-labeled substrate for an additional 24 h. Control cells were evaluated in media containing 25 mM glucose and 4 mM glutamine with either [U-13C]glucose (25 mM) or [U-13C]glutamine (4 mM). This allowed estimation of the roles of both glucose and glutamine as lipogenic precursors. Glutamine-depleted cells were evaluated in media containing [U-13C]glucose (25 mM). Following the 24-h incubation, the experiment was terminated with perchloric acid, and cells were processed for ISA. Total lipids were extracted into 3:2 hexane:isopropanol (9), containing heptadecanoic acid (20 µg per well) as an internal standard for quantification of fatty acids (FA). After solvent evaporation, the residue was treated with BF3/MeOH (14%) to derivatize total FA as methyl esters (11). Methyl esters were dissolved in DMF before analysis by GC-MS. Mass isotopomer analysis focused on palmitate to determine the ISA parameters D, the fractional contribution of the labeled substrate to the lipogenic acetyl-CoA, and g(24 h), the fraction of newly synthesized FA present after 24 h.
DNA microarray experimental protocol.
Cells were seeded onto T25 flasks for the DNA microarrays and onto 96-well breakaway clusters for the hexose isomerase flux indicator assay. Data were taken at the following time points: 0.5, 12, 18, 24, 36, 39, and 48 h. At each time point, the cells were assayed for hexose isomerase flux and for mRNA expression. The protocols for the RNA extraction, labeling of probes, hybridization, and printing of arrays are listed at the following web site: http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi. The accession numbers are as follows: GSE404, GPL285, and GSM5974 through GSM5993. The resulting data were downloaded and formatted in Excel (Microsoft, Redmond, WA) and then analyzed using Matlab (The MathWorks, Natick, MA). The DNA microarrays were performed with duplicate flasks of cells for each condition and each time point. Any unacceptable data points, flagged by scanning software, were eliminated. The data in the duplicate arrays were then merged and averaged to create a union data set. Genes were retained in the union set only if data were available for each of the seven time points. To capture genes with significant changes in gene expression, the data were filtered to eliminate genes that did not have at least one time point with a log2 ratio greater than 0.6 or less than -0.6. These values were selected to insure 95% confidence for significant expression changes. Previous validation studies (not shown) demonstrated that the median coefficient of variation across duplicate arrays was 10.2%.
 |
RESULTS
|
---|
Metabolic alterations during glutamine depletion and repletion.
The effect of glutamine depletion/repletion on the glycolytic flux was examined using 3H2O release from [2-3H]glucose and [3-3H]glucose as described in MATERIALS AND METHODS (Fig. 1). The glycolytic flux at time 0 was 0.30 ± 0.03 nmol/h per 104 cells and did not differ for the two 3H tracers. This indicates that the flux through triose-phosphate isomerase did not differ from that through hexose phosphate isomerase. This agreement is expected in view of the high glycolytic rate in the absence of significant glucose-6-phosphatase activity reported in Hepa1-6 cells (25). Removal of glutamine from the medium led to a gradual decline in the glycolytic rate to values
50% of the control. Restoring glutamine to the medium returned the rates to near the starting values after a delay of
12 h (Fig. 1), demonstrating a reversible change in flux upon glutamine depletion and repletion.

View larger version (17K):
[in this window]
[in a new window]
|
Fig. 1. Effect of glutamine depletion/repletion on the glycolytic flux indicator ratios. The average experimental glycolytic flux indicator was normalized by a 12-h averaged control glycolytic flux indicator to yield the ratio. The hexose isomerase indicator assay used [2-3H]glucose. The triose-phosphate isomerase indicator assay used [3-3H]glucose. * Time points where the average experimental value was significantly different from the average control value (n = 3, P < 0.05).
|
|
To monitor the fate of glucose further into the oxidative pathway, 14CO2 production from 14C-labeled glucose tracers was examined (Figs. 2 and 3). The flux through PDH estimated with [3,4-14C]glucose at time 0 was 0.10 ± 0.01 nmol/h per 104 cells. The flux through PDH declined by 70% in the absence of glutamine. For a more direct estimate of the effect of glutamine on TCA cycle flux, 14CO2 production from [6-14C]glucose was examined at 0, 24, and 48 h of the protocol (Fig. 3). The oxidation of glucose estimated with [6-14C]glucose at time 0 was 0.07 ± 0.01 nmol/h per 104 cells. The TCA cycle activity measured by this indicator declined 70% during glutamine depletion at 24 h and returned to control values by 48 h (Fig. 3). Taken together, these flux indicator assays demonstrate a reversible decline in glycolytic and TCA cycle fluxes induced by glutamine depletion and repletion.

View larger version (13K):
[in this window]
[in a new window]
|
Fig. 2. Effect of glutamine depletion/repletion on glucose oxidation indicator ratios. The average experimental glucose oxidation indicator was normalized by a 12-h averaged control glucose oxidation indicator to yield the ratio. The pyruvate dehydrogenase indicator assay used [3,4-14C]glucose. * Time points where the average experimental value was significantly different from the average control value (n = 3, P < 0.05).
|
|

View larger version (13K):
[in this window]
[in a new window]
|
Fig. 3. Effect of glutamine depletion/repletion on TCA cycle flux indicator ratios. The average experimental TCA cycle flux indicator was normalized by the control TCA cycle flux indicator to yield the ratio. The TCA cycle flux indicator assay used [6-14C]glucose. * Time points where the average experimental value was significantly different from the average control value (n = 6, P < 0.05).
|
|
In addition to lactate and the TCA cycle, a significant fate of glucose and glutamine carbon is polymerization in the lipogenic pathway. The effect of glutamine depletion/repletion on the total rate of lipogenesis was investigated by measuring the rate of incorporation of 3H2O into lipid-soluble compounds. The lipogenic flux at time 0 was 142 ± 20 pmol FA/h per 106 cells. Following the trend of the other flux indicators, the rate of lipogenesis fell by 80% when glutamine was removed from the medium and slowly increased when it was replenished (Fig. 4). To quantify the sources of carbon for de novo lipogenesis, ISA was employed as described in the MATERIALS AND METHODS. ISA estimated the fractional contribution of glucose and glutamine to the lipogenic acetyl-CoA pool (D) and the fraction of newly synthesized FA in esterified lipid after a 24-h incubation, i.e., g(24 h) (Table 1). In control medium, both glucose and glutamine were major contributors to the lipogenic acetyl-CoA pool. Together, they accounted for 70% of the lipogenic carbon. This indicates that 30% of the lipogenic carbons were either from endogenous compounds or from compounds in the medium other than glucose and glutamine. ISA estimated the fractional synthesis of new palmitate in 24 h as
26% of the total cellular esterified pool in control conditions. These results contrasted with the findings in glutamine-depleted medium, where glucose accounted for only 19% of lipogenic carbon and the fractional synthesis of new palmitate was only 8.5% of the total. These findings indicate that glutamine depletion significantly reduced the flux of glucose carbon into lipogenic pathways. The total esterified palmitate for the two conditions following the 24-h tracer study was similar (Table 1), suggesting that esterified palmitate turnover was greater in the presence of glutamine.

View larger version (11K):
[in this window]
[in a new window]
|
Fig. 4. Effect of glutamine depletion/repletion of the rate of lipogenesis determined by 3H2O incorporation. The average experimental lipogenic flux was normalized by a 12-h averaged control lipogenic flux to yield the ratio. * Time points where the average experimental value was significantly different from the average control value (n = 3, P < 0.05).
|
|
To examine whether changes in intracellular metabolites played a role in the observed changes in glycolytic and TCA cycle rate, levels of key TCA cycle intermediates and related amino acids were monitored during glutamine depletion/repletion (Fig. 5). These data indicate that, with the exception of succinate, which did not change over the course of treatment, the metabolite concentrations declined rapidly when glutamine was removed from the medium and returned to basal levels when glutamine was repleted. This finding supports the anaplerotic role of glutamine carbon in replenishing the TCA cycle and the role of glutamate transaminases in maintaining amino acid levels. In summary, the analysis of key metabolite levels indicated that observed changes are consistent with flux changes, and thus metabolite changes could play a key role in altering flux in this model.

View larger version (23K):
[in this window]
[in a new window]
|
Fig. 5. Effect of glutamine depletion/repletion on the metabolite pool sizes. The experimental metabolite pool size was normalized by the control metabolite pool size at each time point to yield the ratio.
|
|
Role of gene expression in flux changes.
Changes in metabolic fluxes may be the consequence of changes occurring at the transcriptional level. To evaluate the significance of new mRNA synthesis, the flux measurements were repeated in the presence of actinomycin D. Cells were treated with actinomycin D for 24 h beginning either at time 0 when glutamine was depleted from the medium or at t = 24 h when glutamine was repleted. One-hour flux assays were conducted at t = 24 or 48 h, following the 24-h actinomycin D exposure. Assays measured glycolytic flux (3H2O production from [2-3H]glucose), PDH flux (14CO2 production from [3,4-14C]glucose), TCA cycle flux (14CO2 production from [6-14C]glucose), and lipogenesis (3H2O incorporation into lipids). Each of these assays indicated that the observed changes in flux produced by glutamine depletion and repletion were altered in the presence of actinomycin D. Three specific patterns are illustrated in Figs. 68. In the presence of actinomycin D, the glycolytic flux did not decline during glutamine depletion (Fig. 6A) and did not recover completely during glutamine repletion (Fig. 6B). This indicates that de novo mRNA synthesis was necessary to allow the glycolytic rate to fall when glutamine was depleted (Fig. 6A) and to allow it to return to normal values when glutamine is repleted (Fig. 6B). However, the maintenance of glycolytic flux in control cells over a 24 h period did not require de novo mRNA synthesis. The glycolytic flux indicator value at time 0 was 0.47 ± 0.04 nmol/h per 104 cells. A similar, but less pronounced pattern was seen in the TCA cycle flux (Fig. 7). The TCA cycle flux indicator value at time 0 was 0.07 ± 0.01 nmol/h per 104 cells. A different pattern was found when the lipogenic flux was monitored with 3H2O incorporation (Fig. 8). Here, de novo mRNA synthesis was required to maintain the normal lipogenic flux over 24 h in control medium, but mRNA synthesis was not required for decreased lipogenesis produced by glutamine depletions (Fig. 8A). De novo mRNA synthesis was also required for cells to respond to glutamine repletion (Fig. 8B). The lipogenic flux indicator value at time 0 was 99.5 ± 7.9 pmol FA/h per 106 cells.

View larger version (19K):
[in this window]
[in a new window]
|
Fig. 6. Effect of actinomycin D on glycolytic flux estimated using [2-3H]glucose over the course of glutamine depletion/repletion. Actinomycin D (Act) was added to control and experimental groups in the first 24 h (A) or in the second 24 h (B) as indicated by a horizontal black bar. Arrow denotes time of flux assay. In comparisons between cells in the absence (open bars) or presence (solid bars) of actinomycin D for each glutamine treatment (n = 6, ±SD), an asterisk denotes a significant difference (P < 0.05).
|
|

View larger version (18K):
[in this window]
[in a new window]
|
Fig. 8. Effect of actinomycin D on lipogenic flux estimated using 3H2O over the course of glutamine depletion/repletion. Actinomycin D was added to control and experimental groups in the first 24 h (A) or in the second 24 h (B) as indicated by a horizontal black bar. Arrow denotes time of flux assay. In comparisons between cells in the absence (open bars) or presence (solid bars) of actinomycin D for each glutamine treatment (n = 6, ±SD), an asterisk denotes a significant difference (P < 0.05).
|
|

View larger version (20K):
[in this window]
[in a new window]
|
Fig. 7. Effect of actinomycin D on TCA cycle flux estimated using [6-14C]glucose over the course of glutamine depletion/repletion. Actinomycin D was added to control and experimental groups in the first 24 h (A) or in the second 24 h (B) as indicated by a horizontal black bar. Arrow denotes time of flux assay. In comparisons between cells in the absence (open bars) or presence (solid bars) of actinomycin D for each glutamine treatment (n = 6, ±SD), an asterisk denotes a significant difference (P < 0.05).
|
|
Microarray analysis of gene expression.
DNA microarray studies were conducted at specified intervals across the glutamine depletion/repletion protocol, as described in the MATERIALS AND METHODS. The gene expression data for duplicate microarrays were merged and averaged to create a union data set. Genes were retained in the union set only if data were available for each of the seven time points. Analysis of this data indicated that 3,185 of the 17,280 genes were expressed at every time point during the glutamine depletion/repletion protocol. To capture genes with significant changes in gene expression, the data were filtered to eliminate genes that did not have at least one time point with a log2 ratio greater than 0.6 or less than -0.6. Of the 3,185 genes, 950 genes were considered unchanged at every time point and condition by this criterion. The remaining 2,235 genes were examined to determine whether genes associated with the observed flux changes were altered significantly and to determine whether any of the 2,235 genes were highly correlated with the glycolytic flux indicator or with the glutamine level in the medium. To investigate the behavior of genes in the pathways affected by glutamine depletion/repletion, expression values for treated cells were compared with controls at the end of glutamine depletion (24 h) and after 12 h of glutamine repletion (36 h) (Table 2). Lipid synthesis increased upon glutamine repletion, and a gene catalyzing a highly regulated step in cholesterol synthesis, HMG-CoA reductase, was found to increase upon glutamine repletion. However, fatty acid synthase, whose expression is often found to correlate with FA synthesis, was not found to change in this study. Genes catalyzing early steps of FA oxidation, acyl-CoA synthetase and acyl-CoA dehydrogenase, increased during glutamine depletion and decreased during the glutamine repletion. These findings are consistent with the fact that lipid oxidation is suppressed when lipogenesis is elevated, due to the actions of malonyl-CoA. Glycolysis increased upon glutamine repletion. Several glycolytic genes increased upon glutamine repletion, including 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase 2, phosphofructokinase, and glucose phosphate isomerase 1 complex. It should be noted that many metabolic genes involved in the pathways affected by glutamine did not change significantly. Among the 950 unchanged genes were citrate synthase, PDH-ß, pyruvate kinase, lactate dehydrogenase 1, triose-phosphate isomerase, and ATP-citrate lyase (data not shown). Acetyl-CoA carboxylase (15) and PEPCK (21) have previously been found to be affected by glutamine oscillations but did not qualify for our 3,185-gene data set. Taken together, our results indicate that the dramatic flux changes associated with glutamine depletion and repletion are accompanied by large changes in only a few of the many enzymes catalyzing the reactions in these pathways. Finally, Table 2 lists the increased expression of ß-actin upon glutamine repletion, consistent with the findings of Husson et al. (12). In addition to characterizing the behavior of genes known to be involved in the pathways affected by glutamine, we used correlational analysis to identify genes whose expression patterns correlated with the observed fluxes. The correlational analysis revealed a larger number of genes that were anticorrelated with glutamine and fluxes than those that were correlated. For example, correlational analysis of the glycolysis flux detected nine correlated genes with a correlation coefficient
0.90 (Fig. 9A). However, 22 anticorrelated genes were detected with a correlation coefficient less than or equal to -0.90 (Fig. 9B). Analysis of gene expression data with the autoscaled glutamine input signal found 16 anticorrelated genes with a correlation coefficient less than -0.90 (Fig. 10). Yet, no correlated genes were found with a correlation coefficient greater than 0.90. Finally, the gene expression data were analyzed by a pattern discovery algorithm, Teiresias (23). The data set of log2 ratios at time points was converted into a binary data set of positive or negative derivatives between time points. Teiresias was then used to discover patterns within this binary data set. Twelve genes were found to have the pattern of three positive derivatives followed by three negative derivatives (Fig. 11). Using the same criteria, Teiresias did not detect any genes exhibiting the expression pattern of three negative derivatives followed by three positive derivatives.

View larger version (35K):
[in this window]
[in a new window]
|
Fig. 9. A: gene expression profiles correlated to hexose isomerase flux indicator with a correlation coefficient 0.90. B: gene expression profiles anticorrelated to hexose isomerase flux indicator with a correlation coefficient less than or equal to - 0.90.
|
|

View larger version (31K):
[in this window]
[in a new window]
|
Fig. 10. Gene expression profiles anticorrelated to autoscaled glutamine concentration with a correlation coefficient less than or equal to - 0.90.
|
|

View larger version (34K):
[in this window]
[in a new window]
|
Fig. 11. Gene expression profiles discovered by Teiresias to have a pattern of 3 positive derivatives followed by 3 negative derivatives.
|
|
 |
DISCUSSION
|
---|
With the completion of the mouse genome sequence and the development of protocols for generating transgenic animals, mouse models are providing important insights into human diseases. The glutamine depletion/repletion protocol described here demonstrated that changes in fluxes could be evaluated in a mouse hepatic cell model at short time intervals over a 48-h period. This was facilitated by the development of 1-h flux assays specifically for this project. The observed flux changes could then be combined with data for metabolite levels and gene expression, providing a prototype for simultaneous monitoring of fluxes, metabolite levels, and gene expression. Although previous studies with mouse hepatoma cell lines have recorded distinct enzyme expression patterns characteristic of these cells (6), quantitative analysis of fluxes comparable to those in human and rat hepatoma cell lines have been lacking. The present study demonstrated that mouse Hepa1-6 cells share metabolic flux characteristics with other transformed cell lines with regard to glutamine metabolism. Consistent with previous findings (1, 20), glutamine is required for high rates of glycolytic flux (Fig. 1). Glutamine is also a major source of carbon for de novo lipogenesis (Table 1), as found with rat hepatoma cells (10). Thus the Hepa1-6 mouse cell line is well-suited for the investigation of physiological regulatory networks that integrate gene expression and functional data.
The glutamine depletion/repletion protocol provided a mechanism for examining changes in metabolic flux in terms of two key sites of controlling flux: control via changes in the levels of substrates and other metabolites, and control via changes in the level of mRNA mediated by transcription. Flux measurements following incubation in actinomycin D provided a tool for examining the necessity of transcriptional changes in this model. The glycolytic flux response observed in response to actinomycin D demonstrated that, in the absence of a change in glutamine, de novo mRNA synthesis was not critical to maintaining flux (Fig. 6). This suggests that the mRNA and/or its protein products required for maintenance of the glycolytic flux are relatively stable, a pattern characteristic of a pathway that is not regulated at the transcriptional level in the short term. The TCA cycle may also be an example of a pathway that is not transcriptionally regulated in the short term. Citrate synthase and PDH-ß were among the 950 genes with no significant gene expression changes during glutamine depletion/repletion. In place of transcriptional changes, alterations in fluxes in these pathways may be the result of control at the enzyme activity or metabolite level, a form of regulation that allows rapid response to changing conditions. The finding that the concentrations of five of the six metabolites correlated with the flux changes (Fig. 5) provides a mechanism for changes in flux as a result of changes in substrate concentration. Although maintenance of glycolytic flux did not require de novo mRNA synthesis, the requirement for mRNA synthesis to effect the changes in flux during glutamine depletion/repletion clearly indicates a role for de novo mRNA synthesis in regulating these fluxes (Fig. 6). The effect of actinomycin D on the lipogenic flux demonstrated that de novo mRNA synthesis was critical to maintaining flux. The lipogenic mRNA and/or its protein products are less stable than glycolytic mRNA and may exert some control over the lipogenic flux at the transcriptional level in the short term. Thus, in developing a more complete description of flux control in this model, the quantitative importance both of metabolite changes and gene expression will be required. The use of actinomycin D in conjunction with the glutamine depletion/repletion protocol provides a model for the analysis of the time course of transcriptional changes modulating metabolic flux.
Changes in gene expression monitored with DNA microarrays indicated activation of gene expression accompanied the decline in metabolic fluxes observed upon glutamine depletion (Figs. 9 and 10). This finding brings into focus the fact that increased transcription of some genes was required to allow cells to respond to the new metabolic conditions created by removing glutamine from the medium. Activation of gene expression in the absence of glutamine was also supported by the finding that actinomycin D prevented, at least partially, the expected decline in glycolytic flux during glutamine depletion as discussed above. Most of the genes found to be activated or anticorrelated with glutamine levels or flux are not known to be directly connected to intermediary metabolism. They were retained following a filter that required a substantial change in gene expression and eliminated most genes due to either to poor signal or small changes in expression. The Teiresias algorithm provided another method other than correlation to identify genes of interest. Of the 12 genes identified with the correct 3-up/3-down pattern, 8 were also found on the list of genes from the glycolytic flux anticorrelational analysis. In other words, eight of the genes identified by Teiresias also had correlation coefficients to the glycolytic flux less than or equal to -0.90. The other four genes had correlation coefficients between -0.71 and -0.89. Teiresias sought out genes that had a desirable pattern, but were not necessarily highly correlated with the flux signal. Although the role of the genes detected here in modulating flux has not been resolved, the ability of this model to examine the relationship between genes and fluxes may be an important tool for future studies. Another finding of note in the analysis of microarray data was that most relevant metabolic genes did not display significant expression changes. Thus this study demonstrates the importance of a physiological approach combining metabolite data and gene expression data to understand regulatory networks controlling flux. We propose the glutamine depletion/repletion model as a prototype for developing physiological regulatory models in integrative systems biology.
 |
ACKNOWLEDGMENTS
|
---|
GRANTS
This work was supported by National Institutes of Health Bioengineering Research Partnership Grant DK-58533.
 |
FOOTNOTES
|
---|
Article published online before print. See web site for date of publication (http://physiolgenomics.physiology.org).
Address for reprint requests and other correspondence: J. K. Kelleher, Dept. of Chemical Engineering, Rm. 66-401, Massachusetts Institute of Technology, 77 Massachusetts Ave., Cambridge, MA 02139 (E-mail: jkk{at}mit.edu).
10.1152/physiolgenomics.00088.2003.
 |
REFERENCES
|
---|
- Ardawi MS and Newsholme EA. Glutamine metabolism in lymphocytes of the rat. Biochem J 212: 835842, 1983.[ISI][Medline]
- Bode BP and Souba WW. Glutamine transport and human hepatocellular transformation. JPEN J Parenter Enteral Nutr 23: S33S37, 1999.
- Bolouri H and Davidson EH. Modeling transcriptional regulatory networks. Bioessays 24: 11181129, 2002.[CrossRef][ISI][Medline]
- Brass EP and Vetter WH. Interleukin-6, but not tumour necrosis factor-alpha, increases lipogenesis in rat hepatocyte primary cultures. Biochem J 301: 193197, 1994.[ISI][Medline]
- Broderick TL, Quinney HA, and Lopaschuk GD. Carnitine stimulation of glucose oxidation in the fatty acid perfused isolated working rat heart. J Biol Chem 267: 37583763, 1992.[Abstract/Free Full Text]
- Darlington GJ, Bernhard HP, Miller RA, and Ruddle FH. Expression of liver phenotypes in cultured mouse hepatoma cells. J Natl Cancer Inst 64: 809819, 1980.[ISI][Medline]
- Des Rosiers C, Di Donato L, Comte B, Laplante A, Marcoux C, David F, Fernandez CA, and Brunengraber H. Isotopomer analysis of citric acid cycle and gluconeogenesis in rat liver. Reversibility of isocitrate dehydrogenase and involvement of ATP-citrate lyase in gluconeogenesis. J Biol Chem 270: 1002710036, 1995.[Abstract/Free Full Text]
- Donnelly M and Scheffler IE. Energy metabolism in respiration-deficient and wild type Chinese hamster fibroblasts in culture. J Cell Physiol 89: 3951, 1976.[ISI][Medline]
- Hara A and Radin NS. Lipid extraction of tissues with a low-toxicity solvent. Anal Biochem 90: 420426, 1978.[ISI][Medline]
- Holleran AL, Briscoe DA, Fiskum G, and Kelleher JK. Glutamine metabolism in AS-30D hepatoma cells. Evidence for its conversion into lipids via reductive carboxylation. Mol Cell Biochem 152: 95101, 1995.[ISI][Medline]
- Holleran AL, Lindenthal B, Aldaghlas TA, and Kelleher JK. Effect of tamoxifen on cholesterol synthesis in HepG2 cells and cultured rat hepatocytes. Metabolism 47: 15041513, 1998.[ISI][Medline]
- Husson A, Quillard M, Fairand A, Chedeville A, and Lavoinne A. Hypoosmolarity and glutamine increased the beta-actin gene transcription in isolated rat hepatocytes. FEBS Lett 394: 353355, 1996.[CrossRef][ISI][Medline]
- Ideker T, Thorsson V, Ranish JA, Christmas R, Buhler J, Eng JK, Bumgarner R, Goodlett DR, Aebersold R, and Hood L. Integrated genomic and proteomic analyses of a systematically perturbed metabolic network. Science 292: 929934, 2001.[Abstract/Free Full Text]
- Kelleher JK and Masterson TM. Model equations for condensation biosynthesis using stable isotopes and radioisotopes. Am J Physiol Endocrinol Metab 262: E118E125, 1992.[Abstract/Free Full Text]
- Krause U, Bertrand L, Maisin L, Rosa M, and Hue L. Signalling pathways and combinatory effects of insulin and amino acids in isolated rat hepatocytes. Eur J Biochem 269: 37423750, 2002.[Abstract/Free Full Text]
- Lowenstein JM, Brunengraber H, and Wadke M. Measurement of rates of lipogenesis with deuterated and tritiated water. Methods Enzymol 35: 279287, 1975.[Medline]
- Mancuso A, Sharfstein ST, Fernandez EJ, Clark DS, and Blanch HW. Effect of extracellular glutamine concentration on primary and secondary metabolism of a murine hybridoma: An in vivo C-13 nuclear magnetic resonance study. Biotechnol Bioeng 57: 172186, 1998.[CrossRef][ISI][Medline]
- Miller WM, Wilke CR, and Blanch HW. Transient responses of hybridoma cells to nutrient additions in continuous culture. 1. Glucose pulse and step changes. Biotechnol Bioeng 33: 477486, 1989.[ISI]
- Newsholme EA, Crabtree B, and Ardawi MS. The role of high rates of glycolysis and glutamine utilization in rapidly dividing cells. Biosci Rep 5: 393400, 1985.[ISI][Medline]
- Portais JC, Voisin P, Merle M, and Canioni P. Glucose and glutamine metabolism in C6 glioma cells studied by carbon 13 NMR. Biochimie 78: 155164, 1996.[CrossRef][ISI][Medline]
- Quillard M, Renouf S, Husson A, Meisse D, and Lavoinne A. Glutamine and regulation of gene expression in mammalian cells. Special reference to phosphoenolpyruvate carboxykinase (PEPCK). Biochimie 79: 125128, 1997.[CrossRef][ISI][Medline]
- Reitzer LJ, Wice BM, and Kennell D. Evidence that glutamine, not sugar, is the major energy source for cultured HeLa cells. J Biol Chem 254: 26692676, 1979.[ISI][Medline]
- Rigoutsos I and Floratos A. Combinatorial pattern discovery in biological sequences: The TEIRESIAS algorithm. Bioinformatics 14: 5567, 1998.[Abstract]
- Tavazoie S, Hughes JD, Campbell MJ, Cho RJ, and Church GM. Systematic determination of genetic network architecture. Nat Genet 22: 281285, 1999.[CrossRef][ISI][Medline]
- Thomas PE and Hutton JJ. Biochemical and tissue culture studies of transplantable mouse hepatomas H-4, H-6, and BW7756. J Natl Cancer Inst 47: 10251031, 1971.[ISI][Medline]
- Wolfe R. Radioactive and Stable Isotope Tracers in Biomedicine. New York: Wiley-Liss, 1992.
- Zielke HR, Ozand PT, Tildon JT, Sevdalian DA, and Cornblath M. Reciprocal regulation of glucose and glutamine utilization by cultured human diploid fibroblasts. J Cell Physiol 95: 4148, 1978.[ISI][Medline]