Cell Handling and Metabolite Extraction
LC-MS/MS data on 167 different metabolites were collected at various times following infection with HCMV or a mock negative control innoculum. Fibroblasts were grown to confluence and then maintained in serum-free medium for 24 h prior to infection. This treatment served to synchronize the cells in G0 (Figure S1
), which reduces cell-to-cell variation in response to infection [4
]. Additionally, the use of serum-free medium removes a complex source of extraneous metabolites from the analysis. Cells were assayed at 4, 24, 48, 72, and 96 h post infection (hpi) to provide coverage during the extended replication cycle of HCMV. In cultured fibroblasts, viral gene expression is first detected at about 2 hpi, viral DNA replication begins between 24–30 hpi, and peak virus yields are achieved between 72–96 hpi, after which the cells die.
At the appropriate time post infection, medium was aspirated from the cells (a thin film of medium remains covering the cells due to surface tension effects, avoiding alteration of the immediate cellular environment) and −75 °C methanol was immediately added to quench metabolic activity and extract metabolites. Given the speed of metabolic reactions, quickly removing medium and immediately quenching metabolism is essential for obtaining reliable results; we have found that even a single wash step leads to substantial changes in metabolome composition. The selection of cold methanol as an extraction solvent was based on systematic studies in bacteria and yeast, which point to its providing relatively good extraction of a broad spectrum of metabolites [5
], while avoiding marked metabolite decomposition and associated formation of decomposition products, which can themselves mimic metabolites [8
]. A contributor to the efficacy of −75 °C methanol as an extraction solvent is likely its disrupting membranes by inducing formation of microscopic ice crystals.
Extracted metabolites were measured via LC-MS/MS, with the LC separation involving hydrophilic interaction chromatography on an aminopropyl resin at basic pH, an approach tailored to quantitation of water-soluble cellular metabolites [9
]. Each metabolite measurement was in the form of a selected reaction monitoring scan event (SRM); formation within the mass spectrometer collision cell of a particular molecular weight product ion from a specific molecule weight parent ion by gas phase collision of the parent ion with argon occurring at a specific chromatographic retention time. These SRMs were designed based on prior knowledge of the retention time and mass spectrometry fragmentation pattern of a purified standard of the metabolite of interest [6
], with a one-to-one correspondence between metabolites and SRMs. An exception to the one-to-one relationship between metabolites and SRMs was the co-elution of certain metabolite isomers yielding identical fragmentation patterns (e.g., different isomers of glucose-6-phosphate, which we refer to collectively as hexose-phosphate).
In many cases, no useful quantitative data were obtained, due to the metabolite never being present in detectable amounts or being present in the medium at such high levels as to interfere with analysis. For 63 metabolites, however, reliable quantitative data were obtained in the form of ion counts from compound-specific SRM chromatograms (for details on the requirements for metabolite inclusion see Materials and Methods
). These 63 metabolites included a variety of key components of central carbon metabolism and nucleotide metabolism, including multiple phosphorylated compounds such as ATP and fructose bisphosphate (FBP).
To ensure that experimental artifacts such as ion suppression did not lead to misinterpretation of metabolite levels, isotope-labeled internal standards were added to every sample. As is typical for electrospray ionization-mass spectrometry data, there are occasional fluctuations in the internal standard signals, generally in the range of ~20% (Figure S2
), which is much smaller than the effects due to viral infection described below. As the fluctuations generally occurred in parallel for the mock versus viral samples and did not follow any systematic pattern with respect to viral versus mock treatment (Figure S2
), there is no evidence of systematic error between mock and viral samples due to ion suppression or other electrospray ionization-mass spectrometry artifacts. Accordingly, experimental error in the LC-MS/MS measurement is very unlikely to have influenced the biological conclusions of the present work.
Data Normalization and Presentation
To enable cross-sample comparison of metabolite (ion count) signals, the observed signal was normalized with respect to the total protein content of the cells used to generate the sample. Protein content, rather than cell number, was used for normalization because HCMV-infected cells increase approximately 2-fold in volume as compared to mock-infected cells. Normalization with respect to protein content should yield signals in units approximating ion counts/cell volume, which for any given metabolite is linearly proportional to its intracellular concentration [9
]. Comparison of ion counts across compounds is complicated by compound-specific ionization efficiencies. Accordingly, signals were further normalized to the initial (t = 4 h) signal in the mock-infected condition.
Time course data for compounds or pathways of particular interest are shown in –, and complete data (including absolute ion counts) on the 63 quantifiable metabolites are provided in Table S1
. RNA expression data were collected by microarray analysis, and data for key metabolic enzymes that changed significantly upon infection are shown in . The primary expression data are provided in full via the Princeton University MicroArray database (http://puma.princeton.edu/index.shtml
). In certain specific cases, we validated our microarray data using quantitative PCR (qPCR) (A).
HCMV-Induced Changes in Glycolytic Metabolites
HCMV-Induced Decreases in Selected Metabolite Pools
Effect of HCMV on RNA Levels of Glycolytic, TCA Cycle, and Pyrimidine Biosynthetic Enzymes
HCMV-Induced Changes in Metabolic Enzyme Activity and Transcript Levels
General Strategies for Data Interpretation
Before addressing particular experimental observations, it is useful to consider the different potential causes of particular metabolite level changes. Consider the following metabolic pathway:
Increased intracellular levels of a given metabolite B can arise from either increased flux from A → B, or decreased flux from B → C. These alternatives can likely best be distinguished using isotope tracers; however, in the absence of such tracer data (which will not be presented here), good estimates of the true cause of a change in the level of metabolite B can often be made based on knowledge of key regulated steps in metabolic pathways. That is, if the reaction A → B but not B → C is known to be a key regulated step, then change in flux through A → B likely accounts for the change in the level of B. For example, in we report an increase in FBP at 48 hpi. As production of FBP via phosphofructose kinase-1(PFK-1) is tightly regulated [10
], whereas consumption of FBP by aldolase is not, the most likely cause of the increase in FBP is increased PFK-1 flux. This conclusion is reinforced by the observation that levels of the aldolase product, dihydroxyacetone phosphate (DHAP), increase roughly in parallel with those of FBP (if the increase in FBP were due to blockade of aldolase, DHAP levels should drop, not increase). It is confirmed by direct measurement of PFK activity from cell lysates, as described further below (B).
The above analysis considers only the two most basic causes of changes in the level of metabolite B: altered consumption or production via a linear pathway. It is worth noting that especially in a highly compartmentalized eukaryotic cell, other situations can arise. For example, imagine that the reactions A → B → C normally occur in a subcellular compartment that upon viral infection begins to leak B into the cytosol, which contains no enzyme that consumes B. In this case, increased levels of B may occur independently of flux increases or decreases via the standard A→ B → C reaction pathway, depending instead on the extent of B leakage. Such a case may describe the pyrimidine pathway intermediate carbamoyl-aspartate, as will be discussed later.
We wish to emphasize that while transport of metabolites between subcellular compartments may play a key role in regulation of mammalian metabolism, our current analytical approach provides only average metabolite levels throughout the full cellular volume. Accordingly, the pathway diagrams in the figures include only covalent, not transport, reactions.
Effect of HCMV Infection on Upper Glycolysis
Glycolysis and the TCA cycle oxidize hexose sugars to form ATP and NADH, the major energy currencies of the cell. While HCMV infection has been previously shown to stimulate glucose uptake in fibroblasts [13
], its detailed effect on intracellular carbon metabolism has not been previously investigated.
We found that early during infection (i.e., 4–24 hpi), there was little difference in measurable glycolytic pathway metabolites between mock- and HCMV-infected fibroblasts (). Beginning at 48 hpi, however, differences emerged: FBP, DHAP, 3-phosphogylcerate, and phosphoenolpyruvate (PEP) increased relative to mock infection ().
The increased levels of glycolytic intermediates during HCMV infection correlated with an increase in the transcript levels of several glycolytic enzymes ( and ). One such enzyme, PFK-1, is the key upstream regulator of glycolytic flow [14
]. To verify our microarray results regarding this important regulatory enzyme, we analyzed its transcript levels over the course of infection by qPCR. Transcript levels were found to be increased at every time point assayed with a peak 6-fold increase at 48 hpi (A). The levels of PFK-1′s product, FBP, also peaked at this time, suggesting increased PFK-1 activity in HCMV-infected cells at 48 hpi. To directly test for such activity, we performed biochemical assays on cell lysates. Addition of fructose-6-phosphate and ATP to lysates of HCMV-infected or mock-infected cells resulted in accumulation of FBP, with the initial rate of accumulation ~4-fold greater for HCMV than mock infection. Since PFK-1 regulates glycolytic flow, the increased levels of downstream glycolytic intermediates during HCMV infection likely reflect, at least in part, this increase in PFK-1 activity.
Glycolytic Efflux during HCMV Infection
Most of the measured glycolytic intermediates increased at 48 hpi relative to their status in mock-infected cells. However, the increase in FBP was transient, presumably due to its rapid conversion to downstream glycolytic intermediates. The terminal glycolytic metabolite, PEP, increased from 48 hpi onward (). This suggested the presence of a bottleneck to glycolytic efflux at the level of pyruvate kinase late in infection. Such a bottleneck could arise from decreased maximum velocity for the pyruvate kinase reaction (e.g., due to decreased pyruvate kinase levels), increased Km
of the enzyme for PEP (e.g., due to allosteric regulation or shift to a higher Km
isozyme), or failure of pyruvate kinase activity to keep pace with increased flux through upper glycolysis. Our microarray data revealed increased expression of enzymes linking FBP to PEP (triose phosphate isomerase, phosphoglycerate kinase, and enolase) but decreased expression of pyruvate kinase (, A, and ). This decrease in pyruvate kinase transcript is consistent with (and may potentially cause) the observed accumulation of PEP. Note that the reactions linking DHAP to PEP are associated with net ΔG0
′ ~zero [17
]. Thus, irrespective of the activities of the intervening enzymes, DHAP and its products must accumulate in parallel with PEP for glycolysis to continue to run in the net forward direction.
The principal carbohydrate product of glycolysis is pyruvate, which was not measured here due to its poor fragmentation in MS/MS. A key metabolite of pyruvate is alanine, which is produced via pyruvate transamination. Levels of alanine increase markedly starting at 48 hpi (), in concert with the increase in FBP and other glycolytic intermediates (). Much like PEP, alanine levels grow steadily from 48 hpi onward. This suggests that overall glycolytic flux (as reflected in part by alanine production) is increased from 48 hpi onward, with shifting enzymatic activity in lower glycolysis altering the balance of FBP versus PEP during this period.
HCMV-Induced Changes in the TCA Cycle
Intriguingly, two common features of human cancers are the activation of phosphofructokinase activity mediated by increased F2,6P accumulation and the expression of a particularly low-activity isozyme of pyruvate kinase [11
]. Thus, increased phosphofructokinase activity and decreased pyruvate kinase activity, which together may favor use of glycolytic intermediates for production of biosynthetic building blocks, appear to be shared between cancer and HCMV infection.
Increased Acetyl-CoA during HCMV Infection
Acetyl-CoA occupies a central position in metabolism. It is the link between glycolysis and the TCA cycle, as well as a hub for fatty acid and amino acid oxidation and de novo lipid biosynthesis. Like alanine, acetyl-CoA can be formed from pyruvate, and like alanine, acetyl-CoA increased dramatically throughout HCMV infection; however, acetyl-CoA is unique in rising even before glycolytic intermediates and ultimately increasing by ~8-fold in infected as compared to uninfected cells (). The strength of the acetyl-CoA response, as well as its initiation prior to evidence of increased glycolytic intermediates, suggests the possibility of increased activity of the key acetyl-CoA producing enzyme complex, pyruvate dehydrogenase. Consistent with this notion, pyruvate dehydrogenase transcript levels were upregulated 2- to 3-fold by the virus both by microarray () and by qPCR ().
TCA Intermediates during HCMV Infection
When acetyl-CoA enters the TCA cycle, it is converted, with the addition of oxaloacetate, into citrate. Throughout infection, the levels of many TCA cycle constituents were increased with respect to mock infection; however, citrate levels were not, although citrate levels did increase in parallel in both mock- and virus-infected samples (). Citrate synthase activity is inhibited by increased levels of NADH and succinyl-CoA [20
], the levels of which are elevated during HCMV infection (), potentially preventing accumulation of citrate (). Alternatively, oxaloacetate may be diverted via aspartate into de novo pyrimidine synthesis, which is activated during HCMV infection (), thus limiting its availability for citrate synthesis; direct oxaloacetate measurement here was precluded by oxaloacetate instability [9
]. More likely, however, citrate synthase flux was increased as a consequence of increased acetyl-CoA levels, but citrate levels were kept steady by increased activity of the citrate-consuming enzyme, aconitase, whose mRNA level increased during HCMV infection (). Consistent with increased aconitase expression, aconitate (4-fold increased at 24 and 48 hpi) and its downstream TCA cycle products, including succinyl-CoA and malate, increased during infection (). Because succinyl-CoA is the entry point for odd-numbered fatty acid chains into central carbon metabolism, its particular strong induction hints at the possibility of increased fatty acid catabolism induced by the virus. Transcriptionally, the expression levels of malic enzyme and malate dehydrogenase both increased (). Malic enzyme converts malate to pyruvate, potentially contributing to increased alanine and acetyl-CoA levels and limiting citrate levels, whereas malate dehydrogenase produces oxaloacetate from malate, thus completing the TCA cycle by providing citrate synthase with oxaloacetate.
HCMV-Induced Changes in Pyrimidine Biosynthesis
The primary recipient of the energy released from citric acid cycle oxidations is the reduced electron carrier NADH. Consistent with elevated glycolysis and citric acid cycle activity, NADH levels increased approximately 2- to 4-fold relative to mock infection at every time assayed after 4 h (). This increase in NADH was unlikely to be due to reduced NADH consumption, as enzymes of the electron transport chain and the ATP-synthesizing proton pump were markedly transcriptionally upregulated (Table S2
). Taken together, these results are most consistent with the hypothesis that HCMV infection stimulates both glycolytic and pyruvate dehydrogenase flux. This could generate the observed high level of acetyl-CoA, which was then used to drive the TCA cycle to produce NADH and eventually ATP, as well as potentially also donating carbon units to lipid biosynthesis, which was not explored here.
Dynamics of Central Carbon Metabolites in the Absence of HCMV
The levels of certain compounds, e.g., citrate, aconitate, and alanine, increased during mock infection (). The levels of other compounds, e.g., FBP and NADH, decreased ( and ). As the cellular medium was not changed during the course of infection, these changes may reflect alteration of the cellular nutrient environment over time, perhaps due to glucose becoming less abundant and glycolytic waste products accumulating. Alternatively, the changes in metabolite levels may reflect long-term adaptations of fibroblasts to tissue culture. One effect of HCMV infection may be to accelerate some of these processes that occur even in its absence; however, acceleration of normal physiological processes is inadequate to account fully for the profound and persistent changes seen in many compounds such as acetyl-CoA and NADH.
HCMV Manipulation of Pyrimidine Biosynthesis
HCMV infection stimulates total RNA and DNA synthesis and increases the total number of cellular ribosomes without degrading host cell chromosomal DNA [2
]. Consequently, the virus must induce substantial de novo nucleotide biosynthesis. The first three steps of pyrimidine biosynthesis are catalyzed in eukaryotes by a homo-trimer of the 230-kDa protein termed carbamoyl phosphate synthetase-aspartate transcarbamoylase-dihydroorotase (CAD), which catalyzes the synthesis of carbamoyl-phosphate, carbamoyl-aspartate, and dihydroorotate [21
]. Carbamoyl-phosphate levels remained relatively similar in mock-infected cells versus HCMV-infected cells until 72 h when they started to fall in infected fibroblasts (). Levels of carbamoyl-aspartate, in contrast, increased dramatically during viral infection starting at 24 hpi (). While levels of carbamoyl-aspartate increased substantially, levels of the CAD transcript remain unchanged during the course of infection (), suggesting that transcriptional activation of the CAD gene is not responsible for the accumulation of carbamoyl-aspartate. The increased carbamoyl-aspartate levels did not appear to arise from blockade in its utilization, as levels of pyrimidine biosynthetic end products such as the ribonucleotides UTP and CTP and the deoxyribonucleotide TTP were generally increased compared to mock infection (), although levels of UMP, which sits in the middle of pyrimidine biosynthesis, were actually slightly decreased.
Interestingly, UTP reportedly serves as a feedback inhibitor of CAD activity [24
] and one would expect that the increased levels of UTP that occur during HCMV infection might inhibit CAD activity, thereby blocking carbamoyl-aspartate accumulation. One possible explanation for the continued high levels of carbamoyl-aspartate in the face of high UTP is that infection might spatially segregate UTP and CAD into different compartments. This might involve interplay between the virus and mitogen-activated protein kinase, which has recently been shown to regulate CAD activity/localization [26
Another challenge in interpreting the carbamoyl-aspartate accumulation is that the three active sites of CAD are linked by a 96-Å long channel which supposedly localizes carbamoyl-phosphate and carbamoyl-aspartate to the interior of the enzyme [27
]. Thus, the presence of free carbamoyl-aspartate may indicate profound perturbation of CAD in infected cells, perhaps as the consequence of an interacting virus-coded protein that results in leakage of carbamoyl-aspartate from the enzyme interior.
Purine Homeostasis during HCMV Infection
In contrast to the aforementioned alterations of the pyrimidine biosynthetic pathway, HCMV infection did not dramatically affect purine metabolite pool levels. AMP, ADP, and ATP metabolite levels did not change significantly with respect to mock infection, with the exception of a slight increase in ADP and ATP at 48 h and a small accumulation of AMP after 48 h in HCMV-infected versus mock-infected fibroblasts (). The present study does not provide substantial insight into de novo purine biosynthesis however, because the only purine intermediate for which an SRM was included, IMP, could not be measured due to interference from naturally occurring heavy isotopes of AMP (see Materials and Methods
). While the levels of purines do not change dramatically, it is likely that increased turnover of the high-energy phosphate bond of ATP occurs, as the RNAs encoding mitochondrial ATP synthases were significantly upregulated (Table S2
Homeostasis of Purine Nucleotides and Amino Acids during HCMV Infection
The dramatic increase in pyrimidine versus purine levels may reflect different biological roles of the two families of metabolites. Dramatic concentration changes in ATP and GTP could alter the balance of a multitude of enzymatic reactions and interactions based on their widespread utilization as signaling constituents and energy currency. As such, a pathogen may be forced to maintain purine levels close to normal physiological levels. Pyrimidines, in contrast, do not have equally wide-ranging functions and thus HCMV may have license to more dramatically influence pyrimidine levels.
HCMV Infection and Amino Acids
With the exception of alanine, HCMV infection did not induce increased levels of other measured amino acids. While the presence of exogenous amino acids in the medium precluded quantification of a number of amino acids, the levels of glutamate were not affected by viral infection (), nor were the levels of asparagine, aspartate, histidine, or proline (Table S1
). Taken together, these data suggest that while HCMV infection strongly impacts the metabolite levels of certain metabolites, it has little effect on many amino acid levels.
HCMV Infection Depletes Cells of a Subset of Metabolites
A number of metabolites were found to decrease rapidly during HCMV infection compared to mock infection of fibroblasts (). These metabolites could be grouped generically into those, such as adenosine and carnitine, for which HCMV infection seemed to accelerate decreases that also occurred during mock infection, and others, like myo-inositol and glycerol phosphate, which decreased only in the HCMV infection condition. Several of the metabolites decreasing during HCMV infection were nucleosides (adenosine and uridine), consistent with use of these metabolites for nucleotide and eventually RNA or DNA biosynthesis. Others were related to lipid metabolism: carnitine, which plays a key role in β-oxidation of long-chain fatty acids [30
]; glycerol-phosphate, a precursor to the glycerol portion of phospholipids, diacylglycerols, and triglycerides; and myo-inositol, a precursor to inositol-containing lipids such as phosphatidyl-inositol. The results for glycerol-phosphate, while unusually noisy, were striking in that decreases were evident already at 4 hpi, making glycerol-phosphate depletion among the earliest known effects of HCMV infection on cellular metabolism. Taken together, these results demonstrate that HCMV infection depletes a subset of precursor metabolites, some very rapidly, most likely utilizing them for its own macromolecular synthesis.
Taurine, an amino acid analog involved in osmoregulation and cell volume control [32
], is not an obvious precursor metabolite, but was nevertheless decreased 4-fold by HCMV infection (, top left panel). Typically, if a cell is subjected to hypo-osmotic conditions, taurine is rapidly transported out of the cell to compensate for the osmotic imbalance and to prevent cell volume increase [32
]. The cytomegaloviruses are named for the cytopathic effect they induce, which involves a significant increase in cell volume. It seems likely that taurine is not utilized by the virus, but instead excreted by the host cell in response to the virus-induced increase in cell volume.
HCMV-Induced Metabolite Levels Far Exceed Those Seen in Actively Growing Cells
HCMV activates a number of cyclin-dependent kinases and induces quiescent G0 cells to enter the G1 compartment of the cell cycle and progress to the G1-S border [34
]. Up to this point, we tested the effects of HCMV infection in confluent, serum-starved G0 cells. Consequently, we were interested if HCMV-induced metabolic changes mirrored the differences between actively growing and quiescent cells, so we tested how the metabolic changes induced by infection of quiescent G0 cells compared to the differences between quiescent and actively growing cells. Toward this end, we harvested fibroblasts that were actively growing and processed them for LC-MS/MS. Metabolite levels in actively growing fibroblasts were compared to 72-hpi, HCMV-infected fibroblasts. The relative levels of acetyl-CoA, alanine, carbomoyl-aspartate, succinyl-CoA, and to a lesser extent PEP, NADH, CTP, and TTP, were dramatically elevated in HCMV-infected fibroblasts compared to actively growing fibroblasts (). Thus, HCMV does not modify the cellular metabolome simply by inducing quiescent cells to enter the G1 phase of the cell cycle. Rather, the virus re-programs the metabolic status of the cell, producing a unique signature of infection.
HCMV Infection Disrupts Normal Homeostatic Metabolite Pool Levels
Conclusions and Future Directions
The network of chemical reactions involving the derivation of energy currency and biosynthetic building blocks from nutrient inputs consists of ~500 different compounds and 700 interconverting reactions [36
]. Remarkably, relatively little is known of how intracellular pathogens modulate these metabolites and their interconversions. We utilized a global approach to address this issue. As commonly occurs with such approaches, we generated many specific biochemical hypotheses that necessitate further inquiry. One such hypothesis, that HCMV infection induces phosphofrucktokinase activity, was verified. Many more remain untested.
Beyond specific biochemical hypotheses, we also began the process of tackling larger biological questions regarding the metabolic interplay between pathogens and their hosts. Can pathogens meet their biosynthetic needs simply by activating cellular growth pathways with a coordinate increase in metabolic activity? This does not appear to be the case for HCMV. Our results indicate that the virus institutes its own metabolic program, altering normal metabolic homeostasis maintained in quiescent and growing cells and considerably stimulating aspects of glycolysis, the citric acid cycle, and pyrimidine biosynthetic metabolic pathways. Future experiments are necessary to determine whether the numerous metabolic differences observed during HCMV infection are necessary for efficient viral replication, and, if so, the means by which these metabolic alterations contribute to the virus growth cycle.