An Overview Of Phosphometabolite Dynamics Biology Essay

Published: Last Edited:

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Comprehensive characterization of diseases-related metabolomic phenotypes and drug effects requires monitoring metabolite levels and their turnover rates from which metabolic fluxes and the status of the whole metabolic system can be determined. Tandem application of stable isotope 18O-assisted 31P NMR and mass spectrometric techniques uniquely allow simultaneous measurements of phosphorus-containing metabolite levels and their respective turnover rates in tissue and blood samples. The 18O labeling procedure is based on incorporation of the 18O atom, provided from H218O, into Pi with each act of ATP hydrolysis and the subsequent distribution of 18O-labeled phosphoryls among phosphate-carrying molecules. Essentially all major phosphometabolites and their turnover rates can be quantified using 18O-assisted 31P NMR spectroscopy and mass spectrometry with or without prior HPLC separation of metabolites classes. This technology permits the simultaneous recording of ATP synthesis and utilization, phosphotransfer fluxes through adenylate kinase, creatine kinase and glycolytic pathways, as well as mitochondrial nucleotide, associated with Krebs cycle, dynamics and glycogen turnover. Another advantage of 18O methodology is that it measures almost every phosphotransfer reaction taking place in the cell including the turnover of small pools of signaling molecules and dynamics of energetic signal communication. Our studies demonstrate that 18O-assisted 31P NMR/mass spectrometry is a valuable tool for phosphometabolomic and fluxomic profiling of transgenic models of human diseases revealing system-wide adaptations in metabolic networks, as well as for biomarker identification and metabolic monitoring of drug toxicity.


3-PG: 3-Phosphoglyceric acid, 6-PG: 6 phosphogluconate, ADP: Adenosine diphosphate, AMP: Adenosine monophosphate, ATP: Adenosine triphosphate, cAMP: Cyclic adenosine monophosphate, CE: Capillary electrophoresis, Cr: Creatine, CrP: Creatine phosphate, DHAP: Dihydroxyacetone phosphate, F6P: Fructose 1,6-bisphosphate, FAD: Flavin adenine dinucleotide, FADH: Flavin adenine dinucleotide reduced, FDP: Fructose 1,6-bisphosphate, G1P: Glucose 1 phosphate, G3P: Glycerol 3 phosphate, G6P: Glucose 6 phosphate, GA3P: Glyceraldehyde 3-phosphate, GC: Gas chromatography, GDP: Guanosine diphosphate, GMP: Guanosine monophosphate, GPC: Glycerophosphocholine, GPE: Glycerophosphoethanolamine, GPS: Glycerol 3-phosphoserine, GTP: Guanosine triphosphate, IMP: Inosine monophosphate, LAC: Lactate, LC: Liquid chromatography, NADP: Nicotinamide adenine dinucleotide phosphate, NADPH: Nicotinamide adenine dinucleotide phosphate reduced, NMR: Nuclear magnetic resonance, PC: Phosphocholine, PCA: Principal component analysis, PEP: Phospho(enol)pyruvic acid, Pi: Inorganic phosphate, PLS DA: Partial least squares discriminant analysis, PPi: pyrophosphate, R5P: Ribose 5-phosphate, TP: Total phosphate.


Metabolomic analyses require comprehensive and simultaneous systematic fingerprinting of multiple metabolites. These are to be identified and quantified along with their cellular and systemic variations in response to diseases, drugs, toxins and lifestyle, as well as in the context of genetic or environmental challenges 1-8. Analytical platforms developed for metabolomics studies allow screening of hundreds of metabolites from complex biological samples with analytical precision, comprehensiveness, and sample throughput 6, 9-12. The physicochemical diversity of metabolites, from ionic inorganic species to hydrophilic carbohydrates, volatile alcohols and ketones, amino and non-amino organic acids, hydrophobic lipids, and complex natural products necessitates application of different complementary analytical techniques 2, 3, 9. Currently no single platform fulfills all requirements for an ideal global metabolite profiling tool. Application of advanced and information rich spectroscopic techniques is typically essential for generation of metabolic profiles required for metabolomic studies 13. The main spectroscopic techniques employed for metabolomic studies are based on NMR spectroscopy (1H, 31P, 13C, 17O and others) and mass spectrometry (direct infusion or combined with GC, LC or CE). Both techniques can give extensive structural and conformational information on multiple chemical classes in a single analytical procedure; however, they have different analytical strengths and weaknesses 1, 11, 13.

Characterization of metabolic phenotype requires knowledge not only of metabolite levels but also of their turnover rates from which metabolic fluxes and, therefore, the dynamic state of a metabolic system can be determined (Figure 1) 14-16. Because many metabolites are present in low concentrations and associated with high flux/turnover rates through the metabolite pools, significant changes in metabolic flux could occur without changes in metabolite concentration 17. Therefore, dynamic metabolomic profiling and flux measurements are essential for a complete understanding of metabolic phenotypes 2, 16-20. (Figure 1 near here)

Stable isotope tracer-based metabolomic technologies allow simultaneous determinations of metabolite levels and their turnover rates with subsequent evaluation of metabolic network dynamics 14, 15, 21, 22. 13C labeling is widely used to track turnover of the carbon backbone of metabolites and label propagation through metabolic networks 23-25. This technique alone, however, does not allow acquisition of a full picture of metabolic dynamics and of the status of the cell energetic system. 18O isotopes are suitable to follow cellular phosphorus turnover and metabolic dynamics of phosphoryls in energetically and signal transduction important biomolecules as well as label distribution through phosphotransfer networks 15, 22, 26-31. 18O is a natural, stable and non-radioactive isotope of oxygen. When tissues or cells are exposed to media containing water with known percentage of 18O, H218O rapidly equilibrates with cellular water, and then 18O from water is incorporated into cellular phosphate metabolites proportionally to the rate of enzymatic reactions involved30. The percentage of 18O incorporation into phosphate metabolites of interest can be determined by 31P NMR or mass spectrometry 15, 32, 33. Incorporation of 18O into phosphoryls as a result of cellular metabolic activity induces an isotope shift in the 31P NMR spectrum due to differences in the shielding effects of 16O versus 18O on the 31P nucleus as well as a shift in the mass spectrum of phosphoryl-containing metabolite species 15, 31, 34. Calculation of percentage of 18O incorporation into phosphate metabolites from the induced isotope shift 31P NMR spectra can be employed to determine turnover rates and phosphotransfer flux through specific energetic circuits (Figure 1A).

The 18O labeling procedure is based on the incorporation of one 18O atom, provided from H218O, into Pi with each act of ATP hydrolysis and the subsequent distribution of 18O-labeled phosphoryls among other phosphate-carrying molecules (Figure 1B). In conjunction with 18O-assisted 31P NMR spectroscopy and mass spectrometry, the 18O labeling procedure provides a versatile methodology for simultaneous measurement of metabolite levels and metabolic fluxes through phosphotransfer systems allowing characterization of different energetic pathways 15, 16, 22, 27-29, 33, 35, 36 (Figure 1A). This includes simultaneous recordings of ATP synthesis and utilization, phosphotransfer fluxes through adenylate kinase, creatine kinase and glycolytic pathways as well as mitochondrial Krebs cycle activity, glycogen turnover and intracellular energetic communication (Figure 1B). Another advantage of 18O methodology is that it can measure almost every phosphotransfer reaction taking place in the cell including important signaling molecules such as cAMP, cGMP and AMP turnovers and their metabolically active pool sizes 22, 30, 37, 38. The 18O-phosphoryl labeling procedure detects only newly generated molecules containing 18O-labeled phosphoryls reflecting their turnover rates and net fluxes through individual metabolic pathways 15, 35, 39. Theoretically up to one-third of all metabolites containing phosphorus 40 and their turnover rates can be quantified using high resolution 31P NMR spectroscopy and mass spectrometry. Thus, this technology permits determination of phosphometabolites and multiple phosphotransfer fluxes within metabolic networks.

All metabolomic studies result in complex multivariate data sets that require visualization software and chemometric methods for interpretation. The aim of these procedures is to produce biochemically based fingerprints that are of diagnostic or other classification value, and to identify potentially complex sets of biomarkers supporting the diagnosis or classification 1, 41-44. Here, multivariate data sets obtained from different analytical techniques and 18O-labeling ratios were combined and interpreted using principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) chemometric techniques to extract latent metabolic information, and enable sample classification and biomarker discovery.

In this chapter, we describe principles and methodology of metabolic profiling and analysis of phosphometabolite turnover rates using stable isotope 18O-assisted 31P NMR and mass spectrometry. This advanced phosphometabolomic platform is valuable tool in studies of intact muscle energetics and phosphoransfer networks, and unique for measurements of intracellular energetic communication and metabolic signal dynamics. Basic concepts of 18O labeling technique are explained and illustrated by several examples. Special focus is placed on sample preparation, calculation of labeling rates and multivariate data analyses.

2. Methodology

2.1. Phosphometabolomic platforms

Phosphorous is an essential element indispensable to life activity, such as genetic inheritance, signal transduction, metabolism and energy conversion 45. Phosphate is the most common fragment by the frequency of occurrence in the metabolome of living organisms 40 . In the Human Metabolome Database ( there are 744 compounds containing "phospho" and 419 with "phosphate" in their structures from 8536 metabolites. Origins of comprehensive analysis of phosphorus containing metabolites can be traced to the Besman's phosphate analyzer where 32P labeling and chromatographic separation and quantification of phosphometabolites was performed 46. Most of phosphorus containing metabolites is highly polar and their separation and analysis represent a challenge. Phosphometabolites can be measured simultaneously by several analytical techniques, including 31P NMR, LC/MS, GC/MS, CE/MS and HPLC 45, 47, 48. Although these methods are generally successful in determining the concentration of metabolites, it is not possible to measure all phosphometabolites using one technique due to stability, concentration of metabolites or the dynamic range of instruments. For example, sugar phosphates are best separated using GC/MS 12, while phospholipids with 1H and 31P-NMR 49 and nucleotides with LC 50.

We established a dynamic phosphometabolomic platform (Figure 2) that includes 18O-assisted GC/MS, 18O-assisted 31P NMR, 1H NMR and HPLC. We are also developing a LC/MS method for quantification of 18O-labeling of mono or oligo-phosphometabolites. 18O-assisted GC/MS technology, which originally was developed in Nelson Goldberg's laboratory 27, 32, 35, 37, allows separation and quantitation of 18O/16O isotope ratios in phosphoryl metabolites with a molecular mass <500 Da. Higher molecular weight phosphates and oligo-phosphates, such as ATP or GTP, can be analyzed after enzymatic transfer of corresponding phosphoryls to glycerol 27, 36. 18O-assisted 31P NMR technique uses 18O-induced shift in 31P NMR spectra to determine the percentage of 18O-labeling of metabolite phosphoryls 15, 31.

(Figure 2 near here)

This technology, which has been used for enzymatic mechanism analyses in vitro 31, 34, is adapted and developed for tracking phosphoryl metabolic dynamics in intact tissues 15, 22. The critical advantage of 18O-assisted 31P NMR technique is that it does not require prior metabolite separation and derivatization; it is stable, and quantitative, and allows simultaneous single-run recordings of multiple metabolite phosphoryls and of separate phosphoryls within one molecule such as α-, β- and g-phosphoryls of ATP 15, 22. However, compared to GC/MS, 18O-assisted 31P NMR is less sensitive, requires a larger amount of sample and longer analysis time. 1H NMR in our studies is used as a complementary technology for quantification of phosphometabolite levels in tissue extracts and biological fluids 22. HPLC using ion-exchange, reversed-phase, hydrophobic and hydrophilic interaction chromatography is a versatile technique for separation and quantification of major phosphometabolite classes 15, 27, 36. The use of triethylammonium bicarbonate (TEAB) buffer, introduced by Khorana 51 is preferential since the volatility of TEAB facilitates sample recovery after HPLC chromatographic separation and makes it suitable for mass spectrometric analysis of phosphometabolites.

2.2. 18O metabolic labeling procedure

18O is a natural, stable and non-radioactive isotope of oxygen. When tissue or cells are exposed with media containing known percentage (20-30%) of 18O, H218O rapidly equilibrates with cellular water and then 18O from water is transferred to cellular phosphate metabolites proportionally to the rate of enzymatic reactions involved. The rate of sequential enzymatic reactions between Pi, g-ATP and CrP are high (Figure 3A) and upon 18O labeling display exponential kinetics with saturation occurring within 2 min 22, 29 (Figure 3B). Therefore labeling of metabolites should be performed within the initial linear phase (0-1 min) of the 18O labeling curve, while for β-ADP and β-ATP, which have lower turnover rates, labeling can be performed within a 5 min time window. After the desired time of exposure with H218O, cell metabolism is instantaneously stopped by immersing cells or tissue into liquid N2.

(Figure 3 near here)

Heart perfusion and 18O phosphoryl labeling. Hearts from heparinized (50 U ip) and anesthetized (75 mg/kg pentobarbital sodium ip) wild-type or transgenic mice are excised and retrogradely perfused with a 95% O2-5% CO2-saturated Krebs-Henseleit (K-H) solution (in mM: 118 NaCl, 5.3 KCl, 2.0 CaCl2, 19 NaHCO3, 1.2 MgSO4, 11.0 glucose, 0.5 EDTA; 37°C) at a perfusion pressure of 70 mmHg. Hearts are paced at 400 beats/min. Hearts are perfused for 30 min and then subjected to labeling with 18O, which is introduced for 30-60 s with the K-H buffer supplemented with 20-30% of 18O-labeled H2O (Isotec). Than hearts are freeze-clamped, pulverized under liquid N2, and extracted in a solution containing 0.6 M HClO4 and 1mM EDTA. Extracts are neutralized with 2 M KHCO3 and used to determine 18O incorporation into metabolite phosphoryls 28, 33.

18O-labeling of cultured cells or isolated cardiomyocytes. Cells are washed with PBS and preincubated with ADS or other medium 52, 53. After 15-min media is removed and replaced with 2 ml of media (for 35 mm dish) enriched with a 20-30% H218O and incubated for 2 min at 37°C. Incubation is terminated by rapid removal of H218O enriched ADS media and immediate addition of ice-cold 0.6 M perchloric acid containing 1 mM EDTA. While on ice, cells are scraped from the surface and transferred along with the perchloric acid to a test tube. Then, acid extracts are neutralized with 2 M KHCO3. The final extracts obtained from cell or heart tissue are analyzed using 18O-assisted GC-MS or 31P-NMR in order to determine 18O labeling ratios in phosphate metabolites of interest and calculate of phosphotransfer rates. Tissue levels of metabolites are analyzed using GC-MS, HPLC, 1H NMR and 31P NMR spectroscopy for metabolomic fingerprinting 15, 16, 22, 28, 29.

2.3. GC/MS analysis of 18O-labeling of metabolite phosphoryls

18O labeling ratios of monophosphates (such as G3P, G6P and G1P) are evaluated using GC-MS after purification with HPLC, because of their low concentration in the sample. Although Pi has high concentration in the sample, it must be separated from other phosphate contained metabolites. Because some are very unstable during GC-MS analysis and those metabolites such as CrP and GA3P are easily degraded and produce Pi, which interfere with free Pi in the sample. Therefore, samples are fractionated and concentrated using HPLC. Consequently the labeling ratio can be determined precisely. (Figure 4 near here)

Cellular phophometabolites are purified and quantified with HPLC (Figure 4A) using a Mono Q HR 5/5 ion-exchange column (Pharmacia Biotech) with triethylammonium bicarbonate buffer pH 8.8 at 1 mL/min flow rate 33, 36, 52. From each sample seven fractions are collected. The first fraction contained G6P, G3P, G1P and CrP and fractions from second to the seventh contained AMP, Pi, ADP, GDP, ATP and GTP, respectively (Figure 4A). Fractions are dried out using vacuum centrifugation (SpeedVac, Savant) and reconstituted with water. Monophosphates are transferred to GC-MS vials for silylation, while oligo-phosphates are subjected to enzymatic reactions in Eppendorf tubes to transfer each phosphoryl to glycerol. The g-phosphoryl of ATP or GTP are transferred to glycerol by glycerokinase, and b-phosphoryls of ATP and ADP are transferred to glycerol by a combined catalytic action of adenylate kinase and glycerokinase. The b-phosphoryls of GTP and GDP are transferred to glycerol by a combined catalytic action of guanylate kinase and glycerokinase. The phosphoryl of CrP is transferred to g-ATP by creatine kinase and then to glycerol with glycerokinase. Samples that contained phosphoryls of g-ATP, g-GTP, b-ATP, b-ADP, b-GTP/GDP as G3P, Pi, G6P, G1P, G3P and CrP, are converted to respective trimethylsilyl derivatives with Tri-Sil/BSA (Pierce) as the derivatization agent 22, 33. The 18O enrichments of phosphoryls are determined with GC-MS operated in the select ion-monitoring mode. GC-MS analysis of Pi, G3P and G6P 18O-labeling is presented in Figure 4B. Left panel represents GC-MS chromatograms of metabolites, while in the right panel oxygen the isotope abundance is shown. Another phosphometabolite G1P can be analyzed in this HPLC fraction too (not shown). Using this approach in a single run the metabolic dynamics of glycolysis and glycogenolysis and mitochondrial substrate shuttle activity can be monitored (Figure 4C). Our data indicate that G-3-P metabolic dynamics is altered in transgenic animal models indicating defects in substrate shuttle and supply of reducing equivalents to mitochondria. This is of importance since G-3-P turnover abnormalities and metabolic arrest are linked to human diseases such as sudden death syndrome. The 18O enrichments of phosphoryls are determined with GC-MS operated in the select ion-monitoring mode. Mass ions (m/z) of selected metabolites monitored as trimethylsilyl derivatives are given in the Table. Monophosphates are able to get labeling of up to three oxygen while Pi and PPi are up to 4 and 7 oxygens, respectively. Mass ions (m/z) of monophosphates corresponding to phosphoryl species of 16O, 18O1, 18O2 and 18O3 are monitored at parent ion (16O) +2, +4 and +6, respectively33, 35. (Table near here)

2.4. 31P NMR analysis of 18O incorporation into metabolite phosphoryls

Samples were pre-cleaned for 1 h with Chelex 100 resin (Sigma) supplemented with internal standard for 31P NMR spectroscopy methylene diphosphonate and concentrated by vacuum centrifugation (Savant) to a volume of 0.3 ml. Concentrated extracts were filtered (centrifuge filter; 0.22 µm, Milipore) and supplemented with 0.1 mL of D2O (Isotec) and 0.1 mL 1 mM EDTA. Samples were cleaned additionally with the Chelex resin by rotation at 4 °C for 12 h. To maximize resolution of 18O induced shifts in 31P NMR spectra and to increase sample stability, perchloric acid-extracted tissue were subjected to extensive chelation to remove divalent cations 15, 22, 28, 29.

31P NMR data acquisition was performed at 202.5 MHz using a Bruker 11 T (Avance) spectrometer in high-quality 5 mm tubes (535-PP-7 Wilmad Glass) at ambient temperature and sample spinning at 20 Hz. 9000 scans were acquired without relaxation delay (acquisition time 1.61 sec) using a pulse width of 10 μs (53° angle) with proton decoupling during data acquisition (WALTZ-16 with 90° angle, pulse width of 506 μs for 1H). Before Fourier-transformation FIDs were zero-filled to 32 K, and multiplied by an exponential window function with 0.3 Hz line-broadening (Figure 5A). Peak areas were integrated using the Bruker software after automatic correction of phase and baseline. Typical line widths at half height of various cellular phosphates in 31P NMR spectra were around 0.0080 ppm (1.5 Hz on 202.5 MHz), significantly less than the 18O-induced shift ranging between 0.0210 and 0.0280 ppm. Internal standard was used to set chemical shifts to be at 16.00 ppm and to determine metabolite levels. The metabolite levels according to internal standard were corrected for NOE (by factors determined in typical sample recorded with and without decoupling) and incomplete relaxation (by factors calculated from T1 times in a typical sample, measured by the inversion-recovery technique) as described 28, 33.

(Figure 5 near here)

A typical 31P NMR spectrum of heart extract is shown in Figure 5A. Incorporation of 18O as a result of cellular metabolic activity induces an isotope shift in the 31P NMR spectrum of phosphoryl containing metabolites 31. Although the 18O-induced isotope shift is rather small (around 0.020 ppm), it can be visualized and quantified using high-resolution NMR spectroscopy (Figure 5B). Incorporation of each 18O induces isotope shifts between 0.0210 and 0.0250 ppm in the 31P NMR spectrum of Pi, CrP, g ATP, b ATP, a ATP, b ADP, a ADP, AMP, PC, G6P and G3P. It should be noted that the isotope shift in the spectrum of b ATP was different for bridging and non-bridging 18O oxygens, 0.0170 and 0.0287 ppm, respectively. Moreover, G6P existed in an equatorial and an axial form, and therefore the 16O and 18O species of G6P were represented with two peaks corresponding to each of the two forms (Figure 5B). During the integration procedure, bridging and non-bridging forms of b-ATP as well as equatorial and axial forms of G6P for particular 16O or 18O species were integrated as a single peak.

2.5. Phosphometabolite analysis by 1H-NMR

1H NMR provides a robust and precise method for metabolite quantification including the number of phosphometabolites. 1H NMR data acquisition were performed at 500 MHz using a Bruker 11 T (Avance) spectrometer at ambient temperature and sample spinning at 20 Hz. 128 scans were accumulated under fully relaxed conditions (12.8 s relaxation delay) with a pulse width of 9 μs (90° angle). FIDs were zero-filled to 32 K, and Fourier-transformed without filtering. Phase and baseline were manually adjusted before integration and deconvolution. Chemical shifts were assigned relative to the trimethylsilyl propionate (TSP) signal at 0 ppm. Metabolite levels such as AMP, ATP, ADP, IMP, CrP, glycolytic intermediates and phospholipids were calculated according to TSP used as internal standard. The identity of metabolites was done using Chenomx NMR Suite software, which provides a pattern recognition technique and efficient method for identifying metabolites in biofluid, and confirmed by standard additions.

2.6. Data analysis and calculations of phosphoryl turnover and phosphotransfer fluxes

Introduction of 18O water in tissues of interest leads to 18O incorporation into cellular phosphates according to the rate of involved phosphotransfer reactions (see Figure 1) 15, 27, 30, 36. Such property allows tracking of high-energy phosphoryl transfer routes, and quantification of respective enzymatic fluxes at different levels of cellular activity 15, 22, 27-30, 33, 36. Up to three 18O atoms can be incorporated in monophosphate (G3P, G6P, G1P and CrP) and phosphate at different position in oligo-phosphate (g, b and a for triphosphates and b and a for diphosphates) and up to four and seven for Pi and PPi, respectively. The percentages of 16O, 18O1, 18O2, 18O3 and 18On are proportional to integrals of their respective peaks in the 31P MR spectrum or in the GC-MS chromatograms 15, 22, 28, 29 ( see Figure 4 and 5). Cumulative percentage of phosphoryl oxygens replaced by 18O in the metabolites is calculated as [%18O1 + 2(%18O2) + 3(%18O3)+ …. n (%18On)]/[n(% 18O in H2O)] 15, 22.

Total cellular ATP turnover can be estimated from the total number of 18O atoms that appeared in the phosphoryl-containing metabolites and orthophosphate 22, 33, 36. The kinetics of 18O-labeled phosphoryl appearance in g-ATP reflects cellular ATP synthesis rate while kinetics of Pi 18O-labeling indicates cellular ATPase activity 33. The Pi/g-ATP 18O-labeling ratio, an index of intracellular energetic communication 54, is calculated using the amount or percentage of 18O-incorporated into Pi and g-ATP. 18O-induced shift in 31P NMR spectra and kinetics of 18O-labeling of Pi and g-ATP are presented in Figure 6. Incorporation of 18O into Pi and g-ATP induces very robust multiple shifts in 31P NMR spectra depending on number of oxygens replaced (Figure 6A). From each shift, the labeling ratio can be calculated at different cycle levels (Figure 6B) or total labeling from the sum of different cycles. Labeling reaches saturation within 2-5 min from which the metabolically active pool size can be determined. At saturation almost 100% of g-ATP and about 80% of Pi are metabolically active (18O labeled) (Figure 6B). Incorporation of one, two, three and four atoms of 18O into phosphoryl reflects Pi«ATP cycling between ATP consumption and ATP production sites (Figure 6C). (Figure 6 near here)

Adenylate kinase phosphotransfer flux can be determined from the rate of appearance of 18O-containing b-phosphoryls in ADP and ATP using a computer model based on Stella software 22, 35 or CWave 55, FiatFlux 56, FluxSimulator 57 and other available software. To obtain AK velocity the total number of 18O-labeled phosphoryls in bADP and bATP produced by the AK catalysis is counted. The pool of metabolically active ADP, obtained from labeling studies, is usually larger than free ADP calculated from the CK equilibrium 32, 58, and is in dynamic equilibrium between free and bound states 59, 60. Best fits to experimental data are obtained using metabolically active (18O-labeled) pool size 90% for b-ATP and 30% for b-ADP 32. Total AMP turnover (AK and non-AK mediated) is estimated from the kinetics of AMP α-phosphoryl (non-AK) and β-ATP/β-ADP phosphoryl (mediated by AK) 18O-labeling. The metabolically active AMP or other phosphometabolite pool size is determined after prolonged (20-30 min) 18O-labeling to establish isotopic equilibrium 32. At saturation, almost 100% of g-ATP and CrP, and about 80% of Pi are labeled and metabolically active. Calculation of a-AMP turnover time is done using the formula: SAt = 1-(2-N), where SAt is specific activity of αAMP 18O-labeling at given time t, and N is equal to the number of turnover cycles observed during incubation period 61, 62. Thus, AK independent turnover time of the AMP pool can be calculated from the expression: T = t/N, where T is the turnover time in s. AK-dependent AMP turnover will be calculated using the formula: dN/dt = ρ(P*/P - N*/N) (1), where N*/N - is the specific 18O-labeling of adenine nucleotide β-phosphoryls; P*/P is the specific 18O-labeling of precursor adenine nucleotide g-phosphoryls and Á is the rate of 18O-labelling in the nucleotide pool per time unit 61, 62.

Creatine kinase phosphotransfer rate is determined from the rate of appearance of CrP species containing 18O-labeled phosphoryls and can be modeled using Stella software 22, 35 and other available software 55-57. Glycolytic flux and glycerol phosphate shuttle is determined from the rate of appearance of 18O-labeled G6P and G3P, respectively 16, 22, whereas glycogen flux is determined from the rate of appearance of 18O-labeled G1P. Activity of NDPK/Succinyl-CoA synthase is determined from gGTP 18O-labeling, while bGTP/GDP 18O-labeling indicate guanylate kinase activity.

2.7. Multivariate statistical analysis

Multivariate data sets obtained from different analytical techniques and labeling ratios were combined and interpreted using principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) methods. Initially, the data are examined with PCA scatter plot of the first two score vectors (t1-t2) in order to reveal the homogeneity of the data, any groupings, outliers, and trends. Then PLS-DA is applied to get additional information, increase the class separation an, simplify interpretation, and find potential biomarkers63, 64. The additional information (significant metabolites in group classification) may get with helping of VIP (variable important in the project), loading and regression coefficients plots. The VIP (variable importance in the projection) values 63, 65, 66, a weighted sum of squares of the PLS weight which indicates the importance of the variable to the whole model, are calculated to identify the most important molecular variables for the clustering of specific groups, while the regression coefficient plots of metabolic variables in the PLS-DA model show variables effect on the groups larger coefficient values (positive or negative) have a stronger correlation with group metabolic profile classification. Examination of the corresponding loading plot indicated those metabolites responsible for the clustering of groups. Metabolites located centre of the plot do not contribute to the clustering of the patient groups, whereas those in the same geographical region of a sample group in the corresponding score plot are responsible for the separation. Attention must be given PLS-DA analysis, because it is a supervised method. Even the two groups are not different from each others; the method is force to separate them67. Therefore the PLS-DA model must be validated. For validation, R2 (the fraction of variance explained by a component) and Q2 (the fraction of the total variation predicted by a component) values are considered as measures of goodness of model and the model robustness, respectively. The value of Q2 ranges from 0 to 1 and typically a Q2 value of greater than 0.4 is considered a good model, and those with Q2 values over 0.5 are robust 63, 68. Additionally, the validation of the PLS-DA model can be performed by comparison to the classification statistics of models generated after random permutations of the class matrix. If the model R2 and Q2 values are higher than those obtained in random permuted models across all iterations, the method is valid. Calculation of the PCA and PLS-DA model parameters was carried out using SIMCA-P+ (v12.0, Umetrics AB, Umea, Malmö, Sweden) and MetaboAnalyst web browser 66.

3. Results

3.1. Phosphometabolomic profiling of transgenic animal models

3.1.1. Adenylate kinase AK1 knockout hearts

Maintenance of optimal cardiac function requires precise control of cellular nucleotide ratios and high-energy phosphoryl fluxes. Within the cellular energetic infrastructure, adenylate kinase has been recognized as an important phosphotransfer enzyme that catalyzes adenine nucleotide exchange (ATP + AMP ⇋ 2ADP) and facilitates transfer of both β- and g-phosphoryls in ATP. In this way, adenylate kinase doubles the energetic potential of ATP as a high-energy-phosphoryl carrying molecule and provides an additional energy source under conditions of increased demand and/or compromised metabolic state. By regulating adenine nucleotide processing, adenylate kinase has been implicated in metabolic signal transduction. Indeed, phosphoryl flux through adenylate kinase has been shown to correlate with functional recovery in the metabolically compromised heart and to facilitate intracellular energetic communication 15, 20-22, 28, 29, 32, 33, 35, 36, 54, 69. Deletion of the major adenylate kinase AK1 isoform, which catalyzes adenine nucleotide exchange, disrupts cellular energetic economy and compromises metabolic signal transduction and ischemia-reperfusion response 16, 28, 29, 69, 70. Here we compare metabolomic phenotypes, phosphometabolite and phosphotransfer dynamics in hearts of wild type and AK1 knockout mice at baseline. Male homozygous AK1 knockout (AK1−/−) mice were compared with age- and sex-matched wild-type controls 16, 29.

In hearts with a null mutation of the AK1 gene, which encodes the major adenylate kinase isoform, total adenylate kinase activity and ATP/ADP β-phosphoryl transfer was reduced by 94% and 36%, respectively. Knock out of the major adenylate kinase isoform, AK1, disrupted the synchrony between inorganic phosphate Pi turnover at ATP-consuming sites and γ-ATP exchange at ATP synthesis sites, as revealed by 18O-assisted 31P NMR 70. This reduced energetic signal communication in the post-ischemic heart 29. Moreover, AK1 gene deletion blunted vascular adenylate kinase phosphotransfer, compromised the contractility-coronary flow relationship, and precipitated inadequate coronary reflow following ischemia-reperfusion 70. This was associated with up-regulation of phosphoryl flux through remaining minor adenylate kinase isoforms and the glycolytic phosphotransfer enzyme, 3-phosphoglycerate kinase 28.

The data from 18O labelling rate, 31P and 1H NMR analysis is transformed into meaningful data through multivariate analysis of global profiling by unsupervised PCA and supervised PLS-DA. Initially, the data were examined with PCA score plot of the first two score vectors (t1-t2) in order to reveal the homogeneity of the data, any groupings, outliers, and trends. As seen in Figure 7A, there is clear separation between groups without any outlier and trends. To improve the visualization, these profiles were displayed as hierarchical cluster analysis (Figure 7B). The heat map represented the unsupervised hierarchical clustering of the data grouped by sample type (rows), which also enabled visualization of the up- or down-regulation of each metabolite (columns). Hierarchical clustering were performed with Spearman's rank correlation for similarity measurement and Ward's linkage for clustering by using the MetaboAnalyst web server 66. As seen in the Figure 7, a very clear clustering obtained between two groups. Then PLS-DA were applied to get additional information, increase the class separation an, simplify interpretation, and to discover potential biomarkers 64. (Figure 7 near here)

Genetic deletion of AK1 removed all but 6% of total myocardial adenylate kinase activity, yet the intracellular adenylate kinase phosphotransfer flux was only halved in AK1 knockout hearts. The reduced adenylate kinase-catalyzed phosphotransfer induced rearrangements in adenine nucleotide and glycolytic metabolism, shifting cellular energetics into an apparently new steady state. These changes produced a differential metabolomic profile of the WT and AK1 -/- KO mice heart as seen in the PCA and PLS-DA score plot (Figure 8A). In order to determine significant metabolites in the group differentiation VIP (variable important in the project), loading and regression coefficient plots were used (Figure 8B, C and D). From these plots, it is concluded that glycolytic and nucleotide metabolism and adenylate kinase flux has been altered significantly. Adenylate kinase fluxomic (b-ATP[18O] and b-ADP[18O] turnover), alanine, glucose, threonine, CrP, GPE and nucleotide levels (ADP, AMP and IMP) were decreased in AK1 -/- mice while 3-PG, pyruvate, Pi, G3P, G6P, g-ATP[18O] and CrP[18O] turnover, glutamate, succinate and F6P all were increased. Alterations in 3-PG, G3P, G6P and F6P metabolites indicate adaptations in glycolytic and substrate shuttle activities while changes in glutamate and succinate levels point to altered mitochondrial Krebs cycle activity. Taken together, these changes indicate a system-wide response of cell energy metabolism to deletion of one significant node in the network. PLS-DA analysis to model the metabolic changes associated with gene deletion, a robust predictive model was produced (R2(X)=0.68; R2(Y)=0.98; Q2=0.89 for the three components) (Figure 8E). This model passed cross-validation according to random 100 permutations of the class matrix. The model R2 and Q2 values on the right were higher than those obtained in random permuted models across all 100 iterations, which indicates validity of the method. Thus, phosphometabolomic profiling of adenylate kinase deficient hearts revealed rearrangements and adaptations in heart energetic system with induced shift in glycolytic and creatine kinase phosphotransfer pathways and substrate utilization networks. (Figure 8 near here)

3.1.2. Creatine kinase M-CK knockout hearts

Creatine kinase (CK)-catalyzed phosphotransfer is the major component of energy transfer and distribution network in the heart, and compromised CK function is a hallmark of abnormal bioenergetics in diseased hearts 39, 71-77. Studies of transgenic animal models have demonstrated an inherent plasticity of the cellular energetic system and the development of cytoarchitectural and metabolic compensatory mechanisms in striated muscles 16, 20, 28, 59, 78-83. These studies have led to the concept that interchangeability and rearrangement of phosphotransfer networks provide an intracellular energetic continuum coupling discrete mitochondrial energetic units with ATP utilization sites 39, 84-86.

Although hearts deficient in the major CK isoforms have no gross basal functional abnormalities, under increased load they cannot sustain normal global ATP/ADP ratios, indicating compromised communication between ATP-consuming and ATP-generating cellular sites 58, 81, 87-89. This renders contractions to be more energetically costly, forcing the heart to operate under less efficient cardiac bioenergetics 58, 89. Such energetic abnormalities reduce the ability of the myocardium to respond to β-adrenergic stimulation 90, and CK-deficient hearts are more vulnerable to ischemia-reperfusion injury 91. In addition, CK-deficient hearts cannot maintain adequate subsarcolemmal nucleotide exchange and have increased electrical instability under metabolic stress 92. It is likely that CK-deficient hearts develop cytoarchitectural and metabolic adaptations that modulate energetic disturbances 82, 93-95. However, the adaptive metabolomic phenotype and rearrangements in the bioenergetic system in CK-deficient hearts are still poorly understood. (Figure 9 near here)

Here, adult wild type mice (strain C57/BL6) and transgenic mice lacking cytosolic CK isoform (M-CK-/-) were used 78, 96. Male homozygous M-CK -/- mice were compared with age- and sex-matched wild-type controls. Hearts were perfused and labeled with 18O as explained in the section 2.2 18O labeling procedure. Metabolic signatures for M-CK knockout hearts were revealed using PLS-DA analysis. As demonstrated in the PLS DA score plot (Figure 9A), a good separation was obtained between wild type and M-CK knockout hearts based on metabolite levels and their turnover/18O-labeling rates and substrate metabolism. In order to determine significant metabolites in group discrimination, VIP, loading and regression coefficient plots were used (Figure 9B, C and D). PLS-DA analysis to model the metabolic changes associated with gene deletion, a robust predictive model was produced (R2(X)=0.59; R2(Y)=0.99; Q2=0.86 for the three components) (Figure 9E).

The CK activity of M-CK-/- hearts was reduced by 71% leading to decreases in CK flux assessed by the rate of appearance of 18O-labeled phosphoryls in PCr of 23%. Yet overall ATP synthesis rate measured as the rate of appearance of 18O-labeled phosphoryls in g-ATP did not differ among wild type and M-CK deficient hearts suggesting robustness of cellular energetic system. The trend to increased g-ATP 18O-labeling and a smaller pool size of metabolically active Pi, together with the decreased Pi/g-ATP 18O-labeling ratio, an indicator of intracellular energetic communication, observed here for M-CK deficient hearts, indicate less efficient phosphotransfer energetics. The VIP results show the importance of parameters of glycolytic metabolism (G6P 18O-labeling), AK phosphotransfer (b-ATP/b-ADP 18O-labeling), Pi/ATPase rate (Pi 18O-labeling, Pi, TP) and adenine nucleotide metabolism and ATP turnover (g-ATP 18O-labeling, ADP and AMP levels) in group classification (Figure 9B). Glycolysis, in addition to the traditional role in ATP production, also catalyzes rapid phosphoryl exchange and has been implicated in intracellular energy transfer and distribution 20, 85. Here, changes in glycolytic phosphotransfer in wild type and M-CK knockout hearts were assessed by monitoring appearance of 18O-labeled phosphoryls in G6P as a result of reaction catalyzed by hexokinase, the entry point into glycolysis. In wild type hearts 18O-labeling of G6P was 8.1±0.5%, which was more than 10% of g-ATP turnover. Deletion of M-CK resulted in an increase of G6P 18O-labeling to 13.3±0.8%, which corresponded to 27% of g-ATP turnover. Thus, glycolytic phosphotransfer is accelerated in M-CK knockout hearts and may represent an important compensation alleviating myocardial energetic disturbances.

These results are consistent with studies of CK-deficient hearts by others. Increased activities of glycolytic enzymes such as pyruvate kinase and GAPDH were also found in hearts of CK knockout animals 94. M-CK deficient cardiomyocytes display a higher sensitivity to glycolytic inhibition manifested in premature opening of ATP-sensitive potassium channels and shortening of action potential as compared to wild type 92, suggesting a greater reliance on glycolytic metabolism. To this end, compensation provided by adenylate kinase and glycolytic phosohotransfers in CK-deficient muscles indicate their integral role in facilitating intracellular high-energy phosphoryl exchange especially under conditions of genetic or metabolic stress. Thus, metabolomic profiling and flux analysis reveal plasticity and restructuring of cellular bioenergetic system in response to genetic deficiency.

4. Conclusions

The 18O-assisted 31P NMR and mass spectrometric techniques provide a versatile methodology allowing simultaneous recordings of multiple parameters of cellular bioenergetics and characterization of metabolic fluxes through different energetic pathways. This includes simultaneous recordings of ATP synthesis and utilization, phosphotransfer fluxes through adenylate kinase, creatine kinase and glycolytic pathways as well as mitochondrial Krebs cycle associated nucleotide turnover and glycogen metabolism. This methodology has also a unique capability to measure intracellular energetic communication by comparing kinetics of Pi 18O-labeling (in ATPase compartment) to that of g-ATP (in ATP synthesis compartment). Integrated kinetic data obtained using 18O-labeling technology provides basis for cardiac system bioenergetics concept where major ATP-consuming and ATP-generating processes are interconnected by phosphotransfer network composed by adenylate kinase and creatine kinase circuits and glycolytic/glycogenolytic network nodes. Metabolomic and fluxomic profiling of phosphotransfer enzyme deficient transgenic animals (AK1-/- and M-CK -/-) using GC/MS, 1H and 18O-assisted 31P NMR indicate metabolic perturbations and adaptations in the whole energetic system.

In summary, 18O labeling technique is capable to monitor phosphotransfer reactions and energetic dynamics in all systems of interest in living tissues. Our studies demonstrate that this approach is valuable for metabolomic and fluxomic profiling of preconditioned and failing hearts as well as transgenic animal models simulating human diseases and diagnosis of mitochondrial energetic deficiency 15, 20, 22, 28, 29. Thus, metabolomic analyses in conjunction with system and network approaches provide new avenues for better understanding of cellular energetic system in health and diseases.

5. Acknowledgments

Supported by National Institutes of Health, Marriott Heart Disease Research Program, Marriott Foundation and Mayo Clinic.