INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES APPENDIX

MYOCARDIAL PERFUSION is very heterogeneous, even in healthy animal (2, 22) and human (34) hearts. This heterogeneity may increase during ischemia (34). Heterogeneous perfusion is matched to metabolism in the normal heart (18, 27), but perfusion-metabolism and perfusion-contraction matching become ineffective during coronary stenosis (4, 12). Transmural profiles of perfusion and metabolites have often been studied, but the fine-scale heterogeneity within each transmural layer is also important (2, 4, 12). However, measurement of local myocardial O₂ consumption (O₂) at sufficient spatial resolution to investigate intramyocardial heterogeneity has been problematic. Furthermore, to study metabolism-perfusion-contraction matching, O₂ should be measurable in 0.1- to 1-g samples as used for flow measurements with labeled microspheres.

Infusion of ¹³C-enriched substrate for the tricarboxylic acid (TCA) cycle leads to ¹³C labeling of glutamate. Estimation of the flux in the TCA cycle by analyzing the time course of the glutamate ¹³C NMR spectrum is possible for the intact heart (7, 8, 15, 36) but cannot yet be applied to several small tissue regions simultaneously. The fractional contribution of various ¹³C-enriched carbon substrates to acetyl CoA can be estimated from the ¹³C NMR multiplets of glutamate in extracts of frozen tissue samples (19, 20) so that several regions can be compared. The new feature of the method described below is that it allows one to simultaneously quantitate the TCA cycle flux and the fractional enrichment of acetyl CoA using frozen tissue samples. To this end the ¹³C-NMR multiplets of extracted glutamate are analyzed after brief, timed coronary infusion of ¹³C-enriched carbon substrate. Pilot studies using a simpler model analysis and labeling protocol (5 min infusion of [2-¹³C]acetate) have shown the validity of the principle of the method in isolated (28, 29) and in situ (31) rabbit hearts.

The goal of the present investigation was to develop a ¹³C double-labeling protocol to estimate O₂ in small myocardial regions. To this end the computer analysis of the ¹³C NMR multiplets of glutamate was refined, and the simultaneous estimation of six metabolic parameters from nine ¹³C-NMR multiplet intensities was investigated by computer simulation. The main focus of the present investigation is to compare this new ¹³C NMR method to quantitate O₂ with "gold standard" O₂ measurements in isolated, perfused hearts, which allow reliable O₂ determination. The correspondence between gold standard and new method establishes that local O₂ can be accurately quantitated by measuring ¹³C NMR multiplet intensities of glutamate on conventional, widely available NMR spectrometers. The method is now available to study intramyocardial O₂ profiles and perfusion-metabolism-contraction matching.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES APPENDIX

First, we describe the experimental protocol for the new ¹³C method and its comparison with conventional (gold standard) measurements of O₂ with O₂ electrodes. A 5.5-min infusion protocol was investigated to measure O₂ accurately (n = 26) and a 7-min infusion protocol (n = 9) to better measure other metabolic parameters. We then describe the new extended computer model applied for the estimation of O₂ from the ¹³C NMR spectrum of extracted glutamate. Finally, computer simulations are described to investigate the analysis method.

Heart Preparation

A 15-mm long glass cannula with a trumpet-shaped end was wedged deep into the coronary sinus to collect venous effluent specifically from the LV free wall. During dissection studies the course of coronary veins in relation to the left and right ventricle and region to be excised had been followed, and Evans blue dye was injected retrogradely to discern the region from which the veins collect venous effluent. The first vein entering the coronary sinus, close to the right atrial orifice, was found to drain both right and LV tissue. The veins entering deeper into the coronary sinus came from the LV region sampled for Wollenberger clamping. As a result of the dissection study, we ensured that the collecting cannula was wedged into the coronary sinus beyond the first vein. To measure venous O₂ tension, a sample flow of venous effluent was drawn by a pump from a side branch of the coronary sinus cannula via tubing with low O₂ permeability. The sample was drawn through a narrow, stirred cuvette over a Clark-type O₂ electrode (Radiometer). Intermittently arterial perfusate was also drawn from a side branch of the aortic cannula through the O₂ measurement cuvette. The O₂ electrode was calibrated by drawing oxygen-equilibrated perfusion buffer from a tonometer via the same tubing to ensure the same conditions as those during the experiment. In the 7-min series (see Experimental Protocol), this cuvette system was not available and O₂ tension was measured in a blood gas machine (ABL, Radiometer). The arterial-to-venous [O₂] difference multiplied with local perfusion, measured with radioactive microspheres (4), yielded regional O₂ for comparison with ¹³C measurements in the same region. The heart was submerged in venous effluent kept at 37°C to minimize O₂ diffusion across the epicardial surface (30). Details of the heart preparation have been described elsewhere (32).

Experimental Protocol

Both protocols started in the same way. To ensure a metabolic steady state, hearts were perfused for a half hour with unlabeled substrate. Thereafter, at t = 0, we switched from unenriched acetate to 99% [2-¹³C]acetate (Sigma), keeping acetate concentration the same. At t = 10 s, radioactive microspheres (¹⁴¹Ce; New England Nuclear; 15 ± 3 µm in diameter; 1-2 ×10⁵ spheres/injection) were injected into the inflow perfusion line over a period of 30 s. The microsphere solution was kept in a vial together with a drop of 0.05% Tween 80. Before injection, the spheres were ultrasonically vibrated for 5 min, agitated with a vortex mixer for at least 5 min, and finally continuously stirred with a home-built mixing device during injection. A 2.4 ml/min sample flow was withdrawn above the aortic cannula during 4 min starting 5 s before injection. There were no changes of heart rate, LV developed pressure, or arrhythmias after the injection of the microspheres. Half of the hearts showed no detectable change in perfusion pressure, the other half showed small increases in perfusion pressure (~5 mmHg). Two hearts, showing a >10 mmHg perfusion pressure change after microsphere injection for unknown reasons, were excluded from the study.

In the 5.5-min group, we switched at t = 4 min from 1.5 mmol/l [2-¹³C]acetate to 1.5 mmol/l [1,2-¹³C]acetate. After a total of 5.5 min of labeled acetate infusion, 90 s after we switched to [1,2-¹³C]acetate, part of the LV free wall was rapidly excised and freeze-clamped between two aluminum blocks, which were cooled to 80°C. The epicardium was pressed against one side and the endocardium against the other side of the freeze-clamp. The frozen tissue was stored at 80°C until further analysis.

To obtain a wide range of O₂, hearts in the 5.5-min group were perfused under a range of conditions: paced at 120 beats/min with LV free wall perfusion 20-32 ml · g dry wt¹ · min¹ (control, n = 6); paced at 210 beats/min, perfusion 40-67 ml · g dry wt ¹ · min¹ (high flow group, n = 4); inotropic stimulation with 320 µg/l dobutamine, flow similar to control (dobutamine group, n = 4). To test the ¹³C method during ischemia, hearts were hypoperfused at 4-19 ml · g¹ · min¹ by lowering pump speed and perfusion pressure (low flow group, n = 5); to investigate whether hypoxia gave similar results as ischemia, arterial perfusate PO₂ was lowered from 620 to 125 mmHg (hypoxia) in two hearts at high flow. To test the resolution of the method under conditions of very low metabolism, hearts were arrested with 20 mmol/l KCl in the perfusate, replacing NaCl (KCl-arrested group, n = 5).

In the 7-min group, the protocol was the same except that 0.5 mmol/l [2-¹³C]acetate was replaced by 0.5 mmol/l [1,2-¹³C]acetate after 5 min, and after an additional 2 min the LV free wall was excised and freeze-clamped. In the 7-min group, the hearts were paced at 125 beats/min (n = 3), at 210 beats/min (n = 4), or perfused with hypoxic perfusate, PO₂ 135 mmHg (n = 2).

Tissue extraction. Each LV free wall sample was freeze-dried for at least 24 h inside a 4-K Modulyo freeze dryer (Edwards). The LV free wall samples were then subdivided into subepicardial and subendocardial halves and weighed. Dry weight is given, except where indicated otherwise. Each tissue sample was homogenized in 4.0 ml of ice-cold 0.6 mol/l perchloric acid. The homogenates were neutralized to pH 7 with a solution containing 3 mol/l KOH and 0.3 mol/l imidazole and then centrifuged (10 min at 4,000 g). The pellets contained the microspheres used for blood flow measurement. Supernatants were freeze-dried for at least 24 h, then dissolved in 0.5 ml D₂O. After final pH readjustment, the supernatant was further diluted with tridistilled water to obtain 2 ml of solution for NMR spectroscopy.

¹³C NMR spectroscopy of extracts. High-resolution ¹³C-NMR spectra were obtained at 100.63 MHz with a Bruker MSL 400 spectrometer. The dissolved supernatants (2 ml) were studied in a 10-mm probe under high-resolution conditions at 25°C with a WALTZ-16 nuclear Overhauser enhancement and broad-band decoupling pulse sequence (9), pulse angle 45°, repetition time 6.5 s, 16-K data points, sweep width 10,000 Hz with 1,200 scans accumulated. The multiplet line intensities of the NMR free induction decay were analyzed in the time domain with the Variable Projection method (computer program MRUI/VARPRO, European Union HCM project) in the frequency-selective mode (11), constraining parameters by prior knowledge on multiplet J-coupling and amplitude ratios (Fig. 1A) (33). Multiplet intensities were expressed in micromole per gram of dry weight by comparison to a standard 50 mmol/l glutamate solution (¹³C at the natural abundance of 1.1%).

View larger version (21K): [in this window] [in a new window]   Fig. 1.   A: glutamate 2-, 4- and 3-carbon resonances in 13C NMR spectrum of extract of subendocardial muscle sample from an isolated perfused rabbit heart. Left ventricular (LV) free wall was freeze-clamped after 4 min of coronary arterial infusion of [2-13C]acetate followed by 1.42 min infusion of [1,2-13C]acetate. Line intensities were estimated by analyzing NMR free induction decay in the time domain. Measured NMR signal and time-domain fit, using 4,096 data points, and residue are displayed in frequency domain (without zero filling). Frequency is relative to midspectrum: 0 Hz corresponds with 50.7 ppm. Multiplet fine structure is clearly visible; numbers at multiplets correspond with B. B: isotopomers giving rise to multiplets. Solid black circle, 13C isotope giving rise to indicated resonance; circle with cross, 13C isotope not visible at this frequency but causes splitting of this peak; open circle, 12C isotope; ?, labeling not relevant for this peak. Peak 9 is G3S superimposed on the central peak of triplet G3T234. For multiplet symbols, see MATERIALS AND METHODS: 13C NMR multiplets and isotope labeling scheme.

¹³C NMR multiplets and isotope labeling scheme. A ¹³C NMR peak is split into multiplets (see Fig. 1) by J coupling to adjacent ¹³C isotopes (19, 20). For instance, a ¹³C-labeled isotope in the glutamate four-carbon position (G4) yields a singlet (G4S) if there is no adjacent ¹³C but yields doublets (D) if ¹³C is present in the fifth or third position (G4D45 and G4D34, respectively); for ¹³C neighbors at both sides, a quartet results (G4Q345). The two-carbon of glutamate (G2) shows similar fine structure; for the three-carbon of glutamate (G3), the two doublets are indistinguishable (G3D23 and G3D34) and two ¹³C neighbors give a triplet (G3T234).

View larger version (18K): [in this window] [in a new window]   Fig. 2.   Scheme of mathematical model for incorporation of 13C in tricarboxylic acid (TCA) cycle and glutamate. 13C from [2-13C]acetate is incorporated in acetyl CoA pool with time constant trans after transport through blood vessels, cell membrane permeation, and activation by enzyme acetyl CoA synthetase. During the first turn of the TCA cycle, 13C from 2 position of acetyl CoA enters 4 position of the 5-carbon intermediate -ketoglutarate, where carbon-skeleton exchanges with 4 position of glutamate at flux Jexch. Carbon from 1 position of acetyl CoA enters 5 position of glutamate. There is also exchange of 13C between oxaloacetate (in 4-carbon TCA pool) and aspartate. Unenriched anaplerotic substrates enter TCA cycle at flux Janap, diluting 13C. Carbon from 4 position of glutamate and 5-carbon pool (-ketoglutarate) randomly enters either 2 or 3 position of oxaloacetate after the next turn of the TCA cycle from which it continues to 3 and 2 position, respectively, of -ketoglutarate and glutamate; carbon in 5 position of glutamate enters either 1 or 4 position of oxaloacetate, and from there goes to CO2 (decarboxylation by isocitrate dehydrogenase) or to 1 position of glutamate. One example of sequential entry of 13C atoms during infusion of [2-13C]acetate is shown: [3,4-13C]glutamate is formed in the second turn of the TCA cycle and [2,3,4-13C]glutamate in the third turn; many other enrichment sequences are possible.

Dynamic Model of TCA Cycle Enrichment

The new model (Fig. 2) consists of 6 metabolite pools: acetyl CoA with 4 isotopomers (2² combinations of ¹²C and ¹³C), 32 isotopomers both in the glutamate and the 5-carbon (-ketoglutarate) pools, 16 isotopomers in the 4-carbon TCA cycle pool (succinyl-CoA, succinate, fumarate, malate, oxaloacetate) and also 16 for aspartate. The 64 isotopomers of 6-carbon intermediates (citrate, cis-aconitate, isocitrate) were combined to 32 pairs, because labeling of the 6-carbon, which disappears in the next reaction step, is irrelevant for glutamate labeling. The time course of enrichment of acetyl CoA after transport through blood vessels and cell membrane permeation (Fig. 2) is characterized by an exponential increase represented by the estimated time constant _trans, 0.5 min as found by ¹⁴C labeling of acetyl CoA (24). This cannot be neglected for short infusion experiments. The equations for the ¹³C enrichment of the metabolite pools are given in detail in the APPENDIX.

Acetyl CoA is usually not fully labeled, but the fractional enrichment of acetyl CoA can be determined from the multiplet pattern (19, 20). F_C2 represents [2-¹³C]acetyl CoA as a fraction of total acetyl CoA after a steady state has been reached during [2-¹³C]acetate infusion. F_C2 equals F_C3, [1,2-¹³C]acetyl CoA as a fraction of total acetyl CoA, because the labeling is assumed not to influence the behavior of acetate. Because unlabeled anaplerotic substrate entering the TCA cycle lowers the [2-¹³C]glutamate-to-[4-¹³C]glutamate ratio (or the [3-¹³C]glutamate-to-[4-¹³C]glutamate ratio), the relative anaplerotic flux (J_anap/J_TCA) is a model parameter that is estimated.

Citrate, aspartate, and glutamate, relatively large metabolite pools, were determined by biochemical assays (3). Literature values for the 5- and 4-carbon intermediate contents, which represent small pools in the rabbit heart that have little effect on the rate of label incorporation, and measured citrate and aspartate for each sample were fixed in the model (38). However, glutamate is estimated from the NMR multiplets as one of the six model parameters, because some glutamate pools may not be reached by ¹³C label on the time scale of the experiment (see below). O₂ can be calculated using the optimized model parameters via O₂(¹³C) = (3  F_C2) · J_TCA, when acetate and glucose are the substrates (see DISCUSSION) (8).

Computer Simulations

Monte Carlo simulation studies with the full model were done for the present study to prove that 4 min of [2-¹³C]acetate plus 1.5 min of [1,2-¹³C]acetate infusion allows simultaneous estimation of six model parameters from nine multiplet intensities and for parameter sensitivity assessment. The 7-min infusion protocol was simulated to investigate how well the ¹³C accessible glutamate pool can be estimated. To NMR peaks, calculated using the model for known parameter combinations, we added simulated NMR noise with Gaussian distribution, using 8.2 ± 3.2% (SD, range 2.8-11.1%) for the relative error in nine NMR line intensities. These SD values are the Cramer-Rao lower bound estimates resulting from the time-domain analysis of measured NMR signals. In this way 25 Monte Carlo runs were done per condition. The parameters were then reestimated using the Marquardt-Levenberg nonlinear least squares optimization routine SENSOP (6) to fit the simulated data sets, including simulated NMR noise.

Statistics

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES APPENDIX

Computer Simulation of Labeling Protocol

1

1

1

1

1

1

1

1

1

1

The time course of development of individual multiplets under our infusion protocol was simulated. Figure 3A shows that the singlet of C4 glutamate (G4S) develops first after infusion is started with [2-¹³C]acetate. The initial delay before the upstroke of G4S is caused by _trans. Glutamate's C3 and C2 are enriched in the next turn of the TCA cycle. The appearance of the G4Q345 and G4D45 multiplets, due to infusion of [1,2-¹³C]acetate in the last 1.5 min, is sensitive to transport delay (Fig. 3A) and therefore allows better estimation of _trans.

View larger version (21K): [in this window] [in a new window]   Fig. 3.   A: simulation of development of 13C NMR multiplets of glutamate by numerically integrating 132 differential equations of model. To predict resonance line intensities in the sample at end perfusion, a 4-min arterial infusion of [2-13C]acetate followed by 1.42 min of [1,2-13C]acetate is simulated. See Fig. 1B and text for nomenclature of multiplets. Solid lines, 4-carbon multiplets; dashed lines, 3-carbon multiplets; dotted lines, 2-carbon multiplets. Multiplet resonance areas are expressed per gram dry weight. B: fit of the TCA cycle model to measured 13C NMR multiplet areas. Model-predicted results are those of A at end of simulation (5.42 min), obtained after nonlinear least-squares optimization of model parameters to fit experimental data. Corresponding estimates of model parameters and correlation coefficient r are also given.

To prove the feasibility of the estimation of many parameters simultaneously, infusion protocols were simulated, and realistic NMR measurement noise was added. The results of simultaneous reestimation of six model parameters from nine NMR multiplets is given in Table 1. The 5.5-min infusion protocol allowed accurate estimation of J_TCA and F_C2, and the 7-min protocol was better for estimation of the ¹³C-accessible glutamate pool. Simultaneous estimation of six metabolic parameters from the glutamate NMR spectrum of one sample extract (one time point) is possible and, despite appreciable spread in some parameters, the two parameters necessary for calculation of O₂, i.e., J_TCA and F_C2, are well estimated. Estimates were independent of the initial values given to the nonlinear least squares optimization routine: four very distinct sets of initial values converged to the same final estimates (n = 25 per set).

Although J_exch could be estimated using our model to fit the 40-min time course data of Yu et al. (38), the simulation predicts relatively poor estimates of J_exch with our estimation at the single time point t = 5.5 min. The short duration of infusion makes the analysis insensitive to J_exch. We analyzed the sensitivity of the estimation of J_TCA and F_C2 to variations in J_exch under the 5.5-min infusion protocol by performing a set of Monte Carlo simulations with different fixed values for J_exch (see Fig. 4). One finds incorrect estimates for J_TCA and F_C2 with increased SD only if J_exch values are set erroneously below 5 µmol · g¹ · min¹ (true value was 10 µmol · g¹ · min¹). With J_exch >5 µmol · g¹ · min¹, we obtained accurate and constant estimates and SD values for J_TCA, F_C2 and also for the other parameters. J_TCA deviated >10% only for J_exch 3 µmol · g¹ · min¹. The experimentally determined J_exch was 28.4 ± 2.3 µmol · g¹ · min¹ (n = 42). It is concluded that the estimation of the key parameters J_TCA and F_C2 is not sensitive to J_exch in our brief perfusion protocol.

View larger version (12K): [in this window] [in a new window]   Fig. 4.   Sensitivity of JTCA and [2-13C]acetyl CoA fraction of total acetyl CoA (FC2) to Jexch. Estimates of JTCA and FC2 are given as a function of value of Jexch, which was fixed in these simulations. Double-labeling experiment was simulated and realistic noise was added (23 data sets). Four metabolic parameters were reestimated, keeping Jexch fixed at different values between 1 and 60 µmol · g1 · min1. A: effect is shown of perturbed values of Jexch (which was 10 µmol · g1 · min1 in original simulation) on estimation of JTCA (10 µmol · g1 · min1 in original simulation, indicated by horizontal dashed line). B: on FC2 (0.9 in original simulation). Error bars give SD.

Experimental Test of ¹³C NMR Measurement of O₂

Estimates of glutamate were compared with the biochemically obtained values. For the 5.5-min labeling protocol, the estimated value of glutamate was poorly correlated (r = 0.476, n = 42, P < 0.01) to the biochemical values. However, for the 7-min protocol a correlation of r = 0.893 (n = 15, P < 0.01) was found with a slope of 0.82 ± 0.11. This better estimation of glutamate after longer labeling with ¹³C agrees with the Monte Carlo simulation result (Table 1), where the 7-min infusion protocol decreased the SD of the estimate by 74% compared with that of the 5.5-min protocol. The biochemically measured content was 6.3 ± 0.6 µmol · g¹ · min¹ greater (P < 0.05, n = 15) than that of the model-derived estimate from the NMR data. The lower estimate of glutamate by the ¹³C method, compared with the biochemical assay, suggests that some intracellular glutamate pools are not accessible to the ¹³C label on such a short time scale. To avoid bias caused by compartmentation, we estimate glutamate content simultaneously with the other metabolic parameters, rather than fixing the biochemical assay value for the model analysis.

Simultaneous analysis of the measured multiplets with the ¹³C model showed that the median of the J_anap-to-J_TCA ratio was 0.056 ± 0.011 (n = 42) in the 5.5-min group and 0.068 ± 0.015 (n = 15) in the 7-min group. J_anap/J_TCA did not correlate with J_TCA. Infused acetate contributed an estimated 91.0 ± 1.4% (n = 42) to the acetyl CoA pool in the 5.5-min group with 1.5 mmol/l acetate and 75 ± 4% (n = 15) in the 7-min group with 0.5 mmol/l acetate. There is no assumption in the method regarding the transport time. The _trans is estimated from the multiplets and was 36.7 ± 2.3 s (n = 42) in the 5.5-min group and 21.6 ± 3.4 s (n = 15) in the 7-min group, in good agreement with the time course of uptake of radioactively labeled acetate in the myocardial acetyl CoA pool (24). The _trans was not correlated with local perfusion (r = 0.270, n = 42, P > 0.05) and local O₂ (r = 0.114, n = 42, P > 0.05).

The total LV free wall O₂, calculated from J_TCA and F_C2, was compared with conventional measurements of O₂ in the LV free wall based on O₂ electrode measurements (Fig. 5), showing good correspondence (r = 0.932, P < 0.0001). The calibration regression line, for the control and hypometabolic groups only, is O₂ (¹³C) = 1.05 · O₂ (Clark-type electrode)  0.77 in µmol · g¹ · min¹ (slope not different from 1, n = 22, P > 0.05). For the high flow group the scatter appears high and before the method is applicable to strongly increased metabolic rates, additional work is required. Approximately 1.6 µmol · g¹ · min¹ of myocardial respiration has been shown to be nonmitochondrial (5). That the O₂ electrode determined O₂ is 2.4 ± 0.5 µmol · g¹ · min¹ (n = 5) during KCl arrest, in the absence of ¹³C-detectable O₂, underscores that the O₂ consumption measured by the ¹³C NMR method is exclusively mitochondrial. The scatter around the regression line is similar to the scatter in the KCl-arrested hearts so that the nonmitochondrial O₂ accounts for a large part of the residual scatter. For the 7-min group the correlation between ¹³C and Clark electrode O₂ measurements was 0.972 (n = 9, P < 0.01) with slope 0.89 ± 0.08. From the high linear correlation it is concluded that the new ¹³C method allows accurate quantitation of local O₂ in myocardial tissue at control and lowered metabolic rates.

View larger version (16K): [in this window] [in a new window]   Fig. 5.   Comparison of new 13C method with "gold standard" measurements. O2 consumption (O2) estimated from 13C NMR measurements for entire LV free wall of isolated rabbit hearts vs. O2 calculated from coronary venous [O2] for same region, measured with a conventional Clark-type electrode and local perfusion measured with radioactive microspheres. Hypoxia, O2 tension in perfusate lowered from ~620 to 125 mmHg, at high perfusion flow. For definition of groups, see MATERIALS AND METHODS. Solid line, linear regression line; dashed lines, 95% confidence bands.

Subendocardial Versus Subepicardial O₂ Consumption

The subendocardial-to-subepicardial (Endo/Epi) O₂ ratio was 1.52 ± 0.19 (n = 21, significantly > 1: P < 0.025) and was not correlated (P > 0.10) to average LV free wall perfusion flow (r = 0.02) or perfusion pressure (r = 0.32). The Endo/Epi O₂ ratio is, however, significantly correlated to the Endo/Epi perfusion ratio (r = 0.63, n = 21, P < 0.01), which was 1.73 ± 0.23 (significantly >1: P < 0.005). Remarkable is that local O₂ was positively correlated (P < 0.05) to glutamate content independently measured by biochemical assay (r = 0.31, n = 42, 5.5-min group; r = 0.67, n = 15, 7-min group), as also reported previously (29). Local O₂ was related to local perfusion flow (R_s = 0.917, n = 38, P < 0.0001) measured with radioactively labeled microspheres in the same samples (Fig. 6). The curvature in the relation indicates that local O₂ extraction is increased at lower local flow.

View larger version (18K): [in this window] [in a new window]   Fig. 6.   O2 (estimated from 13C NMR multiplets) vs. local perfusion in LV free wall of isolated perfused rabbit hearts, measured with radioactively labeled microspheres. LV subendocardial (i.e., from inner muscle layer) and subepicardial (i.e., from outer muscle layer) samples of 49-177 mg dry mass are given separately. Perfusion pressure was lowered to cause hypoperfusion. Quantities expressed per gram dry weight.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES APPENDIX

The present investigation shows that measurement of glutamate enrichment in frozen myocardial samples after brief, timed infusion of ¹³C-labeled substrates makes it possible to quantitate local O₂. The NMR technology required to measure high-resolution spectra in solutions is conventional and widely available. Costs are modest if the enriched substrate is infused intracoronarily, and because the fractional enrichment of acetyl CoA is an estimated model parameter, variation of labeling across the tissue does not cause problems. The transmural spatial resolution of the new method, at present ~2 mm, is already higher than the ~6 mm obtainable with cardiac positron emission tomography (17), and higher sensitivity can be obtained with a 5-mm NMR probe or by investing more NMR time than the ~2 h/sample used here. Much higher sensitivity may be obtainable with the NMR microcoils, which are under development (23). The analysis method is useful in the future to scans of frozen tissue using polarization-enhanced NMR spectroscopy (10). With the application of only modest resources, the method to measure local O₂ already attains similar resolution (50 mg dry wt) as the labeled microsphere blood flow measurement (2).

With the combination of cryospectrophotometric values of hemoglobin saturation in local arterioles and venules with blood flow measured in adjacent tissue samples using radioactively labeled microspheres, the local O₂ has been calculated for subendocardium and subepicardium (35). The values found for the rabbit LV free wall at a heart rate of 230 beats/min were 20.7 ± 2.4 and 15.4 ± 1.3 µmol · g dry wt¹ · min¹ in the subendocardium and subepicardium, respectively. These values are similar as those we found with the ¹³C method, using a 5-min infusion of [2-¹³C]acetate, in the rabbit heart in situ, 17.6 ± 2.8 and 11.9 ± 1.7 µmol · g dry wt¹ · min¹, respectively, at 260 beats/min (31). The values in the isolated perfused control group are also quite similar (Figs. 5 and 6). Comparison with the in situ O₂ of ~18 µmol · g¹ · min¹ shows that we have tested the range from 0-160% of resting O₂ in the rabbit heart. Higher O₂ is not feasible without damaging the isolated heart. O₂ determined during ischemia and hypoxia is on the same linear relation as normoxic controls. It will require more validation to apply the method at strongly increased work loads, which cannot be obtained in isolated rabbit hearts. Thus the new method has been validated here for normal and low metabolic rates and is useful for studies on myocardial ischemia and hypoxia.

Heterogeneity of regional myocardial O₂ has been inferred previously from cryospectrophotometric measurements (39) and from glucose uptake and adenosine measurements (27). To estimate the extent of heterogeneity with the new ¹³C method, one must have an estimate of the variation caused by the method. The scatter of O₂ during KCl arrest, characterized by SD 1.5 µmol · g¹ · min¹, reflects natural variability of extramitochondrial O₂ and variability of the venous O₂ measurement. The scatter around the regression line is characterized by SD 2.4 µmol · g¹ · min¹. It is assumed that the deviations from the line in Fig. 5 are caused by random addition of error in the ¹³C measurement and estimation procedures on the one hand and on the other hand by the natural variation of extramitochondrial O₂ and error in the O₂ electrode measurement. The latter two are jointly given by the SD of O₂ during KCl arrest. The addition rule of variances for the sum of independant variables then enables us to calculate the SD for O₂ determined with the ¹³C measurement method, 1.9 µmol · g¹ · min¹.

As a substitute for O₂, glucose uptake has been determined by the deposition of radioactively labeled deoxyglucose to assess metabolic intensity (21, 26, 27), but not all the glucose is used for aerobic metabolism and other substrates than glucose are often more important for aerobic metabolism in the heart. Our ¹³C method measures O₂, whereas glucose taken up also indicates anaerobic glycolysis. The rate constant of disappearance of radioactive label after infusion of [¹¹C]acetate has been used as a noninvasive measure of aerobic metabolism (1), but this approach does not quantitate O₂ in absolute terms. The ¹³C method measures mitochondrial O₂ accurately because O₂ is stoichiometrically coupled to acetyl CoA turnover in the TCA cycle. The stoichiometric relations, e.g., C₆H₁₂O₆ + 6 O₂  6 CO₂ + 6 H₂O for glucose, are very tight. Because two acetyl CoA molecules enter the TCA cycle per molecule of glucose, the mitochondrial O₂ is exactly three times the number of acetyl CoAs entering the TCA cycle per unit time, which we measure directly with this ¹³C method. One acetyl CoA is derived from acetate, and according to the stoichiometric equation C₂H₄O₂ + 2 O₂  2 CO₂ + 2 H₂O, the O₂-to-acetyl CoA ratio is 2. For hearts perfused with glucose and acetate it follows that O₂ = (3  F_C2) · J_TCA, as corroborated directly by ¹³C data (8). For the fatty acids palmitate and stearate, the O₂-to-acetyl CoA ratio is ~2.9, and fatty acid usage does not lead to appreciable deviation from the value of 3 in the equation. Infused ¹³C-enriched lactate or pyruvate also enters the TCA cycle, leading to appreciable glutamate labeling in the in situ dog heart (13), and the method should also be applicable for these substrates. If pyruvate, lactate, or unsaturated fatty acids are metabolized in appreciable amounts, the stoichiometric number 3 in the equation can be modified.

Although O₂ and TCA cycle flux are very tightly linked, there has been discussion about the P/O ratio (i.e., ATP produced per O₂ consumed) (14). However, even when the mitochondria are uncoupled, the relation between TCA cycle flux and O₂ will not change. A small portion of the O₂ taken up by the beating myocardium is not used for oxidative phosphorylation in the mitochondria but is consumed by other biochemical reactions (5). Our method measures only that part of the O₂ that is directly coupled to the TCA cycle, and extramitochondrial O₂ is not measured. This explains the intercept for measurements with the O₂ electrode (Fig. 5). During KCl perfusion no TCA cycle flux is detected with ¹³C. A large part of the O₂ during arrest is apparently not linked to TCA cycle flux and oxidative phosphorylation.

The ¹³C method gives a somewhat lower O₂ than the venous O₂ electrode, because ¹³C indicates only TCA cycle-linked O₂. With this taken into consideration, the ¹³C method measures mitochondrial O₂ well for normal, hypoperfused, and hypoxic myocardium. Reperfusion after brief ischemia has been shown to lower J_exch, affecting the enrichment of glutamate (16, 37). However, the sensitivity of our method for J_exch is low (Fig. 4), and deviating J_exch influences the calculation of O₂ only under exceptional conditions of very low J_exch.

The ¹³C-labeling pathways included in our model conform with those in previous models (7, 8, 15, 16, 19, 20, 36-38), except for the inclusion of the transport time, which is appropriate for these short labeling protocols (24). Labeling of extraneous pathways does not interfere with the measurement. A low level of labeling is expected of fatty acids, but this is a parallel pathway of low flux not influencing the TCA cycle. Acetate and acetyl CoA cannot be converted to pyruvate in mammalian tissue, and anaplerotic entry into the TCA cycle, which is known to exist from pyruvate, therefore plays no role. Most importantly, the labeling pattern we see in the spectra is consistent with the model, and no peaks resulting from extraneous labeling are seen in the NMR spectra. However, in the future there is room for more extensive modeling, for instance of distinct anaplerotic pathways, which may lead to improved estimation at the higher than normal work loads. Additional information on less abundant intermediates, derived from mass spectrometry, may be used for such extended models.

The double-labeling 5.5-min protocol increases the accuracy of O₂ estimation significantly compared with single-label infusion. The 7-min double-labeling protocol gave better estimates than the 5.5-min protocol of glutamate content and other parameters (Table 1), although the accuracy of the J_TCA estimate was decreased. Thus, depending on the metabolic parameter of interest, a different protocol can be designed using computer simulations.

As a first mechanistic result of the method, we found in this study that in isolated hearts the subendocardial O₂ is significantly higher than the subepicardial O₂. This had previously been inferred for in situ hearts based on cryospectrophotometry (35) and was also found by the simpler version of our ¹³C method in rabbit heart in situ (31). The higher subendocardial energy turnover appears as a robust characteristic of rabbit heart and contributes to the higher subendocardial vulnerability to infarction. The variability of local O₂ we find (Fig. 6) confirms earlier observations obtained with cryospectrophotometry (39) and appears to be bigger than ¹³C measurement error (Fig. 5).

In conclusion, we have shown that the new ¹³C method described here makes measurements of local O₂ in tissue samples feasible and is applicable to resting state and hypoperfused tissue. Local energy turnover can henceforth be measured in studies of metabolism-perfusion-contraction matching. Analysis of the ¹³C NMR multiplet "fingerprint" of frozen tissue samples provides robust information on tissue metabolic rates and other metabolic parameters.